Engineering Fracture Mechanics 68 (2001) 14171432

www.elsevier.com/locate/engfracmech

Fracture testing and nite element modeling of pure titanium

Department of Civil and Environmental Engineering, University of Illinois, Urbana, IL 61801, USA Department of Chemical Engineering and Materials Science, University of California, One Shields Avenue, Davis, CA 95616, USA c Department of Civil and Environmental Engineering, University of California, One Shields Avenue, Davis, CA 95616, USA Received 5 March 1999; received in revised form 11 January 2001; accepted 18 January 2001

Abstract This paper presents the results of fracture experiments and corresponding nite element analyses (FEA) of pure titanium. This investigation was motivated by the desire to develop a JR testing protocol and numerical procedures that are applicable to a titanium/titanium boride layered functionally graded material. Tensile tests and a twodimensional axisymmetric nite element model were used to determine the plasticity data for the titanium. Crack growth experiments were conducted in three-point bending using single edge notched bend specimens. Three-dimensional FEA of crack growth initiation and two-dimensional FEA with automatic crack propagation were performed. Two crack propagation conditions based on experimental data were used: (a) crack length versus load-line displacement and (b) crack length versus crack mouth opening displacement. The subsequent predictions of the non-linear nite element models are in reasonable agreement with the measured value of J at initiation and with the rising JR data during crack propagation. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Titanium; Elastoplastic fracture mechanics; Crack resistance curves; J R; Crack mouth opening displacement; Finite element analysis

1. Introduction In this paper, we present the results of a study that integrates experimental measurements with nite element modeling to describe crack growth initiation toughness and R-curve fracture behavior based on J in commercially pure titanium (CP Ti). This work was motivated by an interest in combining measurements with modeling to characterize the fracture resistance of a functionally graded material (FGM) based on the titanium/titanium boride (Ti/TiB) system [1]. Functionally graded materials are expected to exhibit signicant R-curve behavior when one component is a ductile metal [2], although this has not been veried experimentally. For that reason, we wished to incorporate plasticity into a nite element model that describes the interaction of the crack tip stress and strain elds with the material property gradient. Thus, the primary goal of the present study was to develop numerical techniques to model crack tip stresses and strains in a manner that is useful for crack growth initiation and propagation in a FGM. In addition, pure
*

Corresponding author. Tel.: +1-530-752-0400/7037; fax: +1-530-752-1031/9554. E-mail address: [email protected] (J.C. Gibeling).

0013-7944/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 0 1 3 - 7 9 4 4 ( 0 1 ) 0 0 0 1 8 - 2

1418

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

titanium was used as a model material to enable the nite element method (FEM) results to be calibrated to experimentally measured parameters. It was necessary to measure the fracture properties of CP Ti since no relevant experimental data were available in the literature. Titanium exhibits considerable plasticity, hence the JR response was measured rather than KR behavior that characterizes more brittle solids [3,4]. In addition, it was necessary to determine the true stressstrain behavior of pure Titanium at strains that are characteristic of the fracture deformation process at the crack tip. These data were incorporated into a FEM model, with the goal of developing predictive capabilities regarding the fracture behavior of Titanium. A number of alternatives exist for simulating crack propagation in nite element analysis (FEA) using discrete [57] rather than distributed cracking [8]. For instance, Wawrzynek and Ingraea [6] have developed a general framework for crack propagation by means of automatic remeshing. Lim et al. [7] have also employed a remeshing technique for crack evolution in an elastoplastic medium. Recently, Trdeg et al. [5] have used the code a ard A B A Q U S to model crack propagation in elastoplastic materials by means of a combination of remeshing and nodal relaxation. In this work, we also use the capabilities of A B A Q U S to simulate crack growth initiation and stable propagation. The latter is accomplished here by means of a nodal release technique. The remainder of this paper is organized as follows. First, the experimental procedures are described, including material preparation, the waisted tensile test method to determine the plastic deformation behavior of CP Ti, and the JR testing. Next the computational approach, using the FEM, is described. In particular, a three-dimensional (3-D) model of crack growth initiation and an automatic two-dimensional (2-D) model of stable crack propagation are presented. Subsequently, the JR results of these FEM models are compared with the measured results from fracture tests of the same material. Finally, conclusions are drawn and potential extension of this work is discussed. 2. Experimental methods 2.1. Material preparation Samples for mechanical testing were machined from a rolled Grade 2 CP Ti plate 13.3 mm in thickness. The nominal composition of this material is 99% titanium. Cylindrical tensile specimens with a gage diameter of 6.35 mm and a gage length of 25.4 mm were turned from bars cut from the plate with the longitudinal axes of the specimens aligned with the rolling direction. Initial tensile tests of these specimens were conducted at a displacement rate of 2:54 102 mm/s. The resulting fracture surfaces revealed an elliptical cross-section, with major and minor diameters of approximately 5.15 and 4 mm, aligned with the longer axis normal to the thickness direction of the original rolled plate. This observation demonstrated that deformation was signicantly anisotropic as a consequence of the strong grain orientation due to rolling. In order to observe the grain structure, samples of material were rst ground using silicon carbide paper to 600 grit. They were then polished with 6 and 3 lm diamond on nylon and nal polished on a Buehler Vibramet with colloidal silica on ocked cloth. Stain etching with 5% ammonium biouride was found to be better than Kroll's reagent for revealing the grain structure in optical microscopy. The elongated grains present in the rolled plate are illustrated in Fig. 1(a). Because an isotropic material was desired to facilitate the fracture studies, heat treatment was employed to eliminate the texture introduced by the manufacturing process. In addition, large grain sizes were desired to gain knowledge of the fracture behavior of large grained titanium for future work on the Ti/TiB FGM system. Isotropic properties were obtained after heating for one hour at 1223 K in a vacuum of 4 103 Pa followed by cooling to room temperature over a 12 h period. As a result of the annealing heat treatment, the nominal grain size of the material increased 13-fold from about 30 to 400 lm as illustrated in Fig. 1(b).

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

1419

Fig. 1. Optical micrographs of CP Ti in the (a) as-received condition and (b) following heat treatment. Table 1 Average mechanical properties of grade 2 CP titanium Material Grade 2 CP Ti, heat treated (experimental) Grade 2 CP Ti [11] E (GPa) 106.9 103.4 Yield strength (MPa) 358 280 UTS (MPa) 500 340 % Uniform elongation 11

Correspondingly, the hardness decreased from 87 to 82 on the Rockwell B scale. The fracture surfaces of the annealed specimens exhibited circular cross-sections. The resulting mechanical properties for this CP Ti are summarized in Table 1 along with reference values for Grade 2 titanium obtained from the literature [11]. 2.2. Waisted tensile test method FEM models of crack growth require as input the plastic deformation characteristics at strains beyond the normal limit of uniform tensile deformation under conditions of triaxial loading. In order to determine the true stresstrue strain properties of CP Ti, slightly waisted tensile specimens were machined to the dimensions shown in Fig. 2. Three specimens were heat treated as described previously and tensile tested in

1420

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

Fig. 2. Waisted tensile specimen used to determine the stressstrain behavior of CP Ti. All dimensions are in millimeters.

uniaxial tension at a rate of 2:54 102 mm/s using a MTS 632.11B-21 extensometer with a 25.4 mm gage length to measure axial strain and a MTS 632.18E-20 extensometer to measure diametral strain. The waist in the gage section of the specimen permits proper positioning of the diametral extensometer at the neck and monitoring of diametral contraction during the test up to the point of ductile fracture. The axial extensometer was positioned to span the waisted section. 2.3. 2-D axisymmetric waisted tensile FEM model The true stresstrue strain curve for the CP Ti at large strains and constrained ow cannot be measured directly. Rather, these characteristics were determined using a 2-D axisymmetric FEA of the waisted specimen shown in Fig. 2. This model accounts for the symmetry at the mid-length of the specimen. Fig. 3(a) illustrates the corresponding loading and boundary conditions imposed for these calculations. Fig. 3(b) shows the nite element mesh with 363 nodes and 320 four-noded (Q4) elements. The mesh is rened toward the bottom of the model because of high stress and strain concentrations in the area of the waist. The software P A T R A N [9] was used for pre- and post-processing, and A B A Q U S [10] was used for the analysis. The material model used in all nite element calculations reported here is for an elasticplastic power law hardening material obeying the von Mises ow criterion with associated ow rule and isotropic hardening. An iterative procedure was used to obtain the specic parameters to represent the material behavior. The experimental true stresstrue strain data from the diametral strain data for one test (Fig. 4(a)) were used as the rst input for the material model. The nite element analysis of the waisted tensile specimen was performed to obtain the corresponding engineering stressnominal axial engineering strain behavior. Here, the nominal axial engineering strain for the model was evaluated directly from the output of the axial extensometer. Similarly, the engineering stress was determined by dividing the load by the initial crosssectional area in the waisted section. The experimental and FEA engineering stressnominal axial engineering strain results were then compared (Fig. 4(b)). If the two curves did not match after an iteration, the experimental true stresstrue strain curve was shifted vertically until the experimental and corresponding FEA results for engineering stressnominal axial engineering strain were in agreement. After about 25 iterations, the output shown in Fig. 4 was obtained. The ow behavior determined from this model was used in all subsequent analyses along with a Poisson ratio of m 0:34 [12]. Specically, the results shown in Fig. 4(a) labeled ``FEA (nal)'' were used as the material model for the crack growth analysis by entering the experimental data points individually into the FEA software. We note that this material model describes constrained deformation at strains beyond the normal UTS (in the waist of the tensile specimen). Internally, A B A Q U S represents the material behavior using a standard rate-independent J2 plasticity model accounting for nite strains. 2.4. JR test method The titanium plate was cut into single-edge notched bend (SEN(B)) specimens 13:3 25:53 101:6 mm3 with a starting notch and integral knife edges in accord with ASTM E399-90 [13] using electro-discharge

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

1421

Fig. 3. Tensile test model for Ti (a) loading and boundary conditions imposed in the model, (b) 2-D nite element mesh considering axisymmetry. All dimensions are in millimeters.

machining (EDM). Fig. 5 shows the conguration and orientation with respect to the original plate of the SEN(B) specimens tested in this investigation. The span (88.84 mm) to width (25.53 mm) ratio for this geometry is 3.5, which is less than the recommended ratio of 4.0. Prior to testing, the SEN(B) specimens were heat treated as described previously and cooled overnight to room temperature. Each specimen was pre-cracked in three-point bending using a MTS 810 servo-hydraulic testing machine equipped with a 100 kN model 661.20 force transducer and a MTS 632.03E-30 crack opening displacement gage. Loads were transmitted to the specimen using a bend xture equipped with 12.7 mm diameter roller supports. The precrack was grown in fatigue from an initial machined notch length of 0.4 W to a nal crack length of 0.5 W using the MTS 790.40 TestStar II fatigue crack growth test application. The crack was grown at 4 Hz with a load ratio of 0.1 and a nal Kmax of 17 MPa m1=2 in accordance with ASTM E1820-96 [14]. The crack length was monitored during pre-cracking by the compliance method based on crack mouth opening displacement. Four independent JR tests were conducted in three point bending according to ASTM E1820-96 [14] using the same testing arrangement as for pre-cracking. The specimens were installed in the load frame and cycled 20 times at 2 Hz between 0.45 and 1.98 kN to seat the clip gage. The specimens were then loaded from 0.45 to 4.45 kN at a rate of 0.45 kN/s. In order to determine the initial crack length, the force versus crack mouth opening displacement (CMOD) compliance was then measured three times by reducing the applied load by 1.78 kN and reloading with the testing system operating in load control. Subsequently, the CMOD was increased by 0.0323 mm using CMOD control and the compliance was again measured by the above procedure. After each such increment, the specimens were held for 30 s to allow for crack extension to stabilize, followed by three unloading compliance measurements. The tests were terminated after 3050 CMOD increments. The crack lengths after testing were then measured by compliance to be about 0.56 W. The specimens were post-test fatigue cracked using a load ratio of 0.5 and a Kmax of 24 MPa m1=2 until the cracks were 0.7 W in length and then broken by application of a monotonic load.

1422

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

Fig. 4. Results of experiments and iterative nite element analysis to obtain plasticity data for CP Ti (a) true stress based on instantaneous area in the waist versus true strain determined from the diametral extensometer and (b) engineering stress based on initial cross-sectional area in the waist versus nominal axial engineering strain measured by the axial extensometer.

Fig. 5. Single edge-notched specimen for three-point bend testing. The specimen thickness is 13.3 mm. All dimensions are in millimeters.

Crack lengths were computed from the unloading compliance data using the analysis procedures and compliance calibration equations described in the ASTM standard [14]. The three crack lengths determined from the three unloading compliance measurements for each increment of crack extension were averaged. The initial and nal crack lengths were measured on the fracture surface in accordance with the standard

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

1423

using a nine-point average through the thickness and found to be in agreement within 0.005 W with the compliance measured crack lengths. The initial crack front proles met the straightness criteria of the standard, although the nal crack exhibited slightly greater curvature (15% deviation) at the edges than is permissible (5% deviation). In addition, JR and dR curves were determined from the load and load-line displacement data using the procedures described in the standard [14], where d is the crack tip opening displacement. We note that the load-line displacement data measured using the test machine LVDT were corrected for machine compliance and surface indentation of the specimen at the loading supports. These correction factors were assessed from the load versus load-line displacement behavior of a sandwich specimen composed of CP Ti layers with a sti steel core tested at loads similar to those encountered in the JR tests. 3. Computational approach As noted above, the experiments were conducted in three-point bending using standard SEN(B) specimens. The computational approach involves detailed nite element modeling of the cracked geometry and the calculations have been performed using nite strain assumptions. Thus, it is possible to determine the stress and plastic strain elds in the vicinity of the crack front and their evolution with loading. Initially, crack growth initiation was studied by means of a 3-D FEM model with a stationary crack. Subsequently, a 2-D FEM model for automatic crack propagation was developed. Modeling of the crack tip region for both the 2-D and 3-D congurations is discussed in detail later in this paper. These details can have a signicant inuence on the nite element analysis results. The software P A T R A N [9] was used for pre- and postprocessing, and A B A Q U S [10] was used for the FEA. 3.1. 3-D model of crack growth initiation The three-dimensional nite element model of the SEN(B) test specimen (Fig. 5), takes into account two planes of symmetry, one at mid-length and the other at mid-thickness of the beam, as illustrated in Fig. 6. The corresponding nite element mesh is shown in Fig. 7 and contains 9779 nodes and 8200 eight-noded brick (B8) elements. The elements were integrated using reduced integration with hourglass control. Fig. 7(a) and (b) illustrates that the mesh is rened toward the crack front because of high stress and strain gradients in that region. We note that there are 10 layers of elements along the thickness direction (Fig. 7(c)), with discretization renement as the free surface is approached. From a qualitative point of view, this selective renement was employed because the free-surface approaches a state of plane stress while the middle surface approaches a state of plane strain. An expanded front view of the crack tip region, which was modeled by blunting the sharp tip, is shown in Fig. 8 (cf. Fig. 7(b)). In an attempt to correlate the mesh with the actual geometry of the pre-cracked beam, the blunting has been performed by means of a quarter circle notch of radius 0.005 mm. Guidelines for selecting an appropriate notch radius can be found in reference [10]. Twentyve contours for J-integral calculations (see Fig. 8) were investigated during the incrementaliterative FEA such that rings of elements were dened recursively to surround all previous contours. Here, the equivalent domain integral representation was employed to calculate J since this is the most accurate method for FEA [10,1517]. Theoretically, the J-integral should be independent of the domain. However, J estimates from dierent rings may vary because of the approximate nature of the nite element solution. Thus, a sucient number of contours away from and surrounding the crack tip were necessary to obtain a converged value of J in each increment. In the present study, the J calculations were found to converge by the 20th contour. However, to ensure accurate results, J was always evaluated on the 25th contour, which is suciently far from the crack tip.

1424

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

Fig. 6. Loading and boundary conditions imposed in the CP Ti crack growth model: (a) front view, (b) side view.

The FEA was performed under displacement control until the condition for crack growth initiation was reached. Specically, when the calculated J reached the value of JQ obtained experimentally, the analysis was stopped, and the numerical values of load, load-line displacement and CMOD were compared with the experimental ones. No additional calibration of this model to the experimental data was performed. 3.2. 2-D model of crack propagation Although an approximate representation of the JR behavior of CP Ti could be obtained from the static 3-D model by solving for progressively greater crack lengths, a more thorough treatment must be based on an FEA involving crack propagation. We consider again the edge crack in a CP Ti three-point bend specimen, as illustrated by Fig. 5. In order to describe the stresses, strains and displacements associated with a propagating crack, the notch region was modeled according to the actual geometry of the specimen (compare Figs. 5 and 9) rather than blunting the sharp tip as in the previous 3-D model (Fig. 8). Further, plane strain and nite deformation theory were used in the automatic crack propagation analyses. Due to symmetry, only one-half of the specimen was analyzed. The mesh is composed of 3872 fournoded quadrilateral (Q4) elements, 4019 nodes, plus 4 reference points as illustrated by Fig. 9. The fourth auxiliary reference point (not shown in Fig. 9) is located near the crack tip and used to measure the crack extension. The loading and support points are simulated by analytical rigid surfaces (cf. Fig. 9) in an at-

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

1425

Fig. 7. 3-D FEA model for crack growth initiation considering two planes of symmetry: (a) isometric view, (b) front view and (c) side view.

tempt to resemble the actual conditions in the experiment. The maximum allowable shear stress on the roller support is smax 100 MPa. The purpose of imposing a maximum allowable shear stress at the roller supports is to model essentially rigid boundary conditions in a realistic manner. However, the simulations are insensitive to the exact value chosen unless that value is low enough to permit sliding at the interface. A reasonably ne mesh, necessary to obtain an acceptably smooth load versus crack length relation, is used to model the area in which the plastic zone develops and crack propagation occurs. We note that, for the present problem, the crack path is known in advance, and a ne mesh discretization over the entire region where the crack is assumed to propagate is used (Fig. 9). Initially, a crack length versus load-line displacement condition was used for crack propagation. Because calculations in A B A Q U S are carried out in a time-stepped manner, it was necessary to equate load-line displacement with a pseudo-time for purposes of the FEA calculations. For each increment of load-line displacement (or pseudo-time), the solution was allowed to converge. Essentially, this rst simulation was

1426

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

Fig. 8. Expanded view of crack tip region for the 3-D FEA model shown in Fig. 6 illustrating the quarter-circle notch of radius 0.005 mm.

Fig. 9. Basic mesh for 2-D CP Ti crack propagation nite element model. The loading and the support points are simulated by rigid surfaces and three of the four reference points are indicated by the marks. The fourth reference point is located adjacent to the initial crack tip. All dimensions are in millimeters.

performed to verify the experimental results. Subsequently, a crack length versus CMOD condition was used and compared with the results of the rst method. We note that this second analysis used the crack length versus CMOD results obtained from the model based on the crack length versus load-line displacement simulation as input data. The specimen was subjected to bending loads such that, initially, a well-contained plastic zone developed for the stationary crack. Afterwards, the crack was allowed to propagate. Thus, the FEA was mainly conducted in two phases. The rst phase consisted of pushing the rigid surface 1.09 mm downwards, which was the load-line displacement corresponding to the measured JQ . No crack growth occurred during this stage. In the second phase, the crack was allowed to propagate by means of nodal release along the known crack propagation path. Nodes were released one by one to enable the crack tip to propagate through the mesh. When the fracture condition was met at a node, requiring an increment in crack length for either the current load-line displacement or CMOD, the force at the crack tip node was ramped down according to

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

1427

Fig. 10. Force ramp down curve following release at a node in the 2-D crack propagation model.

the solid curve shown in Fig. 10. The manner in which the forces at the released nodes are ramped down can inuence the convergence of the solution. In the present study, a fth order ramp down has been adopted because it leads to a more simulation ecient solution (fewer iterations) than a linear ramp (dashed line). 4. Results and discussion The experimental results for one JR test are illustrated in Fig. 11(a) and the corresponding data for crack tip opening displacement (dR) are shown on Fig. 11(b). Other tests gave qualitatively similar results. All tests clearly demonstrated the rising JR response expected for this ductile metal. From these results, preliminary values of JQ for the annealed pure titanium were calculated from the JR data and were found to be in the range of 153 to 254 N/mm. We attribute this substantial variability to the coarse grain size of this material (see Fig. 1). Small samples were chosen for this testing due to the limited size of the Ti/TiB

Fig. 11. Results of crack propagation experiments; (a) JR curve showing measured JQ value of 153 N/mm and (b) dR curve showing measured dQ value of 0.170 mm. The construction lines on the plots are dened in ASTM Standard 1820-96 [14].

1428

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

functionally graded plate [1], which was to be tested using the same xtures and protocols. The data from this set of experiments on CP Ti were needed to calibrate the elastoplastic nite element model. ASTM E1820-96 requires the specimen size be such that the specimen thickness, B > 25JQ =rY giving a minimum requirement of B > 28:3 mm. However, the samples used for these tests were of thickness B 13:3 mm, therefore, this specimen geometry is not of sucient size to fully meet the ASTM criteria for valid JIC tests of this material. Nevertheless, the experimental results can be directly compared to the results of the computational models, which were constructed for the same specimen dimensions. Due to the rather large scatter in the experimental data, use of an average J value would unnecessarily complicate comparison between model and experiments. Thus, in this study, the data for the specimen with the lowest value of JQ was used for comparison with the model since this represents the most conservative value. The results of the 3-D FEA for crack growth initiation are presented in Fig. 12 along with appropriate comparison to experimental results. As noted earlier, the calculations were carried out until J JQ corresponding to crack growth initiation. Fig. 12(a)(c) compares experimental and FEA results for load versus CMOD, J versus load, and load versus load-line displacement, respectively. Fig. 12(d) shows the experimental results for three-point bending and FEA results for both three-point and four-point bending. We note the approximately linear portion of the FEA data, which is essentially the same either for the three-point or four-point congurations. In all cases illustrated in Fig. 12, the experimental and numerical results are in reasonable agreement and show similar trends. However, Fig. 12(d) indicates that the calculated J values are greater than the experimental values at large CMOD. This deviation arises because the model does not account for crack blunting, which contributes to the experimental CMOD. Moreover, the size of the plastic zone at advanced stages of loading also aects the results.

Fig. 12. 3-D FEA results for crack growth initiation: (a) load versus crack mouth opening displacement, (b) J-integral versus load, (c) load versus load-line displacement, (d) J-integral versus crack mouth opening displacement.

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

1429

Fig. 13. Results of 2-D FEA of crack propagation based on prescribed crack length versus load-line displacement (FEA LLD) and CMOD (FEA CMOD) simulations illustrating: (a) crack extension versus load-line displacement, (b) load versus load-line displacement, and (c) variation of the reaction force at the rigid loading support as a function of the cumulative crack length (detail).

Results from the 2-D crack propagation FEA model are presented in Fig. 13(a) as accumulated incremental crack extension versus load-line displacement. The increments associated with each nodal release event are apparent in the FEA results. In order to facilitate comparison of the model with the experiments, the measured crack extension versus load-line displacement data were smoothed by tting the points beyond the blunting region with a second order polynomial as shown in Fig. 13(a) (labeled ``best-t'' in the gure). This best-t quadratic was then shifted vertically to a crack extension value of zero (labeled ``oset t''). This shift was necessary because the present technique for nodal release in the crack propagation model does not permit the initial crack tip blunting deformation to be accurately simulated. Finally, crack growth was modeled by requiring that the FEA analysis match the experimental crack length versus loadline displacement data, and the resulting output is also plotted in Fig. 13(a). In a similar manner, the second analysis was completed by demanding that the FEA analysis match the crack length versus CMOD results. In both cases, the crack extension values are associated with true physical propagation and not blunting. Fig. 13(b) illustrates the variation of the reaction force as a function of the displacement at the rigid body reference node. Fig. 13(a) and (b) illustrate results for both the crack length versus load-line displacement (LLD) and the crack mouth opening displacement conditions. Fig. 13(c) shows the reaction force as a function of the accumulated incremental crack extension. These gures reveal no dierences between the FEA results for the crack length versus load-line displacement condition compared to the crack length versus CMOD condition as expected.

1430

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

Fig. 14. Resistance curves: (a) J-integral versus post-blunting crack extension; (b) crack tip opening displacement (d) versus postblunting crack extension.

The rising portions of the JR curve for the experimental and computed data are compared in Fig. 14(a). Similarly, Fig. 14(b) shows dR curves obtained from experiments and numerical simulations. Again, in both plots, the experimental data have been shifted to exclude the apparent crack extension due to blunting, which is not included in the simulation. Both experimental and FEA crack tip opening displacements (d) in Fig. 14(b) have been obtained indirectly. The experimental d values are based on equations from the ASTM standard [14], while the d in the FEA was calculated by means of the equation d dn J =ryield , where ryield is given in Table 1. The constant dn % 0:44 was obtained by matching the experimental crack growth initiation point (post-blunting). Previously, this equation has been applied by Shih [18] to small scale yielding problems, who also showed that dn was in the range 0.40.8 for structural steels. Related considerations of nite deformation have been discussed by McMeeking and Parks [19]. Direct FEA calculations of d considering nite deformation eects and special singular elements at the crack tip have been presented by Dodds [20]. However, special elements were not used in the present FEA simulations. It is worth noting that the experimental and calculated dR curves in Fig. 14(b) can be compared only on a qualitative basis because of inherent dierences between the ASTM equations (based on beam theory) and the FEA (continuum based on nite deformation). Finally, Fig. 15 shows the normal stress contour plot at the last increment of the FEA for the crack length versus load-line displacement condition. This plot illustrates the stress levels in the domain and reects the way in which the crack tip region is modeled for the crack propagation simulations.

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

1431

Fig. 15. Normal stress distribution at last increment of 2-D crack propagation model (stresses are in MPa).

5. Conclusions and extension This study has integrated fracture testing with nite element modeling to describe crack growth initiation and R-curve behavior in CP Ti. The experiments were conducted in three-point bending using subsize SEN(B) specimens. A 3-D FEM model was employed for crack growth initiation and a 2-D FEM model was used for automatic crack propagation using a nodal release technique. In general, the predictions of the non-linear nite element model are in good agreement with the corresponding experimental results. In particular, we have developed the methodology to model stable crack propagation with rising R-curve behavior in an elastoplastic solid. This work constitutes a learning experience which will be helpful when dealing with more sophisticated materials such as FGMs. Thus we will extend the JR testing protocol and numerical procedures to investigate Ti/TiB layered functionally graded beams under three-point bending. The present results for CP Ti also provide useful information regarding the mechanical properties of one layer of the Ti/TiB FGM beam. This challenging extension is currently under investigation [1]. Acknowledgements We gratefully acknowledge support of the research described in this paper from the National Science Foundation (NSF) through grant no. CMS-9713798 (Mechanics and Materials Program). References
[1] Carpenter RD, Liang WW, Paulino GH, Gibeling JC, Munir ZA. Fracture testing and analysis of a layered functionally graded Ti/TiB beam in 3-point bending. Mater Sci Forum 1999;308311:83742. [2] Jin ZH, Batra RC. Some basic fracture mechanics concepts in functionally graded materials. J Mech Phys Solids 1996;44:122135.

1432

G.H. Paulino et al. / Engineering Fracture Mechanics 68 (2001) 14171432

[3] Atkins AG, Chen Z, Cotterell B. The essential work of fracture and JR curves for the double cantilever beam specimen: an examination of elastoplastic crack propagation. Proc R Soc London A 1988;454:81533. [4] Guinea GV, Pastor JY, Elices M. Stress intensity factor compliance and CMOD for a general three-point-beam. Int J Fracture 1998;89:10316.  [5] Trdeg A, Nilsson F, Ostlund S. FEM-remeshing technique applied to crack growth problems. Comput Methods Appl Mech a ard Eng 1998;160:11531. [6] Wawrzynek PA, Ingraea AR. An edge-based data structure for two-dimensional nite element analysis. Eng Comput 1987;3:13 20. [7] Lim IL, Johnston W, Choi SK. A nite element code for fracture propagation analysis within elasto-plastic continuum. Eng Fract Mech 1996;53:193211. [8] Mosalam KM, Paulino GH. Evolutionary characteristic length method for smeared cracking nite element models. Finite Elements Anal Design 1997;27:99108. [9] MSC/PATRAN user's manual, version 7. Los Angeles, CA: The MacNeal-Schwendler Corporation; 1997. [10] A B A Q U S user's and theory manuals, Version 5.7. Pawtucket, RI: Hibbitt, Karlsson and Sorensen, Inc., (HKS); 1998. [11] ASM metals reference book, 3rd ed. Metals Park, OH: ASM International; 1993. p. 50511. [12] Gooch WA, Burkins MS, Palicka R, Rubin J, Ravichandran R. Development and ballistic testing of a functionally gradient ceramic/metal applique. Proceedings of 17th International Symposium on Ballistics, Midrand, South Africa, 1998. [13] Annual book of standards. Standard E399-90: standard test method for plane strain fracture toughness of metallic materials. Philadelphia, PA: American Society for Testing and Materials; 1997. [14] Annual book of standards. Standard E1820-96: standard test method for measurement of fracture toughness. Philadelphia, PA: American Society for Testing and Materials; 1997. [15] Anderson TL. Fracture mechanics fundamentals and applications, 2nd ed. Boca Rotan, FL: CRC Press; 1995. p. 57785. [16] Moran B, Shih CF. A general treatment of crack tip contour integrals. Int J Fract 1987;35:295310. [17] Shih CF, Moran B, Nakamura T. Energy release rate along a three-dimensional crack front in a thermally stressed body. Int J Fract 1987;30:79102. [18] Shih CF. Relationship between the J-integral and the crack opening displacement for stationary and extending cracks. J Mech Phys Solids 1981;4:30526. [19] McMeeking RM, Parks DM. On criteria for J-dominance of crack tip elds in large-scale yielding. ASTM STP 668. American Society for Testing and Materials; 1979. p. 17594. [20] Dodds Jr. RH. WARP3D Static and dynamic nonlinear analysis of solids. Department of Civil and Environmental Engineering, University of Illinois at Urbana, Champaign, 1999.