Seal clearance flow field solution based on physics-informed neural network
Abstract
Abstract The clearance flow field of the liquid annular seal is governed by a set of highly nonlinear PDEs. Conventional approaches like the perturbation method are often subject to its applicability, and the CFD method is subject to computational cost. In this paper, PINN is introduced to solve the governing equations. After mathematically modeling the physical field, a loss function is constructed from the governing equations and boundary conditions, so that the output of the DNN approximates the flow field solution. Due to the meshfree property of neural networks and the physics informing, our method evaluates the clearance flow field effectively, which is validated by the perturbation method, and demonstrates great potential in handling similar problems.
Fields of Study & Subjects
Model Reduction and Neural Networks
Neural Networks and Applications
Turbomachinery Performance and Optimization
Publication details
- Journal / source
- Journal of Physics Conference Series
- Volume
- 3174
- Issue
- 1
- Pages
- 012078–012078
- Publication date
- 1 February 2026
- Language
- EN