Uncertainties Of Wind Pressure And Extreme Value Estimation By Improved Peaks Over Threshold Model For Long Span Roof Cladding

The main structure of Large-span roof is rarely destroyed in the wind damage.However,the damage of building enclosure occurred frequently.To study the uncertainty of wind pressure on long-span roof is of great significance.Windresistance design of building enclosure based on extreme value of wind load,rational probability model is the precondition to forecast the extremal wind load of building enclosure.The current extremum estimation method including classic extreme value theory method which need a large number of samples and the method based on single sample.Classic extreme value theory method can build a more accur ate probability model of extreme wind pressure,but it need to collect a large number of experimental data,leading to higher costs.The method of extremum estimation based on single sample has more practical significance.In engineering,peak factor method based on gaussian assumption is the most widely applied method.For some samples of nongaussian wind pressure,peak factor may underestimate the extreme value of wind load.So research for peaks over threshold model which is adapting different wind pressure distribution is of great significance.Based on wind tunnel test of the rigid model,the uncertainty of columnar roof wind pressure was studied.Used three-parameters generalized pareto distribution model to fit the wind pressure exceeds the threshold.Introduced the change point theory to the threshold selection of peaks over threshold model in extreme value estimation of wind pressure.Finally,based on the classic extreme value theory that have certain assurance rate,the estimation error of different methods based on single sample were compared.The main content of the research is as follows:(1)The main information of the wind tunnel test is presented.Through the analysis of the single sample,the distribution of the wind pressure coefficient and the distribution of the kurtosis and skewness are obtained.Through the analysis of the minority samples,the variation of average wind pressure coefficient,fluctuating wind pressure coefficient,kurtosis coefficient,skewness coefficient and minimu m wind pressure coefficient are discussed.(2)It introduced change point theory and the method of goodness-of-fit test to select the threshold of POT model,improved the traditional methods.Using threeparameters generalized pareto distribution to fit the wind pressure which exceed the threshold and using the probability weighted moment estimation method to estimate distribution parameters,it proves that the refined threshold model method suit the data well.(3)Through examples,the estimation process of peak factor method,improved peak factor method,generalized extreme value method and the improved peaks over threshold model method is illustrated.Based on the classic extreme value theory that have certain assurance rate,the error of the improved threshold model method,peak factor method and improved peak factor method were compared.Founded that no matter of gaussian sample or non-gaussian samples of wind pressure,the error of 99% fractile of refined threshold model is relatively small.Improved threshold model is a more accurate method of extremum estimation.