Influence of Noise in Correlation Function Estimates for Operational Modal Analysis

The modal parameters in Operational Modal Analysis (OMA) are often estimated based on non-parametric signatures of the structure's dynamic response. For time domain OMA methods the non-parametric signatures are often correlation functions (CFs) and the pre-processing step for these methods is thus the estimation of CFs. The present paper demonstrates how measurement noise from sensors and measurement equipment affects the estimated CFs. Furthermore, the influence of the measurement noise on the modal parameter estimates is discussed. It is shown how effects of this noise can easily be avoided by ignoring the first part of the CFs when estimating the modal parameters. This is demonstrated by a theoretical review and on simulated and experimental data. The paper also addresses how to add noise to simulated data, so that it resembles a real-life scenario.