Abstract
The application of extreme-value statistics to the problem of predicting the growth of Type II pits on hot water pipes is described. Optimum combinations of the size and number of unit samples required for reasonable extreme-value prediction were studied by using sets of pit depth data on pipes removed after seven years of field service, covering return periods of up to 1670. Based on the results of the analysis, the following conclusions can be drawn concerning the optimum conditions for obtaining a reasonable extreme-value prediction.
(1) The number and size of unit samples for the extreme-value survey should be determined so that the variance of error of extreme-value estimates is minimized within specific distribution parameters, while the standard deviation of the estimates is limited to one-third of the mode of distribution.
(2) To improve the accuracy of the estimates, increasing the unit sample size may be more effective than increasing the sample number.
(3) As a practical guideline, extreme-value analysis of a sample amounting to 3% of the whole, with a return period of less than 500, is taken to be necessary in order to increase the reliability of extreme-value prediction.
