Prediction of rotor slot width in induction motor using Dyadic wavelet transform and softmax regression

Abstract

Predictive maintenance is required for Induction Motor (IM) to avoid sudden breakdown. In this paper, multimodal sensor data are used for prediction of the rotor slot width variation during runtime of IM. Moreover sensor data are visualized through visual image obtained using Scale Invariant Feature Transform (SIFT) and matching dictionary for various faults in motor such as overload, high speed, Broken Rotor bar (BRB), and unbalance magnetic pull. The multimodal sensor signals data acquired from different parts of three phase induction motor are analysed using various transforms such as Over Complete Rational Dilation Wavelet Transform (ORaDWT), Tuneable Q Wavelet Transform (TQWT), Polynomial Chirplet Transform (PCT) and Dyadic Wavelet Transform (Dyadic WT) for various fault conditions and rotor slot width mentioned during runtime condition of motor. The rotor slot width is predicted using Multiple Linear Regression (MLR), Polynomial Regression (PR), Logistic Regression (LR) and Soft-max Regression (SR) methods through the mean value and energy band of the acquired sensor signal. The Dyadic WT and SR perform better for rotor width prediction. The proposed method such as Dyadic WT and SR provides the prediction accuracy of about 95.2%. From experimental results the rotor slot width expansion more than 2% needs immediate attention to avoid breakdown of motor.

Prediction of Breast Cancer Diseases From Genes Using Pso and Cso Tuned Long Short-term Memory

“…GRU-RNN classifier, tested on dataset [15]. The merits and demerits of various regression algorithms are used for prediction is explained in the articles [16][17][18][19][20]. Ensemble learning approach for cancer classification, utilizes particle swarm optimization for breast cancer prediction and the performance is analyzed [21].…”

A Comprehensive Methodology for CNN Based Fault Identification in Induction Motors – A Case Study for EV’s

“…However, the frequency domain methods do not provide information about motor signals containing transients. To cater for this, the wavelet transforms as a fault diagnostic technique is also used by the researchers enabling them to observe information about both time and frequency of a particular signal [16], [17].…”