Bearing Fault Detection in Induction Motors an‎d Shaft Misalignment Detection in Electro-Pumps Using Electric Signals

1/1/1900 12:00:00 AM

Vibration analysis has traditionally been used for fault detection in induction-motor-driven electro-pumps; however, in industrial environments it suffers from practical limitations such as high cost, restricted sensor placement, an‎d susceptibility to sensor failure. Conse-quently, condition monitoring based on electrical signals has emerged as a cost-effective an‎d non-intrusive alternative. Despite this, conventional spectral analysis of stator current signatures often exhibits low signal-to-noise ratio, strong load dependency, an‎d statistical overlap between healthy an‎d faulty conditions, which can lead to unreliable diagnosis an‎d false alarms. This dissertation presents an integrated electrical-signal-based framework for the detection of localized bearing faults an‎d shaft misalignment in induction motors. The proposed ap-proach exploits the physical electromechanical coupling by which mechanical impacts an‎d torque fluctuations generated by bearing defects an‎d misalignment induce short-time dis-turbances in rotor speed an‎d air-gap flux rotation, which in turn modulate the phase an‎d instantaneous frequency of stator voltages an‎d currents. Instantaneous amplitude an‎d in-stantaneous frequency features are extracted using the analytic signal representation, an‎d their fault-sensitive impulsive behavior is quantified through higher-order statistical indices in the time domain. To enhance fault visibility under noisy industrial conditions, linear-prediction-based an‎d time-synchronization-based noise-cancellation structures are developed for stator current an‎d voltage signals. The resulting residual an‎d difference signals, together with instanta-neous features, are shown to significantly improve the signal-to-noise ratio an‎d to suppress load-dependent deterministic components. Among the investigated indicators, the kurtosis of instantaneous frequency of motor voltage an‎d current, as well as their prediction-error counterparts, provide highly sensitive an‎d statistically separable signatures for both single-point bearing defects an‎d shaft misalignment. The proposed methods are experimentally validated on an 7.5-kW induction motor driving a centrifugal pump under four load levels an‎d multiple fault conditions, including outer-race, inner-race, rolling-element, an‎d cage defects, as well as parallel, angular, an‎d com-bined misalignment. The results demonstrate stable, load-robust, an‎d non-overlapping dis-crimination between healthy an‎d faulty states, confirming that the combination of instan-taneous feature analysis an‎d noise-cancellation-based residual processing constitutes a reli-able an‎d practically applicable framework for condition monitoring of industrial induction-motor-driven systems.

پايش وضعيت، , موتور القايي , خرابي ياتاقان , ناهم‌راستايي محور , دامنه لحظه‌اي , فركانس لحظه‌اي , پيش‌بيني خطي , تحليل پسماند , حذف نويز سيگنال. , Instantaneous Domain

Condition Monitoring , Induction Motor , Bearing Fault , Misalignment , Stator Current , Line Voltage , Signal Processing , Instantaneous Frequency , Linear Prediction , Residue

Asadollah Kalantar

Mir Said Safizadeh