Emergency decision support scheme generation method based on knowledge graph

Abstract

The invention discloses an emergency decision support scheme generation method based on a knowledge graph, and relates to the technical field of engineering safety management. Comprising the following steps: constructing an accident case-emergency plan knowledge graph; establishing a risk-accident scene model library; setting corresponding flow plan templates according to different risks and accident scenes; acquiring various monitoring data of an engineering site, performing fusion analysis on the monitoring data and user input, selecting a corresponding plan template, and generating an emergency plan; establishing an accident investigation report-emergency plan text library, and obtaining historical cases and emergency plans similar to the scene according to fusion analysis; and the generated emergency plan, the historical cases similar to the scene and the emergency plan are split according to task nodes and sent to a task executor, and an information communication mechanism between the emergency decision command center and the action executor is established. According to the invention, an emergency plan can be generated in a targeted manner, a command is accurately sent to a first site, and disaster victims and rescue workers are assisted to carry out self-rescue, mutual rescue and emergency rescue.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering safety management, and particularly to a method for generating an emergency decision support solution based on a knowledge graph. Background Art

[0002] During the construction and operation of engineering projects, emergencies are likely to occur due to various complex and changeable risk factors. Once an accident occurs, it will pose a threat to the lives and property safety of on-site workers. Therefore, the emergency disposal of production safety accidents is crucial for engineering safety management. Production and business operation entities need to compile standardized emergency plans. However, the current emergency plans are mostly formalized and stored in a single text form, with low utilization rate of the text. When an accident occurs, the efficiency of emergency disposal requires high capabilities and experience of decision-making commanders, resulting in inability to respond quickly.

[0003] Currently, there are related methods for structuring emergency plan texts, such as establishing a digital emergency plan database and generating an optimized emergency plan according to problems using algorithm models. In addition, there are also processes for processing emergency plan texts, and generating emergency disposal decision support for each stage according to the node tasks in the process. The known methods have basically achieved the generation of emergency plans for different situations, but are only limited to providing decision support for commanders, unable to implement the function of notifying the first executor in the first place, and still requiring decision-makers to deploy, so that the emergency disposal actions cannot be executed at the fastest speed, which may affect the final accident loss situation.

[0004] Therefore, proposing a method for generating an emergency decision support solution based on a knowledge graph to solve the difficulties existing in the prior art is an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method for generating an emergency decision support solution based on a knowledge graph, which can generate targeted emergency plans, perform process modeling on specific measures, associate the executors of each action in the process, accurately send commands to the first scene, and assist affected personnel and rescue personnel in carrying out self-help and mutual rescue and emergency rescue.

[0006] To achieve the above object, the present invention adopts the following technical solution: A method for generating an emergency decision support solution based on a knowledge graph, comprising:

[0007] S1. For different risks and accident scenarios, collect existing accident investigation reports and emergency plans, perform structured processing, and construct an accident case - emergency plan knowledge graph;

[0008] S2. Establish a risk - accident scenario model library for specific problems of specific risks and accident scenarios;

[0009] S3. Set corresponding process-based pre-plan templates according to different risks and accident scenarios, stipulate the tasks and actions of various types of entities, and set dynamic slots.

[0010] S4. Obtain various monitoring data affecting work safety at the project site, conduct fusion analysis on the monitoring data and user input, determine specific risks or accident scenarios, select the corresponding pre-plan template, fill in slot information according to the monitoring data and other resource data, and generate an emergency plan.

[0011] S5. Establish an accident investigation report - emergency plan text library, conduct fusion analysis on the monitoring data and user input, obtain the corresponding scenario categories, and through natural language processing and knowledge graph retrieval, obtain historical cases similar to the on-site scenario and similar emergency plans.

[0012] S6. Split the generated emergency plan, historical cases similar to the on-site scenario, and emergency plans into structured emergency plan fragments for the action executors according to the task nodes, package the complete emergency plan, the structured emergency plan fragments for each action executor, and the historical cases and emergency plans similar to the on-site scenario, and send them to the task executors to assist the direct executors in emergency disposal. At the same time, establish an information communication mechanism between the emergency decision-making and command center and the action executors until the risk is lifted and the emergency rescue operation is completed.

[0013] For the above method, optionally, the specific steps for constructing the accident case - emergency plan knowledge graph in S1 are as follows:

[0014] S101. Analyze the emergency disposal requirements for different risks and accident scenarios, collect the corresponding accident investigation report and emergency plan text data, uniformly number them, and perform standardized deconstruction on various types of text data.

[0015] S102. Perform entity annotation on the text data, use the BERT-CRF entity extraction model to extract entities from the annotated text data, and obtain the entity label sequence.

[0016] S103. Determine the relationship type between entities according to the entity label sequence to obtain knowledge triples.

[0017] S104. Extract the keywords of the accident investigation report and emergency plan text data through the TextRank algorithm to form triples of text numbers and keywords and triples of text numbers and text summaries.

[0018] S105. Establish a spatial network model for key locations in the target working environment to form triples of each spatial position.

[0019] S106. Store the triples of text numbers and keywords, the triples of text numbers and text abstracts, and each spatial position triple into the graph database to construct an accident case - emergency plan knowledge graph composed of triples.

[0020] For the above method, optionally, the specific steps for establishing the risk - accident scenario model library in S2 are as follows:

[0021] S201. Sort out the accident scenarios and key monitoring indicators that may occur in the target working environment, and establish the corresponding relationship between the accident scenarios and the key monitoring indicators.

[0022] S202. Based on the monitoring data of on - site sensors, set up an accident scenario discrimination model to realize the mapping from on - site monitoring data to accident scenarios.

[0023] S203. For the areas that need video monitoring, set up a video monitoring system and write a video analysis program to discriminate accident scenarios and realize the mapping from the working - site scene to accident scenarios.

[0024] S204. Classify the accident scenarios in the target working environment, number the accident scenarios, sort out the discrimination criteria for the occurrence of different accident scenarios, and correspond them to the omen, fact, and emergency - handling entities in the accident case - emergency plan knowledge graph to construct a risk - accident scenario model library.

[0025] For the above method, optionally, the specific steps for stipulating the tasks and actions of each type of subject and setting up dynamic slots in S3 are as follows:

[0026] S301. According to the standardized structure of the emergency plan, divide the whole plan into seven stages: information reporting, warning, response initiation, emergency handling, emergency support, response termination, and post - handling, and determine the task nodes to be completed in each stage and the signals for the start and end of the tasks.

[0027] S302. Determine the specific behavioral actions required to be completed at each task node and clarify the executor of the actions.

[0028] S303. Establish a Petri - net structure model of the process - based emergency plan template.

[0029] S304. Refine the specific behavioral actions in the task nodes to correspond to the behavioral action entity nodes in the accident case - emergency plan knowledge graph.

[0030] S305. Retrieve the spatial network model in the target working environment from the knowledge graph, calculate the shortest path according to the positions of relevant personnel or units and the forward target positions, and form the analysis results of the disaster - avoidance route and disaster - relief route.

[0031] S306. Set dynamic slots for variables in specific behavioral actions. The variables include: the specific executor of the rescue action, the affected people, rescue supplies, safe locations, accident locations, affected areas, disaster avoidance routes, and disaster relief routes. The names and specific contents of the dynamic slots correspond to the accident case - emergency plan knowledge graph.

[0032] For the above method, optionally, the specific steps for generating an emergency plan in S4 are as follows:

[0033] S401. Access various monitoring data affecting work safety at the project site and set an input end to receive information on possible influencing factors discovered by on-site workers, safety inspectors, and the emergency command center based on experience.

[0034] S402. Integrate and analyze the user input data, on-site monitoring data, and discriminant results in the model library, process them into rule forms, retrieve in the risk - accident scenario model library to match accident scenarios that meet the requirements, and then retrieve the corresponding process - based emergency plan templates according to the matched accident scenarios.

[0035] S403. Read the required information from the accident case - emergency plan knowledge graph according to the slots to be filled in the emergency plan, fill the slots, and generate complete emergency plan information.

[0036] S404. Fill the generated complete emergency plan information into the emergency plan Petri net structure model corresponding to the scenario to generate a process - based emergency plan, which is called the overall emergency plan.

[0037] S405. Extract each action subject from the process - based emergency plan and split the entire process - based emergency plan into emergency plans directly related to each action subject, which are called sub - emergency plans.

[0038] For the above method, optionally, the specific steps for obtaining historical cases and similar emergency plans similar to the on - site scenario in S5 are as follows:

[0039] S501. Organize the relevant accident investigation reports and emergency plan text data of specific accident scenarios, enter them into the unstructured database, and establish an accident investigation report - emergency plan text library.

[0040] S502. Based on the monitoring data and information input by the user, search for similar historical cases and similar emergency plans in the accident case - emergency plan knowledge graph.

[0041] S503. Build a text recommendation model. Use each accident investigation report text in the accident investigation report - emergency plan text library as a sample for model training, use keywords as features, train the text recommendation model, and calculate the similarity between texts through the graph collaborative filtering algorithm to achieve the recommendation of accident investigation reports and corresponding emergency plan texts based on keywords.

[0042] S504. Establish a text similarity analyzer based on Simhash and BERT, and obtain historical cases similar to the on-site scenario and similar emergency response plans according to the input text materials.

[0043] For the above method, optionally, the specific steps in S6 are as follows:

[0044] S601. For each action subject, package the overall emergency response plan, sub-emergency response plans, historical cases similar to the on-site scenario, and similar emergency response plans to form a decision support document.

[0045] S602. Send each decision support document to each task executor respectively.

[0046] S603. Return the information reception situation to form the statistical information of the emergency decision-making command center, and at the same time prepare to receive the next round of information. According to the information returned by each action subject, update the decision support plan and prepare to generate the next round of decision support plans.

[0047] It can be seen from the above technical solutions that compared with the prior art, the present invention provides a method for generating an emergency decision support plan based on a knowledge graph, which has the following beneficial effects: The present invention decomposes the existing plan, splits it by stage, and recombines it according to different scenarios to generate more targeted plans and disposal measures; the generated emergency response plan is presented in a simple visual form, making the emergency response plan easier to use and having a more significant support effect on actual emergency decision-making; the present invention splits the specific action actions in the generated emergency response plan, accurately determines the executor of the actions, and realizes the function of sending action instructions to the corresponding executor, improving the emergency command efficiency and the emergency disposal ability. Description of the Drawings

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0049] Figure 1 It is a flowchart of a method for generating an emergency decision support plan based on a knowledge graph provided by the present invention;

[0050] Figure 2 It is a Petri net model of the emergency disposal process for coal mine water disasters in an embodiment of a method for generating an emergency decision support plan based on a knowledge graph provided by the present invention. Detailed Embodiments

[0051] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0052] In this application, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. The terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the presence of additional identical elements in the process, method, article or device including the element.

[0053] Refer to Figure 1 As shown, the present invention discloses a method for generating an emergency decision support solution based on a knowledge graph, including:

[0054] S1. For different risks and accident scenarios, collect existing accident investigation reports and emergency plans, perform structured processing, and construct an accident case - emergency plan knowledge graph;

[0055] S2. Establish a risk - accident scenario model library for specific problems of specific risks and accident scenarios;

[0056] S3. Set corresponding process - based pre - plan templates according to different risks and accident scenarios, stipulate the tasks and actions of various types of entities, and set dynamic slots;

[0057] S4. Obtain various monitoring data affecting work safety at the engineering site, perform fusion analysis on the monitoring data and user input, determine specific risks or accident scenarios, select the corresponding pre - plan template, and fill the slot information according to the monitoring data and other resource data to generate an emergency plan;

[0058] S5. Establish an accident investigation report - emergency plan text library, obtain the corresponding scenario category according to the fusion analysis of the monitoring data and user input, and obtain historical cases and similar emergency plans similar to the on - site scenario through natural language processing and knowledge graph retrieval;

[0059] S6. Split the generated emergency plan, historical cases similar to the on-site scenario, and the emergency plan into structured emergency plan fragments for the action executors according to the task nodes. Package and send the complete emergency plan, the structured emergency plan fragments for each action executor, the historical cases similar to the on-site scenario, and the emergency plan to the task executors to assist the direct executors in emergency disposal. At the same time, establish an information communication mechanism between the emergency decision-making and command center and the action executors until the risk is lifted and the emergency rescue operation is completed.

[0060] Furthermore, the specific steps for constructing the accident case - emergency plan knowledge graph in S1 are as follows:

[0061] S101. Analyze the emergency disposal requirements for different risks and accident scenarios, collect the corresponding accident investigation reports and emergency plan text data, assign unified numbers, and perform standardized deconstruction on various types of text data.

[0062] S102. Perform entity annotation on the text data, and use the BERT - CRF entity extraction model to extract entities from the annotated text data to obtain entity label sequences.

[0063] S103. Determine the relationship types between entities according to the entity label sequences to obtain knowledge triples.

[0064] S104. Extract the keywords of the accident investigation reports and emergency plan text data through the TextRank algorithm to form triples of text numbers and keywords and triples of text numbers and text summaries.

[0065] S105. Establish a spatial network model for the key locations in the target working environment to form triples of each spatial position.

[0066] S106. Store the obtained triples of text numbers and keywords, triples of text numbers and text summaries, and triples of each spatial position in the graph database to construct an accident case - emergency plan knowledge graph composed of triples.

[0067] Furthermore, the specific steps for establishing the risk - accident scenario model library in S2 are as follows:

[0068] S201. Sort out the possible accident scenarios and key monitoring indicators in the target working environment, and establish the corresponding relationship between the accident scenarios and the key monitoring indicators.

[0069] S202. Based on the monitoring data of on-site sensors, set an accident scenario discrimination model based on thresholds or mathematical models to realize the mapping from on-site monitoring data to accident scenarios.

[0070] S203. For the areas that need video monitoring, such as the areas where the operating status of machines and the working status of workers need to be monitored, establish a video monitoring system, and write a video analysis program to identify accident scenarios and realize the mapping from the work site scenario to the accident scenario;

[0071] S204. Classify the accident scenarios of the target working environment, number the accident scenarios, sort out the discrimination criteria for different accident scenarios, and correspond them to the omen, fact, and emergency disposal entities in the accident case - emergency plan knowledge graph to construct a risk - accident scenario model library.

[0072] Further, in S3, stipulate the tasks and actions of each type of entity and set dynamic slots. The specific steps are as follows:

[0073] S301. According to the standardized structure of the emergency plan, divide the whole plan into seven stages: information reporting, early warning, response initiation, emergency disposal, emergency support, response termination, and post - incident handling, and determine the task nodes to be completed in each stage and the signals for the start and end of the tasks;

[0074] S302. Determine the specific behavioral actions required for each task node and clarify the executor of the actions;

[0075] S303. Establish a Petri net structure model of the process - based emergency plan template;

[0076] S304. Refine the specific behavioral actions in the task nodes and correspond them to the behavioral action entity nodes in the accident case - emergency plan knowledge graph;

[0077] S305. Retrieve the spatial network model in the target working environment from the knowledge graph, calculate the shortest path according to the positions of relevant personnel or units and the forward target positions, and form the analysis results of the disaster - avoidance route and disaster - relief route;

[0078] S306. Set dynamic slots for the variables in the specific behavioral actions. The variables include: the specific executor of the rescue action, the affected personnel, the rescue materials, the safe location, the accident location, the affected area, the disaster - avoidance route, and the disaster - relief route. The names and specific contents of the dynamic slots correspond to the accident case - emergency plan knowledge graph.

[0079] Further, the specific steps for generating the emergency plan in S4 are as follows:

[0080] S401. Connect various monitoring data that affect work safety at the engineering site and set an input end to receive information on possible influencing factors discovered by on - site workers, safety inspectors, and the emergency command center based on experience;

[0081] S402. Integrate and analyze the user input data, on-site monitoring data, and the discrimination results in the model library, process them into a rule form, retrieve in the risk-accident scenario model library to match the accident scenarios that meet the requirements, and then retrieve the corresponding process-based emergency plan template according to the matched accident scenarios;

[0082] S403. Read the required information from the accident case-emergency plan knowledge graph according to the slots to be filled in the emergency plan, fill in the slots, and generate complete emergency plan information;

[0083] S404. Fill the generated complete emergency plan information into the emergency plan Petri net structure model corresponding to the scenario to generate a process-based emergency plan, which is called the general emergency plan;

[0084] S405. Extract each action subject from the process-based emergency plan, and split the entire process-based emergency plan into emergency plans directly related to each action subject, which are called sub-emergency plans.

[0085] Further, the specific steps for obtaining historical cases and similar emergency plans similar to the on-site scenario in S5 are as follows:

[0086] S501. Organize the relevant accident investigation reports and emergency plan text data of specific accident scenarios, enter them into the unstructured database, and establish an accident investigation report-emergency plan text library;

[0087] S502. Based on the monitoring data and the information input by the user, search for similar historical cases and similar emergency plans based on the accident case-emergency plan knowledge graph;

[0088] S503. Build a text recommendation model, use each accident investigation report text in the accident investigation report-emergency plan text library as a sample for model training, use keywords as features, train the text recommendation model, and calculate the similarity between texts through the graph collaborative filtering algorithm to realize the recommendation of accident investigation reports and corresponding emergency plan texts based on keywords;

[0089] S504. Establish a text similarity analyzer based on Simhash and BERT to obtain historical cases and similar emergency plans similar to the on-site scenario according to the input text materials.

[0090] Further, the specific steps in S6 are as follows:

[0091] S601. For each action subject, package the general emergency plan, sub-emergency plans, historical cases similar to the on-site scenario, and similar emergency plans to form a decision support document;

[0092] S602. Send each decision support document to each task executor;

[0093] Return the information reception situation to form the statistical information of the emergency decision-making and command center. At the same time, prepare to receive the next round of information, update the decision support plan according to the information returned by each action subject, and prepare for the generation of the next round of decision support plans;

[0094] The above is a cyclic process. With the investigation of potential hazards and the development of accident rescue, the emergency decision-making center will continuously receive information from all parties, update the emergency decision support plan system, and maintain communication with all parties until the potential hazards are eliminated or the emergency rescue task is completely ended.

[0095] In a specific embodiment, taking the coal mine water disaster accident scenario as an example, establish a knowledge graph of mine water inrush accident cases and emergency plans, structure the unstructured text materials, and assign numbers to each node; establish models such as water inrush source discrimination, water inflow prediction, roof pressure prediction, shortest escape route, and nearest rescue route according to the general laws of coal mine water disaster occurrence and prevention, providing a basis for the selection of specific emergency disposal plans; according to various types of data monitored on-site, including structured data based on sensors, image and video data based on video sensors, inputs from various types of terminals such as computers and PDAs, and the discrimination of the model library, further classify the working environment into various different states such as safe, risky, and accident-occurring, and further subdivide into various different specific scenarios; establish a Petri net structure model of a process-based emergency plan template according to different working, risk, and accident scenarios. Taking the coal mine water disaster emergency disposal stage as an example, such as Figure 2As shown in the figure, where the transition T represents the behavioral action, the place P represents the specific action executor, and the identifier M that triggers the enabling of the transition T is the task signal during each emergency stage; the specific behavioral actions in the refined task nodes correspond to the behavioral action entity nodes in the constructed knowledge graph; slots are set for the variables in the specific behavioral actions, including information such as the specific executor of the rescue action, the affected people, the rescue supplies, the safe location, the accident location, the affected area, the disaster avoidance route, the disaster relief route, etc., which will be affected by the specific situation, and the list of specific responsible personnel corresponding to the actual situation of the unit; taking the on-site disposal measure action as an example, after setting the variable slots for the action "The dispatcher immediately notifies the people in the nearby threatened area to evacuate to a safe location after receiving the report", it becomes "The dispatcher (**) immediately notifies the people in the nearby threatened area (##) to evacuate to a safe location (*)"; if on-site monitoring data is accessed, the monitoring data is accessed into the model library, and after being judged by the model, each component of the emergency plan is integrated corresponding to different scenarios, and a list is constructed in the form of ID numbers to form a plan list; for example: Ln = (1102, 1105, 1207, 1302, 1303,...), where Ln represents the emergency plan list in a certain scenario. The numbers in the list correspond to the specific entities in the knowledge graph on the one hand and the nodes in the emergency plan Petri net on the other hand, so that the decision-making and command center can achieve the rapid retrieval of key information based on the knowledge graph and the rapid generation of a structured emergency plan based on the Petri net; the emergency decision-making and command center finds the executors of each action in the emergency disposal, sends the overall emergency plan and sub-emergency plans generated by the system, relevant historical cases, and plan reference materials to each action executor, and accepts the feedback from each action executor to realize the information interaction between each action executor and the emergency command center, and further generate the next emergency disposal plan until the risk is lifted or the emergency rescue operation ends.

[0096] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the key point of each embodiment is to illustrate the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, it is described relatively simply, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are only illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative work.

[0097] The foregoing description of the disclosed embodiments enables those skilled in the art to practice or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating emergency decision support solutions based on knowledge graph, characterized in that: include: S1. According to different risks and accident scenarios, collect existing accident investigation reports and emergency plans, carry out structured processing, and build an accident case-emergency plan knowledge map; S2. Establish a risk-accident scenario model library for specific issues of specific risks and accident scenarios; S3. Set up corresponding process-based emergency plan templates according to different risks and accident scenarios, define the tasks and actions of each type of subject, and set dynamic slots; S4. Obtain various monitoring data that affect production safety at the project site, integrate and analyze the monitoring data and user input, determine specific risks or accident scenarios, select the corresponding plan template, fill in the slot information based on the monitoring data and other resource data, and generate an emergency plan; S5. Establish an accident investigation report-emergency plan text library, obtain the corresponding scenario categories based on the fusion analysis of monitoring data and user input, and obtain historical cases and similar emergency plans similar to the on-site scenario through natural language processing and knowledge graph retrieval; S6. Split the generated emergency plan, historical cases and emergency plans similar to the on-site scenario into structured emergency plan segments of the action executors according to the task nodes. Package the complete emergency plan, structured emergency plan segments for each action executor, and historical cases and emergency plans similar to the on-site scenario and send them to the task executors to assist the direct executors in emergency response. At the same time, establish an information communication mechanism between the emergency decision-making command center and the action executors until the risk is eliminated and the emergency rescue operation is completed.

2. According to the method for generating an emergency decision support plan based on a knowledge graph according to claim 1, it is characterized in that: The specific steps for constructing the accident case-emergency plan knowledge graph in S1 are: S101. Analyze the emergency response needs of different risks and accident scenarios, collect the corresponding accident investigation reports and emergency plan text data, and unify the numbering, and standardize the deconstruction of various text data; S102, perform entity annotation on the text data, and use the BERT-CRF entity extraction model to perform entity extraction on the annotated text data to obtain an entity label sequence; S103, determining the relationship type between entities according to the entity label sequence, and obtaining a knowledge triple; S104, extracting keywords from the accident investigation report and emergency plan text data by using the TextRank algorithm to form a triple of a text number and a keyword and a triple of a text number and a text summary; S105, establishing a spatial network model for key locations in the target work environment to form spatial location triplets; S106. Store the obtained triples of text number and keyword, triples of text number and text summary, and triples of each spatial position into a graph database, and construct an accident case-emergency plan knowledge graph consisting of triples.

3. According to the method for generating an emergency decision support plan based on a knowledge graph according to claim 1, it is characterized in that: The specific steps for establishing the risk-accident scenario model library in S2 are: S201. Arrange possible accident scenarios and key monitoring indicators in the target working environment, and establish a corresponding relationship between the accident scenarios and key monitoring indicators; S202, setting an accident scenario discrimination model based on the monitoring data of the on-site sensors to achieve mapping from the on-site monitoring data to the accident scenario; S203, for areas that require video surveillance, set up a video surveillance system and write a video analysis program to identify accident scenarios and achieve mapping from work site scenes to accident scenarios; S204. Classify the accident scenarios of the target work environment, number the accident scenarios, organize the criteria for distinguishing the occurrence of different accident scenarios, correspond them with the omens, facts and emergency response entities in the accident case-emergency plan knowledge graph, and build a risk-accident scenario model library.

4. According to the method for generating an emergency decision support plan based on a knowledge graph according to claim 1, it is characterized in that: S3 specifies the tasks and actions of each type of subject and sets dynamic slots. The specific steps are as follows: S301. According to the standardized structure of emergency plans, the plan is divided into seven stages: information reporting, early warning, response initiation, emergency disposal, emergency support, response termination and post-disposal, and the task nodes to be completed in each stage and the signals for the start and end of the task are determined; S302, determine the specific actions that need to be completed at each task node, and specify the executor of the action; S303, establishing a Petri net structure model of a process-based emergency plan template; S304, refine the specific behavior actions in the task node, and correspond to the behavior action entity nodes in the accident case-emergency plan knowledge graph; S305, calling up the spatial network model in the target working environment from the knowledge graph, calculating the shortest path according to the positions of relevant personnel or units and the forward target position, and forming the analysis results of the disaster avoidance route and disaster relief route; S306. Dynamic slots are set for variables in specific behavior actions. The variables include: specific executors of rescue actions, disaster victims, rescue supplies, safe places, accident locations, affected areas, evacuation routes, and disaster relief routes. The names and specific contents of the dynamic slots correspond to the accident case-emergency plan knowledge graph.

5. According to the method for generating an emergency decision support plan based on a knowledge graph according to claim 1, it is characterized in that: The specific steps for generating an emergency plan in S4 are: S401. Access various monitoring data that may affect production safety at the project site, and set up an input terminal to receive information on factors that may cause impacts discovered by on-site workers, safety inspectors, and the emergency command center based on experience; S402: Perform fusion analysis on user input data, on-site monitoring data, and model library discrimination results, process them into rule form, search in the risk-accident scenario model library, match the accident scenarios that meet the requirements, and then retrieve the corresponding process-based emergency plan template according to the matched accident scenarios; S403, according to the slots to be filled in the plan, read the required information from the accident case-emergency plan knowledge graph, fill the slots, and generate complete plan information; S404, filling the generated complete plan information into the plan Petri network structure model corresponding to the scenario to generate a process-based emergency plan, which is called the general emergency plan; S405. Extract each action subject from the process-based emergency plan, and split the entire process-based emergency plan into emergency plans directly related to each action subject, which are called sub-emergency plans.

6. According to claim 1, a method for generating an emergency decision support plan based on a knowledge graph is characterized in that: The specific steps of obtaining historical cases and similar emergency plans similar to the on-site scenario in S5 are: S501. Organize the relevant accident investigation reports and emergency plan text data of specific accident scenarios, enter them into the unstructured database, and establish an accident investigation report-emergency plan text library; S502, searching similar historical cases and similar emergency plans based on the accident case-emergency plan knowledge graph according to the monitoring data and the information input by the user; S503, constructing a text recommendation model, taking each accident investigation report text in the accident investigation report-emergency plan text library as a sample for model training, taking keywords as features, training the text recommendation model, and calculating the similarity between texts through a graph collaborative filtering algorithm to achieve keyword-based accident investigation reports and corresponding emergency plan text recommendations; S504: Establish a text similarity analyzer based on Simhash and BERT to obtain historical cases and similar emergency plans similar to the on-site scenario based on the input text material.

7. The method for generating an emergency decision support plan based on a knowledge graph according to claim 1, characterized in that: The specific steps in S6 are: S601. For each action subject, package the general emergency plan, sub-emergency plans, historical cases similar to the on-site scenario, and similar emergency plans to form a decision-making support document; S602, sending each decision support file to each task executor respectively; S603: Return the information received to form statistical information of the emergency decision-making command center, and prepare to receive the next round of information. According to the information returned by each action subject, update the decision support plan and prepare for the generation of the next round of decision support plan.

Images