부산대학교 도서관

If a centrifugal impeller is in deep stall, it requires a large amount of additional energy to break free. As such, it is crucial to study the prediction method of impeller stall and establish its relationship with force and torque. This study employs a computational fluid dynamics numerical simulation method to analyze the flow field within the impeller of a centrifugal pump, while monitoring the changes in hydraulic torque of the blades. By examining the impeller’s internal flow field characteristics and external (energy performance) characteristics curves under different working conditions, the stall process was classified into three stages: germination, transition and stabilization. The evolution of stall cells under different stall stages was also analyzed. We utilized time-frequency conversion and analysis to extract characteristic parameters from instantaneous hydraulic torque results generated under different working conditions. A more effective model for predicting impeller stall stages was built by virtue of support vector machine algorithm. The proposed model exhibits improved feasibility and effectiveness in comparison to traditional prediction methods. It is expected to have significant implications for stable operation monitoring and safety protection of centrifugal pumps.