Experimental Identification of Bearing Stiffness in a Rotor Bearing System

Abstract

This paper presents the dynamic analysis of a dual-disk rotor supported over a double-row ball bearing system. Finite element model of the rotor is employed to know the approximate range of first few bending frequencies and the corresponding resonance amplitudes as a function of bearing stiffness coefficient. A non-parametric model based on back-propagation neural network is developed in order to assess the bearing stiffness for known values of frequency shifts and resonance amplitudes with respect to rigid bearing support conditions. A scaled prototype of the rotor dynamic test rig with the disks is fabricated and experimental modal analysis is conducted using electrodynamic excitation technique. Frequency modulated sine-sweep tests are conducted in two different conditions: first by laying it over the ballbearing system whose stiffness has to be ascertained and then by mounting it over rigid supports. The non-dimensional frequency shifts and amplitude ratios from the experimentally obtained frequency response curves are supplied as inputs to previously formulated neural network estimator model and the corresponding stiffness of the bearing is reported.