Machinery Fault Diagnosis Based on Feature Level Fuzzy Integral Data Fusion Techniques

Abstract

Fuzzy methods for machinery fault diagnosis are able to classify fault patterns in a non-dichotomous way thereby imitating the way humans process vague information. As an outgrowth of classical set and measure theory, fuzzy measure and fuzzy integral theory has the ability to infer the importance of each criterion and represent certain interactions among them. Based on fuzzy measure and fuzzy integral theory, a novel feature level direct fuzzy data fusion approach for machinery fault diagnosis is presented. Fuzzy analysis method was used to obtain the membership values of each feature for each fault class. The Choquet fuzzy integral data fusion method was employed to produce the diagnostic result using different features. Current and vibration signals from electrical motors were used to validate the method. Results showed that the proposed feature level fuzzy measure and fuzzy integral fusion approach performed very well for electrical motor fault diagnosis.