Identification, tracking and warning of vortex induced vibration using k-means clustering method

Identifying, tracking and warning of the vortexed induced vibration (VIV) are essential for long-span suspension bridges because VIV might cause the driving safety problem. This study proposes an automatic method to identify and track the start and evolution of the VIV. Two feature indices, i.e. the energy concentration coefficient (ECC) and similarity ratio of amplitude (SRA), are utilised to form a feature index vector to identify and track the VIV. The Hilbert transform is utilised to process the vibration signal, based on which the SRA can be extracted. The k-means clustering method is introduced to automatically identify and track the VIV through automatically separating the feature index vectors into several clusters, and multi-level warming is established based on the acceleration root mean square (ARMS) value. The proposed method is validated using the in-field monitoring data of a suspension bridge suffering the VIV. The results indicate that the proposed automatic method can successfully distinguish the VIV from ambient vibration and can identify the forming of the VIV and track the evolution without any manual intervention, even for the VIV with low vibration amplitude. The multi-level warning can also be achieved based on the ARMS value when the VIV is identified.