LAUSR

Abstract Extreme value statistics are one way to determine the maximum design loads for systems in extreme conditions, such as operational loads experienced by ships. Accurate predictions typically require large sample sizes, which are not always possible to obtain. Conversely, small sample sizes lead to more variation in the predictions. Increasing the sample size improves the variance to a desired range. The proposed method aimed to estimate a minimum sample size for an extreme value process by specifying and obtaining an acceptable variance. Minimum sample sizes for extreme value statistics depend on the distribution's behavior, so the method proposed here was designed for use before and during measurements. To test the proposed method, the response of a cantilever fin with a varying angle of attack was measured. The proposed method was able to estimate minimum sample sizes for several distributions. Accuracy was demonstrated by randomly drawing measured and simulated samples.