摘要
传统的结构动力学方程求解方法通常基于微分方程求解、数值方法和模拟技术,但存在计算复杂度高和收敛速度慢的问题。为寻求新的解决途径,文中提出了一种基于循环神经网络(RNN)的新方法,用于模拟和解析结构的动态反应。通过将RNN应用于简单的结构动力学案例,成功实现了对结构动力学方程的预测和模拟。结果表明模型平均相对误差均在10-4,该模型准确捕捉了结构的动态特征,展现了出色的性能。
Traditional methods for solving structural dynamic equations are usually based on differential equations,numerical methods,and simulation techniques,suffering from high computational complexity and slow convergence.In order to seek a new solution path,we propose a new method based on recurrent neural networks(RNN)for model⁃ing and resolving the dynamic response of structures.By applying RNN to a simple case of structural dynamics,we successfully realize the prediction and simulation of structural dynamic equations.The results show that the average relative errors of the model are all in the range of 10-4,and the model accurately captures the dynamic characteristics of the structure and demonstrates excellent performance.
作者
赵铎阳
曾森
ZHAO Duoyang;ZENG Sen(School of Civil Engineering,Qingdao University of Technology,Shandong Qingdao 266525,China)
出处
《低温建筑技术》
2024年第5期69-73,共5页
Low Temperature Architecture Technology
关键词
结构动力学
循环神经网络
微分方程
structural dynamics
recurrent neural network
differential equation
作者简介
赵铎阳(1999-),男,河北保定人,硕士研究生,现从事土木工程信息化方向研究;通信作者:曾森(1983-),男,广州人,副教授,现从事研究土木工程信息化、计算力学方向研究。
