基于振动数据识别的模态参数（频率、振型和阻尼比），可以表征坝体结构真实的动态特性，作为大坝长期运行期间健康状况的评价指标，还可以用于更新大坝分析模型。因此，准确识别模态参数对大坝的健康监测和损伤评估具有重要意义。为实时处理大量连续监测数据，有必要发展模态参数自动识别算法。本文以高拱坝为研究对象，研究模态参数自动识别方法以实时监测坝体结构健康状况。论文的主要工作如下：（1）基于大岗山拱坝无质量地基模型开展SSI算法用户定义参数的敏感性分析，提出相关选择建议生成高质量的稳定图以提高识别精度；使用大岗山拱坝无限地基模型和溪洛渡拱坝的实测数据验证该参数选择建议的合理性；提出采用基于密度的噪声应用空间聚类算法（DBSCAN）解释稳定图实现模态参数的自动识别，并在大岗山拱坝和Baixo Sabor拱坝得到较好运用。（2）提出采用盲源分离（BSS）和奇异谱分析（SSA）提取固有模态函数（IMF），其中BSS算法处理环境激励数据，SSA算法处理地震记录，解决了经验模态分解（EMD）的端点效应和模态混叠现象；研究信号长度对希尔伯特变换（HT）识别模态参数的影响，并提出自动滑移窗口方法（AMW）解决HT的端点效应以提高识别精度。实例分析表明改进后的希尔伯特-黄变换（HHT）算法在提取密集模态方面具有很好的性能。（3）通过将SSI算法、BSS算法和HT算法相结合，提出SSI-BSS-HT自动识别算法，能够处理欠定问题、密集模态、非比例阻尼和时变刚度等问题，并且具有较高的计算效率。构建六自由度系统和十五自由度系统数值模型对其识别性能进行了验证。（4）基于自动识别算法，追踪拉西瓦拱坝的模态参数，揭示了环境因素对模态参数的影响；基于识别的模态参数对拉西瓦拱坝有限元分析模型进行了更新，可以较好地表征拉西瓦拱坝真实工作性态。
The modal parameters (frequency, mode shape, and damping ratio) identified from vibration data can depict the true dynamic characteristics of the dam. They may be used as the evaluation indexes of the health state of dams during the long-term operation monitoring. Meanwhile, they are also essential parameters for updating the seismic analysis model of dams. Therefore, identifying the modal parameters accurately is of great significance for the health monitoring and damage assessment of dams. To process a large amount of monitoring data and monitor the health state of dams in real time, it is interesting to develop automatic identification algorithms for modal parameters. The research content of this thesis is as follows.(1) To generate high-quality stabilization diagram to accurately identify the modal parameters of concrete dams, the sensitivtiy analysis of user-defined parameters in stochastic subspace identification (SSI) algorithm is carried out based on the finite element model with massless foundation of Dagangshan dam, and the corresponding selection suggestions are given. The rationality of the selection suggestions is verified using the finite enement model with unbounded foundation of Dagangshan dam and the measured data of Xiluodu dam. The density-based spatial clustering of applications with noise (DBSCAN) algorithm is adopted to interpret the stabilization diagram to realize automatic identification.(2) The blind source separation (BSS) and singular spectrum analysis (SSA) are adopted to extract the intrinsic mode functions (IMF), avoiding the end effects and mode mixing of empirical modal decomposition (EMD). The influence of data-length on the identification of HT is investigated, and the automatic moving-window (AMW) method is adopted to deal with the end effects of HT. The improved HHT can deal with the closely-spaced modes well and improve the identification accuracy.(3) By combining SSI algorithm, BSS algorithm and HT algorithm, the automatic identification method named SSI-BSS-HT is proposed. This method has high computational efficiency, and can deal with under-determined problems, closely-spaced modes, time-varying stiffness, and non-proportional damping problems. A six-degree-of-freedom and a fifteen-degree-of-freedom system model were constructed to verify its identification performance.(4) The modal tracking of Laxiwa dam is realized based on the proposed automatic identification algorithms, and the influence of environmental factors on modal parameters is investigated. The updated finite element model can better characterize the real working behavior of the dam system.
