Optimizing High‐Temperature Performance of UV-treated polypropylene fibre-reinforced cementitious composites using hybrid machine learning and ensemble AI techniques

The residual compressive strength (RCS) of cementitious composites under high-temperature exposure is a critical parameter in structural fire safety design. Polypropylene fibres (PPF) are widely used to mitigate explosive spalling but often reduce mechanical strength due to poor fibre-matrix bonding. This study proposes a novel hybrid Artificial Intelligence (AI) framework integrating ultraviolet-treated polypropylene fibres (UV-TPPF) with advanced ensemble machine learning to predict and enhance RCS. An experimental dataset of 180 cement paste samples containing 2%, 3%, and 4% UV-TPPF and non-treated PPF was tested at temperatures ranging from 20°C to 350°C. Four standalone AI models were developed and subsequently combined using linear and nonlinear ensemble techniques. Finally, a hybrid Random Forest (RF) ensemble was integrated with the best-performing model to further improve prediction accuracy. The results demonstrated that UV-TPPF significantly improved RCS, with up to 31.9% enhancement at 200°C. The hybrid RF-HW model achieved superior predictive accuracy (NSE = 0.9566, CC = 0.9781), demonstrating the effectiveness of the proposed framework. This research presents a robust AI-driven methodology for optimizing fibre-reinforced cementitious composites under thermal loading, contributing to safer and more durable infrastructure.