Hybrid predictive modeling and heuristic design of surface-modified Biofibre polymer composites for viscoelastic creep resistant applications

The long-term viscoelastic degradation of natural fiber-reinforced polymers remains a significant barrier to their load-bearing structural applications. This study introduces an integrated computational framework combining Artificial Neural Networks (ANN) and a Genetic Algorithm (GA) to model, predict, and optimize the chemical surface treatments of plantain pseudo-stem fiber-reinforced high-density polyethylene (HDPE) composites matrix at 20 percent weight fraction (wt.%) fiber loading. The biofibres were modified via alkaline mercerization in 0.1M, 0.5M and 0.8M Sodium hydroxide (NaOH) and subsequently an acetylation treatment with 5%, 10% and 15% acetic anhydride before thermo-mechanical creep testing under sustained stresses of 35 MPa and 42 MPa across an isothermal gradient of 30℃, 60℃ and 80℃ for 240 hours. The optimized ANN framework successfully mapped the continuous, non-linear creep space across the coupled thermo-mechanical domains. Continuous response surfaces revealed that raising the temperature from 30℃ to 80℃ accelerated primary creep kinetics and increased maximum deformation by 87.5%, highlighting the matrix thermal sensitivity near the glass transition region. Notably, the smooth topology of the predictive profiles confirmed that the dual chemical modification maintained robust fiber-matrix interfacial integrity under all environmental conditions. To transition from empirical observation to intelligent material design, the validated ANN model was coupled with a heuristic GA to autonomously explore the treatment design space. The evolutionary optimization path was mathematically validated via a fourth-degree polynomial regression equation which demonstrates an exceptional coefficient of determination (R²) = 0.99. The framework successfully determined the optimal chemical concentration thresholds required to minimize temporal structural strain, and offers a scalable design methodology for smart manufacturing applications.