Climate-resilient epoxy asphalt mixture design: An intelligent framework

Abstract

Epoxy asphalt mixture exhibits excellent durability, crack resistance and high-temperature stability, making it an ideal choice for climate-resilient pavement materials. In order to expand its application scope and maximize its advantage, it is necessary to propose more advanced mixture design method. This study proposed an intelligent design framework combining machine learning and metaheuristic algorithms for developing epoxy asphalt mixture. First, high-accuracy prediction models of the performance of epoxy asphalt mixture under high and low-temperature environments were established using Extreme Gradient Boosting optimized by Particle Swarm Optimization (PSO-XGBoost). Then, interpretability analysis, including feature importance and accumulated local effects, was conducted based on these models to identify the key design features of epoxy asphalt mixture and determine their empirical value ranges to achieve satisfactory mixture performance. Next, diversified strategies were determined to meet engineering needs, including high performance, low cost and carbon emissions, as well as a comprehensive strategy that incorporates all these objectives. Subsequently, multi-objective optimization models considering these strategies were established, and the optimal solutions were generated based on the Third Generation of Non-dominated Sorting Genetic Algorithm (NSGA-III) and TOPSIS. Finally, the practical feasibility of these solutions was confirmed through laboratory tests. Based on the proposed framework, high-performance, cost-effective, and environmentally sustainable epoxy asphalt mixtures can be obtained. This study sets a new benchmark for future research in the intelligent design of sustainable pavement materials, emphasizing the practical and theoretical implications of integrating advanced computational tools in pavement material science.