• CN: 11-2187/TH
  • ISSN: 0577-6686

Journal of Mechanical Engineering ›› 2022, Vol. 58 ›› Issue (9): 136-146.doi: 10.3901/JME.2022.09.136

Previous Articles     Next Articles

Dynamic Modeling and Analysis of Rigid Multi-body System with Clearance Joints Based on Deep Learning

LIU Zhen1,2, HU Sanbao1,2, HU Junhua3, XIANG Chao3   

  1. 1. Hubei Provincial Key Laboratory of Advanced Technology of Automotive Components, Wuhan University of Technology, Wuhan 430070;
    2. Hubei Provincial Collaborative Center for Automotive Component Technology, Wuhan University of Technology, Wuhan 430070;
    3. Wuhan Second Ship Design and Research Institute, Wuhan 430064
  • Received:2021-05-07 Revised:2022-01-08 Online:2022-05-05 Published:2022-06-23

Abstract: The contact force of the joint in the mechanical system is strongly nonlinear, which is influenced by the amount of clearance in the joint, the material property of the parts of the kinematic pair and the contact state during movement. Traditional theoretical models of friction mainly focus on the universality of the description of friction phenomenon, but it is difficult to accurately describe the nonlinear characteristics of friction in the motion. Based on the data obtained from the physical prototype test, the neural network model of the nonlinear contact force of the clearance joint is established by using the deep learning method. The contact friction data set is generated by the friction experiment, and the contact impact force data set is generated by the mixed model of revolute hinge clearance. After training and testing, a neural network model describing the contact dynamics of revolute clearance hinge is obtained. Combined with Lagrange equation, a "rigid-body-to-clearance-to-rigid-body" dynamic model of a slider crank mechanism with clearance joint is established. The dynamic responses of the key parameters of the system are obtained through simulation and compared with the results of physical tests. The correctness of the clearance joint model based on deep learning method is verified, which provided a feasible idea for the application of deep learning method in the direction of nonlinear system dynamics modelling.

Key words: joint clearance, friction experiment, nonlinear contact force model, multi-body system, dynamics, deep learning

CLC Number: