An improved automated framework for operational modal analysis with multi-stage clustering and modal quality evaluation

Numerous efforts have been made in last decade on fully Automated Operational Modal Analysis (AOMA) with field measurement. Considerable user interaction is required especially when parametric system identification methods are involved with Stability Diagrams (SD) for the modal identification. This paper proposes an improved procedure with multi-stage clustering to address this issue. It is generally applicable to any method that relies on SD for the modal identification. The proposed clustering framework offers advantages over traditional AOMA with reduced reliance on the threshold setting in the hierarchical clustering. It relies on hierarchical clustering to identify only one cluster center. This cluster center is then used as input for Max-min distance clustering to achieve adaptive clustering. The proposed Modal Evaluation Index can be employed to eliminate the effect of subjective bias to small dimensional clusters and to assess the validity of each representative mode identified from the cluster. Measured datasets from the Z24 bridge benchmark and the Yingxian wooden pagoda serve to illustrate the performance and effectiveness of this automation strategy.