Predictive Modeling of Shielding Effectiveness in Al6061 Composites Using Machine Learning and Network Analyzer Data: A Comparative Study of Fly Ash and Aloe Vera Reinforcements

Abstract

This study explores the predictive modeling of shielding effectiveness in Al6061 and its composites, Al6061+10% Fly Ash and Al6061+10% Aloe Vera, using machine learning techniques. The datasets of these composites were obtained through experimental analysis using a network analyzer, ensuring accurate measurements of electromagnetic properties. By leveraging XGBoost, Random Forest, and Linear Regression models, the research evaluates the impact of material modifications on shielding performance. The results demonstrate that ensemble models, particularly XGBoost, consistently outperform Linear Regression, achieving higher R² values and lower RMSE and MAE across all datasets. The addition of 10% Fly Ash significantly enhances predictive accuracy, with XGBoost and Random Forest achieving near-perfect alignment between predicted and actual values. This improvement is driven by the increased importance of transmission coefficients (S21), as highlighted by SHAP analysis. In comparison, Al6061+10% Aloe Vera shows moderate performance improvements, with SHAP results indicating more complex interactions between features. The analysis reveals that while frequency remains a dominant factor, the role of transmission coefficients grows in importance with material reinforcement. The study underscores the limitations of linear models in capturing non-linear dependencies, reinforcing the necessity of ensemble techniques for composite material predictions. These findings provide a pathway for optimizing composite material design through advanced machine learning models, highlighting the potential of Fly Ash for industrial applications and Aloe Vera for sustainable innovation. The integration of SHAP analysis enhances model interpretability, ensuring reliable feature importance assessment and fostering greater trust in machine learning applications for material science. This research contributes to advancing predictive modeling in composite materials, paving the way for more efficient and sustainable material development.