A Hybrid Master-Slave Genetic Algorithm-Neural Network Approach for Modeling a Piezoelectric Actuator

Abstract

This work presents an approach for developing the model of a smart fin dynamics that is activated by a fully-enclosed piezoelectric (PZT) bimorph actuator, which is created by bonding two Macro Fiber Composites (MFCs). Observing the dynamics of the fin indicates that the use of a linear dynamic model does not adequately describe its behavior. An earlier work proposed incorporating a proportional damping matrix as well as Bouc-Wen hysteresis model and backlash operators to create a more accurate model. However, the number of parameters describing the expanded model is large, which limits its use. Therefore, there is a need for a different approach for developing an alternative model of the fin. In this work, a hybrid master-slave Genetic Algorithm (GA)-Neural Network (NN) model is proposed to identify the optimal set of parameters for the damping matrix constants, the Bouc-Wen hysteresis model and the backlash operators. A total of nine sinusoidal input voltage cases that resemble a grid of three different amplitudes excited at three different frequencies are used to train and validate the model. Three input cases are considered for training the NN architecture, connection weights, bias weights and learning rules using GA. The NN consists of three layers: an input layer that has two nodes for the amplitude and the frequency of the input voltage, an output layer that has seven nodes for the backlash, hysteresis, and damping operators, and a hidden layer that is free to have any number of nodes between two and nine. The GA constantly performs natural selection of chromosomes that propagate best compilation of NN parameters. Simulation results show that the proposed model can predict the damping, hysteresis and backlash of the smart fin–actuator system under various operational conditions.