A method for constructing a hazard-bearing body based on a structure-function-behavior model

Abstract

This invention relates to a method for constructing a disaster-bearing entity based on a structure-function-behavior model, comprising the following steps: S1 entity structural semantic modeling; S2 entity functional semantic modeling; S3 behavioral semantic modeling; S4 constructing a semantic fusion module and building urban natural disaster risk factor data and knowledge data, and using the risk factor data and knowledge data to fuse the semantics formed in the models built in steps S1-S3 to realize the construction of the disaster-bearing entity. The method also includes a data and mechanism-driven twin model update method, specifically including: updating the twin model using a change detection model, and update management, wherein the change detection model includes the detection of structural and semantic changes. The method also includes a disaster risk prediction method, which uses a hybrid approach consisting of decision trees and node data formed by urban natural disaster risk factor data and knowledge data. This achieves data analysis and prediction of urban entity disaster situations with visualization of the sum of structure-function-behavior factors.

Description

Technical Field

[0001] This invention relates to a method for constructing disaster-bearing bodies based on a "structure-function-behavior" model, which belongs to the fields of digital twin data and emergency response. Background Technology

[0002] In urban natural disaster risk prevention and control applications, it is typically necessary to have a holistic understanding and unified expression of the coupling relationships among multiple elements of urban natural disaster risk in order to achieve accurate risk assessment and refined prevention and control. Existing technologies mainly adopt traditional GIS spatiotemporal data models, which have many problems: they are difficult to effectively express the complex coupling relationships among multiple elements of urban natural disaster risk, namely "human-machine-object-environment," and cannot meet the analysis needs of the dynamic evolution of multi-granular spatiotemporal objects throughout their entire life cycle; they lack the ability to handle multi-scale ("region-individual-component") and multi-level ("above-ground-surface-indoor-underground") urban natural disaster risk elements, as well as information integration, efficient calculation, and coupled simulation analysis; and digital twin models are difficult to update in a timely and accurate manner according to changes in scene elements and states. Summary of the Invention

[0003] 1. Core Technical Solution

[0004] First, we will establish a semantic expression framework for urban natural disaster risk elements, construct a conceptual model of multi-domain consistent association and a logical model of multi-dimensional information integration, and break through the limitations of traditional GIS spatiotemporal data models.

[0005] Second, we will study the twin modeling technology of urban natural disaster risk factors based on "structure-function-behavior", which covers semantic modeling of entity structure, semantic modeling of entity function, semantic modeling of entity behavior and semantic fusion, to achieve multi-scale and multi-level modeling and information integration expression.

[0006] Third, we will develop a twin model update technology driven by both research data and mechanisms, including automatic change detection technology and update management mechanism, to achieve dynamic updates of the model.

[0007] 2. Technical Solution Content

[0008] To achieve the above method, the present invention provides a method for constructing a disaster-bearing body based on a structure-function-behavior model, comprising the following steps:

[0009] S1 entity structure semantic modeling;

[0010] S2 entity functional semantic modeling;

[0011] S3 behavioral semantic modeling;

[0012] S4 constructs a semantic fusion module and builds urban natural disaster risk factor data and knowledge data. It uses the risk factor data and knowledge data to fuse the semantics formed in the models built in steps S1-S3 to realize the construction of the disaster-bearing body.

[0013] Specifically, entity structure semantic modeling includes:

[0014] The construction of the S1-1 area is used to build the macro-level urban physical structure;

[0015] The construction of S1-2 is used to express the mesoscopic urban entity structure;

[0016] The modeling of S1-3 components is used to represent the internal structure of microscopic urban entities.

[0017] Optionally, the entity structure includes a 3D terrain model, a 3D building model, and a 3D geological model, and the individual unit is formed based on GeoAI's urban entity semantic modeling; the component includes GeoAI-based urban entity semantic modeling and point cloud-based indoor modeling.

[0018] Specifically, entity functional semantic modeling includes:

[0019] S2-1 Extracts entity physical, behavioral, and relational information;

[0020] S2-2 defines semantic description rules to achieve the normalization of entity function semantics of S2-1 information;

[0021] S2-3 performs a semantic quality check on the entity functions of S2-2.

[0022] Specifically, behavioral semantic modeling includes:

[0023] S3-1 Entity Behavior Definition;

[0024] S3-2 Construction of Behavioral Decision Tree;

[0025] S3-3 behavior is to be standardized.

[0026] The methods of semantic fusion include:

[0027] S4-1 Semantic Matching and Alignment;

[0028] S4-2 Semantic merging and disambiguation;

[0029] S4-3 semantic mapping;

[0030] S4-4 Quality Inspection.

[0031] The method also includes a data- and mechanism-driven twin model update method, specifically including: updating the twin model using a change detection model, and update management, wherein the change detection model includes the detection of structural changes and semantic changes.

[0032] Another aspect of this invention is to propose a disaster risk prediction method based on the above method, comprising the following steps:

[0033] Q1 calls upon urban natural disaster risk factor data and knowledge data, constructs data nodes from the urban natural disaster risk factor data and knowledge data respectively, and defines them as the first type of node and the second type of node respectively.

[0034] Q2: Construct the first-type edges between the first-type nodes and the second-type nodes, the relationships between the first-type edges, and the node risk level of the first-type nodes.

[0035] Q3 forms a second type of edge and the second type of edge relationship between any leaf node on the behavior decision tree and the first type of node, thereby visualizing the entity structure in the semantic fusion result, thus constructing a visualization structure-function-behavior hybrid heterogeneous graph on the entity structure formed by the first type of node, the second type of node, leaf node, and the first type of edge and the second type of edge.

[0036] Q4 vectorizes the data in the first type of nodes, the second type of nodes, and the leaf nodes, and aggregates them at the node level and the path level. Thus, by selecting any node in the visualized mixed heterogeneous graph, we can find the data in the top 3-5 neighboring nodes with the highest correlation to the data in that node and the path of the associated disaster, thereby forming a risk level prediction.

[0037] It is understandable that hybrid heterogeneous graphs differ from ordinary heterogeneous graphs in that they integrate the tree nodes of decision trees into the heterogeneous graph. They can be used to analyze behavioral relationships in the decision tree alone, or to study the comprehensive relationship between the decision tree and entity structure, risk factors and disaster knowledge.

[0038] 3. Beneficial effects

[0039] This patent designs a method for constructing disaster-bearing entities based on a structure-function-behavior model, breaking through the technical bottleneck of holistic understanding and unified expression of the coupling relationships among multiple elements of urban natural disaster risk. Based on the structure-function-behavior twin model construction method, through the materialization, abstraction, and semantic unified description of the physical entities of urban natural disaster risk elements such as people, machines, objects, and environment, a consistent and interconnected digital twin entity of urban natural disaster risk elements is established, which integrates geometric models, physical models, behavioral models, and rule models. This achieves high-fidelity twinning of the multi-domain characteristics of the physical entities of urban natural disaster risk elements and automatic updates to dynamic changes.

[0040] Based on the above structure-function-behavior model for constructing disaster-bearing entities, this invention also establishes a hybrid heterogeneous graph, integrating behavioral decision trees into the heterogeneous graph, forming a joint data analysis and risk prediction strategy for entity structure, function, and behavior. Attached Figure Description

[0041] Figure 1 Roadmap for Digital Twin Modeling and Updating of Comprehensive Risk Factors of Urban Natural Disasters

[0042] Figure 2 Research roadmap for entity semantic representation of urban natural disaster risk factors

[0043] Figure 3 A unified conceptual model for digital twin representation of urban natural disaster risk factors.

[0044] Figure 4 A logical model for integrating multi-dimensional information on urban natural disaster risk factors.

[0045] Figure 5 Technical route for twin modeling of urban natural disaster risk factors

[0046] Figure 6 Entity functional semantic modeling technology roadmap

[0047] Figure 7 A data- and mechanism-driven approach to twin model updates.

[0048] Figure 8 The disaster risk prediction method based on the method described in Embodiment 2 of the present invention

[0049] Figure 9a A schematic diagram of a fire at a supermarket next to a motor vehicle lane in Embodiment 3 of the present invention.

[0050] Figure 9b Indicated in the solid structure diagram Figure 9a Diagram illustrating the alarm system at the fire scene. Detailed Implementation

[0051] Example 1

[0052] This embodiment will describe the underlying logic of a method for constructing a disaster-bearing entity based on a structure-function-behavior model. For example... Figure 1 As shown, by combining the entity characteristics of urban natural disaster risk elements, a semantic expression framework for urban natural disaster risk element entities is summarized. A twin modeling technology for urban natural disaster risk elements based on "structure-function-behavior" and a dual-driven update technology of data and mechanism are developed to realize the construction of digital twin data of urban natural disaster risk elements that are integrated on the ground and underground, and static and dynamic.

[0053] This method primarily involves entity model construction, functional information extraction, behavioral description modeling, and semantic fusion of information and behavioral data based on the extracted functional information, ultimately combining risk factor information to construct a twin model. The main technical aspects involved are as follows.

[0054] First, semantic representation of urban natural disaster risk elements.

[0055] like Figure 2 As shown, starting from the holistic cognitive expression and dynamic consistency intelligent application requirements of digital twins, and combining the ideas of systems theory and holism, this paper analyzes the multi-element characteristics, interrelationships, functional characteristics, and physical behaviors of urban natural disaster risks, including "human-machine-object-environment". Based on this, it conceptualizes and abstracts the multi-domain characteristics of urban natural disaster risk elements in terms of "structure-function-behavior", multi-scale in terms of "region-individual-component", and multi-dimensional in terms of "data-model-knowledge", establishing an ontological framework for a unified expression of urban natural disaster risk elements and their relationships. It constructs a unified conceptual model for the entity expression of urban natural disaster risk elements, with multi-level spatiotemporal changes of "process-entity-state" as the core, supporting the construction of "entityization-structuring-semanticization" twin models for different application needs. Finally, it establishes associations through unique spatial identity coding of entities and designs a logical model for the integration of multi-dimensional information of urban natural disaster risk element entities.

[0056] Second, a conceptual model of multi-domain consistency association of urban natural disaster risk elements.

[0057] like Figure 3 As shown, this paper addresses the technical bottlenecks in the holistic understanding and unified expression of the coupling relationships among multiple elements of urban natural disaster risk. Addressing the limitations of traditional GIS spatiotemporal data models, it uses "spatiotemporal change" as the holistic cognitive framework and urban natural disaster risk element entities as the core elements. It systematically analyzes the coupling relationships among multiple elements of urban natural disaster risk, including "human-machine-object-environment," and dissects the dynamic evolution of multi-granularity spatiotemporal objects throughout their entire lifecycle. From the perspectives of multi-domain ("structure-function-behavior"), multi-scale ("region-individual-component"), and multi-level ("above-ground-surface-indoor-underground") and interaction relationships of urban natural disaster risk element entities, it abstracts, summarizes, and unifies the description, developing a unified conceptual model for the unified expression of urban natural disaster risk elements using digital twins with multi-domain spatiotemporal consistency.

[0058] This model expands the three inherent characteristics of spatiotemporal elements—"space," "time," and "attributes"—into "geometric structural objects that express the spatial structure of entities," "behavioral objects that express the behavioral processes of entities," and "functional objects that comprehensively express the physical attributes, mechanistic attributes, and relationships of entities." This is used to integrate multidimensional information of multiple elements of urban natural disaster risk objects through spatiotemporal benchmarks, semantic references, and correlations, supporting a high-fidelity characterization of the multi-domain characteristics and dynamic changes of urban natural disaster risk element entities.

[0059] Geometric Structure Objects: For different types of entities of urban natural disaster risk elements and the need for multi-scale expression of "region-unit-component", spatial features are described for the location, shape, composition, range, geometric path, etc. of elements. It consists of multi-level geometric structure, structural semantics, spatial relationships, etc.

[0060] Behavioral objects: Model the temporal characteristics, behavioral processes and relationships of urban natural disaster risk elements, describe the discrete or continuous changes of urban natural disaster risk elements, and emphasize the expression of spatiotemporal change correlation, spatiotemporal change characteristics-pattern-mechanism.

[0061] Functional objects: semantically describe and express the multi-dimensional inherent attributes of urban natural disaster risk factors. Specifically, this includes: physical state semantics describing the inherent physical properties and characteristics of entities; functional state semantics describing the professional skill status of entities; functional capability semantics describing the magnitude of professional skill capabilities of entities; and relational semantics describing the spatial relationships and connections between entities.

[0062] Third, a logical model for integrating multi-dimensional information on urban natural disaster risk factors.

[0063] like Figure 4 As shown, in order to effectively express the static and dynamic relationships between entities of urban natural disaster risk elements and the relationships between different types of entities, and to achieve "interoperability" between physical space and digital space, this project designed a logical model for multi-dimensional information integration and establishing associations through unique identifiers.

[0064] Example 2

[0065] This embodiment presents an example of a twin modeling technique for urban natural disaster risk factors based on "structure-function-behavior". For example... Figure 5 As shown, in order to meet the needs of multi-scale (regional-individual-component) and multi-level (above-ground-surface-indoor-underground) information integration, efficient calculation, and coupled simulation analysis of urban natural disaster risk factors, this project studies a twin fusion modeling method for coupled, interconnected, and computable urban natural disaster risk factors. This method is based on a structure-function-behavior model to construct disaster-bearing bodies. Figure 1 Within the framework shown, the method includes the following steps:

[0066] S1 entity structure semantic modeling;

[0067] S2 entity functional semantic modeling;

[0068] S3 behavioral semantic modeling;

[0069] S4 constructs a semantic fusion module and builds urban natural disaster risk factor data and knowledge data. It uses the risk factor data and knowledge data to fuse the semantics formed in the models built in steps S1-S3 to realize the construction of the disaster-bearing body.

[0070] Regarding the implementation of semantic modeling of entity structures

[0071] To achieve semantic modeling of entity structures, this study investigates key technologies for modeling at different scales, including large-scale terrain 3D reconstruction, 3D geological body modeling, GeoAI-based urban entity semantic modeling, GeoAI-based building layered and unit-based modeling, and point cloud-based indoor modeling.

[0072] Entity functional semantic modeling designs entity functional semantic models based on the needs of urban natural disaster risk prevention and control. It employs techniques such as direct transformation extraction, computational analysis, and survey data collection to extract physical, behavioral, and relational information of entities. Semantic description rules are defined to provide a foundation for standardized descriptions of entity functions. Quality checks are performed on the entity semantic data, including checks for semantic correctness, consistency, completeness, and standardization. Its technical approach is as follows: Figure 6 As shown.

[0073] Specific design entity functional semantic model

[0074] Based on the needs of natural disaster risk prevention and control, the physical functions of natural disaster risk elements are defined as follows: Table 1:

[0075]

[0076] Semantic information extraction employs techniques such as direct transformation extraction, computational analysis, and survey data collection to extract semantic information about entity functions.

[0077] Semantic normalization is a process that aims to ensure the integrity and consistency of semantics by defining semantic description rules for different functional semantic types and processing the extracted semantic information according to these rules.

[0078] Semantic quality checks include: completeness check: checking whether the semantic output contains extracted functional information; correctness check: checking whether the semantic content in the semantic output file conforms to the entity functional semantic description rules; and consistency check: checking whether the actual data is consistent with the semantic output. For example, if the semantic relationship between an entity and another entity in the semantic output is "belonging," but the semantic relationship between the same entity and another entity in the actual data is "composition," then the actual data is inconsistent with the semantic output.

[0079] Semantic modeling of entity behavior

[0080] The behavior rules of entities are described by decision trees. The behavior state of an entity is determined by the value of the decision attribute, and then the entire decision tree is generated to formally describe the behavior rules of the entity.

[0081] Formal Description of Behavioral Rules: Decision trees provide an intuitive and visual representation for formally modeling the behavioral rules of entities. Through decision trees, the behavioral characteristics and decision logic of entities can be displayed in a tree structure, achieving a standardized expression of behavioral semantics.

[0082] The relationship between entity behavior state and decision attributes: An entity's behavior state can be understood as a relatively singular task or behavioral capability exhibited in the simulated world, such as the opening and closing of a dam gate. State generation rules determine the behavior state exhibited by the entity, and these rules can determine the corresponding behavior state of the entity based on the values ​​of a series of decision attributes.

[0083] The process of building a decision tree includes the following steps:

[0084] Analyze the behavioral characteristics of entities, determine the decision attributes that affect the behavioral state and their value space, and form a set of behavioral states, a set of decision attributes, and a set of decision attribute values;

[0085] Based on expert analysis or decision attribute ranking algorithms, decision attributes are ranked according to their degree of influence to form a decision attribute ranking set.

[0086] The decision attribute for sorting is used as the root node, and the corresponding branches are drawn according to the value of the decision attribute.

[0087] Under each branch, continue to draw the corresponding branches and leaf nodes using the ranked decision attributes as internal nodes, until the entire decision tree is constructed.

[0088] On semantic fusion

[0089] Semantic matching and alignment uses methods such as cosine similarity and edit distance to calculate the semantic similarity between words in order to determine the degree of association.

[0090] Semantic merging and disambiguation involve merging the semantic content of aligned semantics. When semantic conflicts are detected, they are resolved.

[0091] Semantic mapping is the process of converting one semantic representation into another. Its main purpose is to establish connections between different semantic systems, vocabularies, knowledge structures, or language expressions, thereby enabling functions such as knowledge sharing, information integration, and cross-language understanding.

[0092] Quality checks include correctness checks: checking whether the semantic content in the semantic output files conforms to the entity semantic description rules; and consistency checks: checking whether the actual data is consistent with the semantic outputs.

[0093] Twin model update technology driven by both data and mechanism

[0094] Digital twin model enhancement and update processing uses accurately detected scene elements and state change information as input. It conducts research on digital twin model enhancement processing for complex spaces at three granularities: macro-regional scene, meso-level individual elements, and micro-level components, to achieve digital twin model updates. The technical approach implemented is as follows: Figure 7 As shown. This includes:

[0095] (1) Automatic change detection technology

[0096] 1) Structural change detection

[0097] Changes in the spatial structure of point elements: The association and matching of old and new point elements can be calculated using Euclidean distance, that is, taking the base state point element as the center, searching for the latest temporal point elements with the same position or whose Euclidean distance is within a certain threshold range;

[0098] Spatial structure changes of line features: Spatial correlation and matching of old and new line features can be achieved using geometric analysis methods such as length overlap and Hausdorf distance. The Hausdorf distance method calculates the Hausdorf distance between old and new lines, and the length overlap is calculated by overlaying the lines to complete the detection process. The point set of the line feature is simplified to the start point, end point, center point, and several other points obtained on the line according to a percentage of length.

[0099] Spatial structure changes of polygon features: Spatial correlation matching between new and old linear features can be achieved through spatial correlation detection. This involves generating a buffer zone with the new feature A, obtaining candidate features B that intersect with the new feature's buffer zone, and calculating the area ratio of the intersection and union of these two feature buffer zones. Change detection reveals several types of changes to geographic features, including additions (new / old feature 0:1), deletions (new / old feature 1:0), merging (new / old feature n:1), splitting (new / old feature 1:n), aggregation (new / old feature m:n), and geometric and positional changes (new / old feature 1:1).

[0100] 2) Functional semantic change detection

[0101] The functional update of a digital twin model mainly involves processing the semantic content of entity elements. In the specific implementation of the model functional update, accurate detection of scene elements and state change information is used as input, combined with model structure change data and real-time state information to conduct semantic change detection research. This mainly includes semantic change extraction, semantic content determination, and semantic content acquisition.

[0102] (2) Update Management

[0103] In the digital twin model update process, if changes occur due to additions or deletions, the information is directly added or deleted in the database. If the model's spatial and attribute information changes, the old information is deleted first, and then the new information is written. During the model update process, the change information will be stored in a change process database, and the deleted elements will be stored in a history database. The model update information must at least include the identifier of the changed elements, spatial information, attribute information, and time information.

[0104] Data deleted and modified during model updates will be stored in a historical database. To facilitate the retrieval of historical data, a representation of historical information will be established with reference to the spatiotemporal data model. This representation should include identifiers of model deletion and modification changes, spatial information, attribute information, and the start time, end time, and update time of historical data storage.

[0105] Example 3

[0106] This embodiment is a disaster risk prediction method based on the method described in Embodiment 2, such as... Figure 8 As shown, it includes the following steps:

[0107] Q1 calls upon urban natural disaster risk factor data and knowledge data, constructs data nodes from the urban natural disaster risk factor data and knowledge data respectively, and defines them as the first type of node and the second type of node respectively.

[0108] Q2: Construct the first-type edges between the first-type nodes and the second-type nodes, the relationships between the first-type edges, and the node risk level of the first-type nodes.

[0109] Q3 forms a second type of edge and the second type of edge relationship between any leaf node on the behavior decision tree and the first type of node, thereby visualizing the entity structure in the semantic fusion result, thus constructing a visualization structure-function-behavior hybrid heterogeneous graph on the entity structure formed by the first type of node, the second type of node, leaf node, and the first type of edge and the second type of edge.

[0110] Q4 vectorizes the data in the first type of nodes, the second type of nodes, and the leaf nodes, and aggregates them at the node level and the path level. Thus, by selecting any node in the visualized mixed heterogeneous graph, we can find the data in the top 3 neighboring nodes with the highest correlation to the data in that node and the path of the associated disaster (as shown in P1P2L in the figure), thereby forming a risk level prediction.

[0111] Thus, in Figure 9a A fire broke out at a supermarket next to the motor vehicle lane, with black smoke billowing from the entrance. Figure 9b A schematic diagram of the physical structure of the incident site is provided. When the mouse clicks on the alarm dot at the incident site, all relevant risk factor information and knowledge data for disaster analysis will pop up, and a mixed heterogeneous graph of structure, function, and behavior will be visualized, such as... Figure 8 The path P1P2L shown is a path in this heterogeneous graph.

[0112] By searching urban natural disaster risk factor data and knowledge data, one can see whether the supermarket's fire safety facilities are complete, the frequency of pedestrian access to automatic sliding doors, whether the foot traffic frequency exceeds the standard, and whether there are any ignition-inducing behaviors detected by surveillance cameras. It can also be determined whether there are any ignition-inducing factors in other neighboring nodes of the physical structure. This analysis suggests that the original risk level was very high, leading to... Figure 9a The scenario depicted is not unexpected. This demonstrates the crucial role that the method of this invention will play in future disaster prevention efforts.

[0113] Specifically, a Long Short Memory (LSTM) model is constructed by forming node units with the top 3-5 neighbors or predetermined path nodes that are associated with the predicted node. The data nodes are then used to train the model to obtain the risk level prediction for each node.

Claims

1. A disaster risk prediction method for disaster-bearing bodies based on a structure-function-behavior model, characterized in that, Includes the following steps: S1 entity structure semantic modeling; S2 entity functional semantic modeling; S3 Behavioral Semantic Modeling; including: S3-1 Entity Behavior Definition; S3-2 Behavioral Decision Tree Construction; S3-3 Behavior Normalization Process; wherein, the process of constructing the decision tree includes: analyzing the behavioral characteristics of the entity, determining the decision attributes that affect the behavioral state and their value space, forming a behavioral state set, a decision attribute set, and a decision attribute value set; according to expert analysis or a decision attribute ranking algorithm, ranking the decision attributes according to their degree of influence, forming a ranked decision attribute set; using the ranked decision attributes as the root node, drawing corresponding branches according to the value of the decision attribute; under each branch, continuing to use the ranked decision attributes as internal nodes, drawing corresponding branches and leaf nodes, until the entire decision tree is constructed; S4 constructs a semantic fusion module and builds urban natural disaster risk factor data and knowledge data. It uses the risk factor data and knowledge data to fuse the semantics formed in the models built in steps S1-S3 to realize the construction of the disaster-bearing body. It also includes the following steps: Q1 calls up urban natural disaster risk factor data and knowledge data, and constructs data nodes for urban natural disaster risk factor data and knowledge data respectively, defining them as the first type of node and the second type of node respectively; Q2 Construct the first type of edges between the first type of nodes and the second type of nodes, the relationships between the first type of edges, and the node risk level of the first type of nodes; Q3 forms a second type of edge and the second type of edge relationship between any leaf node on the behavior decision tree and the first type of node, thereby visualizing the entity structure in the semantic fusion result, thus constructing a visualization structure-function-behavior hybrid heterogeneous graph on the entity structure formed by the first type of node, the second type of node, leaf node, and the first type of edge and the second type of edge. Q4 vectorizes the data in the first type of nodes, the second type of nodes, and the leaf nodes, and aggregates them at the node level and the path level. Thus, by selecting any node in the visualized mixed heterogeneous graph, we can find the data in the neighboring nodes with the highest correlation to the data in that node and the path of the associated disaster, thereby forming a risk level prediction.

2. The method according to claim 1, characterized in that, in, Entity structure semantic modeling specifically includes: The construction of the S1-1 area is used to build the macro-level urban physical structure; The construction of S1-2 is used to express the mesoscopic urban entity structure; The modeling of S1-3 components is used to represent the internal structure of microscopic urban entities.

3. The method according to claim 2, characterized in that, The entity structure includes a 3D terrain model, a 3D building model, and a 3D geological model. The individual units are formed based on GeoAI's urban entity semantic modeling. The components include GeoAI-based urban entity semantic modeling and point cloud-based indoor modeling.

4. The method according to claim 2, characterized in that, in, Entity functional semantic modeling specifically includes: S2-1 Extracts entity physical, behavioral, and relational information; S2-2 defines semantic description rules to achieve the normalization of entity function semantics of S2-1 information; S2-3 performs a semantic quality check on the entity functions of S2-2; Semantic fusion methods include: S4-1 semantic matching and alignment; S4-2 Semantic merging and disambiguation; S4-3 semantic mapping; S4-4 Quality Inspection.

5. The method according to claim 4, characterized in that, Entity functional semantic modeling employs direct conversion extraction, computational analysis, and survey data collection methods to extract physical, behavioral, and relational information of entities; it defines semantic description rules and performs quality checks on entity semantic data in steps S2-3, including checks on semantic correctness, consistency, completeness, and standardization. Based on the needs of natural disaster risk prevention and control, the physical functions of natural disaster risk elements are defined as follows: ， Semantic matching and alignment use cosine similarity and edit distance methods to calculate the semantic similarity between words to determine the degree of association; Semantic merging and disambiguation are processes that merge the semantic content of aligned semantics. When a semantic conflict is detected, conflict resolution is performed. Semantic mapping establishes connections between different semantic systems, vocabularies, knowledge structures, or language expressions. The quality checks in step S4-4 include: correctness check: checking whether the semantic content in the semantic output file conforms to the entity semantic description rules; consistency check: checking whether the actual data is consistent with the semantic output.

6. The method according to claim 4, characterized in that, The method also includes a data- and mechanism-driven twin model update method, specifically including: updating the twin model using a change detection model, and update management, wherein the change detection model includes the detection of structural changes and semantic changes.

7. The method according to claim 6, characterized in that, Structural change detection includes: Changes in the spatial structure of point elements: The association and matching of old and new point elements can be calculated using Euclidean distance. That is, with the base state point element as the center, search for the latest temporal point elements that are in the same position or whose Euclidean distance is within a certain threshold range. Line feature spatial structure change: Spatial association matching of new and old line features can be performed using geometric analysis methods such as length overlap and Hausdorf distance; the Hausdorf distance method is used to calculate the Hausdorf distance between new and old lines, and the length overlap is calculated by overlaying lines to complete the detection process. The point set of line features is simplified to the start point, end point, center point, and several other points obtained on the line according to the length percentage. Spatial structure changes of surface features: Spatial association matching of new and old line features can be achieved through spatial association detection. That is, a buffer zone is generated with the new feature A, and candidate features B that intersect with the buffer zone of the new feature are obtained. The area ratio of the intersection and union of these two feature buffer zones is calculated. The change types of geographic features obtained through change detection include addition, deletion, merging, splitting, aggregation, geometric shape and location changes. Functional semantic change detection includes: The main function update of the digital twin model is to process the semantic content of the entity elements. When the function update is implemented, the scene elements and state change information are accurately detected as input, and semantic change detection research is carried out in combination with model structure change data and real-time state information. This mainly includes semantic change extraction, semantic content determination, and semantic content acquisition. Update management includes: In the digital twin model update process, if the model is added or deleted, the information is added or deleted directly in the database; if the model's spatial and attribute information changes, the old information of the model is deleted first, and then the new information is written. During the model update process, the change information is stored in the change process database, and the deleted elements are stored in the history database. The model update information includes the change element identifier, spatial information, attribute information, and time information. Data deleted and modified during model updates will be stored in a historical database. To facilitate the retrieval of historical data, a representation of historical information will be established with reference to the spatiotemporal data model. This representation should include identifiers for model deletion and modification changes, spatial information, attribute information, the start time, end time, and update time of historical data storage.

8. The method according to any one of claims 1-7, characterized in that, By forming node units with the top 3-5 neighbors or predetermined path nodes of the predicted node, a Long Short Memory (LSTM) model is constructed. The risk level prediction for each node is obtained by training the data nodes.

Images