Intelligent agent-based fire safety inspection item generation method and system

Abstract

The invention discloses a fire safety check item generation method based on an intelligent agent. The method comprises the following steps: obtaining fire safety check item generation request information; generating a fire safety check item document outline through the intelligent agent, dividing different types of subtasks, and generating fire safety check items corresponding to the different types of subtasks; summarizing and structuring the fire safety check items corresponding to the different types of subtasks to obtain a standardized fire safety check item text; and calculating a text similarity between the standardized fire safety check item text and a known standard text, and outputting the standardized fire safety check item text meeting a preset condition. According to the method, the intelligent agent technology is introduced, automatic generation and intelligent management of fire safety inspection items are achieved, the inspection efficiency is improved, the risk of missing inspection is reduced, real-time monitoring and early warning are achieved, and data recording and analysis are enhanced, so that the overall fire safety management level is improved.

Description

Technical Field

[0001] The present invention relates to the field of fire safety inspection, and more specifically to a method and system for generating fire safety inspection items based on an intelligent agent. Background Art

[0002] With the acceleration of urbanization, the complexity of buildings and the increasing population density, fire safety has become an important part of social public safety. As a key link in preventing fire accidents, fire safety inspection is becoming increasingly important.

[0003] However, traditional fire safety inspection methods mainly rely on manual on-site inspections and record keeping, which is not only inefficient but also easily affected by the inspectors' expertise, experience and subjective judgment, resulting in unstable inspection results and possible omissions. Especially in large or complex building environments, manual inspections face greater challenges and are prone to miss some key safety hazards. In addition, manual inspections cannot provide real-time monitoring and early warning services, making it difficult to respond and handle fires or other emergencies in a timely manner, thus delaying emergency response time and increasing accident losses.

[0004] At the same time, current fire safety inspection methods still have problems with insufficient data recording and analysis. Traditional methods lack effective data recording and analysis mechanisms, making it difficult to form systematic safety assessments and improvement measures, making it difficult to continuously improve and optimize fire safety management.

[0005] In addition, regular manual inspections and subsequent data processing require a lot of manpower and material resources, resulting in high overall costs. They generally lack intelligent and automated support, making it difficult to adapt to the development trend of modern smart cities and smart buildings, limiting the efficiency and accuracy of fire safety inspections.

[0006] Therefore, how to design a fire safety inspection item generation method based on intelligent agent, realize the automatic generation and intelligent management of fire safety inspection items, and improve the efficiency and accuracy of fire safety management work is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0007] In view of this, the present invention provides a method for generating fire safety inspection items based on intelligent agents. By introducing intelligent agent technology, the automatic generation and intelligent management of fire safety inspection items are realized, aiming to improve inspection efficiency, reduce the risk of missed detection, realize real-time monitoring and early warning, strengthen data recording and analysis and other functions, thereby comprehensively improving the efficiency and accuracy of fire safety management work.

[0008] In order to achieve the above object, the present invention adopts the following technical solution:

[0009] In a first aspect, the present invention provides a method for generating fire safety inspection items based on an agent, comprising the following steps:

[0010] S1. Obtaining fire safety inspection item generation request information;

[0011] S2. Based on the fire safety inspection item generation request information, an agent is used to generate a document outline of the fire safety inspection item, and different types of subtasks are divided to generate fire safety inspection items corresponding to the different types of subtasks;

[0012] S3. Summarize and structure the fire safety inspection items corresponding to the different types of subtasks to obtain standardized fire safety inspection item texts;

[0013] S4. Calculate the text similarity between the standardized fire safety inspection item text and the known standard text, and output the standardized fire safety inspection item text that meets the preset conditions.

[0014] In said S1, the fire safety inspection item generation request information includes: user demand information, building basic information and regulatory compliance information;

[0015] User demand information includes specific content of inspection items, goals and objectives, applicable objects, inspection frequency and timing, inspection methods and tools, record and report requirements, exception handling procedures, improvement suggestions, and training and support;

[0016] Basic building information includes unit information, building property information, manager information, responsible person information, other relevant personnel information and historical data information;

[0017] Regulatory compliance information includes laws, local regulations, departmental regulations, local regulations, policy documents, mandatory national standards, non-mandatory national standards, industry standards and local standards.

[0018] Preferably, in S2, generating a document outline of fire safety inspection items by an intelligent agent includes:

[0019] The vertical large model is used to parse the fire safety inspection items to generate request information, perform semantic analysis and intent recognition, and obtain the prompt word content, document subject, outline template, outline keywords, outline style and partial outline content.

[0020] Preferably, in S2, dividing different types of subtasks and generating fire safety inspection items corresponding to different types of subtasks include:

[0021] Generate different types of subtasks by calling professional fire safety knowledge base and existing knowledge base;

[0022] Assign corresponding single agents to different types of subtasks, and combine the API interface and RAG tool to generate fire safety inspection items corresponding to different types of subtasks.

[0023] Preferably, the single agent construction includes:

[0024] Clarify the tasks and functional requirements of a single agent in the generation of fire safety inspection items, including environmental perception, task execution, learning and collaboration capabilities;

[0025] Select a suitable algorithm; the algorithm includes: rule system, machine learning or deep learning;

[0026] Training and optimizing a single agent using relevant data sets including building environment information, historical inspection records, and user feedback;

[0027] The performance of a single agent in the task of generating fire safety inspection items is improved through iterative optimization, and comprehensive testing is performed and integrated into the agent.

[0028] Preferably, the S2 further includes: identifying the fire hazard source of the building unit in combination with the environmental information of the building, and generating a fire risk assessment report based on the primary index and the secondary index, and supplementing the fire safety inspection item types;

[0029] Among them, the building's environmental information includes: temperature, humidity, smoke concentration and equipment status information.

[0030] Preferably, the S3 comprises:

[0031] Merge duplicate items in the checklist and remove redundant content;

[0032] Use XML file format to structure the generated fire safety inspection items, and clarify the name, type, precondition, operation steps and post-action of the inspection items;

[0033] The XML file is parsed through the BPMN 2.0 process engine to generate a visual inspection item flowchart.

[0034] Preferably, in said S4, the preset condition includes a text similarity threshold;

[0035] If the text similarity is greater than the text similarity threshold, the standardized fire safety inspection item text is directly output; otherwise, the vertical large model is called for multiple rounds of dialogue to generate standardized fire safety inspection item text that meets the requirements, and the text similarity is repeatedly calculated until the conditions are met.

[0036] Preferably, the vertical large model construction includes:

[0037] Integrate professional knowledge in the fire protection field, such as regulations, standards, manuals and cases, to build a knowledge base in the fire protection field;

[0038] Select a basic large model for NLP and pre-train it using a common dataset;

[0039] Using the fire protection domain knowledge base to fine-tune the basic large model;

[0040] The fine-tuned basic large model is used to define functional modules, conduct comprehensive tests and integrate into the intelligent agent.

[0041] In a second aspect, the present invention provides an agent-based fire safety inspection item generation system, comprising:

[0042] User request information acquisition module: used to obtain fire safety inspection item generation request information;

[0043] A fire safety inspection item generation module is used to generate request information based on the fire safety inspection items, generate a document outline of the fire safety inspection items through an intelligent agent, divide the items into different types of subtasks, and generate fire safety inspection items corresponding to different types of subtasks;

[0044] Summarization and structured processing module: used to summarize and structure the fire safety inspection items corresponding to the different types of subtasks to obtain standardized fire safety inspection item texts;

[0045] Text similarity calculation and output module: used to calculate the text similarity between the standardized fire safety inspection item text and the known standard text, and output the standardized fire safety inspection item text that meets the preset conditions.

[0046] It can be seen from the above technical solution that compared with the prior art, the technical solution of the present invention has the following advantages:

[0047] Beneficial effects:

[0048] 1. This method automatically generates fire safety inspection items by introducing intelligent agents, realizing the full process intelligence from request information input to standardized inspection item text output. It greatly reduces manual intervention and avoids the inefficiency and frequent errors caused by human factors. The intelligent agent can quickly generate a detailed inspection item document outline based on the input request information and refine it into different types of subtasks, thereby ensuring that each inspection content can be fully covered and accurate, significantly improving the overall efficiency and accuracy of the inspection work.

[0049] 2. The intelligent agent can not only analyze the specific needs of users, but also dynamically adjust the generated inspection items based on the actual conditions of the building (such as building characteristics, environmental information, etc.). This allows the generated fire safety inspection items to accurately match the needs of different scenarios, and can provide targeted customized services for commercial buildings, residential buildings, and special places, thereby greatly improving the practicality and effectiveness of the inspection.

[0050] 3. In the process of generating inspection items, different types of subtasks are summarized and structured, and finally standardized fire safety inspection item texts are output. This step not only ensures the orderliness and clarity of the inspection item content, but also facilitates subsequent data recording, analysis and utilization. The standardized text format helps to unify inspection standards, reduce misunderstandings and errors caused by different formats, and also greatly facilitates subsequent audits and evaluations.

[0051] 4. Before outputting the standardized fire safety inspection item text, this method will calculate the text similarity between the text and the known standard text, and verify it in combination with the preset conditions. This effectively ensures the quality and accuracy of the output text and avoids the risks caused by content deviation or errors. Through the text similarity verification mechanism, the system can promptly discover and correct possible problems, thereby providing users with a more reliable and high-quality fire safety inspection item generation service. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0053] Figure 1 A flow chart of a method for generating fire safety inspection items based on an agent provided in an embodiment of the present invention;

[0054] Figure 2 A schematic diagram of a process for generating fire safety inspection items corresponding to different types of subtasks provided in an embodiment of the present invention;

[0055] Figure 3 A schematic diagram of a process for outputting a text of standardized fire safety inspection items that meet conditions provided by an embodiment of the present invention;

[0056] Figure 4 A framework diagram of an agent-based fire safety inspection item generation system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] Embodiment 1;

[0059] like Figure 1 As shown, this embodiment provides a method for generating fire safety inspection items based on an intelligent agent, comprising the following steps:

[0060] S1. Obtaining fire safety inspection item generation request information;

[0061] S2. Based on the fire safety inspection item generation request information, an agent is used to generate a document outline of the fire safety inspection item, and different types of subtasks are divided to generate fire safety inspection items corresponding to the different types of subtasks;

[0062] S3. Summarize and structure the fire safety inspection items corresponding to the different types of subtasks to obtain standardized fire safety inspection item texts;

[0063] S4. Calculate the text similarity between the standardized fire safety inspection item text and the known standard text, and output the standardized fire safety inspection item text that meets the preset conditions.

[0064] This method introduces intelligent agent technology to achieve automatic generation and intelligent management of fire safety inspection items, improve inspection efficiency, reduce the risk of missed inspections, achieve real-time monitoring and early warning, strengthen data recording and analysis, and thus improve the overall fire safety management level.

[0065] The above steps and related technical features are further described in detail below:

[0066] In this embodiment S1, the fire safety inspection item generation request information includes: user demand information, building basic information and regulatory compliance information; wherein,

[0067] User demand information includes specific content of inspection items, goals and objectives, applicable objects, inspection frequency and timing, inspection methods and tools, record and report requirements, exception handling procedures, improvement suggestions, and training and support;

[0068] Basic building information includes unit information, building property information, manager information, responsible person information, other relevant personnel information and historical data information;

[0069] Regulatory compliance information includes laws, local regulations, departmental regulations, local regulations, policy documents, mandatory national standards, non-mandatory national standards, industry standards and local standards.

[0070] Specifically, user demand information not only includes the specific content, goals and objectives of the inspection items, and clarifies the specific direction of the inspection work; it also lists the applicable objects in detail to ensure that the inspection items are highly matched to different scenarios and needs; at the same time, it clarifies the inspection frequency and timing to reasonably arrange the inspection plan; the description of inspection methods and tools provides specific guidance for the implementation of inspections; the requirements for records and reports ensure the traceability and transparency of the inspection process and results; the planning of exception handling processes provides a rapid response mechanism for dealing with emergencies; in addition, it also considers the collection of improvement suggestions to promote the continuous optimization of inspection work; and the needs for training and support are aimed at improving the professional capabilities of inspectors.

[0071] The basic building information covers a wide range, from the basic information of the unit such as name, address, contact information, etc., to detailed information such as the ownership and historical evolution of the building, and then to the specific circumstances of the manager, responsible person and other relevant personnel, providing a solid background support for the formulation of inspection items; at the same time, the inclusion of historical data information, such as past inspection reports and accident records, provides a valuable data basis for risk assessment and determination of inspection priorities.

[0072] The emphasis on regulatory compliance information ensures that national and local fire safety laws, regulations and standards are strictly followed during the formulation and implementation of inspection items, thereby effectively avoiding legal risks and improving the compliance of fire safety management.

[0073] In this embodiment S2, the outline of the fire safety inspection document generated by the agent includes:

[0074] The vertical large model is used to parse the fire safety inspection items to generate request information, perform semantic analysis and intent recognition, and obtain the prompt word content, document subject, outline template, outline keywords, outline style and partial outline content.

[0075] Specifically, after the vertical large model performs semantic analysis and intent recognition, if the results are vague or ambiguous, the agent should be able to actively initiate multiple rounds of interaction to further clarify the requested information with the user by asking questions or providing more options. For example, when it is identified that the "specific content of the inspection item" is not clear enough, the agent can ask: "Can you describe the specific content you want to include in the inspection item?"

[0076] Considering that users may use the system multiple times to generate check items in different scenarios, the intelligent agent should be able to use contextual association analysis technology to identify and refer to the user's previous generation requests and results in order to better understand the context of the current request and generate an outline that is more in line with user habits and needs.

[0077] Based on the user's historical usage records and preferences, the agent can provide personalized document outline template recommendations. For example, if the user frequently generates fire safety inspection items for high-rise residential buildings, the agent will automatically recommend relevant residential fire safety inspection outline templates the next time it generates them.

[0078] Furthermore, different types of subtasks are divided to generate fire safety inspection items corresponding to different types of subtasks, including:

[0079] Generate different types of subtasks by calling professional fire safety knowledge base and existing knowledge base;

[0080] like Figure 2 As shown, a corresponding single agent is assigned to different types of subtasks, and the fire safety inspection items corresponding to different types of subtasks are generated by combining the API interface and RAG tool.

[0081] Specifically, when dividing subtasks, the agent should automatically prioritize the subtasks according to the importance and urgency of the inspection items. Subtasks involving life safety (such as fire passageway clearance inspection) should be given higher priority to ensure that these key inspection items are generated and executed first.

[0082] Furthermore, when combining the API interface and RAG (Retrieval Augmented Generation) tool, the intelligent agent should continuously optimize its query strategy and retrieval efficiency to ensure that relevant information can be retrieved from the knowledge base quickly and accurately and effectively integrated into the generated inspection items.

[0083] In this embodiment, the construction of a single agent includes:

[0084] Clarify the tasks and functional requirements of a single agent in the generation of fire safety inspection items, including environmental perception, task execution, learning and collaboration capabilities;

[0085] Select a suitable algorithm; the algorithm includes: rule system, machine learning or deep learning;

[0086] Training and optimizing a single agent using relevant data sets including building environment information, historical inspection records, and user feedback;

[0087] The performance of a single agent in the task of generating fire safety inspection items is improved through iterative optimization, and comprehensive testing is performed and integrated into the agent.

[0088] After generating the check items for each subtask, the agent should automatically perform compliance checks to ensure that all check items comply with national or local fire safety regulations and standards. If non-compliant items are found, they should be corrected immediately or the user should be prompted to confirm and modify them.

[0089] Furthermore, it also includes: combining the environmental information of the building, identifying the fire hazard sources of the building, and generating a fire risk assessment report based on the primary and secondary indicators, and supplementing the types of fire safety inspection items;

[0090] Among them, the building's environmental information includes: temperature, humidity, smoke concentration and equipment status information.

[0091] Specifically, the four first-level indicators include: Building characteristics: evaluate the impact of the inherent physical characteristics of the building itself on the fire risk, such as the building's structure, materials, and purpose. Building fire protection: evaluate the fire prevention measures taken during the design and construction of the building, such as fire zoning, fire resistance level, and fire protection facility configuration. Fire management: evaluate the building management's capabilities and measures in fire safety management, such as management systems, emergency plans, and personnel training. Firefighting capabilities: evaluate the firefighting capabilities of fire brigades and facilities in the event of a fire, including response time, fire extinguishing equipment, and evacuation routes.

[0092] Based on the four first-level indicators, 31 second-level indicators are further refined to assess the fire risk of a single building in a more specific and comprehensive manner:

[0093] Building characteristic factors: years of construction, building area, building height, surrounding environment (such as adjacent buildings, road conditions, etc.);

[0094] Building fire prevention factors: fire load (such as the type and quantity of combustibles), storage of flammable and explosive dangerous goods, personnel load (the maximum number of people that can be accommodated in the building), combustion performance of interior decoration materials, safety of electrical lines (such as aging and overload), fire resistance level (fire resistance time of building components), fire separation distance (safe distance between buildings);

[0095] Fire management factors: safe evacuation facilities (such as stairs, exits, signs), emergency lighting systems, configuration and effectiveness of fire extinguishers, completeness of safety signs, fire separation (such as firewalls, fire doors, etc.), rationality of fire zoning, coverage and reliability of automatic fire alarm systems, maintenance records and implementation of fire protection facilities, implementation of fire inspection systems, formulation and implementation of fire safety management systems, professional level and training of management personnel, popularity of fire protection publicity and education, and management systems and implementation of hot work operations;

[0096] Factors affecting fire fighting capabilities: reliability and water volume of the fire water supply system, response time of the fire brigade, reliability and stability of the fire power supply, effectiveness of the automatic fire extinguishing system (such as sprinkler system, gas fire extinguishing system), design and operation of the smoke exhaust system, completeness and operability of the emergency plan, and comprehensive assessment of fire fighting conditions (such as unobstructed passages, complete rescue equipment, etc.).

[0097] The first- and second-level indicators here together constitute a comprehensive framework for assessing the fire risks of individual buildings, ensuring the accurate identification and effective prevention and control of fire risks.

[0098] In this embodiment S3, it includes:

[0099] Merge duplicate items in the checklist and remove redundant content;

[0100] After generating a large number of fire safety inspection items, it is inevitable that there will be duplication or overlap in content. In order to improve the quality and readability of the inspection items, this step first removes duplication and redundant content from these inspection items. In this step, the meaning and purpose of each inspection content are analyzed, and duplicate or highly similar items are identified, and they are merged into a more comprehensive and comprehensive inspection item. This not only reduces the workload of inspectors, but also avoids the waste of resources caused by repeated inspections.

[0101] Furthermore, after completing the removal of duplicate and redundant content,

[0102] Use XML file format to structure the generated fire safety inspection items, and clarify the name, type, precondition, operation steps and post-action of the inspection items;

[0103] The XML file contains a macro view of the entire process, namely the main process and its nested sub-processes. The main process depicts the main trajectory of the business process from the beginning to the end, while the sub-process is refined into the various sub-stages within the main process. Each sub-process consists of a series of interrelated activities (check items). The "activities" or "check items" here are actually specific tasks or operation units to be performed in the process. Their precise definition and arrangement directly affect the effectiveness and efficiency of the entire process.

[0104] The activities (check items) in the XML file have a rigorous structure and rich attributes, which helps to clearly define the key elements of each task. The specific example is as follows:

[0105] Name: The name of the activity (check item) is a concise and clear description of the task, which is easy to understand and identify, for example: "Document Review", "Order Confirmation", etc.

[0106] Type: The activity (check item) type is used to distinguish tasks of different natures, such as manual tasks (performed manually), automatic tasks (automatically performed by the system), script tasks (performed based on program code), etc.

[0107] Precondition: Preconditions specify the prerequisites for the execution of an activity, usually based on the state of certain variables or business rules, and can be expressed in Boolean expressions or other logical forms. This means that the activity can only be started if the "approved by the manager" condition is met.

[0108] Operation steps: Operation steps list in detail the specific sequence of operations required to perform the activity, which may include user operation instructions, system processing logic, or interaction processes with other systems. In the actual XML file, this part of the content may not be directly reflected as a single element, but will be indirectly reflected in the extended element that describes the activity behavior.

[0109] Post-action: The post-action after the activity is executed usually refers to the processing and response to the execution result of the activity, such as updating the database status, triggering the next activity, sending a notification email, etc. Similarly, post-actions can also be expressed through extended attributes or associated elements of XML tags.

[0110] This structured processing not only facilitates subsequent data analysis and processing, but also makes the inspection items more standardized, reducing ambiguity and misunderstandings during the implementation process.

[0111] Furthermore, the XML file is parsed through the BPMN 2.0 process engine to generate a visual inspection item flowchart.

[0112] After the XML file is generated, it will be parsed by the specially designed BPMN 2.0 process engine, which can interpret the structure and logic in the XML file and convert the abstract process model into an actual executable process instance. During this process, users can edit and optimize the activities (check items) in the main process and sub-process according to their needs.

[0113] Under the BPMN 2.0 specification, the XML file generated by the inspection item vertical model fully reflects the meticulousness and complexity of business process design. In terms of file structure, it not only includes the main process at the macro level, but also describes the embedded sub-processes in detail. The main process consists of a series of key components, which can be represented by mathematical set theory and tree diagram models, including:

[0114] 1) Activity node: denoted as T∈T, where T is the set of all activity nodes. Each node represents a specific business operation and has a unique ID identifier, name and type attributes, which can be described by the XML tag [bpmn:task](bpmn:task).

[0115] 2) Event: denoted as E∈E, where E is the set of all events. Events trigger the transition of process states, such as start events, end events, intermediate capture events, etc., which can all be encapsulated using the [bpmn:event](bpmn:event) type tag.

[0116] 3) Gateway: denoted as G∈G, where G contains a set of gateways of different types, such as parallel gateways, inclusive gateways, etc. They play a decision-making role in the process flow and are represented by the [bpmn:gateway](bpmn:gateway) label.

[0117] 4) Data object: denoted as D∈D, where D represents a collection of data objects. Data objects are used to transfer, store and reference business data in the process and are defined using the [bpmn:dataObject](bpmn:dataObject) tag.

[0118] In addition, there is a clear chain of relationships between these components, which can be expressed in the form of a directed graph, such as a sequence flow connecting activity nodes to show the execution order of the business process; and the association relationship may be reflected through a message flow or other types of association lines.

[0119] As for sub-processes, they can be regarded as refinements and extensions of the main process, which are aggregated by a series of interrelated activity nodes (i.e. detection items). Each "activity" or "detection item" is essentially an abstraction of the task or operation unit to be executed in the process, and its detailed information includes:

[0120] 1) Name: used to identify and describe the core content of the activity;

[0121] 2) Type: Indicates the nature of the activity, such as user task, service task, script task, etc.

[0122] 3) Preconditions: The prerequisites for the execution of an activity, which can be defined by constraint expressions or logical judgments;

[0123] 4) Operation Steps: The specific steps for implementing the activity, which can be further divided into subtasks;

[0124] 5) Post-actions: actions that need to be performed or states that need to be changed after the activity ends.

[0125] The above flowchart not only shows the execution order and dependencies of the inspection items, but also clearly presents the key nodes and decision points in the inspection process. Inspectors can intuitively understand the entire inspection process through this flowchart, thereby improving execution efficiency and accuracy.

[0126] In this embodiment S4, Figure 3 As shown, the preset conditions include text similarity threshold;

[0127] If the text similarity is greater than the text similarity threshold, the standardized fire safety inspection item text is directly output; otherwise, the vertical large model is called for multiple rounds of dialogue to generate standardized fire safety inspection item text that meets the requirements, and the text similarity is repeatedly calculated until the conditions are met.

[0128] Specific:

[0129] For a given text A and text B, a variety of methods are used to calculate their similarity, such as cosine similarity, Jaccard similarity, edit distance, etc. In this embodiment, cosine similarity is used as an example for description.

[0130] First, based on the Bag of Words representation method, text A and text B are converted into vectors. The text is segmented and a unique ID is assigned to each word. Then, a vector is created for each text, where the length of the vector is equal to the number of words in the vocabulary, and each element of the vector represents the number of times the corresponding word appears in the text. For text A and text B, the following bag of words vectors are obtained:

[0131] A = [1,1,1,0,0,1,0,0] (catering = 1, kitchen = 1, venue = 1, hotel = 0, restaurant = 0, gas = 1, concentration = 0, security = 0, fire protection = 0, safety = 0)

[0132] B = [0,0,0,1,1,1,1,0] (catering = 0, kitchen = 0, location = 0, hotel = 1, restaurant = 1, gas = 1, concentration = 1, security = 0, fire protection = 0, safety = 0)

[0133] Furthermore, the cosine similarity of the two vectors is calculated. The formula for cosine similarity is:

[0134] similarity=cos(θ)=(A·B) / (||A||||B||)

[0135] Where A·B is the dot product of vector A and vector B, and ||A|| and ||B|| are the Euclidean norms of vector A and vector B, respectively.

[0136] Among them, A·B=(1*0+1*0+1*0+0*1+0*1+1*1+0*1+0*0)=1, ||A||=sqrt(1^2+1^2+1^2+0^2+0^2 +1^2+0^2+0^2)=sqrt(4), ||B||=sqrt(0^2+0^2+0^2+1^2+1^2+1^2+1^2+0^2)=sqrt(4).

[0137] Therefore, text similarity = (1) / (sqrt(4)*sqrt(4)) = 1 / 4 = 0.25.

[0138] Since 0.25<95% text similarity threshold, the output results are hidden, and the vertical large model is called for multiple rounds of dialogue to generate standardized fire safety inspection item texts that meet the requirements, and the text similarity is repeatedly calculated until the conditions are met.

[0139] The vertical large model construction includes:

[0140] Integrate professional knowledge in the fire protection field, such as regulations, standards, manuals and cases, to build a knowledge base in the fire protection field;

[0141] Select a basic large model for NLP and use a general data set for pre-training; in this embodiment, the general data set includes: rich industry background knowledge and practice rules. Such as legal and regulatory texts, industry consensus standards, various operation manuals, and safety management system documents, etc., which can be expressed as K = {k1, k2, ..., kn}, where each ki corresponds to a specific knowledge item, such as a regulatory document or an operating instruction document.

[0142] It also includes manually annotated positive sample data, which plays the role of "textbook" and is the key for the model to learn how to correctly understand and generate the target output. It includes examples with clear labels or corresponding structured representations, such as safety guidance documents for high-altitude operations and their XML representation files that follow the Business Process Modeling Notation (BPMN) 2.0 standard. These paired data are in the form of (Fj, Pj), where Fj is the original text guidance book and Pj is the corresponding standard XML file.

[0143] The fire protection domain knowledge base is used to fine-tune the basic large model; the fire protection domain knowledge base includes: a set of XML files generated by the vertical large model, where the data is a new sample generated based on the existing model, which is used in the fine-tuning stage of the model. The model automatically generates a series of XML files that comply with the BPMN2.0 specification, and also includes the original documents corresponding to these XML files and the optimized XML file versions after manual review and correction. Specifically, this set of data triplets can be written as (O1, R1, E1), where O1 is the XML file generated by the model, R1 is the corresponding real-world original document, and E1 is a more accurate XML file obtained after manual review of O1.

[0144] The fine-tuned basic large model is used to define functional modules, conduct comprehensive tests and integrate into the intelligent agent.

[0145] The agent-based fire safety inspection item generation method in this embodiment realizes the automatic generation and intelligent management of fire safety inspection items by obtaining user request information, intelligently generating document outlines and subtasks, summarizing and structuring inspection items, and calculating text similarity to output standardized text. This method not only improves inspection efficiency and reduces human errors, but also realizes real-time monitoring and early warning of the fire safety status of buildings, and strengthens data recording and analysis, thereby significantly improving the overall fire safety management level.

[0146] Embodiment 2;

[0147] like Figure 4 As shown, this embodiment provides a fire safety inspection item generation system based on an intelligent agent, comprising:

[0148] User request information acquisition module: used to obtain fire safety inspection item generation request information;

[0149] A fire safety inspection item generation module is used to generate request information based on the fire safety inspection items, generate a document outline of the fire safety inspection items through an intelligent agent, divide the items into different types of subtasks, and generate fire safety inspection items corresponding to different types of subtasks;

[0150] Summarization and structured processing module: used to summarize and structure the fire safety inspection items corresponding to the different types of subtasks to obtain standardized fire safety inspection item texts;

[0151] Text similarity calculation and output module: used to calculate the text similarity between the standardized fire safety inspection item text and the known standard text, and output the standardized fire safety inspection item text that meets the preset conditions.

[0152] The agent-based fire safety inspection item generation system in this embodiment significantly improves inspection efficiency and reduces human errors by automatically generating inspection items, realizes real-time monitoring and early warning of the fire safety status of buildings, and ensures the comprehensiveness and pertinence of inspections. In addition, the system can continuously update inspection items according to the latest regulations and standards, provide data-driven decision support, reduce long-term operating costs, and greatly improve the level of fire safety management through easy integration, user-friendly interface and the ability to flexibly respond to various challenges.

[0153] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0154] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one 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. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating fire safety inspection items based on an intelligent agent, characterized in that: The following steps are involved: S1. Obtaining fire safety inspection item generation request information; S2. Based on the fire safety inspection item generation request information, an agent is used to generate a document outline of the fire safety inspection item, and different types of subtasks are divided to generate fire safety inspection items corresponding to the different types of subtasks; S3. Summarize and structure the fire safety inspection items corresponding to the different types of subtasks to obtain standardized fire safety inspection item texts; S4. Calculate the text similarity between the standardized fire safety inspection item text and the known standard text, and output the standardized fire safety inspection item text that meets the preset conditions.

2. The method for generating fire safety inspection items based on an agent according to claim 1, characterized in that: In S1, the fire safety inspection item generation request information includes: user demand information, building basic information and regulatory compliance information; wherein, User demand information includes specific content of inspection items, goals and objectives, applicable objects, inspection frequency and timing, inspection methods and tools, record and report requirements, exception handling procedures, improvement suggestions, and training and support; Basic building information includes unit information, building property information, manager information, responsible person information, other relevant personnel information and historical data information; Regulatory compliance information includes laws, local regulations, departmental regulations, local regulations, policy documents, mandatory national standards, non-mandatory national standards, industry standards and local standards.

3. The agent-based method for generating fire safety inspection items according to claim 1, characterized in that: In S2, the outline of the fire safety inspection document generated by the agent includes: The vertical large model is used to parse the fire safety inspection items to generate request information, perform semantic analysis and intent recognition, and obtain the prompt word content, document subject, outline template, outline keywords, outline style and partial outline content.

4. The agent-based method for generating fire safety inspection items according to claim 1, characterized in that: In S2, different types of subtasks are divided, and fire safety inspection items corresponding to different types of subtasks are generated, including: Generate different types of subtasks by calling professional fire safety knowledge base and existing knowledge base; Assign corresponding single agents to different types of subtasks, and combine the API interface and RAG tool to generate fire safety inspection items corresponding to different types of subtasks.

5. The agent-based method for generating fire safety inspection items according to claim 4, characterized in that: The single agent construction includes: Clarify the tasks and functional requirements of a single agent in the generation of fire safety inspection items, including environmental perception, task execution, learning and collaboration capabilities; Select a suitable algorithm; the algorithm includes: rule system, machine learning or deep learning; Training and optimizing a single agent using relevant data sets including building environment information, historical inspection records, and user feedback; The performance of a single agent in the task of generating fire safety inspection items is improved through iterative optimization, and comprehensive testing is performed and integrated into the agent.

6. The agent-based method for generating fire safety inspection items according to claim 1, characterized in that: The S2 also includes: combining the environmental information of the building, identifying the fire hazard source of the building, and generating a fire risk assessment report based on the primary index and the secondary index, and supplementing the fire safety inspection item types; Among them, the building's environmental information includes: temperature, humidity, smoke concentration and equipment status information.

7. The agent-based method for generating fire safety inspection items according to claim 1, characterized in that: The S3 includes: Merge duplicate items in the checklist and remove redundant content; Use XML file format to structure the generated fire safety inspection items, and clarify the name, type, precondition, operation steps and post-action of the inspection items; The XML file is parsed through the BPMN 2.0 process engine to generate a visual inspection item flowchart.

8. The agent-based method for generating fire safety inspection items according to claim 7, characterized in that: In said S4, the preset condition includes a text similarity threshold; If the text similarity is greater than the text similarity threshold, the standardized fire safety inspection item text is directly output; otherwise, the vertical large model is called for multiple rounds of dialogue to generate standardized fire safety inspection item text that meets the requirements, and the text similarity is repeatedly calculated until the conditions are met.

9. The agent-based method for generating fire safety inspection items according to claim 3 or 8, characterized in that: The vertical large model construction includes: Integrate professional knowledge in the fire protection field, such as regulations, standards, manuals and cases, to build a knowledge base in the fire protection field; Select a basic large model for NLP and pre-train it using a common dataset; Using the fire protection domain knowledge base to fine-tune the basic large model; The fine-tuned basic large model is used to define functional modules, conduct comprehensive tests and integrate into the intelligent agent.

10. A fire safety inspection item generation system based on an intelligent agent, characterized in that: include: User request information acquisition module: used to obtain fire safety inspection item generation request information; A fire safety inspection item generation module is used to generate request information based on the fire safety inspection items, generate a document outline of the fire safety inspection items through an intelligent agent, divide the items into different types of subtasks, and generate fire safety inspection items corresponding to different types of subtasks; Summarization and structured processing module: used to summarize and structure the fire safety inspection items corresponding to the different types of subtasks to obtain standardized fire safety inspection item texts; Text similarity calculation and output module: used to calculate the text similarity between the standardized fire safety inspection item text and the known standard text, and output the standardized fire safety inspection item text that meets the preset conditions.

Images