Distributed physics-informed neural networks via domain decomposition for fast flow reconstruction
arXiv:2602.15883v1 Announce Type: new Abstract: Physics-Informed Neural Networks (PINNs) offer a powerful paradigm for flow reconstruction, seamlessly integrating sparse velocity measurements with the governing Navier-Stokes equations to recover complete velocity and latent pressure fields. However, scaling such models to large spatiotemporal domains is hindered by computational bottlenecks and optimization instabilities. In this work, we propose a robust distributed PINNs framework designed for efficient flow reconstruction via spatiotemporal domain decomposition. A critical challenge in such distributed solvers is pressure indeterminacy, where independent sub-networks drift into inconsistent local pressure baselines. We address this issue through a reference anchor normalization strategy coupled with decoupled asymmetric weighting. By enforcing a unidirectional information flow from designated master ranks where the anchor point lies to neighboring ranks, our approach eliminates gau
arXiv:2602.15883v1 Announce Type: new Abstract: Physics-Informed Neural Networks (PINNs) offer a powerful paradigm for flow reconstruction, seamlessly integrating sparse velocity measurements with the governing Navier-Stokes equations to recover complete velocity and latent pressure fields. However, scaling such models to large spatiotemporal domains is hindered by computational bottlenecks and optimization instabilities. In this work, we propose a robust distributed PINNs framework designed for efficient flow reconstruction via spatiotemporal domain decomposition. A critical challenge in such distributed solvers is pressure indeterminacy, where independent sub-networks drift into inconsistent local pressure baselines. We address this issue through a reference anchor normalization strategy coupled with decoupled asymmetric weighting. By enforcing a unidirectional information flow from designated master ranks where the anchor point lies to neighboring ranks, our approach eliminates gauge freedom and guarantees global pressure uniqueness while preserving temporal continuity. Furthermore, to mitigate the Python interpreter overhead associated with computing high-order physics residuals, we implement a high-performance training pipeline accelerated by CUDA graphs and JIT compilation. Extensive validation on complex flow benchmarks demonstrates that our method achieves near-linear strong scaling and high-fidelity reconstruction, establishing a scalable and physically rigorous pathway for flow reconstruction and understanding of complex hydrodynamics.
Executive Summary
This article introduces a distributed framework for flow reconstruction using Physics-Informed Neural Networks (PINNs), addressing limitations in scalability and pressure indeterminacy. The proposed approach employs domain decomposition, reference anchor normalization, and decoupled asymmetric weighting to achieve near-linear strong scaling and high-fidelity reconstruction. To mitigate computational overhead, the authors implement a high-performance training pipeline accelerated by CUDA graphs and JIT compilation. The method is extensively validated on complex flow benchmarks, demonstrating its potential as a scalable and physically rigorous pathway for flow reconstruction and hydrodynamics understanding.
Key Points
- ▸ The article proposes a distributed PINNs framework for efficient flow reconstruction via spatiotemporal domain decomposition.
- ▸ The approach addresses pressure indeterminacy through reference anchor normalization and decoupled asymmetric weighting.
- ▸ A high-performance training pipeline is implemented using CUDA graphs and JIT compilation to mitigate computational overhead.
Merits
Strength
The proposed framework demonstrates near-linear strong scaling and high-fidelity reconstruction on complex flow benchmarks.
Methodological Innovation
The reference anchor normalization strategy and decoupled asymmetric weighting approach provide a novel solution to pressure indeterminacy in distributed PINNs.
Scalability
The distributed framework enables efficient flow reconstruction in large spatiotemporal domains, overcoming computational bottlenecks and optimization instabilities.
Demerits
Limitation
The article assumes a priori knowledge of the flow domain and governing equations, which may not be feasible in all practical scenarios.
Complexity
The implementation of CUDA graphs and JIT compilation may require significant expertise and computational resources.
Expert Commentary
The article presents a significant contribution to the field of PINNs, addressing critical limitations in scalability and pressure indeterminacy. The proposed framework demonstrates near-linear strong scaling and high-fidelity reconstruction, establishing a scalable and physically rigorous pathway for flow reconstruction and hydrodynamics understanding. However, the article assumes a priori knowledge of the flow domain and governing equations, which may not be feasible in all practical scenarios. Additionally, the implementation of CUDA graphs and JIT compilation may require significant expertise and computational resources. Nevertheless, the proposed framework has the potential to significantly improve flow reconstruction and hydrodynamics understanding in various fields.
Recommendations
- ✓ Future research should focus on developing more generalizable and adaptable PINNs frameworks that can handle uncertainty and incomplete information.
- ✓ The proposed framework should be further validated on a broader range of complex flow benchmarks to demonstrate its robustness and scalability.
