BROWSE
Detail
Conference Paper
PINN application for compressible gas flow transient analysis
Cited 2 time in 
Share 
Share
- Authors
- Sangjoon Lee, Byung Tak Lee, Seok-Kap Ko
- Issue Date
- 2023-10
- Citation
- International Conference on Information and Communication Technology Convergence (ICTC) 2023, pp.1703-1708
- Publisher
- IEEE
- Language
- English
- Type
- Conference Paper
- Abstract
- Simulating transient flow inside pipeline networks has been an important topic in the field of civil engineering safety. The recent development of the Physics-Informed Neural Network (PINN) shed light to solve this problem by the means of machine learning. In this paper, we apply PINN to a transient compressible fluid pipeline flow problem. We show that PINN requires a specific data normalization to preserve all necessary physics information for accurate training. We compare the PINN prediction with the estimation provided by finite difference method (FDM), which previously has been the main tool to solve such problems. As a result, we obtained a mesh-free PINN model with the difference of less than 0.6% for pressure mapping and less than 2.3% for mass flow mapping, compared to the FDM analysis.
- KSP Keywords
- Civil engineering, Compressible gas flow, Data Normalization, Finite difference method, Flow problem, Mesh-free, Pipeline flow, Pipeline networks, Pressure mapping, Transient Analysis, Transient flow