Controlling performance of semiactive suspension with two methods of fuzzy-control and machine-learning

Abstract

Two different methods of Fuzzy-control and Machine-learning are proposed to control the vehicle’s semiactive suspensions. To control performance of semiactive suspensions using machine learning (SS-ML) and semiactive suspensions using fuzzy control (SS-FC), a half-vehicle model has been established to calculate and simulate vibration equations under random road surfaces from ISO A-class to F-class. Via map of control rule data in SS-FC established at roads ISO A-class, B-class, …, and F-class, SS-ML’s Neuro-Adaptive-Learning has been trained for learning these control rules. The results obtained indicate that under the same road surface excitation of ISO C-class, the control performance of SS-FC and SS-ML is equivalent., and the vehicle’s comfort level using both SS-FC and SS-ML is very well improved in comparison with passive suspensions without control (PS-WC) of the vehicle. Under a mixed road surface from ISO A-class to F-class and a change range of the vehicle’s moving velocity from 2.5 m s−1 to 35m s−1 used for simulation, the vehicle’s comfort level using SS-ML is better than vehicle’s comfort level using SS-FC. Especially, the root mean square values of displacements and accelerations in vertical and pitch directions of the vehicle body with SS-ML are smaller than that of SS-FC by 13.4%, 23.2%, 20.7%, and 14.3%, respectively. Therefore, the control performance of SS-ML is better than SS-FC, and it should be used to control the vehicle’s semiactive suspensions for enhancing the comfort level.