Bearing faults account for over 40% of induction motor faults, and for this reason, for several decades, much attention has been paid to their condition monitoring, through vibration measurements and, more recently, through electromagnetic signal analysis. Furthermore, in the last few years, research has been focused on evaluating deep learning algorithms for the automatic diagnosis of these faults. Therefore, the purpose of this study is to propose a novel procedure to automatically diagnose different types of bearing faults and load anomalies by means of the stator current and the external stray flux measured on the induction motor in which the bearings are installed. All the data were collected by performing experimental tests in the laboratory. Then, these data were processed to obtain images (scalograms and spectrograms), which were elaborated by a pre-trained Deep Convolutional Neural Network, modified through the transfer learning technique. The results demonstrated the ability of the electromagnetic signals, and in particular of the stray flux, to detect bearing faults and mechanical anomalies, in agreement with the recent literature. Moreover, the Convolutional Neural Network has been proven to be able to automatically discriminate bearing defects and with respect to the healthy condition.

Electric induction motors are fundamental to industry, where reliability and continuous operation are critical. Though robust, they are prone to faults, particularly in demanding environments such as highway tunnels. Non-invasive diagnostic techniques are widely used for condition monitoring, yet most studies occur under controlled laboratory conditions, limiting their applicability in real-world scenarios. This research investigates the feasibility of applying Motor Current Signature Analysis (MCSA) for monitoring highway tunnel axial fan motors, aiming to determine its effectiveness for real-time diagnostics in industrial environments. Measurements were performed under actual operating conditions, highlighting practical challenges. Data acquisition was implemented remotely from electrical cabins feeding tunnel services, reducing installation complexity and costs compared to in-tunnel measurements. This approach enabled monitoring of all motors in a tunnel using minimal hardware (a single acquisition system equipped with Rogowski sensors) making the solution cost-effective and suitable for periodic measurements. Frequency domain analysis focused on harmonics associated with rotor bar defects and eccentricity, selected for their slow degradation and diagnostic relevance. The magnitude of these harmonics was tracked over time and compared across motors of the same model. Since most of the time the ventilators are de-energized, the periodic measurements can be seen almost as a real-time monitoring, at least for the faults considered, with much lower costs. Results were validated against maintenance reports, confirming bearing faults with eccentricity in two motors, while suspected rotor porosity remained unverified, as expected at low severity. Findings demonstrate that MCSA can provide operational insights for fault detection in tunnel environments, supporting predictive maintenance strategies. A key outcome of this study was selecting and implementing an effective measurement setup for industrial applications, while preparing the base for future machine learning integration to estimate Remaining Useful Life.

The aim of this paper is twofold. On the one hand, the applicability of the well-known diagnostic techniques, based on the motor current and stray flux signature analysis, has been experimentally tested both on small single-phase induction motors and on large on-field three-phase induction motors. On the other hand, the single phasing fault in three-phase induction motors, which has rarely been treated in the literature, has been experimentally analyzed and compared to the behavior of the single-phase motor, in order to find characteristic signatures of this fault. Time and frequency domain analyses were performed on current, axial and radial stray flux, and torque signals. The experimental study highlighted the following insights: i) classical motor current signature analysis cannot be easily applied to permanent-capacitor single-phase machines; ii) stray flux analysis reveals the presence of additional harmonic sidebands, even in healthy three-phase induction motors; iii) a rough analysis of the third harmonic in the current spectrum may be sufficient to diagnose the single phasing fault.

In this work, an investigation on permanent-capacitor single-phase induction machine, healthy three-phase and three-phase induction machine in single phasing condition is presented. The study focuses on the analysis of one-phase current signal, radial and axial stray fluxes, and also torque signal for the single phasing condition. The steady state signals are initially analyzed in the time-domain and then sidebands around the supply harmonic are studied in the frequency-domain. Firstly, new insights are presented for the permanent-capacitor motor, very rarely treated in literature. Then, the presence of a significant harmonic in stray flux signals is highlighted or all machines and conditions. An investigation over a wide frequency range is also carried out proposing a new rule of thumb to locate a supply unbalanced condition looking at third and fifth harmonics. Two Low Frequency Torque Oscillation (LFTO) conditions are considered and evaluated. Finally, the derivative of the phase current is examined, and its interaction with the stray flux signals’ variations in shape and amplitude is evaluated for several conditions. Some correlations between these signals and the current derivative are found, providing interesting diagnostic insights.

In this paper considerations about the time-domain amplitudes and phase shift between diagnostic signals measured in induction machines will be presented. In particular, the line to line voltage, stator current and axial and radial stray fluxes have been considered in various conditions. Since the phase shift between signals is rarely considered in the diagnostics of electrical machines, even in multi-sensor system, this represents a quite strong novelty of the presented method. The machine considered, a small-size induction motor, has been fed by the mains and with an open-loop PWM inverter. Tests with different load conditions and with three severity levels of stator inter-turn short circuits are presented. The phase-shifts between the fundamentals and third harmonics of the signals are considered, giving good diagnostic results.