International Journal of Advanced Engineering Application

Machine Learning-Based Real-Time Fault Detection and Predictive Maintenance in Three-Phase Induction Motors Using Vibration Signal Analysis and Convolutional Neural Network Feature Extraction

Author(s):Rajeev Kumar, Sanjay Bhattacharyya, Priyanka Mehta

Affiliation: Department of Electrical and Electronics Engineering, Rajasthan Technical University, Kota, Rajasthan Department of Mechanical Engineering, North Eastern Regional Institute of Science and Technology, Itanagar, Arunachal Pradesh

Page No: 91-96

Volume issue & Publishing Year: Volume 3, Issue 5, May 2026

Journal: International Journal of Advanced Engineering Application (IJAEA)

ISSN NO: 3048-6807

DOI:

Download PDF Cite this article

Abstract:
Induction motors are the backbone of industrial automation, constituting over 65% of global electrical energy consumption in manufacturing sectors. Unexpected motor failures cause significant production downtime and economic losses estimated at USD 647 billion annually worldwide. Traditional threshold-based fault detection methods suffer from high false-alarm rates and inability to detect incipient faults before catastrophic failure. This study presents a real-time intelligent fault detection and predictive maintenance framework for three-phase squirrel-cage induction motors using tri-axial vibration signals acquired at 25.6 kHz sampling frequency, processed through a custom eight-layer Convolutional Neural Network (CNN) architecture with Short-Time Fourier Transform (STFT) spectrograms as two-dimensional input representations. The framework classifies five motor conditions: healthy operation, bearing inner-race fault, bearing outer-race fault, stator winding inter-turn short circuit, and rotor bar breakage, across three load conditions (25%, 50%, 100% rated load) and two rotational speeds (1450 rpm and 2900 rpm). A dataset of 18,450 signal segments from twelve motor specimens tested on a custom dynamometer test rig was collected at the Motor Testing Laboratory, Rajasthan Technical University. Classification accuracy of 97.3% on the held-out test set surpasses benchmark SVM (89.4%) and Random Forest (91.2%) classifiers. Gradient-weighted Class Activation Maps (Grad-CAM) visualisations confirm that the CNN attends to physically meaningful spectral features — sidebands at bearing characteristic frequencies for bearing faults and stator current harmonics for winding faults. The framework is deployed on an NVIDIA Jetson Nano edge device achieving 12 ms inference latency, enabling real-time monitoring on the factory floor without cloud connectivity.

Keywords: induction motor, fault detection, predictive maintenance, convolutional neural network, vibration analysis, STFT spectrogram, bearing fault, stator winding fault, Grad-CAM, edge deployment

Reference:

• [1] Nandi, S., Toliyat, H. A., & Li, X. (2005). Condition monitoring and fault diagnosis of electrical motors — A review. IEEE Transactions on Energy Conversion, 20(4), 719-729.
• [2] Henao, H., et al. (2014). Trends in fault diagnosis for electrical machines. IEEE Industrial Electronics Magazine, 8(2), 31-42.
• [3] LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444.
• [4] Zhang, W., Peng, G., Li, C., Chen, Y., & Zhang, Z. (2017). A new deep learning model for fault diagnosis with good anti-noise and domain adaptation ability on raw vibration signals. Sensors, 17(2), 425.
• [5] Selvaraju, R. R., et al. (2017). Grad-CAM: Visual explanations from deep networks via gradient-based localization. ICCV 2017.
• [6] Sharma, A., & Mathew, L. (2021). Bearing fault classification using CNN on STFT spectrograms. Journal of Vibration Engineering & Technologies, 9(3), 471-484.
• [7] Verma, N. K., Sevakula, R. K., Dixit, S., & Salour, A. (2016). Intelligent condition based monitoring using acoustic signals for air compressors. IEEE Transactions on Reliability, 65(1), 291-309.
• [8] Case Western Reserve University Bearing Data Center. (2013). Bearing data center seeded fault test data. https://engineering.case.edu/bearingdatacenter
• [9] Bhattacharyya, S., & Das, A. (2022). Edge-deployable CNN for real-time motor fault detection. Proceedings of IEEE INDICON 2022, Nagpur.
• [10] Mehta, P., & Jain, R. (2023). Comparative analysis of machine learning approaches for induction motor fault classification. IETE Journal of Research, 69(5), 2841-2853.
• [11] SKF Bearing Specification Catalogue. (2022). Deep groove ball bearings 6208. SKF India.
• [12] National Instruments. (2020). PXIe-4499 Dynamic Signal Acquisition Module User Manual. NI Corporation.