基于运行模态的桥梁结构损伤识别对比研究

发布时间：2019-04-22 13:21

【摘要】：如何从海量数据中获取结构的特征信息,判断桥梁结构的运营状况是桥梁健康监测系统的主要任务。其中,有限测点下的传感器优化布置是基础,环境激励下的模态参数识别是核心,损伤识别方法则是关键。本文围绕这三个桥梁健康监测系统的共性关键问题,开展深入系统的研究、开发及应用。传感器优化部分,主要阐述8种经典传感器优化算法的基本思想,系统梳理各算法的技术特点和计算流程,总结各算法之间的演变及联系,并集成开发软件,为模态识别提供基础;模态识别部分,详细介绍两种信号预处理方法和6种运行模态参数识别方法的基本理论、归纳总结各算法的技术特点和数值计算方法,以各种算法的数值分析手段为出发点,探讨不同算法的识别精度、适用范围,以及抗噪水平。损伤识别部分,深入研究基于模态的损伤识别方法,主要包括基于频率、基于曲率模态和基于柔度矩阵的损伤识别方法,探讨各种损伤指标的使用范围以及抗噪水平。目前,还没有成套的健康监测数据分析软件,少有的几个商业软件均为黑箱模式、不能定制;更为关键的是,它们需要人工干预,不能满足桥梁健康监测系统集成化、自动化、实时化的要求。为此,本文以MATLAB为平台,集成开发模态分析工具箱和损伤识别工具箱。模态识别工具箱功能涵盖6种经典的模态识别算法,并针对不同算法增设参数调试功能,能够进行实时模态参数识别。损伤识别工具箱涵盖了基于频率、基于曲率模态和基于柔度矩阵三大类的损伤指标计算功能,简洁直观地显示结构损伤位置。用户可以直接使用模态分析工具箱进行实时模态分析和研究,利用损伤识别工具箱判断结构的损伤位置。最后对一座三跨连续刚构的数值模型通过8种传感器优化布置算法进行传感器优化布设,并针对3种混合算法的影响因素进行参数化分析,结果表明EI法效果最差,逐步消减法效果最稳定,混合算法的性能与参数的选取有关;采用6种模态识别算法进行模态识别,结果表明,模态识别方法中,FDD法抗噪性最好,SSI法和ERA法识别精度最高。通过损伤识别的计算发现,损伤指标MFC的抗噪性最好。为进一步验证程序的可靠性和适用性,利用模态分析工具箱对南京长江四桥进行实桥模态分析,结果表明本文开发的工具箱具有很好的实际应用价值。
[Abstract]:It is the main task of the bridge health monitoring system that how to obtain the characteristic information of the bridge structure from the massive data and how to judge the operation condition of the bridge structure. Among them, the optimal arrangement of sensors at finite measurement points is the basis, modal parameter identification under environmental excitation is the core, and damage identification method is the key. This paper focuses on the common key issues of these three bridge health monitoring systems, and carries out in-depth and systematic research, development and application. In the sensor optimization part, the basic ideas of eight classical sensor optimization algorithms are described, the technical characteristics and calculation flow of each algorithm are systematically combed, the evolution and connection of each algorithm are summarized, and the development software is integrated. Provide the basis for modal identification; In the part of modal identification, the basic theories of two methods of signal preprocessing and six methods of operating modal parameter identification are introduced in detail, the technical characteristics and numerical calculation methods of each algorithm are summarized, and the starting point is the numerical analysis methods of various algorithms. The recognition accuracy, application range and anti-noise level of different algorithms are discussed. In the part of damage identification, the damage identification method based on modal is deeply studied, including frequency-based, curvature mode-based and flexibility matrix-based damage identification methods, and the application scope and anti-noise level of various damage indexes are discussed. At present, there is no complete set of health monitoring data analysis software, few commercial software are black box mode, can not be customized; More importantly, they need manual intervention and can not meet the requirements of integration, automation and real-time of bridge health monitoring system. In this paper, MATLAB platform, integrated development of modal analysis toolbox and damage identification toolbox. The function of modal identification toolbox covers six classical modal identification algorithms and adds parameter debugging function to different algorithms to identify real-time modal parameters. The damage identification toolbox includes three kinds of damage index calculation functions: frequency-based, curvature-mode and flexibility matrix. The damage location of the structure can be displayed simply and intuitively. Users can directly use the modal analysis toolbox for real-time modal analysis and research, and use the damage identification toolbox to determine the damage location of the structure. Finally, a numerical model of a three-span continuous rigid frame is set up by eight optimal sensor placement algorithms, and parametric analysis of the three hybrid algorithms is carried out. The results show that the EI method has the worst effect. The effect of step-by-step subtraction method is the most stable, and the performance of hybrid algorithm is related to the selection of parameters. Six modal identification algorithms are used for modal identification. The results show that the FDD method has the best anti-noise property and the identification accuracy of SSI and ERA method is the highest. Through the calculation of damage identification, it is found that the damage index MFC has the best anti-noise ability. In order to further verify the reliability and applicability of the program, the modal analysis toolbox is used to carry out the real bridge modal analysis of the fourth Nanjing Yangtze River Bridge. The results show that the toolbox developed in this paper has a good practical application value.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U446

【相似文献】