.-
(32)
o o 211 -
(33)
o
are the usual displacement coefficients arising from the linear part, a/E, of the strain, E, and are the same as in the first example. In fact, 812 =8 21 =0 due to the selection of the elastic centre as origin of X 1 and X 2 and Eqs. (32) simplify to o 11 X 1 +J(Pqcp,+P, 1cp,)dl·+o 111 - 0 .-
(34)
S22 X 2 + J
-
r2( m2 sin -7T aH°h - I + A. -' -") --. A r m2 - I m
~
I
............
(35)
) which are essentially the same as in (12), but for a small difference in 8 20 arising from the geometry of the cross-section chosen. Furthermore 27Tr 7r1· 3 £7(1 +K)
................................
(36)
..................................
(37)
where
,.
f(E;P 12 + E,P, 2+a0P; 2)ds=O
,-
f(E 1Pil + E,Pn)ds+8 10 =0 ('
,-
c
8 11 = £/' o22 =
Noting that in the actual system no relative movements take place at the cut ends and applying the unit load method as given by Eq. (71a) in Part I we obtain the following two equations f{E;Pq +E,P, 1 +a0Pq)ds=0
J(E 1P 12
2·0
Fig. 5.-Non-linear stress-strain curve
where o 10 =Ja0P; 1ds, o 20 =Ja0P, 2ds . . . . . . . . . . . . . . . . . . . . . . . . . . . .
820 =
y,- E,
or
1·5
('
(27)
_ PnX 1 +P, 2X 2 ..J..(PqX1 +P; 2X 2)
0·5
(24)
Note that M,. is taken about the centroid G of the cross-section. Using the expression for N,, and M,, of Eqs. (2) and (3) in (24) we obtain: for system X 1 =I:
Pq=-h; P, 1 =+r,
0·2
+ E,P, 2 )ds +8 20 =0
}
(29)
Eqs. (36) are identical with (II) in the first example. Non-linear terms of Eq. (34): Using Eqs. (24) and (27) for P,, P, and Eqs. (2) and (3) for M,, N,, the arguments of the cp-function in (34) are conveniently written in the form
c/J,= c/J{AI;h[ -X 1 -
l
(30)
J
• Dinrtional suffices in the strains arE' omitted siner direet bendinte strains only auly!i!iis.
K-A~·2and/-A,~,It2
OCl'ur
in thl'
X 2 r cos 1{1 (1
cp,=c/J{)h[ X1 +X2r cos 1{1
+~1 ~)]}
(I -1' ~)]}
(38)
Substitution of (38) into (34) shows that for a given cp-function we obtain after integration with respect to 1{1 two non-linear equations in the unknown X 1 and X 2 . The solution of these equations determines the complete stress distribution in the ring.
15
e =0.,(1+ cos!JI)/2 0·1
P=O
p=1 P=2
3·0
2{)
Fig. 6.-Second Example: thermal stresses in non-linearly elastic ring. Values of redundancies x 1 and x2
a0.,EI M,o • - h Fig. 7.-Bending moment and normal force distributions in non-linearly elastic ring due to thermal strains
We select now the specific a-E relationship
..................................
E=i[, +C~r-lJ or aE (a)" cp(a)=-3 a.
(40)
where a. is an arbitrary reference stress (a. may be defined as the stress at which the total strain is twice the linear strain component) and n is an odd integer. By varying a. and nit is possible to obtain a versus E diagrams with a wide variety of shapes (see FIG. 5). We assume also that A 1 =A,=A/2 then, K=I/Ar 2=(h/2r)2=f32 The cp-functions of Eqs. (38) take now the form
lr ............
(41)
....................................
(42)
cp;=-p"[x 1 +x 2(1+f3> cos cp.= where
pn[x 1 +x2(1-{J)
Jr1
X!= Mto
and
Jr2r
X2= Mto
COS
,Pr~
j
,Plni
in which Mt •• the moment at ,P=O for fully restrained thermal expansion, is given by (19), and Ea0. h p= 2a. 'fJ=z, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (43)
l
Substituting from Eqs. (25), (26) and (41) in the integrals of (34) we find
!. P; 1 cp;ds=p"~ 1-f [x +x (1 +fJ) cos ,P ]nd,P JP, cp,ds=p"Jz ~J [x +x (1-{J) cos ,P ]"d¢' 1
1
2
1
2
c
I (44)
c
which can be integrated without difficulty since n is an integer. Taking n=3 Eqs. (44) reduce to the simultaneous cubic equations 1 2x 1 +p 2x 1[2x 12 +3(1 +f3 2)x 22 ]-;n =0 2(1
+f3 2)x 2+p 2x 2 [3(2+f3)x 1 2+3(1
+6f3 2 +{3 4)x 2 2] -
-(l +fJ) 2m 2 sin (1rjm) =O m 2 -1
1r
l
which form= l and {3=0 (2r»h) take the particularly simple form
76
2x 2+p 2x 2[6x 1 2 +3x 2 2]-1 ~ 0 The linearly elastic case is represented by p =0. Then x 1 =x 2 = 1/2 which values may also be obtained from Eqs. (13), (14) and (19).
(46)
Numerical Results Eqs. (46), which have only one pair of real roots, have been solved for values of p between 0 and 3 and FIG. 6 shows values of x 1 and x 2 plotted against p. The roots of the simultaneous cubic equations were obtained by an iteration method, successive values of x 1 and x 2 being calculated, using the. previous approximations for the terms in the brackets. For p =0 · 5, successive values of x 1 and x 2 , calculated in this way are: X1 0·5 0·433 0·452 0·449 0·449 x2 0·5 0·390 0·416 0·410 0·412 The initial approximation in this case was the linearly elastic solution; for further values of p, extrapolation of previous results gives satisfactory starting points. Since successive approximations are alternately greater and less than the final values of x 1 and x 2 it is quite easy to improve the natural convergence of the process by a little judicious anticipation. Direct solution ofEqs. (46) was also obtained. Eliminating x 2 an equation cubic in p 2x 1 2 and x 1 is obtained, which can be solved for either parameter and the corresponding values of p. x 1 and x 2 calculated. The bending moment and normal force distribution around the ring is plotted in FIG. 7 for p=O (linear elasticity), p=l·O, and p=2·0, in the non-dimensional form M/Mto and Nr/M 10 • As would be expected, the effect of the non-linear stress-strain curve (in which da/dE decreases with a) is generally to reduce the maximum bending moment produced by a given temperature rise and also to smooth out the variation around the ring. 4. THIRD EXAMPLE Thermal Stresses in a Rectangular Plate
Consider a thin rectangular plate of length 2a and width b with a temperature distribution 0=0.g(y/b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (47) symmetrical about the longitudinal axis of the plate and constant along the length (see FIG. 8). The stresses in the plate are to be found with the assumptions that they are constant across the thickness (plane stress) and that the material is isotropic and follows Hooke's law.
c
c
}
(39)
Additional Notation
(45)
2a, b: length and width respectively of plate x, y: cartesian co-ordinates with respect to the centre of the plate x'=a-x g=x/a, 7J=y/b, 'o;=x'/b: non-dimensional co-ordinates a0 =a0.g(7J): thermal strain
~..F(y/b)
We apply the principle of virtual forces in the form of Eq. (80b) of Part I, i.e.
~,. (Vyy}y.o
JJ[.6. 2F+Ea.6.0]SFdxdy=0
0
b/2 :------'L--- - ' - - - - ' ' - - - ------"'---- ---'---'-------'
t------a--~
_l
-1/2
J
I
I ......................
(48)
a .•• =a.,.=O is set up. To obtain the stress distribution in a plate of finite length, we can apply a stress-system -a.,., at the two ends x= ±a and superimpose the resulting (diffusion) stress-system on (48). We shall, however, adopt here a slightly different approach in order to illustrate the application of our principle of virtual forces in the presence of thermal strains. Guided by the solution (48) for an infinite plate, we assume that it is possible to express the longitudinal stress in the form a.,.,=Ea0J.(TJ)tP(~) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (49) where +1/2
f.(TJ)= Jg(TJ)dTJ-K(TJ)
.. .. .. . .. .. .. . .. . .. .. .. . .. .. . . .. .. ..
(50)
-1/2
and tfo is a function only of x. We know that Eq. (49) cannot, in general, represent the true distribution since the cross-wise distribution will also vary with x. However, for a smooth function g(y) it can be expected that (49) will yield a sufficiently accurate answer. The distribution (49) can be represented by the stress function F=Ea0,)J2J(TJ)t/J(~) .. .. .. . .. . .. .. .. .. . .. .. .. .. .. .. . .. . . .. (51) which gives the stresses o2F do/ au= 0 yz=Ea0.dTJ 2t/J(~)=Ea0.J.t/J(~) . . . . . . . . . . . . . . . . . . . . . . (52) o 2F
bdl dt/J
a.,.=-oxoy=-Ea0•ad1Jd~
o2F b2 d2t/J a •• = 0xz=Ea0.-(i21 de
. . .. . .. . . . . . . . . . . . . . .. . . . . . .
(53)
................................
(54)
The a.,.,, a•• distributions are symmetrical and the a~• distribution antisymmetrical with respect to the x-axis. From Eq. (52) and the conditions (a••)•=±b; 2=0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (55) (a.,.)._ 0 =0 (antisymmetrical distribution) . . . . . . . . . . . . . . . . . . (56) the function I is : '1
I(TJ)=J
'I
J!.(TJ)dTJdTJ .................... ................
where the virtual force system SF is here due to the increment Stfo of the function tfo 8F=Ea0b 218t/J . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (59) Substituting for F and 0 from Eq. (51) and (47) respectively, and noting that d 2g d 4/ dTJ2=-dTJ4 ···················· ···················· ···· (60) we obtain from Eq. (58)
J J[ d~d +I +1'2
d 4 t/J
a2 d 2t/J d 21
a4
d'l]
+27)2 de dTJz+ b'
(61)
I -·1/2
J[ +I
g, /., 1: functions of TJ A, B, C: constants tfo : function of ~ or ' {J 1, 2 =JL±iw: roots of auxiliary equation Ji=J-Lbfa, iii =wbfa The thickness of the plate is taken as unity. The problem can be solved exactly but we shall derive here an approximate solution using the principle of virtual forces in the form discussed in Part I, Section 6E. Consider first a plate of infinite length for which the stress distribution is well known.* Thus, for the symmetrical temperature distribution considered here, the self-equilibrating stress system +1/2
(58)
which may be written, on integrating with respect to TJ
Fig. 8.-Third Example: thermal stresses in rectangular plate; transverse distribution of temperature and stresses
axx=Ea0{J g(TJ)dTJ -g(TJ)
.. .. . . .. . . . . . .. . . . .. . . .. . .. . ..
(57)
-1/2 0
J
d4tfo a2 d2t/J a' A d~c2b2B de+IJ4C(t/J-I) Stfod~=O
.. .. . .. .. .. . ..
(62)
I ([2dTJ, C= I dTJ 4dTJ .. .. .. .. .. .. .. .. -1/2 -1/2 TJ -1/2 Note that A, B and C are all positive. Eq. (62) can only be satisfied for an arbitrary Scp if d4t/J a2 d2tfo a 4 Ad~ 4 -21J2 B d~ 2 +l)4C(t/J-1)=0 .. .. . . .. . .. . .. . .. . . . . . ..
(63)
-I
where the coefficients A, B, C are given by
J+ I;2
A= f2dTJ, B= -
·rl·l/2do/
r+
I''Jy
which is the required differential equation in tfo. The particular solution of (64) is t/J= l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (65) which is, of course, the solution for an infinite plate and the complementary function can be written in the form t/J =K 1 cosh {1 1 ~ + K 2 cosh {1 2 ~ + K 3 sinh {1 1 ~ + K 4 sinh {1 2 ~ . • • • (66) where 8±(82-AC)vz]vz (67) flt·z-b A
·-i![
are the roots of the auxiliary equation a2 a4 A{J4-21)2Bfl2+pC=0 . .. . .. . . .. . . . . .. . . .. . . . . .. . . . . . .. .
(68)
A B C are functions solely of the temperature distribution and are ind~pe~dent of a/b. Thus, the length/width ratio of t~e plate _ente~s only as a common factor in the arguments of the hyperbolic functions m (66). The tfo functions must obviously be symmetrical with respect to the centre of the plate (~ =0) and therefore K 3 =K4 =0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (69) At the ends of the plate (~ = ± l) the stresses axr. a.,. are zero and therefore
tfo= ~~=0
for~=± I
....................................
(70)
The boundary conditions (70) determine the constants K 1, K 2 and the t/J-function is found to be J. t _ _{1 1 sinh {1 1 cosh {1 2 ~ -{1 2 sinh fJz cosh fJ1~ ( 71 ) 'l'(s)-l {J 1 sinh{J 1 cosh{J 2 -{J 2 sinh{J 2 cosh{J1 ·········· In general AC>B2 and the roots {1 1, 2 are therefore complex. They can be written in the form {J 1 , 2 =JL±iw . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . . . . . . . . . . . . . . . (72) where JL=bii=bnc
v/~+~)r 12
8)]
I
~
. .. . .. .. .. . .. .
112 a_ a[l(~ /C w=bw=b 2 VA-A j Substituting (72) into (71) we obtain JL[cosh JLO +t) sin w(l-~)+cosh JL(l-~) sin w(l-r~)] tP = I JL sin 2w +w sinh 2JL w[sinh JLO +0 cosw(l-~)+sinhJL(l-~) cosw(l +~)] JL sin 2w +w sinh 2JL •••••••••••••••••••
• See Timoshenko and Goodier, Th~ory of Elasticity, 2nd Edition, McGraw-Hill, New York 1951, p. 399. •
(64)
0
0
••••••
0
•••
0
••
(73)
(74)
which, at the centre of the plate (~=0) reduces to
77
1 _ 211- cosh IL sin w +w sinh IL cos w (75) IL sin 2w+w sinh 211For the discussion of Eqs. (74) and (75) it is best to consider a particular form of temperature distribution. Take, for example, the function 1 + cos 2mrT} . K(T))= where n IS odd . . . . . . . . . . . . . . . . . . . . . . . . (76) 2 .JJ ) _ 'f'\ 0 -
which is shown in FIG. 8 for n= i. Then " __ cos 2n7TTJ Jo2
I
1 +~~~:1TTJ
l()
lt----+--
(77)
....................................... .
(78)
and the constants A, B, C are, from Eq. (63) 3 1 1 A=128n«1T«' B=32n21T2' C=g ........................... .
(79)
=
1·2 ...---.-----. ----
also iJ.=1Tn[t(v3+ 1)]11 2= 1· 351Tn, =1Tn[t(v'3-1)]11 2=0· 6981Tn .. (80) A perusal of Eqs. (74) and (75) on the basis of the arguments (80) shows that even for the lowest value (n=1) the function !fo, which is zero at g= 1, increases rapidly with decreasing g towards the value 4> =I for an infinite plate. In fact, not only does it reach it but, due to the influence of the trigonometric terms, it exceeds, very slightly (,..._,1·002) the infinite plate value if a> 1 · 2h for n = 1. The above discussion indicates that, for plates with alb> 1, the end effects for g= + 1 and g= - l are practically independent of each other. Thus, for such plates the stress distribution between g= 1 and g=0 may be taken to be the same as in a semi-infinite plate. Introducing the co-ordinate
y
w
x'=a-x
- 0·4 r---+r---t----t---+---+---+-----l
from the free end and putting
x' a-x
-0·2
a
{=b="=o-e>b
we find that for a-+oo, Eq. (74) reduces to
r/>(~)= 1 ;isinw{:wcosw{e-il; ....................... .
-06(81)
which is the appropriate solution for the semi-infinite plate. Note that, for the particular temperature distribution of Eq. (76), ji and w are proportional ton and hence lj>({) is solely a function of n{. The corresponding complete stress system is :
~_tf21 -I.-
cos2n7TTJ[1-ii sin 2
Ea0.- d7J 2 'f'-
w{+w cos w{ -ii:] w
e
·J. • 2n 7TTJ w -2 +11-z . -r -~<· df d'l' sm Ea0. = dTJ d{=- ~ sm we, e •
----w-
Un
u1J11 tP!fo +(l+cos2nry) C.02+;i2[_ -r _ . _,.1-;;· Ea0. =1 d{2= 81r2n2 -w w cos w.,-IL sm we, e '
I
..................................
Jl
(83)
Maximum stresses: For any T} the transverse direct stress (u••) reaches its greatest value at the end of the plate ({=0) and is given by un .~ /c 2 Ea0. =f(fi. +w 2)=f'V A The maximum value (at TJ =0) is (
)
Uww
-2
-2
max.=~ t~ =1·155
27T n For the maximum shear stress (u.,.) YJ=l/4, {=0·218n and thus (u ..) •...,
(uo:w)max. =0· 302 (u.. ).=<>
We mentioned on page 77 that it is possible to find the stress distribution in a finite plate by superposing on (48) the diffusion stress-system arising from -u.,., applied at the two ends x= ±a. In aircraft structural theory such diffusion analyses are often based on the simplifying assumption Ew=O. It is interesting now to compare the results obtained by the
78
Method of Virtual Forces
I
Assuming
£ 11 =0
Fig. 9.-Third Example: therm!'l stresses in rectan1ular plate; lon1itudinal var1ation of stresses
I~
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (82) ji and ware defined in general by Eq. (73) and for the particular temperature distribution of (76) by Eq. (80). The stress system (82) is illustrated in FIGS. 8 and 9 which show the transverse and lengthwise variations respectively of the stresses for the case when n = 1. u .. is the longitudinal direct stress u:u in an infinite plate
u.. = -Ea0. cos f'7TTJ
I
I
la~ter method with t~ose derived from our principle of virtual forces. To this purpose we consider for the temperature function (76) the plate under an end load stress-system
-u = + Ea 0 cos 2n7TTJ 0
..
2
...............................•
(84)
app~ied at x~ ±a. Now, in .a plat~ ~ith Ev.• =O, the s~ress system (84) is an etgenload, I.e. the cross-wise vanat1on of Its stresses IS invariant with x. •
The eigen-value ,\ corresponding to (84) is simply ,\=~1rn/b
and the diffusion coefficient JLD=M(G/E)
112 =21Tnba(G)112 E
The lengthwise variation of the u.,.,-stresses is obviously in this case cosh ILDg cosh IL» Thus, the stresses in a plate subjected to temperature function (76) are when Evv-O• easily found to be ' .!!E-.=_ cos 2n7TTJ(t- cosh IL»g) 2 cosh ILD 112 ~= _ sin 2n7TTJ sinh ILDg Ea00 E 2 COShJLD
Ea0.
..5L = Ea0.
_
(Q) (Q) I+ cos 2n7TTJ
E or for a long plate
2
l
J ... . ....... (8~
cosh IL»g cosh ILD
p. ~~;' Argyris, Dunne, 'The General Theory, etc.', Part V, J,R.Ae.S .. Vol.
Ll, November 1947,
__!!iu_ _
Ea0o-
-(Q) E
112sin
2mr7J .-- • 2 e 11 Ir
a.. GE l+cos2mr7J _-. rae.=-2 e
(86)
1'D·
where
fLD ~~ 27Tn(~) 112
For G/£=0·385 we have: liD = I . 2047TII which is in reasonable agreement with the value of IL in (80). However. the corresponding stresses are very different as may be seen from FIG. (9) for n = 1. It appears that the assumption €•• =0 overestimates considerably the rate of diffusion of the end load system -au in the range 0· 7> ~>0.
5. FOURTH EXAMPLE St. Venant Torsion of a Thin Section Bar In treating the torsion of uniform bars of thin section, approximate expressions for the maximum shear stress and the stiffness may be obtained by assuming the distribution of shear stress across the thickness to be identical with that for a very wide strip of constant thickness equal to that at the point considered in the cross-section. Thus for a section as shown in FIG. 10 we obtain the linear variation of a.x acro!.s the thickness dB a •.,=-2yGdz (90) with the maximum value dB a,.,max.=tGdz and the torsion constant
J=Hy0 3dx
..........................................
(91)
To obtain a more accurate solution we assume still that azr varies linearly across the thickness, but determine the variation over the width by application of the energy theorems enunciated in Part I. A solution is obtained firstly by application of the Principle of Virtual Displacements with an assumed transverse warping distribution to correspond to the linear au distribution and secondly by application of the Principle of Virtual Forces with the assumed stress variation. The two methods provide upper and lower limits respectively for the St. Venant torsion constant J (see Sections 4 and 6 of Part 1). These two analyses are fully worked out for singly symmetrical double-wedge sections and the results illustrated in a number of figures. The closeness of the upper and lower limits prove the methods to be very accurate. This is underlined by the coincidence of the approximate and exact solutions in three cases. Considerable effort has been devoted in the past in developing many approximate methods for the solution of the St. Venant torsion problem: some of them are discussed at the end of this example. Particularly relevant to our present analysis is the approximate solution for the isosceles triangle developed by Duncan, Ellis and Scruton* which is, in fact, identical with our solution of this case based on the Virtual Forces Principle. t
Additional Notation a, al> b, c, !=dimensions of cross-section (see FIGS. 10 and 15) a, a 1 =a/b, a 1 /b respectively Ox, Oy=axes in plane of cross-section (Ox along axis of symmetry of cross-section) Oz=longitudinal axis of bar Yo= y-eo-ordinate of section boundary g, g1, ?]=non-dimensional co-ordinates (see nu. 15 and Eqs. (142) and (143)) B' =dB /dz =rate of twist of bar about z-axis T=applied torque about z-axis J= T/GB' =~St. Venant torsion constant w=wB'=warping displacement of cross-section «!>=stress function /, /1> cf>, 1> 1 =functions of x n=a,,max./GtB'=maxi mum stress parameter • W. J. Duncan, D. L. Ellis and C. Scruton, 'The Flexural Centre and the Centre of Twist of an Elastic Cylinder', Phil. Mag., Series 7, Vol. 16, 1953. p. 231. t After completion and submission for publication of the present analysis the authors' attention was drawn to: E. H. Mansfield, The torsional rigidity o.fsolid cylinders of double-wedge section, A.R.C. Report, 1954, to be published in the R. & M. series. In tbis paper the same upper and lower limits are obtained for the torsion constant of a double-wedge section by a basically identical method. The stress distribution is, however, not discussed.
Fig. 10.-Fourth example; thin solid section in torsion
For the general background of the St. Venant theory of torsion of solid sections, the reader is referred to: Timoshenko and Goodier8 , Theory of Elasticity, 2nd edition, ch. II, McGraw-Hill, 1951. (a) Principle of Virtual Displacements To obtain the linear au variation we assume that the warping displacement w of the cross-section can be written in the form II' _c, ~.()' =yj(X). B' (92) • ' • •• • •. • • . • ' • •. •. •• • • ' • • • •• • •• • • •• • where j(x) is a function of x only and is to be determined.* Then in terms of Eq. (92), the shear strains and stresses are:
a u =G€
g
,,cGB'[~W-)']=,GB'y[ df.-l] ~X ~
a,.=-GE,.=GB'[~ +x] =GB'[f+x]
(93)
....................
(94)
We consider virtual displacements defined by an increment 8/ of the function f. For such virtual displacements B' =dB/dz remains constant and therefore no virtual work is done by the applied torque T. Thus, 8W is zero and Eq. (44) of Part I reduces to the principle of minimum strain energy Eq.(44a) of Section 4C which takes here the form: 8U;=~Jf[au8E,r+a, 11 8E, 11 ]dxdy -~0 . . . . . . . . . . . . . . . . . . . . . . . . (95) where the virtual shear strains due to the increment of are from (93) and (94)
8€,,,~B'y8(:fx) O'_,.t!!.Jt>
-~ . . . . . . . . . . . . . . . . . . . . J
b€,11- 8'8.f
(96)
Substituting from (96), (93) and (94) into (95), integrating with respect to y between ±Yo, we find
1{'~\X -I Y~f/+yo(f+x)b{ }dx=O
(97)
x,
and integrating the first term by parts
(98) Since 8/ is an arbitrary virtual increment off d [y}(df --- --I )] v
Y·/(:fx --
I)= 0 at the limits x
1
and x 2
•••• . ••••• ••••••••
(99)
(100)
giving the necessary boundary contitions at the ends of the cross-section. These latter arise from the virtual work of the stresses on the edges of the strip and are, in fact, the equilibrium conditions there in terms of the assumed transverse stress distribution.
Torsion Constant J To determine J by the Principle of Virtual Displacements, we consider a virtual increment 8B' of the rate of twist of the bar. Then from Eq. (44) of Part I 8U,: -8W=O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (101) where 8W=T8B' . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (102) and 8U;=Jf[a,x8€u+a, 118€:u]dxdy . . . . . . . . . . . . . . . . . . . . . . . . . . (103) In calculating 8U,. due to the increment 80' we note that part of the virtual strains arises from the corresponding increment of the warping displacements which are proportional to B' (cf. Eqs. (92), (93) and (94)). But the increment of strain energy due to any virtual warping displacements is zero (Eq. 95). Hence for the virtual strains giving 8U; in (101) we can omit the part due to warping and take simply ::::::::::
}
.. .. .. .. .. .. .. .. .. . .. . .. .. . .. .
(104)
• See Part I, Section 4E.
79
Using Eqs. (93), (94), (104), (103) and ( 102) in ( 101 ), and integrating with respect toy, we obtain for the torque T due to the shear stresses
J[ ~_Y]\1; x,
T=2GO'
-I )+xyo
(105)
which gives, on integrating by parts the first two terms, x,
oU;*=G(0')2;{
- J{fx(l~os
x,
Substituting from the differential Eq. (99) for Yo(f+x), integrating by parts and noting the boundary conditions (100), we finally obtain (106) x,
a(
df) _ _I_ -~Jx, Yo l - dx dx
J~G0'-3
(107)
x,
~~) +Yoa[Ji(Yo1xo) -1 ]c/> }ocf>dx}
(121) ................................ Since ocf> is zero at the ends of the section the first term vanishes and there remains only
oU;* - -G(0') 2
and therefore the torsion constant is
i2~os~~Scf>+y/Jx·c/>ocf> I
·H {fx(l~o5t)+y.{fx(Yo1x0)--l]c/> }ocf>dx
ow• =OT·O' =G(0') 2· Hyo 3ocf>dx
x,
(122)
From Eq. (119) for S T, the virtual complementary work is (123)
(b) Principle of Virtual Forces To obtain a shear stress system satisfying the internal equilibrium conditions, we express the shear stresses in terms of a stress function
Substituting now Eqs. (122) and (123) in Eq. (116) we conclude that in order to satisfy (116) for any arbitrary ocf>, the function c/> must satisfy the differential equation
The equilibrium conditions on the boundary will also be satisfied if* (Aoq)
and must be zero at the ends of the cross-section.
T=ff[a.wx-a,,y]dxdy
which gives on integration by parts with respect to x and y for the first and second terms respectively and noting the boundary condition (109) (110) T=2JJ
T=G0'Hyo 3 c/>dx
......................................
(112)
............................. .......
(113)
The stresses and strains corresponding to (Ill) are from Eq. (108)
a,,=~C:=-2yc/>·G0'=GEzx
..........................
dyO,J.] G0'de/> ()
. . .. . . ..
(114)
(115)
In the general equation of the Virtual Forces principle (Eq. (62) of Part I, for 0 =0). (116> sui• -ow•=o . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . in which (117) oU;* = ff(E.,oa._,+E •• oa•• ]dxdy . . . . . . . . . . . . . . . . . . . . . . . . . . and the virtual stresses oa•.,, oa•• are obtained by giving to the function c/> a virtual increment ocf>. Such virtual stresses automatically satisfy the internal equilibrium conditions. We note moreover that since
oa•..,= -lySe/> . GO'
].1.
........... .
(124)
APPLICATIONS TO PARTICULAR SECTIONS We now apply the equations developed in the previous section to obtain approximate solutions for the shear stress distribution and torsional stiffness of two particular types of cross-section: first the simple rectangular section and second a double wedge section formed by two isosceles triangles having a common base. Solutions are obtained by both methods and the results compared with one another and also with relevant exact solutions. 1. Rectangular Section (a) Solution by the Method of Virtual Displacements with Assum?d Transverse Warping Variation Consider a rectangular section of width s and thickness t. Taking the origin of axes at the centre of the section, we find for the differential Eq. (99) in this simple case d2[
d~2 -p,2f== ph
whence
J=Jo,=Hyo 3 c/>dx
a[
5 dyo) -I 'I'+ Yo 3 =0 d ( Yodx d (2yo dx +Yo dx T de/>) dx
........................................
where JL 2= 12/t2 • • • • . • • • . . • • . . • . • • • • • • . • • • • • The boundary conditions (100) are here
(125)
. • . • . • • • . • • • . • . •
(126)
dx =I for x= ±s/2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
(127)
and the appropriate solution is therefore 2 sinh JLX f=ji:coshJLs/2-x ················ ···················
(128)
d{
Stress Distribution: Substituting for /in Eqs. (93) and (94) ,
[ cosh JLX
az,=GO · ly cosh JLS/2 -I
J
II
(129) ?~GO'.~ sinh JLX JL cosh JLS /2 •• The distribution of shear stress according to Eq. ( 129) is illustrated graphically in FIG. II.
a
J
I
J ......... .
(118)
sr=Go'Hyo3ocf>dx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
(119)
and from (112), the virtual torque due to ocf> is
Substituting for the strains from (114) and (115) and for the virtual stresses from (ll8), integrating with respect toy between the limits ±Yo, we obtain from Eq. (117) the virtual complementary energy
sui• =G(0'>2H [2~os;t fx
Assumed C1z• distribution
(120) • See Timoshenko and Goodier•, and also Part I, Eqs. ( 5). t See Part I, Section 6E.
80
tis • 0·20 Fig. !I.-Rectangular s~tion; shear stress distribution according to Virtual Displacements solution
1·0
---
Virtual Forces Virtual
m·
......... .........
_jJ
---Virtual Forces
~s--C' 0·4
0·2
0·4
-Virtual Displacements o
tis
0·6
0·8
Exact
Solutions
1·0
Fig. 12.-Maximum shear stress for uniform torsion of rectangular section
The maximum stress is reached at the mid-point of the longer sides and is amax. 1 Gre'=n=l-cosh Oa · · .. · · · · · · .. · · .. · · · · · .. · · · .. · · · · · · <130) where 8a=p.s/2=v3s/t . .. . . . .. . . . . . .. . . . .. . . .. . . . .. . .. . . .. . . (131) Eq. (130) is plotted as the broken line in FIG. 12 with t/s as abscissa. We note that in the extreme case of a square section (s/t=l) n=0·657, which is 2 · 7 per cent less than the exact value 0 · 675. * An obvious error in the detailed stress distribution is the constant a .. with y leading to unbalanced stresses on the boundary.
0·2
0·4
t/s
0·6
0·8
t·O
Fig. 13.-Torsion constant J for uniform rectangular section
Torsional Stiflness: The general expression (I 07) for the torsion constant gives, with the /-function of Eq. (128) 3 +s/2
J=t_J[l- cosh p.x ]dx 3 cosh p.s/2 -s/2
=
s~ll-ta~~ OaJ
(132)
The broken curve in FIG. 13 shows the variation of 3J/st3 with t/s according to (132). In accordance with the general Principle of Virtual Displacements this approximate solution must always overestimate the stiffness. (b) Solution by the Principle of Virtual Forces with Assumed Form of Stress Function The differential Eq. (124) becomes in this case
~~->..2cf>= -,\2
......................................
(133)
where
A2 =10/t2
(134)
and the boundary conditions are cp=O for x=±s/2 .. .. . . . .. . . . .. . . . . . . . . . .. .. .. . .. . .. . From (133) and (135) we find easily cosh Ax tP = 1 cosh As /2 · · · · · · · · · .. · · · · · .. · · · · · · .. · · .. · .. · · .. ·
(135)
( 136)
0
Assumed C1zx distribution
1 .JL
L.........l.... Gt~
tis • 0·20
Fig. 14.-f\ectangular section; shear stress distribution according to Virtual Forces solution
where
8b=As/2=V~s/t
(139)
The full line in FIG. 12 shows the variation of the maximum stress parameter n with tfs. Torsional Stiffness: Eq. (113), in conjunction with the cp function (136), gives for the torsion constant
J=s~l 1 _tan8~eb]
..................................
040
(137)
which is shown as the full line in FIG. 13. We find remarkably close agreement with values given by the exact solution.* Thus for a square section, which can hardly be considered as thin, Eq. (141) gives the value 0·419st3 /3 which is only 0 · 7 per cent below the exact value. In accordance with the general Principle of Virtual Forces our present approximation must underestimate the stiffness.
FIG. 14 illustrates these distributions for a section with s/t=5. We note that although the a,., distribution is generally similar to that of FIG. 11, the a •• distribution now satisfies all the boundary conditions. For the maximum shear stress (at the mid-point of the longer sides) we find amax. 1 Gt8' =n=l-cosh 8b · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (138)
2. Double Wedge Section Since the centre of twist for unrestrained torsion is arbitrary, we assume for convenience that it is at the centre of the common base of the two isosceles triangles which form the section (See FIG. 15). We introduce also the non-dimensional co-ordinates g=(x+a)/a, g1 =(a 1 -x)/a 1 . .. . .. . .. . . . . .. .. . . . . .. . .. .. (142) measured inwards from the ends of the section and 7J=y/b . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (143)
Stress Distribution: Substitution for cp from Eq. (136) in Eq. (114) gives for the stresses =-2y[l-cos\.\x]·G8' cosh 1\S /2 ) , sinh Ax (}' t2 2 ( a •• = 4 -y 1\ cosh >..s/2 · G
a
""'
• Timosbenko and Goodier', foe. cit.
I;
............ · ...... ·
• Timosbeoko and Goodier', /oc. cit.
81
r_·
1---------
c
Fig. IS.-Double wedge section
0
Then for the left-hand part of the section x=a(' -1), y.=,b . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . and for the right-hand part x=a1(l-,1), y.=,1b ·...............................
I _Q_ I Gt .:j
....I..,JIU...,!. L . J •
I - !....... . . . . . . ...........
(144) (l44a)
tic • 0·20
0/C •
0·50
Fig. 16.-5hear stress distribution in symmetrical double wedge section t/C=0·2; approximate solution by Virtual Displacement method
(a) Solution for Assumed Warping Functions: For this type of cross-section we have separate expressions for the function fin the two triangular parts of the section. Writing f and / 1 for the function in the left ( -a
+a.
+
0
0
-a
-a
Iy.3(~ -I )at I - J{~[y.3 (1x -I)
J-y.(J+x) }otdx
a1
IY. 3 (~-t)otii-J{~[y. 3 (~-I)]-y.(f1 +x)}ot1 dx=O 0
0
................................ (145) To satisfy Eq. (145), the functions/and/1 must, as before, satisfy the differential Eq. (99) and the conditions at the ends of the section
[Yo3 (~-t)l=-a =[Y.a(~1 -t)l=+a,
=0
..........
(146)
In addition, however, we see that we must have
[y.3 (~ -I )atJz..o -[y.3 (1x1 -I )ot1].,.. 0 =0
[y.s(~-1 )l_o -[y.s(~I-1 )l=o =0
(150)
Eqs. (148) and (150) are the boundary conditions at the junction of the two parts of t~e section, giving the compatibility and equilibrium requirements respecttvely. With x and Yo from (144), the differential Eq. (99) becomes, for the left-hand part of the section do/ df ' 2iif2+3'a,-3a 2/=3a'(l +a 2)-3a 2a .................. (151) where a=a/b The complete solution of ( 151) is a [ /1-1 -/1-1 f=as_ 1 A' +B' _,(a 2+l)+(aLl) ........... .
J
(152) (153)
where fP=l +3a 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (154) Similarly we find for the right-hand part of the section (substituting from (144a)) the differential equation i:. 2do/1 i:. d/1 !>1 d' 12+Js 1~~ -Ja1o/1 = -3a 1, 10 +a 1 2)-i-3a 12a . . . . . . . . (ISla) and the solution
J
a1 [ /1,-1 -/1,-1 /1=a 12_1 -A1'1 -B1'1 +' 1(a 12+1)-(a 1 2-l) . . where a 1 =a 1 /b, {J 12= I +3a 12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
(I53a)
(154a)
Boundary Conditions: In terms of the non-dimensional co-ordinates, the boundary conditions (146), (148), (150) may be written:
82
l
--
I
.. I ...
1 ..sL I
I
Gt~
(Oyz)y.o
-·~---------~--~~tic • 0·10
(147)
Since we assume a priori that the displacements are compatible, it follows that at the junction of the two regions f,._o=/ltz-o · ·.... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (148) and therefore (149) (ofl-o=(o/1)-.o ···································· which gives in Eq. (I 47)
0
I
a/C • 0·10
Fig. 17.-5hear stress distribution in single wedge section t/c=O·I; approximate solution by Virtual Displacement method
d'
l
d'
(155)
![df] +_!_[df1] =0 a 1 o,= 1 o=l a 1 The first of these gives immediately B=B 1=0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (156) and substitution from (153) and (153a) in the last two of (155) leads to A _ 2a({J1 2 +fJ 1-2)+a 1(fJ2+fJ 1-2) I ({J 1+2)[a({J1-I)+a 1({J-I)] (157) A -2a1(fJ2+fJ-2)+a(fJ12+fJ-2) ............. . 1({J+2) [a(fJ 1-I)+a 1({J-I)] For a doubly symmetrical section (a=a 1), Eqs. (157) reduce simply to A=A 1=2 ................ ·~·.......................... (158) and when a 1 =0, i.e. the section consists of a single isosceles triangle A=2(l +fJ)/3 . . . . . . . . . . . . . . .. . .. . . . . . . . . . . . . . . . . . . . . . (159)
I
Stress Distribution: From Eqs. (94) the stresses corresponding to the general solution (I 53) are a [ A (160) O"zv=-2a2-J' l-2g bG8' ..................... .
fJ-2]
O"zz=-
2[
a 3A ~:.fJ-2] 8 , 2a2-)7J l-2(1+fJ)S bG
................ ..
(161)
The maximum stresses obviously occur on the boundary of the section
(7]=,). Resolving there into tangential and normal components we find
(162)
TABLE I Comparison of Values of Maximum Stress Parameter n with Exact Soluti,.,.ns
and O'zn =
aazy-U.rz
(l +a2)1/2
(fl-2)
a 2A fl-1 , =(a2-l)(l+a2)112 P+t ~ bGB #0
.. . .. . ... ... ..
In an unsymmetrical section (a#a 1) the maximum stress is always on the longer side and hence in general we use in Eq. (164) the greater value of a. In FIG. 18 the broken lines show values of the maximum stress parameter n calculated for the cases a=a 1 and a 1 =0 and plotted against I fa. TABLE I gives some particular results for sections where exact solution is available. Torsional St!ffness:
Eq. (107) takes here the form 1ta=a
J~a( t-1 ~)d~+a 1 J ~~a( I+i I
0
1 1 ~ )d~ 1
0
....... .
(166)
J
6A1
,
(f1t+0(f1t+2)J
(167)
where A, A 1 are given by Eq. (157). For a very thin strip, i.e. when a, a 1 become infinitely large, the limiting form of (167) is (168) where t=2b and c=a+a 1 (see FIG. 15). Eq. (168) is the result given by the rough approximation of Eq. (91); it is exact for the infinitely thin section. For a doubly symmetrical section (a=a 1 ) the values of A and A 1 of Eqs. (158) lead to the simple result 12J <11 +5)([1-1) eta =
Exact Solution
1·00
0·-4n•
Virtual Displacements Solution (a)
1·00 Exact
0·-461 (-3·5%)
0·-4-43 (+l·3%)
Virtual Forces Solution (b)
1·00 Exact
0·-485 (+2·0%)
v'3/-4=0·-433 Exact
Square
Right Angled Isosceles Triangle
Equilateral Triangle a=Vl, a,=O
a=l.a 1 =0
0·32lt
v'3/-4=0·-433•
1=0·375 (+ll·5%)
t. ==0·3125 ( -60~)
• Timoshenko and Goodier'. Ch. II, pp. l66 and ln. t Galerkin, B. G. 8ui/.I'Acad:miedesSciencesdeRussie, p. Ill, 1919; Kolossol, G. Compt. rend., Vol. 178, p. 1057, 1924.
TABLE II et'
l IT With Exact Solutions
Comparison of Values of 1 Infinitely Thin Section
a=a 1 =-l
Equilateral Trian&le a~ v'l, a,==O
Richt Ancled Isosceles Triancle ao:.~ I, a 1 =0
1·00
0·-422•
0·300·
0·1567t
Virtual Displacements Solution (a)
1·00 Exact
.,\=0·-437 (+3·7%)
0·338 (+ll·6%)
... ==0·1B7S (+19·7%)
Virtual Forces Solution (b)
1·00 Exact
~~=0·406
0·300 Exact
,;:,~0·1563
t/c-->0 Exact Solution
Square
(-3·7%)
(-0·16%)
• Timoshenko and Goodier'. Ch. II, pp. l66 and ln. t Galerkin, B. G. 8uii.I'Academiedes Sciences de Russie, p. Ill, 1919; Kolossol, G. Compt. rend., Vol. 178, p. 1057, 1914.
which with the values off and / 1 from (153) and (153a) reduces to 3J aa 2 [ 6A a 1a 12 [ 2ba=aLJ I-~ +a1LJ I
a=a 1 =1
(163)
Expressions analogous to (160)-(163) are obtained for the right-hand part of the section and need not be quoted here. The distribution of shear stress is illustrated in FIGS. 16 and 17 for some particular cases. As in the previous example, our assumed form of warping displacements, yielding a constant u ·~withy, leads to unbalanced shear stresses u zn on the boundary. The maximum u,. stress is given by u,.max. a(a 2 +1) 1 12 [2(a 2 +1) Jl/(fl-2) (164) n= GtO' =J(fi+2)(fi-1) Afi(fi-1) and occurs at fl-2 2(a 2 + I) (165) ~ = AfiCfi -1)
I
Infinitely Thin Section t/c= l/(a 1 +a,)-->0
(169)
Similarly, we find for the isosceles triangle 12J fi2-t eta =(/3+2) 2
(170)
Eqs. (169) and (170) are plotted in FIG. 19 against the ratio 1/a and in TABLE u particular results are given for comparison with the corresponding exact solutions and with the results of the Virtual Forces method. (b) Virtual Forces Solution With Assumed Form of Stress Function
As in 2(a) we have separate expressions for cp and cp 1 in the two triangular parts of the section. Following the argument used there and noting that at the junction of the two parts the two functions must be equal cf>x=o=c/>~o:r=O
· · • • · · • · · · • ·• · ·•..•.•••..•.•.•.•••••...•
(171)
0·8
0·6
c/12 0·4
Virtual Forces
0·2
0·2
Virtual Displacements 0
Virtual
Exact Solutions 0
0·4
b/a
0·6
08
1·0
Fig. 18.-Maximum shear stress for uniform torsion of double wedge section
Forces
Virtual Displacements Exact Solutions
0·8
1·0
Fig. 19.-Torsion constant J for double wedge section: comparison of results by Virtual Displacements and Virtual Forces methods
83
..JL
0 1 _g_ Gt:5
0
Gt.f/
1.. 1.._,......_'""'-'-_._.'-''.._,wl
f/C : 0·10
ale • 0·50
f/C • 0·20
Fig. 20.-Shear stress distribution in symmetrical double wedge section t/c=0·2; approximate solution by Virtual Forces method
a/C = 1·0
Fig. 21.-Shear stress distribution in single wedge section tjc--0· 1: approximate s&lution by Virtual Forces method
and hence the virtual increments are also equal [Sc/Jl.--o=[Sc/J1lx-o
.............................. ......
(172)
we deduce from (121) and (116) the additional condition [ 2yo 5
5
dc/J+
=[2yo5 dc/>1+
4-1..]
4-1..
J
dx Yo '+' x-o 5 dx Yo '+' 1 x-o which may be written more concisely as !!._(Yo5/2J..)J [ dx '+' x-o -[!!_(Yo5/2J.. dx '+' 1)] x-o =0
............... ·.·.
(173)
Eqs. (171) and (173) provide the necessary additional boundary conditions for this section. Eq. (171) is the equilibrium condition (equality of au) and (173) is a compatibility condition. From Eq. (115) we see that it is not possible to satisfy the compatibility condition (equality of *'••) for all values of y because of the discontinuity in dy 0 /dx. Eq. (173) expresses the requirement for the virtual complementary work at the junction to be zero and is thus an averaged compatibility condition. The differential Eq. (124) becomes here, on substituting for x and Yn from (144), for the left-hand part of the section*
ct!J!
c2t!:.P. de+ 5~d~+2~( 1 _
~
__ 2a ~ 2 ....................... . a2)-1..'+'-
(174)
the complete solution of which may be written a2 [ 6-2 c/>=a:2_ 1 1-cg
-ng -6-2] ....................... .
(175)
Similarly we find for the right-hand part of the section a 12
[
cP1=a 12_J 1-c1~1
6,--2
.. 1),-2]
t/c : 0·05
(176)
Boundar.y Conditions: Since must be zero at the ends of the section on boundary)
<-1
+_! [_!!__
g=g 1 =0 (equilibrium
=0 ··············
(173a)
<,~1
For the functions cP and c/> 1 of (175) and (175a) the boundary conditions (171) and (173a) yield _ a 2(S 1 +l)+aa 1 (S 1 +2)+a 1 2 C- 5a(S 1 +2)[a(2S 1 +1)+a 1(2S+1)]
C 1 = 5al(S+2)[a(2S 1 + 1)+a 1(2S+ I)] When a =a 1 Eqs. (178) give simply
I I
J ··········
5
C=Cl=2S+I ....................................... .
and when
r-
_I
Stress Distribution: From Eqs. ( 108) or ( 114) we find, for the function c/J of ( 175), the stresses
b-2] bG8'
l
a2 [ u,.= -2a2 _ 1l] 1-C~ -2a2a_ 1g{ 1-
s;l-l~ 2 -( 1-Dl] 2]
and at the boundary (lJ =g) a2-=r-~ 1-C~
}bG8'
/j-JJ bG8,
Uzn=O
~
(181)
I ~ J
(182)
I J
The maximum value of u .. is given by
D=D 1 =0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (177) Eq. (173) becomes, on substituting for x and Yo from (144) and (144a)
![.!!. (g512c/J)] a d~
ate • 025
I
Fig. n.-Shear stress distribution in unsymmetrical double wedge section t/c=0·05, afc=0·25; approximate solution by Virtual Forces method
Uz.,=2-
S2=(3+5a 2)/2, S 12=(3+5a12)/2
(178)
a(a 2 + 1)1! 2
[
I
]1/(1)-2)
which is easily shown to satisfy the compatibility equation (179)
a 1 =0
............................................... .
Uz 8 max.
(183) n= Gt8' = 52(S+2)(S-I) C(S-1) Similar expressions are ebtained for the right-hand side of the crosssection and need not be given here. Again we find that, in general, the maximum stress occurs on the longer sides of the section. In FIGS. 20, 21 and 22, some typical distributions given by Eqs. (181) and (182) are shown graphically. Note the discontinuity in Uz• at the junction of the two regions. This appears because the variation with y is fixed by the assumed form (Ill) of the stress function which cannot represent adequately the effect of the sudden change of slope of the boundary. It does not arise in the case of the single wedge (a 1 =0) for which we obtain more accurate results. In fact, for an equilateral triangle (a 1 = v'J, S =3) our assumed form for the stress function gives the exact solution. Thus, the stress function* for this case is (184)
(180)
• This equation, _in slightly different form, was used lor the isosceles triangle by Duncan Ellis and Scuton, /oc. Cit. '
84
Oyz'
a(a2+J)ll2 [
where
C=l
~~==------~-~--
r""'==..:J::;;..,..,
Uzy=
(175a)
-D1g1
•
0 2<1>
ox 2 + 0oy2<1>2 +2GB ' =0
..................................
(185)
Values of the maximum stress parameter n are plotted in FIG. 18 against I fa for comparison with the results of the previous method and TABLE 1 gives some particular results for cases where the exact solution has been given. • See Timoshenko and Goodier!, Joe. cit.
0-9
0·8
J
cto/12 0·6
Fig. 23.-Torsion constant j for general double wedge section
Torsional Stiffness:
Using Eq. (115) we find for the torsion constant 3J
I
I
8ba=alg 3 cfodg+at bgt 3cPtdgl
.............. .. ............
which with Eqs. (175) and ( 175a) reduces to
J
a 12 4C a2 [ 12J 8ba=aar::-t 1- 3+2 +atUT!
[
4C 1 1-sl +2
J
......... .
(186)
(187)
where C and C 1 are given by Eqs. ( 178). Eq. (187), which is very similar in form to (167), yields the same limit as a and a 1 become infinitely large. When a=a1 we find a 2(2S+9) 12J 5 ct3 =2 (2S+I)(S+2) 2 and when a 1 =0 a2 12J 5 ct 3 =i ~ Eqs. (188) and (189) are plotted in some particular results are given.
(188)
(189) FIG.
19 as the full lines and on TABLE 11
Accuracy of Results In an approximate solution it is always desirable to be able to assess the order of accuracy achieved. In the two methods applied in this example, the method of Virtual Displacements gives an overestimate whereas the method of Virtual Forces gives always an underestimate for the stiffness. Thus, by using both methods we have succeeded in establishing upper and lower bounds for the stiffness. Some idea of the relative accuracy of the two methods can be obtained from TABLES 1 and n by comparison with results of exact solutions. Note, however, that these are for sections well beyond the intended range of application of the solutions developed here and the basic approximations are not really suitable for such sections. Nevertheless the agreement appears to be quite good. In general the Virtual Forces method appears to give the most accurate solution, in particular for the triangular section where the general formula (189) for J gives the exact value at both ends of the range lfa~o and a= v3. In FIG. 23 we show finally the torsion constant J calculated by the Virtual Forces method for the general double wedge section. Note that FIG. 23 shows minimum values of stiffness. For t/c
85