Grain-Scale Heterogeneity Effect on Mechanistic Modelling of Cleavage Fracture of a Ferritic RPV Steel Forging Material

Abstract

Improving brittle fracture prediction is crucial for structural integrity assessment. In current safety assessments, fracture mechanics treats polycrystalline steels as homogeneous continua. In reality, deformation of structural steels is heterogeneous. Part of this heterogeneity is due to the elastic and plastic anisotropy of their constituent (often randomly orientated) grains. This paper will compare the predicted failure stresses from tensile tests performed on a ferritic pressure vessel steel using the crystal plasticity finite element approach alongside measured carbide distribution and classical Beremin cleavage model. Available tensile data of 22NiMoCr37 steel at low temperature (−91°C and −154°C) were analysed using Bridgman solutions to account for the necking effect on the stress state at the centre of necking where brittle cracking initiates. This stress state imposed on representative volume element (RVE) made up of 10×10×10 randomly orientated grains, whose deformation is simulated using crystal plasticity finite element modelling (CPFEM). Randomly distributed carbides were produced based on the measured carbide size distribution and density for this steel. By assuming carbides as Griffith microcracks, the cleavage fracture stress in each grain can be assessed based the maximum principal stress on the cleavage crystal plane and an assumed surface energy. By repeating the random carbide distribution 1,000 times, brittle fracture probability can be calculated. Detailed examination shows that the above approach is actually a verification of the BEREMIN local approach model for cleavage fracture. The modelling results were compared with the available ductility data at −91°C and the interpolated ductility data at −154°C at the centre of necking. It is foreseen that this approach will lead to improvements in brittle fracture modelling in heterogeneous ferritic steels by introducing realistic surface energies and real defect distributions in specific materials, when used alongside the CPFEM submodelling approach.