DSFA-PINN: Deep Spectral Feature Aggregation Physics Informed Neural Network

Solving parametric partial differential equations using artificial intelligence is taking the pace. It primarily because conventional numerical solvers are computationally expensive and require significant time to converge a solution. However, physics informed deep learning as an alternate learns functional spaces directly provides approximation reasonably fast compared solvers. The Fourier transform approach generalized space among various approaches. This work proposes novel neural network that employs operator fundamental building block spectral feature aggregation extrude extended information. proposed model offers superior accuracy lower relative error. We employ one two-dimensional time-independent well time-dependent equations. three benchmark datasets evaluate our contributions, i.e., Burgers’ equation dimensional, Darcy Flow two Navier-Stokes spatial dimensional with temporal dimension datasets. further case study of fluid-structure interaction used for machine component designing process. computation fluid dynamics simulation dataset generated ANSYS-CFX software system regression behavior fluid. Our method achieves performance on all four employed shows improvements baseline. achieve reduced error by approximately 30%, 35%, 20%.