PREDICTING THE MACROSCOPIC CYCLIC BEHAVIOUR OF POLYCRYSTALLINE STEELS BASED ON MATERIAL MICROSTRUCTURE VIA SURROGATE MODELLING
Crystal plasticity finite element models can simulate the effect of microstructure on the cyclic behaviour of polycrystalline steels and can simulate the resulting local plastic strain. However, such models are computationally expensive and are therefore limited to simulation on small volume elements of material. In this work, a Gaussian process regression model is proposed as a surrogate model to predict macroscopic quantities of interest based on input parameters relating to the cyclic loading and material microstructure. The advantage with relation to computational expense of the surrogate can be leveraged for the purposes of undertaking uncertainty quantification and sensitivity analysis regarding the effect of the model inputs on the output prediction.