Technology Category
Machine Learning: Deep Generative Models & AutoencodersComputer Vision: 3D Computer VisionSearch and Optimization: Sampling/Simulation-based Search
3D variational autoencoder for fingerprinting microstructure volume elements
🤖 AI Summary
Low modeling efficiency and poor generalizability in crystal plasticity (CP) simulations hinder rapid microstructure–stress response prediction. Method: We propose a crystallography-aware 3D variational autoencoder (3D-VAE) framework. First, the fundamental zone (FZ) is embedded into the 3D-VAE preprocessing pipeline to ensure continuity and convergence of orientation representations. Second, voxelized grain orientation fields serve as inputs, enabling generalized low-dimensional encoding of microstructures in a 256-dimensional latent space. Third, CP simulations are coupled with a fully connected surrogate model to establish structure–property relationships. Results: The framework achieves a mean orientation reconstruction error of 9×10⁻³ on the test set and a relative root-mean-square error of 8.9×10⁻⁴ for stress response prediction. It significantly improves computational efficiency and demonstrates strong out-of-distribution generalization across diverse textures, grain sizes, and aspect ratios.
Technology Category
Application Category
📝 Abstract
Microstructure quantification is an important step towards establishing structure-property relationships in materials. Machine learning-based image processing methods have been shown to outperform conventional image processing techniques and are increasingly applied to microstructure quantification tasks. In this work, we present a 3D variational autoencoder (VAE) for encoding microstructure volume elements (VEs) comprising voxelated crystallographic orientation data. Crystal symmetries in the orientation space are accounted for by mapping to the crystallographic fundamental zone as a preprocessing step, which allows for a continuous loss function to be used and improves the training convergence rate. The VAE is then used to encode a training set of VEs with an equiaxed polycrystalline microstructure with random texture. Accurate reconstructions are achieved with a relative average misorientation error of 9x10-3 on the test dataset, for a continuous latent space with dimension 256. We show that the model generalises well to microstructures with textures, grain sizes and aspect ratios outside the training distribution. Structure-property relationships are explored through using the training set of VEs as initial configurations in various crystal plasticity (CP) simulations. Microstructural fingerprints extracted from the VAE, which parameterise the VEs in a low-dimensional latent space, are stored alongside the volume-averaged stress response, at each strain increment, to uniaxial tensile deformation from CP simulations. This is then used to train a fully connected neural network mapping the input fingerprint to the resulting stress response, which acts as a surrogate model for the CP simulation. The fingerprint-based surrogate model is shown to accurately predict the microstructural dependence in the CP stress response, with a relative mean-squared error of 8.9x10-4 on unseen test data.
Problem
Research questions and friction points this paper is trying to address.
Develops 3D VAE for encoding microstructure volume elements
Improves training convergence by handling crystal symmetries
Predicts stress response using microstructure fingerprints
Innovation
Methods, ideas, or system contributions that make the work stand out.
3D variational autoencoder for microstructure encoding
Crystallographic fundamental zone mapping for symmetry
Latent space fingerprints predict stress response
🔎 Similar Papers
No similar papers found.
