Study On Corrosion Signal Processing For Electrochemical Noise

Electrochemical noise (EN) is an in-situ, non-destructive, real-time,high-precision method to study the corrosion mechanism of materials and corrosiondetection. The difficulty of EN is the corrosion mechanism and data analysis ofnuclear materials under high temperature and pressure environment. So it ismeaningful to study the EN experiment and date analysis on nuclear material, whichcan help our country develop nuclear energy and prevent nuclear material failure.A multi-channel EN instrument was designed according to the characteristics ofthe EN signal, and used for a lot of EN experiment. There were three kinds of mainexperiment. Firstly, the normal EN was designed to collect typical signal of Pitting,Uniform corrosion and Passivation under room temperature on304ss. Secondly, theexperiment take place in the mixed solution of H3BO3and LiOH in high temperatureand pressure environment. Thirdly, a simulated nuclear test is designed to study therelationship between the corrosion of pipeline and temperature by detecting the ENthrough all process.Three new methods of data progress were proposed to analysis the EN signal:back-propagation neural network (BP), support vector machine (SVM) and entropyrate. These three new methods can be used to determine the type of corrosion, andused to progress the electrochemical noise from room temperature experiment. Thealgorithm design process and optimization methods were also given in this paper. Theresult shows that back-propagation neural networks and support vector machine had ahigh accuracy close to100%, the entropy rate had a good determination of corrosiontypes.A set of fully functional EN data processing software was carried out based onthe Matlab-GUI. The software, which had complete functions and simple operation,covered the traditional electrochemical data processing methods, data batch function,neural network module, support vector machine module, EMD module and chaostheory module.