Vibration Characteristics Analysis Of A Double-Suction Centrifugal Pump Identified By Intelligent Algorithms

Due to the merits of high flow rate and high lift, double-suction centrifugal pumps are widely used for agricultural irrigation along the Yellow River in northern China. The operational conditions of the pump are highly related to the silt concentration and flow rates. Therefore, it is crucial to identify the vibration characteristics corresponding to different silt concentrations and flow rates to provide a prior guideline for pump condition monitoring. To this end, this paper explores the identification capabilities of typical intelligent models, including convolutional neural network (CNN), deep neural network (DNN), and support vector machine (SVM), to intelligently establish the connection between vibration characteristics and different flow rates and silt concentrations. Experimental studies on a double-section pump covering a range of flow rates and different silt concentrations are conducted. The results show that CNN features higher identification accuracy than DNN and SVM.