F1000Research (Sep 2021)
Detecting industrial motor faults with current signatures [version 1; peer review: 2 approved]
Abstract
Background: A major player in industry is the induction motor. The constant motion and mechanical nature of motors causes much wear and tear, creating a need for frequent maintenance such as changing contact brushes. Unmannered and infrequent monitoring of motors, as is common in the industry, can lead to overexertion and cause major faults. If a motor fault is detected earlier through the use of automated fault monitoring, it could prevent minor faults from developing into major faults, reducing the cost and down-time of production due the motor repairs. There are few available methods to detect three-phase motor faults. One method is to analyze average vibration signals values of V, I, pf, P, Q, S, THD and frequency. Others are to analyze instantaneous signal signatures of V and I frequencies, or V and I trajectory plotting a Lissajous curve. These methods need at least three sensors for current and three for voltage for a three-phase motor detection. Methods: Our proposed method of monitoring faults in three-phase industrial motors uses Hilbert Transform (HT) instantaneous current signature curve only, reducing the number of sensors required. Our system detects fault signatures accurately at any voltage or current levels, whether it is delta or star connected motors. This is due to our system design, which incorporates normalized curves of HT in the fault analysis database. We have conducted this experiment in our campus laboratory for two different three-phase motors with four different fault experiments. Results: The results shown in this paper are a comparison of two methods, the V and I Lissajous trajectory curve and our HT instantaneous current signature curve. Conclusion: We have chosen them as our benchmark as their fault results closely resemble our system results, but our system benefits such as universality and a cost reduction in sensors of 50%.
