Dynamic rate-dependent hysteresis modeling and trajectory prediction of voice coil motors based on TF-NARX neural network

Voice coil motor (VCM) has obvious rate-dependent hysteresis characteristics, which means that when the frequency of the input signal changes, the travel distance and the shape of the hysteresis loop will change significantly. Based on the Nonlinear Auto-Regressive with Exogenous Inputs (NARX) neural network model, a rate-dependent hysteresis model consisting of a transfer function sub-model of the VCM and a NARX neural network sub-model is proposed for VCM in this paper. Different from the commonly used rate-dependent operator model, the proposed model has a relatively simple mathematic format. By introducing the transfer function of VCM, the initial prediction of the amplitude and phase shift is realized dynamically, which ensures the nonlinear fitting effect of the NARX neural network. Comparisons of the model responses with the measured data under different frequencies of input current signals indicate that the proposed model can dynamically describe the nonlinear rate-dependent hysteresis of VCM with very high accuracy. On this basis, the inverse model is designed by adopting the method of direct inverse, and the effectiveness of the inverse model in trajectory tracking is preliminarily verified by simulation.