|
| Machine learning based model for constructing basic solutions of IBIEM control equations |
| Received:October 25, 2023 Revised:February 12, 2024 |
| View Full Text View/Add Comment Download reader |
| DOI：10.7511/jslx20231025002 |
| KeyWord:indirect boundary integral equation method virtual wave source artificial neural network particle swarm optimization solving fluctuation problems |
| Author | Institution |
| 刘中宪 |
天津城建大学 土木工程学院, 天津 ;天津市软土特性与工程环境重点实验室, 天津 |
| 黄珑 |
天津城建大学 土木工程学院, 天津 |
| 孟思博 |
天津城建大学 土木工程学院, 天津 |
| 黄振恩 |
天津大学 建筑工程学院, 天津 |
|
| Hits: 50 |
| Download times: 32 |
| Abstract: |
| The indirect boundary integral equation method(IBIEM)relies on empirical judgment and trial calculations to construct the basic solution of the governing equation when solving wave propagation problems,resulting in unstable solutions for wide-frequency scattering.This paper proposes a model for constructing the basic solution of the IBIEM governing equation using particle swarm optimization and artificial neural networks,replacing empirical judgment with data-driven methods to handle the uncertainty in the construction process.The reliability of the proposed model is validated by simulating the scattering of plane SH waves in a two-dimensional canyon using IBIEM.The results show that the proposed model can effectively predict the optimal placement and quantity of virtual wave sources,while balancing computational efficiency and accuracy,significantly improving the stability and efficiency of IBIEM in solving wave propagation problems.The optimal placement and quantity of virtual wave sources are significantly influenced by the incident wave frequency and the geometric conditions of the site,exhibiting non-monotonic characteristics.Empirical-based solutions have poor reliability,while data-driven prediction models have clear advantages.The proposed method can provide references for solving other types of seismic wave propagation problems using IBIEM. |
|
|
|