Special aspects of threshold approach for detecting faults in induction motor bearings

Bearing damage is a significant contributor to the failure statistics of asynchronous electric motors. Motor vibration information is often used to detect this type of damage. However, obtaining such information requires special equipment (external sensors), and difficulties arise in using this approach in mechanisms operating at an increased level of technological vibrations. Fault detection methods based on stator current analysis avoid these drawbacks. Currently, there are many approaches to solving diagnostic problems using fairly complex methods. The use of relatively simple methods suitable for online use is also of interest. The paper analyzes the possibility of using the simplest threshold approach for detecting motor bearing faults, based on the use of information about the stator current with wavelet analysis for primary signal processing. The possibility of motor bearing fault detection based on the analysis of changes in the standard deviation of the stator current signal in the case of a bearing fault is shown. Special aspects of selecting the standard deviation threshold value are considered and it is shown that the fault recognition system threshold level must be set significantly higher than follows from the three-sigma criterion in order to avoid false operation of the algorithm for recognizing this situation.