Active suspension system control using neural network model to reduce passengers' whole-body vibration

Abstract

Whole-body vibration has negative effects on passengers' comfort, health and safety. The objective of this study is to reduce these negative effects by optimising the control of an active suspension system according to the simulated behaviour of suspension system and collected data on the effects of vibration on passengers' body. Quarter-car model is created to simulate the behaviour of the suspension system in MATLAB. The parameters of suspension system components, both passive and active, are varied together with road profile parameters to obtain the set of correlations between these parameters and vehicle response. This data set is later used to train a backpropagation neural network which will serve as a controller for the active suspension system. Active suspension system controller created this way will operate based on the criteria of lowering the negative effects of whole-body vibration. Desirable vibration parameters are defined according to the collected data from the literature. The ability of the trained neural network to control the suspension system in a desired way is tested by simulating the quarter-car model behaviour in a series of predefined driving cycles. These cycles are created in such way to take into account a different road profiles (including the type and quality of road surface and road gradient) and different traffic conditions (urban and extra urban). When given the desired range of whole-body vibration frequencies, neural network model created for the purpose of this paper is efficient in more than 92% of operating time for all predefined driving cycles. When changing the desired range of vibration frequencies during driving, time needed for suspension system to adapt never exceeds 0.43 seconds. The effectiveness of the presented method is limited by the operating range of suspension system active components. Another limitation is the use of quarter-car model – further investigation will incorporate the tests on physical model, and later, on a real vehicle. The presented paper introduces the use of pre-collected data on vibration effect on human body as a guide for controlling the active suspension system. Additionally, there is a possibility to set the suspension system to conform to individual requests of the passengers. The task of the methodology presented is to reduce the negative effects of vibration induced by road unevenness on passengers' bodies. Neural Network model proved as an efficient tool for the optimisation of suspension system by retaining the vibration amplitudes and frequencies in a range least harmful to the passengers.