Multicavitation States Diagnosis of the Vortex Pump Using a Combined DT-CWT-VMD and BO-LW-KNN Based on Motor Current Signals

Abstract

Vortex pumps play a crucial role in industrial and municipal settings by transferring high-viscosity and particle-laden fluids. However, their performance and reliability are significantly compromised by cavitation. Identifying and diagnosing cavitation promptly is essential for maintaining the proper operation of vortex pumps. The use of current signals as a noninvasive monitoring method has shown great promise in detecting multicavitation states. The proposed method integrates dual-tree complex wavelet transform (DT-CWT) with variational mode decomposition (VMD) to decompose current signals into multiple modes. Subsequently, the Bayesian optimized locally weighted k-nearest neighbor (LW-KNN) algorithm is employed to accurately identify multicavitation states. High-speed photography is also utilized to observe the incipient, developing, and collapsing phases of cavitation. The results indicate that the proposed method achieves a detection accuracy of 96.67% at a flow rate of 40 m3/h, outperforming other flow conditions. The recognition accuracy reaches 98.33% under stable flow conditions, while accuracies of 92.33% and 93.67% are observed for flow rates of 35 and 45 m3/h, respectively. The overall average recognition rate across all tested flow conditions is 94.22%. This methodology not only demonstrates high effectiveness in identifying cavitation states but also offers a reliable and practical solution for fault diagnosis in fluid mechanical systems. It significantly contributes to the improvement of operational efficiency, reliability, and maintenance strategies in industrial and municipal pumping systems.