BEARING FAULT DETECTION FOR ON-LINE QUALITY CONTROL OF ELECTRIC MOTORS

Jirí Vass, Cristina Cristalli

Abstract

This paper is concerned with analysis and classification of vibration signals from universal electric motors. The goal is to reveal manufacturing defects caused by assembly machines composing the motors on a production line. Such machines may under certain conditions use inappropriate force, and hence cause a mechanical shock, resulting in a damage of the motor bearings. The proposed system comprises a preprocessor based on continuous wavelet transform in order to reduce the noise masking the characteristic frequencies of bearing faults. The noise reduction is based on the adaptive Morlet wavelet and soft-thresholding of wavelet coefficients. Individual blocks of the preprocessor are presented and important practical issues are considered, such as segmentation and proper selection of wavelet scales. Identification of defective motors is performed by a simple and effective technique based on autocorrelation function, utilizing prior information on vibration frequency features. Finally, two error measures are designed in order to evaluate influence of a simulated noise level on efficiency of the fault diagnosis system. Achieved results appear to be promising and applicable in automatic quality control.

Keywords

rolling bearing, vibration analysis, continuous wavelet transform

Download

IMEKO-TC10-2005-18.pdf

DOI

-

IMEKO TC

TC10 - Measurement for Diagnostics, Optimization and Control

Back to the proceedings