An agent cluster electromechanical operation and maintenance inspection method based on an industry large model drive

Abstract

An agent cluster electromechanical operation and maintenance inspection method based on an industry large model driver, relates to the technical field of electromechanical equipment operation and maintenance, and is used to improve the accuracy of large models in evaluating the operation state of electromechanical equipment. In the method, the target trigger timestamp is generated by estimating the preparation time consumption including movement and network transmission, so that the specific execution time of the agent is decoupled from the network issuing process of the instruction, thereby reducing the influence of network delay fluctuation on the timing consistency of the linkage task. Subsequently, the observation data meeting the preset synchronization condition is filtered out for uploading through the timing verification of the backhaul data. This method helps to improve the effectiveness of the data submitted to the large model for analysis, thereby providing a higher quality data basis for the evaluation of the operation state of electromechanical equipment.

Description

Technical Field

[0001] This application relates to the field of electromechanical equipment operation and maintenance technology, and in particular to an intelligent agent cluster electromechanical operation and maintenance inspection method based on an industry large model-driven approach. Background Technology

[0002] With the integration of Industrial Internet of Things (IIoT) and Artificial Intelligence (AI) technologies, the operation and maintenance (O&M) of electromechanical equipment is gradually moving towards intelligent and unmanned operation. In large-scale industrial scenarios, electromechanical equipment is diverse, widely distributed, and operates under complex conditions. Traditional single-person manual inspections are insufficient to meet the high-frequency, full-coverage O&M needs. Multi-agent cluster inspection technology driven by large models leverages the semantic understanding, logical reasoning, and task decomposition capabilities of industry-wide large models to assign complex inspection tasks to multiple physical agents (such as drones, inspection robots, and fixed sensor gateways) with different perception and execution capabilities. These agents then collaborate in physical space to achieve full-process automation and intelligence in electromechanical O&M operation and maintenance.

[0003] In relevant multi-agent inspection systems based on large models, a control architecture of centralized decision-making in the cloud and execution on the edge is typically adopted. Specifically, each agent uploads single-modal data such as equipment status and images collected on-site to the large model platform in the cloud in real time via a wireless network. The large model platform acts as a unified scheduling hub, performing comprehensive analysis and logical reasoning on the aggregated data to generate the next inspection action command (e.g., moving to device A, activating the infrared camera, or operating the test button). Subsequently, the cloud platform sends these action commands to the corresponding target agents via the network. After listening to the command network packets sent from the cloud, the control program on the agent's end immediately parses the command content and drives the hardware carriers such as chassis, robotic arms, or sensors to execute the corresponding actions, thereby completing the inspection task planned by the large model.

[0004] However, in the cloud-based centralized decision-making and edge-side execution architecture of related technologies, the instruction sequence generated by the large model is affected by factors such as wireless network fluctuations, bandwidth congestion, and cloud computing queuing during the distribution process. This causes random physical delays in the time when different agents receive the instructions. When performing linkage tests, this timing misalignment will cause the agent responsible for observation to miss the capture window of transient features, reducing the accuracy of the large model in assessing the operating status of electromechanical equipment. Summary of the Invention

[0005] This application provides a method for electromechanical operation and maintenance inspection based on an industry-wide large model-driven intelligent agent cluster, which can improve the accuracy of the large model in assessing the operating status of electromechanical equipment.

[0006] Firstly, a method for electromechanical operation and maintenance inspection based on an industry-wide large model-driven intelligent agent cluster is provided, applied to an edge collaborative controller. This method includes: receiving an electromechanical linkage inspection task, which includes a trigger action command for a first electromechanical device, an observation and acquisition command for a second electromechanical device, and a physical response delay representing the transmission of state changes from the first electromechanical device to the second electromechanical device; obtaining the maximum preparation time for the executing and observing agents, the maximum preparation time being determined based on the movement time of the executing and observing agents to the target device and network transmission delay; and generating a target trigger time based on the current time, the maximum preparation time, and a preset buffer duration. The system sets the target trigger timestamp and the sum of the target trigger timestamp and the physical response delay as the target observation time window. It then sends a trigger action command carrying the target trigger timestamp to the executing agent and an observation acquisition command carrying the target observation time window to the observing agent, so that the executing agent and the observing agent execute the corresponding commands when the local clock reaches the corresponding timestamp. After receiving the observation data with the actual acquisition timestamp returned by the observing agent and the execution confirmation information with the actual trigger timestamp returned by the executing agent, and provided that the time difference between the actual acquisition timestamp and the actual trigger timestamp meets the preset synchronization conditions, the observation data is uploaded as valid linkage feature data.

[0007] By employing the aforementioned technical solution, a target trigger timestamp is generated by estimating the preparation time, including movement and network transmission. This decouples the agent's specific execution time from the network instruction delivery process, thereby reducing the impact of network latency fluctuations on the timing consistency of coordinated tasks. Subsequently, by performing time-series verification on the returned data and selecting observation data that meets preset synchronization conditions for uploading, this approach helps improve the effectiveness of data submitted to large models for analysis, thus providing a higher-quality data foundation for evaluating the operational status of electromechanical equipment.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, the step of obtaining the maximum preparation time of the executing agent and the observing agent specifically includes: obtaining the first movement time of the executing agent to reach the first electromechanical device and the second movement time of the observing agent to reach the second electromechanical device; if it is determined that there is a trajectory intersection node based on the planned movement trajectory of the executing agent and the observing agent, determining the target agent with a longer time to reach the trajectory intersection node, and obtaining the avoidance waiting time required for the priority agent with a shorter time to pass through the trajectory intersection node; adding the avoidance waiting time to the movement time corresponding to the target agent; and adding the maximum value of the first movement time and the second movement time after the superposition process to the network transmission delay to obtain the maximum preparation time.

[0009] By adopting the above technical solution, potential trajectory intersection nodes are identified and the necessary avoidance waiting time is calculated and included in the total time. This allows the estimation of preparation time to reflect the physical interaction between agents, avoiding insufficient preset preparation time due to ignoring avoidance behavior and improving the reliability of time planning in complex path scenarios.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, the step of obtaining the avoidance waiting time required for the priority agent to pass through the trajectory intersection node with shorter time consumption specifically includes: obtaining the moving speed and vehicle length of the priority agent; calculating the passage time for the priority agent to completely leave the trajectory intersection node based on the moving speed and vehicle length; obtaining the communication handshake delay between the priority agent and the target agent; and adding the passage time and the communication handshake delay to obtain the avoidance waiting time required for the priority agent to pass through the trajectory intersection node.

[0011] By adopting the above technical solution, the waiting time is quantified based on the physical size, speed and communication protocol overhead of the intelligent agent, resulting in a more accurate avoidance waiting time.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, the step of determining the sum of the target trigger timestamp and the physical response delay duration as the target observation time window specifically includes: obtaining the physical diffusion coefficient of the state change of the first electromechanical device being transmitted to the second electromechanical device; adding the target trigger timestamp and the physical response delay duration to obtain the reference observation time point; determining the observation redundancy duration based on the physical diffusion coefficient; subtracting the observation redundancy duration from the reference observation time point as the start timestamp of the target observation time window, and adding the observation redundancy duration to the reference observation time point as the end timestamp of the target observation time window.

[0013] By adopting the above technical solution, the physical diffusion coefficient is introduced to characterize the time-broadening characteristics of the response process, and the observation redundancy duration is determined accordingly. This allows the construction of an observation window centered on the expected response time point, enabling the width of the observation window to be adjusted according to the conduction characteristics of different physical phenomena, thus more completely covering the entire time period during which state changes may occur, thereby increasing the possibility of capturing key transient features.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, the step of uploading the observation data as valid linkage feature data when the time difference between the actual acquisition timestamp and the actual trigger timestamp meets a preset synchronization condition specifically includes: calculating the actual time difference between the actual acquisition timestamp and the actual trigger timestamp; calculating the absolute value of the deviation between the actual time difference and the physical response delay duration; and determining that the time difference meets the preset synchronization condition when the absolute value of the deviation is less than or equal to the observation redundancy duration, and uploading the observation data as valid linkage feature data.

[0015] By adopting the above technical solution, the tolerance error (preset synchronization condition) in the verification process is directly adopted using the observation redundancy duration calculated based on the physical diffusion coefficient. This method establishes a verification method with consistent internal logic to determine whether the actual physical response occurs within the pre-set dynamic observation window, further improving the effectiveness of the data submitted to the large model for analysis.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, after calculating the absolute value of the deviation between the actual time difference and the physical response delay duration, the method further includes: determining that the time difference does not meet the preset synchronization condition if the absolute value of the deviation is greater than the observation redundancy duration; extracting data points characterizing the peak value of electromechanical state changes from the observation data, and using the acquisition time corresponding to the data points as the actual response timestamp; calculating the difference between the actual response timestamp and the actual trigger timestamp, and updating the difference to the physical response delay duration of the next electromechanical linkage inspection task.

[0017] By adopting the above technical solution, when the actual time difference exceeds the preset synchronization conditions, an attempt is made to extract the actual state change peak point from the observation data, and to calculate a more realistic physical response delay in this task. This delay data can be used to update the initial parameters of subsequent similar tasks, which helps the edge collaborative controller to gradually improve the accuracy of synchronization planning after multiple runs.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, the step of extracting data points characterizing the peak value of electromechanical state changes from the observation data specifically includes: calculating the data change gradient of the observation data within the target observation time window; and determining the data points corresponding to the zero-crossing points where the data change gradient changes from positive to negative as the data points characterizing the peak value of electromechanical state changes.

[0019] By adopting the above technical solution, an automated method is provided to determine the moment of response peak by calculating the gradient of data change and locating its zero-crossing point from positive to negative.

[0020] In a second aspect, embodiments of this application provide an edge co-controller, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the edge co-controller to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on an edge co-controller, cause the edge co-controller to execute the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on an edge co-controller, cause the edge co-controller to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the edge collaboration controller provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0025] 1. By estimating the preparation time, including movement and network transmission, to generate a target trigger timestamp, the specific execution time of the agent is decoupled from the network command delivery process, thereby reducing the impact of network latency fluctuations on the timing consistency of coordinated tasks. Subsequently, by performing time-series verification on the returned data, observation data that meets preset synchronization conditions is selected for uploading. This approach helps improve the effectiveness of data submitted to large models for analysis, thus providing a higher-quality data foundation for evaluating the operating status of electromechanical equipment.

[0026] 2. By introducing the physical diffusion coefficient to characterize the temporal broadening characteristics of the response process, and determining the observation redundancy duration accordingly, an observation window centered on the expected response time point is constructed. This allows the width of the observation window to be adjusted according to the conduction characteristics of different physical phenomena, more completely covering the entire time period during which state changes may occur, thereby increasing the possibility of capturing key transient features.

[0027] 3. When the actual time difference exceeds the preset synchronization condition, try to extract the actual state change peak point from the observation data, and use it to calculate the physical response delay that is closer to the reality in this task. This delay data can be used to update the initial parameters of subsequent similar tasks, which helps the edge collaborative controller to gradually improve the accuracy of synchronization planning after multiple runs. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating an intelligent agent cluster electromechanical operation and maintenance inspection method driven by an industry large model, as described in this application.

[0029] Figure 2 This is another flowchart illustrating an intelligent agent cluster electromechanical operation and maintenance inspection method driven by an industry large model, as described in this application.

[0030] Figure 3 This is a schematic diagram of the physical device structure of an edge collaboration controller in the embodiments of this application. Detailed Implementation

[0031] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to and includes any or all possible combinations of one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0033] This application provides a method for electromechanical operation and maintenance inspection based on an industry-wide large model-driven intelligent agent cluster, which can improve the accuracy of the large model in assessing the operating status of electromechanical equipment.

[0034] Please see Figure 1 This is a flowchart illustrating an intelligent agent cluster electromechanical operation and maintenance inspection method driven by an industry large model, as described in this application.

[0035] S101. Receive electromechanical linkage inspection task. The electromechanical linkage inspection task includes trigger action command for the first electromechanical equipment, observation and acquisition command for the second electromechanical equipment, and physical response delay time characterizing the transmission of the state change of the first electromechanical equipment to the second electromechanical equipment.

[0036] In this context, the electromechanical linkage inspection task refers to a set of operation instructions generated by a cloud-based large model or a local scheduling system, containing the logic for collaborative operation of multiple devices. The first electromechanical device represents the source-end physical device that needs to be stimulated or operated during the inspection process. Trigger action instructions are used to indicate control commands that instruct the agent to perform specific operations on the source-end physical device, such as pressing, turning, or energizing. The second electromechanical device represents the receiving-end physical device that is physically or electrically related to the source-end physical device. Observation and acquisition instructions are used to indicate data acquisition commands that instruct the agent to activate sensors to capture data on changes in the state of the receiving-end physical device. The physical response delay time refers to the objective time interval required for a change in the state of the source-end physical device to be transmitted to the receiving-end physical device through physical media such as mechanical, thermal, or electromagnetic forces, causing an observable change in its state. For example, if the first electromechanical device is a control switch and the second electromechanical device is a valve, the physical response delay time is the time required for the valve to actually open after the switch is activated.

[0037] To facilitate understanding, let's take a typical industrial scenario of operating a switch and observing the valve's status as an example. The first electromechanical device is the source-end physical device (such as a control switch), and the triggering action command is the instruction instructing the intelligent agent to perform operations such as pressing or tossing. The second electromechanical device is the receiving-end physical device (such as a valve) that is physically or electrically controlled by the source-end device, and the observation and acquisition command is the instruction instructing the intelligent agent to activate a camera or sensor to capture the valve's status. Due to the objective laws of the physical world, it takes a certain amount of time for the valve to actually open after the switch is activated. This objective time interval required for mechanical, thermal, or electromagnetic conduction is the physical response delay.

[0038] Specifically, the edge collaborative controller continuously listens to the communication port to obtain the latest inspection operation schedule. When it receives an electromechanical linkage inspection task, it parses the task data packet, extracts the instruction content for different electromechanical devices, and simultaneously obtains the physical response delay time preset in the task attributes.

[0039] S102. Obtain the maximum preparation time of the executing agent and the observing agent. The maximum preparation time is determined based on the travel time of the executing agent and the observing agent to reach the target device and the network transmission delay.

[0040] In this context, the executing agent refers to an intelligent robot equipped with a robotic arm or other actuator responsible for triggering actions on the first electromechanical device. The observing agent refers to an intelligent machine equipped with a camera or sensor responsible for collecting the status data of the second electromechanical device. Maximum preparation time represents the longest waiting time required for all agents participating in this collaborative task to transition from their current state to a state where they are ready to execute instructions. Movement time refers to the time required for an agent to navigate from its current physical location to the target device's operating location. Network transmission latency represents the communication time required for the edge collaborative controller to send instruction data packets to the agent's receiving end via a wireless network.

[0041] Specifically, because different agents are physically located at different points when they receive a task, and their movement speeds and the signal quality of their network environments vary, the edge collaborative controller needs to uniformly evaluate the arrival time of each agent. It calculates the expected movement time for both the executing agent and the observing agent to reach their respective target work locations, and, considering the current communication latency of the wireless network, selects the agent with the longest estimated time as the maximum preparation time for the entire system. This evaluation process ensures that the system can prevent some agents from acting prematurely and disrupting the synchronization of collaborative inspections before issuing commands.

[0042] In some embodiments, the maximum preparation time of the executing agent and the observing agent can be obtained in various ways. Optionally, the first movement time of the executing agent to the first electromechanical device and the second movement time of the observing agent to the second electromechanical device are obtained. If a trajectory intersection node is determined based on the planned movement trajectories of the executing agent and the observing agent, the target agent with a longer time to reach the trajectory intersection node is identified, and the avoidance waiting time required for the priority agent with a shorter time to pass through the trajectory intersection node is obtained. The avoidance waiting time is added to the movement time corresponding to the target agent. Finally, the maximum value of the first movement time and the second movement time after the addition process is added to the network transmission delay to obtain the maximum preparation time. Optionally, the moving speed and vehicle length of the priority agent are obtained. Based on the moving speed and vehicle length, the passage time for the priority agent to completely leave the trajectory intersection node is calculated. The communication handshake delay between the priority agent and the target agent is obtained. The passage time and the communication handshake delay are added together to obtain the avoidance waiting time required for the priority agent to pass through the trajectory intersection node. Then, the avoidance waiting time is added to the movement time corresponding to the target agent. Finally, the maximum value of the first movement time and the second movement time after the superposition process is added to the network transmission delay to obtain the maximum preparation time.

[0043] S103. Based on the current time, maximum preparation time, and preset buffer duration, generate a target trigger timestamp, and determine the sum of the target trigger timestamp and the physical response delay duration as the target observation time window.

[0044] Here, "current time" refers to the absolute time base of the local system of the edge collaborative controller. The preset buffer duration represents the additional fault tolerance time added to cope with sudden minor obstacles or system jitter during the agent's movement. The target trigger timestamp indicates the future point in time when the executing agent is required to precisely execute the trigger action. The target observation time window refers to the future time period during which the observing agent is required to activate its sensors for continuous data acquisition.

[0045] Specifically, starting from the current time, a span encompassing the maximum preparation time and the preset buffer duration is extended into the future timeline to calculate a future point in time where all agents are in place and ready. This future point is used as the target trigger timestamp. Subsequently, this target trigger timestamp is added to the physical response delay duration carried in the task to calculate the expected time when the state of the receiving device undergoes a physical change. This timeframe is then used to define the target observation time window. By pre-scheduling future execution and observation moments on the timeline, the impact of network delivery delays on the timing of action execution is decoupled.

[0046] S104. Send the trigger action command carrying the target trigger timestamp to the execution agent, and send the observation acquisition command carrying the target observation time window to the observation agent, so that the execution agent and the observation agent execute the corresponding command when the local clock reaches the corresponding timestamp.

[0047] The local clock refers to the hardware or software timer that keeps time running within each agent, and the local clock of each agent has been synchronized with the edge co-controller.

[0048] Specifically, the edge collaborative controller encapsulates the target trigger timestamp and target observation time window calculated by S103 into corresponding control command data packets, and asynchronously sends them to the executing agent and the observing agent via the wireless communication network. Upon receiving the command, the agent does not immediately execute the action, but instead parses the command and stores it in its local memory while continuously polling its local clock. When the local clock's runtime reaches the target trigger timestamp or target observation time window carried in the command, the agent's underlying control system strictly follows the timestamp's indication to drive the hardware mechanism to execute the trigger or observation action, thereby achieving high-precision spatiotemporal coordination.

[0049] S105. After receiving the observation data with the actual acquisition timestamp returned by the observation agent and the execution confirmation information with the actual trigger timestamp returned by the execution agent, the observation data is uploaded as valid linkage feature data if the time difference between the actual acquisition timestamp and the actual trigger timestamp meets the preset synchronization condition.

[0050] The actual acquisition timestamp represents the absolute time when the observing agent actually captures sensor data on-site. Observational data refers to on-site information such as images, sounds, or sensor values ​​collected by the observing agent. The actual trigger timestamp represents the absolute time when the executing agent actually completes the physical trigger action. Execution confirmation information indicates that the executing agent has reported the task as completed. The preset synchronization condition refers to the maximum tolerable error range between the actual execution time difference and the theoretical physical response delay.

[0051] Specifically, the edge collaborative controller performs timing alignment and verification on the asynchronous data transmitted back, extracting the actual acquisition timestamp and the actual trigger timestamp and calculating the time difference between them. This time difference is compared with preset synchronization conditions to verify whether the physical linkage occurs strictly according to the expected spatiotemporal logic. Only when the time difference meets the condition, indicating that the observed data accurately captures the transient physical response caused by the trigger action, will the controller mark the observed data as valid linkage feature data and upload it to the cloud-based large model. This filters out invalid or misaligned data caused by unexpected delays, improving the accuracy of the large model's evaluation.

[0052] In the above embodiments, the target trigger timestamp is generated by estimating the preparation time, including movement and network transmission. This decouples the specific execution time of the agent from the network instruction delivery process, thereby reducing the impact of network latency fluctuations on the timing consistency of the coordinated tasks. Subsequently, by performing timing verification on the returned data, observation data that meets the preset synchronization conditions is selected for uploading. This approach helps improve the effectiveness of the data submitted to the large model for analysis, thus providing a higher-quality data foundation for the evaluation of the operating status of electromechanical equipment.

[0053] However, the above embodiments primarily treat the physical response delay as a single point in time when determining the target observation time window. In some physical scenarios, state transitions are not ideal instantaneous step transitions, and the response signal may exhibit some broadening on the time axis. If the target observation time window is set improperly, or the prior knowledge of the physical response delay is not accurate enough, there is a risk that key features may not be fully captured.

[0054] Please see Figure 2 This is another flowchart illustrating an intelligent agent cluster electromechanical operation and maintenance inspection method driven by an industry large model in this application.

[0055] S201. Receive electromechanical linkage inspection task. The electromechanical linkage inspection task includes trigger action command for the first electromechanical equipment, observation and acquisition command for the second electromechanical equipment, and physical response delay time characterizing the transmission of state change of the first electromechanical equipment to the second electromechanical equipment.

[0056] S202. Obtain the maximum preparation time of the executing agent and the observing agent. The maximum preparation time is determined based on the travel time of the executing agent and the observing agent to reach the target device and the network transmission delay.

[0057] S203. Generate the target trigger timestamp based on the current time, maximum preparation time, and preset buffer duration.

[0058] Step S201 is similar to step S101, step S202 is similar to step S102, and step S203 is similar to step S103, so they will not be described again here.

[0059] S204. Obtain the physical diffusion coefficient of the state change of the first electromechanical equipment to the second electromechanical equipment.

[0060] The physical diffusion coefficient refers to a physical quantity that characterizes the degree to which the response characteristics of a first electromechanical device broaden or diffuse over time when a change in the physical state of the first electromechanical device is conducted to a second electromechanical device in a specific medium or space. Examples include the thermal diffusivity during heat conduction or the attenuation coefficient during fluid pressure transmission.

[0061] Specifically, since state transfer in the physical world is not an ideal instantaneous jump, but rather a gradual establishment process, the edge collaborative controller needs to retrieve the physical diffusion coefficient that matches the current electromechanical equipment combination from the local device ledger or cloud knowledge base. This coefficient reflects the urgency of the state change of the receiving device.

[0062] In some embodiments, the physical diffusion coefficient of the state change of the first electromechanical device transmitted to the second electromechanical device can be obtained in several ways. Optionally, a query request containing the models of the first and second electromechanical devices is sent to the cloud-based large model, the physical diffusion coefficient fed back by the cloud-based large model based on historical operation and maintenance data is received, and the physical diffusion coefficient is cached in local memory. Optionally, a locally configured electromechanical device physical attribute mapping table is read, and the corresponding physical diffusion coefficient is extracted by matching and searching the mapping table according to the topological connection relationship between the first and second electromechanical devices.

[0063] S205. Add the target trigger timestamp to the physical response delay duration to obtain the baseline observation time point.

[0064] Specifically, the edge collaborative controller performs forward extrapolation on the timeline, scalarly adding the target trigger timestamp, representing the start of the action, to the physical response delay, representing the transmission time. This calculation establishes an ideal time-series anchor point, i.e., the baseline observation time point, which represents the core moment when the observing agent is most likely to capture key linkage features.

[0065] S206. Determine the observation redundancy duration based on the physical diffusion coefficient.

[0066] The observation redundancy duration is used to represent the additional time margin required before and after the baseline observation time point in order to accommodate the diffuse distribution of physical response characteristics on the time axis.

[0067] Specifically, the edge collaborative controller uses preset mapping rules to transform the physical diffusion coefficient into a specific time-dimensional parameter. A larger diffusion coefficient indicates a smoother physical response process and a longer duration, allowing the controller to calculate a longer observation redundancy duration; conversely, a smaller diffusion coefficient results in a shorter observation redundancy duration. This step enables the observation window width to adapt to electromechanical devices with different physical conduction characteristics, avoiding missing features due to an overly narrow window or wasting storage and computing power due to an overly wide window.

[0068] In some embodiments, the determination of observation redundancy duration based on the physical diffusion coefficient can be achieved in several ways. Optionally, the physical diffusion coefficient can be input into a preset linear scaling function, and the product of the physical diffusion coefficient and the baseline redundancy coefficient can be calculated. The result of this product can then be used as the observation redundancy duration. Alternatively, a pre-established nonlinear mapping table between the physical diffusion coefficient and the time compensation amount can be looked up. The time compensation amount can be calculated by interpolation in the mapping table based on the currently acquired physical diffusion coefficient, and this time compensation amount can be determined as the observation redundancy duration.

[0069] S207. Subtract the observation redundancy duration from the baseline observation time point to obtain the start timestamp of the target observation time window, and add the observation redundancy duration to the baseline observation time point to obtain the end timestamp of the target observation time window.

[0070] The start timestamp represents the absolute time point at which the observing agent turns on its sensors and begins collecting data; the end timestamp represents the absolute time point at which the observing agent stops collecting data.

[0071] Specifically, the edge collaborative controller takes the reference observation time point as the center, subtracts the observation redundancy duration from the reference observation time point as the start timestamp of the target observation time window, and adds the observation redundancy duration to the reference observation time point as the end timestamp of the target observation time window, thereby constructing a symmetrical target observation time window on the time axis.

[0072] S208. Send the trigger action command carrying the target trigger timestamp to the execution agent, and send the observation acquisition command carrying the target observation time window to the observation agent, so that the execution agent and the observation agent execute the corresponding command when the local clock reaches the corresponding timestamp.

[0073] Step S208 is similar to step S104, and will not be described again here.

[0074] S209. After receiving the observation data with the actual acquisition timestamp returned by the observation agent and the execution confirmation information with the actual trigger timestamp returned by the execution agent, calculate the actual time difference between the actual acquisition timestamp and the actual trigger timestamp.

[0075] Specifically, after receiving the observation data with the actual acquisition timestamp returned by the observation agent and the execution confirmation information with the actual trigger timestamp returned by the execution agent, the edge collaborative controller calculates the actual time difference between the actual acquisition timestamp and the actual trigger timestamp.

[0076] S210. Calculate the absolute value of the deviation between the actual time difference and the physical response delay.

[0077] The absolute value of the deviation is used to represent the degree of deviation between the actual physical conduction time and the theoretically expected physical conduction time of the system.

[0078] Specifically, the edge collaborative controller compares the actual time difference calculated in the previous step with the theoretical physical response delay time initially set for the task, calculates the difference, and takes the absolute value of the difference. This absolute value of the deviation intuitively reflects the magnitude of the error in the time dimension of this coordinated inspection, eliminating interference from the positive and negative directions of the error.

[0079] S211. If the absolute value of the deviation is less than or equal to the observation redundancy duration, determine that the time difference meets the preset synchronization conditions, and upload the observation data as effective linkage feature data.

[0080] Specifically, the edge collaboration controller compares the absolute value of the deviation with a previously determined observation redundancy duration using a threshold. If the absolute value of the deviation does not exceed the coverage range of the redundancy duration, it indicates that although the actual physical response has slight fluctuations, it still falls entirely within the effective window of the observing agent's sensor activation. At this point, the controller determines that the spatiotemporal synchronization collaboration was successful, the observation data is real and valid, and then labels it as valid and uploads it to the cloud-based large model for further analysis.

[0081] In some embodiments, data verification and closed-loop processing can be implemented in various ways. Optionally, if the absolute value of the deviation is less than or equal to the observation redundancy duration, it is determined that the time difference meets the preset synchronization conditions and the observation data is uploaded as valid linkage feature data. If the absolute value of the deviation is greater than the observation redundancy duration, it is determined that the time difference does not meet the preset synchronization conditions. Data points representing the peak value of electromechanical state changes are extracted from the observation data, and the acquisition time corresponding to the data point is used as the actual response timestamp. The difference between the actual response timestamp and the actual trigger timestamp is calculated, and the difference is updated to the physical response delay duration of the next electromechanical linkage inspection task. Optionally, if the absolute value of the deviation is greater than the observation redundancy duration, it is determined that the time difference does not meet the preset synchronization condition. The data change gradient of the observation data within the target observation time window is calculated (for example, by performing first-order difference calculation on the observation data of the discrete time series). The data point corresponding to the zero-crossing point where the data change gradient changes from positive to negative is determined as the data point characterizing the peak value of the electromechanical state change, and the acquisition time corresponding to the data point is used as the actual response timestamp. Finally, the difference between the actual response timestamp and the actual trigger timestamp is calculated, and the difference is updated to the physical response delay duration of the next electromechanical linkage inspection task.

[0082] In the above embodiments, two supplementary designs were introduced. First, a physical diffusion coefficient was introduced to adaptively adjust the width of the target observation time window, enabling it to better match the response characteristics of different physical phenomena and more completely cover the response process. Second, a closed-loop feedback calibration mechanism based on peak detection was added, which can analyze and update the physical response delay duration parameter from the acquired data when the actual response time deviates from the expectation. These designs enable the edge collaborative controller to have parameter correction capabilities, helping it adapt to different device characteristics and improve the effectiveness of its synchronization planning during continuous operation.

[0083] The above describes an electromechanical operation and maintenance inspection method for intelligent agent clusters driven by an industry large model in the embodiments of this application. The following describes an exemplary edge collaborative controller 300 provided in the embodiments of this application.

[0084] Figure 3This is an exemplary hardware structure diagram of the edge collaborative controller 300 provided in this application embodiment. In some embodiments, the edge collaborative controller 300 is a computer device. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores data. The network interface of the computer device is used to communicate with other external terminals or servers via a network connection. In some embodiments, the network interface can be a wired network interface; in some embodiments, the network interface can also be a wireless network interface. When the computer program is executed by the processor, it implements an intelligent agent cluster electromechanical operation and maintenance inspection method based on an industry large model driven by this application embodiment.

[0085] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0086] In some embodiments of this application, a computer-readable storage medium is also provided, including instructions that, when executed on the edge collaborative controller 300, cause the edge collaborative controller 300 to execute an intelligent agent cluster electromechanical operation and maintenance inspection method based on an industry large model driven by an embodiment of this application.

[0087] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0088] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0089] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0090] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for electromechanical operation and maintenance inspection based on an industry-wide large model-driven intelligent agent cluster, characterized in that, Applied to an edge collaborative controller, the method includes: Receive electromechanical linkage inspection tasks, wherein the electromechanical linkage inspection tasks include trigger action commands for the first electromechanical equipment, observation and acquisition commands for the second electromechanical equipment, and physical response delay time characterizing the transmission of state changes of the first electromechanical equipment to the second electromechanical equipment; The maximum preparation time of the executing agent and the observing agent is obtained, and the maximum preparation time is determined based on the movement time of the executing agent and the observing agent to reach the target device and the network transmission delay; Based on the current time, the maximum preparation time, and the preset buffer duration, a target trigger timestamp is generated, and the sum of the target trigger timestamp and the physical response delay duration is determined as the target observation time window; The trigger action command carrying the target trigger timestamp is sent to the execution agent, and the observation acquisition command carrying the target observation time window is sent to the observation agent, so that the execution agent and the observation agent execute the corresponding command when the local clock reaches the corresponding timestamp; Upon receiving the observation data with the actual acquisition timestamp returned by the observation agent and the execution confirmation information with the actual trigger timestamp returned by the execution agent, the observation data is uploaded as valid linkage feature data if the time difference between the actual acquisition timestamp and the actual trigger timestamp meets the preset synchronization condition.

2. The method according to claim 1, characterized in that, The step of obtaining the maximum preparation time for the executing agent and the observing agent specifically includes: The first movement time of the executing agent to reach the first electromechanical device and the second movement time of the observing agent to reach the second electromechanical device are obtained. If a trajectory intersection node is determined based on the planned movement trajectory of the executing agent and the observing agent, the target agent with a longer time to reach the trajectory intersection node is identified, and the avoidance waiting time required for the priority agent with a shorter time to pass through the trajectory intersection node is obtained. The avoidance waiting time is added to the movement time of the target agent; The maximum value of the first and second movement times after superposition processing is added to the network transmission delay to obtain the maximum preparation time.

3. The method according to claim 2, characterized in that, The step of obtaining the avoidance waiting time required for the priority agent with shorter processing time to pass through the trajectory intersection node specifically includes: Obtain the moving speed and vehicle length of the priority agent; Based on the moving speed and the vehicle length, calculate the travel time for the priority agent to completely leave the trajectory intersection node; Obtain the communication handshake delay between the priority agent and the target agent; The passage time is added to the communication handshake delay to obtain the avoidance waiting time required for the priority agent to pass through the trajectory intersection node.

4. The method according to claim 1, characterized in that, The step of determining the sum of the target trigger timestamp and the physical response delay as the target observation time window specifically includes: Obtain the physical diffusion coefficient of the state change of the first electromechanical equipment transmitted to the second electromechanical equipment; Add the target trigger timestamp to the physical response delay duration to obtain the baseline observation time point; The observation redundancy duration is determined based on the physical diffusion coefficient. The starting timestamp of the target observation time window is obtained by subtracting the observation redundancy duration from the baseline observation time point, and the ending timestamp of the target observation time window is obtained by adding the observation redundancy duration to the baseline observation time point.

5. The method according to claim 4, characterized in that, The step of uploading the observation data as valid linkage feature data when the time difference between the actual acquisition timestamp and the actual trigger timestamp meets a preset synchronization condition specifically includes: Calculate the actual time difference between the actual collection timestamp and the actual trigger timestamp; Calculate the absolute value of the deviation between the actual time difference and the physical response delay duration; If the absolute value of the deviation is less than or equal to the observation redundancy duration, the time difference is determined to meet the preset synchronization condition, and the observation data is uploaded as valid linkage feature data.

6. The method according to claim 5, characterized in that, After the step of calculating the absolute value of the deviation between the actual time difference and the physical response delay duration, the method further includes: If the absolute value of the deviation is greater than the observation redundancy duration, it is determined that the time difference does not meet the preset synchronization condition. Extract data points representing the peak values ​​of electromechanical state changes from the observation data, and use the acquisition time corresponding to the data points as the actual response timestamp; Calculate the difference between the actual response timestamp and the actual trigger timestamp, and update the difference to the physical response delay duration of the next electromechanical linkage inspection task.

7. The method according to claim 6, characterized in that, The step of extracting data points characterizing the peak values ​​of electromechanical state changes from the observed data specifically includes: Calculate the gradient of the observed data changes within the target observation time window; The data point corresponding to the zero-crossing point where the data change gradient changes from positive to negative is determined as the data point characterizing the peak value of the electromechanical state change.

8. An edge collaboration controller, characterized in that, The edge co-controller includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the edge co-controller to perform the method as described in any one of claims 1-7.

9. A computer program product containing instructions, characterized in that, When the computer program product is run on the edge co-controller, the edge co-controller performs the method as described in any one of claims 1-7.

10. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are executed on the edge co-controller, the edge co-controller performs the method as described in any one of claims 1-7.

Images