Abstract:Physics-informed neural network (PINN) is a data-driven approach to solve equations. It is successful in many applications; however, the accuracy of the PINN is not satisfactory when it is used to solve multiscale equations. Homogenization is a way of approximating a multiscale equation by a homogenized equation without multiscale property; it includes solving cell problems and the homogenized equation. The cell problems are periodic; and we propose an oversampling strategy which greatly improves the PINN accuracy on periodic problems. The homogenized equation has constant or slow dependency coefficient and can also be solved by PINN accurately. We hence proposed a 3-step method to improve the PINN accuracy for solving multiscale problems with the help of the homogenization. We apply our method to solve three equations which represent three different homogenization. The results show that the proposed method greatly improves the PINN accuracy. Besides, we also find that the PINN aided homogenization may achieve better accuracy than the numerical methods driven homogenization; PINN hence is a potential alternative to implementing the homogenization.

What problem does this paper attempt to address?

This paper attempts to address the issue of low accuracy when using Physics-Informed Neural Networks (PINN) to solve multiscale partial differential equations. Specifically: 1. **Challenges of Multiscale Problems**: - Multiscale problems are common in many scientific and engineering applications, such as biomedical, materials science, and petroleum engineering. - Traditional numerical methods face difficulties in handling multiscale problems, requiring fine grids or complex multiscale finite element methods, which lead to high computational costs. 2. **Limitations of PINN in Multiscale Problems**: - Although PINN performs well in solving many mathematical problems, its accuracy is often unsatisfactory when dealing with multiscale problems. - Traditional PINN has relatively large errors when handling multiscale problems and is sensitive to hyperparameters and randomness during the training process. 3. **Proposed Method in the Paper**: - The paper proposes a Neural Homogenization based PINN (NH-PINN) method to improve the accuracy of PINN in multiscale problems. - Homogenization is a method that approximates multiscale equations as homogenized equations without multiscale properties by solving cell problems and homogenized equations. 4. **Specific Steps**: - **Step 1**: Use PINN to solve the cell problem. The cell problem has periodic boundary conditions, and the paper proposes an oversampling strategy to improve PINN's performance in high-dimensional periodic problems. - **Step 2**: Evaluate the homogenization coefficients. The derivatives of the cell problem solution are obtained through automatic differentiation to get the homogenization coefficients. - **Step 3**: Use PINN to solve the homogenized equation. Since the coefficients of the homogenized equation are usually constants or slowly varying, PINN can accurately solve such equations. 5. **Experimental Results**: - The paper validates the effectiveness of the NH-PINN method through four numerical experiments, showing that NH-PINN significantly improves the accuracy of PINN in handling multiscale problems. - Especially in cases with smaller scales, NH-PINN's accuracy is significantly better than traditional PINN. 6. **Contributions**: - Observed the accuracy decline of traditional PINN in multiscale problems and explained the reasons. - Proposed the NH-PINN method to improve PINN's performance through homogenization. - Proposed an oversampling strategy that significantly improves PINN's accuracy in high-dimensional periodic problems. - Experimental results show that NH-PINN not only improves PINN's accuracy but may also achieve more accurate homogenization than traditional numerical methods. In summary, this paper aims to overcome the accuracy deficiency of PINN in multiscale problems by introducing the homogenization method, thus providing a more effective solution for multiscale problems.

NH-PINN: Neural homogenization based physics-informed neural network for multiscale problems

Densely Multiplied Physics Informed Neural Networks

A hybrid physics-informed neural network based multiscale solver as a partial differential equation constrained optimization problem

MRF-PINN: A Multi-Receptive-Field convolutional physics-informed neural network for solving partial differential equations

Mass-preserving Spatio-temporal adaptive PINN for Cahn-Hilliard equations with strong nonlinearity and singularity

Enhancing PINNs for solving PDEs via adaptive collocation point movement and adaptive loss weighting

Improvement of Bayesian PINN Training Convergence in Solving Multi-scale PDEs with Noise

Hybrid Finite Difference with the Physics-informed Neural Network for solving PDE in complex geometries

Generalized conditional symmetry enhanced physics-informed neural network and application to the forward and inverse problems of nonlinear diffusion equations

A Physics Informed Neural Network for Time-Dependent Nonlinear and Higher Order Partial Differential Equations

A mixed formulation for physics-informed neural networks as a potential solver for engineering problems in heterogeneous domains: Comparison with finite element method

Physics-informed neural networks based on adaptive weighted loss functions for Hamilton-Jacobi equations

Physics-informed neural networks with parameter asymptotic strategy for learning singularly perturbed convection-dominated problem

Physics Informed Neural Networks (PINNs) as intelligent computing technique for solving partial differential equations: Limitation and Future prospects

PINN-MG: A Multigrid-Inspired Hybrid Framework Combining Iterative Method and Physics-Informed Neural Networks

Solution of conservative-form transport equations with physics-informed neural network

BC-PINN: an adaptive physics informed neural network based on biased multiobjective coevolutionary algorithm

Physics-informed Neural Networks (PINN) for computational solid mechanics: Numerical frameworks and applications

Combining physics-informed graph neural network and finite difference for solving forward and inverse spatiotemporal PDEs

iPINNs: incremental learning for Physics-informed neural networks

Physics-informed neural networks combined with polynomial interpolation to solve nonlinear partial differential equations