Sensitive Inter-turn Fault Detection Approach for Induction Motor Under Various Operating Conditions

Gundewar, Swapnil K.,Kane, Prasad V.

DOI: https://doi.org/10.1007/s13369-023-07664-5

IF: 2.807

2023-03-22

Arabian Journal for Science and Engineering

Abstract:Induction motor fault diagnosis and condition monitoring are on the anvil of researchers, who have reported many studies and developments in the literature. However, the reported technique requires intervention to interpret the frequency and time–frequency domain analysis results. Intelligent systems have been reported to overcome these limitations; however, time delays and additional costs are incurred owing to the need for expert interpretation. Hence, this work proposes a methodology based on motor current signal processing in image format to automatically diagnose and classify the inter-turn fault of the three-phase induction motor. The motor current sampled in the time domain by shortening the inter turns in the winding for different load conditions is expressed as three-color monotonous images for each phase. These obtained images are applied for training and testing of convolution neural network, which can classify the faults according to their severity and identify the phase in which the fault prevails. Furthermore, efforts have been made to design the convolution neural network architecture, i.e., selecting hyperparameters such as learning rate, batch size, number of epochs, and momentum rate, and average classification accuracy of 99.38% is obtained. The proposed technique with monotonous coloured images of current signal as an input to CNN is found to be effective in terms of computational time, robustness to identify the faults even in the presence of noise and fault classification accuracy when compared to other published literature.

multidisciplinary sciences