DETECTION OF ROLLER BEARING DEFECTS USING EXPERT-SYSTEM AND FUZZY-LOGIC

A fuzzy expert system, Bearing_Expert, has been developed. This expert system is PC-based. It is menu driven and it is very user-friendly. T his expert system offers specific knowledge of various aspects about b earing monitoring, such as diagnostic methods, defect frequencies, fea ture selection, fuzzy bearing classification etc. By responding to dif ferent questions successively, the user can obtain very specific knowl edge via the shortest path. In order to demonstrate the effectiveness of Bearing_Expert, experiments have been conducted. Radial acceleratio ns have been measured and analysed. Feature selection has been perform ed effectively by the integration of Bearing_Expert and a computer pro gram. The average of the top five values of frequency response in the high frequency region (5-22 kHz) has been found to be the best feature for the detection of roller bearing defects. Fuzzy reasoning has also been used via Bearing_Expert. Piecewise Linear membership functions h ave been developed. A 100% reliability has been achieved for the detec tion of roller bearing defects via fuzzy reasoning. Bearing_Expert is very beneficial for factory automation and preventive maintenance. (C) 1996 Academic Press Limited