Experimental Study on the Dynamic Modeling and Optimization of Suspension Mechanism Using Neural Network

EasyChair Preprint 6627

Abstract

In this paper, vehicle’s suspension dynamics modeling is presented using a nonlinear one quarter car model to optimize ride & handling criteria. Experimental tests have been performed in order to achieve the characteristic damping curve (Force-Velocity) of the shock absorber . An asymmetric two-stage damper model of the shock absorber is incorporated in suspension dynamics model and is compared to that of dynamic simulation software, MD ADAMS. Results of the model, subjected to a half-sine road profile input, have been investigated and a good agreement between the two models has been shown. In the next step, an optimization process is performed on the vibration response of the suspension model in order to obtain the optimized damping characteristic curve . The target of the optimization consists in minimizing the acceleration of the vehicle body as well as the displacement of the vehicle related to the tire. The results show an improvement in the optimized shock absorber’s performance in compare to the ordinary shock absorber.  Finally, due to the long time of the optimization process, a neural network is employed in order to represent the optimization. This network’s input is the experimental suspension system’s coefficient and its output is the optimized values. This neural network shows to be a good replace for the lengthy traditional optimization. The application of this neural network can contribute to the process of design of a shock absorber in industry.

Keyphrases: Suspension mechanism, machine learning, neural network, optimizing vibration modeling, shock absorber

@booklet{EasyChair:6627,
  author    = {Iman Yahyapour and Mehran Ghalami and Elnaz Balazadeh and Mehdi Tale Masouleh and Behzad Moghadam},
  title     = {Experimental Study on the Dynamic Modeling and Optimization of Suspension Mechanism Using Neural Network},
  howpublished = {EasyChair Preprint 6627},
  year      = {EasyChair, 2021}}