Predicting elastic and plastic properties of small iron polycrystals by machine learning

Abstract

Deformation of crystalline materials is an interesting example of complex system behaviour. Small samples typically exhibit a stochastic-like, irregular response to externally applied stresses, manifested as significant sample-to-sample variation in their mechanical properties. In this work we study the predictability of the sample-dependent shear moduli and yield stresses of a large set of small cube-shaped iron polycrystals generated by Voronoi tesselation, by combining molecular dynamics simulations and machine learning. Training a convolutional neural network to infer the mapping between the initial polycrystalline structure of the samples and features of the ensuing stress-strain curves reveals that the shear modulus can be predicted better than the yield stress. We discuss our results in the context of the sensitivity of the system's response to small perturbations of its initial state.

References in corpus (9)

• Machine learning and deep learning
• Machine learning and the physical sciences
• Robust Structural Identification via Polyhedral Template Matching
• Graph Neural Networks for an Accurate and Interpretable Prediction of the Properties of Polycrystalline Materials
• Machine learning plastic deformation of crystals
• Quenched pinning and collective dislocation dynamics
• Machine learning depinning of dislocation pileups
• Machine learning reveals strain-rate-dependent predictability of discrete dislocation plasticity
• Prediction of steel nanohardness by using graph neural networks on surface polycrystallinity maps

Cited by in corpus (3)

• Stress Predictions in Polycrystal Plasticity using Graph Neural Networks with Subgraph Training
• Conditional diffusion model for inverse prediction of process parameters and dendritic microstructures from mechanical properties
• Estimating predictability of depinning dynamics by machine learning