Application of SVM Classification Technique in Single-phase AC Motor Bearing Fault Diagnosis using Motor Current Analysis with MRA - FFT as Feature Extractor and CFFS as Feature Selection Method

Xyrelle Mae Agustin,Ervin Jan Samson,C. Ostia
DOI: https://doi.org/10.1109/ICSGRC57744.2023.10215397
2023-08-05
Abstract:Induction motors dominate a significant part of the industry since it is the most well-known type of motor used in industrial, commercial, and domestic contexts. Considering the harsh conditions in most industrial environments, induction motors are frequently prone to issues such as bearing faults. This study proposes a real-time bearing fault-diagnostic model for a single-phase AC motor using a data acquisition device made from Arduino Uno R3 and SCT-013-030 current sensor. Machine learning algorithms such as Multiresolution Analysis and Fast Fourier Transform (MRA-FFT) as feature extractor, Correlation and Fitness values-based Feature Selection (CFFS) as a feature selection method, and Support Vector Machine (SVM) as classification technique were also applied in this study to evaluate the performance of the proposed bearing fault detection model. Results showed that for the training and validation (80:20), the best accuracy is 89.39% at the SNR level of infinite dB. And for testing the performance of the trained bearing diagnostic model using the confusion matrix, the best accuracy is 88% for the SNR level of infinite dB.
Computer Science,Engineering
What problem does this paper attempt to address?