An intelligent safety supervision method, device, equipment and storage medium
By combining pre-set intelligent agents with production safety documents, and utilizing retrieval enhancement generation technology and multimodal models to comprehensively determine abnormal data, the problem of false alarms in existing technologies has been solved, and the efficiency of handling production safety and emergencies has been improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INSPUR YUNZHOU (SHANDONG) IND INTERNET CO LTD
- Filing Date
- 2025-01-03
- Publication Date
- 2026-07-03
AI Technical Summary
Existing intelligent monitoring systems for safe production use a single threshold for early warning events and risk identification, which is prone to false alarms and thus restricts production safety.
By combining a pre-set intelligent agent with production safety documents, and utilizing retrieval enhancement generation technology, a pre-set large language model and multimodal model, as well as heuristic retrieval algorithms, abnormal data is comprehensively judged, and feedback reports are generated for anomaly handling.
It reduced the workload of those stuck at their posts and improved the safety of enterprise production and the efficiency and accuracy of handling emergencies.
Smart Images
Figure CN122334653A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence technology, and in particular to an intelligent security monitoring method, device, equipment, and storage medium. Background Technology
[0002] Currently, intelligent monitoring systems for safe production mainly include a front-end acquisition module, a transmission network module, a back-end data processing and analysis module, and a visualization management platform module. The front-end acquisition module collects video data and other data (such as the concentration of harmful gases, ambient temperature, speed, and pressure) generated at various production processes and work sites. The transmission network module uploads the data collected from the front end to the back-end data processing and analysis module. The back-end data processing and analysis module performs intelligent analysis on the video data and other data to determine if there are any abnormal behaviors or if thresholds are exceeded. The visualization management platform module displays normal information and alarm information after intelligent analysis. However, the above methods rely on threshold judgments to identify early warning events and risks, resulting in a relatively singular judgment standard. This limits data processing and analysis and may lead to false alarms.
[0003] As can be seen from the above, how to comprehensively judge abnormal data to avoid false alarms is an urgent problem to be solved. Summary of the Invention
[0004] In view of this, the purpose of this invention is to provide an intelligent security monitoring method, device, equipment, and storage medium, capable of comprehensively judging abnormal data to avoid false alarms. The specific solution is as follows:
[0005] Firstly, this application provides an intelligent security supervision method, including:
[0006] When the preset intelligent agent receives abnormal data from the site, it uses retrieval enhancement generation technology to retrieve target search text related to the abnormal data from the production safety documents;
[0007] The target retrieval text is used to perform task planning using a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm to obtain the target task and the corresponding execution action, and the target task is stored in the task list.
[0008] The target task in the task list is executed using a preset execution strategy and the execution action, and the deviation between the execution result of the target task and the expected execution result is determined.
[0009] If the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, a feedback report corresponding to the abnormal data is generated so that anomaly handling can be performed based on the feedback report.
[0010] Optionally, when the preset intelligent agent receives abnormal data from the site, it uses retrieval enhancement generation technology to retrieve target search text related to the abnormal data from production safety documents, including:
[0011] By using intelligent monitoring equipment to monitor various devices on site, abnormal data can be obtained.
[0012] Based on the preset intelligent agent, the abnormal data is obtained, and the text related to the abnormal data is queried from the production safety document using retrieval enhancement generation technology. The corresponding target retrieval text is then generated using text generation technology.
[0013] Optionally, the step of using a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm to perform task planning on the target retrieval text to obtain the target task and the corresponding execution action, and storing the target task in a task list, includes:
[0014] The target retrieved text is semantically understood using a pre-defined large language model and a pre-defined multimodal model to obtain text understanding results;
[0015] Task planning is performed based on text understanding results and heuristic retrieval algorithms to obtain each task. Then, target tasks that meet preset optimal conditions and their corresponding execution actions are obtained from each of the tasks, and the target tasks are stored in the task list.
[0016] Optionally, if the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, a feedback report corresponding to the abnormal data is generated so that anomaly handling can be performed based on the feedback report, including:
[0017] If the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, then the preset state machine is updated based on the target task execution result, so as to generate a feedback report corresponding to the abnormal data based on the preset state machine, so as to perform abnormal processing based on the feedback report;
[0018] The preset state machine is used to store the execution plan of the target task, the execution result of the target task, the number of attempts, the completion status of the target task, and the deviation between the execution result of the target task and the expected execution result of the task.
[0019] Optionally, if the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, a feedback report corresponding to the abnormal data is generated so that anomaly handling can be performed based on the feedback report, including:
[0020] If the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, then subtask planning is performed based on the number of attempts corresponding to the target task, and the subtask planning is executed to obtain the corresponding subtask execution result;
[0021] Determine whether the deviation between the subtask execution result and the expected task execution result is greater than a preset deviation threshold. If the deviation between the subtask execution result and the expected task execution result is not greater than the preset deviation threshold, then generate a feedback report corresponding to the abnormal data based on the subtask execution result, so as to perform abnormal handling based on the feedback report.
[0022] Optionally, after determining whether the deviation between the subtask execution result and the expected task execution result is greater than a preset deviation threshold, the method further includes:
[0023] If the deviation between the subtask execution result and the expected task execution result is greater than a preset deviation threshold, then proceed to the step of using a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm to plan the task for the target retrieval text.
[0024] Optionally, the intelligent security supervision method further includes:
[0025] The task orchestration tool is used to prioritize the target tasks, and the orchestration results are distributed and executed based on a distributed architecture.
[0026] The execution status and execution result corresponding to the target task are saved to a preset storage system so that the preset intelligent agent can perform task management based on the preset storage system.
[0027] Secondly, this application provides an intelligent security monitoring device, comprising:
[0028] The text retrieval module is used to retrieve target retrieval text related to the abnormal data from production safety documents using retrieval enhancement generation technology after the preset intelligent agent receives abnormal data from the site.
[0029] The task planning module is used to plan tasks for the target retrieval text using a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm, so as to obtain the target task and the corresponding execution action, and store the target task in the task list.
[0030] The task execution module is used to execute the target task in the task list using a preset execution strategy and the execution action, and to determine the deviation between the target task execution result and the expected task execution result.
[0031] An exception handling module is used to generate a feedback report corresponding to the exception data if the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, so as to perform exception handling based on the feedback report.
[0032] Thirdly, this application provides an electronic device, comprising:
[0033] Memory, used to store computer programs;
[0034] A processor is used to execute the computer program to implement the aforementioned intelligent security monitoring method.
[0035] Fourthly, this application provides a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned intelligent security monitoring method.
[0036] When a pre-defined intelligent agent receives abnormal data from the field, this application uses retrieval enhancement generation technology to retrieve target retrieval text related to the abnormal data from production safety documents; it then uses a pre-defined large language model, a pre-defined multimodal model, and a heuristic retrieval algorithm to perform task planning on the target retrieval text to obtain target tasks and corresponding execution actions, and stores the target tasks in a task list; it then executes the target tasks in the task list using a pre-defined execution strategy and the execution actions, and determines the deviation between the target task execution result and the expected task execution result; if the deviation between the target task execution result and the expected task execution result is greater than a pre-defined deviation threshold, it generates a feedback report corresponding to the abnormal data, so as to perform anomaly handling based on the feedback report.
[0037] As can be seen from the above, this application first uses a preset intelligent agent to receive abnormal data, and then uses retrieval enhancement generation technology to retrieve target retrieval text related to the abnormal data from production safety documents. In this way, only the execution standards of the equipment layer need to be described in text form, which greatly reduces the supervision cost. Then, a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm are used to plan tasks for the target retrieval text, so that the target task is the task most relevant to the abnormal data. The deviation of the target task execution result is calculated, and a corresponding feedback report is generated based on the calculation result. The preset intelligent agent combined with production safety documents and other factors is used for comprehensive judgment, which not only reduces the workload of those stationed at their posts, but also improves the enterprise's production safety coefficient and the efficiency and accuracy of handling emergencies. Attached Figure Description
[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0039] Figure 1 This application discloses a flowchart of an intelligent safety supervision method.
[0040] Figure 2 This application discloses a specific intelligent security supervision method flowchart;
[0041] Figure 3 This is a schematic diagram of a specific intelligent security monitoring platform disclosed in this application;
[0042] Figure 4 This is a schematic diagram of the structure of an intelligent safety monitoring device disclosed in this application;
[0043] Figure 5 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation
[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0045] Currently, existing intelligent safety monitoring systems rely on threshold judgments to identify early warning events and risks. This simplistic approach limits data processing and analysis, potentially leading to false alarms and other safety concerns. Therefore, this application proposes an intelligent safety supervision method that combines pre-set intelligent agents with production safety documents for comprehensive judgment. This not only reduces the workload of those physically on duty but also improves the enterprise's production safety coefficient and the efficiency and accuracy of handling emergencies.
[0046] See Figure 1 As shown in the figure, an embodiment of the present invention discloses an intelligent security monitoring method, including:
[0047] Step S11: When the preset intelligent agent receives abnormal data from the site, it uses retrieval enhancement generation technology to retrieve target search text related to the abnormal data from the production safety document.
[0048] In this embodiment, intelligent monitoring equipment is used to monitor abnormal data on-site. The abnormal data is acquired through a pre-defined intelligent agent. Search enhancement and generation technology is used to search the production safety document based on the data characteristics of the abnormal data. The search results yield target search text related to the abnormal data. Specifically, when the pre-defined intelligent agent receives abnormal data from the site, it uses search enhancement and generation technology to retrieve target search text related to the abnormal data from the production safety document. This includes: monitoring various devices on-site using intelligent monitoring equipment to obtain abnormal data; based on the abnormal data acquired by the pre-defined intelligent agent, using search enhancement and generation technology to query text related to the abnormal data from the production safety document, and generating corresponding target search text using text generation technology.
[0049] It is understood that various intelligent monitoring devices such as sensors, cameras, infrared detectors, radar detectors, or audio can be used to monitor abnormal data on site. When the preset intelligent agent receives abnormal data on site, it can first preprocess the abnormal data to convert it into a data format more suitable for retrieval. Based on retrieval enhancement generation technology, it retrieves the target retrieval text that best matches the processed data from the production safety document. The preset intelligent agent records the entire retrieval process, which may include the received abnormal data and the retrieved target retrieval text, so as to conduct subsequent analysis based on the recorded data.
[0050] Step S12: Use a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm to perform task planning on the target retrieval text, so as to obtain the target task and the corresponding execution action, and store the target task in the task list.
[0051] In this embodiment, after obtaining the target retrieval text, semantic understanding is performed on the target retrieval text using a preset large language model and a preset multimodal model to obtain text understanding results. Task planning is then performed using the text understanding results and a heuristic retrieval algorithm to obtain target tasks and corresponding execution actions that meet preset optimal conditions from the obtained tasks. Finally, a task list is initialized, and the target tasks are stored in the task list. Specifically, the step of performing task planning on the target retrieval text using a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm to obtain target tasks and corresponding execution actions, and storing the target tasks in the task list, includes: performing semantic understanding on the target retrieval text using a preset large language model and a preset multimodal model to obtain text understanding results; performing task planning based on the text understanding results and a heuristic retrieval algorithm to obtain various tasks; obtaining target tasks and corresponding execution actions that meet preset optimal conditions from each task; and storing the target tasks in the task list.
[0052] Understandably, after obtaining the target retrieval text, a finely tuned preset large language model LlaMA (Large Language Model Meta AI) is loaded to perform semantic understanding on the target retrieval text. If the target retrieval text contains images, speech, or modal information, a preset multimodal model LlaVA (Large Language and Vision Assistant) can be used to perform semantic understanding on the target retrieval text to obtain text understanding results, achieving alignment between non-text content and text content. Then, the Monte Carlo Tree Search (MCTS) algorithm can be used, based on the text understanding results, to plan tasks and corresponding execution actions, select the target task with the highest score among all tasks, and generate the corresponding execution action based on the target task. It should be noted that the preset optimal conditions can be adjusted according to the actual situation, and are not specifically limited here.
[0053] Step S13: Execute the target task in the task list using the preset execution strategy and the execution action, and determine the deviation between the target task execution result and the expected task execution result.
[0054] In this embodiment, after generating the target tasks, a task orchestration tool can be used to prioritize the target tasks, and the target tasks can be distributed based on the orchestration results. The execution status and execution results of the target tasks are stored in a preset storage system so that the preset intelligent agent can manage tasks through the preset storage system. Specifically, it also includes: prioritizing each of the target tasks using a task orchestration tool, and distributing and executing the target tasks based on the orchestration results using a distributed architecture; saving the execution status and execution results of the target tasks to the preset storage system so that the preset intelligent agent can manage tasks based on the preset storage system.
[0055] It is understandable that after generating the corresponding target task, a task orchestration tool is used to orchestrate task priorities and distribute tasks. The task orchestration tool can adopt an asynchronous queue distributed Celery (CeleryDistributed Task Queue) architecture to register tasks and perform distributed task execution. The results and execution status are stored in Redis (Remote Dictionary Server) middleware so that the preset agent can perform task cancellation, task pause, task tracking and other related task status management.
[0056] Step S14: If the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, a feedback report corresponding to the abnormal data is generated so that abnormal processing can be performed based on the feedback report.
[0057] In this embodiment, a preset state machine can be constructed to store the execution plan of the target task, the execution result of the target task, the number of attempts, the completion status of the target task, and the deviation between the execution result of the target task and the expected execution result of the task. This allows for the generation of a feedback report corresponding to the abnormal data based on the deviation between the execution result of the target task and the expected execution result, and for anomaly handling based on the feedback report. Specifically, generating a feedback report corresponding to the abnormal data if the deviation between the execution result of the target task and the expected execution result is greater than a preset deviation threshold, and then performing anomaly handling based on the feedback report, includes: updating the preset state machine based on the execution result of the target task if the deviation between the execution result of the target task and the expected execution result is greater than the preset deviation threshold, and then generating a feedback report corresponding to the abnormal data based on the preset state machine, and performing anomaly handling based on the feedback report. The preset state machine stores the execution plan of the target task, the execution result of the target task, the number of attempts, the completion status of the target task, and the deviation between the execution result of the target task and the expected execution result of the task.
[0058] Furthermore, if the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, then subtask planning is performed based on the number of attempts, and the subtasks obtained from the subtask planning are executed to obtain subtask execution results. It is then determined whether the deviation between the subtask execution result and the expected task execution result is greater than a preset deviation threshold. If the deviation between the subtask execution result and the expected task execution result is not greater than the preset deviation threshold, then a feedback report is generated based on the subtask execution result, and anomaly handling is performed based on the feedback report. Specifically, the step of generating a feedback report corresponding to the abnormal data if the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, so as to perform anomaly handling based on the feedback report, includes: if the deviation between the target task execution result and the expected task execution result is greater than the preset deviation threshold, then performing sub-task planning based on the number of attempts corresponding to the target task, and executing the sub-task planning to obtain the corresponding sub-task execution result; determining whether the deviation between the sub-task execution result and the expected task execution result is greater than the preset deviation threshold; if the deviation between the sub-task execution result and the expected task execution result is not greater than the preset deviation threshold, then generating a feedback report corresponding to the abnormal data based on the sub-task execution result, so as to perform anomaly handling based on the feedback report.
[0059] Understandably, if the deviation between the subtask execution result and the expected task execution result is greater than a preset deviation threshold, then task planning is re-performed, i.e., the process jumps to the step of using a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm to plan the task for the target retrieval text. Specifically, after determining whether the deviation between the subtask execution result and the expected task execution result is greater than the preset deviation threshold, the process further includes: if the deviation between the subtask execution result and the expected task execution result is greater than the preset deviation threshold, then the process jumps to the step of using a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm to plan the task for the target retrieval text.
[0060] In one specific implementation, the deviation between the subtask execution result and the expected task execution result can be calculated using context vectors. If the deviation is greater than 1, subtask planning is performed on the target task based on the number of attempts, resulting in various subtasks. Each subtask is then executed to determine if its execution result meets the expected task result. If the subtask's execution result does not meet the expected result, the target task corresponding to that subtask is replanned. By controlling the number of attempts, task planning and execution capabilities can be enhanced, avoiding task divergence and wasting computational resources.
[0061] As can be seen from the above, this application first uses a preset intelligent agent to receive abnormal data, and then uses retrieval enhancement generation technology to retrieve target retrieval text related to the abnormal data from production safety documents. In this way, only the execution standards of the equipment layer need to be described in text form, which greatly reduces the supervision cost. Then, a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm are used to plan tasks for the target retrieval text, so that the target task is the task most relevant to the abnormal data. The deviation of the target task execution result is calculated, and a corresponding feedback report is generated based on the calculation result. The preset intelligent agent combined with production safety documents and other factors is used for comprehensive judgment, which not only reduces the workload of those stationed at their posts, but also improves the enterprise's production safety coefficient and the efficiency and accuracy of handling emergencies.
[0062] As can be seen from the above embodiments, this application makes intelligent decisions based on a preset intelligent agent to achieve intelligent monitoring of abnormal situations. Therefore, the process of making intelligent decisions based on a preset intelligent agent to achieve intelligent monitoring of abnormal situations is described.
[0063] See Figure 2 As shown in the figure, an embodiment of the present invention discloses a specific intelligent security monitoring method, including:
[0064] The system performs visual inspection on abnormal data acquired by intelligent monitoring equipment to identify target objects within the equipment and classify and label them; or it performs visual segmentation on the abnormal data, dividing it into different semantic regions to determine if any abnormal regions exist; or it performs visual measurement on the abnormal data to measure the size, outline, area, etc., of target objects within the abnormal data, analyzing the shape information of the target objects to determine if any anomalies exist. The obtained abnormal data is then transmitted to a preset intelligent agent, which has an interactive interface, question-and-answer functionality, and report generation capabilities. The preset intelligent agent can also manage and analyze tasks and monitor the system's operation. Furthermore, the system possesses visual inspection and execution supervision capabilities, enabling it to monitor the equipment's status in real time and take appropriate anomaly handling actions.
[0065] Furthermore, Figure 3This is a schematic diagram of an intelligent safety supervision platform provided in this embodiment. The online monitoring layer can use intelligent monitoring equipment to perform safety monitoring on target equipment, environment, parks, factories, personnel, and supply chains. When abnormal data is obtained, the abnormal data is sent to the preset intelligent agent. The preset intelligent agent acts as an industrial safety monitoring agent, which can autonomously execute monitoring tasks and make decisions and actions based on the analysis results. The preset intelligent agent includes using a preset large language model and a preset multimodal model to perform task planning on the target retrieval text related to the abnormal data, and executing the obtained target tasks. It determines whether the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, and generates a corresponding feedback report based on the judgment result, so as to handle the anomaly through the feedback report.
[0066] It is worth mentioning that this embodiment integrates a self-developed preset digital human as the smallest execution unit for the target task, providing services such as notification of responsible personnel, monitoring of task progress, answering of questions, closed-loop tracking of pending issues, and remote control of downstream execution devices, such as camera evidence collection, snapshot capture, and voice broadcasting. In addition, there are various sensory sub-models, such as visual recognition, visual segmentation, 3D point cloud, and speech recognition. Simultaneously, the security monitoring and management system provides open third-party access interfaces for connecting third-party security personnel, email addresses, and WeChat interfaces, facilitating the preset intelligent agent to access resources within the enterprise's defined scope for multi-departmental collaborative security management.
[0067] It is understood that the preset intelligent agent also includes a preset vector library, which consists of enterprise documents, rules and regulations, relevant dynamic knowledge, and relevant external policy documents. This library is used by the preset intelligent agent to search for relevant knowledge so that it can conduct daily interactive question and answer based on the found target knowledge. A preset knowledge graph can be set up, and a fine-tuned preset large language model can be used to extract daily interactive text, abnormal events, production process document parameters, daily handling methods, internal enterprise documents, rule-making and other relevant documents, automatically extract entities, build hierarchical relationships between entities, and form a knowledge network as the enterprise's knowledge accumulation.
[0068] As can be seen from the above, by using a preset intelligent agent, a preset large language model, and a preset multimodal model to perform task planning for the target retrieval text corresponding to abnormal data, the resulting target task is the one most relevant to the abnormal data. The deviation of the target task execution result is calculated, and a corresponding feedback report is generated based on the calculation result. By using a preset intelligent agent in conjunction with production safety documents for comprehensive judgment, the entire process of the safety production monitoring system is streamlined. From incident monitoring, incident judgment, incident push, incident handling and tracking, post-event analysis, statistical prediction, to multi-department collaborative management, the system can handle the entire process independently or flexibly integrate with some existing enterprise systems, reducing the workload of those stuck in fixed positions and thus improving the efficiency and accuracy of the enterprise in handling emergencies.
[0069] Accordingly, see Figure 4 As shown, this application also provides an intelligent safety monitoring device, including:
[0070] The text retrieval module 11 is used to retrieve target retrieval text related to the abnormal data from the production safety document using retrieval enhancement generation technology after the preset intelligent agent receives abnormal data from the site.
[0071] The task planning module 12 is used to plan tasks for the target retrieval text using a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm, so as to obtain the target task and the corresponding execution action, and store the target task in the task list.
[0072] The task execution module 13 is used to execute the target task in the task list using a preset execution strategy and the execution action, and to determine the deviation between the target task execution result and the expected task execution result;
[0073] The exception handling module 14 is used to generate a feedback report corresponding to the exception data if the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, so as to perform exception handling based on the feedback report.
[0074] As can be seen from the above, this application first uses a preset intelligent agent to receive abnormal data, and then uses retrieval enhancement generation technology to retrieve target retrieval text related to the abnormal data from production safety documents. In this way, only the execution standards of the equipment layer need to be described in text form, which greatly reduces the supervision cost. Then, a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm are used to plan tasks for the target retrieval text, so that the target task is the task most relevant to the abnormal data. The deviation of the target task execution result is calculated, and a corresponding feedback report is generated based on the calculation result. The preset intelligent agent combined with production safety documents and other factors is used for comprehensive judgment, which not only reduces the workload of those stationed at their posts, but also improves the enterprise's production safety coefficient and the efficiency and accuracy of handling emergencies.
[0075] In some specific embodiments, the text retrieval module 11 may specifically include:
[0076] The abnormal data acquisition unit is used to monitor various devices on site through intelligent monitoring equipment in order to obtain abnormal data on site.
[0077] The text query unit is used to obtain the abnormal data based on the preset intelligent agent, use retrieval enhancement generation technology to query text related to the abnormal data from production safety documents, and generate corresponding target retrieval text through text generation technology.
[0078] In some specific embodiments, the task planning module 12 may specifically include:
[0079] The semantic understanding unit is used to perform semantic understanding on the target retrieved text using a preset large language model and a preset multimodal model to obtain text understanding results;
[0080] The task planning unit is used to plan tasks based on text understanding results and heuristic retrieval algorithms to obtain each task, and to obtain the target task and corresponding execution action that meet the preset optimal conditions from each task, and to store the target task in the task list.
[0081] In some specific embodiments, the exception handling module 14 may specifically include:
[0082] The state machine update unit is used to update a preset state machine based on the target task execution result if the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, so as to generate a feedback report corresponding to the abnormal data based on the preset state machine, so as to perform abnormal handling based on the feedback report.
[0083] In some specific embodiments, the exception handling module 14 may specifically include:
[0084] The subtask planning unit is used to perform subtask planning based on the number of attempts corresponding to the target task if the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, and to execute the subtask planning to obtain the corresponding subtask execution result.
[0085] The deviation judgment unit is used to determine whether the deviation between the subtask execution result and the expected task execution result is greater than a preset deviation threshold. If the deviation between the subtask execution result and the expected task execution result is not greater than the preset deviation threshold, a feedback report corresponding to the abnormal data is generated based on the subtask execution result so as to perform abnormal handling based on the feedback report.
[0086] In some specific embodiments, the exception handling module 14 may specifically include:
[0087] The task replanning unit is used to jump to the step of using a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm to plan the task for the target retrieval text if the deviation between the subtask execution result and the expected task execution result is greater than a preset deviation threshold.
[0088] In some specific embodiments, the intelligent safety monitoring device may further include:
[0089] The task orchestration unit is used to prioritize each of the target tasks using a task orchestration tool, and to distribute and execute the orchestration results based on a distributed architecture.
[0090] The task management unit is used to save the execution status and execution result of the target task to a preset storage system so that the preset intelligent agent can perform task management based on the preset storage system.
[0091] Furthermore, embodiments of this application also disclose an electronic device, Figure 5 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the intelligent security monitoring method disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.
[0092] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.
[0093] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.
[0094] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including computer programs capable of performing the intelligent security monitoring method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.
[0095] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned intelligent security monitoring method. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.
[0096] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.
[0097] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0098] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.
[0099] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0100] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. An intelligent safety supervision method, characterized in that, include: When the preset intelligent agent receives abnormal data from the site, it uses retrieval enhancement generation technology to retrieve target search text related to the abnormal data from the production safety documents; The target retrieval text is used to perform task planning using a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm to obtain the target task and the corresponding execution action, and the target task is stored in the task list. The target task in the task list is executed using a preset execution strategy and the execution action, and the deviation between the execution result of the target task and the expected execution result is determined. If the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, a feedback report corresponding to the abnormal data is generated so that anomaly handling can be performed based on the feedback report.
2. The intelligent safety supervision method according to claim 1, characterized in that, When the preset intelligent agent receives abnormal data from the site, it uses retrieval enhancement generation technology to retrieve target search text related to the abnormal data from production safety documents, including: By using intelligent monitoring equipment to monitor various devices on site, abnormal data can be obtained. Based on the preset intelligent agent, the abnormal data is obtained, and the text related to the abnormal data is queried from the production safety document using retrieval enhancement generation technology. The corresponding target retrieval text is then generated using text generation technology.
3. The intelligent safety supervision method according to claim 1, characterized in that, The process of using a pre-defined large language model, a pre-defined multimodal model, and a heuristic retrieval algorithm to plan tasks for the target retrieval text, thereby obtaining target tasks and corresponding execution actions, and storing the target tasks in a task list, includes: The target retrieved text is semantically understood using a pre-defined large language model and a pre-defined multimodal model to obtain text understanding results; Task planning is performed based on text understanding results and heuristic retrieval algorithms to obtain each task. Then, target tasks that meet preset optimal conditions and their corresponding execution actions are obtained from each of the tasks, and the target tasks are stored in the task list.
4. The intelligent safety supervision method according to claim 1, characterized in that, If the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, a feedback report corresponding to the abnormal data is generated so that anomaly handling can be performed based on the feedback report, including: If the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, then the preset state machine is updated based on the target task execution result, so as to generate a feedback report corresponding to the abnormal data based on the preset state machine, so as to perform abnormal processing based on the feedback report; The preset state machine is used to store the execution plan of the target task, the execution result of the target task, the number of attempts, the completion status of the target task, and the deviation between the execution result of the target task and the expected execution result of the task.
5. The intelligent safety supervision method according to claim 4, characterized in that, If the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, a feedback report corresponding to the abnormal data is generated so that anomaly handling can be performed based on the feedback report, including: If the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, then subtask planning is performed based on the number of attempts corresponding to the target task, and the subtask planning is executed to obtain the corresponding subtask execution result; Determine whether the deviation between the subtask execution result and the expected task execution result is greater than a preset deviation threshold. If the deviation between the subtask execution result and the expected task execution result is not greater than the preset deviation threshold, then generate a feedback report corresponding to the abnormal data based on the subtask execution result, so as to perform abnormal handling based on the feedback report.
6. The intelligent safety supervision method according to claim 5, characterized in that, After determining whether the deviation between the subtask execution result and the expected task execution result is greater than a preset deviation threshold, the method further includes: If the deviation between the subtask execution result and the expected task execution result is greater than a preset deviation threshold, then proceed to the step of using a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm to plan the task for the target retrieval text.
7. The intelligent security supervision method according to any one of claims 1 to 6, characterized in that, Also includes: The task orchestration tool is used to prioritize the target tasks, and the orchestration results are distributed and executed based on a distributed architecture. The execution status and execution result corresponding to the target task are saved to a preset storage system so that the preset intelligent agent can perform task management based on the preset storage system.
8. An intelligent safety monitoring device, characterized in that, include: The text retrieval module is used to retrieve target retrieval text related to the abnormal data from production safety documents using retrieval enhancement generation technology after the preset intelligent agent receives abnormal data from the site. The task planning module is used to plan tasks for the target retrieval text using a preset large language model, a preset multimodal model, and a heuristic retrieval algorithm, so as to obtain the target task and the corresponding execution action, and store the target task in the task list. The task execution module is used to execute the target task in the task list using a preset execution strategy and the execution action, and to determine the deviation between the target task execution result and the expected task execution result. An exception handling module is used to generate a feedback report corresponding to the exception data if the deviation between the target task execution result and the expected task execution result is greater than a preset deviation threshold, so as to perform exception handling based on the feedback report.
9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the intelligent security monitoring method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, Used to store computer programs, wherein the computer programs, when executed by a processor, implement the intelligent security monitoring method as described in any one of claims 1 to 7.