A domain partitioning based pre-processor for multi-scale modelling of cast aluminium alloys

Abstract

In this paper, a microstructural morphology based domain partitioning MDP methodology is developed for materials with non-uniform heterogeneous microstructure. The comprehensive set of methods is intended to provide a concurrent multi-scale analysis model with an initial computational domain that delineates regions of statistical homogeneity and inhomogeneity. The MDP methodology is intended to be a pre-processor to multi-scale analysis of mechanical behaviour and damage of heterogeneous materials, e.g. cast aluminium alloys. It introduces a systematic three-step process that is based on geometric features of morphology. The first step simulates high resolution microstructural information from low resolution micrographs of the material and a limited number of high resolution optical or scanning electron microscopy micrographs. The second step is quantitative characterization of the high resolution images to create effective metrics that can relate microstructural descriptors to material behaviour. The third step invokes a partitioning method to demarcate regions belonging to different length scales in a concurrent multi-scale model. Partitioning criteria for domain partitioning are defined in terms of microstructural descriptors and their functions. The effectiveness of these metrics in differentiating microstructures of a 319-type cast aluminium alloy with different secondary dendrite arm spacings SDAS is demonstrated. The MDP method establishes intrinsic material length scales for the different SDAS, namely, 23, 70 and 100 µm, and consequently subdivides the computational domain for concurrently coupling macro- and micromechanical analyses in the multi-scale model.