Surrogate Modeling of Fluid Flow at Different Reynolds Numbers Using Physics-informed Deep Operator Network
Please login to view abstract download link
In this study, we demonstrate the ability of a physics-informed deep operator network (PI-DeepONet) to predict fluid flows at different Reynolds numbers through a single training process. By leveraging the operator learning capabilities of Deep Operator Networks (DeepONets), we incorporate the Reynolds number as a direct input parameter of the neural network model, allowing the model to learn and generalize fluid behavior at different Reynolds numbers in a single training process. Furthermore, we embed the Navier-Stokes equations into the loss function following the concept of Physics-Informed Neural Networks (PINNs). This enables learning without the need for large volumen of high-quality training data, or even without any data at all. Compared to traditional data-driven approaches, this data-independent approach enhances model reliability, resulting in a powerful surrogate model for predicting fluid flow across a wide range of Reynolds numbers.
