Modeling a Magneto-Rheological Fluid-Based Brake Via a Neural Network Method

Abstract

Identifying the model of a magneto-rheological (MR) fluid-based brake is extremely important for designing and controlling a haptic device with hybrid actuation. Therefore, in this study, an Elman Recurrent Neural Network (ERNN) is designed to understand and model a characterization of an MR fluid-based rotational brake. Three important factors that affect the MR brake's performance are chosen as inputs: current, speed, and the first derivative of the input current. The proposed network is trained, and the performance of the network is tested with three different experimental scenarios. Then, the effect of these inputs on the system is investigated. According to the results, it can be said that the designed ERNN is a good candidate for modelling an MR brake.