Concurrent Incipient Fault Diagnosis in Three-Phase Induction Motors Using Discriminative Band Energy Analysis of AM-Demodulated Vibration Envelopes.

Three-phase induction motors (TIMs) are widely used in industrial applications, with bearings and rotors representing the most failure-prone components. Detecting incipient damage in these elements is particularly challenging. The associated signatures are weak and highly sensitive to variations, and their identification typically demands sophisticated filters, deep learning models, or high-cost sensors. In this context, the main goal of this work is to propose a new algorithm that reduces the dependence on such complex techniques while still enabling reliable detection of realistic faults using low-cost sensors. Therefore, the proposed Discriminative Band Energy Analysis (DBEA) algorithm operates on vibration signals acquired by low-cost accelerometers. The DBEA operates as a low-complexity filtering stage that is inherently robust to noise and variations in operating conditions, thereby enhancing discrimination among fault classes, without requiring neural networks or deep learning techniques. Moreover, the interaction of concurrent faults generates distinctive amplitude-modulated patterns in the vibration signal, making the AM demodulation-based algorithm particularly effective at separating overlapping fault signatures. The method was evaluated under a wide range of load and voltage conditions, demonstrating robustness to speed variations and measurement noise. The results show that the proposed DBEA framework enables non-invasive classification, making it suitable for implementation in compact and portable diagnostic systems.