多源数据累积的桥梁损伤信息量化提升方法及刚度退化状态辨识研究

Accurate extraction of damage information and quantitative characterization of the overall stiffness degradation state are the bottlenecks and difficulties in bridge structural health monitoring. Complex environmental conditions, time-varying loads and noises can improve the disorder of monitoring data. And these factors can also cover up damage information and reduce the damage sensitivity of the global parameter. Previous studies of the applicant have found that the Mahalanobis distance cumulants are highly sensitive to damage and can be used to characterize the overall stiffness degradation of the bridge structure. This project is proposed on the basis of previous studies in order to improve quantitative damage information as the research train of thought "the damage information accumulation - damage identification improvement" using the advantages of the multiple data types and large quantities of monitoring data. Firstly, damage sensitive components are extracted from various types of monitoring responses. These components are used to establish a global structural stiffness degradation feature vector by the weighted Mahalanobis distance method. The damage information quantitative improvement method and the feature vector of stiffness degradation of bridge structures are established. On this basis, considering the influence factors of vehicle load growth and environmental temperature, the change characteristics of the feature vector relative entropy under different degradation states are revealed. The stiffness degradation state classification identification models are established by using pattern recognition, and finally the method for overall stiffness degradation identification of the bridge structure is proposed. This study is of great significance for the effective identification of early and minor damage of bridge structures and can provide scientific basis for safety assessment and determination strategy of maintenance and reinforcement of bridge structures.

准确提取损伤信息和量化表征整体刚度退化状态是桥梁结构健康监测的瓶颈和难点，复杂的环境条件、时变荷载和噪声等因素可提高监测数据的无序化程度，掩盖损伤信息和降低全局量的损伤敏感度。申请人前期研究发现，马氏距离累积量对损伤具有高度敏感性，可用于表征桥梁结构整体刚度退化。本项目拟在前期研究的基础上，利用结构健康监测种类多、数量大的优势，以“损伤信息量的积累—损伤识别质的提高”为量化提升损伤信息的研究思路，通过提取各类型监测响应的损伤敏感分量进行加权马氏距离量化累积，建立桥梁结构损伤信息的量化提升方法和表征结构整体刚度退化的特征向量；在此基础上，通过特征向量相对熵在刚度退化过程中的变化特征及外荷载和温度的影响分析，利用模式识别建立刚度退化状态分类辨识模型，最终提出桥梁结构整体刚度退化辨识方法。本研究对于有效识别桥梁结构早期微小损伤具有重要意义，可为桥梁安全状态评估和维修加固策略确定提供科学依据。

有效识别结构损伤和准确辨识结构刚度退化状态是保障桥梁使用安全和确定维修加固策略的重要理础，然而，如何提取大量监测数据中的结构损伤信息是桥梁结构健康监测的难点问题。本项目基于多源数据融合的马氏距离累积量提出了结构损伤信息的量化提升方法和结构刚度退化状态的辨识方法。.首先，通过马氏距离累积量建立了结构损伤识别精度提升准则。开发了循环分析程序，以马氏距离累积量为损伤识别向量进行概率密度函数拟合，利用概率密度函数面积变化规律建立了结构损伤识别精度提升准则。研究表明，当损伤工况下的概率密度函数超过无损伤工况的上限面积达到所设定的阈值时，损伤识别的精度可显著提升。.其次，基于多源数据及损伤敏感分量建立了结构损伤信息的量化提升方法。利用经验模态分解建立了损伤敏感分量的筛选方法，分析了损伤前后本征模态函数（IMF）的能量变化规律，通过IMF能量变化比建立了损伤敏感分量的筛选指标。同时，利用结构监测数据类型多的特点，提出了多源数据构建加权马氏距离累积量的结构损伤信息量化提升方法。研究结果表明，结构损伤后监测数据各阶IMF能量变化明显，筛选转移损伤能量占自身能量多的IMF构建损伤敏感分量，能够有效避免损伤信息的掩盖。此外，相对能量变化率为正的多阶IMF共同构造损伤识别向量，能进一步提取结构的损伤识别效果。多源数据构建加权马氏距离作为损伤敏感分量对结构损伤具有更高的敏感度，利用IMF构建加权马氏距离累积量对损伤的识别效果更好。.最后，建立了马氏距离累积量与结构整体刚度退化的关系式，提出了结构整体刚度退化辨识方法，静力试验与动态激励试验刚度退化率对比验证了方法的适用性。此外，多源数据及其IMF构建损伤敏感分量进行结构刚度退化辨识，能够有效的降低刚度退化识别的误差。