Using machine learning to predict the mechanical properties of pet fiber-reinforced concrete

Fiber-reinforced concrete enhances structural integrity by improving post-cracking behavior and ductility, yet traditional fibers can be costly and have a significant environmental impact. Recently, polyethylene terephthalate (PET) fiber, derived from recycled plastic bottles, has been used as a reinforcing material in concrete to enhance its mechanical properties. Utilizing advanced machine learning algorithms, this research analyzes a comprehensive dataset collected from existing literature on the mechanical properties of PET fiber-reinforced concrete (PFRC), including concrete density, water-to-binder ratio, fine and coarse aggregates, and fiber parameters such as volume fraction and aspect ratio of PET fibers. Through the application of Polynomial Regression, Exponential Gaussian Process Regression (GPR), and Cubic Support Vector Machine (SVM) models, the study effectively predicts PFRC’s mechanical behavior, showcasing high R² values indicating the models’ predictive strength: 0.93 for compressive, 0.92 for flexural, and 0.92 for tensile properties. This investigation highlights the exceptional accuracy of Exponential GPR and Cubic SVM models in capturing complex, non-linear relationships within the material, demonstrating machine learning’s capability to advance the sustainable development of construction materials. This study provides significant insights into optimizing the use of recycled fibers for concrete reinforcement.