Enhanced methods of operational modal analysis
University of New Brunswick
In this thesis practical considerations for Operational Modal Analysis (OMA) are discussed and two new methods of analysis are proposed. First, it is shown that although natural frequencies may be accurately estimated, the estimated mode shapes when the input excitation is other than Gaussian white noise contain significant errors, and in most cases, accurate estimation is not feasible. A modal filter system is proposed which implements a nonlinear optimization algorithm to perform OMA on a system subjected to input excitations other than Gaussian white noise. The proposed method is validated using numerical parametric studies on a 4 DOF analytical system subjected to a generalized broadband excitation. Lastly, the application of automated OMA as a solution for practical Structural Health Monitoring is discussed and a novel scale space based frequency domain automated OMA method is proposed. The method is successfully validated with a 4 DOF analytical system and a comparative study is performed using vibration test data from a scale cable stayed bridge model.