In this work the results of a measuring and simulation campaign performed on an experimental induction motor with artificial broken bars are presented, aiming to extract spectral current components useful for fault detection. With nonsinusoidal feeding a lot of frequencies other than (1-2s)f (lowersideband, LSB) and (1+2s)f (upper-sideband, USB) arise, in proximity of low-order current harmonics, that can be good fault indicators. Current signature analysis-based techniques (CSABT) can take advantage by these harmonic current sidebands (HCSBs) since they are less dependent on load, frequency, drive inertia and motor parameters, so producing a better fault estimation. The measures were matched with simulations carried out by using a topologically complete mathematical complex model.
Machine parameters were obtained by 2D and 3D FEA. A comparison between LSB/USB-based indicators (LUSBBIs) and HCSB-based indicators (HCSBBIs) suggests to use the latter whenever a broken bar diagnosis must be attempted for inverter-fed motors.Index Terms-Broken bars, finite elements, harmonic current sidebands, identification, spectral analyses, multiphase model.