Induction motor fault detection using the fast orthogonal search algorithm

The condition monitoring of induction motors is very important to ensure the continuation of many industrial processes. The detection of faults at their inception is key to ensuring safe operation, maintaining maximum efficiency, and minimizing costs associated with downtimes. There are a number of different faults that can afflict induction motors, however the detection of rotor faults is essential to achieving fault tolerant drive systems. Current signature analysis (CSA) has been established as the most effective means of detecting motor faults. CSA is based on extracting known fault harmonics from the stator current of the motor. Previous systems use the fast Fourier transform (FFT) to determine the spectral content of the motor current.;A fine resolution is particularly important when the motor is operating under light-load conditions. This is because for small slip values, the fault signatures will be close to the fundamental frequency. Since the fundamental is much larger in magnitude, its spectra leakage may cover up the smaller fault harmonics. The length of the sampling time required by the FFT when the motor is under light-load conditions is often not possible to achieve.;This thesis uses the fast orthogonal search (FOS) algorithm for the application of rotor fault detection. The FOS algorithm has previously been shown capable of achieving a finer resolution than the FFT for the same record length. The results of this thesis will demonstrate that FOS is able to accurately detect rotor fault signatures using one-eighth the sampling time required by the FFT. Therefore condition monitoring is now possible for motors where lengthy periods of steady-state are unavailable, but a high degree of resolution is necessary. It will also be demonstrated that by properly selecting candidate terms, the required sampling time can be further reduced.;The FFT is inadequate under many conditions because its resolution is directly related to the length of the sampling time. To achieve a fine resolution, a long sampling time is required. For the FFT to work correctly it is necessary that the motor be in steady-state during the sampling time. Due to the non-stationary nature of motors, long periods of steady-state are often unavailable, and a compromise must be made between resolution and the length of the sampling time.