Community fire early warning method combined with knowledge graph

Abstract

The invention discloses a community fire early warning method combined with a knowledge graph, and relates to the technical field of fire early warning, and the method comprises the steps: carrying out the traversing of community region information, carrying out the data collection, and constructing a community fire knowledge graph according to a data collection result; performing feature extraction based on the community fire knowledge graph to obtain a plurality of fire features; constructing a fire risk assessment model to perform fire risk assessment on the community, and generating a plurality of risk assessment results; monitoring the community in real time, and generating a trigger instruction; and starting a fire early warning mechanism according to the trigger instruction, and generating a fire response decision. The technical problems that in the prior art, a traditional fire early warning method is inaccurate in evaluation and cannot effectively prevent fire in advance are solved, and the technical effects of improving the accuracy of fire risk evaluation, improving the accuracy of fire prevention and effectively reducing fire hazards are achieved by combining the knowledge graph.

Description

Technical Field

[0001] The present invention relates to the field of fire warning technology, and in particular to a community fire warning method combined with a knowledge graph. Background Art

[0002] With the acceleration of urbanization, the scale of communities continues to expand, the population density gradually increases, and the fire risk in communities becomes increasingly prominent. Traditional fire prevention and response measures often rely on limited experience and manual inspections, which makes it difficult to comprehensively and accurately assess fire risks and respond effectively in a timely manner. In the information age, a large amount of data is generated and accumulated, but these data are scattered in different systems and departments, lacking effective integration and utilization. At the same time, the occurrence and development of fires are affected by a variety of complex factors, such as the building structure of the community, the configuration of fire-fighting facilities, the activities of personnel, and meteorological conditions.

[0003] The existing technology has the technical problem that the traditional fire warning method is inaccurate in assessment and cannot effectively prevent fires in advance. Summary of the invention

[0004] The present application provides a community fire warning method combined with a knowledge graph, which is used to solve the technical problem that traditional fire warning methods in the prior art are inaccurate in evaluation and cannot effectively prevent fires in advance.

[0005] In view of the above problems, this application provides a community fire warning method combined with a knowledge graph, the method comprising:

[0006] Traverse the community area information to collect data, and build a community fire knowledge graph based on the data collection results; perform feature extraction based on the community fire knowledge graph to obtain multiple fire features; build a fire risk assessment model based on the multiple fire features to conduct fire risk assessment on the community and generate multiple risk assessment results; monitor the community in real time, and generate a trigger instruction based on the real-time monitoring information combined with the multiple risk assessment results; activate the fire warning mechanism according to the trigger instruction, generate a fire emergency plan, combine the community fire knowledge graph for auxiliary response, and generate a fire response decision.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] Traverse the community area information to collect data, and build a community fire knowledge graph based on the data collection results; extract features based on the community fire knowledge graph to obtain multiple fire features; build a fire risk assessment model based on multiple fire features to conduct fire risk assessment on the community and generate multiple risk assessment results; monitor the community in real time, generate trigger instructions, start the fire warning mechanism, generate fire emergency plans, and generate fire response decisions. By combining the knowledge graph, the accuracy of fire risk assessment, the accuracy of fire prevention, and the technical effect of effectively reducing fire hazards can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A flow chart of a community fire warning method combined with a knowledge graph provided in an embodiment of the present application;

[0011] Figure 2 A schematic diagram of the process of constructing a community fire knowledge graph for the community fire warning method combined with a knowledge graph provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] This application provides a community fire warning method combined with a knowledge graph to solve the technical problem that traditional fire warning methods in the prior art are inaccurate in evaluation and cannot effectively prevent fires in advance.

[0013] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0014] Embodiment 1, as Figure 1 As shown, the present application provides a community fire warning method combined with a knowledge graph, the method comprising:

[0015] Step S100: Traverse the community area information to collect data, and build a community fire knowledge graph based on the data collection results.

[0016] Specifically, first, collect the community's geographic and architectural information, such as the community's layout, the type of building (residential, commercial, public facilities, etc.), the height and structure of the building, the age of the building, etc. Secondly, obtain relevant data on fire-fighting facilities, including the location and number of fire hydrants, the distribution of fire extinguishers, the location and smoothness of evacuation routes and emergency exits, etc. Then analyze the population distribution and mobility in the community, such as the number of residents in different areas, population density, and the distribution of special groups (such as the elderly, children, and the disabled). Then, collect information on fire sources and flammable materials in the community, such as the use of electrical equipment, the layout of gas pipelines, and the storage location of flammable and explosive items. Finally, collect historical fire data, including the time, location, cause, and loss of previous fires. After obtaining rich community area information data, start the work of building a community fire knowledge map, classify and organize the collected data, classify geographic and architectural information as infrastructure data, classify fire-fighting facility data as safety facility category, population distribution and mobility as personnel-related data, fire source and flammable material information as risk factor category, and historical fire data as experience reference category. Then, using the construction technology of knowledge graph, create nodes to represent different entities, such as each building, each fire hydrant, and each fire source in the community as an independent node. At the same time, establish the relationship between nodes, such as a building and a nearby fire hydrant have a "proximity" relationship, and a fire source has an "association" relationship with the surrounding flammable materials. Assign attributes to each node and relationship. Building nodes have attributes such as "building structure", "building area", and "purpose of use"; fire hydrant nodes may have attributes such as "location coordinates", "maintenance status", and "water supply capacity"; relationship attributes may include "distance" and "degree of influence". Through the above steps, a comprehensive and detailed community fire knowledge graph is gradually built. This graph can clearly show the complex relationships and attribute characteristics between various elements in the community, provide intuitive and accurate information support for subsequent analysis and decision-making, and can comprehensively and meticulously grasp various fire-related information and relationships in the community, so as to achieve accurate assessment of fire risks.

[0017] Step S200: Perform feature extraction based on the community fire knowledge graph to obtain multiple fire features.

[0018] Specifically, based on the previously constructed community fire knowledge graph, feature extraction is performed. First, features such as the age of the building, structure type (such as brick-concrete, frame, etc.), height, and building area are extracted from the building-related nodes and relationships in the knowledge graph. These features can reflect the fire vulnerability and potential risks of the building. Secondly, the number, type (such as dry powder fire extinguishers, foam fire extinguishers, etc.), distribution density, and maintenance status of fire-fighting facilities are obtained from the nodes and connections of fire-fighting facilities. They have an important indicative role in the ability and effect of fire fighting when a fire occurs. Furthermore, features such as population density and the proportion of special populations (such as the elderly and children) are extracted from the information on population distribution, because population characteristics will affect the difficulty of evacuation and rescue needs when a fire occurs. Then, features such as the type of fire source (electrical fire source, open flame, etc.), the type and quantity of flammable materials are extracted from the fire source and flammable materials, which can directly reflect the possibility of fire and the potential degree of harm. Finally, the seasonal patterns, time period patterns, cause categories and other characteristics of fire occurrence are extracted from historical fire data to provide a reference for predicting future fire risks. Through the feature extraction of the above multiple methods, we can finally obtain multiple fire characteristics that can comprehensively describe the community fire risk status.

[0019] Step S300: constructing a fire risk assessment model based on the multiple fire characteristics to conduct fire risk assessment on the community and generate multiple risk assessment results.

[0020] Specifically, the fire risk assessment model is constructed using the multiple fire features extracted previously. The extracted multiple fire feature data are preprocessed, including processing missing values, outliers, and standardizing the data so that each feature value is in a similar numerical range. The preprocessed fire feature data is input into the random forest model, which is an integrated learning algorithm composed of multiple decision trees. During the training phase, the random forest randomly extracts part of the data and features to construct each decision tree. For each fire feature, such as the age of the building, the distribution density of fire-fighting facilities, etc., the decision tree will split at the node according to the set splitting conditions (for example, based on information gain or Gini index), and gradually grow until the predetermined stop condition (such as the maximum depth of the tree, the minimum number of leaf node samples, etc.) is reached. By constructing multiple such decision trees, a random forest is formed. In the prediction stage, after the new community data is input into the model, multiple decision trees will give prediction results respectively. The final fire risk assessment result is determined by combining the prediction results of multiple decision trees (such as taking the average or majority vote). The multiple risk assessment results generated in this way include clearly pointing out high-risk areas. These high-risk areas are due to the superposition of multiple unfavorable factors, such as old building structures, lack of effective fire protection facilities, dense population distribution, and large amounts of flammable materials stored. At the same time, hidden danger points will also be identified. Hidden danger points are specific locations or situations, such as damaged fire protection equipment, blocked evacuation passages, and faulty electrical lines. Once these hidden danger points are triggered, the risk of fire will rise sharply. For example, if the buildings in a certain area are mostly wooden structures and old, the number of fire hydrants is insufficient, and the area is densely populated and there are warehouses for flammable materials, then this area is likely to be assessed as a high-risk area, and if the door of an evacuation passage is blocked by debris, or a section of wire has signs of aging and short circuit, it may be identified as a hidden danger point. Through these diverse risk assessment results, a comprehensive and accurate decision-making basis can be provided for community fire prevention and management.

[0021] Step S400: Monitor the community in real time, and generate a trigger instruction based on the real-time monitoring information and the multiple risk assessment results.

[0022] Specifically, the community is monitored in real time through a series of monitoring equipment and technologies, including smoke sensors, temperature sensors, video surveillance cameras, etc. installed throughout the community. The real-time monitoring system continuously collects real-time data in the community, such as a sudden increase in smoke concentration in a certain area, abnormal temperature rise, abnormal human activities, etc. At the same time, these real-time monitoring information are combined with multiple risk assessment results generated in the previous steps. The risk assessment results provide the potential fire risk level and characteristics of the community under different circumstances. Through comprehensive analysis and comparison of real-time data and risk assessment results, if certain indicators in the real-time monitoring information exceed the preset threshold or match the high-risk characteristics in the risk assessment results, the system will generate corresponding trigger instructions. For example, if the smoke concentration in a certain area exceeds the safety limit set according to the risk assessment results, the system generates a trigger instruction to start the alarm; if multiple monitoring indicators show abnormalities at the same time and match the assessment characteristics of the high-risk area, a trigger instruction to notify the fire department is generated. The generation of trigger instructions is intended to take corresponding measures in a timely manner to prevent the occurrence of fires or to effectively respond to fires in the early stages of the occurrence of fires.

[0023] Step S500: Initiate a fire warning mechanism according to the trigger instruction, generate a fire emergency plan, combine the community fire knowledge graph to assist in response, and generate a fire response decision.

[0024] Specifically, when the generated trigger instruction is received, the fire warning mechanism is immediately activated according to the instruction. This includes triggering alarm sounds, flashing warning lights, sending text messages or application notifications to relevant personnel, etc., to quickly attract the attention of community residents and managers. At the same time, a fire emergency plan is generated according to pre-established rules and processes. This emergency plan contains a series of pre-set response measures, such as the planning of personnel evacuation routes, the deployment and use of fire-fighting equipment, and contact with external rescue forces. In the process of generating emergency plans, the community fire knowledge graph constructed previously will be combined for auxiliary response. The detailed information such as the community geographical layout, building structure, and distribution of fire-fighting facilities contained in the knowledge graph can provide accurate reference for the formulation of emergency plans. For example, according to the personnel distribution and evacuation channel of a building in the knowledge graph, the optimal evacuation route is determined; according to the location and performance of fire-fighting facilities in the knowledge graph, fire-fighting resources are reasonably deployed. Based on the integration and analysis of the above information, specific fire response decisions are finally generated. These decisions clarify the specific action steps, division of responsibilities, and time arrangements that should be taken in the current fire situation to ensure that a quick, orderly, and effective response can be carried out when a fire occurs.

[0025] In one possible implementation, Figure 2 As shown, step S100 also includes:

[0026] Step S110: traverse the community area information to extract the community three-dimensional structure for modeling and construct a community building model.

[0027] Step S120: Based on the community building model, multiple firefighting equipment locations are identified and combined with historical fire data records to collect data and generate data collection results.

[0028] Step S130: performing entity extraction based on the data collection result to generate multiple fire entity information.

[0029] Step S140: using a relationship extraction algorithm to identify association trajectories of the plurality of fire entity information, performing knowledge fusion on the plurality of fire entity information according to the association trajectories, and constructing the community fire knowledge graph.

[0030] Specifically, a comprehensive search and screening of community area information is initiated. This information comes from a wide range of sources, including but not limited to community planning drawings, geographic information system (GIS) data, aerial images, architectural design documents, etc. When processing architectural construction drawings, advanced computer vision algorithms and image processing technologies are used to accurately identify lines, symbols and text annotations in the drawings. Through complex coordinate transformation and geometric calculations, the two-dimensional information on the drawings is converted into precise three-dimensional coordinate data, so as to obtain the accurate length, width, height and detailed dimensions of each part of the building. For laser scanning data, a powerful point cloud processing algorithm is used to filter, classify and fit the massive point cloud data to reconstruct the complex surface shape and spatial position of the building. At the same time, combined with the terrain data in the geographic information data, the relative height and position relationship of the building in the entire community terrain are determined through the analysis of the digital terrain model (DTM) and the digital elevation model (DEM). Using professional three-dimensional modeling software, such as 3ds Max, SketchUp or programming-based modeling tools, such as the relevant libraries in Python, the extracted three-dimensional structural information is converted into a digital three-dimensional model. In the modeling process, attention will be paid to the restoration and accuracy of details. For complex building structures, methods such as hierarchical modeling and component-based construction can be used to ensure that the model can truly reflect the actual situation of community buildings. At the same time, in order to improve the usability and interactivity of the model, appearance attributes such as materials and textures will be added to the model, and relevant building information and attribute data will be annotated. Finally, a complete, accurate and detailed community building model is constructed, providing an intuitive and reliable basis for subsequent analysis, evaluation and decision-making.

[0031] The focus will be on the community building model constructed earlier. This model depicts the community's architectural structure in detail, covering not only the exterior outline, floor distribution, and internal room layout of each building, but also key structural elements such as stairs, corridors, and ventilation ducts. At the same time, the model also clearly marks the existing firefighting equipment information, such as the specific location of fire hydrants, the storage location of fire extinguishers, and the coverage of automatic sprinkler fire extinguishing systems. Based on such a detailed community building model, more potential firefighting equipment points that need to be added or optimized are identified in accordance with strict firefighting standards and actual emergency needs. At the same time, historical fire data records are comprehensively sorted out and integrated. These records contain extremely rich information, such as the specific building location of each fire (accurate to the building, unit and floor), the source location of the fire (kitchen, bedroom, warehouse, etc.), the cause of the fire (electrical aging short circuit, illegal use of fire and electricity, spontaneous combustion of flammable materials, etc.), the spread path and speed of the fire, the specific extent of casualties and property losses caused by the fire, and the actual use and effect feedback of various fire-fighting equipment in past fires (such as insufficient water pressure in some fire hydrants, expired fire extinguishers not replaced, etc.). When collecting data, the new and old fire-fighting equipment points marked in the community building model will be deeply cross-compared and comprehensively analyzed with the key elements in the historical fire data. For example, for a newly marked fire-fighting equipment point, it will be checked in detail whether there has been a fire in the area in the past. If so, the scale of the fire and the losses caused at that time will be analyzed to evaluate whether the newly marked equipment point can play an effective role in similar fire scenes in the future; for existing fire-fighting equipment points, it will be combined with its actual performance in historical fires to determine whether it is necessary to adjust the position or replace the equipment type. Through such rigorous, comprehensive and in-depth data collection work, a data collection result with rich content and great reference value is finally generated. This result presents in detail the complete structural information of community buildings, the precise distribution of all fire-fighting equipment, and the various internal connections and laws between them and historical fire data.

[0032] Based on the data collection results obtained above, we carry out entity extraction, conduct in-depth analysis and processing of a large amount of text data in the data collection results, and use natural language processing technology to identify and extract key entity information directly related to the fire. For the extraction of the fire location, we accurately locate the specific location from the description, which is the specific building, unit and room number of a community, or the specific store name and floor of a commercial area; the extraction of the fire time will accurately identify the specific year, month, day, hour, minute and second of the fire. When extracting the cause of the fire, we can identify specific cause statements such as electrical failure, improper use of open flames, spontaneous combustion of flammable materials, and arson from complex text descriptions. For the extraction of losses, we will clearly extract the number of casualties, the degree of injury, the specific amount of property losses, the category of damaged items and other detailed information. Through this extraction process, multiple clear, accurate and correct fire entity information is generated. These entity information provides basic and key data support for subsequent data analysis, risk assessment, emergency strategy formulation and other work.

[0033] The information of multiple fire entities is converted into a text sequence form suitable for LSTM model processing. The LSTM model reads each word or token in these text sequences in turn and remembers the previous information. It can learn the long-term dependencies and semantic information between different entities. When processing new fire entity information, the model can predict and identify the association tracks according to the learned patterns and rules. Based on these identified association tracks, the knowledge of multiple fire entity information is fused, and the entity information with similar association tracks is integrated together, such as all entity information that causes fires and certain losses due to electrical short circuits is classified into one category. Finally, based on this fused information, a community fire knowledge graph is constructed. In this graph, nodes represent different fire entities, such as location, time, cause, etc., and edges represent the relationship between them. The weight of the edge can represent the strength or importance of the relationship. Through this LSTM-based processing process, a knowledge graph reflecting the relationship between various elements of community fires can be effectively constructed, providing strong support for fire prevention and response.

[0034] In one possible implementation, step 140 further includes:

[0035] Step 141: performing word segmentation based on the plurality of fire entity information, performing part-of-speech tagging according to the word segmentation results, and generating a plurality of fire entity part-of-speech information.

[0036] Step 142: Determine community fire attribute intervals based on the historical fire data records.

[0037] Step 143: Define a set of relationship types to be extracted based on the community fire attribute interval.

[0038] Step 144: According to the set of relationship types to be extracted, a relationship extraction algorithm is used to perform relationship extraction on the multiple fire entity part-of-speech information to determine multiple relationship triples.

[0039] Step 145: construct an association network based on the multiple relationship triples, and determine the association trajectory according to the association network.

[0040] Specifically, multiple fire entity information is obtained. This information exists in the form of natural language text, including various descriptions and details about the fire. Next, the word segmentation technology in natural language processing is used to segment these texts according to the boundaries of words, and the continuous text is decomposed into independent words or vocabulary units. Then, according to the pre-established part-of-speech tagging rules or models, the part-of-speech of the results obtained by word segmentation is determined and annotated to determine the grammatical category of each word, such as noun, verb, adjective, adverb, etc. Finally, through this series of operations, multiple results that can clearly reflect the part-of-speech of each word in the fire entity information are generated, providing basic and key language feature information for subsequent processing and analysis.

[0041] A comprehensive and in-depth analysis of the collected historical fire data records is conducted. These historical data contain rich information such as the time, location, scale, loss, and cause of the fire. Through comprehensive consideration and statistical analysis of these data, the attribute intervals of community fires are delineated according to different measurement indicators. For example, according to the size of the economic losses caused by the fire, it can be delineated into a light loss interval (such as 0-100,000 yuan), a moderate loss interval (100,000 yuan-500,000 yuan), and a severe loss interval (more than 500,000 yuan). Or according to the scale of the fire, with the burning area as the standard, it can be delineated into a small fire interval (less than 100 square meters), a medium fire interval (100-500 square meters), and a large fire interval (more than 500 square meters). For another example, based on the frequency of fire, a low frequency interval (less than 1 time per year), a medium frequency interval (1-3 times per year), and a high frequency interval (more than 3 times per year) are set. By clearly delineating the attribute intervals of community fires in this way, the various attributes of fires can be more clearly classified and quantified, providing a clear and targeted basis for subsequent analysis and decision-making.

[0042] Since the attribute intervals of community fires are delineated according to historical fire data records, the set of relationship types to be extracted is defined based on these determined intervals. For example, if the intervals of fire loss degree are divided, the defined relationship types include the relationship between light loss and fire cause, the relationship between moderate loss and fire response time, and the relationship between severe loss and building structure. If divided according to the interval of fire scale, the defined relationship types include the relationship between small fires and surrounding environment, the relationship between medium-sized fires and fire equipment, and the relationship between large fires and rescue resource deployment. Or based on the interval of fire frequency, the defined relationship types can be the relationship between low-frequency fires and seasons, the relationship between medium-frequency fires and residents' fire awareness, and the relationship between high-frequency fires and community management measures. By defining the set of relationship types to be extracted based on the community fire attribute intervals, it is possible to specifically mine relationship information that is of great significance to fire prevention and control and response, and provide clear directions and priorities for subsequent risk assessment, prevention strategy formulation, and other work.

[0043] After clarifying the previously defined set of relationship types to be extracted, the relationship extraction algorithm is then used to conduct an in-depth analysis of the multiple fire entity part-of-speech information obtained previously. For example, if the set of relationship types to be extracted contains the relationship between the degree of fire loss and the cause of the fire, the algorithm will look for this corresponding relationship in the fire entity part-of-speech information. During the extraction process, multiple relationship triplets are determined in the form of "entity 1-relationship-entity 2". For example, relationship triplets such as "short circuit of wires-causing-moderate losses" and "illegal use of fire-causing-mild losses" are obtained. In this way, combinations with specific relationships can be accurately extracted from complex fire entity information, providing a basic and valuable data structure for further building an association network and analyzing fire laws.

[0044] Based on the obtained multiple relationship triples, we start to build an association network. In the process of building this network, the entities in each relationship triple are regarded as independent nodes. For example, "short circuit" and "moderate loss" in "short circuit-cause-moderate loss" are two nodes respectively, and the relationship between entities, that is, "cause" in the triple, is the edge connecting these nodes. By integrating multiple such relationship triples, a complex association network is formed. Subsequently, the constructed association network is analyzed in depth, and the connection, distribution characteristics and interaction of nodes and edges in the network are carefully studied. For example, by observing the connection density between nodes and the weight of edges, the association trajectories between multiple fire entities can be identified. These association trajectories can reflect a lot of important information, such as the propagation path of the fire, such as how it spreads from a certain starting point to other areas; or show the scope of the fire, clarify which places are directly or indirectly affected by the fire; and reveal common characteristics, such as certain types of buildings are prone to fires with similar loss levels under certain conditions. By determining these association trajectories, we can have a deeper understanding of the laws of fire occurrence and development, and provide strong support and guidance for fire prevention, monitoring and response.

[0045] In one possible implementation, step 140 further includes:

[0046] Step 146: Align the multiple fire entity part-of-speech information based on the association network to generate an information synchronization data set.

[0047] Step 147: Complete the relationship of the associated trajectory according to the information synchronization data set, update the associated trajectory according to the completion result, perform knowledge fusion on the multiple fire entity information, and construct the community fire knowledge graph.

[0048] Specifically, the previously constructed association network is used as a reference benchmark. Specific data structures, field definitions, and format specifications have been set in the association network. Next, multiple fire entity part-of-speech information is obtained. These information sources are diverse and there are differences in format and content. For each key data element, such as the time, location, objects involved, and relevant personnel of the fire, unified processing is carried out according to the standards set by the association network. After comprehensive sorting and adjustment of multiple fire entity part-of-speech information, they achieve a high degree of consistency and accuracy in data structure, format, and content, and finally generate a mutually matching, synchronized, and complete information synchronization data set. This data set provides a reliable and efficient data foundation for subsequent analysis, calculation, and application, avoiding misunderstandings and errors that may arise due to inconsistent data.

[0049] The association trajectory is processed based on the previously generated information synchronization dataset. Since the information synchronization dataset integrates standardized and consistent data, it can provide more details and supplementary information for the relationships in the association trajectory. For example, the original association trajectory only shows "cause of fire-result-partial property loss", but through the detailed records in the information synchronization dataset, it is found that there is also a hidden relationship such as "partial property loss-impact-resident life". Based on these newly discovered and supplemented relationships, the association trajectory is improved and supplemented. Then, according to the supplemented results, the association trajectory is updated so that it can more comprehensively and accurately reflect the various connections between fire entities. The updated association trajectory is used to perform knowledge fusion on multiple fire entity information. This means that the originally scattered and isolated various information about the fire, such as the time, location, cause, loss, rescue situation, etc. of the fire, are integrated according to the relationships revealed by the association trajectory. Through this deep fusion, a community fire knowledge graph is constructed. In this knowledge graph, the various entities of the fire are nodes, and the relationships between them are edges connecting the nodes, forming a clear, intuitive and comprehensive network structure. Such a knowledge graph can clearly display the complex relationships between fire-related elements and provide comprehensive and in-depth knowledge support for community fire prevention, response and management.

[0050] In one possible implementation, step 147 further includes:

[0051] Step 1471: Perform implicit association identification based on the information synchronization data set to generate multiple implicit association trajectories.

[0052] Step 1472: Evaluate the plurality of implicitly associated trajectories to determine the degree of association of the plurality of implicitly associated trajectories.

[0053] Step 1473: Determine whether the correlation degree of the plurality of implicit correlation trajectories is greater than or equal to a preset correlation threshold.

[0054] Step 1474: extracting the implicit associated tracks whose association degree is greater than or equal to the preset association threshold, performing relationship completion on the associated tracks, and generating the completion result.

[0055] Specifically, the fire entity information in the data set is carefully analyzed, focusing on the two key factors of geographical location and occurrence time. If two fire entities are found to be close to each other in geographical location, such as adjacent buildings in the same community, or in similar locations on the same street. At the same time, their occurrence time is also relatively close, such as occurring within a few days or weeks, then it can be inferred that there may be some kind of internal connection between the two fire entities. This connection may be that the occurrence of one fire has an impact on the other, such as one fire causing firefighting resources to be concentrated, making the adjacent area insufficient to respond when a fire occurs. It may also be that there is a common cause that caused the occurrence of the two fires, such as the aging of the power facilities in the area, which caused fires in different locations at a similar time. Based on such speculation and analysis, these potential hidden associations are identified, sorted and recorded, and multiple hidden association trajectories are generated. Such hidden association identification helps to understand the patterns and laws of fire occurrence more comprehensively and deeply.

[0056] In the process of evaluating multiple implicit association trajectories, multiple factors are considered to determine their degree of association. The number of nodes in the trajectory is a consideration. If a implicit association trajectory involves more fire entity nodes, it means that its impact range is wider and the degree of association is relatively high. Secondly, the strength of the edge, that is, the reliability of the relationship, is crucial. For example, if the association between two fire entities is based on clear and strong evidence, such as a common power failure source causing the fire, then the strength of this edge is greater; while if the association is only inferred based on some more vague or indirect clues, the strength of the edge is relatively weak. In addition, the degree of association between the trajectory and known fire events will also be evaluated. If the implicit association trajectory has a close similarity or direct association with a confirmed and in-depth fire event, then its rationality and importance may be higher. Through the comprehensive analysis and weighing of these factors, each implicit association trajectory is assigned a specific degree of association value or level to quantify its importance in the entire fire association system.

[0057] After the correlation of multiple implicit correlation trajectories is determined, the correlation of each implicit correlation trajectory will be compared with the preset correlation threshold. The preset correlation threshold is a standard value set in advance based on actual needs and experience. If the correlation of a certain implicit correlation trajectory is greater than or equal to the preset threshold, it means that the implicit correlation trajectory has high importance and reliability, which is of great significance for subsequent analysis and decision-making. On the contrary, if the correlation is less than the preset threshold, it is considered that the implicit correlation trajectory is relatively unimportant or unreliable, and will be given less attention or not adopted in subsequent processing. Through such judgment, those truly valuable implicit correlation trajectories can be screened out so that resources can be used more effectively for further analysis and application.

[0058] Those implicit association trajectories whose correlation is greater than or equal to the preset association threshold are screened out. These implicit association trajectories are considered to be of sufficient importance and reliability. Then, these screened implicit association trajectories are used to supplement and improve the existing association trajectories. For example, the original association trajectory only contains direct causal relationships, such as "electrical failure-cause-fire", and through the extracted implicit association trajectory, new relationships such as "fire occurrence-impact-difficulty in evacuating surrounding residents" are discovered, and then added to the association trajectory. After such supplementary operations, the result of relationship completion is finally generated. This completion result makes the association trajectory richer, more comprehensive and accurate, and can more completely present the complex relationship between fire-related entities. Doing so helps to have a deeper understanding of the development process, impact range and potential correlation factors of the fire, and provide stronger support for the subsequent fire analysis, prevention and response strategy formulation.

[0059] In one possible implementation, step 400 further includes:

[0060] Step 410: Delimit a monitoring range based on the community building model, and connect multiple sensor devices to build a sensor monitoring network.

[0061] Step 420: Activate the sensor monitoring network to monitor the community in real time and obtain the real-time monitoring information.

[0062] Step 430: Set multiple risk levels based on the multiple risk assessment results, and calculate warning triggering conditions according to the multiple risk levels.

[0063] Step 440: Determine whether the real-time monitoring information reaches a preset risk level. If so, it is deemed that the early warning trigger condition is met and the trigger instruction is generated.

[0064] Specifically, the constructed community building model is used as the basis. This model contains information such as the detailed layout, structure, functional zoning, etc. of the buildings in the community. Carefully analyze the community building model, consider factors such as the purpose of the building (such as residential, commercial, public facilities, etc.), the density of people, the distribution of fire hazards, etc., so as to clarify which areas are at high risk of fire and need to be monitored. Then, select multiple appropriate sensor devices, which have different functions and can monitor various parameters related to fire, such as smoke sensors for smoke concentration, temperature sensors for measuring temperature, flame sensors for detecting flames, etc. Next, connect these sensor devices to each other through wired or wireless connections to form a network that can work together. In this network, each sensor device acts as a node, and data can be transmitted and shared between them.

[0065] When the sensor monitoring network is built, it will be started or activated to start working. The activation operation involves powering on each sensor device, initializing the settings, and starting the data collection and transmission program. Once the sensor monitoring network is activated, it will monitor the entire community in real time, which means that each sensor device will continuously sense and measure the relevant environmental parameters of its location. For example, the smoke sensor will detect the concentration of smoke particles in the air in real time, the temperature sensor will continuously record the temperature changes of the surrounding environment, and the flame sensor will be ready to capture possible flame signals at any time. These real-time collected and transmitted data together constitute real-time monitoring information. This information contains the immediate environmental status of different locations in the community and reflects whether there are abnormal conditions that may cause fires. Through the analysis and processing of these real-time monitoring information, potential fire risks can be discovered in a timely manner, and appropriate measures can be taken to prevent or respond.

[0066] The operation is performed based on the multiple risk assessment results obtained above, which include fire risk analysis of different areas, different building types, and different time periods in the community. Based on these rich assessment results, the risks are divided into multiple different levels, for example, low risk, medium risk, and high risk. The classification criteria may take into account a variety of factors, such as the probability of fire, the severity of possible casualties and property losses, the type of fire source, and the potential spread rate. The warning trigger conditions are calculated based on the multiple risk levels that have been divided, which means that it is necessary to determine what specific conditions will trigger the warning at each risk level. For example, at the low risk level, the warning is triggered when the temperature continues to rise by more than a certain amount or the smoke concentration reaches a certain lower threshold; at the medium risk level, the warning trigger conditions are more stringent, such as the temperature rises rapidly and the smoke concentration reaches a high level; and at the high risk level, even a slight smoke or a small increase in temperature will trigger the warning, which can ensure that the warning system can issue an alarm in a timely and accurate manner under different risk conditions, thereby improving the community's ability to prevent and respond to fires.

[0067] In one possible implementation, step 430 further includes:

[0068] Step 431: Perform weighted calculation based on the multiple risk assessment results, arrange the calculation results in descending order, and set the multiple risk levels.

[0069] Step 432: Retrieve historical risk event records and match them with the multiple risk levels to generate multiple risk event matching results.

[0070] Step 433: Perform a comprehensive analysis based on the multiple risk event matching results and set the warning triggering condition.

[0071] Specifically, multiple risk assessment results are obtained. These results contain assessment data from multiple different aspects, such as the possibility of fire, the dangerousness of the fire source, the fire resistance of the building, the difficulty of evacuation, etc. Then, a corresponding weight is assigned to each assessment result, and the weight is usually determined based on the degree of influence of these factors on the overall risk. For example, the possibility of fire is assigned a higher weight because it has a greater impact on the risk, while the weight of the fire resistance of the building is relatively low. Weighted calculation is performed on each assessment result after the weight is assigned, each assessment result is multiplied by its corresponding weight, and then these products are added to obtain a comprehensive weighted calculation result. After the weighted calculation is completed, these calculation results are arranged in descending order from large to small. According to the results after arrangement, different intervals are divided to set multiple risk levels. For example, the first 10% of the results can be set as the highest risk level, the next 20% can be set as the second highest risk level, the middle 50% can be set as the medium risk level, the next 15% can be set as the second lowest risk level, and the last 5% can be set as the lowest risk level. In this way, risk levels can be set more scientifically and reasonably, key risk factors can be highlighted, and a basis can be provided for subsequent risk management and early warning.

[0072] Retrieve the records of historical risk events that occurred in the past from the storage system. These records record in detail various key information of each risk event, including the time, location, specific circumstances, losses caused, and related environment and conditions of the event. Compare and match the characteristics and circumstances of these historical risk events with the multiple risk levels set previously. For example, for a historical risk event, if it has a high probability of fire, serious losses, dangerous fire sources, and difficulty in evacuating people, it will be matched to a high risk level; and another event, if its various indicators are relatively mild, will be matched to a medium or low risk level. In the matching process, multiple factors will be comprehensively considered to ensure the accuracy and rationality of the matching. Finally, through such matching operations, multiple risk event matching results are generated.

[0073] Conduct a comprehensive analysis of the matching results of multiple risk events to view the characteristics, rules and influencing factors of risk events at different risk levels. Based on such analysis, set specific warning trigger conditions. For example, if it is found that at a high risk level, as long as the temperature rises by more than 5 degrees Celsius and the smoke concentration reaches a certain value, a serious accident is likely to occur, then this situation is set as the warning trigger condition for the high risk level. In this way, reasonable and effective warning trigger conditions are set for each risk level, so that warnings can be issued in a timely and accurate manner in actual situations.

[0074] In one possible implementation, step 500 further includes:

[0075] Step 510: Based on the trigger instruction, the fire warning mechanism is activated to perform data sensing on the community to generate a fire sensing data set.

[0076] Step 520: Predicting a fire change trend based on the fire sensor data set, and formulating a fire emergency plan according to the fire change trend.

[0077] Step 530: Synchronize the fire emergency plan to the community fire knowledge graph for verification, and generate a plan verification result.

[0078] Step 540: Dynamically adjust the fire emergency plan based on the plan verification result to generate the fire response decision.

[0079] Specifically, when a trigger command is received, it is like a signal being activated, and the fire warning mechanism is immediately activated. At this time, sensors distributed in various key locations in the community, including sensors that specifically monitor smoke concentration, sensors that measure temperature, and devices that capture flame images, all begin to enter a highly sensitive working state. The smoke concentration sensor collects concentration data of smoke particles in the surrounding air at a very high frequency. If the smoke concentration gradually increases, this is one of the early signs of a fire. The temperature sensor continuously senses changes in ambient temperature. Once the temperature rises abnormally, especially rapidly in a short period of time, it means that there is a fire source nearby or the fire is spreading. The flame image acquisition device uses image recognition technology to obtain visual information of the flame in real time. It can clearly record details such as the shape, size, and color of the flame. These real-time data from different types of sensors, such as specific smoke concentration values, accurate temperature readings, and intuitive flame images, are collected and integrated to form a comprehensive, multi-dimensional fire sensing data set.

[0080] The data in the fire sensor data set are preprocessed, including cleaning outliers, processing missing values, and data standardization, and then analyzed using the decision tree algorithm. The decision tree uses factors such as the rate of change of smoke concentration, the rate of temperature rise, and the characteristics of the flame image as input features. For example, if the rate of change of smoke concentration exceeds a certain threshold per minute and the rate of temperature rise exceeds a specific value per second, the decision tree will make branch judgments based on these conditions. By continuously dividing and judging these features, the decision tree can establish a prediction model to predict the changing trend of the fire, and formulate a fire emergency plan based on this prediction result. In this way, based on the analysis of the fire sensor data set by the decision tree algorithm, a targeted and operational fire emergency plan is formulated.

[0081] The previously formulated fire emergency plan is synchronized and integrated with the community fire knowledge graph, which contains detailed information about the community, such as the structure and materials of the building, the location and performance of the fire-fighting facilities, the layout of the evacuation passages, the distribution of residents, etc. When the fire emergency plan is input into the knowledge graph, the plan is fully verified based on the rich information in the graph. For example, check whether there are blockages or obstacles in the evacuation routes planned in the plan, whether the deployment of fire-fighting equipment can cover key areas, whether the division of labor of personnel is reasonable, etc. Through this comprehensive verification and analysis, the plan verification results are finally generated. This result shows that the plan is completely feasible, and it may also point out the shortcomings in the plan, such as the inability of a certain evacuation passage to be used normally, resulting in some residents being unable to evacuate according to the planned route, or the aging of some fire-fighting facilities cannot meet the fire-fighting needs in the plan.

[0082] The generated plan verification results are obtained, which include information such as feasibility assessment of the fire emergency plan, problems found, and suggestions for improvement. Then, the fire emergency plan is dynamically adjusted according to the deficiencies and areas that need to be optimized pointed out in the plan verification results. For example, if the verification results show that a certain evacuation channel is unavailable, it is necessary to replan the evacuation route and select other feasible channels; if it is found that the allocation of fire resources is unreasonable, it is necessary to readjust the deployment of fire equipment and personnel. In the process of adjustment, the actual situation and various constraints will be fully considered to ensure that the adjusted plan is operational and effective. By continuously optimizing and improving the plan, a more reasonable, accurate and practical fire response decision will be generated. This decision will serve as a guideline for taking action when an actual fire occurs, which can more effectively respond to fires, reduce losses and ensure the safety of personnel.

[0083] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0084] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0085] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A community fire warning method combined with knowledge graph, characterized in that: The method comprises: Traverse the community area information to collect data, and build a community fire knowledge graph based on the data collection results; Perform feature extraction based on the community fire knowledge graph to obtain multiple fire features; Constructing a fire risk assessment model based on the multiple fire characteristics to conduct fire risk assessment on the community and generate multiple risk assessment results; Conducting real-time monitoring of the community, and generating trigger instructions based on the real-time monitoring information combined with the multiple risk assessment results; The fire warning mechanism is activated according to the trigger instruction, a fire emergency plan is generated, and the community fire knowledge graph is combined to assist in the response and generate a fire response decision.

2. The community fire warning method combined with knowledge graph according to claim 1, characterized in that: Traverse the community area information to collect data, and build a community fire knowledge graph based on the data collection results. The method includes: Traversing the community area information, extracting the community three-dimensional structure for modeling, and constructing a community building model; Based on the community building model, multiple firefighting equipment locations are identified and combined with historical fire data records to collect data and generate data collection results; Perform entity extraction based on the data collection results to generate multiple fire entity information; A relationship extraction algorithm is used to identify the association tracks of the multiple fire entity information, and knowledge fusion is performed on the multiple fire entity information according to the association tracks to construct the community fire knowledge graph.

3. The community fire warning method combined with knowledge graph according to claim 2, characterized in that: The method of identifying the associated tracks of the plurality of fire entity information by using a relation extraction algorithm includes: Perform word segmentation based on the multiple fire entity information, perform part-of-speech tagging according to the word segmentation results, and generate multiple fire entity part-of-speech information; Delineating community fire attribute intervals based on the historical fire data records; Defining a set of relationship types to be extracted based on the community fire attribute interval; According to the set of relationship types to be extracted, using a relationship extraction algorithm to extract relationship from the plurality of fire entity part-of-speech information, and determining a plurality of relationship triples; An association network is constructed based on the multiple relationship triples, and the association trajectory is determined according to the association network.

4. The community fire warning method combined with knowledge graph as claimed in claim 3 is characterized in that: The knowledge fusion of the plurality of fire entity information is performed according to the association trajectory to construct the community fire knowledge graph, and the method includes: Based on the association network, the plurality of fire entity part-of-speech information are data aligned to generate an information synchronization data set; The relationship of the associated trajectory is completed according to the information synchronization data set, and the associated trajectory is updated according to the completion result to perform knowledge fusion on the multiple fire entity information to construct the community fire knowledge graph.

5. The community fire warning method combined with knowledge graph according to claim 4, characterized in that: The method for completing the relationship of the associated trajectory according to the information synchronization data set includes: Perform implicit association identification based on the information synchronization data set to generate multiple implicit association trajectories; evaluating the plurality of implicit correlation trajectories to determine correlation degrees of the plurality of implicit correlation trajectories; Determining whether the correlation degree of the plurality of implicit correlation trajectories is greater than or equal to a preset correlation threshold; The implicit correlation track whose correlation degree is greater than or equal to the preset correlation threshold is extracted to complete the relationship of the correlation track to generate the completion result.

6. The community fire warning method combined with knowledge graph as claimed in claim 2, characterized in that: The community is monitored in real time, and a trigger instruction is generated based on the real-time monitoring information and the multiple risk assessment results, the method comprising: Delimiting the monitoring range based on the community building model, connecting multiple sensor devices to build a sensor monitoring network; Activating the sensor monitoring network to monitor the community in real time and obtaining the real-time monitoring information; Setting multiple risk levels based on the multiple risk assessment results, and calculating warning triggering conditions according to the multiple risk levels; Determine whether the real-time monitoring information reaches a preset risk level. If so, it is deemed that the early warning trigger condition is met and the trigger instruction is generated.

7. The community fire warning method combined with knowledge graph according to claim 6, characterized in that: Based on the multiple risk assessment results, multiple risk levels are set, and according to the multiple risk levels, an early warning trigger condition is calculated, the method includes: Performing weighted calculation based on the multiple risk assessment results, arranging the calculation results in descending order, and setting the multiple risk levels; Retrieving historical risk event records and matching them with the multiple risk levels to generate multiple risk event matching results; A comprehensive analysis is performed based on the matching results of the multiple risk events to set the warning triggering conditions.

8. The community fire warning method combined with knowledge graph according to claim 1, characterized in that: The fire warning mechanism is activated according to the trigger instruction, a fire emergency plan is generated, and an auxiliary response is performed in combination with the community fire knowledge graph to generate a fire response decision, and the method includes: Based on the trigger instruction, the fire warning mechanism is activated to perform data sensing on the community to generate a fire sensing data set; Predicting a fire change trend based on the fire sensor data set, and formulating a fire emergency plan according to the fire change trend; Synchronize the fire emergency plan to the community fire knowledge graph for verification, and generate a plan verification result; The fire emergency plan is dynamically adjusted based on the plan verification result to generate the fire response decision.

Images