Inter-turn fault detection of induction motors using a method based on spectrogram of motor currents. (December 2022)

• Record Type: Journal Article
• Authors: Ghanbari, Teymoor; Mehraban, Abbas; Farjah, Ebrahim
• Is Part Of: Measurement. Volume 205(2023)
• Publication Date: 2022-12
• Subjects: Fault Diagnosis -- Induction Motor -- Inter-Turn Short Circuit (ITSC) -- Kalman Filter -- Motor Current Signature Analysis (MCSA) -- Frequency Domain Analysis; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.112180
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: ITSC fault in IMs is investigated based on analyzing the time–frequency plane's image of the steady-state modal current signal. The criterion of the detection is considered as cumulative skewness and kurtosis of the image's histogram. A real-time KF estimator is employed for detecting any events and eliminating misleading harmonics from the IM current. A new thresholding method is proposed based on Otsu's image thresholding method. Abstract: Steady-state signatures seem more reliable than incipient fault signatures for detection of inter-turn short circuit (ITSC) fault in induction motors (IMs). An efficient detection approach is proposed based on the deviation of the histogram relevant to the time–frequency plane's image of the steady-state modal current signal from the standard normal distribution. First, the misleading frequency components like 3rd, 5th, and 7th harmonics are excluded using Kalman Filter (KF). Then, from the time–frequency spectrogram of the signal, the converted gray-level image and its histogram are obtained. A considerable deviation of the histogram distribution from a normal distribution is observed in the case of ITSC fault, used for the detection. A cumulative index including normalized skewness and kurtosis of the histogram is used as the criterion of the detection. Finally, a novel threshold setting method based on Otsu's threshold principle is proposed, which efficiently distinct faulty conditions from healthy operations. The required data for assessment of the approach is gathered from different experiments carried out on a test bench, subjected to various fault percentages and different load levels. The results confirm the effectiveness of the proposed methodology.