Safety control method based on knowledge graph

Abstract

The invention discloses a safety control method based on a knowledge graph, and belongs to the field of safety control, and the method comprises the following steps: crawling various types of data according to keywords through a crawler program to construct a fireproof database; carrying out semantic analysis on data in the fireproof database by using a natural language processing technology, and carrying out data processing on the fireproof database after semantic analysis through word vector modeling and entity relationship extraction; constructing an internal large-scale language model by using the processed fireproof database, and performing hierarchical clustering on the internal large-scale language model by adopting a Leiden technology; establishing a multi-dimensional mapping model based on historic building spatial features, cultural relic features and disaster-inducing factors in a hierarchical clustering result, and constructing a multi-modal dynamic association fireproof knowledge graph by using a knowledge graph technology based on the multi-dimensional mapping model; and dynamically selecting a search mode based on a user query keyword, and generating a security control result according to the query keyword based on the search mode.

Description

Technical Field

[0001] The present invention relates to the field of security control technology, and in particular to a security control method based on a knowledge graph. Background Art

[0002] my country boasts a rich cultural heritage, encompassing a vast collection of cultural relics and precious historical sites. Fires in cultural heritage buildings present high risks, significant hidden dangers, destructive potential, and significant firefighting challenges. The enormous costs of monitoring and maintenance have become a major pain point in the management of cultural heritage sites. Once a fire occurs, it inevitably leads to irreparable damage. Traditional fire prevention methods have numerous limitations, making it difficult to achieve precise control and rapid response to complex and ever-changing fire hazards. As cultural heritage preservation efforts continue to advance, the research, development, and application demonstration of intelligent active fire protection systems for cultural heritage buildings are becoming increasingly important.

[0003] A knowledge graph is a series of graphs that display the development and structural relationships of knowledge. It mimics the composition of human structured knowledge through a network structure of nodes (entities) and edges (relationships). Knowledge graphs not only capture raw information but also high-order relationships across multiple documents. They also possess powerful reasoning capabilities and are widely applicable across multiple industries, including fire prevention for cultural relics and ancient buildings. Based on the above, the present invention aims to address the following issues: 1. Traditional fire prevention methods have limitations. Ancient buildings have multiple wooden components, and fire hazards are complex and varied. Current technology cannot meet the fire prevention needs of cultural relics buildings. 2. It is difficult to accurately locate the fire point of ancient cultural relics, making it difficult to control the fire in the first place; 3. Cultural heritage protection requires compliance with the principle of minimum intervention in fire prevention and disposal of cultural heritage buildings, and it is necessary to minimize human intervention in cultural heritage buildings; 4. Cultural relics and ancient buildings are easily affected by various disaster-causing factors such as weather, building materials, building structure, and historical renovations, making it difficult to comprehensively assess the impact level and fire-fighting plans. Summary of the Invention

[0004] In response to the above-mentioned problems, the present invention provides a security control method based on knowledge graph.

[0005] A security control method based on knowledge graph, comprising the following steps: A fire prevention database is constructed by crawling various types of data on ancient building fire prevention, cultural relic fire prevention, fire accidents and material coefficients based on keywords through a crawler program; Using natural language processing technology to perform semantic analysis on the data in the fire prevention database, and performing data processing on the semantically analyzed fire prevention database through word vector modeling and entity relationship extraction; Use the processed fire prevention database to build an internal large-scale language model, and use the Leiden technique to perform hierarchical clustering on the internal large-scale language model; A multi-dimensional mapping model is established based on the spatial characteristics of ancient buildings, cultural relics characteristics and disaster-causing factors in the hierarchical clustering results, and a multi-modal dynamic association fire prevention knowledge graph is constructed based on the multi-dimensional mapping model using knowledge graph technology; Based on the user's query keyword, a search mode is dynamically selected, and based on the search mode, a security control result is generated according to the query keyword.

[0006] Preferably, the fire prevention database is constructed by crawling various types of data on ancient building fire prevention, cultural relic fire prevention, fire accidents and material coefficients according to keywords through a crawler program, including: Identify multiple data sources for obtaining fire protection standards for ancient buildings, fire protection standards for cultural relics, fire accident data, and fire protection coefficients of materials; Based on the particularity of fire prevention of ancient buildings and cultural relics, a multi-level keyword library is constructed; crawling various types of data containing keywords from the multiple data sources according to the multi-level keyword library using a crawler program; Removing duplicated captured data of various types through hash value comparison, storing the data of various types in a database, and retaining the data source and capture timestamp; Experts review the specific parameters of the various types of data in the database, label the reliability level based on the data source, and determine that the database is a fire prevention database.

[0007] Preferably, the use of natural language processing technology to perform semantic analysis on the data in the fire prevention database, and the data processing of the semantically analyzed fire prevention database through word vector modeling and entity relationship extraction, include: Using AI technology to standardize various types of data in the fire prevention database and convert the data in the database into natural language data; Segmenting the natural language data into text units, and distributing the text units in a concept vector space from near to far according to the relevance of the text units; Using word vector similarity retrieval, automatically expanding the synonym library, adjusting the relevance weights of synonyms, and adjusting the distribution of the text units based on the weight adjustment results; Entities are extracted from the text units, the entities are classified, association relationships are constructed between each entity classification, and a structured data system is generated.

[0008] Preferably, the processing of the fire prevention database to construct an internal large language model and performing hierarchical clustering on the internal large language model using the Leiden technique include: Acquire the structured data system in the fire prevention database after preprocessing, and determine the model architecture of a large language model based on the text unit characteristics of the structured data system; The pre-processed structured data is mixed with the original text and input into a large language model to mask the entities in the fire prevention database so that the model can learn to determine domain knowledge associations; Abstracting the parameters of the large language model into graph nodes, with the weights between nodes determined by parameter similarity, to generate a high-dimensional parameter relationship structure; The Leiden algorithm is applied to detect the high-dimensional parameter relationship structure, and highly correlated internal groups are merged to form basic functional modules.

[0009] Preferably, the application of the Leiden algorithm to detect the high-dimensional parameter relationship structure and merge highly correlated internal groups to form a basic functional module includes: Abstracting the internal parameters of the language model into graph nodes, wherein the graph nodes contain the term vector table weight distribution and entity relationships of the text unit; Calculating the correlation of the internal parameters of the language model by cosine similarity, and defining edge weights based on the correlation of the internal parameters of the language model; Use the Leiden algorithm to assign each node to an independent community, traverse the nodes, calculate the modularity gain when each independent community is moved to an adjacent independent community, select the maximum gain direction to iteratively adjust the independent community affiliation, and generate a community set of the independent community; If the edge weight between the two community sets exceeds a preset threshold, they are merged into the same functional module, and the same functional module is determined as the basic functional module; The high-frequency text units of the community set in the basic function module are extracted, matched with the standard function description, and the module label is generated.

[0010] Preferably, the Leiden algorithm is used to assign each node to an independent community, traverse the nodes, calculate the modularity gain when each independent community is moved to an adjacent independent community, select the maximum gain direction to iteratively adjust the independent community affiliation, and generate the community set of the independent community using the following algorithm:

[0011] in, is the edge weight, is the node degree, is the total edge weight, Used to determine whether nodes belong to the same independent community, the algorithm is used to measure the difference between the internal connection density of the independent community and the random distribution, and determine the specific modularity gain based on the difference.

[0012] Preferably, the method of establishing a multidimensional mapping model based on the spatial characteristics of ancient buildings, cultural relics characteristics and disaster-causing factors in the hierarchical clustering results, and constructing a multimodal dynamic association fire prevention knowledge graph based on the multidimensional mapping model using knowledge graph technology includes: Extracting hidden layer representations of the ancient building space features and the cultural relic features from the basic functional modules, and generating standardized feature vectors through graph embedding technology; Extracting the clustering results of disaster-causing factors from the basic functional module, converting the clustering result output into a probability distribution or a risk score vector, and retaining the association weights between the clustering results and spatial features; Build a multi-branch neural network to process the feature vector and the disaster-causing factors separately, realize cross-modal feature interaction, and dynamically adjust the contribution of different disaster-causing factors to the feature vector through the attention mechanism to generate a multi-dimensional model; Mapping the output of the multidimensional model into graph node attributes, establishing a knowledge graph based on the graph nodes, analyzing the risk evolution path through knowledge graph modeling, and applying association rules to mine potential risk coupling relationships; Based on the risk evolution path and the potential risk coupling relationship, a multimodal dynamically associated fire prevention knowledge graph is constructed according to the knowledge graph.

[0013] Preferably, the dynamically selecting a search mode based on a user query and generating a security control result based on the search mode includes: Receiving user query dynamics, performing search pattern matching on the user query dynamics, determining whether the question in the user query dynamics is a holistic question through an intent classification model, and if the determination result is no, directly generating a first query result through the multimodal dynamic association fire prevention knowledge graph based on keywords in the question; Determining that the user query dynamics is a holistic question, and utilizing the internal large language model to search and generate multiple batch node reports based on the holistic question from the knowledge graph of the multimodal dynamically associated fire prevention knowledge graph; Dividing the node report into text blocks of predefined sizes, generating an intermediate response for each text block, wherein the intermediate response includes key information fragments about cultural relics fire prevention, and assigning a weight coefficient to each key information fragment based on its relevance to the overall problem; Ranking the key information segments according to their weight coefficients, filtering the key information segments below a preset threshold, forming an aggregated intermediate response, and generating a second query result using the aggregated intermediate response; The first query result and the second query result are determined as security control results.

[0014] Preferably, the method further includes combining the multimodal dynamic association fire prevention knowledge graph with the cultural relics and ancient building information model, dynamically marking risk information in the three-dimensional model, and automatically generating a cultural relics protection priority list and protection plan based on the risk information, including: Using LiDAR point cloud data, we reconstructed a high-precision geometric model of the ancient building, preserving detailed features such as brackets, beams, and painted patterns. We then created a digital twin that included material properties, structural parameters, and historical repair records. The model was then broken down into component-level units and labeled with data. Aligning the spatial coordinates of the ancient building entity and the ancient building model in the multimodal dynamic associated fire prevention knowledge graph, establishing a bidirectional index relationship, and synchronizing the fire risk data and cultural relic status data in the multimodal dynamic associated fire prevention knowledge graph to the component attribute panel of each component-level unit in real time; Associating the inspection reports and maintenance record documents of previous years in the component property panel to generate a time series change curve; Constructing a risk quantification index system based on building risk, environmental disaster risk, and cultural relic vulnerability indicators, and using the multimodal dynamic association fire prevention knowledge graph to determine the risk weight of each indicator in the risk quantification index system and the correlation coefficient between indicators; Performing a risk quantification assessment on each component-level unit based on the risk quantification indicator system to generate a risk label and risk level for each component-level unit; Based on the corresponding components of the risk labels in the three-dimensional model, dynamically annotating risk information, and color rendering the components based on the risk levels; Performing a processing urgency assessment in the multimodal dynamic association fire prevention knowledge graph according to the risk label and risk level of each component-level unit, and calculating a comprehensive score based on the urgency assessment result; Determining the priority ranking of each of the construction-level units according to the comprehensive score of each of the component-level units, and generating a cultural relics protection priority list based on the priority ranking; According to the risk label of each component-level unit, protective measures are retrieved in the multimodal dynamic association fire protection knowledge graph, and a protection plan for the corresponding component is generated based on the retrieval results.

[0015] Preferably, the method further includes updating the status of cultural relics and ancient building nodes in real time through Internet of Things sensor devices, and dynamically generating the current best safety disposal plan in combination with tourist flow data, including: Determine cultural relics and ancient building environment monitoring nodes, structure detection nodes, and visitor monitoring nodes using the multimodal dynamic association fire prevention knowledge graph; Deploy temperature and humidity sensors, smoke detectors, and ultraviolet sensors at the cultural relics and ancient building environmental monitoring nodes to collect environmental data of the cultural relics and ancient buildings; Deploy strain gauges, inclinometers and vibration sensors at the structural detection nodes to collect structural data of cultural relics and ancient buildings; Deploy infrared radiation devices and smart cameras equipped with behavior recognition algorithms at the tourist monitoring nodes to collect tourist flow data and tourist behavior data in the scenic area; Obtain the current number of visitors and today's reservation data through the ticketing system and predict future passenger flow trends based on the visitor flow data; Unifying the environmental data, the structural data, the visitor flow data and the visitor behavior data into a standard timestamp; Input the current environmental data and the structural data into the multimodal dynamic association fire prevention knowledge graph to determine the real-time risk situation of cultural relics and ancient buildings, adjust the risk weight of tourist-dense areas based on the tourist flow data and future passenger flow trends, and generate a current cultural relics and ancient building risk report; Identify risks in tourist behavior data and generate real-time man-made disaster warnings when high-risk behaviors are found in tourist behavior data; Determine the current risk level based on the current cultural relics and ancient building risk report and the man-made disaster warning, and issue a risk warning on the monitoring display port when the risk level exceeds a preset threshold; Based on a multi-constraint decision-making model, the current cultural relics and ancient building risk report and the man-made disaster warning are used to generate the current best safety disposal plan.

[0016] Through the above technical means, the present invention achieves the following beneficial effects: 1. Based on knowledge graph technology, the possible causes of fires in cultural relics and ancient buildings are integrated and correlated to form a knowledge graph in the field of cultural relics fire safety and establish a knowledge base of emergency plans for cultural relics building safety scenarios. This will help identify and resolve fire hazards in cultural relics and ancient buildings, and meet the fire prevention needs of cultural relics buildings based on new technologies. 2. Use IoT sensor devices to update the status of cultural relics and ancient buildings in real time, solve the problem of accurately locating the fire point of ancient cultural relics and cultural buildings, and dynamically generate the best current safety disposal plan based on tourist flow data to control the fire as soon as it occurs; 3. A rapid response to fires is achieved through an emergency plan knowledge base and rapid response technology. Multi-type sensor networking monitoring and operation and maintenance technology ensures real-time monitoring, paving the way for post-fire handling procedures in cultural relic buildings, meeting the principle of minimum intervention in fire prevention and disposal in cultural relic buildings, and minimizing human intervention in cultural relic buildings. 4. In the field of fire safety of cultural relics buildings, a comprehensive and accurate knowledge map is constructed, and the impact level and fire extinguishing plan of various disaster-causing factors such as weather, building materials, building structure, historical repairs, etc. are comprehensively evaluated to meet the fire prevention needs of cultural relics buildings, and knowledge retrieval and generation of fire prevention plans in related fields are completed.

[0017] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0018] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0020] Figure 1 A schematic diagram of a knowledge graph-based security control method provided by the present invention; Figure 2 Another schematic diagram of a knowledge graph-based security control method provided by the present invention; Figure 3 This is another schematic diagram of a security control method based on knowledge graph provided by the present invention; Figure 4 A schematic diagram of the knowledge graph construction process of a knowledge graph-based security control method provided by the present invention; Figure 5 A schematic diagram of the interface of the auxiliary decision-making software for the knowledge graph-based security control method provided by the present invention; Figure 6 Schematic diagram of the visitor behavior recognition software interface for the knowledge graph-based security control method provided by the present invention. DETAILED DESCRIPTION

[0021] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.

[0022] my country has a rich cultural heritage, including vast cultural relics and precious historical sites. Fires in cultural relics buildings are characterized by high risks, great hidden dangers, strong destructiveness and great difficulty in firefighting. The huge cost of monitoring and maintenance has become the biggest pain point in the management of relevant cultural relics protection units. Once a fire occurs, it will inevitably lead to irreparable losses. Traditional fire prevention methods have many limitations. When dealing with complex and changing fire hazards, it is difficult to achieve accurate prevention and control and rapid response. With the further advancement of cultural protection work, the importance of research and development and application demonstration of intelligent active fire protection systems for cultural relics buildings has become increasingly prominent. Knowledge graph is a series of graphics that shows the process of knowledge development and structural relationships. It imitates the composition of human structural knowledge through a network structure of nodes (entities) and edges (relationships). Knowledge graph can not only capture original information, but also capture high-order relationships across multiple documents, and has strong reasoning capabilities. It can be widely used in multiple industries and fields, including the field of fire prevention of cultural relics and ancient buildings. Based on the above situation, the problems that the present invention needs to solve are as follows: 1. Traditional fire prevention methods have limitations. Ancient buildings have multiple wooden components, and fire hazards are complex and varied. Current technology cannot meet the fire prevention needs of cultural relics buildings. 2. It is difficult to accurately locate the fire point of ancient cultural relics, making it difficult to control the fire in the first place; 3. Cultural heritage protection requires compliance with the principle of minimum intervention in fire prevention and disposal of cultural heritage buildings, and it is necessary to minimize human intervention in cultural heritage buildings; 4. Cultural relics and ancient buildings are easily affected by various disaster-causing factors such as weather, building materials, building structure, and historical renovations, making it difficult to comprehensively assess the impact level and fire-fighting plans.

[0023] A security control method based on knowledge graph, such as Figure 1 As shown, the following steps are included: Step S101: using a crawler program to crawl various types of data on ancient building fire prevention, cultural relic fire prevention, fire accidents and material coefficients based on keywords to build a fire prevention database; Step S102: using natural language processing technology to perform semantic analysis on the data in the fire prevention database, and performing data processing on the semantically analyzed fire prevention database through word vector modeling and entity relationship extraction; In some embodiments, this embodiment uses natural language processing technology to perform deep semantic analysis on unstructured text data in a fire protection database. It uses a pre-trained language model combined with a domain-adapted entity recognition algorithm to extract key information such as building parameters, material properties, and fire protection regulations. It constructs a semantic representation space for fire protection terminology based on word vector modeling, uses an attention mechanism to capture contextual features of concepts such as "fire resistance limit" and "fire protection distance", and establishes triple relationships between building components, fire protection facilities, and hazard factors through entity relationship extraction technology. Step S103: using the processed fire prevention database to construct an internal large language model, and using the Leiden technique to perform hierarchical clustering on the internal large language model; In some embodiments, this embodiment uses a structured fire protection database as training corpus, adopts the Transformer architecture to build a pre-trained language model, learns the deep semantic associations of professional knowledge such as building parameters and fire protection regulations through tasks such as masked language modeling, and uses the Leiden algorithm to perform hierarchical clustering on the model's hidden parameters. By pruning redundant units and strengthening dynamic paths between modules, the model structure and reasoning efficiency are optimized, ultimately forming an interpretable and easily scalable domain-specific model to support fire association analysis and safety decision-making. Step S104: establishing a multi-dimensional mapping model based on the spatial characteristics of ancient buildings, cultural relics characteristics, and disaster-causing factors in the hierarchical clustering results, and constructing a multi-modal dynamic association fire prevention knowledge graph based on the multi-dimensional mapping model using knowledge graph technology; In some embodiments, this embodiment constructs a cross-modal feature mapping network based on the clustering results of ancient building spatial characteristics, cultural relic attributes, and disaster-causing factors. A graph embedding algorithm is used to encode heterogeneous data into a unified vector space. An attention mechanism is designed to dynamically integrate the associated weights of structural risk, cultural relic vulnerability, and disaster probability. A graph neural network is used to implement functions such as fire evolution deduction and risk path tracing. Furthermore, the coupling relationship between protective measures and risk status is dynamically updated through a time series graph, forming an explainable and reasonable intelligent fire prevention decision-making center. Step S105: Based on the user query keyword, dynamically select a search mode, and generate a security control result based on the search mode and the query keyword.

[0024] In some embodiments, this embodiment uses a domain-adapted semantic parsing model to perform intent recognition and entity extraction on user query keywords, and matches user questions to global search and specified search modes respectively. The global search is intended to process holistic questions involving the entire cultural relics fire prevention data corpus, such as "Which ancient buildings are most prone to fire?" or "What are common cultural relics fire prevention measures?", and the specified search focuses on processing queries for specific keywords (such as a certain ancient building or a certain fire prevention measure), such as "What are the fire prevention measures of Jin Temple?" or "How to protect wooden ancient buildings from fire?".

[0025] The construction process of the knowledge graph is as follows: Figure 4 shown.

[0026] The working principle of the above technical solution is: first, crawl various types of data through crawlers to build a fire prevention database; secondly, perform semantic analysis and data processing on the data in the fire prevention database; then use the processed fire prevention database to build an internal large-scale language model, perform hierarchical clustering, and establish a multi-dimensional mapping model based on the spatial characteristics of ancient buildings, cultural relics characteristics and disaster-causing factors in the hierarchical clustering results, and use knowledge graph technology to build a multimodal dynamic association fire prevention knowledge graph; finally, based on the user's query keywords, dynamically select the search mode and generate safety control results according to the query keywords.

[0027] The beneficial effects of the above technical solution are: first, a fire prevention database is constructed by crawling various types of data through a crawler program, and various types of data can be obtained from multiple data sources, thereby improving data quality. Furthermore, semantic analysis and data processing are performed on the data in the fire prevention database, which can unify the database data format and facilitate the construction of a large language model. Furthermore, the processed fire prevention database is used again to construct an internal large language model, and hierarchical clustering is performed. A multi-dimensional mapping model is established based on the spatial characteristics of ancient buildings, cultural relics characteristics and disaster-causing factors in the hierarchical clustering results. The knowledge graph technology is used to construct a multimodal dynamically associated fire prevention knowledge graph. A large language model can be constructed based on the database, and the knowledge graph can be generated using the large language model, so that the knowledge graph has practicality and relevance. Furthermore, based on user query keywords, the search mode can be dynamically selected, and safety control results can be generated based on the query keywords, solving the problem of users obtaining effective and feasible safety control solutions based on knowledge graphs.

[0028] In one embodiment, Figure 2 As shown, the fire prevention database is constructed by crawling various types of data on ancient building fire prevention, cultural relic fire prevention, fire accidents and material coefficients according to keywords through a crawler program, including: Step S201: determining multiple data sources for obtaining ancient building fire protection standards, cultural relic fire protection standards, fire accident data, and material fire protection coefficients; In some embodiments, the multiple data sources include: standards and specifications, such as the "Guidelines for Fire Protection Design of Cultural Relic Buildings" issued by the State Administration of Cultural Heritage and the International Council on Monuments and Sites (ICOMOS) conservation guidelines, and other policy documents; accident cases, including historical fire archives of the Ministry of Emergency Management, accident investigation reports of key cultural relics protection units, and technical analysis reports on major international cultural relics fires; academic research, such as fire protection technology papers and experimental data from platforms such as China National Knowledge Infrastructure and Elsevier; and material parameters, integrating GB8624 building material combustion performance classification data and the NIST material thermodynamics database. Step S202: Building a multi-level keyword library based on the particularity of fire prevention of ancient buildings and cultural relics; In some embodiments, this embodiment targets special scenarios such as ancient wooden structures and masonry cultural relics, constructing a multi-dimensional keyword library covering building attributes, disaster causes, and protection technologies: the first-level classification includes building type, material properties (such as wood carbonization rate, masonry thermal conductivity), and fire prevention factors; the second-level classification details disaster-causing factors (such as weather, lightning strikes, incense, electrical aging), environmental parameters, and protection standards; the third-level classification integrates historical repair records and new protection technologies to form a dynamically updated terminology network to support knowledge graph construction and semantic analysis of risk models; Step S203: crawling various types of data containing keywords from the multiple data sources according to the multi-level keyword library using a crawler program; In some embodiments, this embodiment develops an intelligent crawler program based on a predefined multi-level keyword system to achieve cross-platform data collection. Differentiated crawling strategies are configured for heterogeneous data sources such as standard document libraries, academic databases, and accident reporting platforms. The rule engine is used to automatically annotate the credibility level of data sources. Ultimately, multiple data such as text, tables, and drawings are stored in an architecture database to form a dynamically scalable fire prevention corpus. Step S204: removing duplicate captured data of the multiple types by hash value comparison, storing the multiple types of data in a database, and retaining the data source and capture timestamp; Step S205: Expert review is performed on specific parameters of the various types of data in the database, and the reliability level is marked based on the data source, and the database is determined to be a fire prevention database.

[0029] In some embodiments, this embodiment forms a multidisciplinary expert group including fire protection engineering, cultural relics protection, and materials science to conduct multi-dimensional verification of core data such as material combustion parameters, fire protection code clauses, accident cause chains in the database, etc., verify the version validity of national standard data, verify the integrity of spatiotemporal elements based on accident cases, and perform consistency verification based on the authority of the source.

[0030] The beneficial effects of the above technical solution are: it can obtain corresponding ancient building fire protection standards, cultural relics fire protection standards, fire accident information and material fire protection coefficient data from multiple data sources, and automatically remove the repeatedly captured multiple types of data, and conduct expert review to make the data authentic and reliable, and the content more detailed, which is convenient for subsequent unified processing and enhances the reliability and intelligence of the system.

[0031] In one embodiment, Figure 3 As shown, the method uses natural language processing technology to perform semantic analysis on the data in the fire prevention database, and processes the semantically analyzed fire prevention database through word vector modeling and entity relationship extraction, including: Step S301: using AI technology to standardize various types of data in the fire prevention database and convert the data in the database into natural language data; In some embodiments, this embodiment uses a serialized template to generate structured natural descriptions for table-type parameters; extracts spatial topological features from fire partition annotations in drawing documents and converts them into directional semantic descriptions; and uses a domain-tuned model to perform entity-attribute-value parsing and coherent paragraph reconstruction for unstructured text. It also uses a generative adversarial network to eliminate differences in data source representations and uses a rule engine to correct unit formats and professional terminology, ultimately outputting a standardized corpus that conforms to natural language logic and retains original semantic details. Step S302: Segmenting the natural language data into text units, and distributing the text units in a concept vector space from near to far according to the relevance of the text units; In some embodiments, this embodiment uses a pre-trained language model to perform semantic vector encoding on the segmented text units, combines similarity calculation with domain knowledge graph constraints, constructs a multi-dimensional concept space, identifies semantically similar unit clusters through a hierarchical clustering algorithm, projects high-dimensional vectors onto a two-dimensional plane, forms a topology-preserving distribution mapping, dynamically adjusts unit spacing based on the association strength gradient, and simultaneously separates low-correlation content; Step S303: using word vector similarity retrieval to automatically expand the synonym database, adjust the relevance weights of the synonyms, and adjust the distribution of the text units based on the weight adjustment results; In some embodiments, this embodiment uses a pre-trained word vector model to calculate the cosine similarity of domain terms, sets a threshold to screen highly relevant candidate words, builds a dynamic synonym expansion library, and designs a weighting strategy based on contextual dependencies: increases the association weight of strongly context-bound word pairs, reduces the weight of weakly relevant word pairs, integrates the weight matrix into the text unit vector space through a graph embedding algorithm, and uses a force-directed layout model to adjust the repulsion and attraction between units, so that high-weight association clusters are distributed cohesively and low-weight units are marginalized; Step S304: extracting entities from the text unit, classifying the entities, building associations between each entity classification, and generating a structured data system.

[0032] In some embodiments, this embodiment combines domain dictionaries to extract architectural entities from text units, such as the Hall of Supreme Harmony in the Forbidden City, the East Hall of Foguang Temple, material entities, specification entities and other core elements, and divides them into four categories: architectural ontology, fire prevention measures, disaster factors, and environmental parameters through a predefined ontology library. Dependency syntax analysis and semantic matching are performed to identify cross-category entity associations, use a rule engine to build explicit relationships, and combine graph attention networks to mine implicit associations, thereby realizing multi-level data linkage query and dynamic expansion.

[0033] The beneficial effects of the above technical solution are: by using AI technology to standardize the various types of data obtained, the obtained data is further optimized to facilitate the subsequent use of the data, thereby improving data processing efficiency. The use of AI technology for standardized processing of the original data improves the standardization level and processing speed of the map, thereby improving the loading efficiency of the system.

[0034] In one embodiment, the process of constructing an internal large language model using the processed fireproof database and performing hierarchical clustering on the internal large language model using the Leiden technique includes: Acquire the structured data system in the fire prevention database after preprocessing, and determine the model architecture of a large language model based on the text unit characteristics of the structured data system; In some embodiments, the structured data system in this embodiment includes entities, relationships, attributes, etc., and the characteristics of text units involve the type, structure, and relevance of the data. A local attention mechanism is designed mainly for text attribute fields to strengthen the numerical semantic association. Based on the characteristics of structured data, the association between entities, the multimodality of attributes, etc., the domain-specific model architecture of the large language model is determined; The pre-processed structured data is mixed with the original text and input into a large language model to mask the entities in the fire prevention database so that the model can learn to determine domain knowledge associations; In some embodiments, this embodiment converts structured data into pseudo-natural language descriptions, concatenates them with the original text, and inputs them into the model. A dynamic masking strategy is designed to implement a high-proportion random masking of core entities such as building components and material parameters, with the masking ratio set to 30%-50%. This forces the model to infer masked entities through context, while retaining logical markers such as specification numbers to maintain syntactic constraints. Positive and negative sample pairs are constructed through contrastive learning, strengthening the co-occurrence probability learning of strongly associated entities such as "wood structure-flame retardant treatment" and suppressing erroneous associations such as "metal component-flammability," thereby improving the model's logical reasoning ability for the fire protection knowledge system. Abstracting the parameters of the large language model into graph nodes, with the weights between nodes determined by parameter similarity, to generate a high-dimensional parameter relationship structure; In some embodiments, this embodiment abstracts parameters such as neuron weights and attention heads of the language model into graph nodes, constructs a high-dimensional relational network based on functional similarity, constructs a dense graph for the fully connected layer parameters, filters low-correlation noise edges based on attention mechanism parameters, forms a sparse parameter relationship graph, identifies functionally coupled neuron clusters, such as material combustion prediction modules and fire spread simulation modules, and divides them into groups, thereby achieving interpretable representation and targeted optimization of the model's internal knowledge structure. The Leiden algorithm is applied to detect the high-dimensional parameter relationship structure, and highly correlated internal groups are merged to form basic functional modules.

[0035] The beneficial effects of the above technical solution are: by performing hierarchical clustering on the obtained pre-processed fire prevention database, the fire prevention database is further optimized to facilitate subsequent use of the data and improve data processing efficiency. The Leiden algorithm is applied to detect the high-dimensional parameter relationship structure, and highly correlated internal groups are merged to form basic functional modules, so that data are clustered according to similarity and correlation, thereby improving the effectiveness of this method.

[0036] In one embodiment, the application of the Leiden algorithm to detect the high-dimensional parameter relationship structure and merge highly correlated internal groups to form a basic functional module includes: Abstracting the internal parameters of the language model into graph nodes, wherein the graph nodes contain the term vector table weight distribution and entity relationships of the text unit; In some embodiments, this embodiment abstracts the hidden parameters of the language model into heterogeneous graph nodes. Each node is encoded with multidimensional features, and the basic parameter features include a weight distribution histogram and a gradient update trajectory. The semantic association features are integrated with the embedding representation of the corresponding vocabulary in the term vector table. The structural association features inherit the role weights in the entity relationship triples. The multimodal features of adjacent nodes are aggregated through a graph attention network, and a community discovery algorithm is used to identify functional modules with both numerical similarity and semantic consistency. Calculating the correlation of the internal parameters of the language model by cosine similarity, and defining edge weights based on the correlation of the internal parameters of the language model; In some embodiments, the present embodiment uses cosine similarity to calculate the response correlation of parameters in the activation pattern of training data, and defines the edge weights between nodes based on the response correlation; Use the Leiden algorithm to assign each node to an independent community, traverse the nodes, calculate the modularity gain when each independent community is moved to an adjacent independent community, select the maximum gain direction to iteratively adjust the independent community affiliation, and generate a community set of the independent community; In some embodiments, this embodiment uses the Leiden algorithm to perform an initial partitioning of each node in the network, so that each node constitutes an independent community. All nodes are traversed, and for each node, the modularity gain brought about by merging the independent community in which the node is located into other adjacent independent communities is calculated. By comparing the modularity changes caused by merging different adjacent communities, the merging direction with the largest modularity gain is selected, the community affiliation of the node is adjusted, and the community partitioning is gradually optimized, so that the modularity of the network is gradually improved, and finally a set of stable and optimized community sets is obtained, thereby achieving effective partitioning of the original network structure and community discovery.

[0037] If the edge weight between the two community sets exceeds a preset threshold, they are merged into the same functional module, and the same functional module is determined as the basic functional module; The high-frequency text units of the community set in the basic function module are extracted, matched with the standard function description, and the module label is generated.

[0038] In some embodiments, extracting high-frequency text units refers to finding frequently occurring words or phrases from the data of a community collection; the basic function module refers to a core part of the system, and the community collection refers to user-generated content or data sets; matching standard function descriptions and generating module labels mainly find the most suitable standard description by matching high-frequency text units with a predefined ancient building and cultural relic fire protection function description library.

[0039] The beneficial effects of the above technical solution are: abstracting the internal parameters of the language model into graph nodes, calculating the correlation of the internal parameters of the language model through cosine similarity, so that the graph nodes can be sorted from near to far according to the correlation, and further merging the same functional modules based on the edge weights between community sets, determining the same functional modules as the basic functional modules, thereby improving the clustering degree of the data and improving the practicality and effectiveness of the method.

[0040] In one embodiment, the Leiden algorithm is used to assign each node to an independent community, traverse the nodes, calculate the modularity gain when each independent community is moved to an adjacent independent community, select the direction of maximum gain to iteratively adjust the independent community affiliation, and generate the community set of the independent community using the following algorithm:

[0041] in, is the edge weight, is the node degree, is the total edge weight, Used to determine whether nodes belong to the same independent community, the algorithm is used to measure the difference between the internal connection density of the independent community and the random distribution, and determine the specific modularity gain based on the difference.

[0042] The beneficial effects of the above technical solution are: by measuring the difference between the internal connection density of the independent community and the random distribution, the specific modularity gain is determined, and by iteratively adjusting the independent community affiliation by selecting the maximum gain direction, a community set is generated, which further improves the practicality of the method and the relevance of the user's search results to the question.

[0043] In one embodiment, the method of establishing a multi-dimensional mapping model based on the spatial characteristics of ancient buildings, cultural relics characteristics, and disaster-causing factors in the hierarchical clustering results, and constructing a multi-modal dynamic association fire prevention knowledge graph based on the multi-dimensional mapping model using knowledge graph technology includes: Extracting hidden layer representations of the ancient building space features and the cultural relic features from the basic functional modules, and generating standardized feature vectors through graph embedding technology; In some embodiments, this embodiment captures the semantics of latent space through the neighborhood information aggregation mechanism of graph neural networks, and uses graph embedding technology to map high-dimensional discrete features to low-dimensional continuous vector space, thereby achieving standardized representation of cross-modal features and further generating feature vectors, in which multiple features of cultural relics are extracted, including shape, texture, color, etc. Extracting the clustering results of disaster-causing factors from the basic functional module, converting the clustering result output into a probability distribution or a risk score vector, and retaining the association weights between the clustering results and spatial features; In some embodiments, the clustering results of disaster-causing factors extracted based on the basic functional modules need to be implemented through multidimensional modeling to achieve risk quantification and spatial correlation preservation. First, a hierarchical clustering method is used to group the multi-source disaster-causing factors. The clustering quality is evaluated and the core clusters are screened using the silhouette coefficient. The discrete clustering results are mapped to a continuous probability space to generate a probability distribution function that characterizes the likelihood of disaster occurrence. The weights of the multidimensional disaster-causing factors are determined, and the risk score vector is optimized to construct a hybrid weight matrix. Build a multi-branch neural network to process the feature vector and the disaster-causing factors separately, realize cross-modal feature interaction, and dynamically adjust the contribution of different disaster-causing factors to the feature vector through the attention mechanism to generate a multi-dimensional model; In some embodiments, the disaster factor branch extracts dynamic evolution patterns through time series modeling. To achieve cross-modal interaction, a bidirectional cross-attention mechanism is introduced to construct a dynamic correlation matrix between modalities. Multi-head self-attention is used to calculate the contribution weights of different disaster factors to the knowledge graph. The loss function is used to optimize the collaborative representation between modalities. The generated fusion representation includes the mapping relationship between the multi-dimensional risk attributes of the disaster factor and the knowledge graph. Mapping the output of the multidimensional model into graph node attributes, establishing a knowledge graph based on the graph nodes, analyzing the risk evolution path through knowledge graph modeling, and applying association rules to mine potential risk coupling relationships; In some embodiments, dynamic parameters such as architectural space characteristics, cultural relic vulnerability index, and disaster probability output by the multidimensional model are mapped to knowledge graph node attributes, a network including building entities, protective facilities, and environmental elements is established, and a knowledge graph with spatiotemporal fusion is constructed. A temporal graph convolutional network is used to model the risk evolution path, trace the causal chain, mine cross-modal association rules, identify implicit risk coupling patterns, quantify the association strength through confidence and improvement, match sensing technology to locate high-risk topological structures, and generate a risk management plan including "hazard identification-path deduction-measure recommendation" and other contents; Based on the risk evolution path and the potential risk coupling relationship, a multimodal dynamically associated fire prevention knowledge graph is constructed according to the knowledge graph.

[0044] In some embodiments, a multimodal dynamic association fire prevention knowledge graph is constructed based on the risk evolution path and potential risk coupling relationship, which mainly integrates multi-source heterogeneous data and dynamic reasoning mechanism, integrates multimodal data features, and performs cross-modal semantic alignment of real-time sensor monitoring data and text reports. The association mapping of sensor data and text description is realized through a two-way attention mechanism, and the interpretable reasoning of the knowledge graph is used to reveal the risk transmission chain under the coupling conditions of multiple disasters in cultural relics and ancient buildings.

[0045] The beneficial effects of the above technical solution are: by establishing a multidimensional mapping model based on the spatial characteristics of ancient buildings, cultural relics characteristics and disaster-causing factors in the hierarchical clustering results, the intelligence of the method is improved; by constructing a multimodal dynamic correlation fire prevention knowledge graph according to the knowledge graph based on the risk evolution path and potential risk coupling relationship, the practicality and professionalism of the method are enhanced, and the user experience is improved.

[0046] In one embodiment, dynamically selecting a search mode based on a user query and generating a security control result based on the search mode includes: Receiving user query dynamics, performing search pattern matching on the user query dynamics, determining whether the question in the user query dynamics is a holistic question through an intent classification model, and if the determination result is no, directly generating a first query result through the multimodal dynamic association fire prevention knowledge graph based on keywords in the question; In some embodiments, the system first uses a pattern recognition engine to segment and extract entities from the query text entered by the user. It then uses a pre-trained language model to calculate the semantic similarity between word vectors and domain ontology, identifying core keywords such as "direction of fire spread" or "key points for fire prevention at Jin Temple." A hierarchical intent classification model is then used to analyze the contextual logical structure of the query. A decision boundary threshold is used to determine whether the problem involves holistic analysis requirements. If it is determined to be a local problem, the knowledge graph direct retrieval channel is activated, mapping the keywords to entity nodes in the fire prevention knowledge graph. A graph traversal algorithm is then used to perform multi-hop reasoning along predefined relationship paths such as "building material-combustion threshold" and "meteorological conditions-fire risk level." Determining that the user query dynamics is a holistic question, and utilizing the internal large language model to search and generate multiple batch node reports based on the holistic question from the knowledge graph of the multimodal dynamically associated fire prevention knowledge graph; In some embodiments, when it is determined that the user query involves a holistic question such as the evolution of regional fire conditions or comprehensive fire warning analysis of cultural relics and ancient buildings, the deep knowledge reasoning engine needs to be activated. First, the question semantics are deconstructed through a pre-trained language model, and the global analysis elements are extracted by combining the contextual attention mechanism. = [It realizes adaptive retrieval of large-scale graph data and screens out core nodes related to the global question through vector similarity calculation; Dividing the node report into text blocks of predefined sizes, generating an intermediate response for each text block, wherein the intermediate response includes key information fragments about cultural relics fire prevention, and assigning a weight coefficient to each key information fragment based on its relevance to the overall problem; In some embodiments, the report is divided into equal-length text blocks according to preset semantic paragraph units, and keyword boundaries are identified (such as core concepts such as the fire protection distance of historical buildings). This ensures that each information block contains a complete fire protection logic unit. A semantic parsing engine extracts key information fragments from each text block, such as core parameters such as the critical combustion temperature of wooden buildings and the flame retardant threshold of mural restoration materials. These fragments are then standardized and annotated. The semantic similarity between the information fragments and the user's overall question is calculated based on a pre-trained language model, and the strength of their structural association with the source question is measured using the knowledge graph relationship path length. The weighted information fragments are then screened and reorganized according to a confidence threshold. Ranking the key information segments according to their weight coefficients, filtering the key information segments below a preset threshold, forming an aggregated intermediate response, and generating a second query result using the aggregated intermediate response; In some embodiments, this embodiment uses the matching degree of semantic similarity calculation and user intent as the main ranking factor, and introduces the node centrality index in the knowledge graph as an auxiliary weight coefficient to semantically align heterogeneous data such as text, spatial coordinates, and fire protection code clauses, and convert structured intermediate responses into query results that comply with industry standards; The first query result and the second query result are determined as security control results.

[0047] The beneficial effects of the above technical solution are: by matching search patterns based on user query dynamics, the search pattern corresponding to the user's needs can be accurately matched, thereby improving the intelligence of the method and the comprehensiveness of the retrieval results. Furthermore, when the user's needs and user query dynamics are holistic issues, a large language model can be used to perform a global search in the knowledge graph, thereby improving the practicality of this method and enhancing the user's experience.

[0048] In one embodiment, the method further includes combining the multimodal dynamically associated fire prevention knowledge graph with a cultural relic and ancient building information model, dynamically annotating risk information in the three-dimensional model, and automatically generating a cultural relic protection priority list and protection plan based on the risk information, including: Using LiDAR point cloud data, we reconstructed a high-precision geometric model of the ancient building, preserving detailed features such as brackets, beams, and painted patterns. We then created a digital twin that included material properties, structural parameters, and historical repair records. The model was then broken down into component-level units and labeled with data. In some embodiments, this embodiment uses 3D laser scanning technology to acquire millimeter-level precision point cloud data, fully preserving detailed features such as bracket and mortise and tenon joint structures, beam and frame folding curves, and painted patterns. The ancient building is decomposed into component-level units such as the base, column base, and bracket. Each unit is bound with a digital label containing material parameters such as wood texture and paint layer thickness, and structural mechanical properties such as load-bearing coefficient and internal stress. The resulting digital twin supports component-level reverse disassembly analysis and can simulate structural deformation under different environmental loads. Aligning the spatial coordinates of the ancient building entity and the ancient building model in the multimodal dynamic associated fire prevention knowledge graph, establishing a bidirectional index relationship, and synchronizing the fire risk data and cultural relic status data in the multimodal dynamic associated fire prevention knowledge graph to the component attribute panel of each component-level unit in real time; In some embodiments, this embodiment uses a spatial coordinate system conversion algorithm to match the latitude and longitude positioning data of ancient architectural entities in the knowledge graph with the spatial grid of the model, and uses feature point registration technology to eliminate spatial drift errors between multi-source data. On the knowledge graph side, a three-dimensional model component ID hash value is configured for each bracket and beam node; on the model side, a query interface is embedded in the component attribute panel to support reverse retrieval of related entities in the knowledge graph based on the component ID; Associating the inspection reports and maintenance record documents of previous years in the component property panel to generate a time series change curve; In some embodiments, the time series change curve includes key physical quantities (such as crack width and load-bearing stress) and major maintenance event nodes, and reveals the effectiveness of intervention measures through spatial mapping between event markers and curve fluctuation intervals, supporting multi-scale time window adjustment; Constructing a risk quantification index system based on building risk, environmental disaster risk, and cultural relic vulnerability indicators, and using the multimodal dynamic association fire prevention knowledge graph to determine the risk weight of each indicator in the risk quantification index system and the correlation coefficient between indicators; In some embodiments, this embodiment is based on a weight calculation engine based on a multimodal dynamically associated fire prevention knowledge graph, aggregating node attributes and relationship path features, and determining the strength of the correlation between building ventilation structure and fire spread speed through the cosine similarity between node embedding vectors. It also extracts potential conduction paths between ambient temperature and humidity nodes and wood moisture content nodes, quantifies the nonlinear coupling coefficients of cross-modal indicators, and introduces a dynamic attention mechanism into the weight allocation process. It adjusts the indicator sensitivity weights based on the real-time fire risk warning level, such as automatically increasing the lightning protection system failure risk factor during thunderstorms. Performing a risk quantification assessment on each component-level unit based on the risk quantification indicator system to generate a risk label and risk level for each component-level unit; In some embodiments, this embodiment uses ontology mapping technology to align the risk quantification indicator system with the component attribute data in the three-dimensional model, and adopts a dynamic weighted summation algorithm to calculate the comprehensive risk value: the indicator weights derived from the knowledge graph are combined with real-time monitoring data such as component material aging, ambient temperature and humidity, and the risk level is divided according to the preset differentiated grading standards of the component type to generate a corresponding risk label system. When the sensor detects abnormal fading of the painting or a new fire prevention facility is added to the knowledge graph, the system automatically updates the risk matrix parameters to ensure the spatiotemporal synchronization of the assessment results and the entity status; Based on the corresponding components of the risk labels in the three-dimensional model, dynamically annotating risk information, and color rendering the components based on the risk levels; In some embodiments, the color rendering rule for the components based on the risk level is as follows: red corresponds to high risk, such as wooden components + surrounding flammable materials; yellow corresponds to medium risk: masonry structures + aging circuits; and green corresponds to protected areas; Performing a processing urgency assessment in the multimodal dynamic association fire prevention knowledge graph according to the risk label and risk level of each component-level unit, and calculating a comprehensive score based on the urgency assessment result; In some embodiments, this embodiment performs multi-hop traversal along the spatiotemporal association paths in the knowledge graph to extract key parameters such as the risk transmission intensity, historical disaster loss frequency, and secondary disaster triggering probability of each component node. A time decay function is used to process the timeliness characteristics of risk labels. The weight coefficient of components with continuously increasing risk levels in the past three months is automatically increased to assess the urgency of processing and provide a comprehensive score. Determining the priority ranking of each of the construction-level units according to the comprehensive score of each of the component-level units, and generating a cultural relics protection priority list based on the priority ranking; According to the risk label of each component-level unit, protective measures are retrieved in the multimodal dynamic association fire protection knowledge graph, and a protection plan for the corresponding component is generated based on the retrieval results.

[0049] In some embodiments, risk labels are parsed into semantic feature vectors in the knowledge graph, and graph traversal retrieval is performed along the three-layer relationship path of "risk type-disaster path-protective measures". For conventional risk labels, the preset fire retardant coating construction specifications are matched through the component body; for complex risk labels, the graph neural network is enabled to calculate the cross-modal correlation degree and screen combined measures with synergistic protection effectiveness; the retrieval results are processed by the dynamic priority sorting module, and the protection plan is generated by comprehensively considering the timeliness of the implementation of the measures, the protection value weight and the resource constraints.

[0050] The beneficial effects of the above technical solution are: by combining the multimodal dynamically associated fire prevention knowledge graph with the cultural relics and ancient building information model, a three-dimensional model of the ancient building can be generated in the virtual space, and the three-dimensional model can be annotated, which further improves the flexibility of the method and reduces the intervention in the ancient building entity. Risk information is dynamically annotated in the three-dimensional model, and a cultural relics protection priority list and protection plan are automatically generated based on the risk information, which improves the intelligence level of the method and is conducive to the development of cultural relics and ancient building protection work.

[0051] In one embodiment, the method further comprises: Figure 5 As shown, IoT sensor devices are used to update the status of cultural relics and ancient building nodes in real time, and combined with tourist flow data, the current optimal safety disposal plan is dynamically generated, including: Determine cultural relics and ancient building environment monitoring nodes, structure detection nodes, and visitor monitoring nodes using the multimodal dynamic association fire prevention knowledge graph; In some embodiments, this embodiment matches the latitude and longitude positioning data of the ancient building entity in the knowledge graph with the spatial grid of the model through a spatial coordinate system conversion algorithm, and determines the specific location and type of the sensor placed on the ancient building entity based on the search function of the knowledge graph; Deploy temperature and humidity sensors, smoke detectors, and ultraviolet sensors at the cultural relics and ancient building environmental monitoring nodes to collect environmental data of the cultural relics and ancient buildings; In some embodiments, the environmental data includes temperature and humidity data, fire data, and ultraviolet intensity data, which are used to determine whether the building is in an optimal environmental range and whether a fire has occurred; Deploy strain gauges, inclinometers and vibration sensors at the structural detection nodes to collect structural data of cultural relics and ancient buildings; In some embodiments, strain gauges are used to measure the internal stress of the ancient building, inclinometers are used to determine the inclination of the component, and vibration sensors are used to determine the vibration of the component; Deploy infrared radiation devices and smart cameras equipped with behavior recognition algorithms at the tourist monitoring nodes to collect tourist flow data and tourist behavior data in the scenic area; In some embodiments, infrared radiation devices are used to determine whether there are tourists passing through key locations to prevent tourists from entering prohibited areas. Smart cameras equipped with behavior recognition algorithms are used to determine the number of tourists and can identify dangerous behaviors of tourists. Obtain the current number of visitors and today's reservation data through the ticketing system and predict future passenger flow trends based on the visitor flow data; In some embodiments, a multi-dimensional passenger flow forecasting and analysis framework can be constructed by combining the current number of visitors and today's reservation data obtained in real time from the ticketing system with the time series characteristics of historical tourist traffic. First, the framework connects to the ticketing database to extract structured data streams, including core indicators such as real-time ticket gate pass records, the dynamics of remaining tickets in the online reservation system, and the frequency of refund and ticket change operations. Furthermore, environmental variables such as weather warning information, holiday labels, and traffic control notices from external data sources are simultaneously integrated to generate a passenger flow forecast curve. Unifying the environmental data, the structural data, the visitor flow data and the visitor behavior data into a standard timestamp; In some embodiments, this embodiment is used to time-align all data; Input the current environmental data and the structural data into the multimodal dynamic association fire prevention knowledge graph to determine the real-time risk situation of cultural relics and ancient buildings, adjust the risk weight of tourist-dense areas based on the tourist flow data and future passenger flow trends, and generate a current cultural relics and ancient building risk report; In some embodiments, this embodiment integrates component status data from 3D laser scanning with real-time monitoring results from infrared thermal imaging; uses a knowledge graph inference engine to mine cross-modal risk transmission relationships and generate a risk heat map; the decision-making level combines immediate disposal measures and long-term planning guidance from the emergency plan knowledge base to generate a risk trend report through comparative analysis; Identify risks in tourist behavior data and generate real-time man-made disaster warnings when high-risk behaviors are found in tourist behavior data; In some embodiments, as Figure 6As shown, this embodiment uses the IoT perception layer to collect real-time behavioral features such as tourist trajectory density, abnormal gathering areas captured by thermal imagers, and climbing and crossing-boundary actions identified by smart cameras. It conducts multi-dimensional cross-verification on behaviors such as abnormal tourist stay time, deviation from the preset route, and intrusion into dangerous areas, and issues early warning information in real time. Determine the current risk level based on the current cultural relics and ancient building risk report and the man-made disaster warning, and issue a risk warning on the monitoring display port when the risk level exceeds a preset threshold; In some embodiments, this embodiment uses a knowledge graph inference engine to align structured risk reports with man-made disaster warnings, such as unusual gatherings of tourists in hot spots and illegal carrying of fire sources, in time and space. A dynamic weighting algorithm is used to construct a risk rating algorithm. An adaptive threshold mechanism is introduced into the risk rating calculation to analyze the deviation between historical disaster loss data and monitored values ​​in real time. When the comprehensive risk value exceeds the dynamic safety baseline, a risk warning is triggered. Based on a multi-constraint decision-making model, the current cultural relics and ancient building risk report and the man-made disaster warning are used to generate the current best safety disposal plan.

[0052] In some embodiments, this embodiment aligns the environmental monitoring parameters and structural health indicators in the real-time risk report with the man-made disaster warnings in time and space, constructs a multidimensional decision space that includes constraints such as risk level, emergency resource distribution, and response time requirements, mines the optimal disposal path features in the historical emergency cases contained in the knowledge graph, and outputs the current best safety disposal plan in real time.

[0053] The beneficial effects of the above technical solution are: by updating the status of cultural relics and ancient buildings nodes in real time through Internet of Things sensing equipment, cultural relics and ancient buildings can be monitored in real time based on environmental data, structural data, tourist flow data and tourist behavior data, and the real-time risk situation of cultural relics and ancient buildings can be determined. Furthermore, the current best safety disposal plan can be dynamically generated in combination with tourist flow data, reducing the possibility of man-made disasters. The current cultural relics and ancient buildings risk report and man-made disaster warning can be used to generate the current best safety disposal plan, thereby improving the safety and intelligence of the method.

[0054] Those skilled in the art should understand that the first and second in the present invention simply refer to different application stages.

[0055] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the disclosure herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow from the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the following claims.

[0056] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A security control method based on knowledge graph, characterized in that: The following steps are involved: A fire prevention database is constructed by crawling various types of data on ancient building fire prevention, cultural relic fire prevention, fire accidents and material coefficients based on keywords through a crawler program; Using natural language processing technology to perform semantic analysis on the data in the fire prevention database, and performing data processing on the semantically analyzed fire prevention database through word vector modeling and entity relationship extraction; Use the processed fire prevention database to build an internal large-scale language model, and use the Leiden technique to perform hierarchical clustering on the internal large-scale language model; A multi-dimensional mapping model is established based on the spatial characteristics of ancient buildings, cultural relics characteristics and disaster-causing factors in the hierarchical clustering results, and a multi-modal dynamic association fire prevention knowledge graph is constructed based on the multi-dimensional mapping model using knowledge graph technology; Based on the user's query keyword, a search mode is dynamically selected, and based on the search mode, a security control result is generated according to the query keyword.

2. The security control method based on knowledge graph according to claim 1 is characterized in that: The fire prevention database is constructed by crawling various types of data on ancient building fire prevention, cultural relic fire prevention, fire accidents and material coefficients according to keywords through a crawler program, including: Identify multiple data sources for obtaining fire protection standards for ancient buildings, fire protection standards for cultural relics, fire accident data, and fire protection coefficients of materials; Based on the particularity of fire prevention of ancient buildings and cultural relics, a multi-level keyword library is constructed; crawling various types of data containing keywords from the multiple data sources according to the multi-level keyword library using a crawler program; Removing duplicated captured data of various types through hash value comparison, storing the data of various types in a database, and retaining the data source and capture timestamp; Experts review the specific parameters of the various types of data in the database, label the reliability level based on the data source, and determine that the database is a fire prevention database.

3. The security control method based on knowledge graph according to claim 1 is characterized in that: The method of using natural language processing technology to perform semantic analysis on the data in the fire prevention database, and processing the semantically analyzed fire prevention database through word vector modeling and entity relationship extraction, includes: Using AI technology to standardize various types of data in the fire prevention database and convert the data in the database into natural language data; Segmenting the natural language data into text units, and distributing the text units in a concept vector space from near to far according to the relevance of the text units; Using word vector similarity retrieval, automatically expanding the synonym library, adjusting the relevance weights of synonyms, and adjusting the distribution of the text units based on the weight adjustment results; Entities are extracted from the text units, the entities are classified, association relationships are constructed between each entity classification, and a structured data system is generated.

4. The security control method based on knowledge graph according to claim 1 is characterized in that: The method uses the processed fire prevention database to build an internal large language model, and uses the Leiden technology to perform hierarchical clustering on the internal large language model, including: Acquire the structured data system in the fire prevention database after preprocessing, and determine the model architecture of a large language model based on the text unit characteristics of the structured data system; The pre-processed structured data is mixed with the original text and input into a large language model to mask the entities in the fire prevention database so that the model can learn to determine domain knowledge associations; Abstracting the parameters of the large language model into graph nodes, with the weights between nodes determined by parameter similarity, to generate a high-dimensional parameter relationship structure; The Leiden algorithm is applied to detect the high-dimensional parameter relationship structure, and highly correlated internal groups are merged to form basic functional modules.

5. The security control method based on knowledge graph according to claim 4 is characterized in that: The Leiden algorithm is applied to detect the high-dimensional parameter relationship structure, merge highly correlated internal groups, and form a basic functional module, including: Abstracting the internal parameters of the language model into graph nodes, wherein the graph nodes contain the term vector table weight distribution and entity relationships of the text unit; Calculating the correlation of the internal parameters of the language model by cosine similarity, and defining edge weights based on the correlation of the internal parameters of the language model; Use the Leiden algorithm to assign each node to an independent community, traverse the nodes, calculate the modularity gain when each independent community is moved to an adjacent independent community, select the maximum gain direction to iteratively adjust the independent community affiliation, and generate a community set of the independent community; If the edge weight between the two community sets exceeds a preset threshold, they are merged into the same functional module, and the same functional module is determined as the basic functional module; The high-frequency text units of the community set in the basic function module are extracted, matched with the standard function description, and the module label is generated.

6. The security control method based on knowledge graph according to claim 5 is characterized in that: The Leiden algorithm is used to assign each node to an independent community, traverse the nodes, calculate the modularity gain when each independent community is moved to an adjacent independent community, select the maximum gain direction to iteratively adjust the independent community affiliation, and generate the community set of the independent community using the following algorithm: in, is the edge weight, is the node degree, is the total edge weight, Used to determine whether nodes belong to the same independent community, the algorithm is used to measure the difference between the internal connection density of the independent community and the random distribution, and determine the specific modularity gain based on the difference.

7. The security control method based on knowledge graph according to claim 1 is characterized in that: The method includes establishing a multi-dimensional mapping model based on the spatial characteristics of ancient buildings, cultural relics characteristics and disaster-causing factors in the hierarchical clustering results, and constructing a multi-modal dynamic association fire prevention knowledge graph based on the multi-dimensional mapping model using knowledge graph technology, including: Extracting hidden layer representations of the ancient building space features and the cultural relic features from the basic functional modules, and generating standardized feature vectors through graph embedding technology; Extracting the clustering results of disaster-causing factors from the basic functional module, converting the clustering result output into a probability distribution or a risk score vector, and retaining the association weights between the clustering results and spatial features; Build a multi-branch neural network to process the feature vector and the disaster-causing factors separately, realize cross-modal feature interaction, and dynamically adjust the contribution of different disaster-causing factors to the feature vector through the attention mechanism to generate a multi-dimensional model; Mapping the output of the multidimensional model into graph node attributes, establishing a knowledge graph based on the graph nodes, analyzing the risk evolution path through knowledge graph modeling, and applying association rules to mine potential risk coupling relationships; Based on the risk evolution path and the potential risk coupling relationship, a multimodal dynamically associated fire prevention knowledge graph is constructed according to the knowledge graph.

8. The security control method based on knowledge graph according to claim 1, characterized in that: The dynamically selecting a search mode based on a user query and generating a security control result based on the search mode includes: Receiving user query dynamics, performing search pattern matching on the user query dynamics, determining whether the question in the user query dynamics is a holistic question through an intent classification model, and if the determination result is no, directly generating a first query result through the multimodal dynamic association fire prevention knowledge graph based on keywords in the question; Determining that the user query dynamics is a holistic question, and utilizing the internal large language model to search and generate multiple batch node reports based on the holistic question from the knowledge graph of the multimodal dynamically associated fire prevention knowledge graph; Dividing the node report into text blocks of predefined sizes, generating an intermediate response for each text block, wherein the intermediate response includes key information fragments about cultural relics fire prevention, and assigning a weight coefficient to each key information fragment based on its relevance to the overall problem; Ranking the key information segments according to their weight coefficients, filtering the key information segments below a preset threshold, forming an aggregated intermediate response, and generating a second query result using the aggregated intermediate response; The first query result and the second query result are determined as security control results.

9. The security control method based on knowledge graph according to claim 1 is characterized in that: The method further includes combining the multimodal dynamically associated fire prevention knowledge graph with a cultural relic and ancient building information model, dynamically annotating risk information in the three-dimensional model, and automatically generating a cultural relic protection priority list and protection plan based on the risk information, including: Using LiDAR point cloud data, we reconstructed a high-precision geometric model of the ancient building, preserving detailed features such as brackets, beams, and painted patterns. We then created a digital twin that included material properties, structural parameters, and historical repair records. The model was then broken down into component-level units and labeled with data. Aligning the spatial coordinates of the ancient building entity and the ancient building model in the multimodal dynamic associated fire prevention knowledge graph, establishing a bidirectional index relationship, and synchronizing the fire risk data and cultural relic status data in the multimodal dynamic associated fire prevention knowledge graph to the component attribute panel of each component-level unit in real time; Associating the inspection reports and maintenance record documents of previous years in the component property panel to generate a time series change curve; Constructing a risk quantification index system based on building risk, environmental disaster risk, and cultural relic vulnerability indicators, and using the multimodal dynamic association fire prevention knowledge graph to determine the risk weight of each indicator in the risk quantification index system and the correlation coefficient between indicators; Performing a risk quantification assessment on each component-level unit based on the risk quantification indicator system to generate a risk label and risk level for each component-level unit; Based on the corresponding components of the risk labels in the three-dimensional model, dynamically annotating risk information, and color rendering the components based on the risk levels; Performing a processing urgency assessment in the multimodal dynamic association fire prevention knowledge graph according to the risk label and risk level of each component-level unit, and calculating a comprehensive score based on the urgency assessment result; Determining the priority ranking of each of the construction-level units according to the comprehensive score of each of the component-level units, and generating a cultural relics protection priority list based on the priority ranking; According to the risk label of each component-level unit, protective measures are retrieved in the multimodal dynamic association fire protection knowledge graph, and a protection plan for the corresponding component is generated based on the retrieval results.

10. The security control method based on knowledge graph according to claim 1, characterized in that: The method also includes updating the status of cultural relics and ancient building nodes in real time through IoT sensor devices, and dynamically generating the current optimal safety disposal plan based on tourist flow data, including: Determine cultural relics and ancient building environment monitoring nodes, structure detection nodes, and visitor monitoring nodes using the multimodal dynamic association fire prevention knowledge graph; Deploy temperature and humidity sensors, smoke detectors, and ultraviolet sensors at the cultural relics and ancient building environmental monitoring nodes to collect environmental data of the cultural relics and ancient buildings; Deploy strain gauges, inclinometers and vibration sensors at the structural detection nodes to collect structural data of cultural relics and ancient buildings; Deploy infrared radiation devices and smart cameras equipped with behavior recognition algorithms at the tourist monitoring nodes to collect tourist flow data and tourist behavior data in the scenic area; Obtain the current number of visitors and today's reservation data through the ticketing system and predict future passenger flow trends based on the visitor flow data; Unifying the environmental data, the structural data, the visitor flow data and the visitor behavior data into a standard timestamp; Input the current environmental data and the structural data into the multimodal dynamic association fire prevention knowledge graph to determine the real-time risk situation of cultural relics and ancient buildings, adjust the risk weight of tourist-dense areas based on the tourist flow data and future passenger flow trends, and generate a current cultural relics and ancient building risk report; Identify risks in tourist behavior data and generate real-time man-made disaster warnings when high-risk behaviors are found in tourist behavior data; Determine the current risk level based on the current cultural relics and ancient building risk report and the man-made disaster warning, and issue a risk warning on the monitoring display port when the risk level exceeds a preset threshold; Based on a multi-constraint decision-making model, the current cultural relics and ancient building risk report and the man-made disaster warning are used to generate the current best safety disposal plan.

Images