Study on peak density clustering partition of wind pressure coefficient of shell-shaped roofs
Because of the particularity of the distribution of wind pressure on the surface of the shell-shaped roof,a partition method of wind pressure coefficient based on the peak density clustering of the fast search technology was proposed.Based on the test results of surface wind pressure distribution on a shell-shaped roof,the wind pressure coefficient partition of peak density clustering was carried out.The SD validity index was used to verify the rationality of the number of divisions,and compared with the improved k-means clustering partition results.The results show that the peak density clustering wind pressure coefficient partition can better reflect the wind pressure distribution characteristics of the roof with the relative distance of wind pressure coefficient,local density and comprehensive index as the characteristic parameters,and effectively ensure the class aggregation and class separation.Compared with the improved k-means partition method,the number of partitions obtained by peak density clustering under different wind angles is close to the SD optimal clustering number.The peak density clustering partition results can more accurately reflecting the characteristics of wind pressure distribution on the surface of shell-shaped roofs,fully reflecting the fundamental principle that the local density and relative distance of wind pressure coefficients at measurement points have larger characteristic values,which has better applicability for wind pressure coefficient partitioning of shell-shaped roofs.The maximum negative wind pressure coefficient of peak density clustering partition of shell-shaped roofs ranges from-0.59 to-1.74.
peak densityfast searchclusteringpartition of wind pressure coefficientshell-shaped roof
NETL
NSTL
万方数据
维普
