Data mining method and system applied to hardware product control system

Abstract

The present application provides a data mining method and system for hardware product control systems. By accurately identifying software abnormal control events whose control modes meet specific target characteristics and generating a pattern control event sequence based on this, these software abnormal control events are labeled with accurate fault labels, and events that are highly related to the target fault are screened out to construct a candidate abnormal control event sequence. It not only integrates rich fault information and knowledge points, but also intuitively displays the internal connections and mutual influences between abnormal events in the form of knowledge entities and knowledge links, providing a powerful knowledge support platform for fault diagnosis experts. More importantly, through the refined feature screening and elimination of the basic knowledge network, a highly focused and refined target abnormal knowledge network is finally generated, which directly points to the core abnormal control event, greatly simplifies the troubleshooting process, and shortens the fault recovery time.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a data mining method and system applied to a hardware product control system. Background Art

[0002] In the production process of hardware products, the control system plays a vital role. It is responsible for monitoring and adjusting the operating status of production equipment to ensure the efficiency and stability of the production process. However, with the continuous expansion of production scale and the improvement of automation, the control system is becoming increasingly complex, and the frequent occurrence of software abnormal control events has become one of the key factors restricting production efficiency and quality. Traditional troubleshooting and diagnosis methods often rely on manual experience and on-site debugging, which is not only time-consuming and labor-intensive, but also difficult to fully and accurately capture the root cause of the fault, especially when faced with large amounts of complex software abnormality data.

[0003] Existing data mining technology has demonstrated powerful data processing and pattern recognition capabilities in other fields such as financial analysis and medical diagnosis, but its application in hardware control systems is still in the exploratory stage. Although some preliminary attempts have been able to identify abnormal data points in the system, they often lack in-depth analysis and knowledge integration for specific scenarios of the control system, and cannot effectively guide subsequent troubleshooting and repair work.

[0004] In particular, software abnormal control events in hardware product control systems often do not occur in isolation, and there are complex interactions and causal relationships between them. Traditional data analysis methods often ignore the inherent connections between these events and only conduct isolated analysis based on a single event, which makes it difficult to fully reveal the deep mechanism of the failure. Summary of the invention

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present application, an embodiment of the present application provides a data mining method applied to a hardware product control system, the method comprising:

[0006] Acquire a software abnormal control event sequence of a hardware product control system, and determine a software abnormal control event whose control mode meets the target control mode feature from the software abnormal control event sequence to generate a mode control event sequence; the target control mode feature represents the control mode path information included in the fault session information corresponding to the software abnormal control event in the first industrial control interval after the software abnormal control event establishes an event interaction session with other software abnormal control events;

[0007] Acquire the fault label of the control mode path information corresponding to each mode control event in the mode control event sequence, and determine the mode control event whose corresponding fault label of the control mode path information is the target fault label from the mode control event sequence to generate a candidate abnormal control event sequence;

[0008] Generate a basic abnormal knowledge network based on the candidate abnormal control events in the candidate abnormal control event sequence and the fault knowledge point information of the target fault label; the basic abnormal knowledge network includes knowledge entities corresponding to multiple software abnormal control events and one or more knowledge links;

[0009] The knowledge entities that do not meet the set feature requirements and the knowledge links corresponding to the knowledge entities that do not meet the set feature requirements in the basic abnormal knowledge network are eliminated to generate a target abnormal knowledge network, and the software abnormal control events corresponding to the knowledge entities in the target abnormal knowledge network are output as target abnormal control events.

[0010] On the other hand, an embodiment of the present application also provides an industrial control monitoring system, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to run the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0011] Based on the above aspects, the embodiment of the present application significantly improves the accuracy and efficiency of system fault diagnosis. By accurately identifying software abnormal control events whose control modes meet specific target characteristics, and generating mode control event sequences based on these software abnormal control events, the key links that may cause faults can be quickly locked, providing a clear direction for subsequent analysis. Further, by marking these software abnormal control events with accurate fault labels and screening out events that are highly related to the target faults, a candidate abnormal control event sequence is constructed, laying a solid foundation for the construction of subsequent knowledge networks. The constructed basic abnormal knowledge network not only integrates rich fault information and knowledge points, but also intuitively displays the inherent connection and mutual influence between abnormal events in the form of knowledge entities and knowledge links, providing a powerful knowledge support platform for fault diagnosis experts. More importantly, by fine-tuning the feature screening and elimination of the basic knowledge network, a highly focused and refined target abnormal knowledge network is finally generated, which directly points to the core abnormal control events, greatly simplifies the troubleshooting process, and shortens the fault recovery time. As a result, not only the fault self-diagnosis capability of the hardware product control system is enhanced, but also the operation and maintenance costs are effectively reduced, the production efficiency and system stability are improved, and a solid technical guarantee is provided for intelligent manufacturing and digital transformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic diagram of the execution flow of the data mining method applied to the hardware product control system provided in the embodiment of the present application.

[0013] Figure 2 It is a schematic diagram of the hardware architecture of the industrial control monitoring system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0014] The present application will be described in detail below with reference to the accompanying drawings. Figure 1 This is a flow chart of a data mining method applied to a hardware product control system provided by an embodiment of the present application. The data mining method applied to a hardware product control system is introduced in detail below.

[0015] Step S110, obtaining a software abnormal control event sequence of a hardware product control system, and determining a software abnormal control event whose control mode meets the target control mode feature from the software abnormal control event sequence, and generating a mode control event sequence. The target control mode feature represents the control mode path information included in the fault session information corresponding to the software abnormal control event in the first industrial control interval after the software abnormal control event establishes an event interaction session with other software abnormal control events.

[0016] In detail, the software abnormal control event sequence refers to a series of software abnormal events recorded by the monitoring system during the operation of the hardware product control system. These software abnormal events may involve control program loss of response, parameter abnormality, equipment failure, etc., which can be used as input data for fault diagnosis and provide a basis for subsequent analysis and processing.

[0017] Control mode refers to the specific mode or logical process followed by a specific device or component in a control system when performing a control task, which is used to identify and analyze the state and behavior mode of the system when an abnormal event occurs. The target control mode characteristics refer to a set of specific mode characteristics that the hardware control system should exhibit after an abnormal event occurs. These target control mode characteristics are used to filter out events directly related to the target problem from the abnormal event sequence, and can be used as screening criteria to help identify which abnormal events are most likely to be related to the target problem (such as a specific fault).

[0018] The fault session information refers to a collection of information about the cause, impact, and treatment measures of the event recorded during the occurrence of the abnormal event, which is used to provide detailed context information about the software abnormal control event, which helps to deeply understand the event and formulate corresponding treatment strategies. The control mode path information refers to all control logic and path information from the triggering of the software abnormal control event to the final impact, including sensor signals, controller processing, actuator actions, etc., which is used to reveal how the abnormal event affects the system state and provide a basis for fault location and cause analysis.

[0019] In this embodiment, assume a hardware product manufacturing factory whose control system is managed and monitored by a server. Over a period of time, the server records a series of software abnormal control events, such as:

[0020] Event 1: The pressure control program of the stamping machine suddenly lost response during operation, resulting in an abnormal increase in stamping pressure.

[0021] Event 2: An error occurred in the tool change program of the CNC machine tool and the tool could not be changed correctly.

[0022] Event 3: The speed control module on the conveyor belt experienced frequent fluctuations, resulting in unstable product conveying speed.

[0023] The server first obtains the software abnormal control event sequence, and then defines the target control mode feature as "the control mode path information contains specific key parameter changes within the first industrial control interval after establishing an event interaction session with other software abnormal control events".

[0024] Taking event 1 as an example, the server analyzes the control mode path information in the first industrial control interval (for example, the first 5 minutes of the press machine operation) after the interactive session with other potentially related events (such as the overload protection event of the equipment). It is found that the pressure setting parameters in the pressure control program have undergone abnormal and large changes, which is consistent with the characteristics of the target control mode.

[0025] For event 2, in the first industrial control interval after analyzing its interaction with related events (such as tool inventory management events), no characteristic changes were found in the key parameters in the control mode path.

[0026] For event 3, the server found that within the first industrial control interval after the conveying speed control module interacted with other control modules (such as the product detection module), the feedback adjustment parameters of the speed control frequently underwent abnormal adjustments, which was consistent with the characteristics of the target control mode.

[0027] Through such analysis, the server determines that Event 1 and Event 3 are software abnormal control events whose control modes meet the characteristics of the target control mode, and generates a mode control event sequence.

[0028] Step S120, obtaining the fault label of the control mode path information corresponding to each mode control event in the mode control event sequence, and determining from the mode control event sequence the mode control event whose corresponding fault label of the control mode path information is the target fault label, to generate a candidate abnormal control event sequence.

[0029] In detail, the fault label is the result of classifying and labeling the software abnormal control event, and is used to quickly identify the event type and possible fault cause of the software abnormal control event. The candidate abnormal control event sequence is an abnormal event sequence that has been preliminarily screened, meets the specific fault characteristics, and whose fault label matches the target fault label.

[0030] In this embodiment, for the mode control event sequence (including event 1 and event 3) generated in step S110, the server performs the following operations:

[0031] For event 1, the server determines the relevant control parameters and signal flows based on the corresponding control mode path information. The basic fault description report information corresponding to each control mode path information is obtained from the fault recovery process corresponding to these control mode path information. For example, the obtained report information indicates that the pressure sensor failure causes the pressure control signal to be abnormal. The server classifies and identifies this basic fault description report information and marks it as a fault label of the "sensor failure" type.

[0032] For event 3, the server also obtains relevant information in the fault recovery process corresponding to its control mode path information. Assuming that the information obtained indicates that the speed fluctuation is caused by the parameter setting error in the control algorithm, the server marks its fault label as "control algorithm parameter error".

[0033] Set the target fault labels to "sensor fault" and "control algorithm error". Since the fault label of event 1 is "sensor fault" and the fault label of event 3 is "control algorithm parameter error", both of them meet the target fault labels, the server determines event 1 and event 3 as candidate abnormal control events and generates a candidate abnormal control event sequence.

[0034] Step S130: Generate a basic abnormal knowledge network based on the candidate abnormal control events in the candidate abnormal control event sequence and the fault knowledge point information of the target fault label. The basic abnormal knowledge network includes knowledge entities corresponding to multiple software abnormal control events and one or more knowledge links.

[0035] In detail, the basic abnormal knowledge network is a knowledge network constructed based on candidate abnormal control events and fault knowledge point information, which contains multiple knowledge entities (representing abnormal events) and knowledge links (representing the relationship between events), which is used to provide association information and fault knowledge points between candidate abnormal control events, and support deeper fault analysis and knowledge sharing. In the basic abnormal knowledge network, the knowledge entity represents the basic unit of a specific abnormal event or fault knowledge point, which serves as a node in the basic abnormal knowledge network to store and represent specific abnormal event or knowledge point information. The knowledge link is an edge connecting two knowledge entities, indicating a certain relationship between them (such as linkage triggering, causal relationship, etc.), which is used to reveal the intrinsic connection and mutual influence between abnormal events, and support the diagnosis and analysis of complex faults. The fault knowledge point information of the target fault label refers to the known fault knowledge points associated with the target fault label, including the cause of the fault, the impact, the treatment measures, etc., to provide theoretical support and solution reference for fault diagnosis.

[0036] In this embodiment, in the candidate abnormal control event sequence, there are event 1 (sensor failure of punching machine pressure control) and event 3 (algorithm parameter error of conveyor belt speed control).

[0037] The server first constructs the first abnormal knowledge network based on event 1 and event 3. In the first abnormal knowledge network, a corresponding knowledge entity is created for each candidate abnormal control event. For example, a knowledge entity "stamping machine pressure sensor failure entity" is created for event 1, and a knowledge entity "conveyor belt speed control algorithm error entity" is created for event 3.

[0038] At the same time, the server determines the software abnormal control events that have a linkage trigger relationship with each candidate abnormal control event. Assuming that event 1 has a linkage relationship with the abnormality triggered by the overload protection mechanism of the equipment, a knowledge entity "stamping machine overload protection abnormality entity" is created for this linkage relationship, and a knowledge link is established in the first abnormal knowledge network to reflect the linkage trigger relationship between "stamping machine pressure sensor fault entity" and "stamping machine overload protection abnormality entity".

[0039] For event 3, assuming that it has a linkage relationship with the misjudgment of the product detection module, create a knowledge entity "product detection misjudgment entity" and establish a corresponding knowledge link.

[0040] Then, the server generates a second abnormal knowledge network based on the fault knowledge point information of the target fault label (such as "sensor fault" and "control algorithm error"). It is assumed that the fault knowledge point information such as common causes and solutions of "sensor fault" and "control algorithm error" has been stored in the system. The server creates corresponding knowledge entities for these knowledge point information, such as "common sensor fault cause entity", "sensor fault solution entity", "control algorithm error common type entity", "control algorithm error repair strategy entity", etc. Knowledge links are established in the second abnormal knowledge network to reflect the association between these knowledge entities. For example, there is an association between the "common sensor fault cause entity" and the "sensor fault solution entity", indicating that understanding the cause helps find a solution.

[0041] Finally, the server integrates the first abnormal knowledge network and the second abnormal knowledge network according to the knowledge entity that the corresponding software abnormal control event in the first abnormal knowledge network and the second abnormal knowledge network is the same software abnormal control event, and generates a basic abnormal knowledge network. For example, if the "stamping machine pressure sensor failure entity" in the first abnormal knowledge network is associated with a certain knowledge entity about "sensor failure" in the second abnormal knowledge network, the server integrates the two parts and establishes the corresponding knowledge link, thereby forming a complete basic abnormal knowledge network.

[0042] This basic exception knowledge network includes knowledge entities corresponding to multiple software exception control events (such as "stamping machine pressure sensor failure entity", "conveyor belt speed control algorithm error entity", etc.) and one or more knowledge links (such as links reflecting the linkage relationship between events, links reflecting the association between fault knowledge points, etc.).

[0043] Step S140, removing knowledge entities that do not meet the set feature requirements and knowledge links corresponding to knowledge entities that do not meet the set feature requirements in the basic abnormal knowledge network, generating a target abnormal knowledge network, and outputting software abnormal control events corresponding to knowledge entities in the target abnormal knowledge network as target abnormal control events.

[0044] In this embodiment, the server analyzes the basic abnormal knowledge network. First, the feature vector of the attention event feature of the software abnormal control event corresponding to each knowledge entity is analyzed.

[0045] Taking the "stamping machine overload protection abnormal entity" as an example, the server counts the frequency of actively building event interaction sessions with other software abnormal control events (for example, actively initiating interactive sessions 5 times in the past month), the frequency of passively building event interaction sessions with other software abnormal control events (for example, 10 interactions triggered by other events), the number of fault session information generated (for example, 8 fault session information were generated), the number of fault session information generated containing the target fault tag (for example, 3 of them contained the target fault tag), the number of fault session information received (for example, 12 were received), and the number of fault session information received containing the target fault tag (for example, 4 contained the target fault tag).

[0046] The feature requirements are set as follows: the frequency of active interactive sessions is not less than 10 times / month, and the number of fault session information generated containing the target fault label is not less than 5. Since the active interactive session frequency of the "stamping machine overload protection abnormal entity" is 5 times / month, and the number of fault session information generated containing the target fault label is 3, neither of which meets the set feature requirements, the server determines it as a knowledge entity to be removed.

[0047] The server performs similar analysis and judgment on all knowledge entities, and finds all knowledge entities that do not meet the set feature requirements. Then, these knowledge entities and their corresponding knowledge links are eliminated. The basic abnormal knowledge network after elimination is output as the target abnormal knowledge network.

[0048] Finally, the server outputs the software exception control events (such as event 1 and event 3) corresponding to the knowledge entities in the target exception knowledge network as target exception control events for subsequent more in-depth analysis, processing, and repair.

[0049] Therefore, the embodiment of the present application can effectively screen out key abnormal events from the software abnormal control events of the hardware product control system, and build and optimize the abnormal knowledge network to provide strong support for the system's fault diagnosis and repair.

[0050] Based on the above steps, the embodiment of the present application significantly improves the accuracy and efficiency of system fault diagnosis. By accurately identifying software abnormal control events whose control modes meet specific target characteristics, and generating mode control event sequences based on these software abnormal control events, the key links that may cause faults can be quickly locked, providing a clear direction for subsequent analysis. Further, by marking these software abnormal control events with accurate fault labels and screening out events that are highly related to the target faults, a candidate abnormal control event sequence is constructed, laying a solid foundation for the construction of subsequent knowledge networks. The constructed basic abnormal knowledge network not only integrates rich fault information and knowledge points, but also intuitively displays the inherent connection and mutual influence between abnormal events in the form of knowledge entities and knowledge links, providing a powerful knowledge support platform for fault diagnosis experts. More importantly, by fine-tuning the feature screening and elimination of the basic knowledge network, a highly focused and refined target abnormal knowledge network is finally generated, which directly points to the core abnormal control events, greatly simplifies the troubleshooting process, and shortens the fault recovery time. As a result, not only the fault self-diagnosis capability of the hardware product control system is enhanced, but also the operation and maintenance cost is effectively reduced, the production efficiency and system stability are improved, and a solid technical guarantee is provided for intelligent manufacturing and digital transformation.

[0051] In a possible implementation, the multiple knowledge entities in the basic abnormal knowledge network include a knowledge entity corresponding to each candidate abnormal control event, a knowledge entity corresponding to a software abnormal control event that has a linkage trigger relationship with each candidate abnormal control event, and a knowledge entity corresponding to a software abnormal control event that includes fault knowledge point information of the target fault label in the corresponding fault session information. Any knowledge link in the basic abnormal knowledge network reflects that there is a linkage trigger relationship between the software abnormal control events associated with the two connected knowledge entities, or reflects that the fault session information between the software abnormal control events associated with the two connected knowledge entities includes fault knowledge point information of the target fault label. Step S130 includes:

[0052] Step S131, constructing a first abnormal knowledge network based on the candidate abnormal control events in the candidate abnormal control event sequence. The first abnormal knowledge network includes multiple knowledge entities and one or more knowledge links, the knowledge entity corresponding to each candidate abnormal control event among the multiple knowledge entities in the first abnormal knowledge network and the knowledge entity corresponding to the software abnormal control event that has a linkage trigger relationship with each candidate abnormal control event, and any knowledge link in the first abnormal knowledge network reflects that there is a linkage trigger relationship between the software abnormal control events associated with the two knowledge entities respectively.

[0053] Step S132, generating a second abnormal knowledge network based on the fault knowledge point information of the target fault label. The second abnormal knowledge network includes multiple knowledge entities and one or more knowledge links, wherein the multiple knowledge entities in the second abnormal knowledge network are knowledge entities corresponding to software abnormal control events including the fault knowledge point information of the target fault label in the corresponding fault session information, and any knowledge link in the second abnormal knowledge network reflects that the fault knowledge point information of the target fault label is included in the fault session information between the software abnormal control events associated with the two connected knowledge entities.

[0054] Step S133, integrating the first abnormal knowledge network and the second abnormal knowledge network according to the knowledge entities whose corresponding software abnormal control events in the first abnormal knowledge network and the second abnormal knowledge network are the same software abnormal control events, to generate the basic abnormal knowledge network.

[0055] In this embodiment, still taking the above example as an example, the following candidate abnormal control event sequence is obtained:

[0056] Event 1: The pressure control program of the stamping machine suddenly lost response during operation, resulting in an abnormal increase in stamping pressure.

[0057] Event 3: The speed control module on the conveyor belt experienced frequent fluctuations, resulting in unstable product conveying speed.

[0058] The server constructs the first abnormal knowledge network based on these candidate abnormal control events:

[0059] For event 1, the server determines that the software abnormality control event with which it has a linkage trigger relationship is the overload protection mechanism triggering abnormality of the device. A knowledge entity "stamping machine pressure control abnormality entity" is created for event 1, and a knowledge entity "stamping machine overload protection abnormality entity" is created for the overload protection mechanism triggering abnormality. A knowledge link is established in the first abnormality knowledge network to reflect the linkage trigger relationship between "stamping machine pressure control abnormality entity" and "stamping machine overload protection abnormality entity".

[0060] For event 3, the linkage trigger relationship is the misjudgment of the product detection module. Create a knowledge entity "Conveyor belt speed control abnormal entity" for event 3, create a knowledge entity "Product detection misjudgment entity" for the misjudgment of the product detection module, and establish a corresponding knowledge link in the first abnormal knowledge network.

[0061] Next, the server generates the second abnormal knowledge network based on the fault knowledge point information of the target fault label (such as "sensor fault" and "control algorithm error"):

[0062] For "sensor failure", the acquired knowledge point information includes common pressure sensor failure causes, detection methods, etc. The server creates knowledge entities for these knowledge point information, such as "common pressure sensor failure cause entity", "pressure sensor detection method entity", etc., and establishes knowledge links in the second abnormal knowledge network to reflect the association between them.

[0063] For "control algorithm error", the acquired knowledge point information covers common error types and optimization strategies of the speed control algorithm. The server creates corresponding knowledge entities, such as "conveyor belt speed control algorithm error type entity", "speed control algorithm optimization strategy entity", etc., and establishes knowledge links.

[0064] Then, the server integrates the first abnormal knowledge network and the second abnormal knowledge network to generate a basic abnormal knowledge network:

[0065] For example, the "stamping machine pressure control abnormality entity" in the first abnormal knowledge network is associated with the "common pressure sensor failure cause entity" in the second abnormal knowledge network, because the pressure control abnormality may be caused by a pressure sensor failure. The server integrates these two knowledge entities and establishes a corresponding knowledge link.

[0066] For "conveyor belt speed control abnormal entity" and "conveyor belt speed control algorithm error type entity", since they both involve problems related to conveyor belt speed control, the server also integrates them to build a more complete knowledge link.

[0067] Through such an integration process, the server finally generated a basic abnormal knowledge network. This network includes multiple knowledge entities such as "stamping machine pressure control abnormal entity", "stamping machine overload protection abnormal entity", "conveyor belt speed control abnormal entity", "product detection misjudgment entity", "common pressure sensor failure cause entity", "pressure sensor detection method entity", "conveyor belt speed control algorithm error type entity", "speed control algorithm optimization strategy entity", and knowledge links that reflect the complex associations between them. These knowledge links reflect both the direct linkage relationship between candidate abnormal control events and the inherent connection between them and the fault knowledge point information, providing strong support for subsequent fault analysis and processing.

[0068] In a possible implementation, step S140 includes:

[0069] Step S141, analyzing the feature vectors of the attention event characteristics of the software exception control events corresponding to each knowledge entity in the basic exception knowledge network, the attention event characteristics include at least one of the following characteristics: the frequency of actively constructing event interaction sessions with other software exception control events, the frequency of passively constructing event interaction sessions with other software exception control events, the number of generated fault session information, the number of generated fault session information containing target fault labels, the number of received fault session information, and the number of received fault session information containing target fault labels.

[0070] Step S142, treat the knowledge entities whose feature vectors of the attention event features of the corresponding software abnormal control events in the basic abnormal knowledge network do not meet the set feature requirements as eliminated knowledge entities, and eliminate the eliminated knowledge entities and the knowledge links corresponding to the eliminated knowledge entities, and output the eliminated basic abnormal knowledge network as the target abnormal knowledge network.

[0071] In this embodiment, the basic abnormal knowledge network constructed above includes multiple knowledge entities, such as "stamping machine pressure control abnormal entity", "stamping machine overload protection abnormal entity", "conveyor belt speed control abnormal entity", etc.

[0072] The server starts to analyze the feature vectors of the attention event features of the software abnormal control events corresponding to each knowledge entity.

[0073] Taking the "stamping machine overload protection abnormal entity" as an example, the server counts its related features:

[0074] Frequency of actively establishing event interaction sessions with other software anomaly control events: In the past month, 5 interactive sessions were actively initiated.

[0075] The frequency of passively constructing event interaction sessions with other software abnormality control events: 8 times of interaction triggered by other events.

[0076] Number of fault session messages generated: 10.

[0077] The number of fault session information generated containing the target fault label: 3.

[0078] Number of fault session messages received: 12.

[0079] Number of fault session messages received that contain target fault tags: 4.

[0080] The set feature requirements are: the frequency of active interactive sessions is not less than 8 times / month, and the number of fault session information generated containing the target fault label is not less than 5.

[0081] Since the active interactive session frequency of the "Press Machine Overload Protection Abnormal Entity" is 5 times / month, which is lower than the set 8 times / month, and the number of fault session information generated containing the target fault label is 3, which is lower than the set 5, the server determines the "Press Machine Overload Protection Abnormal Entity" as a knowledge entity to be eliminated.

[0082] Next, the server analyzes the "abnormal conveyor belt speed control entity":

[0083] Frequency of actively establishing event interaction sessions with other software abnormality control events: In the past month, 10 interactive sessions were actively initiated.

[0084] The frequency of passively building event interaction sessions with other software abnormality control events: 15 interactions triggered by other events.

[0085] The number of fault session information generated: 18.

[0086] The number of fault session information generated that contains the target fault label: 6.

[0087] Number of fault session messages received: 20.

[0088] Number of fault session messages received that contain target fault tags: 8.

[0089] Since all the characteristics of the "abnormal conveyor belt speed control entity" meet the set requirements, it is not judged as a discarded knowledge entity.

[0090] After the server completes the above analysis and judgment on all knowledge entities in the basic abnormal knowledge network, the knowledge entities that are judged to be eliminated (such as "stamping machine overload protection abnormal entity") and their corresponding knowledge links will be eliminated.

[0091] After the elimination, the basic abnormal knowledge network, for example, only the knowledge entities such as "stamping machine pressure control abnormal entity" and "conveyor belt speed control abnormal entity" and the knowledge links between them are left, and the server outputs them as the target abnormal knowledge network.

[0092] Through such elimination operations, a more streamlined and effective target abnormality knowledge network is obtained, which can more accurately reflect the key abnormal information in the hardware product control system and provide more targeted support for subsequent fault diagnosis and repair work.

[0093] In a possible implementation, step S120 includes:

[0094] Step A110: determining one or more control mode path information according to the control mode path information corresponding to each mode control event.

[0095] Step A120: Obtain basic fault description report information corresponding to each control mode path information from the fault restoration process corresponding to the one or more control mode path information.

[0096] Step A130, respectively perform fault classification identification on the basic fault description report information corresponding to each control mode path information, and generate a fault label of the control mode path information corresponding to each mode control event.

[0097] In this embodiment, in a hardware product manufacturing factory, the mode control event sequence obtained by the server includes event 1 "the pressure control program of the stamping machine suddenly loses response during operation, resulting in an abnormal increase in the stamping pressure" and event 3 "the speed control module on the conveyor belt frequently fluctuates, resulting in unstable product conveying speed."

[0098] For event 1, the server determines multiple control mode path information such as pressure setting parameters in the pressure control program and pressure sensor signal transmission path based on its corresponding control mode path information.

[0099] Next, the server obtains the basic fault description report information corresponding to each control mode path information from the fault recovery process corresponding to these control mode path information. For example, the report information obtained about the pressure setting parameter is "the pressure setting parameter was abnormally modified at a specific time point", and the report information about the pressure sensor signal transmission path is "the sensor signal transmission is intermittently interrupted".

[0100] Then, the server classifies and identifies these basic fault description report information. By analyzing the report information of "pressure setting parameters are abnormally modified at a specific time point", the server determines that it belongs to the "parameter error" type of fault and generates the fault label of the control mode path information as "parameter error". For the report information of "intermittent interruption of sensor signal transmission", the server classifies it as "sensor fault" and generates the corresponding fault label.

[0101] For event 3, the control mode path information determined by the server includes the output signal path of the conveyor belt speed controller, the feedback signal receiving path, etc.

[0102] Basic fault description report information is obtained from the corresponding fault recovery process. For example, the report information of the output signal path of the conveyor belt speed controller is "the output signal is continuously too high", and the report information of the feedback signal receiving path is "the feedback signal reception is delayed and the value is abnormal".

[0103] The server classifies and identifies faults in these report information, and determines "output signal is continuously too high" as "controller output fault" and generates a corresponding fault label; it classifies "feedback signal reception delay and abnormal value" as "signal transmission fault" and generates a corresponding fault label.

[0104] Through the above processing of the basic fault description report information corresponding to each control mode path information, the server generates accurate fault labels for each control mode path information in Event 1 and Event 3.

[0105] In a possible implementation, step S120 includes:

[0106] Step B110: determining one or more control mode path information according to the control mode path information corresponding to each mode control event.

[0107] Step B120: Obtain video monitoring information corresponding to each control mode path information from the fault recovery process corresponding to the one or more control mode path information.

[0108] Step B130, performing semantic recognition on the video surveillance information corresponding to each control mode path information, and generating target fault description report information of the video surveillance information corresponding to each control mode path information.

[0109] Step B140, respectively performing fault classification identification on the target fault description report information corresponding to each control mode path information, and generating a fault label of the control mode path information corresponding to each mode control event.

[0110] In this embodiment, in a hardware product manufacturing factory, the mode control event sequence obtained by the server includes event 1 "the pressure control program of the stamping machine suddenly loses response during operation, resulting in an abnormal increase in the stamping pressure" and event 3 "the speed control module on the conveyor belt frequently fluctuates, resulting in unstable product conveying speed."

[0111] For event 1, the server determines multiple control mode path information such as the signal transmission path from the pressure sensor to the controller, the calculation processing path of the controller, etc. based on its corresponding control mode path information.

[0112] Then, the server obtains the video monitoring information corresponding to each control mode path information from the fault recovery process corresponding to these control mode path information. Assume that the obtained video monitoring of the signal transmission path from the pressure sensor to the controller shows that at a certain moment, the connection line has obvious jitter and sparks; the video monitoring of the controller's operation processing path shows that the indicator light on the controller flashes abnormally.

[0113] The server performs semantic recognition on these video surveillance information. For the video surveillance information of the signal transmission path from the pressure sensor to the controller, the target fault description report information generated is "physical interference occurs in the signal transmission line"; for the video surveillance information of the controller's operation processing path, the target fault description report information generated is "controller operation processing is abnormal, indicator light status is abnormal".

[0114] Next, the server classifies and identifies the target fault description report information. It classifies "physical interference in the signal transmission line" as "line fault" and generates a fault label for the control mode path information; it classifies "abnormal controller operation and abnormal indicator light status" as "controller internal fault" and generates a corresponding fault label.

[0115] For event 3, the control mode path information determined by the server includes a signal transmission path from the speed sensor to the speed controller, an output adjustment path of the speed controller, and the like.

[0116] The video monitoring information corresponding to these control mode path information is obtained from the fault recovery process. For example, the video monitoring of the signal transmission path from the speed sensor to the speed controller shows that the signal transmission cable has obvious bends; the video monitoring of the output adjustment path of the speed controller shows that the adjustment device does not move smoothly.

[0117] The server performs semantic recognition on these video surveillance information and generates target fault description report information. For video surveillance of the signal transmission path from the speed sensor to the speed controller, the target fault description report information of "signal transmission cable is damaged" is generated; for video surveillance of the output adjustment path of the speed controller, the target fault description report information of "output adjustment device mechanical failure" is generated.

[0118] Subsequently, the server classifies and identifies these target fault description report information, classifies "damaged signal transmission cable" as "cable fault" and generates the corresponding fault label; and determines "output adjustment device mechanical fault" as "mechanical component fault" and generates the corresponding fault label.

[0119] Through the above processing of the video surveillance information, the server generates accurate fault labels for each control mode path information of event 1 and event 3.

[0120] In a possible implementation, step S120 includes:

[0121] Step C110, traversing the control mode path information corresponding to any mode control event in the mode control event sequence, and when it is determined that the control mode path information originates from an operation instruction set for device control, obtaining a control logic program corresponding to the operation instruction set.

[0122] Step C120, execute the control logic program with the help of a simulation environment, and extract the control text log and control status diagram generated by the control logic program during the execution, perform fault classification and identification on the control text log and control status diagram generated by the control logic program during the execution, and generate a fault label for the control mode path information corresponding to any one of the mode control events.

[0123] In the hardware product manufacturing factory, the pattern control event sequence obtained by the server includes event 1 "the pressure control program of the stamping machine suddenly lost response during operation, resulting in an abnormal increase in stamping pressure" and event 3 "the speed control module on the conveyor belt frequently fluctuates, resulting in unstable product conveying speed."

[0124] First, the server traverses the control mode path information corresponding to event 1. After determination, it is found that the control mode path information originates from an operation instruction set for controlling a stamping machine device.

[0125] Then, the server obtains the control logic program corresponding to the operation instruction set. Next, the server executes the control logic program with the help of a specially constructed simulation environment. During the execution process, the control logic program generates a control text log and a control state diagram.

[0126] For example, the control text log records that "the pressure setting value has an abnormal jump at a specific moment", and the control status diagram shows that "the pressure control loop is in an unstable state during a certain period of time."

[0127] The server classifies and identifies faults in these control text logs and control state diagrams. Through analysis, this situation is classified as "pressure control logic failure", thereby generating a fault label for the control mode path information corresponding to event 1.

[0128] Then, the server continues to traverse the control mode path information corresponding to event 3. After determining that it is derived from the operation instruction set of the device control, the server obtains the corresponding control logic program and executes it in the simulation environment.

[0129] Assume that the control text log records "speed adjustment command sending frequency is abnormal", and the control status diagram shows "the deviation between speed feedback and command continues to increase".

[0130] The server analyzes and identifies this information, classifies it as "speed control logic disorder", and generates a fault tag for the control mode path information corresponding to event 3.

[0131] Through the above execution of the control logic program and the processing of the information generated by it, the server accurately generates the corresponding fault labels of the control mode path information for event 1 and event 3.

[0132] In a possible implementation, the method further includes:

[0133] Step D110, based on the abnormal control event in the target industrial control interval, derivative knowledge data is added to determine one or more abnormal control event derivative data, any abnormal control event derivative data includes the abnormal control attribute symbols associated with each of the two software abnormal control events that have a linkage trigger relationship, and the relationship characteristic value of the linkage relationship between the two software abnormal control events that have a linkage trigger relationship, and the linkage relationship includes: having an event interaction session, sharing the same data link point, and being activated by the same control unit.

[0134] Step D120, determines one or more abnormal control event fault knowledge point information based on the fault session information including fault knowledge point information generated within the target industrial control interval, wherein any abnormal control event fault knowledge point information includes the abnormal control attribute symbols associated with each of the two software abnormal control events including the fault knowledge point information in the corresponding fault session information, and the fault knowledge point information included in the corresponding fault session information.

[0135] Step D130, determining the software abnormal control event sequence according to a sequence formed by software abnormal control events corresponding to abnormal control attribute symbols in the one or more abnormal control event derived data and software abnormal control events corresponding to abnormal control attribute symbols in the one or more abnormal control event fault knowledge point information.

[0136] In this embodiment, within the target industrial control range of the hardware product manufacturing factory, the server monitors a series of abnormal control events.

[0137] For example, event A is the pressure control anomaly of the stamping machine, and event B is the temperature control anomaly of the mold cooling system associated with it. Since there is a linkage trigger relationship between event A and event B, that is, the pressure control anomaly leads to the temperature control anomaly of the mold cooling system, the server determines the abnormal control attribute symbols associated with each of the two software abnormal control events. The attribute symbol of event A may be "stamping machine pressure is too high", and the attribute symbol of event B may be "mold cooling temperature is too low". At the same time, the relationship characteristic value of the linkage relationship between them is determined, such as because they share the same data link point, that is, the abnormal pressure data affects the data input of temperature control.

[0138] In addition, the server determines one or more abnormal control event fault knowledge point information based on the fault session information including fault knowledge point information generated in the target industrial control interval. For example, it is mentioned in the fault session information that the pressure control abnormality of the stamping machine is caused by the failure of the pressure sensor, and the temperature control abnormality of the mold cooling system is caused by the damage of the valve of the cooling system. The server obtains the abnormal control attribute symbols associated with each of these two software abnormal control events, namely "pressure sensor failure" and "cooling system valve damage", as well as the corresponding fault knowledge point information, such as common fault types of pressure sensors and maintenance methods of cooling system valves.

[0139] The server determines the software abnormal control event sequence based on the sequence of software abnormal control events corresponding to the abnormal control attribute symbols in the abnormal control event derivative data and the abnormal control event fault knowledge point information. For example, the pressure control abnormality of the stamping machine caused by the "pressure sensor failure" occurred first, and then the temperature control abnormality caused by the "cooling system valve damage" in the mold cooling system due to this abnormality. The server determines this series of events as the software abnormal control event sequence according to the order of occurrence.

[0140] In this way, the sequence of software abnormality control events in the hardware production process is comprehensively and accurately determined, providing a clear basis for subsequent analysis and processing.

[0141] In a possible implementation, the method further includes:

[0142] Step E110 , clustering and splitting the multiple knowledge entities in the target abnormal knowledge network based on the interactive association information between the multiple knowledge entities covered in the target abnormal knowledge network, and generating one or more knowledge entity clusters.

[0143] Step E120 , outputting the software abnormal control event corresponding to the knowledge entity in any one of the one or more knowledge entity clusters as an abnormal control event cluster, and generating one or more abnormal control event clusters.

[0144] Step E130, based on the scale of software abnormal control events covered by each abnormal control event cluster in the one or more abnormal control event clusters, determine the abnormal repair scheduling model corresponding to each abnormal control event cluster, and perform abnormal repair scheduling for the software abnormal control events in each abnormal control event cluster through the corresponding abnormal repair scheduling model.

[0145] In this embodiment, in the hardware product production factory, the server has generated a target abnormality knowledge network. This target abnormality knowledge network covers multiple knowledge entities related to software abnormality control events, such as "stamping machine pressure control abnormality entity", "conveyor belt speed control abnormality entity", "tool wear detection abnormality entity", etc.

[0146] First, the server clusters and splits the knowledge entities based on the interactive association information between the multiple knowledge entities covered in the target abnormal knowledge network. The server extracts the knowledge link features of the target abnormal knowledge network and the structural features and semantic features of each knowledge entity.

[0147] Taking "stamping machine pressure control abnormal entity" as an example, its structural characteristics may include the number of associated software abnormal control events is 3 (such as pressure sensor failure, controller parameter error, overload protection abnormality), the type is pressure control related, and the frequency of occurrence is relatively high; in terms of knowledge link characteristics, the link strength with "overload protection abnormal entity" is relatively high, indicating that the two are closely related, the direction is from "stamping machine pressure control abnormal entity" to "overload protection abnormal entity", and the type is causal relationship; the semantic characteristics are reflected in the professional terms and concepts related to pressure control.

[0148] The server integrates the extracted structural features, knowledge link features, and semantic features to construct a feature vector for each knowledge entity. Then, the server sets the neighborhood radius and minimum number of samples for the density-based clustering algorithm. Assume that the neighborhood radius is set to 0.5 and the minimum number of samples is set to 3.

[0149] Starting from any unvisited knowledge entity, such as the "stamping machine pressure control abnormal entity", with it as the core, calculate the number of other knowledge entities within the set neighborhood radius of 0.5. If the number of knowledge entities in the neighborhood is greater than or equal to the minimum sample number 3, for example, there are also "pressure sensor fault entity" and "controller parameter error entity" in this range, the server will mark "stamping machine pressure control abnormal entity", "pressure sensor fault entity" and "controller parameter error entity" as a cluster, and continue to expand the cluster to visit the unmarked knowledge entities in the neighborhood. If the number of knowledge entities in the neighborhood is less than the minimum sample number, for example, there is only the "stamping machine pressure control abnormal entity" itself, the server will mark the knowledge entity as a noise point.

[0150] The server repeats the above steps until all knowledge entities are visited and marked, and finally generates a clustering result. Then, the server uses the silhouette coefficient and Calinski-Harabasz index to evaluate the quality of the clustering results. The silhouette coefficient measures the similarity of each knowledge entity with other knowledge entities in the cluster to which it belongs and the difference with knowledge entities in other clusters. For example, the "stamping machine pressure control abnormal entity" has a high similarity with the "pressure sensor fault entity" and "controller parameter error entity" in the same cluster, and a large difference with knowledge entities in other clusters, so its silhouette coefficient is high. The Calinski-Harabasz index evaluates the clustering results by calculating the ratio of the between-class variance to the within-class variance. If the between-class variance is large and the within-class variance is small, it means that the clustering effect is good and the Calinski-Harabasz index value is large.

[0151] If the quality of the clustering results meets the set conditions, such as the silhouette coefficient is greater than 0.7 and the Calinski-Harabasz index is greater than 500, the server generates one or more knowledge entity clusters. Assume that three clusters are finally generated, namely:

[0152] Cluster 1: contains "stamping machine pressure control abnormal entity", "pressure sensor failure entity" and "controller parameter error entity".

[0153] Cluster 2: Contains "abnormal conveyor belt speed control entity" and "misjudgment of product detection entity".

[0154] Cluster 3: Contains "Tool wear detection abnormal entity" and "Tool change program error entity".

[0155] Next, the server outputs the software abnormal control event corresponding to the knowledge entity in any one of the one or more knowledge entity clusters as an abnormal control event cluster, generating one or more abnormal control event clusters. For example, the abnormal control event cluster corresponding to cluster 1 is the "stamping machine pressure related abnormal event cluster", the corresponding to cluster 2 is the "conveyor belt speed and detection related abnormal event cluster", and the corresponding to cluster 3 is the "tool related abnormal event cluster".

[0156] Then, the server determines the abnormal repair scheduling model corresponding to each abnormal control event cluster based on the scale of the software abnormal control events covered by each abnormal control event cluster. For the "stamping machine pressure-related abnormal event cluster", since there are many abnormal control events involved and they are more critical, the server determines to adopt a repair scheduling model that prioritizes processing and concentrates resources. For the "conveyor belt speed and detection-related abnormal event cluster", a more balanced repair scheduling strategy is selected based on its scale and impact. For the "tool-related abnormal event cluster", due to its relatively small scale, a more flexible repair scheduling method may be adopted.

[0157] Finally, the server uses the corresponding abnormal repair scheduling model to schedule the software abnormal control events in each abnormal control event cluster. For example, for the "stamping machine pressure-related abnormal event cluster", the server will immediately deploy professional technicians and sufficient maintenance tools to prioritize troubleshooting and repair; for the "conveyor belt speed and detection-related abnormal event cluster", according to the established balance strategy, reasonably arrange maintenance personnel and time; for the "tool-related abnormal event cluster", the repair work is flexibly arranged according to the actual situation to ensure that the problem is solved as soon as possible without affecting the overall production.

[0158] In this way, software abnormality control events in hardware product manufacturing factories can be effectively classified, analyzed and processed, thereby improving the stability and efficiency of the production system.

[0159] In a possible implementation, step E110 includes:

[0160] Step E111, extracting the knowledge link features of the target abnormal knowledge network and the structural features and semantic features of multiple knowledge entities covered in the target abnormal knowledge network, wherein the structural features include the number, type, and occurrence frequency of software abnormality control events associated with the knowledge entity, and the knowledge link features include the strength feature, direction feature, and type feature of the knowledge link, wherein the strength feature indicates the closeness of the association between two knowledge entities, the direction feature indicates the primary and secondary relationship of the association, and the type feature indicates the nature of the association.

[0161] Step E112, integrating the extracted structural features, knowledge link features and semantic features to construct a feature vector for each knowledge entity.

[0162] Step E113, setting the neighborhood radius and minimum number of samples of the density-based clustering algorithm, wherein the neighborhood radius determines the neighborhood range of a knowledge entity, and the minimum number of samples determines the minimum density requirement for an area to become a cluster.

[0163] Step E114, based on the set neighborhood radius and the minimum number of samples, starting from any unvisited knowledge entity, taking the knowledge entity as the core, calculate the number of other knowledge entities in the neighborhood of the set neighborhood radius, wherein if the number of knowledge entities in the neighborhood is greater than or equal to the minimum number of samples, the core knowledge entity and the knowledge entities in its neighborhood are marked as a cluster, and the cluster is continued to be expanded to visit the unmarked knowledge entities in the neighborhood. If the number of knowledge entities in the neighborhood is less than the minimum number of samples, the knowledge entity is marked as a noise point, and the above steps are repeated until all knowledge entities are visited and marked, and a clustering result is generated.

[0164] Step E115, using the silhouette coefficient and Calinski-Harabasz index to evaluate the quality of the clustering results, wherein the silhouette coefficient measures the similarity between each knowledge entity and other knowledge entities in the cluster to which it belongs, as well as the difference between each knowledge entity and knowledge entities in other clusters, and has a value range of [-1,1]. The larger the value, the better the clustering effect. The Calinski-Harabasz index evaluates the clustering results by calculating the ratio of the between-class variance to the within-class variance. The larger the value, the better the clustering effect.

[0165] Step E116: if the quality of the clustering result meets the set conditions, one or more knowledge entity clusters are generated.

[0166] In this embodiment, in a hardware product production factory, the server has already constructed a target abnormal knowledge network, which includes many knowledge entities related to the hardware product production control system.

[0167] First, the server starts to extract the features of the target abnormal knowledge network. For example, for the knowledge entity "stamping machine pressure control abnormal entity", the server analyzes its structural features. It is found that the number of software abnormal control events associated with it is 5, and the types are mainly related to pressure control, such as pressure sensor failure, pressure controller parameter error, etc., and the frequency of occurrence is about 2 times a day.

[0168] In terms of knowledge link characteristics, the knowledge link strength between "stamping machine pressure control abnormal entity" and "pressure sensor failure entity" is relatively high, which means that the two are closely related; the direction characteristic is from "stamping machine pressure control abnormal entity" to "pressure sensor failure entity", indicating that the former may cause the latter; the type characteristic is causal relationship.

[0169] At the same time, the server also extracts its semantic features, including professional terms and technical descriptions related to pressure control.

[0170] Next, the server integrates the extracted structural features, knowledge link features, and semantic features to construct a feature vector for the "stamping machine pressure control abnormal entity." For example, the number, type, and frequency of events in the structural features are represented by specific numerical values, and the strength, direction, and type in the knowledge link features are also encoded with corresponding numerical values. The semantic features are converted into numerical vectors through word vectors and other methods, and then these numerical vectors are combined into a comprehensive feature vector.

[0171] Then, the server sets the parameters of the density-based clustering algorithm. Assume that the neighborhood radius is set to 0.6 and the minimum number of samples is set to 4. The neighborhood radius determines how far around a knowledge entity is considered as its neighborhood, and the minimum number of samples specifies the minimum number of knowledge entities required for an area to become a cluster.

[0172] The server starts with any unvisited knowledge entity, for example, first selects the "Conveyor Belt Speed ​​Control Abnormal Entity" as the core. Then, it calculates the number of other knowledge entities within the set neighborhood radius of 0.6. Assume that within this range, there are "Conveyor Belt Motor Fault Entity", "Speed ​​Sensor Abnormal Entity", and "Product Detection Speed ​​Matching Error Entity", and the number is 3.

[0173] Since 3 is less than the minimum sample number 4, the server marks the "conveyor belt speed control abnormal entity" as a noise point.

[0174] Then, the server selects the next unvisited knowledge entity, such as "tool wear detection abnormal entity", as the core for calculation. Within the neighborhood radius of 0.6, it finds "tool wear excessive entity", "tool replacement frequency abnormal entity" and "tool detection sensor failure entity", with a total of 5 entities.

[0175] Because 5 is greater than or equal to the minimum sample number 4, the server marks the "abnormal entity of tool wear detection" and the three knowledge entities in its neighborhood as a cluster. Then, the server continues to expand this cluster and access the unmarked knowledge entities in the neighborhood. For example, if it finds that the "abnormal entity of tool lubrication system" is also in the neighborhood, it will be included in the cluster.

[0176] The server repeats the above steps until all knowledge entities are visited and labeled, and finally generates a clustering result.

[0177] Next, the server uses the silhouette coefficient and Calinski-Harabasz index to evaluate the quality of the clustering results.

[0178] For the silhouette coefficient, the server calculates the silhouette coefficient of each knowledge entity. For example, for the "tool wear detection abnormal entity", the server compares its similarity with other knowledge entities in the cluster to which it belongs (such as "tool wear excessive entity", etc.), and the difference with other cluster knowledge entities (such as "stamping machine pressure control abnormal entity", etc.). If the similarity is high and the difference is large, its silhouette coefficient is high. Assume that the silhouette coefficient of "tool wear detection abnormal entity" is 0.8.

[0179] For the Calinski-Harabasz index, the server calculates the between-class variance and the within-class variance. Assuming that the between-class variance is large and the within-class variance is small, the Calinski-Harabasz index is calculated to be 800.

[0180] If the quality of the clustering result meets the set conditions, such as the silhouette coefficient is greater than 0.7 and the Calinski-Harabasz index is greater than 600, it is qualified. Since in this calculation, the silhouette coefficient 0.8 is greater than 0.7 and the Calinski-Harabasz index 800 is greater than 600, the set conditions are met.

[0181] Therefore, the server generates one or more knowledge entity clusters. For example, it generates a "tool-related abnormality cluster", including "tool wear detection abnormality entity", "tool excessive wear entity", "tool replacement frequency abnormality entity", "tool detection sensor failure entity", "tool lubrication system abnormality entity"; a "press pressure-related abnormality cluster", including "press pressure control abnormality entity", "pressure sensor failure entity", etc.; a "conveyor belt speed-related abnormality cluster", including "conveyor belt speed control abnormality entity", "conveyor belt motor failure entity", etc.

[0182] Through such comprehensive and detailed processing, the server can effectively cluster and split the knowledge entities in the target abnormal knowledge network, providing valuable structural information for subsequent analysis and processing, which helps to more efficiently manage and solve abnormal problems in the hardware production control system.

[0183] Figure 2 The hardware structure of the industrial control monitoring system 100 provided in the embodiment of the present application for implementing the above-mentioned data mining method applied to the hardware product control system is shown as follows: Figure 2 As shown, industrial control monitoring system 100 may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0184] In one possible design, the industrial control monitoring system 100 can be a single server or a server group. The server group can be centralized or distributed (for example, the industrial control monitoring system 100 can be a distributed system). In some embodiments, the industrial control monitoring system 100 can be local or remote. For example, the industrial control monitoring system 100 can access information and / or data stored in a machine-readable storage medium 120 via a network. For another example, the industrial control monitoring system 100 can be directly connected to the machine-readable storage medium 120 to access the stored information and / or data. In some embodiments, the industrial control monitoring system 100 can be implemented on an industrial control monitoring system. By way of example only, the industrial control monitoring system can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, or the like, or any aggregation thereof.

[0185] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data acquired from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions that the industrial control monitoring system 100 uses to execute or use to complete the exemplary methods described in this application.

[0186] In the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the data mining method applied to the hardware product control system in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0187] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the above-mentioned industrial control monitoring system 100. The implementation principles and technical effects are similar, and this embodiment will not be repeated here.

[0188] In addition, an embodiment of the present application further provides a readable storage medium, in which computer executable instructions are set. When a processor runs the computer executable instructions, the data mining method applied to the hardware product control system as above is implemented.

[0189] It should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof. Similarly, it should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof.

Claims

1. A data mining method applied to a hardware product control system, characterized in that: The method comprises: Acquire a software abnormal control event sequence of a hardware product control system, and determine a software abnormal control event whose control mode meets the target control mode feature from the software abnormal control event sequence to generate a mode control event sequence; the target control mode feature represents the control mode path information included in the fault session information corresponding to the software abnormal control event in the first industrial control interval after the software abnormal control event establishes an event interaction session with other software abnormal control events; Acquire the fault label of the control mode path information corresponding to each mode control event in the mode control event sequence, and determine the mode control event whose corresponding fault label of the control mode path information is the target fault label from the mode control event sequence to generate a candidate abnormal control event sequence; Generate a basic abnormal knowledge network based on the candidate abnormal control events in the candidate abnormal control event sequence and the fault knowledge point information of the target fault label; the basic abnormal knowledge network includes knowledge entities corresponding to multiple software abnormal control events and one or more knowledge links; Eliminate knowledge entities that do not meet the set feature requirements and knowledge links corresponding to knowledge entities that do not meet the set feature requirements in the basic abnormal knowledge network, generate a target abnormal knowledge network, and output software abnormal control events corresponding to knowledge entities in the target abnormal knowledge network as target abnormal control events; The multiple knowledge entities in the basic abnormal knowledge network include knowledge entities corresponding to each candidate abnormal control event, knowledge entities corresponding to software abnormal control events that have a linkage trigger relationship with each candidate abnormal control event, and knowledge entities corresponding to software abnormal control events that include fault knowledge point information of the target fault label in the corresponding fault session information; any one knowledge link in the basic abnormal knowledge network reflects that there is a linkage trigger relationship between the software abnormal control events associated with the two connected knowledge entities, or reflects that the fault session information between the software abnormal control events associated with the two connected knowledge entities includes fault knowledge point information of the target fault label; the basic abnormal knowledge network is generated based on the candidate abnormal control events in the candidate abnormal control event sequence and the fault knowledge point information of the target fault label, including: A first abnormal knowledge network is constructed based on the candidate abnormal control events in the candidate abnormal control event sequence; the first abnormal knowledge network includes a plurality of knowledge entities and one or more knowledge links, a knowledge entity corresponding to each candidate abnormal control event among the plurality of knowledge entities in the first abnormal knowledge network and a knowledge entity corresponding to a software abnormal control event having a linkage trigger relationship with each candidate abnormal control event, and any knowledge link in the first abnormal knowledge network reflects that there is a linkage trigger relationship between the software abnormal control events respectively associated with the two knowledge entities connected; A second abnormal knowledge network is generated based on the fault knowledge point information of the target fault label; the second abnormal knowledge network includes a plurality of knowledge entities and one or more knowledge links, wherein the plurality of knowledge entities in the second abnormal knowledge network are knowledge entities corresponding to software abnormal control events including the fault knowledge point information of the target fault label in the corresponding fault session information, and any one knowledge link in the second abnormal knowledge network reflects that the fault knowledge point information of the target fault label is included in the fault session information between the software abnormal control events respectively associated with the two connected knowledge entities; The first abnormal knowledge network and the second abnormal knowledge network are integrated according to the knowledge entity that the corresponding software abnormal control events in the first abnormal knowledge network and the second abnormal knowledge network are the same software abnormal control events to generate the basic abnormal knowledge network.

2. The data mining method applied to the hardware product control system according to claim 1 is characterized in that: The step of removing knowledge entities that do not meet the set feature requirements and knowledge links corresponding to the knowledge entities that do not meet the set feature requirements in the basic abnormal knowledge network to generate a target abnormal knowledge network includes: Analyze the feature vector of the attention event feature of the software abnormal control event corresponding to each knowledge entity in the basic abnormal knowledge network, wherein the attention event feature includes at least one of the following features: the frequency of actively constructing event interaction sessions with other software abnormal control events, the frequency of passively constructing event interaction sessions with other software abnormal control events, the number of generated fault session information, the number of generated fault session information containing target fault labels, the number of received fault session information, and the number of received fault session information containing target fault labels; The knowledge entities whose feature vectors of the attention event features of the corresponding software abnormal control events in the basic abnormal knowledge network do not meet the set feature requirements are treated as eliminated knowledge entities, and the eliminated knowledge entities and the knowledge links corresponding to the eliminated knowledge entities are eliminated, and the eliminated basic abnormal knowledge network is output as the target abnormal knowledge network.

3. The data mining method applied to the hardware product control system according to claim 1 is characterized in that: The acquiring of the fault label of the control mode path information corresponding to each mode control event in the mode control event sequence includes: Determine one or more control mode path information according to the control mode path information corresponding to each mode control event; Obtaining basic fault description report information corresponding to each control mode path information from the fault restoration process corresponding to the one or more control mode path information; Fault classification and identification are performed on the basic fault description report information corresponding to each control mode path information, and a fault label of the control mode path information corresponding to each mode control event is generated.

4. The data mining method applied to the hardware product control system according to claim 1 is characterized in that: The acquiring of the fault label of the control mode path information corresponding to each mode control event in the mode control event sequence includes: Determine one or more control mode path information according to the control mode path information corresponding to each mode control event; Obtaining video monitoring information corresponding to each control mode path information from the fault restoration process corresponding to the one or more control mode path information; Performing semantic recognition on the video surveillance information corresponding to each control mode path information, and generating target fault description report information of the video surveillance information corresponding to each control mode path information; Fault classification and identification are performed on the target fault description report information corresponding to each control mode path information, and a fault label of the control mode path information corresponding to each mode control event is generated.

5. The data mining method applied to the hardware product control system according to claim 1 is characterized in that: The acquiring of the fault label of the control mode path information corresponding to each mode control event in the mode control event sequence includes: Traversing the control mode path information corresponding to any one mode control event in the mode control event sequence, and when it is determined that the control mode path information is derived from an operation instruction set for device control, acquiring a control logic program corresponding to the operation instruction set; The control logic program is executed with the help of a simulation environment, and the control text log and control state diagram generated by the control logic program during the execution are extracted, fault classification and identification are performed on the control text log and control state diagram generated by the control logic program during the execution, and a fault label of the control mode path information corresponding to any one of the mode control events is generated.

6. The data mining method applied to the hardware product control system according to claim 1 is characterized in that: The method further comprises: Determine one or more abnormal control event derivative data according to the abnormal control event in the target industrial control interval, any abnormal control event derivative data includes abnormal control attribute symbols associated with two software abnormal control events in a linkage trigger relationship, and a relationship feature value of the linkage relationship between the two software abnormal control events in a linkage trigger relationship, wherein the linkage relationship includes: having an event interaction session, sharing the same data link point, and being activated by the same control unit; Determine one or more abnormal control event fault knowledge point information according to the fault session information including the fault knowledge point information generated in the target industrial control interval, wherein any abnormal control event fault knowledge point information includes the abnormal control attribute symbols respectively associated with the two software abnormal control events including the fault knowledge point information in the corresponding fault session information, and the fault knowledge point information included in the corresponding fault session information; The software abnormal control event sequence is determined according to a sequence formed by software abnormal control events corresponding to abnormal control attribute symbols in the one or more abnormal control event derived data and software abnormal control events corresponding to abnormal control attribute symbols in the one or more abnormal control event fault knowledge point information.

7. The data mining method applied to the hardware product control system according to claim 1 is characterized in that: The method further comprises: Clustering and splitting the multiple knowledge entities of the target abnormal knowledge network based on the interactive association information between the multiple knowledge entities covered in the target abnormal knowledge network to generate one or more knowledge entity clusters; Outputting the software abnormal control event corresponding to the knowledge entity in any one of the one or more knowledge entity clusters as an abnormal control event cluster, generating one or more abnormal control event clusters; Based on the scale of software abnormal control events covered by each abnormal control event cluster in the one or more abnormal control event clusters, an abnormal repair scheduling model corresponding to each abnormal control event cluster is determined, and abnormal repair scheduling is performed on the software abnormal control events in each abnormal control event cluster through the corresponding abnormal repair scheduling model.

8. The data mining method applied to the hardware product control system according to claim 7 is characterized in that: The step of clustering and splitting the multiple knowledge entities of the target abnormal knowledge network based on the interactive association information between the multiple knowledge entities covered in the target abnormal knowledge network to generate one or more knowledge entity clusters includes: Extracting knowledge link features of the target abnormal knowledge network and structural features and semantic features of multiple knowledge entities covered in the target abnormal knowledge network, wherein the structural features include the number, type, and occurrence frequency of software abnormal control events associated with the knowledge entity, and the knowledge link features include the strength feature, direction feature, and type feature of the knowledge link, wherein the strength feature indicates the closeness of the association between two knowledge entities, the direction feature indicates the primary and secondary relationship of the association, and the type feature indicates the nature of the association; Integrate the extracted structural features, knowledge link features and semantic features to construct the feature vector of each knowledge entity; Setting the neighborhood radius and minimum number of samples of the density-based clustering algorithm, wherein the neighborhood radius determines the neighborhood range of a knowledge entity, and the minimum number of samples determines the minimum density requirement for an area to become a cluster; Based on the set neighborhood radius and the minimum number of samples, starting from any unvisited knowledge entity, taking the knowledge entity as the core, the number of other knowledge entities in the neighborhood of the set neighborhood radius is calculated, wherein, if the number of knowledge entities in the neighborhood is greater than or equal to the minimum number of samples, the core knowledge entity and the knowledge entities in its neighborhood are marked as a cluster, and the cluster is continued to be expanded to visit the unmarked knowledge entities in the neighborhood; if the number of knowledge entities in the neighborhood is less than the minimum number of samples, the knowledge entity is marked as a noise point, and the above steps are repeated until all knowledge entities are visited and marked, and a clustering result is generated; The silhouette coefficient and Calinski-Harabasz index are used to evaluate the quality of the clustering results. The silhouette coefficient measures the similarity between each knowledge entity and other knowledge entities in the cluster to which it belongs, as well as the difference between each knowledge entity and knowledge entities in other clusters. The value range is [-1, 1]. The larger the value, the better the clustering effect. The Calinski-Harabasz index evaluates the clustering results by calculating the ratio of the between-class variance to the within-class variance. The larger the value, the better the clustering effect. If the quality of the clustering result meets the set conditions, one or more knowledge entity clusters are generated.

9. An industrial control monitoring system, characterized in that: The industrial control monitoring system includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to run the programs, instructions or codes in the memory to implement the data mining method applied to the hardware product control system as described in any one of claims 1 to 8 above.

Images