A method and system for intelligent fault diagnosis of mainboard based on artificial intelligence

By building a task numbering chain and an exception propagation path map, the identification and repair of inter-task linkage exceptions during concurrent operation of the motherboard's multi-tasks is solved, and high-precision fault diagnosis and system stability are achieved.

CN120144410BActive Publication Date: 2025-08-08SHENZHEN COLORFUL YUGONG TECH DEV CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202510600326.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-05-12
Publication Date
2025-08-08
Estimated Expiration
2045-05-12

AI Technical Summary

Technical Problem

The existing motherboard fault diagnosis methods are difficult to effectively capture inter-task linkage abnormalities under the concurrent operation of multi-tasks, and lack dynamic construction of anomalies, resulting in low abnormal positioning accuracy and difficult to support intelligent repair strategies.

Method used

By building a task numbering chain, continuously sampling the running state parameters, calculating the state feature vector offset between adjacent task units, building an abnormal propagation path map, and task replacement and state reconstruction are carried out based on the path risk score to achieve dynamic fault diagnosis.

Benefits of technology

It significantly improves the accuracy and response speed of fault diagnosis, and enhances the robustness and continuous operation capabilities of the motherboard task execution system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120144410B_ABST
    Figure CN120144410B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for intelligent fault diagnosis of a motherboard based on artificial intelligence, which relates to the field of fault diagnosis technology. It includes constructing a task number chain based on the call sequence relationship; continuously sampling the operating state parameters, and constructing a task state feature vector table in combination with the current task number; calculating the state feature vector offset between adjacent task units pair by pair, and if the offset exceeds the set tolerance threshold, adding it to the offset anomaly table as a suspicious mark; through path reasoning, constructing an abnormal propagation path map, and calculating the path risk score based on the structural correlation between task units and the historical state interaction frequency; for abnormal boundaries whose path risk scores exceed the preset value, selecting the optimal alternative number under the structural constraint from the task number chain, and deploying it to the corresponding number position of the current task unit. The present invention realizes the dynamic replacement of fault boundaries and the local reconstruction of system status, and enhances the robustness and continuous operation capability of the motherboard task execution system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to an artificial intelligence-based mainboard intelligent fault diagnosis method and system. Background Art

[0002] With the increasing complexity of computer hardware systems, the motherboard, as a core component of information processing systems, faces a significant challenge. Its operational stability and multitasking capabilities directly impact the performance and reliability of the entire system. Traditional motherboard fault diagnosis methods rely heavily on pre-set logical rules, exception code capture mechanisms (such as BIOS self-test and POST codes), and manual judgment. However, with the widespread adoption of embedded systems, parallel computing, and intelligent devices, the tasks running on the motherboard are increasingly characterized by high concurrency and strong interactions, making fault manifestations more complex and dynamic. In recent years, with the rapid development of artificial intelligence technologies, particularly deep learning, sequence modeling, and intelligent decision-making reasoning, researchers have achieved intelligent monitoring of hardware system operating status through state modeling and abnormal behavior identification. While these methods have improved the automation and accuracy of anomaly detection to a certain extent, most remain at the feature recognition level, lacking a deep understanding of inter-task linkage anomalies and the modeling of their propagation paths.

[0003] Existing fault diagnosis methods still have many shortcomings when dealing with the concurrent operation of multiple tasks in the motherboard. First, most models only perform state sampling and analysis on a single task or module, lack modeling of collaborative behaviors and time series dependencies between tasks, and cannot effectively capture linkage anomalies at task boundaries. Secondly, current anomaly detection is mostly based on a global threshold mechanism for single-point judgment, ignoring the local offset propagation phenomenon caused by structural coupling and state transfer between motherboard running tasks, resulting in low anomaly positioning accuracy and difficulty in supporting intelligent repair strategies. In addition, existing methods often focus on anomaly identification, but lack a reasoning mechanism that is integrated with the task structure model. It is impossible to dynamically construct anomaly path maps based on historical operation data, and thus it is impossible to perform targeted replacement and task continuation.

[0004] These problems seriously limit the adaptability and intelligence level of the fault diagnosis system in multi-task complex environments. Therefore, an artificial intelligence-based motherboard intelligent fault diagnosis solution is needed. Summary of the Invention

[0005] In view of the problems existing in the existing motherboard intelligent diagnosis technology, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is how to effectively improve the intelligence level and response efficiency of system fault identification.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In the first aspect, the present invention provides an artificial intelligence-based motherboard intelligent fault diagnosis method, which includes collecting information on multiple task units running concurrently in the motherboard, and constructing a task number chain based on the calling sequence relationship; during the task unit operation cycle, continuously sampling the operation status parameters, and combining with the current task number, performing synchronous embedding to construct a task state feature vector table; according to the order of the task number chain, calculating the state feature vector offset between adjacent task units one by one, if the offset exceeds the set tolerance threshold, it is determined that there is a task boundary with linkage abnormality, and added to the offset abnormality table as a suspicious mark; calling the historical task number chain and state evolution sequence, through path reasoning, constructing an abnormal propagation path map, and calculating the path risk score based on the structural correlation between task units and the historical state interaction frequency; for the abnormal boundary whose path risk score exceeds the preset value, selecting the optimal alternative number under the structural constraint from the task number chain, deploying it to the corresponding number position of the current task unit, and synchronously updating the task feature vector table to complete task continuation and state reconstruction.

[0009] As a preferred solution of the artificial intelligence-based motherboard intelligent fault diagnosis method described in the present invention, the construction of the task number chain includes: collecting the call relationship logs of the active task units in the motherboard, sorting the call events by timestamps, and constructing a task access sequence; based on the task access sequence, marking the dependency direction identifier between any consecutive task pairs, and setting a context window to identify whether the task pairs are in a logically associated neighborhood, and forming an initial dependency mapping table; normalizing the call frequency of the task pairs in the initial dependency mapping table, and setting a frequency threshold to eliminate low-frequency sporadic dependency paths and generate a task number chain.

[0010] As a preferred solution of the artificial intelligence-based mainboard intelligent fault diagnosis method described in the present invention, the generation of the task state feature vector table includes: periodically sampling the operating status of each task unit and organizing it into a state original matrix; performing serial position encoding on each parameter in the state original matrix, and generating a numbered alignment feature block in combination with the current task number; compressing the alignment feature block of each task to form an embedded vector, and introducing structural position encoding as an additional feature item so that the state vector retains the task chain sequence information; performing time segment splicing on the embedded vector of each task in a continuous sampling period to form a time series feature sequence, and adding it to the corresponding number position of the task state feature vector table.

[0011] As a preferred solution of the artificial intelligence-based motherboard intelligent fault diagnosis method of the present invention, wherein: the pair-by-pair calculation of the state feature vector offset between adjacent task units includes: extracting continuous task pairs from the task number chain and obtaining the corresponding state vector sequence in the task state feature vector table and ; In the vector pair In , the embedded state vector at the corresponding time point is selected and expanded into sub-dimension vectors respectively, and the offset is calculated for all dimensions; and the total offset of the state vector pair is calculated by weighted accumulation.

[0012] As a preferred solution of the artificial intelligence-based motherboard intelligent fault diagnosis method described in the present invention, the construction of the abnormal propagation path map includes: extracting all task pairs marked with abnormalities from the offset abnormality table, adding corresponding nodes in the map, and establishing initial directed edges between the task pairs as potential propagation starting points; according to the structural connection diagram preset by the motherboard, identifying all path segments that are structurally associated with the abnormal task pairs in the map, and calculating the topological distance and coupling tightness index with the abnormal task pairs; for each highly coupled path segment, tracing back the key dimension evolution trend of the feature vector along the path direction to determine whether there is a delayed state mutation. If so, the path segment is marked as an abnormal evolution path; all path segments that meet the topological coupling and delayed transmission characteristics are merged to form an abnormal propagation path map.

[0013] As a preferred solution of the artificial intelligence-based motherboard intelligent fault diagnosis method described in the present invention, the calculation of the path risk score includes: extracting all path edges from the abnormal propagation path map and initializing the scoring item structure for each path edge; calculating the structural dependency of task pairs based on the task unit structure connection information; counting the frequency of task pairs in the cross-influence of feature vectors from the historical state evolution records; using a joint scoring function to generate a risk score for each path edge, and sorting and screening high-risk paths.

[0014] As a preferred solution of the artificial intelligence-based motherboard intelligent fault diagnosis method described in the present invention, wherein: for path segments whose path risk scores exceed the set threshold, the connected task numbers are extracted as abnormally high-risk pairs to form an index sequence of numbers to be replaced; in the system candidate number pool, the set of task numbers that have a structural connection relationship with the current number position to be replaced is screened, and the collaborative stability index with the replaced number in historical execution is calculated, and the task number is selected and replaced according to the maximum principle; the selected task number is written into the task number chain according to the index position, and the task status feature vector table is updated accordingly.

[0015] In a second aspect, the present invention provides an artificial intelligence-based motherboard intelligent fault diagnosis system, which includes:

[0016] The task collection module is used to collect information about multiple task units running concurrently in the mainboard and build a task number chain based on the calling sequence;

[0017] The state embedding module is used to continuously sample the running state parameters during the task unit running cycle, and perform synchronous embedding in combination with the current task number to construct the task state feature vector table;

[0018] The anomaly detection module is used to calculate the state feature vector offset between adjacent task units pair by pair based on the order of the task number chain. If the offset exceeds the set tolerance threshold, it is determined that the task boundary has a linkage anomaly and is added to the offset anomaly table as a suspicious point mark;

[0019] The path reasoning module is used to call the historical task number chain and state evolution sequence, build an abnormal propagation path map through path reasoning, and calculate the path risk score based on the structural correlation between task units and the historical state interaction frequency;

[0020] The task repair module is used to select the optimal alternative number under structural constraints from the task number chain for abnormal boundaries where the path risk score exceeds the preset value, deploy it to the corresponding number position of the current task unit, and synchronously update the task feature vector table to complete task continuation and state reconstruction.

[0021] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the artificial intelligence-based motherboard intelligent fault diagnosis method as described in the first aspect of the present invention are implemented.

[0022] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the artificial intelligence-based motherboard intelligent fault diagnosis method as described in the first aspect of the present invention are implemented.

[0023] The beneficial effects of the present invention are as follows: the present invention introduces an artificial intelligence mechanism to perform intelligent perception, offset detection, and abnormal link reasoning on the operating status of concurrent tasks within the mainboard, effectively realizing the early discovery and precise positioning of potential linkage faults in complex multi-task systems. Through state vector modeling and path graph reasoning technology, the propagation trend of anomalies can be dynamically tracked, thereby significantly improving the accuracy and response speed of fault diagnosis. Compared with traditional rule-based or single-point monitoring methods, the present invention can achieve comprehensive judgment in the multi-dimensional space of structural association and state evolution, and has stronger adaptability and intelligence. Through the replacement task repair mechanism, dynamic replacement of fault boundaries and local reconstruction of system status are realized, enhancing the robustness and continuous operation capability of the mainboard task execution system. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 The figure is a flow chart of an artificial intelligence-based intelligent fault diagnosis method for a motherboard;

[0026] Figure 2 A flowchart for constructing an abnormal propagation path map for an AI-based motherboard intelligent fault diagnosis method;

[0027] Figure 3 This is the structural block diagram of the motherboard intelligent fault diagnosis system based on artificial intelligence. DETAILED DESCRIPTION

[0028] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0029] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0030] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0031] As mentioned in the above background technology, the existing fault diagnosis methods still have many shortcomings when dealing with the concurrent operation of multiple tasks in the motherboard. First, most models only perform state sampling and analysis on a single task or module, lack modeling of collaborative behaviors and time series dependencies between tasks, and cannot effectively capture linkage anomalies at the task boundaries. Secondly, current anomaly detection is mostly based on a global threshold mechanism for single-point judgment, ignoring the local offset propagation phenomenon caused by structural coupling and state transfer between the mainboard running tasks, resulting in low anomaly positioning accuracy and difficulty in supporting intelligent repair strategies. In addition, existing methods often focus on anomaly identification, but lack a reasoning mechanism that is integrated with the task structure model. It is impossible to dynamically construct anomaly path maps based on historical operation data, and thus it is impossible to perform targeted replacement and task continuation. These problems seriously limit the adaptability and intelligence level of the fault diagnosis system in a complex multi-task environment. Therefore, there is a need for an intelligent fault diagnosis solution for the mainboard based on artificial intelligence.

[0032] Figure 1 FIG. 1 is a flow chart of a method for intelligent fault diagnosis of a motherboard based on artificial intelligence according to an embodiment of the present invention. Figure 1 As shown, the artificial intelligence-based intelligent fault diagnosis method for the motherboard includes:

[0033] S1: Collect information on multiple task units running concurrently in the mainboard, and build a task number chain based on the calling sequence.

[0034] Preferably, the construction of the task number chain includes: collecting the call relationship logs of the active task units in the main board, sorting the call events by timestamps, and constructing a task access sequence; based on the task access sequence, marking the dependency direction identifier between any consecutive task pairs, and setting a context window to identify whether the task pairs are in a logically associated neighborhood, and forming an initial dependency mapping table; normalizing the call frequency of the task pairs in the initial dependency mapping table, and setting a frequency threshold to eliminate low-frequency sporadic dependency paths and generate a task number chain.

[0035] Specifically, in the process of constructing the task number chain, it is first necessary to collect information on all the concurrently running task units in the mainboard. The task unit refers to a functional unit that runs between multiple modules inside the mainboard, has clear start and end conditions, and has data processing capabilities. The purpose of information collection is to obtain the running status and call behavior of each task unit, and then form a complete call trace basic data. Currently, conventional information collection methods mostly use periodic polling or unified scheduling management strategies to obtain task status, but due to the high degree of asynchrony in the scheduling method of tasks inside the mainboard, traditional methods are often difficult to capture true concurrent interactive behaviors, especially when dealing with high-speed task switching and temporary resource scheduling, which can easily cause data loss or sequence deviation. The present invention introduces an event-driven and time-triggered dual collection mechanism in the information collection stage, specifically including embedding a call event monitoring module in the mainboard task scheduling engine, generating event nodes for task call, suspension, recovery and other state changes, supplemented by main frequency-level timestamp annotation, to ensure that each call log has accurate start and end records and time series labels, thereby laying the foundation for subsequent call sequence restoration. Furthermore, to prevent non-task-level interference events (such as peripheral device responses and I / O interrupts) from contaminating call log data, the present invention pre-classifies collected events, retaining only the main path call information triggered by the main task scheduling, forming a complete and logically closed set of task call logs. After completing event collection, the system sorts the collected call events in ascending order according to the embedded timestamp field to construct a task access sequence.

[0036] On the basis of constructing the access chain, the dependency relationship of any group of continuous task unit pairs will be identified and marked. This process includes two sub-operations: one is to add dependency direction identification, that is, to determine whether there is a clear resource call, data transfer or scheduling trigger relationship between tasks. If it is established, a direction arrow is added between the task pairs and marked as strong dependency, otherwise it is marked as weak dependency; the second is to set a context window to identify whether the task pairs belong to the same logical execution segment or control block. If the time interval between task pairs is lower than the threshold and appears multiple times in the same context, they are identified as logically associated neighborhoods. Traditional dependency analysis often relies only on static function graphs or module structure graphs for relationship inference, lacks the fusion judgment of dynamic call context, and has a large number of misjudgment or missed judgment problems. The present invention introduces a context window mechanism to dynamically fuse runtime context information, so that dependency judgment is closer to the actual logical path of the task, and significantly improves the accuracy of dependency identification.

[0037] S2: During the task unit operation cycle, the operation status parameters are continuously sampled, and combined with the current task number, synchronous embedding is performed to construct the task status feature vector table.

[0038] It's important to note that in a motherboard's multitasking, concurrent execution environment, to quantify the behavior of each task unit, it's necessary to continuously sample its operating state parameters throughout the task's operating cycle, constructing a set of high-dimensional feature vectors with both timing characteristics and structural signatures. The operating cycle is the complete lifecycle of a task unit, from activation and scheduling to completion and return of control to the scheduler. During this cycle, a task unit undergoes multiple state transitions.

[0039] Preferably, the generation of the task state feature vector table includes: periodically sampling the operating state of each task unit and organizing it into a state original matrix; performing sequence position encoding on each parameter sequence in the state original matrix, and generating numbered alignment feature blocks in combination with the current task number; compressing the alignment feature blocks of each task to form an embedded vector, and introducing structural position encoding as an additional feature item so that the state vector retains the task chain sequence information; splicing the embedded vectors of each task in continuous sampling periods into time segments to form a time series feature sequence, and adding it to the corresponding numbered position of the task state feature vector table.

[0040] Among them, the sequence position encoding refers to the absolute position encoding strategy in the Transformer architecture, and uses sine and cosine functions or linear increment functions to mark the position of each time slice, so that the subsequent processing module can identify the state evolution order corresponding to different time slices. After completing the position encoding, the current task number is introduced into the feature combination of each state sample to form a feature block with task number alignment. It should be noted that the task number is not introduced simply as an additional label, but through the embedding matrix form, the number is converted into a vector representation of a fixed dimension and spliced with the state parameter sequence, thereby ensuring that the feature representation of different tasks at the same sampling time point has a task unique identification. This mechanism effectively prevents semantic confusion of multi-task data under high-frequency concurrent sampling and improves the ability to decouple features between tasks.

[0041] To further compress feature dimensions and improve representation capabilities, a compression operation is performed on the numbered, aligned feature blocks for each task to form an embedding vector of uniform dimensionality. This compression operation can be implemented using linear dimensionality reduction (e.g., PCA) or nonlinear feature mapping (e.g., autoencoders). The specific implementation is flexible based on system resources and performance requirements and is not limited in this embodiment.

[0042] Furthermore, based on the ordered structure of the task number chain, a vector comparison is performed between consecutive tasks i and i+1 in the vector sequence in the task state feature vector table. If the feature compression mismatch rate is detected to be higher than the expansion threshold, a structural direction feature difference term is added to the temporal feature sequence and marked as a correction vector. This correction mechanism has a dual role: first, it serves as a compensation mechanism for structural omissions, correcting the information loss caused by nonlinear compression; second, it serves as a feature correction operation before anomaly detection, significantly improving the consistency and accuracy of vector comparisons.

[0043] It should be noted that by splicing time series segments, the short-term state evolution trend is retained, so that subsequent offset detection can not only perceive instantaneous anomalies, but also has the ability to analyze micro-periodic anomaly changes; and the introduction of structural direction difference as a feature correction amount can effectively compensate for the vector omission phenomenon caused by the structural nonlinear relationship during the embedding compression process, thereby improving the accuracy of anomaly boundary detection.

[0044] For example, the calculation of the characteristic compression mismatch rate can be expressed as:

[0045] ;

[0046] in, For the task and Compression mismatch ratio; and Respectively for tasks and In the compressed map dimension Embedded values on ; is the total number of embedding dimensions; A small constant to prevent the denominator from being zero.

[0047] S3: According to the order of the task number chain, the state feature vector offset between adjacent task units is calculated pair by pair. If the offset exceeds the set tolerance threshold, it is determined that there is a task boundary with linkage anomaly and it is added to the offset anomaly table as a suspicious mark.

[0048] After constructing the task state feature vector table, in order to identify possible behavioral linkage anomalies that may occur during the execution of the task chain, the present invention performs state feature vector offset analysis and mines possible abnormal boundary points by comparing the state change trends between adjacent task units one by one.

[0049] Preferably, the state feature vector offset between adjacent task units is calculated pair by pair, which includes: extracting consecutive task pairs from the task number chain and obtaining the corresponding state vector sequence in the task state feature vector table. and ; In the vector pair , select the corresponding time point The embedded state vector under is expanded into sub-dimension vectors, and the offset is calculated for all dimensions. , where the offset This can be expressed as the absolute value of the vector difference and calculated as the total offset of the state vector pair through weighted accumulation, where the weights are adaptively set based on the importance of the dimensions or information entropy. If the total offset is too large, it indicates a significant transition between the behavioral states of the two tasks, which may constitute a potential linkage anomaly signal.

[0050] Because state offsets are not only reflected in absolute value changes but are also closely related to the direction of the task flow, even large numerical offsets can be considered normal if their direction is consistent with expectations. Therefore, after calculating the total offset, the present invention introduces a directional consistency determination mechanism.

[0051] Furthermore, the calculated total offset is compared with the reverse reference value according to the direction of the task flow. If the directions are inconsistent, the direction correction item is performed on the calculated total offset value to generate a correction value; the correction value is compared with the system tolerance threshold. If the correction value is greater than or equal to the system tolerance threshold, the number pair is added to the offset anomaly table, and the trigger dimension and directionality information are recorded as a suspicious mark.

[0052] For example, the correction value The generation of can be expressed as:

[0053] ;

[0054] in, is the total offset value; is the directional correction factor; is the vector direction difference angle.

[0055] Overall, the present invention realizes the automatic identification of key jump nodes in the execution of task chains, effectively captures the nonlinear offset behavior in the collaborative operation of tasks, and has high precision, strong adaptability and real-time analysis capabilities.

[0056] S4: Call the historical task number chain and state evolution sequence, build an abnormal propagation path map through path reasoning, and calculate the path risk score based on the structural correlation between task units and the historical state interaction frequency.

[0057] After identifying task pairs with abnormal state feature vector offsets in the task number chain, in order to further reveal the potential transmission mechanism and impact range of the anomaly, the present invention proposes a method for constructing an anomaly propagation map based on path reasoning, and implements path risk score calculation on this basis. This method can not only locate the source of the anomaly, but also track the associated task flow paths that it may affect, and ultimately form a dynamic propagation map that describes how the anomaly spreads, with strong fault tracing and early warning capabilities.

[0058] Preferably, the construction of the abnormal propagation path map includes: extracting all task pairs marked with abnormalities from the offset anomaly table, adding corresponding nodes to the map, and establishing initial directed edges between the task pairs as potential propagation starting points; according to the structural connection diagram preset by the main board, identifying all path segments that are structurally associated with the abnormal task pairs in the map, and calculating the topological distance and coupling tightness index with the abnormal task pairs; for each highly coupled path segment, tracing back the key dimension evolution trend of the feature vector along the path direction to determine whether there is a delayed state mutation. If so, the path segment is marked as an abnormal evolution path; all path segments that meet the topological coupling and delayed transmission characteristics are merged to form an abnormal propagation path map.

[0059] Among them, the topological distance and coupling tightness index can be expressed as:

[0060] ;

[0061] in, For the task and The shortest path distance in the mainboard structure diagram, is the structural connection weight between two nodes, is the maximum connection weight among all structural connections, used for normalization.

[0062] Specifically, for each highly coupled path segment, the present invention further introduces a timing backtracking mechanism. This mechanism examines the changing trends of key dimensions in the state feature vector of each task unit as the path is traced back from downstream to upstream, determining whether delayed state mutations exist. This trend determination is performed by fitting and analyzing the first-order derivative (rate of change) and second-order derivative (trend of change) of the state change. If a nonlinear transition or a delayed surge in a key dimension is observed in the downstream task unit along the path, it is considered that there is a potential for delayed propagation of abnormal signals. If the path segment meets the dual criteria of high coupling and significant state mutation, it is added to the candidate set of abnormal propagation paths.

[0063] After the abnormal path graph is constructed, the second phase begins: path risk score calculation. This phase aims to assign a risk level to each propagation edge (i.e., task pair) in the graph to assist in subsequent risk task isolation, scheduling optimization, and response strategy selection.

[0064] The optimal calculation of path risk scores includes: extracting all path edges from the anomaly propagation path graph and initializing the scoring item structure for each path edge; calculating the structural dependency of task pairs based on the task unit structure connection information; counting the frequency of task pairs in the cross-influence of feature vectors from historical state evolution records; using a joint scoring function to generate a risk score for each path edge, and sorting and screening high-risk paths.

[0065] Exemplarily, the structural dependency process of task pairs includes: first, confirm and find all adjacent task sets connected to task i and task j; calculate the number of their common adjacent tasks (recorded as the number of shared adjacencies), as well as the number of unions of the two sets; find out whether there is a direct structural connection (such as wiring, dependency, etc.) between task i and task j, and record its connection strength (which can be expressed as a score or number of times), divide the strength value by the maximum value of the connection strength between all task pairs to obtain a standardized connection ratio; multiply the results of the above two parts by the set weights respectively, and take the weighted sum to obtain the structural dependency of task i and task j.

[0066] Furthermore, the generation of the risk score for each path edge includes: finding the number of cross-influences between task i and task j in the state feature vector (i.e., the frequency of responses to each other's state changes) from the historical state evolution data recorded by the system; then, finding the total number of interactions in which task i participated in all historical state evolutions, dividing the two, and obtaining the state cross-frequency ratio of task i and task j; multiplying the structural dependency and the state cross-frequency ratio by the preset weighting coefficients, and taking the weighted sum to obtain the risk score of the task edge.

[0067] Finally, the risk scores for all path edges are sorted in descending order, screening out task pairs with risk values above a set threshold. These pairs are identified as high-risk propagation paths and recommended for priority processing and tracking. This scoring structure not only helps accurately identify problem propagation paths but also provides quantitative decision-making support for the task scheduling system.

[0068] S5: For abnormal boundaries where the path risk score exceeds the preset value, the optimal alternative number under the structural constraint is selected from the task number chain, deployed to the corresponding number position of the current task unit, and the task feature vector table is updated synchronously to complete task continuation and state reconstruction.

[0069] After constructing the anomaly propagation path map and performing risk scoring analysis, this paper further introduces an intelligent repair mechanism for the task number chain to enhance resilience to potential risks and improve task chain stability. This mechanism replaces identified high-risk task pairs and reconstructs their corresponding state feature vectors in real time, achieving anomaly isolation and closed-loop repair of the task sequence. This mechanism, based on task structure constraints and collaborative stability analysis, ensures that replacement tasks possess not only connectivity but also high operational adaptability.

[0070] Preferably, for path segments whose path risk scores exceed the set threshold, the connected task numbers are extracted as abnormally high-risk pairs to form an index sequence of numbers to be replaced; in the system candidate number pool, the set of task numbers that have a structural connection relationship with the current number to be replaced is screened, and the collaborative stability index with the replaced number in historical execution is calculated, and the task numbers are selected and replaced according to the maximum principle; the selected task numbers are written into the task number chain according to the index position, and the task status feature vector table is updated accordingly.

[0071] Furthermore, the calculation of the collaborative stability index with the replaced number in historical execution involves the following process: First, find historical task records in which the two numbers participated in the execution at the same time (for example, they appeared in the same chain of numbers at the same time) and record the total number of these records. For each execution, extract the state feature vectors of the two numbers and calculate the Euclidean distance between the two vectors. Divide the calculated distance by the reference fluctuation range of the state feature vector (this range can be determined by the root mean square value of the historical feature vectors, denoted as ) to obtain the normalized distance. Average all normalized distances and subtract the average from 1 to obtain the collaborative stability score. The closer the value is to 1, the higher the stability.

[0072] Through the above-mentioned replacement and state reconstruction operations, not only the replacement isolation of high-risk task units is achieved in the task number chain structure, but also the transitional reconstruction of the vector level is achieved at the task operation status level, thereby effectively reducing the possibility of the spread of linkage anomalies and maintaining the operation continuity and structural stability of the entire task chain.

[0073] Furthermore, this embodiment also provides an artificial intelligence-based motherboard intelligent fault diagnosis system, including: a task acquisition module 100, used to collect information from multiple task units running concurrently in the motherboard, and build a task number chain based on the calling sequence relationship; a state embedding module 200, used to continuously sample the running state parameters during the task unit running cycle, and perform synchronous embedding in combination with the current task number to construct a task state feature vector table; an anomaly detection module 300, used to calculate the state feature vector offset between adjacent task units pair by pair according to the order of the task number chain, and if the offset exceeds the set tolerance threshold, It is determined that this is the task boundary with a linkage anomaly and is added to the offset anomaly table as a suspicious mark; the path reasoning module 400 is used to call the historical task number chain and state evolution sequence, and through path reasoning, construct an anomaly propagation path map, and calculate the path risk score based on the structural correlation between task units and the historical state interaction frequency; the task repair module 500 is used to select the optimal alternative number under structural constraints from the task number chain for the anomaly boundary whose path risk score exceeds the preset value, deploy it to the corresponding number position of the current task unit, and synchronously update the task feature vector table to complete task continuation and state reconstruction.

[0074] This embodiment also provides a computer device suitable for the case of an artificial intelligence-based motherboard intelligent fault diagnosis method, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the artificial intelligence-based motherboard intelligent fault diagnosis method proposed in the above embodiment.

[0075] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0076] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for intelligent fault diagnosis of a mainboard based on artificial intelligence proposed in the above embodiment is implemented.

[0077] In summary, the present invention introduces an artificial intelligence mechanism to perform intelligent perception, offset detection, and abnormal link reasoning on the operating status of concurrent tasks within the mainboard, effectively realizing the early discovery and precise positioning of potential linkage faults in complex multi-task systems. Through state vector modeling and path graph reasoning technology, the propagation trend of anomalies can be dynamically tracked, thereby significantly improving the accuracy and response speed of fault diagnosis. Compared with traditional rule-based or single-point monitoring methods, the present invention can achieve comprehensive judgment in the multi-dimensional space of structural association and state evolution, and has stronger adaptability and intelligence. Through the replacement task repair mechanism, dynamic replacement of fault boundaries and local reconstruction of system status are realized, enhancing the robustness and continuous operation capability of the mainboard task execution system.

[0078] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for intelligent fault diagnosis of a motherboard based on artificial intelligence, characterized by: include: Collect information on multiple task units running concurrently in the mainboard and build a task number chain based on the calling sequence; During the task unit operation cycle, the operation status parameters are continuously sampled and combined with the current task number, synchronous embedding is performed to construct the task status feature vector table; According to the order of the task number chain, the state feature vector offset between adjacent task units is calculated pair by pair. If the offset exceeds the set tolerance threshold, it is determined that there is a linkage abnormality task boundary and added to the offset abnormality table as a suspicious mark; Call the historical task number chain and state evolution sequence, build an abnormal propagation path map through path reasoning, and calculate the path risk score based on the structural correlation between task units and the historical state interaction frequency; For abnormal boundaries where the path risk score exceeds the preset value, the optimal alternative number under structural constraints is selected from the task number chain, deployed to the corresponding number position of the current task unit, and the task feature vector table is updated synchronously to complete task continuation and state reconstruction; The construction of the task number chain includes: The call relationship logs of active task units in the mainboard are collected, and the call events are sorted by timestamp to construct a task access sequence. Based on the task access sequence, the dependency direction identifiers are marked between any consecutive task pairs, and a context window is set to identify whether the task pairs are in a logically associated neighborhood, and an initial dependency mapping table is constructed. The call frequency of the task pairs in the initial dependency mapping table is normalized, and a frequency threshold is set to eliminate low-frequency sporadic dependency paths and generate a task number chain.

2. The artificial intelligence-based motherboard intelligent fault diagnosis method according to claim 1, wherein: The generation of the task state feature vector table includes: The running status of each task unit is periodically sampled and organized into a state original matrix; Perform sequence position encoding on each parameter in the original state matrix and generate numbered alignment feature blocks based on the current task number; The aligned feature blocks of each task are compressed to form an embedded vector, and the structural position encoding is introduced as an additional feature item so that the state vector retains the task chain sequence information; The embedded vectors of each task in continuous sampling periods are spliced in time segments to form a temporal feature sequence, and added to the corresponding number position in the task state feature vector table.

3. The artificial intelligence-based motherboard intelligent fault diagnosis method according to claim 1, wherein: The pairwise calculation of the state feature vector offsets between adjacent task units includes: Extract consecutive task pairs from a task number chain and , and obtain the corresponding state vector sequence in the task state feature vector table and , select the corresponding time point The state vector under is expanded into sub-dimensional vectors , calculate the offset for all dimensions ; The total offset of the state vector pair is calculated by weighted accumulation.

4. The artificial intelligence-based motherboard intelligent fault diagnosis method according to claim 3, wherein: The construction of the abnormal propagation path map includes: Extract all task pairs marked as abnormal from the offset anomaly table, add corresponding nodes to the graph, and establish initial directed edges between task pairs as potential propagation starting points; According to the structure connection diagram preset by the mainboard, all path segments that are structurally associated with abnormal task pairs are identified in the diagram, and the topological distance and coupling tightness index with the abnormal task pairs are calculated; For each highly coupled path segment, the evolution trend of the key dimensions of the characteristic vector is traced back along the path direction to determine whether there is a delayed state mutation. If so, the path segment is marked as an abnormal evolution path; all path segments that meet the topological coupling and delayed transmission characteristics are merged to form an abnormal propagation path map.

5. The artificial intelligence-based motherboard intelligent fault diagnosis method according to claim 4, characterized in that: The calculation of the path risk score includes: Extract all path edges from the anomaly propagation path graph and initialize the scoring item structure for each path edge; Calculate the structural dependency of task pairs based on the structural connection information of task units; From the historical state evolution records, count the frequency of task pairs in the cross-effect of feature vectors; A joint scoring function is used to generate the risk score of each path edge, and high-risk paths are sorted and filtered.

6. The artificial intelligence-based intelligent fault diagnosis method for a motherboard according to claim 5, wherein: For path segments whose path risk scores exceed the set threshold, the connected task numbers are extracted as abnormally high-risk pairs to form an index sequence of numbers to be replaced; In the system candidate number pool, select a set of task numbers that have a structural connection relationship with the current number to be replaced, calculate the collaborative stability index with the replaced number in historical execution, and select and replace the task number according to the principle of maximum efficiency; Write the selected task number into the task number chain according to the index position, and update the task status feature vector table accordingly.

7. An artificial intelligence-based motherboard intelligent fault diagnosis system, based on the artificial intelligence-based motherboard intelligent fault diagnosis method according to any one of claims 1 to 6, characterized in that: Also includes: The task collection module is used to collect information about multiple task units running concurrently in the mainboard and build a task number chain based on the calling sequence; The state embedding module is used to continuously sample the running state parameters during the task unit running cycle, and perform synchronous embedding in combination with the current task number to construct the task state feature vector table; The anomaly detection module is used to calculate the state feature vector offset between adjacent task units pair by pair based on the order of the task number chain. If the offset exceeds the set tolerance threshold, it is determined that the task boundary has a linkage anomaly and is added to the offset anomaly table as a suspicious point mark; The path reasoning module is used to call the historical task number chain and state evolution sequence, build an abnormal propagation path map through path reasoning, and calculate the path risk score based on the structural correlation between task units and the historical state interaction frequency; The task repair module is used to select the optimal alternative number under structural constraints from the task number chain for abnormal boundaries where the path risk score exceeds the preset value, deploy it to the corresponding number position of the current task unit, and synchronously update the task feature vector table to complete task continuation and state reconstruction.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the artificial intelligence-based mainboard intelligent fault diagnosis method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the artificial intelligence-based mainboard intelligent fault diagnosis method according to any one of claims 1 to 6 are implemented.

Citation Information

Patent Citations

  • Fault detection method and device for secondary equipment of transformer substation

    CN119598300A

  • Device fault diagnosis model establishment method and apparatus

    WO2025065410A1