Industrial safety production intelligent supervision platform based on intention recognition and field large model

Abstract

The invention discloses an intelligent industrial safety production supervision platform based on intention recognition and a large field model. The intelligent industrial safety production supervision platform comprises a multi-source multi-modal data fusion module, a multi-modal video recognition module, an intelligent algorithm arrangement module, an intention recognition module, a large field model integration module and an emergency response module. The intention recognition technology is combined with an industrial field large model, comprehensive multi-modal data fusion and intelligent decision support are achieved in industrial safety production for the first time, and the method is not only suitable for high-risk industries such as electric power and petrochemical engineering, but also can be applied to safety production management in the fields such as machine manufacturing and food processing; besides, the system platform adopts a modular design and a low-code development environment, a user can flexibly configure a monitoring algorithm and an emergency scheme according to own requirements, and the system is suitable for various industrial environments and production scenes.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent manufacturing and industrial safety management, and specifically relates to an intelligent supervision platform for industrial production safety that integrates safety, fire protection, and electricity based on intent recognition and large domain models. Background Art

[0002] Although businesses and organizations currently accumulate a wealth of knowledge and data related to accident handling, accident levels, and production information during their daily production processes, this knowledge and data is often not effectively utilized and its value is not fully realized. Routine work and data reporting is also not automated, requiring significant manpower. Therefore, a generative large-scale model for the vertical field of industrial safety production is needed to extract the value of existing knowledge and data. This model can then be combined with video algorithm results to quickly provide accident handling recommendations, enable rapid question-and-answer services, or integrate with intent recognition algorithms to quickly generate reports.

[0003] With the continuous upgrading of industrial systems and technological innovation, the safe production environment is becoming increasingly complex. Traditional safety management systems are usually based on fixed processes and preset rules, lacking sufficient flexibility and adaptability, and are unable to keep up with the pace of changes and development of industrial systems; especially in the context of the application of emerging technologies and the emergence of new risk factors, the adaptability of traditional systems is particularly insufficient.

[0004] Although some industrial safety production platforms have introduced automated monitoring and data analysis functions, such as the Chinese patent applications with publication numbers CN119740989A and CN119274142A, these patented technologies still have the following technical limitations: 1. Imperfect multimodal data fusion: Existing platforms cannot effectively correlate spatiotemporal data when processing multi-source data (such as sensors, videos, and equipment operation data), making it difficult to obtain global security situation awareness.

[0005] 2. Untimely emergency response: Most platforms rely on preset fixed rules or single algorithms and lack flexible emergency response mechanisms. This makes it difficult to quickly adjust emergency plans based on actual scenarios, resulting in delayed response times.

[0006] 3. Lack of intelligent decision support: Existing systems fail to integrate expert knowledge and historical data from the industrial field with current production scenarios, and are unable to provide workers with fast and accurate intelligent tools, resulting in weak decision support capabilities.

[0007] 4. Poor model versatility: The algorithm models used in many systems are not optimized for the vertical field of industrial safety production, resulting in poor performance in specific application scenarios and a lack of industry specificity and applicability. Summary of the Invention

[0008] To solve the intelligentization and data fusion problems in industrial safety production supervision, the application provides an integrated safety and emergency management and power industrial safety production intelligent supervision platform based on intention recognition and domain large model, aiming to improve the intelligent supervision ability of industrial safety production and optimize emergency management and accident prevention.

[0009] An integrated safety and emergency management and power industrial safety production intelligent supervision platform based on intention recognition and domain large model, comprising: A multi-source multi-modal data fusion module acquires data of different modalities collected from different sources in the industrial production site, and pre-processes and fuses the data; A multi-modal video recognition module automatically detects abnormal conditions in the video using computer vision and deep learning algorithms, and generates corresponding disposal suggestions; An intelligent algorithm arrangement module uses low-code algorithm arrangement technology to allow users to configure various intelligent algorithms on a graphical interface; An intention recognition module is used to analyze the behavior intention of the operator, identify the potential risk of production operation and give early warning of the risk; A domain large model integration module provides intelligent decision support and rapid knowledge service for users by integrating pre-trained domain large models; An emergency response module generates emergency plans through intention recognition and domain large models, and responds quickly to production accidents.

[0010] Further, the multi-source multi-modal data fusion module first acquires spatio-temporal data collected from multiple sources including sensors, video monitoring devices and production equipment in the industrial production site, which contains multiple modalities such as temperature, humidity, equipment operating status and video images; then the data is cleaned, denoised and standardized to ensure the integrity and consistency of the data, and then a spatio-temporal correlation analysis algorithm is used to fuse the data from different sources, and the fused data is displayed to the management personnel through a visual interface. The platform uses multi-modal data fusion technology, which can effectively integrate data from different data sources and realize all-round situation awareness of the industrial site; this technology improves the monitoring accuracy and response speed of the system, and is suitable for complex industrial environments.

[0011] Further, the multi-modal video recognition module analyzes the surveillance video data in multiple dimensions by introducing CNN (Convolutional Neural Network) and ViT (Visual Transformer Model), and identifies abnormal situations in the video. When an abnormal situation is detected, the module generates an alarm prompt based on the analysis results and further confirms it in combination with other data sources to ensure the accuracy of the alarm. Through multi-modal video recognition technology, the platform realizes real-time analysis and processing of video data in the production site, provides highly accurate scene recognition and monitoring in combination with other data sources; this technology greatly improves the intelligent monitoring capability of the system, and is particularly suitable for high-risk production scenarios.

[0012] Further, the intelligent algorithm arrangement module uses low-code algorithm arrangement technology to enable users to select different time series analysis algorithms through the graphical interface of the platform and drag and drop them into the flowchart, customize the logical order and trigger conditions of these algorithms according to specific needs, form algorithm arrangement of production monitoring processes, and then quickly deploy the arranged algorithms to the industrial production site and adjust and update the algorithm logic and parameters according to actual conditions. The platform has a built-in low-code environment that allows users to perform algorithm arrangement and management of complex time series data through a simple graphical interface; this technology provides the system with high flexibility and scalability, and users can customize safety monitoring algorithms and emergency plans according to actual needs.

[0013] Further, the intent recognition module obtains the action trajectory of the operator and the device operation log, extracts the behavior features from them, and identifies the operator's intent through a deep learning algorithm model. When a high-risk intent is identified, a warning is triggered, and intelligent safety suggestions and emergency handling plans are provided for the operator by analyzing historical operation data and real-time monitoring information in combination with the prediction results of the domain large model. Through intent recognition technology, the system can automatically identify the behavior intent of the operator and the operation state of the device, detect potential risky operations in real time and provide corresponding warnings and suggestions. This technology is an important part of the platform's intelligent emergency response.

[0014] Further, the intent recognition module uses a rule-based intent recognition method to set rules in combination with fixed patterns in historical data to achieve automated intent detection.

[0015] Further, the domain large model integration module integrates an industrial safety production domain large model that learns knowledge including industry experience, accident patterns, and fault types from industry experts and historical production data through pre-training; in daily production monitoring, the domain large model makes analysis and prediction based on current production data to provide intelligent decision support and domain knowledge question and answer services for operators and managers, automatically generates a safety production report, and continuously learns and updates to adapt to the specific needs of the production site and changing industrial environments. The introduction of the domain large model provides highly intelligent decision support functions for the platform, and the system can combine industry-specific historical data and industry knowledge to provide accident prediction, quick knowledge question and answer, emergency handling suggestions, and other services for operators and managers.

[0016] Further, the domain large model integration module uses knowledge graph technology to build an industry-specific knowledge base, providing precise decision support and question and answer systems for safety production through structured representation of expert knowledge and historical events.

[0017] Further, the emergency response module detects an accident that occurs in the first time using intent recognition technology based on real-time monitoring of multi-modal data, and then automatically generates an emergency plan based on the accident type and field data in combination with the domain large model, including operation steps, resource scheduling, and personnel arrangement. The plan content, including operation steps, resource scheduling, and personnel arrangement, is intuitively displayed on the platform interface, and an accident handling report is automatically generated, recording the accident cause, handling process, and subsequent rectification measures, and the report is archived.

[0018] Further, the emergency response module is based on reinforcement learning technology, and the model is trained by simulating different accident scenarios to automatically generate an emergency handling scheme.

[0019] The present application solves the limitations of existing industrial safety production supervision systems through the combination of multi-source data fusion, intent recognition, domain large model, and other technologies, specifically including: 1. Efficient fusion and association of multi-source data: The present application constructs an intelligent platform that can process and associate multi-source spatio-temporal data, enabling global perception and real-time monitoring of the production site.

[0020] 2. Intelligent emergency response and accident prevention: The present application uses intent recognition technology to quickly identify potential risks and abnormal operations in the production scene, providing emergency disposal suggestions for workers and improving accident response speed.

[0021] 3. Intelligent application of domain large model: The present application introduces a large model in the vertical field of industrial safety production, combines industry knowledge and field expert experience, and provides intelligent tools (such as knowledge question and answer, rapid generation of safety report, accident prediction, etc.) for different users (such as staff and administrators), thereby enhancing the practicability and accuracy of the system.

[0022] The present application combines intent recognition technology with industrial field large model, and for the first time realizes comprehensive multi-modal data fusion and intelligent decision support in industrial safety production, which has high foresight and innovation; the platform system adopts modular design and low-code development environment, and users can flexibly configure monitoring algorithms and emergency plans according to their own needs, which is suitable for various industrial environments and production scenes. Therefore, the present application has the following beneficial technical effects: 1. Global perception and intelligent monitoring: The present application realizes global perception of the production site through multi-source data fusion and multi-modal video recognition technology, automatically identifies abnormal behavior and potential risks, and provides real-time monitoring and intelligent response plan.

[0023] 2. Precise decision support: The present application platform combines field large model and intent recognition technology to provide highly accurate intelligent decision support for operators and managers, thereby improving the efficiency and accuracy of safety management.

[0024] 3. Flexible scalability and ease of use: The present application uses low-code algorithm arrangement technology to enable users to quickly adjust system configuration according to actual needs, thereby greatly enhancing the flexibility and scalability of the platform.

[0025] 4. Emergency response speed improvement: The present application uses intent recognition and intelligent algorithms to quickly generate emergency response plans when production accidents occur, thereby significantly improving emergency response speed and reducing accident risk.

[0026] 5. Continuous optimization of domain large model: The domain large model in the present application platform can adapt to changes in different industrial scenes through continuous learning and optimization, and provide continuously updated safety management tools and decision support. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 FIG. 1 is a schematic diagram of the architecture of the present application safety and power integration industrial safety production intelligent supervision platform.

[0028] Figure 2 FIG. 2 is a schematic diagram of the implementation architecture of the intent recognition module in the present application platform. DETAILED DESCRIPTION

[0029] In order to more specifically describe the present application, the technical solutions of the present application are described in detail below in combination with the drawings and specific embodiments.

[0030] The present application integrates multi-source data fusion, intent recognition, domain large model, and multi-modal video recognition technologies to build an intelligent industrial safety production supervision platform, as shown in FIG. 1, which can be applied to various industrial production scenes, especially in high-risk production environments, to improve safety, emergency response capability, and intelligent decision-making level during production. The specific implementation includes several parts: Figure 1 (1) Data access and processing.

[0031] Multi-modal technology is an important research direction in machine learning and computer vision, aiming to integrate data from different types (such as text, images, videos, sensors, etc.) to improve the accuracy of data analysis and decision-making. The present application uses multi-modal data fusion technology to uniformly process multi-source heterogeneous data in industrial production, ensuring that the system can fully perceive the on-site environment.

[0032] The multi-source multi-modal data fusion module in the present application is responsible for cleaning, formatting, and associating the spatio-temporal data from different data sources, thereby achieving comprehensive monitoring of the industrial site. The specific steps are as follows: Data collection: Real-time data is collected from sensors, video monitoring devices, production equipment, etc. in the industrial production site to form a unified spatio-temporal data stream. Data types include temperature, humidity, equipment operating status, video images, etc.

[0033] Data preprocessing: Different types of data are cleaned, denoised, and standardized to ensure data integrity and consistency.

[0034] Data fusion: Using spatio-temporal correlation analysis algorithms, the relationships between data in time and space dimensions are analyzed, and different sources of data (industrial equipment operation data, environmental data, personnel operation data, etc.) are spatio-temporally correlated and fused. The system automatically identifies the implicit relationships between different data sources, forms a complete safety production situation map, and discovers potential safety hazards to support subsequent intelligent analysis and decision-making. For example, the system can analyze temperature sensor and equipment operating status data, combined with the behavior of operators in video monitoring, to identify and warn of safety hazards in real time. Distributed data processing architecture can be considered to use cloud computing for real-time processing and analysis of multi-source data. This architecture can greatly improve the scalability and data processing speed of the platform, and reduce the load of local devices. Under the distributed architecture, the data processing process is more flexible and adaptable, suitable for larger-scale industrial scenes.

[0035] Real-time monitoring and feedback: The fused data is displayed to management personnel through a visual interface and triggers an alarm system in real time when an anomaly occurs, helping to quickly locate the problem area.

[0036] ​(2) Multi-modal video recognition and intelligent monitoring.

[0037] Multi-modal video recognition technology combines video, audio, sensor data and other information sources to achieve comprehensive monitoring of production sites. Through multi-modal fusion, the system can more accurately identify abnormal behavior and emergencies, improving the automation and intelligence level of safety production.

[0038] The multi-modal video recognition module in the present invention uses multi-modal video recognition algorithms to analyze and process video monitoring data in industrial sites in real time. The specific implementation steps are as follows: Video data acquisition and preprocessing: Through various video devices such as high-definition cameras and infrared cameras, real-time monitoring video data of production sites is collected. After denoising and formatting, the video data enters the multi-modal video recognition module.

[0039] Multi-modal analysis: Combining computer vision and deep learning algorithms, the system monitors the equipment status and personnel behavior of production sites in real time, automatically identifies abnormal situations in the video (such as personnel violation, equipment failure, fire, etc.), and performs multi-dimensional analysis of video images through convolutional neural networks and visual Transformer models.

[0040] Abnormal event recognition and alarm: When the system detects an abnormal event, it generates an alarm prompt and disposal suggestion based on the analysis results, and further confirms the accuracy of the alarm by combining other data sources (such as sensor data, equipment running status, etc.).

[0041] (3) Complex time series intelligent algorithm arrangement based on low code.

[0042] Low code (Low-Code) is a software development method that allows users to design, arrange business logic and applications through a visual interface without writing a large amount of code, greatly simplifying the development process, especially suitable for users to quickly customize and deploy complex algorithms and monitoring rules in industrial environments.

[0043] The platform provides an intelligent algorithm arrangement module, which uses low code technology to allow users to flexibly configure algorithms through a graphical interface without writing complex code, greatly improving the ease of use and scalability of the system. The specific implementation steps of this module are as follows: Algorithm arrangement interface: Users can choose different time series analysis algorithms (such as time series anomaly detection, prediction algorithms) through the platform's graphical interface, and drag and drop them into the flowchart to form an algorithm arrangement for production monitoring processes, real-time monitoring and prediction of potential risks in industrial safety production. The platform supports low-code algorithm arrangement environment and complex time series intelligent algorithms, which can learn from historical data to identify patterns of safety hazards in production and make intelligent predictions and alarms.

[0044] Custom algorithm logic: Users can customize the logical order and trigger conditions of multiple algorithms according to specific needs. For example, users can configure the system to automatically trigger an anomaly detection algorithm when the device temperature exceeds a certain threshold, and issue an alarm when a potential fault is detected.

[0045] Algorithm deployment and update: The algorithms arranged by users can be quickly deployed to the production site, and the algorithm logic can be adjusted at any time during production to ensure that the system is flexible and adaptable to different production scenarios.

[0046] (4) Intent recognition technology.

[0047] Intent recognition is an important branch of artificial intelligence and natural language processing in recent years, often used to understand the purpose behind user or system behavior. In this invention, intent recognition technology combines deep learning models to analyze industrial production site operation data and infer operator behavior intent in real time, thereby improving the system's safety warning and emergency response capabilities.

[0048] The platform analyzes the behavior intent of the operator and the potential risks of the production operation in real time through the intent recognition module, and the specific implementation steps are as follows: Data collection and analysis: The module extracts behavior features from the operator's action trajectory, device operation log and other data, and identifies the operator's intent through a deep learning model, such as Figure 2 For example, the system can determine whether the operator is performing a high-risk operation and issue a warning for dangerous behavior. Rule-based intent recognition methods can also be used to set rules based on fixed patterns in historical data to achieve automated intent detection; this approach may lack flexibility, but can achieve higher accuracy through manual configuration in specific scenarios, making it suitable for industrial environments that are not sensitive to new technologies or require strict control.

[0049] Intent recognition and early warning: When the system identifies certain high-risk intent (such as illegal operation during device maintenance), it will immediately trigger an early warning and provide appropriate safety recommendations. For example, when detecting that the operator is overloading the device, the system will recommend reducing the load to avoid device failure.

[0050] Emergency plan automatic generation: In the event of an emergency, by analyzing historical operation data and real-time monitoring information, combined with the prediction results of the domain large model, an emergency handling plan is automatically generated, including specific operation steps and precautions, to help operators take quick action to reduce accident losses and improve the intelligent level of production safety management.

[0051] (5) Integration and application of domain large model.

[0052] Domain large model is an intelligent model that has rapidly developed in vertical industries in recent years. The domain large model in the present invention is a specialized model based on the industrial field. It mainly trains a large number of data sets related to industrial safety production, including historical accident data and equipment operation data, to provide accurate decision support and prediction functions for specific fields. In the industrial environment, the domain large model can quickly identify abnormalities, predict equipment failures, and provide corresponding handling suggestions.

[0053] The present invention integrates the domain large model in the field of industrial safety production into the platform to provide intelligent decision support and fast knowledge services. The specific implementation steps are as follows: Pre-training of large model: The platform pre-trains the domain large model based on industry expert knowledge and historical production data to ensure that it can adapt to the specific needs of the production site. The model can learn industry experience, accident patterns, fault types, and other knowledge from historical data and make predictions based on current production data. Knowledge graph technology can also be used to build an industry-specific knowledge base. Through the structured representation of expert knowledge and historical events, the knowledge graph can provide accurate decision support and question-answering systems for safety production. Compared with the large model, the knowledge graph is more flexible and has lower maintenance costs, making it suitable for scenarios with scarce data.

[0054] Intelligent decision support: In daily production monitoring, the system combines the analysis results of the domain large model to provide intelligent decision support for operators and managers. For example, when the system predicts equipment failure, the large model can suggest maintenance time, required materials, and other information to help workers prepare in advance and reduce equipment downtime.

[0055] Intelligent question-answering and fast report generation: The platform supports domain knowledge question-answering functions. Users can ask questions in natural language to get quick answers about production safety, accident handling, and other aspects. In addition, the domain large model can automatically generate safety production reports based on current data to help managers quickly understand production conditions.

[0056] The platform of the present invention can adapt to changing industrial environments through continuous learning and updating of the domain large model, providing dynamic safety management solutions.

[0057] (6) Emergency response and accident disposal.

[0058] The emergency response module in the present application combines intent recognition technology and domain large model, and can quickly respond and generate disposal scheme when an accident occurs. The specific steps are as follows: Real-time monitoring and accident detection: The platform detects the occurrence of accidents (such as fire, equipment failure, personnel injury, etc.) in the first time through real-time monitoring of multi-modal data.

[0059] Emergency plan generation: After the accident is detected, the system will automatically generate an emergency plan according to the type of the accident, the on-site data and the historical processing experience. The plan content includes operation steps, resource scheduling, personnel arrangement, etc.; the plan will be displayed on the interface of the manager in an intuitive way, which is convenient for quick execution. At the same time, based on reinforcement learning technology, the system can train itself in accident simulation, and automatically generate emergency handling scheme through simulation of different accident situations. This scheme can adapt to more complex emergency situations, but it needs high computing resources and accident simulation data in implementation.

[0060] Accident report and follow-up processing: The system automatically generates an accident handling report, records the cause of the accident, the processing process and the follow-up rectification measures, and archives the report for the manager to review and optimize the future emergency response mechanism.

[0061] (7) Extensibility and application scenarios of the system.

[0062] Modular design: The present application adopts modular design, so that each functional module can be independently expanded. Users can select and deploy different functional modules according to specific needs, such as data fusion, intent recognition, video recognition, etc.

[0063] Wide application scenarios: The platform of the present application is applicable to industrial production scenarios in multiple industries, including power, petrochemical, mechanical manufacturing, mining, etc. It is especially suitable for high-risk scenarios that require real-time monitoring and emergency handling, and can also be applied to safety production management in mechanical manufacturing, food processing and other fields. The platform can provide comprehensive safety supervision and intelligent decision support.

[0064] The above description of the embodiments is for the convenience of those skilled in the art to understand and apply the present application. Those skilled in the art can easily make various modifications to the above embodiments, and apply the general principles described herein to other embodiments without creative labor. Therefore, the present application is not limited to the above embodiments, and those skilled in the art can make improvements and modifications to the present application according to the disclosure of the present application, which should be within the scope of protection of the present application.

Claims

1. An intelligent industrial safety production supervision platform based on intent recognition and domain large model, characterized by: include: The multi-source and multi-modal data fusion module acquires data of different modalities collected from different sources at industrial production sites and performs data preprocessing and data fusion; Multimodal video recognition module, which uses computer vision and deep learning algorithms to automatically detect anomalies in videos and generate corresponding treatment suggestions; The intelligent algorithm orchestration module uses low-code algorithm orchestration technology to enable users to configure various intelligent algorithms on a graphical interface; Intention recognition module, used to analyze operator behavior intentions, identify potential risks in production operations, and issue early warnings for risks; The domain model integration module provides users with intelligent decision support and rapid knowledge services by integrating pre-trained domain models; The emergency response module generates emergency plans through intent recognition and large domain models, and responds quickly to production accidents.

2. The intelligent industrial safety production supervision platform based on intent recognition and domain large model according to claim 1 is characterized by: The multi-source and multi-modal data fusion module first obtains spatiotemporal data collected from multiple sources including sensors, video surveillance equipment, and production equipment at the industrial production site. These data contain multiple modalities such as temperature, humidity, equipment operating status, and video images. The module then cleans, denoises, and standardizes these data to ensure data integrity and consistency. It then uses a spatiotemporal correlation analysis algorithm to fuse the data from different sources and presents the fused data to managers through a visual interface.

3. The intelligent industrial safety production supervision platform based on intent recognition and domain large model according to claim 1 is characterized by: The multimodal video recognition module uses CNN and ViT to perform multi-dimensional analysis on surveillance video data to identify abnormal situations in the video. When an abnormal situation is detected, the module generates an alarm prompt based on the analysis results and combines other data sources for further confirmation to ensure the accuracy of the alarm.

4. The intelligent industrial safety production supervision platform based on intent recognition and domain large model according to claim 1 is characterized by: The intelligent algorithm orchestration module is implemented using low-code algorithm orchestration technology, allowing users to select different timing analysis algorithms through the platform's graphical interface and drag and drop them into the flowchart. The logical order and trigger conditions of these algorithms can be customized according to specific needs to form an algorithm orchestration for the production monitoring process, and then the orchestrated algorithms can be quickly deployed to the industrial production site, and the algorithm logic and parameters can be adjusted and updated according to actual conditions.

5. The intelligent industrial safety production supervision platform based on intent recognition and domain large model according to claim 1 is characterized by: The intention recognition module obtains the operator's movement trajectory and equipment operation log, extracts behavioral features from them, and identifies the operator's intention through a deep learning algorithm model. When a high-risk intention is identified, an early warning is triggered. By analyzing historical operation data and real-time monitoring information combined with the prediction results of the domain big model, the operator is provided with intelligent safety recommendations and emergency response plans.

6. The intelligent industrial safety production supervision platform based on intent recognition and domain large model according to claim 5 is characterized by: The intention recognition module adopts a rule-based intention recognition method, combines fixed patterns in historical data to set rules, and realizes automated intention detection.

7. The intelligent industrial safety production supervision platform based on intent recognition and domain large model according to claim 1 is characterized by: The domain big model integration module integrates the domain big model of industrial production safety. The domain big model learns knowledge including industry experience, accident patterns, and failure types from industry experts and historical production data through pre-training. In daily production monitoring, the domain big model makes analysis and predictions based on current production data, provides operators and managers with intelligent auxiliary decision support and domain knowledge question and answer services, automatically generates production safety reports, and continuously learns and updates to adapt to the specific needs of the production site and the changing industrial environment.

8. The intelligent industrial safety production supervision platform based on intent recognition and domain large model according to claim 7 is characterized by: The domain large model integration module adopts knowledge graph technology to build an industry-specific knowledge base, and provides accurate decision support and question-answering system for safe production through structured representation of expert knowledge and historical events.

9. The intelligent industrial safety production supervision platform based on intent recognition and domain large model according to claim 1 is characterized by: The emergency response module uses intent recognition technology based on real-time monitoring multimodal data to detect the occurrence of accidents at the first time, and then automatically generates emergency plans based on the accident type and on-site data combined with the domain big model. The plan content, including operating steps, resource scheduling, and personnel arrangements, will be intuitively displayed on the platform interface. At the same time, an accident handling report is automatically generated to record the cause of the accident, the handling process and subsequent corrective measures, and the report is archived.

10. The intelligent industrial safety production supervision platform based on intent recognition and domain large model according to claim 9 is characterized by: The emergency response module is based on reinforcement learning technology, uses the model to self-train in accident simulation, and automatically generates emergency treatment plans by simulating different accident scenarios.

Images