Efficient design of harmonic structure using an integrated hetero-deformation induced hardening model and machine learning algorithm

Abstract

Harmonic structured material (HSM) of coarse-grained core and fine-grained shell microstructure in SS304L was designed using a three-dimensional numerical model developed to simulate the unique deformation behavior of heterostructured materials. Hetero-deformation induced (HDI) strengthening feature was implemented into the model by considering the pile-up of geometrically necessary dislocations (GNDs) near grain boundaries. The finite element analysis using statically equivalently synthesized 3D representative volume element (RVE) not only supported the experimental results of tensile stress and HDI stress but also described the local strain partitioning near the core-shell boundaries in the HSM. Based on the developed model, the optimal harmonic microstructure was investigated with the aid of the machine learning (ML) technique. Numerous virtual microstructures were generated based on the real microstructures to enlarge the data pool for machine learning. Bayesian inference was adapted to establish the correlation between the microstructures and the mechanical properties. The optimal microstructure with the greatest combination of strength and ductility was predicted among over 500 candidates. A new HSM designed using the ML-based prediction was successfully manufactured, exhibiting superior mechanical performance compared to any other previously designed heterostructured SS304L.