Hybrid model based on Preisach and support vector machine for novel dual-stack piezoelectric actuator

Abstract

This paper presents a hybrid model for piezoelectric actuators (PAs), which accounts for hysteresis and dynamic effects. The hybrid model is based on discrete Preisach model and Support Vector Machine (SVM). The Preisach model is generally implemented using a numerical technique basing on the first-order reversal curves (FORCs). The accuracy of the Preisach model mainly depends on the number of FORCs. However it is generally difficult to achieve enough number of FORCs in practice. In order to promote the modeling efficiency, firstly in Preisach modeling process, the storage method of prehistory and interpolation approach of output are improved; then SVM is employed in the work to circumvent the deficiency of the Preisach model resulting from the lack of FORCs, which produces a hybrid model based on the improved Preisach model and SVM. This hybrid model is implemented in calibration of a piezoelectric stack actuator and validated experimentally. Finally, a novel dual-stack PA is introduced and calibrated with this hybrid model. The results demonstrate that the presented hybrid model possesses the advantages both of SVM and Preisach model. In comparison with pure Preisach model, the hybrid model has higher accuracy and is easier to be obtained without increasing the number of FORCs. In addition, the inverse model for PA is derived from the hybrid model, which can be employed in active vibration control of structures.