Prediction of pitting corrosion status of EN 1.4404 stainless steel by using a 2‐stage procedure based on support vector machines

Abstract

The excellent properties of EN 1.4404 have made this material one of the most popular types of austenitic stainless steel used for many applications. However, in aggressive environments, this alloy may suffer corrosion. Electrochemical analyses have been extensively used in order to evaluate pitting corrosion behaviour of stainless steel. These techniques may be followed by microscopic analysis in order to determine the resistance of the passive layer. This step requires the human interpretation, and therefore, subjectivity may be included in the results. This work aims to solve this drawback by the development of an automatic model with the capability to predict pitting corrosion status of this material. A combined model based on support vector machines (SVMs) is presented in this work. With the aim to improve the prediction performance, the model considers the breakdown potential values estimated by itself at a first stage. The performance is evaluated based on receiver operating characteristic (ROC) curves. The area under the curve (AUC) and accuracy results (0.998 and 0.952, respectively) demonstrate the utility of the proposed model as an efficient and accurate tool to predict pitting behaviour of EN 1.4404 automatically.

A support vector machine‐based ensemble algorithm for pitting corrosion modeling of EN 1.4404 stainless steel in sodium chloride solutions

“…The results obtained when no pre‐processing technique was considered in the model were 95.50% and 94.75% in terms of accuracy, whereas in term of sensitivity the highest values were 96.74% and 97.38% for polynomial and radial basis function, respectively . According to these results and comparing with the results from Tables and , it can be found that for the SVM ensemble model proposed to predict pitting corrosion behavior of stainless steel in this study, the use of LDA did not improve the quality of the prediction model.…”

“…In this work, according to the studies presented previously, a novel ensemble model based on SVM is presented. The model consists of three phases: in the first one, the breakdown potential value was estimated according to the environmental conditions.…”