Intelligent Inspection Method, System, Electronic Device, Storage Medium and Program Product

Abstract

The present application discloses an intelligent inspection method, system, electronic device, storage medium and program product, relating to the technical field of intelligent inspection, including splitting an inspection task into at least one subtask by a central intelligent agent and assigning it to an associated execution intelligent agent for execution, the execution intelligent agent returning target information obtained during the execution of the subtask, and the central intelligent agent constructing a directed topology graph based on the execution intelligent agents associated with the inspection task, and updating the directed topology graph according to the target information returned by the execution intelligent agent, determining the node with an in-degree of zero in the directed topology graph as the root cause node of the fault. The entire inspection process does not require manual participation. Through the collaborative work among multiple intelligent agents, the intelligent inspection of the production workshop is realized, the faults occurring in the production process can be detected in time, the inspection efficiency and accuracy are greatly improved, and the probability of missed inspection is reduced. Therefore, the technical problems of low efficiency, poor accuracy and missed inspection existing in manual inspection of the production workshop can be solved.

Description

Technical Field

[0001] This application relates to the technical field of intelligent inspection in intelligent manufacturing, and particularly to an intelligent inspection method, system, electronic device, storage medium, and program product. Background Art

[0002] Intelligent manufacturing is the deep integration of advanced manufacturing technology, new-generation information technology, and artificial intelligence technology. Intelligent manufacturing equipment refers to manufacturing equipment with functions of perception, analysis, reasoning, decision-making, and control, which comprehensively applies technologies such as control system design, sensing technology, precision manufacturing technology, and intelligent identification technology. Compared with the traditional production mode, intelligent manufacturing has the advantages of high production rate, high product quality, and high production flexibility.

[0003] Inspection is a key link in intelligent manufacturing. At present, the equipment in the production workshop is usually inspected manually on a regular basis. However, in the face of the massive and complex data in the production workshop, the manual inspection method has problems such as low efficiency, poor accuracy, and missed inspections. Summary of the Invention

[0004] This application provides an intelligent inspection method, system, electronic device, storage medium, and program product to at least solve the problems of low efficiency, poor accuracy, and missed inspections in the manual inspection of the production workshop in the related art.

[0005] This application provides an intelligent inspection method, including:

[0006] Decompose the current inspection task into at least one subtask, and determine the devices to be monitored associated with the at least one subtask;

[0007] Construct a directed topology graph according to the execution agents corresponding to the devices to be monitored, where one device to be monitored uniquely corresponds to one execution agent;

[0008] Send the at least one subtask to the corresponding execution agent, and the execution agent executes the assigned subtask and returns the target information obtained during the execution of the subtask;

[0009] Update the directed topology graph according to the target information;

[0010] When the in-degree of at least one node in the updated directed topology graph is zero, determine the at least one node as the fault root cause node corresponding to the current inspection task.

[0011] This application also provides an intelligent inspection system, including: a central intelligent agent and at least one device, where one device uniquely corresponds to one sub-intelligent agent;

[0012] The central agent is configured to: disassemble the current inspection task into at least one subtask, and determine the devices to be monitored associated with the at least one subtask; construct a directed topology graph according to the execution agents corresponding to the devices to be monitored; send the at least one subtask to the corresponding execution agent, and the execution agent executes the assigned subtask and returns the target information obtained during the execution of the subtask; update the directed topology graph according to the target information; in the case that the in-degree of at least one node in the updated directed topology graph is zero, determine the at least one node as the fault root cause node corresponding to the current inspection task;

[0013] The sub-agent is configured to: execute the assigned subtask and return the target information obtained during the execution of the subtask to the central agent.

[0014] This application also provides an intelligent inspection device, including:

[0015] A task planning module, configured to disassemble the current inspection task into at least one subtask, and determine the devices to be monitored associated with the at least one subtask;

[0016] An intelligent networking module, configured to construct a directed topology graph according to the execution agents corresponding to the devices to be monitored, where one device to be monitored uniquely corresponds to one execution agent;

[0017] A task distribution module, configured to send the at least one subtask to the corresponding execution agent, and the execution agent executes the assigned subtask and returns the target information obtained during the execution of the subtask;

[0018] An update module, configured to update the directed topology graph according to the target information;

[0019] A determination module, configured to, in the case that the in-degree of at least one node in the updated directed topology graph is zero, determine the at least one node as the fault root cause node corresponding to the current inspection task.

[0020] This application also provides an electronic device, including: a memory, configured to store a computer program; a processor, configured to implement the steps of any of the above intelligent inspection methods when executing the computer program.

[0021] This application also provides a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the steps of any of the above intelligent inspection methods are implemented.

[0022] This application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of any of the above intelligent inspection methods are implemented.

[0023] Through this application, the central agent splits the inspection task into at least one subtask and assigns it to the associated executing agent for execution. The executing agent returns the target information obtained during the execution of the subtask, and the central agent constructs a directed topology graph based on the executing agents associated with the inspection task, and updates the directed topology graph according to the target information returned by the executing agent. The node with an in-degree of zero in the directed topology graph is determined as the root cause node of the fault. The entire inspection process does not require manual participation. Through the collaborative work of multiple agents, intelligent inspection of the production workshop is realized, faults occurring during the production process can be detected in a timely manner, the inspection efficiency and accuracy are greatly improved, and the probability of missed inspection is reduced. Therefore, the technical problems of low efficiency, poor accuracy, and missed inspection existing in manual inspection of the production workshop can be solved, and the technical effects of reducing labor costs, improving inspection efficiency and quality, and enhancing the level of intelligent manufacturing can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] To more clearly illustrate the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0025] Figure 1 Shows a schematic diagram of the application environment architecture of intelligent inspection in an exemplary embodiment of the present application;

[0026] Figure 2 Is a schematic flowchart of an intelligent inspection method provided by an embodiment of the present application;

[0027] Figure 3 Is a schematic diagram of a directed topology graph in an exemplary embodiment of the present application;

[0028] Figure 4 Is a schematic flowchart of another intelligent inspection method provided by an embodiment of the present application;

[0029] Figure 5 Is a schematic diagram of the importance factor of a directed edge in an exemplary embodiment of the present application;

[0030] Figure 6 Is a schematic flowchart of yet another intelligent inspection method provided by an embodiment of the present application;

[0031] Figure 7 Is a schematic flowchart of the prediction process for one cycle in an exemplary embodiment of the present application;

[0032] Figure 8 Is a schematic flowchart of still another intelligent inspection method provided by an embodiment of the present application. Detailed implementation manners

[0033] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts belong to the protection scope of the present application.

[0034] It should be noted that in the description of the present application, the terms "include", "comprise" or any other variation thereof are intended to cover a non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article or device. The terms "first", "second", etc. in the present application are used to distinguish similar objects and not to describe a specific order or sequence.

[0035] Intelligent manufacturing equipment refers to manufacturing equipment with functions of perception, analysis, reasoning, decision-making, and control. It has characteristics such as rapid technological updates, capital intensity, and multi-field applications of products. It is a manufacturing industry with strong technical comprehensiveness, integrating multiple fields such as advanced manufacturing, information technology, and artificial intelligence, and comprehensively applying technologies such as control system design, sensing technology, precision manufacturing technology, and intelligent identification technology. Compared with the traditional production mode, intelligent manufacturing has the advantages of high production rate, high product quality, and high production flexibility.

[0036] Patrol inspection is a key link in intelligent manufacturing. Facing the vast and complex data in the factory, the traditional manual patrol inspection method is time-consuming and laborious, and cannot achieve a complete intelligent patrol inspection mode, that is, image recognition and intelligent detection, resulting in high labor costs, low efficiency, and being easily affected by human factors.

[0037] Nowadays, as the scale of automated and intelligent production factories is getting larger and the complexity of production lines is getting higher, the difficulty and cost of manual patrol inspection have increased. Problems such as low efficiency, poor accuracy, and missed inspections have gradually become key factors affecting the intelligent transformation of factories. Therefore, there is an urgent need for an efficient, automatic, and intelligent patrol inspection method to help improve the intelligent level of patrol inspection work in the production workshop, liberate manpower, reduce the false alarm rate, and meet the needs of future intelligent factories. Therefore, how to improve the intelligent level of production manufacturing patrol inspection has become a research hotspot in the current field of intelligent manufacturing.

[0038] Currently, in the related art, intelligent inspection mainly utilizes modern information technologies such as the Internet of Things, big data, and intelligent algorithms to conduct automated and intelligent inspections and monitoring of equipment, facilities, or specific areas, so as to improve the efficiency and quality of inspections. However, the existing intelligent inspection methods mainly combine manual operations with intelligent algorithms. Although the use of intelligent algorithms has improved the inspection efficiency to a certain extent, the inspection process still relies heavily on manual operations. With the increase in the number and types of detection objects, as well as the requirement for real-time performance, it has led to an increase in labor costs, and there are still serious problems with inspection efficiency and quality.

[0039] In view of the above problems, this application provides an intelligent inspection solution. On the one hand, this solution makes full use of the understanding ability of the underlying large model in the intelligent agent to analyze the input data of different modalities and obtain the status index information of the current equipment. Using the large model for information analysis and processing helps improve the inspection efficiency. On the other hand, based on the coordinated work of multiple intelligent agents, it realizes the automatic and intelligent analysis and processing of data such as the environment and equipment in the production workshop. The central intelligent agent constructs a directed topology graph of multiple intelligent agents and dynamically updates the information of nodes and edges in the directed topology graph according to the data reported by multiple intelligent agents, further analyzes the fault transmission path, determines the root cause node of the fault, and thus finds the faulty equipment, achieving intelligent inspection, reducing labor costs, improving the level of intelligent inspection, solving problems such as low efficiency and high misdetection rate in current manual inspections, and can serve the intelligent manufacturing field well and meet the inspection requirements in extreme environments.

[0040] In order to enable those skilled in the art of this technology to better understand the solution of this application, the following further elaborates on this application in combination with the accompanying drawings and specific implementation manners.

[0041] Combined with the specific application environment architecture or specific hardware architecture on which the execution of the intelligent inspection method depends, the specific application environment architecture or specific hardware architecture is described herein.

[0042] Figure 1 Shows a schematic diagram of the application environment architecture of intelligent inspection in an exemplary embodiment of this application, as Figure 1 shown, this application environment architecture includes a central intelligent agent and multiple sub-intelligent agents. Among them, one sub-intelligent agent corresponds to one device uniquely. The sub-intelligent agent can be integrated in the corresponding device or exist as an independent intelligent device. The central intelligent agent is responsible for disassembling, planning, allocating, etc. the inspection tasks according to a certain template definition. The central intelligent agent can be an intelligent inspection robot or a central control device, etc. This application does not limit the existence forms of the central intelligent agent and sub-intelligent agents. As Figure 1 shown, each sub-intelligent agent can interact with the central intelligent agent, and although Figure 1Although not shown in the figure, the sub - agents can also interact through a unified and standardized interface. Each device is equipped with relevant sensors for obtaining the operation index data and internal log data of the device. It should be noted that Figure 1 This application is explained by taking the sub - agents integrated in the device and the central agent connecting 5 devices as an example only, and it should not be regarded as a limitation to this application.

[0043] An agent, that is, an entity with intelligence, refers to an agent that can perceive the environment and take actions to achieve specific goals, has powerful understanding and processing capabilities, and has autonomy, adaptability, and interaction capabilities. An agent perceives changes in the environment (such as through sensors or data input), makes judgments and decisions based on the knowledge and algorithms it has learned, and then executes actions to affect the environment or achieve a predetermined goal. It should be noted that in this embodiment, both the central agent and the sub - agents are agents. Utilizing the powerful information understanding and processing advantages of the agents, intelligent patrol inspection of the production workshop can be realized, faults occurring in the production process can be detected in a timely manner, greatly reducing the labor cost and improving the patrol inspection efficiency and quality.

[0044] The embodiment of this application provides an intelligent patrol inspection method. Combining with the execution process of the intelligent patrol inspection method, the method is described in detail.

[0045] Figure 2 It is a schematic flowchart of an intelligent patrol inspection method provided by the embodiment of this application. This method can be executed by the central agent of the embodiment of this application and can be integrated in an electronic device.

[0046] As Figure 2 shown, this intelligent patrol inspection method includes the following steps:

[0047] Step 101: Decompose the current patrol inspection task into at least one sub - task and determine the devices to be monitored associated with at least one sub - task.

[0048] Among them, the current patrol inspection task can be the currently received patrol inspection task or the timed patrol inspection task that has reached the execution time.

[0049] In this embodiment, the central agent can decompose the current patrol inspection task into at least one sub - task and determine the devices to be monitored associated with at least one sub - task. It can be understood that the devices to be monitored are at least one of all the devices in the current patrol inspection area, and are the devices associated with the decomposed sub - tasks among all the devices. Taking Figure 1 shown as an example, the devices to be monitored are Figure 1 at least one of all the devices in

[0050] As an example, the central agent can utilize the powerful understanding and processing capabilities of the integrated task decomposition large model within itself to decompose the current inspection task according to the task description of the current inspection task, obtaining at least one subtask.

[0051] Step 102, construct a directed topology graph according to the execution agent corresponding to the device to be monitored, where one device to be monitored uniquely corresponds to one execution agent.

[0052] In this embodiment, as Figure 1 shown, one device uniquely corresponds to one sub-agent. Therefore, one device to be monitored uniquely corresponds to one sub-agent. Since the sub-agent corresponding to the device to be monitored is assigned subtasks and executes them, in this embodiment, the sub-agent corresponding to the device to be monitored is referred to as an execution agent. Then, the central agent can construct a directed topology graph according to the determined execution agent to achieve intelligent networking.

[0053] Among them, the directed topology graph is a fully connected directed graph, denoted as G(V, E), where V is the set of nodes, that is, the set of all execution agents, and each node represents an execution agent; E is the set of directed edges, and the directed edges <a1,a2> and <a2,a1> represent two different directed edges corresponding to two identical nodes (a1 and a2).

[0054] For example, assume that the set of decomposed subtasks is denoted as: {t1, t2, ……, t k}, and one subtask is associated with one device to be monitored. Thus, the set of execution agents corresponding to each subtask on the device to be monitored is denoted as: {a1, a2,……, a k}. Taking k = 5 as an example, the directed topology graph generated after networking according to the execution agents corresponding to all devices to be monitored is as Figure 3 shown.

[0055] Step 103, send at least one subtask to the corresponding execution agent, and the execution agent executes the assigned subtask and returns the target information obtained during the execution of the subtask.

[0056] In this embodiment, after the central agent determines the execution agents associated with each subtask, it can send at least one subtask to the corresponding execution agent. After receiving the subtask assigned by the central agent, the execution agent executes the subtask and obtains the target information during the execution of the subtask, and returns the target information to the central agent.

[0057] Among them, the target information can be different types of data. For different types of returned target information, the central agent processes the target information in different ways. Specific explanations will be described in detail in subsequent embodiments and will not be elaborated here.

[0058] Step 104: Update the directed topology graph according to the target information.

[0059] In this embodiment, after the central agent obtains the target information returned by the executing agent, for each terminal node in the directed topology graph, it can find all its corresponding source nodes, determine the importance factor (Importance Factor, IF) from each source node to the terminal node according to the target information, and update it, so as to realize the update of the directed topology graph.

[0060] Step 105: When the in-degree of at least one node in the updated directed topology graph is zero, determine that the at least one node is the fault root cause node corresponding to the current inspection task.

[0061] In this embodiment, after traversing all the nodes in the directed topology graph, the current update of the directed topology graph is completed. At this time, the central agent can detect whether there is a node with an in-degree of zero in the directed topology graph. Here, the in-degree refers to the sum of the number of times a certain node in the directed topology graph is the end point of an edge. In the directed topology graph, each edge has a source node and a terminal node. For any node, the in-degree refers to the number of edges with this point as the terminal node. If there is at least one node with an in-degree of zero in the directed topology graph, all the nodes with an in-degree of zero are determined as the fault root cause nodes corresponding to the current inspection task, and the device corresponding to the fault root cause node is the device detected to have a fault. Thus, the central agent can automatically generate a solution according to the relevant information of the device to be monitored corresponding to the fault root cause node.

[0062] It can be understood that in this embodiment, the process of updating the directed topology graph is repeatedly executed. The executing agent can periodically collect data and return the target information, and the central agent continuously updates the directed topology graph according to the returned target information until the fault root cause node is determined, or until the subtask execution ends and it is determined that all devices are fault-free. Thus, in an alternative embodiment of the present application, the intelligent inspection method may further include: when the in-degree of any node in the updated directed topology graph is not zero, sending an information acquisition instruction to the executing agent so that the executing agent returns new target information in response to the information acquisition instruction; furthermore, the central agent updates the directed topology graph according to the new target information and detects again whether there is a node with an in-degree of zero in the directed topology graph. Repeat the above process until there is a node with an in-degree of zero in the updated directed topology graph, or until the task execution ends.

[0063] As an example, the information acquisition instruction can be a single instruction without carrying other information, which is only used to instruct the execution agent to upload the target information obtained in a new round. After receiving the information acquisition instruction, the execution agent collects data and returns the target information according to the period and sampling frequency configured by itself.

[0064] As another example, the information acquisition instruction can be an instruction carrying the data acquisition period and the sampling frequency within the period, which is used to instruct the execution agent to collect data and return the target information according to the carried period and sampling frequency.

[0065] In this embodiment, when there is no node with an in-degree of zero in the updated directed topology graph, an information acquisition instruction is sent to the execution agent to obtain new target information, and the directed topology graph is updated according to the new target information. Thus, the continuous update of the directed topology graph is realized, which helps to detect faulty devices in a timely manner.

[0066] In the intelligent inspection method of the embodiment of the present application, the central agent splits the inspection task into at least one subtask and assigns it to the associated execution agent for execution. The execution agent returns the target information obtained during the execution of the subtask, and the central agent constructs a directed topology graph according to the execution agents associated with the inspection task and updates the directed topology graph according to the target information returned by the execution agent. The node with an in-degree of zero in the directed topology graph is determined as the root cause node of the fault. The entire inspection process does not require manual participation. Through the collaborative work among multiple agents, the intelligent inspection of the production workshop is realized, and the faults occurring in the production process can be detected in a timely manner, greatly improving the inspection efficiency and accuracy and reducing the probability of missed inspection. Therefore, the technical problems of low efficiency, poor accuracy, and missed inspection existing in manual inspection of the production workshop can be solved, and the technical effects of reducing labor costs, improving inspection efficiency and quality, and improving the level of intelligent manufacturing can be achieved.

[0067] In an alternative embodiment of the present application, in order to realize the collaborative interaction among multiple agents, it is necessary to standardize and define the mode specification of the sub-agent interaction interface. That is to say, in the embodiment of the present application, the central agent and the sub-agents corresponding to at least one device in the current inspection area interact through a unified standard interface. For example, Figure 1 taking the shown architecture as an example, the central agent and all sub-agents adopt a unified standard interface. The sub-agent can receive messages from the central agent through the interface and can also return information to the central agent. Moreover, an execution agent can also collect data from other agents or be collected data by other agents through the interface.

[0068] Among them, in the interface mode definition specification, each agent has four basic attributes: (1) type: used to indicate the agent type, and the type is Agent; (2) name: used to indicate the name of the device corresponding to the agent, which can be customized; (3) description: used to indicate the description information, function introduction, etc. of the device corresponding to the agent, which helps the central agent select the most relevant executing agent when performing task planning); (4) parameters: used to define the parameter information of the agent interface, and the number and type of parameters can be customized according to the device attributes. The required attribute in the parameters represents the list of required parameters.

[0069] Exemplarily, taking the specification in Json format to uniformly define the interface as an example, the agent interaction interface mode definition specification is as follows:

[0070]

[0071] In this embodiment, by configuring a unified specification interface for the central agent and the sub-agents, the central agent and the sub-agents, and between the sub-agents can interact through the unified specification interface, realizing the collaborative interaction between multiple agents and laying a foundation for implementing the solution of this application.

[0072] Further, in an alternative embodiment of the present application, when the central agent determines the devices to be monitored associated with each subtask, it can obtain the device description information of at least one device through the interface with the sub-agents of at least one device; then, based on the device description information, determine the devices to be monitored associated with at least one subtask.

[0073] In the embodiment of the present application, the unified specification of the exchange interface of the agent includes the attribute "description", which is used to indicate the description information, function introduction, etc. of the device corresponding to the agent. Therefore, the central agent can obtain the description information, function introduction, etc. of the corresponding device through the interfaces of each sub-agent, and then, combined with each subtask, can select the most suitable device for executing each subtask as the device to be monitored associated with the corresponding subtask. Thus, by obtaining the device description information of each device and determining the devices to be monitored associated with the subtasks according to the device description information, the accuracy of the determined devices to be monitored can be improved, laying a foundation for improving the accuracy of intelligent inspection.

[0074] In an alternative embodiment of the present application, as Figure 4 shown, based on the foregoing embodiment, step 102 may include the following sub-steps:

[0075] Step 201, traverse the executing agent, determine the currently traversed executing agent and use it as the end point.

[0076] In this embodiment, for each execution agent corresponding to the to-be-monitored devices determined, the central agent traverses each execution agent to determine the currently traversed execution agent, and uses the currently traversed execution agent as the end point.

[0077] Step 202: Use the other execution agents in the execution agent except the currently traversed execution agent as the source nodes of the currently traversed execution agent.

[0078] In this embodiment, for the determined currently traversed execution agent, all the other execution agents except the currently traversed execution agent among all the execution agents are determined as the source nodes of the currently traversed execution agent, and the currently traversed execution agent is the end point.

[0079] Step 203: Construct a directed edge from the source node to the end point, and set an initial importance factor as the edge attribute for the directed edge.

[0080] In this embodiment, for the currently traversed execution agent (i.e., the end point), after determining the corresponding source node, a directed edge from the source node to the end point can be constructed, and an initial importance factor is set as the edge attribute for the directed edge.

[0081] Exemplarily, the edge attribute of each directed edge can be initialized to 1 as the initial importance factor of the edge.

[0082] Among them, the importance factor is used as the attribute corresponding to each edge, and is used to describe the importance degree of the source node s to the end point t in an edge <s, t>. For example, Figure 5 as shown. The importance factor is asymmetric, that is, IF<s, t> and IF<t, s> are essentially different. IF<s, t> represents the importance degree of node s to node t, and IF<t, s> represents the importance degree of node t to node s.

[0083] Step 204: After traversing all the execution agents, obtain a directed topology graph.

[0084] In this embodiment, after traversing all the execution agents, the constructed directed topology graph is obtained. That is to say, when constructing the directed topology graph, each execution agent is a node in the directed topology graph, which is both an end point and a source node of other execution agents.

[0085] In the intelligent inspection method of this embodiment, after determining the execution agents associated with the subtasks, each execution agent is used as an end point, and the other execution agents are used as the source nodes of this end point and a directed edge from the source node to the end point is constructed to obtain a directed topology graph. Thus, the relationship between each execution agent can be clearly shown, which provides convenience for subsequently determining the root cause node of the fault by updating the directed topology graph.

[0086] As described above, for different types of target information, the central agent processes the received target information in different ways, which will be explained separately below.

[0087] In an alternative embodiment of the present application, the target information includes sampling data collected by the first execution agent from the second execution agent at a preset sampling frequency within at least one cycle. Herein, the first execution agent is each of all the execution agents, the second execution agent is the execution agent that interacts with the first execution agent, the number of the second execution agents can be one or more, the first execution agent performs data collection on the second execution agent for at least one cycle, and the sampling frequency within one cycle is preset in advance, and the sampling data is obtained through data collection. It can be understood that the central agent obtains the sampling data respectively returned by all the execution agents. Thus, in this embodiment, as Figure 6 shown, based on the foregoing embodiment, step 104 may include the following sub-steps:

[0088] Step 301, traverse the nodes in the directed topology graph to determine the current traversed node, the target source node corresponding to the current traversed node, and the target edge from the target source node to the current traversed node.

[0089] In this embodiment, when the central agent updates the directed topology graph according to the obtained target information (sampling data), it can traverse each node in the directed topology graph to determine the current traversed node, and according to the directed topology graph, it can determine the source node corresponding to the current traversed node (for the convenience of description and distinction, referred to as the target source node), and determine the directed edge from the target source node to the current traversed node (for the convenience of description and distinction, referred to as the target edge).

[0090] Taking Figure 3 the directed topology graph shown as an example, for the currently traversed node a3, it can be determined that the target source nodes are a1, a2, a4, and a5, and the target edges include <a1,a3>, <a2,a3>, <a4,a3>, and <a5,a3>.

[0091] Step 302, determine the target sampling data of the target execution agent corresponding to the current traversed node from the sampling data.

[0092] In this embodiment, based on the determined current traversed node and target source node, the central agent can filter out the sampling data of the target execution agent corresponding to the current traversed node (for the convenience of description and distinction, referred to as the target sampling data) from all the received sampling data, that is, the sampling data collected by the target execution agent from the execution agents corresponding to each target source node.

[0093] Exemplarily, there is a one-to-one correspondence between each node in the directed topology graph and the executing agent. According to this correspondence and the determined currently traversed node, the target executing agent corresponding to the currently traversed node can be determined, so that the sampling data returned by the target executing agent can be filtered out from the sampling data, that is, the target sampling data is obtained.

[0094] In an alternative embodiment of the present application, when the central agent determines the target sampling data, if the target sampling data corresponding to the target edge cannot be determined from the target information (i.e., the sampling data), the target edge is deleted from the directed topology graph. For example, taking Figure 3 the directed topology graph shown as an example, for the directed edge from the source node a1 to the currently traversed node a3, if the sampling data collected by the target executing agent corresponding to a3 from the executing agent corresponding to a1 cannot be determined from the sampling data, the importance factor of the source node a1 for the currently traversed node a3 cannot be determined, and then Figure 3 the directed edge from a1 to a3 in it is deleted. Thus, the update of the directed edge from the source node with missing sampling data to the currently traversed node in the directed topology graph is realized, avoiding the situation that the directed topology graph cannot be accurately updated due to lack of data, which helps to reduce the probability of missed detection.

[0095] Step 303: Determine the target importance factor of the target source node for the currently traversed node according to the target sampling data.

[0096] In this embodiment, the central agent can determine the importance factor (referred to as the target importance factor for convenience of description and distinction) of each target source node for the currently traversed node respectively according to the determined target sampling data by using its own powerful understanding and processing capabilities.

[0097] Exemplarily, an importance factor prediction model can be pre-trained to predict the importance factor of the source node for the terminal node. The input of the importance factor prediction model is the sampling data of the source node, and the output is the predicted importance factor. For example, the importance factor prediction model can be trained by using Support Vector Machines (SVM). Thus, in this embodiment, the central agent inputs the sampling data corresponding to each target source node into the importance factor prediction model, and the target importance factors of each target source node for the currently traversed node can be obtained.

[0098] In an alternative embodiment of the present application, when the central agent determines the target importance factor of the target source node for the currently traversed node based on the target sampling data, it may first determine the candidate importance factor of the target source node for the currently traversed node within at least one period according to the target sampling data, and then determine the target importance factor of the target source node for the currently traversed node according to the candidate importance factor. That is to say, in this embodiment, for the obtained target sampling data within at least one period, the importance factor of the target source node for the currently traversed node within the corresponding period (referred to as the candidate importance factor) is determined respectively according to the sampling data within each period. For example, the sampling data within one period can be input into a pre-trained importance factor prediction model to obtain the candidate importance factor for the corresponding period. Among them, when there are multiple target source nodes, the sampling data of multiple target source nodes within the same period can be input into the importance factor prediction model simultaneously, and the candidate importance factor of each target source node for the currently traversed node within the corresponding period can be obtained. Then, according to the candidate importance factor of each period, the final target importance factor of each target source node for the currently traversed node is determined. Thus, by first determining the candidate importance factor of the target source node for the currently traversed node within each period respectively, and then determining the target importance factor according to the candidate importance factors of each period, it helps to ensure the reliability and accuracy of the finally determined importance factor, thereby helping to improve the accuracy of fault inspection and patrol.

[0099] As an example, when determining the target importance factor according to the candidate importance factor, if the period is one, that is, the target acquisition data only includes the data collected within one period, the candidate importance factor can be determined as the target importance factor; if the period is multiple, then for the same target source node, the average value of the candidate importance factors of this target source node for the currently traversed node within multiple periods can be determined, and then the obtained average value is determined as the target importance factor of this target source node for the currently traversed node. For example, assuming the period is 3, for a target source node, calculate the average value of the candidate importance factors of this target source node in these 3 periods, and use the obtained average value as the target importance factor of this target source node for the currently traversed node. By calculating the average value of the candidate importance factors of the same target source node in each period as its target importance factor for the currently traversed node when the period is multiple, it helps to ensure the rationality of the determined target importance factor, thereby helping to improve the accuracy of fault inspection and patrol.

[0100] Step 304, update the target edge in the directed topology graph based on the target importance factor.

[0101] In this embodiment, after the central agent determines the target importance factor of the target source node for the currently traversed node, it can update the target edge in the directed topology graph based on the determined target importance factor, so as to realize real-time updating of the directed topology graph according to the data collected by the execution agent, realize the state update between the devices to be monitored, and find the faulty device.

[0102] In an alternative embodiment of the present application, when updating the target edge of the directed topology graph, the target importance factor can be compared with a preset importance factor threshold. The importance factor threshold can be preset according to actual needs. For example, the importance factor threshold is set to 0.2. When the target importance factor is greater than or equal to the importance factor threshold, update the edge attribute of the target edge in the directed topology graph to the target importance factor and retain the target edge. When the target importance factor is less than the importance factor threshold, delete the target edge in the directed topology graph.

[0103] It can be understood that when updating the target edge in the directed topology graph, traverse the target importance factor of each target source node for the currently traversed node one by one. If the target importance factor is greater than or equal to the importance factor threshold, retain the directed edge from the target source node to the currently traversed node in the directed topology graph and update its edge attribute to the corresponding target importance factor; otherwise, delete the directed edge from the target source node to the currently traversed node in the directed topology graph. For example, taking the Figure 3 directed topology graph shown as an example, the target source node is a1, the currently traversed node is a3, and the determined target importance factor of a1 for a3 is 0.5, which is greater than the importance factor threshold of 0.2, then retain the Figure 3 directed edge from a1 to a3; if the target importance factor of a1 for a3 is 0.1, which is less than the importance factor threshold of 0.2, then delete the Figure 3 directed edge from a1 to a3.

[0104] In this embodiment, by comparing the determined target importance factor with the importance factor threshold, only the directed edges in the directed topology graph with a target importance factor greater than or equal to the importance factor threshold are retained, and the directed edges with a target importance factor less than the importance factor threshold are deleted. Thus, only the directed edges with a relatively large correlation are retained in the updated directed topology graph, realizing the reduction and simplification of the directed topology graph, and providing traversal for determining the faulty root cause node through the directed topology graph.

[0105] In the intelligent patrol method of this embodiment, the execution agent returns the sampling data collected at a preset sampling frequency within at least one cycle to the central agent. The central agent determines the target sampling data of the target execution agent corresponding to the current traversed node from the sampling data, and then determines the target importance factor of the target source node for the current traversed node according to the target sampling data and updates the target edge in the directed topology graph. Thus, the execution agent only needs to collect data according to the cycle and sampling frequency, without analyzing and processing the collected data, thereby reducing the algorithm complexity and processing difficulty of the execution agent.

[0106] In an alternative embodiment of the present application, the target information includes an importance factor. Among them, the third execution agent collects data from the fourth execution agent at a preset sampling frequency within at least one cycle, determines the importance factor of the fourth execution agent for the third execution agent according to the collected data, and returns the importance factor to the central agent. The third execution agent is each of all the execution agents, the fourth execution agent is the execution agent that interacts with the third execution agent, the number of the third execution agents can be one or more, the third execution agent performs data collection on the fourth execution agent for at least one cycle, the sampling frequency within one cycle is preset in advance, and the importance factor of each fourth execution agent for the third execution agent is determined according to the collected data. It can be understood that the central agent obtains the importance factors respectively returned by all the execution agents.

[0107] That is to say, in this embodiment, each execution agent collects data from other execution agents, determines the importance factor of other execution agents for itself according to the collected data, and then feeds it back to the central agent. The central agent updates the directed topology graph according to the received importance factor.

[0108] Exemplarily, when each execution agent is used as an end point to calculate the importance factor of other execution agents for it, an artificial intelligence prediction model can be used to complete it. Among them, the artificial intelligence prediction model is pre-trained. Its input is the characteristics of the sampling data of the source node s (that is, corresponding to a fourth execution agent) under the preset sampling frequency conditions within one cycle. The dimension of the characteristics can be at least one of the mean, variance, standard deviation, maximum value, minimum value, median, peak value, skewness, etc. The output is the predicted probability vector of the data characteristics of the end point t (that is, the third execution agent) in the same sampling cycle and sampling frequency, denoted as , The dimension of is , m represents the number of dimensions of the characteristics, and the value of m can be preset according to actual needs. By using the artificial intelligence prediction model to process the input data characteristics, the efficiency of data processing can be improved.

[0109] The calculation formula of the artificial intelligence prediction model is as shown in formula (1). According to the prediction probability vector output by the artificial intelligence prediction model The calculation formula for determining the single-cycle importance factor (denoted as w) of the source node s for the terminal node t is as shown in formula (2):

[0110]

[0111] where Model represents a pre-trained artificial intelligence prediction model; the time span of the sampling period is T, and the sampling frequency within this time is denoted as N T , that is, sampling N times within one cycle; F(s) represents the characteristics of the N T sampling sequence data within the time span T of one cycle; F(t) represents the data characteristics of the predicted terminal node t, and w represents the single-cycle importance factor, that is, the importance factor determined according to the sampling data of one cycle; represents the prediction probability vector transpose vector of.

[0112] For the terminal node t with multiple in-degree edges, the characteristics of the sampling data of the source nodes of each edge can be input into the artificial intelligence prediction model at the same time. At this time, the artificial intelligence prediction model outputs the prediction probability vectors corresponding to each source node respectively. For example, assuming that the terminal node t corresponds to the source nodes s1, s2, s3, and s4, the prediction process for one cycle is as Figure 7 shown. It can be seen that the characteristics of the sampling data corresponding to each source node are input into the artificial intelligence prediction model Model to predict the data characteristics F(t) of the terminal node t in the same sampling period and sampling frequency, and the prediction vector probabilities corresponding to each source node are output , , and .

[0113] Therefore, in this embodiment, when the third execution agent (terminal node) determines the importance factor of each corresponding fourth execution agent (source node) for it, it can collect data from each fourth execution agent according to the number of collection periods and the sampling frequency within each period. Usually, the collection frequency in each period is the same, and then the importance factor of each source node for the terminal node is determined using the collected data. The specific execution process is as follows:

[0114] Step1: Set the number of collection periods as k, and the sampling frequency in the i-th period as N i .

[0115] Step2: Let i = 1.

[0116] Step 3: Collect the data of the i-th cycle from each fourth execution agent, and use the above artificial intelligence prediction model and formula (2) to determine the single-cycle importance factor of each fourth execution agent for the third execution agent, denoted as IF1 = { , , ……, }, where g is the number of fourth execution agents, represents the single-cycle importance factor of the fourth execution agent g in the first sampling cycle. Optionally, the central agent can send the identification information of other execution agents associated with the execution agent, that is, the fourth execution agent corresponding to the execution agent, to the execution agent according to the source node corresponding to each node as the end point in the directed topology graph. The number of fourth execution agents represents the in-degree of the end point in the directed topology graph.

[0117] Step 4: i = i + 1.

[0118] Step 5: Loop and execute Step 3 until i = k to obtain the single-cycle importance factor of each fourth execution agent in the k-th cycle, that is, IF k ={ , , ……, }.

[0119] Step 6: Calculate the final importance factor (denoted as IF) of each fourth execution agent for the third execution agent in k cycles. The calculation formula is as shown in formula (3):

[0120]

[0121] That is to say, for a fourth execution agent, its final importance factor for the third execution agent is the average value of its single-cycle importance factors in k cycles.

[0122] Through the above steps Step 1 - Step 6, the importance factor of each fourth execution agent associated with a third execution agent for the third execution agent can be obtained, and the importance factor can be returned to the central agent. Thus, in this embodiment, as Figure 8 shown, on the basis of the foregoing embodiment, step 104 may include the following sub-steps:

[0123] Step 401, traverse the nodes in the directed topology graph to determine the current traversed node, the target source node corresponding to the current traversed node, and the target edge from the target source node to the current traversed node.

[0124] In this embodiment, when the central agent updates the directed topology graph according to the obtained target information (importance factor), it can traverse each node in the directed topology graph to determine the currently traversed node. And according to the directed topology graph, it can determine the source node corresponding to the currently traversed node (for the convenience of description and distinction, called the target source node), and determine the directed edge from the target source node to the currently traversed node (for the convenience of description and distinction, called the target edge).

[0125] Taking Figure 3 the directed topology graph shown as an example, for the currently traversed node a3, the target source nodes can be determined as a1, a2, a4, and a5, and the target edges include <a1,a3>, <a2,a3>, <a4,a3>, and <a5,a3>.

[0126] Step 402: Determine the target importance factor corresponding to the target edge from the importance factors according to the target edge from the target source node to the currently traversed node.

[0127] It can be understood that a target edge is associated with a target source node and the currently traversed node. According to the two nodes associated with the target edge, that is, the corresponding relationship between each node and the executing agent, the target importance factor corresponding to the target edge can be determined from the importance factors returned by each executing agent.

[0128] In an alternative embodiment of the present application, when the central agent determines the target importance factor, if the target importance factor corresponding to the target edge is not determined from the target information (that is, the importance factor returned by the executing agent), the target edge is deleted from the directed topology graph. For example, taking Figure 3 the directed topology graph shown as an example, for the directed edge from the source node a1 to the currently traversed node a3, if the importance factor of a1 for a3 is not determined from the importance factors, then Figure 3 the directed edge from a1 to a3 in it is deleted. Thus, the update of the directed edge from the source node with missing importance factor to the currently traversed node in the directed topology graph is realized, avoiding the situation that the directed topology graph cannot be accurately updated due to lack of data, which helps to reduce the probability of missed detection.

[0129] Step 403: Update the target edge in the directed topology graph based on the target importance factor.

[0130] In this embodiment, after the central agent determines the target importance factor of the target edge, it can update the target edge in the directed topology graph based on the determined target importance factor, so as to realize the real-time update of the directed topology graph according to the importance factor returned by the executing agent, and realize the state update between the devices to be monitored, so as to find the faulty device.

[0131] In an alternative embodiment of the present application, when updating the target edge of the directed topology graph, the target importance factor can be compared with a preset importance factor threshold, where the importance factor threshold can be preset according to actual needs. For example, the importance factor threshold is set to 0.2. When the target importance factor is greater than or equal to the importance factor threshold, the edge attribute of the target edge in the directed topology graph is updated to the target importance factor, and the target edge is retained. When the target importance factor is less than the importance factor threshold, the target edge in the directed topology graph is deleted.

[0132] It can be understood that when updating the target edge in the directed topology graph, the target importance factor of each target source node for the currently traversed node is traversed one by one. If the target importance factor is greater than or equal to the importance factor threshold, the directed edge from the target source node to the currently traversed node in the directed topology graph is retained, and its edge attribute is updated to the corresponding target importance factor; otherwise, the directed edge from the target source node to the currently traversed node in the directed topology graph is deleted.

[0133] In this embodiment, by comparing the determined target importance factor with the importance factor threshold, only the directed edges in the directed topology graph with a target importance factor greater than or equal to the importance factor threshold are retained, and the directed edges with a target importance factor less than the importance factor threshold are deleted. Thus, only the directed edges with a relatively large correlation are retained in the updated directed topology graph, realizing the reduction and simplification of the directed topology graph, and providing traversal for determining the root cause node of the fault through the directed topology graph.

[0134] The intelligent inspection method of this embodiment is that each executing agent collects data from other executing agents at a preset sampling frequency within at least one cycle, determines the importance factor of the fourth executing agent for the third executing agent based on the collected data, and returns it to the central agent for updating the directed topology graph. Each executing agent independently calculates the importance degree of other executing agents for itself without interference, thereby improving the calculation efficiency of the importance factor and contributing to improving the intelligent inspection efficiency.

[0135] For example, assuming that each executing agent uses a sensor to collect data information of the device to be monitored and uses a support vector machine (SVM) as a prediction model, the complete inspection process is as follows:

[0136] (1) Inspection task description: Conduct a daily routine inspection of the server production workshop.

[0137] (2) The central agent disassembles the inspection task according to the inspection task description, which involves 5 devices to be monitored, and automatically forms a directed topology graph through intelligent networking as Figure 3As shown, initialize the importance factor of each edge to 1. The execution agent on each device to be monitored analyzes and processes the data on the device to generate the data defined by the interface specification.

[0138] (3) Define the acquisition cycle time length as 5 minutes, the acquisition frequency as 20 times, define the importance factor threshold as 0.2, and define the number of acquisition cycles k = 1. Calculate the importance factor of the source node of each edge to the sink node through SVM. The final importance factor after k cycles can be represented by a matrix as follows, where the diagonal elements are 0:

[0139] .

[0140] (4) Update the Figure 3 directed topology graph shown. Assume that in the updated graph, the node with an in-degree of 0 is a3, then a3 is the root cause node of this failure.

[0141] (5) After the central agent analyzes and processes the data and relevant information of the device corresponding to node a3, a fault solution is given.

[0142] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases, the former is a better implementation method.

[0143] The embodiments of the present application also provide an intelligent inspection system, including a central agent and at least one device. Among them, one device uniquely corresponds to one sub-agent. The architecture diagram of this intelligent inspection system is as Figure 1 shown.

[0144] Among them, the central agent is configured to: disassemble the current inspection task into at least one subtask and determine at least one device to be monitored associated with the subtask; construct a directed topology graph according to the execution agent corresponding to the device to be monitored; send at least one subtask to the corresponding execution agent, and the execution agent executes the assigned subtask and returns the target information obtained during the execution of the subtask; update the directed topology graph according to the target information; in the case that the in-degree of at least one node in the updated directed topology graph is zero, determine at least one node as the fault root cause node corresponding to the current inspection task;

[0145] The sub-agent is configured to: execute the assigned subtask and return the target information obtained during the execution of the subtask to the central agent.

[0146] It can be understood that for the specific execution operations of the central agent and the sub-agent, reference can be made to the relevant descriptions of the corresponding embodiments of the foregoing intelligent inspection method, which will not be elaborated here one by one.

[0147] An embodiment of the present application further provides an intelligent inspection device, including:

[0148] A task planning module, configured to disassemble the current inspection task into at least one subtask and determine the devices to be monitored associated with the at least one subtask;

[0149] An intelligent networking module, configured to construct a directed topology graph according to the execution agents corresponding to the devices to be monitored, where one device to be monitored uniquely corresponds to one execution agent;

[0150] A task distribution module, configured to send the at least one subtask to the corresponding execution agent, and the execution agent executes the assigned subtask and returns the target information obtained during the execution of the subtask;

[0151] An update module, configured to update the directed topology graph according to the target information;

[0152] A determination module, configured to determine that at least one node is the fault root cause node corresponding to the current inspection task when the in-degree of at least one node in the updated directed topology graph is zero.

[0153] Optionally, the central agent and the sub-agents corresponding to at least one device in the current inspection area interact through a unified standard interface.

[0154] Optionally, the task planning module is further configured to:

[0155] Obtain the device description information of at least one device through the interface with the sub-agents of the at least one device;

[0156] Determine the devices to be monitored associated with the at least one subtask according to the device description information.

[0157] Optionally, the intelligent networking module is further configured to:

[0158] Traverse the execution agents, determine the currently traversed execution agent and use it as the end point;

[0159] Use the other execution agents in the execution agents except the currently traversed execution agent as the source nodes of the currently traversed execution agent;

[0160] Construct a directed edge from the source node to the end point and set an initial importance factor as the edge attribute for the directed edge;

[0161] When the execution agents are traversed, obtain the directed topology graph.

[0162] Optionally, the target information includes sampling data collected by the first executing agent from the second executing agent at a preset sampling frequency within at least one period; the updating module includes:

[0163] The first traversing unit is configured to traverse the nodes in the directed topology graph, determine the current traversed node, the target source node corresponding to the current traversed node, and the target edge from the target source node to the current traversed node;

[0164] The first determining unit is configured to determine the target sampling data of the target executing agent corresponding to the current traversed node from the sampling data;

[0165] The second determining unit is configured to determine the target importance factor of the target source node for the current traversed node according to the target sampling data;

[0166] The first updating unit is configured to update the target edge in the directed topology graph based on the target importance factor.

[0167] Optionally, the second determining unit is further configured to:

[0168] Determine the candidate importance factor of the target source node for the current traversed node within at least one period according to the target sampling data;

[0169] Determine the target importance factor of the target source node for the current traversed node according to the candidate importance factor.

[0170] Further optionally, the second determining unit is further configured to:

[0171] For the same target source node, determine the average value of the candidate importance factors of the target source node for the current traversed node within multiple periods;

[0172] Determine the average value as the target importance factor of the target source node for the current traversed node.

[0173] Optionally, the target information includes an importance factor; wherein, the third executing agent collects data from the fourth executing agent at a preset sampling frequency within at least one period, determines the importance factor of the fourth executing agent for the third executing agent according to the collected data, and returns the importance factor to the central intelligent agent; the updating module includes:

[0174] The second traversing unit is configured to traverse the nodes in the directed topology graph, determine the current traversed node, the target source node corresponding to the current traversed node, and the target edge from the target source node to the current traversed node;

[0175] The third determining unit is configured to determine the target importance factor corresponding to the target edge from the importance factors according to the target edge from the target source node to the current traversed node;

[0176] A second update unit, configured to update a target edge in the directed topology graph based on a target importance factor.

[0177] Further optionally, the update unit (including the first update unit and the second update unit) is further configured to:

[0178] Compare the target importance factor with a preset importance factor threshold;

[0179] In the case where the target importance factor is greater than or equal to the importance factor threshold, update the edge attribute of the target edge in the directed topology graph to the target importance factor;

[0180] In the case where the target importance factor is less than the importance factor threshold, delete the target edge in the directed topology graph.

[0181] Optionally, the update module further includes:

[0182] A third update unit, configured to delete the target edge from the directed topology graph in the case where the target information corresponding to the target edge cannot be determined from the target information.

[0183] Optionally, the intelligent inspection device further includes:

[0184] An instruction sending module, configured to send an information acquisition instruction to the executing agent in the case where the in-degree of any node in the updated directed topology graph is not zero, so that the executing agent returns new target information in response to the information acquisition instruction;

[0185] The update module is further configured to: update the directed topology graph according to the new target information.

[0186] For the description of the features in the embodiments corresponding to the intelligent inspection device, reference may be made to the relevant descriptions of the embodiments corresponding to the intelligent inspection method, which will not be elaborated here one by one.

[0187] An embodiment of the present application further provides an electronic device, including a memory and a processor, where a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the embodiments of the above intelligent inspection method.

[0188] An embodiment of the present application further provides a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and the computer program is configured to execute the steps in any one of the embodiments of the above intelligent inspection method when running.

[0189] In an exemplary embodiment, the above computer-readable storage medium may include, but is not limited to: various media that can store computer programs such as USB flash drives, read-only memory (ROM for short), random access memory (RAM for short), mobile hard disks, magnetic disks, or optical discs.

[0190] An embodiment of the present application also provides a computer program product. The above computer program product includes a computer program, and when the computer program is executed by a processor, the steps in any of the above embodiments of the intelligent patrol inspection method are implemented.

[0191] An embodiment of the present application also provides another computer program product, including a non-volatile computer-readable storage medium. The non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above embodiments of the intelligent patrol inspection method are implemented.

[0192] Those skilled in the art can further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0193] The above provides a detailed introduction to an intelligent patrol inspection method provided by this application. Specific examples are used in this article to elaborate on the principle and implementation manner of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of this application, several improvements and modifications can be made to this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. An intelligent inspection method, characterized in that: Applied to a central agent, the method comprises: Decompose the current inspection task into at least one subtask, and determine the device to be monitored associated with the at least one subtask; According to the execution agent corresponding to the device to be monitored, a directed topological graph is constructed, wherein one device to be monitored uniquely corresponds to one execution agent; Sending the at least one subtask to the corresponding execution agent, wherein the execution agent executes the assigned subtask and returns target information obtained during the execution of the subtask, wherein the target information is sample data collected by the execution agent or an importance factor determined by the execution agent based on the collected sample data; The directed topological graph is updated according to the target information, wherein an importance factor of a source node in the directed topological graph to a terminal point of the source node is determined according to the target information, and the directed topological graph is updated based on the importance factor, wherein the importance factor is determined based on a product of the predicted probability vector and a transposed vector of the predicted probability vector obtained by predicting the sampling data features of the source node based on a pre-trained artificial intelligence prediction model; When the in-degree of at least one node in the updated directed topological graph is zero, determining the at least one node as a fault root cause node corresponding to the current inspection task; The process of updating the directed topology map is repeatedly executed, the execution agent periodically collects data and returns the target information, and the central agent continuously updates the directed topology map according to the returned target information until the root cause node of the fault is determined, or until the execution of the subtask is completed and it is determined that the monitored equipment is free of faults; The step of constructing a directed topology graph according to the execution agent corresponding to the device to be monitored includes: Traversing the execution agent, determining the current traversal execution agent as the end point; Using other executing agents among the executing agents except the currently traversing executing agent as the source nodes of the currently traversing executing agent; Constructing a directed edge from the source node to the end node, and setting an initial importance factor for the directed edge as an edge attribute; After traversing the execution agents, the directed topological graph is obtained.

2. The intelligent inspection method according to claim 1, characterized in that: The central intelligent agent and the sub-intelligent agent corresponding to at least one device in the current inspection area interact with each other using a unified and standardized interface.

3. The intelligent inspection method according to claim 2, characterized in that: The determining the device to be monitored associated with the at least one subtask includes: Acquiring device description information of the at least one device through an interface with a sub-agent of the at least one device; The device to be monitored associated with the at least one subtask is determined according to the device description information.

4. The intelligent inspection method according to claim 1, characterized in that: The target information includes sampled data collected by the first execution agent from the second execution agent at a preset sampling frequency within at least one cycle; The updating of the directed topological graph according to the target information includes: Traversing the nodes in the directed topological graph, determining a current traversal node, a target source node corresponding to the current traversal node, and a target edge from the target source node to the current traversal node; Determine the target sampling data of the target execution agent corresponding to the current traversal node from the sampling data; Determining, according to the target sampling data, a target importance factor of the target source node to the currently traversed node; Based on the target importance factor, the target edge in the directed topological graph is updated.

5. The intelligent inspection method according to claim 4, characterized in that: The step of determining the target importance factor of the target source node to the current traversal node according to the target sampling data includes: Determining, according to the target sampling data, a candidate importance factor of the target source node to the currently traversed node within the at least one cycle; According to the candidate importance factors, a target importance factor of the target source node to the current traversal node is determined.

6. The intelligent inspection method according to claim 5, characterized in that: There are multiple cycles, and determining the target importance factor of the target source node to the current traversal node according to the candidate importance factor includes: For the same target source node, determining an average value of the candidate importance factors of the target source node to the currently traversed node in multiple cycles; The average value is determined as a target importance factor of the target source node to the current traversal node.

7. The intelligent inspection method according to claim 1, characterized in that: The target information includes an importance factor; The third execution agent collects data from the fourth execution agent at a preset sampling frequency within at least one cycle, determines the importance factor of the fourth execution agent to the third execution agent based on the collected data, and returns the importance factor to the central agent; The updating of the directed topological graph according to the target information includes: Traversing the nodes in the directed topological graph, determining a current traversal node, a target source node corresponding to the current traversal node, and a target edge from the target source node to the current traversal node; According to the target edge from the target source node to the currently traversed node, determining a target importance factor corresponding to the target edge from the importance factors; Based on the target importance factor, the target edge in the directed topological graph is updated.

8. The intelligent inspection method according to claim 4 or 7, characterized in that: The updating of the target edge in the directed topological graph based on the target importance factor includes: Comparing the target importance factor with a preset importance factor threshold; When the target importance factor is greater than or equal to the importance factor threshold, updating the edge attribute of the target edge in the directed topological graph to the target importance factor; When the target importance factor is less than the importance factor threshold, the target edge in the directed topological graph is deleted.

9. The intelligent inspection method according to any one of claims 4 or 7, characterized in that: The method further comprises: When no target information corresponding to the target edge is determined from the target information, the target edge is deleted from the directed topological graph.

10. The intelligent inspection method according to any one of claims 1 to 7, characterized in that: The method further comprises: When the in-degree of any node in the updated directed topological graph is not zero, sending an information acquisition instruction to the execution agent, so that the execution agent returns new target information in response to the information acquisition instruction; The directed topological graph is updated according to the new target information.

11. An intelligent inspection system, characterized in that: It includes a central agent and at least one device, wherein one device uniquely corresponds to one sub-agent; The central agent is configured to: decompose the current inspection task into at least one subtask, and determine the device to be monitored associated with the at least one subtask; construct a directed topological graph according to the execution agent corresponding to the device to be monitored, wherein the execution agent is traversed, the current traversed execution agent is determined and used as the end point, and the other execution agents except the current traversed execution agent in the execution agent are used as the source node of the current traversed execution agent, and a directed edge from the source node to the end point is constructed, and an initial importance factor is set for the directed edge as an edge attribute, and the directed topological graph is obtained after traversing the execution agent; the at least one subtask is sent to the corresponding execution agent, and the execution agent executes the assigned subtask and returns the target information obtained during the execution of the subtask, wherein the target information is the sampling data collected by the execution agent or the importance factor determined by the execution agent based on the collected sampling data; according to The target information updates the directed topological graph, wherein the importance factor of the source node in the directed topological graph to the terminal point of the source node is determined according to the target information, and the directed topological graph is updated based on the importance factor, wherein the importance factor is determined based on the product of the predicted probability vector and the transposed vector of the predicted probability vector obtained by predicting the sampling data features of the source node based on a pre-trained artificial intelligence prediction model; when the in-degree of at least one node in the updated directed topological graph is zero, the at least one node is determined to be the root cause node of the fault corresponding to the current inspection task; wherein the process of updating the directed topological graph is repeatedly executed, the execution agent periodically collects data and returns the target information, and the central agent continuously updates the directed topological graph according to the returned target information until the root cause node of the fault is determined, or until the execution of the subtask is completed and it is determined that the monitored equipment is free of faults; The sub-agent is configured to: execute the assigned sub-task and return the target information obtained during the execution of the sub-task to the central agent.

12. An electronic device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the steps of the intelligent inspection method according to any one of claims 1 to 10 when executing the computer program.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of the intelligent inspection method according to any one of claims 1 to 10.

14. A computer program product, characterized in that It comprises a computer program, which, when executed by a processor, implements the steps of the intelligent inspection method according to any one of claims 1 to 10.

Images