F2010C126
Design and Implementation of an Active Suspension System with Terrain Preview
With a conventional passive suspension system the ride quality is seriously impaired on rough terrain when the suspension working space is fully used up and the bump-stops come into play. This is especially the case with military and specialist off-road vehicles, where the EU directive has become a significant challenge with the requirement for higher speeds. One means of overcoming this problem is the use of some form of adjustable or active element within the suspension system, which can be further improved if combined with terrain preview. In this work a suspension system incorporating an active element and terrain preview is investigated using a whole vehicle model using the simulation tools CarMaker and Matlab/Simulink. The main problem addressed here is the selection of the preview gains, which relate to the terrain characteristics and whether there is a high probability of bump-stop contact. The selection of the preview gains is performed using a Neural Network (NN) that has been trained to identify the probability of bump-stop contact from the terrain profile including random and discrete features. Initially a non-linear model is formed of a single wheel station incorporating an active suspension actuator, consists of a hydraulic cylinder in series with a mechanical spring and damper valve. This model is validated using an experimental active suspension facility. An active suspension system controller is designed using this model and implemented on the test facility using the rapid prototyping tool dSpace. The performance of the suspension system is investigated in both the frequency and time domains to verify its performance. Using the model of the active suspension system, training data is generated using repeat simulations over different types of terrain. This data is then used to train a NN to perform the task of selecting the terrain preview controller gains. The NN selected for this study is the Multi-Layered Perceptron (MLP). The final stage of this work takes the active unit and controller and implements them in CarMaker using the Matlab/Simulink environment to investigate the performance of the controller on a whole vehicle model. The work concludes with a discussion of the results and conclusions are drawn.
This abstract is supplemented by a PDF, which can be viewed here.