【作者】 朱焰煌；

【导师】 滕军；

【作者基本信息】 哈尔滨工业大学 ， 结构工程， 2010， 博士

【摘要】 随着国民经济的不断发展以及建筑科学技术水平的不断提高,各种造型独特、结构复杂的大跨度空间结构不断涌现,以往多应用于桥梁结构上的健康监测技术也逐步地在大跨空间结构中得到应用。参数识别和损伤预警是健康监测的核心技术。环境激励下存在测试数据的信噪比低、结构响应受外界环境因素的影响等问题,这些问题导致一些算法虽然在数值或实验室模型中有效,但却很难直接应用到实际工程中。本文以国家游泳中心水立方结构健康安全监测项目为工程依托,围绕实际工程健康监测中的参数识别及损伤预警相关问题展开研究,主要研究内容如下:(1)研究加速度传感器的优化布置方法。针对大跨空间钢结构的振动特点,研究基于模态能量和分步优化的传感器布置方法。根据结构模态应变能的大小挑选出环境激励下结构的主要贡献模态,即优化时所取的监控模态。根据模态振型形状、特征向量乘积以及模态置信度,采用分步法进行传感器优化布置。以国家游泳中心钢结构为工程背景,对其模态测试时加速度传感器布点进行优化。(2)研究结构健康监测加速度信号的降噪方法。分析有用信号与噪声的模极大值特点,针对小波降噪中分解层数以及奇异值降噪中重构阶次难以确定的问题,研究一种分解层数以及重构阶次的自适应确定法。用数值模拟信号以及实测数据对所提方法进行验证。(3)研究基于小波变换的模态参数识别技术以及模态变异性。针对大跨空间结构具有低频密集模态以及难以实现用力锤或激振器来激励等特点,研究自然激励法与小波变换相结合的模态参数识别方法。分析母小波参数的选择对识别精度的影响,研究基于遗传算法和最小标准差的小波中心频率及带宽的自适应选择方法。通过数值仿真以及国家游泳中心现场实测数据的分析,验证所提方法的有效性。在此基础上进一步分析模态参数的环境变异性。(4)研究考虑不确定性的模型修正方法。针对用于模型修正的模态参数具有不确定性的特点,以及大型复杂结构模型修正时计算量大的问题,研究基于区间理论和多输出支持向量回归机的有限元模型修正方法。利用区间分析法来描述参数的不确定性,用均匀实验设计方法构造样本,将实测模态参数作为输入,多个设计参数作为输出,以支持向量回归机逼近输入输出二者之间的非线性映射关系,然后利用支持向量回归机的泛化推广能力,求解设计参数的目标值,为有限元模型修正提供一种新的探索。利用所提方法对国家游泳中心钢结构模型进行修正。(5)研究运营条件下结构损伤预警方法。研究基于脉冲响应函数和统计判别的损伤预警方法,首先通过虚拟激励技术提取环境激励下结构脉冲响应,然后以脉冲响应作为敏感指标,综合运用主成分分析、支持向量机、控制图等方法,对运营环境下结构损伤进行预警。利用国家游泳中心钢结构数值模型以及实测数据对所提方法的有效性进行验证。更多还原

【Abstract】 With the continuous development of national economy and the construction technology, a lot of large-span spatial structures with unique outline and complex construction are emerging. The health monitoring system that mainly used in bridge structure is applying in large span spatial structures gradually. Parameter identification and damage alarming is the core technology of health monitoring. Because noise ratio of testing data under environmental excitation becomes lower, structural response easy to be affected and other problems, many algorithms are effective in laboratory but failed in practice. This dissertation studies parameter identification and damage alarming method of large span space steel structure under environmental excitation, using National Aquatic Center steel structure health monitoring program as the example to promote practical use of structural health monitoring technology. The detailed research contents of this dissertation are shown as below:Firstly, this dissertation studies optimal placement of acceleration sensor. According to vibration characteristics of large span steel structure, it proposes sensor placement method based on modal energy and step-size optimization. It means that the main contribution mode is selected according to the importance of structural mode responses under environmental excitation, and then applies the optimal sensor placement according to mode shape, feature vector product and confidence level. This dissertation takes National Aquatic Center steel structure as the engineering background to optimize acceleration sensor distribution.Secondly, this dissertation studies the denoising method of vibration signals. Maximum module characteristics of both useful signal and noise are analyzed. In order to overcome the difficulty in determining decomposition level of wavelet denoising and reconstruction order of singular value decomposition, a self-adaptive determination method for decomposition level and reconstruction order is proposed. Numerical signal simulation and the measured data verify the proposed method.Thirdly, this dissertation studies modal parameter identification technology based on wavelet transformation. It proposed a modal parameter identification method with natural incentives and wavelet transformation to solve the problems of low-intensive of large spatial structure and difficulty in motivating with hammer or shaker. This method analyzes the impact of choosing mother wavelet parameter on identification accuracy. According to this analyzing it proposes self-adaptive selection method based on genetic Algorithm and minimum standard deviation of wavelet center frequency and bandwidth. Numerical simulation and field measurement of National Aquatic Center verify the effectiveness of the proposed method.Fourthly, this dissertation studies model updating method considering uncertainty. It proposes finite element model updating method based on range theory and multi-output support vector regression according to the uncertainty of model parameter and too much calculation when updating complex structural model. This method describes the uncertainty of parameters using interval analysis, constructs a sample using uniform experimental designing method, and takes model parameter as input, a number of designing parameters as outputs to get the non-liner mapping relationship between input and output value, then makes use of support vector regression generalization ability to get target value of design parameter, which provide a new exploration for finite element model updating. Numerical models and the National Aquatic Center steel model are used to verify the validity of the method.Lastly, this dissertation studies the early warning of structural damage. It puts forward a damage alarming method based on impulse response function and statistical theory. This method first selects the structural pulse responses under environmental excitation using virtual excitation technique, takes the pulse responses as sensitive indicators, then uses the methods of principal component analysis, support vector machine, control chart, etc. to alarm structural damage. Numerical model of National Aquatics Center steel structure and its measured data verifies the effectiveness of the proposed method.更多还原