Tree-based machine learning approach to modelling tensile strength retention of Fibre Reinforced Polymer composites exposed to elevated temperatures

• DOI: 10.1016/j.compositesb.2023.111132
• Publication Date: 2023-12-19T19:36:50Z
• Authors: Chiara Machello; Keyvan Aghabalaei...; Milad Bazli; Ali Hadigheh; Ali Rajabipour; Mehrdad Arashpour; Hooman Mahdizadeh...; Reza Hassanli

Fibre Reinforced Polymer (FRP) composites are susceptible to degradation at elevated temperatures. Accurate modelling of the tensile performance FRP under high-temperature exposure is crucial for their structural integrity. In this study, tree-based models, namely, decision tree, M5P, and random forest methods, utilised model impact temperatures on strength composite materials. A database 787 experimental results established processed train test regression tree models. The temperature, resin glass transition sample thickness/diameter, duration, ambient cooling, fibre-to-resin ratio, fibre orientation, type, manufacturing process were considered as main parameters affecting retention (TSR) after To improve prediction machine learning, Bayesian optimisation 10-fold cross validation (CV) technique used methods. demonstrated accuracy developed models in predicting TSR Feature contribution analysis showed that temperature exerts most significant TSR, with coming next importance. These followed by thickness, respectively. Resin had least contributions observed variations TSR. Examining high enables development precise predictive design guidelines optimal use across industries.