Automatic Multi-stage Uncertainty-Aware Operational Modal Analysis with Transmissibility Measurements

This chapter introduces a transmissibility-driven multi-stage uncertainty-aware automated operational modal analysis framework that systematically integrates adaptive clustering and Bayesian inference to eliminate manual intervention in mode selection and uncertainty quantification. By leveraging the pole-aligned properties of power spectral density transmissibility, the framework first constructs raw stabilization diagrams through poly-reference least squares complex frequency-domain parameterization, generating axis-resolved stability patterns to bypass eigenvalue tracking ambiguities. A hierarchical clustering algorithm autonomously distinguishes physical modes from spurious modes using a self-adaptive silhouette coefficient-based criterion, replacing heuristic thresholds with data-driven spurious mode rejection via automatic analysis of stabilization features. The identified stable axes of physical modes then trigger Bayesian inference within adaptive frequency bands, propagating measurement uncertainties into posterior distributions of modal parameters, overcoming legacy limitations of manual tuning. Validated on an engineering structure, the framework demonstrates robust automation in modal identification and uncertainty-aware modal parameter tracking, making it particularly suited for structural health monitoring applications requiring scalable, engineer-independent analysis of systems with dense modal populations.