Induction motors are the most commonly used, versatile and popular
electromechanical devices used in industrial applications as they show high reliability, low maintenance
and relatively high efficiency. Although they are highly reliable, they are susceptible to
many faults. Such faults may cause unexpected and untimely production shutdowns. Most
of the faults in the induction motors may be detected in the nascent stages so as to
prevent untimely failures. Motor faults are caused due to mechanical and electrical
stresses. Mechanical faults include bearing faults and rotor damage which occur due to
overloads and abrupt load changes. Electrical faults include the single-turn or multi-turn short
circuits in the same phase or in the different phases or short circuit between any phase and
the stator body, end ring faults and the rotor bar breakages (Thorsen and Dalva,
1995; Benbouzid, 2000; and Siddique et al., 2005). This paper addresses the unbalanced
supply fault, damaged bearing fault and broken rotor bar(s) faults which introduce some form
of unbalance in the stator current.
Several techniques such as vibration analysis, acoustic noise measurement,
torque profile analysis, temperature analysis and magnetic field analysis have been
conventionally used for detecting induction motor faults (Benbouzid, 1999; and
Nandi et al., 2005). These techniques require sophisticated and expensive sensors, additional electrical
and mechanical installations while also accounting for the lower system reliability and
frequent maintenance. Several fault identification and detection techniques analyze the
power spectrum of the stator current to detect motor electrical and mechanical faults.
These techniques look for the magnitude of certain frequency components in the stator
current (called Motor Current Signature Analysis, MCSA) to detect motor faults
(Benbouzid, 2000). These techniques may fail to establish error free fault identification due
to windowing process employed to compute the power spectrum of the current signals.
The analysis of negative sequence components of stator current is based on the detection
of asymmetries produced by the fault in the stator winding. This method fails to detect
faults for induction motors with unbalanced winding, unbalanced power supply voltage,
etc. Recent techniques for the detection of motor faults are based on Artificial
Intelligence (AI) using concepts such as fuzzy logic, genetic algorithm, neural networks and
Bayesian classifiers (Nandi et al., 2005). Most of the AI-based techniques require large datasets
to train them. Wavelet transform analysis is another recent technique to analyze and
detect the non-stationary, non-localized fault signals arising out of motor faults (Eren and
Devaney, 2004). A method using motor internal physical condition based on periodic oscillation
of the rotor magnetic field space vector orientation is another technique used for motor
fault detection and identification. |
