Raocp / PINN-laminar-flow

Physics-informed neural network for solving fluid dynamics problems

PINN-laminar-flow

Physics-informed neural network (PINN) for solving fluid dynamics problems

Reference paper

This repo include the implementation of mixed-form physics-informed neural networks in paper:

Chengping Rao, Hao Sun and Yang Liu. Physics-informed deep learning for incompressible laminar flows.

• This paper has been published by TAML, those who has access to Elsevier database can refer to https://www.sciencedirect.com/science/article/pii/S2095034920300350 for camera-ready version.

Description for each folder

• FluentReferenceMu002: Reference solution from Ansys Fluent for steady flow;

• PINN_steady: Implementation for steady flow with PINN;
• PINN_unsteady: Implementation for unsteady flow with PINN;

Results overview

Steady flow past a cylinder (left: physics-informed neural network; right: Ansys Fluent.)

Transient flow past a cylinder (physics-informed neural network result)

Note

• These implementations were developed and tested on the GPU version of TensorFlow 1.10.0.