Engineering Failure Analysis.pdf

Engineering Failure Analysis 11 (2004) 547–559 www.elsevier.com/locate/engfailanal

Estimating fatigue life under variable amplitude loading through quantitative fractography – A case study J.R. Tarpania,*, C.O.F.T. Ruckerta, M.T. Milana, R.V. Silvaa, A. Rosato Jr.b, R.N. Pereirab, W.W. Bosea, D. Spinellia a

Materials, Aeronautic and Automotive Engineering Department, Engineering School of Saõ Carlos, University of Saõ Paulo, Saõ Carlos-SP, 13.566-590, Brazil b Embraer S/A, Saõ Jose´ dos Campos-SP, 12.227-901, Brazil Received 10 September 2003; accepted 27 September 2003

Abstract Quantitative fractography techniques have been implemented and frequently used in the Failure Analysis Laboratory – LANAF, at the Engineering School of Saõ Carlos, for fatigue life estimations of structural components, whose fracture surfaces are subjected to detailed inspection with the aid of scanning electron microscopes. This work describes one of the recent activities in progress at the LANAF, in which fatigue crack initiation and propagation lives have been estimated for an idealized aeronautical part tested in the laboratory, under variable amplitude loading condition – VAL, i.e., flight simulation testing. Fractographic reconstitution of sub-critical crack growth has been performed through the identification of marking load patterns left in the wake of the propagating crack, which have been correlated to the significant load levels applied during the fatigue test. A semi-automated procedure to estimate fatigue lives under VAL has been developed and implemented by which a virtual marking load pattern is generated and compared to the real pattern determined fractographically. # 2004 Elsevier Ltd. All rights reserved. Keywords: Aluminium alloys; Fatigue markings; Fractography; Marker loads; Quantitative fractography

1. Introduction Fractographic techniques using scanning electron microscopes (SEM) are extensively used in failure analysis of aeronautical components and structures [1–5]. The more advanced procedure employed on the fatigue life estimation of in-service or in-test fractured components undergoing variable amplitude loading – VAL, is depicted in Fig. 1. The technique consists basically of generating a virtual pattern of fatigue load marks provided by the applied loading spectrum, using preferably computational techniques, and compares it with the pattern actually recorded on the fracture surface of the component, obtained through microscopic survey. In summary, to be successful this procedure depends basically on the correct * Corresponding author. Tel.: +55-16-273-9590; fax: +55-16-273-9574. E-mail address: [email protected] (J.R. Tarpani). 1350-6307/$ - see front matter # 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.engfailanal.2003.09.004

548

J.R. Tarpani et al. / Engineering Failure Analysis 11 (2004) 547–559

Fig. 1. Schematic of the failure analysis procedure using quantitative fractography [4].

interpretation of the surface marking pattern printed by the higher cyclic loads (i.e. marker loads) experienced by a flying aircraft or a tested structural component. Fig. 2 shows typical load marking patterns impressed by the action of critical load sequences on the fracture surface of an aeronautical component fatigue tested in the laboratory. In order to estimate fatigue lives spent in both crack initiation and propagation stages, it is necessary to obtain a plot similar to Fig. 3. From the numerical integration of the area under this curve, one can also determine the number of flights as a function of the crack size, as well as the crack growth rate in the structure or component analyzed. The correct determination of fatigue crack initiation and propagation lives has obvious and important implications for several steps of aircraft fabrication and maintenance, such as materials selection, component design, manufacturing processes, non-destructive inspection programs, and the ultimate failure analysis as well. In this work, an idealized aeronautical component was fatigue tested in a laboratory environment in order to simulate its in-service performance under VAL. After testing, the fracture surfaces were thoroughly analyzed in a SEM. The purpose was to determine the resulting crack surface marking patterns, which were compared to those derived from a computer program, which had the simulated flight load history as the input. The methodology described in Fig. 1 has been strictly followed, aimed at determining both crack initiation and propagation lives of the fatigue-tested component.

2. Materials and component tested A 1.5 in.-thick rolled plate of a 7475-T351-aluminium alloy, specification SAE-AMS4204 [6] was investigated in this study. Table 1 presents the chemical composition (wt%) of the alloy as determined via optical emission spectroscopy. Table 1 also supplies the allowable quantities for the elements according to the SAE specification, which are strictly obeyed by the Al-alloy tested. The microstructure of the aluminium alloy is showed in Fig. 4. Texture is found in the rolling direction, L. Wide non-recrystallized regions and segregation can be observed in LS, and particularly, in TS planes.


549

Fig. 2. Typical load marking patterns (arrowed) developed on the fracture surface of a fatigue-tested component: (a) near crack origin; (b) intermediate location; (c) final fracture position.

3. Experimental and analytical An idealized aeronautical component designed by LANAF was fatigue tested under a VAL spectrum expected to occur in the lower wing skin panels of commercial aircraft. Fig. 5(a) presents the main fractured ligament F2 ! F1, which has been analyzed in this study. The black arrow in Fig. 5(a) indicates the location of the fatigue crack site, i.e. a corner crack. The cyclic loading configuration applied to the part is given in Fig. 5(b). A bending moment is introduced in the tested component, whose magnitude and profile depend on the fasteners array employed to connect it up to the auxiliary plates, which in turn are gripped by the servo-hydraulic closed-loop test machine. Therefore, tensile stresses are induced in the surface of the sample that is in close contact with the thicker plates. Consequently, a compression loading is generated in the opposite face of the tested part.

550


Fig. 3. Fatigue crack growth data as determined by quantitative fractography [2]. Table 1 Chemical composition (wt%) of the aluminium alloy tested in this study Element

Weight (%)

Si Fe Cu Mn Mg Cr Ni Zn Ti Ca P Pb Sb Sn Sr V Zr B Cd Co Al Impurities

0.051 (max 0.10) 0.079 (max. 0.12) 1.67 (1.2–1.9) 0.011 (max 0.06) 2.265 (1.9–2.6) 0.253 (0.18–0.25) 0.004 (max 0.05) 5.909 (5.2–6.2) 0.020 (max 0.06) ND (max 0.05) 0.010 (max 0.05) 0.002 (max 0.05) 0.004 (max 0.05) 0.001 (max 0.05) ND (max 0.05) 0.002 (max 0.05) ND (max 0.05) 0.0008 (max 0.05) ND (max 0.05) 0.002 (max 0.05) Balance 0.026 (max 0.15)

Limiting values according to SAE-AMS4204 are provided in parenthesis. ND: non-detected.

Fig. 6 shows two flight block sequences imposed on the component during the laboratory test. In total, 20 blocks of 6000 flights were sequentially applied to the component, therefore performing 120,000 simulated flights.


551

Fig. 4. Tri-dimensional views of the microstructure of the aluminium alloy: (a) lower and (b) higher magnitude.

Fig. 5. (a) Partial view of the VAL-fatigue-tested component. The black arrow points out the approximate location of the fatigue crack site. The full length of the failed ligament F2 ! F1 is 23 mm; (b) schematic of the cyclic loading arrangement imposed on the tested component.

552


Fig. 6. Two loading blocks applied sub-sequentially to the component tested, with the severest flight, numbered 1, occurring once at each applied block of 6000 flights.

After testing, the component’s fracture surfaces were inspected in a SEM operating at a voltage of 20 kV in both secondary and backscattered electrons imaging modes. The chief interest was to elucidate the entire load marking nuances left on the wake of the propagating crack. This task was accomplished by using image magnifications ranging from 10 to 20,000. A computer program was specifically developed to generate virtual marker load patterns from the available flight load history, which were compared to marking patterns determined fractographically. During the programming, special attention was devoted to VAL-variables such as ‘‘maximum tensile stress peaks’’ and ‘‘effective number of tensile stress cycles’’ within a block of simulated flights, since reliable criteria had to be developed concerning the width of virtual marking bands and band spacing as well. Once the software provided the virtual pattern of load marking bands expected in one block of flights, it was straightforwardly compared to that real pattern, in an operation that resembled the reading of bar codes. The procedure was repeated until a minimum matching between virtual and real patterns was achieved. When a successful correlation was reached, the flight blocks counting along the fracture surface could be accomplished manually.

4. Results and discussion 4.1. Metallurgical examination of the crack nucleation site Fig. 7 exhibits the microstructure of the alloy in the fatigue crack initiation site region, which has been indicated by an arrow in Fig. 5(a). The arrow in Fig. 7 points out distinct neighbouring second-phases, respectively black and white colours, located near the fatigue crack nucleation site. The particles were analyzed and identified by energy dispersive X-ray (EDX) analysis. These metallurgical stress concentrators near the surface of the fatigued component suggested that microstructural heterogeneities could have led to the failure of the component in the flight simulation test.


553

Fig. 7. Microstructural aspects of near fatigue crack site region, where the arrow points out two distinct second phases. The crack site location, the borehole position and the fatigue crack surface are identified.

Fig. 8. EDX spectrum of (a) white and (b) black phases located in the neighbourhood of the fatigue crack nucleation site in the borehole F2.

554


Table 2 Chemical composition (wt%) of the intermetallic phases shown in Fig. 7 Element

White phase (wt%)

Black phase (wt%)

O Al Mn Mg Cu Zn

7.03 23.65 0.33 ND 67.76 1.24

0.43 89.47 ND 1.13 1.95 7.03

The copper and zinc contents are highlighted for the white and black phases, respectively. ND: non-detected.

Fig. 9. (a) Top view of the fatigue crack nucleation site pointed out by a white arrow; (b) contact marks due to the friction of the threaded fastener with the internal surface of borehole F2, indicated by black arrows; (c, d) detail of the marks, where white arrows indicate the marks and the black one the crack site position; (e) cross-sectional view of one surface mark generated during the disassembly of the jig, where the white arrow indicates the direction of the movement for the fastener removal.

4.2. Microchemical analysis of the phases Figs. 8(a) and (b) show EDX data for the black and white colours phases depicted in Fig. 7. Table 2 presents their chemical composition, identified as Cu-rich and Zn-rich phases, respectively.


555

4.3. Contact marks on the borehole F2 Fig. 9(a) displays a partial top view of the fracture surface of the component, in which the arrow points out the site of the fatigue crack initiation. The dark stain exhibited in this region occurred due to surface contamination by accidental contact with lubricant oil drained from the mechanical test system. The black arrows in Fig. 9(b) indicate some marks left by the contact of the fastener’s thread screw with the internal surface of the hole identified as F2 in Fig. 5(a). Notice that one of these contact marks lined up exactly with the fatigue crack nucleation site, which is pointed out by the white arrow. However, none of the marks reached the fatigued fracture plane, Fig. 9(c) and (d), in which the crack site is indicated by a black arrow in Fig. 9(d). In fact, a more careful analysis showed that the marks left by the contact between the fastener thread and the borehole F2 were indeed generated during the disassembly of the jig to carry out the procedures of failure analysis. This is clearly verified in Fig. 9(e), which corresponds to a crosssectional view of the part. Here, the white arrow points out the fastener’s withdrawal direction, where one can see the material pulled out from the borehole due to friction with the threaded screw. Fig. 10(a) shows a detailed view of the fatigue crack site of the fractured ligament F2 ! F1 (see also Fig. 5a), where a non-metallic inclusion was located, as indicated by an arrow. This particle was identified by EDX microanalysis as a calcium-sulphur rich phase. The corresponding energy absorption spectrum is provided in Fig. 10(b). Fig. 11 presents three SEM-microfractographs taken from a intermediate position of the fracture path F2 ! F1, at distinct image magnifications. The loading marks printed on the fracture surfaces are numbered in accordance with the flight codes established in Fig. 6, i.e. severest flight sequence which originated this marking pattern. Notice the perfect correspondence between the fractographic marking sequence and the flight sequence applied to the structural component, the latter depicted in Fig. 6. An extra marking, i.e. not generated by the action of a severe flight, is identified in Fig. 11(b). This mark was probably due to an overload experienced by the tested component in the automatic re-start of the mechanical testing system just after the electric energy power supply had been re-established. Fig. 12 shows an overall view of the cracked ligament F2 ! F1, in which the white arrows indicate the sequential positions of occurrence of the severest flight 1. The conclusion, based on this figure is that at least eight full blocks of 6000 flights have been applied to extend the fatigue crack along the whole ligament

Fig. 10. (a) Non-metallic inclusion located at the crack site of the fatigue-fractured ligament F2 ! F1; (b) energy absorption spectrum resulting from EDX microanalysis.

556


Fig. 11. SEM-microfractographs taken from the fracture surface of the fatigue-tested component at three different magnitudes of image: (a) 70, (b) 230, and (c) 700. Arrows indicate the microscopic or local direction of crack growth along the fractured ligament F2 ! F1.

assessed. Still in Fig. 12, the black arrow points out the fatigue crack nucleus position, which was first seen in Figs. 5(a), 7, 9(a), (b), (d) and 10(a). Fig. 13 presents the resulting plot obtained from the fractographic analysis performed in this study, according to the graphic format displayed earlier in Fig. 3. Two distinct methods have been used in order to correlate the experimental data points, namely, general power-law and point-to-point linear fitting. Once the set of data points presented in Fig. 13 had been fitted, the numerical integration of the area under the fitting curves was performed. Fig. 14 plots results from both processes of numerical integration, by which one can estimate fatigue crack initiation and propagation lives of the aeronautical part tested in fatigue. It can be observed that power-law data fitting predicts longer fatigue crack growth periods than a linear fit, and, consequently,


557

Fig. 12. Sequential positions occupied by the severest flight number 1 along the fatigue crack path on the ligament F2 ! F1.

Fig. 13. Fatigue crack growth data determined fractographically from the tested component.

Fig. 14. Fatigue crack growth estimations according to distinct fitting methods: (a) number of flights vs. crack depth; (b) vice-versa.

558


Fig. 15. Mean crack growth rate estimations according to the equivalent flight concept: (a) plotted against crack depth; (b) plotted against number of simulated flights.

shorter crack nucleation periods, which is inferred to be 49,500 flights according to the former method (recalling that total life of the component is 120,000 flights). On the other hand, according to the point-topoint linear data fitting method, about 60,000 flights were necessary to start the crack growing in fatigue. Therefore, from Fig. 14, two limiting situations can be postulated regarding periodic non-destructive inspection and failure prevention of similar in-service components: first, if fatigue crack initiation is the main concern, i.e. fail-safe approach, the most conservative method is the power-law integration; second and conversely, if cracking has already been established in the structural component, propagation becomes of prime interest, i.e. damage-tolerant approach, and linear point-to-point integration is the most conservative method. Fig. 15 plots fatigue crack growth rate in function of crack depth and number of flights, respectively, considering both integration methods shown in Fig. 13. Almost imperceptible differences are established in Fig. 15(a), regarding initial stages of crack growth. Major differences are noticed at higher levels of crack extension only. On the other hand, Fig. 15(b) allows one to clearly differentiate linear vs. power-law methods, as a result of the large differences between fatigue crack growth estimations exhibited in Fig. 14.

5. Concluding remarks This article described one of the recent activities conducted in the Failure Analysis Laboratory (LANAF) of the Engineering School of Saõ Carlos in the field of quantitative fractography. In this study, crack initiation and propagation lives have been estimated for a fatigue tested aeronautical part subjected to VAL, which has been imposed under laboratory controlled conditions, i.e. flight simulation testing. Fractographic reconstitution of sub-critical crack growth has been performed through the identification of marking load patterns left in the wake of the propagating crack, which have been correlated to significant load levels applied during the fatigue test.


559

A semi-automated procedure to estimate fatigue life under VAL has been developed and implemented, by which a virtual marking load pattern is generated and straightforwardly compared to the real one determined fractographically. Implications for the adoption of more or less conservative life estimations methods in the structural integrity program of commercial fleets have been discussed on the basis of both fail-safe and damage-tolerant approaches.

Acknowledgements The authors wish to express their gratitude to Embraer S/A for supplying the material tested and for valuable discussions.

References [1] Abelkis PR. Use of microfractography in the study of fatigue crack propagation under spectrum loading. Fractography in Failure Analysis, ASTM STP (Special Technical Publication) 1978;645:213–34. [2] Wiebe W, Dainty RV. Fractographic determination of fatigue crack growth rates in aircraft components. Can Aeronaut Space J 1983;27(2):107–17. [3] Dainty RV. Use of ‘‘marker blocks’’ as an aid in quantitative fractography in full-scale aircraft fatigue testing: a case study. Fractogr Ceramics Metal Failures, ASTM STP 1984;827:285–308. [4] Goldsmith NT, Clark G. Analysis and interpretation of aircraft component defects using quantitative fractography. Quantat Meth Fractogr, ASTM STP 1990;1085:52–68. [5] Goldsmith NT, Clark G, Barter SA. A growth model for catastrophic cracking in an RAAF aircraft. Eng Failure Anal 1995;3(3): 191–201. [6] Anon. SAE Aerospace Material Specification for Aluminium Alloy Plate 7475-T7351, Designation AMS 4202C, 1989.

Engineering Failure Analysis.pdf

Recommend Documents