Optimized Machine Learning Prediction and RSM Optimization of Mechanical Properties in Boiled Eggshell Filler-Added Biocomposites

This study aims to evaluate the impact of fiber loading (20%, 30%, and 40%), eggshell filler content (5%, 10%, and 15%), and NaOH treatment (0%, 5%, and 10%) on the mechanical properties of roselle fiber polyester composites filled with boiled eggshell powder. Using a complete factorial design, 27 laminates were analyzed to develop regression models correlating tensile, flexural, and impact strengths with these process variables. The key objectives were to identify and predict the significant factors affecting mechanical properties and to optimize these properties using hyperparameter-tuned Artificial Neural Networks (ANNs) and Response Surface Methodology (RSM). ANOVA revealed that filler content and NaOH treatment significantly influence mechanical strength, with filler content being the most impactful, contributing around 80%. This study identified the optimal settings for roselle fiber loading at 32 wt%, filler content at 13 wt%, and NaOH treatment at 6%, achieving remarkable results: a tensile strength of 73.04 MPa, a flexural strength of 139.6 MPa, and an impact strength of 3.04 J. Scanning electron microscopy (SEM) confirmed enhanced fiber-matrix bonding with NaOH treatment. Validation against experimental data demonstrated the predictive accuracy of ANN and RSM models, with the ANN model showing average percentage errors of - 0.756% for tensile strength and 0.50% for flexural strength, and 0.16% for impact strength. The significance of this research lies in its contribution to the development of sustainable and high-performance composite materials from agricultural waste, promoting eco-friendly and innovative applications.