Using Physics-Informed Neural Networks for Solving Navier-Stokes Equations in Complex Scenarios
Using Physics-Informed Neural Networks for Solving Navier-Stokes Equations in Complex Scenarios
Using Physics-Informed Neural Networks for Solving Navier-Stokes Equations in Complex Scenarios
Additional material videos: from left to right: OpenFoam, our Solution,
the estimated error in the flow field (please note that the error has a different scale)
Multi Layer Perceptron vs Modified Fourier Network (Section 6.2)
Multi Layer Perceptron
>
Modified Fourier Network
Results on different layer widths (Section 6.4)
Layer widths = 256
Layer widths = 512
Results on different piece geometries (Section 6.5)
Single Box
Circular Cylinder
Three Boxes
Results with fine-tuning, 60000 epochs (Section 7.1)
Single Box to Circular Cylinder
Single box to Three Boxes
Results with multi resolution learning (Section 7.2)