交通高峰期融合分布式计算与稀疏表示的桥梁荷载与损伤同时识别研究

Frequent excitation of vehicle loads can accelerate the bridge structural damage and even induce a fatal bridge accident. It is imminent to explore a new method for simultaneous identification of bridge external forces and damages. Limiting by the conditions of serial computation and assumption of known initial conditions, existing methods of simultaneous identification are poor timeliness and weak applicability. As a result, existing methods are not suitable for estimating the bridge external forces and damages during rush hour. In this study, theoretical analysis, numerical calculation and model experiment will be employed. Combining the distributed computing and sparse representation, simultaneous identification of bridge external forces and structural damages will be investigated. Analysis model of the bridge will be firstly established by the help of multi-objective swarm intelligence algorithms. Problem of separating bridge external forces and initial conditions will be overcome via two steps. They are sparse decomposition of structural responses and reconstruction of external forces. Then a core method will be proposed for simultaneous identification of bridge external forces and structural damages under unknown initial conditions. Under the guidance of divide and conquer, distributed tasks will be created based on moving time-window technique. Unknown initial conditions are introduced for each task. It can make sure that data for each task is independent. A distributed platform will be designed and built for solving the distributed tasks and fusing the identified results. A novel distributed computing-oriented method will be proposed for simultaneous identification of bridge external forces and structural damages during rush hour. Herein, sparse representation of external forces and sparse distribution of structural damages will be considered. Laboratory experiments will be carried out for verification of the proposed method. Bridge external forces and structural damages can be identified rapidly and simultaneously during rush hour. Some technical supports are provided for structural health monitoring of bridges as well.

车辆荷载的频繁作用会加速桥梁损伤，甚至诱发桥梁事故，开展桥梁荷载与损伤协同识别研究迫在眉睫。受串行计算与初始条件已知的限制，既有桥梁荷载与损伤协同识别方法的时效性与适用性较差，无法满足交通高峰期桥梁荷载与损伤的快速协同识别。本项目拟采用理论分析、数值计算与模型实验有机结合的方法，融合分布式计算与稀疏表示，研究桥梁荷载与损伤协同识别问题。基于群智能算法探讨桥梁多目标优化建模；通过响应稀疏分解与荷载稀疏重构，克服桥梁荷载与初始条件的分离难题，形成未知初始条件下桥梁荷载与损伤协同识别新策略；在“分而治之”思想指导下，基于移动时间窗创建分布式任务，引入未知初始条件实现任务数据对外封闭，设计搭建分布式平台完成任务求解与结果合并；重点提出交通高峰期面向分布式计算的桥梁荷载与损伤稀疏协同识别新方法，并辅以实验验证，以达到交通高峰期快速反演桥梁荷载与损伤的研究目的，为桥梁结构健康监测提供技术支持。

桥梁动荷载与结构损伤是影响桥梁结构安全运营的两个重要因素，对两者有效的反演与识别是桥梁结构健康监测的重要内容。本项目采用理论分析、数值仿真和模型实验相结合的方法，基于稀疏表示和分布式并行计算研究了桥梁荷载反演和损伤识别问题。项目探索了蜻蜓算法及其改进策略在结构损伤优化识别中的应用，融合卷积神经网络与蜻蜓算法提出了桥梁结构损伤识别两步法；提出了结构初始条件的伴随冗余字典表示新方法，解决了动荷载识别方程中初始条件与结构荷载的融合难题，实现了未知结构初始条件下桥梁动荷载的高精度稀疏反演；提出了考虑稳均值约束的移动荷载稀疏正则化识别方法，有效削弱了传感器标定误差的不利影响，提升了桥梁移动荷载的识别精度；发展了动荷载矩阵正则化识别和桥梁移动荷载等效表示理论，建立了移动荷载与固定荷载之间的等效关系，实现了结构动荷载的矩阵正则化反演，并成功将其应用于桥梁移动车重的识别。在此基础上，搭建了分布式并行计算平台，基于时间域分割提出了动荷载并行求解新方法，发展了基于空间域分割的结构荷载与损伤同时识别的并行解决方案，实现了桥梁荷载与损伤的稀疏快速反演，一定程度上为桥梁结构健康监测提供了有益的技术借鉴。至目前，项目共发表了期刊论文6篇（SCI收录5篇），培养硕士研究生已毕业4名，尚有7名硕士研究生在读。