Automated modal identification using principal component and cluster analysis: Application to a long‐span cable‐stayed bridge

Abstract

Obtaining timely information about the health of civil infrastructure is critical to ensuring safe and reliable operation. Structural health monitoring has been proposed as a means to provide such information; however, most structural health monitoring systems provide only raw data, rather than actionable information. It is necessary to develop automated modal analysis strategies that can provide near real-time dynamic information regarding the in-service state of a bridge, which is essential to vibration control, finite element model calibration, and damage detection for safety and serviceability condition assessment. This study presents an automated framework to extract structural modal parameters from the stabilization diagram using a parametric modal identification method such as stochastic subspace identification. The framework focuses on the automated modal analysis issues of an in-service long-span bridge with close-frequency modes. The presented framework is validated using experimental tests of a 1.8-m 18-story laboratory model. Subsequently, data from Sutong Cable-Stayed Bridge are employed to demonstrate its potential usage in the field. Finally, an application of the automated framework is presented to identify and track the modal parameters of the deck of Sutong Cable-Stayed Bridge for 20 days. Results show that the presented framework can successfully extract the structural modal parameters with good accuracy and robustness, hence can provide a reliable technical support for in-service monitoring of long-span bridges.