Application of Extreme Value Statistical Analysis to Actual Localized Corrosion Data from Some Operating Machine Components

Abstract

As a series of study to develop a procedure to make a quantitative evaluation of corrosion data, it was demonstrated that the extreme value statistical analysis can be applied to various actual localized corrosion data acquired from presently operating machine components and that corrosion life prediction can be done by examining the time dependence of estimated maximum depth of penetration. At first a procedure to perform quantitative evaluation of pitting corrosion data was briefly described. Plots of corrosion data on extreme value probability paper were explained and followed by description of linear fitting of the plot, a test for goodness of fit and rejection of an isolated value, and determination of control limits for observed values.
Finally attempts were made to demonstrate the usefulness of this analysis for performing quantitative evaluation of various actual localized corrosion data. The following actual localized corrosion data were obtained from the operating machine components: tuberculation of carbon steel feed water pipes for hydroelectric power stations, pitting and crevice corrosion of cupro-nickel heat exchanger tubes employed for a water cooled chilling machine, and crevice corrosion of joints for type 316 stainless steel capillary tubes used for high speed liquid chromatography.