【作者】 禹丹江；

【导师】 任伟新；

【作者基本信息】 福州大学 ， 结构工程， 2006， 博士

【副题名】理论、实现与应用

【摘要】 土木工程结构模态参数识别-理论、实现与应用的课题研究来源于国家自然科学基金项目(批准号:50378021)。土木工程结构是国家基础设施的重要组成部分,直接影响人民的生活和安全。对土木工程结构进行全面的检测、评估和健康监测,就需要充分了解土木工程结构的动力特征参数。模态参数是决定结构动力特征的主要参数,其识别方法一般可分为传统的模态参数识别方法和环境激励下的模态参数识别方法。环境激励振动试验,具有无需贵重的激励设备,不打断结构的正常使用,方便省时等显著的优点,更加适合土木工程结构的实际使用。环境振动试验不同于传统的基于输入和输出的模态参数识别,仅测得了结构振动响应的输出数据,而真正的输入是没有测量的,是仅基于输出数据的模态参数识别。成为目前工程结构系统识别十分活跃的研究课题,也是一种挑战。本文主要研究了环境激励情况下,土木工程结构的模态参数识别问题。对频域的峰值法和时域的随机子空间识别的理论算法、计算机实现和实际应用进行了深入的研究。完成的主要工作和结论如下: 1.系统地讨论了环境激励情况下模态参数识别频域方法,重点研究了峰值法和频域分解法,对峰值法改进的途径进行了研究,建议采用平均正则化功率谱,并借助传递函数幅角辅助进行峰值选取,使峰值的选取更加客观准确。频域分解法本质上是基于奇异值分解的峰值法,可以比较客观的选择特征频率和识别相近的模态,识别精度高,是目前较先进的频域识别方法。2.详细讨论了时域随机子空间识别基本理论和算法,包括协方差驱动随机子空间识别和数据驱动随机子空间识别。提出了基于稳定图的平均正则化稳定图算法,辅助进行模态参数的自动识别,适应大型土木工程结构分组测试的特点。平均正则化稳定图将不同阶数模型计算的结果综合考虑,提高识别效率和识别精度。分析比较表明,协方差驱动和数据驱动随机子空间方法都可以有效识别结构的模态参数,数据驱动随机子空间方法理论上会比协方差驱动随机子空间方法识别结果更稳定、更精确,但计算时间相对要长些。通过算例详细比较分析了这两种随机子空间识别中不同的加权方法对识别结果的影响。3.基于VC平台开发了土木工程结构模态分析软件MACES,用计算机实现了模态参数识别的频域峰值法,包括不同加权方法的时域随机子空间识别算法,可以方便、快捷和高效地完成大型土木工程结构模态参数识别的全过程。主要功能包括更多还原

【Abstract】 The study on modal parameter identification of civil engineering structures-theory, implementation and application is support by the Natural Science Foundation of China (NSFC), under grant number 50378021. Civil engineering structures are important part of national infrastructures that are directly related to people’s daily life. It is required to understand their dynamic properties during the inspection, assessment and health monitoring of civil engineering structures. Structural modal parameters reflect the dynamic properties that can be obtained by either traditional or ambient vibration identification techniques. Dynamic testing of civil engineering structures under ambient vibration excitation has many advantages, such as no excitation equipment needed, no interruption of structural service conditions and less test time, which is more close to the real working conditions of civil engineering structures. As only output is measured and real input remains unknown in terms of ambient vibration testing, the modal parameter identification will therefore need to base itself on output-only data. The output-only modal parameter identification is different from traditional one that is based on both input and output data and is now very active and challenging research topic in the system identification of engineering structures. The state-of-the-art studies are carried out in this thesis on the modal parameter identification of civil engineering structures under operational conditions. The research is focused on theory, algorithm, implementation and real case applications of peak-picking method in frequency-domain and stochastic subspace identification in time-domain. The main work and conclusions include as follows: 1. The frequency domain methods of structural modal parameter identification under ambient excitations are comprehensively discussed in the thesis. More focus is on the theory and algorithm of pick-peaking (PP) and frequency domain decomposition (FDD) methods. To improve the peak-picking method, the phase angles of transfer functions are proposed to help to select the correct peaks from the average normal power spectrum densities so that the picked peaks are more accurate and objective. The frequency domain decomposition method is basically the singular value decomposition based peak picking technique. It makes the picked peak more objective and can identify the closed-space modes. The frequency domain decomposition method is now the advanced modal parameter identification method in frequency-domain. 2. The theoretical background and algorithms of both covariance-driven and data-driven stochastic subspace identification (SSI) are studied. The average normalized stabilization diagram algorithm based on stabilization diagram is presented in the thesis. The proposed algorithm can help identifying the modal parameters automatically and is more suitable to the measurements of large-scale civil engineering structures where the measurements are normally divided into several setups. It is demonstrated that both covariance-driven and data-driven stochastic subspace identification techniques can effectively identify the structural modal parameters. Theoretically, the data-driven stochastic subspace identification is more stable and accurate than covariance-driven stochastic subspace identification, but it needs much computing time. In addition, the different weighted methods on the decomposed matrix are compared with the help of numerical examples. 3. A VC based modal analysis software for civil engineering structures-MACES is developed where the algorithms of peak picking method in frequency-domain and stochastic subspace identification in time-domain are implemented. The graphical user interface (GUI) software includes the data pre-processing, parameter identification and post-processing that fit well with the particular features of civil engineering structures. With this software, the whole process of modal parameter identification for real structures can be carried out easily and effectively. It is suggested that peak picking method in frequency-domain and stochastic subspace identification in time-domain can be used complementarily. 4. As real applications, the field ambient vibration tests of Qingzhou cable-stayed bridge and Jian concrete filled steel tubular arch bridge are described in detail. The modal parameters of both bridges have been identified by peak picking method in frequency-domain and stochastic subspace identification in time-domain. The identified results from ambient vibration measurements agree well with those calculated from finite element method. It is demonstrated that the ambient vibration response measurements are sufficient enough to identify the most significant modes of large span cable-stayed bridges, in despite of the rather low level of ambient vibration signal captured, the low range (0-1.0Hz) of natural frequencies of interest,and the relatively dense modes of vibration in that range. With the help of the average normalized stabilization proposed in the thesis, the fake modes can be deleted and modal parameters can be identified accurately. 5. The stochastic subspace identification (SSI) algorithm is an advanced technique to perform such an operational modal analysis. However, a white noise assumption of inputs is compulsory, which may limit the SSI application to real civil engineering structures. The white noise assumption is discussed in the thesis in terms of theoretical analysis, numerical simulation and real case verification. It is illustrated that the stochastic subspace identification can be still validated if the non white noise is generated from a linear and time-invariant shaping filter (a simulated structure). Loosening the white noise assumption of inputs in the stochastic subspace identification is extremely important for the modal parameter identification of civil engineering structures. 6. A newly developed signal processing technique, empirical mode decomposition (EMD), is capable of dealing with non-stationary signals. An EMD-based stochastic subspace identification from operational vibration measurements is presented in the thesis. The output only measurements are first decomposed into the modal response functions by using the EMD technique with the specified intermittency frequencies. The stochastic subspace identification method is then applied to the decomposed signals to identify the modal parameters. A case study of the operational measurements from a real bridge is presented to illustrate the applicability of the present technique. It is demonstrated that the stable pole in the stabilization diagrams becomes sole and the vibration characteristics are easily identified for the decomposed signals ignoring the influence of other modal components and fake frequencies due to unwanted noise.更多还原