沥青路表裂缝与结构内部损伤的机理-数据混合建模研究

路表裂缝识别为沥青路面智慧化管养提供大量数据支持，但路表破损率并不能反映结构内部的真实损伤情况。路面结构内开裂损伤的量化普遍采用力学模型辅助有限元模拟的机理驱动方法，但其包含简化假设，且难以处理多变量的耦合性。数据驱动方法能解决多变量问题，但是对裂缝发展动态特性及不确定性缺乏物理、力学定义。因此，本项目以沥青路面Top-Down裂缝为例，拟开展路表裂缝与结构内部损伤关联性的机理-数据混合建模研究，展示两种方法融合在突破技术瓶颈上的优势：基于材料和结构层级的开裂机理研究，提出数据驱动建模中关键参数的选择方法；构建机理模型与数据模型结合的逆向智能算法，解决现场面层结构开裂损伤的量化问题；以大规模的数据分析刻画现场路面开裂行为方式和规律，通过数据建模涵盖多因素间的隐式关联，实现由路表裂缝数据到结构内部损伤的快速、准确计算。为将沥青路面智能巡检从路表破损测量拓展到结构层损伤评价提供理论与技术支撑。

Nowadays automated pavement crack detection could only measure cracks on the surface; those inside the structural layer are not included. The model-driven approach like the numerical modeling plus mechanistic models is the most popular way to quantify the crack damage in asphalt pavements. However, this kind of approach suffers from issues like assumptions, prediction accuracy, and extensive testing requirements due to the complexity involved in a field cracking process. Compared to the model-driven approach, the data-driven approach could consider as many factors as possible, but it lacks of input selection criteria and physical-mechanical definitions of crack propagation characteristics. Therefore, taking the longitudinal top-down cracking in asphalt pavements as an example, this project aims at proposing a hybrid modeling technique to overcome the limitations of using model-driven or data-driven alone. The mechanism investigations on the material- and structure-level provide the input selection criteria for the data-driven modeling process. The data-based analysis would depict the network-level behaviors of top-down cracking in the field. In addition, a mechanism-based model and a data-based model are combined to perform an inverse analysis of the actual crack damage of the asphalt layer in the field. Finally, a data-driven model could be established to connect the surface cracking damage to those inside the asphalt layer, which would promote the development of a more comprehensive automated pavement crack detection technique.