On 11/9/2024, ZHOU WEN, Okamoto/Miwa lab, Department of Nuclear Engineering and Management, (D3), received “The 5th Tome Research Institute Young Researchers Essay Competition” at Tome Co., Ltd. Tome Research Institute.

〈Name of award and short explanation about the award〉

I received an award at the 5th Tome Research Institute Young Researchers Essay Competition for my paper titled “Advancing Fluid Dynamics Simulations: A Comprehensive Approach to Optimizing Physics-Informed Neural Networks.”

This research aimed to enhance the performance of physics-informed neural networks (PINNs) for fluid dynamics problems by proposing improvements in sampling, balancing of loss terms, and optimization of hyperparameters. As a result, it was demonstrated that high-precision fluid analysis can be achieved without relying on traditional CFD simulations.

〈About awarded research〉

Advancing Fluid Dynamics Simulations: A Comprehensive Approach to Optimizing Physics-Informed Neural Networks Abstract　Flow modeling based on physics-informed neural networks (PINNs) is emerging as a potential AI technique for solving fluid dynamics problems. However, conventional PINNs encounter inherent limitations when simulating incompressible fluids, such as difficulties in selecting the sampling points, balancing the loss items, and optimizing the hyperparameters. These limitations often lead to non-convergence of PINNs. To overcome these issues, an improved and generic PINNs for fluid dynamic analysis is proposed. This approach incorporates three key improvements: residual-based adaptive sampling, which automatically samples points in areas with larger residuals; adaptive loss weights, which balance the loss terms effectively; and utilization of the differential evolution (DE) optimization algorithm. Then, three case studies at low Reynolds number, Kovasznay flow, vortex shedding past a cylinder, and Beltrami flow are employed to validate the improved PINNs. The contribution of each improvement to the final simulation results is investigated and quantified. The simulation results demonstrate good agreement with both analytical solutions and benchmarked computational fluid dynamics (CFD) calculation results, showcasing the efficiency and validity of the improved PINNs. This PINNs has the potential to reduce the reliance on CFD simulations for solving fluid dynamics problems.

〈Your impression & future plan〉

I am truly honored to have received an award at the 5th Tome Research Institute Young Researchers Essay Competition. Through this research, I have been able to explore new possibilities in fluid dynamics simulations using physics-informed neural networks (PINNs), which brings me great joy and satisfaction. However, this achievement is just the first step. I aim to further expand the potential for improvements and applications in this field. In particular, I aspire to contribute to enhancing CFD simulation efficiency and advancing fluid analysis technology through AI. I will continue my research endeavors with the hope of producing valuable outcomes not only for academia but also for industry.