Searching for Optimal Patterns of Magnetoelectric Multi-Phase Composites with Machine Learning Method

Abstract

In this paper, a method of searching for optimal patterns of multiphase magnetoelectric composites by using machine learning algorithm (ML) is proposed. Firstly, we use artificial neural network (ANN) and convolutional neural network (CNN) as ML algorithm models with the database established by the two-dimensional finite element method. Secondly, we study the influence of different network parameters (training data density, number of iterations, batch size) on the prediction accuracy, and establish the ML model which can predict the optimal structure of multiphase magnetoelectric composites effectively and accurately. Finally, the predicted results of ML models and the results obtained by the finite element are compared and analyzed to verify the correctness of the model and the effectiveness of the ML methods in searching for optimal patterns of magnetoelectric composites.