Automatic diagnosis of defects in bearings using Fuzzy Logic

Abstract

In order to improve industrial competitiveness, the cost reduction in industrial plant maintenance is becoming increasingly important. A methodology applied to improve reliability in the production and to reduce operational costs is based on predictive maintenance. In this context there is a need for the optimization of diagnosis systems in order to increase precision and to reduce human errors. The automation of diagnosis processes results directly in improved reliability for decision taking. The problem caused by errors in the diagnosis tends to be amplified when large an industrial plant is concerned where a large number of monitored points is needed. Automatic diagnosis systems should be robust to a point where it must operate with a diversified source of information allowing for analysis of different equipment and existing defects and problems. The Fuzzy Logic is applied in the present work since as it is well known, this technique is a flexible tool, which allows the modeling of uncertain, and ambiguous data frequently found in real situations. In the classical logic theory, a element is either part of a given set or not. In fuzzy logic theory, a element can be or not part of a given set and also, can partially be a member of the set thus characterizing the fuzzy sets. This work presents an automatic diagnosis system for the classification of defects in bearings based on fuzzy logic. The system was developed to classify different types of defects in rolling bearings operating under several rotating speeds and load conditions. The experimental data bank used was obtained in a bearing defects simulator apparatus. The acquired signals were analyzed by statistical and spectral techniques for monitoring and diagnosis such as RMS, skewness, kurtosis, and power spectrum density methods The results obtained by using fuzzy logic for classification were conclusive showing that the system is capable to identify and to classify defective bearings.