You are on page 1of 13

International Journal of Fracture 104: 131143, 2000. 2000 Kluwer Academic Publishers. Printed in the Netherlands.

Numerical Calculation of Stress Intensity Factors in Functionally Graded Materials


G. ANLAS, M. H. SANTARE and J. LAMBROS
Department of Mechanical Engineering, University of Delaware, Newark, DE 19716, U.S.A.; Permanent address: Bogazici University, Bebek, Istanbul, Turkey. Received 3 August 1999; accepted 7 January 2000 Abstract. The nite element method is studied for its use in cracked and uncracked plates made of functionally graded materials. The material property variation is discretized by assigning different homogeneous elastic properties to each element. Finite Element results are compared to existing analytical results and the effect of mesh size is discussed. Stress intensity factors are calculated for an edge-cracked plate using both the strain energy release rate and the J -contour integral. The contour dependence of J in an inhomogeneous material is discussed. is calculated and it is shown numerically that J, the strain energy An alternative, contour independent integral J release rate G, and the limit of J as approaches the crack tip (where is the contour of integration) are all approximately equal. A simple method, using a relatively coarse mesh, is introduced to calculate the stress intensity factors directly from classical J -integrals by obtaining lim 0 J . Key words: J -integrals, stress intensity factors, uncracked FGM plate.

1. Introduction There is a class of materials which have continuously varying mechanical properties. Although this type of material exists commonly in nature, there has been recent interest in manufacturing them for specic engineering applications. Such materials are called functionally graded materials because their properties have a spatial gradient. Functionally graded materials (FGMs) are very attractive for demanding applications such as thermal and wear protective coatings, yet their behavior needs to be better understood to fully exploit their characteristics. Analytical work on functionally graded materials goes back as early as the late 1960s when soil was modeled as a nonhomogeneous material by Gibson (1967). More recently, Delale and Erdogan (1983) analytically studied the crack problem in an innite plane where the elastic properties varied exponentially in the direction of the crack. They showed that the asymptotic cracktip stress eld possesses the same square root singularity seen in homogeneous materials. In 1987, Eischen studied the crack-tip-singular behavior of the stress eld in a nonhomogeneous innite plane by using an eigenfunction expansion technique. He veried that the leading term of the asymptotic expansion for stresses was square-root singular. This result was further conrmed by Jin and Noda (1994) for materials with piecewise differentiable property variations. In 1994, Konda and Erdogan studied the behavior of an innite cracked plane with exponential properties gradients in both the in-plane directions. Using a similar technique, Erdogan and Wu (1997) studied various far-eld loadings of an innite FGM strip. They used an exponentially varying Youngs Modulus, E , but kept the Poissons ratio, , constant. This approximation was based on the earlier work of Delale and Erdogan (1983) who showed that the effect of a variation of is negligible. The work by Erdogan and Wu (1997) is one of the

132 G. Anlas, N. H. Santare and J. Lambros few fracture solutions available for a nite width FGM. For this reason, their results are used as the basis of comparison in the current study. For a comprehensive review of fracture of functionally graded materials, see Erdogan (1985). In 1996, Jin and Batra studied the crack tip stress eld, strain energy release rate and stress intensity factors in a ceramic-metal FGM. In two separate studies, Gu and Asaro (1997a, b) analyzed fracture and crack deection in functionally gradient materials. In contrast to the above-described growing body of analytical studies, there are relatively few experimental and numerical investigations of fracture of functionally graded materials. In a recent publication, Gu et al. (1999) discussed a methodology for numerical determination of stress intensity factors in FGMs. They studied the effect of material nonhomogeneity in numerical computations of the J -integral. They concluded that the traditional version of the J -integral can provide accurate results provided that it is evaluated very close to the crack tip, using very small elements ( 105 b, where b is the crack length). Li et al. (2000), show an experimental-numerical hybrid method for evaluating the strain energy release rate in laboratory-scale FGMs. This method doesnt require a highly rened mesh but does require experimental data. Experimental studies have in general been hampered by difculties associated with material fabrication even though some new experimental techniques have been developed recently (see Lambros et al., 1999; Butcher et al., 1999; Parameswaran and Shukla, 1998). The focus of this paper is on the calculation and comparison of the stress intensity factors obtained for a cracked FGM plate by using the several different numerical techniques. However, in the next section we rst, briey discuss nite element solutions for uncracked FGM plates to validate the approximations and assumptions used. Subsequently, a technique similar to that of Gu et al. (1999) is used to evaluate the J -integral numerically. The strain energy release rate G and the J -integral are calculated for an edge-cracked FGM subject to far eld loading. A modied path independent integral, J similar to that proposed by Honein and Herrmann (1997) is computed for the nonhomogeneous case. The results are compared to the analytical solutions of Erdogan and Wu (1997). The relationship between the accuracy of the nite element method (FEM) and mesh renement is also investigated. To our knowledge, -integral this is the rst numerical implementation and assessment of the path independent J for FGMs. 2. Study of uncracked FGM plate The nite element method has been used extensively in solving problems involving homogeneous materials. The error introduced by geometric discretization of the domain, necessary in FEM, has also received close scrutiny. However, when modeling an inhomogeneous material (in this context, dened as one which possesses a continuous spatial variation of E and/or ) a material property discretization is introduced in addition to the geometric one. In this section, we investigate the accuracy of the FEM, when a material property discretization is introduced, by comparing FEM results to simple analytical solutions. (Note that it is possible to avoid material discretization by allowing a specic material property variation when formulating the stiffness matrix. This case is being investigated in a separate work.)

Numerical Calculation of Stress Intensity Factors 133

Figure 1. Edge-cracked plate under uniform traction or displacement loading.

2.0.1. Finite element solution The nite element solutions of a rectangular plate made of a functionally graded material under uniform traction or constant displacement loadings are obtained using ABAQUS (1997). Assume that the material gradient is in the x -direction, and is exponential according to (x) = 1 ex (1)

where is the shear modulus of the material, 1 the shear modulus at x = 0, and is a material constant which represents the length scale over which the properties change (the units for are 1/length). The Poissons ratio, , is taken to be constant. The normal stress values are calculated on the line of symmetry. The results evaluated on the line of symmetry correspond to the results of an innite strip, so they can be compared to those obtained analytically in the paper by Erdogan and Wu (1997). 2.1. U NIFORM
TRACTION LOADING

The plate considered has height twice its width, and is symmetric with respect to its midline, the y = 0 axis. The geometry of the problem is given in Figure 1 with b = 0. Using symmetry, only the upper half is considered in the nite element model. The upper edge is loaded by a uniform traction, yy = 0 . The lower edge has a zero displacement boundary condition

134 G. Anlas, N. H. Santare and J. Lambros


Table 1. Percent error in FEM results for uniform traction loading, 40 elements used in the x -direction. x/ h 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 E2 /E1 = 2 0.37 0.04 0.05 0.05 0.09 0.08 0.05 0.02 0.08 0.18 0.16 E2 /E1 = 5 1.31 0.14 0.08 0.15 0.36 0.47 0.43 0.18 0.28 0.8 3.18 E2 /E1 = 10 2.38 0.52 0.35 0.15 0.61 0.94 0.96 0.59 0.27 1.81 21.28 E2 /E1 = 20 3.86 1.24 0.87 0.06 0.84 1.55 1.76 1.35 0.04 3.65 50.07

in the y -direction to account for symmetry. The plate is divided into a uniform mesh with 40 elements along the x -direction and 10 elements along the y -direction, resulting in 400 elements (900 deg of freedom (dof)). The material is functionally graded, and the modulus of elasticity changes exponentially along the x -axis as, E(x) = E1 ex , where E1 is the elastic modulus at x = 0. From equation (1), it can be shown that E1 = 21 (1 + ). For the nite element solution, this change is modeled discretely by assigning each of the 40 regions along the x -axis the value of E at the centroid of the region, calculated according to the exponential relation given above. The Poissons ratio is taken as = 0.3. Note that the use of a uniform rectangular mesh makes the assignment of material properties quite straightforward. For a radially focused mesh, as is commonly used in fracture problems, the material property discretization may be more involved. A two dimensional continuum element with four nodes and four integration points is used for the plane strain problem. Stresses are calculated at the nodes of the 40 elements on y = 0 (the axis of symmetry). Since each row of elements has a discrete modulus, there will be signicantly different nodal stresses for each node shared by adjacent elements. The reported results for yy /0 are the numerical averages in these cases. The results for E2 /E1 = 2, 5, 10 and 20 are compared to the analytical results from Erdogan and Wu (1997) in Table 1 where E2 is the modulus of elasticity at x = h. For an assumed exponential gradient as in Equation (1), there are two independent parameters needed to establish the material property; either the modulus at two locations, as used here, or the modulus at one point and the length scale . The table shows that in this case, the error involved is less than 1% for most of the nodes. In general, the average error increases with increasing E2 /E1 . The errors at the end points x = 0 and x = h are larger due to the discretization of the material property in the element. There is no adjacent element to mitigate this discrepancy as there is at the other points (recall that the material property assignment is made at the centroid of the elements, not at the nodes). Table 2 shows the same comparison for a 100 by 25 element mesh. It is seen that these results are even closer to the analytical results.

Numerical Calculation of Stress Intensity Factors 135


Table 2. Percent error in FEM results for uniform traction loading, 100 elements used in the x -direction. x/ h 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 E2 /E1 = 2 0.10 0.04 0.03 0.05 0.04 0.08 0.09 0.01 0.04 0.09 0.04 E2 /E1 = 5 0.70 0.31 0.22 0.05 0.26 0.43 0.43 0.23 0.19 0.54 1.43 E2 /E1 = 10 1.35 0.76 0.55 0.05 0.48 0.82 0.93 0.63 0.13 1.39 10.40

2.2. U NIFORM

DISPLACEMENT LOADING

For the case of uniform displacement loading, the boundary condition at the top and the bottom of the plate is an applied constant displacement, V0 , in the y -direction. The numerical results for normalized stress, yy h(1 2 )/E1 V0 for uniform displacement loading are compared to analytical results from Erdogan and Wu (1997), and the percent error is tabulated in Table 3. The comparison of the FEM results to analytical results shows that the error is less than 0.1% in most of the cases, when 400 elements are used to discretize the domain. With the same discretization, the results of constant displacement loading are more accurate than the corresponding ones for uniform traction loading (compare with Table 1). One reason is that the analytical solution for the uniform traction problem used in the comparison, is exact only on the line of symmetry, y = 0. The displacement solution on the other hand, is exact everywhere in the domain. The comparison of analytical and FE results in Table 4 shows that with 100 elements used in the direction of property change, instead of the 40, FE results are almost identical to the analytical solution. 3. Study of Edge-cracked FGM Plate In this section, the problem of a nonhomogeneous nite plate with an edge crack is studied. The problem geometry is shown in Figure 1 for either a uniform traction or a uniform displacement applied in the y -direction. The domain is discretized into uniform meshes of 20 by 20 elements (about 880 dof), 100 by 25 elements (about 5250 dof) and 200 by 50 elements (about 20800 dof) respectively. A four-node, two dimensional plane strain element is used. Stress intensity factors are calculated using the strain energy release rate, G, the J -contour integral as r 0 (we will denote this limiting value as J (1)), and the contour integral J which will be dened subsequently. The numerically evaluated stress intensity factors are then compared to analytical results given by Erdogan and Wu (1997). The effect of mesh size, crack length, , and the relation among G, J (1), and J are discussed.

136 G. Anlas, N. H. Santare and J. Lambros


Table 3. Percent error in FEM results for constant displacement loading, 40 elements used in the x -direction. x/ h 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 E2 /E1 = 2 0.9 0 0 0 0 0 0 0 0 0 0.85 E2 /E1 = 5 2 0 0 0 0 0 0 0.03 0.03 0.02 2 E2 /E1 = 10 2.9 0.08 0.06 0.1 0.04 0.06 0.05 0.04 0.05 0.05 2.84 E2 /E1 = 20 3.8 0.07 0.05 0.04 0.06 0.09 0.07 0.07 0.04 0.08 3.7

Table 4. Percent error in FEM results for constant displacement loading, 100 elements used in the x -direction. x/ h 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 E2 /E1 = 2 0.30 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.35 E2 /E1 = 5 0.80 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.80 E2 /E1 = 10 1.20 0.08 0.00 0.00 0.00 0.03 0.01 0.02 0.00 0.01 1.14

In the general case, when a crack advances at a xed displacement, the strain energy release rate is dened as follows: G= dU dA (2)

where U is the strain energy, and A is the crack surface area. By using a node release technique in the nite element program, dU/dA can be approximated by U/ A. U is the change in strain energy as a result of an increment in crack growth, holding the external boundary conditions constant. The smaller the increment A ( A = t b, where t is the thickness of the plate), the more accurate this result will be. The stress intensity factor can be calculated using the following relation for plane strain (Jin and Batra, 1996),

Numerical Calculation of Stress Intensity Factors 137

Figure 2. J -integral vs. contour, b/ h = 0.2, uniform traction loading.

GEtip , (3) 1 2 where Etip is the Youngs Modulus at the location of the crack tip. For homogeneous materials, an alternative measure of the strain energy release rate is the J -integral, which is dened by Rice (1968), Kl2 = J = W n1 ij ni uj x1 ds, (4)

where in a two-dimensional cracked body, is any path beginning at the bottom crack face and ending on the top crack face. Here W is the strain energy density, uj are the displacement components and ni are the components of the outward unit normal to . In the absence of body forces, thermal strains and crack surface tractions, the J -integral is path independent for homogeneous materials. It is well known that in the case of homogeneous materials, G = J . For a nonhomogeneous material, in general, the J -integral is not path independent. In this study, the contour integral, J , is calculated using ABAQUS. The change of J with contour number is shown in Figure 2 for E2 /E1 = 2, 5, and 10. In this and subsequent gures, the contour numbers represent incrementally larger contours around the crack tip where the size and the increment is governed by the mesh renement. Each contour is placed symmetrically around the crack tip and includes an integer number of elements. For a nonhomogeneous material, G = J because J is path dependent. However, it can be shown that G = J as 0 (Gu et al., 1999). Tohgo et al. (1996) also show that in FGMs the J -integral is path dependent and its value for a path adjacent to the crack tip is identical to that for homogeneous materials. 4. J-integral In the case of nonhomogeneous materials, the strain energy density W is not only a function of (x), but it also depends on x explicitly, i.e., W = W ((x), x), due to material property gradients. This results in an extra term in the classical J -integral which needs to be subtracted as from Equation (4) to obtain the contour independent integral J

138 G. Anlas, N. H. Santare and J. Lambros = J W n1 ij ni uj x1 ds


A

W,1 q dA,

(5)

where is any path beginning at the bottom crack face and ending on the top crack face and A is the area enclosed by that contour. In Equation (5), W,1 denotes partial differentiation of W integral will yield a zero value on any closed with respect to the explicit x dependence. The J contour in a homogeneous as well as an inhomogeneous material, and therefore when used for fracture problems will always be path independent. This result was conrmed analytically in the work of Honein and Herrmann (1997). In the following, we will numerically implement the path-independent J and compare the calculated results with those derived through other methods. is constructed as suggested by The nite element formulation of the contour integral J Li et al. (1985). In the absence of thermal stresses, and crack-face tractions, the following discretized form is obtained from Equation (5): = J
N

(ij ui,1 W 1,i )q1,i W,1 q1 det


A p =1

xk l

p ,
p

(6)

where p are the weight functions of the corresponding Gauss integration points. The quantities in { } are all evaluated at the integration points for each element within the contour chosen. N is the number of integration points per element and q1 is a device used to facilitate the evaluation of a contour integral in nite elements. In general, a nodal value of 0 or 1 is assigned to Q1 and q1 can be dened within an element as follows:
4

q1 =
i =1

Ni Q1i ,

(7)

where Ni are the interpolation functions for the elements. In this study, for the calculation of q1 , a plateau function representation of q1 is used (see Li et al., 1985). Recall that W,1 in equation (12) is not the total derivative but rather the partial with respect to x1 which will only exist if the material property E is an explicit function of x1 , (e.g., E = E(x1 )). If the material property variation is given explicitly, it is simple to analytically derive an expression for W/x1 . In the case of equation (1), W/x1 = W (, x1 ) 5. Calculation of the Stress Intensity Factors We analyze edge-cracked strips with b/ h = 0.4, and E2 /E1 = 2.0 and E2 /E1 = 0.5 (refer to Fig. 1). The stress intensity factors, KI are calculated in three different manners; using the strain energy release rate G, the contour integral J , and the path independent contour integral as follows: J
2 KI (1 2 ) = J (1) lim J. =G=J 0 Etip

(8)

are plotted in Figs. 3 and 4 to show their variation with contour number. The graphs of J and J is contour Clearly, J is contour dependent as was seen previously in Fig. 2. In contrast, J independent after the rst two contours thus numerically verifying the path independence of Equation (5).

Numerical Calculation of Stress Intensity Factors 139

Figure 3. J and J vs. contour, b/ h = 0.4, E2 /E1 = 2.0, 20 by 20 mesh.

from the rst contour are disregarded since they are generally The values for J and J not considered accurate for most nite element meshes. Using the results of the J -integrals calculated from contour 2 to contour 20, and tting a fourth order polynomial to these data points, a value for J (1) can be obtained. Extending the curve to larger number of contours produces very little change. J (1) can be approximated numerically as the intercept of the polynomial curve t at n = 1, i.e. the crack tip. From Fig. 3 we can see that the limiting for a 200 value of J , called J (1), is approximately equal to the path independent integral J by 50 mesh. In addition, the stress intensity factors can be calculated from energy quantities using Equation (14). All results are normalized using the procedure outlined by Erdogan and Wu (1997) and compared to the exact results obtained by them. For uniform traction the I = KI /0 b where 0 is the applied traction and normalized stress intensity factor is K I = KI /0 b where 0 = b the crack length. For the case of applied displacement, 0 , K E1 V0 / h(1 2 ). , we have computed the normalized stress For b/ h = 0.4, using the 20 by 20 mesh and J intensity factor as 1.9244. The exact result calculated by Erdogan and Wu (1997) is 1.9573. Using the same mesh, G computed from U/ A is not reliable because the mesh is very (see Fig. 3 for a comparison of J and J (1)). When coarse. In addition, J (1) is not close to J the 100 by 25 mesh is used however, J and J (1) are almost equal, and G is 4% lower. In this case, the normalized stress intensity factors are 1.9313, 1.9324 and 1.898 respectively. When the even ner, 200 by 50 mesh is used, the normalized stress intensity factors are 1.9458, 1.9461 and 1.931, respectively. A similar computation is carried out for E2 /E1 = 0.5. The plots for, J and J (1) are given in Fig. 5 for a 200 by 50 mesh. Note that in this case J decreases with increasing contour number since now E2 /E1 < 1. The normalized stress intensity factor is 2.231 when calculated using J, and 2.229 when calculated using J (1). The exact result is 2.2598 (Erdogan and Wu, and the use of J (1) gives satisfactory 1997). The results show that, computationally J (1) J results in stress intensity factor calculations with relatively coarse meshes.

140 G. Anlas, N. H. Santare and J. Lambros

vs. contour, b/ h = 0.4, E2 /E1 = 2.0, 200 by 50 mesh. Figure 4. J and J

vs. contour, b/ h = 0.4, E2 /E 1 = 0.5, 200 by 50 mesh. Figure 5. J and J

6. Comparison of numerically calculated stress intensity factors is path independent. However, this quantity can be cumWe have shown numerically that J , which is bersome to compute. We have also shown that, J (1) is a good approximation to J in agreement with Gu et al. (1999) and Tohgo et al. (1996). In the previous section, we have also shown that the stress intensity factors can be approximately calculated using J (1) from a coarse, uniform mesh, when the element size is in the order of 102 b, where b is the crack length. In this section, we study the accuracy of J (1). To generate the numerical results shown in this section, we use a uniform mesh of 100 by 25, with 8-noded elements (15 500 dof). Note that although this mesh is more rened than those used in previous sections, it is still relatively coarse and does not focus on the crack tip. Figs. 6 and 7 show the normalized stress intensity factors for different crack lengths and E2 /E1 = 2.0 and E2 /E1 = 10 respectively. Numerical results are generated using G and J (1), where G is calculated using the node

Numerical Calculation of Stress Intensity Factors 141

Figure 6. Comparison of the normalized stress intensity factors for uniform traction loading, E2 /E1 = 2.0.

Figure 7. Comparison of the normalized stress intensity factors for uniform traction loading, E2 /E1 = 10.

release technique described in section3. The exact results are again taken from Erdogan and Wu (1997). Even with this relatively coarse mesh, we see that J (1) gives very good results. On average, the error increases with increasing . However, the interdependence of parameters on , b/ h, and mesh renement is rather complex. Ideally, we would like to conduct a full eld comparison between the numerical and analytical results. However, the analytical results published to date do not provide this detail. One local eld parameter we do have access to is the crack opening displacements from Erdogan and Wu (1997). The crack surface displacements for uniform traction loading are calculated for an edge crack with b/ h = 0.2 and plotted in Fig. 8 for E2 /E1 = ratios of 2.0, 5.0, and 10.0. These results are in very good agreement with the exact results presented by Erdogan and Wu (1997).

142 G. Anlas, N. H. Santare and J. Lambros

Figure 8. Crack surface displacement for b/ h V (s) = E1 (x, 0)/[2h0 (1 ) ].

0.2. Edge crack under uniform traction loading,

7. Concluding remarks In this study, stresses in the uncracked plate are calculated under both uniform traction and uniform displacement loadings. The results are compared to exact solutions in Tables 14. Traditional nite elements give fairly accurate results for the uncracked case with the assignment of properties at the centroid of each element. In addition, we have used the J contour integral results of ABAQUS in the calculation of stress intensity factors for an edge cracked plate made of a functionally graded material. We , a modied have numerically demonstrated the path independence of the contour integral J J -integral in Equation (11), and shown that J = J (1) where J (1) = lim 0 J . We compared our results for normalized stress intensity factors to the analytical results presented by Erdogan and Wu (1997). This comparison showed that even using a relatively coarse, uniform mesh, and J (1) are very close to the analytical ones. However, for the results obtained from J the same mesh, the normalized stress intensity factor results obtained using the node release technique for G are far less accurate. Clearly, the accuracy of the node release technique can be improved with mesh renement. Nevertheless, signicant mesh renement is not needed to obtain accuracy using J (1), a quantity which can be calculated using many existing nite element codes. The method presented here is quite simple and easy to implement compared with the analytical and other computational methods. Acknowledgement The authors would like to thank NSF for support of this research through grant CMS-9712831. References
ABAQUS, Theory Manual, Hibbitt Karlsson and Sorensen, Version 5.7, 1997. Butcher, R.J., Rousseau, C.E. and Tippur, H.V. (1999). A functionally graded particulate composite: preparation, measurements and failure analysis. Acta Materialia 47(1), 259268.

Numerical Calculation of Stress Intensity Factors 143


Delale, F. and Erdogan, F. (1983). The crack problem for a nonhomogeneous plane. Journal of Applied Mechanics 50, 609614. Eischen, J.W. (1987). Fracture of nonhomogeneous materials. International Journal of Fracture 34, 322. Erdogan, F. (1995). Fracture mechanics of functionally gradient materials. Composites Engineering 5(7), 753770. Erdogan, F. and Wu, B.H. (1997). The surface crack problem for a plate with functionally graded properties. Journal of Applied Mechanics 64, 449456. Gibson, R.E. (1967). Some results concerning displacements and stresses in a nonhomogeneous elastic half space. Geotechnique 17, 5867. Gu, P. and Asaro, R.J. (1997a). Cracks in functionally graded materials. International Journal of Solids and Structures 34(1), 17. Gu, P. and Asaro, R.J. (1997b). Crack deection in functionally graded materials. International Journal of Solids and Structures 34(24), 30853098. Gu, P., Dao, M. and Asaro, R.J. (1999). A simplied method For calculating the crack tip eld of functionally graded materials using the domain integral. Journal of Applied Mechanics 66, 101108. Honein, T. and Herrmann, G. (1997). Conservation laws in nonhomogeneous plane elastostatics. Journal of Mechanics Physics Solids 45(5), 789805. Jin, Z.H. and Batra, R.C. (1996). Some basic fracture mechanics concepts in functionally gradient materials. Journal of Mechanics Physics Solids 44(8), 12211235. Jin, Z.H. and Noda, N. (1994). Crack tip singular elds in nonhomogeneous materials. Journal of Applied Mechanics 61, 738740. Konda, N. and Erdogan, F. (1994). The mixed mode crack problem in a nonhomogeneous elastic plane. Engineering Fracture Mechanics 47, 533545. Lambros, J., Narayanaswamy, A., Santare, M.H. and Anlas, G. (1999). Manufacture and testing of a model functionally graded material. Journal of Engineering Materials and Technology 121(4), 488493. Li, H., Lambros, J., Cheeseman, B.A., Santare, M.H. (2000). Experimental investigation of the quasi-static fracture of a functionally graded material. International Journal of Solids and Structures (to appear). Li, F.Z., Shih, C.F. and Needleman, A. (1985). A comparison of methods for calculating energy release rates. Engineering Fracture Mechanics 21(2), 405421. Parameswaran, V. and Shukla, A. (1998). Dynamic fracture of a functionally gradient material having discrete property variation. Journal of Materials Science 33, 33033311. Rice, J.R. (1968). A path independent integral and the approximate analysis of strain concentration by notches and cracks. Journal of Applied Mechanics 35, 379386. Tohgo, K., Sakaguchi, M. and Ishii, H. (1996). Applicability of fracture mechanics in strength evaluation of functionally graded materials. JSME International Journal Series A 39(4), 479488.

You might also like