High Dielectric Design of Polymer Composites by Using Artificial Neural Network

Abstract

Polymer-based composites with a high dielectric property have shown great potential in electrical energy storage applications. It is important to predict the dielectric constant in designing polymer composites, but it is costly and time consuming. In this study, dielectric properties of various polymer composites have been predicted by using an artificial neural network (ANN) model trained with hundreds of experimentally measured data. Eight variables such as the dielectric constant of matrix, filler, and shell, the diameter of filler, the volume fraction of filler, the dimension of filler, the thickness of shell, and the frequency were considered. To improve the prediction accuracy, hyper parameters of the ANN model were optimized through the hyperband method. Using the ANN model, we demonstrated the correlation between the dielectric constant of polymer composites and the variables. The ANN model predicted the dielectric constant with a coefficient of determination (R2) of 0.97. Furthermore, the ANN model shows good performance to predict dielectric constant at various frequencies (spanning from 100 Hz to 100 kHz). Hence, we present that the AI-based prediction model using ANN method can be helpful in designing the polymer composites with desired properties.