Automated modal parameter identification method for bridges based on cluster analysis

Abstract

To realize the automated identification of modal parameters for bridges, according to the stabilization diagram produced by stochastic subspace identification (SSI), an automated modal parameter identification method for bridges was proposed based on principal component analysis (PCA), k-means clustering method and hierarchical clustering method. First, according to the principal components of the modal validation criteria (MVC) produced by PCA, the false modes in the stabilization diagram were pre-eliminated by using the k-means clustering method. Then, the relationship between the number of the truncated clusters and the number of the effective modes was studied to determine the optimal number of clusters for hierarchical clustering. Finally, an automated modal parameter identification method for bridges was established. The scaled-model tests and field measurements of a railway bridge were carried to verify the proposed method. The results indicate that the false modes in the stabilization diagram can be effectively removed by using the proposed method. The number of the effective modes can be determined. The automation in the process of stabilization diagram produced by SSI can be improved. The automated modal identification of structural modal parameters of bridges based on field measurements is realized.