Modeling of Magneto-rheological Fluid Damper Employing Recurrent Neural Networks

Abstract

Due to inherent nonlinear behaviors of magneto-rheological (MR) fluid dampers, one of challenges for utilizing effectively these devices as actuators to control vibration of mechanical system is to develop accurate models. A recurrent neural networks, with 3 input neurons and 1 output neuron in input layer and out layer respectively and 7 recurrent neurons in the hidden layer, is used to simulate behaviors of automotive MR fluid damper to develop control algorithms for suspension systems. The recursive prediction error algorithms are applied to train the recurrent neural networks using test data from lab where the MR fluid dampers were tested by the MTS electro-hydraulic servo vibrator system. Training of recurrent neural networks has been done by means of recursive prediction error algorithms presented in this paper and data generated from test above-mentioned. In comparison with experimental results of MR fluid damper, the recurrent neural networks are reasonably accurate to depict performances of MR fluid damper over a wide range of operating conditions