Predictive Maintenance of Electric Motor-System Using Vibration Analysis

Abstract

Electric motors are widely used in industrial and mining operations, where their reliability directly affects productivity, safety, and maintenance costs. Unexpected failures can lead to significant downtime, making predictive maintenance an important strategy for improving system performance. This study investigates using vibration analysis to detect common mechanical faults in electric motors, including bearing defects, misalignment, and unbalance. Publicly available datasets, including the Case Western Reserve University (CWRU) bearing dataset and a belt-driven rotating shaft system dataset from the Mendeley Data repository, are used to analyze vibration signals under both normal and faulty conditions. For bearing fault detection, we compare signals from healthy and defective cases using time-domain, frequency-domain, and envelope analysis, allowing us to examine both general vibration behavior and high-frequency impacts. The data are sampled at 12 kHz, with the motor operating at 1797 rpm, enabling identification of localized inner race defects. In addition, vibration data from a belt-driven rotating shaft system are used to study unbalance and misalignment. Normal operation is compared separately with each fault condition using time-domain and frequency-domain methods. The shaft operates at 1238 rpm, and Fast Fourier Transform (FFT) analysis focuses on the shaft rotational frequency (1X) and its harmonics (2X and 3X) for fault identification. The results show that different fault types produce distinct vibration patterns, though the clarity of detection depends on sensor placement and measurement direction, allowing clear fault differentiation. A combination of time-domain features, harmonic analysis, and envelope methods provides a more reliable diagnosis than using a single parameter. Overall, this study demonstrates that vibration analysis can effectively detect both motor-related and externally induced faults, supporting early fault detection and improving predictive maintenance strategies.