A hybrid modeling and optimization framework for piezoelectric actuators: integrating a physics-based dynamic model with data-driven techniques

Abstract Accurate and efficient dynamic modeling is crucial for the stable operation and optimization of piezoelectric actuators. However, existing methods often face a trade-off between model accuracy and computational speed. To address the piezoelectric and contact/friction nonlinearities of the V-shaped piezoelectric actuator (VPEA), this paper proposes a hybrid physics-based and data-driven modeling approach. First, an electromechanical coupling physics-based model of the VPEA is established, encompassing key mechanisms such as stator vibration, stator-rotor contact, and rotor dynamics, to generate sufficient and accurate training data. Subsequently, a backpropagation neural network (BPNN) is constructed, and its initial weights and thresholds are optimized using the grey wolf optimizer (GWO), resulting in a GWO-BPNN prediction model. Experimental comparisons demonstrate the superior performance of this model in predicting output speed and input current, achieving coefficients of determination ( R 2 ) of 0.99 and 0.992, respectively. These results significantly outperform models based on random forest, support vector machine, and BPNN optimized by genetic algorithm (GA) or particle swarm optimization (PSO). Furthermore, the computational response time of the proposed model is reduced by over 99% compared to the physics-based model, with the GWO converging 16.7% faster than GA and PSO. Utilizing this model, the influence of key parameters—including excitation voltage, preload, and external load—on the operational characteristics of the VPEA is analyzed. By further integrating the PSO algorithm for operational parameter optimization, a maximum operational efficiency of 26.57% is achieved, and the high-efficiency operating zone is identified. The results confirm that the proposed hybrid modeling method offers both high prediction accuracy and high computational efficiency, providing an effective solution for performance analysis and operational optimization of nonlinear piezoelectric actuators.