Abstract

The present paper addresses Structural Health Monitoring (SHM) of large bridges by presenting a

strategy for automatic Operational Modal Analysis (OMA) based on Stochastic Subspace

Identification (SSI) combined with the machine Clustering learning algorithm. Data acquired from a

set of accelerometers installed on the 25 de Abril suspended Bridge, located in Lisbon, with over

1km central span, was used to test and validate the strategy. Natural mode information along with

mode complexity and collinearity features were obtained and analysed for the purpose of

automatic OMA, and allowed concluding that the combination of frequencies and mode shape

components were the most effective. Additionally, it was concluded that both partitioning and

hierarchical clustering associated with the SSI in its covariance version are effective in identifying

natural modes and in distinguishing between these and spurious ones in a fully automated

manner. It was also observed that imaginary and non-collinear mode shapes, such as those

observed in the case study used, can be challenging to automatic OMA strategies, but were

overcome by the strategy proposed herein.

Keywords: Suspended bridge, structural health monitoring, operational modal analysis, stochastic

subspace identification, clustering methods.