A fault diagnosis method, program product, electronic device and storage medium

Abstract

The application provides a fault diagnosis method, a program product, an electronic device and a storage medium, and is applied to the technical field of intelligent operation and maintenance, and the fault diagnosis method comprises the following steps: generating a cognitive state object corresponding to each cognitive stage of a cognitive chain which is pre-constructed, wherein the cognitive chain is used for fault processing of a monitored system, the cognitive chain comprises a plurality of cognitive stages which are executed in sequence, the plurality of cognitive stages comprise a health judgment stage and a diagnosis reasoning stage, the health judgment stage is used for judging whether the monitored system has a fault, the diagnosis reasoning stage is used for judging a fault type, the cognitive state object comprises a judgment value of the cognitive stage and a bias value representing uncertainty of the cognitive stage; and delivering the cognitive state object generated by a current cognitive stage to a next cognitive stage, so that the next cognitive stage adjusts its processing logic according to the bias value in the cognitive state object. The method can improve the reliability and accuracy of fault diagnosis.

Description

Technical Field

[0001] This application relates to the field of intelligent operation and maintenance technology, and more specifically, to a fault diagnosis method, a program product, an electronic device, and a storage medium. Background Technology

[0002] In industrial systems, information systems, medical equipment, and complex infrastructure, fault diagnosis and self-healing systems typically employ detection methods based on threshold alarms, expert rules, or machine learning models. In practical applications, these methods often form a unidirectional processing chain of data acquisition, anomaly identification, fault location, and repair actions. However, due to the dynamic nature of complex systems, the heterogeneity of multi-source data, and the diversity of fault modes, the reliability and accuracy of existing fault diagnosis methods are relatively low. Summary of the Invention

[0003] The purpose of this application is to provide a fault diagnosis method, program product, electronic device and storage medium to solve the technical problem that the reliability and accuracy of fault diagnosis methods in the prior art are both low.

[0004] In a first aspect, embodiments of this application provide a fault diagnosis method, comprising: generating a cognitive state object corresponding to each cognitive stage of a pre-constructed cognitive chain, wherein the cognitive chain is used for fault handling of a monitored system, the cognitive chain includes multiple cognitive stages executed sequentially, the multiple cognitive stages include a health judgment stage and a diagnostic reasoning stage, the health judgment stage is used to determine whether the monitored system has a fault, the diagnostic reasoning stage is used to determine the fault type, the cognitive state object includes a judgment value of the cognitive stage and a deviation value characterizing the uncertainty of the cognitive stage; and passing the cognitive state object generated by the current cognitive stage to the next cognitive stage, so that the next cognitive stage adjusts its own processing logic according to the deviation value in the cognitive state object.

[0005] In the above scheme, a cognitive chain is constructed that includes a health assessment stage and a diagnostic reasoning stage. At each stage, a cognitive state object containing assessment and deviation values ​​is generated and passed to the next stage, enabling downstream stages to adjust their processing logic based on upstream deviation values. Therefore, by extending the fault diagnosis process from a single-point decision to a multi-stage cognitive process, the simplification of complex problems into a single response process is avoided, thereby improving the reliability and accuracy of fault diagnosis.

[0006] In an optional implementation, the cognitive state object further includes at least one of confidence level, source layer identifier, timestamp, and upstream dependency chain. In the above scheme, by further introducing at least one of confidence level, source layer identifier, timestamp, and upstream dependency chain into the cognitive state object, the entire cognitive process becomes traceable and interpretable.

[0007] In an optional implementation, the deviation value is calculated as a weighted sum of the internal consistency error and the upstream matching error of the cognitive stage. In the above scheme, the deviation value is quantified as a weighted sum of the internal consistency error and the upstream matching error, transforming the deviation from an abstract concept into a calculable engineering variable. This ensures that the deviation value can truly reflect the uncertainty of the cognitive state at each stage, thereby improving the reliability and accuracy of fault diagnosis.

[0008] In optional implementations, the processing logic is adjusted in at least one of the following ways: when the deviation value is higher than a first threshold, the processing result of the current cognitive stage is downweighted; when the cumulative deviation values ​​of multiple consecutive preceding cognitive stages exceed a second threshold, the output of the current cognitive stage is delayed or suspended; when the cumulative deviation variable of the entire chain exceeds a third threshold, high-risk repair strategies are prohibited; and the algorithm parameters of the current cognitive stage are dynamically adjusted or different processing models are selected based on the deviation value. In the above scheme, downweighting when the deviation value is too high can avoid unreliable judgments affecting subsequent decisions; delaying or suspending the output when the cumulative deviation of multiple stages exceeds the standard can prevent erroneous responses based on unreliable information chains; prohibiting high-risk repair when the cumulative deviation of the entire chain exceeds the threshold can ensure safety under highly uncertain conditions; and dynamically adjusting the algorithm parameters or selecting different models can adaptively match the processing strategy according to the real-time reliability level.

[0009] In an optional implementation, the cognitive chain further includes a causal inference stage and an execution stage. The causal inference stage is used to identify the root cause of the fault, and the execution stage is used to execute the repair operations in the repair strategy. In the above scheme, extending the cognitive chain to both the causal inference and execution stages enables the system not only to determine the fault type but also to identify the root cause and execute repair operations, forming a complete closed loop from anomaly detection to root cause analysis and then to repair action, thereby improving the effectiveness and efficiency of fault handling.

[0010] In an optional implementation, the cognitive chain further includes an observation phase, which monitors the state of the monitored system after the repair operation is performed, generates feedback information, and transmits the feedback information backward to one or more preceding cognitive phases. In the above scheme, extending the cognitive chain to the observation phase and transmitting feedback information backward to preceding cognitive phases allows each phase to adjust its own model or logic based on the actual execution effect, thereby ensuring the long-term stability and reliability of the system.

[0011] In an optional implementation, the cognitive chain further includes a strategy decision-making stage, which generates a repair strategy based on the output of the diagnostic reasoning stage and / or the causal inference stage. In the above scheme, extending the cognitive chain to the strategy decision-making stage, enabling it to generate a repair strategy based on the output of the diagnostic reasoning and / or causal inference stages, achieves intelligent decision-making, thereby improving the reliability and accuracy of fault diagnosis.

[0012] In an optional implementation, the cognitive chain further includes a collection phase, which is used to collect operational data of the monitored system. In the above scheme, extending the cognitive chain to the collection phase enables proactive acquisition of multi-source operational data from the monitored system, providing upstream basis for uncertainty quantification in subsequent stages, thereby avoiding overall judgment errors due to data quality issues.

[0013] In an optional implementation, the fault diagnosis method further includes: in the health assessment stage, determining whether the monitored system has an anomaly based on the collected operational data of the monitored system, and generating a first cognitive state object, wherein the first cognitive state object includes a health status assessment result and a first deviation value characterizing the uncertainty of the health assessment stage; when the health status assessment result indicates an anomaly, entering the diagnostic reasoning stage, matching a fault mode based on the health status assessment result, and generating a second cognitive state object, wherein the second cognitive state object includes a diagnostic reasoning result and a second deviation value characterizing the uncertainty of the diagnostic reasoning stage; in the causal inference stage, identifying the root cause of the fault based on the diagnostic reasoning result, and generating a third cognitive state object. The third cognitive state object includes a causal inference result and a third deviation value representing the uncertainty of the causal inference stage. In the execution stage, a repair operation is performed based on the causal inference result, generating a fourth cognitive state object, which includes the execution result and a fourth deviation value representing the uncertainty of the execution stage. In the observation stage, the state of the monitored system after the repair operation is performed is monitored, generating a fifth cognitive state object, which includes the observation result and a fifth deviation value representing the uncertainty of the observation stage. The observation result is then used as feedback information and transmitted back to one or more of the health judgment stage, diagnostic reasoning stage, and causal inference stage to adjust the processing logic of the corresponding stage. In this scheme, the fault diagnosis process in a complex system is divided into multiple cognitive stages, such as acquisition, health judgment, diagnosis, causal inference, strategy generation, execution control, and observation reflection, thereby avoiding simplifying complex problems into a single response process. Simultaneously, deviation is introduced as an explicit system state in the cognitive system, so that deviation is no longer implicit within a single algorithm but participates in the multi-stage processing process as a perceptible, transmittable, and constrained system operation element. Furthermore, the structured closed-loop reflection mechanism implemented through the observation layer feeds the execution results back into the system to verify the effectiveness of the preceding cognitive path, thereby improving the stability and convergence of the system operation at the system level.

[0014] Secondly, embodiments of this application provide a fault diagnosis device, comprising: a generation module, configured to generate a cognitive state object corresponding to each cognitive stage of a pre-constructed cognitive chain, wherein the cognitive chain is used for fault handling of a monitored system, the cognitive chain includes multiple cognitive stages executed sequentially, the multiple cognitive stages include a health judgment stage and a diagnostic reasoning stage, the health judgment stage is used to determine whether the monitored system has a fault, the diagnostic reasoning stage is used to determine the fault type, and the cognitive state object includes a judgment value of the cognitive stage and a deviation value characterizing the uncertainty of the cognitive stage; and a transmission module, configured to transmit the cognitive state object generated by the current cognitive stage to the next cognitive stage, so that the next cognitive stage adjusts its own processing logic according to the deviation value in the cognitive state object.

[0015] In the above scheme, a cognitive chain is constructed that includes a health assessment stage and a diagnostic reasoning stage. At each stage, a cognitive state object containing assessment and deviation values ​​is generated and passed to the next stage, enabling downstream stages to adjust their processing logic based on upstream deviation values. Therefore, by extending the fault diagnosis process from a single-point decision to a multi-stage cognitive process, the simplification of complex problems into a single response process is avoided, thereby improving the reliability and accuracy of fault diagnosis.

[0016] In an optional implementation, the cognitive state object further includes at least one of confidence level, source layer identifier, timestamp, and upstream dependency chain. In the above scheme, by further introducing at least one of confidence level, source layer identifier, timestamp, and upstream dependency chain into the cognitive state object, the entire cognitive process becomes traceable and interpretable.

[0017] In an optional implementation, the deviation value is calculated as a weighted sum of the internal consistency error and the upstream matching error of the cognitive stage. In the above scheme, the deviation value is quantified as a weighted sum of the internal consistency error and the upstream matching error, transforming the deviation from an abstract concept into a calculable engineering variable. This ensures that the deviation value can truly reflect the uncertainty of the cognitive state at each stage, thereby improving the reliability and accuracy of fault diagnosis.

[0018] In optional implementations, the processing logic is adjusted in at least one of the following ways: when the deviation value is higher than a first threshold, the processing result of the current cognitive stage is downweighted; when the cumulative deviation values ​​of multiple consecutive preceding cognitive stages exceed a second threshold, the output of the current cognitive stage is delayed or suspended; when the cumulative deviation variable of the entire chain exceeds a third threshold, high-risk repair strategies are prohibited; and the algorithm parameters of the current cognitive stage are dynamically adjusted or different processing models are selected based on the deviation value. In the above scheme, downweighting when the deviation value is too high can avoid unreliable judgments affecting subsequent decisions; delaying or suspending the output when the cumulative deviation of multiple stages exceeds the standard can prevent erroneous responses based on unreliable information chains; prohibiting high-risk repair when the cumulative deviation of the entire chain exceeds the threshold can ensure safety under highly uncertain conditions; and dynamically adjusting the algorithm parameters or selecting different models can adaptively match the processing strategy according to the real-time reliability level.

[0019] In an optional implementation, the cognitive chain further includes a causal inference stage and an execution stage. The causal inference stage is used to identify the root cause of the fault, and the execution stage is used to execute the repair operations in the repair strategy. In the above scheme, extending the cognitive chain to both the causal inference and execution stages enables the system not only to determine the fault type but also to identify the root cause and execute repair operations, forming a complete closed loop from anomaly detection to root cause analysis and then to repair action, thereby improving the effectiveness and efficiency of fault handling.

[0020] In an optional implementation, the cognitive chain further includes an observation phase, which monitors the state of the monitored system after the repair operation is performed, generates feedback information, and transmits the feedback information backward to one or more preceding cognitive phases. In the above scheme, extending the cognitive chain to the observation phase and transmitting feedback information backward to preceding cognitive phases allows each phase to adjust its own model or logic based on the actual execution effect, thereby ensuring the long-term stability and reliability of the system.

[0021] In an optional implementation, the cognitive chain further includes a strategy decision-making stage, which generates a repair strategy based on the output of the diagnostic reasoning stage and / or the causal inference stage. In the above scheme, extending the cognitive chain to the strategy decision-making stage, enabling it to generate a repair strategy based on the output of the diagnostic reasoning and / or causal inference stages, achieves intelligent decision-making, thereby improving the reliability and accuracy of fault diagnosis.

[0022] In an optional implementation, the cognitive chain further includes a collection phase, which is used to collect operational data of the monitored system. In the above scheme, extending the cognitive chain to the collection phase enables proactive acquisition of multi-source operational data from the monitored system, providing upstream basis for uncertainty quantification in subsequent stages, thereby avoiding overall judgment errors due to data quality issues.

[0023] In an optional embodiment, the fault diagnosis device further includes: a health assessment module, configured to, during the health assessment phase, determine whether the monitored system has an anomaly based on the collected operating data of the monitored system, and generate a first cognitive state object, wherein the first cognitive state object includes a health status assessment result and a first deviation value characterizing the uncertainty of the health assessment phase; a fault diagnosis module, configured to, when the health status assessment result indicates an anomaly, enter the diagnostic reasoning phase, match a fault mode based on the health status assessment result, and generate a second cognitive state object, wherein the second cognitive state object includes a diagnostic reasoning result and a second deviation value characterizing the uncertainty of the diagnostic reasoning phase; and a causal inference module, configured to, during the causal inference phase, identify the root cause of the fault based on the diagnostic reasoning result, and generate a third... The system comprises a cognitive state object, wherein the third cognitive state object includes a causal inference result and a third deviation value characterizing the uncertainty of the causal inference stage; an execution module, configured to perform a repair operation based on the causal inference result during the execution stage, generating a fourth cognitive state object, wherein the fourth cognitive state object includes the execution result and a fourth deviation value characterizing the uncertainty of the execution stage; and an observation module, configured to monitor the state of the monitored system after the repair operation is performed during the observation stage, generating a fifth cognitive state object, wherein the fifth cognitive state object includes the observation result and a fifth deviation value characterizing the uncertainty of the observation stage, and transmitting the observation result as feedback information back to one or more of the health judgment stage, diagnostic reasoning stage, and causal inference stage to adjust the processing logic of the corresponding stage. In the above scheme, the fault diagnosis process in a complex system is divided into multiple cognitive stages such as acquisition, health judgment, diagnosis, causal inference, strategy generation, execution control, and observation reflection, thereby avoiding simplifying complex problems into a single response process; simultaneously, deviation is introduced as an explicit system state in the cognitive system, so that deviation is no longer implicit within a single algorithm, but participates in the multi-stage processing process as a perceptible, transmittable, and constrainable system operation element. Furthermore, the structured closed-loop reflection mechanism implemented through the observation layer feeds the execution results back into the system to verify the effectiveness of the preceding cognitive path, thereby improving the stability and convergence of the system operation at the system level.

[0024] Thirdly, embodiments of this application provide a computer program product, including computer program instructions, which are read and executed by a processor to perform the fault diagnosis method as described in the first aspect.

[0025] Fourthly, embodiments of this application provide an electronic device, including: a processor, a memory, and a bus; the processor and the memory communicate with each other via the bus; the memory stores computer program instructions executable by the processor, and the processor can execute the fault diagnosis method as described in the first aspect by calling the computer program instructions.

[0026] Fifthly, embodiments of this application provide a computer-readable storage medium that stores computer program instructions, which, when executed by a computer, cause the computer to perform the fault diagnosis method as described in the first aspect.

[0027] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, embodiments of this application are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 A flowchart illustrating a fault diagnosis method provided in an embodiment of this application; Figure 2 A flowchart illustrating another fault diagnosis method provided in this application embodiment; Figure 3 A structural block diagram of a fault diagnosis device provided in an embodiment of this application; Figure 4 This is a structural block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0030] In industrial systems, information systems, medical equipment, and complex infrastructure, fault diagnosis and handling typically involve multi-source data, multi-stage reasoning, and complex decision-making. Fault diagnosis and handling usually employ one of the following technical approaches: threshold- or rule-based fault detection and alarm systems; expert systems based on empirical rules; or anomaly detection and prediction systems based on statistical models or machine learning. In practical applications, these systems typically form the following processing chain: data acquisition, anomaly identification, fault location, and repair actions.

[0031] However, while the aforementioned fault diagnosis and handling methods can play a certain role in simple or stable scenarios, they have the following shortcomings in complex systems: First, existing technologies typically simplify system problems into local anomaly detection or single-point decision-making problems, ignoring the multi-stage cognitive process from data perception to decision execution. Second, on the one hand, cognitive science, information theory, and causal inference have proposed a large number of theoretical models to describe uncertainty, reasoning, and decision-making, but most remain at the theoretical or algorithmic level, making it difficult to guide the overall architecture design of complex systems. On the other hand, engineering systems often rely on developers' experience and rule stacking, lacking unified theoretical constraints, resulting in systems that are neither interpretable nor scalable. Third, existing methods typically treat noise, errors, and uncertainties as local problems, mitigating them by improving the accuracy of single-point algorithms, without considering the propagation and amplification of uncertainty in multi-stage processing links. Finally, most methods, after performing repair actions, only determine whether the action was successful, failing to systematically feed the results back to the preceding cognitive process, thus making it difficult to form a long-term stable self-optimization mechanism.

[0032] Based on the above analysis, this application provides a fault diagnosis method that models the fault diagnosis and self-healing process as a multi-stage cognitive chain structure. This method achieves structured constraints and dynamic stability control of the fault handling process through hierarchical cognitive control. The technical solutions in this application will be described below with reference to the accompanying drawings.

[0033] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a fault diagnosis method provided in an embodiment of this application. This fault diagnosis method can, but is not limited to, be performed by an electronic device. Figure 4 The possible structure of this electronic device is shown below; for details, please refer to the following section. Figure 4 The above-mentioned fault diagnosis methods may specifically include: S101: At each cognitive stage of the pre-built cognitive chain, generate the cognitive state object corresponding to that cognitive stage.

[0034] A cognitive chain refers to a multi-stage, structured fault handling process that mimics human cognitive processes. It is a directed, sequentially executed sequence of stages, each responsible for handling a specific type of cognitive task and passing the result to the next stage. Specifically, the cognitive chain is used for fault handling in a monitored system and can include multiple sequentially executed cognitive stages.

[0035] The cognitive stage is the basic unit constituting the cognitive chain. Each stage corresponds to a specific cognitive function, such as data acquisition, health assessment, and fault diagnosis, used to handle faults in the monitored system. The monitored system refers to the target system monitored and diagnosed using the fault diagnosis method provided in this application. It can be various complex systems such as information technology (IT) systems, industrial control systems, medical equipment, and infrastructure.

[0036] For example, information can be exchanged between different cognitive stages through standardized interfaces.

[0037] It should be noted that no cognitive stage should output only the judgment value while ignoring reliability information.

[0038] In one implementation, the aforementioned multiple cognitive stages may include a health assessment stage and a diagnostic reasoning stage. The health assessment stage is used to determine whether a fault exists in the monitored system, while the diagnostic reasoning stage is used to determine the type of fault.

[0039] It is understood that the above-mentioned cognitive chain may also include other cognitive stages. The embodiments of this application do not specifically limit this. Those skilled in the art can make appropriate adjustments according to the actual situation. For example, the cognitive chain may also include one or more stages such as the collection stage, the causal inference stage, the strategy decision-making stage, the execution stage, and the observation stage.

[0040] A cognitive state object is a unified structured data unit passed between cognitive stages. For example, the cognitive state object corresponding to each cognitive stage may include the judgment value and deviation value for that cognitive stage. The judgment value is the result derived by that cognitive stage based on input data and processing logic, such as health status, fault type, root cause analysis results, etc. The deviation value is a quantitative indicator characterizing the uncertainty of a cognitive stage, reflecting the reliability of the judgment result or the potential error; understandably, the larger the deviation value, the higher the uncertainty and the lower the reliability of the judgment in that cognitive stage.

[0041] In the above S101, the input to the health assessment stage can be operational data collected from the monitored system (e.g., CPU utilization, memory utilization, response time, log information, etc.); the output of the health assessment stage can be a first cognitive state object, including the health status assessment result (i.e., the assessment value, e.g., abnormal or normal) and a first deviation value characterizing the uncertainty of the health assessment stage (e.g., a first deviation value of 0.1 indicates that the health status assessment result output by the health assessment stage is 90% reliable).

[0042] For example, the health assessment phase can use threshold judgment, statistical models or machine learning models to determine whether the monitored system is in an abnormal state; for example, setting a CPU utilization rate of more than 90% as an abnormal threshold, when the collected CPU utilization rate is 95%, it is judged as abnormal.

[0043] The input to the diagnostic reasoning stage can be the first cognitive state object output from the health judgment stage and / or the raw operating data collected from the monitored system; the output of the diagnostic reasoning stage can be the second cognitive state object, including the diagnostic reasoning result (i.e., the judgment value, such as CPU overload) and the second deviation value characterizing the uncertainty of the diagnostic reasoning stage (e.g., a second deviation value of 0.2 indicates that the diagnostic reasoning result output by the diagnostic reasoning stage is 80% reliable).

[0044] For example, in the diagnostic reasoning stage, a fault mode matching algorithm can be used to identify specific fault types, such as fault modes like CPU overload, memory leak, and network packet loss, and their corresponding symptom characteristics; the current symptom is matched with the pattern library, and the fault type with the highest matching degree is selected; multiple fault types are matched simultaneously, and ambiguous states are recorded.

[0045] As one implementation, the diagnostic reasoning phase is triggered only when the health status judgment result output by the health judgment phase indicates that there is an anomaly in the monitored system.

[0046] In one alternative implementation, a cognitive chain for fault handling of the monitored system can be pre-built before performing S101 above.

[0047] S102: Pass the cognitive state object generated in the current cognitive stage to the next cognitive stage so that the next cognitive stage can adjust its own processing logic according to the deviation value in the cognitive state object.

[0048] The next cognitive stage refers to the stage following the current cognitive stage in the sequence structure of the cognitive chain. It receives the cognitive state object output by the current cognitive stage as input.

[0049] Processing logic refers to the specific algorithms, rules, or models executed within each cognitive stage to transform input into output. Specifically, based on the cognitive state object corresponding to the previous cognitive stage, the current cognitive stage can adjust its processing logic, such as selecting different algorithms, adjusting parameters, or performing downgrade processing.

[0050] In the above S102, it should be noted that the specific implementation method for transmitting the cognitive state object is not specifically limited in the embodiments of this application. Those skilled in the art can make appropriate adjustments according to the actual situation. For example, after the current cognitive stage is completed, the cognitive state object can be sent to the next cognitive stage immediately; or, it can be decoupled and transmitted through middleware such as message queues and event buses; or, the cognitive state object can be written into a shared database or cache and actively read in the next stage, etc.

[0051] In one alternative implementation, after executing S102 above, the next cognitive stage can adjust its processing logic based on the deviation value in the cognitive state object.

[0052] In the above scheme, a cognitive chain is constructed that includes a health assessment stage and a diagnostic reasoning stage. At each stage, a cognitive state object containing assessment and deviation values ​​is generated and passed to the next stage, enabling downstream stages to adjust their processing logic based on upstream deviation values. Therefore, by extending the fault diagnosis process from a single-point decision to a multi-stage cognitive process, the simplification of complex problems into a single response process is avoided, thereby improving the reliability and accuracy of fault diagnosis.

[0053] Furthermore, based on the above embodiments, the cognitive chain may also include a causal inference stage and an execution stage, wherein the causal inference stage is used to identify the root cause of the failure, and the execution stage is used to execute the repair operations in the repair strategy.

[0054] In S101 above, the input to the causal inference stage can be the cognitive state object output from the diagnostic inference stage; the output of the causal inference stage can be a third cognitive state object, including the causal inference result (i.e., the judgment value, such as the root cause analysis result) and a third bias value characterizing the uncertainty of the causal inference stage. For example, the causal inference stage can employ causal analysis techniques to identify root causes, such as: constructing a system dependency graph to analyze fault propagation paths; calculating the posterior probability of each potential cause based on a Bayesian network; and quantifying the influence of each factor based on a structural equation model.

[0055] In S101 above, the input to the execution phase can be the root cause directly from the causal inference phase or the repair strategy output from the strategy decision phase; the output of the execution phase can be a fourth cognitive state object, including the execution result (i.e., the judgment value) and a fourth deviation value characterizing the uncertainty of the execution phase. For example, the execution phase can convert the repair strategy into a specific executable command, send the command through an interface with the monitored system, monitor the command execution process, and finally record the execution result.

[0056] In the above scheme, the cognitive chain is extended to the causal inference stage and the execution stage, so that the system can not only judge the fault type, but also identify the root cause of the fault and perform repair operations, forming a complete closed loop from anomaly discovery to root cause analysis and repair action, thereby improving the effectiveness and efficiency of fault handling.

[0057] Furthermore, based on the above embodiments, the cognitive chain may also include: an observation phase, which is used to monitor the status of the monitored system after the repair operation is performed, generate feedback information, and transmit the feedback information in reverse to one or more preceding cognitive phases.

[0058] In S101 above, the input to the observation phase can be a cognitive state object output from the execution phase; the output of the observation phase can be a fifth cognitive state object, including the observation result (i.e., the judgment value) and a fifth deviation value characterizing the uncertainty of the observation phase. For example, the observation phase can monitor the repaired system state, generate feedback information, and then transmit the feedback information in reverse.

[0059] In the above scheme, the cognitive chain is extended to the observation stage and the feedback information is transmitted back to the preceding cognitive stage, so that each stage can adjust its own model or logic according to the actual execution effect, thereby ensuring the stability and reliability of the system in the long term.

[0060] Furthermore, based on the above embodiments, the cognitive chain may also include: a strategy decision-making stage, which is used to generate a repair strategy based on the output of the diagnostic reasoning stage and / or the causal inference stage.

[0061] In S101 above, the input to the strategy decision-making stage can be the root cause from the causal inference stage; the output of the strategy decision-making stage can be a sixth cognitive state object, including the strategy decision result (i.e., the judgment value) and a sixth deviation value characterizing the uncertainty of the strategy decision-making stage. For example, the strategy decision-making stage can collect input information, generate candidate strategies, then assess the strategy risk, and select the optimal strategy.

[0062] In the above scheme, the cognitive chain is extended to the strategy decision-making stage, enabling it to generate repair strategies based on the output of the diagnostic reasoning and / or causal inference stages, thereby achieving intelligent decision-making and improving the reliability and accuracy of fault diagnosis.

[0063] Furthermore, based on the above embodiments, the cognitive chain may also include: a collection phase, which is used to collect operational data of the monitored system.

[0064] In S101 above, the output of the acquisition phase can be a seventh cognitive state object, including the operational data of the monitored system (i.e., judgment values) and a seventh deviation value characterizing the uncertainty of the acquisition phase. For example, the acquisition phase can first configure the data source to determine the data types to be acquired (e.g., performance metrics, logs, events, etc.), then perform data acquisition and evaluate data quality (e.g., data integrity, data timeliness, data consistency, data accuracy, etc.), and finally encapsulate the acquired data and quality information into the aforementioned seventh cognitive state object.

[0065] In the above scheme, the cognitive chain is extended to the collection stage, thereby enabling the proactive acquisition of multi-source operational data of the monitored system. This provides upstream basis for the quantification of uncertainty in subsequent stages, thus avoiding overall judgment errors due to data quality issues.

[0066] Furthermore, based on the above embodiments, the cognitive state object also includes at least one of confidence level, source layer identifier, timestamp, and upstream dependency chain.

[0067] Confidence level is a probability value that characterizes the reliability of a judgment result. Its value typically ranges from [0,1] and can be used to implement automatic degradation control. As one implementation method, the relationship between confidence level and deviation value can be complementary, for example: confidence level = 1 - deviation value.

[0068] The source layer identifier is used to identify which cognitive stage generated the cognitive state object, facilitating traceability and auditing. For example, predefined stage names can be used, such as "collection layer", "health layer", "diagnosis layer", "causal layer", "strategy layer", "execution layer", "observation layer", etc.; or, numerical identifiers can be used, such as 1 representing the collection layer, 2 representing the health layer, etc.

[0069] A timestamp is a record of the point in time when a cognitive state object is generated, used for time series analysis and latency detection.

[0070] The upstream dependency chain is used to record the list of preceding cognitive stages that the current cognitive state object depends on, in order to form a complete cognitive path tracing chain.

[0071] For example, the data passed between different cognitive stages may not be simple result values, but rather a unified structured cognitive state object. As one implementation method, the cognitive state object can be defined as: State object = (State ID, Judgment value, Confidence level, Deviation value, Source layer identifier, Timestamp, Upstream dependency chain). The structure of this cognitive state object follows a cross-layer interface standard.

[0072] Through the aforementioned structural design of cognitive state objects, traceable cognitive paths, auditable judgment reliability, rollbackable decision paths, and automatic degradation control based on confidence levels can be achieved. Therefore, the embodiments of this application explicitly model uncertainty as structural variables, allowing them to participate in system-level control decisions, rather than being confined to the algorithm itself.

[0073] In the above scheme, by further introducing at least one of confidence level, source layer identifier, timestamp and upstream dependency chain into the cognitive state object, the entire cognitive process becomes traceable and interpretable.

[0074] Furthermore, based on the above embodiments, a bias perception mechanism can be introduced at each cognitive stage of the cognitive chain to characterize the uncertainty of stage-based judgments.

[0075] Existing fault diagnosis systems generally consider deviations to be model errors; they view deviations as uncontrollable side effects and fail to model their propagation paths systematically; the generation, accumulation, and dissipation of deviations are considered unobservable and inexplicable, and the system can only passively accept the final result. Based on this, the embodiments of this application redefine deviations as: deviations are a comprehensive manifestation of multi-stage cognitive mismatches.

[0076] For example, the biases in each cognitive stage can be decomposed into: information credibility bias corresponding to the acquisition layer, anomaly judgment bias corresponding to the health layer, fault mode matching bias corresponding to the diagnosis layer, structural reasoning bias corresponding to the causal inference layer, suboptimal decision bias corresponding to the strategy layer, and action execution bias corresponding to the execution layer.

[0077] As one implementation method, the deviation value can be calculated by weighting the internal consistency error and the upstream matching error of the cognitive stage. The internal consistency error reflects the degree of consistency of the internal processing of the cognitive stage, i.e., the difference in results produced by the same input under different conditions; while the upstream matching error reflects the degree of matching between the output of the cognitive stage and the expected upstream input, i.e., whether the judgment of the current cognitive stage is consistent with the information conveyed by the previous cognitive stage.

[0078] For example, define the first Bias in each cognitive stage This definition causes deviation Become a computable engineering variable, rather than an abstract concept: ; in, This indicates the internal consistency error at this cognitive stage. This indicates the upstream matching error. and They are respectively and The weighting coefficients. This is understandable for cognitive stages that heavily rely on prior information (e.g., diagnostic reasoning relies on health judgments). It can be set to a larger value; for cognitive stages with strong self-judgment capabilities (e.g., the acquisition layer based on independent sensors). It can be set to a larger value.

[0079] In the above scheme, the deviation value is quantified as a weighted sum of internal consistency error and upstream matching error, which transforms the deviation from an abstract concept into a calculable engineering variable. This ensures that the deviation value can truly reflect the uncertainty of the cognitive state at each stage, thereby improving the reliability and accuracy of fault diagnosis.

[0080] Furthermore, based on the above embodiments, a cumulative variable of the entire cognitive chain bias can also be calculated. This cumulative variable of the entire cognitive chain bias is a weighted sum of the bias values ​​at each cognitive stage, reflecting the overall uncertainty of the entire cognitive chain.

[0081] For example, the cumulative variable of the whole chain deviation can be calculated using the following formula: ; in, For the first Bias at each cognitive stage For the first Weighting coefficients for each cognitive stage.

[0082] By introducing a cumulative variable of the whole chain deviation, it is equivalent to introducing a hierarchical control model based on deviation propagation constraints. Its core is that deviations are propagable, deviation propagation is subject to structural constraints, and system decisions are controlled by the whole chain deviation, which is different from the traditional single-point confidence model.

[0083] Furthermore, based on the above embodiments, at least one of the following methods is used to adjust the processing logic: Method 1, when the deviation value is higher than the first threshold, the processing result of the current cognitive stage is downweighted; Method 2, when the cumulative deviation values ​​of multiple consecutive previous cognitive stages exceed the second threshold, the output of the current cognitive stage is delayed or suspended; Method 3, when the cumulative deviation variable of the whole chain exceeds the third threshold, the execution of high-risk repair strategies is prohibited; Method 4, the algorithm parameters of the current cognitive stage are dynamically adjusted or different processing models are selected according to the deviation value.

[0084] Regarding Method 1 above, the first threshold is the boundary value for judging whether a single deviation value is too high. When the deviation value in the received cognitive state object is higher than the first threshold, it indicates that the credibility of the previous judgment is low, and the current cognitive stage can reduce the weight of its processing result.

[0085] Regarding method 2 above, the second threshold is the limit value for the cumulative deviation across multiple stages. When the cumulative deviation values ​​of multiple consecutive preceding cognitive stages exceed the second threshold, it indicates that the overall uncertainty of the cognitive chain is too high, and the output of the current cognitive stage should be delayed or suspended to avoid making decisions based on unreliable information.

[0086] Regarding method 3 above, the third threshold is the limit value for the total chain cumulative deviation. The total chain deviation cumulative variable is the weighted sum of the deviation values ​​at each stage, reflecting the overall uncertainty of the entire cognitive chain. When the total chain deviation cumulative variable exceeds the third threshold, it indicates that the system's understanding of the current fault is highly uncertain, and high-risk repair strategies should be prohibited.

[0087] Regarding method 4 above, the current cognitive stage can adaptively adjust the internal algorithm parameters or select different processing models based on the received deviation values ​​to match the current level of uncertainty.

[0088] For example, the constraint rule can be set as follows: if the first Bias in each cognitive stage If the deviation value exceeds the first threshold, or the cumulative deviation value of multiple consecutive preceding cognitive stages exceeds the second threshold, then: the confidence of the downstream layer will automatically decay, or enter the observation mode, or prohibit high-risk strategies; if the cumulative deviation variable of the whole chain of multiple consecutive cognitive stages exceeds the third threshold, then high intervention actions will be prohibited.

[0089] In the above scheme, when the deviation value is too high, the weight is reduced to avoid unreliable judgments affecting subsequent decisions; when the cumulative deviation of multiple stages exceeds the standard, the output is delayed or suspended to prevent erroneous responses based on unreliable information chains; when the cumulative deviation of the entire chain exceeds the threshold, high-risk repair is prohibited to ensure safety under highly uncertain conditions; dynamically adjusting algorithm parameters or selecting different models can adaptively match the processing strategy according to the real-time reliability level.

[0090] Furthermore, based on the above embodiments, traditional fault diagnosis methods assume the system is a static response model, while the embodiments of this application model the system as a discrete-time dynamic system. Specifically, the system state vector can be defined as: ; in, Let be the system state vector. This refers to the system's operating status. This is the cumulative deviation. In order to execute history, This represents the number of repair attempts.

[0091] The system evolution equation is: ; in, This is the current strategy action.

[0092] If the following conditions are met: the same anomaly occurs repeatedly in K windows, and the number of repair attempts is greater than M, If the value does not decrease, the system is determined to be in an oscillation state, and the following actions are taken: increase the trigger threshold, extend the observation period, switch to low intervention mode, or freeze automatic repair.

[0093] If a Lyapunov function exists And satisfy: If the system deviation does not increase monotonically, then during the normal convergence phase, It shows a decreasing trend.

[0094] This application provides a seven-layer closed-loop intelligent system architecture based on cognitive science, which is divided into the following stages based on cognitive nodes from low to high in the cognitive process: acquisition stage (corresponding to perception node), health judgment stage (corresponding to judgment node), diagnostic reasoning stage (corresponding to understanding node), causal inference stage (corresponding to reasoning node), strategy decision-making stage (corresponding to decision-making node), execution stage (corresponding to action node), and observation stage (corresponding to reflection node).

[0095] It is understandable that the above-mentioned stage division is not a business division, but a cognitive function division. Each stage addresses a different cognitive mismatch problem, and the stages form a directed cognitive evolution chain.

[0096] Based on this seven-layer closed-loop intelligent system architecture, please refer to... Figure 2 , Figure 2 A flowchart of another fault diagnosis method provided in this application embodiment, the fault diagnosis method specifically including: S201: In the health assessment phase, based on the collected operational data of the monitored system, determine whether there is any abnormality in the monitored system and generate the first cognitive state object.

[0097] The first cognitive state object includes the health status judgment result and the first deviation value, which represents the uncertainty of the health judgment stage.

[0098] S202: When the health status judgment result indicates an abnormality, enter the diagnostic reasoning stage, match the fault mode according to the health status judgment result, and generate a second cognitive state object.

[0099] The second cognitive state object includes the diagnostic reasoning result and the second deviation value, which characterizes the uncertainty of the diagnostic reasoning stage.

[0100] S203: In the causal inference stage, the root cause of the fault is identified based on the diagnostic reasoning results, and a third cognitive state object is generated.

[0101] The third cognitive state object includes the causal inference result and the third deviation value, which characterizes the uncertainty of the causal inference stage.

[0102] S204: During the execution phase, a repair operation is performed based on the causal inference results to generate a fourth cognitive state object.

[0103] The fourth cognitive state object includes the execution result and the fourth deviation value, which represents the uncertainty of the execution stage.

[0104] S205: During the observation phase, monitor the state of the monitored system after the repair operation is performed, and generate the fifth cognitive state object.

[0105] The fifth cognitive state object includes the observation results and the fifth deviation value that characterizes the uncertainty of the observation stage. The observation results are then used as feedback information to be transmitted back to one or more of the health judgment stage, diagnostic reasoning stage, and causal inference stage to adjust the processing logic of the corresponding stage.

[0106] In the aforementioned scheme, the fault diagnosis process in complex systems is divided into multiple cognitive stages, including data acquisition, health assessment, diagnosis, causal inference, strategy generation, execution control, and observational reflection. This avoids simplifying complex problems into a single response process. Simultaneously, deviations are introduced as explicit system states within the cognitive system, preventing them from being implicit within a single algorithm. Instead, they participate in multi-stage processing as perceptible, transmissible, and constrained elements of system operation. Furthermore, a structured closed-loop reflection mechanism implemented at the observation layer feeds execution results back into the system to verify the effectiveness of preceding cognitive paths, thereby improving the stability and convergence of system operation at the system level.

[0107] Furthermore, based on the above embodiments, before step S204, the fault diagnosis method provided in this application embodiment may further include: S206: In the strategy decision-making stage, the execution strategy is determined based on the causal inference results, and the sixth cognitive state object is generated.

[0108] The sixth cognitive state object includes the decision result and the sixth deviation value, which represents the uncertainty of the strategy decision-making stage.

[0109] Furthermore, based on the above embodiments, prior to S201, the fault diagnosis method provided in this application embodiment may further include: S207: During the data collection phase, the operational data of the monitored system is collected to generate the seventh cognitive state object.

[0110] The seventh cognitive state object includes the operational data of the monitored system and the seventh deviation value, which characterizes the uncertainty of the acquisition phase.

[0111] Furthermore, based on the above embodiments, existing fault diagnosis systems typically only determine whether the repair was successful after performing a repair action, lacking systematic verification of the cognitive path itself. Such systems assume by default that if the execution result improves, the preceding judgments must be correct; if the execution fails, only local strategies or parameters need to be adjusted. This approach ignores a crucial issue: an incorrect cognitive path may produce a superficial success under accidental conditions, thus being incorrectly reinforced.

[0112] Therefore, this application's embodiments verify whether the cognitive path is reasonable, whether the causal structure is stable, and whether the deviation truly converges; the system records the diagnostic path, causal graph structure, and strategy selection logic; during the observation phase, it compares the predicted impact range, the actual impact range, and the deviation change trend. If the result is successful but... If convergence fails, the system marks it as "accidental success" to prevent erroneous paths from being reinforced.

[0113] In summary, the embodiments of this application provide a consistency constraint control architecture on a multi-layered cognitive state space, in which each stage module can be implemented independently and a unified cognitive state interface standard is established. Furthermore, its deviations are structural-level control variables, and the entire chain of deviations drives decision constraints, thereby possessing dynamic stability control capabilities.

[0114] In other words, the embodiments of this application essentially propose a multi-layered cognitive closed-loop control theory for structured control of uncertainty: modeling fault self-healing as a multi-stage cognitive chain structure; defining a unified cognitive state object as a cross-layer interface standard; elevating deviations to structural-level control variables; establishing a cross-layer deviation propagation and accumulation constraint mechanism; introducing dynamic system modeling and convergence control ideas; and introducing a path-level reflection mechanism to avoid false reinforcement.

[0115] The following section uses the cloud security resource pool fault self-healing system as an example to introduce the solution provided in the embodiments of this application.

[0116] During the data collection phase, the system gathers operational data from various monitoring sources, including performance metrics, log information, and call chain tracing data. During this process, the system simultaneously monitors for data latency or missing data, consistency levels between different data sources, and whether metric fluctuations exceed historical stability ranges. The deviation information generated during this phase is used to describe the current system's observability level.

[0117] During the health assessment phase, the system determines the health status of each service and the overall system based on the collected data. When multiple indicators show slight anomalies but do not form a clear fault mode, or when different monitoring signals point to inconsistent conclusions, the system marks the assessment results as low-confidence anomalies and avoids directly entering the forced repair process.

[0118] During the diagnostic reasoning phase, the system matches potential faults based on symptom patterns, such as service instance anomalies, network latency, and resource contention. When multiple fault patterns simultaneously meet the current symptom characteristics, the system identifies a high degree of uncertainty in the diagnosis and passes this uncertainty as bias information forward.

[0119] For the causal inference phase, the system further analyzes the dependencies and call paths between services to attempt to identify possible sources of fault propagation. When a fault may be caused by multiple upstream components, or when the call chain information is incomplete or delayed, the system marks the inference result of this phase as causal uncertainty, rather than forcibly selecting a single root cause.

[0120] During the strategy decision-making phase, when formulating a repair strategy, the system comprehensively considers the current diagnostic results, the accumulated deviation information of each stage, and the potential risks of the repair actions. When the system judges that the current cognitive deviation is high, it will prioritize the repair actions that have the least impact on the business, or trigger a manual confirmation process, or suspend the repair and strengthen monitoring.

[0121] During the execution and observation phases, the system continuously monitors its status after performing repair operations (e.g., service restart, traffic switching, configuration rollback). When the actual results differ from expectations, the system records the execution deviation and uses it as an important basis for subsequent cognitive corrections.

[0122] As can be seen, the embodiments of this application can effectively reduce the probability of false alarms and false repairs, alleviate the interference of alarm storms on operation and maintenance personnel, improve the stability of the system in complex dependent environments, reduce reliance on human experience in the fault handling process, and improve the overall operational reliability of the system.

[0123] Please refer to Figure 3 , Figure 3This is a structural block diagram of a fault diagnosis device provided in an embodiment of this application. The fault diagnosis device includes: a generation module 301, used to generate a cognitive state object corresponding to each cognitive stage of a pre-constructed cognitive chain, wherein the cognitive chain is used to perform fault processing on the monitored system, the cognitive chain includes multiple cognitive stages executed sequentially, the multiple cognitive stages include a health judgment stage and a diagnostic reasoning stage, the health judgment stage is used to determine whether the monitored system has a fault, the diagnostic reasoning stage is used to determine the fault type, and the cognitive state object includes the judgment value of the cognitive stage and a deviation value characterizing the uncertainty of the cognitive stage; and a transmission module 302, used to transmit the cognitive state object generated by the current cognitive stage to the next cognitive stage, so that the next cognitive stage adjusts its own processing logic according to the deviation value in the cognitive state object.

[0124] In the above scheme, a cognitive chain is constructed that includes a health assessment stage and a diagnostic reasoning stage. At each stage, a cognitive state object containing assessment and deviation values ​​is generated and passed to the next stage, enabling downstream stages to adjust their processing logic based on upstream deviation values. Therefore, by extending the fault diagnosis process from a single-point decision to a multi-stage cognitive process, the simplification of complex problems into a single response process is avoided, thereby improving the reliability and accuracy of fault diagnosis.

[0125] Furthermore, based on the above embodiments, the cognitive state object also includes at least one of confidence level, source layer identifier, timestamp, and upstream dependency chain.

[0126] In the above scheme, by further introducing at least one of confidence level, source layer identifier, timestamp and upstream dependency chain into the cognitive state object, the entire cognitive process becomes traceable and interpretable.

[0127] Furthermore, based on the above embodiments, the deviation value is calculated by weighting the internal consistency error and the upstream matching error of the cognitive stage.

[0128] In the above scheme, the deviation value is quantified as a weighted sum of internal consistency error and upstream matching error, which transforms the deviation from an abstract concept into a calculable engineering variable. This ensures that the deviation value can truly reflect the uncertainty of the cognitive state at each stage, thereby improving the reliability and accuracy of fault diagnosis.

[0129] Furthermore, based on the above embodiments, the processing logic is adjusted in at least one of the following ways: when the deviation value is higher than a first threshold, the processing result of the current cognitive stage is downweighted; when the cumulative deviation values ​​of multiple consecutive preceding cognitive stages exceed a second threshold, the output of the current cognitive stage is delayed or suspended; when the cumulative deviation variable of the entire chain exceeds a third threshold, high-risk repair strategies are prohibited; the algorithm parameters of the current cognitive stage are dynamically adjusted or different processing models are selected according to the deviation value.

[0130] In the above scheme, when the deviation value is too high, the weight is reduced to avoid unreliable judgments affecting subsequent decisions; when the cumulative deviation of multiple stages exceeds the standard, the output is delayed or suspended to prevent erroneous responses based on unreliable information chains; when the cumulative deviation of the entire chain exceeds the threshold, high-risk repair is prohibited to ensure safety under highly uncertain conditions; dynamically adjusting algorithm parameters or selecting different models can adaptively match the processing strategy according to the real-time reliability level.

[0131] Furthermore, based on the above embodiments, the cognitive chain also includes a causal inference stage and an execution stage, wherein the causal inference stage is used to identify the root cause of the failure, and the execution stage is used to execute the repair operation in the repair strategy.

[0132] In the above scheme, the cognitive chain is extended to the causal inference stage and the execution stage, so that the system can not only judge the fault type, but also identify the root cause of the fault and perform repair operations, forming a complete closed loop from anomaly discovery to root cause analysis and repair action, thereby improving the effectiveness and efficiency of fault handling.

[0133] Furthermore, based on the above embodiments, the cognitive chain further includes: an observation phase, which is used to monitor the state of the monitored system after the repair operation is performed, generate feedback information, and transmit the feedback information in reverse to one or more preceding cognitive phases.

[0134] In the above scheme, the cognitive chain is extended to the observation stage and the feedback information is transmitted back to the preceding cognitive stage, so that each stage can adjust its own model or logic according to the actual execution effect, thereby ensuring the stability and reliability of the system in the long term.

[0135] Furthermore, based on the above embodiments, the cognitive chain further includes a strategy decision-making stage, which is used to generate a repair strategy based on the output of the diagnostic reasoning stage and / or the causal inference stage. In the above scheme, extending the cognitive chain to the strategy decision-making stage, enabling it to generate a repair strategy based on the output of the diagnostic reasoning and / or causal inference stages, achieves intelligent decision-making, thereby improving the reliability and accuracy of fault diagnosis.

[0136] Furthermore, based on the above embodiments, the cognitive chain also includes a collection phase, which is used to collect the operating data of the monitored system.

[0137] In the above scheme, the cognitive chain is extended to the collection stage, thereby enabling the proactive acquisition of multi-source operational data of the monitored system. This provides upstream basis for the quantification of uncertainty in subsequent stages, thus avoiding overall judgment errors due to data quality issues.

[0138] Furthermore, based on the above embodiments, the fault diagnosis device 300 further includes: a health judgment module, used in the health judgment stage to determine whether there is an anomaly in the monitored system based on the collected operating data of the monitored system, and generate a first cognitive state object, wherein the first cognitive state object includes a health state judgment result and a first deviation value characterizing the uncertainty of the health judgment stage; a fault diagnosis module, used to enter the diagnostic reasoning stage when the health state judgment result indicates the presence of an anomaly, and generate a second cognitive state object by matching a fault mode according to the health state judgment result, wherein the second cognitive state object includes a diagnostic reasoning result and a second deviation value characterizing the uncertainty of the diagnostic reasoning stage; and a causal inference module, used in the causal inference stage to identify the root cause of the fault based on the diagnostic reasoning result. A third cognitive state object is generated, wherein the third cognitive state object includes a causal inference result and a third deviation value characterizing the uncertainty of the causal inference stage; an execution module is used to perform a repair operation based on the causal inference result during the execution stage, generating a fourth cognitive state object, wherein the fourth cognitive state object includes the execution result and a fourth deviation value characterizing the uncertainty of the execution stage; an observation module is used to monitor the state of the monitored system after the repair operation is performed during the observation stage, generating a fifth cognitive state object, wherein the fifth cognitive state object includes the observation result and a fifth deviation value characterizing the uncertainty of the observation stage, and the observation result is transmitted as feedback information to one or more of the health judgment stage, diagnostic reasoning stage, and causal inference stage to adjust the processing logic of the corresponding stage.

[0139] In the aforementioned scheme, the fault diagnosis process in complex systems is divided into multiple cognitive stages, including data acquisition, health assessment, diagnosis, causal inference, strategy generation, execution control, and observational reflection. This avoids simplifying complex problems into a single response process. Simultaneously, deviations are introduced as explicit system states within the cognitive system, preventing them from being implicit within a single algorithm. Instead, they participate in multi-stage processing as perceptible, transmissible, and constrained elements of system operation. Furthermore, a structured closed-loop reflection mechanism implemented at the observation layer feeds execution results back into the system to verify the effectiveness of preceding cognitive paths, thereby improving the stability and convergence of system operation at the system level.

[0140] Please refer to Figure 4 , Figure 4 This application provides a structural block diagram of an electronic device 400, which includes at least one processor 401, at least one communication interface 402, at least one memory 403, and at least one communication bus 404. The communication bus 404 enables direct communication between these components, the communication interface 402 facilitates signaling or data communication with other node devices, and the memory 403 stores machine-readable instructions executable by the processor 401. When the electronic device 400 is running, the processor 401 communicates with the memory 403 via the communication bus 404, and when the machine-readable instructions are invoked by the processor 401, the aforementioned fault diagnosis method is executed.

[0141] For example, the processor 401 in this embodiment of the application can read a computer program from the memory 403 via the communication bus 404 and execute the computer program to implement the following method: In each cognitive stage of a pre-built cognitive chain, a cognitive state object corresponding to that cognitive stage is generated, wherein the cognitive chain is used to perform fault handling on the monitored system, the cognitive chain includes multiple cognitive stages executed sequentially, the multiple cognitive stages include a health judgment stage and a diagnostic reasoning stage, the health judgment stage is used to determine whether the monitored system has a fault, the diagnostic reasoning stage is used to determine the fault type, the cognitive state object includes the judgment value of the cognitive stage and a deviation value characterizing the uncertainty of the cognitive stage; the cognitive state object generated by the current cognitive stage is passed to the next cognitive stage, so that the next cognitive stage adjusts its own processing logic according to the deviation value in the cognitive state object.

[0142] The processor 401 comprises one or more, and can be an integrated circuit chip with signal processing capabilities. The processor 401 can be a general-purpose processor, including a Central Processing Unit (CPU), a Microcontroller Unit (MCU), a Network Processor (NP), or other conventional processors; it can also be a special-purpose processor, including a Neural-network Processing Unit (NPU), a Graphics Processing Unit (GPU), a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. Furthermore, when there are multiple processors 401, some can be general-purpose processors, and others can be special-purpose processors.

[0143] The memory 403 includes one or more, which may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0144] Understandable. Figure 4 The structure shown is for illustrative purposes only; the electronic device 400 may also include components that are more advanced than those shown. Figure 4 The more or fewer components shown, or having the same Figure 4 The different configurations shown. Figure 4The components shown can be implemented using hardware, software, or a combination thereof. In the embodiments of this application, electronic device 400 can be, but is not limited to, physical devices such as desktop computers, laptops, smartphones, smart wearable devices, and in-vehicle devices, or virtual devices such as virtual machines. Furthermore, electronic device 400 is not necessarily a single device; it can be a combination of multiple devices, such as a server cluster, etc.

[0145] This application also provides a computer program product, including a computer program stored on a computer-readable storage medium. The computer program includes computer program instructions. When the computer program instructions are executed by a computer, the computer can perform the steps of the fault diagnosis method described in the above embodiments, such as: S101: In each cognitive stage of a pre-built cognitive chain, generating a cognitive state object corresponding to that cognitive stage. S102: Passing the cognitive state object generated in the current cognitive stage to the next cognitive stage, so that the next cognitive stage adjusts its processing logic according to the deviation value in the cognitive state object.

[0146] This application also provides a computer-readable storage medium that stores computer program instructions. When the computer program instructions are executed by a computer, the computer performs the fault diagnosis method described in the foregoing method embodiments.

[0147] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0148] Furthermore, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0149] Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0150] It should be noted that if the function is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0151] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0152] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A fault diagnosis method, characterized in that, include: In each cognitive stage of the pre-built cognitive chain, a cognitive state object corresponding to that cognitive stage is generated. The cognitive chain is used to handle faults in the monitored system. The cognitive chain includes multiple cognitive stages executed sequentially. The multiple cognitive stages include a health judgment stage and a diagnostic reasoning stage. The health judgment stage is used to determine whether there is a fault in the monitored system. The diagnostic reasoning stage is used to determine the fault type. The cognitive state object includes the judgment value of that cognitive stage and a deviation value that characterizes the uncertainty of that cognitive stage. The cognitive state object generated in the current cognitive stage is passed to the next cognitive stage so that the next cognitive stage adjusts its processing logic according to the deviation value in the cognitive state object.

2. The fault diagnosis method according to claim 1, characterized in that, The cognitive state object also includes at least one of confidence level, source layer identifier, timestamp, and upstream dependency chain.

3. The fault diagnosis method according to claim 1, characterized in that, The deviation value is calculated by weighting the internal consistency error and the upstream matching error of the cognitive stage.

4. The fault diagnosis method according to claim 1, characterized in that, The processing logic can be adjusted in at least one of the following ways: When the deviation value is higher than the first threshold, the processing result of the current cognitive stage is downweighted. When the cumulative deviation value of multiple consecutive previous cognitive stages exceeds the second threshold, the output of the current cognitive stage is delayed or suspended. When the cumulative variable of the whole chain deviation exceeds the third threshold, high-risk repair strategies are prohibited. The algorithm parameters for the current cognitive stage are dynamically adjusted or different processing models are selected based on the deviation value.

5. The fault diagnosis method according to any one of claims 1-4, characterized in that, The cognitive chain also includes a causal inference stage and an execution stage, wherein the causal inference stage is used to identify the root cause of the failure and the execution stage is used to execute the repair operations in the repair strategy.

6. The fault diagnosis method according to claim 5, characterized in that, The cognitive chain further includes an observation phase, which is used to monitor the state of the monitored system after the repair operation is performed, generate feedback information, and transmit the feedback information in reverse to one or more preceding cognitive phases.

7. The fault diagnosis method according to claim 6, characterized in that, The cognitive chain further includes a strategy decision-making stage, which is used to generate the repair strategy based on the output of the diagnostic reasoning stage and / or the causal inference stage.

8. The fault diagnosis method according to claim 7, characterized in that, The cognitive chain also includes a collection phase, which is used to collect the operational data of the monitored system.

9. The fault diagnosis method according to claim 6, characterized in that, The fault diagnosis method further includes: In the health assessment stage, the system is assessed for any abnormalities based on the collected operational data of the monitored system, and a first cognitive state object is generated. The first cognitive state object includes a health status assessment result and a first deviation value that characterizes the uncertainty of the health assessment stage. When the health status judgment result indicates an abnormality, the diagnostic reasoning stage is entered. Based on the health status judgment result, a fault mode is matched to generate a second cognitive state object. The second cognitive state object includes a diagnostic reasoning result and a second deviation value that characterizes the uncertainty of the diagnostic reasoning stage. In the causal inference stage, the root cause of the failure is identified based on the diagnostic inference result, and a third cognitive state object is generated, wherein the third cognitive state object includes the causal inference result and a third deviation value characterizing the uncertainty of the causal inference stage; During the execution phase, a repair operation is performed based on the causal inference result to generate a fourth cognitive state object, wherein the fourth cognitive state object includes the execution result and a fourth deviation value characterizing the uncertainty of the execution phase; During the observation phase, the state of the monitored system after the repair operation is performed is monitored, and a fifth cognitive state object is generated. The fifth cognitive state object includes the observation results and a fifth deviation value that characterizes the uncertainty of the observation phase. The observation results are then transmitted as feedback information to one or more of the health judgment phase, diagnostic reasoning phase, and causal inference phase to adjust the processing logic of the corresponding phase.

10. A computer program product, characterized in that, It includes computer program instructions, which, when read and executed by a processor, perform the fault diagnosis method as described in any one of claims 1-9.

11. An electronic device, characterized in that, include: Processor, memory, and bus; The processor and the memory communicate with each other via the bus; The memory stores computer program instructions that can be executed by the processor, and the processor can execute the fault diagnosis method as described in any one of claims 1-9 by calling the computer program instructions.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a computer, cause the computer to perform the fault diagnosis method as described in any one of claims 1-9.

Images