Research On Vibration Signal Analysis Of Centrifugal Pump Typical Fault Based On Wavelet Transform

Centrifugal pump is widely used in industry and agriculture because of its advantages such as wide range of head application,simple structure,reliable operation and convenient maintenance.In practical application,when the working point deviates from the design working condition or the unit itself has faults,the vibration will increase and the efficiency will decrease.It is very important to analyze the vibration characteristics of centrifugal pump for improving the safe and economic operation of pump unit.The vibration of water pump is induced by many factors,and its external characteristics and internal mechanism are complicated.Therefore,it is of great academic significance and practical value to study the method of vibration feature extraction and fault identification,and to explore a new method for pump vibration mechanism analysis.Firstly,according to the classical characteristics of cavitation-induced vibration of centrifugal pumps and loosening of anchor bolts,based on the wavelet Transform,a single-target Particle Swarm wavelet de-noising Algorithm with root-mean-square error(RMS)as evaluation function and a multi-target particle swarm wavelet de-noising Algorithm with signal-to-noise ratio(snr)and root-mean-square error(RMS)as evaluation index are established respectively,the optimal wavelet base is DB8,the number of layers is 6,and the corresponding threshold of each layer is determined.Secondly,the cavitation test and the vibration test caused by the loosening of Anchor Bolt are designed,and the vibration signals such as the pump shell and motor shell,which are caused by the loosening of Anchor Bolt,are filtered by the wavelet de-noising method,the time domain and frequency spectrum analysis of the processed signal are carried out to obtain the key parameters characterizing the operating characteristics of the pump.Finally,the BP neural network is used to establish the operation identification method of Centrifugal pump.The time-domain and frequency-domain eigenvalues of pump vibration signal are input into the BP neural network classification model of Centrifugal pump to classify the fault state,it provides a strong theoretical support for the practical operation of pumping stations.