Fatigue Life and Fatigue Crack Growth Behavior of Nonthrough Cracksin Friction Stir-Welded 2024-T3 Aluminum Alloy*

Takao OKADA,1)† Shigeru MACHIDA,1) and Naoyuki WATANABE2)

1)Aeronautical Technology Directorate, Japan Aerospace Exploration Agency, Mitaka, Tokyo 181–0015, Japan2)Faculty of System Design, Tokyo Metropolitan University, Tokyo 191–0065, Japan

Friction stir-welded (FSW) joints are evaluated under two welding conditions to investigate their properties for ap-plication to aircraft structures. It is found that the welding condition affects the hardness profile, static strength, and fracturelocation. The tool mark near the burr, the kissing bond, the burr, and the edge of the specimen within the base material arefound to be the origins of fatigue fractures in FSW joints, and the welding condition affects these origins and fatigue life.The kissing bond is not the origin of a fracture when low stress is applied. The fatigue life of a FSW joint is longer than thatof a riveted joint. An evaluation of fatigue crack growth for each case via observation of the fracture surface indicates thatthe crack growth rate when the kissing bond is the origin of the fracture is close to that of the base material. The crackgrowth rate when the tool mark near the burr is the origin of the fracture is different from that of the base material: Over-estimating the stress intensity factor range based on assuming the crack geometry contributes to the difference.

Key Words: Mechanical Property, Friction Stir-Welded Joint, Origin of Fracture, Fatigue Crack Growth, Riveted Joint

Nomenclature

a: crack length in thickness directionb: half specimen widthc: crack length along width directionF: boundary correction factorfw: finite width correction factorf�: angular functiongi: curve fitting function

�K: stress intensity factor range�Keff: effective stress intensity factor range

Kt: stress concentration factorMj: curve fitting functionN: number of fatigue cyclesQ: shape factor of elliptical crackR: stress ratiot: specimen thicknessº: parametric angle of ellipse­: function defined in text·: stress

��: stress range²: normalized crack length

Subscriptsx: x directiony: y directioni: 1 or 2j: 1, 2 or 3

1. Introduction

It is necessary to identify the mechanical behaviors and en-vironmental resistances of the structural configurations usedin aircraft. A riveted joint, which is a typical mechanicallyfastened joint, has been used in aircraft structures for severaldecades. Despite this long period of time, changes to airwor-thiness requirements prompted by incident and accident in-vestigations require increased knowledge of the propertiesof this type of joint, and many research activities have beenconducted to evaluate its mechanical properties.1–3) Xiongand Bedair evaluated the stress carried at each rivet row usingan analytical and numerical procedure, and showed a goodcorrelation between them.4) Rahman et al. evaluated the bound-ary correction factor of the stress intensity factor for a crackemanating from a countersunk rivet hole in order to evaluatethe crack growth behavior under several stress conditions.5)

The results showed that a variety of boundary correction fac-tors can be used for evaluating widespread fatigue damage,and that the boundary correction factor is the highest for acrack close to the boundary condition. Chen et al. obtainedthe crack tip opening angle using elastoplastic finite elementanalysis in order to evaluate the residual strength of a crack ina fuselage panel, and showed good agreement between thetest and analysis results.6) Bakuckas7) and Wang et al.8) eval-uated the multisite damage to a riveted lap joint. Muller,9)

Terada,10) and Li et al.11) evaluated the effects of squeezinga rivet and the interference fit between the rivet and rivet hole,and showed that these parameters affect the fatigue propertiesof riveted joints. One of the authors of this paper has eval-uated the effects of the side edge due to tight clamping andusing stiffeners on riveted joints,12) as well as fatigue crackgrowth up to the first link of up to two adjacent cracks in wideriveted joints with a skin having a thickness of 1.27mm.13)

© 2019 The Japan Society for Aeronautical and Space Sciences+Received 15 September 2017; final revision received 15 August 2018;accepted for publication 3 September 2018.†Corresponding author, okada.takao@jaxa.jp

Trans. Japan Soc. Aero. Space Sci.Vol. 62, No. 1, pp. 20–31, 2019DOI: 10.2322/tjsass.62.20

20

In addition, other types of joints have been developed toreduce the manufacturing cost. Friction stir welding was in-vented in 1991, and it was intended to be applied to primaryaircraft structures. Therefore, friction stir welding has beenthe subject of similar investigations of riveted joints. Satoand Kokawa,14) Rhodes et al.,15) and Sutton et al.16) evaluatedthe microstructure formed by friction stir welding. It was ob-served that the global yielding and ultimate strengths wereaffected by the hardness profile of the joint. The banded mi-crostructure observed affected the fracture process of thejoint. Jones et al.17) investigated the hardness of a frictionstir-welded (FSW) 2024-T351 aluminum alloy and provedthat the microstructure affects the hardness profile of thejoint. The fatigue properties of an FSW joint were evaluatedby Lemmen et al.,18) Ali et al.,19) Dickerson and Przydatek,20)

Uematsu et al.,21) Besel et al.,22,23) and Gemme et al.,24) andthe fatigue crack growth properties have been evaluated byseveral researchers.25–30) Kumagai31) reported the static andfatigue properties, and the resistance to stress corrosioncracking of FSW joints for 2000- and 7000-series aluminumalloys. In addition, the FSW subcommittee in the Japan LightMetal Welding and Construction Association has evaluatedthe properties of FSW joints in order to determine the pre-vailing FSW technique and standardize the FSW proc-ess.32–38) The authors and collaborators have evaluated thestatic and fatigue properties of FSW joints39) and several fa-tigue crack growth properties of through cracks in FSWjoints.40)

A damage-tolerant design mostly requires the fatiguecrack growth behavior of the initial defect, such as a non-through crack. The location of the initial defect is usuallybased on the possibility of fatigue crack growth. In addition,the limit of validity (LOV) corresponds to the average fatiguelife of the structure when it does not incur widespread fatiguedamage in the as-is manufacturing condition. However, mostresearch on the fatigue crack growth of FSW joints has fo-cused on the fatigue crack growth of through cracks in theresidual stress field formed by friction stir welding. There-fore, the crack growth behavior of nonthrough cracks inFSW joints has to be evaluated before enabling applicationto aircraft structures. Moreover, evaluation of the crackgrowth behavior should be based identifying the origin ofthe fracture in the FSW joint. This knowledge is essentialto enable the evaluation of friction stir welding in term ofthe requirements for type certification of aircraft structures,even if the welding quality has been improved accordingto the studies mentioned above.

In this study, hardness, static, and fatigue tests were car-ried out for FSW joints consisting of a 2.0-mm-thick 2024-T3 aluminum alloy under two welding conditions in orderto evaluate their properties for application to aircraft struc-tures. The origins of fracture and the fatigue crack growth be-havior of the joint were identified by observation of the frac-ture surface. A fatigue test of a conventional riveted jointwith the same skin thickness was also conducted in orderto compare its mechanical properties to those of the FSWjoint.

2. Experimental Procedure

2.1. FSW jointsThe FSW panels used for this research were 2024-T3 alu-

minum alloy sheets with a thickness of 2mm and supplied byJapanese aircraft manufacturers. The rolling direction ofthese sheets was parallel to the welding direction, andFSW specimens were cut from the welded panels. Two typesof FSW joints were manufactured for this research. The tooland rig for welding were independently prepared by the man-ufacturers. Each welding condition was selected to be ad-equate for each welding tool on the basis of the experienceof the manufacturers; thus, the detailed welding conditionsare confidential. However, it was confirmed that other weld-ing conditions such as the tilt angle of the welding tool andthe control procedure for the protruding depth of the weldingtool were identical for both welding conditions. On the basisof the cross-section of the FSW joint, the sizes of the pin andshoulder for welding condition A were about 4 and 10mm,and those for welding condition B were 2 and 6mm, respec-tively. The ratios of the tool rotation speed to the tool travel-ing speed for each welding condition are 1.97 and 5.53, re-spectively.

Figure 1 shows the residual stress distributions along andperpendicular to the weld line for both welding conditions.The stresses in x and y directions indicate the stresses perpen-dicular and along the welding direction. To measure the re-sidual stress distribution, strain gauges were attached to bothsides of the jointed panel after welding. First, the strain wasmeasured with gauges attached to the joint. Next, the pieceswith the strain gauges were cut from the joint, and the strainwas measured again. The difference between these strainmeasurements corresponds to the residual stress. The resid-ual stress distribution parallel to the weld line for weld Bwas mostly symmetric with respect to the weld line, whereasthat for weld A exhibited a difference outside of the weld linein the region of compressive residual stress. The residualstress parallel to the welding direction had a peak tensilestress at the center of the weld line and a compressive peakoutside of the weld line. The residual stress distributions per-pendicular to the weld line for each welding condition weremostly symmetric with respect to the weld line. The peak

Fig. 1. Residual stress distributions of the FSW joints.

Trans. Japan Soc. Aero. Space Sci., Vol. 62, No. 1, 2019

21©2019 JSASS

tensile stress was observed outside of the weld line.Metallographic observations of welding conditions A and

B were presented in Okada et al.39) and Okada et al.,40) re-spectively.2.2. Hardness test of the FSW joints

The hardness distribution in the welding area was meas-ured to evaluate the mechanical features perpendicular tothe weld section. A microindentation hardness test procedure(ASTM E-384) was utilized. The hardness profile was meas-ured at the middle in the thickness direction. The indentationforce was 9.8N.2.3. Static test of the FSW joints

The specimen geometry for the static test is shown inFig. 2. Specimens were cut from an FSW panel such thatthe welding direction was perpendicular to the loading direc-tion. Further, the weld line was located at the center of thespecimen. No additional surface treatment was applied tothe specimens. A static test based on ASTM E-8 was con-ducted. Five specimens were tested for each condition, andthe average values of the yield strength, ultimate strength,and elongation were obtained.2.4. Fatigue test of the FSW joints

The specimen geometry used for the fatigue test is shownin Fig. 3. The requirements for preparing the fatigue testspecimens were the same as those for the static test speci-mens. Fatigue tests based on ASTM E-466 were conducted.The specimens were subjected to cyclic loading with a loadfrequency of 10Hz at room temperature and normal relativehumidity. The fatigue test conditions utilized in this test wereas follows: �� ¼ 135, 180, 225, or 270MPa for R ¼ 0:1.The stress range for the FSW joint was calculated usingthe thickness of the plate. This means that the thickness re-duction at the weld section resulting from FSW was not con-sidered in this case. After the fatigue test, the fatigue life wasrecorded, and the fracture surfaces were examined using acharge coupled device (CCD) camera and scanning electronmicroscopy (SEM).

The threshold of the stress intensity factor range and crit-ical stress intensity factor for the base material were approx-imately 3.4MPam0.5 for R ¼ 01) and 98MPam0.5,41) respec-tively.2.5. Fatigue test of the rivet joints

For comparison with the fatigue properties of the FSWjoints, riveted lap joints were fabricated by a Japanese air-craft manufacturer using plates of the same 2mm-thick2024-T3 aluminum alloy used for the FSW joints. A counter-sunk rivet made of a 2117-T4 aluminum alloy was used tofasten the joint. Figure 4 shows the geometry of a single

lap joint specimen riveted by three rows of three fasteners.The width and length of the joint were 60 and 420mm, re-spectively. The diameters of the rivets and rivet holes were3.96 and 4.0mm, respectively. The pitch of the rivets inthe vertical and parallel directions was 20mm. In order to ob-tain the fatigue properties under the as-is condition, no arti-ficial flaws were applied to the specimens. A constant-ampli-tude fatigue load was applied to the lap joints. Five types ofstress ranges, 123.5, 100.0, 79.8, 69.8, and 59.8MPa with astress ratio of 0.05, were utilized for this test. For the rivetedjoint, the stress range was calculated using the thickness andwidth of the joint, and reduction of the section as the result ofthe rivet hole was not considered. The test frequency was5Hz. After the fatigue test, the fatigue life was recordedand the fracture surfaces were examined by the CCD camera.

3. Results and Discussion

3.1. Observations of the welding sectionsFigure 5 shows optical micrographs for each welding con-

dition. The section is etched to see the microstructure, andthe bright area corresponds to the stir zone (SZ) and ther-mo-mechanically affected zone (TMAZ). As shown in thefigure, the heights of the tool mark and burr for welding con-dition A were about 0.05 and 0.45mm, respectively, and thethickness reduction at the weld line was about 0.15mm. Forwelding condition B, the heights of the tool mark and burrand thickness reduction at the weld line were about 0.02,0.22, and 0.06mm, respectively.3.2. Hardness profile

The indentation hardness test results are shown in Fig. 6.The average hardness of the base material was 127. From thisfigure, the hardness in SZ for welding condition A was sig-nificantly increased compared to that of the base material,whereas that for welding condition B was more stable onthe weld line. In addition, there was no region in which thehardness decreased for both welding conditions. The hard-ness value at SZ for welding condition B is also close to thatof the base material.

Fig. 2. Geometry of the static test specimen of an FSW joint.Fig. 3. Geometry of the fatigue test specimen of an FSW joint.

Fig. 4. Geometry of the fatigue test specimen of a riveted joint.

Trans. Japan Soc. Aero. Space Sci., Vol. 62, No. 1, 2019

22©2019 JSASS

Fratini et al. evaluated the hardness profile of the sectionin 2024-T341 FSW joint and obtained W-shaped distribu-tion.28) The hardness profile in FSW precipitation-hardenedaluminum alloys depends on the temper of the alloy, andthe hardness distribution observed is one of the typicalshapes. In the reference, the increased hardness at the centerof the weld was considered to be reprecipitation based on thenatural aging after welding. In addition, severe aging in theheat-affected zone (HAZ) would cause hardness to decrease.As reported in Jones et al.17) and Lemmen et al.,18) the hard-ness of a 2024-T3 FSW joint at SZ is, at most, the same asthat of the base material.

In Okada et al.,39) the authors evaluated the hardness dis-tribution of welding condition A at 0.5, 1.0, and 1.5mm fromthe bottom. The average hardness of the base material at 0.5and 1.5mm from the surface was 137 HV and that in the mid-dle was 127 HV. As shown in Fig. 6, the hardness in SZ andTMAZ for welding condition A is around the average hard-ness of the base material at 0.5 and 1.5mm from the surface.

The pin diameter for welding condition A was about twicethat for welding condition B. The plastic flow during weldingfor welding condition A would be larger than that for weld-ing condition B. The hardness of the SZ and TMAZ wasclose to that of the base material at the surface. This is be-cause a part of the metallic structure around the surfacemoved around the middle during stirring and reprecipitationoccurred around the center of the weld. The FSW panel inthis research is 2.0mm thick and is thinner than the plate

in other references. For thinner plate, heat input during weld-ing is transferred to the backing plate and rig sooner, whichinhibits aging and decreasing hardness in the HAZ.3.3. Static test results

Table 1 summarizes the static test results for the FSWjoints and base material. In order to obtain elongation atthe fracture, we fit ends of the fractured specimen lightly to-gether and measure the distance between the gauge mark.The elongation is obtained as the ratio of the difference be-fore and after the test divided by that of the initial distance.The typical fracture locations and fracture surfaces for thebase material and each welding condition are shown inFigs. 7, 8, and 9, respectively.

As shown in Fig. 7, the entire fracture surface of the basematerial was oriented at an angle with respect to the loadingdirection and was a typical feature of the thin sheet. As statedin Okada et al.,39) the average yield strength of a specimenfor welding condition A is 339.4MPa, 6% higher than thatof the base material, and its ultimate strength is 435.2MPa,10% lower than that of the base material. The elongation atfailure is 8.0% and is 40% of that of the base material.

Observations of the fracture surface for welding conditionA identified a kissing bond on the root side. As shown inFig. 8(d), it became the origin of static failure, and the frac-ture surface was oriented at an angle with respect to the load-ing direction, except for the surface by the kissing bond. Thekissing bond was considered to be one of the key reasons forthe decrease in the ultimate strength. The height of the kiss-ing bond for each specimen was measured from the fracturesurface. The average and variation coefficient of the kissingbond height were 0.152mm and 0.034, respectively. Addi-tionally, the variation coefficient of the yield stress and ten-sile strength were 0.009 and 0.052, respectively. The test

Fig. 6. Hardness profiles of the FSW sections.

Table 1. Static test results.

Yield strength(MPa)

Ultimate strength(MPa)

Elongation(%)

Base material 319.639) 482.039) 20.0Weld A 339.439) 435.239) 8.0Weld B 333.4 451.8 12.0

Fig. 7. Fracture location and surface of tensile test specimen for the basematerial: (a) top view, (b) side view, and (c) fracture surface.

Fig. 5. Morphology of friction stir welding: (a) welding condition A, (b)welding condition B.

Trans. Japan Soc. Aero. Space Sci., Vol. 62, No. 1, 2019

23©2019 JSASS

specimens for each welding condition were cut out from thesame FSW panel. Based on the test results, the variability ofthe property within the FSW panel is considered to be small.For welding condition B, the average yield strength is333.4MPa, 4% higher than that of the base material, andthe ultimate strength is 6% lower than that for the base mate-

rial. The elongation at failure is 12.0% and 60% of that of thebase material. Observations of the fracture surface and frac-ture location indicated that the origin of fracture was not thekissing bond. It was approximately 1.2mm away from thecenter of the weld line and a little outside of the bright areain Fig. 5. Therefore, the fracture origin is considered to be inthe HAZ. Thus, the reduced elongation due to welding con-tributed to reducing the ultimate strength.3.4. Fatigue test results

Figure 10 shows the fatigue test results for the FSW speci-mens and the base material in as-received surface conditions.In Fig. 10, the slope of the S–N curve when the kissing bondis the origin of the fracture is indicated by a thick solid line,and the curve when the tool mark near the burr is the origin ofthe fracture is indicated by a thick dashed line. The fatiguelife of the FSW joint was significantly lower than that ofthe base material. For a stress range of 135MPa, the speci-mens did not fail for 5 million cycles.

Observation of the fracture surface for welding conditionA identified two origins of fractures, the kissing bond andtool mark near the burr, as shown in Figs. 11 and 12. Failuredue to a fracture originating from the kissing bond was ob-served for stress ranges of 270 and 225MPa. The fatigue lifewas longer when the fracture originated at the tool mark nearthe burr.

For a stress range of 180MPa, only failure due to a frac-ture originating by the tool mark near the burr was observed.As shown in Fig. 10, the S–N curve for the tool mark nearthe burr was steeper than that for the kissing bond, and thefatigue life when the kissing bond was the origin of the frac-ture was shorter than that when the tool mark near the burrwas the origin of the fracture.

The fatigue life of welding condition B was higher thanthat of welding condition A. However, it was still lower thanthat of the base material. The origin of the fracture wasaround the burr (Fig. 13) or at the edge of the end point ofthe gauge length (Fig. 14).

As described in Sec. 3.1, the height of the burr and reduc-tion in thickness at the weld line for welding condition B are

Fig. 8. Fracture location and surface of tensile test specimen for weldingcondition A: (a) top view, (b) side view, (c) fracture surface, and (d) frac-ture surface at a higher magnification.

Fig. 9. Fracture location and surface of tensile test specimen for weldingcondition B: (a) top view, (b) side view, and (c) fracture surface.

Fig. 10. Fatigue test results for the FSW and riveted joints.�1Data at �� ¼ 275MPa are in Okada et al.39) �2Several data points at�� ¼ 275 and 225MPa are in Okada et al.39)

Trans. Japan Soc. Aero. Space Sci., Vol. 62, No. 1, 2019

24©2019 JSASS

smaller than those for condition A. Thus, their effects on theorigin of the fracture and fatigue life become less detrimen-tal, and some specimens do not fracture at the burr.

The fatigue test results for the riveted lap joint are indi-cated by the plus (+) symbols in Fig. 10. As shown in thefigure, the fatigue lives of the riveted joints are much shorterthan those of the FSW joints, and the riveted joints fracturedat a stress range of 59.8MPa. The test results do not show thefatigue limits of the riveted joints.

Photographs of the specimens after the fatigue test showthat the origin of the fracture was the manufacturing sideof the joint (Figs. 15 and 16). Inspection of the failed joint

also revealed fretting damage to the matting surfaces, and be-tween the sides of the rivet holes and shanks of the rivets. Fa-tigue cracks grew from this fretting damage in all cases.Therefore, fretting damage was found to be the origin of frac-turing in the riveted joints. After a crack forms from frettingdamage, the surface crack grows in the thickness and widthdirections.

The S–N data for several stress concentration ratios, Kt,are collected in MMPDS-10,42) and some of the data pointsare plotted in Fig. 10. The graph shows that the data forwelding condition A when the origin of the fracture is thetool mark near the burr are close to the data for Kt ¼ 2:0.

Fig. 11. Fracture surface for welding condition A when a kissing bond isthe origin of the fracture (�� ¼ 275MPa, failed at 5,000 cycles): (a) topview, (b) fracture surface between the kissing bond and final failure sur-face, (c) 0.30mm from the bottom surface, and (d) 0.74mm from the bot-tom surface.

Fig. 12. Fracture surface for welding condition A when the tool mark nearthe burr is the origin of the fracture (�� ¼ 225MPa, failed at 86,000cycles): (a) top view, (b) fracture surface between the tool mark nearthe burr and the final failure surface, (c) 0.72mm from the top surface,and (d) 1.20mm from the top surface.

Trans. Japan Soc. Aero. Space Sci., Vol. 62, No. 1, 2019

25©2019 JSASS

As mentioned earlier, the height of the tool mark is 0.45mm,which is quite large compared to the width of the weld line.Thus, it is difficult to obtain the stress concentration factorusing a numerical evaluation. However, an equivalent levelof stress concentration could be predicted by comparingthe S–N data with stress concentration. On the other hand,the equivalent level of stress concentration for welding con-dition B seems to be a certain value between 1.5 and 2.0.

For the riveted joint, the fatigue test results were close tothe MMPDS data for Kt ¼ 5:0. The stress concentration fac-tor of an open hole is 3.0, and that for a filled hole, which isthe case for a riveted joint, is somewhat lower. Observation

of the fracture surfaces identified that the fretting damageformed by relative slip between the matting surfaces wasthe origin of the fracture for the riveted joint. It is well-known that fretting damage to the surface degrades the fa-tigue life of an aluminum alloy. This means that stress con-centration is not the only reason for the lower fatigue life ofriveted joints. Owing to these two characteristics, the fatiguelimit of a riveted joint is lower than that of a FSW joint.

Compared to the FSW joint, the origin of the fracture forthe riveted joints is fretting damage, as identified by observa-tions of the fracture surfaces. Fretting damage is formed dueto a contact force and the relative movement at the matting

Fig. 13. Fracture surface for welding condition B when the burr is the ori-gin of the fracture (�� ¼ 225MPa, failed at 223,000 cycles): (a) topview, (b) fracture surface between the burr and the final failure surface,(c) 1.60mm from the origin of the fracture, and (d) 2.11mm from the ori-gin of the fracture.

Fig. 14. Fracture surface for welding condition B when the edge withinthe base material is the origin of the fracture (�� ¼ 180MPa, failed at299,000 cycles): (a) top view, (b) fracture surface between the edge andthe final failure surface, (c) 1.54mm from the origin of the fracture, and(d) 2.61mm from the origin of the fracture.

Trans. Japan Soc. Aero. Space Sci., Vol. 62, No. 1, 2019

26©2019 JSASS

surfaces of the skins or holes in the skin and rivet shanks dur-ing cyclic loading. As discussed by many researchers,1,9,12)

this is affected by the clearance of the rivet and rivet hole,and the squeezing force of the rivet. The contact force causesfretting stress on the matting surfaces. At areas where thestress applied is lower than the fretting stress, there is no rel-ative slip at the matting surface, and fretting does not occur.When the stress applied is greater than the fretting stress, rel-ative slip arises, and fretting occurs on the matting surfaces.When a higher stress is applied, relative slip occurs over larg-er areas; consequently, the edge of the slip area moves to-ward the centerline of the rivets. This can be confirmed byobservation of the specimen after fracturing occurs. The ori-gin of the fracture is located close to the centerline of the rivetholes at higher stress levels: this was 0.9mm away from thecenter of the rivet hole with a stress range of 100MPa, asshown in Fig. 15. On the other hand, the location of fractureorigin deviates 2.0mm from the centerline in the vertical di-rection at a stress range of 69.8MPa. Furthermore, as shownin Figs. 15 and 16, applying a lower stress results in a smallerarea of the ligament for final failure.

In this research, there are two origins of fracturing in FSWjoints for welding condition A: a kissing bond and a toolmark near the burr. SEM observations reveal striations onthe fatigue surfaces in both cases (Figs. 11 and 12). Stria-tions were observed between 0.25 and 0.82mm from the bot-tom along the thickness direction when the kissing bond wasthe origin of the fracture and between 0.62 and 1.25mm fromthe top along the thickness direction when the tool mark nearthe burr was the origin of the fracture.

The fatigue crack growth behavior of the specimens whenthe origin of the fracture was the kissing bond was obtainedfrom SEM observations, and is indicated by open circles inFig. 17. Five photos in each stress intensity factor range wereobtained. The average for several striation spacing in eachphoto is measured first, and then the average in each stressintensity factor range is calculated and plotted in the figure.The behavior when the tool mark near the burr was the originof the fracture was also evaluated using SEM observations,and is indicated by open squares in Fig. 17.

In Okada et al.,40) the fatigue crack growth rate of the basematerial for the middle tension specimen was obtained, andthe relationship between the fatigue crack growth rate and ef-fective stress intensity factor range was evaluated. The crackopening ratio was also calculated in Okada et al.,40) using theequation proposed by Newman.43) Using the da=dN–�Keff

curves and crack opening ratio, the da=dN–�K curves of thebase material for R ¼ 0:1 was obtained and is shown in asthe solid line in Fig. 17. The stress intensity factor rangefor welding condition A was obtained for the edge crack withequal crack length along the width direction of the specimen.The stress intensity factor was calculated using the equationdescribed in Tada et al.44) This assumption is based on obser-vation of the fracture surface.

�K að Þ ¼ ��ffiffiffiffiffiffi�a

pFð�Þ ð1Þ

F �� � ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2

��tan

��

2

!vuut 0:752þ 2:02�þ 0:37 1� sin��

2

! !30@

1A, cos

��

2

!ð2Þ

Fig. 15. Countersunk riveted joint fractured on manufacturing side (�� ¼100:0MPa, failed at 143,000 cycles): (a) top view and (b) fracture surface.

Fig. 16. Riveted joint fractured on the manufacturing side (�� ¼69:8MPa, failed at 727,000 cycles): (a) top view and (b) fracture surface.

Fig. 17. Fatigue crack growth rates of the FSW specimens.

Trans. Japan Soc. Aero. Space Sci., Vol. 62, No. 1, 2019

27©2019 JSASS

� ¼ a=t ð3Þ

As was shown in Fig. 1, maximum residual stress in theloading direction is approximately 20MPa and is, at most,10% of the applied stress range. The specimen width is sameorder of the width of the weld line. The residual stress causedFSW would be released when the specimen is cut out fromthe panel. Therefore, the stress intensity factor does not in-clude the effect of residual stress in this case.

The da=dN–�K curves for the specimens were obtainedfrom the data, and the fatigue crack growth life was eval-uated. For a specimen that failed from the kissing bond at ap-proximately 5,000 cycles at a stress range of 270MPa, obser-vations of the fracture surface showed that the height of thekissing bond was approximately 0.24mm. The maximumlength of the fatigue crack in this case was approximately0.92mm, and the calculated fatigue life between theselengths based on the da=dN–�K curves is approximately1,800 cycles. Therefore, the fatigue life until the formationof cracks at the kissing bond is about approximately 3,200cycles. On the other hand, for a specimen that failed at ap-proximately 86,000 cycles from the tool mark near the burrfor a stress range of 225MPa, multiple cracks were observedalong the tool mark, and they merged approximately0.62mm from the surface. The maximum length of the fa-tigue crack was approximately 1.3mm. The fatigue crackgrowth life for the crack between 0.62mm and 1.3mm wasalso calculated, and is approximately 1,300 cycles. Belowa crack length of 0.62mm, multiple cracks that form aroundthe tool mark grow simultaneously. Therefore, the fatiguelife developing a 0.62-mm crack from the tool mark is ap-proximately 84,700 cycles.

These results show that the kissing bond separates and be-comes a source of cracks much earlier than the tool mark nearthe burr. As shown in Fig. 10, the kissing bond does not actas the origin of a crack for a stress range of 180MPa; the toolmark near the burr becomes the origin of the fracture in thiscase. At lower stress levels, the kissing bond will not separateduring fatigue loading. Although it is not as strongly joinedas other regions of the weld, the kissing bond is still a phys-ical joint; thus, the changes in the stress around the kissingbond seem to be gradual as long as it is not opened. Onthe other hand, tool marks near a burr causes stress concen-trations around the area under a wider stress level, and theycould act as the origin of fractures at a lower stress levels inthis case.

For welding condition B, two origins of fractures were al-so observed: the burr and the edge of the specimen. Observa-tions of the fracture surface identified a quarter-ellipse shapefor fatigue failure in both cases. Striations were also observedby SEM, as shown in Figs. 13 and 14. The crack growthrates obtained by the striations are also plotted in Fig. 17.The stress intensity factor for the corner crack was calculatedusing the equations proposed by Newman and Raju.45)

�K að Þ ¼ ��ffiffiffiffiffiffi�a

pFða=c; a=t; c=b; �Þ ð4Þ

Fa

c;a

t;c

b; �

!

¼ 1

Q0:5M1 þM2

a

t

!20@ þM3

a

t

!41A g1g2f�fw

ð5Þ

Q ¼ 1þ 1:464a

c

!1:65ð6Þ

M1 ¼ 1:08� 0:03a

cð7Þ

M2 ¼ �0:44þ 1:06 0:3þ a

c

!,ð8Þ

M3 ¼ �0:5þ 0:25a

cþ 14:8 1� a

c

!15ð9Þ

g1 ¼ 1þ 0:08þ 0:4a

t

!20@

1A

0@

1A 1� sin�� �3 ð10Þ

g2 ¼ 1þ 0:08þ 0:15a

t

!20@

1A 1� cos�� �3 ð11Þ

f� ¼ a

c

!2cos�� �2þ sin�

� �20@

1A0:25

ð12Þ

fw ¼ 1� 0:2� þ 9:4�2 � 19:4�3 þ 27:1�4 ð13Þ

� ¼ c

b

ffiffiffia

t

sð14Þ

In these equations, º is a parametric angle of the ellipse.The same as for welding condition A, the effect of residualstress is not included in the evaluation of the stress intensityfactor for welding condition B. Based on observation of thefracture surface, the aspect ratio of the quarter-ellipse fatiguecrack at final failure was measured, and the ratio was used tocalculate the stress intensity factor for the entire length of thecrack.

Figure 17 shows that the crack growth behavior when theorigin of the fracture is the kissing bond for welding condi-tion A is close to that for the base material. The crack growthrates for welding condition B relatively coincides with that ofthe base material. However, the crack growth rate for weld-ing condition A when the tool mark near the burr is the originof the fracture is somewhat different from that of the basematerial, and the difference is larger for longer crack lengths.In this paper, the authors simulate the crack for welding con-dition A as an edge crack from the top or the bottom depend-ing on the origin of the fracture. In the case of the originbeing the tool mark near the burr, the actual crack front alongthe width direction was curved, not straight. The stress inten-sity factor range calculated would be overestimated based onthis assumption. The stress intensity factor range for thesemi-elliptical surface crack of which the width was the sameas the specimen width was calculated using NASGRO,46)

Trans. Japan Soc. Aero. Space Sci., Vol. 62, No. 1, 2019

28©2019 JSASS

and this was compared with the results obtained for the edgecrack. The ratio of the stress intensity factor for the semi-el-liptical surface crack to the edge crack was 0.83 for a 0.62-mm crack and was 0.45 for a 1.26-mm crack. The accuratestress intensity factor has to be obtained using the three-di-mensional finite element method (3D FEM), because thecrack length is not zero at either end, and then the actualcrack shape is not semi-elliptical. However, the stress inten-sity factor range for the actual crack is less than that of theedge crack. This implies that the da=dN–�K curves forwelding condition A in the case when a tool mark near theburr is the fracture origin would shift left and become closeto that of the base material.

In the case when a kissing bond was the fracture origin, thecrack in SZ grew along the thickness direction. As shown inFig. 6 and Okada et al.,39) the hardness in SZ is approxi-mately 140 HV and there is little variability along the thick-ness direction. In other cases, the crack in HAZ grew or thebase material, and the hardness around the surface and in themiddle was 137 and 127, respectively. As shown inFigs. 11–14, no cleavage surface was observed and it is con-sidered a ductile-type striation. As indicated in Fig. 17, thereis little difference in the fatigue crack growth rate. Based onthe results, a difference in hardness would not affect the fa-tigue crack growth feature in this case.

It is well known that the growth rate of a small crack haslarge scatter because of plasticity, metallurgy, roughness ofthe crack surface, and/or oxide/corrosion on the crack sur-face. Therefore, the evaluation of small crack growth usingthe stress intensity factor range based on linear elastic frac-ture mechanics is not appropriate.47–49) In the NASTROReference Manual,46) the scatter of the crack growth ratefor 2024-T3 aluminum alloy at R ¼ 0 was observed for�K < 10MPam0.5. Most of the stress intensity factor rangein this research was above 10MPam0.5. In addition, the min-imum crack length in thickness direction in case the striationswere measured was 0.3mm and it was one order higher thanthe grain size in thickness direction. Therefore, the crackgrowth behavior in this research was evaluated using thestress intensity factor range based on linear elastic fracturemechanics.

The NRCC and JAXA have conducted collaborative re-search work to assess the fatigue life of countersunk rivetedjoints.13) The fatigue test results identified fretting damage tothe matting surfaces of the joint as the origin of fractures thatformed during fatigue loading. The life of the first fatiguecrack observed, that of the first merging of adjacent cracks,and the final fatigue life were 1.95 million, 2.11 million,and 2,113,534 cycles with a stress amplitude of 60.5MPaandR ¼ 0:069. In Li et al.,13) the fatigue life of a riveted jointup to the first merging of two adjacent rivet holes was div-ided in two regions: the fatigue life to form a 0.5-mm cornercrack at a rivet hole in the matting surface, and the fatiguecrack growth life at the 0.5-mm corner crack to first merging.On the basis of fatigue tests of riveted joints and 3D FEManalysis of a joint from manufacturing processes subjectedto fatigue loading, the fatigue life to the formation of a 0.5-

mm corner crack was predicted using the Smith–Watson–Topper (SWT) strain life equation. Further crack growthfrom a 0.5-mm crack to the first merging was predicted usingthe fatigue crack growth characteristics of the aluminum al-loy and the effect of an adjacent crack using an in-houseNRCC code, Can-GROW. Good correlation was obtainedbetween the fatigue test results and the predicted service life.

Observation of the fracture surfaces of the riveted joints inthis study identified fretting damage to the joint as the originof the fracture. Because fretting damage is located aroundrivet holes, it is discretely distributed. The merging of adja-cent cracks was associated with tensile failure of the rest ofthe ligament at a higher applied stress (Fig. 15), whereas itoccurred through the merging of fatigue cracks under a lowerapplied stress (Fig. 16). This difference is only related to themechanism of merging adjacent cracks. The splitting of thefatigue phase into the two regions mentioned above seemsto be adequate given the observations of the fracture sur-faces. This means that the crack growth behavior in a rivetedjoint could be modeled using the behavior in the base mate-rial when applying an appropriate stress intensity factor. Asdescribed in Li et al.,13) detailed modeling of the riveted joint,including the manufacturing process, and the modeling of theeffects of an external load and adjacent crack are imperativefor evaluating the stress intensity factors of cracks in a riv-eted joint.

In 2011, the LOV was mandated as a new requirement fortransport-category aircraft. The LOV is the period of time upto which it has been demonstrated that widespread fatiguedamage (WFD) is unlikely to occur in an airplane structureby virtue of its inherent design characteristics and mainte-nance actions. In order to meet this requirement, a possiblescenario for WFD must first be established. Then, it has tobe demonstrated that WFD based on the scenario does notoccur before the LOV. Although its validation does not nec-essarily require the prediction of WFD, a technique for pre-dicting the occurrence of WFD is imperative for the develop-ment of lighter aircraft structures with a longer operation life.The procedure described above for testing riveted jointswould be one candidate technique for predicting WFD.The test results for the FSW joints obtained in this researchshow several types of fatigue scenarios. However, additionalevaluations comparing the FSW joints to riveted joints wouldbe necessary to comply with the WFD requirement. As ourresearch demonstrates, the possibility of locating the originof fatigue fractures within the base material depending onthe welding condition and higher fatigue limit as comparedto that of the riveted joint would mitigate the number of eval-uations.

4. Conclusion

In order to evaluate the features of FSW joints for the pur-pose of applying them to aircraft structures, hardness, static,and fatigue tests are conducted for two welding conditions.The hardness profile indicates that the welding condition af-fects the hardness distribution within the weld line, and a rel-

Trans. Japan Soc. Aero. Space Sci., Vol. 62, No. 1, 2019

29©2019 JSASS

atively stable hardness profile can be obtained depending onthe welding condition. No typical hardness reduction outsidethe weld line was observed in this research.

Static test results identified that a kissing bond reduces theultimate strength of the FSW joint in cases where one ex-isted. It also reduces the yield stress of the joint. In the casewhere a kissing bond did not form, the heat-affected zone be-came the origin of the fracture, reducing the ultimate strengthof the FSW joint; however, there was no significant reductionin yield stress.

Fatigue tests and observation of the fracture surfaces iden-tified a tool mark near a burr and a kissing bond as the originsof fractures for welding condition A, whereas a burr and anedge within the base material were identified as the origins offractures for welding condition B. The differences in thesizes of the tool mark and burr for each welding condition af-fect the difference in the origin of the fracture for cases inwhich fractures occurred. For welding condition A, a toolmark near a burr can be the origin of a fracture at stressesabove the fatigue limit because the stress concentrationaround it is a source of fatigue failure. On the other hand,the kissing bond is not an origin of fractures at lower stresslevels because it does transfer stress at such stress levels.When the kissing bond is the origin of a fracture, the fatiguelife of the joint is much shorter than that when a tool marknear the burr is the origin of fracture. From an evaluationof fatigue crack growth for each case based on observationsof the fracture surfaces, the crack growth rate when a kissingbond is the origin of the fracture is close to that of the basematerial obtained from the middle tension specimen. Thecrack growth rate when a tool mark near a burr is the originof the fracture is different from that of the base material;overestimating the stress intensity factor range based on as-suming the crack geometry contributes to the difference.

For a riveted lap joint, the locations of the origins of frac-tures in the areas of fretting damage along the loading direc-tion are close to the centerlines of the rivet holes at higherstress levels and are separated from the centerlines at lowerstress levels. When higher stresses are applied, the area ofrelative slip between the matting surfaces is larger, and theedge of the slip area moves toward the centerline of the riv-ets. The FSW joint is found to have a longer fatigue life thanthat of the riveted joint. Although several origins of fractureand failure behaviors demand more resources for their eval-uate, the possibility of locating the origin of fatigue fracturewithin the base material depending on the welding conditionand the higher fatigue limit of an FSW joint compared to thatof the riveted joint would mitigate the drawbacks regardingthe WFD requirement.

References

1) Schijve, J.: Fatigue of Structures and Materials, Kluwer AcademicPublishers, Dordrecht, 2001.

2) Rosenfield, M. S.: Damage Tolerance in Aircraft Structures, ASTMSTP 486, 1970, pp. 164–214.

3) Chang, J. B. and Hudson, C. M.: Methods and Models for PredictingFatigue Crack Growth under Random Loading, ASTM STP 748, 1981,

pp. 53–84.4) Xiong, Y. and Bedair, O. K.: Analytical and Finite Element Modeling

of Riveted Lap Joints in Aircraft Structure, AIAA J., 37 (1999),pp. 93–99.

5) Rahman, A., Bakuckas, J. G., Jr., Bigelow, C. A., and Tan, P. W.:Boundary Correction Factors for Elliptical Surface Cracks Emanatingfrom Countersunk Rivet Holes, AIAA J., 38 (2000), pp. 2171–2175.

6) Chen, C.-S., Wawrzynek, P. A., and Ingraffea, A. R.: ResidualStrength Prediction of Aircraft Fuselages Using Crack-Tip OpeningAngle Criterion, AIAA J., 40 (2002), pp. 566–575.

7) Bakuckas, J. G., Jr.: Full-Scale Testing and Analysis of Fuselage Struc-ture Containing Multiple Cracks, DOT/FAA/AR-01/46, 2001.

8) Wang, L., Chow, W. T., Kawai, H., and Atluri, S. N.: ResidualStrength of Aging Aircraft with Multiple Site Damage/Multiple Ele-ment Damage, AIAA J., 36 (1998), pp. 840–847.

9) Muller, R. P. G.: An Experimental and Analytical Investigation on theFatigue Behavior of Fuselage Riveted Lap Joints. The Significance ofthe Rivet Squeeze Force, and a Comparison 2024-T3 and Glare3,Ph.D. Thesis, Delft University of Technology, 1995.

10) Terada, H.: Structural Fatigue and Joint Degradation, Int. J. Fatigue,23 (2001), pp. s21–s30.

11) Li, G., Shi, G., and Bellinger, N. C.: Residual Stress/Strain in Three-Row Countersunk Riveted Lap Joints, J. Aircraft, 44 (2007),pp. 1275–1285.

12) Terada, H. and Okada, T.: Problems of Laboratory Tests for DurabilityEvaluation of Full-Scale Structures, Int. J. Fatigue, 31 (2009),pp. 1068–1072.

13) Li, G., Renaud, G., Liao, M., Okada, T., and Machida, S.: A Method-ology for Assessing Fatigue Life of a Countersunk Riveted Lap Joint,Adv. Aircr. Spacecr. Sci., 4 (2017), pp. s1–s19.

14) Sato, Y. S. and Kokawa, H.: Distribution of Tensile Property and Mi-crostructure in Friction Stir Weld of 6063 Aluminum, Metall. Mater.Trans. A, 32 (2001), pp. 3023–3031.

15) Rhodes, C. G., Mahoney, M. W., Bingel, W. H., Spurling, R. A., andBampton, C. C.: Effects of Friction Stir Welding on Microstructure of7075 Aluminum, Scr. Mater., 36 (1997), pp. 69–75.

16) Sutton, M. A., Yang, B., Reynolds, A. P., and Taylor, R.: Microstruc-tural Studies of Friction Stir Welds in 2024-T3 Aluminum,Mater. Sci.Eng. A, 323 (2002), pp. 160–166.

17) Jones, M. J., Heurtier, P., Desrayaud, C., Montheillet, F., Allehaux, D.,and Driver, J. H.: Correlation between Microstructure and Microhard-ness in a Friction Stir Welded 2024 Aluminum Alloy, Scr. Mater., 52(2005), pp. 693–697.

18) Lemmen, H. J. K., Alderliesten, R. C., and Benedictus, R.: Fatigue Ini-tiation Behaviour throughout Friction Stir Welded Joints in AA2024-T3, Int. J. Fatigue, 32 (2010), pp. 1928–1936.

19) Ali, A., Brown, M. W., and Rodopoulos, C. A.: Modelling of CrackCoalescence in 2024-T341 Al Alloy Friction Stir Welded Joints, Int.J. Fatigue, 30 (2008), pp. 2030–2043.

20) Dickerson, T. L. and Przydatek, J.: Fatigue of Friction Stir Welds inAluminum Alloys That Contain Root Flaws, Int. J. Fatigue, 25(2003), pp. 1399–1409.

21) Uematsu, Y., Tokaji, K., Shibata, H., Tozaki, Y., and Ohmune, T.: Fa-tigue Behavior of Friction Stir Welds without Neither Welding Flashnor Flaw in Several Aluminum Alloys, Int. J. Fatigue, 31 (2009),pp. 1443–1453.

22) Besel, M., Besel, Y., Alfaro Mercado, U., Kakiuchi, T., and Uematsu,Y.: Fatigue Behavior of Friction Stir Welded Al-Mg-Sc Alloy, Int. J.Fatigue, 77 (2015), pp. 1–11.

23) Besel, Y., Besel, M., Alfaro Mercado, U., Kakiuchi, T., and Uematsu,Y.: Influence of Joint Line Remnant on Crack Paths under Static andFatigue Loading in Friction Stir Welded Al-Mg-Sc Alloy, Fratturaed Integrita Strutt., 35 (2016), pp. 295–305.

24) Gemme, F., Verreman, Y., Dubourg, L., and Wanjara, P.: Effect ofWelding Parameters on Microstructure and Mechanical Properties ofAA7075-T6 Friction Stir Welded Joints, Fatigue Fract. Eng. Mater.Struct., 34 (2011), pp. 877–886.

25) Bussu, G. and Irving, P. E.: The Role of Residual Stress and Heat Af-fected Zone Properties on Fatigue Crack Propagation in Friction StirWelded 2024-T351 Aluminum Alloy, Int. J. Fatigue, 25 (2003),pp. 77–88.

Trans. Japan Soc. Aero. Space Sci., Vol. 62, No. 1, 2019

30©2019 JSASS

26) Lemmen, H. J. K., Alderliesten, R. C., and Benedictus, R.: Macro andMicroscopic Observation of Fatigue Crack Growth in Friction StirWelded Aluminum Joints, Eng. Fract. Mech., 78 (2011), pp. 930–943.

27) Ma, Y. E. and Irving, P.: Residual Stress Effects and Fatigue Behaviorof Friction-Stir-Welded 2198-T8 Al-Li Alloy Joints, J. Aircraft, 48(2011), pp. 1238–1244.

28) Fratini, L., Pasta, S., and Reynolds, A. P.: Fatigue Crack Growth in2024-T351 Friction Stir Welded Joints: Longitudinal Residual Stressand Microstructural Effects, Int. J. Fatigue, 31 (2009), pp. 495–500.

29) Milan, M. T., Bose Filho, W. W., Ruckert, C. O. F. T., and Tarpani, J.R.: Fatigue Behavior of Friction Stir Welded AA2024-T3: Longitudi-nal and Transverse Crack Growth, Fatigue Fract. Eng. Mater. Struct.,31 (2007), pp. 526–538.

30) Pasta, S. and Reynolds, A. P.: Residual Stress Effects on Fatigue CrackGrowth in a Ti-6Al-4V Friction Stir Weld, Fatigue Fract. Eng. Mater.Struct., 31 (2008), pp. 569–580.

31) Kumagai, M.: Application of Friction Stir Welding for Aircraft Struc-ture, J. Light Metal Weld. Constr., 41 (2003), pp. 208–214 (in Japa-nese).

32) Namba, K.: Studies on Characteristics of Friction Stir Welded Joints inStructural Aluminum Alloys: Outline of Research Work, J. Light Met-al Weld. Constr., 45 (2007), pp. 141–150 (in Japanese).

33) Oiwa, N., Kida, K., Iwaki, S., Okada, T., Ishikawa, T., Eguchi, N., andNamba, K.: Imperfections in Friction Stir Welded Zone and Their Pre-cision Non-Destructive Testing—Studies on Characteristics of FrictionStir Welded Joints in Structural Aluminum Alloys (Report 1)—,J. Light Metal Weld. Constr., 45 (2007), pp. 151–162 (in Japanese).

34) Okada, T., Kida, K., Iwaki, S., Ishikawa, T., Eguchi, N., Oiwa, N., andNamba, K.: Metallurgical Features of Friction Stir Welded Zone—Stu-dies on Characteristics of Friction Stir Welded Joints in Structural Alu-minum Alloys (Report 2)—, J. Light Metal Weld. Constr., 45 (2007),pp. 163–175 (in Japanese).

35) Hori, H., Tanikawa, H., Seo, N., and Namba, K.: Tensile Properties ofFriction Stir Welded Joints—Studies on Characteristics of Friction StirWelded Joints in Structural Aluminum Alloys (Report 3)—, J. LightMetal Weld. Constr., 45 (2007), pp. 202–210 (in Japanese).

36) Nagano, Y., Takeda, J., Hashimoto, T., Miyamichi, T., and Namba, K.:Bending Properties of Friction Stir Welded Joints—Studies on Charac-teristics of Friction Stir Welded Joints in Structural Aluminum Alloys(Report 4)—, J. Light Metal Weld. Constr., 45 (2007), pp. 211–218 (inJapanese).

37) Okubo, K., Kamioka, M., and Namba, K.: Fatigue Properties of Fric-tion Stir Welded Joints—Studies on Characteristics of Friction StirWelded Joints in Structural Aluminum Alloys (Report 5)—, J. LightMetal Weld. Constr., 45 (2007), pp. 219–230 (in Japanese).

38) Iwaki, S., Kida, K., Okada, T., Ishikawa, T., Eguchi, N., Oiwa, N., andNamba, K.: Features of Friction Stir Welded Joints though Post-WeldHeat Treatments—Studies on Characteristics of Friction Stir WeldedJoints in Structural Aluminum Alloys (Report 6)—, J. Light MetalWeld. Constr., 45 (2007), pp. 231–241 (in Japanese).

39) Okada, T., Suzuki, M., Miyake, H., Nakamura, T., Machida, S., andAsakawa, M.: Evaluation of Crack Nucleation Site and MechanicalProperties for Friction Stir Welded Butt Joint in 2024-T3 AluminumAlloy, Int. J. Adv. Manuf. Technol., 50 (2010), pp. 127–135.

40) Okada, T., Machida, S., Nakamura, T., Tanaka, H., Kuwayama, K.,and Asakawa, M.: Fatigue Crack Growth of Friction Stir Welded Alu-minum Alloy, J. Aircraft, 54 (2017), pp. 737–746.

41) Niu, C. Y. M.: Airframe, Stress Analysis and Sizing, 2nd ed. HongKong Conmilit Press Ltd., Hong Kong, 1999.

42) Metallic Materials Properties Development and Standardization(MMPDS)-10, MMPDS Coordination Activity, Columbus, 2015.

43) Newman, J. C., Jr.: A Crack Opening Stress Equation for FatigueCrack Growth, Int. J. Fracture, 24 (1984), pp. R131–R135.

44) Tada, H., Paris, P. C., and Irwin, G. R.: The Stress Analysis of CracksHandbook, Del Research, Hellertown, 1973.

45) Newman, J. C., Jr. and Raju, I. S.: Stress-Intensity Factor Equations forCracks in Three-Dimensional Finite Bodies, NASA Langley ResearchCenter, Hampton, 1981.

46) NASGRO Reference Manual, Version 8.01 Final, Southwest ResearchInstitute, 2016.

47) Newman, J. C., Jr. and Edwards, P. R.: Short Crack Growth Behaviourin Aluminum Alloy—An AGARD Cooperative Test Programme,AGARD Report No. 732, 1988.

48) Newman, J. C., Jr., Phillips, E. P., and Swain, M. H.: Fatigue-life Pre-diction Methodology Using Small-crack Theory, Int. J. Fatigue, 21(1999), pp. 109–119.

49) Newman, J. C., Jr.: Advance in Fatigue and Fracture Mechanics Analy-sis for Metallic Aircraft Structures, NASA TM-2000-210084, 2000.

Toshio OgasawaraAssociate Editor

Trans. Japan Soc. Aero. Space Sci., Vol. 62, No. 1, 2019

31©2019 JSASS