Speed Sensor Fault Detection Algorithm Based on Autoencoder Applied for Two-mass System

The paper describes an application of an autoencoder for detecting speed sensor faults. The analyzed drive is equipped with a BLDC motor, which is connected to the load machine through an elastic shaft. The drive is controlled with the Active Disturbance Rejection Control strategy. In this study, experimental verification is conducted. The results show that speed sensor faults are recognized swiftly and correctly. Dynamic states, such as load increase, do not trigger the detection algorithm. Moreover, the diagnosis remains correct even when the drive parameters are changed.