节点文献
桥梁结构模态参数辨识与损伤识别方法研究
Study on Modal Parameter Identification and Damage Recognition Methods for Bridge Structures
【作者】 熊红霞;
【导师】 刘沐宇;
【作者基本信息】 武汉理工大学 , 桥梁与隧道工程, 2009, 博士
【摘要】 大型桥梁工程是国家基础设施的重要组成部分,直接关系到人民的生命和财产安全。对桥梁进行健康监测,通过识别结构动力特征参数的异常变化,及时发现其损伤隐患,建立起预警及适时维修机制,对维护桥梁正常运行,延长其服役期限,避免灾难性事故的发生具有重要的意义。结构参数辨识和损伤识别是健康监测系统的核心技术与理论基础。由于大型工程结构存在体积质量巨大,边界条件复杂,环境因素恶劣、激励信号难以有效测量、测试数据量大、信噪比低等诸多不确定因素,其系统识别技术尚处于发展阶段,有待进一步探讨和完善。现有的参数识别和损伤识别方法在精度、效率、鲁棒性和经济性能指标方面仍存在很多不足,在实际工程应用中还有许多困难需要克服:例如求解复杂结构时收敛速度慢;抗噪性较差;低阶模态时计算精度和计算效率低;测量信息不完整时不易识别等。随着结构健康监测系统的发展,迫切需要寻求新的理论和方法,解决大型土木工程结构在线监测中的参数辨识与损伤识别问题。本文的研究主要围绕进一步发展大型桥梁健康监测与状态评估的核心技术而展开。综合运用粒子群算法、奇异值分解、小波变换、功率谱分析等计算智能工具和现代信号处理技术在结构模态参数识别、物理参数识别以及结构损伤识别领域展开了系统、深入的研究工作。本文的主要研究成果和创新点如下:(1)详细地评述了目前结构模态参数识别、物理参数识别及结构损伤识别方法的理论意义、应用背景、发展现状和已取得的研究成果。系统地分析了在当前健康监测技术中应用较多的几种主要识别方法存在的问题与不足之处,在此基础上,阐述了本文研究工作的思路及主要内容。(2)提出了奇异值分解(Singular Value Decomposition,简称SVD)与小波分析结合的结构模态参数识别方法,将小波变换与SVD滤波相结合,对一个三自由度结构进行数值仿真,利用MATLAB软件编程进行信号的分析处理。研究结果表明SVD与小波分析结合的方法克服了单一小波方法的不足,可明显判别出信号时—频图中反映的模态信息,能够较方便和准确地寻找出结构的小波脊,频率与阻尼比的识别精度较高,结合多个传感器测试得到的信息进行综合判断,获得信息的可靠度更高。(3)针对受环境激励的系统在仅有输出信号时参数识别有困难的情况,提出了基于功率谱奇异值分解的模态参数识别算法,对一座斜拉桥在环境激励试验下的模态参数进行了识别,并将识别的频率数值与有限元计算结果进行了比较。对比分析表明本文提出的功率谱奇异值分解的方法克服了传统的频域峰值法选取模态的主观性,能客观准确地选择特征频率和识别相近的模态,具有处理简单、快速、实用的特点,可在实际工程中推广应用。(4)研究了基于有限元时程分析的桥梁结构模态参数识别方法。以一座独塔斜拉桥为工程对象,采用ANSYS有限元分析软件,以不同的地震动组合输入方式进行了有限元模态分析与非线性时程分析。通过对节点加速度时程响应数据进行分析处理识别出的结构模态参数,与有限元模态分析的计算频率值在低频段非常接近,高频段误差不超过6%,二者吻合较好。研究结果证明了有限元时程分析方法解决地震激励下模态参数识别问题的有效性和实用性。(5)提出了基于粒子群(Particle Swarm Optimization,简称PSO)智能优化算法的结构物理参数识别方法。利用三层框架数值仿真模型,分别模拟其在无噪声、添加0.1%噪声和0.3%噪声以及测量信息不足、模态频率不完备等情况下的刚度参数识别情况。仿真结果证明了本文提出的改进的随机惯性权重PSO算法能够准确地识别出结构中的未知参数,克服了原始PSO算法的早熟现象,计算收敛速度更快,稳定性更好,特别是在解决测量信息不完备和输入信息未知条件下的结构参数识别问题中体现出了前所未有的优越性。(6)提出了基于SVD与改进PSO算法的结构损伤识别方法。将小波变换与SVD滤波相结合对一个简支梁数值模型进行了损伤模拟,并采用改进PSO算法对多自由度结构在各种工况下的损伤参数进行了识别。研究结果表明:将小波变换与SVD方法结合,通过小波系数矩阵的奇异值分解,使得结构损伤突变信号的奇异点放大,可精确地实现对结构损伤位置及损伤程度的诊断;运用本文提出的改进PSO算法可准确地识别出外荷载作用下结构发生损伤的位置,对损伤敏感参数的识别精度更高,收敛速度更快,识别结果也更加稳定,在工程应用上具有可行性。
【Abstract】 Long-span bridges are important part of national infrastructures. During the period of bridge’s service life, different degree levels of damages will occur. Therefore study on techniques for health monitoring of bridges and build a Predicting, warning and timely maintaining system based on identifying the abnormal change of structural dynamic characteristic parameters and inspecting the hidden trouble of structural damages are essential and has great significance to ensure structural safety, prolong structural service life and avoid the occurrence of disastrous accidents.The parameter identification and the damage recognition are one of health monitoring system’s core technologies and theoretical Foundation. The system recognition technology of big-scale Engineering structures is still in the development phase and need further discussing and consummating because of many uncertain factors as follows: huge volume quality, complex boundary condition, bad environmental factor, difficulty of drive signal survey, large quantity of testing data and low Signal-To-Noise Ratio, et al. At present, research in this domain needs to be further strengthened. Existing methods of parameter identification and damage recognition are insufficient in precision, efficiency, robust and economical performance index. They have many difficulties need to overcome in the actual project application: such as slow convergence rate in solving complicated structure, bad noise immunity, low computational accuracy by using low order mode or incomplete survey information et al. New theory and method needs urgently to solve online monitoring parameters identification and damage recognition questions of large-scale civil engineering structures.This dissertation starts from the key problems of further development of the health monitoring system for large bridges. Computation intelligent methods and modern signal processing technologies such as PSO, SVD, WT, and Power Spectrum Analysis are applied synthetically and systematically thorough research is carried out in parameter identification and damage recognition domain. The main research results and innovations are as follows:1. Firstly, the theory significance, the application background, the present development and research results have obtained of structural modal parameter identification, physical parameter identification and damage recognition methods are commented in detail. Main Insufficient and disadvantages of several recognition methods be Applied widely in health monitoring technology are analyzed. Then the studying contents and work achievement are introduced.2. A new method which can identify the structural modal parameters exactly based on Singular Value Decomposition and Wavelet Transform is put forward. The wavelet transform and the SVD filter are unified. And the MATLAB software is using in the value simulation signal processing of a three-degree of freedom structure. The results show that the method based on the SVD and WT has overcome the insufficiency of the sole wavelet method. It can distinguish the modal information in the signal’s time-frequency diagram obviously, extract the Wavelet ridge of the structure conveniently and exactly. And can obtain a high recognition precision of the frequency and the damping ratio by synthetic judgment according to the information of many test sensors.3. In view of the modal parameter identification when only has the output signals of the system under ambient excitations has difficulty, a new method of structural modal parameters identification based on singular Value decomposition of the power spectrum is put forward. The response power spectral density matrix which can obtained through the survey is replaced the frequency response function matrix. The eigenfunction matrix of the system is separated through the singular value decomposition to the power spectral density function. Then the eigenvalue and the eigenvector of the eigenmatrix were solved to realize the system’s modal parameter identification. This method is used in the modal parameter identification of a cable stayed bridge under ambient excitations, and the identification frequency is compared with the finite element computation frequency. The results indicated that this method overcome the subjectivity in modal selection of frequency domain pick-peaking method, choose eigenfrequency and identify close modal accurately and objectively. With the advantages of practical, processing simply and quickly, It can be applied widely in actual engineering projects.4. A method of structural modal parameters identification based on the finite element time-history analysis is put forward. Take a single tower cable-stayed bridge as the project object, using the ANSYS software, the finite element modal analysis and the non-linear time-history analysis has been carried on by different combinatorial seismic input model. The structural modal parameters be distinguished through processing the node acceleration time-history response data is close with the computation value of the finite element modal analysis in the low frequency band, the error of the high frequency band does not surpass 6%, the two results fit well. It may be further shown that the method proposed can solve modal parameter recognition questions under the earthquake drive effectively.5. A new structural physics parameter recognition method based on Particle Swarm Optimization (PSO) algorithm is proposed. The rigidity parameter recognition in different situations (without noise, with 0.1% noise, with 0.3% noise, with insufficient survey information, and with incomplete modal frequency et al.) has simulated separately using a three-tiers Frame value simulation model. The simulation result has proved that the improved PSO algorithm with stochastic inertia weight proposed in this dissertation can distinguish the unknown parameters of structures accurately. It overcomes the precocious phenomenon of primitive PSO algorithm. It has quicker computation convergence rate and better stability, and has more superiority in solving parameters recognition with insufficient survey information and incomplete modal frequency especially.6. A structural damage recognition method based on SVD and proved PSO algorithm is proposed. The WT and the SVD filter are unified to simulate the damage of a simple beam numerical model. And the improved PSO algorithm is used to distinguish the damage parameters under many kinds of operating mode. The findings indicated: The odd and break points of damage signals are amplified through singular value decomposition of the wavelet coefficient matrix, the position and the extent of structural damage can be distinguished precisely. The improved PSO algorithm may distinguish the damage position accurately of structures under the applied loads. It can obtained higher precision of the sensitive parameters of the damage, quicker convergence rate and more stable recognition result. So this method has validity and feasibility in engineering application.
【Key words】 Structural health monitoring; Physics parameter identification; Modal parameter identification; Damage recognition; Wavelet Transform; Singular Value Decomposition (SVD); Power spectrum; Ambient vibration; cable stayed bridge; Particle Swarm Optimization (PSO) algorithm;