Derivation of time-varying mean for non-stationary downburst winds

Yanwen Su, Guoqing Huang, You Lin Xu

Research output: Journal article publicationJournal articleAcademic researchpeer-review

66 Citations (Scopus)


Non-stationary extreme winds cause significant damages to buildings and other structures worldwide. Accurate modeling of these winds is crucial to the evaluation of structural safety. However, the derivation of a reasonable time-varying mean for these extreme winds appears to be not straightforward due to the non-stationarity, which is different from the stationary boundary layer winds. Currently, a variety of techniques have been developed to derive the time-varying mean for non-stationary winds, such as moving average, kernel regression (KR), discrete wavelet transform (DWT) and empirical mode decomposition (EMD). However, these approaches with different parameters may lead to inconsistent time-varying means and ensuing fluctuations. The evaluation of these approaches and corresponding non-stationary wind effects on structures has not been sufficiently addressed in previous research. In this study, two sets of full-scale non-stationary downburst wind records are used as examples to evaluate the performance of three approaches including KR, DWT and ensemble EMD with different time window sizes in deriving the time-varying mean. Based on these evaluations, the recommendations about the selection of the appropriate approach and time window size to derive a reasonable time-varying mean are provided.
Original languageEnglish
Pages (from-to)39-48
Number of pages10
JournalJournal of Wind Engineering and Industrial Aerodynamics
Publication statusPublished - 1 Jun 2015


  • Discrete wavelet transform
  • Downburst wind
  • Empirical mode decomposition
  • Kernel regression
  • Non-stationary
  • Time-varying mean

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Renewable Energy, Sustainability and the Environment
  • Mechanical Engineering

Cite this