Evaluation of Corrosion Failure by Extreme Value Statistics
Toshio Shibata
pp. 115-121
Abstract
Recent advances in the field of corrosion failure evaluation by the extreme value statistics have been reviewed. Historical background of the extreme value statistics, and recent developments in the theory and application to corrosion are described. The extreme value distribution contains three asymptotic distributions which are derived from the initial distributions, but the first type for the maximum value, i.e., the Gumbel distribution and the third type for the minimum value, i.e., the Weibull distribution are mainly used for evaluating the maximum depth of corrosion penetration and the minimum time for corrosion failure, respectively. It is emphasized that the parameter estimation could be accomplished simply by using a minimum variance linear unbiased estimator (MVLUE) method which has been refined recently. Several problems in applying the MVLUE method were discussed with referring experiences in the data analysis.