This work describes an expert system for detection and classification of ball bearing defects such as inner race, outer race, ball defects, ball roundness and race roundness inferiority using the vibration signals obtained from Anderonmeter.
In this work the hypothesis test thoery is introduced to decide whether a defect exists or not, so a decision is accepted or rejected by a specified level of significance. The sample probability density functions, which represent the probability that each defect does not exist, are obtained on the basis of classcifations made by an inspector.
Using the probability density functions and 5% level of significance, the success rate of the expert system is found to be 96.2% for inner race defect, 95.3% for outer race defect, 93.3% for ball defect, 95.3% for ball roundness and 97.1% for race roundness inferiority compared with the inspector's classifications.