Multi-dimensional Parameter-based Rapid Analysis System for the Performance of Prefabricated Buildings

Abstract

This application relates to the field of data analysis, and provides a rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters. Through the configuration acquisition module, the model input module and the analysis module, this application uses the wind speed field cloud map and multiple local maps corresponding to each prefabricated connector collected at regular intervals, combines with the differential neural network model, and analyzes the stability performance of the prefabricated building by calculating the differences of the local maps of each prefabricated connector caused by the real-time wind field, providing a new evaluation method for the stability performance of prefabricated buildings.

Description

Technical Field

[0001] This application relates to the field of data analysis. More specifically, this application relates to a rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters. Background Art

[0002] A prefabricated building transfers a large amount of on-site operation work in the traditional construction method to the factory. Building components and fittings (such as floor slabs and wall panels, etc.) are processed and manufactured in the factory and transported to the construction site, and then assembled and installed on-site through reliable connection methods. A prefabricated building includes multiple different types of basic components and various corresponding prefabricated connectors. Prefabricated connectors include, for example, reinforced concrete connectors, steel structure connectors, and connectors between precast concrete exterior walls and the main structure. Since most current building structures are high-rise configurations, for high-rise buildings, especially for the prefabricated components at the top of high-rise buildings, the wind force is relatively large. Generally, the top of a high-rise building will vibrate or sway under the action of wind force. To avoid excessive rigidity leading to floor fractures, the vibration amplitude should be within the standard range. If the vibration amplitude is too large, there will be stability problems, which will lead to safety hazards. Therefore, a solution for analyzing the performance of prefabricated buildings for high-rise buildings is needed. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, this application provides a rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters to solve the problems raised in the above background art.

[0004] To achieve the above object, this application provides a rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters. The multi-dimensional parameters include multiple local images corresponding to each prefabricated connector and a wind speed field cloud image. The multiple local images are obtained by fixed-angle cameras at multiple fixed orientations photographing the corresponding prefabricated connectors. The rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters includes:

[0005] An acquisition module that acquires the wind speed field cloud image of the prefabricated building to be analyzed and multiple local images corresponding to each prefabricated connector collected at regular intervals. The multiple local images are obtained by fixed-angle cameras at multiple fixed orientations photographing the corresponding prefabricated connectors;

[0006] A model input module that inputs all the local images collected for each prefabricated connector at each sampling moment into the difference neural network model corresponding to the prefabricated connector and outputs the deviation degree from the standard image at each sampling moment;

[0007] An analysis module that, for the prefabricated connector, analyzes and determines the stability performance of the prefabricated building according to the deviation degree at each sampling moment and the wind speed field cloud image.

[0008] Optionally, the differential neural network model includes a channel attention mechanism network, an image cropping layer, a convolutional network, and a comparison layer;

[0009] The input of the channel attention mechanism network is each local graph, and the output is used as the input of the image cropping layer. The image cropping layer outputs an image after removing the feature channels with weight coefficients lower than a set threshold, and inputs it into the convolutional network. The output of the convolutional network is used as the input of the comparison layer, and the comparison layer is pre-configured with a standard graph corresponding to the pre-constructed connector.

[0010] Optionally, the convolutional network specifically includes: multiple convolutional layers and multiple pooling layers. Among them, during the testing process of the differential neural network model, when multiple test data with a difference degree lower than the set threshold are sequentially input into the differential neural network model, some nodes of any one or more convolutional layers are solidified each time until the output difference of the convolutional network before and after solidification is within the set range. At this time, the solidified nodes are output, and the iterative operation of the above steps is performed until all similar training data have been trained or the set solidification ratio is reached, and the final convolutional network is output.

[0011] Optionally, the rapid analysis system for the performance of prefabricated buildings further includes:

[0012] A model pre-optimization module, and the model pre-optimization module includes:

[0013] A removal unit, randomly removing the nodes of any one convolutional layer in the convolutional network, using nodes with zero weights as replacements to replace the removed nodes, and retesting with the same test data before and after removal. If the difference degree of the test results is within the standard range, the convolutional network is updated;

[0014] An update unit, replacing the initial convolutional network with the updated convolutional network, and training the convolutional network with training data to obtain the final convolutional network.

[0015] Optionally, the rapid analysis system for the performance of prefabricated buildings further includes:

[0016] A model pre-optimization module, and the model pre-optimization module includes:

[0017] A deletion unit, randomly removing the nodes of any one convolutional layer in the convolutional network, using nodes with zero weights as replacements to replace the removed nodes, and retesting with the same test data before and after removal. If the difference degree of the test results is within the standard range, the convolutional network is updated;

[0018] The explosion unit randomly uses at least one node in at least one convolutional layer of the updated convolutional network as a standard node, adds at least one node according to the weight of the standard node, and inserts it into the corresponding convolutional layer; wherein the number of deleted nodes is greater than the number of added nodes.

[0019] Optionally, the model pre-optimization module further includes:

[0020] An image slicing layer for slicing the local graph into multiple subgraphs;

[0021] An image splicing layer. For every three adjacent subgraphs, it inserts the first N columns of the first subgraph before the first column of the middle subgraph, and inserts the last M columns of the last subgraph after the last column of the middle subgraph, regenerates the middle subgraph, and replaces the local graph, and inputs it to the convolutional layer.

[0022] Optionally, the model pre-optimization module further includes:

[0023] An image partitioning layer randomly partitions the local graph into multiple regions, the sizes of each region are the same or different, and the size of the largest region is constrained within a preset range;

[0024] An image deletion layer randomly deletes one or more of the regions, performs zero-padding on the deleted region part to form an updated local graph, and replaces the current local graph and inputs it to the convolutional layer. If the change rate of the deviation corresponding to the same test data is higher than the standard threshold, restore the deleted region and perform random deletion again until the change rate of the deviation corresponding to the same test data is lower than the standard threshold, then retain the current updated local graph as the input to the final convolutional layer.

[0025] Optionally, the analysis module includes:

[0026] An alignment unit, for each pre-constructed connector, superimposes with a preset time offset and sampling moment to calibrate the alignment of the sampling moment with the time coordinate in the wind speed field cloud map;

[0027] A search unit, for the aligned sampling moment, searches for the wind field state data corresponding to the moment from the wind speed field cloud map;

[0028] A determination unit, for the aligned sampling moment, analyzes and determines the stability performance of the prefabricated building in combination with the wind field state data and the corresponding deviation data.

[0029] Optionally, the wind field state data includes the wind speed and wind direction of each unit area in the spatial region, and the determination unit includes:

[0030] The comprehensive wind field determination subunit, for each prefabricated connector, determines the comprehensive wind direction and comprehensive wind force within the spatial region according to the spatial region where the prefabricated connector is located and the wind speed and wind direction of each unit region;

[0031] The sorting subunit determines the wind direction influence ranking of all local maps according to the comprehensive wind direction and the shooting angle of the fixed-angle camera;

[0032] The weight coefficient generation subunit dynamically generates the deviation weight coefficient of each local map according to the wind direction influence ranking;

[0033] The determination subunit analyzes and determines the stability performance of the prefabricated building according to the deviation weight coefficient of each local map and the deviation degree of each local map, combined with the comprehensive wind force at the corresponding time point.

[0034] Optionally, the determination subunit includes:

[0035] According to the comprehensive wind force and the initial standard wind force, re-determine the deviation weight coefficient at the corresponding time point. If the comprehensive wind force is greater than the initial standard wind force, reduce the deviation weight coefficient of the local maps corresponding to the first K1 positions in the wind direction influence ranking, where the reduction amplitude is the ratio of the difference between the comprehensive wind force and the initial standard wind force to the initial standard wind force to the proportion of the original deviation weight coefficient, and increase the deviation weight coefficient of the local maps corresponding to the last K2 positions in the wind direction influence ranking, where the increase amplitude is the ratio of the difference between the comprehensive wind force and the initial standard wind force to the initial standard wind force to the proportion of the original deviation weight coefficient; if the comprehensive wind force is less than the initial standard wind force, increase the deviation weight coefficient of the local maps corresponding to the first K1 positions in the wind direction influence ranking, where the increase amplitude is the ratio of the difference between the comprehensive wind force and the initial standard wind force to the initial standard wind force to the proportion of the original deviation weight coefficient, and reduce the deviation weight coefficient of the local maps corresponding to the last K2 positions in the wind direction influence ranking, where the reduction amplitude is the ratio of the difference between the comprehensive wind force and the initial standard wind force to the initial standard wind force to the proportion of the original deviation weight coefficient;

[0036] Perform a weighted average calculation according to the final deviation weight coefficient of each local map and the deviation degree of each local map, and output the final deviation degree of the prefabricated connector;

[0037] Determine the stability performance of the prefabricated building according to the final deviation degrees of all prefabricated connectors, combined with the positions of each prefabricated connector.

[0038] The technical effects of this application:

[0039] A rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters provided by the present application, through a configuration acquisition module, a model input module and an analysis module, through a wind speed field cloud map and multiple local maps corresponding to each prefabricated connection member collected at regular intervals, combined with a differential neural network model, analyzes the stability performance of the prefabricated building by calculating the differences between the local maps of each prefabricated connection member caused by the real-time wind field, and provides a new evaluation method for the stability performance of prefabricated buildings. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic diagram of the scenario in the embodiment of the present application;

[0041] Figure 2 It is a schematic diagram of the system structure in the embodiment of the present application;

[0042] Figure 3 It is a schematic diagram of the processing of the channel attention mechanism network in the embodiment of the present application;

[0043] Figure 4 It is a schematic diagram of the principle of convolution layer curing in the embodiment of the present application;

[0044] Figure 5 It is a schematic diagram of image stitching in the embodiment of the present application;

[0045] Figure 6 It is a schematic diagram of the electronic device in the embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] In the technical field of the present application, when the technical solutions of the embodiments of the present application are described in combination with the drawings, it will be found that the discussed embodiments only constitute a part rather than all of the entire scope of the embodiments of the present application. Given the diversity of the embodiments of the present application, without creative labor input, other related embodiments that can be obtained by those of ordinary skill in the art based on the existing embodiments are all within the protection scope set by the present application.

[0047] Since the prefabricated components are assembled on site, after the construction is completed, it is necessary to evaluate the stability of the prefabricated building, especially the stability evaluation of the top area of high-rise buildings. At this time, the construction equipment has not been withdrawn, and it needs to be withdrawn after the stability is determined to be qualified. At present, most of the evaluations of stability are obtained by long-term observations to obtain the difference amount, and then the difference amount is compared to determine its stability, which takes a long time, often several months, seriously affecting the progress of the project.

[0048] Based on this, it is found that for prefabricated components, their structures are complex, and most prefabricated components are fixedly connected to multiple other prefabricated components. Even after long-term observation, due to the small difference, there are serious problems with measurement accuracy. This application cleverly conducts a local analysis of the connecting parts of the prefabricated components. By installing fixed-angle cameras on the construction equipment that has not yet been removed or on the building body, and then taking local images of the prefabricated connecting parts at various angles and in all directions through the fixed-angle cameras, and comparing the images at the pixel level, the deviation degree can be amplified, and the stability performance of the building can be measured more accurately, without the need for long-term observation.

[0049] As Figure 1 shown in the schematic diagram of the scenario architecture applicable to this application, it should be noted that the building in this application is a high-rise building 1. Generally speaking, the height of the high-rise building 1 can be greater than 60 meters, and this application does not limit this. In the embodiments of this application, the fixed-angle and fixed-orientation cameras 3 can be installed on the construction equipment 2 such as lifting equipment, tower crane equipment, construction elevators, etc., or on the body of the high-rise building 1, such as the walls, beams or other body structures of the high-rise building 1, and this application does not limit this.

[0050] Please refer to Figure 2 shown. The embodiments of this application provide a rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters.

[0051] It should be noted that the multi-dimensional parameters of this application include multiple local images corresponding to each prefabricated connecting part and a wind speed field cloud map. The local images correspond to image visual parameters, and the wind speed field cloud map corresponds to wind field parameters. The multiple local images are obtained by taking pictures of the corresponding prefabricated connecting parts through multiple fixed-angle and fixed-orientation cameras. The positions and configuration angles of the cameras are shown in Figure 1 and will not be elaborated in this application.

[0052] The following will Figure 2 be used to describe the embodiments of this application in detail. As Figure 2 shown, the rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters includes:

[0053] An acquisition module 10 that acquires the wind speed field cloud map of the prefabricated building to be analyzed and multiple local images corresponding to each prefabricated connecting part collected at regular intervals. The multiple local images are obtained by taking pictures of the corresponding prefabricated connecting parts through multiple fixed-angle and fixed-orientation cameras;

[0054] A model input module 20 that inputs all the local images collected for each prefabricated connecting part at each sampling moment into the difference neural network model corresponding to the prefabricated connecting part, and outputs the deviation degree from the standard image at each sampling moment;

[0055] The analysis module 30 analyzes and determines the stability performance of the prefabricated building for the prefabricated connection member according to the deviation degree at each sampling moment and the wind speed field cloud map.

[0056] In an optional embodiment, the differential neural network model includes a channel attention mechanism network, an image cropping layer, a convolutional network, and a comparison layer;

[0057] The input of the channel attention mechanism network is each local map, and the output is used as the input of the image cropping layer. The image cropping layer outputs an image after removing the feature channels with weight coefficients lower than the set threshold and inputs it to the convolutional network. The output of the convolutional network is used as the input of the comparison layer, and the comparison layer is pre-configured with a standard map corresponding to the prefabricated connection member.

[0058] It should be emphasized that the wind speed field cloud map of the present application can be obtained through public channels or tested by corresponding wind speed devices. The wind speed field cloud map can display the corresponding wind speed magnitude through different colors or different grayscales, and display the corresponding wind direction through the flow direction. Since the wind speed field cloud map belongs to the public materials in this field, no further illustration is shown here.

[0059] It should be noted that the implementation process of the channel attention mechanism network of the present application is as follows:

[0060] (1) Squeeze (Fsq): Through global average pooling, the two-dimensional features (H*W) of each channel are compressed into 1 real number, and the feature map is changed from [h, w, c] ==>[1,1,c].

[0061] (2) excitation (Fex): Generate a weight value for each feature channel, construct the correlation between channels through two fully connected layers, and the number of output weight values is the same as the number of channels of the input feature map, [1,1,c] ==>[1,1,c].

[0062] (3) Scale (Fscale): Weight the normalized weights obtained above to the features of each channel. In the present application, multiplication can be used, multiplying by the weight coefficient channel by channel, [h,w,c]*[1,1,c]==>[h,w,c].

[0063] Next, please combine Figure 3 , and the present application uses the SE attention mechanism in EfficientNet to illustrate this process:

[0064] squeeze operation: The feature map undergoes global average pooling, and the feature map is compressed into a feature vector [1,1,c].

[0065] excitation operation: FC1 layer + Swish activation + FC2 layer + Sigmoid activation. Through the fully connected layer (FC1), the channel dimension of the feature map vector is reduced to 1 / r of the original, i.e., [1, 1, c*1 / r]; then it passes through the Swish activation function; then through a fully connected layer (FC2), the feature map of the feature map vector rises back to the original [1, 1, c]; then it is transformed into a normalized weight vector between 0 and 1 through the sigmoid function.

[0066] scale operation: Multiply the normalized weight and the original input feature map channel by channel to generate a weighted feature map.

[0067] It should be noted that the channel attention mechanism network of this application adopts the existing network structure. The technical contribution of this application to the prior art does not lie in providing a specific network architecture, but lies in cleverly selecting the channel attention mechanism network to cooperate with the image cropping layer to crop the input image, thereby reducing the data input volume of the subsequent input to the convolutional network, and then reducing the data processing volume on the premise of ensuring obvious features, achieving the effect of rapid analysis. At the same time, one of the core concepts of this application is to enhance features, thereby increasing the difference in deviation, thus amplifying the difference in stability, and solving the problem that the current stability analysis requires long-term monitoring.

[0068] Next, the convolutional network of this application will be described in detail.

[0069] In an alternative embodiment, as Figure 4 shown, it should be noted that Figure 4 The number of nodes and layers shown in are only for illustration. The actual number of nodes and layers of the convolutional network is much more than those shown in the figure. This application will not elaborate on this. Specifically, the convolutional network specifically includes: multiple convolutional layers and multiple pooling layers. Among them, during the testing process of the difference neural network model, when multiple test data with a difference degree lower than the set threshold are sequentially input into the difference neural network model, each time a part of the nodes of any one or more convolutional layers are solidified until the output difference of the convolutional network before and after solidification is within the set range. At this time, the solidified nodes are output, and the iterative operation of the above steps is performed until all similar training data have been trained or the set solidification ratio is reached, and the final convolutional network is output.

[0070] In the embodiment of this application, in the testing session of the model, by using similar test data as input and "randomly" solidifying some nodes, it is not necessary to converge faster during model training, and the number of model training times can be reduced.

[0071] Furthermore, the set values such as the set range of the present application can be obtained based on experience or experimental verification. For example, the set range of this embodiment can be controlled within 5%. In this way, on the one hand, the number of model training times can be reduced, and only the uncured nodes are updated during the training process. On the other hand, during the subsequent model prediction process, although the cured nodes are randomly selected, they have passed the difference verification, indicating that the importance of these nodes has been confirmed. This shows that relatively speaking, the importance of the uncured nodes is relatively low. At this time, by curing these important nodes, the weights can be prevented from being damaged, ensuring the accuracy of the data.

[0072] It should be noted that for the local views of the prefabricated connectors collected in the embodiments of the present application, there are almost no differences in their overall color and material. Therefore, it is very difficult to highlight the difference degree or deviation degree, and it is difficult to analyze by conventional means. The present application discovers that when some parts of the prefabricated connectors are offset, due to the fixed shooting angle of the camera, it will inevitably introduce the pixels of other objects. For example, when the prefabricated connector is offset 5 mm to the left, correspondingly, in its local view, the 5-mm area on the right may, on the one hand, be the view of other parts of the prefabricated connector. Although there may be no difference in its color and material, there will be differences in its surface morphology, stripes, etc. On the other hand, it may be other prefabricated components or there are no other objects. At this time, there will be significant differences in color, surface, etc. within the 5-mm area on the right. It can be seen from this that when the prefabricated connectors are in different situations, the importance of the feature dimensions is different. The present application combines the channel attention mechanism network and the convolutional network to configure the curing node method, thereby ensuring the importance of the retained feature dimensions. The important feature dimensions and nodes will not be damaged, thus highlighting the important features and further expanding the difference in the deviation degree.

[0073] Furthermore, in other alternative embodiments, the rapid analysis system for the performance of the prefabricated building further includes:

[0074] A model pre-optimization module, and the model pre-optimization module includes:

[0075] A removal unit, which randomly removes the nodes of any convolutional layer in the convolutional network and replaces the removed nodes with nodes having a weight of zero. Then, the same test data is used to perform tests before and after the removal. If the difference degree of the test results is within the standard range, the convolutional network is updated;

[0076] An update unit, which replaces the initial convolutional network with the updated convolutional network and trains the convolutional network using the training data to obtain the final convolutional network.

[0077] It is found in this application that for the local diagram of the prefabricated connector, even when the prefabricated connector vibrates, shifts, etc., the overall surface difference is small. By virtue of this feature, in combination with the curing nodes mentioned in the foregoing embodiments, important nodes are cured with weights to avoid weight damage. Based on the core concept of the foregoing embodiments, the embodiments of this application can further configure the concept of removing some unimportant nodes, that is, by using the same test data for testing. If the difference degree of the test results is within the standard range, it means that the influence of this node is small. Therefore, this node can be configured with a weight of 0, thereby reducing the number of nodes, lightening the model structure. At the same time, since the number of nodes is reduced (a weight of 0 means that the corresponding node can be removed), the speed of model prediction is improved.

[0078] In addition, in the embodiments of this application, another lightweight solution is further provided. Specifically, in another optional embodiment, the rapid analysis system for the performance of the prefabricated building further includes:

[0079] A model pre-optimization module, and the model pre-optimization module includes:

[0080] A deletion unit, which randomly deletes the nodes in any convolutional layer of the convolutional network and replaces the deleted nodes with nodes with a weight of zero. Then, the same test data is used to test before and after the deletion. If the difference degree of the test results is within the standard range, the convolutional network is updated;

[0081] An explosion unit, which randomly uses at least one node in at least one convolutional layer of the updated convolutional network as a standard node, increases at least one node according to the weight of the standard node, and inserts it into the corresponding convolutional layer; where the number of deleted nodes is greater than the number of added nodes.

[0082] In this embodiment, first, the explosion module randomly selects a node as a standard, and then adds this standard node to the convolutional layer. Since the number of deleted nodes is more than the number of added nodes, the overall shows a lightweight characteristic. Further, since the randomly selected node is not an "unimportant node" in the convolutional network, the weight of important nodes can be further increased. While ensuring lightweight, a greater difference in deviation can be ensured, so that the data results are more easily analyzed for stability.

[0083] In a preferred embodiment, this application can configure the standard node to be selected from the cured nodes in the foregoing embodiments. In this way, the selected standard node is the node with the highest importance, thereby further amplifying the difference in deviation.

[0084] Further, the present application also provides a way to amplify the difference in deviation, and it can be understood that all the ways to amplify the difference in deviation in the present application can be freely combined, and the present application does not limit this. In an optional embodiment, the model pre-optimization module further includes:

[0085] An image slicing layer for slicing the local image into multiple sub-images;

[0086] An image splicing layer. For each adjacent three sub-images, insert the starting N columns of the first sub-image before the first column of the middle sub-image, and insert the last M columns of the last sub-image after the last column of the middle sub-image, regenerate the middle sub-image, and replace the local image, and input it to the convolutional layer.

[0087] As Figure 5 shown, in this embodiment, the cooperation of the image slicing layer and the image splicing layer is equivalent to shrinking the original image to one-third. And through splicing, the N columns and M columns on both sides are incorporated into the middle sub-image. Combining the characteristic that the edge features are more prominent when the pre-constructed connecting piece vibrates or deforms, the present application converges the edge N columns and M columns into the middle sub-image, so that there are more pixel positions where the edge features are located and the edge features are more obvious, thereby further amplifying the difference in deviation.

[0088] In addition, in other optional embodiments, the present application can also further combine the test results obtained from the same test data to delete some unimportant sub-regions. Specifically, the model pre-optimization module further includes:

[0089] An image partitioning layer for randomly partitioning the local image into multiple regions, where the size of each region can be the same or different, and the size of the largest region is constrained within a preset range;

[0090] An image deletion layer for randomly deleting one or more of the regions, performing a zero-padding operation on the deleted region part to form an updated local image, and replacing the current local image and inputting it to the convolutional layer. If the deviation rate of change corresponding to the same test data is higher than the standard threshold, restore the deleted region and perform a random deletion operation again until the deviation rate of change corresponding to the same test data is lower than the standard threshold, then retain the current updated local image as the input to the final convolutional layer.

[0091] In the embodiments of the present application, by combining the concept of the same test data, the deleted sub-region is judged. If the deviation degree decreases, it indicates that the sub-region is relatively important, and the deleted region is restored. Iteration is performed until the change rate of the deviation degree is lower than the standard threshold, for example, lower than 5%. This indicates that the change rate of removing or not removing the sub-region is small, indicating that the relative importance of the sub-region is low. Removing this sub-region can greatly improve the model prediction speed and optimize the model structure.

[0092] In an alternative embodiment, the analysis module includes:

[0093] An alignment unit, for each prefabricated connector, superimposes with a preset time offset and sampling moment to calibrate the aligned sampling moment with the time coordinate in the wind speed field cloud map;

[0094] A search unit, for the aligned sampling moment, searches the wind field state data corresponding to the moment from the wind speed field cloud map;

[0095] A determination unit, for the aligned sampling moment, analyzes and determines the stability performance of the prefabricated building by combining the wind field state data and the corresponding deviation degree data.

[0096] In this embodiment, the present application discovers that there is a certain time delay between the real-time wind field reflected by the wind speed field cloud map and the vibration of the prefabricated connector. Therefore, the present application first calibrates a preset time offset by a calibration method. This time offset is mostly within 2s. By first calibrating this time delay, the error caused by the wind speed data delay is reduced.

[0097] It should be understood that since the deviation degree data of the present application is relatively accurate, the error caused by objective factors can be reduced by configuring the time delay calibration, and the accuracy of the data is further improved.

[0098] The following will detail the analysis and determination of the stability performance of the prefabricated building by specifically combining the wind field state data and the corresponding deviation degree data in the present application.

[0099] In an alternative embodiment, the wind field state data includes the wind speed and wind direction of each unit region in the spatial region, and the determination unit includes:

[0100] A comprehensive wind field determination sub-unit, for each prefabricated connector, determines the comprehensive wind direction and comprehensive wind force in the spatial region according to the spatial region where the prefabricated connector is located and the wind speed and wind direction of each unit region;

[0101] A sorting sub-unit, determines the sorting of the wind direction influence of all local maps according to the comprehensive wind direction and the shooting angle of the fixed-angle camera;

[0102] A weight coefficient generation subunit dynamically generates a deviation weight coefficient for each local map according to the sorted wind direction influence.

[0103] A determination subunit analyzes and determines the stability performance of the prefabricated building based on the deviation weight coefficient of each local map, the deviation of each local map, and the combined wind force at the corresponding time point.

[0104] In this embodiment, the determination of the combined wind direction and the combined wind force can be calculated based on a vector algorithm. Specifically, first, the wind direction and the corresponding wind speed data of each unit area are collected. Then, with each wind direction as the direction and the wind speed as the length, a vector is drawn to represent each individual wind force. For example, the north wind can be represented as a vector along the positive y-axis direction, and the east wind can be represented as a vector along the positive x-axis direction. Specifically, the wind direction and the wind speed can be converted into u and v components, where u is the component in the east-west direction and v is the component in the north-south direction. Then, in a common coordinate system, all the wind force vectors are connected end to end to form a closed polygon. Finally, starting from a vertex of the polygon and walking along the edge of the polygon to the opposite vertex, this vector is the resultant wind vector. Finally, the direction and length of the resultant wind vector are read, which are the wind direction and wind speed of the resultant wind. The wind direction can be calculated through inverse trigonometric functions (such as atan2), and the wind speed is the length of the resultant vector.

[0105] It should be noted that the calculation of the combined wind force belongs to well-known technology, and this application does not limit it. At the same time, this application can also adopt other calculation methods for the combined wind field. Further, it should be noted that the unit area of this application can generally be configured as 10 square meters or a larger area, and this application does not limit it. Specifically, in the urban space, there will be shuttle winds between high-rise buildings. Therefore, this application needs to consider the influence of the local wind field to avoid calculation errors caused by uneven local wind field data. Obviously, for example, in an open area, if the wind field does not cause local unevenness, the wind field can be simplified and calculated with the same wind field data to reduce the data processing volume, and this application does not limit it.

[0106] After the combined wind force and the combined wind direction are determined, the influence degree of the local map under the combined wind force and the combined wind direction can be determined by combining the shooting angle of the camera and the position and connection relationship of the prefabricated connectors. Exemplarily, if the combined wind direction is directly facing the surface of the prefabricated connector and the local map is the surface of the prefabricated connector, it can be known that the combined wind force will only make the vibration direction of the prefabricated connector the same as the combined wind direction. Therefore, for the local map, no matter how large the vibration is, there will be no obvious difference in the image. Therefore, this local map can be excluded or its influence can be reduced. The following will elaborate on the specific improvement solutions.

[0107] In this improvement solution, the determination subunit includes:

[0108] Redetermine the deviation weight coefficient at the corresponding time point according to the comprehensive wind force and the initial standard wind force.

[0109] Specifically, if the comprehensive wind force is greater than the initial standard wind force, reduce the deviation weight coefficient of the first K1 local maps in the wind direction influence ranking, where the reduction amplitude accounts for the ratio of the difference between the comprehensive wind force and the initial standard wind force to the initial standard wind force in the original deviation weight coefficient, and increase the deviation weight coefficient of the last K2 local maps in the wind direction influence ranking, where the increase amplitude accounts for the ratio of the difference between the comprehensive wind force and the initial standard wind force to the initial standard wind force in the original deviation weight coefficient; if the comprehensive wind force is less than the initial standard wind force, increase the deviation weight coefficient of the first K1 local maps in the wind direction influence ranking, where the increase amplitude accounts for the ratio of the difference between the comprehensive wind force and the initial standard wind force to the initial standard wind force in the original deviation weight coefficient, and reduce the deviation weight coefficient of the last K2 local maps in the wind direction influence ranking, where the reduction amplitude accounts for the ratio of the difference between the comprehensive wind force and the initial standard wind force to the initial standard wind force in the original deviation weight coefficient;

[0110] Perform a weighted average calculation based on the final deviation weight coefficient of each local map and the deviation of each local map, and output the final deviation of the prefabricated connector.

[0111] Determine the stability performance of the prefabricated building according to the final deviation of all prefabricated connectors and the position of each prefabricated connector.

[0112] In this embodiment, if the comprehensive wind force is large, it indicates that the overall vibration amplitude of the prefabricated connector is large. At this time, the influence of the wind direction on the vibration is small, and there is still an influence even on the local maps with small wind direction influence. At this time, by reducing the deviation weight coefficient of the first K1 local maps in the wind direction influence ranking, since the total weight is 1, the deviation weight coefficient of the last K2 local maps in the wind direction influence ranking is correspondingly increased at this time, so as to ensure that the total weight is dynamically 1. At this time, although the weight in the middle position remains unchanged, if K2 is set to be less than K1, the weight ratio is relatively increased. Therefore, by configuring the sizes of K2 and K1, the weight of the local map with the greatest influence can be relatively reduced, and the weight of the local map with the smallest influence can be increased, which conforms to the characteristics of the influence brought by the comprehensive wind force and makes the prediction result more accurate.

[0113] When the overall wind force is relatively small, the difference of the local map with large wind direction influence is relatively large, while the difference of the local map with small wind force influence is difficult to be obviously reflected in the local map due to the limitation of vibration direction. At this time, a similar concept can be used for configuration. Specifically, the deviation weight coefficient of the local map corresponding to the first K1 positions in the wind direction influence ranking is increased, and the deviation weight coefficient of the local map corresponding to the last K2 positions in the wind direction influence ranking is reduced, so that the weight of the local map with the greatest influence is relatively increased, and the weight of the local map with the least influence is reduced, which is in line with the characteristics of the impact brought by the low overall wind force, making the prediction result more accurate.

[0114] Furthermore, in the embodiment of the present application, a weighted average calculation is performed based on the final deviation weight coefficient of each local graph and the deviation of each local graph, and the final deviation of the prefabricated connector is output. Specifically, the deviation of each local graph can be multiplied by its corresponding weight coefficient, and then all the results are added to obtain the total weighted deviation. The formula can be expressed as:

[0115]

[0116] Where n is the number of local graphs, deviation i is the deviation of the i-th local graph, and weight i is the deviation weight coefficient of the i-th local graph.

[0117] Furthermore, in the embodiment of the present application, if the deviation is large, it means that the stability of the building is poor, so the stability can be evaluated by configuring the deviation to divide the levels.

[0118] For example, the stability performance of the prefabricated building is determined according to the final deviation of all prefabricated connectors and the position of each prefabricated connector, specifically including:

[0119] If the deviation is higher than Y1, the stability is poor; if the deviation is within the range of Y2-Y1, the stability is moderate; if the deviation is between Y3-Y2, the stability is good, among which Y1 is greater than Y2, and Y2 is greater than Y3, and this application does not impose any restrictions on this.

[0120] Of course, the present application may also adopt other methods or set other stability determination standards, but it should be understood that the current stability evaluation includes evaluations of the center of gravity, structural stability, etc., which is not required for the present application. The core concept of the present application is to make judgments based on image deviation, that is, to judge stability directly by the position displacement caused by vibration, and cleverly combines the characteristics that the top of high-rise buildings is greatly affected by wind force, making it more accurate than other existing technologies, and due to the high data accuracy, there is no need for long-term observation, so that rapid analysis can be performed.

[0121] See also Figure 6As shown, the electronic device of this embodiment includes: a processor 11 ( Figure 6 only one processor is shown), a memory 12, and a computer program 13 stored in the memory 12 and executable on the processor 11. When the processor 11 executes the computer program 13, it realizes the functions of each module / unit in the above device embodiments, such as Figure 2 the functions of each module shown.

[0122] Exemplarily, the computer program 13 can be divided into one or more modules / units. The one or more modules / units are stored in the memory 12 and executed by the processor 11 to complete this application. The one or more modules / units can be a series of computer program instruction segments capable of completing specific functions, and these instruction segments are used to describe the execution process of the computer program 13 in the terminal device.

[0123] Those skilled in the art can understand that Figure 6 this is only an example of an electronic device and does not constitute a limitation on the electronic device. It may include more or fewer components than shown in the figure, or combine certain components, or have different components. For example, the electronic device may also include input / output devices, network access devices, a bus, etc. The processor 11 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0124] The memory 12 can be an internal storage unit of the terminal device, such as the hard disk or memory of the terminal device. The memory 12 can also be an external storage device of the terminal device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal device. Further, the memory 12 can also include both the internal storage unit and the external storage device of the terminal device. The memory 12 is used to store the computer program and other programs and data required by the terminal device. The memory 12 can also be used to temporarily store data that has been output or will be output.

[0125] As described above, it is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims described above.

Claims

1. A rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters, characterized in that, The multi-dimensional parameters include multiple local images corresponding to each prefabricated connector and a wind speed field cloud image. The multiple local images are obtained by capturing the corresponding prefabricated connectors with fixed-angle cameras at multiple fixed orientations. The rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters includes: An acquisition module that acquires the wind speed field cloud image of the prefabricated building to be analyzed and multiple local images corresponding to each prefabricated connector collected at regular intervals; A model input module that inputs all the local images collected for each prefabricated connector at each sampling moment into the differential neural network model corresponding to the prefabricated connector, and outputs the deviation degree from the standard image at each sampling moment; The differential neural network model includes a channel attention mechanism network, an image cropping layer, a convolutional network, and a comparison layer; The input of the channel attention mechanism network is each local image, and the output is used as the input of the image cropping layer. The image cropping layer outputs an image after removing the feature channels with weight coefficients lower than the set threshold and inputs it to the convolutional network. The output of the convolutional network is used as the input of the comparison layer, and the comparison layer is pre-configured with a standard image corresponding to the prefabricated connector; An analysis module that analyzes and determines the stability performance of the prefabricated building according to the deviation degree at each sampling moment and the wind speed field cloud image for the prefabricated connector; The analysis module includes: An alignment unit that superimposes each prefabricated connector with a preset time offset and sampling moment to calibrate and align the sampling moment with the time coordinate in the wind speed field cloud image; A search unit that searches for the wind field state data at the corresponding moment from the wind speed field cloud image for the aligned sampling moment; A determination unit that analyzes and determines the stability performance of the prefabricated building by combining the wind field state data and the corresponding deviation degree data for the aligned sampling moment; The wind field state data includes the wind speed and wind direction of each unit area in the spatial area. The determination unit includes: A comprehensive wind field determination sub-unit that determines the comprehensive wind direction and comprehensive wind force in the spatial area according to the spatial area where the prefabricated connector is located and the wind speed and wind direction of each unit area for each prefabricated connector; A sorting sub-unit that determines the sorting of the wind direction influence of all local images according to the comprehensive wind direction and the shooting angle of the fixed-angle camera; A weight coefficient generation sub-unit that dynamically generates the deviation degree weight coefficient of each local image according to the sorting of the wind direction influence; A determination sub-unit that analyzes and determines the stability performance of the prefabricated building by combining the deviation degree weight coefficient of each local image and the deviation degree of each local image with the comprehensive wind force at the corresponding time point.

2. The rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters according to claim 1, wherein The convolutional network specifically includes: multiple convolutional layers and multiple pooling layers. Among them, during the testing process of the difference neural network model, when multiple test data with a difference degree lower than a set threshold are sequentially input into the difference neural network model, some nodes of any one or more convolutional layers are solidified each time until the output difference of the convolutional network before and after solidification is within the set range. At this time, the solidified nodes are output, and an iterative operation is performed until all similar training data have been trained or the set solidification ratio is reached, and the final convolutional network is output.

3. The rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters according to claim 2, wherein, The prefabricated building performance rapid analysis system further includes: A model pre-optimization module, and the model pre-optimization module includes: A rejection unit, randomly rejecting nodes of any one convolutional layer in the convolutional network, and using nodes with a weight of zero as a replacement to replace the rejected nodes. The same test data is used for testing before and after rejection. If the difference degree of the test results is within the standard range, the convolutional network is updated. An update unit, replacing the initial convolutional network with the updated convolutional network, and training the convolutional network using training data to obtain the final convolutional network.

4. The rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters according to claim 2, characterized in that, The prefabricated building performance rapid analysis system further includes: A model pre-optimization module, and the model pre-optimization module includes: A deletion unit, randomly deleting nodes of any one convolutional layer in the convolutional network, and using nodes with a weight of zero as a replacement to replace the rejected nodes. The same test data is used for testing before and after rejection. If the difference degree of the test results is within the standard range, the convolutional network is updated. An explosion unit, randomly using at least one node in at least one convolutional layer of the updated convolutional network as a standard node, increasing at least one node according to the weight of the standard node, and inserting it into the corresponding convolutional layer; where the number of deleted nodes is greater than the number of added nodes.

5. The rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters according to claim 3 or 4, characterized in that, The model pre-optimization module further includes: An image slicing layer, which is used to slice the local image into multiple sub-images; An image splicing layer. For each adjacent three sub-images, the starting N columns of the first sub-image are inserted before the first column of the middle sub-image, and the last M columns of the last sub-image are inserted after the last column of the middle sub-image to regenerate the middle sub-image, and replace the local image, and input it into the convolutional layer.

6. The rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters according to claim 3 or 4, characterized in that The model pre-optimization module further includes: An image partitioning layer, randomly partitioning the local image into multiple regions, the size of each region can be the same or different, and the size of the largest region is constrained within a preset range; An image deletion layer, randomly deleting one or more of the regions, and performing a zero-padding operation on the deleted region part to form an updated local image, and replacing the current local image and inputting it into the convolutional layer. If the deviation rate of change of the corresponding output using the same test data is higher than the standard threshold, the deleted region is restored, and a random deletion operation is performed again until the deviation rate of change of the corresponding output using the same test data is lower than the standard threshold, and the current updated local image is retained as the input of the final convolutional layer.

7. The rapid analysis system for the performance of prefabricated buildings based on multi-dimensional parameters according to claim 6, characterized in that, The determination subunit includes: According to the comprehensive wind force and the initial standard wind force, re-determine the deviation weight coefficient at the corresponding time point. If the comprehensive wind force is greater than the initial standard wind force, reduce the deviation weight coefficient of the first K1 local maps in the wind direction influence ranking, where the reduction amplitude accounts for the ratio of the difference between the comprehensive wind force and the initial standard wind force to the initial standard wind force in the original deviation weight coefficient, and increase the deviation weight coefficient of the last K2 local maps in the wind direction influence ranking, where the increase amplitude accounts for the ratio of the difference between the comprehensive wind force and the initial standard wind force to the initial standard wind force in the original deviation weight coefficient; if the comprehensive wind force is less than the initial standard wind force, increase the deviation weight coefficient of the first K1 local maps in the wind direction influence ranking, where the increase amplitude accounts for the ratio of the difference between the comprehensive wind force and the initial standard wind force to the initial standard wind force in the original deviation weight coefficient, and reduce the deviation weight coefficient of the last K2 local maps in the wind direction influence ranking, where the reduction amplitude accounts for the ratio of the difference between the comprehensive wind force and the initial standard wind force to the initial standard wind force in the original deviation weight coefficient; Perform a weighted average calculation based on the final deviation weight coefficient of each local map and the deviation of each local map, and output the final deviation of the prefabricated connector; Based on the final deviations of all prefabricated connectors, determine the stability performance of the prefabricated building in combination with the position of each prefabricated connector.

Images