Cluster And Discriminate Analysis Of Electrochemical Noise Of Pitting Combined With In Situ Microscope Observation

Electrochemical noise（EN）has been extensively used in the field of corrosion andprotection. It has already been reported that EN had been used to detect corrosion type,nondestructive online monitoring and mechanism research. But there are still manydifficulties for electrochemical noise data processing. The more commonly used methodinclude Time Domain Analysis, Time-domain Statistical Analysis, Frequency DomainAnalysis, Wavelet Analysis, Time-Frequency Joint Analysis and so on, the NeuralNetwork Method which based on Chaos Theory and Cluster and DiscriminantAnalysis(CA and DA) are relatively new ones. Jiayi Huang and Qian Hu et al used CA andDA for the detection of pitting and crevice corrosion, but the method must be amelioratedfurther, especially in the aspect of precision and reliability of the discriminant mode.This paper designed a3-D microscope observation device for the real-timeobservation of pitting of Q235carbon steel, N80pipeline steel and Normal13Cr stainlesssteel in0.5mol/LNaHCO3+（0.05—0.25）mol/L Cl-,0.5mol/LNaHCO3+（0.08—0.30）mol/L Cl-and3.5%—4.0mol/L Cl-solution respectively. Then take the pitting data as thesamples for the discriminant model and made a little corresponding improvement of themethod. After that, this paper established three pitting discriminant modes in different Cl-concentration for each type of materials and a unified mode for materials in the samecorrosive medium. Finally, chose3groups of observation data from every type ofmaterials as verification for the models.The models take standardized potential and current（shown as EStanand|I|Stan）as theCategorical Variables. The three pitting discriminant modes in different Cl-concentrationare as follows:yQ235=-0.7679EStan+0.2631|I|Stan yQ235=-1.9163EStan+0.544|I|Stan-0.216yN80=-0.722EStan+0.3054|I|Stan yN80=-1.7728EStan+1.1246|I|Stan+1.6164y13Cr=-0.7333EStan+0.3325|I|Stan y13Cr=-2.1237EStan+0.9313|I|Stan+1.5963Their prediction time for pitting corrosion are16.98h,0.26h,1.06h respectively, wellmatching with the actual observation result. In addition, the established unified discriminant model is:y=-0.7411EStan+0.3003|I|Stan y=-1.9376EStan+0.8666|I|Stan+0.9989Its prediction time for pitting are16.98h,0.26h,1.09h respectively, well matching with thethree separate models as well, so we could come to a conclusion that the unifieddiscriminant model can used to discriminate the occurrence and development of pitting indifferent concentration of chlorine ions medium as well.