Stainless steel has proved to be an important material to be used in a wide range of applications. For this reason, ensuring the durability of this alloy is essential. In this work, pitting corrosion behaviour of EN 1.4404 stainless steel is evaluated in marine environment in order to develop a model capable of predicting its pitting corrosion status by an automatic way. Although electrochemical techniques and microscopic analysis have been shown to be very useful tools for corrosion studies, these techniques may present some limitationus. With the aim to solve these drawbacks, a three-step model based on Artificial Neural Networks (ANNs) is proposed. The results reveal that the model can be used to predict pitting corrosion status of this alloy with satisfactory sensitivity and specificity with no need to resort to electrochemical tests or microscopic analysis. Therefore, the proposed model becomes a useful tool to predict the behaviour of the material against pitting corrosion in saline environment automatically.

The resistance to localized corrosion of stainless steel in sodium chloride solutions is evaluated at different conditions. The effects of chloride ion concentration, acidity, and temperature on pitting corrosion resistance are analyzed. In order to develop a model capable of predicting pitting corrosion behavior of EN 1.4404 by an automatic way, a support vector machine‐based ensemble algorithm is proposed. An additional step related to feature selection is included in the model in order to improve the prediction capability. According to the excellent prediction results, the support vector machines (SVM) model is used to perform a sensitivity analysis for the purpose of analyzing the influence of the environmental variables on the breakdown potential and the pitting corrosion status modeling of this grade of stainless steel. Based on this analysis, it can be concluded that the breakdown potential is the main variable to be considered in pitting corrosion modeling whereas temperature is the most important one for breakdown potential modeling.