Statistical characterization of pitting corrosion - Part 1: Data analysis

Maximum pit depth is a matter of considerable practical interest. In applied corrosion mechanics it is conventional to represent it by an extreme value distribution, usually Gumbel. However. a detailed examination of recent field data for maximum pit depth of mild steel under marine conditions shows that it can befitted as well or better to a bi-modal probability distribution. similar to that which is typical for the underlying pit depth distribution. A reexamination of some classical maximum pit depth data extracted from the literature for aluminum and stainless steel shows a similar phenomenon. Theoretical arguments suggest one reason for this is that the coupon size typically used may be insufficient to capture the extremes. Also. on the basis of modem understanding of the pitting phenomenon. it is proposed that the formation of pits of maximum depth is consistent only with pits that initiate immediately upon exposure and then experience stable pitting behavior. Pits that undergo metastable pitting are unlikely to become extreme and constitute a different statistical population. Moreover. the depths of maximal pits are likely to be highly dependent. Both factors cast doubt on the conventional use of extreme value theory in representing the uncertainty associated with maximum pit depth.