The objective of this study is to develop a framework using nonlinear autoregressive model process with exogenous inputs (NARX) neural networks (NNs) to identify the dynamic hysteresis of high load capacity suspension systems. Here, a vertical excitation test is used to simulate various terrains at an oscillation frequency range from 0.1Hz to 10Hz by using NARX NNs. The model results are in good agreement with suspension component oscillation responses that manifest as variations in model order selection as the excitation frequency approaches 7Hz. Furthermore, mapping the models into generalized frequency response functions (GFRFs), elucidates any unanticipated couplings between surroundings and mechanical hysteresis within the suspension. The proposed approach's systematic design procedure is advantageous because it provides a cost-efficient method that achieves precise identification of online data.
