Multiscale simulation of spatially correlated microstructure via a latent space representation

TL;DR

This paper introduces a multiscale simulation approach using variational autoencoders to efficiently model spatially correlated microstructures and their effects on material performance, especially for additively manufactured metals.

Contribution

It presents a novel variational autoencoder-based method for encoding and sampling microstructure, capturing spatial correlations and finite-size effects in multiscale simulations.

Findings

Accurately reproduces microstructure in direct numerical simulations

Effectively propagates microstructural uncertainty to performance metrics

Simulates functional gradation with stochastic microstructure

Abstract

When deformation gradients act on the scale of the microstructure of a part due to geometry and loading, spatial correlations and finite-size effects in simulation cells cannot be neglected. We propose a multiscale method that accounts for these effects using a variational autoencoder to encode the structure-property map of the stochastic volume elements making up the statistical description of the part. In this paradigm the autoencoder can be used to directly encode the microstructure or, alternatively, its latent space can be sampled to provide likely realizations. We demonstrate the method on three examples using the common additively manufactured material AlSi10Mg in: (a) a comparison with direct numerical simulation of the part microstructure, (b) a push forward of microstructural uncertainty to performance quantities of interest, and (c) a simulation of functional gradation of a part with stochastic microstructure.