Abstract:As is well known, differential algebraic equations (DAEs), which are able to describe dynamic changes and underlying constraints, have been widely applied in engineering fields such as fluid dynamics, multi-body dynamics, mechanical systems and control theory. In practical physical modeling within these domains, the systems often generate high-index DAEs. Classical implicit numerical methods typically result in varying order reduction of numerical accuracy when solving high-index systems.~Recently, the physics-informed neural network (PINN) has gained attention for solving DAE systems. However, it faces challenges like the inability to directly solve high-index systems, lower predictive accuracy, and weaker generalization capabilities. In this paper, we propose a PINN computational framework, combined Radau IIA numerical method with a neural network structure via the attention mechanisms, to directly solve high-index DAEs. Furthermore, we employ a domain decomposition strategy to enhance solution accuracy. We conduct numerical experiments with two classical high-index systems as illustrative examples, investigating how different orders of the Radau IIA method affect the accuracy of neural network solutions. The experimental results demonstrate that the PINN based on a 5th-order Radau IIA method achieves the highest level of system accuracy. Specifically, the absolute errors for all differential variables remains as low as $10^{-6}$, and the absolute errors for algebraic variables is maintained at $10^{-5}$, surpassing the results found in existing literature. Therefore, our method exhibits excellent computational accuracy and strong generalization capabilities, providing a feasible approach for the high-precision solution of larger-scale DAEs with higher indices or challenging high-dimensional partial differential algebraic equation systems.

What problem does this paper attempt to address?

This paper attempts to address the issues of numerical accuracy degradation and low prediction accuracy encountered when solving high-index differential-algebraic equations (DAEs) systems. Specifically, traditional implicit numerical methods often lead to a reduction in numerical accuracy order when solving high-index DAEs. Although existing physics-informed neural networks (PINNs) can be used to solve DAEs systems, their ability to directly solve high-index systems is limited, with lower prediction accuracy and weaker generalization capability. Therefore, this paper proposes a PINN computational framework based on the Radau IIA method and attention mechanism, aiming to directly solve high-index DAEs and improve solution accuracy through a domain decomposition strategy. ### Main Contributions of the Paper: 1. **Proposed a new computational framework**: Combining the Radau IIA numerical method with neural network structures, particularly enhancing the model's expressive and generalization capabilities by introducing an attention mechanism. 2. **Solved the problem of high-index DAEs**: Traditional numerical methods face accuracy degradation when solving high-index DAEs, whereas the proposed method can directly solve these high-index systems. 3. **Improved prediction accuracy**: Experimental results show that the PINN based on the 5th-order Radau IIA method maintains the absolute error of all differential variables at the order of 10^-6 and the absolute error of algebraic variables at the order of 10^-5 when solving two classic high-index systems, surpassing the results in existing literature. 4. **Adopted a domain decomposition strategy**: By decomposing the time domain into multiple segments, not only is the solution accuracy improved, but the advantages of iterative training are also utilized, reducing the required amount of training data and lowering training costs. ### Experimental Validation: - **Hessenberg-type DAEs system**: By comparing the average absolute errors of different orders of the Radau IIA method (3rd-order, 5th-order, 9th-order, 13th-order), it was found that the 5th-order Radau IIA method achieved the highest accuracy for all differential and algebraic variables. - **DAEs system of the pendulum model**: Similarly, by comparing the average absolute errors of different orders of the Radau IIA method, it was found that the 5th-order Radau IIA method achieved the highest accuracy for all differential and algebraic variables. ### Conclusion: The PINN computational framework based on the Radau IIA method and attention mechanism proposed in this paper can not only directly solve high-index DAEs but also performs excellently in terms of prediction accuracy and generalization capability, providing a feasible solution for high-precision solving of large-scale high-index DAEs or high-dimensional partial differential-algebraic equation systems.

Physical Information Neural Networks for Solving High-index Differential-algebraic Equation Systems Based on Radau Methods

Physics-Informed Neural Networks for Solving High-Index Differential-Algebraic Equation Systems Based on Radau Methods

Solving Partial Differential Equations Using Deep Learning and Physical Constraints

Physics-informed Radial Basis Network (PIRBN): A Local Approximating Neural Network for Solving Nonlinear Partial Differential Equations

DAE-PINN: a physics-informed neural network model for simulating differential algebraic equations with application to power networks

Physics-Informed neural network solver for numerical analysis in geoengineering

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

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

Partial‐differential‐algebraic equations of nonlinear dynamics by physics‐informed neural‐network: (I) Operator splitting and framework assessment

A Priori Error Estimation of Physics-Informed Neural Networks Solving Allen--Cahn and Cahn--Hilliard Equations

An Efficient Framework for Solving Forward and Inverse Problems of Nonlinear Partial Differential Equations Via Enhanced Physics-Informed Neural Network Based on Adaptive Learning

A numerical method to solve PDE through PINN based on ODENet

Differential Transform Method (DTM) and Physics-Informed Neural Networks (PINNs) in Solving Integral–Algebraic Equation Systems

A-PINN: Auxiliary physics informed neural networks for forward and inverse problems of nonlinear integro-differential equations

A General Method for Solving Differential Equations of Motion Using Physics-Informed Neural Networks

Data-Guided Physics-Informed Neural Networks for Solving Inverse Problems in Partial Differential Equations

Physical informed neural networks with soft and hard boundary constraints for solving advection-diffusion equations using Fourier expansions

Multi-Net Strategy: Accelerating Physics-Informed Neural Networks for Solving Partial Differential Equations.

Physics-informed radial basis network (PIRBN): A local approximating neural network for solving nonlinear PDEs

Robust discovery of partial differential equations in complex situations