Mathematical Modeling and ANN-Based Optimization of Electrical Conductivity in Fe₃O₄/rGO Nanocomposites Derived from Corncob Waste

Telekomunikasi Militer

Abstrak

In this study, a mathematical modeling and artificial neural network (ANN)-based optimization approach was developed to predict and enhance the electrical conductivity of Fe₃O₄/rGO nanocomposites synthesized from corncob waste. The Fe₃O₄ nanoparticles were uniformly anchored on reduced graphene oxide (rGO) sheets derived from bio-waste precursors to form a conductive hybrid structure. Experimental data comprising 150 samples were used to train a feed-forward backpropagation ANN model with a 4–8–4–1 architecture. The model effectively captured the nonlinear relationships among Fe₃O₄ content, rGO composition, temperature, and reaction time with high predictive accuracy (R² = 0.9898 and RMSE = 0.0452). The training and validation loss curves confirmed stable convergence of the model. The ANN-based response surface analysis revealed that an optimal combination of 12 wt.% Fe₃O₄, 8 wt.% rGO, and a synthesis temperature of 700 °C resulted in maximum conductivity. This study provides a quantitative and interpretable framework bridging experimental synthesis with data-driven optimization for bio-derived nanocomposites.