Extreme value statistics in flow-induced vibration over long time intervals

Abstract

Cyclical loading on structures can lead to fatigue damage, and damage accumulation can be worsened from peak loading. In order to predict these peaks, extreme value statistics can be applied to limited vibration data. Tests must be conducted for long enough durations so that a representative population can be obtained. In extreme value statistics, the Generalized Extreme Value (GEV) distribution’s parameters are also dependent on measuring a representative sample. As it is often uneconomical to test for long periods of time, the proposed experiment aims to look at the behavior of extreme value statistics over long time records to determine if a “minimum requirement” is possible. A cantilever hydrofoil is vibrated under three flow conditions at two angles of attack. Extreme value statistics are applied to compare parameters and distributions for different record length increments. Statistics of each increment are used to generate return plots for the prediction of repeated tests. Errors are quantified to determine the accuracy of the different record lengths. The results will indicate how testing length influences GEV parameters and prediction in vibration, giving insight into duration requirements for future fatigue tests.Cyclical loading on structures can lead to fatigue damage, and damage accumulation can be worsened from peak loading. In order to predict these peaks, extreme value statistics can be applied to limited vibration data. Tests must be conducted for long enough durations so that a representative population can be obtained. In extreme value statistics, the Generalized Extreme Value (GEV) distribution’s parameters are also dependent on measuring a representative sample. As it is often uneconomical to test for long periods of time, the proposed experiment aims to look at the behavior of extreme value statistics over long time records to determine if a “minimum requirement” is possible. A cantilever hydrofoil is vibrated under three flow conditions at two angles of attack. Extreme value statistics are applied to compare parameters and distributions for different record length increments. Statistics of each increment are used to generate return plots for the prediction of repeated tests. Errors are quantified to d...