Multi-agent-based industrial equipment abnormal event management method, equipment and medium

Abstract

The invention discloses a multi-agent-based industrial equipment abnormal event management method and device and a medium, and relates to the technical field of multi-agents, and the method comprises the steps: carrying out the equipment abnormal event recognition of industrial Internet of Things platform data, so as to determine an equipment abnormal event; based on the original knowledge base, performing multi-dimensional exception decision analysis on the equipment exception event to obtain a task execution instruction; based on the equipment automatic execution task instruction, first work order data is obtained through equipment work order completion processing; according to an operation and maintenance personnel execution task instruction, performing operation and maintenance dialogue completion processing to obtain second work order data; and performing knowledge classification distinguishing knowledge precipitation on the first work order data and the second work order data to determine updated knowledge, and uploading the updated knowledge to the original knowledge base to obtain an updated knowledge base. According to the method, the technical problem that the relevance of the abnormal event management process of the industrial equipment is poor is solved.

Description

Technical Field

[0001] The present application relates to the field of multi-agent technology, and in particular to a multi-agent-based industrial equipment abnormal event management method, equipment and medium. Background Art

[0002] The complexity of abnormal events in industrial equipment requires that management must cover the complete chain of "data collection-analysis and diagnosis-decision execution-experience accumulation". In actual industrial scenarios, equipment abnormalities are usually manifested as multi-parameter coupling phenomena. If only a single link is focused on, it is impossible to locate the root cause or assess the potential impact, leading to misjudgment. In addition, abnormal events that have not formed a closed-loop management are prone to trigger a vicious cycle of "incomplete handling-repeated failures-production downtime". Traditional IoT platforms have significant defects in the management of the entire life cycle of abnormal events: First, the process fragmentation problem is serious, and a single-point model is mostly used, without connecting the links of abnormal analysis, decision generation, work order execution, and knowledge accumulation. Second, knowledge fragmentation leads to inefficient experience reuse, and operation and maintenance knowledge is scattered in isolated carriers such as detection systems, expert notes, and historical work orders, without forming a structured network. Third, decision-making capabilities are difficult to deal with complex anomalies. Traditional algorithms can only handle single device failures, and cannot perform correlation analysis when faced with multi-device linkage anomalies. Summary of the invention

[0003] The embodiments of the present application provide a method, device and medium for abnormal event management of industrial equipment based on multi-agents, which solve the technical problem of poor correlation in the abnormal event management process of industrial equipment.

[0004] In a first aspect, an embodiment of the present application provides an industrial equipment abnormal event management method based on a multi-agent, characterized in that the method includes: obtaining industrial Internet of Things platform data, and identifying equipment abnormal events on the industrial Internet of Things platform data to determine equipment abnormal events; obtaining an original knowledge base, and based on the original knowledge base, performing multi-dimensional abnormal decision analysis on equipment abnormal events to obtain execution task instructions; wherein the execution task instructions include: equipment automation execution task instructions, operation and maintenance personnel execution task instructions; based on the equipment automation execution task instructions, through equipment work order completion processing, obtain first work order data; according to the operation and maintenance personnel execution task instructions, through operation and maintenance dialogue completion processing, obtain second work order data; perform knowledge precipitation on the first work order data and the second work order data by distinguishing knowledge categories to determine updated knowledge, and upload the updated knowledge to the original knowledge base to obtain an updated knowledge base.

[0005] In an implementation manner of the present application, device anomaly event recognition is performed on industrial Internet of Things platform data to determine device anomaly events, which specifically includes: performing data preprocessing on industrial Internet of Things platform data to obtain time-series device status data with dynamic and static attribute mapping; determining device analysis agents based on the time-series device status data with dynamic and static attribute mapping through device static feature analysis; obtaining device dynamic real-time status data through device feature vector comparison according to the device analysis agents; performing two-dimensional device anomaly type recognition on the device dynamic real-time status data to determine device anomaly events; wherein, the two-dimensional device anomaly type recognition includes: industry anomaly type recognition and device type anomaly recognition.

[0006] In an implementation manner of the present application, two-dimensional device anomaly type recognition is performed on device dynamic real-time status data to determine device anomaly events, which specifically includes: performing random forest classification on the device dynamic real-time status data to determine anomaly class mapping; determining device anomaly events based on the anomaly class mapping through anomaly class confidence analysis.

[0007] In an implementation manner of the present application, based on the original knowledge base, multi-dimensional anomaly decision analysis is performed on device anomaly events to obtain execution task instructions, which specifically includes: performing potential impact analysis on device anomaly events, and based on the analysis results of the potential impact analysis, matching the original knowledge base to obtain the first anomaly handling plan; performing root cause traceability analysis on device anomaly events, and according to the analysis results of the root cause traceability analysis, matching the original knowledge base to obtain the second anomaly handling plan; performing severity analysis on device anomaly events, and based on the analysis results of the severity analysis, matching the original knowledge base to obtain the third anomaly handling plan; according to the first anomaly handling plan, the second anomaly handling plan, and the third anomaly handling plan, through multi-source execution step allocation processing, obtaining execution task instructions.

[0008] In an implementation manner of the present application, based on the device automatically executing the task instruction, the first work order data is obtained through device work order completion processing, which specifically includes: performing Internet of Things device execution data processing on the device automatically executing the task instruction to obtain the Internet of Things device execution work order; sending the Internet of Things device execution work order to the execution end, and monitoring the execution process of the execution end to obtain execution anomaly data; determining the first work order update data based on the execution anomaly data through execution status analysis; updating the first work order update data to the Internet of Things device execution work order to obtain the first work order data.

[0009] In an implementation manner of the present application, according to the task instructions executed by the operation and maintenance personnel, through operation and maintenance dialogue completion processing, second work order data is obtained, specifically including: pre-filling a work order for the task instructions executed by the operation and maintenance personnel to obtain a work order for the operation and maintenance personnel to execute tasks; sending the work order for the operation and maintenance personnel to execute tasks to the execution end, and monitoring the execution process of the execution end to obtain execution exception data; based on the execution exception data, determining second work order update data through work order filling interaction guidance; updating the second work order update data to the work order for the operation and maintenance personnel to execute tasks to obtain second work order data.

[0010] In an implementation manner of the present application, knowledge precipitation for knowledge category differentiation of the first work order data and the second work order data is performed to determine updated knowledge, specifically including: integrating the processing process of the first work order data to obtain first integrated knowledge; integrating the processing process of the second work order data to obtain second integrated knowledge; determining first updated knowledge through explicit knowledge precipitation based on the first integrated knowledge; determining second updated knowledge through implicit knowledge precipitation according to the second integrated knowledge, and determining updated knowledge based on the first updated knowledge and the second updated knowledge.

[0011] In an implementation manner of the present application, after uploading the updated knowledge to the original knowledge base to obtain an updated knowledge base, the method further includes: performing an update analysis of the knowledge scope of the new knowledge base to determine tool update data; based on the tool update data, updating the tool types of the preset tool library to obtain an updated tool library.

[0012] In a second aspect, an industrial equipment abnormal event management device based on multi-agent provided by an embodiment of the present application is characterized in that the device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: obtain industrial Internet of Things platform data, and identify device abnormal events for the industrial Internet of Things platform data to determine device abnormal events; obtain the original knowledge base, and perform multi-dimensional abnormal decision analysis on the device abnormal events based on the original knowledge base to obtain execution task instructions; wherein, the execution task instructions include: device automation execution task instructions, operation and maintenance personnel execution task instructions; based on the device automation execution task instructions, obtain first work order data through device work order completion processing; according to the operation and maintenance personnel execution task instructions, obtain second work order data through operation and maintenance dialogue completion processing; perform knowledge precipitation for knowledge category differentiation of the first work order data and the second work order data to determine updated knowledge, and upload the updated knowledge to the original knowledge base to obtain an updated knowledge base.

[0013] In a third aspect, an embodiment of the present application further provides a non-volatile computer storage medium for industrial equipment abnormal event management based on multi-agent, storing computer-executable instructions, characterized in that the computer-executable instructions are set as follows: obtaining industrial Internet of Things platform data, and performing device abnormal event recognition on the industrial Internet of Things platform data to determine device abnormal event; obtaining an original knowledge base, and based on the original knowledge base, performing multi-dimensional abnormal decision analysis on the device abnormal event to obtain execution task instructions; wherein the execution task instructions include: device automation execution task instructions and operation and maintenance personnel execution task instructions; based on the device automation execution task instructions, through device work order completion processing, obtaining first work order data; according to the operation and maintenance personnel execution task instructions, through operation and maintenance dialogue completion processing, obtaining second work order data; performing knowledge precipitation of knowledge category differentiation on the first work order data and the second work order data to determine updated knowledge, and uploading the updated knowledge to the original knowledge base to obtain an updated knowledge base.

[0014] An embodiment of the present application provides a method, device and medium for industrial equipment abnormal event management based on multi-agent. Through abnormal event recognition of dynamic data and static data, multi-dimensional abnormal decision analysis, work order completion processing and automatic update of the knowledge base based on two completed work orders, the technical problem of poor relevance in the process of industrial equipment abnormal event management is solved, the full-process independent planning from abnormal detection, root cause analysis to disposal plan is realized, and the development cost of cross-scenario abnormal management is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The illustrative embodiments and descriptions thereof of the present application are used to explain the present application, and do not constitute an improper limitation of the present application. In the drawings: Figure 1 It is a flowchart of a method for industrial equipment abnormal event management based on multi-agent provided by an embodiment of the present application; Figure 2 It is an overall architecture diagram of a method for industrial equipment abnormal event management based on multi-agent provided by an embodiment of the present application; Figure 3 It is a schematic internal structure diagram of a device for industrial equipment abnormal event management based on multi-agent provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments of this application and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without making creative efforts belong to the scope of protection of this application.

[0017] The embodiments of this application provide a multi-agent-based industrial equipment abnormal event management method, device, and medium. Through abnormal event recognition of dynamic data and static data, multi-dimensional abnormal decision analysis, work order completion processing, and automatic update of the knowledge base based on two types of completed work orders, the technical problem of poor relevance in the process of industrial equipment abnormal event management is solved, and the full-process autonomous planning from abnormal detection, root cause analysis to disposal solutions is realized. Compared with traditional rule chain integration, the development cost of cross-scenario abnormal management is significantly reduced. It not only accurately solves the pain points of traditional systems in terms of operation complexity, knowledge precipitation, cross-scenario adaptation, etc., but also elevates the value of the Internet of Things platform from "device connection" to the "intelligent decision-making" level through the autonomous planning, knowledge evolution, and global optimization characteristics of agents.

[0018] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the drawings.

[0019] Figure 1 It is a flowchart of a multi-agent-based industrial equipment abnormal event management method provided by the embodiments of this application. As Figure 1 shown, a multi-agent-based industrial equipment abnormal event management method provided by the embodiments of this application specifically includes the following steps: Step 101: Obtain industrial Internet of Things platform data, and perform device abnormal event recognition on the industrial Internet of Things platform data to determine device abnormal events.

[0020] Exemplarily, in an industrial scenario, the device objects are rich and diverse, including industrial machines such as numerically controlled machine tools, robots, conveyors, pumps, compressors, and metering devices such as electric meters, water meters, and gas meters. All these devices need to be connected to the Internet of Things platform. During the connection process, the industrial Internet of Things platform data collected can be divided into static data and dynamic data. Abnormal event recognition is performed on the static data and dynamic data respectively, and an index format is established for subsequent rapid query, providing a data basis for multi-dimensional abnormal decision analysis.

[0021] Specifically, for identifying device anomaly events in the industrial Internet of Things (IIoT) platform data to determine device anomaly events, it includes: performing data preprocessing on the IIoT platform data to obtain time-series device status data with dynamic and static attribute mappings; determining device analysis agents based on the time-series device status data with dynamic and static attribute mappings through device static feature analysis; obtaining device dynamic real-time status data through device feature vector comparison according to the device analysis agents; performing two-dimensional device anomaly type identification on the device dynamic real-time status data to determine device anomaly events; where the two-dimensional device anomaly type identification includes: industry anomaly type identification and device type anomaly identification.

[0022] Furthermore, performing two-dimensional device anomaly type identification on the device dynamic real-time status data to determine device anomaly events specifically includes: performing random forest classification on the device dynamic real-time status data to determine anomaly class mappings; determining device anomaly events based on the anomaly class mappings through anomaly class confidence analysis.

[0023] In one embodiment, static data is the inherent and relatively stable attribute information of the device, such as model, manufacturer, design parameters, installation location, etc. Such data is usually determined when the device is put into production or renovated, and is generally only updated during major changes such as equipment overhaul and upgrade. Taking manual entry of static data as an example, when a batch of new CNC machine tools are introduced, technicians can log in to the IIoT platform console, enter the device management module, find the new device entry, and fill in information such as device model, manufacturer, design parameters, installation location, etc. in sequence to complete the manual entry of static data. The static data entry can also be a device self-registration mechanism.

[0024] Dynamic data, on the other hand, is the information generated by the device in real time and changing with the operating state or environment, such as the temperature, vibration value, operating load, anomaly trigger time, and influence range monitored by sensors. This type of data is refreshed at a frequency of seconds or minutes, and can intuitively reflect the current state of the device. During actual acquisition, it is necessary to configure products, devices, communication Topics, and data flow solutions on the IIoT platform console. The device establishes a connection with the platform through various protocols supported by the IIoT platform, such as MQTT, CoAP, HTTPS, etc., and sends the acquired dynamic data to the IIoT platform in real time through the corresponding protocol. For example, the compressor in a chemical plant transmits dynamic information such as the pressure value monitored by the internal pressure sensor and the vibration data collected by the vibration sensor to the IIoT platform periodically through the MQTT protocol.

[0025] First, preprocess the data of the industrial Internet of Things platform, including data cleaning, data formatting, and data standardization processes. Among them, data cleaning aims to remove noise, duplicate values, and invalid data in the data; data formatting is to uniformly convert data in different formats into a standard format recognizable by the system; data standardization is to normalize the data for subsequent analysis. For temperature sensor data, abnormal jump values caused by sensor failures are removed through data cleaning, the data format is unified into the form of timestamp - temperature value, and then standardized processing is carried out. The same applies to the relevant data of other devices.

[0026] Then, store the processed data in the Internet of Things data in chronological order, using the index format of "device - device static attributes - device dynamic attributes", and store the data from different data sources separately for subsequent rapid query and in-depth analysis to determine the time-series device status data mapped by static attributes.

[0027] Finally, perform device anomaly analysis on the time-series device status data mapped by static attributes. Adopt a centralized collaboration mode, the core of which consists of a device analysis agent (an agent for analyzing device static characteristics) and a multi-agent for identifying device anomaly events (an agent for identifying two-dimensional device anomaly types). The device analysis agent, as the central control node of the system, undertakes the distribution of tasks for device static data analysis and anomaly event identification; the device anomaly event identification agent is responsible for in-depth analysis of dynamic data and identification of device anomaly event categories.

[0028] The device analysis agent preferentially extracts device static attribute information from the index format of "device - device static attributes - device dynamic attributes". By analyzing the characteristics of device static data, the device analysis agent accurately locates the corresponding device anomaly event analysis agent and pushes the associated device dynamic data (operation parameters such as flow rate, pressure, and vibration value) in real time. The time-series device status data mapped by dynamic and static attributes is extracted by the static data of the device analysis agent and pushed in association with dynamic data, realizing the agent task allocation of static data under this index structure (dynamic data is pushed in association).

[0029] For dynamic data, construct an industry - device type two-dimensional division system to configure dedicated device anomaly event analysis agents for different types of industrial devices in specific scenarios, realizing two-dimensional device anomaly type identification. After determining the agent task allocation of static data, for industrial devices such as electric motors in the power industry, centrifugal pumps in the water supply industry, and compression pumps in the chemical industry, dedicated device anomaly event analysis agents are equipped. These agents take the dynamic data generated by device operation as the analysis object, such as pressure fluctuation curves, temperature change rates, and current harmonic components.

[0030] By invoking the artificial intelligence algorithm model tool, based on the feature extraction and pattern matching technology in the algorithm model, the specific type of abnormal event is identified.

[0031] Furthermore, the artificial intelligence algorithm model tool can directly transmit parameters to them and use them by invoking the API. In this application, an algorithm developed based on the random forest classification algorithm for the abnormal event types of electric motors in the power industry is invoked by the agent through the API.

[0032] Step 102: Obtain the original knowledge base, and based on the original knowledge base, conduct multi-dimensional abnormal decision analysis on the device abnormal event to obtain the task execution instruction.

[0033] Exemplarily, a centralized collaboration mode is adopted to construct a multi-agent system for device abnormal event decision-making as the executor of multi-dimensional abnormal decision analysis. Its core consists of a severity decision-making agent, a root cause tracing decision-making agent, a potential impact decision-making agent, and a disposal plan decision-making agent. The severity, root cause tracing, and potential impact decision-making agents are responsible for multi-dimensional decision-making on device abnormal events. The disposal plan decision-making agent serves as the system central control node, undertaking to summarize multi-dimensional decisions and generate the final disposal plan for device abnormal events. Through multi-dimensional abnormal decision analysis.

[0034] Among them, the task execution instruction includes: device automatic execution task instruction, operation and maintenance personnel execution task instruction.

[0035] Specifically, based on the original knowledge base, conducting multi-dimensional abnormal decision analysis on the device abnormal event to obtain the task execution instruction includes: conducting potential impact analysis on the device abnormal event, and based on the analysis result of the potential impact analysis, matching the original knowledge base to obtain the first abnormal handling plan; conducting root cause tracing analysis on the device abnormal event, and according to the analysis result of the root cause tracing analysis, matching the original knowledge base to obtain the second abnormal handling plan; conducting severity analysis on the device abnormal event, and based on the analysis result of the severity analysis, matching the original knowledge base to obtain the third abnormal handling plan; according to the first abnormal handling plan, the second abnormal handling plan, and the third abnormal handling plan, through multi-source execution step allocation processing, obtaining the task execution instruction.

[0036] Figure 2 This is the overall architecture diagram of a multi-agent-based industrial device abnormal event management method provided by an embodiment of this application. In one embodiment, when an abnormal event is analyzed, the severity, root cause tracing, and potential impact decision-making agents are enabled to conduct a full-dimensional analysis from risk assessment, cause location to impact deduction, and plan the handling plan accordingly.

[0037] The potential impact analysis is executed by the potential impact decision agent, which deduces the potential impacts of abnormal events on the equipment level, production line level, and enterprise level, formulates countermeasures for the potential impacts, and determines the first abnormal handling plan.

[0038] The root cause tracing analysis is executed by the root cause tracing decision agent, which needs to complete root cause reasoning, and plan the equipment abnormal event handling plan according to the cause of the abnormal event, and determine the second abnormal handling plan.

[0039] The severity analysis is carried out by the severity decision agent, which is responsible for generating the abnormal event severity index and matching the response level based on the analytic hierarchy process and multi-dimensional analysis indicators, outputting the response level, the weight distribution table of each evaluation dimension according to the preset industrial equipment abnormal response specification, and step by step outputting the equipment abnormal event handling plan planned according to the response level to determine the third abnormal handling plan.

[0040] After determining the first abnormal handling plan, the second abnormal handling plan, and the third abnormal handling plan, the disposal plan decision agent conducts the final planning of the abnormal event handling plan, and distributes the processing through multi-source execution steps to obtain the execution task instructions.

[0041] It should be noted that the execution task instructions include: equipment automation execution task instructions and operation and maintenance personnel execution task instructions. That is, after the analysis by the severity decision agent, the root cause tracing decision agent, the potential impact decision agent, and the disposal plan decision agent, the output disposal plan decision is on the one hand for the treatment of the equipment itself (parameter adjustment, equipment status adjustment, etc.); on the other hand, it is the disposal strategy of the operation and maintenance personnel for this abnormality. Through these two types of disposal plan decisions, the problem of incomplete generation results of equipment abnormal disposal plans in the existing technology for complex industrial scenarios is made up for.

[0042] Furthermore, based on the actual operation scenario of the main roadway ventilation system of an intelligent mine, the core equipment is two mutually redundant HT-3000 centrifugal ventilators, which operate according to the "Mine Ventilation Safety Regulations" GB 16423-2020. Its rated power is 132kW, the maximum continuous operation duration is limited to 120 hours, and the motor winding temperature alarm threshold is set at 85°C (Class F insulation level).

[0043] Each ventilator is equipped with multiple types of sensors to monitor the operation status in real time. The vibration sensor monitors the radial and axial vibrations of the pump body, the temperature sensor collects the motor winding and bearing temperatures (range 0 - 150°C, accuracy ±0.2°C), the speed sensor captures the impeller speed (range 0 - 20000r / min, accuracy ±1%), and the current transformer monitors the operating current (range 0 - 100A, accuracy ±0.5%). The sensor data is transmitted to the edge gateway via the Modbus TCP protocol to complete ±3 After outlier cleaning and timestamp alignment processing, it is stored in the InfluxDB time series database.

[0044] When Vent_01 runs continuously for 168 hours, exceeding the scheduled duration by 20%, and the motor winding temperature climbs to 89 °C, triggering the alarm threshold, the Internet of Things platform immediately generates an abnormal event. The decision-making agent for the disposal plan, as the "ultimate formulator of the abnormal event disposal plan", quickly intervenes in the decision-making: First, integrate the results of multi-source analysis to generate an initial step-by-step disposal plan. The following are the plans generated by the severity, root cause tracing, and potential impact decision-making agents respectively: Severity assessment agent: Based on the temperature overrun amplitude and the over-duration operation status, it is determined as a level II emergency event, and an initial disposal plan is generated: 0 - 5 minutes: Start the standby ventilator Vent_02 and gradually reduce the operating power of Vent_01; 5 - 10 minutes: Complete the shutdown of Vent_01 to ensure that the roadway ventilation volume is maintained at a safe standard; 10 - 120 minutes: Organize professional personnel to conduct fault troubleshooting and repair, and closely monitor the operating status of Vent_02 during the repair period.

[0045] Root cause tracing agent: By analyzing the operating duration, temperature curve, and historical maintenance records, it locks the main cause as insulation aging due to long-term motor overload with a probability of 78%, and outputs a disposal plan: 10 - 20 minutes: Conduct an insulation resistance test on the Vent_01 motor to confirm the degree of aging; 20 - 40 minutes: Disassemble the motor and check the wear conditions of components such as windings and bearings; 40 - 100 minutes: Replace the damaged insulation materials and worn components, and reassemble the motor; 100 - 120 minutes: Conduct a performance test on the repaired motor.

[0046] Potential impact analysis agent: Predicts that if not handled in time, a winding short circuit may occur within 4 hours, resulting in the interruption of main roadway ventilation and threatening the safety of underground operations, and formulates a response plan: 0 - 30 minutes: Increase the monitoring frequency of the underground ventilation environment and monitor the concentration of harmful gases in real time; 30 - 60 minutes: Notify the underground operation personnel to make preparations for emergency evacuation and plan the evacuation route; 60 - 120 minutes: If Vent_02 shows abnormalities, start the emergency ventilation plan and activate the standby emergency ventilation equipment.

[0047] The disposal plan decision-making agent immediately retrieves the fault disposal process of the centrifugal ventilator from the "Mine Electromechanical Equipment Maintenance Specification" in the knowledge base, clarifies the maintenance technical standards and safety precautions; calls the API of the Internet of Things platform to send instructions to the DCS system, and completes the seamless switching of Vent_01's safe shutdown and Vent_02 within 5 minutes to ensure uninterrupted roadway ventilation; at the same time, retrieves the suitable motor insulation material (model: JY - 03) and maintenance tool kit through the API of the intelligent warehousing system, links with the personnel scheduling system, accurately assigns an engineer team with mine electromechanical maintenance qualifications according to the maintenance personnel skill matrix, and pushes the AR maintenance guide containing the fault details, disposal steps, and safety regulations through the enterprise internal communication platform.

[0048] Finally, generate and execute the final plan: Complete the ventilator switching within 0 - 5 minutes; Complete the spare parts out-of-storage and transportation within 10 - 25 minutes; Execute the motor insulation replacement and parameter debugging operations within 25 - 90 minutes; Conduct equipment performance testing and ventilation effect verification within 90 - 100 minutes. Each link is strictly implemented in accordance with the industry standard terms to ensure the efficient disposal of abnormal events and minimize safety risks and production losses.

[0049] Step 103: Based on the equipment automation, execute the task instructions, and obtain the first work order data through the completion processing of the equipment work order.

[0050] Exemplarily, the work order file is the core process carrier and management tool, used to systematically record the full life cycle information of equipment faults from occurrence to closure, ensuring that events are traceable, responsibilities are definable, and improvements can be implemented. The decision-making instructions given by the disposal plan decision-making agent; there are two execution methods, namely, executed automatically by the Internet of Things device and executed by the operation and maintenance personnel. It is necessary to generate work orders for these two types of execution methods respectively to complete the management of abnormal events. For the equipment to execute task instructions automatically, it is necessary to complete the processing of the equipment work order to realize the automatic completion of the equipment work order and improve the completion efficiency of the equipment work order.

[0051] Specifically, based on the equipment automation to execute the task instructions, through the completion processing of the equipment work order, the first work order data is obtained, including: performing the execution data processing of the Internet of Things device on the equipment automation to execute the task instructions to obtain the Internet of Things device execution work order; sending the Internet of Things device execution work order to the execution end, and monitoring the execution process of the execution end to obtain the execution exception data; based on the execution exception data, determining the first work order update data through the execution status analysis; updating the first work order update data to the Internet of Things device execution work order to obtain the first work order data.

[0052] In one embodiment, first, a preliminary work order before decision execution is generated. When the agent analyzes the device abnormal event and gives a decision, the device work order generation agent immediately generates a preliminary work order. The work order at this stage includes basic abnormal information (device name, abnormal type, occurrence time) and the agent's decision-making scheme, providing a framework for subsequent execution and recording. After the device receives the instruction from the disposal scheme decision-making agent, it collects execution data in real time through the Internet of Things platform and transmits it to the work order execution monitoring agent. The transmitted content includes: Execution status: whether the instruction is successfully received, starts execution, execution is completed, execution fails; Parameter changes: key parameters of the device before and after execution (such as voltage, temperature, rotation speed); Execution time consumption: the time interval from the instruction issuance to the execution completion.

[0053] The device work order completion process is carried out through the device execution monitoring agent. According to the device execution result, the work order execution monitoring agent automatically supplements the work order content: Execution is successful: supplement the device status data after execution and the parameter recovery situation, and generate a conclusion of "resolved".

[0054] Execution fails: If new abnormalities occur during the execution process or the original disposal scheme fails, the work order execution monitoring agent immediately sends detailed feedback information to the abnormal decision-making agent. Record the failure error code and the device feedback log. The abnormal decision-making agent re-evaluates and adjusts the disposal scheme based on the new data to obtain the first work order data.

[0055] Taking the abnormal temperature of an air conditioner device as an example: The agent detects that the air conditioner temperature continuously exceeds the set threshold, generates a preliminary work order, and the decision is to "increase the cooling power".

[0056] The agent sends an instruction to the air conditioner device. After the device receives it, it executes and transmits the power adjustment process data in real time.

[0057] If the temperature returns to normal, the agent completes the work order and records the temperature data after execution and the power adjustment effect; if the temperature does not drop, supplement the failure information and trigger the on-site processing work order for the operation and maintenance personnel.

[0058] Step 104: According to the task instruction executed by the operation and maintenance personnel, through the operation and maintenance dialogue completion process, obtain the second work order data.

[0059] Exemplarily, for the task instruction executed by the operation and maintenance personnel, multiple rounds of dialogue are required to help the operation and maintenance personnel complete the work order. Through the operation and maintenance dialogue completion process, the update and completion of the work order during the process of the operation and maintenance personnel executing the corresponding instruction are realized, improving the work order completion efficiency of the operation and maintenance personnel.

[0060] Specifically, according to the task instructions executed by the operation and maintenance personnel, through operation and maintenance dialogue completion processing, the second work order data is obtained, including: pre-filling the work order for the task instructions executed by the operation and maintenance personnel to obtain the work order for the operation and maintenance personnel to execute the task; sending the work order for the operation and maintenance personnel to execute the task to the execution end, and monitoring the execution process of the execution end to obtain execution exception data; based on the execution exception data, determining the second work order update data through work order filling interactive guidance; updating the second work order update data to the work order for the operation and maintenance personnel to execute the task to obtain the second work order data.

[0061] Exemplarily, first, obtain the disposal plan output by the exception decision agent, extract the key information, and automatically fill in the basic information of the exception event (event number, occurrence time, involved equipment), severity level, root cause analysis result, steps of the disposal plan, estimated completion time, executor information, required resource list, etc. of the work order to form a draft work order.

[0062] Fields such as response and handling (response time, handler, temporary measures, spare part usage record), fault description (abnormal phenomenon, affected scope, discovery method, historical maintenance record), attachments and signatures (fault photos, test reports, responsible person's signature) need to be reserved for the operation and maintenance personnel to fill in.

[0063] The work order generation agent formats and typesets the draft uniformly, checks whether the required fields are complete and whether the data logic is reasonable, for example, confirms whether the executor matches the problem to be solved. If problems are found, it automatically returns to the exception decision agent to supplement information or correct the plan until the work order information is complete, accurate, and standardized.

[0064] Then, the work order execution monitoring agent sends the work order passed the review to the corresponding operation and maintenance personnel or equipment execution end, notifies them to start executing the task, and starts the execution monitoring process. The operation and maintenance personnel receive the work order through the system interface, and the work order execution monitoring agent guides the operation and maintenance personnel to supplement the information that needs to be manually confirmed in the work order in the form of a dialog box or prompt, such as the details of the actual processing steps, additional temporary measures taken, on-site photos, etc. The agent identifies the key information through multiple rounds of dialogue and automatically fills it into the corresponding fields of the work order.

[0065] Next, after the work order is executed, the operation and maintenance personnel check the content of the work order. After confirming that the execution information, resource usage, processing results, etc. are correct, they perform an electronic signature confirmation and submit the work order. The work order execution monitoring agent verifies the integrity of the submitted content. If there is missing information, it returns a prompt asking for supplementation to ensure the complete recording of the entire work order process, and then archives the final work order. Finally, if a new exception occurs during the execution process or the original disposal plan fails, the work order execution monitoring agent immediately sends the detailed feedback information to the exception decision-making agent. The exception decision-making agent re-evaluates and adjusts the disposal plan based on the new data. Subsequently, the work order generation agent synchronously modifies the work order content according to the updated plan, including disposal steps, resource requirements, etc., to ensure that the work order is consistent with the actual processing flow and obtains the second work order data.

[0066] Step 105: Perform knowledge precipitation for knowledge category differentiation on the first work order data and the second work order data to determine the updated knowledge, and upload the updated knowledge to the original knowledge base to obtain the updated knowledge base.

[0067] Exemplarily, the objects of multi-agent analysis in the industrial Internet of Things will continuously change, and the knowledge base of the agent also needs to be updated. In the prior art, the update of the agent's knowledge base usually only targets the feedback of the agent's analysis results. In the complex environment of the industrial Internet of Things, the update of the knowledge base needs to be based on the analysis results obtained under different conditions as the basis for knowledge precipitation. This application performs knowledge precipitation for knowledge category differentiation on the first work order data and the second work order data to determine the updated knowledge, realizes the special knowledge precipitation for different work order data, and further improves the accuracy and practicality of the knowledge base update, providing better knowledge base data for subsequent agent analysis.

[0068] Specifically, performing knowledge precipitation for knowledge category differentiation on the first work order data and the second work order data to determine the updated knowledge includes: integrating the processing process of the first work order data to obtain the first integrated knowledge; integrating the processing process of the second work order data to obtain the second integrated knowledge; determining the first updated knowledge through explicit knowledge precipitation based on the first integrated knowledge; determining the second updated knowledge through implicit knowledge precipitation according to the second integrated knowledge, and determining the updated knowledge based on the first updated knowledge and the second updated knowledge.

[0069] Further, after uploading the updated knowledge to the original knowledge base to obtain the updated knowledge base, the method further includes: performing knowledge scope update analysis on the new knowledge base to determine the tool update data; performing tool type update on the preset tool library based on the tool update data to obtain the updated tool library.

[0070] In one embodiment, after the exception event is processed, the knowledge update agent summarizes the data of the entire processing process, covering information such as exception event characteristics, disposal plan effects, problems encountered during execution and solutions, and work order execution monitoring records.

[0071] Since the first integrated knowledge is the device policy, explicit knowledge precipitation is carried out on the first integrated knowledge to determine the first updated knowledge; the second integrated knowledge is the operation and maintenance personnel policy, and implicit knowledge precipitation is more needed to determine the second updated knowledge.

[0072] Knowledge update conducts in-depth analysis on the collected data through a large model, mines effective experiences and rules, updates new knowledge, cases, and optimized disposal strategies into the system knowledge base, provides a richer reference basis for subsequent abnormal event handling, and continuously improves the problem-solving ability of the multi-agent system.

[0073] Similarly, the tool library also needs to be updated to meet the actual demand changes in the analysis of device abnormal events in the industrial system.

[0074] The above is the method embodiment proposed in this application. Based on the same inventive concept, the embodiment of this application also provides an industrial device abnormal event management device based on multi-agent, and its structure is as Figure 3 shown.

[0075] Figure 3 It is a schematic internal structure diagram of an industrial device abnormal event management device based on multi-agent provided by the embodiment of this application. As Figure 3 shown, the device includes: At least one processor 301; And a memory 302 communicatively connected to at least one processor; Wherein, the memory 302 stores instructions executable by at least one processor, and the instructions are executed by at least one processor 301, so that at least one processor 301 can: Obtain industrial Internet of Things platform data, and identify device abnormal events for the industrial Internet of Things platform data to determine device abnormal events; obtain the original knowledge base, and based on the original knowledge base, conduct multi-dimensional abnormal decision analysis on the device abnormal events to obtain execution task instructions; wherein, the execution task instructions include: device automatic execution task instructions, operation and maintenance personnel execution task instructions; based on the device automatic execution task instructions, through device work order completion processing, obtain the first work order data; according to the operation and maintenance personnel execution task instructions, through operation and maintenance dialogue completion processing, obtain the second work order data; conduct knowledge precipitation for knowledge category differentiation on the first work order data and the second work order data to determine updated knowledge, and upload the updated knowledge to the original knowledge base to obtain an updated knowledge base.

[0076] Some embodiments of this application provide a corresponding Figure 1 non-volatile computer storage medium for industrial device abnormal event management based on multi-agent, storing computer-executable instructions, and the computer-executable instructions are set as: Obtain industrial Internet of Things (IIoT) platform data, and identify device abnormal events from the IIoT platform data to determine device abnormal events; obtain the original knowledge base, and based on the original knowledge base, conduct multi-dimensional abnormal decision-making analysis on the device abnormal events to obtain execution task instructions; where the execution task instructions include: device automated execution task instructions and operation and maintenance personnel execution task instructions; based on the device automated execution task instructions, through device work order completion processing, obtain the first work order data; according to the operation and maintenance personnel execution task instructions, through operation and maintenance dialogue completion processing, obtain the second work order data; conduct knowledge precipitation with knowledge category differentiation on the first work order data and the second work order data to determine updated knowledge, and upload the updated knowledge to the original knowledge base to obtain an updated knowledge base.

[0077] Each embodiment in this application is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the Internet of Things device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and for the relevant parts, reference can be made to the partial description of the method embodiments.

[0078] The systems and media provided by the embodiments of this application correspond one-to-one with the methods. Therefore, the systems and media also have beneficial technical effects similar to those of their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be elaborated here.

[0079] Those skilled in the art should understand that the embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0080] This application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of this application. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as the combination of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, so that the instructions executed by the processors of the computer or other programmable data processing devices generate means for implementing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0081] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the functions specified in one or more of the processes Figure 1 one or more of the processes and / or blocks Figure 1 specified in one or more of the blocks or blocks.

[0082] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one or more of the processes Figure 1 one or more of the processes and / or blocks Figure 1 specified in one or more of the blocks or blocks.

[0083] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0084] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

[0085] Computer-readable media includes both permanent and non-permanent, removable and non-removable media implemented by any method or technology for storing information. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media such as modulated data signals and carrier waves.

[0086] It should also be noted that the term "comprise", "include" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, commodity or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, commodity or device comprising the element.

[0087] The above are only embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A multi-agent-based industrial equipment abnormal event management method, characterized in that, The method includes: Obtaining industrial Internet of Things (IIoT) platform data, and identifying device anomaly events from the IIoT platform data to determine device anomaly events; Obtaining an original knowledge base, and performing multi-dimensional anomaly decision analysis on the device anomaly events based on the original knowledge base to obtain execution task instructions; wherein, the execution task instructions include: device automation execution task instructions, operation and maintenance personnel execution task instructions; Based on the device automation execution task instructions, obtaining first work order data through device work order completion processing; According to the operation and maintenance personnel execution task instructions, obtaining second work order data through operation and maintenance dialogue completion processing; Performing knowledge precipitation with knowledge category differentiation on the first work order data and the second work order data to determine updated knowledge, and uploading the updated knowledge to the original knowledge base to obtain an updated knowledge base.

2. The method for managing abnormal events of industrial equipment based on multi-agent according to claim 1, characterized in that Identifying device anomaly events from the IIoT platform data to determine device anomaly events, specifically including: Performing data preprocessing on the IIoT platform data to obtain time-series device status data with dynamic and static attribute mapping; Based on the time-series device status data with dynamic and static attribute mapping, determining device analysis agents through device static feature analysis; According to the device analysis agents, obtaining device dynamic real-time status data through device feature vector comparison; Performing two-dimensional device anomaly type identification on the device dynamic real-time status data to determine the device anomaly events; wherein, the two-dimensional device anomaly type identification includes: industry anomaly type identification, device type anomaly identification.

3. The method for managing abnormal events of industrial equipment based on multi-agent according to claim 2, wherein, Performing two-dimensional device anomaly type identification on the device dynamic real-time status data to determine the device anomaly events, specifically including: Performing random forest classification on the device dynamic real-time status data to determine anomaly category mapping; Based on the anomaly category mapping, determining the device anomaly events through anomaly category confidence analysis.

4. A method for managing abnormal events of industrial equipment based on multi - agents according to claim 1, characterized in that, Performing multi-dimensional anomaly decision analysis on the device anomaly events based on the original knowledge base to obtain execution task instructions, specifically including: Performing potential impact analysis on the device anomaly events, and based on the analysis results of the potential impact analysis, matching the original knowledge base to obtain a first anomaly handling solution; Performing root cause tracing analysis on the device anomaly events, and according to the analysis results of the root cause tracing analysis, matching the original knowledge base to obtain a second anomaly handling solution; Performing severity analysis on the device anomaly events, and based on the analysis results of the severity analysis, matching the original knowledge base to obtain a third anomaly handling solution; According to the first anomaly handling solution, the second anomaly handling solution and the third anomaly handling solution, obtaining the execution task instructions through multi-source execution step allocation processing.

5. The method for managing abnormal events of industrial equipment based on multi-agent according to claim 1, wherein, Based on the device automation execution task instructions, obtaining first work order data through device work order completion processing, specifically including: Performing Internet of Things device execution data processing on the device automation execution task instructions to obtain an Internet of Things device execution work order; Send the work order executed by the IoT device to the execution end, and monitor the execution process of the execution end to obtain execution exception data; Based on the execution exception data, determine the first work order update data through execution status analysis; Update the first work order update data to the work order executed by the IoT device to obtain the first work order data.

6. The method for managing abnormal events of industrial equipment based on multi-agent according to claim 1, wherein, According to the task execution instruction of the operation and maintenance personnel, obtain the second work order data through operation and maintenance dialogue completion processing, specifically including: Pre-fill the work order for the task execution instruction of the operation and maintenance personnel to obtain the work order for the task execution by the operation and maintenance personnel; Send the work order for the task execution by the operation and maintenance personnel to the execution end, and monitor the execution process of the execution end to obtain execution exception data; Based on the execution exception data, determine the second work order update data through work order filling interaction guidance; Update the second work order update data to the work order for the task execution by the operation and maintenance personnel to obtain the second work order data.

7. A method for managing abnormal events of industrial equipment based on multi-agent, as claimed in claim 1, wherein Perform knowledge precipitation with knowledge category differentiation on the first work order data and the second work order data to determine the updated knowledge, specifically including: Integrate the processing process of the first work order data to obtain the first integrated knowledge; Integrate the processing process of the second work order data to obtain the second integrated knowledge; Based on the first integrated knowledge, determine the first updated knowledge through explicit knowledge precipitation; According to the second integrated knowledge, determine the second updated knowledge through implicit knowledge precipitation, and determine the updated knowledge based on the first updated knowledge and the second updated knowledge.

8. A method for managing abnormal events of industrial equipment based on multi-agent according to claim 1, characterized in that, After uploading the updated knowledge to the original knowledge base to obtain the updated knowledge base, the method further includes: Perform knowledge scope update analysis on the new knowledge base to determine tool update data; Based on the tool update data, update the tool type of the preset tool library to obtain the updated tool library.

9. An industrial equipment abnormal event management device based on multi-agent, characterized in that, The device includes: At least one processor; And a memory communicatively connected to the at least one processor; Wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: Obtain industrial Internet of Things platform data, and identify device exception events for the industrial Internet of Things platform data to determine device exception events; Obtain the original knowledge base, and perform multi-dimensional exception decision analysis on the device exception events based on the original knowledge base to obtain task execution instructions; wherein, the task execution instructions include: device automation execution task instructions, operation and maintenance personnel execution task instructions; Based on the device automation execution task instructions, obtain the first work order data through device work order completion processing; According to the task execution instruction of the operation and maintenance personnel, obtain the second work order data through operation and maintenance dialogue completion processing; Perform knowledge precipitation with knowledge category differentiation on the first work order data and the second work order data to determine the updated knowledge, and upload the updated knowledge to the original knowledge base to obtain the updated knowledge base.

10. A non-volatile computer storage medium for managing abnormal events of industrial equipment based on multi-agent, storing computer-executable instructions, characterized in that, The computer executable instructions are set to: Obtain industrial Internet of Things platform data, and identify device abnormal events for the industrial Internet of Things platform data to determine device abnormal events; Obtain the original knowledge base, and based on the original knowledge base, conduct multi-dimensional abnormal decision analysis on the device abnormal events to obtain execution task instructions; wherein, the execution task instructions include: device automated execution task instructions and operation and maintenance personnel execution task instructions; Based on the device automated execution task instructions, obtain the first work order data through device work order completion processing; According to the operation and maintenance personnel execution task instructions, obtain the second work order data through operation and maintenance dialogue completion processing; Conduct knowledge precipitation with knowledge category differentiation on the first work order data and the second work order data to determine updated knowledge, and upload the updated knowledge to the original knowledge base to obtain an updated knowledge base.

Images