LightGBM Based Fault Diagnosis of the Canned Motor Pump Based with Stray Flux

The static closed structure and complex working conditions of the canned motor pump severely limit the traditional motor fault detection method. In order to detect the status of the canned motor pump, and the stray flux condition monitoring as a non-invasive detection method reflects the operating status of the motor, it is suitable for canned motor pumps. This paper proposes a fault diagnosis method based on stray flux with LightGBM algorithm. Firstly, the topology and working principle of the motor are introduced, and the corresponding mapping relationship between stray flux and air gap flux is verified. Then, the distribution law of stray flux under different faults is obtained based on the finite element model. Then, it is proposed to rely on the frequency energy distribution decomposed by wavelet packet energy as the eigenvalues and use the LightGBM algorithm to train the data for fault diagnosis. Finally, an experimental testbench was built for different fault experiments and feature extraction and training were conducted, which proved the correctness of the diagnostic method.