PSI - Issue 28

B.W. Williams et al. / Procedia Structural Integrity 28 (2020) 1024–1038 Author name / Structural Integrity Procedia 00 (2019) 000–000

1031

8

that the values of the critical brittle failure stress would be expected to be mesh dependant. Mesh dependency is further addressed in the next section. 4. FEA Model of Charpy Test The tensile response of the round tensile bars was simulated using about 1000 axisymmetric elements in the explicit FEA software Abaqus. The tensile models were used to conform an accurate hardening response of the model at various temperatures. With the damage mechanics model, the effective strain at failure under uniaxial tensile loading conditions was well beyond the point of instability (or necking) measured in the tensile tests. The Charpy response was modelled using approximately 32,000 solid elements with quarter symmetry in the explicit FEA software DYNA3D. The mesh of the Charpy specimen is shown in Figure 6 with elements of size 0.2 mm used along the crack plane.

Figure 6: FEA mesh of Charpy impact test; quarter-symmetry

The damage model was implemented as a user-defined subroutine with details of the subroutine given by Simha et al. (2014) with the exception that the MMC model was used as opposed to the XW damage model. Isotropic deformation and fracture were used. As TC128B displayed anisotropic fracture, as seen in Figure 2, anisotropic fracture models should be studied in future work. In both the tensile and Charpy models, the element size in the location of failure and fracture was 0.2 mm, as was the case in the Paredes et al. (2018) model for TC128. Also, as reported by Paredes et al. (2016), it was found that 0.2 mm was a good compromise between accuracy and computation cost with an explicit solver. Consequently, the damage zone stress/strain values are specific to a size of 0.2 mm. Smaller mesh sizes could be achieved with an implicit solver but stability with element deletion remains a concern. Gesing et al. (2016) detailed the use of the XW damage model for both DWTT and Charpy simulations of X70. It is mentioned that the load-displacement response from the Charpy simulation did not entirely match the experiment response as the XW model was calibrated to match the load-displacement response from DWTT. Consequently, it would not be expected that the Charpy simulations match exactly with the experimental results for the room temperature, quasi-static case, as the MMC-PW was calibrated based on different geometries (notched tension, shear, biaxial, SEB and CT).