Reproduce “An Energy Approach to the Solution of Partial Differential Equations in Computational Mechanics via Machine Learning: Concepts, Implementation and Applications” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Esteban Samaniego, Cosmin Anitescu, Somdatta Goswami, Vien Minh Nguyen-Thanh, Hongwei Guo, Khader Hamdia, Timon Rabczuk, Xiaoying Zhuang
Researchers developed the Deep Energy Method (DEM), which employs Deep Neural Networks to solve Partial Differential Equations in computational mechanics by minimizing the system's total energy functional. This method consistently produced accurate solutions across diverse problems, including linear elasticity and phase-field fracture, often outperforming collocation strategies and simplifying boundary condition handling.
Key insights
- Minimizing the total energy functional provides a natural, physically consistent, and robust loss function for DNNs in computational mechanics, aligning with fundamental physical principles.
- The energy-based formulation (DEM) offers superior stability and accuracy compared to collocation methods (DCM), especially for problems involving discontinuities or complex boundary conditions, as demonstrated in phase-field fracture.
- Leveraging automatic differentiation and optimized algorithms from deep learning libraries (TensorFlow, PyTorch) greatly simplifies the implementation of complex variational formulations and gradient computations for PDE solvers.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.