Power transmission tower model construction method, device, equipment, medium and product

Abstract

The application discloses a power transmission tower model construction method, device, equipment, medium and product, relates to the technical field of computer graphics, and comprises the following steps: acquiring modeling requirements of a target power transmission tower model; constructing a power transmission tower component topology graph and multiple local subgraphs according to the modeling requirements; inputting the power transmission tower component topology graph and the multiple local subgraphs into a pre-constructed graph embedding model respectively to obtain a whole-tower graph embedding vector and multiple local graph embedding vectors; searching from a pre-constructed power transmission tower model library according to the whole-tower graph embedding vector and the multiple local graph embedding vectors to obtain a search result; and constructing the target power transmission tower model according to the search result. The application solves the technical problem that the existing tower model method weakens the component connection relationship and lacks a local search mode when searching a historical model through point cloud, mesh or multi-view search, and the searched historical model is inaccurate.

Description

Technical Field

[0001] This application relates to the field of computer graphics technology, and in particular to a method, apparatus, equipment, medium and product for constructing power transmission tower models. Background Technology

[0002] Transmission towers are crucial supporting structures in transmission line engineering, and their 3D models are widely used in scenarios such as line design, construction verification, project handover, digital twins, operation and maintenance simulation, and asset management. With the increasing digitalization of transmission line engineering, design institutes and power grid companies have accumulated a large number of historical tower 3D models. These historical models contain rich structural design experience, such as tower leg deployment forms, tower body segmentation methods, main and diagonal material arrangement patterns, crossarm structures, suspension point arrangements, and node connection forms. If it is possible to quickly find tower models similar to the requirements of new projects from existing model libraries, and then modify parameters and reuse local structures based on these models, the efficiency of 3D modeling of transmission towers will be significantly improved.

[0003] In the field of 3D model retrieval, existing methods include point cloud retrieval, mesh retrieval, and multi-view image retrieval. Point cloud, mesh, and multi-view methods typically start from the model's shape or surface geometry, suitable for general 3D shape classification and retrieval, but they struggle to fully preserve the precise topological relationships and engineering semantics of CAD models. The engineering significance of transmission towers lies primarily in the topological connections of components and spatial force paths, rather than simply in their similar shapes. Discretizing tower models into point clouds or meshes easily leads to the loss of assembly relationships, connection nodes, hierarchical affiliations, and local structural semantics between components, hindering subsequent model reuse. Furthermore, in transmission tower design, many cases do not require the reuse of the entire tower; instead, only a specific local structure needs to be reused, such as similar crossarms, similar tower legs, similar tower heads, or similar hanging point connection structures. Traditional whole-tower retrieval methods cannot directly meet the needs of local structural-level reuse.

[0004] In summary, existing pole modeling methods, when retrieving historical models, tend to weaken the connection relationships between components by using point clouds, grids, or multi-view retrieval, and lack local retrieval methods, resulting in technical problems such as inaccurate historical models retrieved. Summary of the Invention

[0005] The purpose of this application is to provide a method, apparatus, equipment, medium, and product for constructing transmission tower models, which can solve the technical problem that existing tower model methods tend to weaken the connection relationship of components when retrieving historical models through point clouds, grids, or multi-view retrieval, and lack local retrieval methods, resulting in inaccurate historical models retrieved.

[0006] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for constructing a transmission tower model, including: Obtain the modeling requirements for the target transmission tower model; Based on the modeling requirements, construct a topology diagram of the transmission tower components and multiple local sub-diagrams; The topology diagram of the transmission tower components and multiple local subgraphs are input into the pre-constructed graph embedding model to obtain the whole tower graph embedding vector and multiple local graph embedding vectors. Based on the whole tower diagram embedding vector and multiple local diagram embedding vectors, the search results are obtained from the pre-built transmission tower model library. Each transmission tower model in the transmission tower model library corresponds to the whole tower diagram embedding vector to be searched and multiple local diagram embedding vectors to be searched. Based on the search results, construct a model of the target transmission tower.

[0007] Optionally, based on modeling requirements, a topology diagram of the transmission tower components and multiple local subgraphs are constructed. Specifically, this includes: determining the engineering components and component connection relationships of the target transmission tower model based on modeling requirements; determining graph nodes based on the engineering components; determining connection edges based on the component connection relationships; constructing a topology diagram of the transmission tower components based on the graph nodes and connection edges; and dividing the topology diagram of the transmission tower components to obtain multiple local subgraphs.

[0008] Optionally, the graph embedding model includes a graph convolutional propagation layer, a relational type attention encoding layer, a component attention pooling layer, and a readout layer connected in sequence; wherein: the graph convolutional propagation layer is used to extract the first node features of the topology graph of the transmission tower components and multiple local subgraphs; the relational type attention encoding layer is used to calculate attention weights to determine the second node features, wherein the first node features are fused with the attention weights to obtain the second node features; the component attention pooling layer is used to determine the key node features based on the second node features, wherein the key node features are the features corresponding to nodes whose node importance scores are greater than a preset score threshold; the readout layer is used to determine the whole tower graph embedding vector and multiple local graph embedding vectors based on the key node features.

[0009] Optionally, based on the whole-tower image embedding vector and multiple local image embedding vectors, a search is performed in the transmission tower model library to obtain search results. Specifically, this includes: determining the similarity between the whole-tower image embedding vector and multiple searchable whole-tower image embedding vectors to obtain multiple first similarities; determining the similarity between each local image embedding vector and multiple searchable local image embedding vectors to obtain multiple second similarities; determining the similarity between the engineering parameters in the modeling requirements and the corresponding engineering parameters of each transmission tower model in the transmission tower model library to obtain multiple third similarities; determining multiple comprehensive similarities based on the multiple first similarities, multiple second similarities, and multiple third similarities; and determining the search results based on the multiple comprehensive similarities.

[0010] Optionally, based on the search results, a target transmission tower model is constructed, specifically including: establishing a mapping relationship between the target transmission tower model to be constructed and similar models, wherein the similar models are included in the search results; transferring the parameters of the similar models to the target transmission tower model to be constructed according to the mapping relationship to obtain the model to be verified; verifying the model to be verified according to preset verification rules to obtain verification results; and determining the model to be verified as the target transmission tower model in response to the verification results indicating that the verification is successful.

[0011] Optionally, the construction of the transmission tower model library includes: acquiring multiple historical transmission tower models; performing data standardization processing on the multiple historical transmission tower models to obtain multiple standardized models; completing the connection relationships of the multiple standardized models to obtain multiple completed standardized models; for each completed standardized model, determining the whole tower diagram embedding vector and multiple local diagram embedding vectors corresponding to each completed standardized model; and constructing the transmission tower model library based on the multiple completed standardized models, the whole tower diagram embedding vector, and the multiple local diagram embedding vectors.

[0012] Secondly, this application provides a transmission tower model construction device, comprising: The system comprises the following modules: an acquisition module for acquiring the modeling requirements of the target transmission tower model; a first construction module for constructing a topology diagram of the transmission tower components and multiple local subgraphs based on the modeling requirements; a determination module for inputting the topology diagram of the transmission tower components and multiple local subgraphs into a pre-constructed graph embedding model to obtain a whole-tower graph embedding vector and multiple local graph embedding vectors; a retrieval module for retrieving data from a pre-constructed transmission tower model library based on the whole-tower graph embedding vector and multiple local graph embedding vectors to obtain retrieval results, wherein each transmission tower model in the transmission tower model library corresponds to a whole-tower graph embedding vector to be retrieved and multiple local graph embedding vectors to be retrieved; and a second construction module for constructing the target transmission tower model based on the retrieval results.

[0013] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the transmission tower model construction method described in any one of the above.

[0014] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the transmission tower model construction method described in any one of the above.

[0015] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the transmission tower model construction method described above.

[0016] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method, apparatus, equipment, medium, and product for constructing transmission tower models. The method first constructs a topology graph of tower components with engineering components as nodes and actual connection relationships as edges according to modeling requirements, and then splits it to generate local subgraphs. Unlike traditional methods that only capture the shape of points and meshes and lose the topology of components, this method completely preserves the assembly relationships and local structural information between components. Then, it generates embedding vectors for the whole tower and local components respectively through graph embedding model. Relying on vector quantification of topological similarity, it simultaneously supports global retrieval of the whole tower and accurate retrieval of local structures, making up for the shortcomings of existing methods that can only perform global matching and cannot reuse local templates. The model library pre-stores the dual embedding vectors corresponding to all towers. During comparison, it simultaneously integrates the multi-dimensional similarity screening results of the whole tower and local components, which greatly improves the accuracy of structural matching. Finally, it completes the construction of the target tower model based on the high similarity retrieval results, thereby solving the technical problem that existing tower model methods tend to weaken the connection relationships of components and lack local retrieval methods when retrieving historical models through point clouds, meshes, or multi-view retrieval, resulting in inaccurate historical models retrieved. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is an application environment diagram of a method for constructing a transmission tower model according to an embodiment of this application; Figure 2 A flowchart illustrating a method for constructing a power transmission tower model, provided as an embodiment of this application; Figure 3 A schematic diagram of the architecture of a power transmission tower model building system provided in one embodiment of this application; Figure 4 A flowchart illustrating a method for constructing a power transmission tower model, provided as an embodiment of this application; Figure 5 This is a schematic diagram of an embedding encoding process provided in an embodiment of this application; Figure 6 This is a schematic diagram of a transmission tower model building device provided in one embodiment of this application; Figure 7 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] The transmission tower model construction method provided in this application embodiment can be applied to, for example, Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on another server.

[0022] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, and IoT devices. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers, or it can be a cloud server.

[0023] In one exemplary embodiment, such as Figure 2 As shown, a method for constructing a transmission tower model is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps 201 to 205. Wherein: Step S201: Obtain the modeling requirements for the target transmission tower model.

[0024] In this embodiment of the application, step S201 is a preliminary data acquisition step for modeling and retrieval, used to clarify the engineering constraints and structural design requirements of the target transmission tower to be designed.

[0025] For example, the specific execution method is to read the tower design parameters, local structural reuse requirements, and three-dimensional model base files input by the design terminal; the modeling requirements include global engineering parameters such as voltage level, number of circuits, nominal height, tower type, number of crossarm layers, and number of hanging points, and may also include local modeling requirements for reusing only a single local structure such as tower leg / crossarm / tower head, while collecting the original data of the three-dimensional geometric coordinates, component specifications, and connection assembly relationships of the already formed components of the target tower.

[0026] Understandably, the purpose of step S201 is to define the search scope and reuse targets, providing an engineering basis for the subsequent targeted construction of topology graphs and local subgraphs.

[0027] Step S202: Based on the modeling requirements, construct the topology diagram of the transmission tower components and multiple local sub-diagrams.

[0028] In this embodiment of the application, step S202, based on the modeling requirements obtained in step S201, performs component topology diagram construction and multi-granularity local sub-graph splitting processing, which is different from the existing point cloud, mesh, and multi-view processing methods that only extract the shape geometry data.

[0029] Component topology diagram construction: Abandoning the common modeling logic based on faces and vertices as basic units, this method uses engineering load-bearing components such as main materials, diagonal materials, crossarms, tower legs, hanging points, and node plates as diagram nodes. Assembly relationships between components, such as bolt connections, node plate connections, hierarchical affiliation, and spatial support, are defined as diagram edges. Multi-dimensional node attributes such as component material, cross-section, spatial orientation, height, and whether it is a critical reusable component are simultaneously written, as well as edge attributes such as component angle and connection strength, to form a complete component topology diagram that represents the topological relationships of the tower project. This fully preserves the real assembly connection relationships between components and solves the defect of weakening component connections in existing solutions.

[0030] Multi-granularity local subgraph decomposition: Based on the local reuse requirements in the modeling needs and combined with the tower engineering component division rules, independent local subgraphs such as tower legs, tower body, crossarms, tower heads, and hanging points are segmented from the complete component topology diagram. The components connecting the tower body and foundation within each subgraph are marked as interface nodes, and an interface identifier field is added to the subgraph attributes. Each local subgraph independently stores the node, edge, and attribute set, realizing dual-dimensional data structuring of the whole tower and local structures, providing a data carrier for specialized retrieval of local structures.

[0031] Step S203: Input the topology diagram of the transmission tower components and multiple local subgraphs into the pre-constructed graph embedding model to obtain the whole tower graph embedding vector and multiple local graph embedding vectors.

[0032] In this embodiment, step S203 utilizes an engineering-specific graph embedding model trained offline to complete the vectorized encoding of the graph structure, thereby realizing the digital quantitative representation of the topological structure.

[0033] The pre-built graph embedding model sequentially connects the basic graph convolutional layer, the relation type attention encoding layer, the key component attention pooling layer, and the global readout layer. It completes supervised and self-supervised offline training through a composite loss function that includes tower type classification, whole tower triples, local structure, and engineering rules. It can adapt to the topological features of power transmission truss components and prioritizes learning reusable key structures such as main materials, interfaces, and hanging points.

[0034] During execution, the complete component topology graph generated in step S202 is input into the forward inference link of the graph embedding model. After graph convolution to aggregate basic component features, relational attention to distinguish different assembly connection weights, key component pooling to filter highly reusable nodes, and global readout weighted compression, a fixed-dimensional whole tower graph embedding vector is output. Then, each local subgraph obtained by splitting is input into the graph embedding model with shared weights to generate a local graph embedding vector for each local structure. Understandably, whole-tower and local embedded vectors transform complex graph topology into low-dimensional vectors that can be numerically computed, providing standardized data for subsequent quantitative calculations of similarity and enabling quantifiable comparisons of structural similarity.

[0035] Step S204: Based on the whole tower image embedding vector and multiple local image embedding vectors, retrieve from the pre-built transmission tower model library to obtain the retrieval results. Each transmission tower model in the transmission tower model library corresponds to the whole tower image embedding vector to be retrieved and multiple local image embedding vectors to be retrieved.

[0036] In this embodiment of the application, step S204 completes similarity retrieval based on the pre-encoded model library vector set, simultaneously supporting two modes: whole-tower reuse retrieval and local structure separate retrieval.

[0037] The pre-built transmission tower model library stores multiple historical tower 3D models. Each historical tower data in the library is pre-executed offline using the complete S201–S203 process to generate the corresponding whole tower image embedding vector and all local images to be retrieved, such as tower legs and crossarms, and these are stored persistently. This eliminates the need for repeated online coding and improves retrieval efficiency.

[0038] For example, during the retrieval process, the cosine similarity of the target tower with the towers in the database is calculated using the whole-tower image embedding vector and multiple local image embedding vectors, as well as the weighted cosine similarity of each local structure. The matching degree of engineering parameters such as voltage level and tower type is then superimposed, and a constraint conflict penalty term is introduced to calculate the comprehensive similarity score. The towers in the database are then sorted from highest to lowest score, and the Top-K retrieval results are output. If the modeling requirement is whole-tower reuse, the whole-tower similarity is used as the core weight; if only local structures such as crossarms and tower legs need to be reused, the weight of the corresponding local vector similarity is increased, and local templates are selected separately. This compensates for the shortcomings of existing technologies in lacking local retrieval capabilities, balancing structural topological similarity with engineering applicability.

[0039] Step S205: Based on the search results, construct the target transmission tower model.

[0040] In this embodiment of the application, step S205, based on the high similarity pole / local template retrieval results output in step S204, completes parameterized transfer, structural adaptation and constraint verification, and generates a compliant target 3D pole model.

[0041] In constructing the target transmission tower model, a one-to-one matching relationship is established between the target components and the retrieved template components. Matching scores are calculated based on node embedding similarity, component type, and spatial adjacency, completing the corresponding binding of main materials, crossarms, and hanging points. Then, the template component lengths, cross-sections, endpoint coordinates, and connection topology are migrated to the target frame. Scaling, translation, and rotation transformations are performed on design differences such as nominal height and crossarm length to adapt to the target modeling size requirements. Next, the assembled tower structure undergoes checks on component connectivity, interface matching, number of hanging points, spatial collisions, and parameter ranges, automatically adjusting and correcting structures that do not meet engineering specifications. Finally, after all checks pass, all component parameters and connection relationships are integrated to generate a complete 3D model of the target transmission tower, simultaneously outputting a reuse check report. This achieves automated connection from similarity retrieval to the finished model, eliminating repetitive manual modeling work.

[0042] Based on the description of the above embodiments, it can be understood that the method provided by the present invention first constructs a topological graph of tower components with engineering components as nodes and actual connection relationships as edges according to modeling requirements, and then splits it to generate local subgraphs. This differs from the traditional methods of point clouds and meshes, which only capture the shape and lose the topology of components, and completely preserves the assembly association and local structural information between components. Then, it generates embedding vectors corresponding to the whole tower and the local area through graph embedding models. Relying on vector quantification of topological similarity, it simultaneously supports global retrieval of the whole tower and accurate retrieval of local structures, making up for the shortcomings of existing methods that can only perform global matching and cannot reuse local templates. The model library pre-stores the dual embedding vectors corresponding to all towers. During comparison, it simultaneously merges the multi-dimensional similarity screening results of the whole tower and the local area, which greatly improves the accuracy of structural matching. Finally, it completes the construction of the target tower model based on the high similarity retrieval results, thereby solving the technical problem that existing tower model methods tend to weaken the connection relationship of components and lack local retrieval methods when retrieving historical models through point clouds, meshes or multi-view retrieval, resulting in inaccurate historical models retrieved.

[0043] Optionally, in step S202, a topology diagram of the transmission tower components and multiple local sub-diagrams are constructed according to modeling requirements, specifically including the following steps: Step S2021: Based on the modeling requirements, determine the engineering components and component connection relationships of the target transmission tower model.

[0044] In this embodiment, the design constraints of tower type, voltage, crossarm, tower leg, and suspension point in the modeling requirements are analyzed. All professional engineering components included in the target tower, such as main materials, diagonal materials, auxiliary materials, tower leg components, crossarm components, tower head components, node plates, conductor suspension points, ground wire suspension points, and insulator supports, are screened, and decorative and auxiliary geometric surfaces with no engineering reuse value are eliminated. Then, the component endpoint coordinates and assembly records in the 3D model are read, and the missing explicit connections in the model are automatically filled in by combining the preset endpoint distance threshold. Different connection types, such as common node connections, node plate bolt connections, welded connections, hierarchical subordination, and spatial support, are distinguished, and the connection objects, connection positions, connection strengths, spatial angles, and other related information of each group of components are recorded one by one.

[0045] Understandably, step D2021 completes the engineering screening, completion, and classification of component entities and connection relationships, providing complete and logically consistent basic engineering data for the subsequent generation of graph nodes and connection edges.

[0046] Step S2022: Determine the drawing nodes based on the engineering components.

[0047] In this embodiment, a unique node identifier is generated for each engineering component, and a corresponding multi-dimensional node attribute vector is generated. The node attribute vector contains engineering feature parameters such as component type code, material code, cross-sectional specifications, component length, spatial unit direction vector, installation height, tower section area, and whether it is a key reusable component identifier for main material / hanging point / interface. Different types of engineering components correspond to differentiated node labels, which can distinguish main material nodes, diagonal material nodes, crossarm nodes, tower leg nodes, hanging point nodes, node plate nodes, etc., so that the nodes naturally carry power engineering semantics, ensuring that the subsequent graph embedding model learning focuses on the features of reusable components.

[0048] For example, engineering components in the 3D model of a transmission tower are abstracted as graph nodes. Node types include main member nodes, diagonal member nodes, horizontal member nodes, auxiliary member nodes, tower leg nodes, crossarm nodes, tower head nodes, gusset plate nodes, hanging point nodes, and insulator connection nodes. Each node corresponds to a solid component or an engineering semantic object.

[0049] No. The attribute vector of a component node can be represented as: In the formula, Indicates the first The attribute vector of each node; Indicates component type encoding; Indicates material code; Indicates the cross-sectional specification code; Indicates the length of the component; Represents the spatial direction vector of a component; Indicates the height of the component; Indicates the local region code to which the component belongs; An identifier indicating whether a component is a critical load-bearing component or a reusable formwork component.

[0050] The formula for calculating the spatial direction vector of a component is as follows: In the formula, Indicates the first The unit direction vector of each rod; and These represent the coordinates of the two endpoints of the member; Indicates the length of the rod.

[0051] Step S2023: Determine the connection edge based on the component connection relationship.

[0052] In this embodiment, based on the component connection relationship obtained from step S2021, corresponding connection edges are generated between two graph nodes with assembly association, while distinguishing multiple types of connection edges. Each connection edge corresponds to a unique edge attribute vector, which records information such as connection type code, component spatial angle, connection reliability level, whether it is connected through a node plate, and whether the two components belong to the same local structural unit. Connection edges are divided into six categories: common node connection edges, bolt node plate connection edges, welded connection edges, hanging point connection edges, hierarchical belonging edges, and spatial support edges. Hierarchical belonging edges are specifically used to characterize the hierarchical belonging relationship between the main material and its tower leg, crossarm, and tower head, fully preserving the layered engineering structure of the tower and making up for the deficiency of general geometric diagrams in expressing hierarchical semantics.

[0053] For example, the connection relationships between components are abstracted as graph edges. Edge types include common node connection edges, node plate connection edges, bolted connection edges, welded connection edges, hanging point connection edges, spatial support edges, and hierarchical belonging edges.

[0054] No. The node and the first The edge attributes between nodes can be represented as: In the formula, Represents a node With nodes The edge attribute vector between them; Indicates the type of connection relationship; Indicates the included angle between two components; Indicates the connection strength level or connection reliability level; Indicates whether the connection is via a node plate; Indicates whether they belong to the same local structural unit.

[0055] The formula for calculating the included angle of components is: .

[0056] Step S2024: Construct a topology diagram of the transmission tower components based on the graph nodes and connecting edges.

[0057] For example, the topology diagram of a transmission tower component is represented as follows: In the formula, This diagram shows the topology of the transmission tower components. Represents a set of nodes; Represents the set of edges; Represents the node attribute matrix; Represents the edge attribute matrix; It represents a set of engineering hierarchical relationships, including information on the level to which components belong, such as tower legs, tower body, crossarms, and tower heads.

[0058] Understandably, this step does not use the "face-edge-vertex" low-level geometry of a general model, but instead establishes a "component-connection-hierarchy" topology graph for reuse in transmission tower engineering, so that graph nodes directly correspond to reusable objects in the project. The component topology graph uses power engineering components as the smallest unit, fully carrying engineering information such as component geometry, materials, specifications, stress, assembly, and hierarchy. It avoids the problems of blurred component boundaries and lost connection relationships found in point clouds and mesh schemes, and can accurately represent the true topological similarity of the tower truss structure.

[0059] Step S2025: Divide the topology diagram of the transmission tower components to obtain multiple local sub-diagrams.

[0060] For example, in this embodiment of the invention, after constructing the topology diagram of the entire tower components, the entire tower diagram is divided into multiple local sub-diagrams according to the engineering parts and connection relationships. The local sub-diagrams include tower leg sub-diagrams, tower body sub-diagrams, crossarm sub-diagrams, tower head sub-diagrams, hanging point sub-diagrams, and node connection sub-diagrams.

[0061] Any local structure The subgraph is represented as: In the formula, Represents a local subgraph; Represents the set of nodes in a local subgraph; Represents the set of edges in a local subgraph; Represents the attribute matrix of local subgraph nodes; This represents the edge attribute matrix of a local subgraph.

[0062] When the model already has part labels, the system directly extracts local sub-graphs based on the part field. When the model lacks part labels, the system automatically categorizes them based on component height, spatial orientation, connection density, hanging point location, and component type. For example, the tower leg sub-graph consists of components located at the bottom of the tower, connected to the foundation node, and extending upwards to the main tower body; the crossarm sub-graph consists of components located at a specified height, extending laterally, and containing conductor hanging points; and the tower head sub-graph consists of components located on the upper part of the tower body and connected to the ground wire support or upper crossarm.

[0063] To improve the reusability of local structures, this invention adds interface node markers to the local sub-graphs. An interface node is a node connecting the local structure to other structures of the entire tower. For example, the tower body connection point in the crossarm sub-graph is an interface node, and the foundation connection point and tower body connection point in the tower leg sub-graph are also interface nodes. The interface node attributes are denoted as: In the formula, Represents a node The interface tag. This tag can be used as a node attribute input graph embedding model to enhance the interface matching capability when embedding local structure templates into the target model.

[0064] This step adds local subgraph retrieval in addition to whole tower retrieval and introduces interface node markers, so that the retrieval results are not only structurally similar, but also easier to directly embed into the target tower model.

[0065] Optionally, the graph embedding model includes a graph convolutional propagation layer, a relation type attention encoding layer, a component attention pooling layer, and a readout layer connected in sequence.

[0066] The process involves: a graph convolutional propagation layer to extract the first node features from the topology graph of the transmission tower components and multiple local subgraphs; a relational type attention encoding layer to calculate attention weights to determine the second node features, where the first node features are fused with the attention weights to obtain the second node features; a component attention pooling layer to determine key node features based on the second node features, where key node features are the features corresponding to nodes whose node importance scores are greater than a preset score threshold; and a readout layer to determine the whole tower graph embedding vector and multiple local graph embedding vectors based on the key node features.

[0067] In this embodiment, the graph convolutional propagation layer takes as input the node attribute matrix and normalized self-connected adjacency matrix corresponding to the topology graph or local subgraph of the transmission tower component to be encoded, i.e., the original attribute features of each component and the basic topology adjacency relationship. Then, based on the graph convolutional basic normalized aggregation operation, the original attribute information of each component node and its first-order adjacent components is aggregated to complete the global basic topology feature fusion, generating initial node features without distinguishing connection types, denoted as the first node feature; this layer only relies on the spatial adjacency relationship of components to complete coarse feature extraction, without distinguishing differentiated assembly relationships such as bolted connections and hierarchical affiliation, and only completes the unified aggregation of the bottom layer structural information.

[0068] The relation type attention encoding layer takes into account all first-node features output from the graph's convolutional propagation layer, the set of various connection relationships within the graph, and the edge attribute vector for each edge. Then, for different connection types such as shared nodes, node plates, hanging points, and layer affiliation, independent learnable transformation weights are configured for each type of connection. For any central node and its different types of neighboring nodes, the attention weights after LeakyReLU activation are calculated by concatenating features with edge attribute vectors. The differentiated attention weights are used to weight and sum the first-node features of the corresponding neighbors, superimposed with the node's own transformation features, and the importance information of different assembly connections is fused to generate refined second-node features. That is, the updated second-node features are obtained by fusing the first-node features with relation-based attention weights. The key component attention pooling layer takes into account all second-node features output from the relation type attention encoding layer, the component type corresponding to each node, installation height, topology degree, and main material / interface / hanging point engineering identifiers. Then, the node importance scores are calculated from three dimensions: topology, component attributes, and power engineering rules. The weighted sum is then used to obtain the comprehensive importance score for each component node. The comprehensive importance score is compared with a preset score threshold. Only the node features with scores greater than the threshold are retained as key node features. Secondary auxiliary material nodes with no reuse value are filtered out, and redundant feature data is compressed.

[0069] The readout layer takes as input all key node features and globally learnable readout weight vectors from the attention pooling layer of the component. Then, a weighted global readout mechanism is used to perform weighted summation and aggregation on the selected key node features. If the input is a complete transmission tower component topology diagram, the output is a fixed-dimensional whole tower diagram embedding vector; if the input is a single local subgraph such as a tower leg or crossarm, the output is a local diagram embedding vector of the corresponding dimension.

[0070] Understandably, the original component topology graph is only structured data of discrete nodes and connections, which cannot be directly compared numerically. This model extracts topological features coarsely through graph convolutional propagation layers, then distinguishes assembly connection weights through relational attention layers to obtain fine node features, uses component pooling layers to screen key engineering components, and finally compresses them into standardized embedding vectors by the readout layer. This fully preserves the topology of the tower truss and the semantics of power engineering reuse, overcoming the shortcomings of existing technologies such as point clouds, meshes, and general graph encoding that lose component connections and ignore local reuse value. It provides a reliable quantitative feature foundation for high-precision whole-tower / local dual-dimensional model retrieval.

[0071] For example, the topology graph of the transmission tower components is input into the model. The basic graph convolution propagation formula is: In the formula, Indicates the first Layer node feature matrix; Indicates the first Layer node feature matrix; Represents the graph adjacency matrix; This represents the adjacency matrix after adding self-connections; Represents the identity matrix; express The corresponding degree matrix; Indicates the first Layer-learnable weight matrix; This represents a non-linear activation function. This formula enables each component node to aggregate connection information from adjacent components when updating its features.

[0072] Considering the various connection relationships within transmission towers, this invention preferably employs graph attention encoding with relationship types. Node In the The feature update formula for the layer is: In the formula, Represents a node In the The feature vector of the layer; Represents a node In the The feature vector of the layer; Represents a collection of connection relationship types; Indicates by relation type With nodes The set of adjacent nodes; Indicate neighboring nodes; Represents the feature transformation matrix of the self-node; Representing relation type The corresponding feature transformation matrix; Represents a node In relation types Next pair of nodes Attention weights.

[0073] The formula for calculating attention weights is: In the formula, Representing relation type The corresponding attention parameters; Represents a node With nodes The edge attribute vector between them; This indicates vector concatenation; Represents a node In relationship Any of its neighboring nodes; This represents the activation function of the linear unit with leakage correction.

[0074] To highlight key components such as main materials, crossarm interfaces, hanging points, and tower leg interfaces, this invention introduces a key component attention pooling operation. This operation borrows the idea from graph pooling of retaining key nodes based on their importance, with node importance scores as follows: In the formula, Represents a node The overall importance score; Indicates the topological importance score; This indicates the attribute importance score; Indicates the importance score of the engineering rules; This represents the weighting coefficient.

[0075] Among them, the topology importance score can be obtained based on the node degree and connection relationship type; the attribute importance score can be obtained by the node attributes through the fully connected layer; the engineering rule importance score is used to increase the weight of key nodes such as main materials, interface nodes, hanging point nodes and tower leg foundation nodes.

[0076] The importance of engineering rules can be expressed as: In the formula, Represents a node Is it a main material node? Represents a node Is it an interface node? Represents a node Is it a hanging node? Represents a node Is it a critical node for the tower leg? This represents the corresponding weighting coefficient.

[0077] After passing through multiple layers of neural networks, attention readout is performed on all nodes to obtain the whole tower graph embedding vector: In the formula, Represents the embedding vector of the entire tower diagram; This indicates the number of nodes in the entire tower diagram; Represents a node Readout weights; Represents a node Feature vectors in the last layer of the graph embedding model; This represents the total number of layers in the graph embedding model.

[0078] Read out weights It can be represented as: In the formula, This represents the learnable readout weight vector; This represents any node in the graph.

[0079] For local subgraphs such as tower legs, crossarms, tower heads, and hanging points, local graph embedding vectors are generated using the same or parameter-shared graph embedding model encoder: In the formula, Representing local structure Graph embedding vectors; Representing a local subgraph The set of nodes; Represents a node Readout weights in a local subgraph.

[0080] Optionally, in some embodiments of the invention, the training process of the graph embedding model is as follows: If the model library already contains a tower-type category, similar model pairs, and reuse records, then supervised training is used. The training loss function is: In the formula, Represents the total loss function; Indicates the loss in the tower classification; This represents the triplet loss for overall tower similarity; This represents the loss of local structural similarity. This indicates a loss of consistency in engineering rules; This represents the weighting coefficient of each loss term.

[0081] The loss of the triplet is: In the formula, The graph embedding vector representing the anchor tower model; This represents the positive sample image embedding vector similar to the anchor point model; Represents the negative sample graph embedding vector that is dissimilar to the anchor point model; Represents the vector distance function; This represents the interval parameter.

[0082] To improve the model's generalization ability under limited sample conditions, this application introduces a self-supervised augmentation training operation. Positive samples are generated by slightly perturbing the same tower diagram, such as randomly occluding a few auxiliary material nodes, slightly scaling the length of non-critical components, changing the input order of perturbed components, and keeping the main material and interface nodes unchanged. Diagrams with different tower types or different crossarm arrangements are used as negative samples. The purpose of the self-supervised augmentation training operation is to enable the model to learn stable structural similarities, rather than over-relying on certain non-critical auxiliary components. This operation enhances the model's ability to stably identify main materials, interface nodes, and overall topological relationships, improving retrieval generalization and robustness under limited sample conditions.

[0083] Optionally, in step S204, the retrieval results are obtained from the transmission tower model library based on the whole tower diagram embedding vector and multiple local diagram embedding vectors, specifically including the following steps: Step S2041: Determine the similarity between the whole tower image embedding vector and multiple whole tower image embedding vectors to be retrieved, and obtain multiple first similarity scores.

[0084] In this embodiment, step S2041 is used to complete the global tower topology similarity calculation. Using the whole-tower image embedding vector of the target tower as the global query vector, the embedding vectors of the whole-tower image to be retrieved for each tower in the model library are sequentially traversed. The cosine value of the angle between the two vectors is calculated using the cosine similarity formula, and this is used as the first similarity of the corresponding sample. The first similarity characterizes the degree of similarity between the target tower and the towers in the library in terms of overall truss topology, main material arrangement, and overall tower structure. The closer the value is to 1, the higher the global structural consistency. After traversing all tower samples in the model library, multiple sets of first similarities are generated, each corresponding to a specific tower, to characterize the overall topology matching degree between all towers in the library and the target tower.

[0085] For example, for the model to be retrieved and the first in the model library For each candidate model, the overall similarity is calculated using cosine similarity: In the formula, Indicates the model to be retrieved With candidate models The similarity of the entire tower; Represents the whole-pyramid graph embedding vector of the model to be retrieved; This represents the whole-tower graph embedding vector of the candidate model.

[0086] Step S2042: Determine the similarity between each local graph embedding vector and the multiple local graph embedding vectors to be retrieved, and obtain multiple second similarities.

[0087] In this embodiment, step S2042 is used to calculate the dimensional similarity of local structures such as tower legs, crossarms, and tower heads, supporting the local template reuse retrieval requirements. It iterates through the embedding vectors of each type of local graph of the target tower (crossarm, tower leg, tower head, and hanging point vectors); for the same type of local structure, it calculates the cosine similarity between the target local vector and the embedding vectors of the same type of local graph to be retrieved in the database; it configures weight coefficients according to the reuse importance of different local structures in tower design, and weights the local similarities of the same type to obtain the comprehensive local similarity corresponding to a single base database tower, which is denoted as the second similarity of the sample; after traversing all towers in the database, it outputs multiple sets of corresponding second similarities. The second similarity specifically characterizes the matching degree of local component topology, interface layout, and hanging point structure, solving the technical shortcoming of the prior art's lack of local structure retrieval capabilities.

[0088] For example, the local structural similarity is: In the formula, Indicates the overall similarity of local structures; This represents a collection of local structures, including tower legs, crossarms, tower heads, and hanging points; Representing local structure The weights; Represents the local structure in the model to be retrieved The embedding vector; This represents the embedding vector of the corresponding local structure in the candidate model.

[0089] Step S2043: Determine the similarity between the engineering parameters in the modeling requirements and the corresponding engineering parameters of each transmission tower model in the transmission tower model library, and obtain multiple third similarity scores.

[0090] In this embodiment, step S2043 is used to quantify the degree of engineering condition adaptability and filter invalid tower samples with similar structures but mismatched design conditions. The set of engineering parameters carried by the modeling requirements in step S201 is extracted, including voltage level, loop type, nominal high interval, number of crossarm layers, design wind speed, icing thickness, and tower function type (straight / tension / angle), etc. The corresponding engineering parameters stored for each tower in the model library are read, and the parameter matching is compared item by item. A matching weight is assigned to each parameter: a high score (70-100 points, excluding 70) for complete parameter matching, a medium score (40-70 points, excluding 40) for parameter interval compatibility, and a low score (0-40 points) for parameter conflict. The weighted sum is used to obtain the engineering parameter matching score of a single tower, i.e., the third similarity. All towers in the library are traversed to generate multiple sets of third similarities, used to filter out towers with similar topology but inconsistent engineering usage conditions, improving the engineering practicality of the search results.

[0091] For example, engineering parameter matching is used to avoid retrieving models that are structurally similar but whose engineering conditions are not applicable. Engineering parameter matching can be expressed as: In the formula, Indicates the degree of matching of engineering parameters; Indicates the number of engineering parameters; Indicates the first Weights of individual engineering parameters; This indicates that the model to be retrieved and the candidate models are in the _____th ... The degree of matching in terms of engineering parameters. Engineering parameters include voltage level, tower type, number of circuits, nominal height, number of crossarm layers, number of hanging points, and applicable operating conditions.

[0092] Step S2044: Determine multiple comprehensive similarities based on multiple first similarities, multiple second similarities, and multiple third similarities.

[0093] In this embodiment, taking a single tower sample within the database as a unit, the similarity is fused from three categories: global topology, local structure, and engineering parameters. A constraint conflict penalty term is also added to calculate the final comprehensive similarity. The calculation logic is as follows: In the formula, Indicates the final overall similarity; Indicates a constraint conflict penalty item; This represents the weighting coefficient. Constraint conflicts include inconsistencies in voltage levels, number of mounting points, tower type and function, number of crossarm layers, and number of interface nodes.

[0094] Understandably, a comprehensive similarity score is calculated independently for each tower sample in the model library, forming a comprehensive similarity set that corresponds one-to-one with each sample in the library. The higher the comprehensive similarity score, the more accurately the tower's global structure, local templates, and engineering conditions meet the target modeling requirements.

[0095] Step S2045: Determine the search results based on multiple comprehensive similarities.

[0096] In this embodiment, all tower samples in the model library are sorted in descending order of their comprehensive similarity from high to low, and the top-K tower samples are selected as the final search results. The search results simultaneously output the 3D model file, component parameter table, various local structure templates, engineering parameter descriptions, and detailed similarity information for each tower. Two types of results are distinguished: recommended samples for whole-tower reuse and templates for local structure reuse, for designers to choose from. If there is only a need for local reuse, the second similarity weight can be increased and the order reordered, and the optimal crossarm / tower leg local template can be output separately, taking into account both whole-tower construction and local renovation engineering scenarios.

[0097] Optionally, in step S205, a target transmission tower model is constructed based on the search results, specifically including the following steps: Step S2051: Establish the mapping relationship between the target transmission tower model to be constructed and the similar model, wherein the similar model is included in the search results.

[0098] In this embodiment, after the similarity model is determined, a component mapping relationship is established between the model to be generated and the similar model. The component matching score is: In the formula, Represents the target component node Similar model component nodes Match score; and These represent the embedding vectors of the two component nodes, respectively. Indicates cosine similarity; Indicates whether the component types are consistent; if they are consistent, return 1; otherwise, return 0. Indicates the degree of adjacency matching; Indicates the degree of spatial location matching; This represents the weighting coefficient.

[0099] Step S2052: Based on the mapping relationship, transfer the parameters of the similar model to the target transmission tower model to be constructed to obtain the model to be verified.

[0100] In this embodiment of the application, step S2052 is used to migrate the component length, cross-sectional specifications, node coordinates, connection relationships and local structural templates in the similar model to the target model according to the component mapping relationship.

[0101] For length parameters, scaling can be done proportionally; for coordinate parameters, transformations can be performed through translation, rotation, and scaling; for connection relationships, the topological connection patterns in similar models should be preserved first.

[0102] The formula for local template coordinate transformation is: In the formula, Indicates the original coordinates of the endpoints of the components in the template; Indicates the transformed target coordinates; Represents the rotation matrix; This represents the translation vector.

[0103] Step S2053: Verify the model to be verified according to the preset verification rules and obtain the verification result.

[0104] In this embodiment, model verification is required after parameter transfer. Preset verification rules include component connectivity verification, interface node verification, attachment point quantity verification, component spatial collision verification, parameter range verification, and local structural closure verification.

[0105] Connectivity Verification: Traverse the entire tower's component topology diagram, count the total number of traversable nodes and the total number of nodes in all components. If the structure is determined to have no breaks or isolated components, the verification fails. Interface Matching Verification: Compare the number of interface nodes in the local template with the target design number of interfaces, and check if the spatial deviation of the interfaces is less than the allowable threshold. Hanging Point Quantity Verification: Verify if the number of hanging points on each floor's crossarm matches the target engineering parameter requirements. Spatial Collision Verification: Detect the minimum distance between any two components. If it is less than the preset safety distance, structural interference is determined. Parameter Range Verification: Verify if the main material, diagonal material cross-section, and member length are within the design allowable range. If all verification items meet the standards, the verification pass result is output; if any verification fails, the verification fails, and the conflicting components, conflict locations, and conflict types are marked.

[0106] The component connectivity check requires that the reused whole tower diagram meet the following requirements: In the formula, This indicates the number of nodes that can be traversed starting from any main material node; This represents the total number of nodes in the entire tower diagram. If the two are equal, it indicates that the topology of the entire tower components is connected; if they are not equal, it indicates that there are isolated components or broken structures.

[0107] Step S2054: In response to the verification result indicating that the verification is successful, the model to be verified is determined as the target transmission tower model.

[0108] In this embodiment, the verification result output in S2053 is read. If all preset verification rules are met, there are no structural conflicts, and no parameter violations, the current model to be verified is directly solidified into the final usable 3D model of the target transmission tower. A complete component parameter table, component connection table, reuse source description, and verification report are simultaneously output and archived. If the verification result contains items that fail verification, the process returns to S2052 to readjust the component coordinates, member dimensions, and local template positions, regenerate the model to be verified, and repeat the verification process until all verifications pass.

[0109] Optionally, the construction of the transmission tower model library includes: Step S204a: Obtain multiple historical transmission tower models.

[0110] In this embodiment, step S204a is a data acquisition pre-step for building the model library. Multiple sets of original 3D model files of historical transmission towers stored in the power design platform and historical project archives are read in batches. The models cover various tower types such as straight-line towers, tension towers, and angle towers, and cover different voltage levels, icing conditions, and circuit configurations. Simultaneously, the original engineering parameter tables, component lists, and assembly drawings corresponding to each set of models are read.

[0111] It should be noted that the original historical models were from scattered sources, the modeling software was not uniform, and the data formats were different. They had original defects such as redundant component panels, missing assembly annotations, and incomplete connection relationship records, which need to be uniformly processed.

[0112] Step S204b involves standardizing the data of multiple historical transmission tower models to obtain multiple standardized models.

[0113] In this embodiment, standardized cleaning is performed on the original historical model with inconsistent formats, components, and parameters to eliminate heterogeneous data differences.

[0114] Standardization processing specifically includes three aspects: format unification and conversion, which converts model files output by different 3D modeling software into a unified data storage format and unifies component coordinate references and spatial coordinate systems; redundant geometry removal, which deletes decorative panels and auxiliary lines that are only used for visualization and have no engineering stress reuse value, and retains only standard engineering components such as main materials, diagonal materials, tower legs, crossarms, node plates, and hanging points; and parameter standardization coding, which unifies the coding rules for engineering parameters such as component material, cross section, tower type, voltage, and height, and converts text descriptions into calculable numerical codes.

[0115] After the above standardized cleaning, multiple standardized models with unified data format, simplified components, and standardized parameter coding are output, eliminating heterogeneous interference from the original model.

[0116] For example, models from different sources in the model library are processed uniformly. Data standardization includes: unified coordinate system, unified length units, unified component naming, unified cross-sectional specifications, unified component numbering, and unified connection relationship format. For example, "main pole," "main material," and "tower body main material" are unified as "main material"; "web member," "diagonal member," and "diagonal material" are unified as "diagonal material"; and the model coordinates are unified to a local coordinate system with the tower centerline as the Z-axis and the foundation center as the origin.

[0117] Step S2042c involves completing the connection relationships of multiple standardized models to obtain multiple completed standardized models.

[0118] In this embodiment of the application, the standardized model has problems such as omissions in drawing input, lack of annotation of component assembly connections during modeling, and loss of hierarchical subordinate relationships. Step S2042c is used to identify and complete all component connection relationships.

[0119] The completion logic is as follows: extract the 3D coordinates of all engineering component endpoints and set a threshold for component connection determination; if the spatial distance between the endpoints of two components is less than this threshold, it is determined that the two have an assembly relationship; further distinguish five connection types: common node connection, node plate bolt connection, welding, hierarchical subordination, and spatial support, and automatically generate complete connection records; at the same time, supplement the spatial angle and connection strength attributes between components, complete the missing hierarchical subordination relationships in the model, and establish the hierarchical association edges between the main material and its subordinate tower legs, crossarms, and tower heads. After all connection relationships are completed, a completed standardized model with complete component information and a complete assembly topology is obtained, solving the problem of missing connection information and inability to construct a complete component topology diagram in the original model.

[0120] For example, if explicit connections are missing in the model file, they are automatically filled in using geometric proximity and engineering assembly relationships. For two components... and If the minimum distance between any two endpoints is less than a preset threshold, then a geometric connection is determined to exist between them. The formula for calculating the endpoint distance is: In the formula, Representing components With components Minimum endpoint distance between; Representation of components The One endpoint coordinate; Representation of components The One endpoint coordinate; Indicates Euclidean distance; Indicates the endpoint number of the component. If Then the component is considered With components There is a connection relationship, where The threshold for identifying endpoint connections can be set according to the modeling accuracy.

[0121] Step S204d: For multiple complete and normalized models, determine the whole graph embedding vector and multiple local graph embedding vectors corresponding to each complete and normalized model.

[0122] In this embodiment of the application, graph structure encoding is performed offline for each completion standardization model, and the vector data required for retrieval is pre-stored.

[0123] The specific execution process follows the processing logic of steps S2021-S2025 above. For each completed and standardized model, a component topology graph is constructed, and multiple local subgraphs such as tower legs, tower body, crossarms, and tower heads are divided. The global component topology graph and each local subgraph are sequentially input into the offline-trained graph embedding model, and processed layer by layer through graph convolutional propagation layers, relation type attention encoding layers, component attention pooling layers, and readout layers. The output is a unique, searchable whole-tower graph embedding vector for that tower, as well as multiple sets of searchable local graph embedding vectors corresponding to the tower legs, multi-layer crossarms, tower heads, and hanging point units. All completed and standardized models are traversed, and dedicated global and local vectors are pre-generated and temporarily cached for each historical tower, eliminating the need for repeated encoding during online retrieval and significantly improving retrieval response speed.

[0124] Step S204e: Construct a transmission tower model library based on multiple complete standardized models, whole tower diagram embedding vectors, and multiple local diagram embedding vectors.

[0125] In this embodiment of the application, the cleaned entity model, engineering parameters, and pre-calculated vectors are associated and bound together, and persistently stored to form a callable transmission tower model library.

[0126] Storage association rules: Using the unique tower number as the primary key, each index is synchronously bound to four types of data: ① Complete standardized 3D tower model entity file and component parameter table; ② All engineering working parameters of the tower; ③ Embedded vector of the whole tower image to be retrieved; ④ Embedded vector of various local images to be retrieved; The database establishes a vector index partition to store all embedded vectors separately, supporting fast vector comparison and retrieval; At the same time, it divides the tower type, voltage, and icing level into multi-level classification indexes to facilitate narrowing the search scope.

[0127] Understandably, once the data is entered into the database, a complete power transmission tower model library is obtained. In online design scenarios, only the embedded vector generated by the target tower needs to be input, and multi-dimensional similarity matching can be quickly completed within the vector partition of the model library, outputting the matched historical towers as reuse templates.

[0128] Optionally, in one embodiment of the present invention, the description is based on a 220kV double-circuit straight-line tower in a transmission line project. The design requirements for this target tower are: voltage level 220kV, tower type double-circuit straight-line tower, nominal height 36m, total tower height 43m, three layers of conductor crossarms, one layer of ground wire support, four-legged deployable tower legs, and a space truss structure for the tower body. The overall system architecture is as follows: Figure 3 As shown, the specific process is as follows: Figure 4 As shown.

[0129] The architecture includes a model input and parsing module 301, a model library module 302, a component topology graph construction module 303, a multi-granularity subgraph extraction module 304, a graph neural network embedding module 305, a similarity retrieval module 306, a parameterization reuse module 307, a constraint verification module 308, and a model output module 309.

[0130] The model input and parsing module 301 is used to read the 3D model of the transmission tower to be retrieved or reused, and parse the geometric information, component information, connection relationships, and engineering attribute information in the model. The model library module 302 is used to store existing 3D models of transmission towers, component parameter tables, component connection relationship tables, tower type engineering attribute tables, and local structural templates, providing a data foundation for subsequent retrieval and reuse.

[0131] The component topology diagram construction module 303 is used to convert the 3D model of the transmission tower into a component topology diagram, that is, using engineering components such as main members, diagonal members, crossarms, tower legs, hanging points, and node plates as diagram nodes, and the connection relationships and hierarchical relationships between components as diagram edges. The multi-granularity sub-diagram extraction module 304 is used to extract local sub-diagrams such as tower legs, tower body, crossarms, tower heads, and hanging points from the overall tower component topology diagram, and to mark the interface nodes of local structures.

[0132] The graph neural network embedding module 305 is used to encode the topology graph of the entire tower components and the local subgraphs, generating the whole tower graph embedding vector and the local graph embedding vector. The similarity retrieval module 306 is used to calculate the similarity between the model to be retrieved and the candidate models in the model library based on the graph embedding vectors, and combined with the matching degree of engineering parameters, return the whole tower model or local structure template with high similarity.

[0133] The parameterization reuse module 307 is used to perform component mapping, parameter migration, coordinate transformation, and local structural adjustment based on the search results to generate the target tower model. The constraint verification module 308 is used to verify the connectivity, interface nodes, number of hanging points, spatial collision, and parameter range of the reused model. The model output module 309 is used to output the 3D model of the transmission tower, component parameter information, and verification report after the verification is passed.

[0134] In this embodiment of target tower construction, firstly, the system through Figure 3 The model input and parsing module 301 reads the initial model file of the target tower. This file contains some tower body and leg components, but the crossarm structure still needs to be refined using a historical model. The system also calls... Figure 3 The model library module 102 contains 126 historical tower models, including 48 220kV tower models, 52 110kV tower models, and 26 500kV tower models.

[0135] The parameters of some components of the target tower are shown in the table below:

[0136] Then, the system executes. Figure 4 Step S1 involves data standardization. Component coordinates are standardized to a local coordinate system with the foundation center as the origin; length units are standardized to meters; and component types are standardized as main members, diagonal members, auxiliary members, crossbeam interfaces, hanging points, and node plates. For component M001, its length is: In the formula, The length of component M001 is represented by two three-dimensional coordinates, which represent the starting and ending coordinates of M001, respectively.

[0137] Then, the system executes. Figure 4 In step S2, the component topology graph is constructed. Components M001, M002, M003, X015, X016, etc., are transformed into component nodes; the connection relationships between components are transformed into component connection edges; the relationships between components and engineering parts such as tower legs, tower body, and crossarms are transformed into hierarchical belonging edges. For example, M001 and X015 are jointly connected to node N018, and a shared node connection edge is established between them; H001 and H002 are first-level crossarm interface nodes, and hierarchical belonging edges are established between them and the first-level crossarm semantic nodes.

[0138] In this embodiment, the target tower comprises 214 component nodes, 386 component connection edges, and 54 hierarchical belonging edges. Among these, there are 36 main member nodes, 108 diagonal member nodes, 42 auxiliary member nodes, 12 crossarm interface nodes, 8 hanging point constraint nodes, and 8 node plate nodes. The component topology diagram is denoted as follows: In the formula, Represents the topology of the target tower components; Represents the target tower node set; Represents the set of sides of the target tower; Represents the target tower node attribute matrix; Represents the target tower's side attribute matrix; This represents the set of hierarchical relationships in the target tower project.

[0139] Then, the system executes. Figure 4 Step S3: Local Sub-image Extraction. Based on component height and location labels, the system extracts the tower leg sub-image, tower body sub-image, first-layer crossarm interface sub-image, second-layer crossarm interface sub-image, third-layer crossarm interface sub-image, and tower head sub-image. The first-layer crossarm interface sub-image includes 2 tower body interface nodes, 2 hanging point constraint nodes, and 1 crossarm semantic node, and marks H001 and H002 as interface nodes, i.e. , .

[0140] Then, the system executes. Figure 4 In steps S4 and S5, the neural network embedding encoding is shown, and the encoding structure is as follows: Figure 5 As shown. The node attribute matrix 501, adjacency matrix 502, and edge attribute matrix 503 are input to the relation type attention layer 504, and then pass through the key component attention pooling layer 505 and the readout layer 506 to obtain the whole tower graph embedding vector 507 and the local graph embedding vector 508. In this embodiment, the whole tower graph embedding vector dimension is set to 128 dimensions, the local graph embedding vector dimension is set to 64 dimensions, the number of graph neural network layers is set to 3, and the key component attention pooling retention ratio is set to 0.75. In this embodiment, the graph embedding model has been pre-trained based on historical tower models and reuse records in the model library. In the subsequent similarity retrieval stage, the trained graph neural network encoder is directly called to generate graph embedding vectors and perform similarity calculation.

[0141] Then, the system executes. Figure 4 Step S7: Similarity Retrieval. The tower model to be retrieved and candidate models in the model library are compared using the similarity retrieval module 106 to calculate the overall tower topological similarity, local structural similarity, and engineering parameter matching degree, and a Top-K similar model list is generated. The system returns the top 4 similar models as shown in the table below:

[0142] Based on the search results, the system selected T-220-ZS-34 as the basic model for whole-tower reuse, and simultaneously selected the crossarm structure in T-220-ZS-36A as the source for local replacement. This combined reuse strategy was automatically generated by the similarity retrieval module because T-220-ZS-34 has the highest overall tower topological similarity, while the crossarm of T-220-ZS-36A has a higher degree of local similarity and engineering parameter matching.

[0143] Next, the system executes... Figure 4 Step S8: Parameterized reuse. Based on the search results, the system establishes a component mapping relationship between the target model and similar models. It maps the main tower leg member M001 in the target model to the main member A001 in T-220-ZS-34, M002 to A002, and X015 to the diagonal member AX015. Subsequently, the system performs parameter migration and adjustment on the reused model based on the component mapping relationship: Since the nominal height of T-220-ZS-34 is 34m, while the nominal height of the target tower is 36m, the system adds 2m to the standard section in the middle of the tower body and simultaneously adjusts the coordinates of the endpoints of the main and diagonal members within this standard section; since the crossarm structure of T-220-ZS-36A is closer to the target requirement, the system calls its crossarm template to replace the corresponding crossarm structure of T-220-ZS-34. The parameter adjustment details are shown in the table below:

[0144] The generated tower model was reused for constraint verification. During the first verification, the system found that the distance between the hanging point at the end of the right crossarm of the first floor and the diagonal member of the crossarm was 0.18m, which was less than the preset safety distance of 0.25m. Therefore, the hanging point was moved outward by 0.12m along the crossarm direction. After the second verification, the topology diagram of the entire tower components was connected, the number of hanging points was 8, the number of crossarm interface nodes met the requirements, no spatial collisions were found, and the final model that passed the verification was generated, and a verification report was output.

[0145] As can be seen from this embodiment, the present invention can complete the similarity retrieval of the entire tower based on the topology of the tower components, and further use the retrieval results for parameterized reuse.

[0146] Optionally, in one embodiment of the present invention, the description is based on the same 220kV double-circuit straight-line tower model as the previous embodiment. Unlike the previous embodiment, this embodiment does not reuse the entire tower structure; instead, given that the target tower body is already completed, only the partial structure of the first-layer crossarm is retrieved and reused. The system architecture remains the same. Figure 3 As shown, the method flow remains the same. Figure 4 As shown.

[0147] The target tower has completed modeling of its legs and main body, but the first-layer crossarm has not yet been fully modeled. The design requirements are: the first-layer crossarm is to be installed at a height of 24.0m, symmetrically arranged, with a length of 6.2m on each side, one conductor suspension point on each side, and two interface nodes at the connection between the crossarm and the tower body. The system generates a crossarm query subgraph based on the tower body interface conditions. The node parameters in the crossarm query subgraph are shown in the table below:

[0148] The system first converts nodes A001 to A005 into local subgraph nodes and establishes interface ownership edges, hanging point ownership edges, and crossarm constraint edges. Since A001 and A002 are interface nodes connecting the crossarm to the tower body, the system sets their interface flag to 1, i.e. , A003 and A004 are attachment constraint nodes, not interface nodes, therefore , .

[0149] Then, the system inputs the crossbow query subgraph into the local graph neural network encoder to obtain the local graph embedding vector. In this embodiment, the local graph embedding vector has a dimension of 64, the local graph neural network has 2 layers, and the interface node weight coefficients are... Set to 0.35, the weight coefficient of the attachment node. Set it to 0.30. The purpose of this setting is to make the local search focus more on the interface matching between the crossarm and the tower body, as well as the arrangement of the hanging points.

[0150] Then, the system searches for similar crossarm templates in the local structural template library. The template library contains 86 crossarm templates, including 32 220kV crossarm templates, 36 110kV crossarm templates, and 18 500kV crossarm templates. The system calculates the local similarity between the queried crossarm sub-diagram and each crossarm template in the template library, and sorts them based on the number of interface nodes, the number of hanging points, the crossarm length range, and the applicable voltage level. The search results are shown in the table below:

[0151] The system selected the top-ranked similar local template, ARM-220-03. This template is derived from T-220-ZS-36A, with a single-sided crossarm length of 6.0m, 2 interface nodes, and 1 attachment point per side. Since the target crossarm has a single-sided length of 6.2m, length adjustment is required during reuse.

[0152] The system matches the two interface nodes B001 and B002 of the template ARM-220-03 to the interface nodes A001 and A002 of the target model, respectively, and calculates the translation vector and rotation matrix based on the interface node positions to place the template crossarm onto the target tower body. The coordinate transformation formula is: In the formula, Represents the original coordinates of the endpoints of the components in the template; This represents the coordinates after transformation to the target tower; This represents the rotation matrix calculated based on the orientation of the interface node; This represents the translation vector calculated based on the center position of the interface node.

[0153] After the coordinate transformation, the system adjusts the template parameters. The adjustments are shown in the table below:

[0154] After adjustment, the system embeds the crossarm template into the target model, forming the target model after embedding the crossarm template. The system writes the main crossarm members, diagonal crossarm members, hanging point nodes, and node plates from the template into the target tower component parameter table and updates the overall tower component connection relationship table. The list of newly added components is shown in the table below:

[0155] Finally, the system performs constraint checks. The system performs connectivity checks, interface node checks, hanging point quantity checks, spatial collision checks, and parameter range checks. Specifically, the connectivity check determines whether all newly added components in the crossarm sub-drawing can be traversed from interface nodes A001 or A002; the interface node check determines whether the number of template interface nodes matches the target number of interface nodes; the hanging point quantity check determines whether the number of hanging points on both sides meets the design requirements; the spatial collision check determines whether the minimum distance between NH003 and the tower auxiliary material is greater than the preset safety distance; and the parameter range check determines whether the crossarm main material cross section and diagonal member length are within the allowable range. After all checks pass, the system outputs the final model and a check report.

[0156] As can be seen from this embodiment, the present invention can perform similarity retrieval, template matching, coordinate transformation, parameter adjustment and embedding verification only for the local structure of the crossarm without reusing the whole tower model.

[0157] Based on the description of the above embodiments, it is understood that this application can achieve intelligent retrieval based on the topology of tower structures, replacing traditional keyword, tower type number, and manual experience retrieval. The system supports both whole tower model retrieval and retrieval of local structures such as tower legs, crossarms, tower heads, and hanging points. This application transforms the 3D model of the tower into a component topology diagram, which can preserve the component connection relationships, engineering hierarchy, and local structural semantics, avoiding the problem of topological information loss in point cloud, mesh, or multi-view representations. By learning structural features through graph neural networks, the accuracy of similar model retrieval and the stability of reuse results can be improved. This application can improve the reuse rate of historical tower model libraries, reduce the workload of repetitive modeling, and shorten the 3D design cycle of transmission lines. This method can also be extended to component connection type 3D model retrieval and reuse scenarios such as substation steel structures, communication towers, space trusses, wind turbine towers, and bridge steel structures.

[0158] Based on the same inventive concept, this application also provides a transmission tower model building device for implementing the above-mentioned transmission tower model building method. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more transmission tower model building device embodiments provided below can be found in the limitations of the transmission tower model building method above, and will not be repeated here.

[0159] In one exemplary embodiment, such as Figure 6 As shown, a transmission tower model construction device 60 is provided, comprising: an acquisition module 601 for acquiring the modeling requirements of a target transmission tower model; a first construction module 602 for constructing a topology diagram of the transmission tower components and multiple local sub-graphs according to the modeling requirements; a determination module 603 for inputting the topology diagram of the transmission tower components and multiple local sub-graphs into a pre-constructed graph embedding model to obtain a whole-tower graph embedding vector and multiple local graph embedding vectors; a retrieval module 604 for retrieving from a pre-constructed transmission tower model library based on the whole-tower graph embedding vector and multiple local graph embedding vectors to obtain retrieval results, wherein each transmission tower model in the transmission tower model library corresponds to a whole-tower graph embedding vector to be retrieved and multiple local graph embedding vectors to be retrieved; and a second construction module 605 for constructing the target transmission tower model based on the retrieval results.

[0160] Optionally, the first construction module 602 is further configured to: determine the engineering components and component connection relationships of the target transmission tower model according to the modeling requirements; determine the graph nodes according to the engineering components; determine the connection edges according to the component connection relationships; construct the transmission tower component topology graph according to the graph nodes and connection edges; and divide the transmission tower component topology graph to obtain multiple local subgraphs.

[0161] Optionally, the graph embedding model in module 603 includes a graph convolutional propagation layer, a relational type attention encoding layer, a component attention pooling layer, and a readout layer connected in sequence; wherein: the graph convolutional propagation layer is used to extract the first node features of the topology graph of the transmission tower components and multiple local subgraphs; the relational type attention encoding layer is used to calculate attention weights to determine the second node features, wherein the first node features are fused with attention weights to obtain the second node features; the component attention pooling layer is used to determine the key node features based on the second node features, wherein the key node features are the features corresponding to nodes whose node importance scores are greater than a preset score threshold; and the readout layer is used to determine the whole tower graph embedding vector and multiple local graph embedding vectors based on the key node features.

[0162] Optionally, the retrieval module 604 is further configured to: determine the similarity between the whole tower image embedding vector and multiple whole tower image embedding vectors to be retrieved, to obtain multiple first similarities; determine the similarity between each local image embedding vector and multiple local image embedding vectors to be retrieved, to obtain multiple second similarities; determine the similarity between the engineering parameters in the modeling requirements and the corresponding engineering parameters of each transmission tower model in the transmission tower model library, to obtain multiple third similarities; determine multiple comprehensive similarities based on the multiple first similarities, multiple second similarities, and multiple third similarities; and determine the retrieval results based on the multiple comprehensive similarities.

[0163] Optionally, the second construction module 605 is further configured to: establish a mapping relationship between the target transmission tower model to be constructed and similar models, wherein the similar models are included in the search results; transfer the parameters of the similar models to the target transmission tower model to be constructed according to the mapping relationship to obtain the model to be verified; verify the model to be verified according to the preset verification rules to obtain the verification result; and, in response to the verification result indicating that the verification is successful, determine the model to be verified as the target transmission tower model.

[0164] Optionally, the construction of the transmission tower model library in the retrieval module 604 includes: acquiring multiple historical transmission tower models; performing data standardization processing on the multiple historical transmission tower models to obtain multiple standardized models; completing the connection relationships of the multiple standardized models to obtain multiple completed standardized models; for the multiple completed standardized models, determining the whole tower diagram embedding vector and multiple local diagram embedding vectors corresponding to each completed standardized model; and constructing the transmission tower model library based on the multiple completed standardized models, the whole tower diagram embedding vector, and the multiple local diagram embedding vectors.

[0165] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 7As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores process and result data from the method's execution. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for constructing a power transmission tower model.

[0166] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0167] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0168] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0169] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0170] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0171] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0172] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchain. The processors involved in the embodiments provided in this application may be, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc.

[0173] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0174] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for constructing a transmission tower model, characterized in that, The method for constructing the transmission tower model includes: Obtain the modeling requirements for the target transmission tower model; Based on the modeling requirements, a topology diagram of the transmission tower components and multiple local sub-diagrams are constructed. The topology diagram of the transmission tower components and the multiple local sub-graphs are respectively input into the pre-constructed graph embedding model to obtain the whole tower graph embedding vector and multiple local graph embedding vectors; Based on the whole tower diagram embedding vector and multiple local diagram embedding vectors, the search results are obtained from the pre-built transmission tower model library. Each transmission tower model in the transmission tower model library corresponds to a whole tower diagram embedding vector to be searched and multiple local diagram embedding vectors to be searched. Based on the search results, a model of the target transmission tower is constructed.

2. The method for constructing a transmission tower model according to claim 1, characterized in that, Based on the modeling requirements, the construction of a topology diagram of transmission tower components and multiple local sub-diagrams specifically includes: Based on the modeling requirements, determine the engineering components and component connection relationships of the target transmission tower model; Based on the engineering components, determine the diagram nodes; Based on the connection relationship of the components, determine the connection edge; Based on the graph nodes and the connecting edges, construct the topology diagram of the transmission tower components; The topology diagram of the transmission tower components is divided to obtain the multiple local sub-diagrams.

3. The method for constructing a transmission tower model according to claim 1, characterized in that, The graph embedding model comprises, in sequence, a graph convolutional propagation layer, a relation-type attention encoding layer, a component attention pooling layer, and a readout layer; wherein: The graph convolutional propagation layer is used to extract the first node features of the topology graph of the power transmission tower components and the multiple local subgraphs; The relation type attention encoding layer is used to calculate attention weights to determine the second node features, wherein the first node features are fused with attention weights to obtain the second node features; The component attention pooling layer is used to determine key node features based on the second node features, wherein the key node features are the features corresponding to nodes whose node importance scores are greater than a preset score threshold; The readout layer is used to determine the whole-tower graph embedding vector and multiple local graph embedding vectors based on the key node features.

4. The method for constructing a transmission tower model according to claim 1, characterized in that, The step of retrieving data from the transmission tower model library based on the whole tower diagram embedding vector and multiple local diagram embedding vectors to obtain retrieval results specifically includes: Determine the similarity between the whole tower diagram embedding vector and multiple whole tower diagram embedding vectors to be retrieved to obtain multiple first similarities; Determine the similarity between each local graph embedding vector and the multiple local graph embedding vectors to be retrieved to obtain multiple second similarities; Determine the similarity between the engineering parameters in the modeling requirements and the corresponding engineering parameters of each transmission tower model in the transmission tower model library to obtain multiple third similarity scores; Based on the plurality of first similarities, the plurality of second similarities, and the plurality of third similarities, a plurality of comprehensive similarities are determined; The search results are determined based on the multiple comprehensive similarities.

5. The method for constructing a transmission tower model according to claim 1, characterized in that, The step of constructing the target transmission tower model based on the search results specifically includes: Establish a mapping relationship between the target transmission tower model to be constructed and similar models, wherein the similar models are included in the search results; Based on the mapping relationship, the parameters of the similar model are transferred to the target transmission tower model to be constructed to obtain the model to be verified. The model to be verified is verified according to the preset verification rules, and the verification result is obtained. In response to the verification result indicating that the verification was successful, the model to be verified is determined as the target transmission tower model.

6. The method for constructing a transmission tower model according to claim 1, characterized in that, The construction of the transmission tower model library includes: Obtain multiple historical transmission tower models; Data standardization processing was performed on the multiple historical transmission tower models to obtain multiple standardized models; By completing the connection relationships of the multiple standardized models, multiple completed standardized models are obtained; For the multiple completion and normalization models, determine the whole graph embedding vector and multiple local graph embedding vectors corresponding to each completion and normalization model; The transmission tower model library is constructed based on the multiple complete standardized models, the whole tower diagram embedding vector, and the multiple local diagram embedding vectors.

7. A device for constructing a transmission tower model, characterized in that, The transmission tower model construction device includes: The acquisition module is used to acquire the modeling requirements of the target transmission tower model; The first construction module is used to construct a topology diagram of transmission tower components and multiple local sub-diagrams according to the modeling requirements. The determination module is used to input the topology diagram of the transmission tower components and the multiple local sub-graphs into a pre-constructed graph embedding model to obtain the whole tower graph embedding vector and multiple local graph embedding vectors. The retrieval module is used to retrieve data from a pre-built transmission tower model library based on the whole tower image embedding vector and multiple local image embedding vectors, and obtain retrieval results. Each transmission tower model in the transmission tower model library corresponds to a whole tower image embedding vector to be retrieved and multiple local image embedding vectors to be retrieved. The second construction module is used to construct the target transmission tower model based on the search results.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the transmission tower model construction method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the transmission tower model construction method according to any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the transmission tower model construction method according to any one of claims 1-6.

Images