Asynchronous motor fault detection using machine learning algorithms

BACKGROUND. Machine learning methods are effective advanced means ensuring the operability of various engineering systems, including test systems. As statistics on faults accumulate, test systems based on machine learning algorithms provide higher prediction accuracy and do not require expensive test equipment and skilled personnel. AIM. To develop a test system capable of both determining the fault and assessing its extent with high accuracy. MATERIALS AND METHODS. The subject of the study is a three-phase asynchronous motor with a squirrel cage rotor; machine learning methods are used to achieve the goal. RESULTS. Using the example of interturn faults in the stator winding, the authors demonstrate that it is possible to detect the fault and its extent even at the initial stage (with a few short-circuited turns) with an accuracy of at least 95%. CONCLUSION. Machine learning methods allow to develop effective and affordable test systems that are versatile, highly accurate, and do not require skilled personnel.