Parameter optimization for enhancing mechanical properties of wood-plastic composites using artificial neural network with genetic algorithm and taguchi method

Abstract

This study focused on improving the mechanical properties of wood-plastic composites (WPCs), commonly used in construction materials like flooring and roofing. Poor mechanical performance often results from improper material proportions during fabrication. To address this, the research applied to hybrid approach combining the Taguchi method with artificial neural networks (ANN) and a genetic algorithm to optimize key process parameters: thermoplastic content, fiber content, and fiber size. Nine WPC samples were produced using different fiber sizes (0.425, 1.180, 2.000 mm) and plastic-to-fiber ratios (80:20, 70:30, 60:40 wt.%) via compression molding. Results showed that increasing fiber size and content generally improves flexural strength and hardness, but impact strength decreases with larger fiber size. The highest mechanical performance was achieved at optimized parameters (fiber size: 1.23 mm, fiber content: 31.44 wt.%, thermoplastic: 68.56 wt.%), resulting in flexural strength of 63 N/mm², impact strength of 16.8 J, and hardness of 80 HRB. ANN predictions aligned closely with experimental results, and cross-validation confirmed performance improvements. The cross-validation tests improved the flexural strength by 3.6%, the impact strength by 0.9%, and the hardness by 0.6%. Overall, the study provides a foundational approach to enhancing WPCs for structural applications through AI-driven process optimization.