Investigating the dielectric properties of PMMA/RGO nanocomposites using experimental techniques with artificial neural network ANN Model

Mohamed, R. A.; Elbakry, Mahmoud; Habashy, Doaa; Mohamed, A. S.; Ismail, A. M.

doi:10.21608/ejs.2025.350213.1054

Investigating the dielectric properties of PMMA/RGO nanocomposites using experimental techniques with artificial neural network ANN Model

Document Type : Original Article

Authors

¹ Faculty of Education - Physics department

² faculty of education ain shams university

³ department of physics, faculty of education, ai shams university

⁴ physics, education, Ain shams

⁵ physics, education, Ain Shams

10.21608/ejs.2025.350213.1054

Abstract

The current research introduces a combined investigation using both experimental methods and theoretical model to understand and predict the dielectric behavior of PMMA polymer nanocomposites. Poly (methyl methacrylate) (PMMA)/reduced graphene oxide (RGO) nanocomposite films with varying RGO nano-platelets (NPs) contents are made using the casting process. The dielectric constant ε^', loss ε^'', ac-conductivity σ_ac of PMMA/RGO nanocomposites are investigated in the temperature range (300 K  390 K) and frequency range (100 Hz  1 MHz). σ_ac and the frequency exponent S are interpreted by the correlated barrier hopping CBH theory. The frequency exponent S and charge carrier binding energy W_m in the nanocomposite films exhibit a decrease with increasing temperature and RGO content. ε^', ε^'' and σ_ac of PMMA/RGO nanocomposites depend on both frequency f and temperature T. The study employed ANN as a soft-computing process to model the dielectric behavior of the investigated polymer nanocomposites. The measured experimental datasets served as inputs. The optimized ANN configuration was used to train the model for ε^', ε^'' and σ_ac. ANN simulation results exhibited excellent fitting with the measured experimental data. Notably, the ANN not only accurately predicted experimental measurements (serving as a test step) but also successfully predicted values for unmeasured data points. To evaluate the model's performance, Mean Squared Error MSE was calculated. The consistently low MSE values (below 0.08) indicated a high degree of accuracy. Additionally, the correlation coefficient R provided further confirmation, with its value signifying a strong correlation between the ANN results and their targets.

Keywords