A zero-sequence current analysis approach for rotating machinery fault diagnosis of induction motor drivetrain based on sparse learning

Fault diagnosis of rotating machinery driven by induction motors has received increasing attention. Current diagnostic methods, which can be performed on existing inverters or current transformers of three-phase induction machines, have become a more economical and reliable alternative to vibration diagnostic methods. Existing references mainly focus on the stator current analysis of single-phase, but most single-phase current analysis methods utilize only a fraction of the total information accessible in the three-phase system. Field experience shows that zero-sequence current has more distinctive fault characteristics compared to single-phase current, and therefore it is better to use in mechanical fault detection and diagnostic tasks. This paper proposes a novel sparse learning-based method for zero-sequence current analysis for induction rotating motor drive fault diagnosis. Firstly, it elaborates and compares the effectiveness of zero-sequence current in revealing fault characteristics compared with single-phase current by analyzing the fault diagnosis mechanism of zero-sequence current. Additionally, the method proposes a correlation entropy-enhanced sparse learning method for the problem of high-frequency noise and interference in the zero-sequence current signals, so as to enhance the learning of the features of the noise-containing signals. The diagnostic efficacy of the proposed method has been verified through experiments on two real datasets.