An Automated Operational Modal Analysis (AOMA) Builder Algorithm for Civil Engineering Applications

Due to the growing number of ageing structures and infrastructures that require monitoring to ensure adequate safety, automation in dynamic identification has gained increasing importance. This thesis proposes the development of an Automated Operational Modal Analysis (AOMA) Builder Algorithm designed and tested for civil engineering applications. In this field, determining modal parameters becomes more challenging due to the structures' large size and low frequencies. Moreover, configurable and quantifiable external excitations are often disregarded due to the devices' high cost and weight. Given the unsupervised learning and unmeasured input process, the primary objective is to automate the selection of potential physical modes from the stabilisation diagram obtained using the Covariance-driven Stochastic Subspace Identification (Cov-SSI) method. The automation class includes various selectable clustering algorithms, such as Gaussian Mixture Models (GMM), K-means, Hierarchical Clustering, Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN), Ordering Points to Identify the Clustering Structure (OPTICS), Spectral Clustering, and Affinity Propagation. Once the end-user has selected which algorithms to run, implementing the mapping phase becomes indispensable for obtaining a single result in terms of natural frequencies, damping, and mode shapes. The aim is to merge similar outcomes using representative metrics of the data distributions across all clusters. Additionally, results are evaluated using outlier statistics and various voting and consensus methods to assess the robustness and reliability of the selected algorithms. Finally, the proposed method is validated through a case study involving a tailored numerical model and real accelerometric data from the Heritage Court Tower (HCT).