Exploring Data-Efficient Physics-Informed Neural Networks for Fluid Simulation

Computational Fluid Dynamics (CFD) is crucial in engineering applications such as aerospace and bioengineering, but it often require substantial computational resources. Learning-based simulation methods have the potential to accelerate the computation process but may require a large amount of data. This paper addresses the challenge of data-intensive demands in learning physics-based fluid simulation model by investigating the data efficiency of Physics-Informed Neural Networks (PINNs) for fluid dynamics. It quantifies the minimal data requirements for training PINNs, highlighting their potential to overcome the limitations of traditional CFD methods in resource-constrained environments. By establishing data efficiency benchmarks, our research supports the broader applicability of PINNs across various industries, including aerospace and environmental monitoring.