A Hybrid Modelling Approach Applied to the Evolution of Microstructure during Plane Strain Deformation

Abstract

This paper presents a modelling strategy that combines neuro-fuzzy methods to dene the material model with cellular automata representations of the microstructure, all embedded within a nite element solver that can deal with the large deformations of metal processing technology. We use the acronymnf-CAFE as a label for the method. The need for such an approach arises from the twin demands of computational speed for quick solutions for ecient material characterisation by incorporating metallurgical knowledge for material design models and subsequent process control. In this strategy, the cellular automata hold the microstructural features in terms of sub-grain size and dislocation density which are modelled by a neuro-fuzzy system that predicts the ow stress. The proposed methodology is validated on a two dimensional (2D) plane strain compression nite element simulation with Al1%Mg alloy. Results from the simulations show the potential of the model for incorporating the eects of the underlying microstructure on the evolving ow stress elds. In doing this, the paper highlights the importance of understanding the local transition rules that aect the global behaviour during deformation.