Determination and verification of triaxiality dependent cohesive zone parameters of SA333 Grade 6 steel

Abstract

Cohesive zone models are employed to simulate crack propagation in fracture process zone. It has been demonstrated in the present study that the cohesive zone parameters depend not only on the material, but also on stress triaxiality. To determine cohesive zone parameters, experimental results of 14 three-point bend specimens (TPBB) made of SA333 Grade 6 steel were analyzed using 3D finite element model using WARP3D. Cohesive parameters were determined by carrying out parametric studies of these specimens by varying peak stress ‘T’ to match experimental load-displacement data with the computed data. An exponential traction separation law was used for this purpose. In addition, elastic-plastic finite element analyses (FEA) are conducted to determine the multi-axiality quotient ‘q’ at the crack tip for TPBB specimens and pipe components. This helped to get a relation between peak stress ‘T’ and ‘q’. To extend the range of validity of parameter ‘T’ as a function of ‘q’, experimental results of six piping components with through-wall circumferential crack made of SA333 Grade 6 steel tested earlier were also used similarly. A confidence interval band between q and T is then plotted to study the transferability of cohesive parameters. The variation in ‘T’ for the same value of ‘q’ is attributed to the microstructural changes in the material near crack tip.To test the accuracy of the cohesive parameters determined above, experimental results of six piping components were again used. Using the ‘q’ parameters determined from elastic-plastic FEA on pipes, the upper and lower limits of peak stress ‘T’ from the confidence band determined above were found out. This range of ‘T’ was then used to carry out cohesive zone analyses (CZA) of these piping components to determine load-displacement curves. For each piping component, a set of load-displacement curves were determined by assuming a normal variation of parameter ‘T’ over the maximum and minimum limits. The simulation results were compared with the experimental results and it has been observed that experimental results lie reasonably well within the computed results.