A Hybrid Approach to Model Hysteretic Behavior of PZT Stack Actuators

Abstract

In order to model the hysteretic behavior of piezoceramic actuators, a hybrid model is developed by combining the Preisach concepts with a neural network mapping function. Preisach concepts are utilized in producing a data set for generalization and calculating the final displacements for actuators having nonlocal memory. Although generalization is typically handled by interpolation functions in a traditional Preisach model, these functions can lead to significant errors unless there are sufficient data points around critical regions. In the hybrid model, the generalization of all first-order reversal curves is provided by a single neural network. Since the neural network is essentially a nonlinear mapping function, its functionality is implemented with a fewer number of variables than the Preisach model. In order to account for errors caused by frequency dependency and large input variations, an on-line training technique is also developed. Various comparisons between the outputs of the hybrid model and those of an actual actuator are presented.