CrossMark

Available online at www.sciencedirect.com

ScienceDirect

Procedía Engineering 114 (2015) 124 - 131

Procedía Engineering

www.elsevier.com/locate/procedia

1st International Conference on Structural Integrity

Comparative study of prediction methods for fatigue life evaluation of an integral skin-stringer panel under variable amplitude loading

Sedek J.a*, Ruzek R.a, Raska J.a, Behal J.a

aVZLU (Aerospace Research and Test Establishment), Strength of Structures Dept., Beranovych 130, Prague - Letnany, 199 05, Czech Republic

Abstract

The study deals with an integral skin-stringer wing panel made of AA 7475-T7351. DT solutions for fatigue life prediction are based on computational methods. Crack growth is analysed under the flight by flight loading representative for a commuter aircraft. Crack growth predictions are performed by using both the linear damage accumulation principle and the FASTRAN retardation model. Different solutions of crack and structural geometry, which are expressed by the /5 correction factor related to stress intensity factor assessment, are compared. The predicted crack growth and original data for the integral wing panel are discussed. The influence of geometry correction factors, which are defined by considering different hypotheses of their determination, is not significant for the integrally stiffened panel. In contrast, the retardation effect for the flight by flight loading, investigated alloy and structural part is significant. The FASTRAN retardation model gives underestimated predictions of approximately 10% compared with the experimental data. The reason underlying the increasing difference between the experimental data and numerical predictions of longest crack length propagation is discussed.

©2015 PublishedbyElsevierLtd.This is anopen access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of INEGI - Institute of Science and Innovation in Mechanical and Industrial Engineering Keywords: Crack Growth; FASTRAN; FEM; Integral Skin-stringer Panel; Retardation

1. Introduction

Aircrafts designed in compliance with present airworthiness regulation requirements must conform to the damage tolerance concept (DT). DT requirements, together with a high level of reliability, require sophisticated procedures for service planning and structure monitoring. Operational service economic efficiency and safety are strongly influenced by a number of planned inspections. The fatigue life of an airframe and its components is crucial for

* Corresponding author. Tel.: +420-225-115-172

E-mail address: sedek@vzlu.cz

1877-7058 © 2015 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Peer-review under responsibility of INEGI - Institute of Science and Innovation in Mechanical and Industrial Engineering doi:10.1016/j.proeng.2015.08.050

service inspection intervals determination. The ability of fatigue crack growth prediction is one of the most critical points in light of inspection intervals evaluation. There are many prediction models that are used for fatigue crack growth prediction. The simplest prediction procedures are based on crack growth material characteristics determined by constant amplitude loading with a sinusoidal waveform under different stress ratios. Different formulas for crack growth interpretation are known. Material data regression using formulas of crack growth rate vs. amplitude of the stress intensity factor in its basic form (e.g., Paris [1], Forman [2]) or refined form by the effective stress intensity factor (e.g., Elber [3], Schijve [4]) are consequently directly used for crack growth predictions. This simplest procedure, based on linear damage accumulation, is very fast, but the predicted curves are very often unrealistic, namely, for variable amplitude loading. Then, the reality that each airframe is loaded by variable amplitude loading must be considered. A loading sequence of each individual structure contains a row of very high peaks compared with the loading peaks induced by steady, straight flight. A large plastic zone is induced in the crack tip as a consequence of the high loading peaks. The presence of the large plastic zone in the crack tip induces the retardation of crack propagation. The extent of retardation depends on different factors (e.g., material, mean stresses, peak ordering). A large number of prediction models have been developed to evaluate the retardation effect. The most widely known models and commercial codes that account for an interaction effect are Willenborg [5], Wheeler [6], CORPUS [7], PREFFAS [8], ONERA [9], NASGRO [10], FASTRAN [11] and AFGROW [12], for example. This comparative study shows the significant differences between both procedures. Models utilizing the linear damage accumulation principle and interaction effect evaluation using strip yield models are discussed. The behaviour of 7475-T7351 plate material under flight by flight loading, which is representative for a small commuter aircraft, is investigated. Crack growth predictions in an unstiffened and integrally stiffened wing panel without and with retardation effect evaluation are compared with the experimental data. Results confirm the necessity to use prediction models with interaction effect evaluation.

2. Experimental detail

An experimental program was performed by using an integrally stiffened panel made from a 7475-T7351 aluminium alloy plate with a thickness of 76 mm in the L-T direction. The chemical composition of the plates was 5.5 wt% Zn, 1.5 wt% Cu, 2.3 wt% Mg, 0.2 wt% Cr, 0.12 wt% Fe, 0.1 wt% Si, 0.12 wt% Fe, 0.06 wt% Mn and Ti, and balanced Al. The loading direction was in the longitudinal axis compared with the rolling direction. The integrally stiffened panel represents the lower wing panel of a commuter aircraft. The panel was made by CNC (Computer Numeric Control) milling. The scheme of the panel is shown in Fig. 1 (b). Width (W) of the panel was 550 mm and skin thickness (t) was 3.7 mm. The panel contained six longitudinal stringers with a span of 80 mm and four perpendicular ribs. The thickness of the panel was continuously scaled until the end gripping parts to uniformly distribute the load. A symmetric notch with a total length of 5 mm perpendicular to the longitudinal direction in the centre of the panel was cut. The ribs of the integrally stiffened panel were supported by metal reinforcement to prevent traverse deformation of the panel (skin). The aim of reinforcement was to simulate the stiffness of real ribs. Skin ends of the panel were guided between two stiffened plates to avoid a transverse skin movement. The panel was tested by using an MTS servo hydraulic test machine with a capacity of 1 MN, which was controlled via an MTS FlexTest40 electronic. A test frequency of 2.7 Hz was applied. The test assembly is shown in Fig. 1 (a).

The fatigue test procedure was divided into two major phases. The aim of the first pre-cracking phase was to initiate a short crack; the aim of the second was to monitor crack propagation and define fatigue life. The pre-cracking procedure was conducted under constant amplitude loading with a stress ratio R = 0.1 and a maximal stress in the loading cycle of 72 MPa. The crack propagation phase was conducted under variable amplitude loading. Fatigue loading was force controlled. Crack tips propagation was monitored by using an optical method via OLYMPUS SZ40 microscopes with lenses of 40x magnification on both surfaces of the panel, in agreement with the ASTM E-647 standard. During the first test phase, a fatigue crack length of a = 5.7 mm was pre-cracked.

Fig. 1: (a) Test assembly; (b) 3D model of test specimen; dimensions in mm.

The panel was loaded by using a randomized flight-by-flight sequence, which represented the service loading of a small commuter aircraft. The loading sequence is representative for the bottom skin part of a wing. A maximum stress value of 124.35 MPa in the sequence corresponds to a flight load of 2.8 g. A minimal stress value in the sequence of -26 MPa was applied. The total number of 332 088 loading peaks corresponds to 3000 flight hours. This corresponds to 55.35 cycles per flight.

3. Crack growth prediction

3.1. Methods of modelling crack propagation

Several models for fatigue crack growth predictions were developed in the past [1-11]. A review of several models is discussed in Ref. [13, 14]. The discussed models use cycle by cycle analysis and are significantly different in the principles of damage accumulation and consideration of interaction effect of variable amplitude loading. This study deals with the linear accumulation principle and with the FASTRAN strip yield model, which is implemented into the AFGROW commerce software [12].

The FASTRAN model utilizes the Dugdale model [15] of plastic zone. The original model is modified so that plastically deformed material remaining behind the advancing crack tip is modelled - see Fig. 2.

Fig. 2: Draft of FASTRAN model [11].

The model is composed of three regions around the crack: (1) a linear elastic region containing a circular hole with a fictitious crack of half-length c'+ p, (2) a plastic region of length p and (3) a residual plastic deformation region along the crack surface. The physical crack is of length c'- r. Regions 2 and 3 are composed of rigid-perfectly plastic bar elements in which the local stress and strains are computed. They are affected by the stress state expressed by the constraint factor a, which takes the value of a = 1 for plain stress and a = 3 for plain strain.

The constraint factor a is usually used as a parameter for minimal mean square error estimation of experimental data by the relationship da/dN = f(AKef), but determination by trial and error until the data for all R values collapse to a single curve (or as close as possible) is also mentioned in Ref. [12]. In this work, the constraint factor was analysed in separate sections of crack growth rate data independently according to Ref. [16]. The method is based on the occurrence of shear lips on fracture surfaces of the test specimens. The constraint factor was determined in the flat region and slant region separately. The phenomenon occurs when a stress state at the tip of the crack changes during crack formation [14, 17]. The amplitude of the effective stress intensity factor was determined according to the following assumptions and relations specified in Ref. [11].

During loading, a crack opens at a certain load with a corresponding stress intensity factor Kopen due to the appearance of a compressive region behind the crack tip. The effective value of AK with an opened crack AKejf can be expressed by relations (1).

^Keff = Kmax " Kopen when Kopm > Kmm

AKe// = Kmax - Kmin when Kopen < K^

Fatigue crack growth rate data from tests with a constant amplitude loading for different R ratios use relations (1) reduced to the form of a one-curve relationship of the crack growth rate da/dN versus AKf Kopen is expressed using relations (2), where Fw is the finite width correction, a0 is the flow stress and R is the load cycle asymmetry. The relationships can be used for a maximum load Smax<0.8a0.

K „ K

open - A0 + A1R + A2R2 + A3R3, for R > 0

A0 + A1R, for R < 0 K

K — K if open

open min' jt

_ o if < 0

A0 = (0.825 - 0.34« + 0.05a ) S F

tri OY V

A = (0.415 - 0.071«)-

A2 ~ 1 _ A) _ A1 _ A3 A3 — 2 A0 ^ Aj 1

a- S F

max w 2^0

3.2. Material data

Material characteristics and parameters of the retardation model are used according to Ref. [18]. Material data were generated by using the same AA 7475-T7351 plate with a thickness of 76 mm and L-T direction mentioned in Chapter 2. The 100 mm wide middle tension crack specimens (M(T)) were used in compliance with the ASTM E-647 standard. Three levels of the load ratio R = 0.02, R = 0.2 and R = 0.6 with a sinusoidal waveform were applied. The material data of the crack growth rate were transformed into computational material characteristic by using multiple regression analysis according to relations (1-2) and with respect to shear lips evaluation (range of the transition zone definition). The values of the constraint factors a1 for the beginning and a2 for the end of the transition zone were calculated also by using multiple regression analysis, and the crack growth rate da/dN2 at the end of transition zone was another regression parameter.

3.3. Function of the geometry correction factor of the integrally stiffened panel

In addition to the material characteristics, the geometry correction factor p, which considers the actual geometry of a cracked structure in the general equation for the stress intensity factor (3), is the main factor that can significantly influence crack growth predictions.

where a is a stress function and a is a crack length. [S solutions for a row of geometric configurations can be found in Ref. [19-21], but more complicated geometries and loading conditions must be evaluated via numerical analysis by using the finite element method.

3.3.1. Simplified solution offt function of the panel

Although the analysed integrally stiffened panel contains six stringers, the panel loaded in mode I (tension) can be assumed as a 550 mm wide plate without stiffeners in the first approximation. For this case, the [S function is well-known in the expression of secant function (4), with a maximum error of 1% within the range of 0<a/W<0.8 according to Ref. [14]. This simplest[S relationship is illustrated as the brown curve in Fig. 4.

3.3.2. Numerical solution offt function of the panel

The numerical solution of the [S function of the panel using the finite element method was realized with the commercial finite element code NASTRAN by using the implemented crack element CRACK2D. A surface geometric model of the regular part of the panel was created and subsequently discretized by using 24 000 shell elements of QUAD4 type with a characteristic size of 2 mm. The following presumptions for the [S function "NASTRAN 1" generation were assumed: the stiffeners (stringers) were considered in the model only in the case of a crack propagating in the skin out of the stiffeners region (see top of Fig. 3 (a)). Each individual stiffener was considered to be fully broken in the case of a fatigue crack reaching the stiffener (see top of Fig. 3 (b)). This [S function, determined via the NASTRAN code, is shown as the blue curve in Fig. 4.

K = of3jm

P = 4 secO / W )

NASTRAN 1

crackadvance

step 1

crackadvance

NASTRAN 2

crackadvance

step 1

crackadva

Fig. 3: Illustration of presumptions used for the "NASTRAN 1"and "NASTRAN 2" p function definition: (a) panel configuration before the crack tip achieves a stiffener; (b) panel configuration after the crack tip achieves a stringer.

The second ft function of the panel was determined also by using the NASTRAN code using the same FE model as in the previous solution, but different presumptions were considered. In compliance with the experimental results, the crack growth through the stiffeners was assumed to be a quarter circle crack with axes placed on the skin outer side and stiffener surface. The /? function "NASTRAN 2" generated by using these presumptions is shown as the green curve in Fig. 4.

The solutions of the /? function obtained by the NASTRAN code are a little slower at the initial crack in comparison with the solution of the plate without stringers. From ~l/2 of the stringer spacing, the functional values of fi based on the NASTRAN solutions approach the solution of the middle tension crack specimen (M(T)) solution (SECANT), and at the centre of the stringers spacing, the match is almost achieved. The NASTRAN 1 and NASTRAN 2 solutions are the same beside the stringers, but in the area of the stringers, a qualitative difference appears. The rapid failure of the whole stringer (NASTRAN 1) causes a rapid increase of the /? value, which afterwards approaches the solution without stringers from the top, whereas the gradual failure of the stringer causes the solution to approach the solution without stringers from the bottom (see Fig. 4). The slightly higher values of /? between the second and third stringer in comparison with the solution without stringers is due to the higher stress level resulting from the higher cracked surface of the stringers. The functional values of /? for the NASTRAN 2 solution reflect the changes in the surroundings of the stringers more realistically than the NASTRAN 1 solution, as can be assumed from the presence of process lines on the fracture surfaces.

^ 2 CO.

1,5 1 0,5 0

1 SECANT

m NASTRAN 1 NASTRAN 2

0 50 100 150 200 250

Crack length a [mm]

Fig. 4: /} functions of the panel - plate without stringers (SECANT), NASTRAN 1 and NASTRAN 2.

3.4. Crack growth curves

The prediction of crack growth is realized by using SW AFGROW v.5.1.5.16. The analysis without retardation is also performed to stress the difference of both approaches. The NASGRO equation is used for crack growth rate vs. amplitude of the effective stress intensity factor characterisation. The Paris crack growth rate constant C = 2.51e~10, exponent n = 3.02 and constraint factor a = 2.56 were determined via regression analysis by using the minimal mean square error estimation of the experimental data. Other NASGRO constants p and q are chosen to be equal to zero such that the curvature in the threshold and critical region is neglected (for more information about the NASGRO equation implemented in SW AFGROW, see Ref. [11]). For an analysis that accounted for the retardation effect, the parameters of the FASTRAN model according to Chapter 3.2 are used. Crack growth curves for the performed analyses and experimental test are depicted in Fig. 5. Residual strength is not considered during prediction.

| 200 ro

2 100 u

Cycle [-]

Fig. 5: Crack growth curves of symmetrical crack for /3 solutions of plate without stringers (SECANT), NASTRAN 1 and NASTRAN 2.

During the laboratory experiment, 2 060 000 cycles were reached until failure. Assuming the linear damage accumulation principle, the fatigue life is 12% of the experimental one in the SECANT solution, 21% in the NASTRAN 1 solution and 22% in the NASTRAN 2 solution. Prediction that accounts for the retardation effect is very close to the experimental lifespan. The cycles reached at the crack length at the second stringer are in comparison to the experimental cycles, with a rate of 77% in the SECANT solution, 83% in the NASTRAN 1 solution and 86% in the NASTRAN 2 solution. All predictions are conservative estimations.

4. Discussion

The p functions offer different crack propagation curves when the linear damage accumulation principle is applied. The NASTRAN functions of the geometric correction factor give longer lifespan prediction in comparison with the standard fi function corresponding to the middle tension crack specimen (SECANT function of the geometry factor). The influence of the stiffeners is evident. A lifespan increase of approximately 75% due to stiffeners involvement in the case of the NASTRAN function of the geometry factor was predicted in comparison with the M(T) specimen (see Fig. 5). Nevertheless, all lives predicted by using the linear damage accumulation principle are strongly unrealistic. Experimental data show a more than four-fold longer lifespan in comparison with the prediction according to the linear damage accumulation principle. In contrast, the difference between individual functions of the geometry factor seems to be insignificant when the retardation effect is considered in the prediction. The involvement of the retardation effect in the predictions is obvious (see Fig. 5). There is almost no difference in the predicted lives if different functions of the geometry factor in the FASTRAN retardation model were used, although the individual prediction curves are a little bit different. The difference at the end of the crack growth curves is not relevant from the point of view of subcritical crack increments due to ductile tearing, which apparently occurred at the second stringer according to the experimental observation. The FASTRAN model is not able to cover the phenomenon of ductile tearing and thus the results of the computed crack growth are limited when the conditional value of fracture toughness is exceeded, as referenced in Ref. [18]. The crack length at which the conditional value of fracture toughness could be exceeded by the peak in the loading sequence is drawn in Fig. 5 and marked by Kq.

I • Experiment -SECANT -NASTRAN 1 NASTRAN 2 -----SECANT without reatrd. -----NASTRAN 1 without retard. -----NASTRAN 2 without retard.

ii 1 \ ! ■ J / 7 •

i! # f \

/ / MAX = 124,35 MPa

0 500 000 1 000 000 1 500 000 2 000 000 2 500 000

5. Conclusions

Fatigue life prediction of the integral skin-stringer panel was performed in several variants of solution with dependence on analysis demands, such as y§ function determination or the crack growth model. The retardation effect was either considered or completely omitted in the computational analysis. The retardation effect is an important element in the prediction method. The predicted fatigue life is approximately four times greater when the retardation effect is considered compared with the results of analyses that completely omitted it. The fatigue life is only in good agreement with the results of the detailed analysis where the retardation effect is considered.

The simplified analysis using the plate solution without stringers for y§ function determination seems to be a good assumption as a quick estimation of fatigue life. However, it can be stated that using the NASTRAN functions of the geometry factor showed better agreement with the experimental data; namely, the best results were achieved by assuming gradual fracture of the stringers.

Comparison of the experimental data and numerical analyses demonstrates an indispensable impact of the retardation effect on predicting fatigue crack propagation under variable amplitude loading. However, determination of computational parameters for the retardation model is not trivial. Well-founded material data via laboratory experiments and appropriate post processing are necessary. In contrast, the relevant estimations of crack growth, assessment of fatigue life, service intervals and accompanying saving cost are worth the effort.

Acknowledgements

This work was funded by the Ministry of Industry and Trade of the Czech Republic in the framework of institutional support.

References

[1] P.C. Paris et al., A rational analytical theory of fatigue. The Trend of Engineering. Vol.13, 1961, pp. 9-14

[2] R.G. Forman et al., Numerical analysis of crack propagation in cyclic-loaded structures. J.Basic Engeneering, Trans. ASME Vol.4, 1972, pp. 77-92

[3] W. Elber, The Significance of Fatigue Crack Closure. Damage tolerance in Aircraft Structures, ASTM STP 486, 1971, pp. 230-242

[4] J. Schijve, Some formulas for the crack opening stress level. Eng. Fracture Mechanics, Vol. 14, 1981, pp. 461-465

[5] J. Willenborg et al., A crack Growth Retardation Model Using an Effective Stress Concept, Technical memorandum 71-1-FBR, 1971, AFFDL-TM-71-1-FBR

[6] O.E. Wheeler, Spectrum Loading and Crack Growth, J. of Basic Engeneering, Trans. ASME Vol. 94, Series D, No.1, 1972, pp. 181

[7] A.U. de Koning, A Simple Crack Closure Model for Prediction of Fatigue Crack Growth Rates under Variable Amplitude Loading, ASTM STP 743, 1981, pp. 63-85

[8] D. Aliaga et al., A simple crack closure model for predicting fatigue crack growth under flight simulation loading. Mechanics of Fatigue Crack Closure, ASTM STP 982, 1987, pp. 461-504

[9] G. Baudin, M. Robert, Crack growth model for flight type loading, Proceeding of the 11th ICAF Symposium in the Netherlands, 1981.

[10] NASGRO Reference Manual - Version 4.02. NASA Johnson Space Center and Southwest Research Institute, 2002.

[11] J.C. Newman Jr., FASTRAN II - A fatigue crack growth structural analysis program, NASA Technical Memorandum 104159, Langley Research Center, Hampton, VA 23665, 1992.

[12] J.A. Harter, AFGROW Users' Guide and Technical Manual, Report No. AFRL-VA-WP-1999-3016, Air Force Research Laboratory, WPAFB,OH 45433-7542, 1999.

[13] T. Swift, Verification of Methods for Damage Tolerance Evaluation of Aircraft Structures to FAA Requirements. 12th International Commitee on Aeronautical Fatugue, Toulouse, France, 1983

[14] J. Schijve, Fatigue of Structures and Materials. Kluwer Academic Publishers, Second Edition; 2009.

[15] K.D. Dugdale, Yielding of steel sheets containing slits, J. Mech. Phys. Solids, 8, 1960, pp. 100-104

[16] J. Behal, L. Novakova, Stress state factor evaluation based on a fractographic analysis for use in the crack growth FASTRAN retardation model of the AFGROW computing code. Engineering Failure Analysis 35, 2013, pp. 645-651

[17] D. Broek , Elementary Engineering Fracture Mechanics, 4th edition, Kluwer Academic Publishers, 1982.

[18] J. Behal, P. Homola, R. Ruzek, The definition of initiation fracture resistance Jq and a practical effect of the phenomenon on the retardation of crack growth. International Journal of Structural Integrity. Paper accepted for publication. ISSN: 1757-9864

[19] D.P. Rooke, D.J. Cartwright, Stress intensity Factors. Her Majesty's Stacionary Office, London 1976.

[20] G.C Sih, Handbook of Stress Intensity Factors, Lehigh Universit, Bethlehem, Pensnsylvania, 1973.

[21] Y. Murakami, Stress Intensity Factors Hanbook. Pergamon Press, Oxford, 1987.