Work

PINNs-MPF: An Efficient Physics-Informed Machine Learning-based Solver for Multi-Phase-Field Simulations using Tensorflow

Year: 2025

Type: article

Abstract: This paper introduces PINNs-MPF, a novel Machine Learning-based solver designed for Multi-Phase-Field (MPF) and diffuse interface simulations, offering innovative approaches to address complex challenges in addressing microstructure evolution in polycrystalline materials using Machine Learning. The framework not only surpasses current limitations in handling multi-phase problems but also allows for potential upscaling to tackle more intricate scenarios. Developed in Python, the related code leverages optimized libraries like TensorFlow, showcasing efficiency and potential scalability in materials science and engineering simulations. This framework, integrating advanced techniques such as multi-networking and training optimization, setting a new standard in predictive capabilities and understanding complex physical phenomena. • Efficient Multi-Phase Field Solver: Open-source Python tool using PINNs for complex phase simulations. • High Scalability: Optimized for large-scale problems with adaptive optimization and checkpointing. • Community Access: Available on GitHub and CodeOcean for reproducibility and collaboration.

Source: Software Impacts

DOI: 10.1016/j.simpa.2025.100753

Authors: Seifallah Elfetni, Reza Darvishi Kamachali

Institutions: Federal Institute For Materials Research and Testing, University of Münster

Language: English

Topic: Model Reduction and Neural Networks

Subfield: Statistical and Nonlinear Physics

Field: Physics and Astronomy

Domain: Physical Sciences

Open Access status: gold

Estimated APC paid: $600

Funder: Deutsche Forschungsgemeinschaft