【作者】 刘伟；

【导师】 孙毅； 高维成；

【作者基本信息】 哈尔滨工业大学 ， 固体力学， 2009， 博士

【摘要】 空间网格结构是一种由杆单元和梁柱单元集成的空间三维结构体系,具有受力合理、刚度大、自重轻、抗震性能好、工期短、造价低等优点,尤其能够满足建筑造型丰富、功能齐全等要求,被广泛用于大跨度工业与民用建筑及公共建筑领域,成为反映一个国家建筑科学技术水平的标志。由于环境荷载作用、疲劳效应、腐蚀效应和材料老化等灾害因素影响,空间网格结构容易产生损伤累积和抗力衰减,甚至会发生破坏乃至倒塌的灾难性事故。因此,对大跨度空间网格结构在使用期间的健康监测以及各种灾害影响下的损伤识别进行研究,开发适合于大跨度空间网格结构的健康监测系统,具有重大的理论价值和现实意义。本文结合结构健康监测领域国内外发展现状,对空间网格结构健康监测系统关键技术进行了深入系统的理论研究,包括环境激励下模态参数识别理论、传感器优化布置理论、有限元模型修正理论、空间网格结构损伤识别理论,分别通过了数值仿真验证、实验室模型试验验证和实际工程现场实测验证,最终开发了一套功能强大的空间网格结构健康监测系统。对环境激励下结构模态参数识别理论进行了系统的研究,分别提出了环境激励下低阶时域模态参数识别方法的统一理论和环境激励下高阶模态参数识别方法的统一理论,揭示了不同方法之间的本质联系和不同点,完善了结构模态参数识别的系统化理论。比较了四种适合于环境激励下空间网格结构模态参数识别的方法,指出了它们优缺点及适用范围,验证了两类模态参数识别方法统一理论的合理性和有效性。针对目前传感器优化布置算法应用于空间网格结构的局限性,提出了一种基于改进遗传算法的传感器优化布置方法,以模态变形能和模态置信准则为适应度函数,提出十进制二维数组编码方式以提高数据存储量,为避免相同位置重复布置传感器引入强制变异算子,利用Guyan扩充振型与有限元计算振型均方误差大小评价传感器布置方案优劣,比较了不同测点工况下数据驱动随机子空间算法识别出的结构模态参数,指出了识别结构前若干阶模态所需的最少传感器数目。研究了结构健康监测领域常用的二种传感器优化布置方法,从数学的角度推导了模态动能法与有效独立法的理论相似性,指出了二者在具体应用中的不同之处,促进了传感器优化布置理论的统一化发展。从多自由度系统频响函数矩阵的模态展开式出发,推导了平均位移幅值、平均速度幅值和平均加速度幅值公式,在有效独立法的基础上提出了有效独立-平均加速度幅值法和有效独立-模态动能法,与有效独立法、模态动能法、有效独立驱动点残差法进行了比较研究,验证了两种方法的优越性。针对空间网格结构特点,提出了能够定量评估结构损伤程度的模态应变能损伤识别理论,采用非负最小二乘法求解超静定方程组,实现定量评估损伤程度,即保持了刚度矩阵的稀疏性,又不含除阻尼的任何近似和假设。讨论了测试信息完备和测试信息不完备两种情况下该方法对空间网格结构损伤的识别能力,并与基于最小秩摄动理论的损伤程度评估结果进行了比较研究,验证了方法的有效性。考虑空间网格结构杆件数目远大于节点数目的实际情况,提出了面向节点的基于三层BP神经网络的多阶段损伤识别方法,将标准化固有频率变化率和振型向量变化量范数作为神经网络输入参数,根据各个节点输出的损伤值来判定结构损伤位置和损伤程度,以一平面桁架结构仿真算例验证了方法的合理性。针对钢筋混凝土空间网格结构,提出了直接利用损伤结构模态信息即可进行损伤识别的曲率模态曲面拟合方法,通过曲率模态与曲面拟合值的差异进行损伤定位,研究表明其适合于钢筋混凝土空间网格结构的在线健康监测。设计制作了一空间桁架结构试验模型,对其进行较为充分的静力试验、动力试验和损伤识别试验研究。提出了联合静动力数据的有限元模型修正方法,对初始有限元模型进行了修正,获得了基准模型。按本文提出的方法进行了传感器优化布置和模态参数识别,通过杆端螺栓节点连接失效和节点附加质量来模拟结构杆件和节点损伤,研究结果验证了前文损伤定位方法对空间网格结构损伤识别的有效性。探讨了将该试验模型作为空间网格结构Benchmark标准验证模型的可行性,为不同研究者验证空间网格结构健康监测关键技术提供了统一标准模型。以空间网格结构为对象,系统研究了结构健康监测系统的基本组成和监测内容,开发了基于LabVIEW的数据采集系统和基于Matlab的结构模态分析与损伤识别系统(SMADIS 1.0),实现了动态数据采集及预处理、环境激励下结构模态参数识别、模型匹配和模型修正、损伤定位及程度评估等功能,研究表明基于LabVIEW和Matlab二者结合的软件开发模式值得在结构健康监测领域推广。对一空间管桁架游泳馆屋盖结构进行了现场检测研究,提出了全面系统的检测方案,对其进行了整体结构检测、常规静力检测和环境振动试验,并进行静力核算。依据现场实测结构尺寸、荷载和边界条件建立了三维有限元模型,进行了有限元模态分析。对比了四种适合于空间网格结构的环境激励下结构模态参数识别方法,对该结构的安全状态和工作性能进行了全方位评估,提出了加固改造维修方案,给出了相关结论。更多还原

【Abstract】 Long-span spatial lattice structure which are composed of bar elements and beam-column elements are a kind of three-dimension spatial structure system, and they are the symbol of the national building science and technology for not only their reasonable force, high stiffness, light weight, good earthquake resistance, short construction period and cheap construction cost but also prolific style and complete function. They have been wildly adopted in the fields of long-span architechtures, such as industrial architecture, civil architecture and public building. Spatial lattice structures are inevitable to suffer from natural disasters, such as environmental loading, fatigue, corrosion, aging and so on. And then the damage accumulates and the resistance attenuates during long service period, and finally the damage or even paroxysmal disasters happened. Therefore, intelligent heath monitoring and damage diagnosis for structures during building and service period at various environment become an important technology to study.Detailed researches were carried out for spatial lattice structures health monitoring by combining the status quo of structural health monitoring (SHM) and structural condition assessment. The related important problems on structural health monitoring system, such as modal parameters identification under ambient vibration, optimal sensor placement, finite element model updatting, spatial lattice structure damage detection, were investigated and one spatial lattice structure health monitoring system was developed also.Meanwhile the laboratory test and field investigations also were finished. This dissertation investigated the theories of modal parameters identification and presented the unified approach for the method of low order time domain modal parameters identification under ambient vibration and the method of high order modal parameters identification under ambient vibration respectively. The similarities and differences of different identification methods were presented as well. All these work is helpful to build the whole unified theory of modal parameters identification methods. By comparing the four useful methods for spatial lattice structure modal parameters identification under ambient vibration, the advantages and disadvantages on their application of these methods were given. And also the validity of the two kinds of methods was verified.Considering the problem of optimal sensor placement on spatial lattice structure, an improved genetic algorithm (IGA) was introduced. The method took the modal strain energy (MSE) and the modal assurance criterion (MAC) as the fitness function respectively, and improved the data storage capacity with the decimal two-dimension array coding method and solved the problem of positional repeatability of sensors with the forced mutation operator. Further the evaluation criterion for sensor placement based on the mean square error between the Guyan Expansion mode shapes and the FEM mode shapes was given.With this criterion, the structure modal parameters identified by Data-driven Stochastic Subspace Identification (SSI-DATA) method using the measured data at different optimal sensor placement were compared. Then the minimal number of sensors for the identification of the first few modes were indicated. This dissertation investigated two optimal sensor placement methods used widely in the field of structural health monitoring (SHM). And then the theoretical similarity between the modal kinetic energy (MKE) method and effective independence (EI) were given by mathematical derivation.Additionally, the differences of the applications of the two methods were pointed, which was helpful to the further development of the optimal sensor placement method. Using the model expansion of MDOF system’s FRF matrix, the formulas of average displacement amplitude and average velocity amplitude and average acceleration amplitude were derivated. Further, based on the effective independence method, the effective independence - average acceleration amplitude (EI-AAA) method and effective independence -modal kinetic energy (EI-MKE) method were derivated. Comparing with the effective independence (EI) and the modal kinetic energy (MKE) method and effective independence - driving-point residue (EI-DPR), the advantage of the two proposed methods were verified.An evaluation algorithm based on the modal strain energy theory was given by adopting an overdetermined equation group solved by the non-negative least-squares method to quantitatively identify the damage extent. The method also maintained the sparsness of the stiffness matrix and excluded the approximations and assumptions except the damping ratio. Meanwhile the validity of the method used for the spatial lattice structures was verified by analyzing the two cases of complete and incomplete measured information and comparing with those results obtained by using the minimum rank perturbation theory.Considering the number of the bar is quite larger than that of the node in a spatial lattice structure, a node-oriented multistage damage identification method based on three layers BP neural network was proposed. The method took the normalized natural frequence variation ratio and mode vector norm as the input of neural network, and then determinated the position and size of the damage according to the output of neural network nodes. Then the validity of the method was verified by a plain truss model. Considering the propertities of the concrete spatial lattice structure, the damage identification method directly used the modal information of the damage structure by the surface fitting method of curvature mode was proposed, which localized the damage by the difference between curvature mode and its fitted value. The research results showed that it was suitable for online SHM of the concrete spatial lattice structure.Laboratory work has been done with an autonomous design space truss test model with which the tests of static and dynamic and damage identification have been finished also. A finite element model updating method using the static and dynamic test data was proposed. And with this method, the benchmark model was obtained from the initial model. Analyzing the results using the identification method proposed in the paper, the validity of the damage localization method based on residual modal force was verified once again. And the damage is simulated by the disabled connection nodes and the added node mass. Further the probability of setting the laboratory model as the Benchmark model for the spatial lattice structure was investigated. Finally, the uniform standard test model was given.The dissertation also gave the design approaches of the SHM system for the spatial lattice structure and proposed the system integrated design principles. Also the data acquisition system based on LabVIEW and the structural modal analysis and damage identification system (SMADIS 1.0) based on Matlab were developed, which had the functions such as dynamic data acquisition and its preprocessing, modal parameters identification of the structures under ambient vibration, modal matching, model updating and damage localization and quantification. The application of the two systems verified the developed approach is helpful for the development of the SHM system. Field investigation was carried out on a steel roof of a SPSS structure natatorium in service. The systematic field-work approaches were given. The field visual inspections and static checking were conducted in turn under in-service environmental conditions. Further a three-dimensional finite element model was developed according to its factual geometry properties obtained from the field inspection. By comparing the four useful modal parameters identification methods for spatial lattice structure under ambient vibration, the all-round assessment on structure safe state and service behavior was conducted, then the constructive suggestion of reinforce and maintainance was given, and finally were the relative conclusions.更多还原