Research On Fault Diagnosis Method Of Centrifugal Pump

Water pump is the machine that transports liquid or pressurizes,and plays an extremely important role in many fields,such as defense military industry,agricultural irrigation,industry,municipal administration,and household water supply.The centrifugal pump has the characteristics of simple structure,convenient operation,wide application range,reliable operation and so on,and has become the most extensive one in the manufacture of water pumps.If you ignore the possible failures of the centrifugal pump during actual operation,it will not only cause damage to the centrifugal pump,but also cause paralysis of the production chain and cause serious economic losses.Therefore,it is particularly important to adopt proper diagnostic methods for fault diagnosis of centrifugal pumps.Aiming at the fault diagnosis of centrifugal pumps,this paper has conducted the following researches:(1)Summarize the common failure types of centrifugal pumps,analyze the failure forms of typical components of centrifugal pumps,study the causes and mechanisms of failures,and then adopt different diagnostic methods according to different typical components.(2)The fault of the centrifugal pump impeller is diagnosed by simulation analysis method.By three-dimensional modeling of the centrifugal pump flow path,the flow situation is simulated and analyzed by Ansys Fluent software,the blade failure type is set,and the simulation results under different failure types are compared.The values such as outlet pressure and head obtained after the post-processing are related to the measured value of the instrument in the actual centrifugal pump system,and then determine the type of centrifugal pump impeller failure.(3)The centrifugal pump rotor bearing faults are diagnosed by the method of vibration signal analysis.The centrifugal pump and rotor test bench are built to simulate the type of fault,the vibration signal is extracted by the vibration collector,and the IMF component obtained by VMD decomposition is divided into different components.Obtain the energy value,waveform,pulse,margin,peak value,and kurtosis index to form a feature vector.After LDA dimensionality reduction processing,a feature matrix is formed to prepare for later data components.(4)According to the characteristics of the intelligent recognition method of nuclear extreme learning machine,the genetic extreme algorithm is optimized by genetic algorithm,and the model is trained and tested through samples to verify that the built nuclear extreme learning machine model has good The classification ability can not only identify the faults of the measured centrifugal pump data,but also accurately classify more types of centrifugal pump faults simulated by the rotor test bench.The model has high accuracy and generalization.