A physical wind tunnel pressure test optimization method and storage medium based on CFD-AI

Abstract

The present invention discloses a physical wind tunnel pressure test optimization method and storage medium based on CFD-AI. The method completes the arrangement of wind load measurement points after segmentation and completes surface partitioning based on a parametric modeling platform. A CFD numerical wind tunnel calculation model is established and CFD full wind direction angle calculation is performed to extract the wind pressure and local body coefficient of the building surface measurement points corresponding to all wind direction angles. The local body coefficient is combined with the three-dimensional building model to perform machine learning cluster analysis on the measurement points to form an optimized measurement point arrangement scheme. The surface partition is updated based on the optimized measurement point arrangement scheme and Thiessen polygons, and a physical wind tunnel test is carried out. The wind load analysis results of the physical wind tunnel and the CFD numerical wind tunnel are integrated to mark the areas where the wind pressure exceeds the specification. The overall body coefficient is calculated and output by combining the surface partition area and the local body coefficient. The method aims to improve the problems of high pressure test cost, long test data processing time, and test efficiency and data accuracy that cannot meet the requirements of actual engineering applications in physical wind tunnel tests.

Description

Technical Field

[0001] The present invention belongs to the technical field of composite application of structural wind engineering and CFD technology, and specifically relates to a physical wind tunnel pressure test optimization method and storage medium based on CFD-AI. Background Art

[0002] Physical wind tunnel testing is an important means of evaluating a building's structural safety performance under wind loads. By simulating actual wind conditions, aerodynamic tests are conducted on building models to obtain data on wind pressure distribution on their surfaces.

[0003] Traditional physical wind tunnel testing often relies on engineers' judgment or simple uniform placement strategies when arranging wind pressure measurement points. This approach has several significant limitations: 1) Low test efficiency. Excessive measurement points increase the workload for data collection and processing, extending the test cycle; while too few measurement points may miss important information, affecting the accuracy of test results. 2) High cost. Improper measurement point placement increases the equipment and labor costs of wind tunnel testing, a cost issue that is particularly prominent when testing large or complex building models.

[0004] Computational fluid dynamics (CFD) simulations are currently widely used in architectural wind tunnel simulations, but their integration with physical wind tunnels is rare. Conventional CFD measurement point placement schemes cover a large number of measurement points, but they cannot address the limited number of measurement points required for physical wind tunnels. This makes CFD analysis methods unsuitable for physical wind tunnels. Effectively integrating CFD technology with physical wind tunnels to enhance the accuracy and efficiency of measurement points in physical wind tunnel testing has become a new challenge that needs to be addressed. Summary of the Invention

[0005] The technical problem to be solved by this application is to provide a physical wind tunnel pressure test optimization method and storage medium based on CFD-AI, aiming to improve the problems of high pressure test cost, long test data processing time, and test efficiency and data accuracy that cannot meet the actual engineering application in physical wind tunnel tests.

[0006] Furthermore, the present invention aims to solve the problem that the measurement point arrangement scheme of computational fluid dynamics (CFD) covers a large number of measurement points, but cannot solve the problem that the number of measurement points arranged in a physical wind tunnel is limited.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] A physical wind tunnel pressure test optimization method based on CFD-AI, characterized by comprising the following steps:

[0009] S1. Import the complex building model into NURBS surface modeling software, complete the repair and adjustment of the building model, complete the layout of wind load measurement points after segmentation based on the parametric modeling platform, and complete the surface partitioning based on Thiessen polygons;

[0010] A CFD numerical wind tunnel calculation model is established based on the arrangement of wind load measurement points and surface partitioning after segmentation. Based on the CFD numerical wind tunnel calculation model, CFD full wind direction angle calculation is performed to obtain CFD numerical wind tunnel results.

[0011] S2. Extract the wind pressure at the building surface measurement points corresponding to all wind direction angles from the CFD numerical wind tunnel results, and convert the wind pressure at the building surface measurement points into local shape coefficients;

[0012] S3. Combine the local body coefficient with the three-dimensional building model to establish a measurement point data matrix that includes the body coefficient of the measurement points and the spatial geometry of the measurement points for multiple wind direction angles. Based on the wind pressure results at a certain wind direction angle, different weights are assigned to the spatial coordinates, curvature, normal vector, and wind pressure in the measurement point data matrix of the target building, and the weighted Euclidean distance between the two measurement points is calculated using the assigned weights. Machine learning clustering analysis is performed on the weighted Euclidean distance of the measurement points. After clustering at a single wind direction angle, a sparse measurement point matrix is ​​constructed based on the principle that the wind pressure extreme value area is dense and the non-extreme value area is sparse. The sparse measurement point matrix for all wind direction angles is obtained by integrating the wind direction angles, forming an optimized measurement point layout plan that fully reflects the characteristics of extreme wind pressure.

[0013] Update surface partitions based on optimized measurement point layout and Thiessen polygons, and input them into a physical wind tunnel to conduct physical wind tunnel tests;

[0014] S4. After completing the physical wind tunnel test, output the measurement point shape coefficient results. Based on the wind pressure at the building surface measurement points described in step S2, integrate the wind load analysis results of the physical wind tunnel and the numerical wind tunnel, mark the areas where wind pressure exceeds the specification in accordance with the design specifications, and calculate and output the overall shape coefficient by combining the surface partition area and the local shape coefficient.

[0015] In the above technical solution, the method for repairing and adjusting the model after importing the model in step S1 is specifically as follows:

[0016] Based on the 3D model file imported into NURBS surface modeling software, combined with CAD drawings and project text data, in order to solve the problems of conversion damage, model damage, defects and incompleteness when importing the model, the surface modeling software tool set is used to perform "trim", "merge", "smooth" and other repair operations to ensure that the 3D model is consistent with the project data.

[0017] In the above technical solution, in step S1, the "curvature analysis" and "surface segmentation" tools of the surface modeling software are used to determine the curvature change of the model, and the segmentation strategy is selected according to the degree of curvature change. The fine segmentation and surface adjustment of complex geometric surfaces are completed through the plug-ins provided by the modeling software and the parametric modeling platform.

[0018] In the above technical solution, the method for completing the arrangement of wind load measurement points and surface partitioning in step S1 is specifically as follows:

[0019] On the parametric modeling platform, the number and spacing of control measuring points are set in each surface, and the automatic arrangement of parametric measuring points is completed in combination with wind loads; partitioning is performed according to the geometric points and the geometric model after segmentation adjustment, so that one measuring point corresponds to one partition model, where the partition model is a geometric model obtained by segmenting the original model. When performing geometric partitioning, a three-dimensional Thiessen polygon crystal is first generated on the measuring point, and then the geometric model is segmented using the crystal to obtain the partition model corresponding to the measuring point.

[0020] In the above technical solution, the construction of the Thiessen polygon crystal in step S1 is mainly based on the following steps:

[0021] First, prepare a set of discretely distributed data points, and construct a Delaunay triangulation based on the discrete data points. On the basis of the Delaunay triangulation, draw the perpendicular bisectors of each triangle edge. These perpendicular bisectors divide the screen area into a series of polygons.

[0022] Among them, the Delaunay triangulation is a special triangulation, which has the empty circumcircle property and the maximized minimum angle property, and is the basis for constructing Thiessen polygons.

[0023] In the above technical solution, the method for establishing the CFD numerical wind tunnel calculation model in step S1 is specifically as follows:

[0024] The characteristic dimensions of the target building are analyzed based on the three-dimensional model compiled using NURBS surface modeling software and a parametric modeling platform. The target building site height is used as the basic height. The calculation domain height of the CFD numerical wind tunnel calculation model is determined by adjusting the distance between the target building and the urban building to a corresponding multiple of the basic height. The calculation domain of the CFD numerical wind tunnel is arranged according to the characteristic length of the building and the dimensionless value recommended by the design specification.

[0025] In the above technical solution, in step S1, the pre-processing process of the CFD numerical wind tunnel calculation model is completed according to the climatic conditions of the project location and the physical orientation information provided by the project data.

[0026] In the above technical solution, after completing the pre-processing process in step S1, the CFD numerical wind tunnel calculation model can be used to perform full wind angle calculations on the user's local server or uploaded to a cloud server heterogeneous system for full wind angle calculations as needed. It should be noted that the CFD numerical wind tunnel calculation model is established based on the three-dimensional model and building dimensions provided by the project designer. The model includes detailed architectural drawing information and is a three-dimensional representation of the flat drawings. It is widely used in fields such as building wind resistance design and wind resilience assessment.

[0027] In the above technical solution, the local shape coefficient in step S2 is the ratio of the average pressure or suction caused by the wind acting on a certain or set area of ​​the building surface to the velocity pressure of the incoming wind.

[0028] In the above technical solution, in step S3, the local body coefficients are downloaded and stored in the NURBS surface modeling software and the parametric modeling platform.

[0029] In the above technical solution, the measuring point data in step S3 includes the spatial coordinates of the measuring point, curvature information, the normal vector of the building surface where the measuring point is located, and the wind pressure coefficient and local shape coefficient of the measuring point at multiple wind direction angles.

[0030] In the above technical solution, the specific method for optimizing the measurement point layout using machine learning cluster analysis in step S3 is as follows: First, preprocess the discrete measurement point data under all wind angles. This includes preprocessing the spatial coordinates (X, Y, Z) of the measurement points, curvature information (Curvature), the normal vector of the building surface where the measurement point is located, and the wind pressure coefficient (Pressure Coefficient) at multiple wind directions. This step includes removing outliers, filling in missing values, and normalizing the data to ensure data accuracy and consistency, laying a solid foundation for subsequent analysis.

[0031] In the above technical solution, in step S3, the weighted Euclidean distance between the two measuring points is:

[0032]

[0033] Where d ij is the weighted Euclidean distance between the two measuring points, w1, w2, w3 and w4 are the weight coefficients corresponding to the spatial coordinates, curvature, wind pressure and normal vector respectively, x is with x js are the spatial coordinate components of the two measuring points; s = 1, 2, 3 correspond to the X direction, Y direction and Z direction in space, i and j represent two different measuring points, p ik With p jk is the wind pressure information of the measuring point under the kth wind direction angle; i and j represent two different measuring points; n im With n jmis the normal vector component of the measuring point; m=1, 2, 3 corresponds to the X direction, Y direction and Z direction in space, and i and j represent two different measuring points.

[0034] In the present invention, the weight coefficients w1, w2, w3 and w4 corresponding to the spatial coordinates, curvature, wind pressure and normal vector are preferably 0.5, 0.7, 1.0 and 0.9 respectively.

[0035] In the above technical solution, in step S3, the cluster center is selected, and the specific content is:

[0036] 1) Randomly select a measurement point from the data set as the first cluster center.

[0037] 2) For each measured point, calculate its weighted Euclidean distance to the selected cluster center, and calculate the probability of each data point being selected as the next cluster center based on these distances. Specifically, the farther a data point is from the selected cluster center, the greater its probability of being selected.

[0038] 3) Randomly select a data point as the next cluster center based on the calculated probability distribution, and repeat the cluster center selection process until K cluster centers are selected. Execute the machine learning clustering algorithm based on the selected K initial cluster centers.

[0039] In the above technical solution, preferably K=40 in step S3.

[0040] In the above technical solution, the specific steps of the clustering algorithm in step S3 are:

[0041] 1) For each measurement point in the data set and the obtained K cluster centers or cluster centers, use the weighted Euclidean distance to calculate its distance to each cluster center point and assign it to the cluster with the closest distance;

[0042] 2) Recalculate the center point of each cluster. The new center point is the weighted average of all measurement points in the cluster in the weighted feature space. Specifically, for each feature dimension (spatial coordinates, curvature, wind pressure and normal vector), the data points are weighted and summed according to the corresponding weight, and then divided by the sum of the weights to obtain the weighted average of that dimension. The weighted averages of all dimensions are combined to form the weighted average vector of the cluster center.

[0043] 3) Repeat the steps of assigning measurement points to clusters and updating cluster centers until the cluster centers no longer change or the preset number of iterations is reached.

[0044] After completing the clustering algorithm, the resulting K clusters contain a corresponding number of measurement points, each with similar weighted characteristics (spatial coordinates, curvature, wind pressure, and normal vector). After clustering at a single wind direction angle, a sparse measurement point matrix is ​​constructed based on the principle that wind pressure extremes are dense and non-extremes are sparse. This matrix is ​​then integrated across wind directions to create a sparse measurement point matrix for all wind directions, fully reflecting the characteristics of extreme wind pressures.

[0045] In the above technical solution, in step S3, the retention ratio of the number of measuring points in each cluster under a certain wind direction angle is defined as R, the mean and variance of the wind pressure at each cluster measuring point are analyzed, the extreme wind pressure is determined in combination with the numerical wind tunnel results, and the mean wind pressure of the measuring points in the cluster is analyzed to see whether it meets the following conditions:

[0046] α·p min <p mean <β· pmax ;

[0047] Where p min With p max is the minimum and maximum wind pressure, p mean is the mean wind pressure of the measuring points within the cluster, α and β are the scaling ratios of the minimum and maximum wind pressures, and their values ​​can be adjusted according to actual cases and design requirements;

[0048] When the wind pressure at the measuring points in a cluster satisfies the above formula, it can be considered that the wind pressure at the measuring points in this cluster is far away from the extreme wind pressure. Therefore, the number of measuring points can be appropriately reduced and a smaller measurement point retention ratio R can be set. Otherwise, the wind pressure at the measuring points in this cluster can be considered close to the extreme wind pressure, and a larger measurement point retention ratio R can be set. On this basis, the value of R can be appropriately increased for clusters with larger wind pressure variance. The number of measuring points for each cluster is modified and optimized based on the value of R.

[0049] In this way, a sparse measurement point matrix under a certain wind direction angle is obtained, which accurately retains the extreme wind pressure characteristics and wind pressure gradient characteristics under this wind direction angle. By integrating the results of all wind direction angles, a sparse measurement point matrix of all wind direction angles covering the extreme wind pressure characteristics can be obtained.

[0050] In the above technical solution, in step S3, the modified measurement point distribution is used to further optimize the surface partitioning using the Thiessen polygon principle. This ensures that the measurement points are evenly distributed throughout the monitoring area and fully cover all important wind pressure variation areas, achieving accurate and efficient wind pressure monitoring. Physical wind tunnel testing is conducted based on the optimized measurement point scheme and surface partitioning information.

[0051] In the above technical solution, in step S4, the measuring point body coefficients are optimized based on the output of physical wind tunnel testing, and the wind pressure information on the building surface at all wind angles is derived based on numerical wind tunnel testing. In conjunction with design specifications, the measuring point body coefficients and building surface wind pressure information are compared with the specification limits. The measuring points and areas where wind pressure exceeds the specification are accurately marked, and the information about the measuring point body coefficients and areas with wind pressure exceeding the specification is imported into the 3D building model.

[0052] In the present invention, the design specification may be a domestic specification, or a specification of a certain industry or a specific region.

[0053] In the above technical solution, in step S4, the overall shape coefficient is calculated by combining the surface partition area and the measuring point shape coefficient information through the surface partition weighted average analysis method, thereby obtaining the overall shape coefficient on each surface.

[0054] Specifically, each measuring point P i Has its normal vector (n x , n y , n z ), calculated wind pressure coefficient C of the measuring point pi , local body coefficient μ si The corresponding surface partition area A i , which is obtained by updating the Thiessen polygon in step S3. For each independent surface, if the shape coefficient of the entire surface is averaged, then the overall shape coefficient can be obtained by weighting the partition area as follows:

[0055]

[0056] In the above technical solution, in step S4, taking a high-rise building as an example, for a building interface within a certain height range H, if the density distribution of measuring points is reasonable and the wind pressure magnitude and direction of the measuring points on the partitioned area are unchanged, then the overall shape coefficient of the building in the horizontal and vertical directions within the height range is:

[0057]

[0058] Where μ sa 、μ sc is the overall shape coefficient of the horizontal longitudinal and transverse directions in the height range, μ z is the wind pressure height coefficient corresponding to the center height of the height interval, μ zr is the wind pressure height coefficient corresponding to the building reference point height, L a With L c is the reference length in the longitudinal and transverse directions, and the width of the windward side of the high-rise building in the longitudinal and transverse directions within the height range is taken. i is the angle between the normal direction of the measuring point and the horizontal longitudinal direction, and the normal vector (n x , n y , n z ) is calculated with the specified horizontal direction.

[0059] The overall shape coefficient calculated in this way is a discrete function distributed along the height, which can be applied as the overall shape coefficient of different floors in actual projects.

[0060] In the above technical solution, step S4 organizes and outputs wind pressure exceeding specification areas, measurement point shape coefficients, and the overall building shape coefficient. The overall shape coefficient, combined with the wind pressure cloud map information on the building surface from the numerical wind tunnel, ensures the rigor, accuracy, and validity of wind load information. This provides data support and optimization suggestions for structural wind resistance design.

[0061] The present application also provides a computer-readable storage medium or electronic device having a computer program stored thereon, wherein the computer program implements the steps of any of the above methods when executed by a processor.

[0062] Specifically include: 1) Importing complex building structure models, completing model repair and adjustment in NURBS surface modeling software based on project design data, analyzing its geometric modeling features, completing the layout of wind load measurement points after segmentation based on the parametric modeling platform, and completing surface partitioning based on Thiessen polygons. Establishing a CFD numerical wind tunnel calculation model, and sending the configured model to a local or cloud-based heterogeneous system for CFD full wind direction angle calculation; 2) Extracting the wind pressure at the measurement point from the calculated CFD numerical wind tunnel results, converting it into indicators of wind-resistant design such as surface wind pressure and shape coefficient that are consistent with the specifications, and downloading and storing it on the parametric modeling platform; 3) Optimizing the measurement point layout by combining the measurement point shape coefficient with the three-dimensional building model, performing machine learning clustering analysis on the measurement point shape coefficient, and concentrating the measurement points on areas where the wind pressure is closer based on the clustering results. In areas with extreme wind pressure, physical wind tunnel tests are carried out based on the optimized measurement point layout plan and the updated surface partitions of the Thiessen polygons; 4) after completing the physical wind tunnel test, the measurement point body coefficient results are output, and the building surface wind pressure is derived based on the CFD numerical wind tunnel results in step S2. The wind load analysis results of the physical wind tunnel and the numerical wind tunnel are integrated, and the areas where the wind pressure exceeds the specification are marked in accordance with domestic specifications. The overall body coefficient is calculated and output in combination with the surface partition area and the measurement point body coefficient to supplement the specification value, thereby providing data support and design suggestions for the wind resistance design of the structure.

[0063] Compared with the existing wind pressure measurement point arrangement technology, the beneficial effects of the present invention are:

[0064] (1) Significantly improve test efficiency and data accuracy. The general computational fluid dynamics (CFD) measurement point layout scheme covers a large number of measurement points, which cannot solve the problem of the limited number of measurement points in the physical wind tunnel layout. As a result, the physical wind tunnel cannot adopt the CFD analysis scheme.

[0065] By combining CFD numerical simulation with machine learning cluster analysis, the present invention can accurately predict and evaluate the wind pressure distribution on a building's surface before physical wind tunnel testing. This predictive capability makes the measurement point layout more scientific and reasonable, avoiding the blindness and redundancy of measurement point layout in traditional methods. Therefore, in physical wind tunnel testing, the number of unnecessary measurement points can be significantly reduced, shortening the time for data collection and processing, thereby improving test efficiency. At the same time, due to the more precise measurement point layout, the wind pressure variation characteristics on the building surface can be more comprehensively captured, improving the accuracy and reliability of the test data.

[0066] (2) Reduce test costs. Traditional physical wind tunnel tests are often conservative in the arrangement of measurement points, resulting in an excessive number of measurement points, which increases the test equipment and labor costs. The present invention reduces the number of unnecessary measurement points and reduces the workload of data collection and processing by optimizing the measurement point arrangement. This not only reduces the investment in test equipment, but also reduces labor costs and time costs. In addition, due to the improvement in test efficiency, energy consumption and other related costs during the test process are also indirectly reduced. Therefore, this patent has significant advantages in reducing test costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0068] Figure 1 This is a CFD-AI-based physical wind tunnel pressure test optimization method and storage medium schematic diagram implemented in this application.

[0069] Figure 2 This is a schematic diagram of a three-dimensional building model established in Rhino for the target building selected in an embodiment of the present invention.

[0070] Figure 3 Schematic diagram of the initial arrangement of measurement points established based on the parametric modeling platform Grasshopper in an embodiment of the present invention.

[0071] Figure 4 Schematic diagram of a battery principle for completing surface partitioning based on the Thiessen polygon principle in an embodiment of the present invention.

[0072] Figure 5 This is a schematic diagram of the CFD numerical wind tunnel completed by the pre-processing software urbanFlow for the target building selected in an embodiment of the present invention.

[0073] Figure 6Schematic diagram of a wind direction angle measurement point arrangement scheme after optimization and modification by an adaptive clustering algorithm in an embodiment of the present invention.

[0074] Figure 7 This is a schematic diagram of the measurement point partitioning under a certain unfavorable wind direction angle completed by clustering algorithm optimization and modification and combined with Thiessen polygons in an embodiment of the present invention.

[0075] Figure 8 Schematic diagram of the local shape coefficient of the surface of the target building obtained by numerical wind tunnel calculation under a certain unfavorable wind direction angle in an embodiment of the present invention.

[0076] Figure 9 This is a statistical diagram of the extreme values ​​of the shape coefficient of the measuring points of the target building A considering all wind direction angles in an embodiment of the present invention.

[0077] Figure 10 This is a schematic diagram showing the overall shape coefficient of the target building A under different wind direction angles according to the embodiment of the present invention, where: Figure 10 (a) is the distribution diagram of body shape coefficient in the downwind direction, Figure 10 (b) is the distribution diagram of the crosswind body coefficient. DETAILED DESCRIPTION

[0078] To make the objectives, technical solutions, and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Generally, the components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0079] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present application. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments in the present application without creative work are within the scope of protection of the present application.

[0080] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not require further definition or explanation in subsequent drawings.

[0081] In the description of this application, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, or are the orientations or positional relationships in which the product of this application is typically placed when in use. These terms are intended only to facilitate the description of this application and simplify the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limitations on this application. Furthermore, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0082] Furthermore, terms such as "horizontal," "vertical," and "overhanging" do not necessarily imply that a component must be absolutely horizontal or overhanging, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but rather that it can be slightly tilted.

[0083] It should also be noted that, in the description of this application, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to internal connections between two components. Those skilled in the art will understand the specific meanings of the above terms in this application based on the specific circumstances.

[0084] In this application, unless otherwise expressly specified or limited, a first feature being "above" or "below" a second feature may include the first and second features being in direct contact, or may include the first and second features being in contact not directly but through another feature between them. Moreover, a first feature being "above," "above," and "above" a second feature may include the first feature being directly above or obliquely above the second feature, or may simply mean that the first feature is higher in level than the second feature. A first feature being "below," "below," and "below" a second feature may include the first feature being directly below or obliquely below the second feature, or may simply mean that the first feature is lower in level than the second feature.

[0085] The features and performance of the present application are further described in detail below with reference to the embodiments.

[0086] Example 1

[0087] According to the technical solution of the present invention, a physical wind tunnel pressure test optimization method and storage medium based on CFD-AI are implemented. The principle steps are as follows: Figure 1 As shown, a specific implementation case is given below:

[0088] Taking a numerical simulation project of wind pressure on a super high-rise building as an example, the specific implementation steps of the present invention are introduced as follows:

[0089] Step (1): Based on the CAD drawings and model files provided by the project party, the complex architectural model is imported into the NURBS surface modeling software Rhino. In order to solve the problems such as model damage, incorrect information expression, and excessive model details affecting calculation efficiency during model transmission, the model is repaired and simplified in combination with the CAD drawings. The obtained model diagram is as follows: Figure 2 As shown. Figure 2 The super high-rise building on the middle left is named Building A, and the high-rise building on the right is named Building B.

[0090] Import the organized 3D building model into the parametric modeling software platform Grasshopper, and set each surface of Building A and Building B as an analysis unit Brep. Set the number and spacing of horizontal and vertical measurement points based on the individual surfaces. In this project, the overall spacing is set to about 3m, the horizontal control spacing is maintained at 7 segments, and the vertical spacing is maintained at 30-50 segments according to the surface height. This completes the preliminary setting of the measurement points. Figure 3 As shown. Subsequently, with the help of the SplitGeo battery in the Grasshopper plug-in, the geometric model is partitioned according to the principle of the shortest distance between the measuring points based on the Thiessen polygon principle. The partition process of the case surface is as follows Figure 4 shown.

[0091] Specifically, Figure 4 The partitioning principle is as follows: For each measured point on the surface, a three-dimensional Thiessen polygon crystal is first generated. Any position within the crystal is closest to the measured point within the crystal and farthest from the measured point in the adjacent crystal. This crystal is used to segment the surface, obtaining a partition model corresponding to the measured point.

[0092] Based on the 3D model file provided by the project designer, the NURBS surface modeling software was imported. Combined with the CAD drawings and project documentation provided by the project designer, the surface modeling software's toolset was used to perform repair operations such as "trimming," "merging," and "smoothing" to address issues such as conversion damage, model breakage, and incompleteness that occurred during model import, ensuring that the 3D model matched the project documentation. Furthermore, the surface modeling software's "Curvature Analysis" and "Surface Segmentation" tools provided an intuitive understanding of the model's curvature. Appropriate segmentation strategies were selected based on the degree of curvature change, and the modeling software and the plug-ins provided by the parametric modeling platform were used to complete the fine segmentation and surface adjustments of complex geometric surfaces.

[0093] The method for completing the arrangement of wind load measurement points and surface partitioning is as follows:

[0094] Based on the parametric modeling platform's visual programming capabilities based on nodes and links, measurement point placement rules, such as the number and spacing of control points, were set for each surface. These points were then automatically arranged based on the needs of wind load studies. Subsequent partitioning was performed based on the geometric points and the adjusted geometry model after segmentation. The resulting partition model corresponded to each measurement point. The partition model was a geometric model derived from the original model. Partitioning primarily followed the principle of Thiessen polygons: a 3D Thiessen polygon crystal was generated at the measurement point. This crystal was then used to segment the geometric model, yielding the partition model corresponding to the measurement point.

[0095] Thiessen polygons are an effective spatial partitioning method, primarily used to calculate and analyze specific attribute values ​​within a spatial region based on discretely distributed data points. Each point within a Thiessen polygon is closest to its control point and farther from points within adjacent polygons. This unique spatial partitioning property has led to its widespread application in meteorology, geographic information systems, urban planning, and ecology.

[0096] The definition of Thiessen polygons is briefly as follows:

[0097] Suppose there is a set of discrete points (x i ,y j )(i, j are the number of discrete points), if the area B is divided into k adjacent polygons by a set of straight line ends, such that:

[0098] 1) Each polygon has only one discrete point:

[0099] 2) If any point (x, y) in region B lies inside the polygon containing the discrete point, the following inequality always holds true when i ≠ j:

[0100]

[0101] 3) If a point (x, y) lies on the common edge of two polygons containing discrete points, then the following equation holds: The resulting polygons are called Thiessen polygons.

[0102] The construction of Thiessen polygons is mainly based on the following steps:

[0103] First, prepare a set of discretely distributed data points and construct a Delaunay triangulation based on the discrete data points. The Delaunay triangulation is a special triangulation with the empty circumcircle property and the maximized minimum angle property. It is the basis for constructing Thiessen polygons. Based on the Delauany triangulation, draw the perpendicular bisectors of each triangle edge. These perpendicular bisectors divide the screen area into a series of polygons.

[0104] The specific method for establishing the CFD numerical wind tunnel calculation model is as follows:

[0105] Based on the 3D model compiled using NURBS surface modeling software and a parametric modeling platform, the characteristic dimensions of the target building are analyzed. Specifically, the target building site height is used as the base height. The computational domain height for the CFD numerical wind tunnel model is determined by aligning the distance between the target building and the urban building to a multiple of the base height. The computational domain of the CFD numerical wind tunnel is arranged based on the building's characteristic length and the dimensionless value recommended by the code. The pre-processing of the CFD numerical wind tunnel model is completed based on the project site's climatic conditions and physical orientation information provided by the project documentation. After pre-processing, the CFD numerical wind tunnel model can be used for full wind angle calculations on the user's local server or, if required, uploaded to a cloud server or heterogeneous system for full wind angle calculations. It should be noted that the CFD numerical wind tunnel model is based on the 3D model and building dimensions provided by the project designer. The model includes detailed architectural drawings and is a three-dimensional representation of the plan drawings. It is widely used in building wind resistance design and wind resilience assessment.

[0106] Step (2): The processed three-dimensional building model, measurement points and partitions are stored in the parametric modeling platform Grasshopper, and a numerical wind tunnel is built based on the independently developed building numerical wind tunnel pre-processing software (abbreviated as: urbanFlow). Set Building A and Building B as the target buildings, set the building height of Building A as the basic height H, and use this height as the reference height to set the inlet boundary condition. In combination with the "Standard for Wind Tunnel Test Methods for Building Engineering" (JGJ / T338-2014), the distance from the entrance of the calculation domain to the model is set to 10H, the distance from the model to the exit of the calculation domain is set to 15H, and the distance from the model to both sides of the calculation domain is controlled to 5H to ensure that the geometric model blockage ratio of the building is not greater than 5%. Due to the large calculation area, it is necessary to reasonably partition the entire calculation area, and then generate a grid for each area separately, mainly using a structural grid, and using an unstructured grid for some more complex parts. The grid around the main building group is denser, and the grid away from the main building group is sparser. In the near-wall area and the area where the flow field changes more drastically, the grid is appropriately encrypted to ensure the reliability of the calculation. Under this principle, the grid scheme obtained after grid sensitivity analysis is consistent with the overall calculation domain as shown in the figure. Figure 5 shown.

[0107] It should be noted that this project and the numerical wind tunnel pre-processing software urbanFlow mainly use the open source finite volume method software OpenFOAM for numerical simulation. It is highly customizable and can achieve a high degree of liberalization of parameter setting and secondary development. Therefore, the architectural numerical wind tunnel pre-processing software urbanFlow can be used to automatically establish a parametric numerical wind tunnel according to the OpenFOAM solution file writing format.

[0108] The velocity-inlet boundary condition in OpenFOAM is used at the flow inlet. This boundary condition defines the flow velocity and other scalar flow variables at the flow inlet. This boundary condition is applicable to incompressible flows. The calculations are performed using atmospheric boundary conditions corresponding to representative surface roughness categories as the incoming flow conditions, simulating an exponential distribution of wind velocity profiles in the atmospheric boundary layer, where the velocity varies with height according to an exponential law:

[0109] U z =U0(Z / Z0) α ;

[0110] Where Z0,U0 is the basic wind speed at the height of 10m and 10m, and the ground roughness coefficient is α.

[0111] The turbulence intensity at the inlet varies with height as follows:

[0112]

[0113] Where I 10 is the nominal turbulence constant at 10m height. This project is located in Class A landform, α is 0.12, and the nominal turbulence constant is I 10 = 0.12. The distribution of velocity and turbulence intensity was implemented in the boundary conditions based on the automatic OpenFOAM text file writing function of urbanFlow.

[0114] The outlet uses a fully developed outflow boundary condition (Outflow). The Outflow boundary condition is used when the pressure or velocity at the outflow boundary is unknown. It applies when the flow at the outlet is fully developed. Fully developed means that the flow on the outflow surface is extrapolated from the interior of the region and has no effect on the upstream flow.

[0115] During modeling of this project, the outlet boundary was located behind the building and was much larger than 10 times the height of the building, ensuring that the flow was fully developed. Therefore, the fully developed outflow boundary condition could be used. Symmetric boundary conditions (symmetry) were used at the top and both sides of the computational domain. These are applicable to cases where the flow is symmetrical, i.e., they have mirror symmetry characteristics and can be used to describe free-slip walls in viscous flows. No-slip wall conditions (wall) were used on the building surface and the ground. The wall is a boundary condition used to define fluid and solid areas. The velocity of the no-slip wall is zero, and the velocity of the fluid at the wall is zero. Set the total calculation time and the time step for exchanging the coupled areas, and determine the range of iterations and convergence residuals for different types of areas within the exchange time step. In this case, the calculation step was 0.1s, and the total calculation time was 500s, i.e., the converged solution of the steady-state calculation within 5000 calculation time steps.

[0116] After completing the numerical wind tunnel CFD simulation pre-processing setup in urbanFlow, the building model can be rotated to configure operating conditions for different wind directions. The minimum rotation angle step size is determined based on the project's dominant wind direction type and the meteorological data wind direction time series. In this project, the dominant wind direction in summer is southerly (S), the dominant wind direction in winter is northeast-east-east (ENE), and the dominant wind direction year-round is southeast-east-east (ESE), with a wind direction variation of 22.5°. Therefore, the easterly direction (E) is set to 0°, and the minimum rotation angle step size is 22.5°. This completes the setup for 16 operating conditions, ranging from 0° to 360°. Based on the designated measurement points, sampling points are set at spatial coordinate locations within the computational domain to effectively monitor the time series of field variables, such as wind speed and wind pressure, over the entire computational time. The output data intervals of the sampling points can be adjusted independently according to computational needs. After completing all the above pre-processing setup, OpenFOAM numerical wind tunnel simulations can be performed for all wind directions, either locally or in a heterogeneous cloud-based system.

[0117] It should be noted that in conventional calculations, each case file for the open-source OpenFOAM solver must contain three folders: 0, system, and constant. These folders include the initial and boundary conditions of the computational domain (folder 0), meshing files (blockMeshDict, snappyHexMeshDict, and surfaceFeatureExtractDict), and solution control files (controlDict, fvSchemes, and fvSolution). UrbanFlow, the architectural numerical wind tunnel pre-processing software, developed in C#, can automatically compile OpenFOAM text files within the parametric modeling platform Grasshopper, thereby completing the automated construction of CFD numerical wind tunnels.

[0118] Step (3): Download the wind pressure at the measuring point and convert it into local body coefficient data, and import the results into the NURBS surface modeling software Rhino and the parametric modeling platform Grasshopper. According to actual needs, based on the wind pressure results at a certain wind direction angle, different weights are assigned to the spatial coordinates, curvature, normal vector and wind pressure of the measuring points A and B, and the assigned weights are used to calculate the P of the two points. i With P j The weighted Euclidean distance between them is as follows:

[0119]

[0120] Where d ijis the weighted Euclidean distance between the two measuring points, w1, w2, w3 and w4 are the weight coefficients corresponding to the spatial coordinates, curvature, wind pressure and normal vector, which are 0.5, 0.7, 1.0 and 0.9 respectively in this case. is with x js are the spatial coordinate components of the two measuring points (s = 1, 2, 3 correspond to the X direction, Y direction and Z direction in space, i and j represent two different measuring points), p ik With p jk is the wind pressure information of the measuring point at the kth wind direction angle (i and j represent two different measuring points), n im With n jm is the normal vector component of the measuring point (m=1, 2, 3 corresponds to the X direction, Y direction and Z direction in space, i and j represent two different measuring points);

[0121] Then the cluster center is selected, the specific contents are as follows:

[0122] 1) Randomly select a measurement point from the data set as the first cluster center.

[0123] 2) For each measured point, calculate its weighted Euclidean distance to the selected cluster center, and calculate the probability of each data point being selected as the next cluster center based on these distances. Specifically, the farther a data point is from the selected cluster center, the greater its probability of being selected.

[0124] 3) Randomly select a data point as the next cluster center based on the calculated probability distribution, and repeat the cluster center selection process until K cluster centers are selected. Execute the machine learning clustering algorithm based on the selected K initial cluster centers, in this case K = 40.

[0125] The specific steps of the clustering algorithm are:

[0126] 1) For each measurement point in the data set and the 40 obtained cluster centers (cluster centers), use the weighted Euclidean distance to calculate its distance to each cluster center point and assign it to the cluster with the closest distance;

[0127] 2) Recalculate the center point of each cluster. The new center point is the weighted average of all measurement points in the cluster in the weighted feature space. Specifically, for each feature dimension (spatial coordinates, curvature, wind pressure and normal vector), the data points are weighted and summed according to the corresponding weight, and then divided by the sum of the weights to obtain the weighted average of that dimension. The weighted averages of all dimensions are combined to form the weighted average vector of the cluster center.

[0128] 3) Repeat the steps of assigning measurement points to clusters and updating cluster centers until the cluster centers no longer change or the preset number of iterations is reached.

[0129] After completing the clustering algorithm, the 40 clusters obtained contain a corresponding number of measurement points. Each cluster has similar weighted features (spatial coordinates, curvature, wind pressure and normal vector). After clustering the single wind direction angle, a sparse measurement point matrix is ​​constructed based on the principle of dense wind pressure extreme value areas and sparse non-extreme value areas. The sparse measurement point matrix of all wind direction angles is obtained by integrating the wind direction angles to fully reflect the extreme wind pressure characteristics.

[0130] Specifically, the retention ratio of the number of measuring points in each cluster under a certain wind direction angle is defined as R. The mean and variance of the wind pressure at each cluster measuring point are analyzed. The extreme wind pressure is determined by combining the numerical wind tunnel results. It is also analyzed whether the mean wind pressure of the measuring points in the cluster satisfies the following conditions:

[0131] α·p min <p mean <β·p max ;

[0132] Where p min With p max is the minimum and maximum wind pressure, where the minimum wind pressure is the maximum negative pressure, p mean is the mean wind pressure of the measuring points within the cluster, α and β are the scaling ratios of the minimum and maximum wind pressures, and are uniformly taken as 0.9 in this case.

[0133] When the wind pressure at a cluster's measuring points satisfies the above equation, it can be considered that the wind pressure at that cluster's measuring points is far from the extreme wind pressure. Therefore, the number of measuring points can be appropriately reduced, and a smaller retention ratio R can be set. Otherwise, the wind pressure at that cluster's measuring points is considered close to the extreme wind pressure, and a larger retention ratio R can be set. Furthermore, the value of R can be appropriately increased for clusters with large wind pressure variance. The number of measuring points for each cluster is optimized based on the value of R. In this case, R is uniformly set to 0.1 for clusters that meet the above equation, and 0.5 for clusters that do not. The value of R is appropriately increased for clusters with large variance.

[0134] This gives the simplified measurement points under a certain wind direction angle: Figure 6 As shown in the figure, this measurement point scheme accurately retains the extreme wind pressure characteristics and wind pressure gradient characteristics under the wind direction angle, and the optimized measurement point layout scheme can be obtained by integrating the simplified measurement points under all wind direction angles. Figure 3 The initial measurement point plan reduces the number of measurement points by about 45%.

[0135] Based on the modified measurement point distribution, the Thiessen polygon principle is used to further optimize the surface partitioning. Figure 7 As shown, Figure 7 The measurement points used for surface partitioning are Figure 6 The measurement points are streamlined to ensure that they are evenly distributed throughout the monitoring area and fully cover all important wind pressure variation areas, achieving accurate and efficient wind pressure monitoring. Physical wind tunnel tests are carried out based on the optimized measurement point plan and surface partition information.

[0136] (4) After completing the physical wind tunnel test, the local body coefficient results of the measuring points on the physical wind tunnel building surface and the wind pressure information on the numerical wind tunnel building surface are output. The local body coefficient results of the measuring points can be output as a text file containing the spatial coordinate information of the measuring points and the body coefficient value, or can be imported into the NURBS surface modeling software Rhino, and the measuring point information is saved in the form of point cloud and color map combined with the three-dimensional building model. The color of the measuring point represents the body coefficient value, and the measuring points with wind pressure exceeding the specification are marked in combination with the existing specifications. On this basis, all direction angles can be comprehensively counted to obtain the extreme values ​​of the local body coefficient of the measuring points under all wind direction angles. The surface wind pressure results at a certain wind direction angle and the local body coefficient mechanism results of the measuring point A at all wind direction angles are as follows: Figure 8 and Figure 9 As shown, Figure 8 Areas where negative pressure exceeds the standard are clearly marked. Figure 9 This is a statistical diagram of the extreme values ​​of the shape coefficient of the measuring points of the target building A considering all wind direction angles in an embodiment of the present invention.

[0137] The overall building shape coefficient can be calculated by combining the local shape coefficient results with the surface partitions corresponding to the measuring points. The surface partitions determined based on the Thiessen polygon principle can be used to calculate the area of ​​each partition, thereby obtaining the corresponding relationship between the measuring points and their partition areas. Combined with the shape coefficients of the measuring points, the weighted average method of the surface partition areas is used to calculate the overall shape coefficients of different surfaces, and the distribution diagram of the floor shape coefficients is compiled according to actual needs, as shown below. Figure 10 This result, combined with the wind pressure cloud map information on the building surface from the numerical wind tunnel, can ensure the rigor, accuracy and effectiveness of the wind load information.

[0138] Specifically, each measuring point P i Has its normal vector (n x , n y , n z ), calculated wind pressure coefficient C of the measuring point pi , local body coefficient μ si The corresponding surface partition area A i , which is obtained by updating the Thiessen polygon in step S3. For each independent surface, if the shape coefficient of the entire surface is averaged, then the overall shape coefficient can be obtained by weighting the partition area as follows:

[0139]

[0140] The above example requires further analysis of the overall building shape coefficients of different floors based on demand. Here is a brief explanation:

[0141] For a building interface within a certain height range H, if the density of the measuring points is reasonably distributed, combined with the Thiessen polygon partitioning, it can be assumed that the wind pressure magnitude and direction of the measuring points remain unchanged across the partitioned area. In this example, the height range is represented by the height range of different floors. The overall shape coefficient of the building in the horizontal and vertical directions within the range is:

[0142]

[0143] Where μ sa 、μ sc is the overall shape coefficient of the horizontal longitudinal and transverse directions in the height range, μ z is the wind pressure height coefficient corresponding to the center height of the height interval, μ zr is the wind pressure height coefficient corresponding to the building reference point height, L a With L c The longitudinal and transverse reference lengths can be taken as the longitudinal and transverse windward widths of high-rise buildings within this height range. i is the angle between the normal direction of the measuring point and the horizontal longitudinal direction, and the normal vector (n x , n y , n z ) is calculated with the specified horizontal direction.

[0144] The overall shape coefficient calculated in this way is a discrete function distributed along the height. In this example, it is the overall shape coefficient of different floors, such as Figure 10 shown.

[0145] The data is collated and output, including wind pressure cloud images of the building surface, local shape coefficient measurement points, and the overall building shape coefficient under all wind directions, providing important data support and optimization suggestions for structural wind resistance design. By analyzing and comparing data at different wind directions, the stress characteristics and patterns of the building under different wind loads can be discovered, providing a more scientific and digital basis for structural design and optimization.

[0146] Example 2

[0147] The present invention also provides a computer-readable storage medium or electronic device having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the physical wind tunnel pressure test optimization method based on CFD-AI as described in any one of the above items are implemented.

[0148] The embodiments described above are part of the embodiments of the present application, rather than all of the embodiments. The detailed description of the embodiments of the present application is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

Claims

1. A physical wind tunnel pressure test optimization method based on CFD-AI, characterized by The following steps are involved: S1. Import the 3D building model into NURBS surface modeling software, complete the building model repair and adjustment, complete the wind load measurement point layout after segmentation based on the parametric modeling platform, and complete the surface partitioning based on the Thiessen polygons; establish a CFD numerical wind tunnel calculation model based on the wind load measurement point layout and surface partitioning, and perform CFD full wind direction angle calculation based on the CFD numerical wind tunnel calculation model to obtain the CFD numerical wind tunnel results; S2. Extract the wind pressure at the building surface measurement points corresponding to all wind direction angles from the CFD numerical wind tunnel results, and convert the wind pressure at the building surface measurement points into local shape coefficients; S3. Combine the local shape coefficient with the three-dimensional building model to establish a measurement point data matrix that includes the shape coefficients and spatial geometry of the measurement points for multiple wind direction conditions. Based on the wind pressure results at a certain wind direction, assign different weights to the spatial coordinates, curvature, normal vector, and wind pressure in the measurement point data matrix of the target building, and calculate the weighted Euclidean distance between the two measurement points. Perform machine learning cluster analysis on the weighted Euclidean distance of the measurement points to construct a sparse measurement point matrix and integrate it to obtain a sparse measurement point matrix for all wind direction angles, forming an optimized measurement point layout scheme that comprehensively reflects the characteristics of extreme wind pressure. Update the surface partition based on the optimized measurement point layout scheme and Thiessen polygons, and conduct physical wind tunnel tests. S4. After completing the physical wind tunnel test, output the measurement point shape coefficient results. Based on the wind pressure at the building surface measurement points described in step S2, integrate the wind load analysis results of the physical wind tunnel and the numerical wind tunnel, mark the areas where the wind pressure exceeds the specification, and calculate and output the overall shape coefficient by combining the surface partition area and the local shape coefficient.

2. The physical wind tunnel pressure test optimization method based on CFD-AI according to claim 1 is characterized in that Based on the 3D architectural model file imported into NURBS surface modeling software, combined with CAD drawings and project text materials, the tool set of the surface modeling software is used to perform "trim", "merge" and "smooth" repair operations to address conversion damage, model damage defects and incompleteness that occur when importing the model, ensuring that the 3D model is consistent with the project data.

3. The physical wind tunnel pressure test optimization method based on CFD-AI according to claim 1 is characterized in that In step S1, the curvature change of the model is determined by using the "Curvature Analysis" and "Surface Segmentation" tools of the surface modeling software. A segmentation strategy is selected based on the degree of curvature change. The fine segmentation and surface adjustment of complex geometric surfaces are completed using the plug-ins provided by the modeling software and the parametric modeling platform.

4. The physical wind tunnel pressure test optimization method based on CFD-AI according to claim 1 is characterized in that: The method for completing the arrangement of wind load measurement points and surface partitioning in step S1 is specifically as follows: on the parametric modeling platform, the number and spacing of control measurement points are set in each surface, and the automatic arrangement of parametric measurement points is completed in combination with the wind load; partitioning is performed according to the geometric points and the geometric model after segmentation adjustment, so that one measurement point corresponds to one partition model, wherein the partition model is a geometric model obtained by segmenting the original model. When performing geometric partitioning, a three-dimensional Thiessen polygon crystal is first generated on the measurement point, and then the crystal is used to segment the geometric model to obtain a partition model corresponding to the measurement point.

5. The CFD-AI-based physical wind tunnel pressure test optimization method according to claim 1, characterized in that: The construction of the Thiessen polygon crystal in step S1 is mainly based on the following steps: first, prepare a set of discretely distributed data points, construct a Delaunay triangulation based on the discrete data points, and on the basis of the Delaunay triangulation, draw the perpendicular bisectors of each triangle edge. These perpendicular bisectors divide the screen area into a series of polygons.

6. The CFD-AI-based physical wind tunnel pressure test optimization method according to claim 1, characterized in that: The local shape coefficient in step S2 is the ratio of the average pressure or suction caused by the wind acting on a set area of ​​the building surface to the velocity pressure of the incoming wind.

7. The physical wind tunnel pressure test optimization method based on CFD-AI according to claim 1 is characterized in that: The measurement point data in step S3 include the spatial coordinates of the measurement point, curvature information, the normal vector of the building surface where the measurement point is located, and the wind pressure coefficient and local shape coefficient of the measurement point at multiple wind direction angles.

8. The CFD-AI-based physical wind tunnel pressure test optimization method according to claim 1, characterized in that: In step S3, the retention ratio of the number of measuring points in each cluster under a certain wind direction angle is defined as R. The mean and variance of the wind pressure at each cluster measuring point are analyzed. The extreme wind pressure is determined by combining the numerical wind tunnel results. It is also analyzed whether the mean wind pressure of the measuring points in the cluster meets the following conditions: a·p min <p mean <b·p max ; Where p min With p max is the minimum and maximum wind pressure, p mean is the mean wind pressure of the measuring points within the cluster, α and β are the scaling ratios of the minimum and maximum wind pressures; When the wind pressure at the measuring point in the cluster satisfies the above formula, it is considered that the wind pressure at the measuring point in the cluster is far from the extreme wind pressure. The number of measuring points is appropriately deleted and a smaller measurement point retention ratio R is set. Otherwise, the wind pressure at the measuring point in the cluster is considered to be close to the extreme wind pressure. A larger measurement point retention ratio R is set. On this basis, the value of R is appropriately increased for clusters with larger wind pressure variance. The number of measuring points for each cluster is modified and optimized based on the value of R.

9. The physical wind tunnel pressure test optimization method based on CFD-AI according to claim 1, characterized in that: In step S4, the overall shape coefficient is calculated by combining the surface partition area and the measuring point shape coefficient information through the surface partition weighted average analysis method, thereby obtaining the overall shape coefficient on each surface.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the physical wind tunnel pressure test optimization method based on CFD-AI as described in any one of claims 1 to 9 are implemented.

Images