Research On Fault Diagnosis Technology For Blade Cracks Of Single-stage Single-suction Centrifugal Pump

As the core equipment of drainage system and drive system,the working environment of single-stage single-suction centrifugal pump is very complex,and it is prone to fault,and the probability of blade crack fault is high,which seriously affects the performance of the systems.Therefore,in this paper,according to the mechanism and mode of blade cracks,a fault diagnosis method based on principal component analysis(PCA)and Gaussian mixed model(GMM)which combined signal processing and knowledge was proposed.And it can detect early cracks in advance to prevent their further development.At first,based on the fault mechanism of centrifugal pump blade fault crack,after the theoretical analysis of centrifugal pump vibration and axis trajectory and CFD numerical analysis of centrifugal pump flow,the signals needed for centrifugal pump fault diagnosis were confirmed.This paper explained the mathematical principles of PCA and GMM,and established a fault diagnosis model for centrifugal pump blade cracks by combining the two methods.And Expectation-maximization(EM)algorithm was used to converge the model.In order to verify the diagnostic effect of the model,the training and testing program of fault diagnosis model were confirmed and a centrifugal pump set test bench was designed,and various working conditions of the centrifugal pump were simulated.The working conditions of the centrifugal pump were detected by vibration,flow rate,axis trajectory and current sensors.Subsequently,experiment was conducted to collect the running data of the centrifugal pump under various working conditions.Finally,the training set and test set were constructed by using the collected data,and the feature parameters of PCA and GMM models were obtained by training.The classifier based on these parameters was very effective in diagnosing the original data of centrifugal pump crack fault.