Research on automatic identification of modal parameters for transmission tower structures based on two-stage clustering under ambient excitation

Modal parameters play a crucial role in damage identification and load recognition. To address the issues of poor automation, difficulty in identifying and eliminating spurious modes, and challenges in recognizing torsional mode shapes in the classical stochastic subspace identification （SSI） method, a two-stage automatic modal parameter identification method is proposed. First, the stochastic subspace identification method is used to process the time-domain dynamic response. Then, the extracted sample features are subjected to Fuzzy C-Means （FCM） clustering to remove spurious modes. Subsequently, a new truncation threshold calculation method is employed to hierarchically cluster the extracted structural modes based on frequency, obtaining modal orders. A second hierarchical clustering is then applied to eliminate local modes, followed by a projection method to identify torsional modes. A numerical simulation and full-scale experimental verification were conducted using a transmission tower as a prototype. The results demonstrate that the proposed method achieves a frequency identification error of no more than 1%, with mode shape correlation coefficients exceeding 0.99. The method not only effectively removes spurious modes but also automatically identifies both translational and torsional mode shapes of the transmission tower structure.