You are on page 1of 10

THE EFFECT OF COMBINED RESIDUAL AND MECHANICAL STRESS ON CREEP RELAXATION H. Yazdani Nezhad and N.P.

ODowd Department of Mechanical & Aeronautical Engineering Materials & Surface Science Institute University of Limerick, Limerick, Ireland

ABSTRACT The evaluation of combined primary and secondary stress effects on creep deformation and fracture is an important issue for structural integrity assessments. In this work the elastic-plastic creep behaviour of mechanically loaded cracked structures in the presence of residual stress has been studied. The transient crack tip parameter, C (t), has been evaluated in single edge notch bend, SEN(B), and tension, SEN(T), specimens at different residual and mechanical stress levels, using the nite-element method. The effect of elastic-plastic material response on stress relaxation has been studied and results of the nite-element predictions are compared with an approximate analytical solution. KEYWORDS creep, redistribution time, nite element, primary and secondary load, residual stress, stress relaxation 1 INTRODUCTION Residual stress effects on creep behaviour play an important role in structural integrity and life assessment of engineering steel structures (see e.g. [1] and [2]). In this work, nite-element (FE) studies have been carried out to examine the stress redistribution and the associated fracture mechanics parameters during creep under combined mechanical and residual stress. In [3] creep relaxation behaviour of axially loaded cylinders under combined mechanical and thermal loading was examined. A similar approach is adopted in this work where single edge notch bend, SEN(B), and single edge notch tension, SEN(T), geometries have been considered under a combination of primary (mechanical) and secondary (residual) stress. Particular attention has been paid to the redistribution time, i.e. the time for the stress distributions to reach steady state with crack tip stress and strain elds characterised by the parameter C , independent of the initial residual stress distribution. 2 THEORETICAL BACKGROUND Under transient creep conditions, the C (t) parameter can be used to characterise the crack tip stress and strain distribution (see e.g. [4]), C (t) = u ( )dy t ds, W x 0 1 (1)

where ( ) = W d (2)

is the strain energy rate density. C (t) is a path dependent integral and thus is evaluated over paths very close to the crack tip ( 0). Under constant stress and at long times (steady state conditions) the value of C (t) becomes both time and path independent and is denoted by C , C = u ( )dy t ds. W x (3)

As discussed in [3] and examined numerically in this work, the value of C is independent of secondary stresses, as self-equilibrating stresses will relax over long times, but the value of C (t) and the time to reach steady state will be sensitive to the value of residual stress. The average creep strain rate, , can be related to the stress, , during high temperature deformation using the power law relation = B n , (4)

where B and n are material constants. A redistribution time, tred may be dened, (see [3, 5, 6]), as a characteristic normalising time, which represents the approximate time for C (t) to reach its steady state value of C , J0 tred = . (5) C Here J0 is the initial J value, which in this work, includes the contribution of secondary (residual) stress. Under small scale yielding conditions, J0 is obtained from the stress intensity factors, (K p + K s )2 , (6) J0 = E where K p and K s are (initial) primary and secondary stress intensity factors prior to creep, respectively, and E is the plane strain Youngs modulus, where E = E/(1 2 ). The redistribution time, tred , may also be used to provide an estimate of C (t) during the redistribution period (see e.g. [4]), C (t) (1 + t/tred )n+1 = , C (1 + t/tred )n+1 where as proposed in [7] =1 AC , BJ0 (8) (7)

The parameter is dependent on the degree of initial plasticity and has the value of unity for elastic behaviour and is zero under widespread plasticity. The constant A in eqn. (8), depends on the elastic-plastic response of the material. In this work, a power law elastic-plastic material response has been assumed, such that, = /E, y y (/y )N , > y (9)

where is the total (elastic plus plastic strain), y is the material yield stress, y = y /E is the yield strain and N is the hardening exponent. Based on this material response, A in eqn. (8) is given as 1 A = N 1 , (10) E y 2

Equation (7) has been used successfully by a number of authors to predict C (t) under mechanical loading only, e.g. [8]. In this paper, we focus on the application of eqn. (7) under combined mechanical and residual stress. Following [1] the magnitude of the residual stress eld is quantied using the dimensionless parameter , = Ks , K p /Lr (11)

where Lr is a measure of the mechanical (primary) load, Lr = P/PL , (12)

with P the applied load, and PL a limit load, which quanties the proximity to plastic collapse. It should be noted that K p is a linear function of Lr and therefore is not dependent on Lr . In other words, depends only on the value of K s and the specimen geometry. For widespread plasticity, the initial elastic-plastic J value, J0 , can be estimated based on the reference stress method as proposed in [3], J0 = (K p + V K s )2 2 f (Lr ), E (13)

where V is a multiplying factor and can be dened using FE analyses, or otherwise, for a particular geometry by Kp J p+ s 1 , (14) V = s K Jp and
p Jel f (Lr ) = , (15) Jp where J p and J p+s are the J -integral values under primary stress only and combined stresses respectively, and the subscript el denotes a (linear) elastic analysis. Figure 1 presents eqn. (15) in the form of a failure assessment diagram (FAD) for a range of primary loads for both planestrain and plane-stress conditions. As seen the FAD curves are almost independent of crack size (a/W ) for both N = 3 and 10.

3 COMPUTATIONAL APPROACH The modelling framework is described schematically in Fig. 2: a stress distribution is rst introduced by elastic-plastic bending of the uncracked specimen (step 1); unloading of the specimen introduces a residual (secondary) stress distribution (step 2) and a crack is inserted in the body (step 3). In step 4, the mechanical (primary) loading is applied and stress redistribution allowed to occur due to creep.

3.1

Finite-Element Model

The FE mesh used is illustrated in Fig. 3. As shown, concentric elements have been constructed around the crack tip, which results in an accurate extraction of crack tip parameters. Two dimensional 4-node quadratic elements have been used to discretise the model. Linear hybrid plane strain elements (CPE4H in ABAQUS v6.9 [9]) have been used.

1.5

1.5

plane strain elastic-plastic SEN(B)

plane stress elastic-plastic SEN(B)

1 1

1 1

f (Lr )

0.5

a/W a/W a/W a/W

= 0.07 = 0.20 = 0.30 = 0.40

f (Lr )

N= 3

N= 3

0.5

a/W a/W a/W a/W

= 0.07 = 0.20 = 0.30 = 0.40

N = 10

N = 10

0.5

1.5

0.5

1.5

Lr (=P/PL )
(a)

Lr (=P/PL )
(b)

Figure 1: Comparison of f (Lr )-Lr based on the reference stress estimates for different a/W ; (a) plane strain, (b) plane stress 4 RESULTS

4.1

Evaluation of redistribution time, tred and J -integral

Estimating the redistribution time dened by eqn. (5) requires the evaluation of the initial J value, J0 . This could be obtained (i) directly from the FE analysis (after step 3) or (ii) by combining the secondary (residual) stress intensity factor, K s , and the primary stress intensity factor, K p , obtained, e.g. from a handbook solution, and using eqn. (6), or (iii) by using the elastic-plastic relation (13). In this work J0 has been obtained from a FE analysis using the modied J expression, introduced in [10], which provides a path independent J value when residual stresses are present. Figure 4 compares normalised values of J , due to residual stress only (extracted after unloading step 3, Fig. 2a), obtained using the path independent J expression in [10] and the standard J denition used in ABAQUS, [9]. Here the value of J has been normalised by ay y , where a is crack length. It may be seen that the standard value of J shows signicant path dependence while the value obtained using the approach in [10] is almost path independent over 40 domains.

4.2

Creep response under combined primary and secondary loading

4.2.1 Elastic creep analysis In the analyses discussed in this section, the effect of plasticity on stress relaxation is ignored (step 4, in Fig. 2 is an elastic-creep analysis). Results are presented in Fig. 5(a) and 5(b), for the SEN(B) geometry with Lr = 0.3, in terms of normalised C (t) and time normalised by redistribution time, tred , dened using eqn. (5). The symbols in the gure are from the FE analysis and the solid line is the result from eqn. (7) with

(a)

(b)

Figure 2: Combined loading strategy for elastic-plastic analysis; (a) schematic loaded model, (b) loading steps including primary and secondary loads

Figure 3: Typical FE model used in analysis

1.5

= 0.33

a/W=0.07 SEN(B) N =3

0.5

J/(ay y )

= 0.13
0

0.5

results by the method in [10]


1

results by ABAQUS

1.5

10

20

30

40

Domain around crack tip

Figure 4: Comparison of J values due to residual stress from the method by [10] and ABAQUS, for two levels of residual stress, = 0.13, 0.33. = 1. The value of C (t) is an average obtained directly from the FE analysis for the rst ve rings of elements adjacent to the crack tip (excluding the rst ring) to represent the contour, , approaching zero. The value of C is taken to be the path and time independent C (t) value at very large times, t/tred >> 1. The mechanical load is quantied by Lr and the residual stress by , as dened in Section 2. The results show that, for the SEN(B) geometry with n = 3, eqn. (7) provides a good estimate for C (t) during stress redistribution for combined mechanical loading and residual stress (Fig. 5a). However, for n = 10, eqn. (7) provides a non-conservative estimate of C (t) (underestimates the FE value). It may also be noted that eqn. (5) provides an accurate and conservative estimate of the redistribution time for both values of n. Figures 5(c) and 5(d) present the results for the SEN(T) geometry, for n = 3 and n = 10 respectively. The agreement between the numerical and theoretical values is less good than for the SEN(B) geometry for n = 3 value though the prediction remains conservative for n = 3. As for the SEN(B) case (Fig. 5b), agreement is less good between the numerical model and theoretical equation for n = 10, with non-conservative predictions for C (t) for = 0.45. 4.2.2 Elastic-plastic creep analysis In this section, the effect of plasticity on the creep redistribution is incorporated in the analysis. Elastic-plastic-creep results for combined primary and secondary loading under bending load are presented in Fig. 6. The results for mechanical loading only ( = 0) when n = N = 3 show good agreement between the FE analysis and the prediction by eqn. (7) (compare the solid line and square symbols in each gure). The FE result is provided along with the prediction from eqn. (7). Note that if obtained from eqn. (8) is less than zero, then is set to be zero (see [7]) which implies

10

10

= 0.00 = 0.13 = 0.33 Eqn. (7), = 1

= 0.00 = 0.18 = 0.45 Eqn. (7), = 1

C (t)/C

n=3 Lr = 0.3 a/W = 0.07 SEN(B) elastic creep

C (t)/C

n = 10 Lr = 0.3 a/W = 0.07 SEN(B) elastic creep

1
0.01 0.1 1 10

1
0.01 0.1 1 10

t/tred
(a)
10 10

t/tred
(b)

= 0.00 = 0.13 = 0.33 Eqn. (7), = 1

= 0.00 = 0.18 = 0.45 Eqn. (7), = 1

C (t)/C

n=3 Lr = 0.3 a/W = 0.07 SEN(T) elastic creep

C (t)/C

n = 10 Lr = 0.3 a/W = 0.07 SEN(T) elastic creep

1
0.01 0.1 1 10

1
0.01 0.1 1 10

t/tred
(c)

t/tred
(d)

Figure 5: Comparison of evolution of C (t) under elastic creep conditions at different for a/W = 0.07; (a) SEN(B), n = 3, (b) SEN(B), n = 10, (c) SEN(T), n = 3, (d) SEN(T), n = 10

10

10

(a)
8

FE 0.00 0.13 0.33

Eqn. (7)

(a)
8

= 0.00 = 0.18 = 0.45 Eqn. (7), 1

C (t)/C

n=N =3 Lr = 0.3 a/W = 0.07 SEN(B) elastic-plastic creep

C (t)/C

n = N = 10 Lr = 0.3 a/W = 0.07 SEN(B) elastic-plastic creep

1
0.01 0.1 1 10

1
0.01 0.1 1 10

t/tred
10 10

t/tred
Eqn. (7)

(b)
8

FE 0.00 0.13 0.33

(b)
8

= 0.00 = 0.18 = 0.45 Eqn. (7), 1

C (t)/C

C (t)/C

n=N =3 Lr = 0.5 a/W = 0.07 SEN(B) elastic-plastic creep

n = N = 10 Lr = 0.5 a/W = 0.07 SEN(B) elastic-plastic creep

1
0.01 0.1 1 10

1
0.01 0.1 1 10

t/tred
4
10

t/tred
FE Eqn. (7) =0 =0

(c)

0.00 0.13 0.33

(c)
8

FE 0.00 0.18 0.45

Eqn. (7) =0 =0

C (t)/C

C (t)/C

n=N =3 Lr = 1.0 a/W = 0.07 SEN(B) elastic-plastic creep

n = N = 10 Lr = 1.0 a/W = 0.07 SEN(B) elastic-plastic creep

1
2

1
0.01 0.1 1 10
0.01 0.1 1 10

t/tred

t/tred

Figure 6: Comparison of evolution of C (t) under elastic-plastic creep conditions in SEN(B) at different for n = N = 3 and a/W = 0.07; (a) Lr = 0.3, (b) Lr = 0.5, (c) Lr = 1.0 8

Figure 7: Comparison of evolution of C (t) under elastic-plastic creep conditions in SEN(B) at different for n = N = 10 and a/W = 0.07; (a) Lr = 0.3, (b) Lr = 0.5, (c) Lr = 1.0

that C (t) immediately reaches C (see eqn. (7)). These results show that under elastic-plastic creep conditions with primary stress only ( = 0) eqn. (7) provides a good estimate for C (t) during stress redistribution for n = N = 3. Also the time to reach steady state, i.e. when C (t) reaches C , is closely represented by tred . For combined primary and secondary stress under elastic-plastic creep conditions the theoretical curves depends on as depends on through J0 . The C (t) predictions from eqn. (7) are non-conservative (underestimate the FE results) for most non-zero values examined for n = N = 3. (It should be noted that in [3] some non-conservatism was reported for the cases of combined thermal and mechanical loading in the elastic-plastic analysis of cylindrical geometries.) Figure 7 shows the results for n = N = 10. Note that in this case (high value of n and N ), in eqn. (7) is approximately 1, except for Lr = 1.0 (Fig. 7c), so a single line is obtained for the examined values of . Similar trends are seen as for n = 3, with the prediction of eqn. (7) providing increasingly non-conservative estimates of C (t) as Lr and are increased, e.g. Fig. 7(c). Though not shown here, similar trends have been seen in the elastic-plastic creep analysis of the SEN(T) geometry. 5 CONCLUSIONS Creep relaxation in the presence of residual stress under elastic-plastic creep and elastic-creep conditions was examined in this work. The C (t) estimation scheme, in general, underestimates the value obtained from the FE analysis for the elastic-plastic creep analysis with the predicted effect of combined residual and mechanical loading from eqn. (7) being signicantly larger than is observed in the FE analysis. Similar trends, though less dramatic, were seen in [3]. Work is ongoing to provide a more accurate estimate of stress relaxation under combined residual and mechanical loading for elastic-plastic creep conditions. ACKNOWLEDGMENTS Financial support for this work has been provided by Science Foundation Ireland under grant 08/RFP/ENM1477. Helpful discussions on numerical methods with Dr Dongfeng Li at the University of Limerick, Drs Catrin M Davies and Simon Kamel and Prof. Kamran Nikbin from Imperial College London are gratefully acknowledged. References [1] R6, Revision 4: Assessment of the Integrity of Structures Containing Defects Gloucester, UK: British Energy Generation Ltd., 2009. [2] R5, Revision 2: An Assessment Procedure for the High Temperature Response of Structures Gloucester, UK: British Energy Generation Ltd., 2003. [3] Song, T.K.; Kim ,Y.J.; Nikbin, K.; Ainsworth, R.A.: Estimation of the Transient Creep Parameter C (t) under Combined Mechanical and Thermal Stresses J. Engng. Fract. Mech. 77 (2010) No. 4, pp. 685704

[4] Webster, G.A.; Ainsworth, R.A.: High Temperature Component Life Assessment London, UK: Chapman & Hall, 1994 [5] Ainsworth, R.A.; Budden, P.J.: Crack Tip Fields under Non-steady Creep Conditions -1. Estimates of the Amplitude of the Fields J. Fat. Fract. Engng. Mater. Struct. 13 (1990) No. 3, pp. 263276 [6] Davies, C.M.; ODowd, N.P.; Nikbin, K.M.; Webster, G.A.: An Analytical and Computational Study of Crack Initiation under Transient Creep Conditions Int. J. Solids Struct. 44 (2007) No. 6, pp. 18231843. [7] Joch, J.; Ainsworth, R.A.: The Effect of Geometry on the Development of Creep Singular Fields for Defects under Step-load Controlled Loading J. Fat. Fract. Engng. Mater. Struct. 15 (1992) No. 3, pp. 229240. [8] Kim, Y.J.; Dean, D.W.; Budden, P.J.: Finite Element Analysis to Assess the Effect of Initial Plasticity on Transient Creep for Defects under Mechanical Loading Int. J. Pres. Ves. Pip. 78 (2001) No. 11-12, pp. 10211029 [9] ABAQUS v6.9. Users Manual. Inc. and Dassault Systems. [10] Lei, Y.: J -integral Evaluation for Cases Involving Non-proportional Stressing J. Engng. Fract. Mech. 72 (2005) No. 4, pp. 577596 [11] Ainsworth, R.A.; Sharples, J.K.; Smith, S.D.: Effects of Residual Stresses on Fracture Behaviour - Experimental Results and Assessment Methods J. Strain Anal. 35 (2000) No. 4, pp. 307316 [12] ODowd, N.P.; Nikbin, K.M.; Wimpory, R.C.; Biglary, F.R.; ODonnell, M.P.: Computational and Experimental Studies of High Temperature Crack Initiation in the Presence of Residual Stress J. Pres. Ves. Tech. 130 (2008) No. 4, 041403 [13] Turski, M.; Bouchard, P.J.; Steuwer, A.; Withers P.J.: Residual Stress Driven Creep Cracking in AISI Type 316 Stainless Steel J. Acta Materialia 56 (2008) No. 14, pp. 35983612 [14] Hossain, S.; Truman, C.E.; Smith, D.J.; Peng, R.L.; Stuhr, U.: A Study of the Generation and Creep Relaxation of Triaxial Residual Stresses in Stainless Steel Int. J. Solids Struct. 44 (2007) No. 9, pp. 30042020 Corresponding author: Hamed.Yazdani-Nezhad@ul.ie

10

You might also like