A multi-loop hysteresis model of piezo actuator based on LS-SVM

In order to approximate the behavior of hysteresis nonlinearity which often severely limits the performance of system and improve the prediction accuracy of classical Preisach model (CPM), a least squares support vector machine (LS-SVM) based model is presented in this paper. According to the geometric characteristics of multi-loop hysteresis, the input voltage sequences was combined with output stroke sequence and expand to a matrix as the input of LS-SVM. Then the multi-valued mapping of hysteresis was transformed to one-to-one mapping which enable LS-SVM to approximate the behavior of hysteresis. The prediction accuracy is improved effectively on the condition of finite samples and the default of CPM was avoided. Results of simulation and experiments show that the proposed hysteresis model can exactly describe and predict the multi-valued hysteresis feature compare with bilinear interpolation and has the better generalization ability.