Operational Modal Analysis (OMA) extracts modal parameters of systems that are excited by unknown ambiental excitation, being broadly used in the monitoring of civil structures. In the past decades, research has been made to enable the application of OMA in rotating machines, dealing with challenges such as harmonic excitation, nonlinearities, and the lack of proper excitation. With the popularization of machine learning techniques, many researchers have been using these tools to overcome some challenges in this research field. Clustering techniques, that can group information about datasets without prior knowledge of their characteristics, has been used along with statistical methods to automate OMA so that it can be used for condition monitoring. Recently, automatic OMA (AOMA) was applied to rotating machinery data. This paper evaluates one of these AOMA algorithms, successfully tested with data from a test rig with a rotor supported by hydrodynamic bearings, in a more complex dataset, with data from a rotor supported by magnetic bearings and influenced by gas seal. The results show that the proposed algorithm can extract modal parameters close to the ones extracted by EMA and by the mathematical modeling of the test rig, being robust even when a more complex system is analyzed.