On the detection of corrosion pit interactions using two-dimensional spectral analysis

Abstract

A statistical methodology for detecting pits interactions based on a two-dimensional spectral analysis is presented. This method can be used as a tool for the exploratory analysis of spatial point patterns and can be advanced as an alternative of classical methods based on distance. One of the major advantages of the spectral analysis approach over the use of classical methods is its ability to reveal more details about the spatial structure like the scale for which pits corrosion can be considered as independent. Furthermore, directional components of pattern can be investigated. The method is validated in a first time using numerical simulations on random, regular and aggregated structures. The density of pits, used in the numerical simulations, corresponds to that assessed from a corroded aluminium sheet. In a second time, this method is applied to verify the independence of the corrosion pits observed on the aforementioned aluminium sheet before applying the Gumbel theory to determine the maximum pit depth. Indeed, the property of independence is a prerequisite of the Gumbel theory which is one of the most frequently used in the field of safety and reliability.