A Workflow for Utilizing OpenFOAM Data Structure in Physics-Informed Deep Learning Training

Yijin Mao,Yuwen Zhang

2024-08-11

Abstract:This study presents a novel methodology for integrating physics-informed loss functions into deep learning models using OpenFOAM's comprehensive data structures. Leveraging the robust and flexible capabilities of OpenFOAM's data structure for handling complex geometries and boundary conditions, it is demonstrated how to construct detailed loss functions that accurately embed physics constraints and potentially enhance the training and performance of neural networks in handling industrial-level complicated geometry for computational fluid dynamics (CFD) simulations. The present work primarily focuses on the 1D Burger equation to showcase the detailed procedure of constructing initial loss, boundary loss, and residual loss. While the computational geometry employed here is relatively simple, the procedure is sufficiently general to illustrate its applicability to more complex computational domains. The results show the trained operator former (OFormer) neural network can successfully predict the simulation results subject to the OpenFOAM'data structure composed loss. This framework potentially opens new avenues for using deep learning to tackle complex industrial simulation challenges, promising significant advancements in the accuracy and practicality of CFD applications.

Fluid Dynamics

What problem does this paper attempt to address?

The paper aims to address the following issues:

1. **Combining Physical Information Loss Functions with Deep Learning Models**: The paper proposes a novel approach that integrates OpenFOAM's data structure with physical information loss functions in deep learning models. By leveraging OpenFOAM's capability to handle complex geometries and boundary conditions, detailed loss functions are constructed to accurately embed physical constraints, thereby improving the training and performance of neural networks in computational fluid dynamics (CFD) simulations of industrial-grade complex geometries.

2. **Validating the Method's Effectiveness**: Through a specific case of the 1D Burger equation, the paper demonstrates how to use OpenFOAM's data structure to construct initial loss, boundary loss, and residual loss, and validates the proposed framework's effectiveness in predicting CFD simulation results.

3. **Extending to More Complex Computational Domains**: Although the computational geometry used in the paper is relatively simple, the method is general enough to be applied to more complex computational domains, showcasing its potential in practical industrial applications.

4. **Enhancing the Accuracy and Practicality of CFD Simulations**: By integrating OpenFOAM's data structure into deep learning models, the accuracy and practicality of CFD simulations are improved, providing a new approach to addressing complex industrial simulation challenges.

In summary, the core objective of this research is to enhance the performance of deep learning models in handling complex CFD simulations by combining OpenFOAM's data structure with physical information loss functions, and to validate its effectiveness and generality in practical applications.