Bias and sampling errors in estimation of extremes of non-Gaussian wind pressures by moment-based translation process models

Qingshan Yang, Xinzhong Chen, Min Liu

Research output: Contribution to journalArticle

6 Scopus citations

Abstract

Accurate estimation of extremes of non-Gaussian wind pressures is important for cladding design. The translation process model approach is often used to calculate extreme value distribution of a non-Gaussian process. The translation model can be given as moment-based Hermite polynomial function or derived from three-parameter Gamma distribution with mapping of cumulative distribution functions. This study investigates bias and sampling errors of these two approaches. The prediction error or bias is examined using wind pressure data of a saddle type roof. A refined three-parameter Gamma distribution is proposed and its improved performance is demonstrated. This study also presents formulations for estimating sampling errors of the first four statistical moments. By using non-Gaussian moment closure technique, the sampling errors of statistical moments are estimated from the first four moments. The sampling errors in the estimated extremes of wind pressures are further predicted analytically based on First Order Second Moment (FOSM) method. The proposed framework is different from those presented in literature, where the sampling errors of skewness and kurtosis are calculated by multiple time history samples which are often not available. The accuracy and effectiveness of the proposed framework are examined using the wind pressure data with a very long duration.

Original languageEnglish
Pages (from-to)214-233
Number of pages20
JournalJournal of Wind Engineering and Industrial Aerodynamics
Volume186
DOIs
StatePublished - Mar 2019

Keywords

  • Bias
  • Extreme
  • Moment-based translation process models
  • Non-Gaussian wind pressure
  • Peak factor
  • Sampling errors

Fingerprint Dive into the research topics of 'Bias and sampling errors in estimation of extremes of non-Gaussian wind pressures by moment-based translation process models'. Together they form a unique fingerprint.

  • Cite this