Open AccessFinite Element-Neural Network Hybrid Modeling of Complex Dynamic Systems
Author(s) -
Jianwen Fu,
Hongbin Zhao,
H T Zhang,
Shaojuan Su,
Chenxin Ji,
F Chen
Publication year - 2021
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/701/1/012085
Subject(s) - finite element method , artificial neural network , nonlinear system , computer science , displacement (psychology) , system dynamics , vibration , reduction (mathematics) , control theory (sociology) , structural engineering , engineering , artificial intelligence , acoustics , psychotherapist , psychology , physics , geometry , mathematics , control (management) , quantum mechanics
This study involves a new hybrid modeling method for finite element-neural network complex dynamic systems. Dynamic modeling of components with linear-elastic properties is obtained by using the finite element method (FEM) in complex dynamic systems, while the dynamic characteristics of the nonlinear elements were described by neural networks. They may be connected by force or displacement to generate the dynamic hybrid model of the entire system, which effectively combined the characteristics of finite element modeling and neural network modeling. The relevant characteristics include clear physical structure, high modeling precision, small network scale, and fast training speed. The mode reduction method is applied to reduce the computational load for the response prediction of the hybrid model. Simulations of hybrid modeling for dynamic characteristics in the vibration isolation systems of steel wire rope are performed to demonstrate the effectiveness of the proposed method. This result provides an effective approach for modeling the complex dynamic system with an elastic structure, which includes nonlinear element components.