The rotor thermal sensitivity often affects the dynamic behavior of power unit generators. Owing to this phenomenon, increments of field current and other process parameters that are related to it may cause a shaft thermal bow and significant changes in the synchronous vibration. This symptom can also be caused by many other common malfunctions that affect rotating machines. Therefore, diagnostic techniques aimed at identifying the actual fault are very useful for optimizing maintenance activities. The thermal sensitivity of generator rotors can be deemed as a fault because it is commonly caused by a local deterioration of the winding insulation as well as by jamming phenomena between conductors and rotor slots, caused by friction forces due to the different thermal expansions of these components. This paper shows the results obtained applying a diagnostic method, based on multiple linear regression models, which has been developed for the analysis of generator vibrations caused by thermal sensitivity. Nevertheless, nonlinear relationships between vibration and process parameters have also been taken into account. The capabilities of this diagnostic technique have been validated using the analysis of experimental data collected in a power plant. The results of this investigation are shown and discussed in the paper.