Data-driven prediction of mechanical properties in recycled fibre-reinforced polymer composites: Integrating machine learning with material–processing feature importance analysis

The performance of polymer components made from recycled or reused materials is strongly influenced by material composition, processing routes, and manufacturing parameters. This study presents an integrated experimental–computational framework to optimize and predict the tensile properties of glass fibre-reinforced recycled polypropylene (GF-rPP). The material was produced using twin-screw extrusion and injection moulding. Glass fibre content, recycled polypropylene proportion (rPP), virgin polypropylene content, additive content, screw speed, extruder flow rate, and cooling conditions were systematically varied. These factors were used to establish quantitative links between thermomechanical processing and mechanical performance. A comprehensive experimental dataset was analysed using four machine learning (ML) models. The Artificial Neural Network (ANN) achieved the highest predictive accuracy (R2 > 0.85) for both Young's modulus and tensile strength. Feature-importance analysis showed that glass fibre content was the most influential factor for stiffness and elongation at break. However, rPP content was the most influential factor on tensile strength. Among processing parameters, extruder flow rate had the greatest impact, while other parameters played smaller roles. This combined experimental and ML-based approach provides a powerful method for optimizing the performance of recycled composites. It enables data-driven material selection and process tuning. Overall, the methodology supports the sustainable development of high-performance polymer composites by enhancing material efficiency and product performance.