Predicting Water Absorption in <i>Yucca Treculeana</i> L./HDPE Biocomposites: Optimization Using GA-ANN and RSM Techniques

This study utilized cellulose fibers from waste leaves of Yucca treculeana L. to reinforce a high-density polyethylene (HDPE) matrix, aligning with a sustainable production approach that recovers agricultural waste. After complete extraction, the fibers underwent a mild chemical treatment (3% sodium bicarbonate for 4 hr) to improve bonding with the polymer matrix and remove surface contaminants. The primary aim was to examine how different fiber percentages (10, 15, 20, and 30%) affect the dynamic behavior of biocomposites, with particular focus on their absorption kinetics and associated diffusion mechanisms. The absorption characteristics were comprehensively analyzed and modeled using two innovative techniques: response surface methodology (RSM) and artificial neural networks (ANN), enhanced by genetic algorithm optimization. Results indicate that the ANN method surpasses RSM in accuracy and robustness, achieving very high correlation coefficients (0.9979 in training, 0.9777 in testing, and 0.9986 in validation) between experimental results and model predictions. This demonstrates the potential for natural fiber/HDPE composites in manufacturing industrial biocomposites, utilizing plant waste, and supporting more environmentally friendly production processes.