| With the rapid development of building technology and the increasing improvement of the performance of building materials,long-span roof structures are more widely used in industrial and civil buildings,such as transportation hubs,convention and exhibition centers,industrial factories,stadiums and goods warehouses.Long-span roof structure has many advantages,such as diverse structure,beautiful shape,scientific force,large internal space and so on.In addition,long-span roof structure is a typical wind-sensitive structure,it has become the key to reasonable structural design to accurately obtain the distribution law of wind load on the roof surface.GB 50009-2012"Code for load of Building structures"stipulates that wind tunnel test is needed to obtain the design wind load of long-span roof structures with complex shape.However,in the actual design process,it is difficult to take into account the wind loads of all the measuring points.In order to improve the calculation efficiency of structural design,the wind loads of several measuring points on the surface of the structure obtained from the wind tunnel test are usually treated as a small number of regional wind loads.The structural design calculation is carried out according to the representative value of the wind load in the corresponding area.In this thesis,the improved K-means clustering algorithm is used to study the wind load zoning of various forms of roof structures,and the reasonable results of wind load zoning are obtained,which is convenient for the design and calculation of roof structures.The details include:(1)In view of the lack of theoretical basis for the selection of classification number k value when K-means clustering algorithm is used for wind load zoning of roof structure,both artificial given and initial clustering centers must be randomly selected,which will lead to too many classification cases and the deficiency of large workload in selecting the optimal classification.In this thesis,the basic idea of the elbow method was introduced,and the accurate identification of the optimal classification number kst was realized through the curve of the relationship between the classification number k and the sum of squares of wind load errors at the corresponding measuring points;On the basis of the random selection of the first initial clustering center,the basic idea of roulette method was introduced to complete the efficient selection of the remaining initial clustering centers;According to the principle of intra-class compactness and inter-class dispersion,through the intra-class compactness index S and inter-class dispersion index D,the optimal wind load zoning result was obtained with the help of SD validity index.(2)Taking the long-span cantilevered roof structure of the Tennis Center of Beijing Olympic Park as an example,the wind load on the surface of the roof was obtained through the wind tunnel test,and the most unfavorable wind load on the upper and lower surface of the roof under the full wind angle was calculated by using the improved algorithm,and the amount of calculation of the optimal zoning result could be obtained by comparing the improved algorithm,which proved the high efficiency of the improved algorithm.(3)The surface wind load data of the roof structure of Beijing Olympic Park Tennis Center were obtained through wind tunnel test.The net wind pressure coefficients of 360 measuring points on 12 roof units were calculated by improved K-means clustering algorithm,and the final zoning results were obtained to provide support for the wind resistance design of the roof structure.(4)The wind loads on the surfaces of three kinds of common roofs(cylindrical,plane and spherical)were obtained by CFD numerical simulation,and the wind loads on the surfaces of three kinds of long-span roof structures were calculated by using the improved K-means clustering algorithm,and the wind load zoning results of three kinds of roof structures under the corresponding wind angles were obtained,which provided a reference for the design of long-span roof structures. |