复杂变化环境下大跨桥梁基于协整和非线性模型修正的损伤识别

大跨桥梁不仅自身结构特性复杂，且面临着变化的环境和荷载条件。因此，分离非线性环境因素影响，判断结构损伤与否并估计损伤程度成为亟待解决的工程难题。本项目将：(1)结合模糊C均值聚类方法，通过降低监测变量间的非线性相关程度改进协整方法；(2)利用核主成分分析(KPCA)处理变量间非线性关系的能力，通过引入滑窗方法建立动态更新的KPCA模型，实现结构损伤与否的在线判断；(3)采用变分模态分解识别分段等效线性结构特征参数，将结构非线性部分在脉冲响应函数意义下线性化，分离结构的局部非线性部分；(4)结合尾等效线性化和频响函数综合法，利用基于子结构的非线性有限元模型修正，建立准确的多参数影响-基准非线性子结构模型；(5)通过考虑环境因素影响的非线性子结构模型修正和卡尔曼滤波相结合实现损伤位置、程度的在线识别。该项研究不仅具有重要的理论意义和科学价值，而且对于将损伤识别应用于重大工程结构具有重要的意义。

Large-span bridges are not only complicated in their own structural properties but operating under changing environmental and loading conditions. Therefore, two essential issues emerge. The first is to separate the influence of environmental factors in case the relationship among monitored structural variables is nonlinear, and the second is to judge damage exists or not and further to evaluate damage severity. In this project, fuzzy C-means clustering algorithm which can decrease the level of nonlinear correlation of monitored variables is to be used to improve cointegration method. Due to the ability to deal with the nonlinear relationship among the monitored variables, kernel principal component analysis (KPCA) is planned to be used. By combining with sliding window method, dynamic KPCA models are to be established and updated constantly, thus online detections of structural damage will be achieved. Variational mode decomposition is adopted to estimate characteristic structural parameters and tail equivalent linearization method is employed to linearize nonlinear parts of a structure in the sense of impulse response functions, and thus the nonlinear substructure method is separated. Based on nonlinear sub-structural finite element model updating method, accurate multiple-parameter multiple-phase baseline substructure model will be established and introduced as baseline models for structural damage identification in the following phase. Finally, on the given accurate linear substructures, Kalman filter model updating method considering the influence of environmental factors is combing with nonlinear sub-structural finite element model updating to realize online identification of damage location and severity. The proposed research not only has its scientific merits and theoretical significance but is important in the practical application of structural damage identification in critical engineering projects.