Sampling Errors in the Estimation of Peak Wind-Induced Internal Forces in Low-Rise Structures

Abstract

Peak values in time histories of wind effects may be obtained by using the entire information inherent in the time series of the wind effect, a method entailing the estimation of the probability distribution of the peak through the application of the classical Rice procedure extended for non-Gaussian time histories. We present estimates of the sampling errors inherent in this method. These are needed for structural reliability calculations and for decisions on the requisite length of wind tunnel pressure records, especially for database-assisted design. If based on the analysis of 1-h-long records generated by Monte Carlo simulation, typical sampling errors in the estimation of peaks of time histories corresponding to windstorms of 1-h duration are about 5%. If based on 30- or 20-min records, they are about 1.5 times or twice as large, respectively. Consideration of the sampling errors in reliability calculations entails an estimated increase in the requisite safety margins with respect to wind loading of roughly 2, 3, and 5% if 1-h, 30-min, or 20-min records are used, respectively.