Non-Gaussian Extreme Wind Pressure and Peak Factor Predictions for a Long-Span Cantilevered Roof

By conducting wind tunnel pressure measurement test based on a rigid model of a long-span cantilevered roof， the estimation of non-Gaussian extreme wind pressure on the roof surface and the prediction of peak factors were studied. The time series of wind pressure on the roof surface was obtained through experiments， and the distribution characteristics of the mean wind pressure coefficient， fluctuating wind pressure coefficient， skewness， and kurtosis on the roof surface were analyzed under typical wind directions of 0 °， 45 °， and 90 °. The Gaussian peak factor method and the revised Hermite series method were used to calculate the peak factor at typical measuring points， respectively. The non-Gaussian characteristics and wind pressure fitting at typical measuring points were evaluated based on five probability density functions， and the minimum wind pressure estimation was obtained using the revised Hermite series method and the existing extreme wind pressure evaluation method， respectively. Finally， the peak factor was predicted using a general regression neural network. The results showed that the minimum wind pressure on the roof surface was significant under typical wind directions of 0 °， 45 °， and 90 °， and the long tail of the wind pressure distribution was measured in the negative direction； the Gaussian peak factor method was found to frequently underestimate the peak factor of non-Gaussian wind pressure； the revised Hermite series method estimated the peak factor more accurately and performed best in wind pressure fitting， especially in the negative pressure long tail section； the revised Hermite series method obtained better minimum estimates； the general regression neural network based on the fourth-order statistics of wind pressure time history exhibited good prediction performance.