Simulation of a High-Speed Train against a Turbulent Air Flow using Computational Fluid Mechanics Method and Multi-Layer Feed-Forward Neural Network Algorithm
In this study, the aerodynamic performance of a high-speed train against a turbulent air flow is examined numerically from two approaches. First, using computational fluid dynamics, the parameters of aerodynamics and fluid flow are analyzed and then, using Multi-Layer Feed-Forward Neural Network (ML...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | fas |
| Published: |
Semnan University
2021-12-01
|
| Series: | مجله مدل سازی در مهندسی |
| Subjects: | |
| Online Access: | https://modelling.semnan.ac.ir/article_5737_79ef57a03d85a64dcfb97fa86e20fae3.pdf |
| Summary: | In this study, the aerodynamic performance of a high-speed train against a turbulent air flow is examined numerically from two approaches. First, using computational fluid dynamics, the parameters of aerodynamics and fluid flow are analyzed and then, using Multi-Layer Feed-Forward Neural Network (MLFFNN) Algorithm, a prediction and comparison with the obtained values from the CFD analysis are presented. To achieve this, using Reynolds-Averaged Navier-Stokes (RANS) method with 𝑘-𝜔 (SST) turbulence model, an incompressible turbulent air flow around a high-speed train model by OpenFOAM CFD Software is simulated. In this research, some of the significant and key parameters of fluid flow and aerodynamics as velocity, pressure, streamlines, flow structure, pressure coefficients, drag, lift and side forces for some yaw angles of wind movement and velocity changes are analyzed and compared. In the following, the Multi-Layer Feed-Forward Neural Network which is modified with various data is applied for prediction of the output of the problem. Accordingly, the aerodynamic drag, lift and side forces for the yaw angles of wind movement and velocity changes by this algorithm method are obtained and compared with the obtained results from CFD analysis. The comparisons indicate an appropriate similarity between the CFD data and the used MLFFNN one. |
|---|---|
| ISSN: | 2008-4854 2783-2538 |