Online Data-Driven Fault Diagnosis of Dual Three-Phase PMSM Drives Considering Limited Labeled Samples
Luhan Jin, Yao Mao, Xueqing Wang, Linlin Lu, Zheng Wang
Abstract
In practical applications, it is time-consuming and laborious to collect sufficient labeled samples for the data-driven fault diagnosis of motor drives. To solve this issue, this article proposes an online data-driven semisupervised diagnosis method for electrical faults of dual three-phase permanent-magnet synchronous motor drive on the condition of limited labeled samples. In the proposed method, the combination of currents on harmonic space and the rotor position are extracted as the input features which have lower dimensions and computational complexity. Then, the optimal key feature points and safety margin of each fault mode are calculated offline using a large number of unlabeled samples with the aid of limited labeled samples. Finally, online fault diagnosis in the proposed method is fulfilled using feature matching, which can accurately identify each fault mode and indicate the unknown faults timely with lower computational complexity. The proposed method has been online verified with the experiments.