Physics-Informed Data-Driven Simulation of Constrained Multibody Systems

EasyChair Preprint 13605

Abstract

We describe a framework that can integrate prior physical information, e.g., the presence of kinematicconstraints, to support data-driven simulation. Unlike other approaches, e.g., Fully-connected NeuralNetwork (FCNN) or Recurrent Neural Network (RNN)-based methods that are used to model the sys-tem states directly, the proposed approach embraces a Neural Ordinary Differential Equation (NODE)paradigm that models the derivatives of system states. A central part of the proposed methodology is itscapacity to learn the multibody system dynamics from prior physical knowledge and constraints com-bined with data inputs. This learning process is facilitated by a constrained optimization approach, whichensures that physical laws and system constraints are accounted for in the simulation process. The mod-els, data, and code for this work are available at https://github.com/jqwang2373/PNODE-for-MBD.

Keyphrases: Multibody System Dynamics, Neural Ordinary Differential Equation, Physics-informed data-driven simulation, constrained optimization

@booklet{EasyChair:13605,
  author    = {Jingquan Wang and Shu Wang and Huzaifa Mustafa Unjhawala and Jinlong Wu and Dan Negrut},
  title     = {Physics-Informed Data-Driven Simulation of Constrained Multibody Systems},
  howpublished = {EasyChair Preprint 13605},
  year      = {EasyChair, 2024}}