A semantic enhanced graph-tree hybrid storage and dynamic updating method for smart highway

Abstract

The application relates to the technical field of highway engineering digitization, and discloses a semantic enhancement graph-tree hybrid storage and dynamic updating method for smart highways. The method comprises the following steps: constructing a semantic enhancement engineering object type tree; constructing an attribute semantic feature tree and fusing attribute types based on unit consistency, value range similarity and name vector semantic similarity; mapping the engineering object type tree and the attribute feature tree into a unified graph type layer structure and establishing a multidimensional hybrid index; constructing an instance layer on the basis of the type layer, realizing type inheritance through an "instance" edge, storing attribute numerical information through a "value" edge, realizing graph-tree hybrid storage and local dynamic updating. Through decoupled modeling of a type system and an attribute system and an attribute value edge storage mechanism, the number of attribute nodes is reduced, data organization density and query efficiency are improved, and unified semantic retrieval and local rapid updating across projects and stages are supported.

Description

Technical Field

[0001] This invention relates to the field of smart highway construction, and in particular to a semantically enhanced graph-tree hybrid structure data storage and dynamic update method for managing, organizing and updating digital highway data, which can be applied to smart highway platforms, engineering data management platforms and digital twin systems, etc. Background Technology

[0002] With the comprehensive advancement of smart highway and smart expressway construction in my country, efficient storage and management of massive amounts of data generated across stages such as design, construction, monitoring, and maintenance has become increasingly important. Throughout the entire lifecycle of a project, the amount of data increases dramatically, with diverse sources, complex formats, and frequent updates. Although the current highway engineering information model standards stipulate the classification and coding of engineering objects, data storage and updates have not yet formed a good solution in the practical application of smart highways, and mainly face the following key challenges: (1) The classification and semantic system of engineering objects are not unified, and type conflicts often occur; (2) There are problems such as synonyms, inconsistent units, and overlapping value domains in attribute fields; (3) There are significant hierarchical relationships among engineering objects, but there are cross-level associations between attributes; (4) There is a lack of reasonable data organization models that can take into account both the expression of complex associations of highway full-cycle data and efficient data retrieval performance; (5) When data updates occur, the traditional data structure needs to be reconstructed as a whole.

[0003] Therefore, there is an urgent need for a data modeling approach that differs from traditional single-structure data models, addressing core issues such as semantic conflicts of types, heterogeneous attributes, and the sparsity and poor query performance of graph-structured data, in order to adapt to the real-world demands of smart highway construction, which involves large scale, numerous objects, complex attributes, and frequent updates. Summary of the Invention

[0004] To address the aforementioned issues, the present invention aims to provide a semantically enhanced graph-tree hybrid storage and dynamic update method for smart highways. Starting with a tree structure, it integrates the characteristics of graph and tree structures to construct a standardized graph-tree hybrid type-layer data organization form. Furthermore, it utilizes instance nodes and edges to organize information, thereby increasing data storage density and improving the efficiency of managing, retrieving, and updating digital data for smart highways.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: a semantically enhanced graph-tree hybrid storage and dynamic update method for intelligent highways, comprising the following steps:

[0006] S1. Construct a semantically enhanced engineering object type tree;

[0007] S2. Construct an attribute semantic feature tree and perform attribute type fusion based on unit consistency, value range similarity, and name vector semantic similarity;

[0008] S3. Map the engineering object type tree and attribute feature tree into a unified graph type layer structure and establish a multi-dimensional hybrid index;

[0009] S4. Build an instance layer on the basis of the type layer, realize type inheritance through instance edges, and store attribute numerical information through value edges, so as to realize graph-tree hybrid storage and local dynamic updates.

[0010] Optionally, step S1 includes:

[0011] S1.1 Analyze the digital design and delivery data of highway engineering projects and extract the types of engineering objects from different disciplines, such as roadbed_retaining wall, bridge_pier, tunnel_system anchor, etc.

[0012] S1.2 Based on the extracted basic entity objects, establish a tree structure according to the hierarchical relationship of projects, specialties, facilities, sub-facilities, and components;

[0013] S1.3. Add enhanced semantic information to each type node. In addition to classification coding, tree structure hierarchy coding, and name, add semantic labels and route feature labels to form a semantically enhanced engineering object type layer.

[0014] Optionally, step S2 includes:

[0015] S2.1. Parse the digital highway data, extract the attribute content, merge and remove duplicates of those with identical names and those with different names but the same meaning, and perform normalization preprocessing on the original names of the attribute objects to obtain a preliminary list of attribute types.

[0016] S2.2 Establish attribute type grouping, including identification information, location information, size information, engineering information, etc., and establish a tree structure according to the hierarchical relationship of project, profession, attribute grouping, and attribute type;

[0017] S2.3 Calculate the semantic similarity between attribute types for the unit, attribute value, and field name in sequence;

[0018] S2.4 After merging, add a standardized name, the original field name before merging, unit, data type, value range (such as maximum value, minimum value, value range, etc.) and validation rules to each attribute type node to achieve automatic semantic fusion of heterogeneous attributes.

[0019] Further, step S2.3 includes:

[0020] ① Establish a knowledge base for unit conversion rules, including relationships between directly convertible units and relationships between equal units, and calculate unit consistency similarity. If the units can be directly converted, then =1; if the units belong to the same category but cannot be directly converted, then =0.7; If the units are not of the same type and cannot be converted, then... =0; reserved The attribute type;

[0021] ② Calculate the similarity of numerical ranges Extract the attribute value or random value v within the numerical range, and calculate it using formula (1). If it is a non-numerical type, directly take 0:

[0022] Formula (1)

[0023] In the formula: The value is between 0 and 1; The value of attribute 1; This is the value of attribute 2.

[0024] ③ Calculate the semantic similarity of field names The field names are converted into vectors A or B using an embedding model, and then calculated using formula (2):

[0025] Formula (2)

[0026] In the formula: The values ​​are between 0 and 1; A is the vector representation of field name 1; B is the vector representation of field name 2.

[0027] ④ Calculate the overall similarity using formula (3) :

[0028] Formula (3)

[0029] In the formula: , , The weights are configurable.

[0030] ⑤ When the overall similarity When the similarity threshold is exceeded, they are determined to have the same semantic attribute and are merged into a single "standard attribute type node".

[0031] Optionally, step S3 includes:

[0032] S3.1. Based on the diagram structure, transform the tree structure generated in steps S1 and S2, and maintain the hierarchical logic by establishing containing edges.

[0033] S3.2 Write the tree structure node information from steps S1 and S2 into the graph node attributes;

[0034] S3.3 Construct a hybrid index of type layers, including: building a coding index based on classification coding and tree structure coding, building a semantic index based on semantic tags, building a route index based on route feature information tags, and building an attribute feature index based on attribute name, unit, value range, etc.

[0035] Optionally, step S4 includes:

[0036] S4.1 Read the engineering objects in the digital highway data, locate the corresponding type layer node according to the classification code, create a new engineering object instance layer node, and write data such as instance ID, classification code, route number, station number information, construction stage, and spatial coordinates.

[0037] S4.2 Establish an instance edge association between the project object type node and the instance node, and store the sequence code and version information of the instance object in the instance edge;

[0038] S4.3 Parse the attribute content of the highway digitization data, and create value edge association instance nodes and attribute type nodes according to the attribute content name;

[0039] S4.4 Write the attribute value content field on the value side, including basic fields such as attribute name, unit, value, version information, and other descriptions. Validate the written data through the value range and validation rules of the type node.

[0040] S4.5 When engineering changes or data updates occur, the affected subgraph is located by calculating the index of the instance node. Only the content of the instance node and the directly associated "instance" edge and "value" edge is updated, without rebuilding the entire graph structure, thus achieving efficient dynamic data updates.

[0041] As described above, in step S1, this invention constructs a semantically enhanced engineering object type tree to hierarchically organize smart highway data. While retaining the traditional classification coding and naming system, it introduces semantic tags and route feature information to achieve semantically enhanced expression of engineering object types, providing a structural foundation for subsequent type matching and accurate retrieval. In step S2, addressing the issue of insufficient standardization of attributes in multi-stage and multi-source data, an independent attribute type tree is constructed. Based on unit consistency, value range similarity, and semantic similarity of name vectors, a comprehensive judgment is made to achieve automatic semantic fusion and standardization of attribute types. This step improves the alignment capability of heterogeneous attribute data while maintaining the consistency of the tree structure. In step S3, the engineering object type tree and attribute type tree are uniformly mapped to a semantically enhanced graph type layer, and a multi-dimensional indexing mechanism including hierarchical relationships, semantic information, route features, and attribute rules is constructed. This integrates the hierarchical clarity of the tree structure with the associative flexibility of the graph structure, forming a stable type layer structure. In step S4, an instance layer structure is built on top of the aforementioned type layer. Type inheritance is achieved through "instance" edges, and attribute numerical information is stored through "value" edges. Attribute types are abstracted into type layer nodes, so that attribute values ​​no longer exist as independent nodes. This structurally avoids the problem of node size bloat, improving data organization density and query efficiency. Simultaneously, a layered mechanism with a stable type layer and an updatable instance layer is adopted to enable rapid updates of local subgraphs during engineering changes, enhancing the ability to continuously evolve large-scale data.

[0042] In summary, this invention forms a graph-tree hybrid storage mode of "type layer + instance layer + value edge", which supports unified semantic retrieval across projects and levels, as well as data positioning and dynamic iteration for specific engineering sections. It is suitable for high-density storage and efficient management of massive data in smart highways.

[0043] Compared with existing technologies, this invention has the following beneficial effects: ① It achieves semantic decoupling of type and attribute, and constructs a unified semantic system. This invention constructs an engineering object type tree and an attribute type tree respectively, and performs semantic enhancement and standardization processing to achieve structural decoupling of the type system and attribute system, avoiding semantic conflicts between different professions and different stages. On this basis, it can realize unified semantic retrieval and data alignment analysis across projects, professions, and stages, improving data reuse capability and comparability. ② It proposes a triple attribute fusion algorithm to achieve automatic standardization of heterogeneous fields. Through a comprehensive fusion mechanism of unit consistency judgment, numerical range similarity calculation, and name vector semantic similarity calculation, it realizes automatic semantic merging and standardization of attribute fields, effectively reducing the cost of manual rule maintenance and improving the automatic access capability of multi-source heterogeneous data. ③ It constructs a graph-tree hybrid type layer structure, taking into account hierarchical expression and complex association management. This invention maps the engineering structure tree and the attribute structure tree into a unified graph type layer. While retaining the logical clarity of the tree hierarchy, it introduces the flexible association capability of the graph structure and constructs a multi-dimensional hybrid index mechanism to unify the expression of hierarchical and semantic relationships between engineering objects. ④ Employing an attribute-value edge storage mechanism significantly improves data density and query efficiency. By writing attribute values ​​into the edge attributes of "value" edges and abstracting attribute types into type-level nodes, the system avoids building each attribute value as an independent node. This structurally reduces the size of graph nodes and increases the data density of the graph structure, maintaining high query and storage efficiency even in scenarios with millions of instances. ⑤ Constructing a layered mechanism with a stable type layer and a variable instance layer enhances system scalability and update efficiency. The type layer structure remains stable, while instance layer data is updated independently. When engineering changes occur, only instance nodes and their associated edges need to be updated to complete local data iteration, without rebuilding the overall structure. This significantly reduces update complexity and improves the system's adaptability to the continuous expansion and evolution of large-scale engineering data. Attached Figure Description

[0044] Figure 1 The project object type tree structure;

[0045] Figure 2 The structure of the project object type diagram;

[0046] Figure 3 This is for querying the properties of the reinforcing steel bars in the bridge abutment cap beam. Detailed Implementation

[0047] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should be noted that the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0048] A semantically enhanced graph-tree hybrid storage and dynamic update method for smart highways includes the following steps:

[0049] S1. Construct a semantically enhanced engineering object type tree, including:

[0050] (1) Analyze the digital design delivery data of highway engineering and extract the types of engineering objects of different specialties, such as roadbed_retaining wall, bridge_pier, tunnel_system anchor, etc.;

[0051] (2) Based on the extracted basic entity objects, establish a tree structure according to the hierarchical relationship of project, profession, facility, sub-facility, and component;

[0052] (3) Add enhanced semantic information to each type node. In addition to classification coding, tree structure hierarchy coding and name, add semantic tags to describe the role and function of the engineering object, such as structural bearing, drainage, structural reinforcement, slope reinforcement, slope protection and other functional semantics, to realize the connection with structural calculation, hydrological calculation and slope stability analysis applications. For example, concrete, steel bars, rock and soil and other material semantics to realize the association with the corresponding material library. Add route feature tags to describe the relationship between the engineering object and the route spatial system, including line name, station range, left and right lane positions, chain break information, coordinate system transformation parameters, etc., to enhance the depth of digital expression of type nodes and finally form a semantically enhanced engineering object type layer.

[0053] S2. Construct an attribute semantic feature tree and perform attribute type fusion based on unit consistency, value range similarity, and name vector semantic similarity, including:

[0054] (1) Parse the digital highway data, extract the attribute content, merge and deduplicate those with the same name and those with different names but the same meaning, realize the normalization preprocessing of the original name of the attribute object, and obtain a preliminary list of attribute types;

[0055] (2) Establish attribute type grouping, including identification information, location information, size information, engineering information, etc., and establish a tree structure according to the hierarchical relationship of project, profession, attribute grouping, and attribute type;

[0056] (3) Calculate the semantic similarity between each attribute type for the unit, attribute value, and field name in turn. The specific method is as follows:

[0057] ① Establish a knowledge base for unit conversion rules, including relationships between directly convertible units and relationships between equal units, and calculate unit consistency similarity. If the units can be directly converted, then =1; if the units belong to the same category but cannot be directly converted, then =0.7; If the units are not of the same type and cannot be converted, then... =0; reserved The attribute type;

[0058] ② Calculate the similarity of numerical ranges Extract the attribute value or random value v within the numerical range, and calculate it using formula (1). If it is a non-numerical type, directly take 0:

[0059] Formula (1)

[0060] In the formula: The value is between 0 and 1; The value of attribute 1; This is the value of attribute 2.

[0061] ③ Calculate the semantic similarity of field names The field names are converted into vectors A or B using an embedding model, and then calculated using formula (2):

[0062] Formula (2)

[0063] In the formula: The values ​​are between 0 and 1; A is the vector representation of field name 1; B is the vector representation of field name 2.

[0064] ④ Calculate the overall similarity using formula (3) :

[0065] Formula (3)

[0066] In the formula: , , The weights are configurable, and values ​​of 0.3, 0.5, and 0.2 are generally recommended.

[0067] ⑤ When the overall similarity When the similarity threshold is exceeded, they are determined to have the same semantic attribute and are merged into a single "standard attribute type node".

[0068] (4) After merging, add a standardized name, the original field name before merging, unit, data type, value range (such as maximum value, minimum value, value range, etc.) and verification rules to each attribute type node to achieve automatic semantic fusion of heterogeneous attributes.

[0069] S3. The dual-tree mapping is transformed into a unified type layered graph structure, and a multi-dimensional hybrid index is established, including:

[0070] (1) According to the diagram structure organization form, the tree structure generated in step S1 and step S2 is transformed, and the hierarchical logic is maintained by establishing "containment" edges;

[0071] (2) Write the tree structure node information from steps S1 and S2 into the graph node attributes;

[0072] (3) Construct a hybrid index of type layers, including: establishing a coding index based on classification coding and tree structure coding, establishing a semantic index based on semantic tags, establishing a route index based on route feature information tags, and establishing an attribute feature index based on attribute name, unit, value range, etc.

[0073] S4, instance layer construction and edge attribute storage, including:

[0074] (1) Read the engineering objects in the digital highway data, locate the corresponding type layer node according to the classification code, create a new engineering object instance layer node, and write the instance ID, classification code, route number, station information, construction stage, spatial coordinates and other data.

[0075] (2) Establish an "instance" edge association between the project object type node and the instance node, and store the sequence code and version information of the instance object in the "instance" edge;

[0076] (3) Parse the attribute content of the highway digitization data, and create "value" edge association instance nodes and attribute type nodes according to the attribute content name;

[0077] (4) Write the attribute value content field next to “Value”, including basic fields such as attribute name, unit, value, version information, and other descriptions. Validate the written data through the value range and validation rules of the type node.

[0078] (5) When engineering changes or data updates occur, the affected subgraph is located by calculating the index of the instance node, and only the content of the instance node, the directly associated "instance" edge and "value" edge is updated. There is no need to rebuild the entire graph structure, thus achieving efficient dynamic data updates.

[0079] In existing technologies, directly writing attributes onto nodes lacks unified management when dealing with massive amounts of complex attributes. This invention addresses this by establishing a type layer for attributes, abstracting the massive attribute content and compressing it into a controllable node region. This lays the foundation for storing instantiated data using value edges. The resulting attribute type layer, with its controllable and standardized nodes, leverages all data analysis capabilities specific to nodes. By connecting project objects to attribute types via value edges, not only are project objects given concrete attribute semantics, but the corresponding data is also stored. This approach ensures the type layer remains unchanged, supports massive instance expansion, and only affects localized areas during updates.

[0080] Example 1:

[0081] A semantically enhanced graph-tree hybrid storage and dynamic update method for smart highways includes:

[0082] S1. Construct a semantically enhanced engineering object type tree:

[0083] (1) Parse the digital data of a bridge, as shown in Table 1, and extract the bridge entity objects, including beam bridge, bridge deck system and ancillary works, concrete T beams, etc.

[0084] (2) Establish a tree structure according to the hierarchical relationship of projects, specialties, facilities, sub-facilities, and components, such as Figure 1 As shown;

[0085] (3) Add enhanced semantic information to each type of node, as shown in Table 2, including node type, name, encoding, semantic interpretation, etc.

[0086] S2. Construct the attribute semantic feature tree:

[0087] (1) Parse the digital highway data, extract the attribute content, and perform preprocessing and semantic similarity calculation, as shown in Tables 3 and 4;

[0088] (2) Merge the attribute content and construct a tree structure of attribute content, as shown in Table 5.

[0089] S3. Dual-tree mapping is a unified type layer graph structure:

[0090] (1) According to the diagram structure organization form, the tree structure generated in step S1 and step S2 is transformed, and the hierarchical logic is maintained by establishing "containment" edges;

[0091] (2) Write the tree structure node information from steps S1 and S2 into the graph node attributes, such as Figure 2 As shown;

[0092] (3) Construct a type-level index, including ID, code, description, level, name, etc., as shown in Table 6.

[0093] S4. Instance layer construction and edge attribute storage:

[0094] (1) Read the engineering objects in the highway digitization data, locate the corresponding type layer node according to the classification code, create a new engineering object instance layer node, write the instance data, and use edge storage to store attribute information, such as Figure 3 As shown in Table 7.

[0095] (2) When engineering changes or data updates occur, the affected subgraph is located by calculating the index of the instance node, as shown in Tables 8 and 9. Only the content of the instance node, the directly associated "instance" edge and "value" edge is updated, without rebuilding the entire graph structure, thus achieving efficient dynamic data updates.

[0096] Table 1. Digital data of a certain bridge (partial)

[0097]

[0098] Table 2 Semantic Enhancement Information

[0099]

[0100] Table 3 Semantic Similarity Calculation

[0101]

[0102] Table 4. Comprehensive Similarity Calculation

[0103]

[0104] Table 5 Attribute Information Type Tree Structure

[0105]

[0106]

[0107] Table 6 Type Information

[0108]

[0109] Table 7 Main Beam Bottom Width Dimensions

[0110]

[0111] Table 8 Type Data Index Calculation

[0112]

[0113] Table 9 Instance Data Index Calculation

[0114]

[0115] The above description is merely a specific embodiment of the present invention, and the scope of protection of the present invention is not limited thereto. Any transformations or substitutions that can be conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A semantically enhanced graph-tree hybrid storage and dynamic update method for intelligent highways, characterized in that, Includes the following steps: S1. Construct a semantically enhanced engineering object type tree; S2. Construct an attribute semantic feature tree and perform attribute type fusion based on unit consistency, value range similarity, and name vector semantic similarity; S3. Map the engineering object type tree and attribute feature tree into a unified graph type layer structure and establish a multi-dimensional hybrid index; S4. Build an instance layer on the basis of the type layer, realize type inheritance through instance edges, and store attribute numerical information through value edges to achieve graph-tree hybrid storage and local dynamic updates; Step S3 includes: S3.

1. Based on the diagram structure, transform the tree structure generated in steps S1 and S2, and maintain the hierarchical logic by establishing containing edges. S3.2 Write the tree structure node information from steps S1 and S2 into the graph node attributes; S3.3 Construct a hybrid index of type layers, including: building a coding index based on classification coding and tree structure coding, building a semantic index based on semantic tags, building a route index based on route feature information tags, and building an attribute feature index based on attribute name, unit, and value range; Step S4 includes: S4.1 Read the engineering objects in the digital highway data, locate the corresponding type layer node according to the classification code, create a new engineering object instance layer node, and write the instance ID, classification code, route number, station number information, construction stage, and spatial coordinate data. S4.2 Establish an instance edge association between the project object type node and the instance node, and store the sequence code and version information of the instance object in the instance edge; S4.3 Parse the attribute content of the highway digitization data, and create value edge association instance nodes and attribute type nodes according to the attribute content name; S4.4 Write the attribute value content field on the value edge, including the basic field, and validate the written data through the value range and validation rules of the type node; S4.5 When engineering changes or data updates occur, the affected subgraph is located by calculating the index of the instance node, and only the instance node, directly associated instance edge and value edge content are updated.

2. The semantically enhanced graph-tree hybrid storage and dynamic update method for intelligent highways according to claim 1, characterized in that, Step S1 includes: S1.1 Analyze the digital design delivery data of highway engineering and extract the engineering object types of different specialties, including roadbed_retaining wall, bridge_pier, and tunnel_system anchor; S1.2 Based on the extracted basic entity objects, establish a tree structure according to the hierarchical relationship of project, profession, facility, sub-facility, and component; S1.

3. Add enhanced semantic information to each type node. In addition to classification coding, tree structure hierarchy coding, and name, add semantic labels and route feature labels to form a semantically enhanced engineering object type layer.

3. The semantically enhanced graph-tree hybrid storage and dynamic update method for intelligent highways according to claim 1, characterized in that, Step S2 includes: S2.

1. Parse the digital highway data, extract the attribute content, merge and remove duplicates of those with identical names and those with different names but the same meaning, and perform normalization preprocessing on the original names of the attribute objects to obtain a preliminary list of attribute types. S2.2 Establish attribute type grouping, including identification information, location information, dimension information, and engineering information, and establish a tree structure according to the hierarchical relationship of project, profession, attribute grouping, and attribute type; S2.3 Calculate the semantic similarity between attribute types for the unit, attribute value, and field name in sequence; S2.4 After merging, add a standardized name, the original field name before merging, unit, data type, value range and validation rules to each attribute type node to achieve automatic semantic fusion of heterogeneous attributes.

4. The semantically enhanced graph-tree hybrid storage and dynamic update method for intelligent highways according to claim 3, characterized in that, Step S2.3 includes: ① Establish a knowledge base for unit conversion rules, including relationships between units that can be directly converted and relationships between equal units, and calculate the consistency similarity of units. If the units can be directly converted, then =1; if the units belong to the same category but cannot be directly converted, then =0.7; If the units are not of the same type and cannot be converted, then... =0; reserved The attribute type; ② Calculate the similarity of numerical ranges Extract the attribute value or random value v within the numerical range, and calculate it using formula (1). If it is a non-numerical type, directly take 0: Official (1), In the formula: The value is between 0 and 1; The value of attribute 1; The value of attribute 2; ③ Calculate the semantic similarity of field names The field names are converted into vectors A or B using an embedding model, and then calculated using formula (2): Official (2), In the formula: The value is between 0 and 1; A is the vector representation of field name 1; B is the vector representation of field name 2; ④ Calculate the overall similarity using formula (3) : Official (3), In the formula: , , Configurable weights; ⑤ When the overall similarity When the similarity threshold is exceeded, they are determined to have the same semantic attribute and are merged into a single standard attribute type node.

5. The semantically enhanced graph-tree hybrid storage and dynamic update method for intelligent highways according to claim 1, characterized in that, In step S4.4, the basic fields include attribute name, unit, value, and version information.

Images