A novel approach to model magneto-rheological dampers using EHM with a feed-forward neural network

Abstract

This paper proposes a novel method for modelling magneto-rheological (MR) dampers. It uses an elementary hysteresis model (EHM) with a feed-forward neural network (FNN) to capture hysteresis characteristics of an MR damper, and another FNN to determine the current gain. These parts can be trained separately, thus reducing the size of the training dataset. The inputs of the proposed model include velocity, acceleration, and current to estimate the generated damping force. Unlike previous FNN models, this model does not require force sensor inputs. Simulation results show the high performance of the proposed EHM-based FNN when compared to conventional methods such as a recurrent neural network.