roblcms 'l:2fr-'lo IJ"fr+'Jo / on p. 197, 'vhen dislocations and ~concentrated forces and couplcs 9 E are to be includcd. ºk--r"h+--.x 1 9 64. Hyperbolic Boundarles. I Notches. It "'as sho>vn in .i\.rt. 60 '7"9o 'l:ff-IJo that the curves '11 = constant in elliptic coordinates are hyperbolas, and in Art. 62 that t.he range of" may be taken as O to 211", that of ~ Fra. 130. being O to oo. Let 'l1o be thc constant valuc of '11 along the hypcrbolic are B.4. of Fig. 130. It ,vill be bet\vecn O and 1f /2, f'ince bolh x a.nd y are positive along BA. Along the other hrJf of t.his branch of the hyperbola, BC,
1 1 1 1
,.,.,.----
-
--- '
f
,,
---
' ,_
K. '\Yolf, Z. lech. Physik, 1922, p. 1130. 'H. Neubcr, lngenieur-Archiv, vol. 5, p. 242, 1934. This solution and several others relating tu cllipses and hyperbolas are givcn in Neuber's book "Kerbspannungslehrc," Bcrlin, 1\J37. i P. S. Symonds, J. Applú:d ,ifecho.nics (Tro.ns. A.S.,1!.E.), vol.13,p. A-183, 1946. A solution in fi.nite form is givcn by A. E. Oreen, ibid., vol. 14, p. A-246, 1947. 4 N. I. l\1uschcli8vili, Zeit. o.ngew. Jfalh. 2l
20;)
PROBLEJ.!S l1V CURV!Lfll/EAR COOllDf,VATES
THEORY OF ELASTICITY
204
+
0,
Y,(z) = -jA.i 1,
0
•
'
x(z)
=
-t.4i 1 - Bci sinh 1
1vhcre A and B are real constants, und z f'(') -- - 2c sinh iA I'
X' (z)
=
=
e cosh \.
21 Ai. 1 -
-
(' A 2
These give
+ B )· coth 1 i
E~uation \'
(103) of Art. 59 shcnvs that the hypcrbolic boundary 11 be free from force proyidcd the function
f('J +'Vi(') +
(a)
(b) = 110
kl
x'(')
is constant along it, or equivalently if the conjugatc of this function is con1>tant. 1'hc conjugate is, from I~qs. (a) and (h),
A 11-zlA.coshf isinh.7- ('2-4+B) ·icotht ()n t.hc hyperbola 7J expression becomcs
Â'IJo -
= '11"
\Ye have f
1-
=
2i11
º'
t.4. sin 27/o - (t
0
(d)
and \\-"ith this the
+ -}.4. + B)i coth 1
~vhich i1> a constant if the quantity in parentheses is made to vanish.
Thu::1 B
=
-A
co;; 2 'lo
(e)
'fo find thc rcsultant force t.ransmittcd 11·e may apply Eq. (103) of \rt. 59 to thn narro'v secLion EOB, Fig. 130, more prc('iscly to the Io,ver part of t.he limiting ellipse ~ = O bet\veen thc hyperbolas = and _ . . . _ 7J '17íl 7J - 11" - 1Jo· 0 n t 1us elhpse !; becomes i7J, 1 becomes -i7/, and \t·e have from Eq. (103), (e) and (d)
1
+ R) cot '17]~.:;;:-~·
F~ - i"Fu = i"[A11 - (A = i[A(:ir - 2'170
+ 2 cot 7Jo) + 2B cot 7Jo]
'Th·Is pro bl ()til (also the case of shear loadin") \Yll.S solvcd b~c A A G 'ffitb "·J·.1"11, TechRA · ept. rronau( Research Comm. (Great Britain) 1027-1928 vol II 668· andH:-;;bz "h , ,.,p., h · - eu er, . o.ngew. 1rial . Mech., vol. 13 p. 439 1933· or "Kcrba ,e re," p. ;)5, Berlin, IV38. ' , ' pannungs-
THEORY OF ELASTICITY
206
PROBLEMS IN CURVILINEAR COORDINATES
Since A and B were taken as real, F,. is zero and, using Eq. (e)
Fu
-A(11" - 211 0
=
+ sin 21'1o)
which determines A 'vhen the total tension Fv is assigned. The stress and displacement components are easily found from Eqs. (109), (110), ( 111). The first gives 4A cosh I; sin 11 u~+u~= cosh 21; - cos 217
- -e-
The value of 11 1 along the hyperbolic boundary is found by setting 11 = 110 in this expression. It has a maximum, -2A/c sin 1)0, at the waist ,vhcre I; = O. Ncuber 1 has exprcssed this as a function of the radius of curvature of thc hyperbola at the "''aist. He has solved, by anothcr mcthod, the problems of bcnding and shear of the plate as well as tension. 66. Bipolar Coordinates. Problems involving two nonconcentric circular boundaries, including thc spccial case of a circular hole in a semi-infinite plate, usually require the use of thc bipolar coordinates t, 11, defined by (a) z = ia coth -!r r = t i11
207
It may be seen from Fig. 131 that 61 - 62 is the angle between the two l~nes joinin~ the "poles" -ia, ia to the typical point z, when this point hes to the nght of the y-axis, and is m.inus this angle when the point lies to the left. It follo\\'S that a curve 1/ = constant is an are of a circle passing through the poles. Several such circles are drawn in Fig. 131. From Eqs. (e) it is clear that a curve ~ = constant will be a curve for ,~·hich r1/r2 = constant. Such a curve is also a circle. It surrounds the pole ia if r1/r2 exceeds unity, t.hat is, if tis positive. It surrounds the other pole -ia if ~is negative. Several such circles are drawn in
+
+
Replacing coth ir by (eit e-!t)/(etr - e-Ir) and solving the first equation for et, it is easily shown that this is equivalent to
r=
y
z +ia log--.
(b)
z - ia
+
The quantity z ia is represented by the line joining the point -ia to the point z in the xy-plane, in the sense that its projections on the axes give the real and imaginary parts. The sarne quantity may be represented by r 1e1º1 ,vhere r 1 is the length of the linc, and 01 the angle it makes ,vith the x-axis (Fig. 131). Similarly z - ia is the line joining the point ia to the point z, and may be represented by r28' 9• (Fig. 131). Then Eq. (b) becomes
~ + i 11
= log
(~ eie•e-'º')
= log
~ + i(01 -
r,
= log-,
"
Fig. 131. They form a family of coaxal circles with the two poles as limiting points. The coordinate '1 changes from 1r to -'Ir on crossing the segmcnt of the y-axis joining ~he poles, its range for the ''rhole plane being -'Ir to 'Ir. Stresses and d1splaccments -.,vill be continuous across this segment jf they are r~presented by periodic functions of 11 \vith period 2'1r. Separat1on of real and imaginary parts in Eq. (a) leads t 0 t X=
92)
a sin 11 ' cosh ~ cos 1)
y
=
a sinh ~ cosh ~ C08 '1
(d)
Differentiation of Eq. (a) yields
so that ~
Fio. 131.
1 . h 2_.\ 1 --iacosec 2 2
(,)
1 Loc. cit. For a cornparison of Neubcr's rcsults v;ith photoelastic and fatigue teats of notched platcs and grooved shafts sce R. E. Peterson and A. M. Wahl, J. Applied Mechanics, vol. 3, p. 15, 193fi, or S. Timoahenko, "Strength of MaterialB," 2d ed., vol. 2, p. 340. Sec also M. "'.\f. }'rocht, "Photoelasticity ," voL 2.
and 2 ia
e
_ dz/dt - dZ/df
=
-
'nh211 - .i cosech2 - .i 2 2
s1
'Soo the deriva.tion of Eq. (e) in Art. 54.
(<)
(!)
'l'HEORY OF ELASTICITY
PROBLE.lJS IN CURVILIJ.IEAR COORDINATES
66. Solutions in Bipolar Coordinates. W e no'v consider the problcm of a circular disk with an eccentric hole, subject to pressure Po round the outsi
This may be expresscd in bipolar coordinates by mcans of Eqs (d) of Art. 65, the result being ·
208
cosh n,\,
v.·ith n an integcr, sincc these are in fact periodic functions of 1/ of period 2:ir. So also are their derivatives \Vith respect to z, since d,\/dz has thc sarne propcrty [Eq. (e), •.\.rt. 65]. Jf such functions are i:atroduced into Eqs. (103), (104), applied to any circle t = constant in the material, the corresponding force and couple 'vill be zero, in virtue of the periodicity. This must hold for the complete solution, for equilibrium of the plate 'vithin the circle. We shall requirc also thc function x(z) = aD,\, D being a constant. Considering this in Eqs. (103), (104) as above, we find that the momcnt of Eq. (104) v.'ill be zero only if D is real. We therefore take it to be so. Considering the displacernent equation (99) we find that this function, as \vell as the functions (a) used as either !f(z) or x(z), will give displacements free from discontinuity. The state of uniform all-round tcnsion or compression, which vrill be part of the solution, is obtained from the complex potential ,Y(z) = Az \vith A real. The corresponding real stress function is, from Eq. (97), =
A(x2
+ y2)
'The original solutíon, in terms of the real stress function, is dueto G. B. Jeffery, 1'rans. Roy. Soe. (Lu11don), series ./\., vol. 221, p. 265, 1921.
2
cosh~+cos1/
cosh
~
(b)
cos 11
(;onsidering functions of the form (a) ""'·ith n = 1 '\Ve ob th · th t serve . at s1nr·c ~ s ress 1str1but1on in thc prcscnt problem is symmetrical about :·~e y-ax1s, we must choose them so that the corresponding stress func..1ons have the sarne symmetry. Thus vle may take
a· . .
'
f(z) = iB cosh .\,
,
x(z) = B' sinh
.1
(e)
\\'ith B, B' real, and
r.
f(z) = iC sinh x(z) = C' cosh .1 1vith C, C' real. 'fhe real stress function corresponding to (e) is, from Eq. (97), 11inh t co;;h t cos 1/ - sinh t sin2 1/ cosh t - cos 11 B' r;inh t cosh t cos cosh ~
aH
+
(a)
sinh n.\
q, = Re(zAz) = Azz
,
''"
20!)
(d)
sinh t co;; 2 co;; 1/
1/ -
1/
~fc 1•re c~oose B' = aB thc tcrms in sin2 11 , cos2 1/ in the numcrators Co~e 1ndependent of 7/, and thc complete numerator depends on 1/ ~~·ly 1~ t~c term in cos 1/, just as. does the function (b). The sarne thing IS t1~e o_f the complcx potcnt1als (d), if v.cc choosc C' = aC. U8 obta1n s1mpler, more reRtrict.ed functions v.·hich turn out to be ad!Jquatc for the present problem. 'faking thereforc
'''e
!f(z) = iR cosh .\,
x(z) = aB sinh .1
(e)
\\"C fi d b co n Y_means of Eqs. (109), (110) and (a), (f) of Art. 65, that the rresponding stress componenhi are given by
a(u- _ ~
a(u-r; .+ u-~) = 2R(~ sinh t cos
1/ -
sinh 2 t cos 2'11)
u-~ + Ztr<~! = -2B[s1nh 2t - 2 sinh 2t cosh t cos 11 + sinh 2t cos 21/ - i(2 cosh 2~ cosh t sin 11
s·1 "]
-
m1 arly the functions
f(z)
=
iC sinh .\,
x(z)
=
cosh 2t sin 211)]
aC co1:1h
.1
(fJ (g) (h)
THEORY OF ELASTJCITY
210 yield
a(ui: +O'~) = -2C(l - 2 cosh ~ cos 1/ a(u~-u;+2iTi:~) =
+
PROBLEMS 11\' CURVILlfoiEAR COORDlll'ATP.'15
+
and therefore
+ cosh 2~ cos 21))
2C[- cosh 2~ 2 cosh 2~ cosh ~ COSf/ - cosh 2~ cos 21} i(2 sinh 2~ cosh !'; sin 1J - sinh 2~ sin 271)] The stress components arising from
(J)
are given by ui:+u~=O
=
D[sinh
2~
(l)
The 1:1tate of uniform all-round tension given by (m)
,P(z) = Az
yields
.,., + u~ = ui:
=
4A, u~ =
(n)
2.4,
The solution of our problem can be obtaincd by superposition of the states of stress rcpresentcd by the complex potentials (e), (h), (k), and (m). Collecting thc terms rcpresenting T<~ i~ Eqs. (g), (j), and (l)_vve find that thc vanishing of ri;~ on the boundar1es ~ = ~º' ~ = ~' requ1res
D - 2R cosh JJ - 2B cosh
2~º 2~1
These with Eqs. (p) complete the determination of the complex potentiali:!. When there is interna} pressurc Pt only (p0 = O) the peripheral ;;tress at the bole is found to be ((T~)<-b = -pi+ 2pi(sinh 2 ~1
- '2 sinh ~ cos 1/ - i(2 cosh ~ sin 1/ - sin 211)]
- 2C sinh - 2C sinh
2~. = 2~1 =
O O
(a)
Solving these for B and C in terms of D we have 2C = -D sinh (ti+ to) cosh (~i t.)
(p
2A
+ Da sinh t. tanh (ti 2
2A - D sinh 2 ti tanh a
(~1
-
t.) t.)
=
-po -pi
(~i
- t.)
+ cos 11]
,\n expression for the maximum valuc of this has already been given on page 60. A general series form of stress function in bipolar coordinates \Vas givcn by G. B. Jeffery. 2 Its equivalent complcx potentials are easily found, and involve the functions considered here together \vith the bipolar analogues of the simple functions quoted in the Problems on page 197, when dislocations and concentrated forces are included. It has been applied to the problems of a semi-infinitc plate \vith a concentrated force at any point, 3 a semi-infinite region -..vith a circular bole, under tcnsion parallel to the straight edge or plane boundary,4 and undcr its o'\\-·n "'·eight, 5 and to the infinite plate with two holes,6 ora bole formed b:r tv•o intersecting circles. 7 Soiutions have been given for the circular disk subject to concentrated forces at any point, 8 to its own \veight when suspended at a point, 9 or in rotation about an eccentric axis, 10 with and "'·ithout 1 i the use of bipolar coordinates, and for the eiTect of a circular hole in a semiinfinite plate with a concentrated force on the straight edge.12 1
An exhaustive discussion of the maximum value is given by Cokcr and Filon,
loc. Cil,
• Loc. cit.
E. ~felan, Z. angew. ,ilfath.•ilfech., vol. 5, p. 314, 1925. 'Sce p. 82. 'R. D. ~Iindlin, Proc. A.S.C.E., p. Gl9, 1939. • T. Põschl, Z. angrno. Math. Jlech., voL 1, p. 174, 1921, and vol. 2, p. 187, 1922. Also C. \Vcber, ibid., vol. 2, p. 267, 1922; E. 'Veinel, ibid., voL 17, p. 276, 1937; Chih BingLing, J. Applied.Phys., vol.19, p. 77, 1948. 7 Chih Bing Ling, ibid., p. 405, 1948. 8 R. D. ~:lindlin, J. Applied ,\.fechanics (Trans. A.S.M.E.), voL 4, p. A-115, 1937. • R. D. Mindlin, J. Applied Physics, vol. 9, p. 714, 1938. '"R. D. Mindlin, Phil. JJag., series 7, vol. 26, p. 713, 1938. 11 B. Sen, Buli. Cale ui/a Jf alh. Soe., vol. 36, pp. 58 nnd 83, 1944. u A. Barjansky, Quart. Applied ,lfath., vol. 2, p. 16, 1944. 3
'l'he normal stress 0-1; can be found by subtracting thc real part of Eq. (g) from Eq. (f) and similarly for the othcr pairs. On the boundary t = t. it is to take the valuc - p., and on the boundary t = ti the value - pi. Using the values of B and C given by Eqs. (p) these conditions lead to the two equations
+ sinh 2 t.)- 1(cosh ti - cos lJ)[sinh ti coth
1
2B =D cosh (ti+ to), cosh (t1 - to)
+ + +
1 p,, sinh 2 ti pi sinh 2 t,, sinh 2 ~i sinh 2 ~. (p. - p1) coth (~1 - t.) -a sinh 2 ti sinh 2 ~.
- 2
(k)
x(z) = aDt
211
'l'HEORY OF ELAS1'ICITY
212
Other Curvilinear Coordinates. z
yielding x y
= (e< = (e<
=
er
The cquation
+ abe-t + ac e- i 3
3
+ abe-<) cos 7J + ac e-•< cos 317 3
- abc-<) sin
71 -
ac c- 3< sin 311
"''here a, b, e are constants, gives a family of cui:ves ~ ~ constant '"h~ch can be made to include various oval shapcs, 1nclud1ng a square with rounded corners. The effect of a bole of such shape in a plat: under t.ension has becn evaluated (b~y means of the real stress funct10.n) by M. Greenspan.1 B~y means of a generalization of these .coord1nates A. E. Green2 has obtained solutions for a triangular bole w1~h roundcd comers, and, by means of another coordinate transformat1on, for an exactly rectangular bole. ln the latter case thc perfcctly sharp corners introducc infinite stress conccntration. The curvilinear coordinatcs given by
+ ia 1cii + ia2e'~r + · · · + ia,.c'"f to a bcing real constants ' have been applicd a ... ,,. a1,., . by C. Weberl f z = ~
the semi-infinite plate -..vith a serrated boundary, 3 as 1n the cxamp e o evenly spaced semicircular notchcs \vhich is -..vorke~ out. "\\rhen the distance betwecn notch ccnters is t'vice the notch d1amcter, thc s~ress concentration, for tension, is found to be 2.13. Thc value for a s1ngle notch is 3.07 (see page 89). A method for determining thc complex potcntials from the boundary conditions, '\\>·ithout the neccssity of guessing their form in advance, has been dcvelopcd by N. 1Iuscheli!lvili. ·~ 4
1
Quart. Applied J[ath., vol. 2, p. 60, 101-l.
See a.lso 'il, !llorkovin, ibid., p. 350,
1945. • Proc. R01J. Soe. (London), series A, vol. 184, p. 231, 1945. • e. '\\'eber, angew. ,\.fath. i'[ech., vol. 22, p. 29, 1942. • Math. 1 1nn., vol..107, pp. 282-312, 1932. Also Z. angew. ,\.falh. ~ech., vol.13, 264 1933 An account of this mP-thod is given by I. S. Sokolnikofi, Lectures ~~ th~ 'fhco~y of Elasticity, Bro\Vn Univcrsity, 194~ (mimeographed notes). ~ It is used by !-.Iorkovin (sec footnote 1). !\Iost of the work of ~- 1. !\Iu~che~~ iilvili and his associates is in Ilussian. His book "Singular Integra.! Equat1?~ 8 (2d ed.), whieb contains solutions of severa\ two-dimensional probl1:1ns of clast1city, in pa.rticular 1nixed boundury valne problcms, has becn translated by the Aeronautical Research Laboratories, Dept. of Supply and Dcvelopme~t, Co_mmonwealth of Australia (Translation No. 12, 1949). Another tr_anslatio~ ed1ted by J. R. 11. Radok voas published hy P. Xoordhoff, N.V.,_ Groningen, Netherlands, 1953. :\IuscheliShvili gives his mcthods in ''Some Bas1c Problems of the :\íathema.tical Theory of F.lasticity," 3d ed., :\{o~co"·, l\}19, translated by J. R. !\L Radok, ublished by P. Xoordhoff, Groningt:n, :-.'ctherlands, 1\)53. Results of th~ ty~ !otcd under footnotes 1 and 2 were obtained eo.rlier by Russio.n authors c1ted U1
z.
the later book.
CHAPTER 8
3
ANALYSIS OF STRESS AND STRAIN 1N THREE DIMENSIONS 67. Specifi.cation of Stress at a Point. Our previous discussions wcre limited to t"'·o-dimensional problems. Let us consider now the general case of stress distribution in thrce dimensions. It was sho,vn (sce Art. 4) that the stresses acting on the six sides of a cubic element can be described by six components of stress, namely the three normal stresses
,, 0
a small distance from O, so that L~~<;z;1~,:"'q-:'.'~'>~>,,<:-:;..:-;.=-~-y this latter plane together with the _,..-Ty~----< coordinate planes cuts out from B _,. X the body a ·very small tetrahedron OiBCDO. Since the stresses vary F1a. t::i2. continuously over the volume of the body, the stress acting on the plane BCD will approach the stress on the parallel plane through O as the element is made infinitesimal. In considering the conditions of equilibrium of the elemental tetrahedron the body forces can be neglcctcd (see page 4). Also as the element is very small we can neglect the variation of the stresses over the sides and assume that the stresses are uniformly distributed. The forces acting on the tetrahcdron can thercforc be detcrmincd by multiplying the stress components by the areas of the faces. If A denotes the area of the face BCD of the tetrahedron, then the areas of the three other faces are obtained by projecting A on the three coordinate planes. If N is the normal to the plane BCD, and we ''1rite cos (Nx) = l,
cos (Ny) = m, 213
cos (Nz) = n
(a)
i:
THEORY
214
()F
ELASTIC/TY
ANALYSJS OF S1'RESS Al'VD STRAIN
the areas of the three othcr faces of the tetrahedron are
Am,
Al,
Substituting
An
.r,. =
z
If ..,.,.e denote by X, Y, the three components of stress, parallel to the coordinate axes, acting on the inclined face BCD, then the co_m~onent f force acting on the face BCD in the ilirection of the x-ax1s is AX. ~so the components of forces in the x-direction acting on the three other faces of the tetrahedron are -Aúr,,, -Amrxy, -A:ir:u. The corresponding equation of cquilibrium of the tetrahedron is AX - Alu,, - AmT"ll - Anr,,, =O
In the sarne manner t\vo other equations of equilibrium a~e obtained by projccting the forces on the y- and z-axes. Aftcr cancehng th~ factor .1, these equations of equilibrium of thc tetrahedron can be \\11tten X
=
Y
= ,,,.,,[
u,,l
Z
=
+ -rcym + r.,,n + .r11m + ,,,,11n
T,,.l
+
1' 11
,m
(112)
+ u.n
Thus the components of stress on any plane, defined by the ~irection cosines l m n can easily be calculatcd from Eqs. (112), prov1ded the six com~on~nt~ of stress .r,,, u 11 , u,, ,,,.,,, 1'11,, ,,,'" at thc point O are known. 68. Principal Stresses. Let us consider now the normal compo~ent of stress u,, acting on the plane BCD (Fig. 132). Using the notat1ons (a) for the direction cosines we find
u,. = Xl
+ Ym + Zn
or substituting t.he values of X, Y, Z from Eqs. (112),
'
.r,. = .r,]i
+ u,,m2 + u,n2 + 2r ,mn + 2r~,ln + 2r.,,lm 11
The coordinates of the end of this =
mr,
z
=
nr
. (<)
(d)
r
(114)
As the plane BCD rotates about the point O, the end of the vcctor r always lics on the surlace of the second degree given by Eq. (114). This surlace is complctely defined by the stress condition at thc point O, and, if the directions of the eoordinate axes x, y, z are changed, the ::;urlace -..vil\ remain entirely unchanged and only the components of ::;tress u,,, u,,, u,, ru,, r,,,, r.,,, and hcnce the coefficients in Eq. (114), will alter. It is '\\"ell knovrn that in the case of a surlace of the second dcgree, snch as given by Eq. (114), it is al\vays possible to find for the axes x, y, z 1>uch directions that the terms in this cquation containing the products of coordinatcs vanish. This means that \Ve can always find thrce perpendicular planes for \\-"hich r11,, r,~, r.,, vanish, i.e., the resultant ::;tresses are perpendicular to the planes on '\\'hich they act. We call these stresses the principal stresses at the point, their directions the principal axes, and thc planes on 'vhich they act principal planes. It can be i:ieen that the stress at a point is completely defined if the directions of the principal axes and the magnitudes of the three principal stresses are given. 69. Stress Ellipsoid and Stress-director Surface. If the coordinate axes x, y, z are taken in the directions of the principal axes, calr:ulation of the stress on any inclined plane becomes very sim pie. Thc shearing stresses r 11,, 1',:c, r.,, are zero in this case, and Eqs. (112) become ..Y
=
.r,J,
y
= .r,,m,
Z
= (1',n
(115)
Putting the values of l, m, n from these equations into the 'vell-known relation l2 + m 2 + n2 = 1, -..ve find
x2 +
1
(116)
The plus-or-minus sign in Eq. (d) applies according as"~ is terurlle or compres~ eive, and correepondingly in Eq. (114). '\\'hcn ali three principal stre8.'les have the sarne sign, only one of the alternative signs is needed, and the ~urface is an ellipsoid. When the principal streeses are not all of the sarne sign, both signs llJ'e Deeded and the surface, now represented by both Eqs. (114), consists of a hyperboioid of two sheets, together with a hyperboloid of one sheet, with a <:ommon. aaymptotic cone, 1
(b)
± !k'
from (b), and the values of l, m, n from (e) in Eq. (113), we find 1
(113)
The variation of .rn v.'ith the direct.ion of the normal N can be representcd geometrically as follo\vs. Let. us put in the direction of N a vector whose }cngth, r, is invcrsel3r proportional to the square root of the absolute value of the stress u,,, i.e.,
in vrhich k is a constant factor. vector will be y X= [r,
215
216
THEORY OF ELASTICITY
ANALYSIS OF STRESS A/\'D S1'RAIJ.'
This means that, if for each inclined plane through a point O the stress is represcnted by a vector from O "\\'i.th the components X, Y, Z, the ends of all such vectors lie on the surface of the ellipsoid given by Eq. (116). Tbis ellipsoid is called the stress ellipsoid. !ts semiaxes give the principal stresses at the point. From this it can be concluded that the maximum stress at any point is the largest of the three principal stresses at this point. If t\VO of the three principal stresses are numerically equal the stress ellipsoid becomes an ellipsoid of revolution. If these numerically equal principal stresses are of the sarne sign the resultant stresses on all planes through the axis of symmetry of the ellipsoid v.·ill be equal and perpendicular to the planes on which they act. ln this case the stresses on any t\VO perpendicular planes through this axis can be considered as principal stresses. If all three principal stresses are equal and of the sarne sign, the stress ellipsoid becomes a sphere and any three perpendicular directions can be taken as principal axes. When one of the principal stresses is zero, the stress ellipsoid reduces to the area of an ellipse and the vectors representing the stresses on all the planes through the point lie in the sarne plane. This condition of stress is called plane stress and has already been discussed in previous sections. "\Vhen t'vo principal stresses are zero \Ve have the cases of sirnple tension or cornpression. Each radius vector of the stress ellipsoid represents, to a certain scale, the stress on one of thc planes through the center of the cllipsoid. To find this plane ·we use, togethcr with the stress ellipsoid (116), the stress direcWr surface defincd by the equation x1 y~ it (117) -+-+-=1 CT~
cr,
U;
The stress represented by a radius vector of the stress ellipsoid acts on the plane parallel to the tangent plane to the stress-director surface at the point of its intersection with the radius vector. This can be shown as follows. The equation of the tangent plane to the stress-director surface (117) at any point Xo, yo, Zo is ~+YY•+ZZo=l
"•
u,,
u,
(a)
Denoting by h the length of the perpendicular from the origin of coordinates to the above tangent plane, and by l, m, n the dircction cosines of this perpendicular, the equation of this tangent plane can be -..vrittrn in the form
lx+my+nz=h
(b)
Com paring (a) and (b) we find "~ =
x,h
T'
217
Substituting these valucs in Eqs. (115) we find X= zoh,
y = yoh,
Z
zoh
=
i.e., the components of stress on thc plane with direction cosines l m t' lt th ___,. ' ,narepropor 1ona o e coo.1u1nates Xo, Yo, z 0• Hcnce the vector representing thc stress gocs through the po1nt xo, y 0, zo, as \VaB stated above.'
70. Determination of the Principal Stresses. If the stress comP?nen.ts for thrce coordinate planes are kno\vn, ,ve can determine the d1rect1ons and m~gn.itudes of the principal stresses by using the propert! that the principal stresses are perpendicular to the planes on 'vh1ch thcy act. ~et l, m, n be the direction cosincs of a principal plane and S the magnitude of the principal stress acting on this plane Then the components of this stress are: . X~
Sl,
y
= ~'\m,
Z
=
Sn
Substituting in Eqs. (112) Vi'e find (S -1'rYl
u,)l -
+ (S
-r,,,l -
1'.rvm -
1',,,n =
O
- u 11 )m - 1'11 ,n =O
r 11 ,m
+ (S -
u,)n =
(a)
O
These are three homogeneous linear equations in l m n Th ·11 · 1 · diff ' ' · eym g1ve ~o ut~ons erent from zero only if the determinant of these equat1on~ is zero. Calculating this determinant and putting it equal to zero g1ves us the follo"'·ing cubic equation in S:
S
3 -
(u~
+ u,, + u,)S + (u~ull + UyU, + U;rfY 2
-
(U#:vfT,
+ 2ry,T,,1'Xll -
U,1'11 ,
2
-
z
-
T 2 Y•
u,,r,,, 2 -
-
2
u,1'"112)
2\S
-1'Xll/
=O
1'he three roots of ~his .equation give the values of the three principal stres1>es .. By subst1tut1ng each of these stresses in Eqs. (a) and using 2 the. relat1on l + m 2 n 2 = 1, we can find thrce sets of direction cos1nes for the three principal planes.
+
ln~!
may be notcd t~at ~q. {118) for determining the principal strCJ!ses must he pendcnt oi thc d1rect1ons of the coordinates x y z· h~noe the f to · thc~i · th' · ' • • " ac rs 1n parends. in is equation should remain constant for any change of directions of coor inates. Hencc the coeflicienta of Eq. (118) (a) cr, +u, +u,
..
..
+ + "·"· - ,.._,i . - .- ' - -, ' + - u,.-,.2 - cr"r,.• - u..-,,•
(b) """• u,,u, (e) cr.cr,,u. 2.-,,.-~,.-,,
1 Another method of r tº h . developed by O M h ~f~es~n ~ng t e stress ata po1nt, by using circle~, has bccn A. F" · .. º r, ec I11Sche JV1eschn.nik," 2d ed., p. 192, 1914. See also opp1 and L. Foppl, "Drang und Z\';an " l 1 d g, vo · 'p. 9, an H. r-.1. Vi·estergaard Z. angew M th Jl,f h 1 4 5 made i ~r ª : ec "".º · , .P· 20, 1924. Applications of híohr's circles we~ n 1scusa1ng two-d1mens1onal problems (scc Art. 9). T
(,)
(118)
218
THEORY OF ELASTICITY
ANALYSIS OF-STRESS AND STRAli\'
do not vary with changing directions of the ooordinates. This means that the sum u~ +
71. Determination of the Maximum Shearing Stress. Let x, y, z be the principal axcs so that t:r~, u11 , d, are principal stresses, and let l, m, n be the direction ensines for a given plane. Then, from Eqs. (115), the square of thC' total stress on this plane is
'fhe square of the normal component of thc stresR on the sarne plane is, from Eq. (113), (a) Then the square of the shearing stress on thc sarne plane must be
DIRECTION Cos!NEB FOR
l -
o
m-
o
n-
±1
o -±1 -o
p
"
219
N EB OJ;' Tma%. AYD Tm;•.
o
±1
±
v1
±
-v1
±
v1 o
o
±
v1
o
o
±VI
±VI
The fi~st .t~ree columns give the directions of t-he planes of coordinates, co1nc1d1ng, as was assumed originally, with the principal planes F~r.these planes the shearing stress is zero, i.e., expression (b) is~ ~ 1 ~~mu~. . The three remaining columns give planes through each º. al e principal the angles bet"•een the t wo other pr1n. S b axes · bisecting · c1p ax_es. u st1tut1ng the direction cosines of thesc three planes into ehxpresshion (b) we find the follo"\\·ing values of the shearing stresses on t ese t ree planes: r
=
±k(u11
-
u,),
r =-+ )' T i q.,( -q,
7 =
±y'(q_,
-
q 11 )
(119)
!his sho"·s that the maximum shearing stress acts on the plane bisectWe shall now eliminate one of the direction cosines, say n, from this equation by using the relation li+ mz
+ n2
=
1
and then determine l and m so :i.s to make 1" a maximum. After substituting n 2 = 1 - 12 - m 2 in expression (b), calculating its dcrivatives with respect to l and m, and equating these dcrivativcs to zero, \Ve obtain the following equations for determining the direction cosines of the planes for which ris a maximum or minimum: l[(u,, - u,)l 2 m[(u,, - u,)l 2
+ (u
11 -
+ (u
1 -
q,)m 2 u,)m 2
-
j-(ux - u,)] =O i(u11 - u,)] =O
(e)
Onesolutionoftheseequationsisobtainedbyputtingl = m =O. We can alBo obtain solutions different from zero. Taking, for instance, l = O, we find from the second of Eqs. (e) that m = ±VI; and taking m = O, we find from the first of Eqs. (e) that l = ± ..y}. There are in general no solutions of Eqs. (e) in which l and m are both different from zero, for in this case the expressions in brackets cannot both vanish. Repeating the above calculations by eliminating from expression (b) m and then l, \ve finally arrive at the following table of direction cosines making 1" a maximum or minimum:
~ng the angle between the largest and the smallest principal stresses and
is equa1 to half the difference betwcen these t"'º principal
stress~~.
If the x-, y-, z-a.xes in Fig 132 t th d. · if OB = OC = OD so that th represen e rrect1ons of principal stress, and direction cosines l ' - m = n ~ ~ormal N to thc slant face of t~c tetrahcdron has Eq. (a), or (113), as /v'3, t-he normal stress on th1s face is given by (d)
This is called thc "mean stress."
The shear stress on thc face is given hy Rq. (b) as
TI = j-(,,-,1 _ "•'
+ .r,')
_ j-(.r,
+ "• + ,,-,)!
+ (o,
- .r,)•
+ (o-,
_ ,,-,)1]
+ (
_
+ (
-
This can also be written T!
= j-[(u,, -
.r,)•
and also, by using (d), as TI = j-[(.r, _
This shear stres.s is callcd the " t h d 1 it act.s is one face of a. regular º~e: ~ ~a she~ stress;" beca use the face on v.·hich frequently in the theory of plastici:y.e ron w1th verbces on thc axes. lt occurs
72. Homogeneous Def ti tions, such as o . o~ º-?· We consider only small deformaments of the pa~~c~:s ~f :n:~~eerinJ ~trud ctu.res. The small displaceorme 0 Y \\'lll usually be resolved into
il,,
••
THEORY OF ELASTICITY
ANALYSIS OF STRESS AND STRAI!of
components u, v, w parallel to the coordinate axes x, y, z, respectiv:e!Y· It will be assumed that these components are ·yery small quant1t1es varying continuously over the volume of the bo~y. . Consider, as an example, simple tension of a pr1smat1cal bar ~edat the upper end (Fig. 133). Let e be the unit elongation of the bar in the x-direction and ve the unit lateral contraction. Then the components of displacement of a point with coordinates x, y, z are
simple tension.. Planes and straight lincs remain plane and straight after deformat1on. Parallel planes and parallel straight lincs remain parallel after deformation. A sphere becomes after deformation an cllipsoid. Thls kind of deformation is called ~mogeneous defvrmaÍion. I~ '\\il~ be.shown later that in this case the deformation in any given d1rect1on is the sarne at ali points of the deformed body. Thus two geometrically similar and similarly oriented elements of a body remain gcometrically similar after distortion. ln more general cases the deformation varies over the volume of a deformed body. For instance, when a beam is bent, the elongations and contractions of longitudinal fibers depend on thcir distances frorn the neutral surface; thc shearing strain in elements of a twisted circular shaft is proportional to their distances z from the axis of the shaft. ln such cases of nonhomogeneous deformation an analysis of the strain in the o, neighborhood of a point is necessary. 73. Strain at a Point. ln discllilsing strain in the neighborhood of a point O of a deformed body (Fig. x 134), let llil consider a small linear Fin. 134. element 001 of length r, ""ith the direction cosines l, m, n. The small projections of this element on the coordinate axes are
220
W
U =EX,
1 1 1 1 1
1 1 1 1 1 1
= -VEZ
Denoting by x', y', z' the coordinates of the point aíter deformation, x' = x + u = x(l +E), y' = y +V = y(l - ve), z' = z w = z(l - ve) (a)
+
If '\VC consider a plane in the bar before deformation such as that givcn by thc equation
1
(b)
ax+by+cz+d=O
the points of this plane \Vill still be in a plane after deforrnation. The equation of this new plane is obtained by substituting in Eq. (b) the values of x, y, z from Eq. (~). It can easily be proved in this manner that parallel planes rema1n parallcl after deformation and parallel lines remain parallel. If '"e considera spherical surface in the bar before deformation such as given by the equation (e) x2+y2+z2=r2 FlG. 133.
this sphere becomes an ellipsoid aftcr deformation, the equation of v:hich can be found by substituting in Eq. (e) the expressions for x, y, z obtained frorn Eqs. (a). This gives x'2 r 2 (1 e)2
+
+
y'2 z'2 r2(1 - ve) 2 + r 2 (1 - ve) 2
~
1
(d)
Thus a sphere of radius r defonns into an ellipsoid with semiaxes r(l e), r(l - ve), r(l - ve). The simple extension, and lateral contraction, considered above, represent only a particular case of a more general type of deformation in which the components of displacement, u, v, w, are linear functions of the coordinates. Proceeding as before, it can be shown that this type of deformation has all the properties found above for the CHise of
+
lix = rl,
8y = rm,
liz = rn
221
(a)
Thcy represent the coordinates of the point 0 1 with respect to the x-, y-'. z-axes through O as an origin. If u, v, w are the components of the dtsplac:men: of the point O during deformation of the body, the correspond1ng d1splacements of the neighboring point 0 1 can be represented as follows: Ui =
u
vX
av
av
+ ÔX 8x + ây
av
+ ÔZ 8z aw aw aw W + ih: 8x + Ô'y 8y + 7fZ 8z
Vi = V W1 =
au + .au , - - 8x + -au 8y + - 8z ây ÔZ 8y
(b)
!!r~ ~ed ~ere that the quantities Bx, 8y, 8z are small, and hence the
8 '\Ytt~ h1gher po'\\·ers and products of these quantíties can be 1 ::~ ected. in (b) as small quantities of higher order. The coordinates he po1nt 01 become, after deformation
'
222
A1VALYSIS OF STRESS Ail/D STRAJN
THEORY OF ELASTTCITY
au
au
au
av
av
av
aw
aw
aw
notations
b+u,-u-b+MM+~•+.•
uw az
•+~-·-•+k·+~•+.•
(e)
liz+wi-w= õz+ axóx+ (}ylly+ az óz
+ u)i =
(ax + :~ õx + :; ôy + ~~ óz) + (liy
+:
óx
2
+ :; liy + :~ õz)
au+&ay ax -
2
+ (lJz + ~~ lix + ~~ IJy + ~~ lJz)
2
or, dividing by r 2 and using Eqs. (a),
+ t)2
= [
l(1
1n, =
2
+(z~~+m~~+n(1+~~)]
E
= l~
2
(d)
+
2 aw au ao ÔX + m ()y + n ()z + lm 2
(ª" + ª') + (ª" + aw) + mn(º' + aw) iJy
[n
ÔX
()z
ÔZ
ÕX
::• . . . (:;
+ :;;), ...
+ E,,m + E,n2 + -y"'lllm + -y,,.ln + -y,,.mn 2
ôy
are
known.
óy+111-11
r(l
=
+ •) · =
l
(
1 - •
011
l ôx
'") ôu ++ m+ nôa ÔZ ôyôz
+m(1
- •
'') '' + Oy + n ;}Z
(j)
iJw)
Taking another element r' through the sarne point with direction cosines l' m' n' the magn1tudes . ' , of thcse cosincs, after deformation, are given by equations' analogous to (J). The cosine of the angle between thc two elements after deformation is cos (rr') = li/,'
(120)
Using. the
+ m mi' + n 1
1
n 1'
Considüring thc t:'longations • and •' in these two directions as s1nall quantitie~ and using Eqs. lj), we find cos (rr') "" (ll'
Hence the elongation of an element r can be calculated provided the expressions
2
óz+w 1 -w Ow ôw ( n,= r(l+•) =lax+may+n l-•+az
Remembering that E and the derivatives âu/ax . .. âw/az are small quantities -..vhose squares and products can be neglected, and using i2 m2 n2 = 1, Eq. (d) becomes
+
= EJ
l • - ilz+u,-u (l + • ) T
(1 + !~) + n !~]
7 "'
ln investigating 1:1train around a point O it is necessary sometimes to knov• the change in the anglc between two linear elcments through the point. Using Eqs. (e) an
+ ~i) + m :; + n 3;;] + [z :; + m
"Y"'ll,
E,
(121) The physical meaning of such quantities as E% • • • , 1'~· . has already been discussed (sce Art. 5), and it '\Vas shown that E%, E,,, E, are unit elongations in the x-, y-, z-directions and "Y"'ll, 'Yn, 'Y,,, the three unit t1hcar strains related to the sarne directions. We now see that the elongation of any linear element through a point O can be calculated from :Eq. (121), provided '\•le kno\v the six strain components. ln the particular case of homogeneous deformation the components u, v, w of displaccment are linear functions of the coordinates, and from Eqs. (e) the components of strain are confltant over the volume of the body, i.e., in this case each element of the body undergoes the sarne strain. E
2
(1
=
av aw az+ay=
Eq. (120) can be presented in the form
It v..-ill be noticed that these coordinates are linear functions of the initial coordinates lix, liy, õz; hence the deformation in a very small element of a body at a point O can be considered as homogeneous (Arl. 72). Let us consider the elongation of the element r, dueto this deformation. The square of the length of this element after dcformation is equal to the sum of the squares of the coordinates (e). Hence, if tis the unit elongation of the element, we find (r
223
+ nnn' + nn')(l + "l'u•(mn'
- ' - •')
+ m'n)
+ 2(•.ll' + •umm' + ~.nn')
+ ,,~,(nl' + n'l) + "l'~u(lm' + l'm)
lf the directions of r and r' are perpendicular to each othcr, then ll' +mm'+ nn' =O
imd Eq. (122) gives the shearing strain between these directione.
(122)
224
ANALYSIS OF STRESS AND STRAIN
THEORY OF ELASTICITY
74. Principal Axes of Strain. From Eq. (121) a geometrical interpretation of the variation of strain ata point can be obtained. For this purpose let us put in the direction of each linear element such as r (Fig. 134) a radius vector of the length
R
~
-k-
(a)
vi;\
Then, proceeding as explained in Art. 68, it can be shovrn that the ends of all these radii are on the surface given by the equation (123)
'.fhe shape and orientation of this surface is completely determined by the state of strain at the point and is independent of the directions of coordinates. Jt is always possible to take such directions of orthogonal coordinates that the terrns with products of coordinates in Eq. (123) disappear, i.e., the shearing strains for such directions become ze_ro. These directions are called principal axes of strain, correspond1ng planes the pincipal planes oj strain, and the corresponding strains the principal strains. From the above discussion it is evident that the principal axes of strain remain perpendicular to each other after deformation anda rectangular parallelepiped with the sides parallel to the princip~l planes remains a rectangular parallelepiped after deformation. ln general it will have undergone a small rotation. .,, If the x-, y-, and z-axes are principal axes of strain, then Eq. (123) becomes ± k2 = Ep.t; 2 + E~y2 E,Z2
+ E, remains constant when the system of coordinates is rotated. Th!s sum has, as we know, a simple physical meaning; it representa the un1t volume expansion due to the strain at a point. 75. Rotation. ln general during the deformation of a body any element is changed in shape, translated and rotated. On acco~nt of the shear strain the edges do not rotatc by equal amounts and it is necessary to consider ho'v the rotation of the ""·hole elem:nt can be ~pecified. Any rectangular element could have been brought into its tina.! f?rm, position, and orientation in the follo\ving three steps, heg:tnn1ng \VIt.h the element in the undeformed body: 1. The strains E,,, Ei;, E,,, 'YZll, -y11,, 'Y,,, are applied to the element and the clement is so oriented that the directions of principal strain have not rotated. 2. 'fhe element. is translated until its center occupies its final position. 3. The element is rot.at.ed into its final orientation. 'fhe rotat.ion in st.cp 3 is evidently the rotation of the directions of principal strain, and is ~herefore independent of our choice of x-, y-, z-axes. It must be poss1ble to cvaluate it \Vhen t.he displacements u v w are given. On the other hand it is clearly independent of the str~i~ components. Since the translation of t.he element is of no interest to us here w.e may consider the displacement of a point 0 1, as in Art. 73 and Fig. Í34, relative to the point O, the cent.er of the element. This rela.tive displacement is given by Eqs. (b) of Art. 73 as E,, ' . Ei;
+
U1 -
ln this case the elongation of any linear element with the direction eosines l, m, n becomes, from Eq. (121),
V1 -
(124)
and the shcaring strain corresponding to two perpendicular directions r and r' becomes, from Eq. 122, 'Yrr'
= 2(E,Jl'
+ Ei;mm' + E,nn')
W1 -
au ÓX + -au fiy + -au ÓZ ax ay àz av av av - fix + - fiy + - fiz
U = -
v = W
àx àw = fix ax
ày aw fiy ày
+-
az aw •z az
(a)
+-
Introdncing the notation (e) of Art. 73 for the strain components and also thc notation 1 '
(125)
It can thus be seen that the strain ata point is completely determine~ if vte kno\V the directions of the principal axcs of strain and the magn1tudes of the principal cxtensions. The determination of t~e principal axes of strain and the principal extensions can be done in the. sarne manner as explained in Art. 70. It can also be shown that the sum
225
21
(ªw ª") =W:., ày-az
1
21
(ª" ª") =w, ax-ày
(126)
. A glancc at Fig. 6 will show that av/ax and -au/ay, occurring in the expreaaion · rotations · · clcmenta O' A' O' B' from their . . for w,, are th e e Jock wIBc of the line origina.! pos1"f1ons OA , OB . Thus ''" 1s · the average of thcse rotations ' and " • h ave · lhe yz- and :i:z-planes. "-- sign1 · -6 cance 1n , ~, a s1mi.uu ·•
226
THEORY OF ELASTICITY
ANALYSIS OF STRESS AND STRAilil
we can write Eqs. (a) in the form U1 -
W1 -
+ hey óy + hn 8z + tv 6y + h .., ôz hzo lix + f7.,, 5y + ~. ôz -
U = f,. ÔX
Vi - V = W =
hzv õx
w, 8y w., o5z w,, 8x
+ w., ôz + w, ÔX + w., liy
(b)
which express the relativedisplacement in t'vo parts, one depending only on the strain components, the other depending only on the quantitites
w,,, w,, "'··
We can no\v show that w,,, w11 , w, are in fact the components of the rotation 3. Consider the surface given by Eq. (123). The square of the radius in any direction is inversely proportional to the unit elongation of a linear element in that direction. Equation (123) is of the form F(x,y,z) = constant (e) If we consider a neighboring point x face, we bave the relation
+ dx, y + dy, z + dz on the sur(d)
The shiít dx, dy, dz is in a direction whose direction cosines are proportional to dx, dy, dz. The three quantities iJF/ax, aF/ay, aF/az also specify a direction, since we can take direction cosines proportional to them. The left-hand side of Eq. (d) is then proportional to the cosine of the angle between these t\\·o dircctions. Since it vanishes, the two directions are at right angles, and since dx, dy, dz representa direction in the tangent plane to the surface at the point x, y, z, the direction represented by iJF / ax, iJF / iJy, iJF / az is normal to the surface given by Eq. (e). Now F(x,y,z) is in our case the function on thc right-hand side of Eq. (123). Thu•
aF i)x
aF
= 2EzX
+ 'Yri1Y + 'Y.Z
+ 2~y + 'Y11.Z aF i)z = 'Y~ + 'Y11•Y + 2E.Z
iJy = 'Y~
(•)
The surface given by Eq. (123) being drawn with the point O (Fig. 134) as center, we may identify llx, lly, llz in Eqs. (b) with x, y, z in Eqs,
(e).
227
. W e consi~er now the special case when "'"• "'11• wz are zero. Then the r1ght.hand s1des of Eqs. (e) are the sarne as the right-hand sides of E . qs. (b) but for a factor 2. Consequently th dº 1 . e isp acement g1ven by Eqs (b) is normal to the surface givcn by Eq ( 123) C .d . h . · O (F. 13 ) . · · ons1 er1ng t e point i ~g. 4 as a po1nt on the surface, this means that the dis laceme of 011s nor~al to thi:' surface at 01. Hence if 001 is one of th: princip~~ axes of stra1n, t~at. IS, one of the principal axes of the surface, the displaccment of 0_1 is in the direction of 001, and therefore 001 does not rotatc. The d1splacement in question will correspond to st 1 ln arder. to complete the displacement we must restore t? Eq.s (b) the terms 1n w w11 E u t th ese t erros correspond to a small rigid· • ~. : "'•· body rotat1on having components "' b t h C "' Wy, W, a OU t e X- y- z-axes onsequently t.hcse quantities, given by (126) exprnes th ' t' ,. f. step3 tht" th · ' eroa1ono a IS, e ro~at.1on of the principal axes of strain at the point O. They are callcd s1mply the components of rotation. Problem 1. '\\'hat is thc equation of the type j(x y ) ·which hecomes 11. sphere ;,, y'' z'• _'.. ~~ of .-\rt. i2? What kind of surface ifi it? ª
+
+
;t"0't:fet hhomogoneous e snrfMe with centcr at O deformation
GElo/ERAL
THEORE~fS
229
The two other equations of equilibrium are obtained in the sarne manner. After dividing by ()x ôy ôz and proceeding to the limit by shrinking the element do,,.,·n to the point x, y, z, 've find
+ àrn + àr,,, +X ay az iJU~ + iJTX11 + iJT~z + y ày àx àz iJr;, + iJT,,~ + iJTyz + z
àr;,, ax
CHAPTER 9 GENERAL THEOREMS 76. Differential Equations of Equilibrium. ln t~e discussion of Art. 67 ,ve considered the stress ata point of an elast1c bod?': Let us
consider now the variation of the stress as we chan~~ th_e pos1t1on of the · t For this purpose the conditions of equ1hbr1um of a small ;;~:a~gular parallelepiped with the sides 8x, õy, fiz (~ig. 135) ~ust be studied. The componcnts of stresses acting on the s1des of this small element and their positive directions are indicated in the figure. Here we take into account the small z changcs of the components of
2!_::'.:~~~~~~~::i
'
stress due to the small increases Ox óy 8z of the coordinates. ' ' 1(Z".xz'1 ..,,..,, I Thus designating the mid-points -n'\' ; ; ; - ,1, 1 /-,3 -.-(0"..,3 "' of the sides of t he e1emen t bY 4 ôZ. !r.r~•·~~~- ~,..h 1, 2, 3, 4, 5, 6 as in Fig. 135, we ,.,.;.:::,, Y distinguish between the value ,,"' --d( of u,, at point 1, and its value at ôy point 2, 'vriting these (u".)1 and 135 1'1"· · (u,,) 2 respectively. The symbol u,, itself denotes, of course, the value of this stress component at t~e point z In calculating the forces acting on the element "'·e con&der the x, y, . ui . 1 . th t sides as very small, and the force is obtained by m t1p y1ng e s ress at the centroid of a side by the area of this side. . It should be noted that the body force acting on t~e el~ment~ which was neglected as a small quantity of higher order m d1scuss1ng. the equilibrium of a tctrahedron (Fig. 132), must now be taken into account, because it is of the sarne order of magnitude as the tei:ms ~ue to variations of the stress components, which "'·e are now co~tdenng. If "'"e let X y z denote the components of this force per un1t volume f the elem~nt' then the equation of equilibrium obtained by summing o ' d . . all the forces acting on the element in the x- irect1on is
S1!
l
1 r;.,.
r~Js
)'i'h
az
àx
=O = Ú
(127)
= Ú
ày
Equations (127) must bc:;;atisficd at ali points throughout the volume of t.hc body. The stresses vary OYer thc volume of the body, and lvhen '\'e arrive at the surlace they must be such as to be in equilibrium "'iLh the external forces on the surlace of the body. These conditions of equilibrium at the surlace can be obtaincd from Eqs. (112). Taking a tetrahedron OBCD (Fig. 132), so that the side BCD coincides ,vith thc surlace of the body, and denoting by X, Y, Z thc componentsofthe surlace forces per unit area at this point, Eqs. (112) become
+ Txym + r""n + + r;,n + T,,,/, + Tu,m
X y
=
u,,l
=
Uym
Z
=
Ty,n
TX)Jl
(128)
in ,vhich l, m, n are thc direction cosines of the externai normal to the surface of the body at the point under consideration. If the problem is to determine the st.ate of stress in a bod_y submit.tcd to the action of given forces it is necessary to solve Eqs. (127), and the solution must be such as to satisfy the boundary conditions (128). These equations, containing six components of stress, u,,, .•• , Ty,, are not sufficient for the detcrmination of these components. The problem is a stat.ically indeterminate one, and in order to obtain the solution \Ve must proceed as in the case of t"-·o-dimensional problems, i.e., the elastic deformations of the body must also be considered. 77. Conditions of Compatibility. It should be noted that the six components of strain at each point are completely determincd by the three functions u, v, w, representing thc components of displacement. fience the components of strain c:annot. be taken arbitrarily as functions of x, y, z but are subject to relations ''"hich follow from Eqs. (2) (see page 6). Thus, from Eqs. (2), â'v = ax 2 ay'
a2"Y,,,,, àx ày
l,.,
THEORY OF ELASTIGITY
230
GENERAL l'HEOREMS
Substituting these expressions in (e), we obtain
from which (a)
Two more relations of the sarne kind can be obtained by cyclical interchange of the letters x, y, z. Calculating the derivatives é)2E~
iJ'Yu =
ay
(1
+ v) (ªa2:: + :2~·) - '(!2?z + a20) Y iJy
ax ay az' a'lu + ~.
ih,,,
=
ay az
ay=
ax ay
(j2~
az2
ax 2
+
iJ2E,
az
iJx2
ax ay
ay2
a E, + a 2
= a2'Y:w' =
2 E,,
àz2
iJ2')'yz 1
ay az =
+ v) ày a2r,,, àz
(d)
iJq, àr= -az-ax-z
2
a 'Y··, ax az
2
Ô
2 E, =
ay az
~ ( - iJ'Yll• + Ô'Yzz + iJ'Y:
2 -iJ2E11 = !!.__ (ª'Yll• -
ax az Ô2E,
2 ax
ay
ay
=
ax
Ô')'.v;
ay
+ Ô')'cy) az
az ax
ay
a2E,, + a2t, az ay
a
_
1
+ v)o-
1
"(,,.
11 -
+ v)q, 2(1 + v}r z
E,= -p[(l
E
r,,,
=
2
u, _ à2u11
_ Õ
iJz 2
iJy az
i) ax
_
iJy 2
(àr,, Tz
=
0 20".,
iJy i)z
_
ax2
iJ 2u11 ày2
_
à 2u,
+ àr""') ay
iJZ
-
az -
àY iJy
ax
ay -
az
Substituting this in Eq. (d) and using to simplify the writing, the symbol '
2
a2
a2
0 v2=·-+-+ax2 ay 2 az2 '\ove find
(1
ª'ª) _,(v•e _ª'ª)
+ ,) (v•e - v••. _iJx
2
àx 2
(l
az) (')
+ •)(ªX_ aY _ az ax ay
;wo analogous .e~~atio!ls can be obtained fro1n thc t\VO other condi1ons ~f compat1b1hty of the type (e). Adding together all thrce equations of the type (e) \Ve find
116] (1 -
11
1 Proofs that these six equations are sufficient to ensure the existence of a displacement corrcsponding to a given set of functions ~,, . . . , 1'~ 11 • • , may be found in .~. E. H. Lovc, "Mathematical Theory of Elasticity," 4th ed., p. 49, and 1. S. Sokolnikoff, ":\lathe1nat.ical Theory of Elasticity," p. 24, 1946. '
y
+ ax _ aY az
i)z2
~ ve]
ax-
or, by using the first of Eqs. (127),
az
From Eqs. (3) and (4), using the notation (7), we find ~ = E[(l
Õ
ay az
2 -
2
2
2
(129)
(Õ')'11, + 1'.1-y,, _ Ô"frJJ)
.!!._
ay
Different~ating the first of these equations with respect to z and the sccond w1th respect to y, and adding them togethcr, V.'e find
2 2 iJ r,,,
These diffcrential relations 1 are called the conditions oj campatibility. By using Hooke's la'\'\' [Eqs. (3)] conditions (129) can be transformed into relations between the components of stress. Take, for instance, the condition 2
'Yv• =
ar'"'
ih,,, i)q!I -~---
Ô 2E~ = !!.__ ( - Õ'Y'll• + Ô"fr• + Ô')'r11) (b) 2 ay az ax ax õy iJz Two more relations of the kind (b) can be obtained by interchange of the letters x, y, z. We thus arrive at the following six differential relations betwcen the components of strain, ,vhich must be satisfied by virtue of Eqs. (2): ay 2
2(1
~he right si?~ o~ this equation can be transformed by using the equations of equ1hbr1um (127). From these equations we find
we find that
+ a2E"
=
2
Q3U
ay az
a2E~
231
,)v'B ~
-(l
Substºt · this expression for 1 uting Vtu.,
+ ,) v2e in
(ªx + aY + az) ax
ày
()z
Eq. (e),
+ _1_ a e2 _ _ -'-(ªx + av + az)- 2 ax I+11ax 1 - v àx ay az ax 2
(fl
1 233
GENERAL THEOREMS
THEORY OF ELASTICITY
232
We can obtain three equations of this kind, corresp~n~ing to the fir~t In the sarne manner the rema1n1ng thrce cond1th ree of E qs. (12") "· ·'d tions (129) can be transformed into equations of thc follow1ng k1n : (g)
Jf thcre are no body forces or if the body forces are constant, Eqs. (f)
The second derivatives for the two othcr components of displacement v and w can be obtained by cyclical interchange in Eqs. (a) of the letters x, y, z. Now u, v, w can be obtaincd by double integration of these second derivatives. The introduction of arbitrary constants of integration ,\·ill result in adding to the valucs of u, v, w linear functions in x, y, z, as itisevidentthatsuc:hfunctions can be addcd to u, v, w 'vithout affecting such equations as (a). To havc the strain components (2) unchanged by such an addition, the additional linear functions must have the form
and (g) beco me
(1
õ'e + v)V2cr~ + JX2-
=
O,
(1
(1
+ v)V 21T11 + a'e Jy2
=
0,
(1
(1
+ v)V cr, + õ'e Jz
=O,
2
2
u'=a+by-cz v'=d-bx+ez w'=f+cx-ey
à'8 + ,)V r + -ay az ~ O 2
'li•
à'8 = o + ,)v2.,.~· + __ ax az à'8 (1 + v)V z-v + -ax ay ~ O
(130)
2T
We sec that in addition to the equations of cquilibrium (127) and thc boundary conditions (128) the stress componcnts in an isotropic bo~y must satisfy the six conditions of compatibility (f) and (g) or.the s1x conditions (130). This system of equations is generally suffic1ent for determining the stress components '\vithout ambiguity (see ".'--rt. _82). The conditions of compatibility coniain only second denvat1ves of thc stress components. IIence, if the cxternal forces are such th~t- the equations of cquilibrium (127) together '\\'ith the bounda17 condit1ons (128) can be satisfied by taking the stre~s components et~her as constants or as linear functions of the coord1nates, the equat1ons of compatibility are satisfied identically and this stress system is the c~rrect solution of thc problem. Several examples of such problems will bc considercd in Chap. 10. 78. Determination of Displacements. When thc components ?f stress are found from the previous equations, the components of stra1n can be calculated by using IIooke's la"-· [Eqs. (3) and (6)]. Then ~qs. (2) are used for the determination of the displacemcnts ~· v, w. D1~er entiating Eqs. (2) \Vith respect to x, y, z wc can o_bta1n 18 equat~ons containing 18 second derivativcs of u, v, w, from '\'h1ch ~ll these der1vatives can bc determined. For u, for instance, we obta1n i.1 2u
i.IE~
i.lx 2
i.lx'
i.!2u
i.IE,,,
i.lx i.ly
i.ly'
(b)
This means that tho displacements are not entirely determincd hy the stresses and sLrains. On the displacements found from the diffcrcntial Eqs. (127), (128), (130) a
'"º
u,,, = }.e
+ 2G àu i.lx
(a)
and from (6} (b)
We
find (X+ G) i.le
i.lx
2 + G (i.1ax2 u +
2
i.1 u
ay2
2 + i.lz2 â u) +X
=O
The two other equations can be transformed in the sarne manner.
234
'
THEORY OF ELASTICITY
Then, using the symbol vi (see page 231), the equations of equilibrium (127) become (X+ G) ôe + GV 2 u +X= O
GENERAL THEOREMS
80. General Solution for the Displacements. I t is easily verified by substitution that the differential equations (132) of equilibrium in tcrms of displaccment are satisfied byl
ox
(X+G)ôe+av 2v+Y=O
ay
>1 - • (>o+ + +
U
(131)
X>1
a ôx
º'
W
=
•
+ X>1 + Yc/>2 + z
c/>a - a iJz (cf>o
\Vhere 4a = 1/(1 - v) and the four functions q, 0, c/>1, q, 2,
8
(X+G)ª +GV2u =0
ax
Differentiating these equations, the first \vith respect to x, t~e second with respect to y, and the third \vith respcct to z, and add1ng them together, we find · the volume expansion e satisfies the differential equation i.c., (133)
'fhe sarne conclusion holds also \vhcn body forces are constant throughout the volume of the body. Substituting froro such equationi,; as (a) and (b) into the houndary conditions (128) we find
+a(::l+ ::m + n)
V 2c/>o =
(132)
ª'
+:;m +
z
+ Xc/>1 + Y>2 + Zcf>3)
and, when there are no body forces,
X= Ael+G(::z !~n) :~ .. . . . . . . . . . . .. . . . . .. . . . . . . . . . . . . . . .
Y>2
•
V =
(X+G)ôe+GV 2w+Z=O
(>+G)º' +GV'v ~o ªY (X+G)ôe+GV 2w=O
235
(134)
Equations (131) together with the boundary conditions (134) define completely the three functions u, v, w. From these the components of strain are obtained from Eqs. (2) and the components of stress fro~ Eqs. (9) and (6). Applications of these equations will be shown m Chap. 15.
o,
v2q,1
=
o,
V2<1>2 =
o,
lt can be sho,vn that this solution is general, and that any one of the four functions may be dropped 'vithout loss of generality. This form of solution has been adapted to curvilinear coordinates by Neuber, and applied by him in the solution of problems of solids of revolution 2 gencrated by hyperbolas (the hyperbolic groovc on a cylinder) and cllipses (cavity in the form of an ellipsoid of rcvolution) transmitting tension, bending, torsion, or shear force transverse to the axis with accompanying bcnding. 81. The Principie of Superposition. The solution of a problem of a given elastic solid v.·ith given surface and body forces requires us to dLtermine stress components, or displacements, \Vhich satisfy the d1fferential equations and the boundary conditions. If we choose to v.·ork v.-·ith stress components 've have to satisfy: (a) the equations of equilibrium (127); (b) the compatibility conditions (129); (e) the boundary conditions (128). Let u,, . . . , -rxv . . . , be the stress components so determined, and duc to surface forces X, Y, Z, and body forces X, Y, Z. Let u,/ . . . , r"'71' ••• be thc stress components in the sarne elastic solid dueto surfacc forces X', :f', Z' and body forces X', Y', Z'. Thcn 1
This solution was given indcpendently by P. F. Papkovitch, Compt. rend., vol. 195, pp. 513 and 754, 1932, and by H. Keuber, Z. angew. Jl,fath. Mech., vol. 14, p. 203, 1934. Other general solutions wcre given by B. Galerkin, Compt. rend., voI. 190, p. 1047, 1930, and by Boussinesq and Kelvin-see Todhunter and Pear110n, "Hist-Ory of Ela.sticity," vol. 2, pt. 2, p. 268. Sce also R. D. Mindlin, Bull. Am. Math. Soe., 1936, p. 373. 1 H. Neuber, "Kerbspannungslehre." This book also contains solutions of two-Qimensional problems. See Chap. 7 above.
236
TIIEORl' OF ELASTICITY
+
GENERAL THEOREMS
the stress components u,, u/, . . . , T:11 + • • • , v.rill represent the stress dueto the surface forces X X', ... , and the body forces X+ X', . . . . 'fhis holds bccause all the differential equations and boundary conditions are linear. Thus adding the first of Eqs. (127) to the corresponding equation iJu/
ax
+
1 Tri1 ,
+ ÔT,,.,/ + iJT,,.' +X' ay
ª'
231
Then for the first solution \Ye have such equatioIIB as
au,,
1
ax
ay
X
~o
+ iJr""' +ar=' +X =
rr:/l
az
=O
+ r""'m + r,,,'n
and also the conditions of compatibility. For the second solution \Ve have and similarly from the first of (128) and its counterpart addition
"\\'8
have by
The compatibility conditions can be comhined in the sarne manner. The complete set of cquations sho,vs that "~ u/, . . . , Tcy Tzu' . . . , satisfy all the equations and conditions determining the stress duc to forces X+ X', . . , X+ X', . . . . This is the principle of superpottition. . ln dcriving our cquations of cquilibrium (127) and boundary condttions (128) v;e madc no distinction bet,veen the position and forn;i of the element bcfore loading, and its position and form after loadtng. As a consequence our equations, and the conclusions dra\vn from them, are valid only so Iong as the small displacements in the deformation do not affcct substantially the action of the external forces. There are cases ho\vcver, in \Vhich the deformation must be takcn into account. The~ the justification of the principle of superposition given above fails. Thc beam under simultaneous thrust and lateral load affords an cxample of this kind, and many others arise in considering the elastic stability of thin-\~·alled structurcs. 82. Uniqueness of Solution. We consider no\Y \Vhether our equations can have more than one solution corresponding to given surface and body forces. . . reprcsent a solut.ion for loads X ... 1 Let "~' . . . , T"" 1 .. , rx/' . . . represent a second solution for X •.. , and let 11,/ the same loads X . . 'X ..
+
+
and also the conditions of compatibility . By subtraction \Ve find that the stress distribution given by the diffcrcnces rr,/ - rrx'', . . . , T:ri1' - r,,/', satisfics the equations iJ(rr/ - rr,/
1 )
ax
O
+ ?~r""'
(rr,/ - rrx")l
- rxv")
~o
ay
+ (rxu
1
-
Txy
11
)m
+ (r,.,'
- r,,,'')n
in \vhich all externai forces vanish. The conditions of compatibility (129) \Vill also be satisficd by the corresponding strain componcnts E,/ -
11 E,, ,
•••
> '"f:o11' -
"/xy", . . . .
Thus this stress distribution is one \vhich corresponds to zero surface and body forces. The \York done by thcse forces during loading i.>: zero, and.it follo\vs that fffV 0 dx dy dz vanishes. But, as Eq. (85) sho..,·s, Vo is positive for all states of strain, and therefore the integral can vanish only if V 0 vanishes at all points of the body. This requires that each of the strain components t "'' - Er " , • • · > "'1>:11' ~ ,.,.1>:11" 1 • • should be zero. The two states of strain t/ . . , "1>:11' • • • , and
238
THEORY OF ELASTICITY
E/ 1 .. , 'Y:rv" , and consequently the t\vO states oi stl'eSS u/ . . . , r""' . . • , and u:r" . . . , r:rv" • . . , are thercfore identical. 'Ihat is, the equations can yicld only one solution corresponding to given loads.1 The proof of uniqueness of solution was bascd on the assumption that the strain energy, and hence stresses, in a body disappear when it is freed of external forces. Howevcr there are cases \Vhen initial stresses may exist in a body \Vhile externai forces are absent. An example of this kind ,~·as encountered in studying the circular ring (see ltrt. 39). lf a portion of the ring between two adjacent cross sections
is cut out, and the ends of the ring are joincd again by welding or other mcans, a ring with initial stresses is obtained. 2 Severa! examples of this kind wcre discusscd in considering two-dimensional problems. We can also have initial stresses in a simply connectcd body dueto some nonelastic deformations during the process of forroing the body. \Ve may have, for instance, considerablc initial stresses in large forgings due to nonuniform cooling and also in rolled metallic bars due to the plastic flow produced by cold v:ork. For determining these initial stresses the equations of elasticity are not sufficient, and additional information regarding the process of forming thc body is necessary. It should be notcd that in all cases in which thc principlc of superposition can be used the deformations and stresses produced by externai forces are not affected by initial stresses and can be calculated in exactly the sarne manner as if there wcre no initial stresses. Then the total stresses are obtained by superposing the stresses produced by externa} forces on the initial stresses. ln cases when the principle of supcrposition is not applicable, the stresses produced by external loads cannot be detcrmined -..vithout knowing the initial stresses. We cannot, for instance, calculate bending stresses produccd by lateral loads in a thin bar, if the bar has an initial axial tension or compression, -..vithout kno"ving the magnitude of this initial stress. 1 This thcorem is dueto G. Kirchhoff. See his Vorle:rungen über Math. Phys., Mechanik. • The ring represents the simplest example of multiply-connected bodies. ln the case of such bodies general equations of clasticity, cxpressed in terms of stress components, are not sufficicnt for determining stresses, and to get a complete solution an additional invcstigation of displacementa :is necessary. The :6rst investigations of this kind were made by J. H. Michell, Proc. London Math. Soe., vol. 31, p. 103, 1899. See also L. N. G. Filon, Brit. Assoe. Advancement Sei. Rept., 1921, p. 305, and V. Volterra, Sur l'équilibre des corps élastiques multiplement connexés, Ann. école norm., Paris, series 3, vol. 24, pp. 401-517, 1007. Furtber references on initial stresses are given in the papcr by P. Neményi, Z. angfl'W. Mat/i. Mech., vol. 11, p. 59, 1931. ·
GENERAL THEOREMS
239
83. The Reciproca! Theorem. Limiting ourselves to the t\.vodimensional case let us consider the plate under two different loading conditions, and denote by X,, Y 1, X 1, and Y1 the components of the boundary and the volume forces in the first case' and by X 2, Y2, X 2, • and Y2 1n the second case. For the displacements, thc strain componcnts, and stress components in the two cases we use the notation ,,,,,,d,,,,,, Ui, Vi, Ez, E~, 'Yr'll, u,,,
T
JX1u2ds
=
+ JY1t•2ds + JJX1u
2
dxdy
+ JJY 1v
2
dxdy
(a)
v.·here the first t\\'O integrals are cxtcnded around the entire boundary of the plate and the second t'"º ovcr the entire area of the plate. Substitutingfor X1 its expression from Eqs. (20), page 23, we can represcnt the first term on the right~hand side of Eq. (a) as follows:
JX1u2ds
=
fl-
+ Jmrr'll'u
2
ds
(b)
Proceeding now as explained on page 164 ,vo get
f
f
lux'u2 ds
=
mr"J011'u2 ds
=
f f ª;;' f f ª;;' U2
dx dy
U2
dx dy
Jf ~:2 + Jf ~;2 r~'
+
u,/ dx dy
dx dy
s_ubstituting this in (b) V.'e find that the first and thc third t.erms of (a} gi:ve us
JX,u,a, + JJ
X,u,Jxdy
~ JJ (~;' + ª;;' + x,)u,dxdy 2 + f f (~~ u,,'+~;2 riv')dxdy (e)
Similarly the second and the fourth tcrms give
f
Y1v2ds
+JJY1v2dxdy
=
f f (ª;;'+a;;'+ + f f (:~ + :: 2
1
2
y 1)v 2 dxdy
rxu') dx dy
(d)
Obser-v-ing now that the first terms on the right-hand side of equations ( e) and (d) va.nish in virtue of equilibrium equations (18), and sub-
240
=
=
JJ (~,/'u,,' + ~11({1 1 + E1
ff
1r.T,;" ""'+ r111"U11 '
241
and the elongation of the bar, produced by two forces P in Fig. 1300, is
stituting in Eq. (a), \\'e obtain
T
GENERAL THEOREMS
THEORY OF ELASTICITY
ô= vPh
AE
'Yrr1"r,,/) dx dy -
l!lly
" ' rf.,
vu,,''u,,'
Exactly the sarne result is obtained if v.'e calculate the -~vork donc by the forces of the second statc of stress on the displacements of the first state. Thus '\'\'e can conclude, comparing t\vo states of stress of an elastic body, that thc \Vork done by the forces of the first state on the corresponding displaccncnts of the second is equal to the \Vork done by the forces of the second state on the corrcsponding displacements of t.he first. This represcnts the reciproca! theorem. It can be easily extendcd also to bodies in motion or in vibration. It is only necessary to add (ai the inertia forces to the external loads. 1'he reciprocal theorem finds an important application in the theory of structures in Q~!b} the construction of influence lines. It also -thas useful applications in thc theory of Fio. 136. elasticity. Take as a simple example thc case of a prismatical bar compressed by two equal and opposite forces' P, Fig. 136a. The problcm of finding the stresses produced by these forces is a complicated one; but assume that "\Ve are interested not in the stresses but in the total elongation ô of the bar. This qucstion can be answered at once by using the theorem. For this purposc \Ve consider in addition to the given stress condition represented in Fig. 136a the simple central tension of thc bar shuwn in Fig. 136b. For this second case we find
and is independent of the shape of the cross section. As a second example let us calculatc the rcduction ~ in volume of an clastic body produced by t'.VO cqual and opposite forces P, Fig. 137a. As a sccond case of stress \VC take the sarne body submitted to the action of uniformly distributed presp surc p. ln this latter case "\Ve \Vill have at each point of the body a uniform compression in all directions of thc magnitude (1 - 2v)p/E [sec fb) Eq. (8), page 9J and the distance Fro. 137. l between the points of application A and R V>'ill be diminished by thc amount (l - 2v)pl/E. The reciproca} theorem applied to the t'"º stress conditionsl of Fig. 137 váll then give
p. (1 -;>)pl ~ ~p and the reduction in the volume of the body is therefore ~ ~
Pl(l - 2,)
E
Qh
84. ApproJCimate Character of the Plane Stress Solutions. It was pointed out on pagc 25 that th.i set of cquations ·wc found sufficient for plane stress problems undcr the assumptions madc (,,., = T,, = T"' = O,""'""'.-," independent of z) did ~ot cnsure satisfaction of ali the conditions of compatibility. Thcse assumptions lmply that '" •u, ••, 1'~• are independent of z, and that 1 .,, '"'are zero. The :first of the conditions of compatibility (129) '\ovas included in the plane stress theory, as Eq. (?1). It is easily verified that the other tive are satisfied only if., is a linear func~1on of x and y, v..J1ich is the exception ratbcr tban the rule in the plane stress soluttons obtained in Chaps. 3 to 7. Evidently thcse solutioru; cannot be exact, but We shall now see that tbey are closc approximations for thin plates. Let us seek exact solutions of the thrcc-din1ensional cquations for which2
We may suppose that the forces are distributed over a small area. to avoid singularities. Idcally conccntrated forces in two-dimen~ion_al problema usUtLlly result in infinite displaeement, indieating that the actual d1splacement. dependa on the distributio11
F?r. other applications of this kind see •.\..E. H. Lovc, "l\Iathcmatical Theory of El ast1c1ty," 4th ed., pp. 174-176, 1927. T •A. Ciebsch, "Elasticitii.t," Art. 39. See also ..\..E. li. Love, "l\lathPmatical heory of Elasticity," 4th ed., p. 145, 1927.
the lateral contraction, cqual to ô, tional area of the bar. equation
= v
11" vrhere A is the cross-sec-
Then thc reciproca! theorem gives us the
p. "AE = Qô 1
1
THEORY OF ELASTICITY
242
taking body force as zero. Sueh Holutions must satiRfy the equations of equilibrium (127) a.nd thc compstibility eondit.ions (130). Sinoe ,,,, .,.~,,'••are zero, the third, fourth, :uul fifth of Eqs. (130) (reading by columns) give
'-az (ªº) az - o' which rnean that respect to z,
-o
-~- ('.1~) az
iJy
oo/az is a const!i.nt.
a_ (~') - o ax az
'
\Yriting this k, 've l1ave, by int.egration with (")
S=kz+eo
\•1here fio i~ so far an arbitrary function of x and y. The tloird of Bqs. (127) is identieally satisfied, and the first two reduee to the two-dimC'nsionul forms c)q,
"• = ª'• iJxs'
ª'•
ª"
iJl
Similarly the second
ª' (''• ax• az• +1+' .. 0 •) =O,
but
v•e =o
(•)
=O
(d)
and thereíore, from (a) V1'00
vohere
il'o;b
V1 2o;b = kz
+ 0o
(e)
where o, is a function of x and y satisfying Eq. (d). the first of Eqs. (130) becomes
+ v)v• ª'4> + a•e. ay• ax•
But =
,, (
Integrating this equation twice with
+A +Bx +Cy +
(i)
whcre A, B, C are functions of z obtained by repeated integration of a, b, e, and q,1 , q, 0 are functions of x and y, a.s yet arbitrary. li we evaluate u., u~, Tz" from (i) by means of the formulas (b), the terms A
+Bx+Cy
make no difference. We may therefore set A, B, and C equal to zero, corresponding to taking a, b, e, zero in (h). lf we rei;trict ourselve.s to problems iu which the stress distribution is symmetrical about thc midd\e plane of tho platc, z = O, the term \{> 1z must ali:!O be zero. So also must k in Eq. (a). Then (i) reduces to 4> = o - -1 -' - e,zt 2 l ~
(J)
+
However \b and 0 0 are related hy (e) in v.·hich we can now take k substituting (j) in (e) and using (d), vre have
=o
O.
(k)
=
o
(j)
ª'•) ay• ª' (60 + ''•) az•
ay• "•'4> + az•
(h)
=a+bx+cy
Using (a) and thc first of (b),
=
where Eq. (e) has been used in the last step. .-\lso, on aecount of (d), vre can replace /iY.c' in Ul by -a•e0 /ay•. Then Ul becomes
Q 20 0
a•e, ª' (e, + ª'•) (1 + v) ay• àz' ---
ª"
~,.1
..
+ 1 +.,0 0
-21 1 +' "0,z'
Also, since u, is zero, and us and ""are given by the first t11.'o of Eqs. (b), we can 'vrite V1'4> = 0, and thercfore, using (a)
(1
ª' (ª'• +1 +' .. 0 •) = 0
ax ay az•
Thesc, with (g), show that all three second derivatives with respect to :t and y of the function (of x, y, and z) in brackets vanish. Thus this function 1nust be linear in x and y, and V.'e can v.·rite
\b = (b)
)
This equation may be used in plaee of the :first of (130). and Jast can be rcplaccd by
where a, b, ande are arbitrary functions-of z. respeet to z, V.'e find
which are satisfied, as before, by = _,
'
(g)
àz'
+ -ay- o , ax
'1s
(ª'•
ª' az• +1+ .. 0 • =O ay•
(JT,y
-
243
GENERAL THEOREMS
-·
and therefore, from (d), (1)
The remaining equations of (130) are satisficd on account of Eq. (a) and the vanishin.g of "" Tn, Tv•· ~Ve can now obtain a stress distribution by choosing a functi.on \bo of z and y wh1ch satisfios Eq. (!), fiuding 0 0 from Eq. (k), and \b from Eq. (j). Thc strc.'IBes are then found bythe formulas (b). Each will consist of two parts, the first dcrived from \bo in Eq. (j), the seeond from the term -
~ 1 ~.,
e,z•.
ln view of Eq. (1),
the first part is exactly like the plane stress components detennined in Chaps. 3 to
1,I
244
'l'HEORY OF l!.'LASTICITY
7. The second part, being proportional to z•, may be made as email as we plea.ee compared with the first by restricting ourselves to plates ·which are sufficiently thin. Hence the conclusk'!l. that our solutions in Chaps. 3 to 7, which do not satisfy ali the compatibility conditioilll, are nevertheless good approximations for thin plates. The "exact" solutions, represented by stress functions oi the form (j), will requfre that the stresses at thc boundary, as cLscwhere, havc a parabolic variation ovcr the thickness. However any change from this distribution, so iong as it does not alter the intensity of force per unit length of boundary curve, will only alter the stress in the immediate neighborhood of tbe edge, by Saint-Venant's principie (page 33). The type of solution considcrcd abovc will alwaya represent the actual stress, and the components "'''Tu, T,, will in fact bc llero, cxccpt close to the edges. Problems 1. Show that •• = •• =
'Y··
k(y' =
+ z'),
l'••
=
'Y••
=
k'xyz
o
where k, k' are small constants, is nota possible state of strain. 2. A solid is heated nonuniformly to temperature T, a function of x, y, and z. If it is snpposed that each element has unrcstraincd thcrinal cxpansion, the strain components v;ill be
•·=•.=•.=aT,
1'zu = 1'•• = 1'•• =
O
where a is thc const:i.nt cocfficient of thcrmal cxpansion. Prove that this can only occur when Tis a linear function of x, y, and z. (The stress and consequent further strain arising 'vhen Tis not linear are discussed in Chap. 14.) 3. A disk or cylinder of the shape shown in Fig. 137a is compressed by forces P at C and D, along CD, causing extension of AB. lt is then compressed by forces P along AB (Fig. 137a) causing extension of CD. Show that these extensions are equal. 4. ln the general solution of Art. 80 v•hat choice of the functions
CHAPTER 10 ELEMENTARY PROBLEMS OF ELASTICITY IN THREE DIMENSIONS 85. Uniform Stress. ln discussing the equations of equilibrium (127) and thc boundary conditions (128), it \Vas stated that the true solution of a problem must satisfy not only Eqs. (127) and (128) but also the compatibility conditions (sec Art. 77). These latter conditions contain, if no body forces are acting, or if the hody forces are constant, only second derivatives of the stress components. lf, therefore, Eqs. (127) and conditions (128) can be satisfied by taking the stress components cither as constants oras linear functions of the coordinates, Fru. 138. the compatibility conditions are sat-.. isfied identically and these stresses are the correct solution of the problem. AB a very simple example we may take tension of a prismatical bar in the axial direetion (Fig. 138). Body forces are neglected. The equations of equilibrium are satisfied by taking u,, = eonstant,
(a)
It is evident that boundary conditions (128) for the lateral surface of the bar, \vhich is free of externa} forces, are satisfied, bflcause all stress components, except cr"', are zero. The boundary conditions for the ends reduce to u,
=X
(b)
i.e., we have a uniform distribution of tensile stresses over cross sections of a prismatical bar if the tensile stresses are uniformly distributed over the ends. ln this case solution (a) satisfies Eqs. (127) and (128) and is the correct solution of the problem because the compatibility conditions (130) are identically satisfied. lf the tensile stresses are not uniformly distributed over the ends, solution (a) is no longer the eorrect solution beeause it does not satisfy the boundary conditions at the ends. Thc true solution becomes more complicated because the stresses on a cross section are no longer uni245
246
THEORY OF ELASTICITY
PROBLEMS OF ELASTICITY IN THREE DIMENSIONS
formly distributed. Examples of such nonuniform distribution occurred in the discussion of tv;o-dimensional problcms (see pages 51 and 167). As a second example consider the case of a uniform hydrostatic compression with no body forces. The equations of equilibrium (127) are satisfied by taking
The displacements u, v, w can now be found by integrating Eqs. (e), (d), and (e). Integration of Eq. (e) gives
fl~
=
Uu
=
-p,
T"11
=
Tn
=
71/Z
(e)
= Q
The ellipsoid of stress in thls case is a sphere. Any three perpendicular directions can be considcred as principal directions, and the stress on any plane is a normal compressive stress equal top. The surface conditions (128) will evidently be satisficd if the pressure p is uniformly distributed over the surface of the body. z 86. Stretching of a Prismatical Bar by Its Own Weight. lf p(l is the v.'eight per unit volume of the bar (Fig. 139), the body forces are
X= Y =O, l
_____ _,
z
=
(a)
-pg
The differential equations of equilibrium (127) are satisfied by putting
;y
u~ = CT11 = Tr11 = T11• = r,,,, =
o
(b)
}'10. 139.
i.e., by assuming that on each cross section we have a uniform tension produced by the \\·eight of the lo\\·er portion of the bar. It can easily be seen that the boundary conditions (128) at the lateral surface, ,vhich is free from forces, are satisfied. The boundary conditions give zero stresses for the lo'\\·er end of the bar, and, for the upper end, the uniformly distributed tensile stress u. = pgl, in which l is the length of the bar. The compatibility equations (130) are also satisfied by the solution (b), hcnce it is the correct solution of the problem for a uniform distribution of forces at the top. lt coincides with the solution which is usually given in elementary books on the strength of materials. Let us consider now the displacements (see Art. 78). From Hooke's law, using Eqs. (3) and (6), we find aw
(J,
pgz
(e)
f,=az=E=E E,,
=
Eu
=
au
ax
ilv
= iJy =
pgz
-v E
au av au aw av aw ~-~-Tu--+---+---+--o •ayaxazaxazay
(d)
(<)
w= pg'' 2E+wo
(f)
\\·here Wo is a function of x and y, to be determined later. (f) in the second and third of Eqs. (e), \Ve find
ilwº+ªu=O
ax
az
from "'·hich
aw,
2
ax
+ iluo = ax
2
Remembering that satisfied only if
Uo
_
11
and
ax
and y only.
2
_ z il wo
E
0
ay
x
pgz1
ay 2
Substituting expres-
+ ôv
0
=
ay
pgz
-vE
(h)
do not depend on z, Eqs. (h) can be
Vo
duo = ilv 0 =
(g)
'
in which Uo and Vo are functions of sions (g) into Eqs. (d), we find
-zil Wo
Substituting
'
u=-z-+u,
ax
247
(k)
'
Substituting expressions (g) for u and v into the first of Eqs. (e), we find
_ 2z d W~ ax ôy 2
and, since
Uo
and
Vo
+ ÔUo + ÔVo ay
= Ü
ilx
do not depend on z, \Ve must have il
2 Wu
iJx ay
=
o
(l)
'
From Eqs. (k) and (l) general expressions can now be written for the fu~ctions uo, vo, Wo. It is easy to sho\v that ali these equations are sat1sfied by Uo={)y+r,1
Vo=-{)x+,.. 1 'pg Wo = 2E (x2 + y2)
+ ax + (3y + 'Y
249
THEORY OF ELASTICITY
PROBLEMS OF ELASTICI1'Y /!1í THREE DIMENSIONS
in which a, {3, ")',ó, Ili, 'Y1 are arbitrary constants. Now, from Eqs. (f) and (g), the general expressions for the displacements are
parallel to the z-axis before deformation become inclined to this axis after deformation, and the form of the bar after deformation is as indicated in Fig. 139 by dotted lines. Cross sections of the bar perpendicular to the z-axis after deformation are curved to the surface of a paraboloid. Points on the cross section z = e, for instance, after deformation will be on the surface
248
(m) w = pgz2
2E
+ !:.EfJ. (x2 + y2) + ax + /3Y + 'Y 2E
'fhe six arbitrary constants must bc detcrmined from the conditions at the support. The support must be snch as to prevent any movement of the bar as a rigid body. To prevent a translatory motion of the bar, let us fix the ccntroid A of the upper end of the bar so that u = v = w = O for x = y = O and z = l. To eliminate rotation of the bar about axes through the point A, parallel to the x- and y-axcs, let us fix an element of the z-axis at A. Then au/az = i!v/i!z =O at that point. The possibility of rotation about the z-axis is eliminated by fixing an elemental area through A, parallel to the zx-plane. Then âv/iJx =O at the point A. Using Eqs. (m) the above six conditions at the point A beco me
-al
+ Ó1 =
O,
-fjl
O,
a=
+ 'Yl =o,
{3
=o,
pgl2+-y=0
2E i;
=o
z
=
e
+w
2
= e
'l'his surface is perpendicular to ali longitudinal fibers of the bar, these being inclined to the z-axis aJter deformation, so that thcre is no shearing strain ')'ey or 'Y:u· 87. Twist of Circular Shafts of Constant Cross Section. 1'he elcmentary theory of twist of circular shafts states that the shearing stress T at any point of the cross scction (Fig. 140) is A perpendicular to the radius r and proportional to the. ,--$--,-.x length r and to the angle of t\vist fl per unit lcngth of the shaft: (a) T = Ger where G is the modulus of rigidity. Resolving this stress into t'"º components parallel to the x- and y-axes, v.·e find X
T 11 ,
= G8T · ~ = Gflx
T:u
=
'
Hence 51
=o,
'Y1 =
O,
pgl2
- 2E
and the final expressions for the displacements are
---p;
vpgyz
E: U!
' + vpg (x2 + y2)
= pgz~
2E
2E
It may beseen thatpoints on the z-axis have only vertical displacements w=-pg(z2-z2)
2E
Other points of the bar, on account of lateral contraction, have not only vertical but also horizontal displacements. Lines which were
z
(b)
-Ger · ~ = -GfJy
The elementary theory also assumes that tTx
vpgxz
VP{j + pgc 2E + 2E (x2 + y2)
=
tT11
=
=
TZJI
= Ü
We can sho,v that this elementary 80lution is the exact solution under certain conditions. Since the stress components are a.II either linear functions of thc coordinates or zero, the equations of compatibility (130) are satisfied, and it is only necessary to con1:iidcr thc cquations of equilibrium (127) and the boundary eonditions (128). Substituting the abovc cxpressions for stress components into Eqs. (127 J \Ve find that thcse equations are satisfied, provided there are no body forces. The lateral surface of thc shaft is frec from forces, and the boundary conditions (128), rcmcmbcring that for the cylindrical surface cos (Nz) = n =O, reducc to Ü =
Trz COS
(Nx)
+T
11 , COS
(Ny)
(e)
250
THEORY OF ELASTICITY
PROBLEMS OF ELASTICITY IN TIIREE
For the case of a circular cylinder we have also
cos (Nx)
=
X
_,
'
cos (Ny)
Dl~!E!-.'SIO!-íS
251
cross section and the xz-plane in the principal plane of bending, the strells components given by the usual elementary theory of bending are (d)
= '}!_
'
Ex
rr. =
Substituting these and expressions (b) for the stress componcnts into Eq. (e) it is cvident that this cquation is satisfied. It is also evident that for cross sections other than circular, for \Vhich Eqs. (d) do not hold, thc stress components (b) do not satisfy the boundary condition (e), and therefore solution (a) cannot be applied. 'l'hcse more complicated problems of t\vist -..vill be considered !ater (see Chap. 11). Considering DO\\' thc boundary conditions for the ends of the shaft, we see that the surface shearing forces must be distributed in exactly the sarne manner as the stresses T,,. and 7~, over any intcrmediate cross section of the shaft. Only for this case is the stress distribution given
(a)
R'
in which R is the radius of curvature of the bar after bending. Substituting expressions (a) for the stress components in the equations of equilibrium (127), it is found that these equations are satisfied if there are no body forces. The boundary conditions (128) for the lateral sutface of the bar, which is free from externa} forces, are also satisfied. The boundary conditions (128) at the ends require that the surface forces must be distributed over the cnds in the sarne manner as the stressesª•· Only under this condition do the stresses (a) represent the exact solution of the problem. The bending moment M is given by the equation
ç.~I- __-.-_F--3"! f-~-1~ª y
j>---c ---i.Í
l-it.--1
X
[
(a}
(61
in \Vhich lv is the moment of inertia of the cross section of the beam ·with respect to the neutral axis parallel to the y-axis. From this equation \Ve find
Fio. 141.
by Eqs. (b) an exact solution of the problem. But the practical application of the solution is not limited to such cases. From SaintVenant's principie it can be concluded that in a long t"''isted bar, ata sufficient distance from the ends, the stresses depend only on the magnitude of the torque Jf, and are practically independent of the manner in \Vhich the forces are distributcd over the ends. The displacements for this case ean be found in the sarne manner as 1n the previous article. Assuming the sarne condition of constraint at t.he point A as in the previous problem "'·e find
u
=
-Oyz,
v = ()xz,
\vhich is a 'vell-known formula of the elementary theory of bending. Let us consider no,~· the displacements for the case of pure bending. Using Hooke's law and Eqs. (2) we find, from solution (a),
aw
x
fz=az=R
(e) (d)
w=O
This means that the assumption that cross sections remain plane and radii remain straight, "'·hich is usually made in the elementary derivation of the theory of t\~·ist, is correct. 88. Pure Bending of Prismatical Bars. Considera prismatical bar bent in one of its principal planes by two equal and opposite couples M (Fig. 141). Taking the origin of the coordinates at the centroid ôf the
(b)
By using these diffcrential cquations, and taking into consideration the fastening conditions of the bar, the displacements can be obtained in the sarne manner as in Art. 86. From Eq. (b) we find
"
w=R+wo
252
PROBLE~fS
THEORY OF ELASTIC/TY
in "\Vhich Wo is a function of x and y only.
Thc second and third of
Eqs. (d) give
au
ª'
from which
u
z2 -2R-
~
ÕWo
Z
-R-Tx' 2
awo
7ix+uo,
-z éPw 0
ax
2
+ auo =
vx,
_
ax
These equations must be sat.isfied for any value of z, hence 2 à Wo =
o
ilx2
,
J2wo =O ây'
(f)
'l'hcse conditions are satisfied by taking all the arbitrary constants equal to zero. Then =
1
2R [z2
-
z2
'x'
ZR
+ f1(y),
2
2z J wo _ df1(Y) _ ôf2(x)
axay
ôy
+ vy
ax
=
O
R
Noting that only the first term in this equation depends on clude that it is necessary to havc 2 Õ Wo
ax ay
=
z, we con-
o
(h)
Mz2 2EI'/I
This is the sarne deflection curve as is given by the elementary theory of bending. Let us consider now any cross section z = e, a distancc e from the left cnd of the bar. After dcformation, the points of this cross section will be in the plane
"
z=c+w=c+R
the elcmentary theory. To examine thc deformation of the cross scction in its plane, consider the sides y = ± b (Fig. 141b). After bcnding we have
wo=mx+ny+p f,(y)
~
;;'!; + ay + ~
f2(x) = -ax
+ f3
in \vhich m, n, p, li!, (3, y are arbitrary constants. the displacemcnts now become
The expressions for
z2 vx2 vy2 -2R-mz-2R+2R+a?J+Y ny -nz - R - ax + /3
"
v=w=O
i.e., in pure bending the cross section remains plane as is assumed in
'
These equations and Rqs. (f) require that
=
2R
(g)
Now substituting (e) and (g) into the first of Eqs. (d), 've find
v
- y2)],
U=--~---,
-
U=
+ v(x2
'fo get the deflection curve of the a.xis of the bar \Ve substitute in the above Eqs. (h) x = y = O. Then
und by intcgration Uo =
ôu_av_av_ õz-az-ax- 0
u=v=w=O,
u
R
253
The arbitrary constants are determined from the condition.s of fastening. Assuming that the point A, the centroid of the left end of the bar, together \\•ith an element of the z-axis and an element of the xz-plane, are fixed, we have for x = y = z = O
(')
Hcre u 0 and v0 denote unkno\vn functions of x and y, which 'vill be determincd !ater. Rub:;;tituting expressions (e) in Eqs. (e),
OF ELASTICITY IN THREE DIMEl\f810A'S
w= R+mx+ny+p
The sides become inclined as sho\vn in the figure by dotted lines. The other two sides of the cross section x = ±a are represented after bending by the equations X
= ±a+ u = ±a -
1 R [c2 2
+ v(a2
- y2)]
They are therefore bent to parabolic curves, V>'hich can be replaced with sufficient accuracy by an are of a circle of radius R/v, when the
THEORY OF ELASTICITY
PROBLEMS OF ELAST!CITY IN THREE DIME1\'SJO,\'S
deformation is small. ln considering the upper or lower sides of the bar it is evident that "\.Vhile the curvature of these sides after bending is convex down in thc lengthv;rise dircction, the curvature in the crossv,ise direction is convex up\vard. Contour lines for this anticlastic surlace will be as sho"\\'Il in Fig. 142a. By taking x and u constant in the first of Eqs. (h) we find that the equation for the contour lines is z2 - vy2 = constant
paths of the two rays is equal to an uneven number of half wave lengths of the light. The picture shov.·n in Fig. 142b, rcpresenting the hyperbolic contour lines, V.'as obtained hy this means. 89. Pure Bending of Plates. The result of the previous article can be applied in discussing the bending of plates of uniform thickness. If stresses 11% = Ez/R are distributed over thc edges of the plate parallel to the y-axis (Fig. 143), the surfacc of the plate v.·ill hecome 1 an A-,~:::=-==~==-=-=====-=-~«11-x _;,,.,.. 1 "''-;7 anticlastic surface, thc curvature of >vhich in planes parallel to the xz-plane is 1 / R and in the perpendicular direction is - v/ R. FIG. 143. If h denotes the thickness of thc platc, .L1f1 the bending moment per unit length on the edges parallel to the y-axis and
254
They are therefore hyperbolas with the asymptotes z2-vy2=0
FIG. 142a.
From this equation the angle a (Fig. 142a) is found from 1 2 tan a=-
'
This equation has been used for detcrmining Poisson's ratio 11. 1 If the upper surfacc of the beam is polished and a glass plate put over it, there will be, after bending, an air gap of variable thickness between the glass plate and the curved surface of the beam. This variable thickness can be measured optically. A beam of monochromatic
the moment of inertia per unit length, the relation between from the previous article, is 1
II
_\ _____ _ FIG. 142b.
light, say yello"-' sodium light, perpendicular to the glass plate, will be reflected partially by the plate and partially by the surface of the beam. The two refleeted rays of light interfere ,vith each other at points where the thickness of the air gap is such that the difference between the 1 A. Cornu, Compt. rend., vol. 69, p. 333, 1869. See also R. Straubel, Wied. Ann., vol. 68, p. 369, 1899.
Mi Elu
I2M1 =
Ehª
.Af 1
255
and R,
(a)
When V.'e have bending momcnts in t"-'O perpendicular directions (Fig. 144), the curvatures of the deftection surface may be obtained by superposition. Let 1/R 1 and X l/R2 be the curvatures of the deflection surface in planes parallel to the coordinate planes zx and zy, respectively; and lct frf 1 and M 2 be the bending z moments per unit length on the FIG. 144. edgcs parallel to the y- and xaxes, respcctivcly. Then, using Eq. (a) and applying the principie of superposition, we find 1 12 Ri= Ehs (M1 - vM2) 1
12
Ri= Ehª (.A.-12 - vM 1) 1
(b)
It is assumed that dellectioirn are small in comparison with the thickness of the plate.
257
THEORY OF ELASTICITY
PROBLEMS OF ELASTICITY IN THREE DIMEA'SfO,\'S
The moments are considered poRitive if they produce a deflection of the plate which is convex do\\'Il. Solving Eqs. (b) for M 1 and M 2, we find
The formulas (136) are used in the theory of plates when the bending moments are not uniform, and are accompanied by shear forces and surface pressures. For these circumstances they can be deduced from the general equations of Chap. 9 as approximations valid when the plate is thin. The elementary theory of bending of bars can bc related to the general equations in a similar manner. 1
256
M,
=
Eh' ( 1 12(1 - v 2) R1
Eh' M 2 = 12(1 -
For small deflections
\Ve
v 2)
1+ 1)
(
R2
R,
(o)
v R1
1
can use thc approximations
à'w - ay2
1
à'w
1
1) + v R2
- ax 2 '
R,
'l'hen, ""'·riting (135) v.·e find
=-D(::~+ v!~) M2 = -D(:~~+ v ::~) M1
(136)
The constant D is callcd the jlexural rigidity of a plate. ln the particular case when thc plate is bent to a cylindrical surface vrith generators parallel to the y-axis we have a2w/ay 2 =O, and, from Eqs. (136),
M, (137)
For the particular case in which M1 = M2 = M, we have 1
1
1
R,
R,
R
'l'he plate is bent to a sphcrical surface and the relation between the curvature and the bending moment is, from Eq. (e), Eh' M ~ 12(1 - v)
1
ºR
D(l + ,) =--li-
We shall have use for thcse rcsults later.
(138)
J. N. Goodicr, Tran8. Roy. Soe. Can., 3d ser., sec. III, vol. 32, p. 65, 1938.
259
TORSION
CHAPTER 11 TORSION 90. Torsion of Prismatical Bars. It has already been shown (Art. 87) that the exact solution of the torsional problem for a circular shaft is obtained if "'e assume that the cross sections of the bar remain plane and rotate 'vithout any distortion during t\'.'ist. This theory, developed by Coulomb,1 was applied latcr by Navier 2 to prismatical bars of noncircular cross scctions. Making the above assumption he arrived at the erroneous conclusions that, for a given torque, the angle of tvrist of bars is inversely proportional to the centroidal polar moment of inertia of the cross section, and that the maximum shearing stress occurs at the points most remate from the centroid of the cross section.ª It is easy to see that the above assumption is in contradiction with the boundary conditions. Take, for instance, a bar of rectangular cross section (Fig. 145). From Navicr's assumption it follows that at any point A on the boundary the shearing stress should act in the direction perpendicular to FIG. 145. the radius 0.4.. Resolving this stress into two components T~· and -r~,, it is evident that thcrc should be a complementary shearing stress, equal to T 11 ,, on thc clement of the lateral surface of the bar at the point A (see page 4), \Vhich is in contradiction \Vith the assumption that the lateral surface of the bar is free from external forccR, the twist being produccd by couples applied at the ends. A simple experiment >vith a rectangular bar, represented in Fig. 146, sho>\'S that the cross sections of the bar do not remain plane during torsion, and that the distortions of rectangular elements on the surface of the bar are greatest at the middles of the sides, i.e., at the points which are nearest to the axis of the bar.
The correct solution of the problem of torsion of prismatical bars by couples applied at thc ends was given by Saint-Venant. 1 He used the so-callcd semi-inverse method. That is, at the start he made ccrtain assumptions as to the deformation of the twisted bar and showed that \Vith these assumptions he could satisfy the equations of cquilibrium (127) and the boundary conditions (128). Then from the uniqueness of solutions of the elasticity equations (Art. 82) it follows that the assumptions made at thc start are correct and the solution obtained is thc exact solution of thc torsion problem. Consider a prismatical bar of any cross section t>visted by couples applied at the ends, Fig. 147. Guided by the solution for a circular shaft (page 249), Saint-Venant assumes that the deformation of the twisted shaft consists (a) of rotations of cross sections of the shaft as in thc case of a circular shaft and (b) of warping of the cross sections ,,,.hich F1G. 146. is the sarne for ali cross sections. Taking the origin of coordinatcs in an end cross section (Fig. 147) we find that the displacements corresponding to rotation of cross sections are
,.,
=
-8zy,
V
= 8z:x
(a)
where 8z is the angle of rotation of the cross section at a distance z from the origin. The warping of cross sections is defined by a function
w
z
~
e;;(x,y)
(b)
F1G. 147.
With the assumed displacements (a) and (b) we calculate the components of strain from Eqs. (2), which give E:
= Eu = E, = 'Yt11 =
i!w 7 •• -_ iJ:x
iJw 'Yu·=oy
1
"Histoire de l'académie," 1784, pp. 229-269, Paris, 1787. 1 Navier, "Résumé des leçons sur l'applieation de la mécanique,'' 3d ed., Paris, 1864, edited by Saint-Venant. • These conclusions are correct for a thin elastie layer, corresponding to a alice of the bar between two eross sections, attached to rigid pia.te~. See J, N, Qoodi.er, J, Appli~d Phyil., vol. 13, p. 167, 1942.
u
Y'
1
0
(º"' ) +az=IJ (º"' iJy+x) i!u + iJz
= 8 iJx - Y
(e)
i!c
M tm. savants étrangers, vol. 14, 1855. See also Saint-Venant's note to Navier's book, loc. cil., and 1. Todhunter and K. Pearson, "History of the Theory of Elaaticity," vol. 2.
The corresponding components of stress, from Eqs. (3) and (6), are do;
=
CTu
=
ífz
=
T"ll
=
and Eq. (e) bccomcs
)dy -(ª"'ay +x)dx ~O
( ay, ax y ds
O
=ao(:~ - y) Tuz = G8 (~: + x)
r::
(d)
It can be seen that with the assumptions (a) and (b) regarding the deformation, thcrc \Vill be no normal stresses acting bet\\·een the longitudinal fibcrs of the shaft or in the longitudinal direction of those fibers. There also ,,,jll be no distortion in the planes of cross sections, since E,,, Eu, 1'Xll vanish. V!i'c have at each point pure shear, defined by the components Txz anel Tuz· The function >Jt(x,y), defining warping of cross section, must now be determined in such a way that equations of equilibrium (127) will be O,__ _ _ __,__,x satisfied. Substituting expres~iions (d)
in these equations and
neglecting body forces we find that the function Y, must satisfy the equation
ª"' + ª"' ~ o ax2
iJy2
(139)
Consider nowthe boundary conditions (128). For the lateral surface of the bar, which is free from externai forces and has normais perpendicular to the z-axis, \Ve have X = Y = Z = Oand cos (N z) = n = O. The first two of Eqs. (128) are identically satisfied and the third gives (•)
which means that the resultant shearing stress at the boundary is directed along the tangent to the boundary, Fig. 148. It \Vas shown before (see page 258) that this condition must be satisficd if the lateral surfacc of the bar is free from externai forces. Considering an infinitesimal element abc at the boundary and assuming that s is increasing in the direction from e to a, we have
d'
m
=
dx
cos (Ny) = - ds
(140)
ds
1'hus each problcm of torsion is reduced to the problcm of finding a function Y, satisfying Eq. (139) and thc boundary condition (140). An alternativc procedure, which has the advantagc of leading to a simpl('r boundary condition, is as follows. In view of the vanishing of ff,,, ffy, ff,, Tri1 [Eqs. (d)], the equations of equilibrium (127) reduce to iJT:, = Ü
ª'
ªTy, = Ü
,
ª'
iJrn
ax
'
+ iJTy, ay
=O
The first two are alrcady satisfied sincc r'"" and r 11., as given by Eqs. (d), are independent of z. The third means that we can express r,.. and Ty, as T,..
ª"'ay
ª"'ax
=-1
(141)
v;•here 4> is a function of x and y, callcd the stressfunction. 1 From Eqs. (141) and (d) \Ve havc
ª'ay ~ ª' (ª"'ax - )·
-ªº~ao(ª"'+x) ax iJy
Y
(f)
Eliminating Y, by differentiating the first with respect to y, the second with respect to x, and subtracting from the first, we find that the stress function must satisfy the differential equation
Fio. 148.
l = cos (Nx) = dy,
261
TORSION
THEORY OF ELASTICITF
260
o2q, a2q, ax2+ay2=F
(142)
/<' = -2G8
(143)
where The boundary condition (e) becomes, introducing Eqs. (141), aq, dy
+ aq, dx
dy ds
ax ds
=
dq, =
ds
0
(144)
This shows that the stress function
It was introduced by L. Prandtl.
See Ph'IJrik. Z., vol. 4, 1903,
262
THEORY OF ELASTIClTY
TORSION
consists in finding the function cf> which satisfies Eq. (142) and is zero at the boundary. Several applications of this general theory to particular shapes of cross sections will be shown later. Let us consider now the conditions at the ends of the twisted bar. The normals to the end cross sections are parallel to the z-axis. Hence l = m = O, n = ± 1 and Eqs. (128) become
X=
(g)
±Txz,
in which the + sign should be taken for the end of the bar for which the external normal has the direction of the positive z-axis, as for the lower end of the bar in Fig. 147. We see that over the ends the shearing forces are distributed in the sarne manner as the shearing stresses over the cross sections of the bar. It is easy to prove that these forces give us a torque. Substituting in Eqs. (g) from (141) and observing that cf> at the boundary is zero, we find
263
forces, and sets up at the ends the torque given by Eq. (145). The compatibility conditions (130) need not be considered since the stress has been derived from the displacements (a) and (b). Thus all the equations of elasticity are satisfied and the solution obtained in this manner is the exact solution of the torsion problem. It was pointed out that the solution requires that the forces at the ends of the bar should be distributed in a definite manner. But the practical application of the solution is not limited to such cases. From Saint-Venant's principle it follows that in a long twisted bar, ata sufficient distance from the ends, the stresses depend only on the magnitude of the torque Mt and are practically independent of the manner in which the tractions are distributed over the ends. 91. Bars with Elliptical Cross Section. Let the boundary of the cross section (Fig. 149) be given by the equation x2 y2 a2 b2 - 1 = O (a)
+
FIG. 149. Then Eq. (142) and the boundary condition (144) are satisfied by taking the stress function in the form
Thus the resultant of the forces distributed over the ends of the bar is zero, and these forces represent a couple the magnitude of which is M1 =
ff
(Yx - Xy) dxdy = -
f f ~:xdxdy
-JJ~:
Integrating this by parts, and observing that cf> we find Mt = 2ff ct>dxdy
+ b2y2 -
x2 e/>= m ( a2 in which m is a constant.
)
1
(b)
Substituting (b) into Eq. (142), we find a2b2 m = 2(a2 + b2) F
Hence y dx dy
(h)
_
a
e/> - 2(a2
= O at the boundary, (145)
bF
+ b2)
(x + 2
a2
y2
)
(e)
b2 - 1
The magnitude of the constant F will now be determined from Eq. (145). Substituting in this equation from (e), we find
~t = a~ ! ~ 2 2 2
each of the integrals in the last member of Eqs. (h) contributing one half of this torque. Thus we find that half the torque is due to the stress component Txz and the other half to Tyz· We see that by assuming the displacements (a) and (b), and determining the stress components T.,., Tyz from Eqs. (141), (142), and (144), we obtain a stress distribution which satisfies the equations of equilibrium (127), leaves the lateral surface of the bar free from externa!
2 2
Smce
!! ~2d ~
X
( :2
ff
2
x dxdy
dy = I 'V= 7rbaª 4'
+:
2
ff
2
y dxdy -
ff
dxdy)
(d)
THEORY OF ELASTICITY
265
TORSION
in which
we find, from (d),
A = 7rab, from which (e)
F= Then, from (e),
(f)
Substituting in Eqs. (141), the stress components are 7
"''
2M1Y = - 7rab 3 '
(146)
The ratio of the stress components is proportional to the ratio y/x and hence is constant along any radius such as OA (Fig. 149). This means that the resultant shearing stress along any radius OA has a constant direction which evidently coincides with the direction of the tangent to the boundary at the point A. Along the vertical axis OB the stress component 'Tyz is zero, and the resultant stress is equal to T.,., Along the horizontal axis OD the resultant shearing stress is equal to Tyz· It is evident that the maximum stress is at the boundary, and it can easily be proved that this maximum occurs at the ends of the minor axis of the ellipse. Substituting y = b in the first of Eqs. (146), we find that the absolute value of this maximum is
are the area and centroidal moment of inertia of the cross section. Having the stress components (146) we can easily obtain the displacements. The components u and vare given by Eqs. (a) of Art. 90. The displacement w is found from Eqs. (d) and (b) of Art. 90. Substituting from Eqs. (146) and (148) and integrating, we find x
w
= M 1 (b2 - a2)xy (150) 7ra 3b3G
-L-
This shows that the contour lines lôrque Frn. 150. for the warped cross section are hyperbolas having the principal axes of the ellipse as asymptotes (Fig. 150). 92. Other Elementary Solutions. ln studying the torsional problem, SaintVenant discussed several solutions of Eq. (142) in the form of polynomials. To solve the problem let us represent the stress function in the form
+ F4 (x2 + y2)
q, = >1
(a)
Then, from Eq. (142),
a2 <1>1 ax2
a2<1>1
+ a:J1
(b)
=O
and along the boundary, from Eq. (144), 'T max.
2Mi = 7rab2
(147) >1
For a = b this formula coincides with the well-known formula for a circular cross section. Substituting (e) in Eq. (143) we find the expression for the angle of twist a2 + b2 (148) 8 = Mt . 7ra 3b3G The factor by which we divide torque to obtain the twist per unit length is called the torsional rigidity. Denoting it by C, its value for the elliptic cross section, from (148), is (149)
;
~Ili,; 1
'
+ F4 (x + y 2
2
)
=
constant
(e)
Thus the torsional problem is reduced to obtaining solutions of Eq. (b) satisfying the boundary condition (e). To get solutions in the form of polynomials we take the function of the complex variable (x
+ iy)n
(d)
The real and the imaginary parts of this expression are each solutions of Eq. (b) (see page 182). Taking, for instance, n = 2 we obtain the solutions x2 - y2 and 2xy. With n = 3 we obtain solutions xª - 3xy2 and 3x2y - yª. With n = 4, we arrive at solutions in the form of homogeneous functions of the fourth degree, and so on. Combining such solutions we can obtain various solutions in the form of polynomials. Taking, for instance,
q,
= F 4- (x2
+ y2) + >1
= F -
2
[12 (x2 + y2) - 2a-1 -
(xª -
3xy2)
+ b]
(e)
TORSION
THEORY OF ELASTICITY
266
· l tºon of Eq· (142) in the form of a polynomial of the degree weobtamasou1 . · third · l · with constants a and b which will be adjusted later. Th1s p_o~ynomial is~ so _ut10n of the torsional problem if it satisfies the boundary cond1t1~n (144), i.e., if the boundary of the cross section of the bar is given by the equation
!
2
(x2
+
y2) - _!_ (x• - 3xy2)
2a
+b =
(f)
O
By changing the constant b in this equation, we obtain various shapes of the cross · l section. . Taking b = - /1 a• we arrive at the solution for the eqmlatera1tnang e. tion (f) in this case can be presented in the forro (x - y13 y - ja)(x
+ V3 y
E
qua-
267
By changing a, Saint-Venant obtained the family of cross sections shown in Fig. I52a. Combining solutions in the form of polynomials of the fourth and eighth degrees, Saint-Venant arrived at the cross section shown in Fig. 152b.
On the basis of his investigations, Saint-Venant drew certain general conclusions of practical interest. He showed that, in the case of singly connected boundaries and for a given cross-sectional area, the torsional rigidity increases, if the polar moment of inertia of the cross section decreases. Thus for a given amount of material the circular shaft gives the largest torsional rigidity. Similar conclusions can be drawn regarding the maximum shearing stress. For a given torque and cross-
- ja)(x +ia) =O
which is the product of the three equations of the sides of the triangle shown in Fig. 151. Observing that F = -2GIJ and substituting
q,
= -GIJ
U
(x 2
+ y 2)
-
ia (x 3
-
3xy 2 )
-
;
2 1a ]
(g)
into Eqs. (141), we obtain the stress components Txz and Tu•· Along the x-axis, Txz = O, from symmetry, and we find, from (g),
a 3
y
T
Frn. 151.
(b)
= 3GIJ (2ax _ x 2)
2a
yz
(h)
3
GIJa
(k)
=2
At the corners of the triangle the shearing stress is zero (see Fig. 151). Substituting (g) into Eq. (145), we find G1Ja 4
Mt = · - - 15 v3
3 IJGlp 5
(l)
= -
Taking a solution of Eq. (142) in the form of a polynomial of th~ fourth degree containing only even powers of x and y, we obtain the stress funct10n >
= -GIJ
[~ (x• + y•)
_
~ (x4
_ 6z•y•
+ y4) + ~ (a
- 1)
J
The boundary condition (144) is satisfied if the boundary of the cross section is given by the equation
x'
;l
!.li. IH.
+ y'
- a(x' - 6x y
2 2
+y +a 4
)
1 = O
y
Frn. 152.
The largest stress is found at the middle of the sides of the triangle, where, from (h), T'max.
/J \ '
sectional area the maximum stress is the smallest for the cross section with the smallest polar moment of inertia. Comparing various cross sections with singly connected boundaries, Saint-Venant found that the torsional rigidity can be calculated approximately by using Eq. (149), i.e., by replacing the given shaft by the shaft of an elliptic cross section having the sarne cross-sectional area and the sarne polar moment of inertia as the given shaft has. The maximum stress in all cases discussed by Saint-Venant was obtained at the boundary at the points which are the nearest to the centroid of the cross section. A more detailed investigation of this question by Filon 1 showed that there are cases where the points of maximum stress, although always at the boundary, are not the nearest points to the centroid of the cross section. 1
L. N. G. Filon, Trans. Roy. Soe. (London), series A, vol. 193, 1900. Paper by G. Polya, Z. angew. Math. Mech., vol. 10, p. 353, 1930.
See also the
268
THEORY OF ELASTICITY
TORSION
Taking n = 1 and n = -1 in expression (d), and using polar coordinates r and y,, we obtain the following solutions of Eq. (b):
bar, subjected to a uniform tension at the edges anda uniform lateral pressure. li q is the pressure per unit area of the membrane and S is the uniform tension per unit length of its boundary, the tensile forces acting on the sides ad and bc of an infinitesimal element abcd (Fig. 154) give, in the case of small deflections of the membrane, a resultant in the upward direction -S(a 2z/ax 2 ) dx dy. ln the sarne manner the tensile
1
q, 1 = r cos Y,,
"'' =-;;. cos"'
Then the stress function (a) can be taken in the form
F 4
q, = - (x 2
+
y 2)
-
in which a and b are constants.
Fa r cos if; 2
-
+ Fb• - 2 a-r cos if;
(r 2
FIG. 153.
(m)
- - b2
It will satisfy the boundary condition (144) if at the boundary of the cross section we have
y
F 4
269
-
2a(r 2
-
b2 )
(
1 -
-
2
cos"'
b2 ) - r -
ª ~os"')
=
=O
O
(n)
(o)
which represents the equation of the boundary of the cross section shown in Fig. 153. 1 By taking r 2 - b2 =O
we obtain a circle of radius b with the center at the origin; and by taking Frn. 154.
1 _2acosif;=O r
we have a circle of radius a touching the y-axis at the origin. shearing stress is at the point A and is
The maximum
forces acting on the other two sides of the element give the resultant -S(éJ 2z/éJy 2 ) dx dy and the equation of equilibrium of the element is q dx dy
Tmax.
= G0(2a - b)
(p)
When b is very small in comparison with a, i.e., when we have a semicircular longitudinal groove of very small radius, the stress at the bottom of the groove is twice as great as the maximum stress in the circular shaft of radius a without the groove.
93. Membrane Analogy. ln the solution of torsional problems the membrane analogy, introduced by L. Prandtl, 2 has proved very valuable. Imagine a homogeneous membrane (Fig. 154) supported at the edges, with the sarne outline as that of the cross section of the twisted 1
This problem was discussed by C. Weber, Forschungsarbeiten, No. 249, 1921. Physik. Z., vol. 4, 1903. See also Anthes, Dinglers polytech. J., p. 342, 1906. Further development of the analogy and applications in various cases are given in the papers by A. A. Griffith and G. I. Taylor, Tech. Rept. Adv. Comm. Aercr nautics, vol. 3, pp. 910 and 938, 1917-1913. 2
a2z
a2z
+ S :>?. dx dy + S vX éJy
2
dx dy
=
O
from which (151) At the boundary the deflection of the membrane is zero. Comparing Eq. (151) and the boundary condition for the deflections z of the membrane with Eq. (142) and the boundary condition (144) (see page 261) ~or the stress function >, we conclude that these two problems are identical. Hence from the deflections of the membrane we can obtain values of > by replacing the quantity -(q/S) of Eq. (151) with the quantity F = -2GO of Eq. (142). Having the deflection surface of the membrane represented by cont~ur .line~ (~ig. 15~), several important conclusions regarding stress distnbut10n m tors10n can be obtained. Consider any point B on the
271
THEORY OF ELASTICITY
TORSION
membrane. The deflection of the membrane along the contour line through this point is constant, and we have
From Eq. (145) it can be concluded that double the volume bounded by the deflected membrane and the xy-plane (Fig. 155) represents the torque, provided q/S is replaced by 2G8.
270
az = O as
The corresponding equation for the stress function ac? = as
(ªe? dy + ac? dx) ay ds iJx ds
=
Txz
dy ds
Tyz
e? is
dx = O ds
This expresses that the projection of the resultant shearing stress at a point B on the normal N to the contour line is zero and therefore we may conclude that the shearing stress x at a point B in the twisted bar is in the direction of the tangent to the contour line through this point. The 1 z 1 curves drawn in the cross section of 1 1 a twisted bar, in such a manner that the resultant shearing stress at any ~--+-++7'-l~X point of the curve is in the direction of the tangent to the curve, are called lines of shearing stress. Thus the N contour lines of the membrane are the Y lines of shearing stress for the cross Frn. 155. section of the twisted bar. The magnitude of the resultant stress r at B (Fig. 155) is obtained by projecting on the tangent the stress components Txz and Tyz· Then
s~
T
=
Tyz COS
(Nx) -
Txz COS
ª
Let us consider now the equilibrium condition of the portion mn of the membrane bounded by a contour line (Fig. 155). The slope of the membrane along this line is proportional at each point to the shearing stress r and equal to r · q/S · l/2G8. Then denoting by A the horizontal projection of the portion mn of the membrane, the equation of equilibrium of this portion is
f s(r~ 2~8)ds
(Ny)
Substituting Txz
It may be observed that the form of the rnernbrane, and therefore the stress distribution, is the sarne no rnatter what point in the cross section is taken for origin in the torsion problern. This point, of course, represents the axis of rotation of the cross sections. It is at first sight surprising that the cross sections can rotate about a different (parallel) axis when still subjected to the sarne torque. The difference, however, is merely a matter of rigid body rotation. Consider, for instance, a circular cylinder twisted by rotations about the central axis. A generator on the surface becomes inclined to its original direction, but can be brought back by a rigid body rotation of the whole cylinder about a diarneter. The final positions of the cross sections then correspond to torsional rotations about this generator as a fixed axis. The cross sections remain plane but become inclined to their original planes in virtue of the rigid body rotation of the cylinder. ln an arbitrary section there will be warping, and with a given choice of axis the inclination of a given elernent of area in the end section is definite, aw/ax and aw/ay being given by Eqs. (d) and (b) of Art. 90. Such an elernent can be brought back to its original orientation by a rigid body rotation about an axis in the end section. This rotation will change the axis of the torsional rotations to a parallel axis. Thus a definite axis or center of torsional rotation, or center of torsion, can be identified provided the final orientation of an elernent of area in the end section is specifiedas for instance if the elernent is completely fixed.
or Tyz
__ ª
cos (Nx)
dx dn
=-1
cos (Ny)
= dy dn
we obtain r
= _
(ªe? dx + ac? dy) ax dn ay dn
= _ ddnc/J
Thus the magnitude of the shearing stress at B is given by the :naximum slope of the membrane at this point. It is only nece~s~ry m the expression for the slope to replace q/ S by 2G8.. From th1s it can be concluded that the maximum shear acts at the pomts where the contour lines are closest to each other.
= qA
frds = 2G8A
(152)
From this the average value of the shearing stress along a contour line can be obtained. By taking q = O, i.e., considering a membrane without lateral load. we arrive at the equation (153)
which coincides with Eq. (b) of the previous article for the function c/J 1 • Taking the ordinates of the membrane at the boundary so that z
+ 4F (x 2 + y 2)
= constant
(154)
272
273
THEORY OF ELASTICITY
TORSION
the boundary condition (e) of the previous article is also satisfied. Thus we can obtain the function cJ> 1 from the deflection surface of an unloaded membrane, provided the ordinates of the membrane surface have definite values at the boundary. It will be shown later that both loaded and unloaded membranes can be used for determining stress distributions in twisted bars by experiment. The membrane analogy is useful, not only when the bar is twisted within the elastic limit, but also when the material yields in certain portions of the cross section. 1 Assuming that the shearing stress remains constant during yielding, the stress distribution in the elastic zone of the cross section is represented by the membrane as before, but in the plastic zone the stress will be given by a surface having a constant
deflection of the membrane from the elementary formula for the parabolic deflection curve of a uniformly loaded string 1 (Fig. 156b), qc2
0
(a)
= 8S
From the known properties of parabolic curves, the maximum slope, which occurs in the middle portions of the long sides of the rectangle, is equal to (b)
The volume bounded by the deflected membrane and the xy-plane, calculated as for a parabolic cylinder, is qbc 3 128
2 3
V= -cob = S
X
~...__~=:__..=Ô..!. 2Ô
z
(b)
FIG. 156.
maximum slope corresponding to the yield stress. Imagine such a surface constructed as a roof on the cross section of the bar and the membrane stretched and loaded as explained before. On increasing the pressure we arrive at the condition when the membrane begins to touch the roof. This corresponds to the beginning of plastic flow in the twisted bar. As the pressure is increased, certain portions of the membrane come into contact with the roof. These portions of contact give us the regions of plastic flow in the twisted bar. Interesting experiments illustrating this theory were made by A. Nádai. 2 94. Torsion of a Bar of Narrow Rectangular Cross Section. ln the case of a narrow rectangular cross section the membrane analogy gives a very simple solution of the torsional problem. N eglecting the effect of the short sides of the rectangle and assuming that the surface of the slightly deflected membrane is cylindrical (Fig. 156), we obtain the 1 This was indicated by L. Prandtl; see A. Nádai, Z. angew. Math. Mech., vol. 3, p. 442, 1923. See also E. Trefftz, ibid., vol. 5, p. 64, 1925. 2 See Trans. A.S.M.E., Applied Mechanics Division, 1930. See also A. Nádai, "Theory of Flow and Fracture of Solids," 1950, Chaps. 35 and 36.
(e)
Now using the membrane analogy and substituting 2G8 for q/S in (b) and (e), we find 'Tma.x. = cGO, (d) from which Mt 8 (155) = ibc 3G Mt (156) 'Tmax. = -/rbc2 From the parabolic deflection curve (Fig. 156b)
z=
4ó2 (~ c 4
x2)
and the slope of the membrane at any point is
dz dx=
8ox
q
-7= -Sx
The corresponding stress in the twisted bar is 'Tyz
= 2G8x
The stress distribution follows a linear law as shown in Fig. 156a. Calculating the magnitude of the torque corresponding to this stress distribution we find 'Tmax. 2 b 1b -4 e. 3 e. = 6 e2'Tmox.
1
See S. Timoshenko and D. H. Young, "Engineering Mechanics," p. 35.
274
THEORY OF ELASTICITY
TORSION
This is only one half of the total torque given by Eq. (156). The second half is given by the stress components Txz, which were entirely neglected when we assumed that the surface of the defl.ected membrane is cylindrical. Although these stresses have an appreciable magnitude only near the short sides of the rectangle and their maximum values are smaller than Tmax. as calculated above, they act at a greater distance from the axis of the bar and their moment represents the second half of the torque Mt. 1 It is interesting to note that the Tmax. given by the first of Eqs. (d) is twice as great as in the case of a circular shaft with diameter equal to e and subjected to the sarne twist 8. This can be explained if we consider the warping of the cross sections. The sides of cross sections such as nn 1 (Fig. 157) remain normal to the longitudinal fibers of the bar at the corners, as is shown at the points n and n 1. The {aJ (b) total shear of an element such as abcd conFrG. 157. sists of two parts: the part 'Y1 dueto rotation of the cross section about the axis of the bar and equal to the shear in the circular bar of diameter e; and the part 'Y 2 due to warping of the cross section. ln the case of a narrow rectangular cross section 'Y 2 = ')' 1, and the resultant shear is twice as great as in the case of a circular cross section of the diameter e. Equations (155) and (156), obtained above for a narrow rectangle, can also be used in the cases of thinwalled bars of such cross sections as shown in Fig. 158 by setting b equal to the developed length of the cross section. This follows from the fact that, if the thickness e of a slotted tube (Fig. 158a) is small in comparison with the diam(a) (ó) eter, the maximum slope of the memFIG. 158. brane and the volume bounded by the membrane will be nearly the sarne as for a narrow rectangular cross section of the width e and of the sarne length as the circumference of the middle surface of the tube. An analogous conclusion can be made also for a channel (Fig. 158b). It should be noted that in this latter
case a considerable stress concentration takes place at the reentrant corners, depending on the magnitude of the radius r of the fillets, and Eq. (156) cannot be applied at these points. A more detailed discussion of this subject will be given in Art. 98. 96. Torsion of Rectangular Bars. Using the membrane analogy, the problem reduces to finding the defl.ections of a uniformly loaded rectangular membrane as shown in Fig. 159. These deflections must satisfy the Eq. (151)
1 This question was cleared up by Lord Kelvin; see Kelvin and Tait, "Natural Philosophy," vol. 2, p. 267.
275
(a) and be zero at the boundary. The condition of symmetry with respect to the y-axis and the boundary conditions at the sides x = ±a of the a rectangle are satisfied by taking z in the form of a series, Z=
1
~ L.
mrx y
b n COS
2(L
(b)
b
n
_J_
b
n=l,3,5,...
1
in which b1, b3, . . . are constant coeffi.cients l_._____. and Y1, Y3, . . . are functions of y only. Suby stituting (b) in Eq. (a), and observing that the Frn. 159. right side of this equation can be presented in the form of a series,1
\:"
q 4 n-l n11"X S n11" ( -1) 2 cos -2-a
f...
(e)
n=l,3,5, ...
we arrive at the following equation for determining Yn: 2 2 4 n-1 y11_n11"Y= q ( -2 n 4a 2 n -Sn11"bn -l)
(d)
from which
Yn =A sinh n211"Y a
+B
cosh n211"Y a
+ Sl6qa2
n~~n
(-l)n;1
() e
From the condition of symmetry of the defl.ection surface of the membrane with respect to the x-axis, it follows that the constant of integration A must be zero. The constant B is determined from the 1
B. O. Peirce, "A Short Table of Integrals," p. 95, 1910.
277
TORSION
THEORY OF ELASTICITY
276
condition that the defiections of the membrane are zero for y = ± b, i.e., (Yn)u=•b =O, which gives
y = 16qa2 (-l)n; [l _ cosh (n7ry/2a)] " Sn 37r 3 b,. cosh (n7rb/2a) 1
very narrow rectangle, b/a becomes a large number, so that the sum of the infinite series in (157) can be neglected, and we find =
Tmax.
(f)
and the general expression for the defiection surface of the membrane, from (b), becomes ., 1 16qa2 ~ _!__ (-l)n; [l _ cosh (n7ry/2a)] cos n7rX z = S7rª ~ n3 cosh (n7rb/2a) 2a
2G8a
This coincides with the first of the Eqs. (d) of the previous article. ln the case of a square cross section, a = b; and we find, from Eq. (157), 1 2GOa { 1 - : 2 [ cosh (7r/2) + 9 cosh\37r/2) + · · ·]} Tmax. = =
2G8a [ 1 - -82 ( -17r 2.509
+9
1 X 55.67
.)]
+
= 1.351G8a (158)
n= 1,3,5, ...
Replacing q/ S by 2G8, we obtain for the stress function ., 1 q, = _32_G_8_a2 ~ _!_ (-l)n; [l _ cosh (n7ry/2a)] cos n7rx 3 7ra n cosh (n7rb/2a) 2a
ln general we obtain Tmax.
(g)
in which k is a numencal factor depending on the ratio b/a. values of this factor are given in the table below.
n=l,3, , ••.
yz
ôcp
16G8a
ÔX
1r2
~
=--=--
~
_!_(-l)n; n2
1 [
1 _cosh (n7ry/2a)]sinn7rx (h) cosh (n7rb/2a) 2a
n= 1,3,5, ...
Assuming that b > a, the maximum shearing stress, corresponding to the maximum slope of the membrane, is at the middle points of the long sides x = ±a of the rectangle. Substituting x = a, y = O in (h), we find ., 16G8a 1 Tmax. =~ [ - cosh
L
n= 1,3,5, ...
~2
(~7rb/2a)]
ba
-b a
k
ki
k2
1.0 1.2 1.5 2.0 2.5
0.675 0.759 0.848 0.930 0.968
0.1406 0.166 0.196 0.229 0.249
0.208 0.219 0.231 0.246 0.258
k
3 4 5 10
ki
0.985 0.997 0.999 1.000 1.000
00
Mi=2
t
f~af~b cpdxdy=64~:ª2f~af~b{
~3(-l)n;l
n=l,3,5, ..•
7r
2:
.,
n=l,3,5, ...
n 2 cosh
~n7rb/2a)
n=l,3,5, .•• 1
1
I 1
1
1
+ 3 4 + 54 + . . .
7r4
= 96
B • O • P errce, · " A Short Table of Integrals," p. 90, 1910.
1
n4
n=l,3,5, .••
~
7r6
or, observing that
~
~
64G8(2a)4 (157)
The infinite series on the right side, for b > a, converges very rapidly and there is no difficulty in calculating Tmax. with sufficient accuracy for any particular value of the ratio b/a. For instance, in the case of a
~
l _ cosh (n7ry/2a)] c s n7rx} d d _ 32G8(2a)3(2b) 0 cosh (n7rb/2a) 2a x Y 7r4
.,
16G8a = 2G8a - - -2 -
0.267 0.282 0.291 0.312 0.333
Let us calculate now the torque M 1 as a function of the twist o. Using Eq. (145) for this purpose, we find
[
we have
k2
0.263 . 0.281 0.291 0.312 0.333
or, observing that
Tmax.
Severa!
TABLE OF CONSTANTS FOR TORSION OF A RECTANGULAR BAR
The stress components are now obtained from Eqs. (141) by differentiation. For instance,
T
(159)
= k2G8a
_!_5 tanh n7rb n
2a
we have
The ~unction.
"' '\"''
(
M1 =
1
3 G0(2a)
3
(2b)
(
1 -
7192 ba
_!_ tanh n7rb)
~
n•
2
ª
(160) q,,
= GIJ
n=l,3,5, .•.
~
2
[r coscos a2..Y + ª 2
Ê)
q, =
~IJ [-r• (l
= 0.1406G0(2a)
96. Additional Results. By using infinite series as in the previous article, the torsional problem can be FIG. 160. solved for several other shapes of cross sections. ln the case of a sector of a circle 1 (Fig. 160) the boundaries are given by "1 = ± a/2, r = O, r = a. We take a stress function in the form
y
> = >1
+4
+ y•)
G1Jr 2 = <1>1 -
2
This problem was discussed by Saint-Venant, Compt. rend., vol. 87, pp. 849 e.nd 893, 1878. See also A. G. Greenhill, Messenger of Math., vol. 9, p. 35, 1879. Another method of solution by using Bessel's function was given by A. Dinnik, Bull. Don Polytech. Inst., Novotcherkassk, vol. 1, p. 309. See also A. Fõppl and L. Fõppl, "Dre.ng und Zwang," p. 96, 1928. 1
A
(1.'.)~ n7r1/IJ a cos ~
n
This expression is zero at the boundaries:
To make it vanish also along the circular boundary r = a, we must put
l
A,. cos
n1rf
=
l _ cos 2..Y cos
Ol
Ol
n = 1,3,5, ...
(163)
where k 2 is a numerical factor the values of which can be taken from the table on page 277.
(x 2
~ L,.
+ az
from which we obtain, in the usual way,
in which k1 is a numerical factor depending on the magnitude of the ratio b/a. Several values of this factor are given in the table on page 277. Substituting the value of (J from Eq. (163) into Eq. (159), we obtain the maximum shearing stress as a function of the torque in the form (164)
F
_ cos 2..Y) cos a
"
ln general the torque can be represented by the equation
Mi = kiG0(2a) 3 (2b)
(r)'.;;: cos n7Mf;] -a
A,. -a
n=l,3,5, ...
(162)
4
Taking a solution
we arrive at the stress function
ln the case of a square, a = b; and (160) gives
Mt
l"
2
n=l,3,5, ..
The series on the right side converges very rapidly, and M 1 can easily be evaluated for any value of the ratio a/b. ln the case of a narrow rectangle we can take n7rb 1 t anh 2a Then (161) M = G0(2a) 3 (2b) (1 - 0.630 1
279
TORSION
THEORY OF ELASTICITY
278
16a•
A,.=
7
n+l 2
(-l)
(
1 2a) (
n n+-;
2a)
n--;-
The stress function is therefore
q,
=
GIJ 2
[-r• (l _
21/;) + 16a•a• 7rs
cos cos a
"'
l:
n=l,3,5, ..
Eq. (145), we find M, = 2f f
~ubstituting i~to a~tor
a=
1r
1r
1r
27r
4
3
2
3
k= k, = k. =
0.0181
0.0349 0.452 0.490
0.0825
11'
37r
57r
2
3
27r
------ ---
...... ......
. ..... . .....
0.148 0.622 0.652
0.296 0.719 0.849
0.572 1 0.672 1
..... . ....
. .... . ....
0.878•
...... . .....
'These figures have been corrected b M A' S G lnequalities in Mathematical Phyics" P y261. p1~seut. Uee . . P?lya and G. Szegõ, "lsoperimetric • Th' fi . • · • rmce on mvers1ty Press, 1951. lB gure has been corrected by Dinnik, Zoe. cit.
280
THEORY OF ELASTICITY
TORSION
The maximum shearing stresses along the circular and along the radial boundaries are given by the formulas k 1Ga8 and k.Ga8, respectively. Several values of ki and k2 are given in the table on page 279. The solution for a curvilinear rectangle bounded by two concentric circular ares and two radii can be obtained in the sarne manner. 1 ln the case of an isosceles right-angled triangle 2 the angle of twist is given by the equation
lar case to the determination of the stress function satisfying the differential equation (142) and the boundary condition (144). ln deriving an approximate solution of the problem it is useful, instead of working with the differential equation, to determine the stress function from the minimum condition of a certain integral, 1 which can be obtained from consideration of the strain energy of the twisted bar. The strain energy of the twisted bar per unit length, from (88), is
M,
8 = 38.3ªª'
in which a is the length of the equal sides of the triangle. The maximum shearing stress is at the middle of the hypotenuse and is equal to
M,
Tmax.
v =
2~ JJ
2
+ Ty, 2) ax ay
=
281
2~ JJ[(::)2 + (::)2] ax ay
lf we give to the stress function q, any small variation ô
= 18.02 O,i
By introducing curvilinear coordinates several other cross sections have been investigated. Taking elliptic coordinates (see page 193) and using conjugate functions ~ and 'h determined by the equation
x
+ iy
=
e cosh
(~
+ iri)
we arrive at cross sections bounded by confocal ellipses and hyperbolas. 3 using the equation•
and the variation of the torque is, from Eq. (145), By
we obtain cross sections bounded by orthogonal parabolas. Solutions have been found for many other sections,6 solid and hollow, including polygons, angles, cardioids, lemniscates, 6 and circles with one or several eccentric holes.7 When the section can be conformally mapped into the unit circle a solution can always be written down in terms of a complex integral. 8
97. Solution of Torsional Problems by Energy Method. 9 We have seen that the solution of torsional problems is reduced in each particuSaint-Venant, Zoe. cit. See also A. E. H. Love, "Theory of Elasticity,'' 4th ed. p. 319, 1927. 2 B. G. Galerkin, Bull. acad. des sei. de Russ., p. 111, 1919; G. Kolosoff, Compt. rend., vol. 178, p. 2057, 1924. 8 A. G. Greenhill, Quart. J. Math., vol. 16, 1879. See also L. N. G. Filon, Trans. Roy. Soe. (London), series A, vol. 193, 1900. •E. W. Anderson and D. L. Holl, Iowa State Coll. J. Sei., vol. 3, p. 231, 1929. 6 A compilation is given by T. J. Higgins, Am. J. Phys., vol. 10, p. 248, 1942. 6 References to papers giving exact solutions for such sections, too numerous to include here, may be found by consulting Applied Mechanics Reviews, Science Abstracts A, Mathematieal Reviews, and Zentralblatt für Meehanik. Most of the references on p. 331 refer to or include the corresponding torsion problem. 7 See C. B. Ling, Quart. Applied Math., vol. 5, p. 168, 1947. 8 Due to N. I. Muschelisvili. See I. S. Sokolnikoff, "Mathematical Theory of Elasticity,'' Chap. 4, 1946. 9 For a survey, with references, of this and other approxi!llate Il'leth<;>ds 11ee T; J. Hi~gins, J. Applied Phys., vol. 14, p. 469 1 1943, .· 1
2ff ôq, dx dy Then by reasoning analogous to that used in developing equation (91) on page 164, we conclude that
2~ ô or ô
ff [(::r + (::rJ ff f f g[(::r + (::YJ - 2aoq;} ax ay = 20
ôq, dx ay
dxdy =
º
Thus the true expression for the stress function q, is that which makes zero the variation of the integral
u
=
ff g[(::r + (::YJ -
2aoq,} dx dy
(165)
.w~ come also to the sarne conclusion by using the membrane analogy and the prmc1ple of virtual work (Art. 48). If S is the uniform tension in the membrane the i_ricr~ase in strain energy of the membrane due to deflection is obtained b; mult1plymg the tension S by the increase of the surface of the membrane. ln this manner we obtain
~s JJ[ (:;)2 + (!;)2] dxdy 1
This method was proposed by W. Ritz, who used it in the solution of problems of bending and vibration of rectangular plates. See J. reine angew. Math., vol. 135, 1908, and Ann. Physik, series 4, vol. 28, p. 737, 1909. 2 If ôq, is taken equal to zero at the boundary, no forces on the lateral surface of the bar will be introduced by variation of q,.
282
THEORY OF ELASTICITY
TORSION
where z is the deflection of the membrane. If we take now a virtual displacement of the membrane from the position of equilibrium, the change in the strain energy of the membrane due to this displacement must be equal to the work done by the uniform load q on the virtual displacement. Thus we obtain
Take as an example the case of a rectangular cross section1 (Fig. 159). The boundary is given by the equations x = ±a, y = ±b, and the function (x 2 - a 2)(y 2 - b2) is zero at the boundary. The series (a) can be taken in the form
~ Sô
JJ[(:;)' + (::rJ ax ay = JJ
.283
(e)
q ôz ax ay
in which, from symmetry, m and n must be even. Assuming that we have a square cross section and limiting ourselves to the first term of the series (e) we take ·
and the determination of the deflection surface of the membrane is reduced to finding an expression for the function z which makes the integral
e/>
= ao(x 2 - a 2) (y2 - a2)
(d)
Substituting this in (165) we find from the minimum condition that a minimum. If we substitute in this integral 2GO for q/S, we arrive at the integral (165) above.
ªº
M1 = 2Jf e/> dx dy = ~-jGOa 4 = 0.1388(2a)4GO Comparing this with the correct solution (162) we see that the error in the torque is about li per cent. To get a closer approximation we take the three first terms in the series (e). Then, by using the condition of symmetry, we obtain
(a)
au
ªªº
=0
'
au âa2
e/>
(b)
Thus we obtain a system of linear equations from which the coefficients a 0, a 1, a 2, . . . can be determined. By increasing the number of terms in the series (a) we increase the accuracy of our approximate solution, and by using infinite series we may arrive at an exact solution of the torsional problem. 1 1 The condition of convergency of this method of solution was investigated by Ritz, loc. cit. See also E. Trefftz, "Handbuch der Physik," vol. 6, p. 130, 1928.
= (x2 - a2)(y2 - a2)[ao
+ a1(x2 + y2)]
(e)
Substituting this in (165) and using Eqs. (b), we find
ªº
5 259 GO
=
8 · 277 ·
a2'
ª
5 3
1
=
35
GO
8 · 2 · 277 · a4
Substituting in expression (145) for the torque, we obtain Mi
=0 . . . '
8 a2
The magnitude of the torque, from Eq. (145), is then
ln the approximate solution of torsional problems we replace the above problem of variational calculus by a simple problem of finding a minimum of a function. W e take the stress function in the form of a series
in which cf>o, cp 1, cp 2, . . . are functions satisfying the boundary condition, i.e., vanishing at the boundary. ln choosing these functions we should be guided by the membrane analogy and take them in a form suitable for representing the function cp. The quantities ao, a 1, a 2, . . . are numerical factors to be determined from the minimum condition of the integral (165). Substituting the series (a) in this integral we obtain, after integration, a function of the second degree in a 0, a 1, a 2, . . . , and the minimum condition of this function is
5GO
=
= VCi~~
+i
· ! · fü)GOa 4 = 0.1404G0(2a)4
This value is only 0.15 per cent less than the correct value. A much larger error is found in the magnitude of the maximum stress Substituting (e) into expressions (141) for the stress components w~ find that the error in the maximum stress is about 4 per cent, and to get ª better accuracy more terms of the series (e) must be taken. e It ~an be seen from the membrane analogy that in proceeding as xplamed above we generally get smaller values for the torque than the correct value. A perfectly fiexible membrane, uniformly stretched at a
1
See S. Timoshenko, Bull. lnst. Ways of Gommunication, St. Petersburg, 1913 nd Proc. London Math. Soe., series 2, vol. 20, p. 389, 1921. '
284
THEORY OF ELASTICITY
cf>
=
00
00
2:
2:
m11"X
n7r'Y
cos 2a cos '2b
amn
285
TORSION
tho boundary and uniformly loaded, is a system with an infinite number of degrees of freedom. Limiting ourselves to a few terms of the series (e) is equivalent to introducing into the system certain constraints, which reduce it to a system with a few degrees of freedom only. Such constraints can only reduce the flexibility of the system and diminish the volume bounded by the deflected membrane. Hence the torque, obtained from this volume, will generally be smaller than its true value. E. Trefftz suggested 1 another method of approximate determination of the stress function cf>. With this method the approximate magnitude of the torque is larger than its true value. Hence by using the Ritz and the Trefftz methods together the limits of error of the approximate solution can be established. ln using Ritz's method we are not limited to polynomials (e). We can take the functions cf>o, cf> 1, c/>2, ••• of the series (a) in other forms suitable for the representation of the stress function cf>. Taking, for instance, trigonometric functions, and observing the conditions of symmetry (Fig. 159), we obtain (f)
where a = b/a. Substituting in (f) we obtain the exact solution of the problem in the forro of an infinite trigonometric series. The torque will then be
00
00
~
~
128G8b 2
32ab
Lt Lt 11'4mn(m2a2 + n2) . mn11"2 m=l,3, ... n=l,3, ...
(g)
This expression is brought into coincidence with expression (160) given before if we observe that oo
1 m2
~
Lt
n=l,3,5, ..•
n 2 (m 2a 2
tanh ma11" - ma11" 2 2 = 96m 2 -i(ma1r/2) 3 11' 4
l
+n
2 )
As another example, in the case of a narrow rectangle, when b is very large in comparison with a (Fig. 159), we may take, as a first approximation, cf> = GO(a 2 - x 2) (h)
n=l,3,5, ... m=l,3,5, •..
Substituting in (165) and performing the integration, we find that
2: 00
u = 1r~b
~ Lt
a2mn
(m2 n2) (i2 + b2
cf> = Ge(a 2
m=l,3,5, ... n=l,3,5, .•.
2: m=l,3,5, ...
2:
n = 1,3,5, ...
amn
x 2)[1
fj=!
a
Equations (b) become
?r2ab 4
-
(m2a2 + n2) - 2G8. mn11'2 I6ab ( -1) mtn-1 = O b2
and we find 128G8b 2 ( -1) 11" 4mn(m 2a 2
m+n_ 1 2
+n
-
e-ll(b-yl]
(k)
and choose the quantity /j in such a manner as to make the integral (165) a minimum. ln this way we find
00
-2G8
which coincides with the solution discussed before (Art. 94). To get a better approximation satisfying the boundary condition at the short sides of the rectangle, we may take
2
)
1 E. Trefftz, Proc. Second Intern. Congr. Applied Mech., Zürich, 1926, p. 131. See also N. M. Basu, Phil. Mag., vol. 10, p. 886, 1930.
@_
'\}2
Dueto the exponential term in the brackets of expression (k) we obtain a stress distribution which practically coincides with that of the solution (h) at all points a considerable distance from the short sides of the rectangle. Near these sides the stress function (k) satisfies the boundary condition (144). Substituting (k) into equation (145) for the torque, we find
287
THEORY OF ELASTICITY
TORSION
which is in very good agreement with Eq. (161) obtained before by using infinite series. A polynomial expression for the stress function, analogous to expression (e) taken above for a rectangle, can be used successfully in all cases of cross sections bounded by a convex polygon. If
ln the particular case when m = !, p = q = 1, a = a 1, we have y = ±aif;(x/b) = ± v'x/b[l - (x/b)J, and we obtain
286
are the equations of the sides of the polygon, the stress function can be taken in the forro
cp
= ( a1X
+ biy + C1) (a2X + b2y + C2)
· · · (anX
+ b,.y + Cn) "1; "1;amnXnym
and the first few terms of the series are usually sufficient to get a satisfactory accuracy. The energy method is also useful when the boundary of the cross section (Fig. 161) is given by two curves 1 Y
= aif; (~)
A
= -
Go 11 a 2' 1 + 13 b2
f
c
a
e
where
a~
Frn. 161.
+ a1if;)
Substituting into the integral (165) we find, from the equation dl/dA =O, GO A where a=
Io1 i/;
3
fo
Mi = -A b(a
8= 1
+ a1)ª (1 if;ª dt 3
}o
i Such problema were discussed by L. S. Leibenson. See his book "Variational Methods for Solving Problema of the Theory of Elasticity," Moscow, 1943. See also W. J. Duncan, Phil. Mag., series 7, vol. 25, p. 634, 1938.
(cJ
3Mt (b1c1 3 + 2b2c2 3)G
(a)
LuftJahrt-forsch., vol. 20, 1944, tranlated as N.A.C.A. Tech. Mem. 1182, 1948. A more elaborate formula, taking account of the increased stiffness resulting fron;- the junctions of the rectangles, was developed on the basis of soap film and tors10n tests by G. W. Trayer and H. W. March, Natl. Advisory Comm. Aeronaut., Rept. 334, 1930. 3 Comparison of torsional rigidities obtained in this manner with those obtained by · t s is · g1ven · . ex?enmen for severa! types of rolled sections and for various dimensions m the paper by A. Fõppl, Sitzber. Bayer. Akad. Wiss., p. 295, München, 1921. See also Bauingenieur, series 5, vol. 3, p. 42, 1922. 2
From Eq. (145) we find the torque
bL
is obtained with sufficient accuracy from Eq. (155) by putting, instead of b, in this equation the developed length of the center line, 2 namely, b = 2a - e. ln the case of a channel section (Fig. 162b) a rough approximation for the angle of twist is obtained by taking for the fianges an average thickness c2, subdividing the cross section into the three rectangles, and substituting in Eq. (155), b1c1 3 + 2b 2c23 instead of bc 3, i.e., assuming that the torsional rigidity of the channel is equal to the sum of the torsional rigidities of the three rectangles. a Then
3
i/; dt
G8ba 3 11 ª 2 1 + 13 b2
Frn. 162.
(#/dt) 2 dt 1
DJ (b)
(a)
cp = A(y - aif;)(y
= 0.0736
An approximate solution, and a comparison with tests, for sections bounded by a circle and a chord has been given by A. W eigand. 1 Numerical methods are discussed in the Appendix. 98. Torsion of Rolled Profi.le Sections. ln investi~ating the torsion of rolled sections such as angles, channels, and 1-beams, the formulas derived for narrow rectangular bars (Art. 94) can be used. If the cross section is of constant thickness, as in Fig. 162a, the angle of twist
and
The boundary conditions will be satisfied if we take for the stress function an approximate expression
Mi
'
ll
l1u,1.
288
To calculate the stress at the boundary at points a considerable distance from the corners of the cross section we can use once more the equation for a narrow rectangle and take T
narrow rectangle
T1
= G8c. Using this, we obtain from (d) dT dr
= _
2T1
r
(d')
e
A T1r T=---
r
(b)
The sarne approximate equations can be used for an I-beam (Fig. 162c). At reentrant corners there is a considerable stress concentration, the magnitude of which depends on the radius of the fillets. A rough approximation for the maximum stress at 1 these fillets can be obtained from the 1 membrane analogy. Let us consider a 1 cross section in the forra of an angle of 1 1 constant thickness e (Fig. 163) and with 1 radius a of the fillet of the reentrant comer. r / Assuming that the surface of the membrane e _______ ....-v1'at the bisecting line 00 1 of the fillet is L " approximately a surface of revolution, with FIG. 163. axis perpendicular to the plane of the figure at O, and using polar coordinates, the Eq. (151) of the de:flection surface of the membrane becomes (see page 57) 1
ºª
+ !T
from which, by integration,
= e(}(]
Then, from Eq. (a), we obtain for the :flanges of the channel
r
289
TORSION
THEORY OF ELASTICITY
(f)
e
where A is a constant of integration. For the determination of this constant, let us assume that the shearing stress becomes zero at point 01 at a distance c/2 from the 3.Sr-r--.----,.--~-boundary (Fig. 163). Then, from (f),
A _ a+ (c/2)
T1[a
+ (c/2)] e
3.0 t-t--+---+---+------l
_
- 0 2.5
and
t---t---+---+----1-----1
"l:mux
'1
2.0 t - - - \ - - + - - - 1 - - - - - 1 - - - - 1
1
/,-(
2
dz dr 2
+ ! dz
= _
r dr
!1 S
(e)
Remembering that the slope of the membrane dz/dr gives the shearing stress T when q/S is replaced by 2G8, we find from (e) the following equation for the shearing stress:
dT 1 dr+ -;:r = -2G8
(d)
The corresponding equation in the arms of the angle at a considerable distance from the corners, where the membrane has a nearly cylindrical surface, is dr
-
dn
= -2G8
(e)
in which n is the normal to the boundary. Denoting by T 1 the stress at the boundary we find from (e) the previously found solution for a
Substituting in (f) and taking r =a, we find Tmax. =
T1
(1
+ 4: )
(g)
1.0 ~-_,__--.1-_---l._
o
0.5
1.0
J.5
__J
2.0
For a = !e, as in the Fig. 163, we a/e have Tmax. = l.5T1. For a very Frn. 164· small radius of fillet the maximum stress becomes very high. Taking, for instance, a = O.lc we find Tmax. = 3.5T1. More accurate and complete results can be obtained by numerical calculations based on the method of finite differences (see Appendix). A curve of Tmax./r1 as a function of a/e obtained by this methodI is shown in ~ig. 164 (curve A), together with the curve representing Eq. (g). It w1ll be seen that this simple formula gives good resultswhen a/ e is less than 0.3. 99. The Use of Soap Films in Solving Torsion Problems. We have ~een that the membrane analogy is very useful in enabling l,lS to visualize the stress distribution over the cross section of a twisted bar. T~ B~ J. H. Ruth, J. Applíed Mechan_ícs (~rans. A.S.M.E.), vol. 17, p. 388, 1950. e nse of the curve towards the nght is required by the limiting case as the fillet radius is increased in relation to the leg thickness. References to earlier attempts to solve this problem including soap-film measurements are given by I. Lyse and B. G. Johnston 1 Proc, A.S.Ç,E. 1 19351 p. 4691 and in. the above paper.
290
THEORY OF RLASTICITY
Membranes in the form of soap films have also been used for direct measurements of stresses. 1•2 The films were formed on holes cut to the required shapes in flat plates. To make possible the direct determina·· tion of stresses, it was found necessary to have in the sarne plate a circular hole to represent a circular section for comparison. Submitting
FIG. 165.
both films to the sarne pressure, we have the sarne values of q/S, 3 which correspond to the sarne values of G8 for the two bars under twist. Hence, by measuring the slopes of the two soap films we can compare the stresses in the bar of the given cross section with those in a circular 1 See papers by Taylor and Griffi.th, loc. cit.; also the paper by Trayer · an~ March, loc. cit. . . . . 2 A survey of this and other analogies for torsion, with references, is g1ven by T. J. Higgins, Experimental Stress Analysis, vol. 2, no. 2, p. 17, 1945. a It is assumed that the surface tension is the sarne in both films. This was proved with sufficient accuracy by the tests.
'i:.
l i:'I:,
l,I
TORSION
291
shaft under the condition that they have the sarne angle of twist O per unit length and the sarne G. The corresponding ratio of the torques is determined by the ratio of the volumes between the soap films and the plane of the plate. For obtaining the contour lines of the films the apparatus shown in Fig. 165 was used. 1 The aluminum plate with the holes is clamped between the two halves of the cast-iron box A. The lower part of the box, having the form of a shallow tray, is supported on leveling screws. The mapping of contour lines is done by using the screw B passing through a hole in a sheet of plate glass sufficiently large to cover the box in any possible position. The lower end of the screw carries a hard steel point whose distance from the 6rfllrfl glass plate is adjustable by the ---;-_-_-_-_ _ 1,_s.s_.• s ffl__lfrfl~~r11l~r11lr11_,, screw. The point is made to - . .· " /•" ( ( , - -:. -=..:r:...-1-j..--;:- -"-._ ·1rfl /"" approach the film by moving the ', ' 1 ,. ,....;.t-, ,, 't or11 ,.. '-~,', ~ I j \ l I / / glass pia te until the distortion of "\~\'-....:,~'/"~..{.:--' the image in the film shows that '~~~--/~~/." ,,,/,{' contact has occurred. The record is \ 1 11 made on a sheet of paper attached to /\/ 1 v, the board E, which can swing about 1 1 a horizontal axis at the sarne height : 1 1 1 as the steel recording point D. To mark any position of the screw, it is Fra. 166. only necessary to prick a dot on the paper by swinging it down on the recording point. After the point has been made to touch the film at a number of places, the dots recorded on the paper are used for drawing a contour line. By adjusting the screw B this can be repeated for as many contour lines as may be required. When these lines have been mapped, the volume and the corresponding torque can be obtained by summation. The slopes and the corresponding stresses are obtained by measuring the distances between neighboring contour lines. The slope can be obtained optically with much more accuracy by projecting a beam of light on to the surface of the film and measuring the angle of ~he_ reflected ray. The normal to the film is then half way between the mc1dent and the reflected rays. A special instrument was constructed for this purpose by Griffith and Taylor. Figure 166 represents an e~ample of contour lines obtained for a portion of an 1-beam (wooden ~mg spar of an airplane). From the close grouping of the contour hnes at the fillets of the reentrant corners and at the middle of the Upper face, it follows that the shearing stresses are high at these places. The projecting parts of the flange are very lightly stressed. The 1 See the paper by Taylor and Griffi.th, loc. cit.
292
THEORY -OF ELASTICITY
TORSION
rnaxirnurn stress in the middle portion of the web is practically constant along the side of the web and equal to that in a narrow rectangle for the sarne angle of twist. The application of soap-filrn rneasurernents to such cross sections as ellipses and rectangles, for which exact solutions are known, shows that the rnaxirnurn stress and the torque can be rneasured with an accuracy of 1 or 2 per cent. At the points of great stress concentration, as in the case of fillets of very srnall radii, an accuracy of the sarne order is not readily obtained. 1 100. Hydrodynamical Analogies. There are several analogies between the torsional problern and the hydrodynamical problern of the rnotion of fluid in a tube. Lord Kelvin 2 pointed out 0.-----t--X that the function [c/> 1 see Eq. (a), Art. 92] which is sornetirnes used in the solution of torsional problerns is identical with the stream function of a certain irrotational rnotion of y "ideal fluid" contained in a vessel of the sarne FIG. 167. cross section as the twisted bar. Another analogy was indicated by J. Boussinesq. 3 He showed that the di:fferential equation and the boundary condition for determining the stress function e/> (see Eqs. 142 and 144, Art. 90) are identical with those for determining velocities in a laminar motion of viscous fluid alonga tube of the sarne cross section as the twisted bar. 4 Greenhill showed that the stress function e/> is mathematically identical with the strearn function of a motion of ideal fluid circulating with uniform vorticity, 5 in a tube of the sarne cross section as the twisted bar.6 Let u and v be the components of the velocity of the circulating fluid ata point A (Fig. 167). Then from the condition of incompressibility of the ideal fluid we have
õu+õv =O õx
õy
(a)
i See the paper by C. B. Biezeno and J. M. Rademaker, De Ingenieur, No. 52, 1931. See also papers by P. A. Cushman, Trans. A.S.M.E., 1932, H. Quest, . Ingenieur-Archiv., vol. 4, p. 510, 1933, and J. H. Huth, Zoe. cit. 2 Kelvin and Tait, "Natural Philosophy," pt. 2, p. 242. a J. Boussinesq, J. math. pure et appl., series 2, vol. 16, 1871. 4 This analogy was used by M. Paschoud, Compt. rend., vol. 179, p. 451, 1924. See also Bull. tech. Suisse Rom. (Lausanne), November, 1925. · 6 The analytical expression for vorticity is the sarne as for rotation "'• discussed on p. 225, provided u and v denote the components of the velocity of ~he fl~id. . .· G A. G. Greenhill, Hydromechanics, an article in the Encyclopaedia Bntanmca, llth ed., 1910.
''l
irr lllr!
293
The condition of uniform vorticity is
õv
õu
-õx - -õy = constant
(b)
By taking U
= Õc/> ay'
V= -
Õc/>
õx
(e)
we satisfy Eq. (a), and from Eq. (b) we find
a2c1>
a2q,
2
2
ax + oy
= constant
(d)
which coincides with Eq. (142) for the stress function in torsion. At the boundary the velocity of the cir<»ulating fluid is in the direction of the tangent to the boundary and the boundary condition for the hyd~odynamical problern is the sarne as the condition (144) for the tors1?nal problem.. Hence the velocity distribution in the hydrodynam1cal problern is mathematically identical with the stress distribution in torsion, and some practically important conclusions can be drawn by using the known solutions of hydrodynamics. As a first example we take the case of a small circular hole in a twisted circular shaftl (Fig. 168). The effect of this hole on the stress distribution is similar to that of introducing a stationary solid Fw. 168. cy~inder of the sarne dia~eter as the hole into the stream of circulatmg flmd ?f the hydro~y~an.nc~l mode_I. Such a cylinder greatly changes the veloc1ty of the flmd m its immediate neighborhood. The velocities at th~ front and rear points are reduced to zero, while those at the side pomt~ m and n ~re doubled. A hole of this kind therefore doubles the shearmg stress m the portion of the shaft in which it is located A small semicircular groove on the surface parallel to the length of the shaft (Fig. 168) has the sarne effect. The shearing stress at the bottom of the groove, the point m, is about twice the shearing stress at the surface of the shaft far away from the groove. Th~ s~me hydrodynamical analogy explains the e:ffect of a small hole of elliptic cross section or of a groove of semi-elliptic cross section If ~~e 0~ the principal axes a of the small elliptical hole is in the r~dial irect10n and the other principal axis is b, the stresses at the edge of the hole at the ends of the a-axis are increased in the proportion ( 1 1
See J. Larmor, Phil. Mag., vol. 33, p. 76, 1892,
+ ~): 1.
294
THEORY OF ELASTICITY TORSION
The maximum stress produced in this case thus depends upon the magnitude of the ratio a/b. The effect of the hole on the stress is greater when the major axis of the ellipse is in the radial direction than when it runs circumferentially. This explains why a radial crack has such a weakening effect on the strength of a shaft. Similar effects on the stress distribution are produced by a semi-elliptic groove on the surface, parallel to the axis of the shaft. From the hydrodynamical analogy it can be concluded also that at the projecting corners of a cross section of a twisted bar the shearing stress becomes zero, and that at reentrant corners this stress becomes theoretically infinitely large, i.e., even the smallest torque will produce yielding of material or a crack at such a comer. ln the case of a rectangular keyway, therefore, a high stress concentration takes place at the reentrant corners at the bottom of the keyway. These high stresses can be reduced by rounding the corners. 1 101. Torsion of Hollow Shafts. So far the discussion has been limited to shafts whose cross sections are bounded by single curves. Let us consider now hollow shafts whose cross sections have two or more boundaries. The simplest problem of this kind is a hollow shaft with an inner boundary coinciding with one of the stress lines (see page 270) of the solid shaft, having the sarne boundary as the outer boundary of the hollow shaft. Take, for instance, an elliptic cross section (Fig. 149). The stress function for the solid shaft is
295
the shaft. Then, from the above conclusion regarding the direction of the sh~arin~ st~esses, it follows that there will be no stresses acting across th1s cylmdncal surface. W e can imagine the material bounded by this surface removed without changing the stress distribution in the outer portion of the shaft. Hence the stress function (a) applies to the hollow shaft also. For a given angle 8 of twist the stresses in the hollow shaft are the sarne as in the corresponding solid shaft. But the torque will be smaller ~y the amount whic~ in the case of the solid shaft is carried by the port10n of the cross sect10n corresponding to the hole. From Eq. (148) we see that the latter portion is in the ratio k4: 1 to the total torque. Hence, for the hollow shaft, instead of Eq. (148), we will have
+
Mt a 2 b2 8 = - -4- -3- 1 - k 7ra b3G
and the stress function (a) becomes
= _ >
M1 7rab(l - k 4 )
(x2a + 1)2y2 2
)
1
The formula for the maximum stress will be 2M1
T ma.x.
1
= 7rab2 1 - k4
ln the membrane analogy the middle portion of the membrane cor-
(a)
The curve x2 (ak) 2
y2
+ (bk)
2
= 1
(b)
is an ellipse which is geometrically similar to the outer boundary of the cross section. Along this ellipse the stress function (a) remains constant, and hence, for k less than unity, this ellipse is a stress line for the solid elliptic shaft. The shearing stress at any point of this line is in the direction of the tangent to the line. Imagine now a cylindrical . surface generated by this stress line with its axis parallel to the axis of , 1 The stresses at the keyway were investigated by the soap-film method. See the paper by A. A. Griffith and G. I. Taylor, Tech. Rept., Natl. Advisory Comm. · Aeronaut., vol. 3, p. 938, 1917-1918. The sarne problem was discussed by the' photoelastic method. See the paper by A. G. Solakian and G. B. Karelitz, Trana. A.S.M.E., Applied Mechanics Division, 1931.
res~onding to the hole in the shaft (Fig. 169), must be replaced b~ the horizontal plate CD. We note that the uniform pressure distributed A r---a~ B x over the portion CFD of the roem~:=t''"'.: 1 ?f1q.1~ brane is statically equivalent to the l 1 ~ : : : : - D 11 pressure of the sarne magnitude 11 uniformly distributed over the plate CD and the tensile forces S in the membrane acting along the edge of this plate are in equilibrium with the ~niform load on the plate. Hence, m the case under consideration the ;y same experimental soap-film method Frn. 169. ~~;efore can be employed because the replacement of the portion 0 fi ~ the memb~a.ne .by the plate CD causes no changes in the congurat10n and eqmhbnum conditions of the remaining portion of the membrane.
297
THEORY OF ELASTICITY
TORS_-:'ON
Let us consider now the more general case when the boundaries of the holes are no longer stress lines of the solid shaft. From the general theory of torsion we know (see Art. 90) that the stress function must be constant along each boundary, but these constants cannot be chosen arbitrarily. ln discussing multiply-connected boundaries in twodimensional problems it was shown that recourse must be had to the expressions for the displacements, and the constants of integration should be found in such a manner as to make these expressions singlevalued. An analogous procedure is necessary in dealing with the torsion of hollow shafts. The constant values of the stress function along the boundaries should be determined in such a manner as to make the displacements single-valued. A sufficient number of equations for determining these constants will then be obtained. From Eqs. (b) and (d) of Art. 90 we have
A physical significance for Eq. (e) was discussed before [see Eq. (152), page 271]. It indicates that in using the membrane analogy the level of each plate, such as the plate CD (Fig. 169), must be taken so that the vertical load on the plate is equal and opposite to the vertical component of the resultant of the tensile forces on the plate produced by the membrane. If the boundaries of the holes coincide with the stress lines of the corresponding solid shaft, the above condition is sufficient to ensure the equilibrium of the plates. ln the general case this condition is not sufficient, and to keep the plates in equilibrium in a horizontal position special guiding devices become necessary. This makes the soap-film experiments for hollow shafts more complicated.
296
(e)
Let us now calculate the integral (d)
fr ds along each boundary. components we find
f
T
ds =
Using (e) and resolving the total stress into its
J ~: + ~;) J(~: + :;
= G
(Txz
Tyz
dx
ds
dy) - 8G
J
(y dx - x dy)
To remove this difficulty the following procedure may be adopted.1 We make a hole in the plate (Fig. 165) corresponding to the outer boundary of the shaft. The interior boundaries, corresponding to the holes, are mounted each on a vertical sliding column so that their heights can be easily adjusted. Taking these heights arbitrarily and stretching the film over the boundaries we obtain a surface which satisfies Eq. (142) and boundary conditions (144), but the Eq. (e) above generally will not be satisfied and the film does not represent the stress distribution in the hollow shaft. Repeating such an experiment as many times as the number of boundaries, each time with another adjustment of heights of the interior boundaries and taking measurements on the film each time, we obtain sufficient data for determining the correct values of the heights of the interior boundaries and can finally stretch the soap film in the required manner. This can be proved as follows: If i is the number of boundaries and 4'1, 4'2, ••• , cf>; are the film surfaces obtai?ed with i different adjustments of the heights of the boundaries, then a funct1on (f) e/> = m1c/>1 + m2c/>2 + · · · + m;cf>;
m.,
in which m1, provided that
, m; are numerical factors, is also a solution of Eq. (142),
m1
+ m2 + · · · + m;
= 1
The first integral must vanish, from the condition that the integration is taken round a closed curve and that w is a single-valued function. Hence, fr ds = 8Gf(x dy - y dx)
Observing now that the shearing stress is equal to the slope of the membrane and substituting (f) into Eqs. (e), we obtain i equations of the forro '
The integral on the right side is equal to double the area of the hole. Then (e) fr ds = 2G8A
from which the i-factors m1, m2, • • . , m; can be obtained as functions of O. Then _the true stress function is obtained from (f).2 This method was applied by Griffith and Taylor in determining stresses in a hollow circular shaft havmg a
Thus we must determine the constant values of the stress function along the boundaries of the holes so as to satisfy Eq. (e) for each boundary. This equation is also valid for any closed curve drawn in the cross section, as may be seen by reexamining the proof.
f ~:
ds = 2GoA;
1 A. A. Griffith, and G. I. Taylor, Tech. Rept. Natl. Advisory Comm. Aeronaut. Vol. 3, p. 938, 1917-1918. ' 2 Griffith and Taylor concluded from their experiments that instead of constantpressure films it is more convenient to use zero-pressure films (see p. 272) in studying the stress distribution in hollow shafts. A detailed discussion of the calculation of factors m1, m2, ••• is given in their paper.
298
THEORY OF ELASTICITY
TORSION
keyway in it. It was shown in this manner that the maximum stress can be considerably reduced and the strength of the shaft increased by throwing the bore in the shaft off center.
The torque in the shaft with one or more boles is obtained using twice the volume under the membrane and the flat plates. To see this we calculate the torque produced by the shearing stresses distributed over an elemental ring between two adjacent stress lines, as in Fig. 169 (now taken to represent an arbitrary hollow section). Denoting by othe variable width of the ring and considering an element such as that shaded in the figure, the shearing force acting on this element is To ds and its moment with respect to O is rTo ds. Then the torque on the elemental ring is
dM 1 = f TTo ds
It is inversely proportional to the thickness of the wall and thus greatest where the thickness of the tube is least. To establish the relation between the stress and the torque M 1 we apply again the membrane analogy and calculate the torque from the volume ACDB. Then Mt = 2Ah = 2AoT (b) in which A is the mean of the areas enclosed by the outer and the inner boundaries of the cross section of the tube. From (b) we obtain a simple formula for calculating shearing stresses, (166)
(e)
in which the integration must be extended over the length of the ring. Denoting by A the area bounded by the ring and observing that T is the slope, so that To is the difference in levei h of the two adjacent conAjoe;;::: 1 X tour lines, we find, from (e), j' jC z dMt = 2hA (d)
For determining the angle of twist 8, we apply Eq. (152).
fl
Dílf
·~-1~·~
299
T
ds
M1J Bds = 2G8A
= 2A
(e)
from which 1
; 1
b
i.e., the torque corresponding to the
elemental ring is given by twice the volume shaded in the figure. :Y The total torque is given by the Fm. 170. sum of these volumes, i.e., twice the volume between AB, the membrane AC and DB, and the flat plate CD. The conclusion follows similarly for severa! holes. 102. Torsion of Thin Tubes. An approximate solution of the torsional problem for thin tubes can easily be obtained by using the membrane analogy. Let AB and CD (Fig. 170) represent the leveis of the outer and the inner boundaries, and AC and DB be the cross section of the membrane stretched between these boundaries. ln the case of a. thin wall, we can neglect the variation in the slope of the membrane across the thickness and assume that AC and BD are straight lines. This is equivalent to the assumption that the shearing stresses are uniformly distributed over the thickness of the wall. Then denoting by h the difference in levei of the two boundaries and by o the variable thickness of the wall, the stress at any point, given by the slope of the membrane, is h 1' = õ
r
Then
(167)
_,, --ô
l1-'-=========::::::...J i-1,---c----..l
ln the case of a tu be of uniform thickness ' constant and (167) gives
o .is
fa}
ô
(168) (h)
;n which s is the length of the center line of the ring section of the tube.
'
Fm. 171.
If the tu~e has reentrant comers, as in the case represented in Fig. 171, a cons1derable stress concentration may take place at these corners. The maximum stress is larger than the stress given by Eq. (166)
and depends on the radius a of the fillet of the reentrant comer (Fig. l 71b). ln calculating this maximum stress we shall use the membrane analogy as we did for the reentrant comers of rolled sections (Art 98) · of the membrane at the reentrant comer may be taken · · .The equat10n in the form d 2z
dr 2 1
+ _! dz
_ _ q
r dr -
S
E t" V D qua ions (166) and (167) for thin tubular sections were obtained by R. Bredt · .I., vol. 40, p. 815, 1896. '
301
THEORY OF ELASTICITY
TORSION
Replacing q/S by 2G8 and noting that r = -dz/dr (see Fig. 170), we find (d)
Substituting r = a we obtain the stress at the reentrant comer. This is plotted in Fig. 172. The other curve 1 (A in Fig. 172) was obtained by the method of finite differences, without the assumption that the membrane at the comer has the form of a surface of revolution. It confirms the accuracy of Eq. (i) for small fillets-say up to a/ô = i. For larger fillets the values given by Eq. (i) are too high. 3.5 Let us consider now the case when the cross section of a tubular member has more than two bound3.0 1--t~-+---+---+-------l aries. Taking, for example, the case shown in Fig. 173 and assuming that the thickness of the wall is
300
Assuming that we have a tube of a constant thickness ô and denoting by ro the stress at a considerable distance from the comer calculated from Eq. (166), we find, from (e), 2G8 = ros A
Substituting in (d), (e)
'Cmax
~
2.0 1--~~---+---+-------l
The general solution of this equation is
r =
Q + rosr r
.....
(f)
2A
11 1 1
Assuming that the projecting angles of the cross section have fillets with the radius a, as indicated in the figure, the constant of integration e can be determined from the equation ra+õ }a T dr
1.o oL__ _Jo.L.s---'-1.o=---1...L.s==-..12.o a/
~~'--------li e
C::----'----r'
.D Frn. 173.
(g) . .
very small, the shearing stresses in each portion of the wall, from the membrane analogy, are
which follows from the hydrodynamical analogy (Art. 100), viz.: if an ideal fiuid circulates in a channel having the shape of the ring cross section of the tubular member, the quantity of fiuid passing each cross section of the channel must remain constant. Substituting expression . (f) for r into Eq. (g), and integrating, we find that
(k)
e=
T 0
=
ToÔ
ô 1 - (s/4A)(2a +ô) log. (1 +ô/a)
in which h 1 and h 2are the levels of the inner boundaries CD and EF.2 The magnitude of the torque, determined by the volume ACDEFB
.IS
and, from Eq. (f), that T
= roô 1 - (s/4A)(2a +ô) + r 0sr r
log. (1 +ô/a)
For a thin-walled tube the ratios s(2a and (h) reduces to
+
(l)
(h):
2A
ô)/A, sr/A, will be small,
'
where A1 and A 2 are areas indicated in the figure by dotted lines. Further equations for the solution of the problem are obtained by applying Eq. (152) to the closed curves indicated in the figure by 1
!
Huth, loc. cit. lt is assumed that the plates are guided so as to remain horizontal (see p. 297).
302
THEORY OF ELASTICITY
dotied lines. Assuming that the thicknesses ô1, ô2, 03 are constant and denoting by 8 1, 8 2, 8 3 the lengths of corresponding dotted curves, we find, from Fig. 173, T181+1"383 = T282 -
1"383
2G8A1
= 2G8A2
(m)
+ 02Ôa81A22 + Ô1Ô28a(A1 + A2) 2] M1[ôa81A2 + 018a(A1 + A2)] 2 2[01Ôa82A1 2 + 02Ôa81A2 2 + 01Ô28a(A1 + A2) ] M1(ô1s2A1 -
02s1A2)
obtained above for rectangular bars (Art. 95). ln discussing these stresses, let us consider first the case of a very narrow rectangle 1 and assume that the dimension a is large in comparison with b. If cross sections are free to warp, the stresses, from Art. 94, are (a) Tz• = -2GOy, Ty• = 0 and the corresponding displacements, from Eqs. (a), (b), and (d) of Art. !lO, are
By using the last of the Eqs. (k) and Eqs. (l) and (m), we find the stresses T1, T2, Ta as functions of the torque: 2[ô1Ô382A1 2
303
TORSION
(n) (o)
(p)
u = -(Jyz,
v = (Jxz,
w = -8xy
(b)
To prevent the warping of the cross sections, designated as displacement w, normal stresses "• must be distributed over the cross sections. We obtain an approximate solution by assuming that "• is proportional to w and that it diminishes with increase of distance z from the middle cross section. These assumptions are satisfied by taking
2P
(e)
ln the case of a symmetrical cross section, 81 = 82, 01 = ô2, Ai = A2, and Ta = O. ln this case the torque is taken by the outer wall of the tube, and the web remains unstressed. 1 To get the twist for any section like that shown in Fig. 173, one substitutes the values of the stresses in one of the Eqs. (m). Thus 8 can be obtained as a function of the torque M t· 103. Torsion of a Bar in Which One Cross Section Remains Plane. ln discussing torsional problems it has always been assumed that the torque is applied by means of shearing stresses distributed over the ends of a bar in a definite manner, obtained from the solution of Eq. (142) and satisfying the boundary condition (144). If the distribution of stresses at the ends is different from this, a local irregularity instress distribution resulta and the solution of Eqs. (142) and (144) can be applied with satisfactory accuracy only in regions at some distance from the ends of the bar. 2 A similar irregularity occurs if a cross section of a twisted bar is prevented from warping by some constraint. We encounter problems of this kind occasionally in engineering.ª A simple example is shown in Fig. 174. From symmetry it can be concluded that ·the middle cross section of the bar remains plane during torsion. Hence the stress distribution near this cross section must be different from that 1 The small stresses corresponding to the change in slope of the membrane across the thickness of the web are neglected in this derivation. 2 The local irregularities at the ends of a circular cylinder have been discussed by F. Purser, Proc. Roy. Irish Acad., Dublin, vol. 26, series A, p. 54, 1906.. See also K. Wolf, Sitzber. Akad. Wiss., Wien, vol. 125, p. 1149, 1916, and A. Tlmpe, Math. Annalen, vol. 71, p. 480, 1912. a Torsion of I-beams under such conditions was discussed by S. Timoshenko, z. Math. Physik, vol. 58, p. 361, 1910. See also C. Weber, Z. angew. Math. Mech., -vol. 6, p. 85, 1926.
ji I'
lllllL
in which m is a factor to be determined later. Due to the factor e-m• the stress "• diminishes with increase of z and 2P becomes negligible within a certain distance depending upon the magnitude of m. The remaining stress components must now be chosen in such a manner as to z satisfy the differential equations of FIG. 174. equilibrium (127) and the boundary conditions. It is easy to prove that these requirements are satisfi.ed by taking
"• = "11 = o
r, 11 = -1;Em3 8e-m•(a 2
r., = iEm 2 8e-m•(a 2 r 11, = iEm 28e-mz(b 2
-
x 2)(b2 - y•) x 2 )y - 2G8y y 2)x
-
(d)
For large values of z this stress distribution approaches the stresses (a) for simple torsion. The stress component r, 11 becomes zero at the boundary x = ±a and Y = ±b; r,. and r 11, are zero for x = ±a and y = ±b, respectively. Hence the boundary conditions are satisfied and the lateral surface of the bar is free from forces. For determining the factor m, we consider the strain energy of the bar and calculatem to make this energy a minimum. By using Eq. (84) on page 148, we fi.nd 1 1·._l V = 2G 1
f f a -a
b [ Tzy 2 -b
+ T,, 2 + r
11,
2
+ Z(l 1+ v) •
2
] dx dy dz
Substituting from (d), and noting that for a long barwe can with sufficient accuracy put (le-mzdz =
}o 1
!
m
&e S. Timoshenko, Proc. London M ath. Soe., series 2, vol. 20, p. 389, 1921
304
THEORY OF ELASTICITY
TORSION
12 V = 1- E0 2a 3b3 { -3m + (1 + 11) [ -2 a 2b2m5 + -1 (a2 + b2)mª + -l ]} (e) 9 25 5 (1 + 11)2 a2
the z-axis and use polar coordinates r and 8 for defining the position of an element in the plane of a cross section. The notations for stress components in such a case are
we get
The minimum condition gives us the following equation for determining m:
305
which, for a narrow rectangle, reduces approximately to
m2 =
(1
5
(f)
+ 11)a2
Substituting this value of m in (e) and (d), we find the stress distribution for the case when the middle cross section of the bar remains plane. For calculating the angle of twist f, we put the potential energy (e) equal to the work done by the torque Mt, Mtf =V 2
from which the angle of twist is
f
=
3M, 16Glab 3
[z _ v5(1 + 6
11)
ª
J
(g)
z
Comparing this result with Eq. (155) on page 273, we conclude that by preventing the middle cross section from warping we increase the rigidity of the bar with respect to torsion. The effect of the local irregularity in stress distribution on the value of f is the sarne as the influence of a diminution of the length l by
a
v5<1+11)
6
Taking 11 = 0.30, this reduction in l becomes 0.425a. We see that the effect of the constraint of the middle cross section on the angle of twist is small if the dimension a is small in comparison with l. The twist of a bar of an elliptic cross section can be discussed in an analogous manner. 1 Of greater effect is the constraint of the middle cross section in the case of torsion of a bar of I cross section. An approximate method for calculating the angle of twist in this case is obtained by considering bending of the flanges during torsion. 2
104. Torsion of Circular Shafts of Variable Diameter. Let us consider a shaft in the form of a body of revolution twisted by couples applied at the ends (Fig. 175). We may take the axis of the shaft as 1 A. Fõppl, Sitzber. Bayer. Akad. Wiss., Math.-phys. Klasse, München, 1920, p. 261. 2 See S. Timoshenko, Z. Math. Physik., vol. 58, p. 361, 1910; or Strength of Materials, vol. 2, p. 2871 1941,
,j
Fw. 175.
formulas obtained previously for two-dimensional problems (Art. 28), we find the following expressions for the strain components: Er
'Yr8
au ar au av V = -r ae + -ar - -,r
= -,
Eq
av + r--, ae au aw - +-, az ar u r
= -
'Yrz =
aw
Ez
'Yz8
= az =
av + aw az r ae
(169)
Writing down the equations of equilibrium of an element (Fig. 175), as was done before for the case of two-dimensional problems (Art. 25),
THEORY OF ELASTICITY
TORSION
and assuming that there are no body forces, we arrive at the following differential equations of equilibrium: 1
To satisfy the compatibility conditions it is necessary to consider the fact that r,o and ro, are functions of the displacement v. From Eqs. (a) and (d) we find
306
aur ar
+!
arr8 + arrz + ~ = O r ae az r arrz + ! aro. + au. + Trz = o ar r ae az r arro + ! auo + arez + 2Tr8 = O ar r ae az r
Tr8 = G'Yr8 = G (170)
Hence, of all the stress components, only rro and ro. are different from zero. The first two of Eqs. (170) are identically satisfied, and the third of these equations gives (b)
This equation can be written in the form
a
+ aza (r ro.) = o
(e)
2
aq, r 2r r o= - -az,
T21 aq, az
ª"'
(e)
_ - 0
(f)
Toz = G'Yoz = G av = Gr !__ (~) .!_ az az r - r2 ar
(1-
!__ 3 aq,) ar r ar
e{>
of r (d)
i These equations were obtained by Lamé and Clapeyron; see Crelle's J., vol. 7, 1831.
+~ (1- aq,) az r az 3
or (g)
Let us consider now the boundary conditions for the function e{>. From the condition that the lateral surface of the shaft is free from externai forces we conclude that at any point A at the boundary of an axial section (Fig. 175) the total shearing stress must be in the direction of the tangent to the boundary and its projection on the normal N to the boundary must be zero. Hence
dz dr Tro ds - T8z ds
=
O
where ds is an element of the boundary. Substituting from (d), we find that aq,dz + aq,dr =o (h) az ds ar ds from which we conclude that e{> is constant along the boundary of the axial section of the shaft. Equation (g) together with the boundary condition (h) completely determines the stress function e{>, from which we may obtain the stresses satisfying the equations of equilibrium, the compatibility equations, and the condition at the lateral surface of the shaft. 1 1
It is seen that this equation is satisfied by using a stress function and z, such that
(ªVar - !:'r .) = Gr !__ar (~)r = -
From these equations it follows that
ln the application of these equations to the torsional problem we use the semi-inverse method (see page 259) and assume that u and w are zero, i.e., that during twist the particles move only in tangential directions. This assumption differs from that for a circular shaft of constant diameter in that these tangential displacements are no longer proportional to the distance from the axis, i.e., the radii of a cross section become curved during twist. ln the following pages it will be shown that the solution obtained on the basis of such an assumption satisfies all the equations of elasticity and therefore represents the true solution of the problem. Substituting in (169) u = w = O, and taking into account the fact that from symmetry the displacement v does not depend on the angle 8, we find that av av V (a) 'Yr8 = - - -1 "/Oz = az Er = E9 = E, = 'Yrz = O, ar r
ar (r 2rro)
307
This general solution of the problem is dueto J. H. Michell, Proc. London M ath. S~c., vol. 31, p. 141, 1899. See also A. Fõppl, Sitzber. Bayer. Akad. Wiss., Münc en, vol. 35, pp. 249 and 504, 1905. Also the book "Kerbspannungslehre" by H. _Ne~ber, which gives solutions for the hyperboloid of revolution, and for a cavity m the forro of an ellipsoid of revolution by a different method. Reviews ' of Mecthe i·iterature on the subject have ?ee_n given b~ T. Põschl, z. angew. Math. h., vol. 2, p. 137, 1922, and T. J. Higgms, Experimental Stress Analysis vol. 3 no. 1, p. 94, 1945. ' '
308
THEORY OF ELASTICITY
TORSION
The magnitude of the torque is obtained by taking a cross section and calculating the moment given by the shearing stresses ro.. Then
[ª 21C'r2re. dr = 27f' }o[ª ª
(k)
where a is the outer radius of the cross section. The torque is thus easily obtained ü we know the difference between the values of the stress function at the outer boundary and at the center of the cross section. ln discussing displacements during twist of the shaft let us use the notation i/; = v/r for the angle of rotation of an elemental ring of radius r in a cross section of the shaft. This ring can be considered as the cross section of one of a number of thin elemental tubes into which the shaft is subdivided. Then i/; is the angle of twist of such a tube. From the fact that the radii of the cross section become curved, it follows that i/; varies with r and the angles of twist of elemental tubes are not equal for the sarne cross section of the shaft. Equations (e) can now be written in the form
meridian, and the corresponding shearing stress in the axial section of the shaft has the direction normal to the meridian. At the boundary this stress is tangent to the boundary and the meridians are normal to the boundary of the axial section. If we go from the surface 1/t = constant to an adjacent surface the rate of change of if; along the boundary of the axial section of the shaft is dif; / ds, and in the sarne manner as for a cylindrical shaft of circular cross section (Art. 87) we have di/; r = Grds
where
dr
= rro-
r
ds
3 ) + ~ (r 3ai/;) ~ ar (r ª"' ar é)z é)z
=O
or
or 2
+ ~ oi/! + a2i/; r ar
oz 2
+ ro.ds
=
o
r
/\ I 1 ( 1 I li 1
z.
+ z2)l
Frn. 176.
is constant at the boundary of the axial section and equal to cos a. Any function of this ratio will satisfy the boundary condition (h). ln order to satisfy also Eq. (g) we take
ar
from which
a2i/;
dz
z
(r2
az
Gr 3 oi/; = ª
az
(n)
is the resultant shearing stress at the boundary. It is seen that the value of this shearing stress is easily obtained if we find by experiment the values of dif;/ds.1 Let us consider now a particular case of a conical shaft2 (Fig. 176). ln this case the ratio
Grª ai/; = - acp
ar
309
o
(l)
A solution of this equation gives us the angle of twist as a function of r and z. If we put (m) i/; = constant in this solution, we obtain a surface in which all the points have the sarne angle of twist. ln Fig. 175, AA1 represents the intersection of such a surface with the axial section of the shaft. From symmetry it follows that the surfaces given by Eq. (m) are surfaces of revolution and AA 1is a meridian of the surface going through the point A. During twist these surfaces rotate about the z-axis without any distortion, exactly in the sarne manner as the plane cross sections in the case of circular cylindrical shafts. Hence the total strain at any point of the meridian AA 1 is pure shearing strain in the plane perpendicular to the
= e { (r2
where e is a constant.
~ z2)t - ~ [ (r2 ~ z2)t T}
(o)
Then by differentiation we find
1 ocp crz rez = - - = - ~-~-,-r2 ar (r 2 + z2 )1
(p)
The constant e is obtained from E q. (k) . Su b st"t i u t"mg (o) m t h"is equation we find
Mt
e=_
27r(-i - cos a
+t
cos 3 a)
~o calculate the angle of twist we use Eqs. (e), from which the expression for i/;, satisfying Eq. (l) and the boundary condition, is e i/; = 3G(r 2 1
+ z )f 2
(q)
Such experiments were made by R. Sonntag Z. angew. Math Mech vol 9 p. 1, 1929. , . ., . . 2
See Fõppl, Zoe. cit.
THEORY OF ELASTICITY
TORSION
It will be seen that the surfaces of equal angle of twist are spherical surlaces with their center at the origin O. The case of a shaft in the form of an ellipsoid, hyperboloid, or paraboloid of revolution can be discussed in an analogous manner. 1 The problems encountered in practice are of a more complicated nature. The diameter of the shaft usually changes abruptly, as shown in Fig. 177a. The first investigation of such problems was made b! A. Fõppl. C. Runge suggested a numerical method for the approx1mate solution of these problems, 2 and it was shown that considerable stress concentration takes place at such points as m and n, and that the magnid tude of the maximum stress for a shaft n a of two different diameters d and D (Fig. 177a) depends on the ratio of the radius a of the fillet to the diameter d D of the shaft and on the ratio d/D. ln the case of a semicircular groove (6) of very small radius a, the maximum faJ FIG. 177 · stress at the bottom of the groove (Fig. 177b) is twice as great as at the surface of the cylindrical shaft without the groove. ln discussing stress concentration at the fillets and grooves of twisted circular shafts, an electrical analogy has proved very useful. 3 The general equation for the flow of an electric current in a thin homogeneous plate of variable thickness is
Let us assume that the plate has the sarne boundary as the axial section of the shaft (Fig. 178), that the x- and y-axes coincide with the z- and r-axes, and that the thickness of the plate is proportional to the cube of the radial distance r, so that h = ar 3• Then Eq. (r) becomes
310
i_
ax
(h
àif;) àx
+ i_ (h ªifi) ày ày
=
o
(r)
in which h is the variable thickness of the plate and if; the potential function. 1 See papers by E. Melan, Tech. Blãtter, Prag, 1920; A. N. Dinnik, Bull. D~ Polytech. Inst., Novotcherkask, 1912; W. Arndt, Die To~sion v_on We~e1:1 m1t achsensymmetrischen Bohrungen und Hohlrãumen, D1ssertat10n, G?ttm?en, 1916; A. Timpe, Math. Annalen, 1911, p. 480. Further references are g1ven ma review by Higgins, loc. cit. . 2 See F. A. Willers, Z. Math. Physik, voI. 55, p. 225, 1907. Artother approximate method was developed by L. Fõppl, Sitzber. Bayer. Akad. Wiss., München, vol. 51, p. 61, 1921, and by R. Sonntag, Z. angew. Math. Mech., voI. 9, p. 1, 1929. a See paper by L. S. Jacobsen, Trans. A.S.M.E., vol. 47, p. 619, 1925, and. the survey given by T. J. Higgins, loc. cit. Discrepancies between results obtamed from this and other methods are discussed in the latter paper. For further comparisons and strain-gauge measurements extending Fig. 179 to 2a/d = 0.50 see A. Weigand, Luftfahrt-Forsch., vol. 20, p. 217, 1943, translated in N.A.C.A. Tech. Mem. 1179, September, 1947.
a2y; az 2
+ ~ ay; + a2y; r ar
311
=o
éJr 2
This coincides with equation (l), and we conclude that the equipotential lines of the plate are determined by the sarne equation as the lines of equal angles of twist in the case of a shaft of variable diameter. Assuming that the ends of the plate, corresponding to the ends of the shaft, are maintained at a certain difference of potential so that the current flows along the z-axis, the equipotential lines are normal to the
FIG. 178.
lateral sides of the plate, i.e., we have the sarne boundary conditions as for lines of constant angle of twist. If the differential equations and the boundary conditions are the sarne for these two kinds of lines, the lines are identical. Hence, by investigating the distribution of potential in the plate, valuable information regarding the stress distribution in the twisted shaft can be obtained. The maximum stress is at the surface of the shaft and we obtain this stress by using Eq. (n). From this equation, by applying the electrical analogy, it follows that the stress is proportional to the rate of drop of potential along the edge of the plate. Actual measurements were made on a steel model 24 in. long by 6 in. wide at the larger end and 1 in. maximum thickness (Fig. 178). The drop of potential along the edge mnpq of the model was investígated by using a sensitive galvanometer, the terminals of which were connected to two sharp needles fastened in a block at a distance 2 mm. apart. By touching the plate with the needles the drop in potential over the distance between the needle points was indicated by the galvanometer. By moving the needles along the fillet it is possible to find the place of, and measure, the maximum voltage gradient. The ratio of this maximum to the voltage gradient ata remote point m (Fig. 178a) gives
312
THEORY OF ELASTICITY
TORSION
the magnitude of the factor of stress concentration k* in the equation
of stress concentration for any particular case can be found with sufficient accuracy.
'l'max.
= k
l6M1
Problems
1rd3
The results of such tests in one particular case are represented in Fig. 178c, in which the potential drop measured at each point is indicated 9
3.2 2.8
2.4
2.0
1.6
~
,\ i\' 1{l=zoo
'\
\' ~
\,,'il
/
~
r---...
!J=LSO.
/ 'f}=!.JJ ~ ~ ,q=l.20
'~
1.2
1. Show by considering the equilibrium of the whole bar that when ali stress components vanish except T%•> Tu•, the loading must consist of torsional couples only [cf. Eqs. (h), Art. 90]. 2. Show that > = A (r 2 - a 2 ) solves the torsion problem for the solid or hollow circular shaft. Determine A in terms of GO. Using Eqs. (141) and (145) evaluate the maximum shearing stress and the torsional rigidity in terms of M, for the solid shaft, and verify that the results are in agreement with those given in any text on strength of materiais. 3. Show that for the sarne twist, the elliptic section has a greater shearing stress than the inscribed circular section (radius equal to the minor axis b of the ellipse). Which takes the greater torque for the sarne allowable stress? 4. Use Eq. (g) of Art. 92 and Eq. (145) to evaluate the torsional rigidity of the equilateral triangle, and thus verify Eq. (l), Art. 92. 5. Using the stress function (m) of Art. 92 expressed in rectangular coordinates, fi.nd an expression for Tu• along the middle line Ax of Fig. 153, and verify that the greatest value along this line is the value given by Eq. (p). 6. Evaluate the torsional rigidity of the section shown in Fig. 153. Is it appreciably different from that of the complete circular section when the groove is small? y 7. Show that the expression for the stress function > which corresponds to the parabolic membrane of Art. 94 is
n
-r- ~ ~· j~=l.09-~
f'... r--o..
0.8
q,
0.4
OD4
0.08
0.12 2a
0.16
313
o.zo
0.24
d
* Small variations in radius r {Eq. (n)] can be neglected in this case.
-Go ( x2 -
1
~)
ln a narrow tapered section such as the triangle y shown in Fig. 180, an approximate solution can be Frn. 180. FIG. 181. obtained by assuming that at any levei y the membrane has the parabolic form appropriate to the width at that level. Prove that for the triangular section of height b approximately.
FIG. 179.
by the length on the normal to the edge of the plate at this point. From this figure the factor of stress concentration is found to be 1.54. The magnitudes of this factor obtained with various proportions of shafts are given in Fig. 179, in which the abscissas represent the ratios 2a/d of the radius of the fillet to the radius of the smaller shaft and the ordinates the factor of stress concentration k for various values of the ratio D/d (see Fig. 177). By interpolating from these curves the factor
=
M1
=
f#obc. 3
8•. Using the method indicated in Prob. 7, find an approximate expression for the
~rs~onal rigidity of the thin symmetrical section bounded by two parabolas shown lll
Fig. 181, for which the width e ata depth y below the center is given by
e=
Co
(1 - ~)
9•. Show that the method indicated in Prob. 7 gives for a slender elliptical sect1on the approximate stress function ti> = -Gob2
e::: + a•
y2 b•
1)
314
THEORY OF ELASTICITY
the ellipse being that of Fig. 149 with b/a small. Art. 91 approaches this as b/a is made small. Derive the approximate formulas Mt = trab'GO,
TORSION
Show that the exact solution of 2M1
Tmax,
18. At any point of an axial section of a shaft of variable diameter, line elements ds and dn (at right angles) in the section are chosen arbitrarily as shown in Fig. 184. The shear stress is expressed by components Ta, r,. along these. Show that
= 2GOb = trab 2
1 éJcp
for the slender elliptical section, and compare with the corresponding formulas for the thin rectangular section of length 2a and thickness 2b. 10. Apply the method given at the end of Art. 97 to find an approximation to the torsional rigidity of the section described in Prob. 8. 11. A section has a single hole, and the stress function cf> is determined so that it vanishes on the outside boundary and has a constant value c/>H on the boundary of the hole. By adapting the calculation indicated on page 262 for Eq. (145), prove that the total torque is given by twice the volume under the cp-surface plus twice the volume under a flat roof at height cf>H covering the hole (cf. page 298). 12. A closed thin-walled tube has a perimeter l and a uniform wall thickness li. An open tube is made by making a fine longitudinal cut in it. Show that when the maximum shear stress is the sarne in both closed and open tubes, Mtopen Mtclosed
=
/li
Oopen
6A'
Oc!osed
2A =
satisfies Eq. (g) of Art. 104 only if the constant A is
-! [cf. Eq. (o)].
Ta
=Ti éJn'
Tn
= GréJif;
éJn
and deduce the boundary condition satisfied by Y,. Show without calculation that the function given by Eq. (q) of Art. 104 satisfies this boundary condition for a conical boundary of any angle. 19. Verify that Eq. (q) of Art. 104 gives correctly the function Y, corresponding to the function cf> in Eq. (o). z 20. If the theory of Art. 104 is modified by discarding the FIG. 184. boundary condition cf> = constant the stress will be due to certain "rings of shear" on the boundary, as well as end torques. Considering the uniform circular shaft, describe the problem solved by cf> = Czr 4 where C is a constant, for O < z < l. 21. Prove that the relative rotation of the ends of the (conical) tapered shaft shown in Fig. 185 due to torque M1 is
U
and that the ratio of the torsional rigidities is /2li2 /12A 2, A being the area of the "hole." Evaluate these ratios for a circular tube of FIG. l8Z. 1 in. radius, T\ in. wall thickness. 13. A thin-walled tube has the cross section shown in Fig. 182, with uniform wall thickness li. Show that there will be no stress in the central web when the tube is twisted. Find formulas for (a) the shear stress in the walls, away from the corners, (b) the unit twist o, in terms of the torque. 14. Find expressions for the shear stresses in a tube of the section shown in Fig. 183, the wall thickness li being uniform. 15. ln discussing thin-walled closed sections, it was assumed that the shear stress is constant across the wall thickness, corresponding to constant membrane slope across the thickness. Show that this cannot be strictly true for a straight part of the wall (e.g., Fig. 171a) and that in general the correcFIG. 183. tion to this shear stress consists of the shear stress in a tube made "open" by longitudinal cuts (cf. Prob. 12). 16. The theory of Art. 104 includes the uniform circular shaft as a special case. What are the corresponding forms for the functions cf> and Y,? Show that these functions give the correct relation between torque and unit twist. 17. Prove that z Az3 )i cf> = R + R,3 where R = (r 2 + z2 ~
315
M 2tr(f - cos a
+!
1(1 b31)
cos 3 a) • 3G ;ã -
If a and b are both made large, with b - a = l, and a is made small, the above result should approach the relative rotation of the ends of a uniform shaft of length Z, and radius aa, dueto torque M 1• Show that it does so.
Fm. 185.
Fm. 186.
22. Use the functions given by Eqs. (o) and (q) of Art. 104 to find, in terms of M,, the relative rotation of the ends of the hollow conical shaft shown in Fig. 186. The ends are spherical surfaces of radii a, b, center O.
.,, 1
317
BENDING OF PRISMATICAL BARS
CHAPTER 12 BENDING OF PRISMATICAL BARS 106. Bending of a Cantilever. ln discussing pure bending (Art. 88) it was shown that, ü a prismatical bar is bent in one of its principal planes by two equal and opposite couples applied at the ends, the deflection occurs in the sarne plane, and of the six components of stress only the normal stress parallel to the axis of the bar is different from zero. This stress is proportional to the distance from the neutral axis. Thus the exact solution coincides in this case with the elementary theory of bending. ln discussing bending of a cantilever of narrow rectangular cross section by a force applied at the end (Art. 20), it was
FIG. 187.
shown that in addition to normal stresses, proportional in each cross section to the bending moment, there will act also shearing stresses proportional to the shearing force. Consider now a more general case of bending of a cantilever of a constant cross section of any shape by a force P applied at the end and parallel to one of the principal axes of the cross section 1 (Fig. 187). Take the origin of the coordinates at the centroid of the fixed end. The z-axis coincides with the center line of the bar, and the x- and y-axes coincide with the principal axes of the cross section. ln the solution of the problem we apply Saint-Venant's semi-inverse method and at the very beginning make certain assumptions regarding stresses. We assume that normal stresses over a cross section at a distance z from the fixed end are distributed in the sarne manner as in the case of pure bending: P(l - z)x (a) "· = I i This problem was solved by Saint-Venant, J. mathémat. (I.tiouville), series 2, vol. 1, 1856. 316
'
We assume also that there are shearing stresses, acting on the sarne cross sections, which we resolve at each point into components r,,, and Tyz· We assume that the remaining three stress components"z, " 111 r zy are zero. It will now be shown that by using these assumptions we arrive at a solution which satisfies all of the equations of the theory of elasticity and which is hence the exact solution of the problem. With these assumptions, neglecting body forces, the differential equations of equilibrium (127) become ih,,. iJz iJr,,,
Tx
=o
+
=O
dTy,
' iJry, _
az
(b)
Px
(e)
Ty--7
From (b) we conclude that shearing stresses do not depend on z and are the sarne in all cross sections of the bar. Considering now the boundary conditions (128) and applying them to the lateral surface of the bar, which is free from external forces we find that the first two of these equations are identically satisfied 'and the third one gives T,,.l + Tyzm = 0 From Fig. 187b we see that l
= cos
dx
(Nx) = dy
m = cos (Ny) = - ds
ds'
in which ds is an element of the bounding curve of the cross section. Then the condition at the boundary is T:z;z
dy ds -
T 11z
dx ds = O
(d)
Turning to the compatibility equations (130), we see that the first three of these equations, containing normal stress components, and the last equation, containing Tru, are identically satisfied. The system (130) then reduces to the two equations
V2Tyz =
o1
~ V T
-
""
-
-
p
~-~
J(l
+ v)
(e)
Thus the solution of the problem of bending of a prismatical cantilever ofanY cross sect10n · red uces to findmg, · for ""'• and ,,11., functions of x and ?.which satisfy the equation of equilibrium (e), the boundary condition (d), and the compatibility equations (e).
BENDING OF PRISMATICAL BARS
THEORY OF ELASTICITY
318
106. Stress Function. ln discussing the bending problems we shall again make use of a stress function q,(x,y). It is easy to see ~hat the differential equations of equilibrium (b) and (e) of the previous 1trticle are satisfied by taking aq, Px 2 - aq, (171) Tyz = Txz = ày - 2[ + f(y), àx in which q, is the stress function of x and y, and f(y) is a function of Y only, which will be determined later from the boundary condition: Substituting (171) in the compatibility equations (e) of the prev1ous article, we obtain 2 a q, a q,) àx ox 2 + ày 2 = O
(
(a 2q, a2q,\ ay ax 2 + ayi} = 1 + a
v
P
vY
2
df dy 2
-
From these equations we conclude that 0 2q,
ax 2
a2q,
v
Py·
+ iiif = 1 + v I
df
(a)
- dy + e
where e is a constant of integration. This constant has a very simple physical meaning. Consider the rotation of an element of area in the plane of a cross section of the cantilever. This rotation is expressed by the equation (see page 225) _ õv _ àu 2Wz - ÔX ày The rate of change of this rotation in the direction of the z-axis can be written in the following manner:
~(~-~)=~(~ ~)-~(~+~)=~-~ az ax ay ax az + ay ay az ax ax ay and, by using Hooke's law and expressions (171) for the stress components, we find
a ( ) - 1 (ªTyz az 2w• - G ax
-
OTxz) ay
= -
.! G
(ª
2
2
1/>2
ax
+ ààytf>2 + df) dy
Substituting in Eq. (a), à
2 w,) = 1 -G az C
V
Py
+ vT + e
319
If the x-axis is an axis of symmetry of the cross section, bending by a force P in this axis will result in a symmetrical pattern of rotation w. of elements of the cross section (corresponding to anticlastic curvature), with a mean value of zero for the whole cross section. The mean value of àw,/àz will then also be zero, and this requires that e in Eq. (b) be taken as zero. If the cross section is not symmetrical we can define 1 bending without torsion by means of the zero mean value of aw,/az, again of course requiring the zero value for e. Then Eq. (b) shows that aw,/ àz vanishes for the elements of cross sections at the centroids-that is, these elements along the axis have zero relative rotation, and if one is fixed the others have no rotation-about the axis. With e zero Eq. (a) beeomes a2q, + a2q, = _v_ Py _ df (172) ax2 oy 2 1+ v I dy Substituting (172) in the boundary condition (d) of the previous article we find 2 aq, dy + aq, dx = aq, = [Px _ f( ) ] dy (173) ày ds àx ds as 21 y ds From this equation the values of the function q, along the boundary of the cross section can be calculated if the function f(y) is chosen. Equation (172), together with the boundary condition (173), determines the stress function q,. ln the problems which will be discussed later we shall take function f(y) in such a manner as to make the right side of Eq. (173) equal to zero. 2 q, is then constant along the boundary. Taking this constant equal to zero, we reduce the bending problem to the solution of the differential equation (172) with the condition q, = O at the boundary. This problem is analogous to that of the deflection of a membrane uniformly stretched, having the sarne boundary as the cross section of the bent bar and loaded by a 1.Jontinuous load given by the right side of Eq. (172). Severa! applications of this analogy will now be shown. 107. Circular Cross Section. Let the boundary of the cross section be given by the equation x2 + y2 = r2 (a) 1 J. N. Goodier, J. Aeronaut. Sei., vol. 11, p. 273, 1944. A different definition Was proposed by E. Trefftz, Z. angew. Math. Mech., vol. 15, p. 220, 1935. 2 See S. Timoshenko, Bult. Inst. Engineers of Ways of Communications, St. Petersburg, 1913. See also Proc London Math. Soe., series 2, vol. 20, p. 398, 1922.
THEORY OF ELASTICITY
BENDING OF PRISMATICAL BARS
The right side of the boundary condition (173) becomes zero if we take
The maximum shearing stress is obtained at the center (y = O), where
P (r2 _ y2) f (y) = 21
(3 + 2v)Pr2 (r..,.)max. = 8(1 + 11)!
320
(b)
Substituting this into Eq. (172), the stress function cjJ is then determined by the equation a2q, a2q, 1 + 2v Py ( ) àx 2 + ày 2 = 1 + v T e and the condition that q, = O at the boundary. Thus the stress function is given by the deflections of a membrane with circular boundary of radius r, uniformly stretched and loaded by a transverse load of intensity proportional to 1 + 2vPy
-T+vT
It is easy to see that Eq. (e) and the boundary condition are satisfied in this case by taking (d) e/> = m(x2 + y2 - r2)y where m is a constant factor. This function is zero at the boundary (a) and satisfies Eq. (e) if we take (1
m
+ 2v)P
= 8(1
+
v)I
Equation (d) then becomes
q,
= (1
+ 2v)P (x2 + y2 _
8(1
+ v)I
r2)y
(3 + 2v)P ( 2 _ 8(1 + v)I r (1 + 2v)Pxy " 11• = 4(1 + v)I
'Tzz
=
X
2 _ 1 - 2v 2) 3 + 2v Y
(r
(1 + 2v)Pr 2 (-rzz)ll=±r = 4(1 v)J
+
= ±r) is (h)
It will be seen that the magnitude of the shearing stresses depends on the magnitude of Poisson's ratio. Taking v = 0.3, (g) and (h) beco me p p (k) (r,,,) 11=±r = 1.23 A (-r,,,)max. = 1.38 A' where A is the cross-sectional area of the bar. The elementary beam theory, based on the assumption that the shearing stress r,,, is uniformly distributed along the horizontal diameter of the cross section, gives
4P
Tzz
= 3A
The error of the elementary solution for the maximum stress is thus in this case about 4 per cent. 108. Elliptic Cross Section. The method of the previous article can also be used in the case of an elliptic cross section. Let x2
a2 + b2y2 -
f(y) = -
(174)
(a)
1 =O
be the boundary of the cross section. vanish if we take
The vertical shearing-stress componen:, -r,,, is an even function of x and y, and the horizontal component -r11, is an odd functi.on of the sarne variables. Hence the distribution of stresses (17 4) g1ves a resultant along the vertical diameter of the circular cross section. Along the horizontal diameter of the cross section, x = O; and we find, from (174), (3 + 2v)P 2 _ 1 - 2v y2), .,11, = O (j) 'Tzz = 8(1 + v)J 3 + 2v
(g)
The shearing stress at the ends of the horizontal diameter (y
(e)
The stress components are now obtained from Eqs. (171):
321
The right side of Eq. (173) will
~(~:y2 - a2)
(b)
Substituting into Eq. (172), we find
a2q, ax2
a2q,
Py (ª2 b2 + 1
+ ay2 = T
" )
(e)
+v
This equation together with the condition q, = O at the boundary determines the stress function q,. The boundary condition and Eq. (e) are satisfied by taking (1
q, = 2(1
+ v)a + vb 2 2 2
+
v)(3a
2
P(
+ b )·I
x
2
a2
+ b2 y
2
-
ª
2)
y
(d)
.------··-·_.....,.;...;..;_..,,
f
1NST1TUTUl flüUTEMll'C i
\
11!~1~;~c~~~~R~U.
~
322
THEORY OF ELASTJCITY
BENDING OF PRISMATJCAL BARS
When a = b, this solution coincides with solution (e) of the previous article. Substituting (b) and (d) in Eqs. (171), we find the stress components
109. Rectangular Cross Section. The equation for the boundary line in the case of the rectangle shown in Fig. 188 is
T:u
T11•
+ +
+ +
2(1 v)a 2 b2 P [ 2 = (1 v)(3a2 b2). 2I a (1 + v)a 2 + vb 2 Pxy = - (1 + v)(3a 2 + b2 ) l
_
x2 _ (1 - 2v)a 2 y 2] 2(1 v)a2 b2
+
+
(175)
T,,, Tyz
= Ü
+
+
b2 P [a 2 _ (1 - 2v)a 2 y 2] b2) 21 2(1 v)a 2 b2
+
+
(Tzz)max. =
Pa
2
2J [
1 -
a
+3a2vb + /(1 + v)] b2
2
2
If b is very small in comparison with a, we can neglect the terms containing b2 /a 2, in which case {Tzz)max. =
If we substitute into Eq. (173) the consta.nt Pa2 /2J for f(y), the expression Px 2 /21 - Pa 2 /21 becomes zero along the sides x = ±a of the rectangle. Along the vertical sides y = ± b the derivative dy/ds is zero. Thus the right a side of Eq. (173) is zero· along the boundary line and we can take
3J = 3 A
2
('Tzz)max. = 1
V
-1
+
V
If
p A
The stress at the ends of the horizontal diameter (y = case is 4v P T,,. = 1 +V A
±
b) for this
The stress distribution along the horizontal diameter is i~ this,case :rery far from uniform and depends on the magnitude .of Pmsson s ratio v. Taking v = 0.30, we find '''
p
(Tzz)max. = 1.54 A'
This equation, together with the boundary condition, determines completely the stress function. The problem reduces to the determination of the deflections of a uniformly Frn. 188. stretched rectangular membrane produced by a continuous load, the intensity of which is proportional to
4P
Pa 2
which coincides with the solution of the elementary beam theory. b is very large in comparison with a, we obtain
p
(Tzz)x=o, y==b =
0.92 A
· bthe The maximum stress is about 14 per cent larger than t h at given Y elementary formula,
··'~:li.:!
(a)
(b)
The maximum stress is at the center (y = O) and is given by equation
'
(x2 - a2)(y2 - b2) =O
r
For the horizontal axis of the elliptic cross section (x = O), we find 2(1 + v)a 2 + = (1 v)(3a 2
323
Py
+ vT
The curve mnp in Fig. 188 represents the intersection of the membrane with the yz-plane. From Eqs. (171) we see that shearing stresses can be resolved into the two following systems: (1) (2)
T
11 '
li
= - Ôcp
(e)
ÔX
The first system represents the parabolic stress distribution given by the usual elementary beam theory. The second system, depending on the function
324
THEORY OF ELASTICITY
BENDING OF PRISMATICAL BARS
axis of symmetry, the stress 7'zz is not uniform as in the elementary theory but has maxima at the ends, m and p, anda minimum at the center n. From the condition of loading of the membrane it can be seen that cp is an even function of x and an odd function of y. This requirement and also the boundary condition are satisfied by taking the stress function q, in the form of the Fourier series,
From this we find the following formulas for the center of the cross section (y = O) and for the middle of the vertical sides of the rectangle:
m=oon=oo
q,
i..{
=
m= ao n= oo
\' (2m + l}irx . n1f'Y ~ A2m+l,n cos sm b 2a
\'
2
(d)
-
m=O n=l
Substituting this into Eq. (b) and applying the usual method of calculating the coefficients of a Fourier series, we arrive at the equations A 2m+l,nr2ab [ (
2
m
d )2 + (~)2]
1
P8b
V
+ r-;a ~ "
p
b'
(-l)m-1
"
(2m
+ 1) [ (2m + 1)2 :;2 + n2]
The summation of these series is greatly simplified if we use the known formulas
1
2
--vI
+
1
p
A2m+l,n = -
p
v
= -
1
f
ª
f
-a
b
-b
p
+PT r4(2m +
y cos
(2m
+ l)n sm . n1f'Y -b- dx dy
2a 8b( - l)m+n-1
l)n [ (2m2d
n="'
1y + (~)2]
"(-1)" =
4 n2 n-1
Substituting in (d), we find m="' n= "''(
v
cp
= -
l
2:
p 8b3
+ PI 7
m=O
n=l
-
m="'
l) m+n- 1
(2m
COS
(2m
+ l)n Slil. -nry
2a
m="' n="'
4
b
m=O
+ l)n [ (2m + 1)2 :;2 + n2 J
{- l)m+n-1
(
"
cos n1f'Y
(2m
+
(-l)m 1)[(2m 1) 2
+
+
2Pl 3 y = EJ.,.. 4
in which
n="' sin ~ sin n-irz l l n 2(n 2 + k2) n=l
2:
k2=~ EJ.,..2
.
ande is the distance of the load P from the left support (Fig. 112). Substituting now e =O and Pc = M, we arrive at the following deflection curve produced by the couple M applied at the left end '
2
, 1 l:h.i
P8b +v vT-;a f:( "
f{ (2m + 1) [ (2m + 1) 4~b 2 + n "
2
2
)*
_ r 3 1 - sechi k 2) - 32 . j(kr/2)2
~ This formula can be obtained in the following manner: Using the trigonometric series (h) (p. 155) for the case of a tie rod loaded by the transverse force p and the direct tensile force S, we find that
Having this stress function, the components of shearing stress can be found from Eqs. (e). Let us derive the corrections to the stress given by the elementary theory along the y-axis. It may be seen from the deflection of the membrane (Fig. 188) that along this axis the corrections have the largest values, and therefore the maximum stress occurs at the middle points of the sides y = ±b. Calculating the derivative aq,/ay and taking x = O, we find that
- 1
325
2Ml2
]
y = Ehª ·
nL:="' sm. -z n-irz n(n 2 n=l
+ k2)
Then
By using the membrane analogy useful approximate formulas for calculating these shearing stresses can be derived. If a is large in comparison with b (Fig. 188) we can assume that, at points sufficiently distant from the short sides of the rectangle, the surface of the membrane is practically cylindrical. Then Eq. (b) becomes
n= oo
V
- 1
+
3P b2 v 2A. a2
[
1
4
3 + 11'2
~ f1
n2
( - l)n ] cosh n:a
[2 41: 1 ] n= oo
2
(Tzz")z=O, y=h
3P b V 2A a2
V
= 1+
3-
(176)
a 2 q, _
h n'll'a n 2 cos -b
11'2
Point
2
1
2
4
X=
0, y = 0
Exact Approximate
0.983 0.981
0.940 0.936
0.856 0.856
0.805 0.826
X=
0, y = b
Exact Approximate
1.033 1.040
1.126 1.143
1.396 1.426
1.988 1.934
(e)
Substituting in Eqs. (e), the stresses along the y-axis are
+1+
_ 21 p [ a2
V
2
V
(
2 b )] y - 3
(f)
It will be seen that for a narrow rectangle the correction to the elementary formula, given by the second term in the brackets, is always small. If b is large in comparison with a, the deflections of the membrane at points distant from the short sides of the rectangle can be taken as a linear function of y, and from Eq. (b) we find
a q, v Py ax = 1 +V T 2
2
>
= 1
Py
V
+ v 21 (x2
-
(g)
a2)
Substituting in Eqs. (e), the shearing stress components are -
Tzz -
l
p ( 2 +1 V . 21 a -
X
2)
p
V
,
Tyz
= - 1
+ T xy V
At the centroid of the cross section (x = y = O),
and the deflection at the middle is m= oo
(-1)"'
\'
L
V p = - - - (yª - b2y) 1+V61
1' xz -
3p /2A given by the elementary formula. In the first lines of the table above numerical factors are given by which the approximate value of the shearing stress 3P /2A must be multiplied in order to obtain the exact values of the stress. 1 The Poisson's ratio vis taken equal to onefourth in this calculation. It is seen that the elementary formula gives very accurate values for these stresses when a/b ~ 2. For a square cross section the error in the maximum stress obtained by the elementary formula is about 10 per cent.
2Ml 2
>
1
-= b
0 = Ef1r3
+V T
and we find
in which A = 4ab is the cross-sectional area. These series converge rapidly and it is not difficult to calculate corrections Tzz" for any value of the ratio a/b. These corrections must be added to the value 1
Py
v
dy 2 - 1
n=l
a
327
BENDING OF PRISMATICAL BARS
THEORY OF ELASTICITY
326
(2m
+ 1)[(2m + 1) + k 2
2
1
(a) Tzz
]
Pa 2
= l +V 2['
Tyz
=
Ü
m=O
The sarne deflection obtained by integration of the differential equation of the deflection curve is Ml2
ô = 2EJ-ir2k2
(
br)
1 - sech 2
(b)
The above formula follows from comparison of (a) and (b). . • i The figures of this table are somewhat different from those g~ven by Sa~nt Venant. Checking of Saint-Venant's results showed that there is a numerical error in his calculations.
ln comparison with the usual elementary solution the stress at this point is reduced in the ratio 1/(1 + v). To satisfy the boundary condition at the short sides of the rectangle we take, instead of expression (g), the following expression for the stress function: V Py (x 2 - a 2)[1 - e-Cb-'l/lm] (h) > = - - 1 +V 21
328
THEORY OF ELASTICITY
BENDING OF PRISMATICAL BARS
in which m must be determined from the condition of minimum energy (see Art. 97). ln this manner we find
329
given in the form (b - TJ)/2a in the last column, b - 11 being the distance of the maximum point from the comer.
1
m
= 2aVW
b -
With this value for m, and by using Eq. (h), we can calculate with sufficient accuracy the maximum shearing stress which occurs at the middle of the short sides of the rectangle. lf both sides of the rectangle are of the sarne order of magnitude we can obtain an approximate solution for the stress distribution in a polynomial form by taking the stress function in the form
q,
= (x2 - a2) (y2 -
b2) (my
+ nyª)
(k)
Calculating the coefficients m and n from the condition of minimum energy we find 1
(rz.).,_o,
a
3P/2A
(Tu.) ~-a, 11-"I 3P/2A
2a
o
1.000 1.39(4) 1.988 2.582 3.176 3.770 5.255 6.740 8.225 15.650
0.000 0.31(6) 0.968 1.695 2.452 3.226 5.202 7.209 9.233 19 .466
0.000 0.31(4) 0.522 0.649 0.739 0.810 0.939 1.030 1.102 1.322
2 4 6 8 10 15 20 25 50
11 -1>
b -.,
11_0. Additional ~esul~s. Let us consider a cross section the boundary of which cons1sts of two vertical s1des y = ±a (Fig. 189) and two hyperbolas1 (1
+ v)x2
- vy2 = a2
(a)
It is easy to show that this makes the right side of Eq. (173) on page 319 zero at the boundary if we take f(y) = : ;
The shearing stresses, calculated from (k), are
Pa2 (T zz) z=0, 11=0 = 'if 2
(Tzz)x=o,
y=b
= ~~
+ ma 2b2 - 2a2b2(m
(1 ~ py2 + 1~ p)
Substituting into Eq. (172), we find éJ2>
+ nb 2)
(l)
The approximate values of the shearing stresses given on the second lines of the table (see page 326) were calculated by using these formulas. It will be seen that the approximate formulas (Z) give satisfactory accuracy in this range of values of a/b. If the width of the rectangle is large in comparison with the depth maximum stresses much larger than the value 3P /2A of the elementary theory are found. Moreover ü b/a exceeds 15 the maximum stress is no longer the component Tzz at x =O, y = ±b, the mid-points of the vertical sides. It is the horizontal component 7'11, at points x = a, y = ± TJ on the top and bottom edges near the corners. V alues of these stresses are given in the table 2 on page 329. The values of TJ are 1 See Timoshenko, loc. cit. 'E. Reissner and G. B. Thomas, J. Math. Phys., vol. 25, p. 241, 1946.
ax2
X
éJ2
+ ay2
Frn. 189.
=O
This equation and the boundary condition (173) are satisfied by taking Then the shearing-stress components, from Eq. (171), are
r.,.
= 2pl ( - z2
T11•
= 0
> =
O .
+ _v_ y2 + ~) l+v
l+v
At each point of the cross section the shearing stress is vertical. The maximum of this stress is at the middle of the vertical sides of the cross section and is equal to Ttll3x.
=
Pa 2
2f
The problem can also be easily solved if the boundary of the cross section is given by the equation 1
n~
~),
(± = ( 1 a >X > -a (b) 1 This problem was discussed by F. Grashof, "Elastizitãt und Festigkeit "p 246 1878. ' . ,
For
li
=
Then an approximate expression for the stress function is
i, this cross-section curve has the shape shown in Fig. 190.
By taking
q,
1
f(y) =
Pa2 [ 1 'iI - (
± by);]
2
+
a2q, _ _ 11_Py + Pa ay2 - 1 + 11 I - 2bl 11
!-1 (± 1!)• b
. equat•ion and the boundary condition are satisfied by Th1s taking
1
Pa 2 11 :Z:.: _ 1 q, = 2(1 11)! [ y
(a•
+
+
)-
b (
± 11 ) -+1] v b
(y - a)[x
~
FIG. 190.
Txz
= 2(1
p
+ 11)! (a
2 -
x
'
Tyz
(1~11)Ixy
= -
(e)
y
dx = dy
/
I
FIG. 191.
th d (Art 109) we may arrive By using the energy me . o .. th s Let us consider for instance, the · t 1 t"onmmanyo ercase · ' . b at an app:oxima e so. u ~- 191 The vertical sides of the boundary are g1ven y cross sectt~on sh~wn+mb a~~· the ~ther two sides are ares of the circl~. the equa 10n Y - - • x2
+ y2
- r• =O
The right side of Eq. (173) vanishes if we take
f(y) =
{z. (r• -
y2)
- (2a
+ y)
2
+ By + · · ·) 3
+ y) tan a]
=
O
tan 2 a
P - (2a I
tan 2 a
1
Substituting from (e) and integrating, we arrive at the equation of the boundary, x 2) '
r 2 )(Ay
+ y) tan
l-a X
2
a
(a)
FIG. 192.
An approximate solution may be obtained by using the energy method. particular case when
r
I
p (2a 21
a2q, a2q, -ax - 2 = 1-+.,,-v Py - 2 + ay I
Tyz
-
+ y• -
Equation (172) for determining the stress function q, then becomes
. ( ) for stress components may be derived. The equafrom which the express10ns e b f df the condition that at the boundary tion of the boundary can now e oun ~om the direction of shearing stress coincides w1th the tangent / ' to the boundary. Hence I \
y = b(a 2
+ (2a + y) tan a][x f(y) =
. d·ifferent way In discussing stresses in a · t the sarne result m a · . h W e can arnve a . which is large in comparison w1th the dept ' we rectangular beam t_he w1dthl ot~ f r the stress function [Eq. (g), Art. 109] the used as an approximate so u ion o expression - - " - Py (x2 - a•) > - 1 +V 2[
Txz
b2)(x2
The right side of Eq. (173) is zero if we take
Substituting in Eqs. (171) we find 2)
= (y• -
in which the coefficients A, B, . . . are to be calculated from the condition of minimum energy. Solutions for many shapes of cross section have been obtained by using polar and other curvilinear coordinates, and functions of the complex variable. These include sections bounded by two circles, concentric 1 or nonconcentric,2 a circle with radial slits, 3 a cardioid, 4 a limaçon, 5 an elliptic limaçon,• two confocal ellipses, 7 an ellipse and confocal hyperbolas, 8 triangles and polygons 9 including a rectangle with slits, 10 anda sector of a circular ring. 11 111. Nonsymmetrical Cross Sections. As a first example let us consider the case of an isosceles triangle (Fig. 192). The boundary of the cross section is given by the equation
· · (173) · h s i e q, must be constant th left side of the boundary cond1t1on vams e ' .. , e along the boundary. Equation (172) becomes
a2q, ax2
331
BENDING OF PRISMATICAL BARS
THEORY OF ELASTICITY
330
'
(d) '
= 1-"+V = 3~
In the (b)
1 A solution is given in A. E. H. Love's "Mathematical Theory of Elasticity," 4th ed. p. 335, and in I. S. Sokolnikoff's "Mathematical Theory of Elasticity," p. 253. 2 B. R. Seth, Proe. Irulian Acad. Sei., vol. 4, sec. A, p. 531, 1936, and vol. 5, p. 23, 1937. 3 W. M. Shepherd, Proe. Roy. Soe. (London), series A, vol. 138, p. 607, 1932; L. A. Wigglesworth, Proe. London Math. Soe., series 2, vol. 47, p. 20, 1940, and Proe. Roy. Soe. (London), series A, vol. 170, p. 365, 1939. ' W. M. Shepherd, Proe. Roy. Soe. (London), series A, vol. 154, p. 500, 1936. •D. L. Holl and D. H. Rock, Z. angew. Math. Meeh., vol. 19, p. 141, 1939. 6 A. C. Stevenson, Proe. London Math. Soe., series 2, vol. 45, p. 126, 1939. 7 A. E. H. Love, "Mathematical Theory of Elasticity," 4th ed., p. 336. 8 B. G. Galerkin, Bull. I nst. Engineers of W ays of Communieation, St. Petersburg, vol. 96, 1927. See also S. Ghosh, Bull. Calcutta Math. Soe., vol. 27, p. 7, 1935. 9 B. R. Seth, Phil. Mag., vol. 22, p. 582, 1936, and vol. 23, p. 745, 1937. 'º D. F. Gunder, Physies, vol. 6, p. 38, 1935. 11 M. Seegar and K. Pearson, Proe. Roy. Soe. (London), series A, vol. 96, p. 211,
1920.
332
~
BENDlNG OF PRlSMATICAL BARS
an exact solution of Eq. (a) is obtained by taking for the stress function the expresision
The caf'e shown in Fig. 194 can be treated in a similar manner. Assume, for example, that the cross section is a parabolic segment and that the equation of the parabola is x 2 = A(y +a) Then we take p j(y) = 21 . A(y +a) y---+--r---1
> =
p 61
[x
2 -
!
3
(2a
+ y)
2
J(y -
a)
The stress components are then obtained from Eqs. (171): âc/>
Tz•
= ây -
âc/>
Px 2
2T
Tu• = - âx =
P + 61 (2a + y)
2
=
2 y'ãp 2 27 4 [-x
ª
+ a(2a + y)]
2 v3P
(e)
27a4 x(a - y)
Along the y-axis, x = O, and the resultant shearing stress is vertical and is represented by the linear function (Tzz)z-o
=
2v3P
27a 3 (2a
+ y)
The maximum value of this stress, at the middle of the vertical side of the cross section, is (d)
By calculating the moment with respect to the z-axis of the shearing forces given by the stresses (e), it can be shown that in this case the resultant shearing force passes through the centroid C of the cross section. Let us consider next the more general case of a cross section with a horizontal axis of symmetry (Fig. 193), the lower and upper portions of the boundary being given by the equations = ..p(y) X = -..p(y) X
for for
X
>O
X
Then the function FrG. 193.
[x
+ ..p(y)][x -
With this assumption the stress function has to satisfy the differential equation
and be constant at the boundary. The problem is reduced to that of finding the deflections of a uniformly stretched membrane when the intensity of the load is given by the right-hand side of the above equation. This latter problem can usually be solved with sufficient accuracy by using the energy method as was showa in the case of the rectangular cross section ( page 328).
r:''
~l,
With this expression for J(y) the first factor on the righthand side of Eq. (173) vanishes along the parabolic portion of the boundary. The factor dy/ds vanishes along the straightline portion of the boundary. Thus we find again that the FIG. 194. stress function is constant along the boundary and the problem can be treated by using the energy method.
112. Shear Center. ln discussing the cantilever problem we chose for z-axis the centroidal axis of the bar and for x- and y-axes the principal centroidal axes of the cross section. We assumed that the force P is parallel to the x-axis and at such a distance from the centroid that twisting of the bar does not occur. This distance, which is of importance in practical calculations, can readily be found once the stresses represented by Eqs. (171) are known. For this purpose we evaluate the moment about the centroid produced by the shear stresses T:x:z and T 11,. This moment evidently is
M, =
Jf
(T:x:zY -
T11zX)
dx dy
(a)
Observing that the stresses distributed over the end cross section of the beam are statically equivalent to the acting force P we conclude that the distance d of the force P from the centroid of the cross section is d=
..p(y)] = z2 - [..p(y)]2
vanishes along the boundary and in our expressions for stress components (171) we can take p J(y) = 21 [..p(y)] 2
':
333
THEORY OF ELASTIClTY
IM.I p
(b)
For positive M, the distance d must be taken in the direction of positive ln the preceding discussion the assumption was made that the force is acting parallel to the x-axis. When the force P is parallel to the y-axis instead of the x-axis we can, by a similar calculation, establish the position of the line of action of P for which no rotation of centroidal elements of cross sections occurs. The intersection point of the two lines of action of the bending forces has an important significance. If a force, perpendicular to the axis of the beam, is applied at that point we can resolve it into two componentfl parallel to the x- and y-axes and on the basis of the above discussion we conclude that it does not produce rotation of centroidal elements of y.
.;.· I
334
THEORY OF ELASTICITY
cross sections of the beam. This point is called the shear ccntersometimes also the center of flexure, or flexural center. If the cross section of the beam has two axes of symmetry we can conclude at once that the shear center coincides with the centroid of the cross section. When there is only one axis of symmetry we conclude, from symmetry, that the shear center will be on that axis. Taking the symmetry axis for y-axis, we calculate the position of the shear center from Eq. (b). Let us consider, as an example, a semicircular cross section 1 as shown in Fig. 195. To find the shearing stresses we can utilize the solution developed for circular beams (see page 319). In that case there are no stresses acting on the vertical diametral section ·xz. W e can imagine the beam divided by the xz-plane into two halves each of which represents a semicircular beam bent by the force P /2. The stresses are given by Eq. (174). FIG. 195. Substituting into Eq. (a), integrating, and dividing M, by P /2, we find for the distance of the bending force from the origin O the value 2M, 8 3 + 4v e= p = 15-ir T+v r
This defines the position of the force for which the cross-sectional element at point O, the center of the circle, does not rotate. At the sarne time an element at the centroid of the semicircular cross section will rotate by the amount [see Eq. (b) page 318] w
= vP(lHI- z) ·O .424r
where 0.424r is the distance from the origin O to the centroid of the semicircle. To eliminate this rotation a torque as shown in Fig. 195 must be applicd. The magnitude of this torque is found by using the table on page 279, which gives for a semicircular cross section the angle of twist per unit length Mt (} = 0.296Gr4 1 See S. Timoshenko, Bull. Inst. Engineers of Ways of Communications, St. Petersburg, 1913. It seems that the displacement of the bending force from the centroid of the cross section was investigated in this paper for the first time.
'
'
' 1
BENDING OF PRISMATICAL BARS
335
Then t~e condition that centroidal elements of cross sections do not rotate gives M1(l - z) _ vP(l - z) 0.296Gr4 EI · 0.424r and M _ vP · 0.296r4 · 0.424r 1 2(1 + v)I This tor.que will be produced by shifting the bending force p /2 toward the z-axis by the amount ô = 2Mt
= Sv • 0.296 · 0.424r
P
2(1
+ v)'ll"
This quantity ~ust be subtracted from the previously calculated diso of
tanc~ e to obtam t?e distance of the shear center from the center the circle. Assummg v = 0.3, we obtain
e- ô
= 0.548r - 0.037r = 0.511r
ln sections as in Fig. 193 the shearing-stress components are aq,
= ()y -
Tu
P
2J [x 2
aq,
if2(y)],
-
Tyz
Hence M, =
f f (~:y + ~=x)dxdy- ~! f
= - -
ax
[x 2 -1/;2(y)]ydxdy
(e)
Integrating by parts and observing that cp vanishes at the bound :r = ±if;(y), we obtain ary
f f G:
Y
ff[x 2
+ :: x) dx dy
ff
= -2 cp dx dy = Íif;ª(y) - 21/;ª(y) = --ll/;ª(y) = -tf yif;ª(y) dy
f[x2 - y;2(y)] dx if; 2 (y)]y dx dy I = Jfx 2 dxdy = ifif; 3 (y) dy
Substituting in (e) and dividing by p we find d =
l~·I
=
1- ~ f ! p
cp dx dy
+ f yif;ª(y) dy' f if; 3(y) dy
. if; (y) and using the membrane analogy for finding q, we can ai Kn owmg wafys calculate 1 with sufficient accuracy the position of the shear cen·
t er or these cross sections.
"~ E~ai:1ples of such calculations can be found in the book by L s L "b . 19;,iational Methods for Solving Problems of the Theory of Elasti~it~,"e~;::;::
336
BENDING OF PRISMATICAL BARS
THEORY OF ELASTICITY
The question of the shear center is especially important in the case of thin-walled open sections. Its position can be easily determined for such sections with suffi.cient accuracy by assuming that the shearing stresses are uniformly distributed over the thickness of the wall and are parallel to the middle surface of the wall. 1 The location of the shear center in the cross section is determined by the shape of the section only. On the other hand the location of the center of twist (see page 271) is dependent on the manner in which the bar is supported. By choosing this manner of support suitably the axis of twist can be made to coincide with the axis of shear centers. It can be shown that this occurs when the bar is so supported that the integral f f w 2 dx dy over the cross section is a minimum, 2 w being the warping displacement of torsion (indeterminate by a linear function of x and y before this condition is applied). ln practice the fixing will usually disturb the stress distribution near the fixed end-as for instance when it prevents displacements in the end section completely. ln that case, if we regard the bending force as a concentrated load at the shear center, producing zero rotation, the reciproca! theorem (page 239) shows that a torque will produce zero deflection of the shear center. This indicates that the center of twist will coincide with the shear center. 3 The argument is of an approximate character since the existence of a center of twist depends on absence of deformation of cross sections in their planes, and this will not hold in the disturbed zone near the fixed end.
113. The Solution of Bending Problems by the Soap-film Method. The exact solutions of bending problems are known for only a few special cases in which the cross sections have certain simple forms. For practical purposes it is important to have means of solving the problem for any assigned shape of the cross section. This can be accomplished by numerical calculations based on equations of finite differences as explained in the Appendix, or experimentally by the soapfilm method, 4 analogous to that used in solving torsional problema (see page 289). For deriving the theory of the soap-film method we use Eqs. (171), (172), and (173) (see Art. 106). Taking v
f(y) = 2(1
Py2
+ v) T
1 References may be found in S. Timoshenko, "Strength of Materials," 2d ed., vol. 2, p. 55. 2 R. Kappus, Z. angew. Math. Mech., vol. 19, p. 347, 1939; A. Weinstein, Quart. Applied Math., vol. 5, p. 79, 1947. 8 See R. V. Southwell, "lntroducÚ~n to the Theory of Elasticity," p. 29; W. J: Duncan, D. L. Ellis, and C. Scruton, Phil. M ag., vol. 16, p. 201, 1933. 4 This method was indicated first by Vening Meinesz, De Ingenieur, p. 108, Holland, 1911. It was developed independently by A. A. Griffith and G. I. Taylor, Tech. Rept. Natl. Advisory Comm. Aeronaut., vol. 3,p. 950, 1917-1918. The results given here are taken from this paper.
337
Eq. (172) for the stress function is ()2cf> ax2
()2cf>
+ ay2
(a)
=O
This is the sarne equation as for an unloaded and uniformly stretched membrane (see page 271). The boundary condition (173) becomes ocf>
as
= [Px 2 _ 21
Py 2] dy
v
2(1
+
v)
T
ds
(b)
Integrating along the boundary s we find the expression _ P
e/> -
y
f
x 2 dy
v
Pyª
- 2- - 2 (1 + v) 3T
+ constant
(e)
from which the value of e/> for every point of the boundary can be calculated. f (x 2/2) dy vanishes when taken around the boundary, since it rep.resents the moment of the cross section with respect to the y-axis, wh1ch passes through the centroid of the cross section. Hence e/>, calculated from (e), is represented along the boundary by a closed curve. Imagine now that the soap film is stretched over this curve. Then the surface of the film satisfies Eq. (a) and boundary condition (e). Hence the ordinates of the film represent the stress function cf> at all points of the cross section to the scale used for representation of the function cf> along the boundary [Eq. (e)]. The photograph 196a illustrates one of the methods used for construction of the boundary of the soap film. A hole is cut in a plate of celluloid, of such a shape that after the plate is bent the projection of the edge of the hole on the horizontal plane has the sarne shape as the boundary of the cross section of the beam. The plate is fixed on vertical studs and adjusted by means of nuts and washers until the ordinates along the edge of the hole represent to a certain scale the values of e/> given by expression (e). The photograph 196b illustrates another method for construction of the boundary by using thin sheets of annealed brass. 1 The small corrections of ordinates along the edge of the hole can be secured by slight bending of the boundary. . The analogy between the soap-film and the bending-problem equations holds rigorously only in the case of infinitely small deflections of the memb:ane. ln experimenting it is desirable to have the total range of the ordmates of the film not more than one-tenth of the maximum 1
See the paper by P. A. Cushman, Trans. A.S.M.E., 1932.
339
THEORY OF ELASTICITY
BENDING OF PRISMATICAL BARS
horizontal dimension. If necessary the range of the funct.ion along the boundary can be reduced by introducing a new function c/>1, instead of cf>, by the substitution cf> = cf> 1 + ax + by (d)
The reduction of the range of the function c/>1 at the boundary can usually be effected by a proper adjustment of the constants a and b. When the function cf> 1 is obtained from the soap film, the function cf> is calculated from Eq. (d). Having the stress function cf>, the shearingstress components are obtained from Eqs. (171), which have now the form ocf> Px 2 V Py 2 Tzz = ay - 2I + 2(1 + v) T (e) ocf>
338
Tyz
FIG. 196b.
where a and b are arbitrary constants. It may be seen that the function cf> 1 also satisfies the membrane equation (a). The val~es of the function cf> 1 along the boundary, from Eqs. (e) and (d), are g1ven by cf>
_
1 -
f_ J
f
2
x dy _ v Pyª - ax - by 2 2(1 v) 31
+
+ constant
=
-
OX
The stress components can now be easily calculated for every point of the cross section provided we know the values of the derivatives àcf>/ày and àcf>/àx at this point. These derivatives are given by the slopes of the soap film in the y- and x-directions. For determining slopes we proceed as in the case of torsional problems and first map contour lines of the film surface. From the contour map the slopes may be found by drawing straight lines parallel to the coordinate axes and constructing curves representing the corresponding sections of the soap film. The slopes found in this way must now be inserted in expressions (e) for shear-stress components. The accuracy of this procedure can be checked by calculating the resultant of ali the shear stresses distributed over the cross section. This resultant should be equal to the bending force P applied at the end of the cantilever. Experiments show that a satisfactory accuracy in determining stresses can be attained by using the soap-film method. The results obtained for an I-section 1 are shown in Figs. 197. From these figures it may be seen that the usual assumptions of the elementary theory, that the web of an I-beam takes most of the shearing force and that the shearing stresses are constant across the thickness of the web, are fully confirmed. The maximum shearing stress at the neutral plane is in very good agreement with that calculated from the elementary theory. The component r 11, is practically zero in the web and reaches a maximum at the reentrant corner. This maximum should depend on the radius of the fillet rounding the reentrant corner. For the proportions taken, it is only about one-half of the maximum stress r,,, at the neutral plane. The lines of equal shearing-stress components, giving the ratio of these components to the average shearing stress P /A, are shown in the figures. The stress concentration at the reentrant corner has been studied for 1 ln this case of symmetry only one-quarter of the cross section need be investigated.
BENDING OF PRISMATICAL BARS THEORY OF ELASTICITY 340 the case of a T-beam. The radius of the reentrant corner was increased in a series of steps, and contour lines were mapped for each case. It was shown in this manner that the maximum stress at the corner equals the maximum stress in the web when the radius of the fillet is about onesixteenth of the thickness of the web.
center line is built in, u and du/dz are zero when z = O and hence constants e and d in Eq. (b) are zero. ' The cross sections of the beam do not remain plane. They become ':arped, owing to the action of shearing stresses. The angle of inclinat10n of an element of the surface of the warped cross section at the centroid to the defl.ected center line is
:JC
:JC
X
341
'
!
l 1
Confour lines of solilp-film
Lines of eqw.il :;heQr stress
.1
t:yz=nf
1
1
+--
y
1
T:cz=m:. y
-i-.
y
FIG. 197.
114. Displacements. When the stress components are found, the displacements u, v, w can be calculated in the sarne manner as in the case of pure bending (see page 250). Let us consider here the defl.ection curve of the cantilever. The curvatures of this line in the xz- and yz-planes are given with sufficient accuracy by the values of the derivatives a2u; az2 and a2v/ az 2 for X = y = o. These quantities can be calculated from the equations
a2u az2 = a2v az2 -
a'Yx•
ae,
1 ar,,,
1 au.
Tz - ax = GTz - E ax = a'Y11z
Tz -
P(l - z)
EI
(a)
ae. - o ay -
We see that the center line of the cantilever is bent in the xz-plane in which the load is acting, and the curvature at any point is proportional to the bending moment at this point, as is usually assumed in the elementary theory of bending. By integration of the first of Eqs. (a), we find
(b)
and can be calculated if the shearing stresses at the centroid are known 11~. Further Investigations of Bending. ln the foregoing article~ we d1scussed the problem of bending of a cantilever fixed at one end and loaded by a transverseforce on the other. The solutions obtained are the exact solutions of the bending problem, provided the externa! forces are distributed over the terminal cross sectione in the sarne manner as the stresses u., T,,., r 11, found in the solutions. If this condition is not fulfilled there will be local irregularities in the stress distribution near the ends of the beam, but on the basis of Saint-Venant's principie we can assume that at a sufficient distance from the ends, say at a distance larger than cross-sectional dimensions of the beam our solutions are suffic!en~ly accurate. By using the sarne principie ;e may extend the app!1cat10n of the above solutions to other cases of loading and supportmg of beams. We may assume with sufficient accuracy that the stresses at any cross section of a beam, at sufficient distance from the loads, depend only on the magnitude of the bending moment and the .s~earing force at. this cros~ section and can be calculated by superposit10n of the solut10ns obtamed before for the cantilever. If the bending forces are inclined to the principal axes of the cross sec~ion. of the ~eam! they can always be resolved into two components actmg m the d1rect10n of the principal axes and bending in each of the two principal planes can be discussed separately. The total stresses and disp!~cements will then be obtained by using the principie of superposit10n. . N ~ar the points of application of externa! forces there are irregularities m stress distribution which we discussed before for the particular c~se of. a narrow rectangular cross section (see Art. 36). Analogous ~1scuss1on for other shapes of cross section shows that these irregularit1es are of a local character. 1
where e and d are constants of integration which must be determined from the conditions at the fixed end of the cantileyer. If the end of the
1'
~ '
'
1
See L. Pochhammer's, "Untersuchungen über das Gleichgewicht des elasStabes," Kiel, 1879. See also a paper by J. Dougall, Trans. Roy. Soe. (Edinburgh), vol. 49, p. 895, 1914. •
tisc~en
342
THEORY OF ELASTICITY
The problem of bending is solved also for certain cases of distributed load.1 It is shown that in such cases the central line of the beam usually extends or contracts as in the case of the narrow rectangular cross section (see Art. 21) already discussed. The curvature of the center line in these cases is no longer proportional to the bending moment but the necessary corrections are small and can be neglected in practical problems. For instance, in the case of a. cir?ular beam bent by its own weight, 2 the curvature at the fixed end 1s g1ven by the equation 1 _ M [l _ 7
r-
EI
+ 12v + 4v 6(1 + v)
2
a 2]
l2
in which a is the radius of the cross section, and l the length of the cantilever. The second term in the brackets represents the correction to the curvature arising from the distribution of the load. It is small, of the order of a 2 /l 2 • This conclusion holds also for beams of other shapes of cross section bent by their own weight. 3 1J. H. Michell, Quart. J. Math., vol. 32, 1901; also K. Pearson, ibid., vol. 24, 1889, and K. Pearson and L. N. G. Filon, ibid., vol. 31, 1900. 1 This problem is discussed by A. E. H. Love, "Mathematical Theory of Elasticity," 4th ed., p. 362, 1927. . . . a The case of a cantilever of an elliptical cross sect1on has been discussed by J. M. Klitchieff, Bull. Polytech. Inst., St. Petersburg, p. 441, 1915.
CHAPTER 13 AXIALLY SYMMETRICAL STRESS DISTRIBUTION IN A SOLID OF REVOLUTION
116. General Equations. Many problems in stress analysis which are of practical importance are concerned with a solid of revolution deformed symmetrically with respect to the axis of revolution. The simplest examples are the circular cylinder strained by uniform internai or externai pressure, and the rotating circular disk (see Arts. 26 and 30). For problems of this kind it is often convenient to use cylindrical coordinates [see Eqs. (170), page 306]. The deformation being symmetrical with respect to the z-axis, it follows that the stress components are independent of the angle O, and all derivatives with respect to () vanish. The components of shearing stress -r,9 and -r9, also vanish on account of the symmetry. Thus Eqs. (170) reduce to
au, + OTrz +
(177)
The strain components, for axially symmetric.al deformation, are, from Eqs. (169), Er
=
au iJr'
u
E9
= r'
iJw
Ea ':"'
az'
'Yrz
aw = -au +az iJr
(178)
It is again of advantage to introduce a stress function cp. It may be verified by substitution that Eqs. (177) are satisfied if we take
= -a ( V V2cp i)z
u9 =
i. az
(v
V 2q,
a2q,) - -ar 2 - !
] u. = -a [ (2 - v) V 2 cp - -
Tn
= -a [ (1 ar
az
- v) 343
i '
,\ ...
a2q,] V2cp - -·.
az2
(179)
345
AXIALLY SYMMETRICAL STRESS DISTRIBUTION THEORY OF ELASTICITY
344
This equation holds for any value of 8, hence
provided that the stress function cf> satisfies the equation
+ ! !!_ + ~)(ª cf> + ! acp + ªazcf>) 2 = V V cf> = O (~ ar 2 r ar az 2 ar 2 r ar 2
2
2
2
(180)
2
a2 1 a 1 a2 a2 2 ar 2 + ar+ T2 ae + az 2
r
(a)
which corresponds to Laplace's operator
a2 ax2
a2
a2
+ ay2 + az2
in rectangular coordinates [see Eq. (d), page 57]. It should be noted that the stress function cf> does not depend on 8, so that the third term in (a) gives zero when applied to cf>. W e now transform the compatibility equations (130) (see page 232) to cylindrical coordinates. Denoting by 8 the angle between r and the x-axis we have [see Eq. (13)]
""' =
2
unaffected by the presence of "•' Then v2"
"'
=
a- 2
( ar 2
+ <19 sin + <19 cos
2
8 8
2
2
2
2
(ur cos 8
-
+ <19 sin
2
V
2
2
r
COS
28(<1r - <19)
(e)
+--ar r
ar
Substituting (e) and (d) in the first of Eqs. (130), we obtain
2
T2 (ur
-
+ [ ( -ara2 + -r1 -àra + -àza 2) 2
!
~ '
.
1
LI r:i
2
<19
ar2 =
1 ae
1
+1+ r p
ar =
o
(e)
o
a2e
ax az = ar az cos
8)
U sing the symbol e for the sum of the three normal components of stress and applying Eq. (b) on page 57, we obtain for a symmetrical stress distribution a 2e a2e 2 ae sin2 8 (d) = -cos 8 2
a2) r( + r ar + àz2 ~r
p
(J -
T;;)
cos 8
Substituting in the fifth of Eqs. (130), we obtain
2 2 + r- -ara + -r1 -aoa 2 + -aza )
a
a2e
V' 2Txz = V2(Trz cos O) = ( V 2Trz
a2 + r1 àra + az2 a2) ( 2 o+ . 2 8) O"r cos <19 sm
1
(u, -
a2e
1
+1+
The sarne result is obtained by considering the second of Eqs. (130), so that Eqs. (e) take the place of the first two equations of the system (130) for the case of a symmetrical deformation. The third equation of (130) retains the sarne form in cylindrical coordinates. Consider now the remaining three equations of the system (130), containing shearing-stress components. ln the case of symmetrical deformation only the shearing stress Trz is different from zero, and the stress components Txz and r 11,, acting on a plane perpendicular to the z-axis, are obtained by resolving r,, into two components parallel to the x- and y-axes, Tu = Trz COS O, T11z = Trz Sin 8 We have also
(b)
= ( ar2
a2 - ar2
2
)
(u, - <1 9)
T2
Trz.
1
ax2
2
O"r -
( àr 2 + r ar + àz 2 " 9 + T2
v2 denotes the operation
The symbol
a2 1 a a2 ) ( àr2 + r ar + az2 a2 1 a a
·
- <19)
1
+1+
+ -r22 (
u9)
p
aar22e] cos2
8
. + 1-+1-v -1r -ªº] sm àr
2
8= O
2 Trz -
1
1
a2e
T2 Trz + 1 + p ar az
=
o
(f)
The sarne result is obtained by considering the fourth of Eqs. (130). The last equation of the system (130) can also be transformed to cylindrical coordinates by substituting Txy
=
j.(
ln this way we find (1
+ v)V
2
1 (u, - u 9) sin 20] 2
[-
+ sin2 28 (~ - ! !!.) O = ar 2 r ar
O
This equation follows at once from Eqs. (e) on subtracting one from the other. Hence the compatibility equations (130), in the case of a deformation symmetrical with respect to an axis are in cylindrical ' ' coord.mates,
346
THEORY OF ELASTICITY
V2u r V 2uo
2
1 ae +1 +-11 -or = O 2
-
r2 (u r
- uo)
+ r-2 (ur -
O'B)
-
2
2
1 lél8 +--- = 1+11ror
Trz -
T2 Trz +
1
1
+
(g)
ae
1
1
2
Ü
2
+ 1 + 11
v2u, V
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
o
oz2 =
a2e _
o
11 or oz -
It can be shown that all these equations are satisfied by the expressions for the stresses given in Eqs. (179) when the stress function satisfies Eq. (180). We see that the discussion of problems involving stress distributions symmetrical about an axis reduces to finding in each particular case the solution of Eq. (180), satisfying the boundary conditions of the problem. 1 ln some cases it is useful to have Eq. (180) in z polar coordinates R and ift (Fig. 198) instead of Frn. 198. cylindrical coordinates r and z. This transformation can easily be accomplished by using the formulas of Art. 25. We find
a2 a2 02 1 a ar 2 + az 2 = oR 2 + R oR 1 éJ
1
r or = R sin"'
1
a2
2
a
( éJ . oR Slll ift
1 ( 0R 2 + R oR + R 2 ctn ift
Another way of solving these problems is to consider explicitly the displacements. By using Eqs. (178) the stress components can be represented as functions of the displacements u and w. Substituting these functions in Eqs. (177) we arrive at two partia! differential equations of the second order containing the two functions u and w. The problem is then reduced to the solution of these two equations. 117. Solution by Polynomials. Let us consider solutions of the Eq. (181), which are at the sarne time solutions of the equation
a q, 2
éJR2
2
aq,
aq,
1
1
+ R éJR + R2 ctn "' ª"' + R2
a q, 2
ay,.2 = o
(182)
A particular solution of this latter equation can be taken in the forro
(a) in which '11n is a function of the angle Y,. only. Substituting (a) into Eq. (182) we find for '11,. the following ordinary differential equation: o ( sm . Y,. 0'11 sin1 ift éJY,. oi/;")
+ n (n + 1)'11n
= O
(b)
Thisequationcan besimplified byintroducinganewvariable, x =cosi/;. Then
+ R 2ay,. 2 cos Y,. o ) 1 éJ oift = R oR
+ ----n:-
ctn Y,. o
+ J[2 ª"'
Substituting in Eq. (180),
a2
347
Substituting in Eq. (b), we obtain n 2 éJ'1' n (1 - x 2) é)2'1f éJx 2 - x a;-
a 1 a2 ) ay,. + R 2 ay,. 2 1
(
a2q,
éJR 2
2
aq,
+ R éJR
aq,
1
a2q,)
+ R2 ctn ift éJift + R2 éJift2
+ n ( n + 1)'11,. = O
(183)
We shall solve this equation by series. 1 Assuming that = O (181)
We shall apply severa! solutions of this equation in succeeding articles to the investigation of particular problems involving axial symmetry. 1 This method of expressing all the stress components in terms of a single stress function, which satisfies Eq. (180), is given in detail by A. E. H.Love, "Mathematical Theory of Elasticity," 4th ed., p. 274, 1927. Another method of expressing the problem in terms of a stress function has been given by J. H. Michell, Proc. London Math. Soe., vol. 31, p. 144, 1900. The relation between the stress function of two-dimensional problems and the stress function discussed in this chapter has been considered by C. Weber, Z. angew. Math. Mech., vol. 5, 1925.
'11 n = a1xm1
+ a2xm• + aaxm• +
and substituting in Eq. (183), we find n(n + l)(a1xm1 + a2xm• + aaxm• + .. -m1(m1 - l)a1xm 1- 2 + m2(m2
(e)
·) = m1(m1 + l)a1xm1 + l)a2xm• - m2(m 2 - l)a 2xm,-2 + . . . (d)
ln order that this equation may be satisfied for any value of x, there must be the following relations between the exponents m 1, m 2, m 3, 1 This is known as Legendre's equation. A complete discussion can be found in A. R. Forsyth, "A Treatise on Ditierential Equations," p. 155, 1903.
349
THEORY OF ELASTICITY
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
It follows that the series (e) is arranged in descending powers of x.
Ao, Ai, ••• are arbitrary constants. These polynomials are also solutions of the Eq. (181). From these solutions we can get new solutions of Eq. (181) which will no longer be solutions of Eq. (182). If R"'iJ!,. is a solution of Eq. (182), it can be shown that R"+2'iJ!,. is a solution of Eq. (181). Performing the operation indicated in the parentheses of Eq. (181),
348
The magnitude of m 1 will now be determined by equating coefficients of xm1 in (d). Then n(n
+ 1)
- m 1(m 1 + 1) = (n - m1)(m1
+n+
1) =O
This gives for m 1 the two solutions m1 = -(n
+ 1)
(e)
For the first of these solutions, m1
=
n,
a2 2 a 1 a 1 a2 ) ( aR2 + R aR + R2 ctn if; ay; + R2 ay;2
= 2(2n ma= n - 4,
m 2 = n - 2,
The coefficients a 2, a 3, • • • in Eq. (d) are found by equating to zero the coefficients of each power of x. Taking, for instance, the terms containing xm,-2r+2, we find for the calculation of the coefficient a, the
+
l)ar = (m 1 - 2r
+ 2)(m1
from which, by substituting m1
a, = -
- 2r + 3)a, - (m 1 - 2r =
+ 4)(m1 -
2r
The series (e) can now be put in the form
n(n - 1) n-2 il',. = a1 [ x" - 2(2n - 1) x
+
l)(n - 2)(n - 3) 2·4(2n-1)(2n-3)
n(n -
which represents a solution of Eq. (183). in (a) and remembering that X=
cos if;,
c/>2 = B2(r 2 + z2) c/>s = BaZ(r 2 + z 2) cp4 = B4(2z 2 - r 2)(r2 + z2) 3 cp5 = B &(2z 3r 2z) (r 2 + z2)
+ 3)a,_1
a..-i
Rx = z,
xn-4 _
•• ·],
(f)
Substituting this solution
R =
vr + 2
Substituting in Eqs. (179), we find
2
we find, for n equal to O, 1, 2, 3, . . . , the following particular solutions of Eq. (182) in the form of polynomials:
+ + +
+
. . . . . . . . . ..........
(185)
118. Bending of a Circular Plate. Severa! problems of practical interest can be solved with the help of the foregoing solutions. Among these are various cases of the bending of symmetrically loaded circular plates (Fig. 199). Taking, for instance, the polynomials of the third degree from z (184) and (185), we obtain the stress FIG. 199. function 2 3 2 3 (a) cf> = a 3 (2z - 3r z) + b3 (r z + z )
Z
cf>o =Ao cf> 1 = A1z cf> 2 = A2[z 2 - i(r 2 z2)] cf>a = A 3 [zª - fz(r 2 z2)] cp4 = A4[z4 - ~z2(r2 + z2) + /r,(r2 + z2)2] cp5 = A&[z5 - 1jzª(r2 z2) + /rz(r2 z2)2]
(g)
+
n,
(n - 2r + 4)(n - 2r + 3) 2(r - 1)(2n - 2r + 3)
+ 3)R"'iJ!,.
Repeating the sarne operation again, as indicated in Eq. (181), we obtain zero, since (g) is a solution of Eq. (182). Hence R"+ 2'iJ!,. is a solution of Eq. (181). It is seen that multiplying solutions (184) by R2 = r2 z 2, we can obtain the following new solutions:
equation n(n
Rn+2'iJ!,.
(184)
+
+
= 6aa (lOv - 2)ba, <1'9 = 6aa (lOv - 2)bs = -12aa (14 - lOv)ba, Trz =O
+
(186)
The stress components are thus constant throughout the plate. By a suitable adjustment of constants a 3 and ba we can get the stresses in a plate when any constant values of u. and ur at the surface of the plate are given. Let us take now the polynomials of the fourth degree from (184) and (185), which gives us cf> = a 4 (8z 4 --- 24r2z2 + 3r4) + b4 (2z 4 + r 2z2 - r 4 ) (b)
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
THEORY OF ELASTICITY
350
Substituting in Eqs. (179), we find u, Uz Trz
= 96a4Z + 4b4(14v - l)z = -192a4Z + 4b4{16 - 14v)z = 96a4r - 2b4(l6 -
(187)
l4v)r
Here q denotes the intensity of the uniform load and 2c is the thickness of the plate. Substituting the expressions for the stress components in these equations, we determine the four constants a 6, b6 , a 4, b. Using these values, the expressions for the stress components satisfying conditions (d) are
u, = q
Taking
= Trz = O,
Ur
= 28(1
+ v)b4Z
= !-a5(16z 6
120z 4r 2
-
+ 90z r
2 4
-
5r6 ) + b6 (8z 6
2
=
q (-
-
2lz 2r 4
+ 3r&)
Substituting in (179),
Trz
= Seª (c 2
= aa(960rz 2
-
240r3)
48 · 22(2 - v)]r2z}
(r
6[(
-672
+ 48 · 22v)z r + (432 2
12 · 22v)r 8]
= a),
= 0
for
Trz Trz
+ v~ _ 8
3(3
C3
+ v) r 2z
_
~2
+ vz
e+
3(3
+ v) a2z] C3
(188)
[2 +8 v zseª - ~8 2+5 v:e + 3(3 32+ v) a2z] c
(f)
32
c3
8
5
32
3
obtained from (187) by taking b4 = O, and a uniform tension in the z-direction "• = b, which can be obtained from (186). Thus we arrive at expressions for the stress components containing four constants a6, b6 , a4, b. These constants can be adjusted so as to satisfy the boundary conditions on the upper and lower surfaces of the plate (Fig. 199). The conditions are for for for
= q [2
(ur)r=o = q
= -192a4Z,
"·=o -q "· = =0
=O
and at the center of the plate we have
To these stresses we add the stresses u,
z2)
-
Then the final expression for u, becomes
+b
(e)
2
J~. u.z dz
= a 6 (320z 3 - 720r2z) + b6[64(2 + llv)z 3 + (504 - 48 · 22v)r2z] 3 u, = a 6 ( -640z 3 + 960r 2z) + b6 {[-960 + 32 · 22(2 - v)]z
+ [384 -
4
8e
It will be seen that the stresses u, and r,, are distributed in exactly the sarne manner as in the case of a uniformly loaded beam of narrow rectangular cross section (Art. 21). The radial stresses u, are represented by an odd function of z, and at the boundary of the plate they give bending ~oments ~niformly distributed along the boundary. To get the solut10n for a Simply supported plate (Fig. 199), we superpose apure bending stress (e) and adjust the constant b4 soas to obtain for the boundary
u,
Trz
c3
4c 3
3qr
16z4r2
-
u,
(e)
If z is the distance from the middle plane of the plate, the solution (e) represents pure bending of the plate by moments uniformly distributed along the boundary. To get the solution for a circular plate uniformly loaded, we take the stress function in the form of a polynomial of the sixth power. Proceeding as explained in the previous article, we find
q,
[2 +8v; _ 3(332+ v) r z _ ~ :] ~ + ~.:e - !) e
we have
351
z =e z= -e z =e z = -e
(d)
The el.ementary theory of bending of plates, based on the assumptions that lmear el.ement~ of the plate perpendicular to the middle plane (z = O) remam stra1ght and normal to the deftection surface of the plate 1 during bending, gives for the radial stresses at the center 3(3 Ur
=
+ v) a z 2
32
C3 q
(g)
1 This assun_iption is analogous to the plane cross sections hypothesis in the theory of bendmg of beams. The exact theory of bending of plates was developed by J. H. Michell, Proc. London M ath. Soe. vol. 31 1900 and A E H Lo "Mathematical Theory of Elasticity," 4th ed., p. 46S, 1927. · · · ve,
352
THEORY OF ELASTICITY
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
Comparing this with (f), we see that the additional terms of the exact solution are small if the thickness of the plate, 2c, is small in comparison with the radius a. It should be noted that by superposing pure bending we eliminated bending moments along the boundary of the plate, but the radial stresses are not zero at the boundary but are 32+v 2+vz (ur)r=a = q ( - 8 - C3 - S - 5 3
z)
ê
we find that the last three equations, containing shearing-stress components, remain the sarne as in the system (130), and the first three equations become [see Eqs. (e), Art. 116] 2pw2 2 1 a•e 2 (u, - uo) + -+ 1- 2 = - -r vâr 1-v 2 1 1 ae 2 1 V 2uo + 2 (u, - uo) + -1-- - = pw r +vrâr -1-v v•a. + _1_ a•e = - 2vpw• 1 + v az• 1- v
v•u, -
(h)
The resultant of these stresses per unit length of the boundary line and their moment, however, are zero. Hence, on the basis of SaintVenant's principle, we can say that the removal of these stresses does not affect the stress distribution in the plate at some distance from the edge. By taking polynomials of higher order than the sixth for the stress function, we can investigate cases of bending of a circular plate by nonuniformly distributed loads. By taking, instead of solution (f) on page 348, the other solution of Eq. (182), we can also obtain solutions for a circular plate with a hole at the center. 1 All these solutions are satisfactory only if the deflection of the plate remains small in comparison with the thickness. For larger deflections the stretching of the middle plane of the plate must be considered. 2
353
(b)
We begin with a particular solution of Eqs. (189), satisfying the compatibility equations. On this solution we superpose solutions in the forro of polynomials (184) and (185) and adjust the constants of these polynomials so as to satisfy the boundary conditions of the problem. For the particular solution we take the expressions uo = Cr 2
+ Dz
2,
.... =o
(e)
It can be seen that these expressions satisfy the second of the equations of equilibrium. They also satisfy the compatibility equations which contain shearingstress components [see Eqs. (f) and (g), Art. 116]. It remains to determine the constants A, B, C, D, so as to satisfy the remaining four equations, namely the first of Eqs. (189) and Eqs. (b). Substituting (e) in these equations, we find
+
A = pw 2 (1 3v) 6v '
B=
pw•
D = _ pw2 (1
e= o,
-3,
+
2v)(l 6v(l - 11)
+ v)
The particular solution is then 119. The Rotating Disk as a Three-dimensional Problem. In our previous discussion (Art. 30) it was assumed the stresses do not vary through the thickness of the disk. Let us now consider the sarne problem assuming only that the stress distribution is symmetrical with respect to the axis of rotation. The differential equations of equilibrium are obtained by including in Eqs. (177) the centrUugal force. Then âu, + + pw r ar Tz + - - ruo ÔTn
a.... + au. + ar az r
7'rz
2
=
= _ pw 2 r• _ pw 2 (1
3
u. ==
pw 2 (1
Q 7'ra
(189)
o
z =o
(a)
A number of solutions for a circular plate symmetrically loaded have been discussed by A. Korobov, Bull. Polytech. Inst., Kiew, 1913. Similar solutions were obtained independently by A. Timpe, Z. angew. Math. Mech., vol. 4, 1924. 2 See Kelvin and Tait, "Natural Philosophy," vol. 2, p. 171, 1903.
+ 3v) r 6v
pw 2 (1
_
=
where p is the mass per unit volume, and w the angular velocity of the disk. The compatibility equations also must be changed. Instead of the system (130) we shall have three equations of the type (f) (see page 231) and three equations of the type (g). Substituting in these equations the components of body force, 1
6v(l - v) 2
(190)
+ 2v)(l + v)
611(1 -
-
+ 2v)(l + v) zl
v)
Z
2
=O
This solution can be used in discussing the stresses in any body rotating about an axis of generation. ln the case of a circular disk of constant thickness we superpose on the solution (190) the stress distribution derived from a stress function having the forro of a polynomial of the fifth degree [see Eqs. (184), (185)], > =
a5(8z 6 - 40r 2z3
+ 15r 4z) + b6(2z6
- r•z3 - 3r•z)
(d)
Then, from Eqs. (179), we find
+
+
+ +
a, = -as(l80r 2 - 240z 2) b6 [(36 - 54v)r• (1 18v)6z•] u, = -a.(-240r 2 480z 2 ) b.[(96 - 108v)z2 (-102 54v)r•] 2 09 = a,(-60r 240z 2 ) b5 [(6 108v)z• (12 - 54v)r•] r., = 480a5rZ - b.(96 - l0811)rz
+
+
+
+
+
+
+
(e)
llSTtTUlllL %UTEMlth t 1~ -1 i S O A l< A 81BllOTetA CENTRA~~ ~-
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
THEORY OF ELASTICITY
354
Adding this to the stresses (190) and determining the constants make the resultant stresses r., and
=
-pw
2
+ + 3- 8 +- r +8 3v) r + 2(1 v(l + v) •] - v) z
[v(l v) 2 2(1 - v) z
2
[(1
ª• and b1 soas to
V
-e
2
r=a
+
(192)
! ]
To eliminate the resultant radial compression along the boundary, i.e., to make
(! º
ef>1 = A1(r2 + z 2)-t ef>2 = A2z(r 2 z2)-i 2 2 ef>s = A 3 [z (r + z2)-t - i(r 2 + z2)-i]
355
=O
which are also solutions of Eq. (181). By multiplying expressions (192) by r2 + z2 (see page 349), we obtain another series of solutions of Eq. (181), namely, ef>1 = B1(r 2 + z2)! ef>2 = B2z(r 2 + z 2)-l (193)
we superpose on (f) a uniform radial tension of magnitude pw• (
8
3
+ v)a + pw 2
2
+
v(l v) ~ 2(1 - v) 3
Then the final stresses are
-
• [3
+ v (a• 8
[3
+V 2 8 a
2
tT,
=
O,
+ v6(1(1 + v) (e• -3z')] - v) 1 + 3v 2 + v(l + v) (c2 8 r 6(1 - v)
- r•) -
'Trz
(191)
3z')]
q, =
120. Force at a Point of an Inde:finitely Extended Solid. ln .discussing this problem we use again Eq. (182) on page 347. By taking m 1 = -(n 1) [see Eq. (e), page 348], we obtain the second integral of Eq. (183) in the form of the following series:
+
n -
a
[
-(n+l) X
B(r 2
+ z )l 2
=O
Comparing this with the previous solution (55), we have here additional terms with the factorl (e• - 3z'). The corresponding stresses are small in the case of a thin disk and their resultant over the thickness of the disk is zero. If the rim of the disk is free from external forces, solution (191) representa the state of stress in parts of the disk some distance from the edge. . . The stress distribution in a rotating disk having the shape of a fiat ellipso1d of revolution has been discussed by C. Chree. 2
'11 _
Each of the solutions (192) and (193), and any linear combination of them, can be taken as a stress function, and, by a suitable adjustment of the constants A1, A 2, • • • , B 1, B 2, • • • , solutions of various problems may be found. For the case of a concentrated force we take the first of the solutions (193) and assume that the stress function is
+ (n + l)(n + 2) x-
.. ·]
Taking n equal to -1, -2, -3, . . . , we obtain from this the following particular solutions of Eq. (182): 2 1 These terms are of the sarne nature as the terms in z found in Art. 84. Equations (191) represent a state of plane stress since
where Bis a constant to be adjusted later. Substituting in Eqs. (179), we find the corresponding stress components
u, = B[(l u9 = B(l u, = -B[(l Trz = -B[(l
2v)z(r 2 + z2)-i - 3r 2z(r 2 + z2)-t] 2v)z(r 2 + z2)-i - 2v)z(r 2 + z2)-f + 3z 3 (r 2 + z 2)-t] - 2v)r(r 2 + z2)-i + 3rz 2(r2 + z2)-1]
(194)
All these stresses approach infinity when we approach the origin of coordinates, where the concentrated force is applied. To avoid the necessity of considering infinite stresses we suppose the origin to be the center of a small spherical cavity (Fig. 200), and consider forces over the surface of the cavity as calculated from Eqs. (194). It can be shown that the resultant of these forces represents a force applied at the origin in the direction of the z-axis. From the z 2 condition of equilibrium of a ring-shaped Fm. ºº· element, adjacent to the cavity (Fig. 200), the component of surface forces in the z-direction is
Z
= -(r., sin 1/1
+ u, cos 1/;)
duced at any point by the force P applied at the origin O are determined by Eqs. (194) and (195) of the previous article. By using the sarne equations, the stresses produced by the force P at 0 1 can also be calculated. Remembering that the second force is acting in the opposite direction and considering the distance d as an infinitely small quantity, any term f(r,z) in expressions (194) should be replaced by -[f + (àf/àz)d]. Superposing the stresses produced by the two forces and using the symbol A for the product Bd, we find
Using Eqs. (194) and the formulas sin if! = r(r we find that
Z
=
2
+
z 2 )-~,
cos if! = z(r
B[(l - 2v)(r2 + z2)- 1
2
+ z )-i 2
+ 3z2(r + z2)2
2 ]
The resultant of these forces over the surface of the cavity is 2
"
fo 2 Z VT2+Z2 · di/! · 211"r
= 8B11"(1 -
v)
The resultant of the surface forces in a radial direction is zero, from symmetry. If P is the magnitude of the applied force, we have
P = 8B11"(1 - v)
p
811"(1 - v)
(195)
into Eqs. (194), we obtain the stresses produced by a force P applied at the origin in the z-direction. 1 This solution is the three-dimensional analogue of the solution of the two-dimensional problem discussed in Art. 38. Substituting z =O in Eqs. (194), we find that there are no normal stresses acting on the coordinate plane z = O. The shearing stresses over the sarne plane are P(l - 2v) B(l - 2v) Tra = r2 811"(1 - v)r2
(a) '~-----4-r
These stresses are inversely proportional to the square of the distance r from the point of application of the load. 121. Spherical Container under Internal or Extemal Uniform Presz sure. By superposition we can get FIG. 201. from the solution of the previous article some new solutions of practical interest. W e begin with the case of two equal and opposite forces, a small distance d apart, applied to an indefinitely extended elastic body (Fig. 201). The stresses proThe solution of this problem was given by Lord Kelvin, Cambridge and Dublin See also his "Mathematical and Physical Pap~rs," vol. 1, P· 37. From his solution it follows that the displacements correspondmg to t?e stresses (194) are single-valued, which proves that (194) is the correct solut1on of the problem (see Art. 82). 1
Math. J., 1848.
'1r
à [(1 - 2v)z(r2 + z2)-i - 3r 2z(r 2 àz à -A- [(1 - 2v)z(r2 + z2)-i] àz
= -A -
"º =
Substituting
B =
357
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
THEORY OF ELASTICITY
356
"• = A Tn
à
+ z2)-t] (196)
+ z2)-i + 3zª(r2 + z2)-t] 2v)r(r 2 + z2)-i + 3rz 2(r2 + z2)-t]
àz [(1 - 2v)z(r2 à
= A àz [(1 -
Let us consider (Fig. 201) the stress components UR and TR>/I acting ata point M on an elemental area perpendicular to the radius OM, the length of which is denoted by R. From the condition of equilibrium of a small triangular element such as indicated in the figure we find 1 uR TR>/I
=
u, sin 2 i/!
+ u, cos 2 i/! + 2r,. sin i/! cos i/!
= (u, - u,) sin if! cos if!
(a)
- T,,(sin 2 if! - cos 2 if!)
Using (196), and taking sin if! = r(r 2
+ z2)-!
= !...,
cos i/! = z(r 2
R
+ z2)-!
z = -
R
we obtain
-
+ v)A [ - Slll . 2 ,f, + 2(2 - v) 2 ,f,] 1 + " cos 2(1 + v)A . .t, .t, Rª sm cos
2(1
Rª
't'
'f'
't'
(b)
'f'
The distribution of these stresses is symmetrical with respect to the z-axis and with respect to the coordinate plane perpendicular to z. Imagine now that we have at the origin, in addition to the system of two forces P acting along the z-axis, an identical system along the r-axis and another one along the axis perpendicular to the rz-plane. 1 The stress components u 9, acting on the sides of the element in the meridional sections of the body, give a small resultant of higher order and can be neglected in deriving the equations of equilibrium.
358
THEORY OF ELASTICITY
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
By virtue of the symmetry stated above, we obtain in this way a stress distribution symmetrical with respect to the origin. If we consider a sphere with center at the origin, there will be only a normal uniformly distributed stress acting on the surface of this sphere. The magnitude of this stress can be calculated by using the first of Eqs. (b). Considering the stress at points on the circle in the rz-plane, the first of Eqs. (b) gives the part of this stress due to the double force along the z-axis. By interchanging sin 1/; and cos 1/1, we obtain the normal str~ss round the sarne circle produced by the double force along the r-axis. The normal stress dueto the double force perpendicular to the rz-plane is obtained by substituting 1/; = 11" /2 in the sarne equation. Combining the actions of the three perpendicular double forces we find the following normal stress acting on the surface of the sphere:
Then (pi - Po)a 3b3 a 3 - b3 D = Pob 3 - pia 3 a 3 - b3 Pob 3 (R 3 - a 3)
=
(197)
<1R
(e)
The combination of the three perpendicular double forces is called a center of compression. We see from (e) that the corresponding compression stress in the radial direction depends only on the distance from the center of compression and is inversely proportional to the cube of this distance. This solution can be used for calculating stresses in a spherical container submitted to the action of interna! or external unif orm pressure. Let a and b denote the inner and outer radii of the sphere (Fig. 202), and Pi and Po the internal and the externa! uniform pressures. Superposing on (e) a uniform tension or com2 2 · Fm. pression in all directions, we can take a general expression for the radial normal stress in the form
º
<1R
=
e Rª +D
(d)
C and D are constants the magnitudes of which are determinedfrom the conditions on the i~ner and outer surfaces of the container, which are
e
- +D= aª
-p.,
Pia (b R 3 (a 3 3
+
3
-
-
R 3) b3)
The pressures Po and Pi also produce in the sphere normal stresses
+
2
4(1 - 2v)A = Rª
359
+
2
(e)
Using expression (197) for
"t
Pob 3 (2Rª + a 3) = 2R 3 (a 3 - b3)
Pia 3 (2R 3 + b3) 2R 3 (a 3 - b3 )
(198)
If Po = O, then p;a 3 (2R 3 "t
= 2R 3 b3
-
+b
3)
a3
It will be seen that the greatest tangential tension in this case is at the inner surface, at which
(ut)max.
p; 2a 3
+ b3
= 2 bª - aª
Ali these results are due to Lamé. 1 z 122. Local Stresses around a Spherical Fm. 203. Cavity. As a second example consider the stress distribution around a small spherical cavity in a bar submitted to uniform tension of magnitude S (Fig. 203). 2 ln the case of a solid bar in tension, the normal and shearing components of stress acting on a spherical surface are 1
Lamé, "Leçons sur la théorie . . . de l' élasticité," Paris, 1852. Solution of this problem is dueto R. V. Southwell, Phil. Mag., 1926; see also J. N. Goodier, Trans. A.S.M.E., vol. 55, p. 39, 1933. The triaxial ellipsoidal cavity is considered by E. Sternberg and M. Sadowsky, J. Applied Mechanics (Trans. A.S.M.E.), vol. 16, p. 149, 1949. 2
360
THEORY OF ELASTICITY = S cos 2 if;,
TR>/I
(a)
= -S sin if; cos if;
To get the solution for the case of a small spherical cavity of radius a, we must superpose on the simple tension a stress system which has stress components on the spherical surface equal and opposite to those given by the Eqs. (a), and which vanishes at infinity. Taking from the previous article the stresses (b), dueto a double force in the z-direction, and the stresses (e), dueto a center of compression, the corresponding stresses acting on the spherical surface of radius a can be presented in the following form: 2 (1 +. v)A sinif;cosif; _ 2 ( 1 + v)A (-1 + 5 - "cos2 tr..jl 1 = (b) a• 1 + 11 a 11 un" = !!_, TR>/1 =O (e) a•
y;),
where A and B are constants to be adjusted !ater. It is seen that, combining stresses (b) and (e), the stresses (a) produced by tension cannot be made to vanish and that an additional stress system is necessary. Taking, from solutions (192), a stress function
u. =
us = T,.
=
~~ (1
~~
~~ (1
+ 35 sin 2 if; cos
2
111
A(l - 2v) 5(1 - 2v)S a 3 r• = 2(7 - 5v) Ti From Eq. (e) of the previous article,
u. = u,
TR·'·
"
=
(l)
S(l - 5v) a 3 2(7 - 5v) Ti
(m)
=
S [1
+ 2(7 4 -
511 a
3
5v) Ti
+ 2(7 -9
6
a 5v) Ti
J
(n)
At r = a, we find 27 - 15v (u.)max. = 2(7 - 5v) S
Taking v = 0.3, (u,)max. =
=
f
24 a
Sin
if;
COS
if;
(e)
+ 11)A - 2(5 - 11) :'!:.. cos2 if; + ~ - 12; + 36f cos2 if; a• a a a a 2(1 + 11)A . 24C . Sill if; COS if; + Sill if; COS if;
(o)
i!S
The maximum stress is thus about twice as great as the unüorm tension S applied to the bar. This increase in stress is of a highly localized character. With increase of r, the stress (n) rapidly approaches the value S. Taking, for instance, r = 2a, v = 0.3, we find u, = 1.0548. ln the sarne manner we find, for points in the plane z = O,
Combining stress systems (b), (e), (e) we find CTR
(h)
u/'
, + u,,, + u,,,, + S
if;)
J;~ (-3 sin if; cos if; + 7 sin if; cos 3 if;)
TR.jl
s = 2(7 - 5v)
The total stress on the plane z = Ois
+ 35 cos' if;)
if;),
ã6
7 - 511 '
B
1
2
=
From Eqs. (196), for z =O,
- 5 cos 2 if;)
f (-1 + 3 cos
e
S(l - 5v)
ã3
The complete stress at any point is now obtained by superposing on the simple tension S the stresses given by Eqs. (d), the stresses (196) due to the double force, and the stresses dueto the center of pressure given by Eqs. (e) and (e) oí the previous article. Consider, for instance, the stresses acting on the plane z = O. From the condition of symmetry there are no shearing stresses on this plane. From Eqs. (d), substituting if; = 7r/2 and R = r, 9C 9Sa 6 (k) u,' = Ti" = 2(7 - 5v)r6
(d)
12 a
B
5S = 2(7 - 511)'
tr.' 11 = (ut).-o == - 2r 3
Using now Eqs. (a) oí the previous article, the stress components acting on a spherical surface of radius a are
un"' =
A
ã3
2
- 5 cos 2 if; - 5 sin 2 if;
(3 - 30 cos 2 if;
írom which
+ z )--i
the corresponding stress components, from Eqs. (179), are u, =
361
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
_ 3C _ A(l - 2v) _ .!!.._ (u9) •-• - r• r• 2rª
= 2(1 = -
U>
a6
a•
Superposing these stresses on the stresses (a), the spherical surface of the cavity becomes free from forces if we satisfy the conditions 2(1
+ 11)A + !!_ a•
a•
-2(5 - 11) :'!:_3 a _ 2(1 + 11)A a3
_ 12C
=
a5
+ 36C = a• + ~4C = a'
Substituting from Eqs. (h) and taking r = a, we find that the tensile stress along the equator (if; = 7r /2) of the cavity is 15v - 3 (u9)•-D, r-a = 2(7 - 5v) 8
At the pole of the cavity (if; =O or if; = 7r) we have
0
u, -S S
(g)
= u8 =
2(1 - 2v)A 12C a• - 7
-
+
B _ 3 15v S 2a 3 = 2(7 - 5v)
Thus the longitudinal tension S produces compression at this point. Combining a tension 8 in one direction with compression S in the perpendicular iirection we can obtain the solution for the stress distribution around a spherical
362
THEORY OF ELASTICITY
cavity in the case of pure shear. 1 shearing stress is
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
It can be shown in this way that the maximum
'Tmax.
=
15(1 - v) S 7 - 5v
(p)
The results of this article may be of some practical interest in discussing the effect of small cavities 2 on the endurance limit of specimens submitted to the action of cyclical stresses.
123. Force on Boundary of a Semi-infinite Body. Imagine that the plane z = O is the boundary of a semi-infinite solid and that a force P is acting on this plane along the z-a:xis (Fig. 204). 3 It was shown in Art. 120 that the stress distribup tion given by Eqs. (194) and r-----r---::::oon--.:::--------,r (195) can be produced in a semiinfinite body by a concentrated force at the origin and by shearing forces on the boundary plane z = O, given by the equation Trz
= -
B(l - 2v)
r2
(a)
To eliminate these forces and arrive at the solution of the FIG. 204. problem shown in Fig. 204, we use the stress distribution corresponding to the center of compression (see page 358). ln polar coordinates this stress distribution is
363
Assume now that centers of pressure are uniformly distributed along the z-a:xis from z = O to z = - oo. Then, using the principle of superposition, the stress components produced in an indefinitely extended solid are, from Eqs. (199), ur
= A
J."' (r 2 -
iz 2)(r 2
Trz
u9
J."' (z àr 2
3
-
2 )
2
~ [~ - ~ (r 2 + z2)-l -
= cr. = A
+ z )-i! dz
(r 2
+ z)-i! dz 2
A
("'
1 !."'
= - - A
2
•
(r 2
+ z )-i dz 2
~ z(r 2 + z )-i!
=
= zA } z rz(r 2 + z2)-!! dz = 2 r(r2
2
2
(200)
+ z2)-i!
A[lr
= - -
z(r 2 + z2)-i]
-2 -
-z (r2 r2
+ z2)-l J
Considering the plane z = O we find that the normal stress on this plane is zero, and the shearing stress is (b)
.z
dcr11 R
A
CJ'R
= R3'
O't
= dR 2
+
1 A
CJ'R
= - 2R3
in which A is a constant. ln cylindrical coordinates (Fig. 204) .we have the following expressions for the stress components: crr = cr11 sin 2 1f; +
+
cr9
+
-B(l - 2v)
2
1 This problem was discussed by J. Larmor, Phil. Mag., series 5, vol.·33; 1892.. See also A. E. H. Love, "Mathematical Theory of Elasticity," 4th ed., p. 252, 1927. 2 Such cavities are, for instance, present in a weld, and fatigue experiments show that cracks usually begin at these cavities. 3 The solution of this problem was given by J. Boussinesq, see "Application des potentiels . . . ,'' Paris, 1885. The solution for a force at an interna! point of the semi-infüüte body was found by R. D. Mindlin, Physics, vol. 7, p. 195, 1936.
+ -A2
=O
from which
A
+
=
It appears now that by combining solutions (194) and (200), we can, by a suitable adjustment of the constants A and B, obtain such a stress distribution that the plane z = O will be free from stresses and a concentrated force P will act at the origin. From (a) and (b) we see that the shearing forces on the boundary plane are eliminated if
=
2B(l - 2v)
Substituting in expressions (200) and adding together the stresses (194) and (200), we find
= B { (1 - 2v) [~ cr. = -3Bz 3 (r 2 + z2)-t ur
~ (r + z2)--l] - 3r2z(r2 + z2)-t} 2
1 z ue = B(l - 2v) [ - r2 + r2 (r 2 -r,.. = -3Brz 2 (r 2 + z2)-t
.,
+ z2)-l + z(r2 + z2)-f J
(e)
364
3ô5
THEORY OF ELASTICITY
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
This stress distribution satisfies the boundary conditions, since -r., = O for z = O. It remains now to determine the constant B so that the forces distributed over a hemispherical surface with center at the origin are statically equivalent to the force P acting along the z-axis. Considering the equilibrium of an element such as shown in Fig. 200, the component in the z-direction of forces on the hemispherical surface is Z = -(Trz sin i/; +
The stress is thus inversely proportional to the square of the distance from the point of application of the load P. Imagine a spherical surface of diameter d, tangent to the plane z = O at the origin O. For each point of this surface,
"• =
For determining B we obtain the equation
,,. P = 2'11"
,..
{2 -
lo Zr(r
+
2
z2)l dtf
=
67rB
r 2 + z2 = d 2 cos 2 t/I
Substituting in (202) we conclude that for points of the sphere the total stress on horizontal planes is constant and equal to 3P/27rd 2. Consider now the displacements produced in the semi-infinite solid by the load P. From Eqs. (178) for strain components,
{2
lo cos 2 tf sin tf dif; = 27rB
u = Esr =
from which p
Finally, substituting in (e) we obtain the following expressions for the stress components due to a normal force P acting on the plane boundary of a semi-infinite solid:
=;:. {
(1 - 2v)
3P
"• = - -27r
z 3 (r 2
[~ - ~ (r 2 + z2)-l] - 3r2z(r2 + z2)-t}
_ /
i..:
-
V
2
+
[ z(r
2 + z2)-l - 1
+ -1 -1- 2v r 2z(r2 + z
)-t]
2
(203)
aw 1 az = Ez = E [u. - v(
+ z2)-l + z(r + z2)-i }
(201)
aw ar = 'Yrz -
2
3P z2 3P cos 2 i/; Trz - 27r (r2 + z2)2 = 2r (r2 + z2) 2 _
+ u.)]
(1 - 2v)(l + v)P 2'll"Er
2
au az
=
+
au az
E
Substituting for the stress components, and for the displacement u the. values found above, we obtain
Hence the direction of the resultant stress passes through the origin O. The magnitude of this resultant stress is _
- v(u,
For determining vertical displacements w, we have, from Eqs. (178),
This solution is the three-dimensional analogue of the solution for the semi-infinite plate (see Art. 33). If we take an elemental area mn perpendicular to z-axis (Fig. 204), the ratio of the normal and shearing components of the stress on this element, from Eqs. (201), is
~
u =
+ z )-i
us = -p (1 - 2v) { - -12 + -z2 (r 2 2'11" r r 3P 2 2 -r., = '11" rz (r + z2)-t 2
Er [us
Substituting the values for the stress components from Eqs. (201),
B=2'11"
(e)
(202)
+ v)r z(r + z2)-i - [3 + v(l - 2v)]z(r2 + z2)-I} P(l + v) [2(1 - v)r(r + z2)-t + 3rz (r + z2)-t] 27rE
aw = _!__ {3(1 az 27rE aw = ar
2
2
2
2
2
from which, by integration,
For the boundary plane (z = O) the displacements are
(w)z=O
=
P(l - v2) E 7r r
(205)
366
THEORY OF ELASTICITY
367
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
showing that the product wr is constant at the boundary. Hence the radii drawn from the origin on the boundary surface, after deformation, are hyperbolas with the asymptotes Or and Oz. At the origin the displacements and stresses become infinite. To eliminate the difficulties in applying our equations we can imagine the material near the origin cut out by a hemispherical surface of small radius and the concentrated force P replaced by the statically equivalent forces distributed over this surface.
124. Load Distributed over a Part of the Boundary of a Semiinfinite Solid. Having the solution for a concentrated force acting on the boundary of a semi-infinite solid we can find the displacements and the stresses produced by a distributed load by superposition. Take, as a simple example, the case of a uniform load distributed over the area of a circle of radius a (Fig. 205), and consider the deflection, in the direction of the load, of a point M on the surface of the body ata distance r from the center of the circle. Taking a small element of the loaded area shown shaded in the figure, bounded by two radii including the angle dlf; and two FIG. 205. ares of circle with the radii s and s + ds, all drawn from M, the load on this element is qs dlf; ds and the corresponding deflection at M, from Eq. (205), is
from which dlf; = a cos 8 d8 = r cos 1f;
r
a cos 8 d8 a2 1 - - sin 2 r2
~
(}
Substituting in Eq. (a) and remembering that 8 varies from O to ~;2, when 1f; changes from O to 1/; 1, we find w = 4(1 -
v2)q
· ~E
[
(~
}o "
1- ª
f
r
2 2
(~ a 2 cos 2 (} d(} = }o r Vl - (a2 /r 2) sin2 8 sin 2
(}d(} -
4(1 _ 11 2)qr ~E
(1 - ~) ri vl r
2
}o
J
d(}
(a 2 /r 2 ) sin 2
(206)
(}
The integrais in this equation are known as elliptic integrals, and their values for any value of a/r can be taken from tables.1
---'--·r
(1 - 112)q . s d1f; ds = (1 - v2)q d1f; ds ~E
~E
s
(aJ
The total deflection is now obtained by double integration, w = (1
~;2)q
FIG. 206.
JJ
dlf;ds
To get the deflection at the boundary of the loaded circle we tah r = a in Eq. (206) and find · '
Integrating with respect tos and taking into account the fact that the length of the chord mn is equal to 2 a 2 - r 2 sin 2 "'we find
v
w=
4(1 - v2 )q
~"''
~E
o
va
2,
2
-
r sin
2
"'ª"'
(a)
in which 1f;1 is the maximum value of lf;, i.e., the angle between r and the tangent to the circle. The calculation of the integral (a) is simplified by introducing, instead of the variable lf;, the variable angle 8. From the figure we have asin8=rsinlf;
(w ) r=a
_ 4(1 - v2 )qa -
.
~E
(207)
. If th~ point M is within the loaded area (Fig. 206a), we again cons1der the deflection produced by a shaded element on which the load qs ds dlf; acts. Then the total deflection is
w
= (1
~;2)q
ff
ds dlf;
S~e for instance, E. Jahnke and F. Emde, "Funktionentafeln," Berlin, 1909; . or Pe1rce, "Short Table of Integrais," 1910. 1
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
THEORY OF ELASTICITY
368
The length of the chord mn is 2a cos 8, and i/t varies from zero to ,,,.;2, so " w = 4(1 - v2)q Íc2 a cos () dift
,,,.E
"
4(1 - v2)qa Íc2 ~1 ,,,.E o
r2 .
--S i n 2 ·'· 'I' a2 .
d·'· 'I'
(208)
Comparing this with the deflection at the boundary of the circle we find that the latter is 2/,,,. times the maximum deflection. 1 It is interesting to note that for a given intensity of the load q the maximum deflection is not constant but increases in the sarne ratio as the radius of the loaded circle. By using superposition the stresses can also be calculated. Consider, for example, the stresses at a point on the z-axis (Fig. 206b). The stress"• produced at such a point by a load distributed over a ring area of radius r and width dr is obtained by substituting in the second of Eqs. (201) 2,,,.r drq instead of P. Then the stress"• produced by the uniform load distributed over the entire circular area of radius a is =
- loª 3qr drz (r 2 + z 3
2
)-t
= qz 3 [(r2 + z2
)-i1:
= q
[-l + (a2 ~ z2)il]
= qr dq, ,,,. dr { (1
(b)
The solution of this problem was given by Boussinesq, Zoe. eit. See also H. Lamb, Proe. London Math. Soe., vol. 34, p. 276, 1902; K. Terazawa, J. Coll. Sei., Univ. Tokyo, vol. 37, 1916; F. Schleicher, Bauingenieur, vol. 7, 1926, and Bauingenieur, vol. 14, p. 242, 1933. A complete investigation of this problem, also of the case in which the load is distributed over a rectangle, is given in the paper by A. E. H. Love, Trans. Roy. Soe. (London), series A, vol. 228, 1929. Special properties of the deformation and stress in the general case are pointed out by S. Way, J. Applied Meehanies (Trans. A.S.M.E.), vol. 7, p. A-147, 1940.
1 - 2v) [ T2
- T2z (r 2 + z 2)-! ]
dr (1 - 2v) [ - -1 d
+ -r2z (r2
- 3r 2z(r 2 + z2)-t }
+ z )-l + z(r 2
+ z )-i! J
2
(e)
2
The normal stresses produced on the sarne planes by the elemental loads at points 3 and 4 are
)-t]
du/' = qr dq, ,,,. dr (1 - 2v) [ - .!__ r2
+ !..r2 (r 2 + z2)-i + z(r + z2
dr { (1 - 2v) [ T2 1 d
ri
-
2
z (r 2 + z2)-t ] - 3r 2z(r 2 + z2)-t }
(d)
By summation of (e) and (d) we find that the four elemental loads, indicated in the figure, produce the stresses
ª"• = d
qr dq, dr
,,,.
[-2(1
2v)z(r 2 + z2)-! - 3r2z(r 2 + z2)-t]
+ v)z(r 2 + z2)-il + 3z (r 2 + z2)-t]
(e)
3
To get ihe stresses produced by the entire load uniformly distributed over the area of a circle of radius a we must integrate expression (e) with respect to
"' =
This stress is equal to -q at the surface of the body and gradually decreases with increase of distance z. ln calculating the stresses"' and 1
area (Fig. 206b) with the loads qr dq, dr. The stresses produced by these two elemental loads at a point on the z-axis, from the first and third of Eqs. (201), are
d
Thus the deflection can easily be calculated for any value of the ratio r/a by using tables of elliptic integrais. The maximum deflection occurs, of course, at the center of the circle. Substituting r = O in Eq. (208), we find 2(1 - v2 )qa (209) (w)max. = E
<1z
"º at the sarne point, consider the two elements 1 and 2 of the loaded
o
or, since a sin O = r sin i/t, we have W=
369
=
"º
2
=
~loª [-2(1 + v)z(r 2 + z )-f + 3z (r + z )-!]r dr 2
3
2
[-(l + 211) + y2(1a2++v)zz2 _ ( va2z+ z2)3]
2
(f)
For the point O, the center of the loaded circle, we find, from Eqs. (b) and (f), "· = -q,
<1r
=
<18
=
q(l
+ 2v) 2
Taking v = 0.3, we have "• = "º = -0.8q. The maximum shearing stress at the point O, on planes at 45 deg. to the z-axis, is equal to O.lq. Assuming that yielding of the material depends on the maximum
THEORY OF ELASTICITY
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
shearing stress, it can be shown that the point O, considered above, is not the most unfavorable point on the z-axis. The maximum shearing stress at any point on the z-axis (Fig. 206b), from Eqs. (b) and (f), is
in which m is a numerical factor depending on a, A is the magnitude of this area, and P is the total load.
370
TABLE OF FACTORS
1
q [1 - 2v 2 (ue - CTz) = 2 - 2 - + (1
+ v)
z
ya2
+ z2
-
3(
z
2 ya2
).
+ z2
3 ]
+ z2
Rectangles with various a =
1.5 m=
3
from which 2(1 + v) 7 - 2v
(h)
Substituting in expression (g), Tmax.
=
ª[
2
l -; v +
~ (1 +
(k)
v) y2(1 + v)]
Assuming v = 0.3, we find, from Eqs. (h) and (k),
z = 0.638a,
Tmax.
= 0.33q
This shows that the maximum shearing stress for points on the z-axis is at a certain depth, approximately equal to two-thirds of the radius of the loaded circle, and the magnitude of this maximum is about onethird of the applied uniform pressure q. For the case of a uniform pressure distributed over the surface of a square with sides 2a, the maximum defiection at the center is
- 81
Wmax. -
;;:
ogn
(- !
..,
+ 1) qa(l E-
v2)
= 2 .24 qa(l E-
v2)
= 1.90
qa(l -
E
v2 )
!i
l
'
1
ili
2
3
5
10
100
0.94
(211)
0.92
0.88
0.82
0.71
0.37
Several values of the factor m are given in the table. It will be seen that for a given load P and area A defiections increase when the ratio of the perimeter of the loaded area to the area decreases. Equation (212) is sometimes used in discussing defiections of foundations 1 of engineering structures. ln order to have equal defiections of various portions of the structure the average pressure on the foundation must be in a certain relation to the shape and the magnitude of the loaded area. It was assumed in the previous discussion that the load was given, and we found the displacements produced. Consider now the case when the displacements are given and it is necessary to find the corresponding distribution of pressures on the boundary plane. Take, as an example, the case of an absolutely rigid die in the forro of a circular cylinder pressed against the plane boundary of a semi-infinite elastic solid. ln such a case the displacement w is constant over the circular base of the die. The distribution of pressures is not constant and its intensity is given by the equation 2 p 21ra ya 2 - r 2
(213)
in which P is the total load on the die, a the radius of the die, and r the distance from the center of the circle on which the pressure acts. This distribution of pressures is obviously not uniform and its smallest value is at the center (r = O), where
p qmin.
= 2'1T'a2
i.e., it is equal to half the average pressure on the circular area of contact. At the boundary of the sarne area (r = a) the pressure becomes 1
See Schleicher, Zoe. cit.
0.95
q =
Analogous calculations have also been made for uniform pressure distribution over rectangles with various ratios, a = a/b, of the sides. All the results can be put in the form 1 P(l - v2) (212) Wave,. = m E VA 1
0.96
(210)
The defiection at the corners of the square is only half the deflection at the center, and the average deflection is Wavor.
'fi
- - - - - -- - - - - - - - -
= .!_ v~2c~1-+~v)
z =a
EQ. (212)
Square
This expression becomes a maximum when
z
IN
(g) Circle
ya2
m
371
2
See Schleicher, Zoe. eit. This solution was given by Boussinesq, loe. cit.
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
373
THEORY OF ELASTICITY
372
infinite. In actual cases we shall have yielding of material along the boundary. This yielding however is of local character and. does not substantially affect the distribution of pressures (213) at pornts some distance from the boundary of the circle. The displacement of the die is given by the equation
w =
P(l - v2)
(214)
2aE
We see that, for a given value of the average ~nit press~re on the boundary plane, the deflection is not constant but rncreases rn the sarne ratio as the radius of the die. For comparison we give also the average deflection for the case of a uniform distribution of pressures [Eq. (208)]:
_ Joª w27rr dr Waver. -
= _!Q_ P(l - v2) = 0. 54 P(l 37!"2
11"ª2
aE
~, v2)
aE
/z
= 2Ri'
Z2
=
2R2
+ z2
=
1 r2 ( 2Ri
r 2 (Ri
1 )
+ 2R2
=
In the case of contact between a ball anda spherical seat (Fig. 208b), Ri is negative in Eq. (b), and
z
r2 (R1 - R2) 2 -
Zi =
(e')
2R1R2
If the bodies are pressed together along the normal at O by a force P, there will be a local deformation near the point of contact producing contact õver a small surface with a circular boundary, called the surface of contact. Assuming that the radii of curvature R1 and R2 are very large in comparison with the radius of the boundary of the surface of
/ IZ2
~~ Y/.
(6)
faJ
ZJ Zz
Frn. 208.
contact, we can apply, in discussing local deformation, the results obtained before for semi-infinite bodies. Let w1 denote the displacement dueto the local deformation in the direction Zi of a point such as M on the swface of the lower ball (Fig. 207), and W2 denote the sarne displaceme:rit in the direction z2 for a point such as N of the upper bali. Assuming that the tangent plane at O remains immovable during local compression, then, due to this compression, any two points of the bodies on the axes z1 and z2 at large distances 1 from O will approach each other by a certain amount a, and the distance betweén two points such as M and N (Fig. 207) will diminish by ~ - (w1 + w2). If finally, dueto local compression, the points M and N come i:o.side the surface of contact, we have (d)
and the mutual distance between these points is Zi
(e)
(215 )
This average deflection is not very much different from the displacement (214) for an absolutely rigid die. . 125. Pressure between Two Sphencal Bodies in c.ontact. The r~sul~s of ~he 2 R2 previous article can be used rn d1scussrng the pressure distribution between two bodies in contact. i W e assume that at Zi the point of contact these bodies have spherical surfaces with the radii Ri and R 2 (Fig. 207). If there is no pressure between the bodies we have contact at one point O. The distances from the plane tangent at O of points such as M Frn. 207. and N, on a meridian section of the spheres at a very small distance 2 r from the axes Zi and z2, can be represented with sufficient accuracy by the formulas r2 r2 (a) Zi
In the particular case of contact between a sphere and a plane (Fig. 208a), Ri = oo; and Eq. (b) for the distance MN gives
+
R2) 2RiR2
(b)
1 This problem was solved by H. Hertz, J. Math. (Crelle's J.), vol. 92, 1881. See also H. Hertz, "Gesammelte Werke," vol. 1, p. 155, Leipzig, 1895. 2 r is small in comparison with R1 and R2.
in which fJ is a constant depending on the radii R 1 and R 2 and given by Eq. (b), (e), or (e'). Thus from geometrical considerations we find for 1 Such distances that deformations due to the compression at these points can be neglected.
THEORY OF ELASTICITY AXIALLY SYMMETRICAL STRESS DISTRIBUTION
375
any point of the surlace of contact, Wi
+ W2
= a - {3r 2
(e)
Let us now consider local deformations. From the condition of symmetry it can be concluded that the intensity q of pressure between the bodies in contact and the corresponding deformation are symmetrical with respect to the center O of the surface of contact. Taking Fig. 206a to represent the surlace of contact, and M as a point on the surface of contact of the lower ball, the displacement Wi of this point, from the previous article, is Wi = (1 ;E;i2) f f
q ds dift
+ W2
= (ki
+ k2)Í f q ds dift
This equation will be fulfilled for any value of r, and hence the assumed pressure distribution is the correct one if the following relations exist for the displacement a and the radius a of the surlace of contact: 1r2a
(f)
in which vi and Ei are the elastic constan,ts for the lower ball, and the integration is extended over the entire area of contact. An analogous formula is obtained also for the upper ball. Then Wi
or
a= (ki
+ k2)qo2
a= (ki
+ k2)
(g)
qo. ~'Iraª = P
a 3
(216)
from which qo
From Eqs. (e) and (g), (ki
+ k2)ff q ds d1/t
= a - {3r 2
(h)
Thus we must find an expression for q to satisfy Eq. (h). It will now be shown that this requirement is satisfied by assuming that the distribution of pressures q over the contact surlace is represented by the ordinates of a hemisphere of radius a constructed on the surface of contact. If q0 is the pressure at the center O of the surlace of contact, then qo = ka
1
1!
in which k = q0 / a is a constant factor representing the scale of our representation of the pressure distribution. Along a chord mn the pressure q varies, as indicated in Fig. 206 by the dotted semicircle. Perlorming the integration along this chord we find
fqds=~ºA -
r 2 sin 2 ift).
=
3P 27ra2
(218)
i.e., the maximum pressure is li times the average pressure on the surface of contact. Substituting in Eqs. (217) and taking, from Eq. (b),
f3 =Ri+ R2 2RiR2
we find for two balls in contact (Fig. 207) 3
a=
31r P(k1
4
+ k2)RiR2
Ri+ R2
(219)
Assuming that both balls have the sarne elastic properties and taking 11 = 0.3, this becomes 3 P RiR2
ª
in which A is the area of the semicircle indicated by the dotted line and is equal to; (a 2
~
The value of the maximum pressure qo is obtained by equating the sum of the pressures over the contact area to the compressive force P. Then, for the hemispherical pressure distribution this gives
in which
'
(217)
1r2qo
ª
Substituting in Eq. (h), we find that
=
l.IOY
= ·1 2·3·
.
E Ri + R2
3 p2
E2
Ri
+ R2
(220)
RiR2
The corresponding maximum pressure is qo =
~ .!:.__2 2 1ra
, 1ili:: 1
1
= 0.388 3 PE2
R1 + R2)2 Ri 2R2 2
(221)
376
377
THEORY OF ELASTICITY
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
ln the case of a ball pressed into a plane surface, and assuming the sarne elastic properties of material for both bodies, we find, by substituting R1 = oo in Eqs. (220) and (221),
The other principal stress, acting in the circumferential direction, is numerically equal to the above radial stress but of opposite sign. Hence along the boundary of the surface of contact, where normal pressure on the surface becomes equal to zero, we have pure shear of the amount qo(l - 2v)/3. Taking v = 0.3, this shear becomes equal to 0.133qo. This stress is much smaller than the maximum shearing stress calculated above, but it is larger than the shearing stress at the center of the surface of contact, where the normal pressure is the largest. Many experiments have been made which verify the theoretical results of Hertz for materiais which follow Hooke's law and stress within the elastic limit.1 126. Pressure between Two Bodies in Contact. More General Case. 2 The general case of compression of elastic bodies in contact may be treated in the sarne manner as the case of spherical bodies discussed in the previous article. Consider the tangent plane at the point of contact O as the xy-plane (Fig. 207). The surfaces of the bodies near the point of contact, by neglecting small quantities of higher order, can be represented by the equations 3 2 A2xy Z1 = A1x Aay 2 (a) 2 Z2 = B1x + B2xy + Bay 2
3~
a= 1.109 '\}E-'
a
ª[P2
= 1.23 '\} EJ2Ri--, 2
ª{PE2
qo = 0.388 '\}R;!
(222)
By taking R 1 negative we can write down also equations for a ball in a spherical seat (Fig. 208b). Having the magnitude of the surface of contact and the pressures acting on it, the stresses can be calculated by using the method developed in the previous article. 1 The resulta of these calculations for points along the axes Oz 1 and Oz2 are shown in Fig. 209. The maximum pressure qo at the center of the º-.-i-..-,,..-,..:..::;.~~---,-~:_q surface of contact is taken as a unit of stress. ln measuring the distances along the z-axis, the radius a of the surface of contact is taken as the unit. The greatest stress is the compressive stress cr, at the center of the surface of contact, but the two other principal stresses crr and crs, at the sarne point, are equal to 1+2v - -cr,. Hence the maximum 2 shearing stress, on which the yielding z of such material as steel depends, is Fm. 209. comparatively small at this point. The point with maximum shearing stress is on the z-axis at a depth equal to about a half of the radius of the surface of contact. This point must be considered as the weakést point in such material as steel. The maximum shearing stress at this point (for v = 0.3) is about 0.3lq0 • ln the case of brittle materiais, such as glass, failure is produced by maximum tensile stress. This stress occurs at the circular boundary of the surface of contact. It acts in a radial direction and lias the magnitude · (1 - 2v) cr.. = qo 3 1 Such calculations were made by A. N. Dinnik, Bull. Polyteck. Inst., . KieW, 1909. See also M. T. Huber, Ann. Physik, vol. 14, 1904, p. 153; S. Fuchs, Physik. Z., vol. 14, p. 1282, 1913; M. C. Huber and S. Fuchs, Physik. Z., vol. 15, p. 298, 1914; W. B. Morton and L. J. Close, Phil. M ag., vol. 43, p. 320, 1922.
+
+
The distance between two points such as M and N is then z1 + z2 = (A1 + B1)x 2 + (A2 + B2)xy + (As + B 3)y2
(b)
· We can always take for x and y such directions as to make the term containing the product xy disappear. Then (e)
in which A and B are constants depending on the magnitudes of the principal curvatures of the surfaces in contact and on the angle between the planes of principal curvatures of the two surfaces. If R1 and Ri' denote the principal radii of curvature at the point of con1
References to the corresponding literature can be found in the paper by See also "Handbuch der physikahschen und technischen Mechanik,'' vol. 3, p. 120. 2 This theory is due to Hertz, loc. cit. Tangential force and twisting couple at the contact are considered by R. D. Mindlin, J. Applied Mechanics (Trans. A.S.M.E.), vol. 16, p. 259, 1949. 3 It is assumed that point O is not a point of singularity on the surfaces of the bodies, but the surface adjacent to the point of contact is rounded and may be considered as a surface of the second degree.
?· Berndt, Z. tech. Physik, vol. 3, p. 14, 1922.
378
THEORY OF ELASTICITY
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
tact of one of the bodies, and R2 and R2' those of the other, i and if; the angle between the normal planes containing the curvatures 1/R 1 and 1/R2, then the constants A and B are determined from the equations
The problem now is to finda distribution of pressures q to satisfy Eq. (g). H. Hertz showed that this requirement is satisfied by assuming that the intensity of pressures q over the surface of contact is represented by the ordinates of a semi-ellipsoid constructed on the surface of contact. The maximum pressure is then clearly at the center of the surface of contact. Denoting it by qo and denoting by a and b the semiaxes of the elliptic boundary of the surface of contact the magnitude of the maximum pressure is obtained from the equation
A
+B
1(1
= 2 Ri
1) + Ri'1 + R21 + R2' 2
2
1 [( 1 1) B - A = 2 Ri - Ri'
+
(
1
1)
R2 - R2'
+ 2 (~i
(d)
_ ;i,)
(~2 _
; 2,) cos 21/;
r
It can be shown that A and B in Eq. (e) both have the sarne sign, and it can therefore be concluded that all points with the sarne mutual distance Zi + Z2 lie on one ellipse. Hence, if we press the bodies together in the direction of the normal to the tangent plane at O, the surface of contact will have an elliptical boundary. Let a, wi, w2 have the sarne meaning as in the previous article. Then, for points on the surface of contact, we have
P = from which
+ W2
=
a - Ax 2
-
fJ
'~
where q dA is the pressure acting on an infinitely small element of the surface of contact, and r is the distance of this element from the point under consideration. The integration must be extended over the entire surface of contact. Using notations (216), we obtain, from (e) and (f), qdA (g) (ki k2) -r- =a - Ax 2 - By2
+
!!
1 The curvature of a body is considered as positive if the corresponding center of curvature is within the body. ln Fig. 207 the curvatures of the bodies are positive. ln Fig. 208b the spherical seat has a negative curvature.
37r P(ki
+ k2)
/37r P(ki
+ k2)
3
T (A+ B)
m 3
b = n
By 2
This is obtained from geometrical considerations. Consider now the local deformation at the surface of contact. Assuming that this surface is very small and applying Eq. (205), obtained for semi-infinite bodies, the sum of the displacements Wi and W2 for points of the surface of contact is 2 2 Wi + W 2 = (1 - Vi + 1 - V2 ) q dA (f) 7rE1 7rE2 r
(223)
We see that the maximum pressure is 1-f times the average pressure on the surface of contact. To calculate this pressure we must know the magnitudes of the semiaxes a and b. From an analysis analogous to that used for spherical bodies we find that
(e)
Wi
~7rabqo
3 p qo = - 2 7rab
ª= or
ff q dA =
379
'\[4 (A+ B)
(224) .
in which A + Bis determined from Eqs. (d) and the coefficients m and n are numbers depending on the ratio (B - A): (A B). Using the notation
+
B-A cos8=A+B
(h)
the values of m and n for various values of () are given in the table below.i 8 =
30°
35º
40°
45°
50°
55°
60º
65º
70º
75º
80º
85º
90º
- - -- -- -- -- -- -- -- -- -- -- -- -- -m= 2.731 2.397 2.136 1.926 1.754 1.611 1.486 1.378 1.284 1.202 1.128 1.061 1.000 n= 0.493 0.530 0.567 0.604 0.641 0.678 0.717 0.759 0.802 0.846 0.893 0.944 1.000
Considering, for instance, the contact of a wheel with a cylindrical rim of radius Ri = 15.8 in. and of a rail with the radius of the head R2 = 12 in., we find, by substituting Ri' = Rl = oo and if; = 7r/2 into Eqs. (d), A
+B 1
= 0.0733,
B - A
= 0.0099,
cos 8
= 0.135,
8
= 82°15'
The table is taken from the paper by H. L. Whittemore and S. N. Petrenko, U. S. Bur. Standards, Tech. Paper 201, 1921.
AXIALLY SYMME1'RICAL STRESS DISTRIBUTION
THEORY OF ELASTICITY
380
exists pure shear. The magnitude of this shear for the ends of the major axis (x = ±a, y = O) is
Then, by interpolation, we find from the above table that m = 1.098,
n
= 0.918
Substituting in Eqs. (224) and taking E = 30.10 6 p.s.i. and v = 0.25, 1 we find . b = 0.00792 ~ a = 0.00946 ~'
T
T
b = 0.0792 in.,
area of contact 7rab = 0.0236 sq. in.
and the maximum pressure at the center is q0 = ;
:;,b = 63,600 p.s.i.
Knowing the distribution of pressure, the stresses at any point can be calculated. 2 It was shown in this manner that the point of maximum shearing stress is on the z-axis at a certain small depth z1, depending on the magnitude of the semiaxes a and b. For instance: Z1 = 0.47a, when b/a = l; and z1 = 0.24a, when b/a = 0.34. The corresponding values of maximum shearing stress (for v = 0.3) are Tmax. = 0.31qo and Tmax. = 0.32q 0 respectively. Considering points on the elliptical surface of contact and taking the x- and y-axes in the direction of the semiaxes a and b respectively, the principal stresses at the center of the surface of contact are u,. = -2vqo - (1 - 2v)qo a
Uz
+b b
a -2vqo - (1 - 2v)qo - a+ b = -qo
(k)
For the ends of the axes of the ellipse we find u,. = -uy and Tzy = O. The tensile stress in the radial direction is equal to the compressive stress in the circumferential direction. Thus at these points there 1 H 11 is increased from 0.25 to 0.30 the semiaxes (224) decrease about 1 per cent and the maximum pressure qo increases about 2 per cent. s Such investigations have been made by Prof. N. M. Belajef, see Bull. Inst. Engineers of Ways of Communication, St. Petersburg, 1917, and "Memoirs on Theory of Structures," St. Petersburg, 1924; see also H. R. Thomas and V. A. Hoersch, Univ. Illinois Eng. Expt. Sta., Bull. 212, 1930, and G. Lundberg and F. K. G. Odqvist, Proc. Ingeniõrs Vetenskaps Akad., No. 116, Stockholm, 1932.
= (1 - 2v)qo (!_2 e
and for the ends of minor axis (x
For a load P = 1,000 lb., a = 0.0946 in.,
381
= (1 - 2v)qo -f32
e
(!e arctanh e - 1) = O, y = ± b)
(l)
is
e)
( 1 - f3- arctan e {3
(m)
where {3 = b/a, e = (1/a) v' a 2 - b2 • When b approaches a and the boundary of the surface of contact approaches a circular shape, the stresses given by (k), (l), and (m) approach the stresses given in the previous article for the case of compression of balis. A more detailed investigation of stresses for ali points in the surface of contact shows1 that for e < 0.89 the maximum shearing stress is given by Eq. (l). For e > 0.89 the . ):') ' 0.Silo . . stress is . ath o max1mum sh earmg t e center Jl"''r'F"!':::::c:::l""ici-r-e:p.iiF."1 of the ellipse and can be calculated from Eqs. (k) above. By increasing the ratio a/b we obtain narrower and narrower ellipses of contact, and at the limit a/b = CX) we arrive at the case of contact of two cylinders with parallel axes. 2 The surface of contact is now a narrow rectangle. The distribution of pressure q along the width of the surface of contact (Fig. 210) is represented by a semi-ellipse. If the x-axis is perpendicular to the plane of z the figure, b is half the width of the FIG. 210. surface of contact, and P' the load per unit length of the surface of contact, we obtain, from the semi-elliptic pressure distribution, P' = *'1-bqo from which 2P' (225) qo =~ 7rb 1
See Belajef, loc. cit. A direct derivation of this case, with consideration of tangential force at the contact, is given by H. Poritsky, J. Applied Mechanics (Trans. A.S.M.E.), vol.17, p. 191, 1950. 1
AXIALLY SYMMETRICAL STRESS DISTRIBUTION THEORY OF ELASTICITY
382
The investigation of local deformation gives for the quantity b the expression 4P' (ki + k2)RiR2 (226) b = Ri+ R2 in which Ri and R 2 are the radii of the cylinders and ki and k2 are constants defined by Eqs. (216). If both cylinders are of the sarne material and v = 0.3, then
b = 1.52
=
i.osR {P'R
P'(Ri
=
7í 2
(ki
+ R2)
+ k2)RiR2
(a)
= -P,
(228)
and we find, from Eqs. (a), that ã =
_pm'
+
m2
m1m2
(229)
Substituting b from Eq. (226) into Eq. (225), we find
qo
dv,
m,dt
in which m 1 and m 2 denote the masses of Frn. 211. the spheres. Let a be the distance the two spheres approach one another due to local compression at O. Then the velocity of this approach is
For the case of contact of a cylinder with a plane surface, b = 1.52 '\)E-
127. Impact of Spheres. The results of the last two articles can be used .in investigating impact of elastic bodies. Consider, as an example, the impact of two spheres (Fig. 211). As soon as the spherei;, in their motion toward one another, come in contact ata point 0, 1 the compressive forces P begin to act and to change the velocities of the spheres. If v, and v2 are the values of these velocities, their rates of change during impact are given by the equations
(227)
In the case of two equal radii, Ri = R2 = R,
b
383
Investigations show that the duration of impact, i.e., the time during which the spheres remain in contact, is very long in comparison with the period of lowest mode of vibration of the spheres. 2 Vibrations can therefore be neglected, and it can be assumed that Eq. (219), which was established for statical conditions, holds during impact. Using the notations A
(230)
(b)
n = we find, from (219),
/16
R1R2
\'g,,. 2 (k1 + k.) 2 (R1 + R.)' P = nal
(e) (d)
and Eq. (b) becomes (e)
If the materials of both cylinders are the sarne and v "" 0.3, (231)
q0 = 0.418 I~
Multiplying both sides of this equation by à, j-d(à) 2
= -nn 1ai da
from which, by integration, (/)
ln the case of contact of a cylinder with a plane surface,
[FE
qo = 0.418 '\)R-
(232)
Knowing qo and b, the stress at any point can be calc~lated. ~hese calculations showi that the point w:ith maximum sheanng _stress lS .ºn the z-axis at a certain depth. The variation of stress co~ponents w~th the depth, for v = 0.3, is shown in Fig. 210. !he u:iax1mum shearmg stress is at the depth zi = 0.78b and its magmtude lS 0.304qo. 1
See Belajef, loc. cit.
where v is the velocity of approach of the two spheres at the beginning of impact. If we substitute à = O in this equation, we find the value of the approach at the instant of maximum compression, a 1, as
ª'
=
(~4nn, ~)i
(g)
With this value we can calculate, from Eqs. (219), the value of the maximum compressive force P acting between the spheres during impact, and the corresponding radius a of the surface of contact. 1 We assume motions along the line joining the centers of the spheres. 2 Lord Rayleigh, Phil. Mag., series 6, vol. 11, p. 283, 1906.
384
THEORY OF ELASTICITY
385
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
For calculating the duration of impact we write Eq. (/) in the following forro:
(à) This solution can be taken in the forro t/> = f(r) sin kz
or writing a/a 1 = z and using Eq. (g), we find that
dt=~ V
in which f is a function of r only. Substituting (b) into Eq. (a), we arrive at the following ordinary differential equation for determining f (r):
dz
yl -
(z)t
d2f dr 2
from which the duration of impact is
t = 2a1
r
v }o
1
dz
yl -
(b)
(z)i
= 2.94 ~ v
(233)
ln the particular case of two equal spheres of the sarne material and radius R, we have, from (g),
+ !.r ~ dr.
k'f =
o
(e)
We take an integral of this equation in the forro of a series, (d)
Substituting this series in Eq. (e) we find the following relation between the consecutive coefficients:
(234)
from which where p denotes the mass per unit volume of the spheres. We see that the duration of impact is proportional to the radius of the spheres and inversely proportional to (v)A. This result was verified by severa! experimenters.1 ln the case of long bars with spherical ends, the period of the fundamental mode of vibration may be of the sarne arder as the duration of impact, and in investigating local compression at the point of contact these vibrations should be considered. 2 128. Symmetrical Deformation of a Circular Cylinder. For a circular cylinder submitted to the action of forces applied to the lateral surface and distributed symmetrically with respect to the axis of the cylinder, we introduce a stress function q, in cylindrical coordinates and apply Eq. (180). 3 This equation is satisfied if we take for the stress function q, a solution of the equation 1 M. Hamburger, Wied. Ann., vol. 28, p. 653, 1886; A. Dinnik, J. Russ. Phys.Chem. Soe., vol. 38, p. 24,2, 1906, and vol. 41, p. 57, 1909. Further references to the literature of the subject are given in "Handbuch der physikalischen und technischen Mechanik," vol. 3, p. 44~, 1927. 2 See p. 452. Longitudinal impact of rods with spherical surfaces at the ends has been discussed by J. E. Sears, Proe. Cambridge Phil. Soe., vol. 14, p. 257, 1908, and Trans. Cambridge Phil. Soe., vol. 21, p. 49, 1912. Lateral impact of rods with consideration of local compression was discussed by S. Timoshenko, Z. Math. Physik, vol. 62, p. 198, 1914. s The problem of the deformation of a circular cylinder under the action of forces applied to the surface was discussed first by L. Pochhammer, Crelle's J., vol. 81, 1876. Several problems of symmetrical deformation of cylinders were discussed by C. Chree, Trans. Cambridge Phil. Soe., vol. 14, p. 250, 1889. See also the paper by L. N. G Filon, Trans. Roy. Soe. (London), series A, vol. 198, 1902, which contains solutions of several problems of practical interest relating to symmetrical deformation in a cylinder.
Substituting these in the series (d), we have
f(r) = ao ( 1
k 2r 2
k'r'
k 6r 6
+ 22 + 22 . 42 + 22 . 42 . 6' +
(e)
The second integral of Eq. (e) can also be obtained in the forro of a series, and it can be shown that this second integral becomes infinite when r = O, and hence should not be considered when we are discussing deformation of a solid cylinder. The series in the parentheses of Eq. (e) is the Bessel function of zero order and of the imaginary argument ikr. 1 ln the following we shall use for this function the notation J 0(ikr) and write the stress function (b) in the forro
(!)
t/>1 = aoJ o(ikr) sin kz
Equation (180) also has solutions different from solutions ~f Eq. (a). One of these solutions ·can be derived from the above function J 0 (ikr). By differentiatfon,
dJo(ikr) ikr ( d(ikr) = - 2 1
k 2r 2
k'r'
k6r6
)
+ 2 · 4 + 2 · 42 · 6 + 2 · 42 · 62. 8 + · · ·
(g)
This derivative with negative sign is called Bessel's function of the first order and is denoted by J,(ikr). Consider now the function
f 1(r)
d
.
.
.
= r dr Jo(ikr} = -ikrJ,(ikr) =
ktr 2
2
(
2
1
k2r k'r' +n + 2 .42. 6 +· ..
~
)
(h)
1 Discussion oÍ the · differential equation (e) and of Bessel's functions -~an bl f?und,,in.the following books: A. R. Forsyth, "A Treatise on Differential Equa~ tions, and A. Gray and G. B. Mathews, "A Treatise on Bessel Functions:" ~umerical tables for Bessel's functions can be found in E. Jahnke and F. Erode, Funktionentafeln mit Formeln und Kurv~n," Berlin, 1909.
386
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
THEORY OF ELASTICITY
and the shearing forces by the series
By differentiation it can be shown that
- + -r1 -drd -
d2 ( dr 2
2,,.z B . 31rz+ . ,,.z+B. B tSlllT 2Slll-z-+ 3Slll-z-
k 2 ) f1(r) = 2k2Jo(ikr) · -
Then, remembering that J 0 (ikr) is a solution of Eq. (e), it follows that f1(r) is a solution of the equation
~2 + !.! (dr r dr
k•)
(d'f1 dr•
+ !r dfi dr
> = f(r)
- k•fi)
and proceed as before, we find, instead of expression (;j), the stress function
(i)
Combining solutions (f) ,and (i), we can take the stress function in the form sin kz[aoJo(ikr)
+ a1(ikr)J1(ikr)J
(j)
Substituting this stress function in Eqs. (179) we find the following expressions for the stress components:
+ +
in which F 1(r), . . • , F.(r) are certain functions of r containing J o(ikr) and J 1(ikr). By using tables of Bessel functions, the values of F1(r), • . . , F,(r) can easily be calculated for any value of r. Denoting by a the externa! radius of the cylinder, the forces applied to the surface of the cylinder, from Eqs. (k), are given by the following values pf the stress components:
+
By a suitable adjustment of the constants k, a 0 , a1, various cases of symmetrical loading of a cylinder can be discussed. Denoting the length of the cylinder by l and taking k = n,,. l
aoF1(a) aoFa(a)
+ a1F2(a)
+ a1F.(a)
-A,. =O
=
we obtain the values of the constants ao and a1 for the case when normal pressures A,. cos (n,,.z/l) act z on the lateral surface of the cylinder. The éase FIG. 212. when n = 1 is represented in Fig. 212. ln an analogous manner we can get a solution for the case when tangential forces of intensity B,. sin (n,,.z/l) act on the surface of the cylinder. By taking n = 1, 2, 3, •.. , and using the superposition principle, we arrive at solutions of problems in which the normal pressures on the surface of the cylinder can be represented by the series
,,.z
2,,.z A 1 cos T +A, cos - 1-
cos kz
=o
>2 = a1 sin kz(ikr)J1(ikr)
=
(n)
If we take for the stress function >, instead of expression (b), the expression
Hence a solution of Eq. (180) can be taken in the form
q,
387
+ Aa cos 31rz -l-
+
(m)
= cos kz[boJo(ikr)
+ b1(ikr)J1(ikr)]
(o)
By a suitable adjustment of the constants k, bo, b1, we obtain the solution for the case in which normal pressures on the cylinder are represented by a sine series and the shearing forces by cosine series. Hence, by combining solutions (j) and (o), we can get any axially symmetrical distribution of normal and shearing forces over the surface of the cylinder. At the sarne time there will also be certain forces distributed over the ends of the cylinder. By superposing a simple tension or compression we can always arrange that the resultant of these forces is zero, and their effect on stresses at some distance from the ends becomes negligible by virtue of Saint-Venant's principle. Several examples of symmetrical loading of cylinders are discussed by L. N. G. Filon in the paper already mentioned.1 We give here final results from his solution for the case shown in Fig. 213. A cylinder, the length of which is equal to ,,.a, is submitted to the tensile action of shearing forces uniformly distributed over the shaded portion of the surface of the cylinder indicated in the figure. The distribution of the normal stress ,,., over cross sections of the cylinder a is of practical interest, and the table below gives the z ratios of these stresses to the average tensile stress, FIG. 213. obtained by dividing the total tensile force by the cross-sectional area of the cylinder. It can be seen that local tensile stresses near the loaded portions of the surface diminish rapidly with increase of distance from these portions and approach the average value.
z
r =O
0.05Z O.lOZ 0.15Z 0.20Z
0.689 0.673 0.631 0.582 -0.539
o
1 Loc. cit. p. 176, 1944.
r
= 0.2a r = 0.4a r = 0.6a 0.719 0.700 0.652 0.594 0.545
0.810 0.786 0.720 0.637 0.565
0.962 0.937 0.859 0.737 0.617
r=a
1.117 1.163 1.344 2.022 1.368
See also G. Pickett, J. Applied Mechanics (Trans. A.S.M.E.), vol. 11,
THEORY OF ELASTICITY
388
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
Another application-of the general solution of the problem in terms of Bessel's functions is given by A. N ádai in discussing the bending of circular plates by a force concentrated at the middle 1 (Fig. 214). 129. The Circular Cylinder with a Band of Pressure. 2 When a short collar is shrunk on a much longer shaft the simple shrink-fit formulas, valid when collar and shaft are of equal lengths, are not accurate. A much better approximation is obtained by considering the problem, FIG. 214. indicated in Fig. 215a, of a long cylinder with a uniform 3 normal pressure p acting on the band ABCD of the surface. The required solution can evidently be obtained by superp?sing the e~ects of the two pressure distributions indicated in Fig. 215b. The bas1c problem 18 therefore that of pressure p/2 on the lower half of the cylindrical surface and -p/2 on
389
From Eqs. (179) we find that the shear stress will be Trz
=
lo
00
[pklo'(kr) - k 2rli'(kr) - kl 1 (kr)
- 2k(l - v)Io'(kr)]k2f(k) cos kz dk
(e)
where primes denote differentiation with respect to kr. This must vanish at the surface r = a. Putting r = a in the expression in square brackets, and equating this bracket to zero, we obtain an equation for p which gives p = 2(1 - 11)
+ ka Io(ka) J,(ka)
(d)
The remaining boundary condition is
=E
= -
for
r =a, z >O
for
r =a, z
(e)
E
The value of
= -
lo
00 [
(1 - 211 - p)lo(kr)
+ (kr + fr) l
1
(kr)
Jk /(k) sin kz dk 3
(f)
We now make use of the fact that'
,,.
(b}
(aJ
2 for z >O
FIG. 215.
1
l!.
O for z =O
the upper half, the length of the cylinder being infinite, and its solution will now be given. . . We begin with the stress function given by Eq. (o) of Art. 128, wntmg I o(kr) for J 0 (ikr) and i/1(kr) for J1(ikr). We also write bo = pbi. Then cf> = [pl 0 (kr) -
krl 1(kr)]b1 cos kz
bi = f(k) dk
Putting this in (a) and adding up all such stress functions we obtain a more general stress function in the form cf> =
lo
[pl 0 (kr) - krl 1 (kr)]f(k) cos kz dk
'!!.
f
,,. lo
00
~
sinkzdk= k (
Efor z >o O for z =O
(h)
- '!!.for z
(b)
A. Nádai, "Elastische Platten," p. 315, 1925. M. V. Barton, J. Applied Mechanics (Trans. A.S.M.E.), vol. 8, p. A-97, 1941. A. W. Rankin, ibid., vol. 11, p. A-77, 1944. a The pressure in the shrink fit is not uniform in the axial direction. 2
p/,,., we obtain
- [ (1 - 211 - p)lo (ka)
00
We shall now see how it is possible to select the function f(k) so that this stress function will give the solution to our problem. i
If we multiply this by
(a)
This satisfies Eq. (180) no matter what value is given to k. lf we consider k to take a range of values we can allow b1 to depend on k and an increment dk by writing
(g)
- ~for z
+ (ka +:a) l
1 (ka)
Jk•j(k) =; ·~
(i)
and this equation determines/(k). The stress components are then found from the stress function (b) by means of the formulas (179), and will be integrais of the sarne ~eneral nature as that of Eq. (f), which gives rrr. Values, obtained by numerical i~tegration, are given by Rankin in the paper cited on page 388. The curves in Fig. 216 show the variation of the stresses in the axial direction for various radial distances, and also the surface displacements. 1
See for instance 1. S. Sokolnikoff, "Advanced Calculus," lst ed., p. 362.
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
THEORY OF ELASTICITY
390
They are reproduced from the paper of Barton (see page 388) and were obtained by a different method using Fourier series. From these curves results can be obtained for the problem of Fig. 215 by superposition, as explained at the beginning of this article. Curves for the stresses and displacement for pressure bands of several widths are given in the papers cited. When the width is equal to the radius of the cylinder the tangentiaJ stress q9 at the surface and at the middle of the pressure band reaches a value about 10 per cent higher than the applied pressure, and is, of course, compressive. The axial stress(/• in the surface just outside the pres1.0\ (ª9,,0.f,.,0.6
IT9
p
-0.5 f-1.IG~~--r-, -0.4 \~ -0.31-#1/,I'-+'-+---+--<
-0.2 flll.--+---1-----1--__,
-2a -~ -a -ª-0.1 --+--+---+--l, a 2
391
130. Twist of a Circular Ring Sector. This problem is of practical interest in connection with the calculation of stresses in close-coiled helical springs. Consider a ring sector under the action of two equal and opposite forces P along the axis through the center of the ring and perpendicular to the plane of the ring (Fig. 217). These forces produce the sarne torque M 1 = PR in all cross sections of the ring. lf the cross-sectional dimensions of the ring are small in comparison with the radius R, formulas derived for the torsion of prismatical bars can be used with sufficient accuracy in calculating the stresses. ln the case of heavy helical springs the cross-sectional dimensions are no longer small, and the difference in length of outer and inner circumferential fibers must be considered. ln this manner it can be shown that at inner points, such as i, the r O shearing stress .is considerably larger than that given by the theory of torsion of straight bars.1 For a more rigorous solution of the problem we apply the z general equations of the theory of elasticity in dº 2 [E FIG. 217. . d . 1 cy1m rica coor mates qs. (170), page 306]. Assumin~ that in this case of torsion only the shearing-stress components 78, and rre are d1fferent from zero (Fig. 218), we find, from Eqs. (170), (a)
r/a
-0.318 --------------0.3
Consider now the compatibility equations (130).
--=a9 ----=a6
-0.2
1"yz
-----=0.4
tj-z
p-0.I
From Fig. 219 we find
- - - =0.8
--:/j.j.\!\---------- =0.2 a ~ 2a -~ -a
O t-"""'~~~'-t4l>::::=;;;l;R_.l.__-+z
Txy
= T9z COS 8 = Tr9(COS 2 O -
sin 2 O) =
Tr9
cos 28
-2a -~
0.1 fw.!z-.a/uo
u 0 =-pa(J-v)/2E
,.crtf8t]J'J"'z -!f f
-Za -~ -a
O
~
a
-za -J,f -a
Za
FIG. 216.
sure band reaches a tensile value of about 45 per cent of the applied pressure. The shear stress Trz attains a greatest value, equal to 31.8 per cent of the applied pressure, at the edges of the pressure band AB and CD in Fig. 215 and just below the surface. When the pressure is applied all over the curved surface of the cylinder, of any length, we have simply compressive (fr and q9 equal to the applied pressure, and (/• and Trz zero. Solutions have been obtained in a similar manner for a band of pressure in a hole in an infinite solid, 1 and for a band of pressure near one end of a solid cylinder. 2 1 G. J. Tranter, Quart. Applied Math., vol. 4, p. 298, 1946; O. L. Bowie, ibid., vol. 5, p. 100, 1947. 2 C. J. Tranter aod J. W. Craggs, Phil. Mag., vol. 38, p. 214, 1947.
y FIG. 218.
FIG. 219.
Substituting in the fourth and sixth of Eqs. (130) and remembering that
e=
(fr
+ + =o (JQ
(/z
1 An elementary theory of twist of a ring sector was given by V Roever V D [ vol · 57'. 1913 · ' · · " . · See also M. P1lgram, Artill. Monatshefte, 1913. ·An experimental determmat1on of maximum stress by measuring strain at the surface of the coil was made by A. M. Wahl, Trans. A.S.M.E., 1928. 2 Th' . d ue to O. Gõhner, Ingenieur-Archiv, vol. 1, p. 619, 1930· vol. 2 isso1u t'10n is PP- 1 and 381, 1931; vol. 9, p. 355, 1938. ' '
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
THEORY OF ELASTICITY
392 we find
(b)
The remaining four of the compatibility equations [see Eqs. (g), page 346] are satisfied by virtue of our assumption that u, =
?·
GR 2acJ>
az'
7
'TrB =
GR 2 ôc/> j:2 ar
= -
'T8z
(e)
where G is the modulus of rigidity and R the radius of the ring. in Eqs. (b), we find
a (ª2"' a2cJ> ar ar 2 + az 2 a (ª2"' a2cJ> ôz ar 2 + az 2
~ ª"')
r ar ~ ª"') r ar
=
o
=
o
Substituting (e)
ªar2 + ªaz2 2
2
>
"' -
ª"'
~r ar + 2c
Then, as the number of terms in the series (g) increases, the sum of Eqs. (h) approaches more and more closely Eq. (e), and the series (g) approaches the exact solution for the stress function cJ>. Consider now the boundary conditions. The resultant shearing stress at the boundary (Fig. 218) must be in the direction of the tangent to the boundary, hence 'TrB
cos (NJ;) -
"B•
cos (Nr) = o
or, by using Eqs. (e),
This shows that cJ> must be constant at the boundary, and we satisfy this condition hy taking solutions of Eqs. (h) such that c/>o, c/>1, e/>., ••• are zero at the boundary. Having obtained c/>o, cj> 1, ••• the successive approximations for the stress components are now obtained from Eqs. (e). Introducing the new variables J; and r, these equations can be represented in the following form: (i)
f m which we conclude that the expression in the parentheses must be a constant. ~:noting this constant by -2c, the equation for determining the stress function e/> is
U sing now the expansion
(d)
= O
and the series (g), we find as the first approximation
We introduce now, instead of coordinates r and z (Fig. 218), new coordinates J; and
r.
J; = R
-r,
a2cJ> a1;2
+
a2c/> ar2
(Tez)o = G
r=z
and Eq. (d) becomes
i;) ª"' +
3
+ R ( 1 -n,
ai;
(e)
1
J;
--J; = 1
1
1-R
1;2
+ R + R2
.
we shall now solve Eq. (e) by successive approximations. e/> = c/>o
and determine c/>o, c/>1, c/>2,
(1 + 2R1;) ª"'º + ª"' ar ar 2 G [ (1 + 1;) ª"'º + ª"' R
1
(,,.,e)i = G [ (,,.e.)i
Considering J;/R as a small quantity, and using the expansion
!:
~~o
For the second approximation we find from Eqs. (i) 2c = O
I' 1
393
(f)
Assume
+ c/>1 + c/>2 + · · ·
(g)
. in such a manner as to satisfy the equations
=
(k)
1
ai;
]
ai;
]
For the third approximation, (,,.,e)2 = G [ ( 1
+ ~ +~:)ªa"'/ + ( 1 + ~) ~t1 + ªa~2 ]
(,,.e.). = G [ ( 1
+ ~ + ~:) ª~º + ( 1 + ~) ªa~' + ªa~2 J
(l)
We apply this general discussion to the particular case of a ring of circular cross section of radius a. The equation of the boundary (Fig. 218) is
(h)
1; 2
+r
2
-
a•=
o
(m)
and the solution of the first of Eqs. (h), satisfying the boundary condition, is
c/>o = 1.
''
'li
~ '
e
2 (!;2 +
r2 - a•)
l~Sil Tv Púl\T~ f----··---·---= ,., ' T l\JI..
'
1~-11 ~O
~i
\
R /4..
'
~i~~\OTECI\ Cf.N'f'RÀ~_.
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
THEORY OF ELASTICITY
394
horizontal diameter of the cross section of the ring (Fig. 21S) from the second of the Eqs. (q), we find
The first approximation for the stress components, from Eqs. (j), is he)o
= -cGr,
(n)
(re,)o = -cG~
This is the sarne stress distribution as for a circular shaft. value of the torque is M1 = - Jf Crrer + -r.e~) d~ dr
(-rezh = -cG (t
The corresponding (o)
For the inner point i, Ç
Substituting from Eqs. (n),
+ ~ f. + 8R
e= 2(M,)o G7ra 4
T
' re
=
r
O, and
2 + a2ç) SR 4R2
13 _E+ 3a
16 R 2
= a, and we have
(-re,); = -cGa (1
dha 4 (M1)0 = - -, 2
r= o
395
+ 4R ~ !: + 17 ~) 16R 2
For the outer point O, t = -a, and
To get the second approximation we use the second equation of (h). tuting for >o the expression above we find
Substi-
(-ro,)o = cGa (1 -
~!: + 17 ª) l6R2 2
4R
Using Eq. (r), the values of these stresses become The solution of this equation, satisfying the condition that >1 vanishes at the boundary, is >1 = 3- e~ - (~2 + r2 - a2) SR Substituting this in Eqs. (k) we find the second approximation for the stress components
-cG
(r +Hf)
-cG
1e a 2 [ + -SR- - -SR (r t
a2)
J
. (TOz, )
! j: l
+
a2<1>2 ar2
+
~ .!:__ <,2 S R2 '
+
o
e
+ 5r2
- 15a2)(e
+ r2
17 a 2 4R + l6R2 1 3 a2 + l6R2 = ::: (
1 -
~ ~ + ~ ~:)
= _
1 1a 1 ( 2PR .1 - (a/R) 4 R i6 'Ira• _ ~ (a/R)2 1 16 1 - (a/R)2
+
+
ª)2
R
(235)
1:he distribution of shearing stresses along the horizontal d~ameter for a particular case, a/R = t, is shown in Fig. . 220. For comparison the first approximation obtained by applying the formula for a circular shaft is shown by a dotted line.2 T~e method described has also been applied to the tors1on problem for ring sectors of elliptic and rectangular FIG. 220. ~ross s~ctio~s. a • For a square cross section with sides of length 2a, the third pproximatwn g1ves for the stress at the inner point
The solution of this equation satisfying the boundary condition is
>2 = - 64R2 (t2
+ ~ _i:
(p)
(-re,);
ar2 - a 2) = a
'Iraª
1
---~-.::~:.::...
The calculatio.n of further approximations shows that the final expression for the greatest shearmg stress can be put in the form 1
1
a2<1>2 a~2
-2M1
--
(ro,)o
Substituting >o and >1 in the third of Eqs. (h), we find
1
-
- a2)
By using Eqs. (l) we find the third approximation for the stress components,
(q)
Substituting these expressions for the stress compoilents into Eq. (o) the corresponding torque is 2 cG7ra 4 ( 3 a ) (r) (M1)2 = -2- 1 16 R2
+
By determining from this the constante, and substituting it in expressions (q), we tan find the stress components as funct10ns of the applied torque (M,)2. Along the
(-re, ) •
0.6PR = - -(1 aª
2) + 1.20 -Ra + O.56 .!!... R•
(236)
131. Pure Bending of a Circular Rin S approximations used in th . . g ector. The method of successive e prev1ous artIC1e can be applied also in discussing pure This formula was communica . t ed t 0 S • T'1moshenko m . a letter from O Go"'h Th e e ementar 1 t' · 1 before (see p · 391) gi've for (-ro,· ); va1ner. . h are m . wh 1c d Y so u 1ons ment10ned . ues G'·h goo . agreement w1th the results calculated from Eq (235) 3 o ner, Zoe. cit. · · 1
2
AXIALLY SYMMETRICAL STRESS DISTRIBUTION
THEORY OF ELASTICITY
396
bending of a sector of a circular ring. 1 If two equal and opposite couples M are applied at the ends of a circular ring sector in the plane of the center line of the ring (Fig. 221), they produce strain symmetrical with respect to the z-axis, and the shearing stresses Tr8 and T8• in the meridional cross sections of the ring are zero. The remaining four stress components must satisfy the equations of equilibrium for the case of symmetrical strain (see Art. 116)
aur + ar., + Ur - 11'8 ar az r ar., + au, + Trz ar az r
z Fm. 221.
w1:Jere
4M e= ra•E To get the second approximation we consider ~as small in comparison with R and neglect the products of ~IR and of small corrections in the stresses as small quantiiies of higher order. Equations (d) and (e) then become a(u~\
r
2
- (ur -
r•
2
V'ue
+ T2 (ur
1 a•e +-1+var
ue)
2
a(r~l)l
(a)
+ _1_ a•(o)i 1 + a~ + ~ ( 1 + 1 ~ v) cE
tl(u )
t
1
V'u•
1
a•e
T2 Tra + 1 + V aTaz
l
2
V
1
= O
+ l + v ----af2
A(rt1)1
+
1
+v
1
= O
= O
(e')
a•(e)i
A(u1\
(b)
a•e
+ 1 + v az•
1
V2Trz -
A(u 6\
o
+ 1 + V r ar = o
(d')
a(u!)I
~+---ar =0
= Q
1 ae
1
- ue)
=
cE~
a(rt!)t
~+-----ar-T=O
= O
and the corresponding compatibility equations [see Eqs. (g), Art. 116] V 2u -
397
a•(o)i
a~ar
= 0
=o
where the symbol A means a•/a~• + a•/ar•. We introduce now a stress function c/n. By taking
= 0
Taking, as an example, a ring of constant circular cross section and introducing, instead of r and z, the new coordinates (Fig. 218) (e) r=z ~ = R - r,
(u ) = cE (~•
~
2R cE
1
a•q,1
R ~
(u1)1 =
Eqs. (a) and (b) become
+ r• _ a •) + cE a•q,1 R ar• (g)
cE a•q,1 (r~1) 1 = - R a~ af
(cl)
we satisfy Eqs. (d'). Substituting (g) in Eqs. (e'), we find that the stress function c/>1 should satisfy the equation A Ac/>1 = - 1 1
!~
(e)
+ 2v +V
The boundary conditions for c/>1 are obtained from Eqs. (12). the expression for (u~)i is zero at the circular boundary and
z =dr,
,,
·
'!
ds
(h) As the first term in
m=
we find that As a first approximation we take the sarne stress distribution as occurs in pure
bending of prismatical bars.
Then
(u~)o = (u!)o = (r;i-)o =O
(f)
(u6) 0 = -cE~ 1;
1
Gõhner, loc. cit.
,•
j,_
ds
(ª"'l) ar =o'
!:_ ds
(ª"'l) =o a~
Thus aq,i(ar and aq,ifa~ are constant along the boundary, and we can assume that c/>1 and dq,ifdn are zero at the boundary. Equation (h) together with these boundary conditions completely determines the stress function
THEORY OF ELASTICITY
398
tions for the deflection of a plate clamped at the edges and unif~rmly loa~ed.. ln the case of a circular plate we know the deflection surface. Th1s deflect1on g1ves us the expression for the stress function - 1 + 2v (~2 + r2 - a2)2 >l 64(1 + v)
(k)
CHAPTER 14
Substituting in Eqs. (g) we find the following expressions for the stress components: _ cE {(7 + 6v)(~2 _ a2) + (5 + 2v)l2} (,,.t ) 1 - 16R(l + v) __ cE(l + 2v) ( 3 ~ 2 + r2 _ a2) (,,., ) 1 16R(l + v) cE 1 + 2v (r~1)1 = SR 1 + v ~t
(t)
Snbstituting these expressions in Eqs. (e'), we find éP(u 8\
~ = -
éJ 2 (u 9) 1
--;)f2 =
cE(4 + 5v + 2v 2 ) 2R(l + v)
cE(3v + 2v 2) 2R(l + v)
THERMAL STRESS 132. The Simplest Cases of Thermal Stress Distribution. One of the causes of initial stresses in a body is nonuniform heating. With rising temperature the elements of a body expand. Such an expansion generally cannot proceed freely in a continuous body, and stresses due to the heating are set up. ln many cases of machine design, such as in the design of steam turbines and Diesel engines, thermal stresses are of great practical importance and must be considered in more detail. The simpler problems of thermal stress can easily be reduced to problems of boundary force of types already considered. A.B a first example
éJ2(u9)1 =O
atar Integrating these and adjusting the consta~ts of integr~tion :ot"as l~o ma~::~:~~ tribution of normal stresses over cross sect1ons of the rmg s a ma Y eqm the bending moment M, we find 4M [ (8 + 10v + 4v2W - (6v + 4v2)r2 - (2 + v)a2] (m) (u 8) 1 = - 'll"a• t + 8(1 + v)R Taking
t
=
O and t = a, the stress at the inner point i (Fig. 218) is 2 4M{ 6-+ 9v + 4v (u9)1 = - 'll"aª 1 + R 8(1 + v)
1
~
(n)
Calculations of further approximations result in the following expression for the stress at the inner point1 (~ = a): 2 4M [ a 0.64(a/R) ] 0'9 = - 'll"aª 1+0.87R+1 - (a/R)
(p)
The elementary theory of bending of curved bars, based ~n t~e as~umpti~n that cross sections remain plane and neglecting the stresses ur, g1ves m th1s case ,,. = - 4M [1 + 0.75 _Ra + 0.50 ;: + . . . ]
8 'll"aª letter from O. Gõhner. . h k · This formula was communicated to S . T imos en o m ª 2 See S. Timoshenko, "Strength of Materials," 2d ed., vol. 2, p. 73.
1
r e
_l
1---t
t--1 (a)
X
~ 12
1
-tja:E?J(ó)
Frn. 222.
For v = 0.3, the above equation becomes
+ 0.87 -Rª)
l
o
y
ª}
-4M ( 1 (,,. 8) 1 = - 'll"aª
T
let us consider a thin rectangular plate of uniform thickness in which the temperature T is an even function of y (Fig. 222) and is independent of x and z. The longitudinal thermal expansion aT will be entirely suppressed by applying to each element of the plate the longitudinal compressive stress u,,' = -a.TE (a) Since the plate is free to expand laterally the application of the stresses (a) will not produce any stresses in the lateral directions and to maintain the stresses (a) throughout the plate it will be necessary to distribute compressive forces of the magnitude (a) at the ends of the plate only. These compressive forces will completely suppress any expansion of the plate in the direction of the x-axis dueto the temperature rise T. To get the thermal stresses in the plate, which is free from externa! forces, we have to superpose on the stresses (a) the stresses 399
400
401
THERMAL STRESS
THEORY OF ELA8TICITY
produced in the plate by tensile forces of intensity aTE distributed at the ends. These forces have the resultant
from which
-11" = -3 3 e
J_:º aTE dy
2c
f
+e aETydy
-e
u,." = -3y3 2e
'
f
+e aETydy
-e
Then the total stress is and ata sufficient distance from the ends they will produce approximately uniformly distributed tensile stress of the magnitude -1 2c
f
+e aTEdy
-e
so that the thermal stresses in the plate with free ends at a considerable distance from the ends will be u,. = -1 2c
f
+e aTEdy - aTE
(b)
-e
Assuming, for example, that the temperature is distributed parabolically and is given by the equation T = To
2
(e)
This stress distribution is shown in Fig. 222b. N ear the ends the stress distribution produced by the tensile forces is not uniform and can be calculated by the method explained on page 167. Superposing these stresses on the compressive stresses (a), the thermal stresses near the end of the plate will be obtained. If the temperature T is not symmetrical with respect to the x-axis, we begin again with compressive stresses (a) suppressing the strain E,.. ln the nonsymmetrical cases these stresses give rise not only to a resultant force -
J_~º aET dy but also to a resultant couple
-
J_~º aETy dy,
and in order to satisfy the conditions of equilibrium we must superpose on the compressive stresses (a) a uniform tension, determined as before, and bending stresses u",. = uy /e determined from the condition that the moment of the forces distributed over a cross section must be zero. Then +e uy 2d y +o aETy dy = O
f
-o
e
f
-o
f
+e
-e
aET dy
+ 233ye
f
+o
-e
aETy dy
aET
IJ":c
2 aToE - aToE ( 1 - yc2) u,. = "3
~
+ 21e
(d)
ln this discussion it was assumed that the plate was thin in the z-direction. Suppose now that the dimension in the z-direction is large. We have then a plate with the xz-plane as its middle plane, and a thickness 2c. Let the temperature T be, as hefore, independent of x and z, and soa function of y only. The free thermal expansion of an element of the plate in the x- and z-directions will be completely suppressed by applying stresses u,., u, obtained from Eqs. (3), page 7, by putting E,. = E, = -aT,
(1 - ~:)
we get, from Eq. (b),
1,
u,. = -aET
= u. = - 1 - v
(e)
The elements can be maintained in this condition by applying the distributions of compressive force given by (e) to the edges (x = constant, z = constant). The thermal stress in the plate free from externa! force is obtained by superposing on the stresses (e) the stresses dueto application of equal and opposite distributions of force on the edges. If T is an even function of y such that the mean value over the thickness of the plate is zero, the resultant force per unit run of edge is zero, and by Saint-Venant's principle (Art. 18) it produces no stress except near the edge. If the mean value of Tis not zero, uniform tensions in the x- and z-directions corresponding to the resultant force on the edge must be superposed on the compressive stresses (e). If in addition to this the temperature is not symmetrical with respect to the xz-plane we must add the bending stresses. ln this manner we finally arri~e at the equation Ux
= Uz
1 = - 1a TE - v + 2c(l -
v)
f
+e
-e aTE dy
3y !+e - v) -e aTEy dy
+ 2cª(l
(f)
THEORY OF ELASTICITY
402
which is analogous to the Eq. (d) obtained before: By 1_1si~g E~. (f) we can easily calculate thermal stresses in a pla~e, li the d1stnbut10n of temperature T over the thickness of the plate is known. Consider as an example, a plate which has initially a uniform temperature To d which being cooled down by maintaining the surfaces y = ±e ata constant an t T i By Fourier's theory the distribution of temperature at any t empera ure ,. instant tis 3 T = Ti (To - Ti) (e-P1 1 cos e--Pat cos ;;' (g)
is
+~
~~
- 3•p p = n•p 1 pi, pa i, · · · ' " ' ing in Eq. (f), we find ·
in
w
h" h lC
_ _ 4aE(To - T1) [e-Pit u, - u, 11"(l _ v)
e~71"
_
•
•
~
•
cos 71"Y) 2c
'
+ · · ·)
are certain constants.
For y =
At the middle plane y
=
+ ~ e-p,t (;11"
l -
5 - cos ; ; )
+ · · ·]
(h)
-
cos 71"Y) 2c
O we obtain compressive stresses
1~
from which
1
'!
(l _71"~)
2c
1
',,1
~ 1 1
'
.
1\
1
y
= '2 (T 1 + T 2) + 2 (T 1 + T 2) e
(i)
This problem was discussed by Lord Rayleigh, Phil. Mag., series 6, vol. 1,
p. 169, 1901.
li
pas rs
-1
paª
(m)
= -2r 3
= - p,
and the increase of this radius, due to pressure p, is
If the surfaces y = ±e of a plate are maintained at two different temperatures T 1, T 2, a steady state of heat flow is esta?lished afte: a certain time and the temperature is then given by the lmear funct10n 1
(l)
At the radius r = a we obtain
y = ±0.560c
T
(k)
C
The thickness of the plate does not enter in this formula, but in the case of a thicker plate a greater difference of temperature between the two surfaces usually exists. Thus a thick plate of a brittle material is more liable to break due to thermal stresses than a thin one. As a last example let us consider a sphere of large radius and assume that there occurs a temperature rise T in a small spherical element of radius a at the center of the large sphere. Since the element is not free to expand a pressure p will be produced at the surface of the element. The radial and the tangential stresses due to this pressure at any point of the sphere ata radius r > a can be calculated from formulas (197) and (198) (see page 359). Assuming the outer radius of the sphere as very large in comparison with a we obtain from these formulas
~-cos71"Y=0
71"
V
aETi
The points with zero stresses are obtained from the equation
1'
!
aT y - --Ti-
= (
4aE(To - Ti) -p 1t ~ 11"(1 - v) e 71"
_ _ _ 4aE(To - T1) e-p,t u, - u, 11"(1 - v)
[,
and Eqs. (e) give
The maximum stress is
±e we have tensile stresses = u, =
(j)
+ !3 6-p,t ( 3271" + cos ª;y) e
_ _ 4aE(To - T1) • 6 -p 1t e~ u, - u, 11"(1 - v) 71"
u,
Substitution in Eq. (f) shows that the thermal stresses are zero, 1 provided, of course, that the plate is not restrained. If the edges are perfectly restrained against expansion and rotation, the stress induced by the heating is given by Eqs. (e). For instance li T2 = -Ti we have from (i)
Substitut-
After a moderate time the first term acquires dominant importance, and we can assume
403
THERMAL STRESS
Âr
= (ae,)r=a = ; [
+
~; (1
+ v)
This increase must be equal to the increase of the radius of the heated spherical element produced by temperature rise and pressure p. Thus 1 In general, when T is a linear function of x, y, z, the strain corresponding to free thermal expansion of each element, viz.,
'Yzu = 'Yz• = 'Yu• =
O
satisfies the conditions of compatibility (129) and there will be no thermal stress.
we obtain the equation
aTa from which
1; (1 -
2v)
~; (1
=
+ v)
2 aTE 31 - JI
(n)
p=---
Substituting in equations (m) we obtain the formulas for the stresses outside the heated element
= -
2 aTEa 3 3 (1 - v)r 3'
1 aTEa 3
405
THERMAL STRESS
THEORY OF ELASTICITY
404
= -3 (1 - v)r 3
(o)
It was pointed out on page 95 that a normal stress on a straight boundary prod~ces a like normal stress parallel to and at the boundary. ~ence the tens10ns aET will produce tensile stress aET in the x-direct10n. Both normal stresses die away as we proceed into the plate nor,mal to th~ edge. ?n s_uperposing these stresses on the compressive stress (a) m ~he y-direct10n, we obtain curves 1 for
133. Some Problems of Plane Thennal Stress. Suppose that a strip of thin plate (Fig. 223) is nonuniformly heated so that the temperature T is a function of the longitudinal coordinate x only, being
L.1 I
-
1
aET
/
# -~
+I I
I
-
/
l_T e 2
+/
I
B
1.-=1 e
+
fb)
~
,/
Frn. 223.
Edge +
uniform across any given cross section. If the plate is cut into strips such as AB (Fig. 223), these strips expand vertically by different amounts. Due to the mutual restraint there will be stresses set up when they are in fact attached as in the plate. Considering the unattached strips, their vertical expansion is suppressed if they are subjected to compressive stress
1~ ,,' 1
'1
u11 = -aET
FIG. 224.
If the
~emperature
T is a periodic function of x the application of
~~:~ens10ns aET presents a problem of the type considered in Art. 23.
T = To sin ax
(a)
by applying such stress at the ends A and B of each strip. The strips fit together as in the unheated plate. To arrive at the thermal stress we must superpose on (a) the stress due to the application of equal and opposite forces, i.e., tension of intensity aET, along the edges y = ±e of the strip. If the heating is confined to a length of the strip short in comparison with its width 2c, such as CDFE in Fig. 223, the effect of the tensions aET wi1l be felt only in the neighborhood of CD on the top edge, and of EF on the bottom edge. Each of these neighborhoods can then be considered as presenting a problem of the type considered in Art. 34.
C.L.
(b)
we find from Eqs. (k) of Art. 23, putting A = B = -aETo in accordance with Eq. (f), rrx
= -2aETo (ac cosh ac - sinh ac) cosh ay -
sinh 2ac rr 11
Txu 1
+ 2ac
ay sinh ay sinh ac · sin ax
= 2aETo (ac cosh ac + sinh ac) cosh ay - ay sinh ay sinh ac .
=
sinh 2ac + 2ac · sm ax ay - ay cosh ay sinh ac 2aETo ac cosh ac sinh sinh 2ac + 2ac . cos ax
J. N. Goodier, Physics, vol. 7, p. 156, 1936.
406
THEORY OF ELASTICITY
Together with the compressive stress
lf u denotes the radial displacement we have, from Art. 28, Er
du,
dr
+
= O
r
(a)
obtained from Eq. (40), page 58, by putting R = O. The shear stress is zero on account of the symmetry. The ordinary stress-strain relations, Eqs. (52), page 66, for plane stress, require modification since now the strain is partly dueto thermal expansion, partly due to stress. If Er represents the actual radial strain, Er - aT represents the part due to stress, and we have
T,9
Er -
aT =
E1 (ur
-
V
(b)
and similarly E9 - aT = 1
'
~
E1 (u9
-
V
(e)
O'r
=
E
l _ v2 [er
+
VE9 -
(1
.!}_ dr
E
i
u = (1
l .
.
. i
+ v)a dT dr
(g)
+
11·
v)a r
a
Tr dr
e + C1r + _: r
(h)
wh~re t~e lower limita in the integral can be chosen arbitrarily. For a disk with a hole it may be the inner radius. For a solid disk we may take it as zero. The stress components are now found by using the solution (h) in Eqs. (f), and substituting the results in Eqs. (d). Then E
1 ('
1 {'
= aE·T2 },. Trdr - aET
v2
[
C1 (1
+ 1 _E v
2
[
+ v)
C 1 (1
- C2 (1 - v)
~]
+ v) + C (1- v)~] 2
(i) (j)
?1,
The consta?ts C2 are determined by the boundary conditions. For a sohd d1sk, we take a as zero, and observing that
11· r
Tr dr= O
O
(d)
C1 = (1 - v) (e)
~2 fb b
}o
Tr dr
The final e"pressions for the stresses are consequently
~la' Tr dr) = aE(-T + ; lab Trdr +~la' Trdr)
i
d(ru)] _ ( 1 r dr -
we see from Eq. (h) that C2 must vanish in order that u may be zero at the center. At the edge r = b we must have O'r = O and therefore from Eq. (i) '
1 The problem wa.s discussed by J. P. Den Ha.rtog, J. Franklin Inst., vol. 222, p. 149, 1936, in connection with the therma.1 stress produced in the process of welding.
;I
+ v)a dr
Integration of this equation yields
11nd with these Eq. (a) becomes d dT r dr (E,+ VEo) + (1 - v)(er - Eo) = (1 + v)ar dr
(f)
dT
(1
=
[!
r-+O
= -1 _ v2 [E9 + VEr - (1 + v)aT]
=r
which may be written
lim-
+ v)aT], <19
u
E,
r
Solving (b) and (e) for u,, u 9 we find
1
du dr 1
=
Substituting these in (e) we obtain d 2u 1 du u dr2 + dr - T2
134. The Thin Circular Disk: Temperature Symmetrical about Center. When the temperature T does not vary over the thickness of the disk, we may assume that the stress and displacement dueto the heating also do not vary over the thickness. The stresses u, and uo satisfy the equation of equilibrium
407
THERMAL STRESS
aE (:2
lab Tr dr 2
-
(237) (238)
408
11'
1 +-v · a · u = 1- v r
These give finite values at the center since
lim ..!. (' Tr dr = !2 To
aE
,,_.o r 2 } o
where To is the temperature at the center. 135. The Long Circular Cylinder. The temperature is taken to be symmetrical about the axis, and independent of the axial coordinate z.1 We shall suppose first thàt w, the axial displacement, is zero throughout, and then modify the solution to the case of free ends. We shall now have three components of stress,
aT =
E1 [
- v(
E, -
aT =
E1 [<1,
-
+ <1,)]
(239)
v(
But since w = O, e, = O, and the third of Eqs. (239) gives
+
- aET
(a)
On substituting this into the first two of Eqs. (239), these equations become 1 - p2 ( p ) Er (1 + v)aT = --E-
!i
li i•
LJ '"
1.
It may be seen at once that these equations can be obtained from the corresponding equations of plane stress, Eqs. (b) and (e) of the preceding article, by putting, in the latter equations, E/(1 - v2) for E, v/(1 - v) for v, and (1 + v)a for a. . Equations (a) and (f) of the preceding article re1Il~in_ yal_i~ ...he~E;J, The solution for u,
409
THERMAL STRESS
THEORY OF ELASTICITY
a
1 1r
•
'2
a
·
Tr dr
+ C 1r + -Cr
Tr dr
E +1+
aET
( P
2
(e)
C1 C2) 1 - 2v - T2
(d)
E- ( --C1 + -C +1 + v 1 - 2v r
2)
2
(e)
and, from Eq. (a), _ aET 1 - v
+ (1
2vEC1 - 2v)
+ v) (1
(f)
Normal force distributed according to Eq. (f) must be applied to the ends of the cylinder in order to keep w = O throughout. If we superpose a uniform axial stress
lb o
Trdr
(g)
The resultant of the axial stress (f) is
{b
__
}o
21íaE [b 1 _ v }o Trdr
+
(l
+
2vEC1 2 v)(l _ 2 v) 7rb
and the resultant of the uniform axial stress C 3 is C 3 • 7íb 2 • The value of e 3 making the total axial force zero is therefore given by _ • 27íaE {b 2 C 3 7íb - 1 _ v }o Tr dr - (l
2vEC1 2 _ 2 v) 7íb
+ v)(l
(h)
THEORY OF ELASTICITY
410
The final expressions for u,
are, from Eqs. (e), (d), (e), (f), (g),
1+ u =-v · a [ (1 - 2v) -r2 1 -
b
V
Íub Tr dr + -r1 O
f
r
O
Tr dr ]
(1 (b Tr dr - f21Jr Tr dr) aE- (1 =- fb Tr dr + 1Jr Tr dr - T ) r
=
1
ª_E v
1 -
=
1
V
b2
}
ª! v (;:
(240)
2
O
Íob
i
A ..Jo
(ti.
n=l
e-p.t
(i)
2 A,. "" {J ..Ji (tJ,.)
k
fJ,.2
cP 'bi
in which k is the thermal conductivity, e the specific heat of the material, and P the density. Substituting series (i) into Eq. (241) and taking into account the fact that 8
we find that
1 u _ 2aETo r -
1 -
P
~
L.,
e-p.t {_!_ _ - \ ~ Ji[P,.(r/b)]} fJ,.2 tJ,. r J i(fJ,.)
(k)
n=l
i It is assumed that the surface of the cylinder suddenly assumes the temperature zero. If the temperature of the surface is Ti instead of zero, then To - Ti must be put instead of To in our equations. 2 See Byerly "Fourier Series and Spherical Harmonics," p. 229. The calculation of thermal stresses for this case is given by A. Dinnik, "Applications of Bessel's Function to Elasticity Problems," pt. 2, p. 95, Ekaterinoslav, 1915. See also C. H. Lees, Proc. Roy. Soe. (London), vol. 101, p. 411, 1922. a See E. Jahnke und F. Emde, "Funktionentafeln," p. 165, Berlin, 1909.
li
ll·:· .
1
~ j
+ _!_ ~ Ji[tJ,.(r/b)] {J,. 2 r
J i(fJ,.)
_ Jo[{J,.(r/b)]}
(l)
fJ,.J i(tJ,.)
~
L.,
n=l
e-P.•
{.!.2 _Jo[tJ,.(r/b)]} 1(fJ,.) {J,.
{J,.J
(m)
Formulas (k), (l), and (m) represent the complete solution of the problem. Several numerical examples can be found in the papers by A. Dinnik and C. H. Lees, mentioned above. 1
Figure 225 represents 2 the distribution of temperature in a steel cylinder. It is assumed that the cylinder had a uniform initial temperature equal to zero and that beginning from an instant t = O the surface of the cylinder is maintained ata temperature T 1 • The temperature distributions along the radius, for various values of t/b 2 (tis measured in seconds and 601---,\--~...+.......i~:I--+-'~ u b in centimeters), are represented k; 50 l---+-\-~~---t...---1------t by curves. It will be seen from ~401----+--'\l---'t--~--f~f--t--=,....~::::I Eqs. (i) and (j) that the tempera301--+-----+->.rt-3'cl---i-'"'*ture distribution for cylinders of 201--+-+-.>o,f--1-'-.+------+---+---"i=-t various diameters is the sarne if the time of heating tis proporo tional to the square of the diameter. From thefigure, theaverage temperature of the whole cylinder FIG. 225. and also of an inner portion of the cylinder of radius r can be calculated. Having these temperatures we find the thermal stresses from Eqs. (241), (242), and (243). If we
5
and the constants p,. are given by the equation =
e-,..,{_!_ {J,. 2
u = 2aETo • 1 - "
in which J 0 (tJ,.r/b) is the Bessel function of zero order (see page 385), and the {J's are the roots of the equation J 0 (fJ) = O. The coefficients of the series (i) are
p,.
~
1 - " L.,
(242) (243)
n
= 2aETo
Substituting series (i) in Eq. (243) we find
Take for example a long cylinder with a constant initial temperature To. If, beginni~g from an i~stant t = O, the lateral surface of the cyl~nder is ~a~tained ata temperature zero,i the distribution of temperature at any mstant tis given by the series 2
T =To
(241)
O
Tr dr - T)
ln the sarne manner, substituting series (i) in Eq. (242), we obtain
n=l
0
b2
411
THERMAL STRESS
1 Temperature distribution in solids during heating and cooling was discussed by Williamson and Adams, Phys. Rev., series 2, vol. 14, p. 99, 1919. An experimental investigation of the effects of fire and water on columns has been made by Ingberg, Griffin, Robinson, and Wilson. See U. S. Bur. Standards, Tech. Paper 184, 1921. 2 The figure is taken from A. Stodola, "Dampf- und Gasturbinen," 6th ed., p. 961, 1924.
413
THEORY OF ELASTICITY
THERMAL STRESS
take a very small value for t, the average temperatures, mentioned above, approach zero and we find at the surface
surface is zero, the temperature T at any distance r from the center is represented by the expression
412
O'r
=O,
ue =
O'z
= -
T =
aET1 1 -
p
This is the numerical maximum of the thermal stress produced in a cylinder by heating. It is equal to the stress necessary for entire suppression of thermal expansion at the surface. During heating this stress is compression, during cooling it is tension. In order to reduce the maximum stresses it is the usual practice to begin the heating of shafts and rotors with a somewhat lower temperature than the final temperature T 1, and to increase the time of heating in proportion to the square of the diameter. Cylinder with a Concentric Circular Hole. 1 The radius of the hole being a, and the outer radius of the cylinder b, the constants C1, C2 in Eqs. (e), (d), (e) are determined so that ur will be zero at these two radii. Then
aE 1 (b d - 1 - P • b2 }a Tr r
E
+1+P
(
av~f;g(b/a) [ - log ~ - (b2 ~ a2) ( 1 - ~) log ~] "º = 2(1 - ªv~f;g(b/a) [ 1 - log~ - (b2 ~ a2) ( 1 + ~) log~] = 2(1 -
Ur
_ Uz
-
(247)
2
aETi [ b 2a l b] 2(1 - v) log(b/a) 1 - 21 og T - (b 2 - a 2 ) og
a
If Ti is positive, the radial stress is compressive at all points and becomes zero at the inner and outer surfaces of the cylinder. The
C1 C2) 1 - 2v - b2 =O FIG. 226.
EC2 aE a2 (b 1 + v = 1 - v b2 - a2 j ª Tr dr EC1
+ v) (1
- 2v)
= ~ __1_ 1 -
P
b2
a2
-
stress components ue and u, have their largest numerical values at the inner and outer surfaces of the cylinder. Taking r = a, we find that
fb Trdr }
(ue)r=a = (uz)r=a =
a
Substituting these values in (d), (e), and (f), and adding to the last the axial stress C 3 required to make the resultant axial force zero, we find the formulas aE 1 2 - 2 (b (' ) (244) 2 j ª Tr dr ur = l _ v f2 b2 _ j ª Tr dr
aE 1 - V
0"8
=
O"z
=1-
(r aª 1 (r + a {b {' f2 b2 - a2 J Tr dr + J Tr dr -
aE ( V
2
2
a
2
b2 - a2
J(ba Tr dr -
a
T
)
(245) (246)
Consider, as an example, a steady heatjlow. If Ti is the temperature on the inner surface of the cylinder and the temperature on the outer 1
(n)
Substituting this in Eqs. (244), (245), and (246), we find the following expressions for the thermal stresses: 1
and from these
(1
Ti log ~ log(b/a) r
See R. Lorenz, Z. Ver. deutsch. Ing., vol. 51, p. 743, 1907.
b ( 1 - b2 - a2 log a 2(1 - v) log-
2b
b)
(248)
(l -
a 2 log ~)
(249)
aET,
2
a
For r
=
b we obtain (ue).....o = (u,)r=b =
aETi b 2(1 - v) log a
b2
2
~
2
The distribution of thermal stresses over the thickness of the wall for a particular case a/b = 0.3 is shown in the Fig. 226. If T, is positive, the stresses are compressive at the inner surface and tensile at the outer surface. In the case of such materials as stone, brick or concrete . are weak in tension, cracks are likely to start on the' outer surface' wh1ch of the cylinder under the above conditions. 1
Charts for the rapid calculation of stresses from Eqs. (247) are given by L. Barker, Engineering, vol. 124, p. 443, 1927.
414
THEORY OF ELASTICITY
THERMAL STRESS
If the thickness of the wall is srnall in cornparison with the outer radius of the cylinder, we can sirnplify Eqs. (248) and (249) by putting
ln the foregoing discussion it was assumed that the cylinder is very long and that we are considering stresses far away from the ends. N ear the ends, the problem of thermal-stress distribution is more complicated due to local irregularities. Let us consider this problem for the case of a cylinder with a thin wall. Solution (250) requires that the normal forces shown in Fig. 227a should be distributed over the ends of the cylinder. To find the stresses in a cylinder with free ends we must superpose o~ the stresses (250) the stresses produced by forces equal and oppos1te to those shown in Fig. 227a. ln the case of a
b
-=l+m a
'
b log a
=
m2 m - 2
and considering m as a srnall quantity. (as),._
=
(a,)r=a
=-
(ae).-õ = (a,),=1> =
3
+ -m3 -
Then
2 (~E!_iv) ( 1 + i-)
2(~E!_iv) ( 1 -
l)
(248') (249')
If the ternperature at the outer surface of the cylinder is different frorn zero, the above results can be used by substituting the difference between the inner and the outer ternperatures, Ti - T 0 , in all our equations instead of Ti. In the case of a very thin wall we can rnake a further sirnplification and neglect the terrn m/3 in cornparison with unity in Eqs. (248') and (249'). Then aET, (as)._ = (a,),._ = - 2(1 - v) (250) aET; (ae)r=b = (a,),=1> = 2(1 - v) and the distribution of therrnal stresses over the thickness of the wall is the sarne as in the case of a flat plate of thickness 2c = b - a, when the ternperature is given by the equation (Fig. 222)
T
=
415
T,y (b - a)
and the edges are clarnped, so that bending of the plate, due to nonuniforrn heating, is prevented [see Eq. (k), Art. 132]. If a high-frequency fluctuation of ternperature is superposed on a steady heat flow, the therrnal stresses produced by the fluctuation can be calculated in the sarne rnanner as explained for the case of flat plates (see Art. 132). 1 1 Thermal stresses in cylinder walls are of great practical importance in the design of Diesel engines. A graphical solution of the problem, when the thickness of the wall of the cylinder and the temperature vary along the length of the cylinder, was developed by G. Eichelberg, Forschungsarbeiten, No. 263, 1923. Some information regarding temperature distribution in Diesel engines can be found in the following papers: H. F. G. Letson, Proc. Mech. Eng., p. 19, London, 1925; A. Nãgel, Engineering, vol. 127, pp. 59, 179, 279, 466, 626, 1929.
z
z
(aJ
(ó) FIG. 227.
thin wall of thickness h these forces can be reduced to bending moments M, as shown in Fig. 227b, unüormly distributed along the edge of the cylinder and equal to aET; h2 M= (o) 2(1 - v) 6 per unit length of the edge. To estimate the stresses produced by these moments, consider a longitudinal strip, of width equal to unity cut out from the cylindrical shell. Such a strip can be treated as aba; on an elastic foundation. The deflection curve of this strip is given by the equation 1 u
Me-fJ•
= 2132D (cos {3z
- sin {3z)
(p)
in which
fJ
=
D= ~_E_hª_ 12(1 - v2)
(q)
ª.nd e is the middle radius of the cylindrical shell. Having this deflection curve, the corresponding bending stresses a, and the tangential 1
See S. Timoshenko, "Strength of Materials,'' 2d ed., vol. 2, p. 166.
THEORY OF ELASTICITY
THERMAL STRESS
stresses uo can be calculated for any value of z. The maximum deflection of the strip is evidently at the end z = O, where
On:account of the symmetry there will be three non-zero stress components, the radial component u,, and two tangential compon'ents G't as in Art. 121, and these must satisfy the condition of equilibrium, i~ the radial direction, of an element [Fig. 202, Eq. (e), page 359]
416
M acTi~ (u)z=O = 2[j2D = 2 .y3 (1 - 11)
du, dr
The corresponding strain component in the tangential direction is EO
u e
= - =
aTi~ 2v3 (1 - 11)
(r)
uo = E Eo
+
llG'z
(~ .y3 -
li
+ 1)
G't)
o
=
(a)
Er - aT =
E1 (u,
(b)
- 211u 1)
1 Et - aT = E [ut - 11(u,
+u
(e)
1)]
and, u being the radial displacement, we have
Adding this stress to the corresponding stress calculated from Eqs. (250), the maximum tangential stress for a thin-walled cylinder at the free end is aET, (uo)max. = 2 (l _ li)
G'r -
The stress-strain rela tions are
The stress component in the tangential direction at the outer surface of the wall is then obtained, using Hooke's law, from the equation aET, ~ 11aET, = 2 V3 (1 - 11) 2(1 - 11)
+ r2 (
417
(251)
Assuming v = 0.3, we find aETi (uo)max. = 1.25 2 (l _ li) Thus the maximum tensile stress at the free end of the cylinder is 25 per cent greater than that obtained from Eqs. (250) for the stress at points remote from the ends. From Eq. (p) it can be seen that the increase of stress near the free ends of the cylinder, since it depends on the deflection u, is of a local character and diminishes rapidly with increase of distance z from the end. The approximate method of calculating thermal stresses in thinwalled cylinders, by using the defl.ection curve of a bar on an elastic foundation, can also be applied in the case in which the temperature varies along the axis of the cylindrical shell. 1 136. The Sphere. We consider here the simple case of a temperature symmetrical with respect to the center, and soa function of r, the radial distance, only. 2 S. Timoshenko and J. M. Lessells, "Applied Elasticity," p. 147, 1925, and C. H. Kent, Trans. A.S.M.E., Applied Mechanics Division, vol. 53, p. 167, 1931. 2 The problem was solved by Duhamel, loc. cit.; F. Neuman, Abhandl. Akad. Wiss., Berlin, 1841; see also his "Vorlesungen über die Theorie der Elastizitiit der
du Er = dr'
Et =
ru
(d)
From (b) and (e) we find E
= (1 + ll)(l _
=
211 ) [(1 - 11)Er +
2VEt -
(1 + 11)aT]
E
(1 + ll)(l _ 211 ) [Et + llEr - (1 + 11)aT]
(e)
(j)
Substituting these in (a), then replacing E,, Et by the values given in (d) we find the differential equation for u 2
du dr 2
+ ~ du
r dr
+
_ ~~ _ 1 11 • a dT r 2 - 1 - 11 dr
(g)
which can be written d dr
[1T2
d (r 2u) dr
The solution is 1+11 u = - · a · -1 1 - 11 r2
dT J=1+11 --·a1 - 11 dr
i' a
Tr 2 dr
er2 + C 1r + ~
(h)
where C1 and C2 are constants of integration to be determined Iater ~rom boundary conditions, and a is any convenient Iower limit for the
mtegral, such as the inner radius of a hollow ·sphere.
1
festen Kôrper," LeiP;zig, 1885; J. Hopkinson, Messenger of Math., vol. 8, p. 168, 187_9. Nonsymmetr~cal te~peratures were considered by C. W. Borchardt, Monatsber. Akad. Wiss., Berlm, 1873, p. 9. ·
THERMAL STRESS
THEORY OF ELASTICITY
418
This solution can be substituted in Eqs. (d), and the results used in Eqs. (e) and (f). Then
2aE . !_ {' Tr 2 dr+ EC1 _ 2EC2 . !_ 1 - v r3 } a 1 - 2v 1 + v r3
(j)
[The mean temperature of the whole sphere + (-Í the mean temperature within the sphere of radius r) - jT]
W e shall now consider several particular cases. Solid Sphere. ln this case the lower limit a of the integrais may be taken as zero. We must have u = O at r = O, and, from Eq. (h) this requires that C2 = O, since
If the distribution of temperature is known, the calculation of stresses in each particular case can be carried out without difficulty. 1 An interesting example of such calculations was made by G. Grünberg 2 in connection with an investigation of the strength of isotropic materiais subjected to equal tension in three perpendicular directions. If a solid sphere at a uniform initial temperature To is put in a liquid at a higher temperature T 1, the outer portion of the sphere expands and produces at the center of the sphere a uniform tension in all directions. The maximum value of this tension occurs after a time
11 1
= ~ . !_ {' Tr 2 dr 1 - v r3
}
+
a
lim -1 r->O r2
f' O
+
EC1 1 - 2v
EC2 . !_ _ aET 1 v r3 1 - v
+
Tr 2 dr = O
The stress components given by (7:) and (J) will now be finite at the center since lim !_ [' Tr2 dr
,........o r 3
}o
= To
2
t = 0.0574
3
where To is the temperature at the center. The constant C1 is determined from the condition that the outer surface r = b is free from force, so that 11, = O. Then from Eq. (i), putting "• = O, a = O, C2 =O, r = b, we find
EC1 = 2aE .l_ 1 - 2v 1 - v b3
fb Tr
}o
2
dr
11,
111
2 (1 fb 1 (' ) (' 2 ) ª_E v (2bª }o[b Tr dr+ 1 :;:a }o Tr dr - T E
= 1 '.:__ v b3 } Tr 2 dr - :;:a } 0 Tr 2 dr 0 =
(252)
2
1
The mean temperature of the sphere within the radius r is 4ir
r ;
Tr 2 dr
1ír 3
=
3 3r
J.r o
Tr 2 dr
Therefore the stress"• at any radius ris proportional to the difference between the mean temperature of the whole sphere and the mean tem-
b ~p
(k)
Here b is the radius of the sphere, k the thermal conductivity, e the specific heat of the material, and p the density. The magnitude of this maximum tensile stress is 3 11r
and the stress components become
~.
2aE 3(1 - v) multiplied by the expression
= _
r
ll.
The tangential stress at any point is
(i)
"
I'
perature of a sphere of radius r.
419
=
aE = 0.771 2 (l _ v) (T1 - To)
(l)
The maximum compressive stress occurs at the surface of the sphere at the moment of application of the temperature T 1 and is equal to aE(T1 - T 0)/(l - v). This is the sarne as we found before for a cylinder (see page 412). Applying Eqs. (k) and (l) to the case of steel, and taking b = 10 cm. and T 1 - To = lOOºC., we find O'r = 0' 1 = 1,270 kg. per square centimeter, and t = 33.4 sec. Sphere with a Hole at the Center. Denoting by a and b the inner and outer radii of the sphere, we determine the constants C 1 and C 2 in (i) 1 Several examples of such calculations are given in the paper by E. Honegger, Festschrift Prof. A. Stodola, Zürich, 1929. A table for calculating the distribution of temperature during cooling of a solid sphere is given by Adams and Williamson, loc. cit. 1 G. Grünberg, Z. Physik, vol. 35, p. 548, 1925. ª It was assumed in the analysis that the surface of the sphere takes at once the temperature T 1 of the fluid.
THEORY .OF ELASTICITY
420
and (j) from the conditions that
EC1 1 - 2v
_
f
2aE ..!_ 1 - v b3
_ 2EC2 . _!__ = O 1 + v a3
Tr 2 dr
b
+
a
EC i
2EC
_
1 - 2v
1
+
2...!_b = O 3
V
=
2aE [ r r=-;; (bª
3
a
-
3
_ aª)rª
f
3
= 2aE [ 2r +aª
1 - v 2(b
a 3)r 3
3 -
1 ('
2
2
ª Tr dr - ;;:a l ª Tr dr b
fb Tr2 dr+__!__3
la
2r
-
~ T]
(253)
2
Thus the stress components can be calculated if the distribution of temperature is given. Consider, as an example, the case of steady heat flow. We denote the temperature at the inner surface by Ti and the temperature at the outer surface we take as zero. Then the temperature at any distance r from the center is T =
__I_!!__ (~ b- a r
1)
b = a(l
for r =a,
= b,
for r
]
f' Tr2 dr la
ln the case of a spherical shell of small thickness we put
= aETi bª
~aª [a + b -
= aETi
ab
1 - v
2(~E!_iv) (1 + ~ m) = 2(~E!_,v) (1 - ~ m)
If we neglect the quantity im we arrive at the sarne values for the tangential stresses as we obtained before for a thin cylindrical shell [see Eqs. (250)] and for a thin plate with clamped edges. 137. General Equations. The differential equations (132) of equilibrium in terms of displacements can be extended to cover thermal stress and strain. The stress-strain relations for three-dimensional problems are 1 E,. aT = E [
(m)
Ey -
aT =
1 E [uu
- v(
+
Ez -
aT =
1 E [
- v(
+
1 - v b3
-
a3
[a + b -
'Tyz
'YY•
+ ab + a2) + a2~2] r _..!._ (b2 + ab + a2) - a2b21 2r 2rª
_! (b2
The stress
> O,
aET.;
e=
It becomes a
= - 2(1 - v)
b(b - a)(a + 2b) bª - a 3
1
E (1
- 2v)8
=
Àe
+ 2GE,,
When r = a, we
(n)
When r = b, we obtain
aET.; a(b - a)(2a + b)
(b)
= 7"G"''
'Yzz
+ 3aT
Using this, and solving for the stresses from Eqs. (a), we find
increases as r increases.
=a'
(a)
Equations (b) are not affected by the temperature because free thermal expansion does not produce angular distortion in an isotropic material. Adding Eqs. (a) and using the notation given in (7) we find
r
We see that the stress
neglect~
= -
Substituting this in expressions (253), we find
+ m)
where m is a small quantity. Substituting in (n) and (o) and ing higher powers of m, we obtain
Solving for C 1 and C 2 and inserting the results in (i") and (j) we find
421
THERMAL STRESS
aET 1
_
2
v
(e)
Substituting from this and Eqs. (ô) into the equations of equilibrium (127), and assuming there are no body forces, we find
(À (o)
-
+ G) ax ae + G v2u -
_!!§___ aT =
1 - 2v
ax
o
... ' INSTITU~~UL
(254)
-~-·-
T 1 l· -'
1 :,
·-···" ,..;.._____..
PGülSWiit O
~,
R A.
BIBLIOT~A C~NtR~L~
lu
422
THEORY OF ELASTICITY
THERMAL STRESS
These equations replace Eqs. (131) in the calculation of thermal stresses. The boundary equations (128), after substituting from Eqs. (e) and (6) and assuming that there are no surface forces, become
superposing hydrostatic pressure (f) on the stresses produced by body forces (d) and surface forces (e). The sarne conclusion may be reached in another way. Imagine that the body undergoing nonuniform heating is su bdivided into infinitely small elements and assume that the thermal strains e,, = Ey = Ez = aT of these elements are counteracted by applying to each element a uniform pressure p, the magnitude of which, from Eq. (8), is given by (f). In this way thermal strain is removed, and the elements fit one another and form a continuous body of the initial shape. The pressure distribution (f) can be realized by applying certain body forces and surface pressures to the above body formed by the elements. These forces illust satisfy the equations of equilibrium (127) and the boundary conditions (128). Substituting in these equations
aE r
1 - 2v l
(íJu
íJu
= >..el + G ax l + ay m
) + íJu az n + G (ªu l + íJv m + íJw n) ax ax ax
(255)
Comparing Eqs. (254) and (255) with Eqs. (131) and (134) it is seen that terms
-
aE ar , _____
1 - 2v ax
aE
ar
- 1 - 2v íJy'
aE
ar
- 1 - 2v Tz
take the places of components X, Y, Z of the body forces, and terms
aEr l 1 - 2v '
aEr 1 - 2v m,
aE ar, y = ______ 1 - 2v ay
z=
_
aE
ar
(d)
1 - 2v az
and normal tensions
aEr 1 - 2v
(e)
distributed over the surface. If the solution of Eqs. (254) satisfying the boundary conditions (255) is found, giving the displacements u, v, w, the shearing stresses can be calculated from Eqs. (b) and the normal stresses from Eqs. (e). From the latter equations we see that the normal stress components consist of two parts: (a) a part derived in the usual way by using the strain components, and (b) a "hydrostatic" pressure of the amount
aEr 1 - 2v
=
=
aEr
= -p = - 1 - 2/
Txy = Tzz = Tyz =
Ü
(g)
we find that, to keep the body formed by the elements in its initial shape, the necessary body forces are
aEr ---n 1 - 2v
replace components X, Y, Z of the surface forces. Thus the displacements u, v, w, produced by the temperature change r, are the sarne as the displacements produced by the body forces
aE ar, X= - _____ 1 - 2v íJx
423
(f)
proportional at each point to the temperature change at that point. Thus, the total stress produced by nonuniform heating is obtained by
y =~ar, 1 - 2v íJy
Z=~ar 1 - 2v íJz
(h)
and that the pressure (f) should be applied to the surface. We now assume that the elements are joined together, and remove the forces (h) and the surface pressure (f). Then the thermal stresses are evidently obtained by superposing on the pressures (f) the stresses which are produced in the elastic body by the body forces X = _
~ar, 1 - 2v
dX
y = _ ~ar, 1 - 2v ay
z
=
aE aT - 1 - 2v í)z
and a normal tension on the surface equal to
aEr 1 - 2v These latter stresses satisfy the equations of equilibrium
íJu,, + íJr,,,, + íJTxz _~ar= O ax ay az 1 - 2v ax auy + ªTxy + ªTyz _ ~ar= Q ay ax az 1 - 2v ay au, + íJTxz + aTyz _ ~ ar = Q í)z ax ay 1 - 2v az
(256)
assume that u, = 'f"x, = 'f"uz = O. We may also regard each clement as free to expand in the z-direction. It will be sufficient, to ensure fitting of the elements, to suppress the expansion in the x and y directions only. This requires
and the boundary conditions
u,n
+ Tzzl + Tyzm
aET 1 - 2v aET 1 - 2v m aET --n = 1 - 2v
aT Y = - 1 - 2vây
(i)
and a normal tensile stress of amount
aET on the curved surface only, as a problem in plane strain (e, = O). This problem is cf the type considered at the end of Art. 16, except that we have to co~vert Eq. (32) from plane stress to plane strain by replacing 11 by 11 /1 - 11. Thus mstead of Eqs. (31) and (32) we shall have
aET
1-Tv =
and éJ4.p é);c4
éJ4.p
â 2 >
(k)
ax2'
éJ4.p _ _
-5'!!}__
-
1 - "
+ 2 é);c2é)y2 + é)y4
(â 2T + â T) 2
é);c2
q'll :::::::
aET
-
(m)
--•
1 - "
X=~aT,
aE
aT X= - 1 - 2vâx'
=
Substituting these in the equations of equilibrium (18), page 22, we find that the required body forces are
Plane strain will occur in a long cylindrical or prismatic body wh~n the temperature although varying over a cross section, does not vary along lmes parallel to the ~xis of the cylinder or prism (the z-axis). Then Tis indepen~ent of z. Beginning again with the stresses (g) which result in zero stram, the necess~ry body forces are given by (h) where now Z =O, and the pressure (f) must be apphed to the surface, including the ends. Then supposing the elements joined together, we remove the body forces, and the surface pressure on the curved surface only, keeping the axial strai~ e. zero. The effects of this removal are obtained by solving the problem of applymg body
aE
U::e
(257)
together with the compatibility conditions discussed in Art. 7!. The solution of these equations, superposed on the pressure (f), g1ves the thermal stresses in a body undergoing temperature change.
force
425
THERMAL STRESS
THEORY OF ELASTICITY
424
(l)
é)y2
The required stress function is that which satisfies Eq. (l) and gives the normal boundary tension (j). The stresses are then calculated from (k). On these we have to superpose the stresses (g). • The axial stress u, will consist of the term from (g) together w1th 11(
1 -
(n)
"é):.
and the normal pressure aET / (1 - 11) should be applied to the edges of the piate. Removing these forces, we conclude that the thermal stress consists of (m) together with the plane stress due to body forces
X= -~aT, 1 - 11 âx
y = _
aE aT 1 - "ây
(o)
and to normal tensile stress aET /(l - 11) applied round the edges. The determination of this plane stress again presents us with a problem of the type considered in Art. 16. We have only to put in Eqs. (31) and (32)
V= aET 1 - "
this being the potential corresponding to the forces (o). When the edges are fixed the problem reduces to finding the stress due to the body forces (o). A method for solving this problem for the rectangular plate wa11 explained on page 156. 138. Initial Stresses. The method used above for calculating thermal stresses can be applied in the more general problem of initial stresses. Imagine a body subdivided into small elements and suppose that each of these elements undergoes a certain permanent plastic deformation or change in shape produced by metallographical transformation. Let this deformation be defined by the strain components I
'Yx11 '
I
'Yxz,
(a)
We assume that these strain components are small and are represented by continuous functions of the coordinates. If they also satisfy the compatibility conditions (129), the elements into which the body is subdivided fit each other after the permanent set (a) and there will be no initial stresses produced. Let us consider now a general case when the strain components (a) do not satisfy the compatibility conditions so that the elements into which the body is subdivided do not fit each other after permanent set, and forces must be applied to the surface of these elements in order to make them satisfy the compatibility equations. Assuming that after the permanent set (a) the material remains perfectly elastic, and applying Hooke's Iaw, we find from Eqs. (11) and (6) that the permanent set (a) can be eliminated by applying to each elcment the surface forces -(Xe'
+ 2Gez'),
... ,
where
e' = e.:/
+ e./ + e/
(b)
426
THEORY OF ELASTICITY
THERMAL STRESS
The surface forces (b) can be induced by applying certain body and surface forces to the body formed by the small elements. These forces must satisfy the equations of equilibrium ,(127) and the boundary conditions (128). Substituting the stress components (b) in these equations we find that the necessary body forces are
ºx
X = 0 (M'
+ 2Ge.') + :y CG-rxv') + :z (G'Yxz') (e)
and the surface forces are
(d) By applying the body forces (e) and the surface forces (d) we remove the initial permanent set (a) so that the elements fit one another and forma continuous body. We now assume that the elements into which the body was subdivided are joined together and remove the forces (e) and (d). Then evidently the initial stresses are obtained by superposing on the stresses (b) the stresses which are produced in the elastic body by the body forces
(e)
and the surface forces
X
= (M'
+ 2Ge.')l +
Thus the problem of determining the initial stresses is reduced to the usual system of equations of the theory of elasticity in which the magnitudes of the fictitious body and surface forces are completely determined if the permanent set (a) is given. ln the particular case when •.' = •/ = e.' = aT and 'Yxv' = 'Yxz' = 'Yu•' = O, the above equations coincide with those obtained before in calculating thermal stresses. Let us consider now the reversed problem when the initial stresses are known and it is desired to determine the permanent set (a) which produces these stresses. ln the case of transparent materials, such as glass, the initial stresses can be investigated by the photoelastic method (Chap. 5). ln other cases these stresses can be determined by cutting the body into small elements and measuring the strains which occur as the result of freeing these elements from surface forces representing initial stresses in the uncut body. From the previous discussion it is clear that the initial deformation produces initial stresses only if the strain components (a) do not satisfy the compatibility equations; otherwise these strains ma)' exist without producing initial stresses. From this it follows that knowledge of the initial stresses is not sufficient for determining the strain components (a). If a solution for these components is obtained, any permanent strain system satisfying
427
the ~o.~patibility equations can be superposed on this solution without affecting the m1tial stresses.' . lnitia! st:esses, producing doubly-refracting properties in glass, present great ~1fficult1es m m~nufacturing optical instruments. To diminish these stresses it is the ~sual practice to anneal the glass. The elastic limit of glass at high temperatures is v~ry lo:", ai:d t~e material yields under the action of the initial stresses. If a su~c1ent t~me is g1ven, the yielding of the material at a high temperature result~ m a cons1derabl~ release fro_m init~al stresses. Annealing has an analogous effect l"?" the case of vanous metalhc castmgs and forgings. Cuttmg of large bodies into smaller pieces releases initial stresses along the surfaces of cutting an~ diminishes the total amount of strain energy due to initial stresses, but. the magmtude of the maximum initial stress is not always diminished by such c_uttmg.. ~or example, suppose a circular ring (Fig. 228) has initial stresses sy mme_tncally d1stnbuted with respect to the center and the initial stress component 1
+
139. Two-dimensional Problems with Steady Heat Flow. In ste~dy he~t flow par_all~l to the xy-plane, as in a thin plate or in a long cylmder with no vanat10n of temperature in the axial (z-)direction the temperature T will fü\tisfy the equation '
a2T ax2
a2T
+ ay2
=O
(a)
Consider a cylinder (not necessarily circular) in a state of plane strain, with Ez = 'Yxz = "fyz = O. The stress-strain relations in Carte1 The f~c_t _that perma:ient. set (a) i~ not completely determined by the magnÍ· tudes of imtial stresses is d1scussed m detail in the paper by H. Reissner; sef
Z. angew. Math. Mech., vol. 11, p. 1, 1931.
2 . Several examples of the calculation of initial stresses in portions cut out from a circular plate are '.1iscussed in the paper by M. V. Laue, Z. tech. Physik, vol. 11, f~ 385, 19~0. Var.1ous methods of calculating residual stresses in cold-drawn bes are d1scussed m the paper by N. Dawidenkow Z. Metallkunde vol 24 p 25 1932. ' ' . ' . •
INSTITUTUL
poumilo \
Tlr,,-:i~O.\RA -~· ·---· _,,... ......... ~
t
1
428
THERMAL STRESS
Sian coordinates are analogous to Eqs. (a) and (b) oí Art. 135 in the case of plane strain. Corresponding to Eqs. (b) we shall have
the outside it will tend to uncurl, and the slit will open up. There will be a discontinuity of displacement between the two faces of the slit. Thus the displacement should be represented by discontinuous functions of IJ. The cross section is solid, i.e., singly connected, and Eqs. (d) give the stress correctly for plane strain. But if the tube has no slit (Fig. 229a) discontinuities of displacement are physically impossible. This indicates that the assumed temperature distribution will in fact give rise to stress components u.,, u11 , rey, representing the stress produced by suitably drawing together again the separated faces of the
E., -
Ey -
(1
+ v)aT =
(1
1 -r + v)aT = -
1 - p2 ( u., -Ep2 (
u11
-
1 1
_JI _p
11
u11) (b) )
11
u.,
We now inquire whether it is possible to have u.,, u11 and Putting u., = u11 = O in Eqs. (b) we find E.,
=
(1
+ v)aT,
e11 = (1
+
v)aT
rey
zero. (e)
and of course "fey = O. Such strain components are possible only if they satisíy the conditions of compatibility (129). Since e, = O and the other components of strain are independent of z, all of these conditions except the first are satisfied. The first reduces to Õ
(a)
(b)
+ Õ E11 = Ü 2
2 E.,
ày 2
Frn. 229.
àx 2
slit tube and joining them (cf. Art. 39 and Fig. 82).
But on account of Eqs. (e) and (a), this equation also is satisfied. We find, therefore, that in steady heat fl.ow the equations oí equilibrium, the boundary condition that the curved surface be free from force, and the compatibility conditions, are all satisfied by taking
=
=
Tey
=
Ü1
= -aET
(d)
For a solid cylinder the above equations and conditions are complete, and we can conclude that in a steady state of two-dimensional heat conduction there is no thermal stress except the axial stress u, given by Eqs. (d), required to maintain the plane strain condition e, = O. ln the case of a long cylinder with unrestrained ends we obtain an approximate solution valid except near the ends by superposing simple tension, or compression, and simple bending so as to reduce the resultant force and couple on the ends, due to u,, to zero. For a hollow cylinder, however, we cannot conclude that the plane strain problem is solved by Eqs. (d). It is necessary to examine the corresponding displacements. It is quite possible that they will prove to have discontinuities, analogous to those discussed on pages 68 and 120. For instance suppose the cylinder is a tube and a longitudinal slit is cut, as indicated in Fig. 229b. If it is hotter on the inside than on
'
l
The component
u, will also be affected by this operation. To investigate this question further, we rewrite Eqs. (e) as ÔU = .' ÔX
b
429
THEORY OF ELASTICITY
where e' = (1
+ v)aT.
ÔIJ
= .'
ôy
'
(e)
Since 'Yzu = O we can write (f)
and ÔV _
ÔU
ÔX
ôy
= 2..,.
"'• being a component of rotation (see page 225). ôu ôy =
1
-w.,
(g)
Equations (f) and (g) yield
ÔIJ ÔX
= "'•
(h)
and these with (e) give (i)
Equations (i) are Cauchy-Riemann equations, discussed in Art. (55). They show that e' + iw, is an analytic function of the complex variable x + iy. Denoting this function by Z, we have Z =e'+ iw, (j)
THERMAL STRESS
lf u 1, v1, u 2, v2 are the values of u, v at two points 1, 2 in the cross section of the cylinder the difierences u2 - u,, v2 - v, can be expressed as
Consider, for example, a hollow circular cylinder of outer radius b with a concentric bore of radius a. If the temperature T, at the inner surfac~ is uniform ª1:1d ~he temperature at the outer surface is zero, the temperature T at any radiu~ r IS given by Eq. (n) of Art. 135. We may write this as
V2 -
tl1
Íi
=
i
2
(Ôfl iJv dy ) -ax dx + ay
where the integrais are taken along any curve joining the two points and lying entirely in the material. Multiplying the second by i and adding to the first we find U2 -
U1
. + i(t12
-
t11)
[2 [ºu iJX dx
= }i
. (º" iJv )] + iJu iJy dy + i iJX dz + iJy dy
(k)
side is the sarne as /
2 1
(<'
u2 -
u1
+ i(v2
Z =
Thus Eq. (k) becomes
2
-
(l)
have (w.h - (w.)1 =
Ji
2
e:
Writing B for (1
Ji
2 (-
~ '
\11
1
1
+ iw, = (1 + 11)aT + iw, + 11)aA log r + iw, = (1 + 11)aA log z
+ v)aA, we have from Eq. u1
+ i(t12 -
t11)
= B
Ji
2
(l)
1og z dz = B [ z(log z - 1) ]~
(o)
rial. It gives the rela.tive displacement of the two points when the temperature is given by Eq. (m) and the stress by Eqs. (d). Applying it to a circular path of radius r starting at 1 (Fig. 229), going round the hole, and ending at 2, we have, since 81 = O, 82 = 2,,., [z(log z - l}U = rei2r (log r
~~ dx + ~: dy)
The relation between thermal stress in steady heat flow and dislocational stress was established by N. Muschelisvili, B ull. Elec. Tech. Inst., St. Petersburg, vol. 13, p. 23, 1916, and independently by M. A. Biot, Phil. Mag., series 7, vol. 19, p. 540, 1935. Thermal stresses in a hollow circular cylinder, and in a square cylinder with a circular hole, have been determined photoelastically by this method by E. E. Weibel, Proc. Fifth Intern. Cong. Applied Mechanics, Cambridge, Ma.ss., 1938, p. 213.
(n)
T; log (b/a)
1:'hls equa~ion applies t~ any curve between points 1 and 2 lying wholly in the mate-
+ ~~· dy)
Since .' is proportional to T, this integral is proportional to the amount of heat flowing per unit time, per unit axial distance, across the curve joining points 1 and 2. lf this is a closed curve (w.h - (w,)1 must vanish, and therefore the total heat flow across the curve must be zero. lf a pipe has heat flow from inside to outside or vice versa this condition is not fulfilled and the stress is not correctly given by Eqs. (d). But if the pipe is slit, as indicated in Fig. 229b, the displacement or rotation at point 2 can difier from that at point 1, for instance if the heating causes the slit to open up. The simple state of stress given by Eqs. (d) is then correct. To arrive at the state of stress which exists in the pipe when it is not slit, we have to superpose 1 the stress due to closing the gap. The determination of this dislocational stress involves problema of the types illustrated by Figs. 45 and 82. 1
A= -
+ i · 2,,.)
- reio (log r
+ i. O)
= i. 2,,.r
Inserting this in Eq. (o) we find
and using Eqs. (i) this becomes {w.h - (w,)1 =
dx
E
= (1
U2 -
The displacements will be single-valued when this integral vanishes for a complete circuit of any closed curve, such as the broken-line circle in Fig. 229a, which can be drawn in the material of the cross section. We shall use this result later in solving a thermal-stress problem of the hollow circular cylinder. Not only the displacements, but also the rotation w, must be continuous. We
(m)
The constant term -A log b in Eq. (m) can be ignored, since a uniform change of temperature does not cause thermal stress. Then, since log z = log r + i8,
2
Ji Z dz. v,) = Ji Z dz
+ iw,)(dx + i dy) or
T = -A log b +A log r where
and it is easily verified from Eqs. (e) and (h) that the integral on the right-hand
:1,
431
THEORY OF ELASTICITY
430
U2 -
'U1
= 0,
V2 -
V1=B.211"T
(p)
Th~
relative ~placement is not zero, and therefore it is necessary to consider the c:ylmder as slit, so th~t the point 2 can move away from the point 1 by the vertical dISplacement 211"TB (Fig. 229b). The movement of the upper face of the slit relative to the lower face is equivalent to a rotation 2,,.B in the clockwise sense about the center of the cylinder. However, B is negative if T, is positive. Then the slit opens to a gap subtending an angle -2,,.B at the center. The problem of closing such a gap was solved on page 69 for the case of plane stress. The solution can be converted to plane strain by the substitutions given on page 34. The stress com~one~ts which result, combined with the axial stressª" = -aET from Eqs. (d), are 1dent1cal with those given by Eqs. (247). Inside and outside temperatures which vary round the boundary circles can be represented by Fourier series
T, = Ao +A, cos 8 + Az cos 28 + · To= Ao'+ Ai' cos 8+ Az' cos 28 +
· · + B, sin 8 + B sin 28 + ... · · · + Bi' sin 8 + B sin 28 + . . 2
2'
(q)
The t~ermal stress due to the several terms can be treated separately that due to theuniform.terms Ao, Ao'being covered bythepreceding case, with T; '= Ao - Ao'. Correspondmg to the terms cos 8, sin 8; cos 28, sin 28; etc., the function z will have terms proportional to z, z-1 ; z2, :r2 ; etc. '1')
432
THERMAL STRESS
THEORY OF ELASTIOITY
Now f zn dz taken round a complete circle of radius r vanishes unless n = -1, for we have
J
zn dz = =
J
rnein8re•8i d!J= irn+l
[2.-
irn+i }o
[cos (n
fo
2 ,..
plarie strain, of course. stress of Eqs. (d) are
ei
The stress components to be combined with the axial
1- %) (~ - 1)
=
K
cos IJ • r (
u9
=
K
cos IJ • r ( a- b4 r
Tr9
=
K
. IJ • r ( 1 - T2 a ) sm
2 2
+ l)IJ + i sin (n + l)IJ] d!J
a• + b + --r 2
2
(bTa -
2
This is clearly zero unless n
+1
= O,
in which case we have where
Thus the only term in (r) which will produce a non-zero integral on the right of Eq. (l) is the term z- 1• It follows that the terms in cos 21J, sin 21J, and higher harmonics in the temperature series (q) do not produce any relative displacement of the twc faces of the slit in the slit tube. The net heat flow from inside to outside corresponding to such terms is zero, and the only stress they produce is that given by Eqs. (d). The terms in (q) which give rise to a term in z- 1 in Z are those in cos IJ, sin IJ. It is sufficient to consider cos IJ only, since the effects of the sin IJ terms can be deduced from those of cos IJ by changing the initial line IJ = O. Accordingly we consider only (t) T, = coso, To =Ai' cos IJ
A1
C1
=
~(A1 _Ai') b2 - a 2 a b '
C _ A{b - A1a 2 b2 - a 2
The values aré
1)
The axial stress is given by Eq. (d), with T determined from Eqs. (u) and (v), so long as axial expansion or contraction is prevented. If the ends are free the axial stress due to remova! of the force and couple on each end must be considered. 140. Solutions of the General Equations. Any particular solution we can obtain of Eqs. (254) will reduce the thermal-stress problem to an ordinary problem of surface forces. The solution for u, v, w will lead by means of Eqs. (a) and (b) of Art. 139, using Eqs. (2), to values of the stress components. The surface forces required, together with the nonuniform temperature, to maintain these stresses, are then found from Eqs. (128). The removal of these forces in order to make the boundary free, so that the stress is due entirely to the nonuniform temperature constitutes an ordinary problem of surface loading. ' One way of finding particular solutions of Eqs. (254) is to take
ª"'
U=-•
V
ax
ª"'
= af,
W=()z
ay
(a)
where f is a function of x, y, z, and also of time t if the temperature varies with time. Using Eqs. (5) and (10) we can write Eqs. (254) in the form ~ -ax + (1
(v)
The term Ci/z in (u) corresponds to the value
2
2 2
" = 2(1 - v)
The problem of determining the steady temperature distribution corresponding to these boundary values is solved by taking the temperature T as the real part of a function
and determining C1 and C2 so that the conditions (t) are satisfied.
3)
(A1a - Ai') a b T b' - a'
-aE
(s:
433
- 2v)V 2 u = 2(1
+ v)a -aT ax (b)
au . 8mce e = ax
av aw + ay + az' Eqs. (a) lead to e =
V"f, and Eqs. (b) become
(1+ v)a Ci
z
(1 - v)
for the function Z. Inserting this in Eq. (l) and making use of (s), we find that the displacement discontinuity is given by U2 -
U1
+ i(v2 -
V1)
= i • 2,,.(1
+ v)aC1
and therefore U2 -
U1
= 0,
This means the top face of the slit in Fig. 229 moves down by 2,,.(1 + v)aC1 into. the space occupied by the bottom face and material below it. Physically this is impossible, of course, and is prevented by forces between the faces sufficient to create a counteracting displacement. The stress set up by this counteracting displacement is determined as explained at the end of Art. 39, in the present case for
~ª V2f ux
= (1
+ v)a aT ax (e)
a/ay and a/az replacing a/ax in the second and third of these equations. All three equations are evidently satisfied if we take the function 1 f as a solution of the equation V2f 1
= l
+ v aT
1 -
(d)
V
Functions of this kind were used by E. Almansi in the problem of the sphere. See (1) Atti reale accad. sei. Torino, vol. 32, p. 963, 1896-1897; (2) Mem. reale accad. sei. Torino, series 2, vol. 47, 1897.
434
THEORY OF ELASTICITY
THERMAL STRESS
Solutions of equations of this type are considered in the theory of potential. t A solution can be written down as the gravitational potential of a distribution of matter of density - (1 + v)aT /4,..(1 - v), which is 2 ,..(l+ _v)a >/t = - 4(1 v)
!!!
1 d~ d11 dr T (~,11.r)?
ox 2
+ õ">/t
õy 2
+"
>/t = 1,.. · 1 _ ,, · a
1
= l
+ " aT
(J)
1 - "
JJ
where
r' = l(x - ~)2
T(~,11)
+ (y -
log r' d~ d11
(g)
11 )2]i
u. = -E-2 [ºu 1 - v ox
éJv + v -oy
(1
+ v)aT ]
E [º"-+v--(l+v)aT ou ] u,=-1 - 11 2 éJy ox 2(1
+
ov) é>y
2 1 - 11 (éJ u + 1 + li éJx2 +
o'u) ay2
=
oT
2a é>x
These are satisfied by
U=~
ox
provided that ..y is a solution of
ax ª""'2
+
º""
é>y' = (1+11)aT
ou) + 11) (º" oX + éJy .
E
1 See, for instance, "Theory of the Potential," by W. D. MacMillan, New York, 1930. 2 This potential function was used by C. W. Borchardt in the problem of the sphere. See Monatsber. kõnigl. Preuss. Akad. Wiss., Berlin, 1873, p. 9. 3 J. N. Goodier, Phil. Mag., vol. 23, p. 1017, 1937. The semi-infinite solid is considered by R. D. Mindlin and D. H. Cheng, J. Applied Phys., vol. 21, pp. 926, 931, 1950. 4 N. O. Myklestad, J. Applied Mechanics (Trans. A.S.M.E.), 1042, p. A-131.
éJ>/t é)y
{J)
r--a---.--e
(k)
fb fª-a2"log 1 [(x -b
~)2 + (y - 11)2] d~d11
(l)
The displacements are obtained by differentiation according to (j) and then the stress con:ponents can be found from (h). The results for u. and "•• at points such as P outs1de the hot rectangle can be reduced to
u•-- E 1 (·'· ·'· ) a T -,..1-,..2, 2,..
(h)
V=
Comparing with Eq. (fJ we see that a particular solution is given by the logarithmic potential (g) with the factor 1 - .,, in the denominator omitted. This gives the complete solution for local heating in an infinite plate, where the stress and deformation must tend to zero at infinity. As a first example of this kind we conp sider an infinite plate at temperature zero except for a rectangular region ABCD of Frn. 230. sides 2a, 2b (Fig. 230) within which the temperature is T and uniform 1 The required logarithmic potential is -' · 1 (1+11)aT >/t = 211"
For a thin plate, with no variation of T through the thickness, we may assume plane stress, with u, = -r., = "•• = O, and u, v, u., u,, "•• independent of z. We have then the stress-strain relations [cf. Eqs. (d) of Art. 134]
"• 11 =
.!!.. (ºu é>x ox
(i)
A particular solution is given by the logarithmic potential
2
Substituting these in the two equations of equilibrium (18) (with zero body force) we find the equations
(e)
where T(~,11,r) is the temperature at a typical point ~. 11, r at which there is anelement of volume d~ d11 dr, and r' is the distance between this point and the point x, y, z. Equation (e) gives the complete solution of the thermal-stress problem of an infinite solid at temperature zero except for a heated (or cooled) region. 3 The cases of such a region in the form of an ellipsoid of revolution and a semi-infinite circular cylinder, uniformly hot, have been worked out. 4 For the ellipsoid the maximum stress which can occur is aET /1 - "• and is normal to the surface of the ellipsoid at the points of sharpest curvature of the generating ellipse. This value occurs only for the two extreme cases of a very flat or very elongated ellipsoid of revolution. Intermediate cases have smaller maximum stress. For a spherical region the value is two-thirds as great. When Tis independent of z, and w = O, we shall have plane strain, with ,Y, u, and v independent of z. Equation (d) becomes
o">/t
435
. T. 11 = E a T 1o 1 g rira 411" T2T4
(m)
the angles "11, "12 and the distances r1, r2, ra, r, being those indicated in Fig. 230. The angles are tho.se subtended at P by the two sides AD, BC of the rectangle paral~el t? the x-ax1s. The expression for"" is obtained from the first of Eqs. (m) by usmg mstead of "11 and >/t2 the angles subtended at P by the other two sides AB DC of the rectangle. ' The value of u. just below AD and just to the left of A is
Ea.T · 21,.. ( .... - arctan ~)
~nd is greatest for a rectangle infinitely long in the y-direction, when it becomes vEaT. Both normal stress components change sharply on turning a comer of the 1 Goodier, Zoe. cit,
436
THEORY OF ELASTICITY
THERMAL STRESS
rectangle. The shear stress rz11 approaches infinity as a comer is approached. These peculiarities are, of course, a consequence of the ideally sharp corners of the heated rectangle. If the heated areais elliptical1 instead of rectangular, the ellipse being
where the real or imaginary parts of einB may be taken to bt · · F E ( ) . o am cos no or sm rom q. o the funct1on f corresponding to this temperature term will be l +" 1 if;... = - 1 - " a.K 82 T ...
Ea.T
u
+ (b/a)
which approaches Ea.T for a very slender ellipse. If the heated areais circular it becomes iEaT. The stress
+ (a/b)
1
and approaches zero for a very slender ellipse. The method of the present article becomes particularly simple when the temperature varies with time, and satisfies the differential equation of heat conduction2 -
éJt
=
T
(n)
where K is the thermal conductivity divided by the specific heat and by the density. Differentiating Eq. (d) with respect to t and then substituting for éJT /éJt from Eq. (n) we find that the function f must satisfy the equation
ª"'éJt
V2 -
+"
2T = --a.KV l
1- "
We may therefore take
The integral of this which is appropriate for a temperature which approaches zero as time goes on is
h.
1 +" if;=---a.K 1 - " t
Tdt
(o)
as may be verified by substitution in Eq. (d), making use of Eq. (n). Consider for instance a long circular cylinder (plane strain) which is cooling or being heated toward a steady state of heat conduction. The temperature is not symmetrical about the axis, but is independent of the axial coordinate z. The temperature is then representable by a series of terms of the form (p) 1 2
I
L 1
,L::
Goodier, Zoe. cit. See, for instance, "Theory of Heat Conduction," by Ingersoll and Zobel.
(q)
ª"'éJr
=-1
ª"'ao
V =1- -
r
The strain components follow from the resulta of Art. 28, page 65. The stress components can then be found from the plane strain formulas (a) and (b) of Art 135, together with the last of Eqs. 52, page 66 for the shear stress · Wh h 1 • ' Tr9. en sue a so ut1on has been obtained it will, in general, be (ound that it gives non-zero bound~ry forces (crr, rr8) on the curved surface of the cylinder. The e~ects of removmg these are found by solving an ordinary plane strain problem usmg the general stress function in polar coordinates given in Art. 39.1 ' This proble~ is worked out for a hollow cylinder, with temperature correspond mg to Eq. (p), m the paper by J. N. Goodier cited above. •
K'\7 2
•
u and v here being the radial and tangential components The · 1 is zero in vlane strain. • axia component w
1
aT
no
A ser!es of such terms, corresponding to the series for T, will represent a particular solut10~ of the general eq~ations (b). The displacements may be calculated accordmg to Eqs. {a), or their polar equivalents,
the value of the stress u11 just outside the ellipse, near an end of the major axis, is 1
437
PROPAGATION OF WAVES IN ELASTIC SOLID MEDIA
439
two adja.cent cross sections mn and m 1n 1, the difference of forces acting on the s1des mn and m 1n 1 is AE
CHAPTER 15 THE PROPAGATION OF WAVES IN ELASTIC SOLID MEDIA 141. ln the preceding chapters it was usually assumed that the elastic body was at rest under the action of external forces, and the resulting problems were problems of statics. There are cases, however, in which motion produced in an elastic body by suddenly applied forces or by variable forces should be considered. The action of a suddenly applied force is not transmitted at once to ali parts of the body. At the beginning the remote portions of the body remain undisturbed, and deformations produced by theforce are propagated through the body in the forro of elastic waves. If the dimensions of the body are large, the time taken by the waves to traverse the body becomes of practical importance and should be considered. We have such problems, for instance, in discussing the effect of impact or waves produced by earthquakes. The investigation of the propagation of waves in an elastic medium is the subject of the following discussion. 1 We begin with the simple problem of the propagam m1 tion of longitudinal waves in a long ~ prismatical bar. " ~ x 142. Longitudinal Waves in Prisn n1 matical Bars. Taking the axis of the Frn. 231. bar for x-axis (Fig. 231) and assuming that cross sections of the bar remain plane during deformation, the unit elongation at any cross section mn, due to a longitudinal displacement u, is equal to ôu/ôx and the corresponding tensile force in the bar is AE(ôu/ôx), where A is the cross-sectional area. 2 Considering an element of the bar between the 1 Probleros of steady vibrations of elastic bodies (standing waves) have been discussed in S. Tirooshenko, "Vibration Probleros in Engineering," N ew York, 1928. 2 It is assuroed that we have here a sirople tension in the x-direction and the elongation au/ax is accoropanied by lateral contraction of the aroount v(au/ax). lnertia forces corresponding to rootion of particles in lateral direction are neglected in our derivation. This approxiroate solution is accurate enough so long as the length of waves is not sroall in coroparison with cross-sectional diroensions of the bar. ln the case of short waves, rootion of particles in the direction perpendicular ·· to the axis of the bar should be considered. See Lord Rayleigh, "Theory o{
438
(~xu + ~2~ dx) u ux
- AE ôu = AE a2u dx ax ax2
and the equation of motion of the element is ô 2u
ô2u
Ap dx ôt 2 = AE ôx 2 ax
or (258) n
,f=l=3--F-l~
X
(b) Frn. 232.
in which
p
is the mass per unit volume of the bar and
e=~
(259)
. It can be shown by substitution that any functionf(x + ct) is a solution of Eq. (258). Any function fi(x - ct) is also a solution and the general solution of the Eq. (258) can be represented in the fo;m u = f(x
+ ct) + fi(x
- ct)
(260)
~~~ solution ~as a .very simple physical interpretation, which can t Y be explamed m the following manner. Consider the second . er~ on the right side of Eq. (260). For a definite instant t this term is a unction of x only and can be represented by a certain ~urve such ~tnp C!ig. 232a), the shape of which depends on the function f er an mterval of time dt, the argument of the function f i becom:~ Sound " Ch
Q uart ' J M ap. th
p. 7o3 Í942~ 1948, ' '
7 L ; . Pochharoroer, J. Math. (Crelle's J) vol 81 1876 C Ch 1 · ' · ' ; . ree ~.D21 '.1886, and vol. 24, 1890; J. Prescott, Phil. Mag., vol. 33: ' ' . ll.vies, Trans. Roy. Soe. (London), series A, vol. 240, p. 375,
il'
440
THEORY OF ELASTICITY
PROPAGATION OF WAVES IN ELASTIC SOLID MEDIA
x - c(t !l.t). The function f 1 will remain unchanged provided that simultaneously with the increase of t by Ât the abscissas are increased by an amount ÂX equal to e At. This means that the curve mnp, constructed for the moment t, can also be used for the instant t + At, if it is displaced in the x-direction by the distance Ax = e At, as shown by the dotted line in the figure. From this consideration it can be seen that the second term of the solution (260) represents a wave traveling in the direction of the x-axis with a constant speed e. In the sarne manner it can be shown that the first term of the solution (260) represents a wave traveling in the opposite direction. Thus the general solution (260) represents two waves traveling along the x-axis in two opposite directions with the constant velocity e given by Eq. (259). This velocity depends only on the modulus E and the density of the material of the bar. In the case of steel, for instance, we can assume e = 16,850 ft. per second. The functions f and f 1 should be determined in each particular case from the initial conditions at the instant t = O. For this instant we have, from Eq. (260), (u)t=o = f(x) f1(x) (a) = c[f'(x) - f1'(x)]
will be compressed and the remaining portion will be at rest in an unstressed condition. The velocity of wave propagation e should be distinguished from the velocity v, given to the particles in the compressed zone of the bar by the compressive forces. The velocity of the particles v can be found by taking into account the fact that the compressed zone (shaded in the figure) shortens due to compressive stress u by the amount (u/E)ct. Hence the velocity of the left end of the bar, equal to the velocity of particles in the compressed zone, is
+
(b)
The velocity e of wave propagation can be found by applying the equation of momentum. At the beginning the shaded portion of the bar was at rest. After the elapse of the time t it has velocity v and momentum Actpv. Putting this equal to the impulse of the compressive force, we find Aut = Actpv (e)
+
Using Eq. (b), we find for e the value given by Eq. (259)1 and for the velocity of particles we find
(~7) 1=0
v=--
VEP
Assume, for instance, that the initial velocity along the length of the rod is zero and there is an initial displacement given by the equation
(261)
It will be seen that, while e is independent of the compressive force the velocity v of particles is proportional to the stress u. ' If, instead of compression, a tensile force is suddenly applied at the end of the bar, a tension is propagated along the bar with the velocity e. The velocity of particles again is given by Eq. (261). But the direction of this velocity will be opposite to the direction of the x-axis. Thus in a compressive wave the velocity v of particles is in the sarne direction as the velocity of wave propagation, but in a tension wave the velocity v is in the opposite direction from that of the wave. From Eqs. (259) and (261) we have
(u)1=0 = F(x) Conditions (a) are satisfied by taking
f(x) = f1(x) = iF(x) Thus in this case the initial displacement will be split into halves which will be propagated as waves in .two opposite directions (Fig. 232b). The velocity of propagation of tT waves in prismatical bars can be ª--t+lll-Hll+rllttil l l+rllttil 11+-I_ __,_{ x obtained from elementary considerations. Assume that a uniformly ct distributed compressive stress is 233 · FIG. suddenly applied to the left end of a prismatical bar (Fig. 233). It will produce at the first instant a-uniform compression of an infinitely thin layer at the end ofthe bar. This compression will be transmitted to the adjacent layer, and so on. A wave of compression begins to travel along the bar with a certaill velocity e, and, after a time interval t, a portion of the bar of length ct
441
i-- ---J
u =
E~
(262)
The stress in the wave is thus determined by the ratio of the two v:el.ocities and by the modulus E of the material. If an absolutely rigid body, moving with a velocity v, strikes longitudinally a prisThis elem~ntary derivation of the formula for the velocity of wave propagation due _to Babmet; see Clebsch, Théorie de l'élasticité des corps solides, traduite par Samt-Venant, p. 480d, 1883. 1
•
18
PROPAGATION OF WAVES IN ELASTIC SOLID MEDIA
THEORY OF ELASTICITY
442
matical bar, the compressive stress on the surface of contact at the first instant is given by Eq. (262). 1 If the velocity v of the body is above a certain limit, depending on the mechanical properties of the material of the bar, a permanent set will be produced in the bar although the mass of the striking body may be very small. 2 Consider now the energy of the wave shown shaded in Fig. 233. This energy consists of two parts: strain energy of deformation equal to
Assume that a compression wave is moving along the bar in the x-direction and a tension wave of the sarne length and with the sarne magnitude of stress is moving in the opposite direction (Fig. 235). When the waves come together, tension and compression annul each m
1
·i..-i-1
Actu 2 2E and kinetic energy equal to
Actpv
2
-2-
Actu
u{l 11111111
It will be seen that the total energy of the wave, equal to the work done by the compressive force Acr acting over the distance (cr /E) ct, is half potential and half kinetic. e ~
o;{l 11111111
111liIBi111 l}o-~
~mmrl~~L.J..Ll..JI11IL.J. Ll. JI11
llL..-}uz-
llll 1111·11111
x
(a)
fb)
~
J 11111111
; k
1rl
I: 1
1
X
(C)
1
1
'
'
'~Ut · ( 1: ,,I• L
This conclusion is due to Thomas Young; see his "Course of Lectures on Natural Philosophy . . . ," vol. 1, pp. 135 and 144, 1807. 2 It is assumed that contact occurs simultaneously at all points of the end section of the bar.
1 n
11111111 l}cr X
i
1111111 ll}o-
--=-'-cc ....__..
e-
(6)
l
am
J
X
'-·------.Jln
(e)
FIG. 235.
other, and in the portion of the bar in which the two waves are superposed we have zero stress. At the sarne time the velocity of particles in this portion of the bar is doubled and equal to 2v. After passing, the waves return to their initial shape, as shown in Fig. 235b. At the mid~le cr~ss section mn there will be at all times zero stress and we may cons1der it as a free end of a bar (Fig. 235c). Then comparing Figs. 235a and 235b it can be concluded that in the case of a free end a compressive wave is reflected as a siinilar tension wave, and vice versa.
FIG. 234.
Equation (258), governing the wave propagation, is linear, so that, if we have two solutions of the equation, their sum will also be a solution of this equation. From this it follows that in discussing waves traveling along a bar we may use the method of superposition. If two waves traveling in opposite directions (Fig. 234) come together, the resulting stress and the resulting velocity of particles are obtained by superrosition. If both waves are, for instance, compressive waves, the resultant compression is obtained by simple addition, as shown in Fig. 234b, and the resultant velocity of particles by subtraction. After passing, the waves return to their initial shape, as shown in Fig. 234c. 1
(a}
. 1
1
2
= ""2E
443
~.
u-011111111
rn
.....S
11111111 l}u
1
~t-+-a=l=a ,j, .....Q__
111111111
}
n.
t-l
-
x
(a}
e
l
111111111
1
fh) X
(e)
FIG. 236.
If two identical waves, moving toward one another (Fig. 236a), come together, there will be doubled stress and zero velocity in the portion of t?e bar in which the waves are superposed. At the middle cross sect1on mn we always have zero velocity. This section remains immova-
445
THEORY OF ELASTICITY
PROPAGATION OF WAVES IN ELASTIC SOLID MEDIA
ble during passage of the waves and we may consider it as a fixed end of the bar (Fig. 236c). Then, from comparison of Figs. 236a and 236b, it can be concluded that a wave is reflected from a fixed end entirely unchanged. Up to now we have considered waves produced by constant forces. The stress u and the velocity of particles v were constant along the length of the wave. In the case of a variable force, a wave will be produced in which u and v vary along the length. Conclusions obtained before regarding propagation, superposition, and reflection of waves can be applied also in this more general case. 143. Longitudinal Impact of Bars. If two equal rods of the sarne material strike each other longitudinally with the sarne velocity v (Fig. 237a), the plane of contact mn will·not move during the impact1 and
tions of the moving bars are equal and are directed in each bar away from the surface of contact. The magnitude of these velocities must be equal to (v1 - v2)/2 in order to have the absoluta velocities of particles of the two bars at the surface of contact equal. After an interval of time equal to l/c, the compression waves arrive at the free ends of the bars. Both bars are at this instant in a state of uniform compression, and the absolute velocities of all particles of the bars are
444
1
?
1
X
(Ó)
The compression waves will then be reflected from the free ends as tension waves and at the instant t = 2l/c, when these waves arrive at the surface of contact of the two bars, the velocities of bars 1 and ~ become V1 + V2 + V1 - V2 = Vi 2 2 Thus the bars, during impact, exchange their velocities.
l---1--I Frn. 237.
two identical compression waves start to travei along both bars with equal velocities e. The velocities of particles in the waves, superposed on the initial velocities of the bars, bring the zones of waves to rest, and at the instant when the waves reach the free ends of the bars (t = l/c), both bars will be uniformly compressed and at rest. Then the compression waves will be reflected from the free ends as tension waves which will travei back toward the cross section of contact mn. In these waves the velocities of particles, equal to v, will now be in the direction away from mn, and when the waves reach the plane of contact the bars separate with a velocity equal to their initial velocity v. The duration of impact in this case is evidently equal to 2Z/c and the compressive stress, from Eq. (261), is equal to v v'1fP. . Consider now a more general case when the bars 1 and 2 (Fig. 237b) are moving2 with the velocities v1 and v2(v1 > v2). At the instant of impact two identical compression waves start to travel along both bars. The corresponding velocities of particles relative to the unstressed por1 It is assumed that contact takes place at the sarne instant over the whole l!lurface of the ends of the rods. . 2 Velocities are considered positive if they are in the direction of the x-aXJS.
-
v,, -..m
z l-t.1
1
º2
.~
--=pn n
1xraJ
'2
1
l2
1111111111111111
1·
l
(ó)
21.1--I FIG. 238.
If the above bars have different lengths, Z1 and Z2 (Fig. 238a), the conditions of impact at first will be the sarne as in the previous case. But after a time interval 2Zi/c, when the reflected wave of the shorter bar 1 arrives at the surface of contact mn, it is propagated through the surface of contact along the longer bar and the conditions will be as shown in Fig. 238b. The tension wave of the bar li annuls the pressure hetween the bars, but they remain in contact until the compression wave in the longer bar (shaded in the figure) returns, after reflection, to the surface of contact (at t = 2Z2/c). ln the case of two bars of equal length, each of them, after rebounding, has the sarne velocity in all points and moves as a rigid body. The total energy is the energy of translatory motion. ln the case of the bars of different lengths, the longer bar, after rebounding, has a
PROPAGATION OF WAVES IN ELASTIC SOLID MEDIA
THEORY OF ELASTICITY
446
traveling wave in it, and in calculating the total energy of the bar the 1 energy of this wave must be considered. Consider now a more complicated problem of a bar with a fixed end struck by a moving mass at the other end 2 (Fig. 239). Let M be the mass of the moving body per unit area of the cross section of the bar and v0 the initial velocity of this body. Considering the body as absolutely rigid the velocity of particles at the end of the bar at the instant of impact (t = O) is vo, and the initial compressive stress, from Eq. (261), is
=
Vo
(a)
VEP
Owing to the resistance of the bar the velocity of the moving body and hence the pressure on the bar will gradually decrease, and we obtain a
___.-1-t--.i~I
1~1--1 ~rn 11111 mn-i_Lo;..::.º--x 'l--ct=-....i
lfü
1111
ftffIDll 111~
(a)
(6)
(e)
FIG. 239.
compression wave with a decreasing compressive stress traveling along the length of the bar (Fig. 239b). The change in compression with the time can easily be found from the equation of motion of the body. Denoting by u the variable compressive stress at the end of the bar and by v the variable velocity of the body, we find
M
~ + u =O
M
257 and 376, 1867. : This problem was discussed by several authors. The final solution was given by J. Boussinesq, Compt. rend., p. 154, 1883. A history of the pr?blem can. be found in "Théorie de l'élasticité des corps solides" Clebsch, tradmte par SamtVenant, see note of par. 60. The problem was also discussed by L. H. Donnell. By using the laws of wave propagation he simplified th~ solution a~d e~t~n?ed it to the case of a conical bar. See Trans. A.S.M.E., Apphed Mechamcs D1VIS1on,
du
v'EPat + cr =o from which (e)
This equation can be used so long as t < 2Z/c. When t = 2l/c the compressive wave with the front pressure uo returns to the end of the bar which is in contact with the moving body. The velocity of the body cannot change suddenly, and hence the wave wi11 be reflected as from a fix~d end and the compressive stress at the surface of contact ~uddenly mcreases by 2cro, as is shown in Fig. 239c. Such a sudden 1~crease of pressure occurs during impact at the end of every interval of time T = 2l/c, and we must obtain a separate expression for u for each one of these intervals. For the first interval, O < t < T, we use Eq. (e). For the second interval, T < t < 2T, we have the conditions represented by ~ig. 239c, and the compressive stress u is produced by two waves movmg away from the end struck and one wave moving toward t~is end. We designate by s 1 (t), s 2 (t), s 3 (t), . . . the total compress1ve s~ress produced a~ the end struck by all waves moving away from. th1s end, after the mtervals of time T, 2T, 3T. . . . The waves commg back toward the end struck are merely the waves sent out during the preceding interval, delayed a time T, due to their travel across the bar and back. Hence the compression produced by these waves at the end struck is obtained by substituting t - T for t in the expres~ion .for the compression produced by waves sent o~t during the precedmg mterval. The general expression for the total compressive stress during any interval nT < t < (n l)T is therefore
+
<1
The question of the lost kinetic energy of translatory motion in the case of longitudinal impact of bars was discussed by Cauchy, Poisson, and finally by Saint-Venant; see Compt. rend., p. 1108, 1866, and J. mathémat. (Liouville), PP·
1930.
or, substituting for v its expression from Eq. (261),
(b)
1
447
= Bn(l)
+ Bn-1(t -
T)
(d)
The velocity of particles at the end struck is obtained as the difference between the velocity dueto the pressure sn(t) of the waves going away and the velocity due to the pressure Bn-i(t - T) of the waves goin~ toward the end. Then, from Eq. (261), 1
V
= ...jEp [sn(t) -
Sn- 1(t
- T)]
(e)
T~e relation between sn(t) and Bn-1(t - T) will now be obtained by usmg the equation of motion (b) of the striking body. Denoting by a
THEORY OF ELASTICITY
448
PROPAGATION OF WAVES IN ELASTIC SOLID MEDIA
the ratio of the mass of the bar to the mass of the striking body, we have
ª=
v'IffP
lp M'
-x;r-
clp _ 2a = Ml - T
frorn which
croT ( 2 ) C=--1+4ae ª 4a
(f)
Substituting in Eq. (k),
Using this, with (d) and (e), Eq. (b) becomes 81
!!:_ [8,.(t) - 8n-1(t - T)] dt
2at
()
T)]
iT, 2 1 ;
=e
2at
4
2at
eT 8n-1(t - T) -
4a
2at
(l) 81,
instead of
2at
; eT 8n-1(t - T)
2at
;t [eT 8,.(t)] = ;t [e T 8n-1(t - T)] - Te
=
t)]
T
d8,._ 1 (t - T) dt
+ 2;
8,.(t)
[ 1 + 4a ( 1 + croe -2a(i-1) T
Proceeding further in the sarne rnanner and substituting into Eq. (g), we find
T
from wbich
80
8n-1,
2at + 2.!!. e'F s,.(t)
or
=
=O
2at
Multiplying by eT d8,. t dt
+ 2Ta [8,.(t) + 8,.-1(t -
449
T
8n-1(t - T)
2at[J ~ 8n-1(t - T) - 4; e-T e T 8n-1(t - T)dt
1
+e
(g)
in which is a constant of integration. This equation will now be u~ed for deriving expressions for the consecutive values 8~, 82. • • • • Durmg the first interval O < t < T the compressive stress 1s given by Eq. (e),
e
and we can put
2at
80
Substituting this for s 1 (t) = croe
8n-1
in Eq. (g),
-2a(~-1) _
=croe
(h)
= croe -T
-2a(~-1)
4a T e
2 ;!
(J
(i - 4at) T
CT 0e2a
dt
+e)
Continuing in the sarne way,
e 4a e-;at
(k)
T
The constant of integration C is found from the conditi~n that ats~~~ instant t = T the compressive stress at the end struck mcreases denly by 2cr 0 (Fig. 239c). Hence, using Eq. (d),
Í croe _2;t] L
+ 2cro t= T
= [croe -2aG-1)
+ croe-2ª(~-1) (i
- 4;t) 4a
-C-e T
_2at] T
1-2'
83
=
82
t)
[ 1 + 2 · 6a ( 3 - T + croe -2a(i-a) T 2
+ 2 · 3 · 4a
2
(3 -
Tt )
2 3 8a + 2~
3
(3 -
Tt
)ª].
(n)
and so on. fo Fig. 240 the functions 8o, 8 1, 8 2, ••.• are represented graphically for cro = 1 and for four different ratios, 1 a = l, 1, }, 1. By using these curves the cornpressive stress cr at the end struck can easily 1 These curves were calculated by Saint-Venant and F1amant. rend. 1 pp. 127, 214, 281, and 353, 1883,
See Compt.
450
THEORY OF ELASTICITY
PROPAGATION OF WAVES IN ELASTIC SOL/D MEDIA
be calculated from Eq. (d). In Fig. 241 this stress is represented graphically for O"o = 1 and for a = -l, -!, 1. It changes at intervals T, 2T, . . . by jumps. The maximum value of this stress depends on the ratio a. For a = -! and a = 1 the stress has its maximum value at t = T. ln the case of a = -!, the maximum stress occurs at t = 2T.
F' · om . ig. 240 it can be concluded at once that for 2 X r:6~6(J':• i'xt~e max1mum compressive stresses are 2 X 1.7520"0, ' ·368<1'o, and 2 X 1.1350"0, respectively. ln Fig. 242
mum value of
a =
i
i
i
451
Fr
8
•
ti'
1
2t
o
6 T
2 FIG. 241.
The instant when <1' becomes equal to zero indicates the end of the impact. It will be seen that the duration of the impact increases when a decreases. Calculations of Saint-Venant give the following values for this duration:
a= 2t
r=
1
1
6
4
~
1
1
7.419
5.900
4.708
3.068
For a very small a the time of contact can be calculated from the elementary formula
fI e \{;;
t = 'lrl
the values of <1' / f · max. O"o or vanous values of the ratio - l/M · 1 For comparison there is 1 h a - P are g1ven. a so s own the lower parabolic curve calculated from the equation <1'
(p)
which is obtained by neglecting the mass of the rod entirely and assuming that the duration of the impact is equal to half the period of simple harmonic oscillation of the body attached to the rod. Functions 81, 82, 83, . . . calculated above can also be used for determining the stresses in any other cross section of the bar. The total stress is always the sum of two values of 8 [Eq. (d)], one value in the resultant wave going toward the fixed end and one in the resultant wave going in the opposite direction. When the portion of the wave corresponding to the maximum value of 8 (the highest peak of one of the curves in Fig. 240) arrives at the fixed end and is refl.ected there, both of the waves mentioned above will have this maximum value: the total compressive stress at this point and at this instant is as great as can occur during the impact. From this we see that the maximum stress during impact occurs !J.t the fixed end and is equal to twice tl:le m,azj...
=
(q)
which can be obtained at once i l mass of the rod entirely and nt~n ethement~ry way by neglecting the . . equa mg e stram energy f th dt h kmetic energy of the st 'k' b d o e ro o t e bolic curve2 defined byrt1hmg ºt!· The dotted line shown is a parae equa lon
(r)
It will be seen that for lar l f approximation. ge va ues o
.
1/a, It always gives a very good
The theory of impact deveio ed b . that contact takes place at th P a. ove is based on the assumption the end of the rod Th' d~ ~a~e I~stant over the whole surface of . ls con lt10n lS d1fficult to realize in practice and 1 1
Se T~p:pers by Saint-Venant and Flamant, 'loc. cit. urve was proposed by Boussinesq; see Compt. rend., p. 154, 1883.
'
452
THEORY OF ELASTICITY
453
PROPAGATION OF WAVES IN ELASTIC SOLID MEDIA
experiments do not agree satisfactorily with the theory. 1 A much better agreement with the theory was obtained by using helical springs instead of rods. 2 ln such a case the velocity of propagation of longitudinal waves is small, and the time T taken by the wave to travel across the rod and back is large in comparison with the time required for flattening small unevennesses of the ends. Another way of making experiments under definite conditions is to use rods with spherical ends and to consider the local deformation, which can be found from Hertz's formula 3 (see page 372). 144. Waves of Dilatation and Waves of Distortion in Isotropic Elastic Media. ln discussing the propagation of waves in an elastic medium it is of advantage to use di:fferential equations in terms of displacements [Eqs. (131), page 234]. To obtain the equations of small motion from these equations of equilibrium, it is only necessary to add the inertia forces. Then, assuming that there are no body forces, the equations of motion are ae a2u (>. + G) ax + G V2u - p at 2 = o
ae + G) ay + G V2v ae (>. + G) az + G V w -
(À
2
p p
a2v at2 =
o
. Consider now the case when the deformation roduced
:~:c~t t:c~~:~::~:~sb(s::~~~;S) The rotation ~f an ele!:n:h:;:::~
7ii2 = o
1
a2
av
at 2
au
ax - ay
aw
=o,
ª"'ax
Then
V=-,
ay
(À
=O
+ 2G) v2u - P a2u =
..... ..... .....
See W. Voigt, Ann. Physik, vol. 19, p. 44, 1883, and vol. 46, p. 657, 1915. For a complete review of literature on impact see the article by T. Põschl, "Handbuch der Physik," vol. 6, p. 525, Berlin, 1928. 2 Such experiments were made by C. Ramsauer, Ann. Physik, vol. 30, p. 416, 1909. a Such an investigation was made by J. E. Sears, Trans. Cambridge Phil. Soe., vol. 21, p. 49, 1908. See also J. E. P. Wagstaff, Proe. Roy. Soe. (London), series A, vol. 105, p. 544, 1924; and W. A. Prowse, Phil. Mag., vol. 22, p. 209, 1936.
'
i1\1,
(b)
ª"'az
=
(e)
v2u
O (265)
T~s: are equations for irrotation~~ ~a~es". ~r ~a~es of dilatation.
obtain~~\general cas~ of propagat10n of waves in an elastic medium is For both forro
~n~~p~~p~:~!s ~~;es of ~istortifon an_d waves of dilatation.
equa ions o mot1on have the common
- 2 = a2v2.1. at "'
These are equations for waves called waves of distortion.
•
w
ae a a= -dX v2cf> = X
2
e = V >,
in which
i',
au - aw - -o az ax -
=o,
ª"'
U=-,
(264)
1
av
ay - az
These equations are satisfied ü the dis l t from a single function cf> as follows: p acemen s u, v, w are derived
a2
Assume first that the deformation produced by the waves is such that the volume expansion is zero, the deformation consisting of shearing distortion and rotation only. Then Eqs. (263) become GV2u -
d
Substituting these in Eqs. (263), we find that
+ ay2 + (jz2
a2u p-
-
;::t~~ ~:~::;:r:e deformation is i"rrotati"onal ca:xt:;ef~~:x~:· re~r:~
in which e is the volume expansion, and the symbol V2 representa the operation
. a2 v2 = ax2
(a)
Anal?~ous expressions give the rotation about th
(263)
a2w
~ (:: - ~;)
=
w.
a=
c1
=~À~ 2G
(266)
(267)
for the case of waves of dilatation, and a =
C2
= /;
(268)
for the case of waves of distortion. We shall are velocities of propagation of waves f d"l t t~ow show that c1 and c2 o i a a ion and of distortion .
\'
.
THEORY OF ELASTICITY
454 146. Plane Waves. If a disturbance is produced at a point of an elastic medium, waves radiate from this point in all directions. Ata great distance from the center of disturbance, however, such waves can be considered as plane waves, and it may be assumed that all particles are moving parallel to the direction of wave propagation (longitudinal waves), or perpendicular to this direction (transverse waves). ln the first case we have waves of dilatation; in the second, waves of distortion. Considering longitudinal waves, if we take the x-axis in the direction of wave propagation, then v = w = O and u is a function of x only. Equations (265) then give
PROPAGATION OF WAVES IN ELASTIC SOLID MEDIA
455
Agam we have an equation of the s ª:~ f orm as before, and we can conelude that waves of distortion with the velocity are emg propagated along the x-axis C2
=~
C1
~
or, by (269), C2
=
1 - 2v
(270)
2(1 - v)
For v = 0.25, the above equation gives
(a) Any function
f(x - c2t)
This is the same equation as we had before in discussing longitudinal waves in prismatical bars [see Eq. (258), page 4391 except that the
(e)
~~~~~-l~~~~~--i
quantity
is replaced by the quantity C1
=~À +p2G
Substituting for À and G their expressions in terms of E and Poisson's ratio (see pages 10 and 9), this latter quantity can be represented in the forro (1
E(l - v) v)(l - 2v)p
+
i
,1
is a solution of Eq .e(b) and with the velocity Ta{ep;esents a wave traveling in the x-direction 2· e, or example, solution (e) in the form •
(269)
It can be seen that c1 is larger than e. This result is due to the fact that lateral displacement in this case is suppressed, while in the case of a bar it was assumed that longitudinal strain is accompanied by lateral contraction or expansion. The ratio ci/c depends on the magnitude of Poisson's ratio. For v = 0.25, ci/c = 1.095; for v = 0.30, ci/c = 1.16. All the conclusions obtained before regarding the propagation and superposition of longitudinal waves can also be applied in this case. Consider now transverse waves. Assuming that the x-axis is in the direction of wave propagation, and the y-axis in the direction of transverse displacement, we find that the displacements u and w are zero and the displacement v is a function of x and t. Then, from Eqs. (264),
,1
Frn. 243.
(b)
21!"
v = Vo sm T (x - c2t)
(d)
!he wave has in this case a sinusoidal f is l and the amplitude v Th 1 . orm. The length of the wave o. e ve omty of transverse motion is iJV -
at -
27rC2
- -l- Vo cos
21!"
T (x -
C2t)
(e)
It.IS zero when the displacement (d) is
value when the displacement is zero. the wave is 'Y"ll
= ax av =
27rVo
l
. T~ max1m~m and. has its largest e shearmg stram produced by
21!" cos T (x - C2t)
(f)
It will be seen that the maximum d" t . absolute value of the velo "t ( ) is ort10n(~) and the maximum of the c1 Y e occur ata g1ven po" t . l . m s1mu taneously. . e can represent this kind of wa W (Fig. 243) be a thin thread of l ve.propa~at10n as follows: Let mn an e astic medmm. When a sinusoidal
456
· THEORY OF ELASTICITY PROPAGATION OF WAVES IN ELASTIC SOLID MEDIA
wave (d) is being propagated along the x-axis, an element A of the thread undergoes displacements and distortions, the consecutive values of which are indicated by the shaded elements 1, 2, 3, 4, 5 . . . . At the instant t = O, the element A has a position as indicated by 1. At this moment its distortion and its velocity are zero. Then it acquires a positive velocity and after an interval of time equal to i l/c2 its distortion is as indicated by 2. At this instant the displacement of the element is zero and its velocity is a maximum. After an interval of time equal to i;l/c2 the conditions are as indicated by 3, and so on. Assuming that the cross-sectional area of the thread is equal to unity, the kinetic energy of the element A is P dx (ªv)2 = P dx. 4'r2c22 vo2 cos2 211' (x - c2t)
at
2
2
z2
z
and its strain energy is 1
G dx 4'r 2vo 2
2 Gy""' 2 dx = 2
-z-
2 -
211' cos 2 T (x - c2t)
Remembering that c2 2 = G/p, it can be concluded that the kinetic and the potential energies of the element at any instant are equal. This is the sarne conclusion as we had before in discussing longitudinal waves in prismatical bars (see page 442). ln the case of an earthquake both kinds of waves, those of dilatation and those of distortion, are propagated through the earth with velocities c1 and c2. They can be recorded by a seismograph, and the interval of time between the arrival of these two kinds of waves gives some indication regarding the distance of the recording station from the center of disturbance. 1 146. Propagation of Waves over the Surface of an Elastic Solid Body. ln the previous article we discussed propagation of waves in an elastic medium at a distance from the surface. On the surface of an elastic body is it possible to have waves of a different type, which are propagated over the surface and which penetrate but a little distance into the interior of the body. They are similar to waves produced on a smooth surface of water when a stone is thrown into it. Lord Rayleigh, who was the first to investigate these waves, 2 remarked: The waves produced in the earth by reciprocating engines are discussed in the following papers: L. Mintrop, Dissertation, Gõttingen, 1911; A. Heinrich, Dissertation, Breslau, 1930; G. Bornitz, "Übei· die Ausbreitung von Bodenschwingungen," Berlin, 1932. See also E. Reissner and H. F. Sagoci, J. Applied Phys., vol. 15, p. 652, 1944; H. F. Sagoci, ibid., p. 655. 'See Proc. London Math. Soe., vol. 17, 1887. 1
!
~ '
1
1
'I•
iill!L.,
457
"It is not improbable th t th rf important part in eartha e su ace. waves here investigated play an Diverging in two dimen~~::e;~tn~~n the collisio~ of elastic solids. tance from the source . y, . ey must acqmre at a great dis~ contmually mcreasing preponderance " Th st u d Y of record s of se1smic · e At a great distance from t:eª:~~r~~p~~:t~ ~aylei;.h's expectation. these waves may be considered as a ' . e .orma ion produced by that the body is bounded by th 1wo-d11~~ns10nal one. We assume sense of the y-axis in the directio: o:ne th ~' an~ take the positive the positive direction of the x-axis in th:rd. ~.mtenor of the body and Expressions for the displacements are obt:;:;d1on of wa;e. pro~agation. waves [Eqs. (265)] and d" t t" by combmmg dilatation both cases that w = O th1es orl Ito.n wafvEes [Eqs. (264)]. Assuming in , sou 1on o qs (265) re t" of dilatation can be taken in the form . presen mg waves
f J
U1
=
se-ru sin (pt - sx)
= -re-ru cos (pt - sx) (a) Th · expressions indicates that for real p. •t• e exronential factor in these waves rapidly diminishes with incre~:~ :;et~: ~:;t~ r th~~mplitude of pt - sx of the trigonometrical functions h y. e argument propagated in the x-direction with the velo:i;;,.ws that the waves are •
•
'
V1
m whwh p, r, and s are constants
e3 - P -
8
(271)
Substituting expressions (a) into E s (265) . tions are satisfied if q · _ ' we find that these e_qua. p2 r2 = 8 2 _ -...:.P_.___ Ã +2G or, by using the notation pp2 p2 À + 2G = C12 = h2 (b) we have r2 = s2 - h2 We take solutions of E (264) . (e) form qs. ' representmg waves of distortion in the tt2 = Abe-bu sin (pt _ ) . . . sx ' V2 = - Ase-bu cos (pt - sx) (d) in wh1chA is a const t d b . . . the volume ex ansio an an a P?sit1ve number. It can be shown thàt that Eqs. (26r) are :a~~;~:T~ndmg to the displacements (d) is zero and
b2 =
82
_PP
G
2
458
THEORY OF ELASTICITY
459
PROPAGATION OF WAVES IN ELASTIO SOLID MEDIA
or, by using the notation
Using the notation (e)
we obtain b2
and remembering that
= s2 - k2
(f) C2
ª:
) d (d) d taking u = Ui + U2, V = Vi + V2, we Combittlng solutions (a an A soas to satisfy the boundary now determine the constants f 'th pb~d;' is free from external forces, conditions. The 1:o~ndary do y =e O. Substituting this in Eqs. (134) hence, for Y =O, X:- O an _ = -1 we obtain on page 234 and taking l = n - 0 ' m ' ôu+ôv=O ôy ÔX
Eq. (l) becomes a
6 -
(~: _
2) (r2 _ 82)
4
+ 8(3
(g)
11
3a
or
6
+ A(b2 + 82)
=O
+ 2(r2 + Abs)
=O
(h)
a
2
= 4,
k2 ). h2 - 2 = G
= 16 (1 - ::)
(i - ~:)
d (271) all the quantities of this equation can By using Eqs. (b), (e), an. . f f dilatation, c2 of waves of be expressed by the veloc1t1es c1 o waves o bt . distortion, and ca of surface waves' and we o am
(~:: _ •· ·-··" ,
r
2
= 16 ( 1 _
11)
24a 4
+ 56a
2 -
= 2
32
+ 8)
2 +-,
=
211 11)
J= O
(m)
O
=O
2
a 2 =2--
V3
V3
Of these three roots only the last one satisfies the conditions that the 2 2 quantities r and b , given by Eqs. (e) and (f), are positive numbers. Hence, =
Taking, as an extreme case,
from (b) and (e). Eliminating the constant A f rom Eqs . (h) and using (e) and (f), we obtain (28 2 _ k2) 2 = 4brs 2 (k) or, by (e) and (f),
y
-
a2
C3
2
211
1 -- 211) - - a 2 - 16 [ 1 - 1 1 - 11 2(1 = 0.25, we obtain
2
where
(~: -
1 -
= 2(1 -
(a - 4)(3a 4 - 12a 2 The three roots of this equation are
. . d" t that the shearing stresses, and the The first of these equat1ons m ica esth surface of the body, are zero. h th rmal stresses on e . e no . f or u and v in these equat10ns we second t at the Substituting above express1ons 2rs
8a
Taking, for example,
>.e+2a::=o
find that
C1
2 2
and we find
a
6 -
aC2
11
8a 4
C3
=
= 0.9194 / ;
i,
Eq. (m) becomes
+ 24a
2
-
16 = O
= 0.9553 / ;
ln both cases the velocity of surface waves is slightly less than the velocity of waves of distortion propagated through the body. Having a, the ratio between the amplitudes of the horizc.,ntal and vertical displacements at the surface of the body can easily be calculated. For 11 = t, this ratio is 0.681. The above velocity of propagation of surface waves can also be obtained by a consideration of the vibrations of a hody bounded by two parallel planes. 1 1
~::) ( 1 - ~::)
(l)
See H. Lamb,Proc. Roy. Soe. (London), Series A, vol. 93,p.114, 1917. SI. Timoshenko. Phil. Mag., vol. 43, p. 125, 1922.
Seealso
INSTITUTUL PQLITEMlll
T1l'·i 1::. O
A BIBLIOTECA CENTRALl A R
APPENDIX THE APPLICATION OF FINITE-DIFFERENCE EQUATIONS IN ELASTICITY 1. Derivation of Finite-difference Equations. We have seen that the problems of elasticity usually require solution of certain partial differential equations with given boundary conditions. Only in the case of simple boundaries can these equations be treated in a rigorous manner. Very often we cannot obtain a rigorous solution and must resort to approximate methods. As one of such methods we will discuss here the numerical method, based on the replacement of differential equations by the corresponding finite-difference equations. 1 If a smooth function y(x) is given by a series of equidistant values Yo, Y1, Y2, . . . for x = O, x = õ, x = 2õ, . . . , we can, by subtraction, calculate the first differences (Ã1Y)=o = Y1 - Yo, (Ã1Y)z=a = Y2 - Y1, (Ã1Y)=u = Ya - y2, . . . . Dividing them by the value õ of the interval, we obtain approximate values for the first derivatives of y(x) at the corresponding points: (1) 1 It seems that the first application of finite-difference equations in elasticity is dueto C. Runge, who used this method in solving torsional problems. (Z. Math. Phys., vol. 56, p. 225, 1908.) He reduces the problem to the solution of a system of linear algebraic equations. Further progress was made by L. F. Richardson, Trans. Roy. Soe. (Lond-On), series A, vol. 210, p. 307, 1910, who used for the solution of such algebraic equations a certain iteration process, and so obtained approximate values of the stresses produced in dams by gravity forces and water pressure. Another iteration process and the proof of its convergence was given by H. Liebmann, Sitzber. Bayer. Akad. Wiss., 1918, p. 385. The convergence ofthis iteration process in the case of harmonic and biharmonic equations was further discussed by F. Wolf, Z. angew. Math. Mech. vol. 6, p. 118, 1926, and R. Courant, Z. angew. Math. Mech., vol. 6, p. 322. The finite-differences method was applied successfully in the theory of plates by H. Marcus, Armierter Beton, 1919, p. 107; H. Hencky, Z. angew. Math. Mech., vol. 1, p. 81, 1921, and vol. 2, p. 58, 1922. ln recent times the finite-differences method has found very wide application in publications by R. V. Southwell and his pupils. See R. V. Southwell, "Relaxation Methods," •rnls. 1, II, and III.
43:1
463
THEORY OF ELASTICITY
APPENDIX
U sing the first differences we calculate the second differences as follows:
ln this way every torsional problem reduces to the finding of a set of numerical values for the stress function q, which satisfy equation (5) at every nodal point within the boundary of the cross section and become constant along the boundary. As the simplest example, let us consider a bar of a square cross section a X a, Fig. 2, and use a square net with mesh side ó = ta. From symmetry we conclude that in this y case it is sufficient to consider only a one-eighth of the cross section, shaded in the figure. If we determine '\:e.__ the values a, fJ, 'Y of the function e/> 4~ /J at the three points shown in Fig. 2, a r we shall know e/> at all nodal points of the net within the boundary. Along the boundary we can asEume it equal to zero. Thus the problem reduces to the calculation of three Frn. 2. quantities a, fJ, 'Y, for which we can write three equations of the forro (5). Observing the conditions of symmetry we obtain
462
(Ã2y),.o=a = (Ã1y)~ - (Ã1Y)x=o = Y2 - 2y1
+ Yo
With second differences we obtain the approximate values of second derivatives such as ,li
r-41
r ~
If we have a smooth function w(x,y) of two 2 variables, we can use for approximate calcula.1 o f tions of partial derivatives equations similar to Eqs. (1) and (2). Suppose, for example, that 4 x we are dealing with a rectangular boundary, Frn. 1. Fig. 1, and that the numerical values of a function w at the nodal points of a regular square net with mesh side ó are known to us. Then we can use as approximate values of the partial derivatives of w ata point O the following expressions aw _,,,, ax
a2w ():x;2
aw _,,,,
Wo
W1 -
'
ó
2wo õ2
W1 ""'
+ Wa
ay a2w iJy2 ""'
'
W2 -
2wo õ2
+ W4
2a
differences. Take as a first example the torsion of prismatical bars. The problem can be reduced, as we have seen,1 to the integration of the partial differential equation
ªax2 + ªay2 2
2
>
!
i
>
(4)
= -2G8
in which e/> is the stress function, which must be constant along the boundary of the cross section, 8 is the angle of twist per unit length of the bar and G is the modulus of shear. Using formulas (3) we can transfo;m the above equation into the finite-difference equation
~ (c/>1 1
+ 4'2 +
See Eq. (142), p. 261.
c?a
+
Ç4 - 4c?o)
l
(3)
ln a similar manner we can derive also the approximate expressions for partial derivatives of higher order. Having such expressions we can transform partial differential equations into equations of finite :' i:
7
Wo
ó W2 -
--
-2G8
(5)
2fJ - 4a = - 2G8ó 2 'Y - 4fJ = -2G8ó 2 4fJ - 4'Y = -2G8ó 2
+
Solving these equations, we find a = l.375G8ó 2,
fJ = l.750G8ó 2,
'Y
= 2.250G8ó 2
fhe required stress function is thus determined by the above numerical values at all nodal points within the boundary and by zero values at the boundary. To calculate partial derivatives of the stress function we imagine a smooth surface having as ordinates at the nodal points the calculated numerical values. The slope of this surface at any point will then give us the corresponding approximate value of the torsional stress. Maximum stress occurs at the middle of the sides of the boundary square. To get some idea of the accuracy which can be obtained with the assumed small number of nodal points of the net let us calculate torsional stress at point O, Fig. 2. To get the neces;ary slope we take a smooth curve having at the nodal points of the x-axis the calculated ordinates fJ, 'Y, fJ. These values, divided by -fG8ó 2, are given in the
APPENDIX THEORY OF ELASTICITY
464
Using now seven ordinates along the x-axis and calculating 1 the slope at point O, we obtain the maximum shearing stress
f the table below The remaining lines of the table give . second 1me o · · d th the values of the consecutive finite differences.1 The reqmre smoo curve is then given by Newton's interpolation formula: x=
o
=
7
A2 =
-5
-4
Â3 =
1
-1
=
-2
õ
2õ
3õ
(~:)_ 0 =
------------5
------------ - - -- - - - - - - Â4
0.661G8a
The error of this result is about 2 per cent. Having the results for ll = ia and ll = ta a better approximation can be obtained by extrapolation.1 It can be shown 2 that the error of the derivative of the stress function q,, due to the use of finite difference rather than differential equations, is proportional to the square of the mesh side, when this is small. If the error in maximum stress for ll = ia is denoted by Â, then for ll = ia it can be assumed equal to Á(i) 2. Using the values of maximum stress calculated above we obtain  from the equation
4õ
---- - - - - -- o 7 9 o 7 q, = -----2 -2 -7 A1
465
 - Â(i) 2 = O.OI5G8a
from which
q, =
+ x ~1 + x(x
Â
+ x(x
- ll) 1 . 2 2. ll2
+ x(x
Â
Â3
- ll) (x - 2/l) 1 . 2 . 31ls
The more accurate value of the stress is then
Â4
- ll)(x - 21l)(x - 31l) 1 . 2. 3. 41l4
0.646G8a
Taking the derivative of q, and substituting for Â1, Â2, . . 1.. thl_eidr values from the table mu t1p 1e by G81l 2 /4 we obtain, for x = O, '"
r-&-.J
~a
llJ
r
rx.1
P1
"-1
o
(ª"'ax )""""º =
"Jí
::JC
FIG. 3.
124G81l = 0.646·Ga8 48
Comparing this result with the correct value given on page 277, we see that the error in this case is about 4.3 per cent. To get better accuracy we have ' to use a finer net. Taking, for example, ll = a/6, Fig. 3, we
have to solve six equations and we obtain 2 ,
a = 0.952 X 2G81l 2 a 1 = 2.125 X 2G81l ,
f3 = 1.404 X 2G81l 2, f3i = 2.348 X 2G81l
2
,
'Y 'Yi
=
=
2
1.539 X 2G81l 2 2.598 X 2G81l
. . d of the set of quantities We consider here the di:fferences as all ex1stmg a.t. one en and use them in Newton's formula 1
= 0.027G8a
+ 0.027G8a = 0.673G8a
which differs from the exact value 0.675G8a by less than t per cent. 2. Methods of Successive Approximation. From the simple example of the preceding article it is seen that, to increase the accuracy of the finite-difference method, we must go to finer and finer nets. But then the number of equations which must be solved becomes larger and larger and the time required for their solution will be so great that the method becomes unpractical. 3 The solution of the equations can be greatly simplified by using a method of successive approximations. To illustrate this let us consider the equation 4
(6) 1 The calculation of derivatives of an interpolation curve is greatly simplified by using the tables calculated by W. G. Bickley. These tables are given in the book "Relaxation Methods in Theoretical Physics," by R. V. Southwell. 2 See L. F. Richardson, loc. cit. 3 ln the previously mentioned paper by C. Runge a system of 42 equations was used and, due to the simplicity of these equations, the solution was obtained without much difficulty. 4 It was shown on p. 265 that torsional problems can be reduced to the solution of this equation with prescribed values of q, at the boundary.
.. The corresponding finite-difference equation, from Eq. (5), will be
H cJ>1
<7)
+ c/>2 + cJ>3 + cJ>4)
It shows that the true value of the function cJ> at any n~dal point O of the square net is equal to the average value of the_ ~unct10n at the f?ur adjacent nodal points. This fact will ~ow ~e ut1hzed for calculat1?n of the values of q, by successive approx1mat10ns. Let us t~ke agam, as the simplest example, the case of a square boundary'. Fig. 4, and assume that the boundary values of q, are such as shown m_the figure. From the symmetry of these values with respe~t to t~e vertical central axis we conclude that q, also will be symmetncal with respect to the sarne axis. Thus we have to cal!OOO 121JO 1000 culate only six nodal values, a, b, b1 a 2 b2 of q,. This can easily 1000 1000 1000 1 92S 988 ' done ' 'by 'writing and solvmg . s1x . 92S be 899 962 899 800 800 e íb equations (7), which are simple in a 900 900 900 806 this case and which give >a = 854, 8SO 806 7SO 800 7SO 600 c/>b = 914, c/>a1 = 700, c/>b, = 750, 600 CL ª.t 800 b1 800 .. 800 >a,= 597, c/>b, = 686. 1 Instead of 65"/ 7J8 6S/ 622 711 400 this we can proceed as follows: 622 400 Cz lbz W e assume some values of cJ>, for ª2 instance those given by the top ' . 600 800 600 numbers in each column wntten FIG. 4. in Fig. 4. To get a better
ª
approximation for e/>, we use for each nodal point Eq. (7). point a we take, as a first approximation, the value >a' = -l-(800
+ 1,000 + 1,000 + 900)
Considering
= 925
ln calculating a first approximation for point b we ~se the c~lculated value
+
+
+
Màking similar calculations for ali inner nodal points, we obtain t~e first approximations given by the second (from the top) n~mbers l~ each column. U sing these numbers we can calculate t e secon
1\
j
r: !'
approximations such as cJ>a" = -l-(800 c/>b" = t(899
+ 1,000 + 988 + 806) = 899 + 1,200 + 899 + 850) = 962
............
1
·~,, '
dU
467
APPENDIX
THEORY OF ELASTICITY
466
We m11lrn th~ c~lc~la~ion with three figures only and neglect decimals.
The second approximations are also written in Fig. 4 and we can see how the successive approximations gradually approach the correct values given above. After repeating such calculations 10 times we obtain in this case results differing from the true values by no more than one unit in the last figure and we can accept this approximation. Generally, the number of repetitions of the calculation necessary to get a satisfactory approximation depends very much on the selection of the initial values of the function q,. The better the starting set of values the less will be the labor of subsequent corrections. It is advantageous to begin with a coarse net having only few internai nodal points. The values of cJ> at these points can be obtained by direct solution of Eqs. (7) or by the iteration process described above. After this we can advance to a finer net, as illustrated in Fig. 5, in which heavy lines represent the coarser net. Having the values of cJ> for the nodal points, shown by small circles, and / applying Eq. (7), we calculate the '-// values for the points marked by crosses. / / Using now both sets of values marked / / / by circles and by crosses and applying again Eq. (7) we obtain values for FIG. 5. points marked by small black circles. ln this way all values for the nodal points of the finer net, shown by thin lines, will be determined, and we can begin the iteration process on the finer net. Instead of calculating the values of q,, we can calculate the corrections 1/; to the initially assumed values q, 0 of the function q,. 1 ln such a case q, = cJ>º + 1/;
''
''
Since the function it, and we obtain
cJ>
satisfies Eq. (6) the sum q, 0
''
'
'
+ 1/; also must satisfy (8)
At the boundary the values of cJ> are given to us, which means that there the corrections 1/; are zero. Thus the problem is now to find a function 1/; satisfying Eq. (8) at each internal point and vanishing at 1 This method simplifies the calculations since we will have to deal with comparatively small numbers.
the boundary. Replacing Eq. (8) by the corresponding fi_nite-difference equation we obtain for any point O of a square net (Fig. 1) 0 1[t 1 +1[t 2 +1/ts + 1/14 - 41/to = -(c/>1º + c/>2° + c/>3° + c/>4° - 4c/>o ) (8') The right-hand side of this equation can be evaluated for each internai nodal point by using the assumed va~ues e/>º of the function e/>.. Thus the problem of calculating the correct1ons 1[t reduces to ~he solu_t10n of a system of equations similar to Eqs. (5) of the precedmg art1cle, and these can be treated by the iteration method. . . 3. Relaxation Method. A useful method f?r treatmg d1fference equations, such as Eqs. (8') in the p~eceding art1cl~, was developed by R. v. Southwell and was called by h1m the relaxati~n m_ethod. Southwell begins with Prandtl's membrane analogy, 1 wh1ch is based on the fact that the differential equation (4) for torsional problems has the sarne form as the equation
aw 2
ax2
aw 2
+ ay2
q
= -
S
(9)
for the deflection of a uniformly stretched and laterally _lo_a~ed me~n this equation w denotes deflection from the m1tially honb rane · l · · f th d" zontal plane surface of the membrane, q is the mtens.1ty o e istributed load, and S is the constant tensile force per umt length ~f the boundary of the membrane. The problem is to find the de~ect10n w as a function of x and y which satisfies Eq. (9) at every pomt of the membrane and which vanishes at the boundary.. Let us derive now the corre~ sponding finite-difference equa... ~ ~ Uj tion. For this purpose we replace ° 1 the membrane by the squar~ net 3 o of uniformly stretched strmgs, qô2 Fig. 1. Considering point O and Fw. 6. denoting by Sô the tensile force in the strings, we see that the st~ings. 0-1 and O-~ exert on the node O, Fig. 6, a force in the upward direct10n, equal to
Só ( Wo
~ W1 + Wo ~ Ws)
(10)
A similar expression can be written for the forc~ exerted by t~e two other strings, 0-2 and 0-4. Replacing the contmuous load actmg on i
2
See p. 268. We consider the deflections as very small.
469
APPENDIX
THEORY OF ELASTICITY
468
the membrane by concentrated forces qô 2 applied at nodal points, we can now write the equation of equilibrium of a nodal point O as follows: (11) This is the finite-difference equation, corresponding to the differential equation (9). To solve the problem, we have to find such a set of values of the deflections w that equation (11) will be satisfied at every nodal point. We begin with some starting values woº, w1º, w2º, w3°, w4°, . . . of the deflection. Substituting them into Eqs. (11) we shall usually find that the conditions of equilibrium are not satisfied and that, to maintain the assumed deflections of the net, we need to introduce supports at the nodal points. The quantities such as
Ro = qô 2 + S(w1º
+ w2º + W3° + W4° -
4woº)
(12)
will then represent the portions of the load transmitted to the supports. W e call these forces residual forces, or residuals. Imagine now that the supports are of the screw-jack type, so that a controlled displacement may be imposed at any desired nodal point. Then by proper displacements of the supports we can ultimately make all residual forces (12) vanish. Such displacements will then represent the corrections which must be added to the initially assumed deflections woº, w1º, w 2°, ... to get the true values of w. The procedure which Southwell follows in manipulating the displacements of the supports is similar to that developed by Calisev 1 in handling highly statically indeterminate frames. We first displace one of the supports, say support O, Fig. 6, keeping the other supports fixed. From such equations as (11) we can see that to a downward displacement w0' will correspond a vertical force - 4Swo' acting on the nodal point O. The minus sign indicates that the force acts upward. Adjusting the displacement so that , Ro that is, (13) Ro - 4Swo' = O, Wo = 4S we make the residual force (12) vanish and there will no longer be a pressure transmitted to the support O, but at the sarne time pressures Swo' will be transferred to the adjacent supports and their residual forces will be increased by this amount. Proceeding in the sarne way ! K. A. Calisev, Tehnicki List, 1922 and 1923, Zagreb. German translation in Pubs. lntern. Assoe. Bridge and Structural Engineering, vol. 4, p. 199, 1936. A similar method was developed in this country by Hardy Cross.
471
THEORY OF ELASTICITY
APPENDIX
with all the other supports and repeating the procedure several times, we can reduce all residual forces to small quantities, which can be neglected. The total displacements of the supports, accumulated in this procedure, represent then the corrections which must b~ added with the proper signs to the starting values woº, w1°, w2°, ••• morder to obtain the true deflections of the stretched square net. To simplify the calculations required by the procedure described, we first put Eq. (11) in nondimensional form by substituting
The denominator 1,000 is introduced into Eq. (17) for the purpose of making the iJ;'s such large numbers that half a unit of the last figure can be neglected. Thus we have to deal with integer numbers only. To make our example as simple as possible we will start with the coarse net of Fig. 2. Then we have to find values of iJ; only for three internal points for which we already have the correct answers (see page 463). W e make our square net to a large scale to have enough space to put on the sketch the results of all intermediate calculations (Fig. 7). The calculation starts with assumed initial values of i/;, which we write to
470
qó2
(14)
w=sl/I ln this way we obtain
-~
1 + (ih+ 1/12
+ 1/13 + 1/14 -
(15)
41/;o) =O
= 1
+ (i/; 1º + 1/12º + i/;3º + 1/14º -
41/;oº)
(16)
which in this case are pure numbers. Our problem now is to add to the assumed values i/;o 0, 1/11°, 1/12°, . such corrections as to annul all residuals. Adding to i/;oº a correct1on 1/lo' we add to the residual To the quantity -41/;o', and to the residuals of the adjacent nodal points the quantities i/;o'. Taking i/;o' = To/4 we shall annul the residual at the nodal point O and shall somewhat change the residuals at the adjacent nodal points. Proceeding in the sarne way with all nodal points and re?eating the proced~~e many times we shall in due course reduce the residual forces to neghg1ble values and so obtain the values of iJ; with sufficient accuracy. The corresponding values of w will then be obtained from Eq. (14). ~ To illustrate the procedure let us consider the problem of torsion of a square bar, already discussed in Art. 1. ln ~his c~se we have the differential equat;on (4). To bring it to nond1mens10nal form let us put
2
(17)
q, = 2G8ó iJ; 1,000
The finite-difference equation (5) then becomes 1,000
+ (i/; 1 + 1/12 + i/;a + 1/14 -
41/;o)
-2 -J -6
700
where i/;o, iJ; 1, • . . are pure_ numbers. The problem then reduces to finding such a set of values of 1/1 that Eq. (15) will be satisfied at all inner points of the net. At the bo~nd ary iJ; is zero. To get the solution we proceed in the manner descr1b~d above and take some starting values i/;oº, i/;1°, 1/12°, • . • • They w1ll not satisfy the equilibrium equations (15) and we shall have residuals To
o
=
O
(18)
-6
-12
900
:}$ o
-12 -z -J
-6 700
!l =i
:-
-2 -J -6
-12
900
-~ =~ -X -:~
=~
-'.
-1
~i
-1 -);
-2 -3
o
-)( -:~
+i
-2
o
~
-~
-:;: -2
: s-!'}
-J
-6 700
so -~
-12
!~ --fi -~
=i -z :5 --1
-:~ -J -~ -6 +~
-:~
-6 -12
~
1100
900
_,
-~ -:~ -~ -~
-2 -J
-~
-~ - -~ -~
-~ -~ -~
--;~ j
+i 7õ~ =~
-~ +~
o
--1-~ -12 -i -6 :li:;~ :5 o
-12
900
l r
ti=-
a Fm. 7.
the left above each nodal point. The values 700 900 and 1 100 are ' ' ' intentionally taken somewhat different from the previously calculated correct values. Substituting these values together with the zero values at the boundary into the left-hand side of Eq. (18) we calculate the residual forces for all nodal points. These forces are written above each nodal point to the right. The largest residual force, equal to 200, occurs at the center of the net, and we start our relaxation process from this point. Adding to the assumed value 1,100 a correction 50, which is written in the sketch above the number 1,100, we eliminate entirely the residual force at the center. Thus we cross out the number 200 in the sketch and put zero instead. Now we have to change the residuals in the adjacent nodes. W e add 50 to each of those residuals and write the new values -50 of the residuals above the original values as shown in the figurP,. This finishes the operation with the central point of th~
472
THEORY OF ELASTICITY
APPENDIX
11et. We have now four symmetrically located nodal points with residuais -50 and it is of advantage to make corrections to all of them simultaneously. Let us take for all these points the sarne correction, equal to -12. 1 These corrections are written in the sketch above the initial values, 900. With these corrections the values 12 X 4 = 48 must be added to the previous residuais, equal to -50, and we will obtain residuais equal to -2, as shown in the sketch. At the sarne time the forces -12 will be added to the residuais in the adjacent points. Thus, as it is easy to see, -12 X 4 = -48 must be added to the residual at the center and -12 X 2 = -24 must be added at the points closest to the corners of the figure. This finishes the first round of our calculations. The second round we again begin with the point at the center and make the correction -12, which eliminates the residual at this point and adds -12 to the residuais of the adjacent points. Taking now the points near the corners and introducing corrections -6, we eliminate the residuais at these points and make the residuais equal to - 26 at the four symmetrically located points. To finish the second round we introduce corrections -6 at these points. The sketch shows three more rounds of calculations which result in further 'reduction of the residuais. · The required values of if; will be obtained by adding to the starting values all the corrections introduced. Thus we obtain
get a better approximation we must advance to a finer net. Usmg the method illustrated in Fig. 5, we get starting values of y,, for a square net with mesh side õ = -§-a. Applying to these values the standard relaxation process the values of if; for the finer net can be obtained and a more accurate value of the maximum stress can be calculated. With the two values of maximum stress found for 0 = ia and õ = ia a better approximation can be found by extrapolation as explained in Art. 1 (see page 465). ' 4. Triangular and Hexagonal N ets. ln our previous discussion a square net was used,. but sometimes it is preferable to use a triangular or hexagonal net, F1gs. 8a and 8b. Considering the triangular net,
700 - 6 - 3 -
-!
= 688.5,
900 - 12 - 6 - 3 - 2 - 1 1,100
+ 50 -
:g0 oeo 5
2
=
i.311oeo 2 =
12 - 6 - 3 - 2 = 1,127
o.08610ea 2
2
= l.752G1Jõ 2 = 0.1095GIJa 2
5~~7 G1Jõ
2
= 2.254G1Jõ 2 = 0.1409G1Ja 2
which are in very good agreement with the results previously obtained (see page 463). It is scen that Southwell's method gives usa physical picture of the iteration process of solving Eqs. (15) which may be helpful in selecting the proper order in which the nodes of the net should be manipulated. • We take the correction -12, instead of work with·integer numbers.
-!ii1-
(aJ
(6) Fw. 8.
~~~ GIJõ 1
!º
= 876,
Equation (17) then gives for cf> the values 6
473
= -12.5, since it is preferable to
Fig. 8a, .we se~ that the distributed load within the hexagon shown by dotted lmes will be transferred to the nodal point O. If 0 denotes the mesh side, the side of the above hexagon will be equal to õ/-\!'3 and thc area of the hexagon is V3 õ2/2, so that the load transferred to each nodal ~oint will be V3 o2q/2. This load will be balanced by forces in the strm~s 0-1, 0-2, . . . , 0-6. To make the string net correspond to the uniformly stretched membrane, the tensile force in each string must .be equal to the tensile force in the membrane transmitted through one s1de of the hexagon, i.e., equal to So/V3. Proceeding now as in the preceding article, we obtain for the nodal point O the following equation of equilibrium:
,.
or w1
w e introduce
+ w + · · · + Ws 2
3 qó 2
-
6wo
+2S
=O
(19)
a nondimensional function 1f; defined by the equation 3 qó 2
w
+ 1/; + . . · + 1f; 2
6 -
61/;o
+1
= O
(21)
tion methods. . . d th ln the case of a hexagonal net, Fig. 8b, the load d1str:bute over e equilateral triangle, shown m the figure by dotted lines, will be transferred to the nodal point O. Denoting by ó the length of the mesh side we see that the side of the 2 triangle will b~ ó .y3 and its area 3 V3 ó / 4. 2 The corresponding load is 3 V3 qó / 4. This load will be balanced by the tensile forces in the three strings, 0-1, 0-2, 0-3. To make the string net correspond to the ~ a .\ stretched membrane we take the tensile Fm. 9. forces in the strings equal to Só .y3. The equation of equilibrium will then be w 1 + W2 + wa - 3wo. Só .y3 + 3 or W1
+ W2 + W3
-
3wo
6cf>o = 3G8ó 2 = GBa 3
2
and
Such an equation can be ·written for each internal n~dal p.oint, and for the solution of these equations we can, as before, use iterat10n orrelaxa-
ó
there will be only one internai point O of the net, and the values of the required stress function cf> are zero at all adjacent nodal points 1, 2, ••. , 6 since these points are on the boundary. The finite-difference equation for point O is then obtained from Eq. (19) by substituting cf>o for w 0 and 2G8 for q/ S, which gives
(20)
=2s1/I
and the finite-difference equation becomes 1/;i
475
APPENDIX
THEORY OF ELASTICITY
474
4
-- o
(22)
To get the finite-difference equations for torsional problems we have to substitute in Eqs. (19) and (22) 2G8 instead of q/ S. . As an example let us consider torsion of a bar the cross. sect10n ~f which is an equilateral triangle, i Fig. 9. The rigorous solut10n for th1s . case is given on page 266. Using the relaxation method it is natural to select for th1s c~se a triangular net. Starting with a coarse net we take th~ mesh s1de ô equal to one-third of the length a of the side of the triangle. Then This example is discussed in detail in Southwell's book, referred to above.
G8a 2 18
(23)
= --
Let us now advance to a finer net. To get some starting values for such a net let us consider point a - the centroid of the triangle 1-2-0. Assume that this point is connected to the nodal points O, 1, and 2 by the three strings a-0, a-1, a-2 of length ó/y'3. Considering the point a as a nodal point of a hexagonal net, Fig. 8b, substituting into Eq. (22) o/0 for ó, 2G8 for q/S, and taking W1 = W2 = o, W3 = >o, Wo = cf>a, we obtain cf>
ª
=
!3 (cf>o + G8ó 2
2 )
= G8a 27
2
(24)
The sarne values of the stress function may be taken also for the points b, e, d, e, and f in Fig. 9. To get the values of the stress function at points k, l, m, we use again Eq. (22) and, observing that in this case W1 = w2 = wa = O, we find >k
2
V3 qó =o
+ 34 sq õ2
cf>o
=
cf>m
=
>1
G8a 2
= 54
(25)
ln this way we find the values of cf> at ali nodal points marked by small circles in Fig. 10. It is seen that at each of the nodal points a, e, and e there are six strings ~·----a 10 as required in a triangular net, Fig. 8a. Fm. · But at the remaining nodes the number of strings is smaller than six. To satisfy the conditions of a triangular net at all internai points we proceed as indicated by dotted lines in the upper portion of Fig.10. ln this way the cross section will be divided into equilateral triangles with sides õ = a/9. From symmetry we conclude that it is sufficient to consider only one-sixth of the cross section, which is shown in Fig. 1la. The values of cf> at the nodal points O, a, b, and k are already determined. The values at the points 1, 2, and 3 will now be determined, as before,
l
476
. E (22) and the values of cp at three adjacent points. by usmg q. point 1, for example, we will get 3 cf>o + c/>b + cf>a - 3c/J1 + 4 . 2GO g - O
For
(ª)2. -
and substituting for obtain
,i..ª'
,i..b,
,i..o
'I'
'I'
'I'
cf> 1
(26)
= "''\ · G8a 2
left-hand side of Eq. (28), we find the corresponding residuais
Ro = 1/11º
the previously calculated values, we
t d Ali these values are written ln a similar way c/>2 and c/>a are ca1c~1a e . . . . lla 1 They down to the left of the corresponding nodal pomts ~n Fig. . O will now be taken as starting values in the relaxat10n process.
1
+
c/> 2
+ ... + cp 6 -
6cf>o
ª2 + 3G8 81
=O
To bring it into purely numerical form we introduce the notation cf>
and obtain
Goa2i/I* = 486
1/1 = 486cf>
Goa2
or
+ 1/16
(27)
= O (28) E (27) are written to the The starting values of 1{;, calcu1ate d f rom q. ' 1 . t the . t ·n F1º rr 1lb Substituting these va ues m o left of t h e nodal pom s 1 ,.... · r
1/11 + 1/12 + . . .
The constant factor Goa2 is
- 61/;o + 18
omitt~d in_thfe figu;e.
* The number 486 is introduced m th1s ormu a so integers only.
61/;oº
+ 18
(29)
+/
.+/.
+/
+!
+/
+/
+!
+/
-11
N
+!
-f/
+!
+!
+/
-f/
FIG. 12.
ln the case of torsion, Eq. (19) will be replaced by the equation c/>
+!
I
(e)
fbJ Fra. 11.
+ •hº + · · · + 1/;6° -
The residuais, calculated in this way, are written to the right of each nodal point in Fig. llb. The liquidation of these residuals is begun with point a. Giving to this point a displacement i/;a' = -2 we add [see Eq. (29)] +12 to the residual ata and -2 to the residuals at all adjacent points. Thus the residual at a is liquidated and a residual -2 appears at point b. We are not concerned with residuals at the boundary, since there we have permanent supports. Considering now the point e and introducing there a displacement +2 we bring to zero the residual there and add +2 to the residuals at b, d, and e. All the remaining residuals will now be brought to zero by imposition of a dis-
1
(aJ
477
APPENDIX
THEORY OF ELASTICITY
that we may work with
placement - 2 at point f. Adding to the starting values of if; all recorded corrections, we obtain the required values of 1{;, and from Eq. (27) we obtain the values of cp. These values, divided by G8a 2, are shown in Fig. llc. They coincide with the values which can be obtained from the rigorous solution (g) on page 266. 5. Block and Group Relaxation. The operation used up to now in liquidating the residuais consisted in manipulation of single nodal points, considering the rest of the points as fixed. Sometimes it is better to move a group of nodal points simultaneously. Assume, for example, that Fig. 12 represents a portion of a square net and that we give to all points within the shaded area a displacement equal to unity while the rest of the nodal points remain fixed. W e can imagine that all nodal points of the shaded area are attached to an absolutely rigid weightless plate and that this plate is given a unit displacement, perpendicular to the plate. From considerations of equilibrium (Fig. 6),
I' 478
we conclude that the displacement described will produce changes of residuais at the end points of the strings attaching the shaded plate to the remaining portion of the net. If O and 1 denote the nodes at the ends of one string, the contributions to the residuals due to displacement Wo and W1 are Ro = -Sõ Wo õ
and
W1
Ri= Sõ Wo
1
-
W1
õ
If now we keep point 1 fixed and give to point O an additional displacement .6wo, we get the increments of the residual forces .6Ro = - S .6wo,
Introducing dimensionless quantities according to our previous notation, qõ2 R qõ2 = r,
we find .6ro = -
w=si/I
.61/lo,
and the liquidation of the residuals by subsequent point relaxation will proceed more ra pidly. Instead of giving the fictitious plate a displacement perpendicular to the plate and constant for all the nodal points attached to the plate, we ~an r~tate the plate about an axis lying in its plane. The correspondmg d1splacements of nodal points and changes of residuals can be readily calculated. So we can liquidate not only the resultant residual force sustained by the fictitious plate but also the resultant moment about any axis chosen in the plane of the plate. We can also discard the notion of the fictitious plate and assign to a group of points arbitrary selected displacements. If we have some i~ea of the shape o~ the deflection surface of the net we can select group d1splacements wh1ch may result in acceleration of the liquidation process. 6. Torsion of Bars with Multiply-connected Cross Sections. It was shown 1 that in the case of bars with multiply-connected cross sections the stress function q, must not only satisfy Eq. (4), but along the boundary of each hole we must have
- Jan
aq, ds = 2GOA
We see that unit increment in 1{10 produces changes in the residuais equal to .6ro
= -1,
M1 =
479
APPENDIX
THEORY OF ELASTICITY
+1
These changes are shown in the figure. The residuais of the rest of the nodal points of the net remain unchanged. If n denotes the number of strings attaching the shaded plate to the rest of the net, the unit displacement of the plate results in diminishing by n the resultant of the residual forces of the shaded portion of ~v~lL the net. Choosing the displacement so that the resultant vanishes we get residual forces which are self-equilibrat" _ .......~-'-'-'-'-'-lf-='/./ ing and as such lend themselves more t. .2 -4.8 readily to liquidation by subsequent FIG. 13• point relaxation of the normal kind. ln practical applications it is advantageous to alternate sequences of block displacement with sequences of point relaxation. Assume, for example, that the shaded area in Fig. 13 represents a portion of the triangular net. The number n of strings attaching this portion to the rest of the net is 16 and the resultant of the residuais shown in the figure is 8.8. Consequently an appropriate block displacement in this case will be 8.8/16 = 0.55. After such a displacement the resultant of the residual forces, acting on the shaded portion of the net, vanishes
(30)
where A denotes the area of the hole. ln using the membrane analogy the corresponding equation is
-S
aw
Jan
-ds = qA
(31)
which means that the load uniformly distributed over the area of the hole 2 is balanced by the tensile forces in the membrane. N ow applying finite-difference equations and considering a square net, we put Sõ for the tension in the strings, Wo for the deflection of the boundary of the hole, and Wi for the deflection of a nodal point i adjacent to the hole. Instead of Eq. (31) we then have
Sõ
2:
(wi
~ Wo) + qA =
O
or (32) See p. 296. The hole is represented by a weightless absolutely rigid plate which can move perpendicularly to the initial plane of the stretched membrane. 1
2
where n is the number of strings attaching the area of the hole to the rest of the net. The equilibrium equation (11) is only a particular case of Eq. (32) in which n = 4. W e can write as many equations (32) as there are holes in the cross section. These equations together with Eqs. (11) written for each nodal point of the square net are sufficient for determining the deflections of all nodal points of the net, and of all the boundaries of the holes. Consider as an example the case of a square tube, the cross section of which is represented in Fig. 14. Taking the coarse square net, shown in the figure, and considering the conditions of symmetry, we observe that it is necessary in this case to calculate only five values, a, b, e, d, and e, of the stress function. The necessaryequationswill be obtained by using Eq. (32) and the four Eqs. (11) written for the nodal e d e b a points a, b, e, d. Substituting 2GO FIG. 14. for q/S and observing that n = 20
/
ye
/
and A
=
16ó 2 we write these equations as follows:
relaxation procedure is used for all the points. Very often the points cl_ose to ,th~ boundary are connected with it by shorter strings. Due to d1fference m lengths of strings some changes in equilibrium equations (11) _and (19)_ should be introduced. The necessary changes will now be d~scusse~ m co~n~ction with the example shown in Fig. 15. A flat specimen with semicircular grooves is submitted to the action of tensile forces uni~or~y distributed over the ends. Suppose that the difference of prmc1pal stresses at each point has been determined by the photoelastic method, as explained in Chap. 5, and that we have to determine the sum of the principal stresses, which, as we have seen Frn. 15. (page 465), must satisfy the differential equation (6). For the points at the boundary one of the ~wo principal stresses is known, and using the results of the photoelastic tests, the second principal stress can be calculated, so that the sum of these two stresses along the boundary is known. Thus we have to solve the differential equation (6) the values of e/> along the boundary being known. ln using the finit~-difference method and taking a squar~ net we conclude, from symmetry, that only one-quarter of the spec1men should be considered. This portion 4/00
20e - 8b - Se - 4d = 2b - 4a = a - 4b +e+ e = b - 4c d e = 2c - 4d e =
+ + +
16 · 2GOó -2GOó 2 2 -2GOó 2 -2GOó - 2GOó 2
481
APPENDIX
THEORY OF ELASTICITY
480
4100
4JOO
460(}
S400
6700
8SOfJ
/OSfJO
1!41JO
2
These equations can be readily solved and in this way we obtain
4100
4100
S/(J(J
54/JO
8400
!!lfJO
!2JOO
A IJ600
4100
4100
4()00
J900
J900
4200
J800
JSOO
.J.SOO
2600
2
e = 1,170. 2Goó 2
488
and also the values a, b, e, and d. These values obtained with a coarse net, do not give us the stresses ' . . with sufficient accuracy and an advance to a finer net is necessary. The results of such finer calculations, made by the relaxation method, 1 can be found in Southwell's book. 7. Points Near the Boundary. ln our previous examples the nodP.l points of the net fall exactly on the boundary and the sarne standard 1
R. V. Southwell, "Relaxation Methods in Theoretical Physics,'' P· 60, Oxford
University Press, New York, 1946.
4100
J800
3200
2100
1000 Frn. 16.
with _the boundary values of cf> is shown in Fig. 16. Considering point A of this figure, we see that three strings at that point have standard length ó ~hile the fourth is shorter, say of length mó (m :::::: 0.4 in this case). Th1s n:11:1st .be taken ~nto consideration in the derivation of the equation of eqmhbrmm of pomt A. This equation must be written as follows: Só (cf>a - c/>1 ó
+ cf>a -ó
c/>2
+ cf>a
- c/>a 13
+ cf>a m/3 - cf> 4)
=
O
482
THEORY OF ELASTICITY
APPENDIX
or cf>1
+ cf>2 + "'ª + _!_m
(3
+ !) m "'ª
=
o
ln applying to point A the standard relaxation process and giving to cf>a an increment equal to unity, we will introduce the changes in residuals shown in Fig. 17a. This pattern must be used in liquidating residuals at point A. ln considering point B we see that there are two shorter strings. Denoting their lengths by mô and nõ we find that in the liquidation of residuais at B the pattern shown in Fig. 17b should be used. Introducing these changes at the points near the boundary and using the standard relaxation process at all other points the vall!es of
+1 o----<+rl-.,..(J_.f_.m_')--o+/ +J
+L m
~ n
fb) Fm. 17.
In a more general case, when we are dealing with Eq. (9) and there is externa! load at the nodal points, we denote by mô, nõ, rõ, sô the lengths of the strings at an irregular point O of a square net and we assume that the load transmitted to this point is equal to qô2
T
qô2
T
(m
+ n + r + s).
i)4cf>
iJx4
+2~
ax2 ay2
The equation of equilibrium then will be
()4> -
+ ày4
- O
(34)
and the boundary conditions (20) on p. 23 which . thº b ' , in is case, ecome a2q, Y a2q,
a2<1> àx ày
uX
àx ày
-
l-a2 - m - - =X a29 mTZ-l--=Y
'í-il''!'*)
483
With the changes discussed · thº · extended to cases in which we hm .1s artl1c1e t~e relaxation process is . ave irregu ar pomts near the boundary 8 Biharm . E . oruc quation. W e have seen ( 26) th . . of two-dimensional problema of elasticit . pate at m the case f y, m t e absence of volume forces and with · by a stress funct~~:e:, :~~:hª!a~~~~sº~~edab~yh, the s~resses a:e defined i armomc equat10n
7
(35) 8
2
6
' (m
+ n + r + s) + s [W1 + ~n + War + W4s m
-
Wo
(_!_ m
+ !n + !r + !)) = o s
(33)
For m = n = r = s = 1 this equation coincides with our previous Eq. (11) derived for a regular point. Using Eq. (33) the proper pattern, similar to those shown in Fig. 17, can be developed in each particular case. An equation, similar to Eq. (33), can be derived also for a triangular net. This example was treated by R. Weller and G. H. Shortley; see J. Applied Mecluinics, 1939, p. A-71. Since the boundary values of >, obtained from photoelastic tests, are known only with a comparatively low accuracy, the values of > at internal points are given with no more than two or three significant figures,
t
ô Knowing the forces distributed along _j_ 9 3 o ~ the boundary we may calculate
we conclude that
(:;~). = f (~~) = t. [(~:~), - (::~). + (~~).] 2
1
~ 'ô4 (6cf>o - 4<1>1 - 4q, 3 t
+
q, 5
+ cpg)
We consider here only simply connected regions.
485
APPENDIX THEORY OF ELASTICITY
484 Similarly we find
4 ô > ~ _!_ (6>o - 4
20> - 8(>1 + >2 + >a+ >4) + 2(>6 + >s + >10 + >12) - O (36) o + q, 5 + q, 1 +
statisfi;~ ~~~v:~: ;~~:~~~n~a~!::~t~ew:t~~~
This equation mfusht be the boundary o t e P1ª e. function > we integrate Eqs. (35). dy l =cosa= ds
,.
· 23) th t Observing (Fig. 20, p. a m = sin
and
we write Eqs. (35) in the following forro: 2 dy a2 q, dx a <1> = !!:__ ds ay2 + ds ax ay ds ay dx a2 q, dy a <1> _ - ds éJx2 - ds ax ay 2
and by integration we obtain _ aq, ax
.
Eq. (39) will introduce a third constant, say C, so that the final expression for > will contain a linear function Ax + By + C. Since the stress components are represented by the second derivatives of >, this linear function will not affect the stress distribution and the constants A, B, C can be taken arbitrarily. From the boundary values of > and its A e first derivatives we can calculate the approximate values of > at the nodal points of the net adjacent to the boundary, sucli Frn. 19. as points A, C, E in Fig. 19. Having, for example, at point B the values >B and (aq,/ax)B, we obtain
=
:: =
f f
_
>A = >B -
= -dx/ds
a
(ª"') x (ª"') = y =
(37)
!!:__
ds ax
Yds
(38)
Xds
To find > we use the equation aq, _ aq,dx as - ax ds
+ aq,dy ay ds
(ª"') O ÔX
B
Similar formulas can be written also for point E. A better approximation for these quantities can be obtained later when, by further calculation, the shape of the surface representing the stress function q, becomes approximately known. Having found the approximate values of > at the nodal points adjacent to the boundary, and writing for the remaining nodal point within the boundary the equations of the forro (36), we shall have a system of linear equations suffi.cient for calculating all the nodal values of
which, after integration by parts, gives >
=
aq, X ÔX
aq,
+ y éJy
-
J( asax + a2q,
X
y
a2q, ) ds éJy
(39)
ÔS
Substituting in this equation the values of thedderivat;ves g;v;n bÍt E (37) d (38) we can calculate the boun ary va ues o . sh~~ld be ::ted th~t in calculating the first derivatives \38), twt~ co~ stants of integration, say A and B' will appear and the mtegra wn m
1 This is one of many cases discussed by P. M. Varvak, "Collection of Papers on Structural Mechanics,'' Kiev Structural Institute, vol. 3, p. 143, 1936 (Russian). A solution, also numerical, of a similar problem is given in the book by K. Beyer, "Die Statik im Eisenbetonbau,'' 2d ed., vol. 2, p. 733, Berlin, 1934. 2 The system is obtained by rotating clockwise by 'Ir the axes used in Fig. 20,
p. 23.
The constants of integration will be calculated from the conditions that for the point x = 0.4a, the common point of the two parts of the boundary, the values of q, and aq,/ax must have the sarne values for both parts. Hence
and integration gives aq, -=A, ax
q,
=
+ B,
Ax
h ·s which as mentioned A B C are constants along t e x-axi ' ' O H ere ' ' b. ·1 We assume A = B = C = . before can be chosen ar itrari y. .d f the Then ~ vanishes along the unloaded portion of the bottom s1 e o J
y
~1
2
0.8a'L-~-1-~-+-i, 1
11111
11
111
lllll!P
i-0.728 'tf/.óB +J.IB +J.óB 2
1 -4.BB -a72B 6
s
4
8
7
li
'º
14
/J
3
2
-4 BB -a72B 15"
~JS·
O,
and we find C1 = 1.6pa,
C2 = -0.32pa2
The stress function
ª
= -0.4pa,
"' = -0.4pax FIG. 20.
t f q, with respect to the y-axis. if rml distributed load of plate, which ensures the symme ry o From x = 0.4a to x = 0.5a there acts a un o y intensity 4p and Eqs. (38) give
-J
4pdx = -4px aq, iJy
+ C1
=o
The second integration gives q, = -2px2 + Cix
(b)
ª
ay
=o
(e)
From this it follows that c/J remains constant along the vertical side of the plate. This constant must be equal to -0.02pa 2, as calculated above for the lower comer. Along the unloaded portion of the upper side of the plate the first derivatives of cp remain constant and will have the sarne values (e) as calculated for the upper comer. Thus the stress function will be
.
aq, ax -
-0.4pa
Along the vertical side of the plate there are no forces applied and, from Eqs. (38), we conclude that along this side the values of aq,/dx and of aq,/dy must be the same as those at the lower comer, i.e.,
t/Ji. -4.BB -0.728 12
+ C1)-o.4a =
(4')z=0.5a = -0.02pa 2,
6
-4.88 i-a728 9
( -4px
At the comer of the plate we obtain
'
-4.88 -a72B J
4'
487
APPENDIX
THEORY OF ELASTICITY
486
+e
Since at the upper comer to the left c/J must have the previously calculated value, equal to -0.02pa 2, we conclude that C = O.l8pa 2 and the stress function is (d)
+ C2 ltiSTtTUTUL PUll'TEMm T l ~ ' i _: O .\ R A atBUOTPr.~ f'~NTnA 1 l
THEORY OF ELASTICITY 488 For x = 0.4a these values must coincide with the values (e). C 1 = C2 = O, and the stress function must have the forro
489
APPI!JNDIX
Hence,
equations! we have t? wr_ite in this symmetrical case the Eqs. (36) for the 15 pomts shown m Fig. 20. The solution of these equations gives for > the values shown in the table below.
px2
-2+ª
1
For x = 0.4a it must have a value equal to that obtained from Eq. (d). We conclude that C = O.lpa 2 and px2
q, = - -y The stress function is respect to the y-axis. values of q, and its first the boundary all these With the notation
+ O.lpa2
(e)
represented by a parabola symmetrical with This finishes the calculation of the boundary derivatives, since for the right-hand portion of values are obtained from symmetry.
-
2
3
4
Fig. 20. Next, by extrapolation, we calculate the values of > for the nodal points taken outside the boundary. Starting again with the bottom side of the plate and observing that aq,/ay vanishes along this side we can take for the outside points the sarne values >1a, >14, >15 as for the inside points adj acent to the boundary . 1 We proceed similarly along the upper side of the plate. Along the vertical side of the plate we have the slope
-0.4pa and we can, as an approximation, obtain the values for the outside points by subtracting the quantity
0.4pa · 2ó
=
2
0.4:tt
=
4.8B
from the inside points adjacent to the boundary, as shown in Fig. 20. N ow we can start the calculation of q, values for the inside nodal points of the net. Using the method of direct solution of the difference i
This manner of extrapolation, used in Varvak's paper, is different from that
described on p. 485.
6
7
8
9
10
11
12
13
14
15
- - - - -- - -
.p/B 3.356 2.885 1.482 2.906 2.512 1.311 2.306 2.024 1.097 1.531 1.381 0.800 0.634 0.608 0.396
Le:t us calculate _the normal stress u,, along the y-axis. The values of tl_lls stress are g1ve:°" by the second derivative a2q,/ay2. Using finite d1fferences we obtam for the upper point (y = a) (u,,)y_, ,,,,, (3.356 - 2 · 3.600 ó2
+ 3.356)B
0.488pa2 36ó2
-0.488p
For the lower point (y = O) we find (
we can now write all the calculated boundary values of > as shown in
5
-- -- ---- -- -- -- -- -- -- -
+ 0.634)B =
(0.634 - O 02
l.268p
I_f we c?ns~der. the plate as a beam on two supports and assume a lmear d1stnbut10n of u,, over the middle cross section (x = O) we find (u,,)max. = 0.60p. We can see that for a plate of such proportions the usual beam formula gives a very unsatisfactory result. To solve the finite-difference equations (36) by iteration, we assume some starting values q, 1, q, 2, • • • , >u for the stress function. Substituting these into Eqs. (36) we obtain residual forces for ali interna! nodal points which can be liquidated by a relaxation process. The proper pattern, as obtained from Eq. (36), Frn. 21. is shown in Fig. 21, in which the ch~nges in residuais due to unit change of q, 0 are given. Jn applying this method to the square plate discussed above it must be observed that .t~e > val~es along the boundary are restricted by the boundarv condit10ns, wh1ch means that the residual forces at points on th~ boundary need not be liquidated. W e can next advance to a finer net, obtaining starting values of q, from the results of the calculation on the coarse net.
490
THEORY OF ELASTICITY
APPENDIX
ln the case of a nonsymmetrical loading such as shown in Fig. 22a, we can split the load as shown in Figs. 22b and 22c into symmetrical and antisymmetrical loadings. ln both latter cases we have to consider one-half of the plate, since q,(x,y) = q,( -x,y) for the symmetrical case and q,(x,y) -q,( -x,y) for the antisymmetrical loading. y
491 we a rigorous solution of the problem and in t• l t ' prac ica cases we must u 11 sua. y resor to approximate methods. Usmg the finite-difference method we shall tak sidering a nodal oint O F. ' e a square net. Conin Eq (40) b F' ig. 24, we can treat the second derivatives . as e ore. or the first derivative we can take
r
( ôq,)
,,,
Ôrr=ro
l
2
+
('1>1 - tl>o Ó
tf>o - q, 3) = q, 1 - q, 3 Ó
?,ó
a (b)
(a)
(cJ
4
r
Frn. 22.
ô
The work can be further reduced by considering also the horizontal axis of symmetry of the rectangular plate. The load shown in Fig. 20 can be resolved into symmetrical and antisymmetrical cases as shown in Fig. 23. For each of these cases only one-quarter of the plate should be considered m calculating numerical values of the stress function.
b
z
z
FIG. 24.
FIG. 25.
Then the finite-difference equation, corresponding to Eq. (40), is '1>1 -t----+---.x '
i,
9. Torsion of Circular Shafts of Variable Diameter. In this case, as we have seen (page 307), it is necessary to find a stress function wLlch satisfies the differential equation
a2 q, ôr 2 -
-
;r~ (f/>1
4f/>o -
- f/>3)
=
O
(41)
n::
i
FIG. 23.
1,
+ '1>2 + f/>3 + 'Í>4
3 aq,
a2 q,
r ôr + ÔZ 2 = o
(40)
at every point of the axial section of the shaft, Fig. 24, and is constant 1J.long the boundary of that section. Only in a few simple cases have
The problem then is to find such a set of 1 be satisfied at every nodal point of the ::~ of '1>.that Eq. (41) will assumed constant value at the bound . > will be equal to the either by direct solution of Eqs. (41) :;~·y oTnhe1sfptrhob.ltem ct~n be treated A o e l era 10n methods . s an exa~ple, let us consider the case shown in Fig 25 l th . ::;:;:d~!t~ap1~ change in the .
3 dq,
-
r dr =o
'42)
The general solution of this equation is
= Ar4
+B
(43)
492
THEORY OF ELASTICITY APPENDIX
\ 1
and the corresponding stresses are (see page 306)
l 2 dq,
r u= •
r dr
= 4Ar
Tr6
'
:·
S~
= Ü
To
(1 _~) 2ro
W1 õ
Comparing this result with the Coulomb solution, we find W1 - 2wo 4A
= M1 lp
_Mi. 4
(44)
These expressions vanish at the axis of the shaft and assume at the boundary a common value M t/211". Since
+
2
(45)
0
This equation is of the sarne forro as that for deflections to a cylindrical forro of a membrane with tension J--ó' varying inversely as r 3• To show this let us consider three consecutive points of the net, Fig. 26. The corresponding deflections we denote .J Frn. 26. by Wa, Wo, W1. The tension at the middle of the strings 3-0 and 0-1 will be
d'__,
E=r:7r Sõ
,,,,,
So(
(ro _ ~y ro :
: \
3
1
+ So (l + ~) w 3 -
Wo
roª
+ Wa
2ro
23õTo (w1 -
-
wa)
Wo
0
=
=O
O
This is the sarne as Eq. (45).
where M 1 is the applied torque and IP is the polar moment of inertia of the shaft. Omitting the constant B in the general solution (43) as having no effect on the stress distribution, we find for the stress function at suffi.cient distances from the fillet the expressions
>1
493
The equation of equilibrium for the point O is then
+
3õ) 2ro
b
Sim~arly, in the general case, observing that the tension in the me rane oes not depend on z, we obtain the equation mW1
+ W2 + Wa + W4
- 4wo -
;r~
(w1 - wa) =O
(46)
which agrees with Eq (41) It · h function as the defle. t" . f is seen t at w~ can calculate the stress c IOn o a membrane w1th if . having constant deflection Mi/~ along the b dnonun dorm ten~10n (44) at th · t 1 oun ary an deflect10ns . e pom s at arge distances from the fillets W startmg values for w at the d 1 . · e assume some . hand sides of Eqs (46) d nol a lpomts, su~st1tute them into the left. . ' an ca cu ate the residuais N ow th bl is to liquidate ai! these residuais by the rei t• . e pro em 26 th b axa wn process From Fig we see at Y giving to point O adis 1 · · · residuais at points 1 and 3 the quantitiesp acement umty we add to the
s
To 3
(1
-
3õ)
2r0
and
.§__3
(1 + 3õ)
2r0 which indicates that th tt f in Fig 27 It . f e pa e;n or the relaxation process is as shown · · varies rom pomt t o pom · t wit · h vanation . of the radial To
4/
---~--1--41--~--1-'1!!.. 'O
2r0
+! Frn. 27.
distance ro. Calculations of this kind were carried out by R. V. Southwell and D. N. de G. Allen.1
and i Proc. Roy. Soe. (London) series A ol Well's book "Relaxation Meth d 8 . Th' v .. 183, pp. 125-134. See also South0 m eoret1cal Physics," p. 152.
'
1
AUTHORINDEX A
Cauchy, A. L., 446 Cheng, D. H., 434 Chree, C., 69, 354, 384, 439 Churchill, R. V., 49 Chwalla, E., 172 Clapeyron, B. P. E., 306 Clebsch, A., 241, 441, 446 Close, L. J., 376 Coker, E. G., 120, 123, 131, 204, 211 Comu, M. A., 254 Coulomb, C. A., 258 Courant, R., 461 Cox, H. L., 172 Craemer, H., 53 Craggs, J. W., 390 Cross, Hardy, 469 Cushman, P. A., 292, 337
Adams, L. H., 378, 411 Airy, G. B., 26 Allen, D. N. de G., 493 Almansi, E., 433 Anderson, E. W., 280 Anthes, 268 Arndt, w., 310 B
Babinet, 441 Barjansky, A., 211 Barker, L. H., 413 Barton, M. V., 150, 388, 390 Basu, N. M., 284 Bay, H., 50 Belajef, N. M., 380 Bemdt, G., 377 Beschkine (Beskin) L., 172 Beyer, K., 53, 485 Bickley, W. G., 83, 465 Biezeno, C. B., 292 Billevicz, V., 64 Biot, M. A., 51, 430 Bisshopp, K. E., 73 Bleich, F., 46, 52 Borchardt, c. w., 417, 434 Bornitz, G., 456 Boussinesq, J., 85, 101, 235, 292, 362, 368, 446, 451 Brahtz, J. H. A., 98 Bredt, R., 299 Brewster, D., 131, 143 Byerly, 49, 410
D Davidenkoff, N., 427 Davies, R. M., 439 Den Hartog, J. P., 143, 406 Dinnik, A., 278, 310, 376, 384, 410, 411 Donath, M., 73 Donnell, L. H., 84, 204, 446 Dougall, J., 341 Drucker, D. C., 143 Duhamel, J. M. C., 408, 416 Duncan, W. J., 286, 336
E Eichelberg, G., 414 Elliott, H. A., 201 Ellis, D. L., 336
e
F Fillunger, P., 45, 125 Filon, L. N. G., 46, 50, 101, 120, 122, 123, 131, 204, 211, 238, 267, 280, 384, 387
Calisev, K. A., 469 Carothers, S. D., 45, 95 Castigliano, A., 162 1'
495
Fischer, A., 73 Flamant, 85, 449 Fõppl, A., 63, 217, 278, 287, 304 Fõppl, L., 217, 278, 307, 310 Forsyth, A. R., 347, 385 Frocht, M. M., 131, 140, 142, 206 Fuchs, S., 376 G
,11
!'
AUTHOR INDEX
THEORY OF ELASTICITY
496
Galerkin, B., 159, 235, 280, 331 Girkmann, K., 102 Gõhner, O., 391, 395, 398 Golovin, H., 63, 73, 78 Goodier, J. N., 50, 52, 84, 109, 120, 150, 171, 191, 257, 258, 319, 359, 405, 434 Goursat, E., 185 Grammel, R., 73 Grashof, F., 43, 63, 107, 329 Gray, A., 385 Green, A. E., 85, 204, 212 Greenhill, A. G., 278, 280, 292 Greenspan, M., 212 Griffin, H. K., 411 Griffith, A. A., 161, 268, 291, 294, 297, 336 Grübler, M., 73 Grunberg, G., 419 Gunder, D. F., 331
H Hadji-Argyris, J., 177 Hamburger, M., 384 Hearle, H., 73 Heinrich, A., 456 Hencky, H., 461 Hengst, H., 83 Hertz, H., 109, 372, 377, 379, 452 Hetényi, M., 144, 145 Higgins, T. J., 280, 290, 307 Hodkinson, B., 73 Hoersch, V. A., 380 Hoff, N. J., 20, 150 Holl, D. L., 280, 331 Holzer, H., 72 Honegger, E., 419 Hopkinson, J., 417
Hosokawa, Y., 123 Howland, R. C. J., 83, 102, 115 Hsu, M. C., 84 Huber, M. T., 149, 376 Ruth, J. H., 289, 301
Lode, W., 150 Lorenz, R., 412 Love, A. E. H., 115, 120, 230, 241, 280, 331, 342, 346, 351, 362, 368 Lundberg, G., 380 Lyse, I., 289
497
o Odqvist, K. G., 380 Osgood, W. F., 49 Osgood, W. R., 44 p
I
M Ingberg, S. H., 411 Ingersoll, L. R., 436 Inglis, C. E., 98, 170, 198
J Jacobsen, L. S., 310 Jeffery, G. B., 60, 82, 208, 211 Jing, K. S., 83 Johnston, B. G., 289
K Kappus, R., 336 Karelitz, G. B., 294 Kármán, T. von, 43, 102, 172, 173 Kelvin, Lord, 235, 292, 356 Kelvin and Tait, 274, 292, 352 Kent, C. H., 416 Kirchhoff, G., 238 Kirsch, G., 80 Klitchieff, J. M., 342 Knight, R. e., 83, 84 Kolosoff, G., 189, 196, 198, 280 Korobov, A., 352 L
Lamb, H., 102, 368, 459 Lamé, G., 59, 192, 306, 359 Larmor, J., 293, 362 Laue, M. V., 427 Lees, C. H., 410, 411 Leibenson, L. S., 286, 335 Lessells, J. M., 416 Letson, H. F. G., 414 Leven, M. M., 131, 144 Levy, M., 45, 125 Levy, s., 82 Liebman, H., 461 Ling, C. B., 83, 84, 89, 211, 280
MacMillan, W. D., 434 McPherson, A. E., 82 March, H. W., 287, 290 Marcus, H., 461 Marguerre, K., 51 Martin, H. M., 73 Mathar, J., 123 Mathews, G. B., 385 Maunsell, F. G., 89 Maxwell, J. C., 131 Meinesz, V., 336 Melan, E., 115, 116, 211, 310 Mesmer, G., 50 Mesnager, A., 29, 97, 131, 142 Michell, J. H., 85, 93, 97, 109, 111, 116, 120, 238, 307, 342, 346, 351 Mindlin, R. D., 82, 143, 191, 211, 235, 362, 377, 434 Mintrop, L., 456 Mises, R. von, 149 Miura, A., 98 Mohr, O., 14, 93, 217 Morkovin, V., 212 Morton, W. B., 376 Muller, I., 45 Murphy, G., 20 Muscheliiívili, N., 185, 204, 212, 280, 430 Myklestad, N. O., 434
N Nádái, A., 272, 388 Nãgel, A., 414 Navier, 258 Neményi, P., 238 Neuber, H., 204-206, 235, 307 Neumann, F. E., 416 Nishimura, G., 84
Papkovitch, P. F., 235 Paschoud, M., 292 Pearson, K., 43, 45, 63, 331, 342 Peterson, R. E., 206 Petrenko, S. N., 379 Pickett, G., 170, 387 Pigeaud, M., 50 Pilgram, M., 391 Pochhammer, L., 341, 384, 439 Poisson, S. D., 446 Pollard, C., 45 Polya, C., 267 Poritsky, H., 89, 381 Põschl, T., 198, 211, 307, 452 Prandtl, L., 63, 261, 268, 272 Prescott, J., 439 Prowse, W. A., 452 Purser, F., 302
Q Quest, H., 292
R Rademaker, J. M., 292 Ramsauer, e., 452 Rankin, A. W., 388 Rankine, W. J. M., 43, 107 Rayleigh, Lord, 383, 402, 438, 456 Reíssner, E., 125, 172, 177, 328, 456 Reissner, H., 122, 427 Résal, H., 63 Ribiêre, C., 63, 78 Richardson, L. F., 45, 461 Ritz, W., 158, 159, 281, 282 Robinson, w. e., 411 Rock, D. H., 331 Roever, V., 391 Runge, C., 461
THEORY OF ELAS7'ICITY
498
s Sadowsky, M., 83---85, 95, 96, 359 Sagoci, H. F., 456 Saint-Venant, 33, 150, 259, 265, 267, 278-280, 316,326, 446, 449, 450 Scblechtweg, H., 73 Scbleicher, F., 149, 368 Schnadel, G., 172 Schulz, K. J., 83 Scruton, C., 336 Searle, G. F. C., 146 Sears, J. E., 384, 452 Seegar, M., 331 Seewald, F., 45, 50, 102 Sen, B., 211 Seth, B. R., 331 Sezawa, K., 84 Shepherd, W. M., 98, 331 Shortley, G. H., 482 Smekal, Adolf, 161 Smith, F. C., 82 Sneddon, I. N ., 201 Snively, H. D., 89 Sokolnikoff, I. S., 212, 230, 280, 331 Solakian, A. G., 294 Sonntag, R., 309, 310 Southwell, R. V., 336, 359, 461, 465, 468, 469, 480, 493 Sternberg, E., 85, 359 Stevenson, A. C., 83, 198, 202, 331 Stodola, A., 69, 72, 411 Stokes, G. G., 99 Straubel, R., 254 Strauch, F., 122 Symonds, P. S., 204
T
!
! ,
Takemura, K., 123 Taylor, G. I., 268, 291, 294, 297, 336 Terazawa, K., 368 Thibodeau, W. E., 84 Thomas, G. B., 328 Thomas, H. R., 380 Timoshenko, S., 8, 44, 60, 77, 81, 82, 121, 123, 149, 155, 166, 167, 206, 273, 283, 302-304, 319, 334, 336, 384, 415, 416, 459
Timpe, A., 29, 44, 50, 63, 116, 120, 204, 302, 310, 352 Todhunter and Pearson, 63, 235, 259 Tranter, C. J., 125, 390 Trayer, C. W., 287, 290 Trefftz, E., 272, 282, 284, 319 Tuzi, Z., 83
SUBJECT INDEX A
V Varvak, P. M., 485 Vlasov, V. Z., 150 Voight, W., 452 Volterra, V., 120, 238
w Wagstaff, J. E. P., 452 Wahl, A. M., 206, 391 Wang, C. K., 83 Wang, P. S., 83 Watson, G. N., 157, 179 Way, S., 368 Weber, C., 211, 212, 268, 302, 346 Weibel, E. E., 139, 140, 430 Weigand, A., 287, 310 Weinel, E., 211 Weinstein, A., 336 Weller, R., 482 Westergaard, H. M., 217 Whittaker, E. T., 157, 179 Whittemore, H. L., 379 Wieghardt, K., 121 Wigglesworth, L. A., 331 Willers, F. A., 310 Williamson, E. D., 411 Wilson, C., 85, 99, 131 Wilson, R. E., 411 Wilson, T. L., 81 Winkler, E., 63 Wolf, F., 461 Wolf, K., 45, 83, 204, 302 Wood, L. A., 84 Wylie, C. R., 89
y Young, D. H., 273 Young, Thomas, 442
z Zobel, O. J., 436
Additional curvature of beams, 106 Additional deflection due to shearing force, 39, 106 Analytic functions, 181 Analyzer, 132 Angle of twist per unit length, 249, 259, 264, 273,280, 287, 295, 299 Anisotropy, 1 Annealing, remova! of stresses by, 427 Anticlastic surface, 254 Axes, principal, of strain, 224 of stress, 14, 215 Axially symmetrical stress distribution, 343 in cylinder, 384 B
Balls, compression of, 372 Beams, bending of (see Bending, of beams) concentrated force acting on, 99 continuously distributed load on 39 44, 49, 342 , ' curvature of, 43, 106, 340 additional, 106 (See also Curvature of beams) deflection of, 38, 42, 43, 340 distortion of cross sections of, 38, 341 shearing stress in, 36, 41, 44, 104, 320, 322,326,329,330,332 Bending, of beams, of narrow rectangular cross section, by concentrated force, 99 by distributed load, 44, 46 by own weight, 42, 53, 342 by terminal load, 35 by uniform load, 39 of particular forros of section, circular, 319
Bending, of beams, of particular forros of section, elliptic, 321 other forros, 329, 331, 332 rectangular, 323 solution of problem of, by soapfilm method, 336 'houndary conditions in, 317 of circular cylinder, 319 of circular plate, 349 of curved bar, by force at end, 73 cylindrical bar of any cross section in, 250, 316 of prismatical bars, 316 determination of displacements in, 340 pure (see Pure bending) ring section in, 395 semi-inverse method in, 316 triangular prism in, 331 Bending moment, relation of, to curvature, 340 Body forces, 3 Boundary conditions, in bending, 317 in terms of displacements, 234 in three-dimensional problems, 229 in torsion, 260 in two-dimensional problems, 22
e Cantilever, deflection of, 38, 340 distortion of cross section of, 38, 341 of particular forros of cross section, circular, 319 elliptic, 321 other forros, 329 rectangular, 323 narrow, 35 stresses in, 35, 316 Castigliano's theorem, 162 Cauchy-Riemann equations, 181 499
500
THEORY OF ELASTICITY
Cavity, ellipsoidal, 235 spherical, 359 Center, of compression, 358 flexural, 334 shear, 334 of twist, 271, 336 Central line, extension of, 43, 342 of prismatie bar, 316 Circular cylinder, bending of, 319 strained symmetrically, 384 Circular disk, under forces in its plane, 107 rotating, 69, 352 thermal stresses in, 406 Circular bole, effect of, on stresses, in plates, 78 in rotating disk, 71 in shaft, 293 Circular plate, bending of, 349 Circular ring, compressed by two opposite forces, 121 general solution for, 116 initial stresses in, 120 Circular shaft, in torsion, 249 of variable cross section, 304, 490 Coefficient of thermal expansion, 399 Cold-drawn tubes, residual stresses in, 427 Compatibility conditions, in cylindrical coordinates, 344 in three-dimensional problems, 230 in two-dimensional problems, 23 Complex potentials, 187 Complex variable, functions of, 179 Components, of strain, 5, 6 of stress, 4 in curvilinear coordinates, 195 Compression, of balis, 372 by concentrated force, of disk, 107 of rectangular plate, 49, 51 of wedge, 96 of rollers, 381 of solid bodies, 377 Concentrated force (see Transmission) Concentration of stress at hole (see Hole) Conditions of compatibility (see Compatibility conditions)
SUBJECT INDEX
Conduction of heat, stresses dueto, 402, 410,412,420,427,436 Cone in torsion, 309 Constants, elastic, 6, 7 (See also Hooke's law) Contact, surface of, 373 Contact stresses between bodies in compression, 372, 377 Continuously distributed load on beam, 39, 44, 49, 342 Cooling, nonuniform, stresses due to, in plates, 402 in shafts, 411 Coordinates, bipolar, 206 curvilinear orthogonal, 192 cylindrical, 305, 306, 343 elliptic, 193 polar, 55 spherical, 346 Cracks, reduction of strength dueto, 161 in tension member, 184 in torsion member, 294 Curvatura of beams, 43, 340, 342 additional, 106 effect of shearing force on, 43 relation of, to bending moment, 340, 342 Curved bar, bending of, by force at end, 73 pure, 61, 395 deflections of, 66, 76 stresses in, 63, 75 Curvilinear orthogonal coordinates, 192 components of stress in terms of, 195 Cylinder, band of pressure on, 388 circular, with eccentric bore, 60, 208 symmetrical deformation in, 384 thermal stresses in, 408, 427, 436 thick, under pressure, 59 Cylindrical body of any cross section, in bending, 250, 316 in tension, 245, 246 in torsion, 258 Cylindrical coordinates, 305, 306, 343
D Dams, stresses in, 45 Deflection, additional, due to sliearing force, 39, 106
501
Deflections of foundations, 371 Energy, of elastic waves, 442, 456 (See also Beams) of strain, 146 Deformation, homogeneous, 219 Equilibrium, general equations of, 228 irrotational, 453 in case of nonuniform heating 421 plastic (see Plastic deformation) 423 , , Differential equations of equilibrium, in cylindrical coordinates, 306, 343 in terms of displacements, 233 in polar coordinates, 55 in three dimensions, 229, 306, 343 in terms of displacements, 233 in two dimensions, 21, 55 Expansion, thermal, 399 Dilatation, waves of, 452 volume, 9 Disk, rotating, 69, 352 Extensional vibrations of rods, 438 thermal stresses in, 406 Eyebar, stress in, 122 of variable thickness, 72 Displacements, determination of, in F bending of prismatical bars, 340 in semi-infinite body, 365 Failure, stress at, 149 in semi-infinite plates, 89, 95 Fatigue cracks, 81 in three-dimensional problems, 232 Fillets, stress concentration at, in bendin torsion of prismatical bars, 259 ing and tension of plates, 140, in two-dimensional problems, 34, 142 36, 42, 66, 76 in shafts of variable diameter, 310 Distortion, of cross sections, of bent in torsion of prismatical bars 288 beam, 38, 341 299 , , of twisted bar (warping), 259, 265 Film (see Soap-film method) strain energy of, 149 Finite-difference equations, 461 waves of, 452 boundary conditions, 485, 488 Double force, acting on infinite body, 356 points near boundary, 480, 485 acting on plate, 114 Flanges, effective width of, 171 Flexural center, 334 E (See also Center) Earthquake, waves of, 456 Flexural rigidity of plates, 256 Effect, of circular hole on stress disFoundations, deflection of, 371 tribution, 78 pressure distribution on, 371 of shearing force on deflection, 39, Fourier series, application of, in bend43, 102 ing problems, 324 Effective width of wide beam flanges, deflection curve represented by, 171 155 Elasticity, 1 in torsional problems, 275, 284 Electric current, analogy with twisted in two-dimensional problems, 46, shaft of variable diameter, 310 53, 117 Electric-resistance strain gauge, 19 Fringe value, 134 Ellipsoid, stress, 215 Elliptic coordinates, 193 G two-dimensional problems in, 197 Elliptic cylinders, under bending, 321 General equations of equilibrium, in under torsion, 263 case of varying temperature, 421, Elliptic hole, in plate, 84, 197, 201 423 in twisted shaft, 294 in cylindrical coordinates, 306, 343
General solution, for displacements, 235 of two-dimensional problems, in polar coordinates, 116 for a wedge, 123 Groove, hyperbolic, 235 semicircular, in shaft, 268, 293 H
_Harmonic function, 182 Heat fiow (see Conduction of heat) Heating, stresses due to nonuniform (see Thermal stress) Helical springs, stresses in, 391 Hertz' problem, 372, 377 Roles, circular, in plate, 78 eccentric, 60, 208, 211 elliptic, 84, 197, 201 stress concentration at, 80, 81, 84 in twisted shaft, 293, 294 of various shapes in plate, 212 Hollow shaft, torsion of, 294 Homogeneous deformation, 219 Homogeneous material, 1 Hooke's law, 6 Hydrodynamical analogies, 191, 292 Hyperbolic groove, 235 Hyperbolic notches, 204
503
SUBJECT INDEX
THEORY OF ELASTICITY
502
Isochromatic lines, 134 Isoclinic lines, 134 Isotropic materials, 1 L
Mohr's circle, 14 Multiply-connected bodies, 120, 238, 301, 479
N
Lamé's problems, 59, 359 Laplace's equation, 181, 182 Layer, elastic, on rigid smooth base, 51 Least work, application of, 167 principle of, 166 Light, polarized, measurement of stresses by means of, 131 Lines, isochromatic, 134 isoclinic, 134 of shearing stress in torsion, 270 Local stresses, at circular hole, 80, 81 at fillets, in bending and tension, 140, 142 in torsion, 238, 299 at spherical cavity, 359 Localized character of stress distribution at hole, 81, 359 Longitudinal vibrations of prisms, 438 Longitudinal waves, 438, 454 M
Membrane analogy, application of, in photoelasticity, 143 with bent beam, 319 I determination of stresses, in beams by,336 Identical relations between strain comof rectangular cross section, 329 ponents (see Compatibility condiwith twisted shaft, 268 tions) measurement of torsion stresses by, Impact, duration of, 384, 450 290 longitudinal, of bars, 444 solution of stresses, in thin twisted with rounded ends, 452 tubes by, 298 of spheres, 383 in various forms of cross section stresses produced by, 442, 449 by, narrow rectangular, 272 transverse, of bars, 384 rolled profile, 287 Inclusions, 84 use of, in calculation of torque, 271 Inftuence line, 91 Mesh side, 462 Initial stress, 68, 120, 238 Middle plane of plate, 351 general equations for determination Modulus, of rigidity, 9 of, 425 in shear, 9 in glass plates, 427 in tension, 6 in rings, 68, 120 of volume expansion, 10 Invadants of stress, 217 Young's, 10 Irrotational deformation, 453
Net, hexagonal, 473 square, 462 triangular, 4 73 Neutral surface, 42 Newton's interpolation formula, 464 Nicol prism, use of, in photoelastic work, 137 Notches, hyperbolic, 204 semicircular, 89, 212, 481
o Orthogonal curvilinear coordinates, 192 p
Photoelasticity, method of stress measurement by, 131 three-dimensional, 143 Plane harmonic functions, 182 Plane strain, 11 compatibility equations for, 23 Plane stress, 11, 216, 241 equations of equilibrium for, 21 stress at point in case of, 13 stress function for, 26 Plane waves, 454 Planes, principal, of stress, 14, 215 Plastic deformation, initial stresses due to, 425 in twisted shafts, 272 Plates, bent by couples, 255 circular, symmetrically loaded, 349 quarter-wave, 136 Poisson's ratio, 7 determination of, 254 Polar coordinates, in three-dimensional problems, 346 in two-dimensional problema, 55 Polariscope, circular, 135 plane, 132 Polarized light, use of, in stress measurements, 131
Polarizer, 132 Polynomials, solution by, of axially symmetrical stress-djstribution problems, 347 of torsional problems, 265, 286 of two-dimensional problems, 29 Potential energy, 153 minimum of, 153 Pressure, distributed over surface of contact, 374, 375, 379 hydrostatic, 10 in spherical container, 356 stresses produced by, 246 in thick cylinder, 59 produced by rigid die, 96, 371 Principal axes, of strain, 224 of stress, 14, 215 Principal directions, 14 Principal planes, 215, 224 Principal strain, 19, 224 Principal stress, 14, 214 determination of, 142, 217 Principie, of least work, 166 of Saint-Venant, 33, 150 of superposition, 203 of virtual work, 151 Prism, bending of, 316 torsion of, 258 vibrations of, longitudinal, 438 Propagation of waves, in bars, 438 in solids, 452 over surface of body, 456 Pure bending, of curved bars, 61, 395 of plates, 255 of prismatical bars, 250 Pure shear, 8
R Radial displacement, 65 Radial strain, 65 Radial stress, 55 distribution of, 85 Rayleigh waves, 456 Reciprocai theorem, 239 Rectangular bar, in bending, 323 in torsion, 272, 275, 303 Reentrant corner, stress concentration at, in I-beams, 339
504
THEORY OF ELASTICITY
Reentrant comer, in plates, 140 in shafts, 288, 294 in tubes, 299 Relaxation, block and group, 477 Relaxation method, 468 Residual stresses, in cold-drawn tubes, 427 ( See also Initial stress) Residuais, 469 Rigid-body displacement, superposable upon displacement determined by strain, 233 Rigidity, modulus of, 9 torsional, 264 Ring (see Circular ring) Ring sector, in bending, 395 in twist, 391 Rolled profile sections, in torsion, 287 Rollers, compression of, 107, 381 Rotating disk (see Disk) Rotation, components of, 225 Rupture, hypothesis concerning conditions of, 149
s Saint-Venant, principie of, 33, 150 problem of, 259, 316 Sector of circle, torsional problem for, 278 Seismograph, 456 Semicircular notches, 89, 212, 481 Semi-infinite body, boundary of, concentrated force on, 362 distributed load on, 366 Semi-infinite plate, 85, 91 Semi-inverse method, in bending, 316 in torsion, 259 of shafts of variable diameter, 306 Shaft, torsion of (see Torsion) of variable diameter, 304 Shear, pure, 8 Shear center, 334, 336 Shear lag, 177 Shearing strain, 5, 223 Shearing stress, 3 components of, 4 distribution of, in beams of narrow rectangular cross section, 51
Shearing stress, distribution of, in beams of particular forms of sections, circular, 320 elliptic, 322 I-beams, 339 rectangular, 326 effect of, on deflection of beams, 39, 106 lines of, 270 maximum, 15, 218 Single-valued displacements, 68, 119 Soap-film method, solution by, of bending problems, 336 of torsional problems, 289 Solution, uniqueness of, 236 Spherical bodies under compression, 372 Spherical cavity in infinite solid, 359 Spherical container, under externai or internai pressure, 356 thermal stresses in, 419 Spherical coordinates, 346 Strain, plane, 11, 23 at point, 17, 221 compatibility of, 23, 230, 344 components of, 5, 223 identical relations between components of (see compatibility of, above) Mohr circle of, 19, 20 in polar coordinates, 65 principal axes of, 224 principal planes of, 224 surface, measurement of, 19 radial, 65 Strain energy, 146 of volume change and distortion, 149 Strain-energy methods, 146 applications of, 157, 167, 280 Strain gauge, electric-resistance, 19 Strain rosette, 20 Stream function, 292 Strength theory for brittle materiais, 161 potential energy as basis of, 149 Stress, axially symmetricaJ. distribution of, 343 components of, 4, 55 in terms of curvilinear orthogonal coordinates, 192
rrf ·.~
LL ..
SUBJECT INDEX Stress, due to temperature change, 399 in eyebar, 122 at failure, 149 invariants of, 217 measureroent of, by photoelasticity method, 131 normal and tangential, 3 plane (see Plane stress) at point, 13, 213 principal, 14, 214 determination of, 142, 217 radial, 55 distribution of, 85 shearing (see Shearing stress) Stress concentration, at circular hole, 80, 81 at elliptic hole, 84 at fillet of shaft of variable diameter 310 ' at fillets in tension merober, 140 at reentrant corner in torsion 288 299 ' ' (See also Reentrant comer) at spherical cavity, 359 use of hydrodynaroical analogy in deterroination of, 292 Stress-director surface, 215 Stress ellipsoid, 215 Stress functions, 26 56 86 89 116 183 ' ' ' ' ' ' 261, 318, 343, 483 Stress-optical coefficient, 133 Stress-strain relation (see Hooke's law) String, deflection of, 153 Superposition, principie of, 235 Surface of contact, 373 Surface energy, 161 Surface forces, 3 Syroroetrical and antisymroetrical loadings, 490 Syrometrical stress distribution about . axis, in circular cylinder, 384 ~n three diroensions, 343 m two diroensions, 58
T Teroperature, stresses due to nonuniform distribution of, 399
__ Li
•iil!I505
Teroperature fluctuation, stresses produced by, in cylindrical shells 414 ' in plates, 399 Tension of prismatical bars, by concentrated forces, 51 by distributed forces, 167 by gravity force, 246 unüorm, 245 Thermal expansion, coefficient of, 399 Therroal stress, 399 in cylinders, 408, 427 in disks, 406 general equations for, 421 solutions of, 433 in infinite solid 434 in long strips, a99, 404 in plates, 399, 435 ~n spheres, 403, 416 m steady heat flow, 412, 427 Thin tubes, torsion of, 298 Tore, incoroplete, bending of, 3!l5 torsion of, 391 Torque, 262 calculation of, by use of membrane analogy, 271 Torsion, boundary conditions in 260 circular shaft in, 249 ' cone in, 309 cylindrical body of cross section · 258 in, elliptic cylinders in, 263 of hollow shaft, 294 lines of shearing stress in, 270 local stresses at fillets in 238 299 of prismatical bars, 258, '462 ' approxiroate method in investigating, 280 displacements in, 259 fillets in, 288, 299 stress function for, 261 of various forros of cross section, circular, 249 elliptical, 263 other forros, 266, 278, 280, 286 rectangular, 272, 275 rolled profile, 287 seroi-inverse method in, 259
THEORY OF ELASTICITY
506
Velocity, of waves, of dilatation, 453 Torsion, of shafts of variable diameter, of distortion, 453 304, 490 Virtual displacement, 151 of thin tubes, 298 Virtual work, application of, in torof tore, 391 sional problems, 281 Torsional rigidity, 264 principle of, 151 Trajectories of principal stresses, 134 Transmission of concentrated force act- Vorticity, 292 ing, on beam, 99 w on disk, 107 on infinite body, 354, 362 W arping of cross sections of prisms in on infinite plate, 112 torsion, 259, 265 on semi-infinite plate, 85 Waves, of dilatation, 453, 454 on strip, 115 of distortion, 452, 454 on wedge, 96 longitudinal, 454 Transverse waves, 454 in prismatical bars, 438 Triangular prism, in bending, 331 plane, 454 in torsion, 266, 280, 474 propagation of (see Propagation of Tubes, submitted to internal and exterwaves) nal pressure, 59 Rayleigh, 456 torsion of, 298, 480 refiection of, 443 Twist, center of, 271, 336 superposition of, 442 of circular ring sector, 391 surface, 456 of shafts (see Torsion) transverse, 454 velocity of (see Velocity) Wedge, loaded, at end, 96, 98 along faces, 123 Uniqueness of solution, 236 Weight, bending of beam by own, 42, Unit elongation, 5 53, 342 tension of bars by own, 246 V Work, virtual, principle of, 151
u
Variable diameter, shaft of, 304 Velocity, of surface waves, 459 of wave propagation in prismatical bars, 440
Y Young's modulus, 10