Random Forest Regression Analysis for Estimating Dielectric Properties in Epoxy Composites Doped with Hybrid Nano Fillers

Abstract

Bisphenol-A resin epoxy resin composites doped with various concentrations of inorganic hybrid nanofillers, TiO2 + ZnO, were thoroughly examined in the research described in this report for their microwave dielectric relaxation spectroscopy. Ultrasonic dispersion techniques were used to disperse the TiO2 + ZnO hybrid nanofiller into the Bisphenol-A resin. X-ray diffraction (XRD) was employed to ascertain the structural characteristics of the TiO2 and ZnO nanoparticles (NPs) and TiO2 + ZnO hybrid NPs doped neat epoxy (epoxy + hardener). A Vector network analyzer was used to measure the dielectric properties of the hybrid NPs doped neat epoxy in the frequency range from 200 MHz to 20 GHz. Our findings offer important new perspectives on the polarization mechanisms and structural properties of these composite materials. The complex relationship between frequency and dielectric characterization led to measurement complexities that often come with extensive time and cost requirements. This paper presents the random forest regression model as an effective method for predicting the frequency-dependent dielectric constants in composite materials. The main objective of this study was to examine the performance metrics, such as the adjusted R-squared score, root mean squared error (RMSE), and mean absolute error (MAE), across various values of the minimum node size parameter (n min). This study showcases the effectiveness of the RF regression model to accurately and efficiently predict the frequency-dependent dielectric constants in composites. As a result, it eliminates the requirement for laborious and expensive laboratory testing.