基于非高斯仿真的风压系数极值计算方法

Prediction of wind pressure peak factor with non-gaussian simulation

以多变量相关非高斯过程仿真方法为基础,发展了一种基于单次采样的多变量非高斯仿真极值计算方法。首先介绍开孔屋盖的风洞试验概况和多变量相关非高斯过程仿真的基本理论,对屋盖上一组测点风压进行了非高斯仿真,结果表明基于谱修正的多变量相关非高斯过程仿真方法得到的时程在功率谱密度,相干函数,高阶矩三方面与目标值接近,仿真效果较好,然后采用经典极值理论对多次仿真的非高斯时程进行极值计算,将该方法得到的峰值因子与以往常用方法的结果进行比较,结果表明:Davenport峰值因子法高估气流分离区左偏风压的正峰值因子60%,低估负峰值因子43%;Sadek-Simiu峰值因子法低估了高峰度风压的峰值因子50%;而基于单次样本进行仿真的非高斯仿真峰值因子法,其估计的开孔屋盖的峰值因子最为准确,与观察峰值因子总体上最为接近。

Based on the Non-Gaussian Simulation of multivariate stochastic processes method, one prediction method for wind pressure extreme value was proposed. The wind pressure time histories of several opening roofs were simulated with multivariate Non-Gaussian simulation method based on wind tunnel test data. It was shown that power spectral density, coherence, deviation, skewness and kurtosis of simulated Non-Gaussian time histories are very close to the destination values. Then the peak factors of wind pressures on the opening roofs were predicted from the multivariate non-gaussian simulation time histories for several times with the typical extreme value theory, and the results were compared with those several general methods. It was shown that Davenport method overestimates the positive peak factor by 60%, its skewness is negative, and it underestimates the negative peak factor by 43%; Sadek-Simiu method underestimates the peak factor by 50%, it has a higher kurtosis; the proposed method can predict the peak factor effectively, and the overall error is smallest.