Performance of a Classifier Based on Time-Domain Features for Incipient Fault Detection in Inverter Drives

• Published in: IEEE transactions on industrial informatics Vol. 15; no. 1; pp. 3 - 14
• Main Authors: Bandyopadhyay, Indrayudh, Purkait, Prithwiraj, Koley, Chiranjib
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Circuit faults; Classifiers; Current distribution; Diagnostic software; Diagnostic systems; Fault detection; Feature extraction; Induction motor (IM) drive; Induction motors; Insulated gate bipolar transistors; Inverters; Maintenance costs; mean current vector (MCV); recursive feature elimination (RFE); Resistance; spurious resistance; support vector machine (SVM) classifier; Support vector machines; Time domain analysis; time-domain features
• ISSN: 1551-3203; 1941-0050
• DOI: 10.1109/TII.2018.2854885

In the field of industrial informatics and fault diagnostic applications, an important topic is about detecting and locating fault in an industrial machinery in its incipient stage. This enables avoiding severe faults that are often followed by unscheduled process interruptions and escalated maintenance cost. Inverter-driven induction motor drives, among other irregularities, suffer from degradation of switching device (such as insulated gate bipolar transistor (IGBT)) junctions giving rise to spurious resistance at those junction points. The present research attempts to identify those incipient yet progressive faults from motor current signature by a time-domain computation technique for feature extraction followed by the use of support vector machine classifier. Additionally, it is observed that spurious resistance faults in an inverter can be grouped in a particular pattern using mean current vector distribution. Different experimental results confirm that the proposed method is robust to change in motor parameters and load while giving satisfactory classification accuracy.