Arrangement design method and device for building structural components based on heterogeneous graph neural network

Abstract

The invention provides a building structure component arrangement design method and device based on a heterogeneous graph neural network, and relates to the technical field of building structure component arrangement design, and the method comprises the steps: obtaining a to-be-processed building drawing; extracting the position information of the building component and the position information of the potential structural component from the to-be-processed building drawing, and generating a heterogeneous graph based on the position information of the building component and the position information of the potential structural component; the heterogeneous graph is used for representing a building structure as a heterogeneous graph with different types of nodes and different edge connection relationships; and inputting the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a building structure component arrangement design strategy. According to the building structural member arrangement design method and device based on the heterogeneous graph neural network provided by the invention, the corresponding structural member arrangement design strategy is generated by using the heterogeneous graph and the pre-trained heterogeneous graph neural network model, so that the efficiency of building structural member arrangement design is improved.

Description

Technical Field

[0001] This application relates to the technical field of architectural structure member layout design, and in particular, to a method and device for architectural structure member layout design based on a heterogeneous graph neural network. Background Art

[0002] During the architectural structure design process, the layout of structural members is a very important step. Engineers often rely on professional knowledge and experience to determine the positions of members, but manual layout is time-consuming, laborious, inefficient, and difficult for beginners to quickly master. The design threshold is relatively high, which is not conducive to the intelligent transformation of structural design.

[0003] To solve the above problems existing in the related technologies, in recent years, researchers have begun to explore the use of artificial intelligence technology for the layout design of architectural structure members. However, these methods usually design the layout of different types of members separately, lacking consideration of the overall structure layout, and unable to ensure the safety and economy of the design results.

[0004] Therefore, there is an urgent need for an efficient and intelligent method for structural member layout design to overcome many limitations in traditional designs. Summary of the Invention

[0005] The purpose of this application is to provide a method and device for architectural structure member layout design based on a heterogeneous graph neural network, which uses a heterogeneous graph and a pre-trained heterogeneous graph neural network model to generate corresponding structural member layout design strategies, so as to improve the efficiency of architectural structure member layout design, achieve the efficient design of architectural structure members, and realize the automated transformation of architectural structure design.

[0006] This application provides a method for architectural structure member layout design based on a heterogeneous graph neural network, including: Obtain the building drawing to be processed; extract the position information of building components and the position information of potential structural components from the building drawing to be processed, and generate a heterogeneous graph based on the position information of the building components and the position information of the potential structural components; the position information of the potential structural components is used to represent that a horizontal structural component can be arranged at the position where the potential structural component is located without affecting the function of the building; the heterogeneous graph is used to represent the building structure as a heterogeneous graph with different types of nodes and different edge connection relationships; the edge connection relationship is determined according to the position relationship between the minimum line segments used to represent each building component and each potential structural component; input the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a building structure component layout design strategy; wherein, the building structure component layout design strategy includes: whether to arrange structural components for each heterogeneous node in the heterogeneous graph, and the type of the arranged structural components; the heterogeneous graph neural network model is: trained with a training sample data set composed of building drawings, the corresponding generated heterogeneous graphs, and the corresponding actual structural component layout strategies.

[0007] Optionally, the extracting the position information of building components and the position information of potential structural components from the building drawing to be processed includes: extracting the position information of building components from the building drawing to be processed and extracting the position information of potential structural components from the building drawing to be processed by using a preset rule coding method; wherein, the preset rule coding method includes: extending the line segment used to represent the building component until it intersects with the outer contour line of the building, or extending any line segment used to represent the building component until it intersects with the line segment of other building components; the potential structural component is represented by the extended line segment obtained after intersection.

[0008] Optionally, the generating a heterogeneous graph based on the position information of the building components and the position information of the potential structural components includes: constructing different types of nodes based on the minimum line segment used to represent the building component, the minimum line segment used to represent the potential structural component, and the intersection points of all the minimum line segments; the different types of nodes include: door nodes, window nodes, wall nodes, potential structural nodes, and intersection nodes; the edge connection relationship includes: the edge connection relationship between different types of nodes and the edge connection relationship between the same type of nodes; according to the position relationship between each minimum line segment, determine different types of edge connection relationships, and use different types of nodes as heterogeneous nodes, and based on the different types of edge connection relationships, determine the edge relationships of each heterogeneous node to generate the heterogeneous graph.

[0009] Optionally, input the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a building structure component layout design strategy, including: performing message passing and feature aggregation on the nodes within the edge relationships of each type in the heterogeneous graph to update the node features of different types of nodes, and aggregating the nodes obtained from different types of edge connection relationships to update the node features again; converting the node features of different types of nodes into structural layout features through a multi-layer perceptron network, and performing information extraction based on the structural layout features to determine whether structural components are arranged for each heterogeneous node in the heterogeneous graph and the types of the arranged structural components.

[0010] Optionally, the heterogeneous graph neural network model is trained based on the following steps: obtaining a plurality of building drawings and the corresponding actual structural component layout strategies for each building drawing, and constructing a heterogeneous graph corresponding to each building drawing; using the plurality of building drawings, the corresponding actual structural component layout strategies for each building drawing, and the heterogeneous graph corresponding to each building drawing as a training sample set to train the network model to be trained; using the difference between the predicted layout type of the structural components and the actual layout type of the structural components as the training loss during the training process, and iteratively optimizing the network model to be trained through a preset loss function to obtain a pre-trained heterogeneous graph neural network model.

[0011] This application also provides a building structure component layout design device based on a heterogeneous graph neural network, including: An acquisition module, configured to acquire a building drawing to be processed; a generation module, configured to extract the position information of building components and the position information of potential structural components from the building drawing to be processed, and generate a heterogeneous graph based on the position information of the building components and the position information of the potential structural components; the position information of the potential structural components is used to represent that a horizontal structural component can be arranged at the position where the potential structural component is located without affecting the function of the building; the heterogeneous graph is used to represent the building structure as a heterogeneous graph with different types of nodes and different edge connection relationships; the edge connection relationship is determined according to the position relationship between the minimum line segments used to represent each building component and each potential structural component; a design module, configured to input the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a building structure component layout design strategy; wherein, the building structure component layout design strategy includes: whether structural components are arranged for each heterogeneous node in the heterogeneous graph and the types of the arranged structural components; the heterogeneous graph neural network model is: trained with a training sample data set composed of a building drawing, the corresponding generated heterogeneous graph, and the corresponding actual structural component layout strategy.

[0012] Optionally, the generating module is specifically configured to extract the position information of building components from the to-be-processed building drawings and extract the position information of potential structural components from the to-be-processed building drawings by using a preset rule coding method; wherein, the preset rule coding method includes: extending the line segments for representing building components until they intersect with the outer contour line of the building, or extending any line segment for representing a building component until it intersects with the line segments of other building components; the potential structural components are represented by the extended line segments obtained after intersection.

[0013] Optionally, the generating module is specifically configured to construct different types of nodes based on the minimum line segments for representing the building components, the minimum line segments for representing the potential structural components, and the intersection points of all the minimum line segments; the different types of nodes include: door nodes, window nodes, wall nodes, potential structural nodes, and intersection nodes; the edge connection relationships include: edge connection relationships between different types of nodes and edge connection relationships between the same type of nodes; the generating module is specifically further configured to determine different types of edge connection relationships according to the positional relationships between the minimum line segments, use the different types of nodes as heterogeneous nodes, and determine the edge relationships of the heterogeneous nodes based on the different types of edge connection relationships to generate the heterogeneous graph.

[0014] Optionally, the design module is specifically configured to perform message passing and feature aggregation on the nodes inside the edge relationships of each type in the heterogeneous graph to update the node features of different types of nodes, and perform feature aggregation on the nodes obtained from the different types of edge connection relationships to update the node features again; the design module is specifically further configured to convert the node features of different types of nodes into structural layout features through a multi-layer perceptron network, and perform information extraction according to the structural layout features to determine whether structural components are arranged on the heterogeneous nodes in the heterogeneous graph and the types of the arranged structural components.

[0015] Optionally, the device further includes: a sample construction module and a training module; the sample construction module is used to obtain a plurality of building drawings and the actual structural component layout strategies corresponding to each building drawing, and construct the heterogeneous graph corresponding to each building drawing; the training module is used to use the plurality of building drawings, the actual structural component layout strategies corresponding to each building drawing, and the heterogeneous graph corresponding to each building drawing as a training sample set to train the to-be-trained network model; the training module is further used to use the difference between the predicted layout type of the structural components and the actual layout type of the structural components as the training loss during the training process, and iteratively optimize the to-be-trained network model through a preset loss function to obtain a pre-trained heterogeneous graph neural network model.

[0016] The present application also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the method for designing the layout of building structure components based on a heterogeneous graph neural network as described in any one of the above.

[0017] The present application also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the steps of the method for designing the layout of building structure components based on a heterogeneous graph neural network as described in any one of the above.

[0018] The present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the method for designing the layout of building structure components based on a heterogeneous graph neural network as described in any one of the above.

[0019] For the method and device for designing the layout of building structure components based on a heterogeneous graph neural network provided by the present application, first, obtain a building drawing to be processed; then, extract the position information of building components and the position information of potential structure components from the building drawing to be processed, and generate a heterogeneous graph based on the position information of the building components and the position information of the potential structure components; the position information of the potential structure components is used to represent that a horizontal structure component can be arranged at the position where the potential structure component is located without affecting the function of the building; the heterogeneous graph is used to represent the building structure as a heterogeneous graph with different types of nodes and different edge connection relationships; the edge connection relationship is determined according to the position relationship between the minimum line segments used to represent each building component and each potential structure component; finally, input the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a design strategy for the layout of building structure components; wherein, the design strategy for the layout of building structure components includes whether to arrange structure components for each heterogeneous node in the heterogeneous graph and the types of the arranged structure components; the heterogeneous graph neural network model is trained with a training sample data set composed of building drawings, the corresponding generated heterogeneous graphs, and the corresponding actual structure component layout strategies. In this way, a corresponding design strategy for the layout of structure components is generated by using the heterogeneous graph and the pre-trained heterogeneous graph neural network model to improve the efficiency of the design of the layout of building structure components, realize the efficient design of the layout of building structure components, and realize the automatic transformation of building structure design. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0021] Figure 1 It is a schematic flow chart of the building structure component layout design method based on the heterogeneous graph neural network provided by this application; Figure 2 It is a schematic diagram of the layout design of the reinforced concrete frame-shear wall structure components based on the heterogeneous graph neural network provided by this application; Figure 3 It is a schematic flow chart of the method for obtaining potential structural components by rule coding provided by this application; Figure 4 It is a schematic diagram of the heterogeneous graph representation method of the building structure provided by this application; Figure 5 It is a schematic diagram of the architecture of the heterogeneous graph neural network model provided by this application; Figure 6 It is a schematic flow chart of the training process of the heterogeneous graph neural network provided by this application; Figure 7 It is a schematic diagram of the structure of the building structure component layout design device based on the heterogeneous graph neural network provided by this application; Figure 8 It is a schematic diagram of the structure of the electronic device provided by this application. Detailed implementation manners

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below with reference to the accompanying drawings in this application. Obviously, the described embodiments are some but not all of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments in this application belong to the scope of protection of this application.

[0023] The terms "first", "second", etc. in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of this application can be implemented in an order different from those illustrated or described herein, and the objects distinguished by "first", "second", etc. generally belong to the same category, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims means at least one of the connected objects, and the character " / " generally means an "or" relationship between the associated objects before and after.

[0024] The building structure component layout design method based on the heterogeneous graph neural network provided by the embodiments of this application will be described in detail below with reference to the accompanying drawings through specific embodiments and their application scenarios.

[0025] As Figure 1As shown, a method for architectural structural member layout design based on a heterogeneous graph neural network provided by an embodiment of the present application may include the following steps 101 to 103: Step 101: Obtain the architectural drawing to be processed.

[0026] Exemplarily, before performing the architectural structural member layout design, it is first necessary to obtain the architectural design CAD floor plan, that is, the above-mentioned architectural drawing to be processed. As Figure 2 shown, first, architectural components (including doors, windows, walls, and room type information) are extracted from it. Then, the method for architectural structural member layout design based on the heterogeneous graph neural network provided by the present application is used for predictive design to obtain a structural member layout plan (including the layout design of beams, columns, and shear walls).

[0027] Step 102: Extract the position information of architectural components and the position information of potential structural members from the architectural drawing to be processed, and generate a heterogeneous graph based on the position information of the architectural components and the position information of the potential structural members.

[0028] Among them, the position information of the potential structural members is used to represent that the position where the potential structural members are located can arrange horizontal structural members without affecting the function of the building; the heterogeneous graph is used to represent the architectural structure as a heterogeneous graph with different types of nodes and different edge connection relationships; the edge connection relationship is determined according to the position relationship between the minimum line segments used to represent each architectural component and each potential structural member.

[0029] Exemplarily, first obtain the architectural drawing, generate potential structural members through a rule encoding method based on building information. Then, construct heterogeneous nodes based on architectural components and potential structural members, and generate the corresponding heterogeneous graph; finally, based on a pre-trained heterogeneous graph neural network model, predict whether to arrange structural members for the heterogeneous nodes and the types of the arranged structural members.

[0030] Specifically, in the above step 102, the step of extracting the position information of architectural components and the position information of potential structural members from the architectural drawing to be processed may further include the following step 102a: Step 102a: Extract the position information of architectural components from the architectural drawing to be processed and extract the position information of potential structural members from the architectural drawing to be processed using a preset rule encoding method.

[0031] Among them, the preset rule encoding method includes: extending the line segment used to represent the architectural component until it intersects with the outer contour line of the building, or extending any line segment used to represent the architectural component until it intersects with the line segment of other architectural components; the potential structural members are represented by the extended line segments obtained after intersection.

[0032] Exemplarily, in the actual design process, there are often certain differences between the locations of building components and structural components. To cover all possible structural layout positions, a regular coding method is first used to obtain potential structural components. For example Figure 3 As shown, taking a reinforced concrete frame-shear wall structure as an example, since shear wall components must be arranged at the positions of building walls, and column components are arranged at the intersections of the floor plan, the potential structural components are the positions where beam components can be arranged. To enable the potential structural components to cover more real layout positions, a preset regular coding method is adopted. First, the outer contour of the building is found. Secondly, all doors, windows, walls, and room line segments (in architectural drawings, different types of line segments are used to represent building components such as doors, windows, and walls) are extended to intersect with the outer contour (or, the building components intersect with each other) to serve as potential structural components (such as Figure 3 the dashed lines shown, which are used to represent potential structural components). All building components and potential structural component line segments are segmented according to the intersections to obtain the minimum line segments, which is convenient for the subsequent establishment of a heterogeneous graph.

[0033] Specifically, the step of generating a heterogeneous graph based on the position information of the building components and the position information of the potential structural components in step 102 may further include the following steps 102b1 and 102b2: Step 102b1: Construct different types of nodes based on the minimum line segments representing the building components, the minimum line segments representing the potential structural components, and the intersections of all the minimum line segments.

[0034] Among them, the different types of nodes include: door nodes, window nodes, wall nodes, potential structure nodes, and intersection nodes; the edge connection relationships include: edge connection relationships between different types of nodes and edge connection relationships between the same type of nodes.

[0035] Step 102b2: Determine different types of edge connection relationships according to the positional relationships between the minimum line segments, and use the different types of nodes as heterogeneous nodes and determine the edge relationships of each heterogeneous node based on the different types of edge connection relationships to generate the heterogeneous graph.

[0036] Exemplarily, as Figure 4 shown Figure 4Each line segment in it represents a building component, and the points in the middle of the line segments represent nodes. For example, a thick line can represent a window, and the point in the middle of the thick line can represent a window node. Five types of heterogeneous nodes are constructed according to building components and potential structural components. Door nodes, window nodes, and wall nodes are constructed based on the minimum line segments of doors, windows, and wall components respectively, potential structural nodes are constructed based on the minimum line segments of potential structural components, and at the same time, all line segment intersections are constructed as intersection nodes. The edge connection relationship of the heterogeneous graph is constructed according to the positional relationship of the line segments. There is an edge connection relationship between the intersection node and other line segment nodes connected to it, and there is also an edge connection relationship between different line segment nodes connected to the intersection node. Since the constructed heterogeneous graph is an undirected graph, there are ten different types of edge relationships: door-window, door-wall, door-potential structural component, door-intersection, window-wall, window-potential structural component, window-intersection, wall-potential structural component, wall-intersection, potential structural component-intersection. At the same time, in order to further strengthen the information transmission, self-loops are added to each type of node, that is, five types of edge connection relationships: door-door, window-window, wall-wall, potential structural component-potential structural component, intersection-intersection. Based on this method, the building structure can be represented as a heterogeneous graph with five different types of nodes and fifteen different edge relationships.

[0037] Exemplarily, for the above different types of nodes, different characteristic attributes can be assigned to them, including: position, length, direction, and the types of surrounding rooms. Among them, the position attribute is divided into absolute position and relative position. The absolute position refers to the coordinate position of the midpoint of the component on the plane after centering processing of the standard floor; the relative position refers to the relative coordinate position of the midpoint of the component after normalization processing, as well as the total lengths of the standard floor in the horizontal and vertical directions.

[0038] Step 103: Input the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a building structure component layout design strategy.

[0039] Among them, the building structure component layout design strategy includes: whether structural components are arranged for each heterogeneous node in the heterogeneous graph, and the types of the arranged structural components; the heterogeneous graph neural network model is trained with a training sample data set composed of building drawings, the corresponding generated heterogeneous graph, and the corresponding actual structural component layout strategy.

[0040] Exemplarily, after obtaining the heterogeneous graph, the heterogeneous graph can be input into a pre-trained heterogeneous graph neural network model for prediction to judge whether structural components are arranged for each heterogeneous node and the types of the arranged structures.

[0041] Specifically, the above step 103 may further include the following steps 103a1 and 103a2: Step 103a1: Perform message passing and feature aggregation on the nodes within the edge relationships of each type in the heterogeneous graph, update the node features of different types of nodes, and perform feature aggregation on the nodes obtained from different types of edge connection relationships, and then update the node features again.

[0042] Step 103a2: Convert the node features of different types of nodes into structural layout features through a multi-layer perceptron network, and perform information extraction based on the structural layout features to determine whether structural members are arranged for each heterogeneous node in the heterogeneous graph and the types of the arranged structural members.

[0043] Exemplarily, as Figure 5 shown, it is the model architecture of the heterogeneous graph neural network provided by the embodiment of the present application. This model uses heterogeneous graph technology to process the information transmission, integration, and update processes. For different types of connection edges, information communication and aggregation can be achieved through graph neural network technology, and various methods such as GCN, GAT, and GraphSAGE can be used.

[0044] Exemplarily, in the information aggregation stage, for each node type, the model sums and integrates the information collected from multiple edges to ensure that data from different sources can be effectively fused. Subsequently, this integrated information will be further processed and feature extracted through a multi-layer perceptron network. And through the residual connection technology, the problem of gradient disappearance that may occur in deep networks is alleviated, and the expression and generalization capabilities of the model in complex tasks are enhanced. After multi-level iterative updates, this heterogeneous graph neural network can finally accurately predict whether structural members need to be arranged and identify the specific types of each member.

[0045] Exemplarily, after obtaining the building structure member layout design strategy output by the model, the layout requirements can be checked for each member according to the structural design specifications, and the layouts that do not meet the specifications can be adjusted and modified.

[0046] Optionally, in the embodiment of the present application, building drawings and corresponding heterogeneous graphs can be used to train the model, and then a heterogeneous graph neural network model can be obtained.

[0047] Exemplarily, before the above step 103, the building structure member layout design method based on the heterogeneous graph neural network provided by the embodiment of the present application may further include the following steps 104 to 106: Step 104: Obtain a plurality of building drawings and the corresponding actual structural member layout strategies for each building drawing, and construct a heterogeneous graph corresponding to each building drawing.

[0048] Step 105: Use the multiple architectural drawings, the corresponding actual structural member layout strategies for each architectural drawing, and the heterogeneous graph corresponding to each architectural drawing as a training sample set to train the network model to be trained.

[0049] Step 106: During the training process, use the difference between the predicted layout type of the structural member and the actual layout type of the structural member as the training loss, and iteratively optimize the network model to be trained through a preset loss function to obtain a pre-trained heterogeneous graph neural network model.

[0050] Exemplarily, as Figure 6 shown, based on actual engineering projects, 300 buildings of reinforced concrete frame structure, reinforced concrete frame-shear wall structure, and reinforced concrete shear wall structure were selected as the data set. First, use the Figure 3 method shown to generate potential structural members, use the Figure 4 method shown to convert CAD drawings into heterogeneous graph representations, use different types of nodes to represent doors, windows, walls, potential structural members, and intersections respectively, and store the features of various components into the corresponding nodes. Use the Figure 5 heterogeneous graph neural network to learn the design patterns of the dimensions of building structural members. Among them, the nodes of doors, windows, walls, and potential structural members are used to predict whether to arrange structural members, and whether the arranged structural member is a shear wall or a beam. The intersection nodes are used to predict whether to arrange columns. Therefore, the training process uses the cross-entropy loss function to measure the deviation between the model prediction layout and the true layout, and uses this as the basis for optimizing the model performance.

[0051] The building structure component layout design method based on the heterogeneous graph neural network provided by the embodiment of the present application. First, obtain the building drawing to be processed; then, extract the position information of building components and the position information of potential structure components from the building drawing to be processed, and generate a heterogeneous graph based on the position information of the building components and the position information of the potential structure components; the position information of the potential structure components is used to represent that a horizontal structure component can be arranged at the position where the potential structure component is located without affecting the function of the building; the heterogeneous graph is used to represent the building structure as a heterogeneous graph with different types of nodes and different edge connection relationships; the edge connection relationship is determined according to the position relationship between the minimum line segments used to represent each building component and each potential structure component; finally, input the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a building structure component layout design strategy; wherein, the building structure component layout design strategy includes whether to arrange structure components for each heterogeneous node in the heterogeneous graph and the type of the arranged structure components; the heterogeneous graph neural network model is trained with a training sample data set composed of building drawings, the corresponding generated heterogeneous graphs, and the corresponding actual structure component layout strategies. In this way, a corresponding structure component layout design strategy is generated by using the heterogeneous graph and the pre-trained heterogeneous graph neural network model to improve the efficiency of building structure component layout design, realize the efficient design of building structure components, and realize the automated transformation of building structure design.

[0052] It should be noted that for the building structure component layout design method based on the heterogeneous graph neural network provided by the embodiment of the present application, the execution subject can be a building structure component layout design device based on the heterogeneous graph neural network, or a control module in the building structure component layout design device based on the heterogeneous graph neural network for executing the building structure component layout design method based on the heterogeneous graph neural network. In the embodiment of the present application, the building structure component layout design device based on the heterogeneous graph neural network is described by taking the building structure component layout design device based on the heterogeneous graph neural network as an example to execute the building structure component layout design method based on the heterogeneous graph neural network.

[0053] It should be noted that in the embodiment of the present application, the building structure component layout design method based on the heterogeneous graph neural network shown in each of the above method drawings is exemplarily described by taking one of the drawings in the embodiment of the present application as an example. Specifically, when implemented, the building structure component layout design method based on the heterogeneous graph neural network shown in each of the above method drawings can also be implemented in combination with any other drawings that can be combined as shown in the above embodiments, which will not be elaborated here.

[0054] The following describes the device for designing the layout of building structure components based on the heterogeneous graph neural network provided by this application. The following description can be correspondingly referred to the method for designing the layout of building structure components based on the heterogeneous graph neural network described above.

[0055] Figure 7 FIG. 4 is a schematic structural diagram of the device for designing the layout of building structure components based on the heterogeneous graph neural network provided by an embodiment of this application. As Figure 7 shown, it specifically includes: An acquisition module 701, configured to acquire a building drawing to be processed; a generation module 702, configured to extract the position information of building components and the position information of potential structural components from the building drawing to be processed, and generate a heterogeneous graph based on the position information of the building components and the position information of the potential structural components; the position information of the potential structural components is used to represent that a horizontal structural component can be arranged at the position where the potential structural component is located without affecting the function of the building; the heterogeneous graph is used to represent the building structure as a heterogeneous graph with different types of nodes and different edge connection relationships; the edge connection relationship is determined according to the position relationship between the minimum line segments used to represent each building component and each potential structural component; a design module 703, configured to input the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a design strategy for the layout of building structure components; wherein, the design strategy for the layout of building structure components includes: whether to arrange structural components for each heterogeneous node in the heterogeneous graph, and the types of the arranged structural components; the heterogeneous graph neural network model is: trained with a training sample data set composed of building drawings, the corresponding generated heterogeneous graphs, and the corresponding actual layout strategies of structural components.

[0056] Optionally, the generation module 702 is specifically configured to extract the position information of building components from the building drawing to be processed and extract the position information of potential structural components from the building drawing to be processed by using a preset rule encoding method; wherein, the preset rule encoding method includes: extending the line segment used to represent the building component until it intersects with the outer contour line of the building, or extending any line segment used to represent the building component until it intersects with the line segment of other building components; the potential structural component is represented by the extended line segment obtained after intersection.

[0057] Optionally, the generating module 702 is specifically configured to construct different types of nodes based on the minimum line segments for characterizing the building components, the minimum line segments for characterizing the potential structural components, and the intersection points of all the minimum line segments; the different types of nodes include: door nodes, window nodes, wall nodes, potential structural nodes, and intersection points nodes; the edge connection relationships include: the edge connection relationships between different types of nodes and the edge connection relationships of the same type of nodes; the generating module 702 is further specifically configured to determine different types of edge connection relationships according to the positional relationships between the minimum line segments, use the different types of nodes as heterogeneous nodes, and determine the edge relationships of the heterogeneous nodes based on the different types of edge connection relationships, so as to generate the heterogeneous graph.

[0058] Optionally, the design module 703 is specifically configured to perform message passing and feature aggregation on the nodes inside the edge relationship of each type in the heterogeneous graph, update the node features of different types of nodes, and perform feature aggregation on the nodes obtained from different types of edge connection relationships, and update the node features again; the design module 703 is further specifically configured to convert the node features of different types of nodes into structural layout features through a multi-layer perceptron network, and perform information extraction according to the structural layout features to determine whether structural components are arranged for the heterogeneous nodes in the heterogeneous graph and the types of the arranged structural components.

[0059] Optionally, the device further includes: a sample construction module and a training module; the sample construction module is configured to obtain a plurality of building drawings and the actual structural component layout strategies corresponding to each building drawing, and construct a heterogeneous graph corresponding to each building drawing; the training module is configured to use the plurality of building drawings, the actual structural component layout strategies corresponding to each building drawing, and the heterogeneous graphs corresponding to each building drawing as a training sample set to train a network model to be trained; the training module is further configured to use the difference between the predicted layout type of the structural component and the actual layout type of the structural component as a training loss during the training process, and iteratively optimize the network model to be trained through a preset loss function to obtain a pre-trained heterogeneous graph neural network model.

[0060] The building structure component layout design device based on the heterogeneous graph neural network provided by this application first obtains the building drawings to be processed; then, extracts the position information of building components and the position information of potential structural components from the building drawings to be processed, and generates a heterogeneous graph based on the position information of the building components and the position information of the potential structural components; the position information of the potential structural components is used to represent that a horizontal structural component can be arranged at the position where the potential structural component is located without affecting the function of the building; the heterogeneous graph is used to represent the building structure as a heterogeneous graph with different types of nodes and different edge connection relationships; the edge connection relationship is determined according to the position relationship between the minimum line segments used to represent each building component and each potential structural component; finally, inputs the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a building structure component layout design strategy; wherein, the building structure component layout design strategy includes whether to arrange structural components for each heterogeneous node in the heterogeneous graph and the type of the arranged structural components; the heterogeneous graph neural network model is trained with a training sample data set composed of building drawings, the corresponding generated heterogeneous graph, and the corresponding actual structural component layout strategy. In this way, a corresponding structural component layout design strategy is generated by using the heterogeneous graph and the pre-trained heterogeneous graph neural network model to improve the efficiency of building structure component layout design, realize the efficient design of building structure components, and realize the automatic transformation of building structure design.

[0061] Figure 8 Schematic diagram of the physical structure of an electronic device is exemplified, such as Figure 8As shown in the figure, the electronic device may include: a processor 810, a communications interface 820, a memory 830, and a communication bus 840. Among them, the processor 810, the communications interface 820, and the memory 830 communicate with each other through the communication bus 840. The processor 810 may call logic instructions in the memory 830 to execute a method for designing the layout of building structure components based on a heterogeneous graph neural network. The method includes: First, obtain the building drawing to be processed; Then, extract the position information of building components and the position information of potential structure components from the building drawing to be processed, and generate a heterogeneous graph based on the position information of the building components and the position information of the potential structure components; The position information of the potential structure components is used to represent that a horizontal structure component can be arranged at the position where the potential structure component is located without affecting the function of the building; The heterogeneous graph is used to represent the building structure as a heterogeneous graph with different types of nodes and different edge connection relationships; The edge connection relationship is determined according to the position relationship between the minimum line segments used to represent each building component and each potential structure component; Finally, input the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a design strategy for the layout of building structure components; Among them, the design strategy for the layout of building structure components includes: whether to arrange structure components for each heterogeneous node in the heterogeneous graph, and the type of the arranged structure components; The heterogeneous graph neural network model is: trained with a training sample data set composed of building drawings, the corresponding generated heterogeneous graphs, and the corresponding actual structure component layout strategies. In this way, a corresponding design strategy for the layout of structure components is generated by using the heterogeneous graph and the pre-trained heterogeneous graph neural network model to improve the efficiency of the design of the layout of building structure components, achieve the efficient design of the layout of building structure components, and realize the automated transformation of building structure design.

[0062] In addition, when the logic instructions in the above-mentioned memory 830 are implemented in the form of software function units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0063] On the other hand, the present application also provides a computer program product. The computer program product includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the method for designing the layout of building structure components based on a heterogeneous graph neural network provided by the above-mentioned various methods. The method includes: First, obtain the building drawing to be processed; Then, extract the position information of building components and the position information of potential structure components from the building drawing to be processed, and generate a heterogeneous graph based on the position information of the building components and the position information of the potential structure components; The position information of the potential structure components is used to represent that a horizontal structure component can be arranged at the position where the potential structure component is located without affecting the function of the building; The heterogeneous graph is used to represent the building structure as a heterogeneous graph with different types of nodes and different edge connection relationships; The edge connection relationship is determined according to the positional relationship between the minimum line segments used to represent each building component and each potential structure component; Finally, input the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a design strategy for the layout of building structure components; Wherein, the design strategy for the layout of building structure components includes: whether to arrange structure components for each heterogeneous node in the heterogeneous graph, and the types of the arranged structure components; The heterogeneous graph neural network model is: trained with a training sample data set composed of building drawings, the corresponding generated heterogeneous graphs, and the corresponding actual structure component layout strategies. In this way, a corresponding design strategy for the layout of structure components is generated by using the heterogeneous graph and the pre-trained heterogeneous graph neural network model, so as to improve the efficiency of the design of the layout of building structure components, realize the efficient design of the layout of building structure components, and realize the automatic transformation of building structure design.

[0064] In another aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is configured to execute the above-described building structure component layout design method based on a heterogeneous graph neural network. The method includes: First, obtain a building drawing to be processed; After that, extract the position information of building components and the position information of potential structure components from the building drawing to be processed, and generate a heterogeneous graph based on the position information of the building components and the position information of the potential structure components; The position information of the potential structure components is used to represent that a horizontal structure component can be arranged at the position where the potential structure component is located without affecting the function of the building; The heterogeneous graph is used to represent the building structure as a heterogeneous graph with different types of nodes and different edge connection relationships; The edge connection relationship is determined according to the positional relationship between the minimum line segments used to represent each building component and each potential structure component; Finally, input the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a building structure component layout design strategy; Wherein, the building structure component layout design strategy includes: whether to arrange a structure component for each heterogeneous node in the heterogeneous graph, and the type of the arranged structure component; The heterogeneous graph neural network model is: trained with a training sample data set composed of a building drawing, the corresponding generated heterogeneous graph, and the corresponding actual structure component layout strategy. In this way, a corresponding structure component layout design strategy is generated by using the heterogeneous graph and the pre-trained heterogeneous graph neural network model to improve the efficiency of building structure component layout design, achieve efficient design of building structure components, and realize the automated transformation of building structure design.

[0065] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative effort.

[0066] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, also by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A building structure component layout design method based on a heterogeneous graph neural network, characterized in that Including: Obtain the building drawing to be processed; Extract the position information of building components and the position information of potential structural components from the building drawing to be processed, and generate a heterogeneous graph based on the position information of the building components and the position information of the potential structural components; The position information of the potential structural components is used to represent that the position where the potential structural components are located can arrange horizontal structural components without affecting the function of the building; the heterogeneous graph is used to represent the building structure as a heterogeneous graph with different types of nodes and different edge connection relationships; the edge connection relationship is determined according to the position relationship between the minimum line segments used to represent each building component and each potential structural component; Input the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a building structure component layout design strategy; Wherein, the building structure component layout design strategy includes whether to arrange structural components for each heterogeneous node in the heterogeneous graph and the type of the arranged structural components; the heterogeneous graph neural network model is trained with a training sample data set composed of building drawings, the corresponding generated heterogeneous graphs and the corresponding actual structural component layout strategies.

2. The method according to claim 1, wherein The extracting the position information of building components and the position information of potential structural components from the building drawing to be processed includes: Extract the position information of building components from the building drawing to be processed and extract the position information of potential structural components from the building drawing to be processed by using a preset rule encoding method; Wherein, the preset rule encoding method includes extending the line segment used to represent the building component to intersect with the outer contour line of the building, or extending any line segment used to represent the building component to intersect with the line segment of other building components; the potential structural component is represented by the extended line segment obtained after intersection.

3. The method according to claim 2, wherein The generating a heterogeneous graph based on the position information of the building components and the position information of the potential structural components includes: Construct different types of nodes based on the minimum line segment used to represent the building component, the minimum line segment used to represent the potential structural component, and the intersection points of all the minimum line segments; the different types of nodes include door nodes, window nodes, wall nodes, potential structural nodes and intersection nodes; the edge connection relationship includes the edge connection relationship between different types of nodes and the edge connection relationship between the same type of nodes; Determine different types of edge connection relationships according to the position relationship between each minimum line segment, and use different types of nodes as heterogeneous nodes and determine the edge relationships of each heterogeneous node based on different types of edge connection relationships to generate the heterogeneous graph.

4. The method according to claim 3, characterized in that Inputting the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a building structure component layout design strategy includes: Perform message passing and feature aggregation on the nodes inside the edge relationship of each type in the heterogeneous graph, update the node features of different types of nodes, and perform feature aggregation on the nodes obtained from different types of edge connection relationships, and update the node features again. The node features of different types of nodes are converted into structural layout features through a multi-layer perceptron network, and information extraction is performed according to the structural layout features to determine whether structural members are arranged for each heterogeneous node in the heterogeneous graph and the types of the arranged structural members.

5. The method according to any one of claims 1 to 4, characterized in that The heterogeneous graph neural network model is trained based on the following steps: Obtain a plurality of architectural drawings, as well as the actual structural member layout strategies corresponding to each architectural drawing, and construct a heterogeneous graph corresponding to each architectural drawing; Use the plurality of architectural drawings, the actual structural member layout strategies corresponding to each architectural drawing, and the heterogeneous graphs corresponding to each architectural drawing as a training sample set to train the network model to be trained; During the training process, use the difference between the predicted layout type of the structural member and the actual layout type of the structural member as the training loss, and iteratively optimize the network model to be trained through a preset loss function to obtain a pre-trained heterogeneous graph neural network model.

6. An architectural structure component layout design device based on a heterogeneous graph neural network, characterized in that, The device includes: An acquisition module for acquiring an architectural drawing to be processed; A generation module for extracting the position information of architectural components and the position information of potential structural components from the architectural drawing to be processed, and generating a heterogeneous graph based on the position information of the architectural components and the position information of the potential structural components; the position information of the potential structural components is used to represent that a horizontal structural component can be arranged at the position where the potential structural component is located without affecting the function of the building; the heterogeneous graph is used to represent the building structure as a heterogeneous graph with different types of nodes and different edge connection relationships; the edge connection relationship is determined according to the position relationship between the minimum line segments used to represent each architectural component and each potential structural component; A design module for inputting the heterogeneous graph into a pre-trained heterogeneous graph neural network model to obtain a design strategy for the layout of building structural components; Wherein, the design strategy for the layout of building structural components includes: whether structural members are arranged for each heterogeneous node in the heterogeneous graph and the types of the arranged structural members; the heterogeneous graph neural network model is: trained with a training sample data set composed of an architectural drawing, the corresponding generated heterogeneous graph, and the corresponding actual structural member layout strategy.

7. The device according to claim 6, wherein The generation module is specifically configured to extract the position information of architectural components from the architectural drawing to be processed and extract the position information of potential structural components from the architectural drawing to be processed by using a preset rule coding method; Wherein, the preset rule coding method includes: extending the line segment used to represent the architectural component until it intersects with the outer contour line of the building, or extending any line segment used to represent the architectural component until it intersects with the line segment of other architectural components; the potential structural component is represented by the extended line segment obtained after intersection.

8. The device according to claim 7, wherein The generation module is specifically configured to construct different types of nodes based on the minimum line segment used to represent the architectural component, the minimum line segment used to represent the potential structural component, and the intersection points of all the minimum line segments; The different types of nodes include: door nodes, window nodes, wall nodes, potential structure nodes, and intersection nodes; the edge connection relationships include: edge connection relationships between different types of nodes and edge connection relationships between nodes of the same type; Specifically, the generation module is further configured to determine different types of edge connection relationships according to the positional relationships between the minimum line segments, use different types of nodes as heterogeneous nodes, and determine the edge relationships of the heterogeneous nodes based on the different types of edge connection relationships, so as to generate the heterogeneous graph.

9. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the building structure component layout design method based on the heterogeneous graph neural network according to any one of claims 1 to 5 are implemented.

10. A computer-readable storage medium, characterized in that, A computer program is stored thereon. When the computer program is executed by a processor, the steps of the building structure component layout design method based on the heterogeneous graph neural network according to any one of claims 1 to 5 are implemented.

Images