鞍形屋盖风压系数模糊聚类分区研究

Wind pressure zones on a saddle roof based on fuzzy C-means clustering

以鞍形屋盖为研究对象,针对其破坏主要始于屋盖边缘等部位的现象和关于鞍形屋盖风压分区无据可循的现状,围绕鞍形屋盖风压系数分区问题,采用模糊C均值聚类根据风易损程度将鞍形屋盖表面划分为多个区域。分区过程中,针对模糊C均值聚类对初始变量具有依赖性的问题进行了改进,包括采取预先限制聚类数目最大值与采用有效性指标确定最优聚类数目的方式。基于此,给出了鞍形屋盖表面的最不利平均、极值风压系数分区图,并计算了各区域的分区风压系数。

It is known that destruction of a saddle roof mainly begins at its edge and now there is no basis to understand how the wind pressure is partitioned on a saddle roof. Here,a saddle roof was divided into multiple regions with the fuzzy C-means clustering method according to the wind vulnerability level. In the process of partitioning,the method of fuzzy C-means clustering was improved in order to get rid of the dependency of clustering on initial variables.The improving measures included limiting in advence the maximum number of clusters and making use of the validity index to determine the optimal clustering number. Based on these measures,the mean and peak of the most unfavorable wind pressure coefficient distribution zones on the saddle roof were obtained and the wind pressure coefficient on each zone was calculated.