Sensitivity analysis of carbon fiber reinforced asphalt pavements: Experimental study and data driven predictive model using machine learning

It is well known that the impact of fibers on reinforcing asphalt pavement (AP) has been extensively investigated. However, the optimal fiber length and content considering various temperatures is still an unresolved issue in this field. To reach this goal, this research aims to study the impact of carbon fiber (CF) reinforcement on the fracture energy and toughness of AP under varying environmental conditions. Several mix designs considering different CF length (10–30 mm) and content (1–3.5 %) at 5 °C and ambient temperatures were evaluated to determine the optimal values. The results demonstrated that while increasing CF length and content generally improves fracture energy and toughness, there is an optimal threshold beyond which additional CF content yields diminishing returns. Notably, the sample with a 1.5 % replacement ratio and 15 mm CF length exhibited superior performance, significantly enhancing the AP’s resistance to fracture due to effective stress distribution and crack bridging. Conversely, samples with the higher fiber content and length showed decreased fracture toughness, due to fiber clustering and reduced workability. Additionally, experimental results indicated higher values at 5 °C for fracture energy and toughness. Advanced machine learning techniques were also utilized to develop predictive models to accurately simulate the fracture behavior of CF-reinforced AP mixtures. Among all models, Gaussian process regression demonstrated superior performance and exhibited lower error in predicting experimental data. Moreover, a sensitivity analysis of this model was conducted to determine how the length and content of CF, as well as temperature, influence fracture energy and toughness, guiding the optimization of CF integration in AP design. These findings provide valuable insights for engineering durable and resilient asphalt pavements capable of withstanding diverse climatic stresses, ultimately contributing to more sustainable and cost-effective infrastructure solutions.