泄流激励下水工闸门运行模态参数自动识别

为了能够自动快速准确的识别泄流激励下水工闸门的模态参数，解决系统稳态图自动化定阶困难和真实模态筛选的问题，该研究提出了一种改进的势能聚类（Potential-based Hierarchical Agglomerative，PHA）的协方差驱动随机子空间（Covariance-driven Stochastic Subspace Identification，COV-SSI）水工闸门自动模态参数识别方法。首先采用小波阈值的方法对采集信号进行去噪处理。再通过奇异值加权判断指标（Singular-value Weighted Index，SWI）进行系统阶次及稳态图最大阶次的自动化确定，剔除稳态图中的噪声模态，进一步使用势能聚类方法实现结构模态参数的自动化识别。将该方法应用于泄流激励下弧形闸门试验模型中，并与锤击法测试结果进行对比分析，结果表明，所提识别方法在无需人工激励条件下，能够实现系统阶次和稳态图最大阶次的自动化确定，剔除稳定图中虚假极点。通过与试验对比，该方法平均相对误差为3.5%，可用于泄流激励下水工闸门结构运行模态参数的自动智能化识别。

Automatic identification of the operational modal parameters for hydraulic gates under flow release excitation

Abstract: Structural health monitoring (SHM) of hydraulic steel gates is one of the most important technologies for the safety of water conservancy and agriculture. Among them, the modal parameter identification of SHM can provide the key information for the gate vibration control, model correction, and damage identification. However, the calculation peaks can be generated by the overestimation of the model order, while the noise spikes can be introduced by the measurement noise in the process of modal parameter identification of hydraulic gate operation. There is also a great interference with the modal parameters. Some manual participation can be required in the modal model grading and modal selection during steady-state graph recognition. In this study, an improved potential energy clustering (PHA) covariance-driven stochastic subspace (COV-SSI) hydraulic gate automatic modal parameter identification was proposed to automatically identify the operational modal parameters of hydraulic gates under the flow release excitation. The modal parameters of the hydraulic gate were identified using only the output response signal under the structural drainage excitation. The dynamic characteristics of the structure were also revealed under the real boundary and load working conditions. Firstly, the vibration signals were collected to process by noise reduction using the wavelet threshold method. The spurious modes were then reduced to optimize the quality of vibration signals due to environmental noise. Secondly, the Toeplitz matrix was constructed using the COV-SSI. The response signals were then obtained to decompose by singular value decomposition (SVD). The system order n and the maximum order nmax of the steady-state graph were automatically determined by the singular value weighted judgment index (SWI). The noise mode was also eliminated in the steady-state graph. Then, the PHA was used to realize the automatic identification of structural model parameters. Finally, the accuracy of the improved model was verified by the numerical calculation of the two-degree-of-freedom mass-spring-damping system, followed by an experimental model of an arc gate under drainage excitation, and the comparison with the hammering test. The results show that the new identification was automatically determined the system order and the maximum order of the steady-state diagram without the artificial excitation, and the false poles in the steady-state diagram. An automatic identification was also realized for the model parameters of hydraulic gates under drainage excitation. The maximum and average relative errors of the test were 8.5% and 3.5%, respectively, compared with the hammering method. Therefore, the COV-SSI and potential clustering can be expected to identify the online modal parameters of hydraulic gates. The automatically fixed-order stability diagrams were reduced the influence of human factors, indicating the better identification of modal parameters in the hydraulic gates under the discharge excitation. This finding can provide a promising application for the health monitoring and safety analysis during the hydraulic gates in service. Especially, the artificial forces and excitation signal measurement were greatly reduced for the modal parameters.