A Study On Extreme Wind Pressure: POT Model

The safety of structures under wind loads has been a key subject in engineering.In calculating the strength of structure and local stability, we need the extreme valueof wind pressure. Since the load code provides only the distribution of mean value ofshape coefficients, in general, the wind tunnel test is carried out to achieve theextreme value. Using proper method, we can get a reliable extreme value. So it is veryimportant to choose a suitable extreme value calculating method. The most widelyused method is the peak factor method, which is based on the assumption of Gaussiandistribution. However, actual time series of wind pressure does not always obey theGaussian distribution, wind pressure at the corner of the model shows remarkablenon-Gaussianity. In this case, the peak factor method tends to obtain an unsafeextreme value. Therefore, to carry out further research on the method of calculatingextreme value has a lot meaning.This thesis makes use of the data obtained from the wind tunnel test of Urumchihigh speed railway station project. It focuses on the features of wind pressure at thelong-span roof. By use the peaks over threshold model method, classical ex tremevalue method and peak factor method, the extreme wind pressure is calculated in threedifferent methods. Through comparison, it proves that the peaks over threshold modelmethod suit the data best. The main innovation of this thesis is that it puts forward anew extreme value calculating method which can be used in any distribution of windpressure. Besides, the relationship among the number of exceeded samples, skewnessand kurtosis is found, which proves the peak factor method is not adoptable fornon-Gaussian wind pressure.Firstly, the fundamental information about the wind tunnel test is introduced.Many high speed railway stations are built in recent years. The main station buildingis usually designed over the whole reception-departure yard at these new stations. Asa result, there is a hole in the center of station building. After preliminary analysis,the roof is consistently under negative pressure in all kinds of wind direction. Ingeneral, high negative pressure is found at the windward side, then decreases fasttoward the center of the roof. The contour line of wind pressure is changed in highfrequency at each edge of the roof, while the negative pressure remains stable at thecenter part. Secondly, how the peaks over threshold model method, classical extreme valuemethod and peak factor method are used is illustrated through an example. Accordingto Gaussian and non-Gaussian samples, the extreme wind pressure obtained throughthese three different methods is compared with each other. The re sult shows that thethree methods tend to calculate an agreed extreme wind pressure when the sampleobeys Gaussian distribution. However, there is some difference among the threemethods with regard to non-Gaussian sample, and it proves that the generalize d Paretodistribution fitted the extreme value best.Finally, after the comprehensive analysis of skewness coefficient and kurtosiscoefficient, it demonstrates the non-Gaussianity is remarkable at each edge of the roof.What is more, wind pressure sample at roof center obeys Gaussian distribution well.At last, the peaks over threshold model method is adopted to calculate extreme windpressure. In a word, the maximum value of wind pressure at roof edge may be positiveor negative under different wind directions, but the minimal value is far greater thanthe maximum value as an absolute value, and the central part of the roof isconsistently under high negative wind pressure.