Prediction of the electromechanical properties of a piezoelectric composite material through the artificial neural network

This paper presents an approach to evaluate the effective properties of the 1–3 composite material formed of piezoelectric fibre and matrix material using an artificial neural network. Firstly, finite element method (FEM) is used to determine the effective elastic, dielectric, and piezoelectric characteristics of composite materials, and the results are compared with classic micromechanical models. Predicting the electromechanical properties of a composite material using the FE method is computationally costly and therefore a computational approach based on an artificial neural network (ANN) is developed to reduce the time-consuming nature of traditional micromechanical models. A large volume of training data is necessary for good prediction from ANN models. To accomplish this, 280 sample data sets are generated utilizing 10 distinct piezoelectric fibre materials, 4 distinct matrix materials, and their combinations with varying volume fractions. An effective ANN model with the Tanh activation function and 12 neurons in the hidden layer was found to have a 99.999% prediction accuracy across all data sets when compared to the FEM model. The current study demonstrates that an artificial neural network can significantly reduce the time and cost involved in designing hybrid piezo-composite materials.