Induction Motor Fault Diagnosis Using Graph-Based Semi-Supervised Learning

摘要

In this paper, a graph-based semi-supervised learning (GSSL) method is proposed for fault diagnosis of direct online induction motors using stator current and vibration signals. A 0.25 HP induction motor under healthy, single- and multi-fault conditions is tested in the lab. Three-phase stator currents and three-dimensional vibration signals of the motor are recorded simultaneously under steady-state operation in each test. Features for machine learning are extracted from the raw experimental stator current and vibration data using the discrete wavelet transform (DWT). Three GSSL algorithms, local and global consistency (LGC), Gaussian field and harmonic function (GFHF), and greedy-gradient max cut (GGMC), are used in the paper. It is found that both stator current and vibration signals perform well for one individual fault diagnosis using GSSL algorithms, but for classification of a combination of five different faults, the stator current outperforms the vibration signal significantly. Among the three GSSL algorithms, GGMC shows better classification results over LGC and GFHF for both stator current and vibration signals.