An Unsupervised Automated Method to Diagnose Industrial Motors Faults

• Main Authors: Sheikh, M.A., Saad, N.R., Mohd Nor, N.B., Tahir Rakhsh, S., Irfan, M.
• Format: Conference or Workshop Item
• Published: Institute of Electrical and Electronics Engineers Inc. 2019
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85062097438&doi=10.1109%2fESARS-ITEC.2018.8607646&partnerID=40&md5=b73e8cbf70c23af9ce3f656455054ae4; http://eprints.utp.edu.my/23661/

In industries, induction motors are wilding used due to its large scale utilization, about 90 of the total industrial power is consumed by induction motors. Although induction motor have rugged structure, but still they are most oftenly subjected to unexpected mode of failure because of long operational duty without any predictive maintenance. Consequently, motors have to face various faults, among these faults the bearing faults are consider the major problem, if bearing faults are unnoticed at incipient stage, they will result in catastrophic damage to the motor. Therefore, predictive condition monitoring techniques should be introduced which continuously monitor the health of the bearing. The main objective of this paper is to identify and classify bearing localized and distributed defects at Inner for the various levels of fault severity. The conventional statistical analysis based on motor current signature analysis (MCSA), instantaneous power analysis (IPA), and vibrational analysis were unable to classify or discriminate bearing localized defects. Moreover, does not shed any light on the segregation of distributed defects. Hence, this paper present a new method known as autonomous fault identification and fault segregation based on current analysis. The proposed technique use current analysis method. The well- known combination of Park Vector Analysis (PVA) and Artificial Intelligence (AI) has been carried out to identify and classify the exact class of the bearing faults. In addition, the method is justified through hardware test rig, which was design through this research. Moreover, the results shows that the features of PVA, which are utilized by AI are not only sensitive to diagnose the faults but at the same time they are capable enough to segregate each class of bearing fault. The proposed method can be used a condition monitoring index, which may be generalized to other rotational machines in any industry. © 2018 IEEE.