The Research Of Acoustic Emission Characterization Of 2.25Cr-1Mo Based On Clustering And Bp Neural Network

The hydrogenation reactor,operated at high temperature high pressure and hydrogen exposure,is the critical equipment in the oil refining industry.2.25Cr-1Mo steel,used as the main material of a hydrogenation reactor,has very strict requirement on forming,welding and heat treatment temperature controlling,to be worse,confronted with H2 and H2S which are corrosion medium.Once break,it will cause great destruction.Acoustic emission(AE)technology is a new non-destructive testing method and is used in pressure vessel detection and structure integrity assessment.That applying AE technology to test the hydrostatic test of the hydrogenation reactor can detect dangerous source probably caused by fracture or leakage.The research on AE characteristics of the material 2.25Cr-1Mo's plastic deformation and crack propagation fracture will provide a basis for AE online detection of the hydrostatic test of the hydrogenation reactor.Noise can not be avoided during the process of AE test,it will thus influence the structure integrity assessment.Therefore,filtering out the noise during the detection process and studying the AE characteristics of the material 2.25Cr-1Mo,s different deformation stage are significant in theory and application.The main works are listed as fellow:(1)Research on the method of de-noising in acoustic emission signal.AE Source Signal is always disturbed with noises.To solve this problem,a new method of de-noising was studied,which is based on wavelet analysis to find the main frequency band of AE Source Signal and reconstructing wavelet coefficients with the main frequency bands.(2)Research on sound source identification based on clustering and neural network.Clustering of AE signals based on regular characteristic parameters could not have met the requirement of source identification.Thus,an idea that extracts new parameters from three aspects that can reflect the characteristics of waveform frequency distribution,shape and intensity was raised,and two methods were studied:waveform spectrum characteristic was extracted based on wavelet analysis and waveform shape characteristics were determined by using of a A%floating threshold.The method of source indentify based on clustering and neural network was applied to analysis the data collacted from the designed AE experiments to determine weather the method is feasible and reliable.(3)Research on acoustic emission characteristics of material 2.25Cr-1MoWe worked on two kinds of specimens of hydrogenation reactor material 2.25Cr-1Mo-one with crack 2mm in depth and another without crack,subjected to three-point bending,awaiting damage modes in the material.Then wavelet and pattern recognition methods were applied to analysis the AE signals collected during the total procedure,and Parameter distribution characteristics and spectral characteristics of AE signal from plastic deformation at yield stage and crack signals at destructed stage were finally acquired.