Large-span roof wind pressure partitioning method and system based on information entropy weighted canopy-k-means algorithm

Abstract

The present application relates to a kind of large-span roof wind pressure partitioning method and system based on information entropy weighting Canopy-K-means algorithm, the method is by introducing information entropy and within-class standard deviation as weight factor, the threshold of Canopy algorithm is adaptively modified, to obtain optimal threshold value automatically, subsequently using Canopy algorithm determines initial clustering center.In addition, by introducing clustering effectiveness index, the optimal cluster number value is accurately determined.Therefore, the algorithm can effectively reduce the blindness of initial value selection such as threshold, clustering center and cluster number center, reasonably and efficiently obtain the optimal wind pressure partitioning result of large-span roof surface, and has important engineering practical value.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of building wind resistance, and particularly relates to a large-span roof wind pressure partitioning method and system based on an information entropy weighted Canopy-K-means algorithm. BACKGROUND

[0002] The wind pressure distribution on the surface of a large-span roof is usually complex and uneven, and reasonable and efficient partitioning of the wind pressure on the surface is an important prerequisite for accurately determining the representative value of the wind load, and this process is crucial to the wind resistance safety of the large-span roof.

[0003] As a typical unsupervised clustering method, the K-means algorithm can effectively mine the internal structural features of a data set, and has been widely used in the wind load partitioning of large-span roofs. However, the K-means algorithm is too dependent on the number of clusters K, and an unreasonable K value may lead to the phenomenon of "under-clustering" or "over-clustering" in the clustering results, thereby affecting the accuracy of the wind pressure partitioning results. In addition, the K-means algorithm is very sensitive to the initial clustering center, and the clustering effect on non-convex data sets is poor, and different initial centers may cause the algorithm to fall into a local minimum value. SUMMARY

[0004] To solve the above problems, the application provides a large-span roof wind pressure partitioning method based on an information entropy weighted Canopy-K-means algorithm. The method introduces information entropy and intra-class standard deviation as weight factors to adaptively correct the threshold value of the Canopy algorithm to automatically obtain the optimal threshold value, and then uses the Canopy algorithm to determine the initial clustering center. In addition, the method introduces a clustering effectiveness index to accurately determine the optimal cluster number value. The application can effectively reduce the blindness of the selection of initial values such as the threshold value, the clustering center and the cluster number center, and reasonably and efficiently obtain the optimal wind pressure partitioning result on the surface of a large-span roof, and has important engineering practical value.

[0005] To achieve one of the above application purposes, an embodiment of the application provides a large-span roof wind pressure partitioning method based on an information entropy weighted Canopy-K-means algorithm, which comprises,

[0006] 0006. Step 1, standardizing the data set of the wind pressure coefficient of the large-span roof, and calculating the distance sequence between the standardized wind pressure coefficient of each measuring point and the global mean value based on the standardized data set;

[0007] 0007. Step 2, using the distance sequence obtained in step 1 to construct a comprehensive evaluation index based on information entropy weighting, and determining the optimal threshold value of the Canopy clustering based on the optimal principle of the comprehensive evaluation index;

[0008] 0008.Step 3, using the optimal threshold obtained in Step 2, the measurement points are clustered by Canopy;

[0009] 0009.Step 4, using the cluster center measurement points obtained in Step 3 as the initial cluster center of K-means, iterative calculation is performed to determine the optimal cluster number;

[0010] 0010.Step 5, based on the optimal cluster number obtained in Step 4, the corresponding K-means clustering result is the wind pressure partition result of the large-span roof.

[0011] As a further improvement of an embodiment of the present application, the standardization processing in Step 1 includes:

[0012] 0012. ,

[0013] 0013.In the formula, C pi is the wind pressure coefficient of measurement point i, i = 1, 2, …, n, n is the total number of measurement points, mean{C pi} represents the global mean of the data set, represents the standard deviation of the data set.

[0014] As a further improvement of an embodiment of the present application, the calculation of the distance sequence between the wind pressure coefficient of each measurement point and the global mean in Step 1 includes,

[0015] Calculate the Euclidean distance between the standardized wind pressure coefficient of each measurement point in the standardized data set and the global mean:

[0016] ,

[0017] In the formula, D i represents the Euclidean distance between the standardized wind pressure coefficient of measurement point i and the global mean, C is the standardized wind pressure coefficient of measurement point i,

[0018] Sort the Euclidean distance to obtain the distance sequence between the standardized wind pressure coefficient of each measurement point and the global mean.

[0019] As a further improvement of an embodiment of the present application, the distance sequence obtained in Step 1 is used to construct a comprehensive evaluation index based on information entropy weighting, and based on the optimal principle of the comprehensive evaluation index, the optimal threshold of Canopy clustering includes,

[0020] Using the distance sequence obtained in Step 1, calculate the candidate threshold set T = {t1, t2, …, t j}, j = 1, 2, …, m, t j j represents the candidate threshold value, and m represents the number of candidate threshold values;

[0021] Using the candidate threshold value t i The normalized data set is divided into two groups: Group1(t j ) and Group2(t j ):

[0022] , ,

[0023] In the formula, M is the sampling interval; The Euclidean distance between the normalized wind pressure coefficient of measuring point i and the global mean value is represented by i = 1, 2, …, n, and n is the total number of measuring points;

[0024] Based on the information entropy weighted comprehensive evaluation index, the clustering quality under different candidate threshold values is evaluated: respectively

[0025] ,

[0026] In the formula, J(t i ) represents the comprehensive evaluation index of t j , H(t i ) represents the information entropy of t j , p1 and p2 represent the proportion of Group1(t j ) and Group2(t j ) respectively, , n1 and n2 represent the total number of measuring points corresponding to Group1(t j ) and Group2(t j ) respectively, and λ is the weight coefficient, , , , is the normalized wind pressure coefficient of measuring point i1, is the normalized wind pressure coefficient of measuring point i2, and are the global mean values of Group1(t j ) and Group2(t j ) respectively;

[0027] Based on the comprehensive evaluation index, the optimal threshold value of Canopy clustering is determined as:

[0028] ,

[0029] ,

[0030] In the formula, T1 and T2 are the first and second optimal thresholds for Canopy clustering, and α is the weight coefficient.

[0031] As a further improvement to one embodiment of the present invention, in step 4, the cluster center measurement points obtained in step 3 are used as the initial cluster centers for K-means to perform iterative calculations to determine the optimal number of clusters, including:

[0032] Step 4.1: Set the initial value K0 for the number of K-means clusters. If K0 is less than or equal to the number of cluster center measurement points obtained in Step 3, select the K0 measurement points that are at the top of the distance sequence from the cluster center measurement points obtained in Step 3 as K-means cluster centers, perform K-means clustering, and calculate the corresponding clustering effectiveness index; otherwise, based on the cluster center measurement points obtained in Step 3, randomly select measurement points to supplement K0 as K-means cluster centers, perform K-means clustering, and calculate the corresponding clustering effectiveness index.

[0033] Step 4.2: Let K0 = K0 + 1, return to step 4.1, and continue until the maximum number of clusters is satisfied;

[0034] Step 4.3: Select the number of K-means clusters corresponding to the optimal clustering effectiveness index as the optimal number of clusters.

[0035] As a further improvement of one embodiment of the present invention, the clustering effectiveness index is the Dunn index, and the number of candidate clusters corresponding to the largest Dunn index is selected as the optimal number of clusters.

[0036] As a further improvement to one embodiment of the present invention, the Dunn index is:

[0037] ,

[0038] ,

[0039] In the formula, C is the set of all clusters; c a and c b The minimum distance between them c a Middle measurement point i a Standardized wind pressure coefficient With c b Middle measurement point i b Standardized wind pressure coefficient The distance between them.

[0040] To achieve the above-mentioned one of the purposes of the application, an embodiment of the application further provides a large-span roof wind pressure partition system applying the method, the system comprising a data processing module, a Canopy threshold module, a Canopy clustering module, a K-means clustering module and a partition module.

[0041] The data processing module is used for standardizing a data set of wind pressure coefficients of the large-span roof, and calculating a distance sequence between the standardized wind pressure coefficient of each measuring point and a global mean value based on the standardized data set.

[0042] The Canopy threshold module is used for constructing a comprehensive evaluation index based on information entropy weighting by using the distance sequence obtained by the data processing module, and determining an optimal threshold of Canopy clustering based on an optimal principle of the comprehensive evaluation index.

[0043] The Canopy clustering module is used for clustering the measuring points by using the optimal threshold obtained by the Canopy threshold module.

[0044] The K-means clustering module is used for performing iterative calculation by taking the cluster center measuring point obtained in step 3 as a cluster center of K-means, to determine an optimal cluster number.

[0045] The partition module is used for outputting the K-means clustering result corresponding to the optimal cluster number as a large-span roof wind pressure partition result.

[0046] To achieve the above-mentioned one of the purposes of the application, an embodiment of the application further provides an electronic device comprising a memory and a processor, the memory storing a computer program capable of running on the processor, and the processor implements the steps in the large-span roof wind pressure partition method based on the information entropy weighted Canopy-K-means algorithm when executing the program.

[0047] To achieve the above-mentioned one of the purposes of the application, an embodiment of the application further provides a storage medium storing a computer program, and the computer program implements the steps in the large-span roof wind pressure partition method based on the information entropy weighted Canopy-K-means algorithm when executed by a processor.

[0048] Compared with the prior art, the information entropy weighted Canopy-K-means algorithm is used in the application, the optimal threshold is adaptively determined by introducing an information entropy weighted comprehensive index, and the initial clustering center is further determined by means of the Canopy algorithm. In addition, the optimal clustering number value is accurately determined based on a clustering effectiveness index, the blindness and sensitivity problems of the Canopy-K-means algorithm in the selection of parameters such as the threshold, the clustering center and the clustering number are effectively solved, and the accuracy and reliability of the clustering analysis process are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 Flow chart of the large-span roof wind pressure partition method based on the information entropy weighted Canopy-K-means algorithm.

[0050] Figure 2a Geometric diagram of a flat roof.

[0051] Figure 2b Geometric diagram of a cylindrical roof.

[0052] Figure 2c Geometric diagram of a spherical roof.

[0053] Figure 3a Geometric diagram of the top of a flat roof.

[0054] Figure 3b Geometric diagram of the top of a cylindrical roof.

[0055] Figure 3c Geometric diagram of the top of a spherical roof.

[0056] Figure 4a Layout of measuring points and wind direction angle of a flat roof.

[0057] Figure 4b Layout of measuring points and wind direction angle of a cylindrical roof.

[0058] Figure 4c Layout of measuring points and wind direction angle of a spherical roof.

[0059] Figure 5a Change curve of the information entropy weighted index of a flat roof under a 0° wind direction angle.

[0060] Figure 5b Clustering partition diagram of the surface wind pressure coefficient of a flat roof under a 0° wind direction angle.

[0061] Figure 5c Real partition diagram of the surface wind pressure coefficient of a flat roof under a 0° wind direction angle.

[0062] Figure 6a Change curve of the information entropy weighted index of a flat roof under a 45° wind direction angle.

[0063] Figure 6b Clustered zoning map of wind pressure coefficient on flat roof surface under 45° wind direction angle.

[0064] Figure 6c Real zoning map of wind pressure coefficient on flat roof surface under 45° wind direction angle.

[0065] Figure 7a Information entropy weighted index variation curve of cylindrical roof under 0° wind direction angle.

[0066] Figure 7b Clustered zoning map of wind pressure coefficient on cylindrical roof surface under 0° wind direction angle.

[0067] Figure 7c Real zoning map of wind pressure coefficient on cylindrical roof surface under 0° wind direction angle.

[0068] Figure 8a Information entropy weighted index variation curve of cylindrical roof under 30° wind direction angle.

[0069] Figure 8b Clustered zoning map of wind pressure coefficient on cylindrical roof surface under 30° wind direction angle.

[0070] Figure 8c Real zoning map of wind pressure coefficient on cylindrical roof surface under 30° wind direction angle.

[0071] Figure 9a Information entropy weighted index variation curve of spherical roof.

[0072] Figure 9b Clustered zoning map of wind pressure coefficient on spherical roof surface.

[0073] Figure 9c Real zoning map of wind pressure coefficient on spherical roof surface. DETAILED DESCRIPTION

[0074] The application will be described in detail below with reference to specific embodiments shown in the attached drawings. These embodiments do not limit the application, and changes in structure, method, or function made by those of ordinary skill in the art based on these embodiments are included in the scope of the application.

[0075] The embodiments of the present application will be described in detail below with reference to the drawings. The embodiments described below are examples for explaining the present application and should not be construed as limiting the present application.

[0076] The application is based on a large-span roof wind pressure partitioning method based on information entropy weighted Canopy-K-means algorithm, as shown in the technical scheme, as follows: Figure 1

[0077] Step 1, standardize the large-span roof wind pressure coefficient data set to eliminate the influence of wind pressure level differences between different measuring points on the clustering results. Based on the standardized data, the distance sequence between the wind pressure coefficient of each measuring point and the global mean value is calculated.

[0078] Step 2, use the distance sequence obtained in step 1 to construct a comprehensive evaluation index based on information entropy weighting to quantify the clustering division quality. Then, based on the principle of optimal comprehensive evaluation index, the optimal threshold T1 and T2 of Canopy clustering are adaptively determined by traversing the candidate threshold interval.

[0079] Step 3, use the optimal threshold T1 and T2 obtained in step 2 to perform Canopy clustering division on the measuring points. Calculate the distance between the measuring points and the existing cluster centers, and assign them to the corresponding clusters in turn until all measuring points are assigned.

[0080] Step 4, based on the cluster center measuring points obtained in step 3, perform iterative calculation as the initial cluster center of K-means. Update the cluster center in each iteration process and judge the convergence condition until the stable clustering calculation result is obtained. At the same time, introduce the cluster validity index (such as Dunn index, silhouette index, etc.) to quantitatively analyze the division effect under different cluster numbers to determine the optimal cluster number K.

[0081] Step 5, output the final clustering result, i.e. the final wind pressure coefficient partitioning map of the large-span roof surface, to provide a reliable basis for wind resistance design and safety evaluation of large-span roof structures.

[0082] Specifically, step 1 calculates the global mean and standard deviation of the wind pressure coefficient of all measuring points based on the input wind pressure coefficient data set, which are defined as mean{C pi} and :

[0083]

[0084]

[0085] wherein C pi is the wind pressure coefficient of each measuring point; n is the total number of measuring points. Then, the input data set is standardized by Z-score:

[0086]

[0087] wherein, ​This represents the standardized wind pressure coefficient at each measuring point. Finally, the Euclidean distance D between each measuring point in the standardized dataset and its global mean is calculated. i After sorting, a distance sequence is formed for subsequent analysis. The specific mathematical formula is as follows:

[0088]

[0089] In the formula: This is the global mean of the standardized dataset for wind pressure coefficients.

[0090] Specifically, step 2 is based on the generated distance sequence D i Calculate the candidate threshold set T = {t1, t2, …, t}. j}, j=1, 2, …, m. Then, using the candidate threshold t j The standardized dataset is divided into two groups, Group1 and Group2, for subsequent cluster analysis. The specific mathematical formula is as follows:

[0091]

[0092]

[0093]

[0094] In the formula: M is the sampling interval; n1 and n2 are the total number of measurement points in the corresponding datasets of Group1 and Group2, respectively. Based on the candidate threshold t j Based on the partitioning results, the proportions p1 and p2 of the two datasets are calculated respectively, and the corresponding information entropy H(t) is further calculated. j The specific mathematical formula is as follows:

[0095]

[0096]

[0097] Subsequently, the within-class standard deviations of the two datasets were calculated. and And calculate the mean of the two. The specific mathematical formula is as follows:

[0098]

[0099]

[0100]

[0101] Based on this, a comprehensive evaluation index J(t) based on information entropy weighting is constructed. jThe comprehensive evaluation index J(t) is used to evaluate the clustering quality at different thresholds. This is achieved by traversing the candidate threshold set. j The threshold corresponding to the minimum point is taken as the optimal threshold T1. Then, the optimal threshold T2 is determined based on threshold T1. The specific mathematical formula is as follows:

[0102]

[0103]

[0104]

[0105] In the formula: λ is the weighting coefficient, which is generally taken as 0.5; T1 and T2 are the optimal thresholds; α is the weighting coefficient, which is generally taken as 0.8.

[0106] Specifically, in step 3, the optimal thresholds T1 and T2 are first set as the thresholds for the Canopy algorithm. Then, some unlabeled measurement points are used as cluster centers, and the Euclidean distance d between all measurement points and the cluster center is calculated. i The specific mathematical formula is as follows:

[0107]

[0108] In the formula: This represents the wind pressure coefficient corresponding to the cluster center measurement point. Unmarked measurement points whose distance to the current cluster center is less than T1 are merged into a single cluster. Measurement points whose distance is less than T2 are marked to prevent them from being repeatedly selected as new cluster centers. This selection process is repeated until all measurement points are successfully assigned. Finally, the center measurement points of all clusters are output.

[0109] Specifically, in step 4, the cluster centers generated by the Canopy algorithm are used as the initial cluster centers for the K-means algorithm, and the initial number of clusters K0 is input. Then, the distance from each data point to each initial cluster center is calculated, and each data point is assigned to the nearest cluster (c...). k In the next step, the cluster center position is updated based on the current distribution of measurement points within the cluster to continuously reduce the sum of squared errors (SSE) of the clusters. The iteration terminates and the final clustering result is output when the SSE no longer changes or the objective function converges. The specific mathematical formula is as follows:

[0110]

[0111]

[0112] In the formula: c k Let k be the k-th cluster, k = 1, 2, 3, …, K0; μ k For c k The wind pressure coefficient corresponding to the center measuring point.

[0113] The candidate cluster number set K = {K0, K0+1, K0+2, …, K0+N} (N is a preset value) is looped, and the cluster validity index under different K values is calculated, for example, the Dunn index. When the Dunn index is selected, the K value corresponding to the maximum value of the Dunn index is taken as the optimal cluster number, and the specific mathematical formula is as follows:

[0114]

[0115]

[0116] In the formula, Dunn is the Dunn index; C is the set of all clusters; δ is the minimum distance between clusters, represents the minimum distance between c a and c b , represents the distance between the normalized wind pressure coefficient of measuring point i a in c a and the normalized wind pressure coefficient of measuring point i b in c b .

[0117] Specifically, step 5 outputs the final clustering result based on the optimal cluster number, that is, the wind pressure coefficient partition map of the large-span roof surface, which provides a reliable basis for the wind resistance design and safety evaluation of the large-span roof structure.

[0118] Embodiment

[0119] To verify the accuracy and applicability of the information entropy weighted Canopy-K-means clustering algorithm proposed in the present application in the calculation of wind pressure partition on the surface of a large-span roof, three typical large-span roof structures, namely a flat roof, a cylindrical roof and a spherical roof, are selected as research objects, which cover the main forms of common large-span roofs. Based on the wind tunnel test of the rigid model, the average wind pressure coefficient dataset of all measuring points on the roof surface is obtained, and the clustering partition calculation is performed on the wind pressure distribution on the roof surface under the wind direction angles of 0°, 30° and 45°. Among them, the most unfavorable wind direction angle of the flat roof and the cylindrical roof is 45° and 30° respectively. Considering the wind tunnel test conditions and the characteristic size of the roof structure, the geometric scale ratio of the large-span flat roof and the spherical roof model is 1:100, and the geometric scale ratio of the cylindrical roof model is 1:200. The geometric size and measuring point arrangement of the large-span roof model are shown in Figures 2a to 2c , Figures 3a to 3c and Figures 4a to 4c , wherein the number of measuring points of the flat roof, the spherical roof and the cylindrical roof is 413, 1092 and 729 respectively.

[0120] ​​(1) Wind pressure zoning based on information entropy weighted Canopy-K-means clustering algorithm

[0121] In the clustering calculation process, the wind pressure coefficient data sets of the three large-span roof surface measuring points of the flat roof, the cylindrical roof and the spherical roof are taken as input variables to achieve the optimal clustering zoning as the final goal. According to steps 1 to 5, the average wind pressure coefficients of the surfaces of the above three roofs are calculated. After obtaining the clustering zoning results through the above steps, the clustering zoning results and the true zoning results under different wind direction angles (0°, 30° and 45°) are compared and analyzed. Among them, the true zoning results comprehensively refer to the zoning principles of ASCE / SEI 7-22 of the United States and BS EN1991-1-4:2005 of Europe and the actual distribution characteristics of the wind pressure cloud map.

[0122] Taking a large-span flat roof as an example, Figure 5a and Figure 6a The curves of the information entropy weighted comprehensive index with the threshold T1 under 0° and 45° wind direction angles are given respectively. According to the minimum entropy principle, the information entropy weighted comprehensive index reaches the minimum value at T1=0.79 and T1=0.017 under 0° and 45° wind direction angles respectively, so they are taken as the optimal thresholds of this example. As shown in Figure 5b and Figure 5c It can be seen that under 0° wind direction angle, the clustering zoning results obtained by the present application have high consistency with the true zoning at key positions such as the airflow separation zone (I) and the airflow reattachment zone (IV) of the windward leading edge of the roof, which shows that the wind pressure zoning obtained by the clustering algorithm can accurately reflect the wind pressure distribution characteristics of the roof surface. Comparing the zoning results under 45° oblique wind direction, as shown in Figure 6b and Figure 6c The clustering algorithm proposed in the present application can accurately identify the conical vortex influence area (I~III) of the windward leading edge of the roof, and the zoning results are basically consistent with the true zoning results.

[0123] Taking a large-span cylindrical roof as an example, Figure 7a and Figure 8a The curves of the information entropy weighted comprehensive index under 0° and 30° wind direction angle conditions are given respectively. As shown in the figure, under 0° and 30° wind direction angles, the information entropy weighted comprehensive index reaches the minimum value at T1=0.065 and T1=1.81 respectively, and the threshold is taken as the optimal threshold of the subsequent Canopy-K-means algorithm. Comparing Figure 7b and Figure 7c It can be seen that under 0° wind direction angle, the clustering algorithm proposed in the present application divides the cylindrical roof into windward area (I and II), top airflow acceleration area (IV) and wake area (V), which has good consistency with the true zoning results. The clustering zoning and the true zoning under 30° wind direction angle are as shown in Figure 8b and Figure 8cAs shown, the clustering algorithm of the present application can accurately identify the high wind suction value influence area (V) of the windward front edge of the top of the cylindrical roof under 30° wind direction and the middle part, and better reflect the action law of the turbulent flow of the roof surface characteristics under oblique wind direction.

[0124] To further verify the applicability of the algorithm of the present application to large-span curved roof, a spherical roof is selected for example analysis. The curve of the information entropy weighted comprehensive index with the threshold T1 is as shown in Figure 9a As shown, the information entropy weighted comprehensive index takes the minimum value at T1=1.23, and this T1 value is taken as the optimal threshold value of the Canopy-K-means algorithm. By comparing Figure 9b and Figure 9c It can be seen that the algorithm of the present application can accurately identify the flow characteristic region consistent with the true partition, such as the windward area (I) of the spherical roof, the high wind suction value influence area (IV) of the top and the wake area (V) and the like.

[0125] (2) Effectiveness verification of clustering algorithm

[0126] In order to verify the superiority of the clustering algorithm proposed in the present application, an external clustering effectiveness evaluation index, the adjusted Rand index (Adjusted Rand Index, ARI), is introduced to quantitatively evaluate the matching degree of the clustering partition result and the true partition result, and the specific mathematical formula is:

[0127]

[0128]

[0129] In the formula, RI is the Rand index; a1 is the number of similar measuring points classified into the same cluster; a2 is the number of dissimilar measuring points classified into different clusters; a3 is the number of similar measuring points classified into different clusters; and a4 is the number of dissimilar measuring points classified into the same cluster. The value range of the adjusted Rand index ARI is [-1, 1], and the larger the value is, the more consistent the clustering partition result and the true partition are, and the more accurate and reliable the clustering algorithm is.

[0130] Based on this, the Canopy-K-means algorithm based on information entropy weighting is compared with the traditional K-means algorithm, the standard Canopy-K-means algorithm and the density-based Canopy-K-means algorithm. Table 1 to Table 3 respectively give the adjusted Rand index (ARI) values of flat roof, cylindrical roof and spherical roof under different clustering algorithms. The results show that the Canopy-K-means algorithm based on information entropy weighting obtains the highest adjusted Rand index value under different wind direction conditions of the three roofs, and the matching degree of the clustering result and the real partition result is significantly better than that of the other three clustering algorithms, further verifying that the algorithm has significant superiority in the accuracy and reliability of wind pressure partition of large-span roof.

[0131] Table 1 ARI values of flat roof clustering partition results under different clustering algorithms

[0132] Clustering algorithm Invention algorithm Density-based Canopy--means Canopy--means --means 0° 0.97661 0.81574 0.74663 0.4129 45° 0.98831 0.98087 0.53983 0.41494

[0133] Table 2 ARI values of cylindrical roof clustering partition results under different clustering algorithms

[0134] Clustering algorithm Invention algorithm Density-based Canopy--means Canopy--means --means 0° 0.96626 0.60310 0.57004 0.43819 30° 0.94143 0.69897 0.43708 0.35034

[0135] Table 3 ARI values of spherical roof clustering partition results under different clustering algorithms

[0136] Clustering algorithm Invention algorithm Density-based Canopy--means Canopy--means --means 0° 0.83422 0.44045 0.38503 0.32511

[0137] In summary, the effectiveness and applicability of the clustering algorithm are verified by selecting three typical large-span roofs and multiple wind direction conditions. The wind pressure partition results obtained by the algorithm are generally consistent with the real partition results, and can accurately reflect the areas affected by characteristic turbulence such as separation vortex, conical vortex and wake. In addition, compared with the traditional clustering algorithm, the clustering algorithm of the application obtains the optimal value in the external effectiveness index (adjusted Rand index), further verifying the superiority of the algorithm. Therefore, the algorithm can accurately and reliably partition the wind pressure distribution of large-span roof structure under different wind direction angles, and provide reference for wind resistance design and safety evaluation of roof structure.

[0138] The application also provides a large-span roof wind pressure partition system based on the Canopy-K-means algorithm based on information entropy weighting, which comprises a data processing module, a Canopy threshold module, a Canopy clustering module, a K-means clustering module and a partition module.

[0139] The data processing module is used for standardizing the data set of the wind pressure coefficient of the large-span roof, and calculating the distance sequence between the standardized wind pressure coefficient of each measuring point and the global mean value based on the standardized data set.

[0140] The Canopy threshold module is configured to construct a comprehensive evaluation index based on information entropy weighting by using the distance sequence obtained by the data processing module, and determine the optimal threshold of Canopy clustering based on the optimal principle of the comprehensive evaluation index.

[0141] The Canopy clustering module is configured to perform Canopy clustering on the measuring points by using the optimal threshold obtained by the Canopy threshold module.

[0142] The K-means clustering module is configured to perform iterative calculation by taking the clustering center obtained by the Canopy clustering module as the initial clustering center of K-means, and determine the optimal clustering number.

[0143] The partition module is configured to output the K-means clustering result corresponding to the optimal clustering number as the large-span roof wind pressure partition result.

[0144] The application further provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program capable of running on the processor, and the processor implements the steps in the large-span roof wind pressure partition method based on the information entropy weighted Canopy-K-means algorithm when executing the program.

[0145] The application further provides a storage medium storing a computer program, wherein the computer program implements the steps in the large-span roof wind pressure partition method based on the information entropy weighted Canopy-K-means algorithm when executed by a processor.

[0146] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described modules can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0147] The modules described as separate components can or can not be physically separated, and the components shown as modules can or can not be physical modules, that is, they can be located in one place or distributed on multiple network modules. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment scheme.

[0148] In addition, the functional modules in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or 2 or more modules can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of hardware plus software functional module.

[0149] The integrated module realized in the form of the software function module can be stored in a computer readable storage medium. The software function module is stored in a storage medium, and includes a plurality of instructions for enabling a computer system (which can be a personal computer, a server, or a network system, etc.) or a processor to execute part of steps of the method described in each embodiment of the present application. The storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media capable of storing program codes.

[0150] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A wind pressure zoning method for large-span roofs based on the information entropy weighted Canopy-K-means algorithm, characterized in that: include, Step 1: Standardize the dataset of wind pressure coefficients for large-span roofs, and based on the standardized dataset, calculate the distance sequence between the standardized wind pressure coefficients at each measuring point and the global mean. Step 2: Using the distance sequence obtained in Step 1, construct a comprehensive evaluation index based on information entropy weighting, and determine the optimal threshold for Canopy clustering based on the optimal principle of the comprehensive evaluation index. Step 3: Using the optimal threshold obtained in Step 2, perform Canopy clustering on the measurement points; Step 4: Using the cluster center measurement points obtained in Step 3 as the initial cluster centers of K-means, perform iterative calculations to determine the optimal number of clusters; Step 5: Based on the optimal number of clusters obtained in Step 4, the corresponding K-means clustering result is the wind pressure zoning result for the large-span roof. The step involves using the distance sequence obtained in step 1 to construct a comprehensive evaluation index based on information entropy weighting, and then determining the optimal threshold for Canopy clustering based on the principle of optimal comprehensive evaluation index. Using the distance sequence obtained in step 1, calculate the candidate threshold set T = {t1, t2, …, t}. j }, j=1, 2, …, m, t j Let j represent the candidate threshold, and m represent the number of candidate thresholds; Using candidate threshold t i The standardized dataset is divided into Group1(t) j ) and Group2(t j Two groups: ， ， In the formula, M is the sampling interval; The Euclidean distance between the standardized wind pressure coefficient at measuring point i and the global mean is represented by i = 1, 2, ..., n, where n is the total number of measuring points. A comprehensive evaluation index based on information entropy weighting is used to evaluate the clustering quality under different candidate thresholds: (respectively) ， In the formula, J(t) i ) represents t j Comprehensive evaluation indicators H(t) i ) represents t j Information entropy, p1 and p2 represent Group1(t) respectively. j ) and Group2(t j The proportion of ) , n1 and n2 represent Group1(t) respectively. j ) and Group2(t j The total number of measurement points corresponding to ) , where λ is the weighting coefficient. , , , The standardized wind pressure coefficient at measuring point i1 The standardized wind pressure coefficient at measuring point i2 and Group1(t) j ) and Group2(t j The global mean of ). Based on comprehensive evaluation metrics, the optimal threshold for Canopy clustering is determined as follows: ， ， In the formula, T1 and T2 are the first and second optimal thresholds for Canopy clustering, and α is the weight coefficient.

2. The method according to claim 1, characterized in that: The standardization process in step 1 includes: ， In the formula, C is the standardized wind pressure coefficient at measuring point i. pi Let C be the wind pressure coefficient at measuring point i, where i = 1, 2, ..., n, and n is the total number of measuring points. pi } represents the global mean of the dataset. This represents the standard deviation of the dataset.

3. The method according to claim 1, characterized in that: Step 1, which calculates the distance sequence between the wind pressure coefficient at each measuring point and the global mean, includes... Calculate the Euclidean distance between the standardized wind pressure coefficient at each measuring point in the standardized dataset and the global mean: ， In the formula, D i This represents the Euclidean distance between the standardized wind pressure coefficient at measurement point i and the global mean. The standardized wind pressure coefficient at measuring point i. This represents the global mean of the standardized dataset. The Euclidean distances are sorted to obtain a distance sequence between the standardized wind pressure coefficient and the global mean for each measuring point.

4. The method according to claim 1, characterized in that: In step 4, the cluster center measurement points obtained in step 3 are used as the initial cluster centers for K-means to perform iterative calculations to determine the optimal number of clusters, including... Step 4.1: Set the initial value K0 for the number of K-means clusters. If K0 is less than or equal to the number of cluster center measurement points obtained in Step 3, select the K0 cluster center measurement points obtained in Step 3 that are at the top of the distance sequence as the K-means cluster centers, perform K-means clustering, and calculate the corresponding clustering effectiveness index. Otherwise, based on the cluster center measurement points obtained in step 3, randomly select measurement points to supplement K0 K-means cluster centers, perform K-means clustering, and calculate the corresponding clustering effectiveness index; Step 4.2: Let K0 = K0 + 1, return to step 4.1, and continue until the maximum number of clusters is satisfied; Step 4.3: Select the number of K-means clusters corresponding to the optimal clustering effectiveness index as the optimal number of clusters.

5. The method according to claim 4, characterized in that: The clustering effectiveness index is the Dunn index, and the number of K-means clusters corresponding to the largest Dunn index is selected as the optimal number of clusters.

6. The method according to claim 5, characterized in that: The Dunn index is: ， ， In the formula, C is the set of all clusters; c a and c b The minimum distance between them c a Middle measurement point i a Standardized wind pressure coefficient With c b Middle measurement point i b Standardized wind pressure coefficient The distance between them.

7. A large-span roof wind pressure zoning system applying the method described in any one of claims 1-6, characterized in that: It includes a data processing module, a Canopy thresholding module, a Canopy clustering module, a K-means clustering module, and a partitioning module; The data processing module is used to standardize the dataset of wind pressure coefficients for large-span roofs, and based on the standardized dataset, calculate the distance sequence between the standardized wind pressure coefficients of each measuring point and the global mean. The Canopy threshold module is used to construct a comprehensive evaluation index based on information entropy weighting using the distance sequence obtained by the data processing module, and to determine the optimal threshold for Canopy clustering based on the optimal principle of the comprehensive evaluation index. The Canopy clustering module is used to perform Canopy clustering on the measurement points using the optimal threshold obtained by the Canopy thresholding module; The K-means clustering module is used to iteratively calculate the optimal number of clusters by using the cluster center measurement points obtained by the Canopy clustering module as the cluster centers of K-means. The partitioning module is used to output the K-means clustering result corresponding to the optimal number of clusters, which serves as the wind pressure partitioning result for large-span roofs.

8. An electronic device, comprising a memory and a processor, characterized in that: The memory stores a computer program that can run on the processor, and when the program is executed on the processor, it implements the steps in the large-span roof wind pressure zoning method based on the information entropy weighted Canopy-K-means algorithm as described in any one of claims 1-6.

9. A storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, it implements the steps in the large-span roof wind pressure zoning method based on the information entropy weighted Canopy-K-means algorithm as described in any one of claims 1-6.

Images