PROBABILISTIC MODELLING OF LONG-TERM PITTING CORROSION OF STRUCTURAL STEELS UNDER MARINE IMMERSION CONDITIONS

摘要

Long-term corrosion of steel to marine immersion conditions is strongly influenced by pitting caused by the corrosive action of the metabolites of sulphate reducing bacteria and not by long-term cathodic oxidation processes. This means that mean and extreme pit depth as a function of time for longer exposures are not described by a simple function of form x(t) =At~B as commonly assumed. As shown earlier, a more accurate description of expected pit depth is a piecewise function that changes with the process controlling corrosion. To represent the uncertainty associated with pit depth growth and to predict the probability of perforation of a given metal thickness it is conventional to use extreme value theory and in particular the Gumbel distribution. However, this has been shown recently to be not entirely appropriate, even for relatively short-term observations when oxidation is the main corrosion mechanism- For longer term corrosion it may be shown that the Frechet extreme value distribution is more appropriate. This is a further departure from convention. Both the new mean growth law formulation and the Frechet distribution have serious implications for the prediction of long-term pitting corrosion failure. This is demonstrated with a simple example.