Abstract
Engineering design problems often involve solving parametric PartialDifferential Equations (PDEs) under variable PDE parameters and domaingeometry. Recently, neural operators have shown promise in learning PDEoperators and quickly predicting the PDE solutions. However, training theseneural operators typically requires large datasets, the acquisition of whichcan be prohibitively expensive. To overcome this, physics-informed trainingoffers an alternative way of building neural operators, eliminating the highcomputational costs associated with Finite Element generation of training data.Nevertheless, current physics-informed neural operators struggle withlimitations, either in handling varying domain geometries or varying PDEparameters. In this research, we introduce a novel method, the Physics-InformedGeometry-Aware Neural Operator (PI-GANO), designed to simultaneously generalizeacross both PDE parameters and domain geometries. We adopt a geometry encoderto capture the domain geometry features, and design a novel pipeline tointegrate this component within the existing DCON architecture. Numericalresults demonstrate the accuracy and efficiency of the proposed method. All thecodes and data related to this work are available on GitHub:https://github.com/WeihengZ/Physics-informed-Neural-Foundation-Operator.