Intelligent Diagnosis of Closed-Loop Motor Drives Using Interior Control Signals Under Industrial Low Sampling Rate Conditions

Abstract

Interior control signals derived from motor controllers have gained increasing attention in closed-loop motor drive systems for interturn short-circuit fault diagnosis. Mainstream diagnosis methods generally rely on the extraction of control signals within experimental settings featuring high sampling rates, such as 10 kHz or 40 kHz. However, in practical engineering, the industrial sampling rate of control signals typically reaches only 1 kHz or even lower. This limitation makes it challenging for control signals to intuitively distinguish between healthy and faulty states. To address this practical constraint, an intelligent diagnosis method, termed the prior knowledge integrated contrastive diagnosis model (PK-CDM), is proposed. First, space voltage vectors of interior control signals are extracted as inputs of the PK-CDM to detect the interturn short circuit in a closed-loop motor drive system. Second, the physical variation regularity of space voltage vectors is formulated as the prior diagnostic knowledge to compensate for the lack of information under low sampling rate conditions. Finally, a contrastive pretraining strategy is employed to facilitate the construction of the PK-CDM at an industrially low sampling rate. Experimental results demonstrated that the proposed PK-CDM solves the issue of information loss under industrial low sampling rate conditions by integration of prior diagnostic knowledge with a contrastive learning strategy, thereby yielding superior diagnostic accuracy compared to other state-of-the-art (SOTA) methods.