Automatically Extracting Bridge Frequencies Using SSA and K-Means Clustering from Vehicle-Scanned Accelerations

Abstract

This paper removes the subjective judgment by proposing a technique for automatically identifying the bridge frequencies using singularity spectrum analysis and [Formula: see text]-means clustering. First, the vehicle-bridge contact-point acceleration newly derived from is used to avoid the adverse effect brought by vehicle’s frequency. Next, the contact signal is processed by singular spectrum analysis to remove the noises. Then, the first several principal components are processed by the FFT to extract peak frequencies. Finally, the [Formula: see text]-means clustering algorithm is adopted to group the peak frequencies into clusters, with their centroids denoting the bridge frequencies. They are verified against those extrapolated from the 1st bridge frequency. Moreover, the current technique was assessed by studies on factors including pavement roughness, vehicle velocity, ongoing traffic, and vehicle stiffness, along with a field test on a bridge to demonstrate its capability. One feature of the current method is that the errors for higher frequencies predicted are not higher than lower frequencies.