Abstract
Magnetorheological (MR) semi-active dampers have proven to be a practical evolution from conventional passive dampers. Their shares keep rising in the markets of automobiles and structures due to their small size, relatively cheap price and fast response time. The main challenge towards effective usage of MR dampers is modeling their highly nonlinear hysteresis behavior. Several attempts have been done to model the behavior from the simple Bingham model to the Phenomenological Bouc-Wen model, through attempting to use non linear problem solving techniques such as Neural Networks, fuzzy logic and lazy learning to provide more accurate models. In this paper, a non-parametric, neural networks based inverse model is proposed. The model results will be validated against experimental data. The results show smaller errors than most of the currently available models. It also solves problems such as speed of response, generality and the ability to invert of the model, all necessary to practically incorporate the model in an automobile control system.
Keywords: Magnetorheological, MR damper, inverse model, neural networks, frequency