内嵌物理知识网络驱动的振动离散模型求解方法研究  

Solving research on vibration discrete models based on PINN

作　　者：赵科炜 何国林[1] ZHAO Kewei;HE Guolin(School of Mechanical and Automotive Engineering,South China University of Technology,Guangzhou 510641,China)

机构地区：[1]华南理工大学机械与汽车工程学院,广州510641

出　　处：《重庆理工大学学报(自然科学)》2025年第3期172-176,共5页Journal of Chongqing University of Technology:Natural Science

基　　金：国家自然科学基金项目(52075182)。

摘　　要：为探究内嵌物理知识神经网络(physics-informed neural networks,PINN)在动力学系统求解方向的应用,从振动力学角度出发,对网络输入输出以及损失函数进行修改,引入残差网络,实现传统PINN的改进。以二自由度振动离散系统动力响应分析问题和系统识别问题求解为例,验证了改进PINN在振动力学模型正反问题精确求解方面的有效性,可为采用神经网络求解复杂动力学模型的相关研究提供参考。Currently,with the growing sophistication of mechanical equipment,solving dynamic model becomes ever more pressing.Fortunately,machine learning provides new ways for solving complex problems.To explore the applications of Physics-informed neural networks(PINNs)in solving dynamic systems,from the perspective of vibration mechanics,the input/output and loss functions of the network are first modified.Then,the residual network is introduced to improve traditional PINNs.Taking the system dynamic response analysis and system identification of the two-dimensional vibration discrete system as an example,our results show the improved PINN accurately solves forward and inverse problems of vibration mechanics model,providing some references for solving the complex dynamic model by employing the neural network.

关 键 词：内嵌物理知识网络 系统动力响应分析 系统识别 

分 类 号：TH113.1[机械工程—机械设计及理论]