Characterization of pitting corrosion of stainless steel using artificial neural networks

Abstract

In this work, different classification models were proposed to predict the pitting corrosion status of AISI 316L stainless steel according to the environmental conditions and the breakdown potential values. In order to study the pitting corrosion status of this material, polarization tests were undertaken in different environmental conditions: varying chloride ion concentration, pH and temperature. Two different techniques were presented: k nearest neighbor (KNN) and Artificial Neural Networks (ANNs). The parameters for the classifiers were set based on a compromise between recall and precision using bootstrap as validation technique. The ROC space was presented to compare the classification performance of the different models. In this frame, Bayesian regularized neural network model proved to be the most promising technique to determine the pitting corrosion status of 316L stainless steel without resorting to optical metallographic studies.