Study on Automatic Generation of Suspension Model using Machine Learning

Abstract

With the recent development of computers, simulation has been effectively used in the design and development of automobiles. Simulation requires modeling of the object. A large computational load was required when solving a system with complex characteristics in detail or when multiple phenomena are involved. In this study, a method to generate a model with a small computational load using machine learning was investigated for the suspension mechanism, which affects the driving performance of an automobile. A multibody analysis of the suspension was conducted, and the obtained input-output relationships were modeled by machine learning. The multibody model and the machine learning model were incorporated respectively into a vehicle motion analysis to confirm the validity and analysis efficiency of the machine learning model. The predictions of the machine learning model generally agreed with the results of the multibody model, and the machine learning model was able to reproduce the vehicle body motion from the predictions of the machine learning model and significantly reduce the analysis time, indicating the effectiveness of the machine learning model.