The probabilistic model of roller bearing vibrational signals in the form of periodically corre-lated random processes (PCRP) is considered in the paper. The properties of given approach are shown and the advisability of its using in the tasks of roller bearing unit vibrodiagnostics is analysed. The task of estimating the periodical changeable PCRP spectral density and its Fourier components with apriori unknown correlation period is formulated. The coherent method of PCRP spectral density statistical estimating are elaborated. The examples of elabo-rated methodology of coherent spectral estimation application for vibrodiagnostic of technical state of the highspeed spindle rolling bearing units are considered. Keywords: periodically correlated random processes, spectral density, vibrations, bearing, coherent.