基于反馈控制的双随机结构统计损伤识别方法研究

Uncertainties both in the excitations and properties of structures can seriously affect the damages detecting results, Using traditional deterministic methods can't detect the damages of engineering structures effectively. In order to solve the problem, based on the analysis of double randomness caused by both loads and properties of structures, a new statistical structural damages identification method under feedback control is proposed in this research. First, in order to suppress the affects of uncertainties and enhance the sensitivity of damage indicator to element damages at the same time, the feedback control is used to assign the poles of system under detection intentionally. Then, the advanced theory and methodology of random structural analysis and random vibration are utilized to solve the time domain response of controlled structure under double randomness vibration. Finally, the responses of closed-loop system are analysed by time series analysed and damages can be identified by the parameters of time series model. It is the application and development of interdisciplinary that the project combines the subjects of engineering mechanics, statistics, control theory, and theory of signal analysis and procession and so on. The study results of this projects will promote the theory and method of random vibration science, and provide a new way for the research of structural damage detection. Accordingly, the research of the project will produce significant values in theory and practical engineering in the field of Structural Health Monitoring.

基于结构振动的损伤检测方法中，环境载荷、结构本身等的不确定性会严重影响识别结果。传统的确定性方法无法对实际工程结构的损伤进行有效识别。针对此问题，本项目在同时考虑载荷和结构本身的双重不确定性的基础上，引入反馈控制，提出了一种具有较高准确性和稳定性的统计损伤识别新方法。首先，通过反馈控制合理配置系统极点，抑制不确定因素的干扰，提高闭环系统损伤特征量的灵敏度；然后，借助随机结构分析与随机振动理论的最新研究成果，对双随机下受控结构的振动问题进行分析和研究；最后，对闭环系统时域响应进行统计分析建模，通过模型参数识别损伤。项目研究融合了力学、统计学、控制理论、信号分析与处理等多学科的理论和方法，是学科交叉的应用和发展。其成果不仅有助于丰富和发展随机振动理论，并为结构损伤检测研究开辟新的思路和途径。因此，本项目研究具有重要的科学意义以及充分的理论和应用价值。

本项目针对双随机结构的损伤识别问题展开了研究。为了提高损伤指标的灵敏度同时抑制不确定性因素影响，引入了反馈控制，提出了一种具有较高准确性和稳定性的统计损伤识别新方法。建立了AR系数对单元刚度损伤系数的灵敏度方程，并通过数值算法对该方程的正确性进行了验证；提出了通过反馈控制合理配置待识别结构极点从而提高AR系数对单元刚度的灵敏度，在此基础上提出了一种将反馈控制与控制图相结合的结构健康监测方法；针对IASE-ASCE提出的Benchmark问题进行了较系统深入的研究，将反馈控制用于Benchmark损伤识别问题，取得了较好的结果；将局部均值分解（Local Mean Decomposition, LMD）方法引入结构健康监测领域，利用LMD进行双随机结构模态参数识别，以Benchmark实验数据为例，对LMD在信号分解过程中出现的模态混叠和虚假模态问题进行了研究。研究成果有助于丰富和发展随机振动理论，并为结构健康监测研究开辟了新的思路和途径。