Peak factor estimation of non‐Gaussian wind pressure on high‐rise buildings

Abstract

Summary A vast quantity of measurements of wind‐induced non‐Gaussian effects on buildings call for the burgeoning development of more advanced extrema estimation approaches for non‐Gaussian processes. In this study, a well‐directed method for estimating the peak factor and modeling the extrema distribution for non‐Gaussian processes is proposed. This method is characterized by using two fitted probability distributions of the parent non‐Gaussian process to separately fulfill the estimations of the extrema on long‐tail and short‐tail sides. In this method, the Johnson transformation is adopted to be the probabilistic model for fitting the parent distribution of the non‐Gaussian process due to its superior fitting goodness and universality. For each dataset, two Johnson transformations will be established by two parameter estimation methods to individually estimate the extrema on two sides. Then a Gumbel assumption is applied for conveniently determining the non‐Gaussian peak factor. This method is validated through long‐duration wind pressure records measured on the model surfaces of a high‐rise building in wind tunnel test. The results show that the proposed method is more accurate and robust than many existing ones in estimating peak factors for non‐Gaussian wind pressures.

Evaluation of XIMIS for Assessing Extreme Pressure Coefficients

“…Reflecting the necessity of the determination of the extreme values of the combination of multiple variables in structural designs, Folgueras et al (2016) suggested extended Davenport peak factor, which can estimate the extreme values of the resultant of the linear combination of multiple Gaussian and non-Gaussian random variables that come from a common agent. Ma and Xu (2017) suggested new peak factor calculation method by employing Johnson transformation to fit the marginal distribution of the non-Gaussian process and estimating extrema on long-tail and short-tail sides separately.…”

“…The Hermite-Davenport Method of Yang et al (2013) From the extensive review of Cp peak estimation methods in section Previous Peak Estimation Methods, it is clear the non-Gaussian peak factor obtained by moment-based Hermite polynomial translation model has received the attention of many researchers. Among several studies dealing with this model, the method suggested by Yang et al (2013), which will be denoted as "YGP method" hereafter, will be selected for our calibration of XIMIS method because of its relatively simple use and accuracy examined by several researchers (Peng et al, 2014;Liu et al, 2017;Ma and Xu, 2017;Song et al, 2019).…”

Parallel Statistical and Machine Learning Methods for Estimation of Physical Load

“…The similar approach in different ways was widely used for analysis of distributions and was successfully applied in various fields of science, including computer science, physics, materials science, finance, geoscience, etc. [29][30][31][32][33].…”

Random Process Peaks Prediction: A Literature Overview

“…However, the three parameter Gamma distribution specific relationship between skewness and kurtosis holds. Once the combinations of skewness and kurtosis of actual samples deviate far away from the relationship, remarkable fitting errors will be incurred [74]. Ma and Xu, 2017 [74], proposed an update of the Gamma method through the Johnson transformation to solve the fitting error.…”