Experimental Characterization and Computer Vision-Assisted Detection of Pitting Corrosion on Stainless Steel Structural Members

Pitting corrosion is a prevalent form of corrosive damage that can weaken, damage, and initiate failure in corrosion-resistant metallic materials. For instance, 304 stainless steel is commonly utilized in various structures (e.g., miter gates, heat exchangers, and storage tanks), but is prone to failure through pitting corrosion and stress corrosion cracking under mechanical loading, regardless of its high corrosion resistance. In this study, to better understand the pitting corrosion damage development, controlled corrosion experiments were conducted to generate pits on 304 stainless steel specimens with and without mechanical loading. The pit development over time was characterized using a high-resolution laser scanner. In addition, to achieve scalable and automatic assessment of pitting corrosion conditions, two convolutional neural network-based computer vision algorithms were adopted and implemented to evaluate the efficacy of networks to identify existence of pitting damage. One was a newly trained convolutional neural network (CNN) using MATLAB software, while the other one was a retrained version of GoogLeNet. Overall, the experimental results showed that time is the dependent variable in predicting pit depth. Meanwhile, loading conditions significantly influence pit morphology. Under compression loading, pits form with larger surface opening areas, while under tension loading, pits have smaller surface opening areas. Deep pits of smaller areas are dangerous for structural members, as they can lead to high stress concentrations and early stress corrosion cracking (SCC). Furthermore, while the training library was limited and consisted of low-resolution images, the retrained GoogLeNet CNN showed promising potential for identifying pitting corrosion based on the evaluation of its performance parameters, including the accuracy, loss, recall, precision, and F1-measure