抄録
The applicability of extreme value statistics to localized corrosion was studied based on a actual case. The localized corrosion under tubercles on inner surface of the steel channel in which river water flowed was investigated. Their depths obeyed the distribution of exponential type, while the maximum depths did the double exponential distribution. When the similar data appeared in the previous papers, extreme value statistics was often assumed to be applicable. However, according to the theory of extreme value statistics, the maximum depths are necessary to be sampled from a single population.
Then, it was evaluated by calculating the confidence intervals for the population mean whether the data belong to the same population or not. Moreover, the theoretical distribution of maximum depths was calculated and was compared with the measured one. The results of these analyses showed that extreme value statistics was inapplicable to the present case. Thus, the present case study demonstrates that extreme value statistics is not always applicable to the maximum depths of localized corrosion, even if the depths obey the distribution of exponential type and the maximum depths do the double exponential distribution.
