The Research On Pump State Acoustic Emmission Detection Based On EMD

Pump is the common rotating machinery equipment.Research on pumpstate monitoring,it can save cost,reduce energy consumption and improveequipment operation safety reliability.AE is a potential new state detectionmethod.Using AE technique,research on extracting feature of pump state basedon EMD.It is one of the key technology of pump state classification.It getsome IMF components after using EMD to decompose AE signal of pumpstate.It can easily produce end effect for using cubic spline interpolationfunction to calculate average of upper and lower envelopes.The paperproposes end effect processing new method based on half cycle mirrorextention; because condition of decomposition terminate is inadequate,it willproduce over decomposition,and leads to false model component generation,itproposes new method of false modal components elimination based onKPCA.For shortcoming of EMD,new method about extracting acoustic emissionsignal feature of pump state based on half cycle mirror extention and KPCA, the simulation and experimental results prove the method can restrain endeffect and decrease false mode components.Using Gauss RBF kernel function to construct state classifiers ofSVM.IMF energy entropy is AE signal feature vectors of pump state.Thepump state can be classified,and the classification result is better than everbefore.