Assessing the Effects of Riveting Induced Residual Stresseson Fatigue Crack Behaviour in Lap Joints by means of

Fractography

C.D. Ransa, R.C. Alderliestenb, P.V. Straznickya

aDepartment of Mechanical and Aerospace Engineering, Carleton University, 1125 Colonel By Drive, K1S5B6, Ottawa, Canada

bFaculty of Aerospace Engineering, Delft University of Technology, Kluyverweg 1, 2629 HS, Delft, theNetherlands

Abstract

The interference fit provided by solid rivets induces a residual stress field beneficial to thefatigue performance of riveted lap joints. Attributing benefits in crack growth nucleation andgrowth behaviour within this residual stress field is non-trivial. Once fatigue cracks becomevisible on the surface of a joint they have already grown beyond the region of beneficial residualstress. In order to circumvent this problem, fractographic techniques were employed to evaluatepostmortem the influence of the rivet squeeze force and resulting residual stress field on crackbehaviour. Results demonstrated that within the range of rivet squeeze forces studied, a 3-foldreduction in crack growth rate is achievable at high rivet squeeze forces, representing a markedimprovement in damage tolerance of the joint.

1

2

1 Introduction

The hole filling properties of solid rivets are a key contributor to the fatigue performance of rivetedlap joints. As a rivet is installed, it first expands to fill the hole then further expands, cold workingthe surrounding sheet material and generating a beneficial residual compressive stress field. Thenature of this residual stress field plays an important role in the nucleation and growth of cracks inthe vicinity of the rivet hole.

Studying the effects of riveting induced residual stresses on the fatigue performance of lap jointsrequires consideration of several factors. The region of beneficial residual stresses is localizedaround the rivet hole and is obstructed by the rivet head, reducing the effectiveness of conventionalcrack detection and measurement techniques. The secondary bending stresses resulting from rivetrotation produce highly elliptical crack fronts, causing the crack to remain below the visible sheetsurface through most of the region of interest. As a result, overall fatigue lives, crack time-to-detection lives, and growth rates for crack tips beyond the region of residual stress are relied on asmeasures; however, such secondary effects can only provide a limited understanding of the residualstress field. Ideally, knowledge of the impact on nucleation and crack growth within the region ofresidual stress is required.

Fractography provides an alternative means for studying the effects of residual stress on cracknucleation and growth. Postmortem examination of fracture surfaces from failed lap joints, typicallyusing a scanning electron microscope (SEM), can provide direct evidence of crack growth behaviourin a region that is otherwise not observable during the fatigue process.

Two methods are often employed in extracting crack growth information in fractographic exam-inations.

The first method examines the spacing between individual fatigue striations on a fatigue surface[2, 6, 9]. By measuring the spacing at a specific point, the crack growth rate at that point canbe determined directly. This method provides an accurate and flexible means to determine crackgrowth rates at specific points on the fracture surface, but does not provide an efficient means forvisualizing crack front shapes.

The second method involves identifying striation patterns resulting from load variations withinthe fatigue spectrum [3,5,7,10]. These variations may naturally occur in the fatigue spectrum or canbe introduced with periodic under- or over-loads known as marker loads. These patterns outline theshape of the crack front giving an indication of shape and location of the crack at a given momentin the spectrum. If these patterns can be associated with a period of time within the spectrum, thenthe crack growth history can be reconstructed.

The present investigation aims to study the crack growth behaviour within the region of residualstress in riveted lap joints. To this end, a series of fatigue tests using a marker load spectrumdeveloped by Piasick and Willard [10] were undertaken. Fractography methods were employed toreconstruct crack growth behaviour for these tests allowing the effects of rivet squeeze force and theresulting residual stress state on fatigue performance to be evaluated.

The work presented in this paper represents the second part of a larger research effort to betterunderstand the role of rivet installation on fatigue performance in riveted lap joints. The first partof this research effort focused on quantifying the effects of rivet installation on the stress state inriveted lap joints. Using finite element methods, the riveting process was simulated in order toquantify the effect of rivet installation force (squeeze force or FSq) on the resulting residual stressfield [11–13]. Follow-up finite element studies examined the effects of rivet installation on the stressstate in loaded lap joints [11,13]. Particular attention was placed on secondary bending stresses due

3

to their role in crack nucleation. Details on these studies can be found in publications by the presentauthors [11–13].

2 Fatigue Crack Growth Tests

Two sets of fatigue tests were carried out in order to investigate the effects of rivet installation on thefatigue performance of riveted lap joints. The first set was undertaken to quantify the effects of rivetsqueeze force and rivet type on overall fatigue life. The second set was designated for fractographiccrack growth reconstruction. Due to the time consuming nature of such reconstructions, these testswere not run until final fracture. Rather, once cracks of a given size were present, the tests werestopped and the specimen disassembled for fractographic analysis.

2.1 Test Specimen Geometry

Fatigue tests were conducted on 2-row riveted lap joint specimens as illustrated in Figure 1. Speci-mens were manufactured with 2024-T3 Alclad sheet and 2117-T4 rivets. Alclad refers to a thin layerof pure aluminum placed on the sheets for corrosion resistance. Specimens containing countersunk(NAS1097AD4-4) and universal (MS20426AD4-4) rivets were made. Specimens manufactured us-ing rivet squeeze forces of 1500, 2000, 2500, and 3000 lbf, corresponding to D/Do ratios of 1.4, 1.5,1.6 and 1.7 respectively, were studied. These D/Do ratios fall within a range of generally accepteddriven rivet head geometries for aerospace applications [8].

The orientation of the rivets in the two rivet rows was reversed giving a true anti-symmetricgeometry. This modification, although not practical in a commercial application, is useful for ex-perimental investigations as it ensures the same conditions are present in the two fatigue criticallocations. This anti-symmetric configuration results in two specimen types, designated type 1 andtype 2 (Figure 1). For the type 1 configuration, crack nucleation is assured to occur next to a man-ufactured rivet head (countersunk or universal rivet head), simulating failure within the outer orexposed sheet in a practical joint. For specimen type 2, failure at the driven rivet head is ensured aswould be the case for failure of the inner joint sheet. Use of these two specimen types permits theeffects of rivet installation on inner and outer sheet failure to be studied independently.

2.2 Fatigue Test Spectrum

In order to facilitate the fractographic reconstructions of crack growth, a specialized test spectrumwhich would introduce distinctive features on the fracture surface was needed for the current tests.For this purpose, a constant amplitude spectrum that included periodic underload cycles to createband features on the fracture surface was used. This spectrum, illustrated in Figure 2 and used inprevious fractography studies [3,5,10], introduced patterns of 10, 4, and 6 bands which could easilybe identified on the fracture surface.

A maximum stress level, Smax, of 100 MPa was specified. To prevent damage of the markerbands on the fracture surface, a stress ratio R = Smax/Smin = 0.1 was specified to minimize contactbetween the opposing fracture surfaces during unloading. All tests were carried out at a frequencyof 10 Hz on an MTS 318.10 load frame equipped with a 10 kN load cell and a TestStar IIsTM

controller.Although the marker band spectrum was only required for the set of tests designated for crack

growth reconstructions, the spectrum was used in all tests. To account for the difference in crack

4

growth at the 75% Smax cycles, a correction factor derived using Elber’s plasticity induced crackclosure concept [4] was used. This approach was developed by Fawaz [5] and applied in additionstudies completed by de Rijck [3]. For the loading case used in this investigation, de Rijck [3]determined that the ratio of the crack growth rate for the 75% Smax and Smax cycles is 9.22% in2024-T3 and verified this for open-hole fatigue specimens. For the derivation of this factor, thereader is encouraged to refer to the work of de Rijck [3].

Given the 9.22% crack growth correction factor, one program of the marker band fatigue spec-trum, which consists of 20800 cycles, reduces to 13538 baseline cycles. This correction factoris based on differences in crack growth and does not necessarily account for differences in cracknucleation between the baseline and marker load cycles. All fatigue lives and crack growth rateswithin this paper will be presented with the applied correction factor unless otherwise noted.

2.3 Test Matrix

Two series of fatigue tests were undertaken in this investigation: a series to determine overall fatiguelife, and a series for fractographic crack growth reconstruction (designated F and SEM respectively).Furthermore, two specimen configurations (Type 1 and 2 in Figure 1) and specimens containing tworivet types were tested. In order to distinguish between the various specimen configurations and testtypes undertaken, the following specimen naming convention was used:

CSK3.0-T1-F-01 Specimen number Specimen series (F or SEM) Specimen type (T1 or T2) FSq in 1000’s lbf Rivet type (U or CSK)

Where the specimen series designators F and SEM refer to the overall fatigue life and fractographicreconstruction series respectively. The specimen type designators refer to the type numbers illus-trated in Figure 1 and the rivet type designators of U and CSK refer to the rivet types MS20426AD4-4 and NAS1097AD4-4 respectively. All specimens were made with 1 mm thick 2024-T3 Alcladsheet.

All of the tests completed are summarized in the test matrix given in Table 1. There are a lessernumber of tests conducted on Type 2 specimens as these tests were performed only to verify that theouter sheet failure which occurs in the Type 1 specimens is more critical. Additionally, as difficultieswere encountered with fractographic reconstructions for specimens containing universal type rivets(MS20426AD4-4), a larger number of tests were conducted on countersunk rivet (NAS1097AD4-4)specimens.

2.4 Crack Growth Reconstruction

Crack growth reconstructions were performed using a JEOL JSM-6400 scanning electron micro-scope. The reconstruction procedure was as follows. A photographic survey of the fracture surfacewas taken in the SEM facility at 450x magnification, with a probe voltage of 20 kV and a workingdistance of 21 mm. The resulting images from the survey were manually assembled to generate asingle high resolution image of the fracture surface. From this high resolution composite image, thelocations of the advancing crack front could be identified from the pattern of 10-, 4-, and 6-band

5

Table 1: Test matrix for 10-4-6 marker band fatigue tests

FSq(lb f ) Rivet Type Specimen Type No. F-Specimens No. SEM-Specimens

1500 NAS1097-AD4-4 1 6 -2000 NAS1097-AD4-4 1 6 32500 NAS1097-AD4-4 1 10 -3000 NAS1097-AD4-4 1 9 3

1500 MS20426-AD4-4 1 4 -2000 MS20426-AD4-4 1 4 12500 MS20426-AD4-4 1 4 -3000 MS20426-AD4-4 1 4 1

1500 NAS1097-AD4-4 2 3 -2000 NAS1097-AD4-4 2 3 -2500 NAS1097-AD4-4 2 3 -3000 NAS1097-AD4-4 2 3 -

1500 MS20426-AD4-4 2 3 -2000 MS20426-AD4-4 2 3 -2500 MS20426-AD4-4 2 2 -3000 MS20426-AD4-4 2 2 -

markers visible on the fracture surface (Figure 3). Due to the smaller striation spacing, blocks con-taining the lower load cycles (75% Smax) show up as dark lines or featureless bands at the givenmagnification. Once all the crack fronts were identified, the life of the specimen for each frontcould be determined by identifying the interface between fatigue and static failure (associated withthe final life of the specimen, see Figure 3) and counting backwards in blocks of cycles for eachcrack front. In some instances, it was difficult to detect particular crack fronts; however, it was easyto determine and account for such missing fronts due to the alternating 10-4-6 marker band patternexpected. Final crack growth measurements were made relative to the hole edge as observed on thefracture surface. For cases where cracks did not intersect the rivet hole, crack front positions weremeasured relative to the rivet centre line.

Several factors limited the extent to which the crack growth reconstructions could be performed.First, as crack length increases, the morphology of the fracture surface changes. At large cracklengths, the fracture surface becomes dominated by micro voids which impede the detection of themarker bands. Second, the reconstruction process requires many hours to complete. For both thesereasons, testing of fatigue specimens destined for crack growth reconstruction were stopped once acrack of approximately 2-3 mm was visible. As each specimen contained 6 rivet locations wherecracking could occur, the rivet containing the largest cracks was always selected for crack growthreconstruction.

3 Overall Fatigue Life Results

The overall fatigue life results for both specimen configurations (Type I and II) manufactured withthe universal and countersunk rivet types are shown in Figure 4. Type I and II specimen configu-rations correspond to outer sheet and inner sheet failure respectively. As different population sizes

6

were used to determine the mean fatigue life, statistical confidence intervals based on an ANalysisOf VAriance (ANOVA) [14] study are included.

Inner sheet failures are undesirable in airframe skin joints as access to the inner surface ofthe fuselage is necessary during inspections to detect such cracks. For this reason, only a limitednumber of tests were performed on the Type II specimen configuration with the primary objectiveof establishing if inner or outer sheet failure would likely occur first for a given rivet squeeze force.Outer sheet failure was favoured for rivet squeeze forces equal to or greater than 2000 lbf for bothrivet types. At high squeeze forces, the disparity in compressive residual stresses between the innerand outer sheet promote failure in the outer sheet. At lower squeeze forces, this disparity is reducedand the probability of failure in the inner sheet increases. This was observed in the universal rivetspecimens where inner sheet failure became critical for the lowest squeeze force examined. At thissqueeze force, outer sheet failure remained critical for the countersunk rivet specimens as a resultof the added stress concentration provided by the machined countersink. The remaining discussionon fatigue life will refer to outer sheet failure only.

For the universal rivet specimens, a small increase in mean fatigue life was observed with in-creasing squeeze force. Taking into account the scatter observed in the tests (indicated by the 90%confidence interval bars in Figure 4), however, the mean fatigue lives observed for squeeze forcesbelow 3000 lbf are not statistically distinct according to the ANOVA analysis. This relative in-sensitivity of overall fatigue life to rivet squeeze force results from the location of fatigue damageobserved in the universal rivet specimens as shown in Figure 5. Fatigue cracks nucleated above therivet hole and grew around it. This location corresponds to the region of maximum secondary bend-ing stress for universal rivet joints observed in a finite element study by the present authors [11, 13]and marked in the Figure 5. This location for crack nucleation and growth is also beyond the re-gion of beneficial residual stress observed in finite element simulations [12, 13] for the majority ofrivet squeeze forces examined (indicated by the plastic zone radii in Figure 5). This explains why astatistically distinct increase in fatigue life was only observed once the squeeze force became largeenough to induce a beneficial residual stress state at the location of cracking.

Conversely, the results for the countersunk rivet specimens show a strong dependency betweensqueeze force and overall fatigue life. In contrast to the universal rivet specimens, fatigue damageoccurs close to the rivet row centre line as shown in Figure 6. Cracks nucleate at the hole edge andcontinue to grow along the rivet row through the region of residual compressive stress. For higherrivet squeeze forces, the region of beneficial residual stress through which the crack has to growincreases, increasing the overall fatigue life.

4 Crack Growth Reconstruction Results

Fractographic crack growth reconstructions were completed in order to quantify the effect of rivetinstallation on the behaviour of fatigue cracks. Based on the overall fatigue life results presented inthe previous section and the required hours for each reconstruction, the scope of the crack growthreconstructions was reduced to two squeeze forces: 2000 lbf and 3000 lbf. Discussion of the resultswill be divided by rivet type.

7

4.1 Countersunk Rivet Joint Reconstructions

The complete set of crack front plots generated for countersunk riveted specimens is shown inFigures 7 and 8 for rivet squeeze forces of 2000 and 3000 lbf respectively. For the 3000 lbf results,only one third of the crack fronts (corresponding to the 10-band markers) are plotted for clarity.Thus, the number of cycles between plotted crack fronts must be noted when comparing the plotsfor the two squeeze forces.

Several key differences in crack growth behaviour are evident when comparing the crack frontplots for the two examined squeeze forces.

First, crack nucleation location appears to be influenced by the rivet squeeze force. For thelower squeeze force, crack nucleation occurs at or very near the rivet hole location. At the highersqueeze force, the nucleation location is shifted approximately 0.5 – 1 mm away from the rivet hole.The reason for this shift is a change in the dominate nucleation mechanism. At the lower squeezeforce, crack nucleation is dominated by the presence of a notch (rivet hole) and its associated stressconcentration. At the higher squeeze force crack nucleation shifts to a region of heavy frettingdamage. Within this region, cold-welding followed by disbonding of the asperities of the contactingjoint surfaces occurs under cyclic loading, resulting in the formation of surface micro-cracks. Givensufficient bulk stress, these micro-cracks can propagate into the substrate, forming a fatigue crack.

Several factors contribute to the shift in crack nucleation location and mechanism. First, atthe higher rivet squeeze force, larger residual compressive stresses are generated at the hole edge[12,13]. These compressive stresses reduce the peak tensile stress at the hole edge, thus reducing itsfatigue criticality. Second, the rivet clamping force increases with rivet squeeze force, exacerbatingthe fretting process. Although fretting and fretting fatigue are beyond the scope of the presentinvestigation, a complementary investigation is being completed by Brown [1] on fretting fatiguein riveted lap joints. Preliminary results from Brown have confirmed the crack nucleation locationobserved at the higher squeeze force correlates with the edge-of-contact between the contactingsheets, where fretting damage and micro-crack densities are expected to be highest.

The crack reconstructions also provide an indication of when crack nucleation occurs. Althoughthe exact time of nucleation cannot be established, crack front locations can be associated withspecific times in the fatigue life, allowing the period over which nucleation occurred to be narroweddown. Results are shown in Table 2. The difficulty encountered with this method is that it is basedon the earliest detectable crack front that can be accounted for. In some instances, additional crackfronts may have been present but too small or obscured to detect, or discontinuities in the fracturesurface thwarted efforts to associate a fatigue life with a particular crack front. As a result, althoughthe results in Table 2 seem to indicate that an increased squeeze force delays crack nucleation, thisconclusion can not be made. What can be said, however, is that comparing the detection lives inTable 2 to the overall average fatigue lives in Section 3, crack nucleation was found to occur duringthe first 35% of the average overall fatigue life.

An additional difference in crack growth behaviour evident when comparing the crack frontplots in Figures 7 and 8 is crack shape. At the higher rivet squeeze force, significant crack growthoccurs along the faying joint surface while the crack is still a part-through crack. This results inelongated crack fronts that deviate from the classical quarter elliptical front shapes at the lowersqueeze force. The more favourable residual compressive stress field at the higher squeeze forcecombined with secondary bending in the joint greatly reduces the rate of propagation of the crackthrough the thickness of the sheet.

Crack growth data from the reconstructions is summarized in terms of crack length and crack

8

Table 2: Nucleation period obtained from SEM reconstructions

Specimen Crack (kcycles) Specimen Crack (kcycles)

CSK2.0-T1-SEM-01 A1-left 26.2 CSK3.0-T1-SEM-01 B3-left 103.7A1-right 26.2 B3-right 32.0

CSK2.0-T1-SEM-02 B3-left 41.6 CSK3.0-T1-SEM-02 B2-left 81.2B3-right 41.6 B2-right 81.2

CSK2.0-T1-SEM-03 B1-left 32.0 CSK3.0-T1-SEM-03 A1-left 54.2B1-right 40.6 A1-right 67.7

average 34.7 70.0

growth rate in Figures 9 and 10. The highly elongated crack front shapes observed at the higherrivet squeeze force made determining the crack length at the free joint surface difficult. As a result,only the faying surface crack length behaviour is shown. These results clearly show the effects ofriveting induced residual stresses on fatigue crack growth. Excellent repeatability was observedin the crack growth behaviour amongst specimens manufactured with the same rivet squeeze force.Furthermore, a distinct difference in crack growth rate was observed for the two rivet squeeze forces.A 3-fold reduction in crack growth rate was observed for the higher squeeze force as a result of themore favourable residual stress field induced during rivet installation [11–13].

4.2 Universal Rivet Joint Reconstructions

The universal rivet joint reconstructions were less successful. The reconstruction process relied onbeing able to identify crack front locations and associate a fatigue life and crack length with theselocations. Identifying the crack front locations was not problematic. Associating a fatigue life witheach crack front location, however, was hampered by the presence of multiple fatigue cracks andfracture plane intersections. Defining a robust measure for crack length in the presence of multiplefatigue cracks was also difficult. For these reasons, only a limited number of reconstructions wereperformed for universal rivet joints.

The above described difficulties are illustrated in the reconstruction results shown in Figures 11and 12. Figure 11 best illustrates the difficulty of defining a robust measure for crack length. Themain fracture surface was formed by two distinct fatigue cracks (centred at approximately y = -1 and0.7 mm) that eventually linked up to form a larger crack. During this link-up process, the shape andpropagation rate of the crack front is dependant on the proximity and number of cracks involved.Simply tracking the location of the faying surface crack front location does not account for thesefactors, making it difficult to compare such results between different reconstructions. Reconstruc-tions were further hampered by the presence of crack plane intersections. Such discontinuities inthe fracture surface prevent the necessary accounting in determining the fatigue life associated witheach identified crack front. As an example, in Figure 12, although crack fronts could be identifiedbetween y = -3 - 2.5 mm, associated fatigue lives could not be determined for the fronts due to thediscontinuities in the fracture surface at y = -3 and 2.5 mm.

Despite the limited quantitative data that could be obtained from the universal rivet joint recon-struction, some qualitative observations could be made. The presence of multiple fatigue cracksand fracture plane intersections indicate that fretting fatigue and the formation of multiple micro-cracks is a dominant crack nucleation mechanism for the universal joints. Although fretting was

9

also observed as a nucleation mechanism for the countersunk rivet joints, the presence of the rivethole along the fracture plane for these joints prevents a large number of micro-cracks from growing.Typically, a micro-crack would link-up with the rivet hole and become the dominant crack. Forthe universal rivet joints, the rivet hole does not lie along the fracture plane. As a result, multiplemicro-cracks tend to grow and subsequently coalesce into larger fatigue crack.

5 Conclusions

Two series of fatigue tests have been carried out in order to quantify the effects of riveting inducedresidual stresses on the fatigue performance of riveted lap joints. This work was carried out as afollow-up to finite element studies carried out by the present authors on quantifying the effects ofrivet installation on the stress field in loaded and unloaded lap joints [11–13]. The first test seriesfocused on quantifying the effect of rivet installation on the overall fatigue performance. Based onthe results from this series, the following conclusions can be made:

• In general, fatigue life increased with increasing rivet squeeze forces for both countersunkand universal rivet joints for the sheet thickness and joint geometry examined.

• The influence of rivet squeeze force on fatigue life is dependant on the location of cracking.For the universal rivet case, little improvement in fatigue performance was observed for thelower range of squeeze forces due to crack nucleation occurring away from the rivet holeoutside the region of beneficial residual stresses.

• Conversely, cracking in the case of countersunk joints occurred at or near the joint net sectionwhere it is highly influenced by residual stresses. Thus, the countersunk joints displayed agreater sensitivity between rivet squeeze force and fatigue performance.

The second series of fatigue tests were used to study the effects of riveting induced residualstresses on crack growth. Based on the results of this test series, the following conclusions can bemade:

• The SEM reconstruction techniques employed in this study provided little quantifiable data interms of the fatigue performance of universal riveted joints. The presence of multiple fatiguecracks and multiple fracture planes impeded the reconstruction process.

• For countersunk rivet joint specimens, the reconstructions demonstrated that use of forcecontrolled riveting techniques resulted in repeatable fatigue crack growth behaviour for thesame rivet squeeze force.

• Reductions in crack growth rate as great as 3-fold were observed when the rivet squeeze forcewas increased from 2000 to 3000 lbf. For the rivet/joint geometry studied, these squeezeforces produce upset rivet head geometries which fall within typical riveting practice stan-dards [8].

• Crack nucleation was observed to occur within the first 35% of the average overall fatiguelife. Although exact nucleation times could not be established, a time period within which itoccurred was established from the earliest detectable fracture surface marker.

10

type 1 type 2

1.0 TK 2024-T3 Alclad SHEET

*ALL UNITS IN MM

76.2

241.3

25.4

25.4

12.7 38.1 12.7

12.7

88.9

A1

A2

A3

B1

B2

B3

Figure 1: Fatigue specimen geometry (adapted from coupon design presented in [8])

400 cycles 40 cycles

2640 cycles 1760 cycles 4400 cycles

4000 cycles

4000 cycles

4000 cycles

75% Smax Smax

S

1 Program = 20800 cycles

Figure 2: Schematic of modified 10-6-4 marker-band spectrum

11

Crack Crack propagation propagation directiondirection

10

4

6

(a) visible marker bands

Static failureFatigue failure

10-band

(b) fatigue-static failure interface

Figure 3: Examples of SEM reconstruction features

FSq (lbf)

corr

ecte

dfa

tigu

elif

e(k

cycl

es)

1000 1500 2000 2500 3000 35000

50

100

150

200

250

300

350

2-row 2024-T3 splice with MS20470AD4-4s = 25.4mm, p = 38.1mm, t = 1.0mmDo = 3.2mm, Marker-band spectrum

outer sheet failure

inner sheet failure

error bars show 90%confidence levels

> 600 kcycles

(a) universal joint

FSq (lbf)

corr

ecte

dfa

tigu

elif

e(k

cycl

es)

1000 1500 2000 2500 3000 35000

50

100

150

200

250

300

350

2-row 2024-T3 splice with NAS1097AD4-4s = 25.4mm, p = 38.1mm, t = 1.0mmDo = 3.2mm, Marker-band spectrum

outer sheet failure

inner sheet failure

error bars show 90%confidence levels

(b) countersunk joint

Figure 4: Overall fatigue lives for inner and outer sheet failure

12

Figure 5: Typical location of fatigue damage observed in universal rivet joint fatigue specimens.FE results [12, 13] showing outer sheet faying surface plastic zone (circular regions) and region ofmaximum bending included for comparison.

Figure 6: Typical location of fatigue damage observed in countersunk rivet joint fatigue specimens.FE results [12, 13] showing outer sheet faying surface plastic zone (circular regions) and region ofmaximum bending included for comparison.

13

dist

ance

from

hole

edge

(mm

)

z(mm)

00.

51

1.5

22.

53

3.5

44.

55

5.5

0

0.51

4-b

and

mar

ker

6-b

and

mar

ker

10

-ban

dm

arke

r

CS

K2.

0-T

1-S

EM

-01

Cra

ck:A

2le

ftel

liptic

alcu

rve

fit

63.1

59.0

54.2

49.5

45.5

40.6

36.0

32.0

27.0

kcyc

le(c

orr

ecte

d)

dist

ance

from

hole

edge

(mm

)

z(mm)

00.

51

1.5

22.

53

3.5

44.

55

5.5

0

0.51

4-b

and

mar

ker

6-b

and

mar

ker

10

-ban

dm

arke

r

CS

K2.

0-T

1-S

EM

-01

Cra

ck:A

2ri

ght

ellip

tical

curv

efit

kcyc

les

(co

rrec

ted

)

63.1

59.0

54.2

49.5

45.5

40.6

36.0

32.0

27.1

dist

ance

from

hole

edge

(mm

)

z(mm)

-5.5

-5-4

.5-4

-3.5

-3-2

.5-2

-1.5

-1-0

.50

0

0.51

10

-ban

dm

arke

r4-

ban

dm

arke

r6-

ban

dm

arke

r

CS

K2.

0-T

1-S

EM

-02

Cra

ck:B

3le

ftel

liptic

alcu

rve

fit

kcyc

les

(cor

rect

ed)

63.1

67.7

59.0

54.2

49.5

45.5

40.6

36.0

32.0

27.1

dist

ance

from

hole

edge

(mm

)

z(mm)

00.

51

1.5

22.

53

3.5

44.

55

5.5

0

0.51

10

-ban

dm

arke

r4-

ban

dm

arke

r6-

ban

dm

arke

r

CS

K2.

0-T

1-S

EM

-02

Cra

ck:B

3ri

ght

ellip

tical

curv

efit

67.7

49.5

59.0

54.2

45.5

40.6

36.0

27.1

32.0

63.1

kcyc

les

(co

rrec

ted

)

dist

ance

from

hole

edge

(mm

)

z(mm)

-5.5

-5-4

.5-4

-3.5

-3-2

.5-2

-1.5

-1-0

.50

0

0.51

10-b

and

mar

ker

4-ba

ndm

arke

r6-

band

mar

ker

CS

K2.

0-T

1-S

EM

-03

Cra

ck:B

1le

ftpo

lyno

mia

lcur

vefit

kcyc

les

(co

rrec

ted

)

81.2

76.6

72.6

67.7

63.1

59.0

54.2

49.5

32.0

dist

ance

from

hole

edge

(mm

)

z(mm)

00.

51

1.5

22.

53

3.5

44.

55

5.5

0

0.51

10

-ban

dm

arke

r4

-ban

dm

arke

r6

-ban

dm

arke

r

CS

K2.

0-T

1-S

EM

-03

Cra

ck:B

1rig

htpo

lyno

mia

lcur

vefit

kcyc

les

(co

rrec

ted)

76.6

81.2

72.6

67.7

63.1

59.0

54.2

49.5 36

.0

Figu

re7:

Cra

ckfr

ontr

econ

stru

ctio

nfo

raco

unte

rsun

kriv

etjo

intw

ithF S

q=

2000

lbf

14

dist

ance

from

hole

edge

(mm

)

z(mm)

-5.5

-5-4

.5-4

-3.5

-3-2

.5-2

-1.5

-1-0

.50

0

0.51

10-b

and

mar

ker

CS

K3.

0-T

1-S

EM

-01

Cra

ck:B

3le

ftpo

lyno

mia

lcur

vefit

kcyc

le(u

ncor

rect

ed)

162.

514

8.9

135.

4

121.

8

108.

3

dist

ance

from

hole

edge

(mm

)

z(mm)

00.

51

1.5

22.

53

3.5

44.

55

5.5

0

0.51

10-b

and

mar

ker

CS

K3.

0-T

1-S

EM

-01

Cra

ck:B

3rig

ht

poly

nom

ialc

urve

fit

162.

5

kcyc

les

(co

rrec

ted)

135.

412

1.8

108.

3

94.8

54.2

148.

9

67.7

40.6

81.2

dist

ance

from

hole

edge

(mm

)

z(mm)

-5.5

-5-4

.5-4

-3.5

-3-2

.5-2

-1.5

-1-0

.50

0

0.51

10-b

and

mar

ker

CS

K3.

0-T

1-S

EM

-02

Cra

ck:B

2le

ftpo

lyno

mia

lcur

vefit

kcyc

les

(cor

rect

ed)

162.

514

8.9

135.

412

1.8

108.

3

94.8

81.2

dist

ance

from

hole

edge

(mm

)

z(mm)

00.

51

1.5

22.

53

3.5

44.

55

5.5

0

0.51

10-b

and

mar

ker

CS

K3.

0-T

1-S

EM

-02

Cra

ck:B

2ri

ght

poly

nom

ialc

urve

fit

kcyc

les

(cor

rect

ed)

162.

5

148.

913

5.4

121.

8

108.

394

.8

81.2

dist

ance

from

hole

edge

(mm

)

z(mm)

-5.5

-5-4

.5-4

-3.5

-3-2

.5-2

-1.5

-1-0

.50

0

0.51

10-b

and

mar

ker

CS

K3.

0-T

1-S

EM

-03

Cra

ck:A

1le

ftpo

lyno

mia

lcur

vefit

kcyc

les

(co

rrec

ted)

162.

514

8.9

135.

4

121.

8

108.

3

94.8

81.2

67.7

54.2

dist

ance

from

hole

edge

(mm

)

z(mm)

00.

51

1.5

22.

53

3.5

44.

55

5.5

0

0.51

10-b

and

mar

ker

CS

K3.

0-T

1-S

EM

-03

Cra

ck:A

1rig

htpo

lyno

mia

lcur

vefit

kcyc

les

(co

rrec

ted)

162.

5

148.

9

135.

412

1.8

108.

394

.8

81.2

67.7

Figu

re8:

Cra

ckfr

ontr

econ

stru

ctio

nfo

raco

unte

rsun

kriv

etjo

intw

ithF S

q=

3000

lbf(

6-an

d4-

band

mar

kers

rem

oved

forc

lari

ty)

15

kcycles (corrected)

c 1(m

m)

0 50 100 150 2000

1

2

3

4

5

6

7

All SEM Specimens with NAS1097AD4-4 rivets

FSq = 2000lbf(8.9kN)

FSq = 3000lbf (13.3kN)

Figure 9: Crack growth curves from SEM reconstructions of countersunk rivet joints

c1 (mm)

da/d

N(µ

m/c

ycle

)

0 1 2 3 4 5 6

10-2

10-1

100

CSK2.0-T1-SEM-01 A2-leftCSK2.0-T1-SEM-01 A2-rightCSK2.0-T1-SEM-02 B3-leftCSK2.0-T1-SEM-02 B3-rightCSK2.0-T1-SEM-03 B1-leftCSK2.0-T1-SEM-03 B1-rightCSK3.0-T1-SEM-01 B3-leftCSK3.0-T1-SEM-01 B3-rightCSK3.0-T1-SEM-02 B2-leftCSK3.0-T1-SEM-02 B2-rightCSK3.0-T1-SEM-03 A1-leftCSK3.0-T1-SEM-03 A1-right

FSq = 2000lbf (8.9kN)

FSq = 3000lbf (13.3kN)

Figure 10: Crack growth rates from SEM reconstructions of countersunk rivet joints

16

distance from rivet centre, y (mm)

z(mm)

-4 -3 -2 -1 0 1 2 3 40

1

10-band marker4-band marker6-band marker

y

Figure 11: Crack front reconstruction for U2.0-T1-SEM-01 (arrows indicate intersecting fractureplanes, shaded area indicates rivet hole location)

y (mm)

z(mm)

-6 -5.5 -5 -4.5 -4 -3.5 -3 -2.50

0.5

1

10-band marker4-band marker6-band marker

y (mm)

z(mm)

2 2.5 3 3.5 4 4.50

0.5

1

10-band marker4-band marker6-band marker

Figure 12: Crack front reconstruction for U3.0-T1-SEM-01 (arrows indicate intersecting fractureplanes)

17

References

[1] A. M. Brown and P. V. Straznicky. Modelling fretting contact stresses in a single lap splice.In Proceedings of The Fith International Symposium on Fretting Fatigue, ISFF5, Montreal,Canada, April, 2007.

[2] P. F. P. de Matos, P. M. G. P. Moreira, I. Nedbal, and P. M. S. T. de Castro. Reconstructionof fatigue crack growth in Al-alloy 2024-T3 open-hole specimens using microfractographictechniques. Engineering Fracture Mechanics, pages 2232–2246, 2005.

[3] R. de Rijck. Stress Analysis of Fatigue Cracks in Mechanically Fastened Joints. PhD disser-tation, Delft University of Technology, 2005.

[4] W. Elber. The significance of fatigue closure. In Damage Tolerance in Aircraft Structures,number ASTM STP 486, pages 230–242, 1971.

[5] S. A. Fawaz. Fatigue Crack Growth in Riveted Joints. PhD Dissertation, Delft University ofTechnology, 1997.

[6] N.M. Grinberg. Spacing of fatigue striations and crack growth rate. Soviet Materials Science,21(2):153–160, 1985.

[7] P. M. G. P. Moreira, P. F. P. de Matos, and P. M. S. T. de Castro. Fatigue striation spacing andequivalent initial flaw size in al 2024-t3 riveted specimens. Theoretical and Applied FractureMechanics, 43:89–99, 2005.

[8] R. P. G. Muller. An Experimental and Analytical Investigation on the Fatigue Behaviour ofFuselage Riveted Lap Joints: the Significance of the Rivet Squeeze Force and a Comparisonof 2024-T3 and Glare 3. PhD dissertation, Delft University of Technology, 1995.

[9] I. Nedbal, J. Siegl, and J. Kunz. Relation between striation spacing and fatigue crack growthrate in al alloy sheets. Advances in Fracture Research. Proceedings of the 7th InternationalConference on Fracture (ICF7), pages 3483–3491, 1989.

[10] R. S. Piascik and S. A. Willard. The characterization of wide spread fatigue damage in thefuselage riveted lap splice. Technical Report NASA-TP-97-206257, NASA, 1997.

[11] C. D. Rans. The Role of Rivet Installation on the Fatigue Performance of Riveted Lap Joints.PhD dissertation, Carleton University, 2007.

[12] C. D. Rans, R. C. Alderliesten, and P. V. Straznicky. Riveting process induced residual stressesaround solid rivets in mechanical joints. Journal of Aircraft, 44(1):323–329, 2007.

[13] C. D. Rans, R. C. Alderliesten, and P. V. Straznicky. Effects of rivet installation on resid-ual stresses and secondary bending in a riveted lap joint. In Proceedings of The 48th AIAAStructures, Structural Dynamics, and Materials Conference, Waikiki, USA, Apr. 2007.

[14] D. B. Schiff and R. B. d’Agostino. Practical Engineering Statistics. Wiley-Interscience, 1996.