Modelling Heat Diffusion In Voronoi-Based Polycrystalline Materials Using Ai
Nishi Bhemani, Thulsiram Gantala, Krishnan Balasubramaniam · e-Journal of Nondestructive Testing · 2026
Not cached yet. This record was fetched live for this request and is stored after the lookup. Reopening the page should serve the cached copy.
e-Journal of Nondestructive Testing · NDT.net GmbH & Co. KG · article · 2026
Abstract
This article presents a deep-learning–based framework for simulating heat diffusion in polycrystalline materials. The microstructure is represented using Voronoi tessellations, where grain size and orientation define the 2D CAD geometry for numerical heat diffusion simulations. Accurately modelling anisotropic heat transport using conventional finite element (FE) methods is computationally expensive due to complex microstructural features. To overcome this limitation, a Data-driven Simulation-Assisted-Physics learned AI (DPAI) model is developed and trained using FE-generated datasets.
The proposed encoder–decoder architecture employs four ConvLSTM cells to learn the physics of heat diffusion from microstructure-informed thermal fields. The trained model is validated under varying anisotropy conditions and demonstrates a significant reduction in computational time while maintaining reasonable accuracy compared to FE simulations.
Open the resolved locator · 10.58286/33366
Metadata licence: OpenAlex metadata (CC0)