A Convolutional Neural Network for Electrical Fault Recognition in Active Magnetic Bearing Systems

Active magnetic bearings are complex mechatronic systems that consist of mechanical, electrical, and software parts, unlike classical rolling bearings. Given the complexity this type system, fault detection is a critical process. This paper presents new easy way to detect faults based on use dictionary machine learning. The was built starting from signatures consisting images obtained signals available in system. Subsequently, convolutional neural network trained recognize such signature images. objective study develop classifier most frequent soft electrical affect position sensors actuators. proposed method permits, computationally convenient can be implemented real time, determination which component has failed what kind failure occurred. Therefore, identification system allows determining countermeasure adopt order enhance reliability performance assessed by means case concerning turbomachine supported two active for oil gas field. Seventeen classes were considered, reached an accuracy 93% test dataset.