Page 286 - 《振动工程学报》2026年第5期
P. 286
第 39 卷第 5 期 振 动 工 程 学 报 Vol. 39 No. 5
2026 年 5 月 Journal of Vibration Engineering May 2026
考 虑 风 压 相 关 性 的 建 筑 表 面 风 压 分 区 方 法 研 究
李寿科 , 李友新 , 郭 凡 , 杨泽佳 , 苏培林 , 陈 宁 , 刘 敏 2
1
1
1
1
1
1
(1. 湖南科技大学结构抗风与振动控制湖南省重点实验室,湖南 湘潭 411201; 2. 重庆大学土木工程学院,重庆 400044)
摘要:建筑结构荷载规范和风洞试验均有风压分区需求。聚类算法能有效实现风压分区,但传统分区方法常存在区内风压相
关性不明确、分类数特征不明显、初始聚类中心选取随机等问题。为此,本文考虑风压相关性,采用 K-means 方法、SOM(self-
organizing maps)聚类方法和层次聚类方法,对建筑表面风压系数分区开展研究。建筑表面风压系数分区算例表明,基于分类
数 k 与误差平方和及轮廓系数关系曲线,能够快速确定分区最优分类数;将测点风压相关性与分区方法结合,能较好地反映结
构表面风压分布相关特征;综合考虑计算工作量、轮廓系数、CH(Calinski-Harabasz)指数和 DB(Davies-Bouldin)指数等分区效
果评价指标,采用 K-means 方法、SOM 方法对建筑表面风压系数进行分区,均能得到较理想的分区结果,并可有效应用于高层
建筑和大跨空间结构的风压分区。
关键词: 风荷载;风压分区;相关系数;聚类算法;风洞试验
+
中图分类号:TU312 .1;TU311.41 文献标志码:A DOI:10.16385/j.cnki.issn.1004-4523.202406002
Wind pressure zoning method for building surfaces
incorporating wind pressure correlation
1
1
1
1
1
1
LI Shouke ,LI Youxin ,GUO Fan ,YANG Zejia ,SU Peilin ,CHEN Ning ,LIU Min 2
(1.Hunan Provincial Key Laboratory of Structures for Wind Resistance and Vibration Control,Hunan University of Science and Technology,
Xiangtan 411201,China;2.School of Civil Engineering,Chongqing University,Chongqing 400044,China)
Abstract:Both building structural load codes and wind tunnel tests require wind pressure zoning. Clustering algorithms can effectively achieve
wind pressure zoning, but traditional zoning methods often suffer from issues such as unclear wind pressure correlations within zones, indistinct
classification features, and random selection of initial cluster centers. To address this, this paper considers wind pressure correlations and
employs the K-means method, the SOM (self-organizing maps) clustering method, and hierarchical clustering to study the zoning of wind
pressure coefficients on building surfaces. Case studies on wind pressure coefficient zoning for building surfaces demonstrate that the optimal
number of clusters can be rapidly determined based on the relationship curve between the number of clusters k, the sum of squared errors, and
the profile coefficient. Integrating the correlation of wind pressures at measurement points with the zoning method effectively captures the
correlation characteristics of wind pressure distribution on the building surface. Taking into comprehensive consideration evaluation metrics for
partitioning effectiveness—such as computational effort, the silhouette coefficient, the CH (Calinski-Harabasz) index, and the DB (Davies-
Bouldin) index—the application of both the K-means and SOM methods for partitioning wind pressure coefficients on building surfaces yields
satisfactory results. These methods can be effectively applied to wind pressure partitioning for high-rise buildings and large-span spatial
structures.
Keywords:wind load;wind pressure zoning;correlation coefficient;clustering algorithm;wind tunnel test
在结构抗风设计中,通常需对结构表面风荷载 分区要求。在风压分区方法中,学者们认为采用数
进行分区,建筑结构荷载规范 GB 50009—2012 [1] 和 学上的聚类算法可有效进行风压分区,将风压分布
ASCE 7—2016 [2] 中,考虑风荷载分布特征,对建筑 特征相似区域聚合在一起,形成一个聚类,从而得到
表面围护结构局部平均风压系数和极值风压系数进 分区风压。常用的风压分区方法为 K-means 聚类算
[3]
行了分区,建筑结构风洞试验后期数据处理中也有 法,李丹煜等 、张建等 [4] 基于 K-means 聚类算法,采
收稿日期:2024-06-03;修订日期:2024-11-05
基金项目:国家自然科学基金资助项目(52378510,52308501,52378509);中国国家留学基金资助项目(202309480002);
湖南省自然科学基金省企联合基金资助项目(2024JJ9070)

