Input-output and output-only modal analysis in the frequency domain: a comparison.

Vibrations are an omnipresent phenomena and generally their presence is sought to be limited since they may cause malfunction, noise, wear, discomfort and destruction. In the last century, methods and technologies regarding to dynamic analysis developed significantly. Modal analysis constitutes an important section of these developments and all the numerous approaches it has can be grouped in two major categories: Experimental Modal Analysis (EMA) and Operational Modal Analysis. Because of the sheer number of modal analysis approaches and them being constantly developed it is hard to come by up-to-date comparisons of different methodologies under different circumstances. In this thesis, modal parameter extraction method Rational Fraction Polynomials method in z-domain (RFP-z) is elaborated and evaluated by creating a numerical model to run digital simulations. Other than a comparison of EMA and OMA accuracy, different parameters and their effect on accuracy are contrasted as well since the performance of the extraction may change depending on factors such as noise, excitation typology and frequency. Results indicate with numerical evidence that EMA is significantly more accurate and should be utilized if possible, especially for the analysis of critical parts. The effect of the presence of noise can be mitigated for well-separated modes by increasing the model order of RFP-z, but it affects the extraction accuracy of closely placed modes in a critical and unsalvageable way. Results also highlight that RFP-z method works much more accurately when the excitation applied to the structure is arbitrary in nature, rather than harmonic. In case of hybrid excitations, the “random” contribution should dominate the load behaviour. Another finding is that if of steep peaks are present in the frequency response function (FRF), it should be estimated with a higher number of spectral lines, otherwise the low resolution causes a loss of information ergo an inaccurate modal parameter extraction. The effect of RFP-z maximum model order is very situational and a general conclusion cannot be reached in that regard.