| Cavitation will reduce the efficiency of centrifugal pump,induce vibration,pressure pulse and noise,and seriously affect its safe operation and service life.Therefore,it is of great significance to monitoring and study the cavitation state of centrifugal pump.As an advanced detection method,acoustic emission(AE)technology has no interference to the operation of the detection object and is sensitive to the state change.It can effectively avoid the interference of low-frequency signals and can be used for the detection of the cavitation state of centrifugal pump.In this paper,based on the acoustic emission detection technology,the cavitation state of centrifugal pump is monitored.The Multi-resolution singular value decomposition(MRSVD)method and the improved variational mode decomposition(IVMD)method are used to the denoising and the features extraction of AE signals of centrifugal pump cavitation.Combined with the extreme learning machine(ELM)method,the extracted characteristic values are trained and learned to realize the recognition of different states of centrifugal pump cavitation.This paper mainly carried out the following work:(1)Using the acoustic emission acquisition system independently developed by our research group,the AE signals of centrifugal pump under different cavitation conditions are obtained on the pump test bench,and the parameters such as inlet and outlet pressure,flow rate and flow of centrifugal pump are recorded in the test process,and the characteristic curve of centrifugal pump head changing with NPSH is established.(2)Aiming at the noise of the collected AE signals of centrifugal pump cavitation,the MRSVD method is used to the denoising of AE signals of centrifugal pump cavitation.This method can realize the singular value decomposition of signals in multiple levels and multiple spaces based on the principle of matrix two recursion construction,decompose the signal into approximate signals and multiple detail signals,and finally retain the approximate signals as the denoising signals.Simulation and experimental results show that the MRSVD method can effectively remove the noise components in the signal and retain the real and effective useful signal.(3)For the feature extraction of AE signals of centrifugal pump cavitation,the IVMD method is used to extract the feature of AE signals of centrifugal pump cavitation.In VMD,There are problems of artificially experience setting of the decomposition layers and the penalty factor.The envelope entropy difference coefficient and the artificial bee colony algorithm are used to optimize the two parameters of VMD,and the optimal parameters are selected.Then it is applied to the feature extraction of AE signals of centrifugal pump cavitation,and the component with larger correlation coefficient is selected as the feature component of cavitation.The results show that the absolute energy value,relative energy value and envelope entropy value of the characteristic components of each operating point show certain rules in the process of cavitation state of centrifugal pump from scratch,from weak to strong,and the cavitation state of centrifugal pump corresponds.(4)For the identification of different states of centrifugal pump cavitation,the ELM method is used to identify different states of centrifugal pump cavitation.The eigenvectors representing cavitation different states are used as the input of the ELM to identify different states of centrifugal pump cavitation.The results show that the recognition rate of this method is as high as 85%,and it can recognize different states of centrifugal pump cavitation. |
