Probabilistic investigation of error propagation in frequency domain decomposition-based operational modal analysis

Abstract

Each operational modal analysis (OMA) technique may produce significant errors during the identification procedure due to the applied methodology, environmental/operational conditions, and instrumentation. Consequently, those errors can adversely affect the quality of identified parameters. In this context, this study aims at providing a comprehensive discussion on the propagation of predictions errors in the frequency domain OMA. To mitigate the prediction errors those considered to be induced by modeling and measurement errors, an extended formulation is presented based on a recently developed Modified Frequency and Spatial Domain Decomposition technique. A comprehensive investigation is presented for the probabilistic modeling of output power spectral density (PSD), considering prediction errors. Numerical and real data applications are conducted to show the effectiveness of the proposed methodology.