Maximising mechanical properties of aluminium alloys by microstructural optimisation using a coarsened surrogate model

Surrogate-based microstructural optimization was applied to describe the relationship between local microstructural patterns and particle damage in wrought 2024 aluminium alloy. A support vector machine was used to realise high-accuracy optimisation from a limited number of high-computational-cost image-based simulation results. The methodology integrated thoroughgoing microstructural quantification, a couple of coarsening processes, and surrogate modelling. The following three objective functions were defined: the maximum principal stress in particles, the equivalent plastic strain, and the stress triaxiality in the matrix. A number of design parameters were comprehensively prepared that quantitatively expressed the size, shape, and spatial distribution of particles and pre-existing micro pores in numerous ways. The number of design parameters was then reduced from 86 to 4 for each objective function during the coarsening process. The surrogate model provided the dependency of particle damage for the size, shape, and spatial distribution of particles and micro pores in the form of a multi-dimensional response surface.