Extreme value analysis for assessing structural reliability of welded offshore steel structures

• Model maximum pit depth with time calibrated to long-term observations.
• Long-term corrosion parameters variability is quantified by extreme value analysis.
• Parameter contribution to overall failure probability is shown to vary with time.
• Long-term pit depth uncertainty is better represented by a Frechet EV distribution.

Natural deterioration of mild steel exposed to marine environment compromises the long-term integrity, serviceability and safety of new and existing infrastructure and increases the risk of structural failure. Welded structures are known to be prone to even higher risks as a result of adverse effects of pitting corrosion in weld-heated areas. A bi-modal model has been shown recently to be a better description for the long-term development of the maximum depth of pits. Also, the statistics of pit depth have been shown to be better represented, for long term exposures, by the Frechet extreme value distribution. Both new developments present challenges for structural reliability analysis. Herein a linearization is used to represent long-term development of pit depth. It is shown that data for maximum pit depths can be separated into those with Gumbel statistics and those for which a Frechet distribution is more appropriate. An example is given for the reliability analysis of a welded pipeline subjected to localized corrosion. The effect of random variable uncertainty is assessed using a sensitivity study. Results show the considerable influence on the probability of failure of pit diameter and the parameters describing the pitting corrosion model.