Experimental Characterization and Computer Vision-Based Detection of Pitting Corrosion on Stainless Steel

Abstract

Pitting corrosion is a form of corrosive damage that leaves cavities in metallic materials, which can significantly compromise structural safety and reliability. For instance, stainless steel, as a versatile structural material, can subject to pitting corrosion, regardless of its high corrosion resistance. The pit growth typically exhibits an initial high growth rate during nucleation, followed by an eventual saturation limit, which will ultimately lead to material failure. Although pitting corrosion is a prevalent damage mode, there is limited data on pit development under different corrosive conditions. Also, it remains challenging to efficiently detect pitting corrosion on large-scale water resources infrastructure. Therefore, this study aims to not only experimentally characterize the evolution of pit morphologies (i.e., depth and surface opening area), but also adopt computer vision techniques to detect pitting corrosion. In particular, an accelerated corrosion experiment was designed and implemented to introduce pitting corrosion to stainless steel in a controlled manner. The three-dimensional pit morphologies were measured using a high-precision laser scanner. Different statistical models were implemented to characterize the evolution of pit morphologies. Furthermore, this study also trained and compared several different computer vision algorithms to identify a promising structural health monitoring approach to detect pitting corrosion in a scalable and non-contact manner.