You are on page 1of 7

International

Journalof
International Journal of Fatigue 29 (2007) 1744–1750
Fatigue
www.elsevier.com/locate/ijfatigue

Internal stress effects on fatigue crack initiation at notches


D. Kujawski *, S. Stoychev
Department of Mechanical and Aeronautical Engineering, College of Engineering and Applied Sciences, Western Michigan University,
Kalamazoo, MI 49008-5343, United States

Received 11 October 2006; received in revised form 30 November 2006; accepted 4 December 2006
Available online 27 December 2006

Abstract

In this paper the effects of the internal stresses on crack initiation at notches is investigated. Internal stresses are induced by applying
two pre-loading cycles at a given load ratio, R = SMIN/SMAX, to the ‘virgin’ notched specimen. After pre-loading, blocks of 106 cycles
with a constant nominal stress range, DS, slightly larger than the nominal threshold stress range, DSth, are applied with increasing min-
imum stress, SMIN, in each subsequent block until a crack initiates. The lowest SMIN above which a crack initiates is recorded as SIN. The
process is repeated for a wide range of positive and negative load ratios in order to obtain SIN/SMAX vs. load ratio curve. This curve for
positive load ratios surprisingly resembles the KPR/KMAX plot reported in the literature for a cracked specimen. The results are explained
in terms of compressive internal stress and the associated local clamping action at the notch cyclic plastic zone. Both the analytical and
finite element calculations show a fairly good agreement with the experimental results.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Crack initiation; Load interaction; Internal stress; Clamping force

1. Introduction fracture surfaces oxidation [2]. Therefore, existing crack


closure models contain a number of ‘‘adjustable’’ parame-
There is a general agreement in the fatigue community ters which have to be altered and tuned in order to fit load
that the load interaction effects are associated with plastic- history effects such as R-ratios, overloads or spectrum
ity-induced phenomena. However, there is no consensus loadings.
whether these phenomena operate in front and/or behind The plasticity-induced phenomena, however, are present
crack tip. The majority of the researchers believe that the not only in the crack wake, but in front of the propagating
plasticity-induced crack closure [1], operating in the crack crack as well. In particular, when an elastic–plastic material
wake, is solely responsible for the observed load interaction is loaded, a forward plastic zone is developed due to mate-
effects. Then, it is a custom practice to assume that any rial yielding, inducing crack-tip compressive stresses after
load history could be fully accounted for by means of unloading. There is a growing group of researchers who
an effective stress intensity factor (SIF) range, DKeff = believe that those compressive stresses are responsible for
Kmax Kop, where Kmax and Kop are the stress intensity load interaction effects whereas the effects associated with
factors calculated for the maximum, Pmax, and the crack crack wake closure are small or negligible. Schijve [3] and
opening load, Pop, respectively. It is also recognized that Jacoby [4] were the first researchers who postulated that
a number of other mechanisms besides plasticity may con- the load interaction effects are controlled by plasticity
tribute to DKeff, including surface roughness, crack branch- induced compressive stresses in the active ‘process zone’
ing or deflection, fracture surface mismatch, debris, and ahead of the crack tip. They suggested that retardation
(or acceleration) in fatigue crack growth (FCG) rate after
an overload (or underload) is governed by the compressive
*
Corresponding author. Fax: +1 616 387 3358. stress gradient ahead of the crack tip. Ling and Schijve
E-mail address: daniel.kujawski@wmich.edu (D. Kujawski). [5] demonstrated that retardation of the FCG following

0142-1123/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijfatigue.2006.12.004
D. Kujawski, S. Stoychev / International Journal of Fatigue 29 (2007) 1744–1750 1745

overload could be entirely eliminated by intermediate heat Fig. 1 shows a block diagram of the investigation that is
treatment, which removes the residual stress gradient, set presented and discussed in this paper.
by the prior overload. Also, Kujawski and Ellyin [6] uti-
lized the local stress gradient in front of the crack in analyt- 2. Experiments
ical modeling of the load ratio effects on the FCG behavior.
More recently, in a number of articles Sadananda and 2.1. Specimen
Vasudevan [7–12] laid up a framework for unified model,
(for short SV model) for FCG analysis by accounting for Fig. 2 depicts a single-edge notch geometry that is cho-
plasticity-induced internal stresses and their role in damage sen for the specimen. The radius of the notch root is
process, which takes place in front of the crack. They pos- 0.28 mm and the depth about 5.3 mm. The specimen is flat
tulated that FCG behavior is two-parameter phenomenon and the small thickness of t = 2.54 mm dictates that plane
and should be analyzed using the total values of DK and stress condition prevails. During testing, the notch is mon-
Kmax, which account for both an applied load and internal itored for a crack initiation using a digital camera with res-
stress contributions. The present authors demonstrated olution better than 10 lm.
that a two-parameter fatigue crack driving force,
Dj = (Kmax)p (DK)1 p in terms of the total values, is able 2.2. Testing procedure
to account for load interaction effects (R-ratios and block
loadings) on FCG behavior without the need to invoke First, the threshold stress range, DSth, is determined
the crack closure concept [13–16]. using the procedure illustrated in Fig. 3a and b. In this pro-
In general, there is a consensus among researchers that
load interaction effect on FCG in metallic materials is asso-
ciated with plasticity induced phenomena. However, there Load
is no broad agreement whether these phenomena operate
in the wake or front of the crack or both. It is often argued
that this inherent controversy exists since it is difficult to r = 0.28 mm
Al 2324–T39
separate plasticity-induced phenomena operating in the
90°
crack wake from those in the crack front.
The objective of this paper is to overcome the above dif-
ficulty. This is accomplished by analyzing the plasticity-
induced internal stresses at a notched specimen in terms Camera
of crack initiation rather than crack propagation. Notched 5.3 mm
specimens eliminate the possibility of closure beyond 2.5mm
doubt, leaving the internal stress effects of the prior load
history as a primarily factor affecting the crack initiation 45 mm
phenomenon. Internal stresses at the notch are induced
by applying two pre-loading cycles with a constant maxi-
mum stress, SMAX, at a given load ratio, R = PMIN/
PMAX = SMIN/SMAX. Different R values result in different
internal stresses at SMIN. It should be noted that upper case
subscripts are used with reference to the pre-loading cycles. Fig. 2. Specimen geometry.

Experiments FEA simulation Analytical modeling

Initiation stress ANSYS COMSOL

Internal stress Clamping force method

Comparison and conclusions Initiation stress

Fig. 1. Block diagram of the notch study.


1746 D. Kujawski, S. Stoychev / International Journal of Fatigue 29 (2007) 1744–1750

SMAX = constant c shows the procedure for the crack initiation tests adopted
a
from Doker and Bachmann [17].
ΔSth The purpose of this procedure is to determine the stress
S level SIN above which fatigue damage starts to accumulate
resulting in a crack initiation at the notch root. First, the
internal stresses are introduced in the ‘virgin’ specimen
by applying only two pre-load cycles with a given load
b ratio (R = 0 shown in Fig. 3c). It is assumed that those
two cycles alone cannot cause crack initiation. SMAX is
the same as that used for DSth determination. Then, a stress
range somewhat larger than DSth is applied in blocks of 1
Crack length

million cycles each. The Smin in the first block is approxi-


crack = notch mately equal to the SMIN during pre-loading. The notch
root is examined after each block and SMIN is increased
by d @ 0.05SMAX (keeping DSth = constant) if there is no
initiation

1e6
crack initiation. The test is stopped when a crack is
observed. The SIN is defined as the average value of Smin
for the last two blocks.

2.3. Experimental results


c
The raw test results from all experiments are presented
S ΔSth in Tables 1 and 2.
δ It can be seen that SMAX in Table 1 and DS in Table 2
SIN are not exactly constant for each specimen. These small
variations in stresses compensate for the slightly different
Cycles
notch sizes in different specimens. The adjustment was
Fig. 3. Procedures for determination of the threshold stress range, DSth (a done by keeping constant the corresponding stress intensity
and b), and the crack initiation stress, SIN (b and c).

Table 1
Preloading histories of the test specimens
cedure the applied range of nominal stress, DS, is increased
Specimen number
in small steps, keeping the maximum stress, SMAX, = con-
1 2 3 4 5
stant (each step is applied for 106 cycles at frequency of
a b
20 Hz). The next step is applied only if and after the crack R or UR 0 0.25 0.5 0.2 0.39
SMAX [MPa] 171 170 165 170 168
does not initiate within the previous step. The value of DSth
Notch [mm] 5.27 5.37 5.80 5.33 5.49
is defined as the stress range when the crack is initiated. a
R = SMIN/SMAX for SMIN > 0.
After DSth is determined, series of crack initiation tests b
UR = SMIN/r0 for SMIN < 0, where r0 is the yield strength (430 MPa).
with different internal stresses are performed. Fig. 3b and

Table 2
Load histories of the test specimens
Specimen number
1 2 3 4 5
a b
R or UR 0 0.25 0.5 0.2 0.39
Stress [MPa] DS Smin DS Smin DS Smin DS Smin DS Smin
Block number 1 25.2 1.4 25.1 44.4 24.5 88.1 24.1 86.3 24.9 167.7
2 18.1 53.2 97.1 62.9 157.8
3 35.3 72.9 105.9 38.9 139.6
4 52.5 82.5 114.7 15.3 121.3
5 69.7 91.2 8.7 103.2
6 85.1
7 66.9
8 48.7
a
R = Smin/Smax for Smin > 0.
b
UR = Smin/r0 for Smin < 0, where r0 is the yield strength (430 MPa).
D. Kujawski, S. Stoychev / International Journal of Fatigue 29 (2007) 1744–1750 1747

factors (calculated by assuming that crack length = notch


1 MAX
size).
The actual experimental crack initiation results for the 0.8
test with preload ratio R = 0.25 are shown in Fig. 4. It
KPR – crack (Lang)
can be seen from Image 1, which was taken after the 4th 0.6
loading block (or after 4 million cycles) that there is no

SIN / SMAX or KPR / KMAX


0.4
crack at the notch root. Image 2, however, which is taken
after the 5th block (or after additional 1 million cycles), 0.2
shows clearly that crack was formed (Fig. 5 depicts the
amount of crack propagation in the last loading block). 0
This indicated that SMIN of the last block has to be bigger
-0.2
than SIN since the crack was initiated and propagated.
Therefore, the corresponding stress level for crack initia- -0.4
tion, SIN, should be situated somewhere between Smin4
and Smin5. This testing procedure for SIN determination -0.6 SIN - notch
is repeated for other load ratios and the results are shown

MIN
in Fig. 6 in term of SIN/SMAX vs. load ratio. The upper and -0.8
lower bounds for the error bars are equal to the minimum
-1
stresses, Smin, corresponding to the last two loading blocks -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
(i.e. crack and on crack blocks) at the particular UR = SMIN / σ 0 R = SMIN / SMAX
experiment.
Fig. 6. Experimental results: notched specimen vs. cracked specimen.

For comparison, Fig. 6 shows also the results from Lang


1 2 [18] for cracked specimen in terms of the normalized stress
intensity factor for crack propagation KPR/KMAX. Note
that both methods for determination of SIN and KPR are
based on Doker and Bachmann’s procedure [17]. For neg-
ative load ratios, SIN results are plotted in terms of
UR = SMIN/r0 instead of R in order to make comparison
with Lang’s data. The upper and lower KPR curves corre-
1.1 spond to constant amplitude and overload tests, respec-
180 tively. It can be seen that the notched and cracked
1 Smax
0.9
160 specimen results are different only for negative load ratios.
Smin This difference comes from the fact that under compression
0.8 140
120
the crack is closed and the applied stress is redistributed
0.7
over larger area including crack wake. The notch, on the
S [MPa]
S / SMAX

0.6 100
other hand, cannot close and SIN decreases rapidly with
0.5 80 decreasing UR.
0.4
1
2 60 So, if crack closure is causing the difference between
0.3
40 notched and cracked specimen for negative load ratios,
0.2
why do the results correlate so well for positive load ratios?
0.1 no crack R = 0.25 20
Does that mean that there is no significant crack closure
0 0 effect for R > 0? Also, how is it possible to explain the
0 1 2 3 4 5 6
notched specimen results using crack closure if there is
N X 10 6 [cycles]
no crack to start with? In order to answer these questions
Fig. 4. Experimental results for R = 0.25. a numerical and analytical investigation of the effect of

Fig. 5. Cracked specimen after the last load step (R = 0.25).


1748 D. Kujawski, S. Stoychev / International Journal of Fatigue 29 (2007) 1744–1750

the compressive internal stresses ahead of the notch on SIN 1.5


will be presented next.
1 10
3. Finite element simulations
0.5
Stresses ahead of the notch in the crack initiation exper-

σ / σ0
0
iments described in the previous section are simulated by
1 10
finite element (FE) analysis. The FE model is shown in -0.5 1

S / SMAX
Fig. 7. Due to symmetry only half of the specimen is mod-
eled. In the experiment, the hydraulic grips of the testing -1 2
machine prevent the specimen’s ends in the gripping area 0 1
Time
from rotating. In the FE-model this was achieved by apply- -1.5
ing uniform displacement at the edges as shown in Fig. 7. 0 0.5 1 1.5 2 2.5 3
The displacement is then increased until the sum of the Distance from the notch tip [mm]
reaction forces at the loading locations becomes equal to Fig. 8. Normalized internal stress profiles, calculated using FEA.
the load that is applied in the real test.
The model shown in Fig. 7 was created with a custom
designed graphical user interface which was written in Mat- The same simulation is carried with the commercially
Lab 7. The model is than meshed and solved using this available FE solver – ANSYS. The only difference is the lay-
interface with COMSOL 3.2b commercial software. This ered mesh which has a built in algorithm in ANSYS. This
was done in order to streamline the calculation and analy- mesh consists of 3275 elements. Despite the obvious differ-
sis of the results. ences in the meshing and the numerical solvers, the results
The smallest mesh element has 1 lm and is located at the from COMSOL and ANSYS are practically identical.
notch root. The rest of the elements are created in circles Therefore the rest of the simulations and the analysis of
around the notch root. The elements in every next circle the results are done using only one solver (COMSOL).
grow by 10% until they reach 3.5 mm in size. After that Fig. 8 depicts the normalized stresses in the notch plane sim-
the element size is kept constant. The procedure outlined ulated for the second preloading cycle with R = 0 using ten
above results in 6095 elements. The material model, used loading substeps. The cyclic and monotonic plastic zones
for the simulation is also shown in Fig. 7. It follows closely can be easily identified from curves 1 and 10, respectively.
the actual material response without being too complicated
in order to speed up the calculations. The strain-hardening 4. Clamping force method
slope is determined from a least squares fit of the experimen-
tal cyclic stress–strain curve of 2324-T39 aluminum alloy. Each stress profile from the FE simulation is analyzed
Preliminary, FE results indicated that there was no signifi- according to the ‘clamping force’ method. The ‘clamping
cant difference in the stress–strain field induced by pre-load- force’ is defined as the local compressive force, generated
ing cycles around the notch depending on the bi-linear or by the stresses in the ‘process zone’. It can be easily calcu-
multi-linear material models used. Therefore, for the pur- lated by integrating the stress profile and multiplying by the
pose of this paper, a very simple model is chosen (Fig. 7). thickness. For positive load ratios, the process zone corre-
sponds to the cyclic plastic zone (Fig. 9). For negative load

Applied displ. σ 530MPa


430MPa 1.5
σmax

72GPa 2σ 0 1

ε σin
0.5
σmin
σ / σ0

0
S MAX

-0.5
SIN

MIN
-1
Process zone 0 Time
-1.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4
COMSOL
ANSYS Distance from the notch [mm]

Fig. 7. Model for the FEA simulation. Fig. 9. SIN determination from the FE results for R = 0.25.
D. Kujawski, S. Stoychev / International Journal of Fatigue 29 (2007) 1744–1750 1749

1.5
1 MAX
σmax S MAX

1 Time
0
SIN 0.8
MIN

0.5 0.6
at S = 0
0.4
σ / σ0

0
0.2

SIN / SMAX
Notch
-0.5 tests
σin 0
σmin
-1 -0.2

Process zone
-1.5
-0.4
0 0.2 0.4 0.6 0.8 1 1.2 1.4
-0.6
Distance from the notch [mm] Model
-0.8

N
Fig. 10. SIN determination from the FE results for UR = 0.39.

MI
-1
-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
600 UR = SMIN / σ 0 R = SMIN / SMAX
500
Fig. 12. SIN – experiments and prediction.
400
Clamping force [N]

300 5. Conclusions
200
SIN
100 The effects of internal stresses induced by preloading
0 cycles on fatigue crack initiation stress at notches, SIN, were
-100 studied experimentally and numerically. The experimental
-200
data are analyzed and modeled in terms of the associated
clamping force action in the notch ‘process zone’. Based
-300
on the results the following conclusions can be drawn:
-400
0 0.2 0.4 0.6 0.8 1
S / Smax
 The ‘clamping force’ method used for determination of
SIN from the internal stresses provides fairly good repre-
Fig. 11. Clamping force as a function of the applied stress for R = 0. sentation of the experimental results.
 For positive load ratios there is no significant difference
ratios, the process zone is equal to the plastic zone size at between SIN/SMAX for notched and KPR/KMAX for
zero load (Fig. 10). cracked specimens.
It is assumed that damage can be accumulated in the  Unlike the crack, the notch does not close at negative
material only if the ‘clamping force’ is positive. Therefore, load ratios and therefore SIN/SMAX is significantly smal-
SIN is defined as the nominal stress level for which the ler than KPR/KMAX.
‘clamping force’ is equal to zero.
Fig. 11 is constructed in order to determine SIN more Finally, the predictions in Fig. 12 can be improved fur-
accurately. The ‘clamping force’ is calculated using the pro- ther by using more elaborated material model for the FE
cedure outlined above and is plotted as a function of the simulations. This however could increase the calculation
normalized applied stress. This function can be easily inter- time significantly and is not necessary in order to prove
polated to find the intersection with the abscissa which rep- the applicability of the clamping force method.
resents the crack initiation stress, SIN, corresponding to
zero ‘clamping’. Acknowledgement
The ‘clamping force’ method for determination of SIN
was applied to all load ratios from the experimental study This study is supported by the Office of Naval Research
and the results are shown in Fig. 12. The correlation is under Grant N00014-04-1-0718.
fairly good except in the case of UR = 0.39. The error
References
is probably caused by the fact that it is more difficult to
visually detect the crack when the material is under bulk [1] Elber W. The significance of fatigue crack closure. In: Damage
compression. Therefore, the actual stress for crack initia- tolerance in aircraft structures. ASTM STP 486. Philadelphia, PA:
tion could be lower than that recorded. American Society for Testing and Materials; 1970. p. 230–42.
1750 D. Kujawski, S. Stoychev / International Journal of Fatigue 29 (2007) 1744–1750

[2] Suresh S. Fatigue of materials. Cambridge: Cambridge University [11] Sadananda K, Vasudevan AK. Short crack growth and internal
Press; 1991. stresses. Int J Fatigue 1997;19:S99–S109.
[3] Schijve J. Fatigue crack propagation in light alloy sheet material and [12] Sadananda K, Vasudevan AK, Holtz RL, Lee EU. Analysis of
structures. Advances in aeronautical sciences, vol. 62. Oxford: overload effects and related phenomena. Int J Fatigue 1999;21:
Pergamon Press; 1962. p. 387–408. S233–46.
[4] Jacoby G. Application of microfractography to the study of crack [13] Kujawski D. A new (DK+KMAX)0.5 driving force parameter for
propagation under fatigue stresses, AGARD Report 541, Paris; 1966. crack growth in aluminum alloys. Int J Fatigue 2001;23:
[5] Ling MR, Schijve J. The effect of intermediate heat treatments and 733–40.
overload induce retardations during fatigue crack growth in an Al- [14] Kujawski D. DKeff parameter under re-examination. Int J Fatigue
alloy. Fatigue Fract Eng Mater Struct 1992;15:421–30. 2003;25:793–800.
[6] Kujawski D, Ellyin F. A fatigue crack growth model with load ratio [15] Dinda S, Kujawski D. Correlation and prediction of fatigue crack
effects. Eng Fract Mech 1987;28(4):367–78. growth for different R-ratios using KMAX and DK+ parameters. Eng
[7] Vasudevan AK, Sadananda K, Loaut N. Two critical stress intensities Fract Mech 2004;71(12):1779–90.
for threshold crack propagation. Scripta Metall 1993;28:65–70. [16] Stoychev S, Kujawski D. Analysis of crack propagation using DK and
[8] Vasudevan AK, Sadananda K, Loaut N. A review of crack closure Kmax. Int J Fatigue 2005;27:1425–31.
fatigue crack threshold and related phenomena. Mater Sci Eng [17] Doker H, Bachmann V. Determination of crack opening load by
1993;188A:1–22. use of threshold behavior. In: Newman Jr JC, Elber W, editors.
[9] Sadananda K, Vasudevan AK. Analysis of fatigue crack closure and Mechanics of fatigue crack closure. ASTM STP 982. Philadel-
threshold. In: Fracture mechanics, vol. 25. ASTM STP 1220; 1993. p. phia: American Society for Testing and Materials; 1988. p.
484–501. 247–59.
[10] Vasudevan AK, Sadananda K. Classification of fatigue crack growth [18] Lang M. A model for fatigue crack growth, Part I: phenomenology.
behavior. Met Trans A 1995;26A:1221–34. Fatigue Fract Eng Mater Struct 2000;23:587–601.

You might also like