ABSTRACT Ball bearing faults are a major contributor to failures in squirrel‐cage induction motors (SCIMs), while their early detection remains challenging under noisy and variable load industrial conditions. This paper presents a hybrid signal‐processing framework for incipient bearing fault diagnosis based on a deliberately designed and problem‐driven sequencing of established techniques. Discrete wavelet transform (DWT) was first used for targeted noise suppression, followed by velocity‐based envelope analysis to enhance fault‐related modulations, and wavelet packet transform (WPT) was used for precise isolation of fault‐sensitive frequency bands. Fault presence and severity are evaluated using standard deviation and energy indicators selected for their robustness and consistency with ISO 10816 guidelines. The proposed framework is validated using the Case Western Reserve University (CWRU) benchmark dataset, laboratory measurements and vibration data from real industrial motors. The results demonstrate reliable sensitivity to early stage bearing defects across varying operating conditions, without reliance on data‐driven training or complex classifiers. Owing to its interpretability, moderate computational complexity and compatibility with both direct and inverter‐fed SCIMs, the proposed approach provides a practical and industry‐oriented solution for condition monitoring and predictive maintenance.
Agah et al. (2026) studied this question.