A fast partition method for wind pressure coefficient of large-span roof based on modified GK clustering

Structures with large-span roofs are particularly vulnerable to wind action, and evidences have shown that wind-induced damage of large-span roofs occurs mostly on edges and corners. This implies that different regions of large-span roofs may exhibit different degree of sensitivity to wind. It is therefore of substantial importance to partition the roof surface into different regions based on their respective vulnerability, which allows a better wind-resistant design. In this study, the concept of cluster analysis was extended for wind pressure coefficient (WPC) partitioning of large-span roofs, in which a fast partition method based on the modified GK clustering algorithm was employed to partition the WPC on two engineering cases, i.e., a cantilever flat roof and a curved roof. Its reliability was examined via comparative analysis of WPC distribution and the optimal zoning of WPC derived by cluster analysis. The optimal number of clusters was identified using clustering validity indices. The comparisons were overall satisfactory, both in quantitative and qualitative sense. The optimal zoning results of WPC reflect well the characteristics of WPC distribution. It is expected that the methodology applied in this study can be well used to aid the wind-resistant design of other similar structure systems.