Crack Tip Stresses and Their Effect On S

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Available online at www.sciencedirect.

com

Engineering Fracture Mechanics 75 (2008) 2469–2479


www.elsevier.com/locate/engfracmech

Crack-tip stresses and their effect on stress intensity factor


for crack propagation
S. Stoychev, D. Kujawski *

Department of Mechanical and Aeronautical Engineering, College of Engineering and Applied Sciences,
Western Michigan University, Kalamazoo, MI 49008-5343, United States

Received 2 February 2007; received in revised form 27 July 2007; accepted 31 July 2007
Available online 8 August 2007

Abstract

In this paper, analytical and numerical simulations of the crack-tip stresses are presented. Analytical calculations are
performed utilizing modified Rice and elastic equations. 2D finite element analyses (FEA) are conducted in parallel using
ANSYS. Results from both methods are compared and discussed in terms of the crack-tip compressive residual stresses.
Then, the effect of the compressive residual stresses is quantified in terms of the stress intensity factor for crack propaga-
tion, KPR, using ‘clamping force’ method. The comparison between the calculated KPR values and those obtained exper-
imentally by Lang demonstrates a fairly good agreement. Both the present results and Lang data independently support a
two-parameter, KmaxTH and DKTH, description of the threshold condition for fatigue crack propagation.
Ó 2007 Elsevier Ltd. All rights reserved.

Keywords: FEA; Residual stresses; Crack-tip clamping; Two-parameter threshold

1. Introduction

It has been observed that fatigue crack growth (FCG) behavior depends on two fracture mechanics param-
eters, namely DK and Kmax. The range DK accounts for fatigue damage whereas Kmax for the crack-tip stretch-
ing and associated material separation. In general, FCG rate can be represented as a three-dimensional plot,
da/dN versus DK and Kmax. Traditionally however, FCG data are graphed as two-dimensional plots of crack
growth rate da/dN versus DK for a given load ratio, R = Pmin/Pmax = Kmin/Kmax [1]. In such approaches, the
effect of Kmax is not accounted directly but indirectly through R-ratio. It can be noted that Kmax is an intrinsic
fracture mechanics parameter whereas the R-ratio is not.
Hence, in a number of articles Sadananda and Vasudevan [2–7] proposed a unified description of the FCG
behavior using DK and Kmax parameters that account for both an applied load and internal/residual stress
contributions. They postulated [7] that contribution of the internal/residual stresses is the missing link to
connect the long and the short crack growth description. In their formulation the threshold condition is

*
Corresponding author.
E-mail address: daniel.Kujawski@wmich.edu (D. Kujawski).

0013-7944/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfracmech.2007.07.019
2470 S. Stoychev, D. Kujawski / Engineering Fracture Mechanics 75 (2008) 2469–2479

represented by L-shape curve in terms of DKTH and KmaxTH instead of DKTH alone, which gives an incomplete
picture. For example, it is observed that DKTH values may differ depending on testing methodology. This is
because each test method introduces it own, in situ generated, crack-tip stresses that affect the individual
D KTH which pertains to a different segment of the total L-shape curve [8]. In order for a crack to propagate,
both KmaxTH and DKTH must be satisfied simultaneously. The unified approach, however, is a conceptual
framework rather than an actual FCG prediction method. Subsequently, Kujawski and co-workers utilized
a composed fatigue crack driving force, Dj = (Kmax)p(DK)1p, which provides a tool for practical implemen-
tation of the unified approach [9–13]. The composed fatigue crack driving force, Dj, was able to account for
load interaction effects due to R-ratios, block loadings and overloads.
In this study, we are concerned about the crack-tip stresses (perpendicular to the crack plane) set up due to
plasticity. In particular, at zero applied load, these stresses are commonly known as the crack-tip residual
stresses. In general, residual stresses may exist due to previous material processing or can be in situ generated.
For example, a welded or quenched material may contain residual stresses. By measuring da/dN for this mate-
rial and comparing it with the annealed wrought material data the effect of residual stresses on FCG behavior
can be quantified. On the other hand, there is no common agreement how to quantify the effect of the crack-tip
residual stresses (in situ generated) on FCG behavior – in particular, whether these stresses operate behind (in
the wake) or in front of the crack or both. It is often argued that this inherent controversy exists since it is
difficult to separate plasticity-induced phenomena operating in the crack wake (crack closure) from that at
the crack front (residual stresses) [10]. Recently, the present authors quantified experimentally and modeled
the effects of residual compressive stresses induced by preloading cycles on fatigue crack initiation at notches
using a newly developed ‘clamping force’ method [14]. Any uncertainty associated with crack wake effect was
avoided by considering a notched component rather than cracked one.
In this paper, the effect of the in situ generated compressive stresses at the crack-tip is modeled by adopting
the ‘clamping force’ method. The actual effect is quantified in terms of the stress intensity factor for crack
propagation, KPR. The results of such modeling however, depends on the accuracy at which these compressive
stresses are determined. Therefore, in the first part of the paper analytical estimations for the crack-tip stresses
are compared with finite element analysis for positive and negative load ratios, R. The analysis is limited to
stationary cracks. This simplification allows the analytical model to be suitable for a cycle-by-cycle
application.

2. Analytical calculations

2.1. Positive load ratios

For a given Kmax, the maximum stress profile along the crack plane in the plastic zone is calculated using
the modified Rice equation [15] and by the linear-elastic solution elsewhere:
 1þn
n0
K 2max 0

rmax ¼ r0 0 2 
Rice
ð1 þ n Þpr0 ðX þ q Þ
ð1Þ
K max
rmax ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Elastic
2pðX þ q Þ

In Eq. (1), q is the ‘process zone’ size. Both stress and strain are assumed to be approximately constant in this
zone. The size of q is determined from the condition that the strain at the crack-tip ‘process zone’ cannot be
larger than the ultimate strain of the material. r0 is the yield stress, n 0 is the stain hardening exponent, and X is
the distance from the crack-tip. Fig. 1 shows how the elastic solution is ‘offset’ in the direction of crack prop-
agation by the amount needed to coincide with the yield point from Rice equation (approximately half of the
plastic zone size, X0/2 = (1/p)(Kmax/r0)2/2.
Fig. 2 illustrates how the procedure outlined above is used to determine stress distribution ahead of the
crack-tip for cyclic loading. The stress range, Dr, is calculated using the same procedure where Kmax is substi-
tuted by DK and r0 by 2r0. The minimum stress profile, rmin, is calculated by subtracting Dr from rmax.
S. Stoychev, D. Kujawski / Engineering Fracture Mechanics 75 (2008) 2469–2479 2471

Stress
Elastic
~X0/2

Rice
σ0
Final stress profile
ρ*

Distance from the crack tip -x

Fig. 1. Analytical calculation of the crack-tip stresses.

4 σ 2 4
P
2
σ1 = 0
3 σ2 = σmax2
1
t σ3 = σ2 - Δσ3
3 σ4 = σmax4
1
ε
Fig. 2. Calculation of crack-tip stresses procedure for complex load histories.

In general, Eq. (1) is needed only if applied load exceeds all the peaks from the past load history. In all other
cases, Dr is calculated instead, and added (or subtracted) to the stress profile from the last peak [12].

2.2. Negative load ratios

The method outlined in previous section provides the minimum stresses without any consideration of the
crack faces contact phenomena. Under negative loads, however, the crack faces may contact each other.
Therefore the compressive portion of the applied load would be distributed along both uncracked and cracked
but being in contact cross-section of the specimen. This calls for a different procedure for calculation of the
stresses ahead of the crack-tip when Pmin < 0.
Fig. 3 depicts the stress profiles for the maximum stress, rmax, and stress range, Dr (full lines) resulting from
the maximum load followed by unloading to zero. For any negative load ratio but the same maximum applied
load, the maximum stress profile does not change. The stress range however, should be increased everywhere
by the applied minimum stress jSminj (dashed line). This simple superposition assumption is valid only for the
elastic stresses when the stress range does not exceed 2r0. Therefore, a new yield point, X0, corresponding to
Dr = 2r0 has to be determined. Theoretically, the stress range in this plastic zone cannot be calculated directly
using DK value, because the crack is closed under compressive load and the stress intensity factor can not be
defined. In order to solve that problem, the Rice equation (Eq. (1)) will be modified to the following form:
1
Dr ¼ C n0
ð2aÞ
ðX þ qÞ1þn0
! n0
1þn0
DK 2
C ¼ 2r0 2
ð2bÞ
ð1 þ n0 Þpð2r0 Þ
The constant, C, in Eq. (2) must be recalculated to allow for Rice and elastic solutions to intersect in the new
yield point, X new
0 which is determined from Fig. 3:
2472 S. Stoychev, D. Kujawski / Engineering Fracture Mechanics 75 (2008) 2469–2479

Fig. 3. Procedure for calculation of the stress distribution ahead of the crack-tip for negative applied load ratios.

Fig. 4. Calculation of the minimum stress profile under negative applied loads.

1
Dr ¼ 2r0 ¼ C n0
ð3aÞ
ðX new
0 þ qÞ1þn0
n0
C ¼ 2r0 ðX new
0 þ qÞ1þn0 ð3bÞ
Finally, the constant, C, in Eq. (3b) is substituted in Eq. (2a). This operation preserves the slope of the Rice
equation (on a log–log plot) and changes only the vertical position in order to remove the discontinuity in the
stress distribution at plastic zone boundary. Eq. (3) is just a simple phenomenological modeling, which will be
checked in the following section with the results of the finite element analysis.
Fig. 4 illustrates a step-by-step calculation of the stress distribution ahead of the crack-tip for R = 3. The
stresses at point 3, S3, cannot be used for the stress intensity factor (SIF) calculations because the crack is
closed and the SIF cannot be defined. Therefore the specimen has to be unloaded back to point ‘4’ till applied
K = 0. As a result, the stresses in the whole cross section of the specimen (including the plastic zone) will relax
elastically by the amount of the applied minimum stress, S3 (r4 = r3 + jS3j).

3. Finite element analysis

Finite element analysis (FEA) was conducted in 2D with ANSYS using plane stress assumption. The geo-
metry was meshed with triangular elements (Plane2). The crack wake was meshed with contact element
S. Stoychev, D. Kujawski / Engineering Fracture Mechanics 75 (2008) 2469–2479 2473

couples (Target160–Conta172). This allowed modeling a possible crack wake contact effect. The mesh was
refined around the crack-tip in order to capture the steep stress gradient expected at that region. The smallest
elements were had a 0.05 mm radius, which is 0.13% of the plastic zone size under maximum load. For the
center crack specimen used in this study the total number of elements was 1487. Due to the symmetry condi-
tion only a quarter of the center crack specimen (shown in Fig. 5) was modeled. By removing the symmetry
constraints on the vertical edge of the center crack specimen an edge crack geometry was modeled.
The next step of the finite element modeling requires a proper characterization of the material stress–strain
curve. In the present simulations three possibilities for the material stress–strain behavior have been consid-
ered namely, a bilinear model and two multilinear models with stresses up to 2r0 and 5r0 as are shown
in Fig. 6. Such high ultimate stresses were chosen in order to simulate the high stress gradient at crack-tip

Fig. 5. Finite element models for a center crack specimen.

Multilinear up to 2σ0
2 a
Multilinear up to 5σ0
1.5
6
5 b
σ / σ0

1 4

Bilinear
σ / σ0

3
2

0.5 1
0
0 2 4
Strain
0
0 0.05 0.1 0.15
Strain

Fig. 6. Material models for FEA.


2474 S. Stoychev, D. Kujawski / Engineering Fracture Mechanics 75 (2008) 2469–2479

Table 1
Material constants and material models
Bilinear Multilinear up to 2r0 Multilinear up to 5r0
e r [MPa] e r [MPa] e r [MPa]
1 0 0 0 0 0 0
2 0.006 430 0.006 430 0.006 430
3 0.303 2150 0.009 473 0.024 600
4 0.013 516 0.044 700
5 0.018 559 0.118 900
6 0.024 602 0.180 1000
7 0.031 645 0.372 1200
8 0.041 688 0.687 1400
9 0.052 731 1.169 1600
10 0.065 774 1.869 1800
11 0.081 817 2.843 2000
12 0.099 860 3.792 2150
Material constants
n0 0.2512
H 0 [Mpa] 1538.3
E [Mpa] 71000
t 0.33
r0 [MPa] 430

vicinity. The bilinear model consists of two slopes – the elastic modulus below and 10 times smaller slope
above yielding. The discrete points for the multilinear models are calculated using the Ramberg–Osgood equa-
tion with the material constants given in Table 1 (for Al 2324-T39 alloy).
Fig. 6 shows that the bilinear model is in close agreement with the other two only when the stresses are
lower than twice the yield stress. This means that although easy to implement and still widely used, the bilinear
model is expected to introduce error close to the crack-tip, in particular within the cyclic plastic zone. The
stresses at the cyclic plastic zone (‘process zone’) are used in the ‘clamping force’ method [14] for the residual
stress intensity factor calculation (described in the next section). Therefore, only the multilinear models will be
used in the finite element simulations.

4. FEA results versus analytical calculations

Both FEA and analytical simulations were conducted using a simple loading history shown in Fig. 7a. The
stress peaks were applied in the following sequence 136–192–136 MPa (with R = 0 and the corresponding
overload ratio OL = 1.41). This allows for proper sizes of the plastic zone under maximum loading (2–4–
2 mm). These plastic zone sizes are sufficiently small to satisfy the small scale yielding conditions and, at
the same time they are relatively large in comparison to the finite element size.
It can be seen in Fig. 7b that the stress profile corresponding to the overload (point 3) is ‘capped’. This is a
direct consequence of the assumption (used in ANSYS) that the stress–strain curve becomes nearly horizontal
after the last discrete point in the multilinear model is reached. The multilinear model shown in Fig. 7c is used
in the rest of the paper.
A careful inspection of Fig. 7b and c, indicates that there is a clear trend in the FEA results which some-
what overestimate the analytical predictions. Despite this, it can be concluded that analytical results correlate
favorably with that from FEA. It can be noted, that similar simulations were conducted using the edge crack
geometry. The applied stresses were adjusted in order to keep the SIF the same as the ones for the center crack
simulations. FEA results for the edge crack geometry (not shown) were essentially the same as that for the
center crack depicted in Fig. 7.
In order to confirm that Fig. 7 is not just an isolated case, another overload history with R > 0 (R = 0.5 and
OL = 1.41) was simulated (Fig. 8). The peak values of the applied stress remained the same, only the valleys
were changed to 68 MPa. The results show that the time consuming FEA calculations can be substituted with
S. Stoychev, D. Kujawski / Engineering Fracture Mechanics 75 (2008) 2469–2479 2475

a
3

Load
1 5

2 4
Time

3.0 3.0

2.5 b 2.5 c
ANSYS
2.0 Analytical 2.0

1.5 1.5
1
3 1 3
σy / σ0
σy / σ0

1.0 1.0
5 5
0.5 0.5

0.0 0.0

2 4 2 4
-0.5 Multilinear power law up -0.5 Multilinear power law up
to 2*So to 5*So
-1.0 -1.0
0 1 2 3 4 5 0 1 2 3 4 5
Distance ahead of the crack tip [mm] Distance ahead of the crack tip [mm]

Fig. 7. Effect of the material model.

analytical calculations without introducing a significant error. This conclusion was challenged further by
studying a negative load ratio depicted in Fig. 9.
Fig. 9 shows the load history and the predicted stresses for R = 1. The maximum applied stress was
192 MPa. Three cycles were simulated in order to check if any transient effects are present. The FEA results
did not show any significant transient effects and therefore only the last cycle is presented. In particular,
Fig. 9a and b shows only the negative (unloading and reloading) part of the cycle since the positive part
was already reported in Figs. 3 and 4. It can be seen from Fig. 9b that during reloading from Pmin the
FEA predicts higher stresses than expected from the analytical model in particular for the elements very close
to the crack-tip. This is expected because the analytical model assumes that the crack closes completely below
zero load. The FEA simulation of crack mouth opening indicates that the crack-tip is open even at Pmin as it is
depicted in Fig. 10. It should be noted however, that additional finite element simulations of propagating
crack indicated that crack was closed to its tip. Therefore, the assumption used in analytical calculation agrees
better with actual situation for propagating than stationary crack.

5. Calculation of Kres

The effect of the compressive stresses ahead of the crack-tip will be quantified adopting the recently devel-
oped ‘clamping force’ method [14]. First, the profiles of the in situ generated stresses ahead of the crack-tip
2476 S. Stoychev, D. Kujawski / Engineering Fracture Mechanics 75 (2008) 2469–2479

3.0

2.5

3
Load

2.0
ANSYS
Analytical
1 5 1.5

σy / σ0
3
1.0
5
2 4 0.5

2
0.0
4
Time -0.5

-1.0
0 1 2 3 4 5

Distance ahead of the crack tip [mm]

Fig. 8. Overload simulated with both FEA and analytical formulas (R = 0.5).

0.5
a
1
0 2
3
4
5
6

-0.5
σy / σ0

-1
Load

1
2 Time
-1.5 ANSYS 3
Analytical 4
5
6

-2
0 1 2 3 4 5
X [mm]

1
b ANSYS
0.5 Analytical

5
0 4
3
2
1
σy / σ0

-0.5

-1
Load

5
4 Time
-1.5 3
2
1

-2
0 1 2 3 4 5
X [mm]

Fig. 9. Correlation between the analytical model for calculation of the internal stresses and the FEA results.

during cyclic loading (R = 0) are shown in Fig. 11. The cyclic plastic zone (CPZ) is define as the zone adjacent
to a crack-tip where Dr = rmax  rmin P 2r0. Within the CPZ the compressive residual stresses, rres = rmin,
S. Stoychev, D. Kujawski / Engineering Fracture Mechanics 75 (2008) 2469–2479 2477

9.E-05
Load
8.E-05 1
1 2
7.E-05 3
2

6.E-05 3 Time

5.E-05 20
Y [m]

4.E-05

3.E-05

2.E-05

1.E-05

0.E+00 20
-5.E-03 -4.E-03 -3.E-03 -2.E-03 -1.E-03 0.E+00
X [m]

Fig. 10. Crack mouth opening from the FEA model (R = 1).

Fig. 11. Clamping force procedure for Kres calculation for R = 0.

‘clamp’ locally the material and prevent it from separation. In order to determine when the clamping action
vanishes the following incremental re-loading procedure from Kmin for R P 0 and from K = 0 for R < 0 (point
4 in Fig. 4) is utilized. Then, the average stress within the CPZ is determined by integrating the corresponding
stress profile within the CPZ. The clamping action would vanish when the average stress within the CPZ is
equal zero.
The applied stress intensity factor that corresponds to zero clamping action is called Kres. Then, the stress
intensity factor for crack propagation is calculated as
K PR ¼ K min þ K res for R P 0 ð4aÞ
K PR ¼ K res for R < 0 ð4bÞ
2478 S. Stoychev, D. Kujawski / Engineering Fracture Mechanics 75 (2008) 2469–2479

0.9

0.8 Lang
(experiments)
0.7

Kpr / Kmax
0.6

0.5

0.4 Simulation

ax
0.3

m
/K
in
0.2

Km
0.1

0
-0.5 -0.3 -0.1 0.1 0.3 0.5 0.7 0.9
R or UR

Fig. 12. Stress intensity factor for crack propagation – predictions and experiments.

Fig. 12 shows the KPR values calculated from Eq. (4) and normalized by Kmax versus load ratio, R. For
comparison, experimental KPR data from Lang [16] are also shown in the figure. Note that, for negative load
ratios, KPR results are plotted in terms of UR = Smin/r0 instead of R in order to make comparison with Lang’s
data (for Al 7475-T7351 alloy). The upper and lower KPR curves reported by Lang correspond to constant
amplitude and overload tests, respectively. It can be seen that the agreement between experimental and sim-
ulated values for KPR is fairly good.

6. Discussion

Lang [16] has demonstrated experimentally that KPR values are associated with residual stresses ahead of
the crack-tip and that the effect is not due to plasticity induced closure. He used the following formula for KPR
calculations:
K PR ¼ K max  DK TH ð5Þ
where Kmax is the stress intensity factor below which fatigue crack does not propagate under cyclic stress inten-
sity range somewhat large than DKTH. In terms of the present approach, KPR given by Eq. (5) represents the
minimum value of the applied stress intensity factor when the local clamping action of the compressive stresses
within the ‘process zone’ vanishes. In other words, Kmax in Eq. (5) corresponds to KmaxTH in the unified ap-
proach proposed by Sadananda and Vasudevan [2–8]. Thus, Eq. (5) can be rewritten in the following form:
K PR ¼ K maxTH  DK TH ð6Þ
Hence, both the present analytical results and Lang’s experimental data independently support the two-
parameter threshold condition postulated by Sadananda and Vasudevan in terms of both KmaxTH and DKTH
values.

7. Conclusions

Simulations of crack-tip stresses from a plane stress FEA and an analytical method are presented. Compar-
ison between the two methods shows that the analytical model predicts the stresses fairly well without the need
of extensive calculations that are typical for the elastic–plastic FEA. As such, the analytical approach seems to
be sufficient for determination of the in situ generated residual stresses ahead of the crack-tip. Then, the effect
of these residual stresses on the KPR has been quantified by the means of the ‘clamping force’ methods. The
comparison between the simulated KPR values and those obtained experimentally by Lang demonstrates a
fairly good agreement. Both the present results and Lang data strongly support the existence of a two-param-
eter threshold condition in terms of KmaxTH and DKTH.
To the best of our knowledge, this is the first analytical attempt to model the effects of in situ generated
compressive residual stresses on KPR.
S. Stoychev, D. Kujawski / Engineering Fracture Mechanics 75 (2008) 2469–2479 2479

Acknowledgements

This study is supported by the Office of Naval Research under Grants N00014-04-1-0718 and N00014-05-1-
0872. The authors would like to thank Mr. B. Taylor for supplying Fig. 10.

References

[1] Suresh S. Fatigue of materials. Cambridge: Cambridge University Press; 1991.


[2] Vasudevan AK, Sadananda K, Loaut N. Two critical stress intensities for threshold crack propagation. Scripta Metall 1993;28:65–70.
[3] Sadananda K, Vasudevan AK. Analysis of fatigue crack closure and threshold. Fracture mechanics. ASTM STP 1220, vol. 25. 1993.
p. 484–501.
[4] Vasudevan AK, Sadananda K. Classification of fatigue crack growth behavior. Met Trans A 1995;26A:1221–34.
[5] Sadananda K, Vasudevan AK. Short crack growth and internal stresses. Int J Fatigue 1997;19:S99–S109.
[6] Sadananda K, Vasudevan AK, Holtz RL, Lee EU. Analysis of overload effects and related phenomena. Int J Fatigue
1999;21:S233–46.
[7] Vasudevan AK, Sadananda K, Glinka G. Critical parameters for fatigue damage. Int J Fatigue 2001;23:S39–53.
[8] Vasudevan AK, Sadananda K. Transient stress effects due to testing methodology on fatigue crack growth thresholds. Int J Fatigue
2007;29:1985–9.
[9] Kujawski D. A new (DK+Kmax)0.5 driving force parameter for crack growth in aluminum alloys. Int J Fatigue 2001;23:733–40.
[10] Kujawski D. DKeff parameter under re-examination. Int J Fatigue 2003;25:793–800.
[11] Dinda S, Kujawski D. Correlation and prediction of fatigue crack growth for different R-ratios using Kmax and DK+ parameters.
Engng Fract Mech 2004;71(12):1779–90.
[12] Stoychev S, Kujawski D. Analysis of crack propagation using DK and Kmax. Int J Fatigue 2005;27:1425–31.
[13] Stoychev S. Load interaction effects on fatigue crack growth. PhD dissertation, Western Michigan University, etd-0519105-163119,
05/2005.
[14] Kujawski S, Stoychev S. ‘Internal stress effect on fatigue crack initiation at notches’. Int J Fatigue 2007;29:1744–50.
[15] Rice JR. Stress due to a sharp notch in a work-hardening elastic–plastic material loaded by longitudinal shear. J Appl Mech
1967;34:298–387.
[16] Lang M. Methodology for fatigue crack growth, part I: Phenomenology, fatigue and fracture of engineering. Mater Struct
2000;23:581–601.

You might also like