Utilização de filtros adaptativos para detecção de falhas em mancais de rolamentos

Abstract

The analysis of vibrations constitutes one of the powerful tools destined for predictive maintenance. In this context, and with the objective of contributing to the improvement of the process of detection and diagnosis of ball bearing faults, the use of the technique of adaptive filtering applied to vibration signals is focused in this work, through the minimization of the average quadratic error between the output of the filter and a reference signal considered as desired, with and without delay applied to the input signal (LMS), and the cancellation of the gradient of the objective function in terms of the error signal (RLS), based on Kalman filtering. These vibration signals are considered to be unknown regarding the behavior of their statistical properties. Initially, these methods were applied to mathematical models of imperfections contaminated with noise in the outer race (stationary), inner race, balls and cage of the bearing, with different fault intensities. Later, the technique was used with vibration signals from ball bearings mounted on an experimental test rig, operating under adverse conditions, as bent shaft and lack of lubrication, what favored the sprouting of defects on the cage of axial self-aligning ball bearings. The tracking of the evolution of the defects for detection and diagnosis was done by the analysis of statistical parameters such as RMS level, crest factor, K factor, sixth order central statistical moment and envelope technique, for both cases where the method was used. The results show that for both, simulated and test rig case, the adaptive filtering technique contributed to the improvement of the performance of the methods used for detection and diagnosis of faults, especially the envelope method, in which case reasonable estimates for the signals of the defective component were obtained, and the influence of other vibratory sources was minimized.