Determination of Corrosion Types from Electrochemical Noise by Artificial Neural Networks

In this paper, a novel approach for distinguishing the type of corrosion from Electrochemical Noise (EN) signals is presented. A database containing numerous sets of original EN data is established，then the database is divided into training sets and test sets randomly. The EN data sets are used as the inputs of the Artificial Neural Networks (ANN) and the types of corrosion as the outputs of the ANN. A feature vector is extracted from each EN data set. Subsequently, two kinds of artificial neural networks with different neural structures, the Back Propagation (BP) and the Support Vector Machine (SVM), are constructed by training feature vector. The test sets are used to test the accuracy of the two neural networks. The result shows the ANN is a very accurate and effective way to distinguish the type of corrosion and the SVM is more accurate than the BP.