Three-dimensional elastic-plastic j-integral calculations for semielliptical surface cracks in a tensile plate

Engineering Fracture Mechanics Printed in Great Britain.

Vol. 29, No. 1, pp. 81-87,

0013-7944/88 Pergamon

1988

$3.00+ .OO Journals Ltd.

THREE-DIMENSIONAL ELASTIC-PLASTIC J-INTEGRAL CALCULATIONS FOR SEMIELLIPTICAL SURFACE CRACKS IN A TENSILE PLATE G. P. NIKISHKOVI’ and S. N. ATLURIS Center for the Advancement of Computational Mechanics, Georgia Institute of Technology, Atlanta, Ga 30332, U.S.A. Abstract-Using a finite element method, in conjunction with an equivalent domain integral approach, the elastic-plastic crack-tip J-integral is computed for semielliptical surface flaws in a plate under a remote tensile load. The computed elastic-plastic values of the J-integral are compared to the elastic ones. It is concluded that a linear elastic fracture mechanics approach can be used for semielliptical cracks of relative depth of a/r = 0.25 and aspect ratio of c/a 5 3, up to load levels of 0.9cr, with an error of no more than 10% in the (equivalent) elastic-plastic stress intensity factor. A simple empirical equation for evaluating the elastic-plastic J-integral

for surface flawsis proposed.

INTRODUCTION

of surface flaws is of great importance in structural integrity analyses. Usually these cracks have a shape which is close to that of a semiellipse. Often a semielliptical crack is considered as a hypothetical defect in a structure during the design procedure. The postulated “maximum” flaw as in the ASME code[l] can serve as an example. The standard procedure for fracture analysis consists of calculating some well-defined fracture mechanics parameter and comparing it to the critical value defined through material testing. If the load level is not high such that the plastic zone size is small as compared to the crack depth, then the linear elastic stress intensity factor K can be used as a well-established fracture parameter. Elastic stress intensity factors are available from handbooks for many practical crack configurations. Even for three-dimensional semielliptical cracks, there are reliable solutions and approximate equations for stress intensity factor evaluation[2,3]. The J-integral is a widely recognized fracture parameter in nonlinear elastic-plastic fracture mechanics. Some very useful numerical solutions for the elastic-plastic J-integral are presented in [4], but they are limited to two-dimensional bodies. Even though elastic-plastic analyses of surface flaws in a cylindrical pressure vessel was carried out in [5], it is somewhat difficult to use the results of this publication for other (even simple) crack configurations. In this brief paper, we present the results of three-dimensional elastic-plastic J-integral calculations for several semielliptical surface cracks in a tension plate. We propose a simple empirical equation which enables one to estimate the value of the elastic-plastic J-integral by using the well-known solutions for the elastic stress intensity factor.

THE ASSESSMENT

NUMERICAL

PROCEDURE

The boundary value problem is solved by a finite element displacement method. Quadratic 20-node brick type elements are used for discretization. Nonlinear material behavior is modeled by means of the flow theory of plasticity with an isotropic hardening. An “initial-stress” iterative scheme was adopted for the solution of nonlinear equations; both the residual force and the relative norms of the incremental displacement were controlled during the convergence process. A sky-line, out-of-core solver was used to solve the system of linear algebraic equations. The values of J-integral along the crack front were computed by an equivalent domain TVisitingAssociate Professor; permanent address: Moscow Institute of Engineering Physics, Moscow 115409. SRegents’ Professor and Director. 81

82

G. P. NIKISHKOV

and S. N.

ATLURI

function

Fig. 1. Integration volume V around a crack front segment.

integral (EDI) method[6] Jf=-I,(Wg-

~ij~~)dV-~“(~-~(lr,~))

sdV +

(1)

The integration volume V is a disc centered around the crack front segment, with a cylindrical surface A, and side surfaces Al and A2 respectively (see Fig. 1). In eq. (1) xj is the crack front coordinate system, mijare stresses, Uiare displacements, W is the stress work density, nj are the components of normal vector to the surface, s is an arbitrary continuous function which is equal to zero on the surface A, and f is the area under the S-function curve along the segment of the crack-front under consideration.? The “equivalent domain integral” procedure is similar to the virtual crack extension method[7,8] in essence. However, there are two important differences between the two approaches: (i) first, the elimination of the concept of a virtual crack extension during the development of eq. (1) leads to the possibility of using any continuous s-function. For 2O-node elements it is reasonable to have a s-function that is quadratic, linear or even constant along the’ crack front within each crack-front segment shown in Fig. 1. (ii) Secondly, the a priori transformation of displacements and stresses to the crack front coordinate system simplifies the numerical procedure considerably. As a consequence of this, it is easy to have closed form expressions for the area under the s-function curve along the considered segment of the crack-front for any choice of s-function. The details of formulation and numerical implementation of the EDI-method can be found in [6]. CRACK

CONFIGURATIONS

AND FINITE

ELEMENT

MODELS

Semielliptical surface cracks in a tensile plate are considered as examples of defects in engineering structures. The geometric parameters of a plate with a crack are defined in Fig. 2. The crack of length 2c, and depth a is located in a plate of thickness t, width 2b. and height 2h. The aspect ratios c/a = 1S, 2 and 3 were adopted for the present analyses (the latter coincides with ASME maximum postulated flaw[l]), and in all cases the relative crack depth was a/t = 0.25 and the ratio c/b was 0.2. A typical finite element mesh consisting of 128 quadratic isoparametric elements and 743 nodes is shown in Fig. 3. Singular quarter-point degenerate elements were used in the vicinity of crack front. The three-dimensional mesh was created by moving a two-dimensional mesh along the elliptical front and copying this mesh at prescribed elliptical angles 8. Some geometric changes were affected in the mesh, after this procedure, depending on the shape of the plate. TEquation (1) can be further simplified by taking such an s which vanishes on Al + AZ. This eliminates the third term on the right hand side of eq. (I).

J-integral

calculations for semielliptical surface cracks

83

Fig. 2. Semielliptical surface crack in a plate subjected to uniform far-field stress.

Fig. 3. Finite element mesh used in surface crack analysis.

RESULTS AND APPROXIMATIONS The calculations were performed for all the above mentioned cases of semielliptical cracks. First, the elastic stress intensity factors along the crack front were computed, and some of them were compared to the Benchmark solutions[9]. The maximum absolute relative error was 3.5%. Then the elastic-plastic problems were solved for a linear hardening material (Fig. 4) with yield stress aY and two hardening coefficients Ep = 0 and Ep = O.lE.Tensile stress v was applied incrementally to the point of maximum load level CT= CT,,, with the size of the increment decreasing progressively. The results of the elastic-plastic J-integral calculations are presented in Fig. 5 for Ep = 0, and in Fig. 6 for Ep = 0.2E, in the form of an “equivalent” elastic-plastic stress intensity factor

84

G. P. NIKISHKOV

and S. N. ATLURI

Fig. 4. Postulated streess-strain

relation.

Kerr normalized by the elastic stress intensity factor K,, i.e.

The values of this ratio, as calculated from the present ED1 method, at the deepest point of the crack (point A in Fig. 2), are shown by circles in Figs 5 and 6. The dashed line in Fig. 5(a) relates to the two-dimensional finite element elastic-plastic J-integral calculations. It is evident

c/a =I 5

_----

/

/

/

1.6

c/a

14-

-3

E, 20

0.5

0.6

07

08

0.9

v IV,

Fig. 5. The equivalent elastic-plastic

stress intensity factor an elastic-perfectly

plastic material.

J-integral calculations for semielliptical surface cracks

I.4

-

85

E, =O.lE

r” $

t.a-

.*<

0.7

0.6

0.5

06

09

8 1 By.

I 6

I6

c/a

22

c/a

I4-

“3

Es-O,

E, = 0. I E

1E

Y I2-

‘, Y

06

0.5

I

I

I

I

06

0.7

0.8

0.9

0.6

05

I

I

I

I

06

07

06

09

o/oy

Fig. 6. The equivaient

elastic-plastic

stress intensity factor for a material with linear hardening E,=O.lE.

that two-dimensional J-integral estimation is too conservative. The soiid lines in Figs 5 and 6 represent the least square approximation of computed values, by the empirical relation: I( ep= K

J

$

e

= 1+ b[(o/a;)

- 0.71”

for cr > 0.75, for P 5 0.7~3,

K&P= K where the values of the coefficients

b and m are given in Table 1.

Table 1. Coefficients of approximation

E,IE

cla

0.1

m

2 3

742.2 241.0 146.0

6.241 5.293 4.596

1.5 2 3

173.8 52.39 53.28

5.374 4.313 4.240

1.5

0

b

(2)

G. P. NIKISHKOV and S. N. ATLURI

86

I

I

-

-Y \ -J

, 04-

o.21 0

t5

45

30

60

75

90

8

0

15

30

45

60

75

90

0

15

30

45

8

60

75

90

a

Fig. 7. J-integral variation along the front of semielliptical surface cracks.

5

a/t = 0.25 a=1 c =3

45 4

u = 0.9

35

I

Tension

U”

c/a=3 3 7

P

25

‘L

,--8

2 I 5

,

,

,

,

0.5

I

I.5

2 I

= 1.26”

( (gjrfrf;;*“,

,

,

I 05

I

2.5

a

3

35

4

4.5

5

55

6

65

7

75

6

I t

Fig. 8. Plastic zone contours for the crack with c/a = 3 at the load level w = 0.90;.

65

9

X

J-integral caiculations for semiell~p~cal surface cracks

87

The maximum absolute relative error of the approximation given in eq. (2) is 5%. It is interesting to note that the elastic-plastic value of the J-integral for a crack with c/a = 1.5, in an elastic-perfectly plastic material, at a far-field load level of (+= 0.5cr,, is only 6% less than its corresponding elastic value. This fact can be explained by the redistribution of stresses and the J-integral values along crack front due to plasticity. The comparison of the J-integral variation along the crack front for elastic and elastic-piastic materials at a foad bvel of cf = oY is shown in Fig. 7. It can be seen that the largest redistribution is for the crack with c/a = 1.5 and for the material with E,, = 0. The general tendency is that of a decrease of the relative values of the J-integral at the surface (0 = 0) during the increase of load level. Finally, Fig. 8 presents contours of the plastic zones at some planes normal to the crack front, for the flaw with c/ca = 3 at the load level D = O-9+ The planes are located at ellipticat angles f .3”, 9”, 27.4”, 62.4” and 82X?,“;then symmetrically to the minor axis of an ellipse. The plastic zone has the largest size at the surface. The minimum size of the plastic zone relates to the angle of 27.4”. The plastic zone at the deepest point of the crack is smaller than the plastic zone at the surface but larger than the plastic zone at the angle of 27.4.

Several semielliptical surface cracks in a tensile plate were analyzed by means of an elastic-plastic finite element method. The values of the elastic-plastic integral Je,, were computed by an “equivalent domain integral” method and compared with its elastic values J,. It is interesting that the ratio J,/J, can be less than one, for cracks with small c/a ratios at Iow load levels. The main conclusion is that it is possible to use the hnear eIastic fracture mechanics estimations up to a load level of o= 0.9a, for cracks with a/t = 0.25 and c/a 5 3. The maximum absolute relative error in the estimation of an equivalent elastic-plastic stress intensity factor KeP in this case is 10% for a crack with c/a = 3, in an elastic-perfectly plastic material; whife for an elastic-plastic material with Iinear hardening EP = 0.1 E, the error has the value of about 6%. The results of the present calculations were approximated by a simple empirical equation, which relates the values of elastic and elastic-plastic J-integral at the maximum depth point, for surface flaws of a semielliptical shape. Acknowledgements-The Hurstauthor acknowledges the financial support of the International Research & Exchange Board, New York. The partial support of the Oftice of Naval Research, and the encouragement of Drs Y. Rajapakse and Alan Kushner are gratefully acknowledged. It is a pleasure to thank MS Cindi Anderson for her assistance in the preparation af this transcript.

REFERENCES [iI ASME Boiler and Pressure Vessel Code, Section III, Appendix G. The American Society of Mechanical Engineers, New York (1974).

EZIJ. C. Newman, Jr and 1. S. Raju, Stress-intensity factor evaluations for cracks in three-dimensional

finite bodies subjected to tension and bending (NASA-TM-85783). Report N84-23925 (1984). r3.l M. Isida, H. Noguchi and T. Yosida, Tension and bending of finite thickness plate with a semielliptical surface crack. Int. .I. Fmcfure 26, 157-188 (1984). r41 V. Kumar, M. D. German and S. F. Shih, An engineering approach for elastic-plastic fracture analysis. EPRI Topical Report NP-1931, Palo Alto, California (1981). PI W. W. Wiikening, H. G. deI_oreti and M. Ba&iqmlsky~ E~asti~~l~tic an&ses of surfsee ffaws in a reactor YWX~. f. PressureVessel T&m& Htti, 247-254 (1984). Ed G. P. Nikisbk~v and S. N. Atluri, Calculation of fracture mechanics parameters for arbitrary three-dimensional crack by equivalent domain integral method. Int. J. number. Meth. Engng, in press (1987). I.71H. G. delorenzi, Energy release rate calculations by the finite element method. Engng Fracture Mech, 21,129-143 (1985). PI F. Z. Li, C. F. Shih and A. Needleman, A comparison of methods for calculating energy release rates. Engng Fmcmre Me&. 21,405-421 (1985). 181A Criticai Evaluation of Numerical Solutions to the “Benchmark” Surface Flaw Problem, Benchmark Editorial Committee (Edited by J. J. McGowan), SESA (1980). (Received 6 February 1987)