An analysis method, device, equipment and medium for device failure

Abstract

The application discloses an equipment fault analysis method and device, equipment and medium, and relates to the technical field of fault root cause diagnosis. The method comprises the following steps: acquiring full life cycle data and fault description information of a fault equipment, comprehensively analyzing the equipment fault based on the full life cycle data, and determining a traceable full life cycle evidence chain to avoid data islands. Furthermore, a time sequence knowledge graph is constructed, the fault description information is matched with the time sequence knowledge graph to determine a cause-effect path, accurate tracing is realized based on the time sequence cause-effect relationship, physical verification is performed on the cause-effect path, the final output target fault root cause is realized through physical constraint screening, physical consistency and the full life cycle evidence chain are met, and the accuracy of fault root cause positioning is effectively improved.

Description

Technical Field

[0001] This application relates to the field of fault root cause localization technology, specifically to a method for analyzing equipment faults, a device for analyzing equipment faults, an electronic device, and a readable storage medium. Background Technology

[0002] As the level of electronic and intelligent technology in smart devices continues to improve, the complexity of equipment fault analysis is increasing day by day.

[0003] In related technologies, on the one hand, equipment fault analysis often relies on a single data source or human experience for judgment, such as troubleshooting based on user descriptions, limited on-site operating parameters, or historical maintenance records, making it difficult to accurately pinpoint the cause of the fault. On the other hand, equipment fault analysis is usually based on simple rule matching, making it difficult to extract the causal path of the fault, resulting in a high false positive rate. Summary of the Invention

[0004] This application provides a method, apparatus, device, and medium for analyzing equipment faults, aiming to improve the problem that existing equipment fault analysis methods rely on a single data source and simple rule matching, resulting in a high misjudgment rate in root cause localization.

[0005] This application discloses a method for analyzing equipment failures, the method comprising: Obtain full lifecycle data and fault description information for the faulty device; Determine the full lifecycle evidence chain corresponding to the full lifecycle data; Based on the full lifecycle data and the full lifecycle evidence chain, a temporal knowledge graph for the faulty device is constructed. The temporal knowledge graph includes root cause nodes and temporal causal relationships between the root cause nodes. The fault description information is matched with the temporal knowledge graph, and the initial root cause node and the initial causal path corresponding to the initial root cause node are extracted from the temporal knowledge graph based on the matching result and the temporal causal relationship. The initial causal path is physically verified, and the target failure root cause that satisfies the full life cycle evidence chain is selected from the initial causal path based on the physical verification results.

[0006] In some feasible implementations, the full lifecycle evidence chain includes a deterministic evidence chain and a candidate evidence chain; determining the full lifecycle evidence chain corresponding to the full lifecycle data includes: Obtain the association primary key used to associate the entire lifecycle data; The associated primary key is compared with the full lifecycle data; If the full lifecycle data has the associated primary key, then a deterministic chain of evidence for the faulty device is constructed based on the associated primary key; If at least one of the associated primary keys is missing from the full lifecycle data, then candidate data of the candidate object is obtained, and a candidate evidence chain for the faulty device is constructed based on the candidate data.

[0007] In some feasible implementations, constructing a candidate evidence chain for the faulty device based on the candidate data includes: The full lifecycle data is compared with the candidate data to obtain the model consistency, time correlation, and attribute consistency between the faulty device and the candidate object; Based on the model consistency, the time correlation degree, and the attribute consistency, the candidate association edge score of the candidate object is calculated; If the candidate association edge score is greater than the preset candidate score threshold, then based on the candidate data, a candidate association edge for the faulty device with the missing primary key is constructed. Obtain the defined associated edges of the faulty device for the associated primary key; Based on the determined associated edges and the candidate associated edges, a candidate evidence chain for the faulty device is constructed.

[0008] In some feasible implementations, after constructing the candidate evidence chain for the faulty device based on the determined associated edges and the candidate associated edges, the method further includes: Obtain the score of the determined associated edge; Based on the determined associated edge score and the candidate associated edge score, the second association reliability score of the candidate evidence chain is calculated.

[0009] In some feasible implementations, the physical verification of the initial causal path includes: Obtain a physical model for the initial causal path; The physical model is used to perform physical verification on the initial causal path, and the physical verification result of the initial causal path is obtained.

[0010] In some feasible implementations, the step of using the physical model to perform physical verification on the initial causal path and obtaining the physical verification result of the initial causal path includes: The physical model is used to analyze the initial causal path to obtain the physical residual improvement and the direction of observed change of the initial causal path; If the improvement in the physical residual is greater than the preset residual threshold, and the observed change direction is consistent with the predicted change direction of the physical model, then the physical verification of the initial causal path passes. If the improvement in the physical residual is less than or equal to a preset residual threshold, or if the observed change direction is inconsistent with the predicted change direction of the physical model, then the physical verification of the initial causal path fails.

[0011] In some feasible implementations, the physical verification result further includes the number of directional consistency windows, the number of valid verification windows, and the total number of windows in the initial causal path; the step of filtering the target fault root cause that satisfies the full lifecycle evidence chain from the initial causal path based on the physical verification result includes: When the physical verification of the initial causal path passes, the target residual improvement amount corresponding to the faulty device is obtained, and the physical verification score of the initial causal path is calculated based on the target residual improvement amount, the physical residual improvement amount, the number of directional consistency windows, the number of effective verification windows, and the total number of windows. The association reliability score of the full lifecycle evidence chain is obtained by calculation based on the full lifecycle evidence chain. Based on the physical verification score and the correlation reliability score, the root cause score of the initial causal path is calculated. Based on the root cause score, target failure root causes that satisfy the full life cycle evidence chain are selected from the initial causal path.

[0012] In some feasible implementations, the calculation of the root cause score of the initial causal path based on the physical verification score and the correlation reliability score includes: When the whole lifecycle evidence chain is a deterministic evidence chain, the first association reliability score of the deterministic evidence chain is calculated, and the root cause score of the initial causal path is calculated based on the first association reliability score corresponding to the deterministic evidence chain and the physical verification score. When the full lifecycle evidence chain is a candidate evidence chain, if the second association reliability score corresponding to the candidate evidence chain is greater than or equal to the preset association threshold, the root cause score of the initial causal path is calculated based on the physical verification score and the second association reliability score.

[0013] In some feasible implementations, the step of filtering target failure root causes that satisfy the full lifecycle evidence chain from the initial causal path based on the root cause score includes: Based on the root cause score, a target causal path and the target fault root cause corresponding to the target causal path are selected from all the initial causal paths.

[0014] This application also discloses a device for analyzing equipment failures, the device comprising: The data acquisition module is used to acquire full lifecycle data and fault description information corresponding to the faulty equipment; The evidence chain determination module is used to determine the full lifecycle evidence chain corresponding to the full lifecycle data. The graph construction module is used to construct a temporal knowledge graph for the faulty device based on the full lifecycle data and the full lifecycle evidence chain. The temporal knowledge graph includes root cause nodes and temporal causal relationships between the root cause nodes. The root cause path determination module is used to match the fault description information with the time series knowledge graph, and extract the initial root cause node and the initial causal path corresponding to the initial root cause node from the time series knowledge graph based on the matching result. The root cause localization module is used to physically verify the initial causal path and, based on the physical verification results, select the target fault root cause that satisfies the full lifecycle evidence chain from the initial causal path.

[0015] In some feasible implementations, the full lifecycle evidence chain includes a deterministic evidence chain and a candidate evidence chain; the evidence chain determination module is specifically used for: Obtain the association primary key used to associate the entire lifecycle data; The associated primary key is compared with the full lifecycle data; If the full lifecycle data has the associated primary key, then a deterministic chain of evidence for the faulty device is constructed based on the associated primary key; If at least one of the associated primary keys is missing from the full lifecycle data, then candidate data of the candidate object is obtained, and a candidate evidence chain for the faulty device is constructed based on the candidate data.

[0016] In some feasible implementations, the evidence chain determination module is specifically used for: The full lifecycle data is compared with the candidate data to obtain the model consistency, time correlation, and attribute consistency between the faulty device and the candidate object; Based on the model consistency, the time correlation degree, and the attribute consistency, the candidate association edge score of the candidate object is calculated; If the candidate association edge score is greater than the preset candidate score threshold, then based on the candidate data, a candidate association edge for the faulty device with the missing primary key is constructed. Obtain the defined associated edges of the faulty device for the associated primary key; Based on the determined associated edges and the candidate associated edges, a candidate evidence chain for the faulty device is constructed.

[0017] In some feasible implementations, the apparatus further includes: The reliable correlation calculation module is used to obtain the score of the determined correlation edge; and to calculate the second correlation reliability score of the candidate evidence chain based on the determined correlation edge score and the candidate correlation edge score. In some feasible implementations, the root cause localization module includes: The physical verification submodule is used to obtain a physical model for the initial causal path; and to perform physical verification on the initial causal path using the physical model to obtain the physical verification result of the initial causal path.

[0018] In some feasible implementations, the physical verification submodule is specifically used for: The physical model is used to analyze the initial causal path to obtain the physical residual improvement and the direction of observed change of the initial causal path; If the improvement in the physical residual is greater than the preset residual threshold, and the observed change direction is consistent with the predicted change direction of the physical model, then the physical verification of the initial causal path passes. If the improvement in the physical residual is less than or equal to a preset residual threshold, or if the observed change direction is inconsistent with the predicted change direction of the physical model, then the physical verification of the initial causal path fails.

[0019] In some feasible implementations, the physical verification result further includes the number of directional consistent windows, the number of valid verification windows, and the total number of windows in the initial causal path; the root cause localization module includes: The verification score calculation submodule is used to obtain the target residual improvement amount corresponding to the faulty device when the physical verification of the initial causal path passes, and calculate the physical verification score of the initial causal path based on the target residual improvement amount, the physical residual improvement amount, the number of direction-consistent windows, the number of effective verification windows, and the total number of windows. The reliability score calculation module is used to calculate the correlation reliability score of the full life cycle evidence chain based on the full life cycle evidence chain. The root cause score calculation submodule is used to calculate the root cause score of the initial causal path based on the physical verification score and the correlation reliability score. The root cause localization submodule is used to filter out target fault root causes that satisfy the full life cycle evidence chain from the initial causal path based on the root cause score.

[0020] In some feasible implementations, the root cause score calculation submodule is specifically used for: When the whole lifecycle evidence chain is a deterministic evidence chain, the first association reliability score of the deterministic evidence chain is calculated, and the root cause score of the initial causal path is calculated based on the first association reliability score corresponding to the deterministic evidence chain and the physical verification score. When the full lifecycle evidence chain is a candidate evidence chain, if the second association reliability score corresponding to the candidate evidence chain is greater than or equal to the preset association threshold, the root cause score of the initial causal path is calculated based on the physical verification score and the second association reliability score.

[0021] In some feasible implementations, the root cause localization submodule is specifically used for: Based on the root cause score, a target causal path and the target fault root cause corresponding to the target causal path are selected from all the initial causal paths.

[0022] This application also discloses an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; The memory is used to store computer programs; When the processor executes a program stored in the memory, it implements the method described in any of the embodiments of this application.

[0023] This application also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the methods described in this application.

[0024] The embodiments of this application have the following advantages: In this embodiment, the following steps are taken: acquiring full lifecycle data and fault description information corresponding to the faulty device; determining a full lifecycle evidence chain corresponding to the full lifecycle data; constructing a temporal knowledge graph for the faulty device based on the full lifecycle data and the full lifecycle evidence chain, the temporal knowledge graph including root cause nodes and temporal causal relationships between the root cause nodes; matching the fault description information with the temporal knowledge graph, extracting initial root cause nodes and initial causal paths corresponding to the initial root cause nodes from the temporal knowledge graph based on the matching results and the temporal causal relationships; physically verifying the initial causal paths, and filtering out target fault root causes that satisfy the full lifecycle evidence chain from the initial causal paths based on the physical verification results. In this embodiment, full lifecycle data is acquired to comprehensively analyze equipment failures and determine a traceable full lifecycle evidence chain to avoid data silos. Furthermore, a temporal knowledge graph is constructed, and the failure description information is matched with the temporal knowledge graph to determine the causal path, thereby achieving accurate tracing based on temporal causal relationships. The causal path is also physically verified, and physical constraints are used to filter and finally output the target failure root cause, satisfying physical consistency and the full lifecycle evidence chain, effectively improving the accuracy of failure root cause location. Attached Figure Description

[0025] Figure 1 This is a flowchart of a method for analyzing equipment failures provided in an embodiment of this application; Figure 2 This is a flowchart of a time-series knowledge graph construction method provided in an embodiment of this application; Figure 3 This is a schematic diagram of a temporal causal path provided in an embodiment of this application; Figure 4 This is a schematic diagram of a scenario for fault root cause localization provided in an embodiment of this application; Figure 5 This is a structural diagram of a device for analyzing equipment failures provided in an embodiment of this application; Figure 6 This is a block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0028] As an example, with the continuous improvement of the electronic and intelligent levels of smart devices, the complexity of equipment fault analysis is increasing day by day.

[0029] In related technologies, on the one hand, equipment failure analysis often relies on a single data source or human experience for judgment, such as troubleshooting based on user descriptions, limited on-site operating parameters, or historical maintenance records. Furthermore, data from design, production, operation, and service stages are stored in a scattered manner, lacking effective correlation mechanisms and making it impossible to fully trace the root cause of the failure, thus hindering accurate fault location. On the other hand, equipment failure analysis is usually based on simple rule matching, making it difficult to extract causal paths and lacking physical verification of these paths, resulting in a high false positive rate.

[0030] Based on this, this application provides a method, apparatus, device, and medium for analyzing equipment faults, addressing the problem of high misjudgment rates in existing equipment fault analysis methods that rely on a single data source and simple rule matching. It relates to the field of fault root cause localization technology. By acquiring full lifecycle data for comprehensive equipment fault analysis and determining a traceable full lifecycle evidence chain, it avoids data silos. Furthermore, it constructs a temporal knowledge graph and matches fault description information with the temporal knowledge graph to determine causal paths, achieving accurate tracing based on temporal causal relationships. It also performs physical verification of the causal paths and uses physical constraints for filtering to ultimately output the target fault root cause, satisfying physical consistency and the full lifecycle evidence chain, effectively improving the accuracy of fault root cause localization.

[0031] Reference Figure 1 This application provides a flowchart of a method for analyzing equipment failures, which includes the following steps: Step 101: Obtain the full lifecycle data and fault description information corresponding to the faulty device; In this embodiment, the faulty device can be a home appliance, industrial equipment, or other device exhibiting malfunctions, such as a refrigerator to be diagnosed. For the faulty device requiring diagnosis, its entire lifecycle data and fault description information are acquired. The entire lifecycle data can be understood as the device data spanning its entire lifecycle, from design and production to operation and maintenance. The fault description information is a natural language description of the malfunctioning device, such as: the refrigerator to be diagnosed has poor cooling performance. The fault description information can be obtained from user repair requests or extracted from work orders included in the entire lifecycle data. This embodiment, by acquiring the entire lifecycle data of the faulty device, breaks down data silos between different stages of equipment design, production, operation, and service, enabling comprehensive tracing of the root cause of the fault and laying a data foundation for subsequent operations.

[0032] As an example, this application embodiment divides the entire lifecycle data into design domain, production domain, operation domain, and service domain. The design domain refers to data from the equipment design phase, such as the model of the refrigerator to be diagnosed, BOM (Bill of Materials) version, insulation layer thickness, material calibration parameters, target thermal resistance, etc. The production domain refers to data from the equipment production phase, such as the refrigerator's serial number (SN), production batch, foaming density, process parameters, etc. The operation domain refers to operational data generated during equipment operation, such as the refrigerator's internal temperature, ambient temperature, compressor current, and uptime. The service domain refers to data from the after-sales service phase, such as the refrigerator's work order number, fault symptoms (which can also be understood as fault description information), repair or quality confirmation results, etc.

[0033] Step 102: Determine the full lifecycle evidence chain corresponding to the full lifecycle data; In this embodiment, a full lifecycle evidence chain for the faulty device is determined based on its full lifecycle data. This chain links data scattered across the design, production, operation, and service domains to the faulty device, forming a traceable evidence chain and providing a traceable data foundation for subsequent operations. The full lifecycle evidence chain can be a deterministic chain based on the faulty device's full lifecycle data, ensuring high reliability; alternatively, it can be a candidate chain combining its own full lifecycle data with data from other candidate objects, ensuring data comprehensiveness.

[0034] Step 103: Based on the full lifecycle data and the full lifecycle evidence chain, construct a temporal knowledge graph for the faulty device. The temporal knowledge graph includes root cause nodes and temporal causal relationships between the root cause nodes. In this embodiment, after obtaining the full lifecycle data and evidence chain of the faulty device, a temporal knowledge graph for the faulty device is constructed based on the full lifecycle data and evidence chain. The temporal knowledge graph is used to capture the temporal causal relationships between nodes, which can also be understood as the association in both time and causality dimensions. Specifically, the full lifecycle data is converted into different types of nodes in the temporal knowledge graph. Nodes include root cause nodes, which represent candidate causes that can lead to device failure. Furthermore, edges are established between nodes based on the full lifecycle evidence chain, that is, the relationships between different nodes are determined, forming a temporal knowledge graph with temporal causal relationships. The causal path of the root cause of the failure can then be formed in the temporal knowledge graph based on the root cause nodes and their temporal causal relationships.

[0035] Step 104: Match the fault description information with the time-series knowledge graph, and extract the initial root cause node and the initial causal path corresponding to the initial root cause node from the time-series knowledge graph based on the matching result and the time-series causal relationship. In this embodiment, the fault description information is a natural language description of the fault phenomenon in the device, such as poor cooling effect. The fault description information is then matched with nodes in a temporal knowledge graph, such as through semantic matching, to determine the node corresponding to the fault description information. Furthermore, based on the temporal causal relationships in the temporal knowledge graph, nodes with temporal causal relationships are matched to form an initial causal path containing the initial root cause node. This initial causal path can also be understood as a candidate causal path. Since the same fault phenomenon may correspond to multiple candidate root causes, multiple initial causal paths can be extracted. This embodiment extracts multiple candidate causal paths through a temporal knowledge graph and fault description information for subsequent accurate location of the fault root cause, without relying on human experience or simple rule matching, effectively improving the accuracy of fault root cause location.

[0036] Step 105: Perform physical verification on the initial causal path, and based on the physical verification results, select the target failure root cause that satisfies the full life cycle evidence chain from the initial causal path.

[0037] In this embodiment, each extracted initial causal path undergoes physical verification to determine whether it passes the physical verification, generating a physical verification result for the initial causal path. This physical verification can be performed using a physical model corresponding to the initial causal path. The physical verification result is used to filter initial causal paths. For initial causal paths that pass the physical verification, further filtering is performed using a full lifecycle evidence chain. The final filtered target causal path is then output to determine the target root cause of the fault corresponding to the target causal path. This embodiment verifies candidate root causes of faults through physical constraints, combining temporal causal reasoning from a temporal knowledge graph with physical verification. Furthermore, filtering is performed using the reliability of the full lifecycle evidence chain, ensuring that the final determined target root cause of the fault simultaneously satisfies physical consistency and temporal causal relationship. Moreover, the full lifecycle evidence chain associated with the target root cause is traceable and highly reliable, effectively improving the accuracy of fault root cause localization.

[0038] This application embodiment acquires full lifecycle data to comprehensively analyze equipment failures and determine a traceable full lifecycle evidence chain, avoiding data silos. Furthermore, it constructs a temporal knowledge graph and matches the failure description information with the temporal knowledge graph to determine the causal path, achieving accurate source tracing based on temporal causal relationships. It also performs physical verification on the causal path and filters it through physical constraints to finally output the target failure root cause, satisfying physical consistency and the full lifecycle evidence chain, effectively improving the accuracy of failure root cause location.

[0039] In some feasible implementations, referring to Table 1, embodiments of this application obtain cross-domain full lifecycle data and associate the full lifecycle data based on the key fields of each domain for subsequent fault root cause localization.

[0040]

[0041] Table 1 Specifically, equipment instances are associated with production batches, operation windows, and service work orders via equipment serial numbers (SN); production batches are used to trace the foaming process; and BOM versions are used to obtain insulation layer design parameters and material calibration versions. These association identifiers provide the field foundation for subsequently determining the entire lifecycle evidence chain.

[0042] In some feasible implementations, the full lifecycle evidence chain includes a deterministic evidence chain and a candidate evidence chain; determining the full lifecycle evidence chain corresponding to the full lifecycle data includes: obtaining the association primary key used to associate the full lifecycle data; comparing the association primary key with the full lifecycle data; if the full lifecycle data has the association primary key, then constructing a deterministic evidence chain for the faulty device based on the association primary key; if the full lifecycle data lacks at least one of the association primary keys, then obtaining candidate data for the candidate object, and constructing a candidate evidence chain for the faulty device based on the candidate data.

[0043] In this embodiment, the full lifecycle evidence chain includes a deterministic evidence chain and a candidate evidence chain. In determining the full lifecycle evidence chain, firstly, the associated primary keys used to associate the full lifecycle data are obtained. For example, the associated primary keys include the device serial number (SN), production batch, BOM version, and work order number. The associated primary keys are then compared with the full lifecycle data. If the full lifecycle data of the faulty device has all associated primary keys, it indicates that a deterministic evidence chain for the faulty device can be constructed based on its own full lifecycle data. This directly constructs the deterministic evidence chain based on the associated primary keys, resulting in high reliability and strong traceability. If the full lifecycle data of the faulty device lacks at least one associated primary key, it indicates that the faulty device cannot completely construct an evidence chain based on its own full lifecycle data, as some associated edges are missing. Therefore, candidate data for candidate objects needs to be obtained, and candidate associated edges for the faulty device are constructed based on the candidate data. These candidate associated edges are then combined with the existing deterministic associated edges of the faulty device to form a candidate evidence chain. This application's embodiments establish corresponding deterministic evidence chains or candidate evidence chains for both complete and missing data throughout the entire lifecycle of data. This provides a reliable data foundation for subsequent construction of time-series knowledge graphs, extraction of initial causal paths, and physical verification, avoiding root cause misjudgments due to missing primary keys or incorrect associations.

[0044] As an example, based on full lifecycle data, the deterministic chain of evidence for a faulty device is constructed as follows: LINK={Device Serial Number SN}-{Production Batch}-{BOM Version}-{Work Order Number}; Among them, the equipment serial number (SN) is used to associate equipment instances with production, operation, and service records; the production batch is used to trace the foaming process; the BOM version is used to obtain the insulation layer design parameters and material calibration version; through the SN, production, operation, and service records can be associated with the same equipment instance to form a deterministic chain of evidence.

[0045] For example, when the device serial number SN of the refrigerator to be diagnosed is SN123456, the production batch is 20240315-B05, the BOM version is BC-V2, and the associated work order is TK001, the deterministic evidence chain of the refrigerator to be diagnosed is LINK={SN123456}-{20240315-B05}-{BC-V2}-{TK001}; where the device instance and the production batch, the production batch and the BOM version, the device instance and the running window, and the device instance and the service work order are all traceable and deterministic associated edges.

[0046] In some feasible implementations, constructing a candidate evidence chain for the faulty device based on the candidate data includes: comparing the full lifecycle data with the candidate data to obtain model consistency, time correlation, and attribute consistency between the faulty device and the candidate object; calculating the candidate association edge score of the candidate object based on the model consistency, time correlation, and attribute consistency; if the candidate association edge score is greater than a preset candidate score threshold, constructing candidate association edges for the faulty device with respect to the missing primary key based on the candidate data; obtaining the definitive association edge for the faulty device with respect to the associated primary key; and constructing a candidate evidence chain for the faulty device based on the definitive association edge and the candidate association edge.

[0047] In this embodiment, candidate objects are missing related records used to complete the faulty equipment. For example, candidate objects can be production batch records, service work order records, etc. When the full lifecycle data of the faulty equipment lacks a related primary key, for each candidate object, based on the same fields in the candidate data and the full lifecycle data, model consistency, time correlation, and attribute consistency are calculated. These consistency factors are then weighted and fused to obtain a candidate association edge score. If the candidate association edge score is greater than a preset candidate score threshold, a candidate association edge for the missing primary key is constructed. This candidate association edge is then combined with the confirmed association edge corresponding to the existing related primary key of the faulty equipment to obtain a candidate evidence chain for the faulty equipment. This embodiment, by completing the evidence chain, avoids the inability to perform equipment fault analysis due to missing data.

[0048] As an example, when a faulty device is missing its serial number (SN), production batch, BOM version, or work order number, for each candidate object (c), the model similarity, time correlation, and attribute consistency are calculated using a set of fields (M) that are available to both the candidate object and the faulty device. The model similarity score Sim is calculated using the following formula. model (c): Sim model (c)=Σ[m j×I j (c)] / Σm j ; Where j is the index of the attribute field being compared (such as model, structural configuration, BOM application scope, etc.); m j The weight of field j represents its importance in model similarity. For example, the weights for model, structural configuration, and BOM applicability are 0.50, 0.30, and 0.20, respectively. j (c) is an indicator function that takes the value 1 if it matches the candidate object in the field, and 0 if it does not. If one party is unknown (missing) in the field, then the field is not included in the numerator and denominator.

[0049] Sim for time correlation time (c): The time correlation degree is calculated using three different methods depending on the type of the correlated object. Specifically: (1) Between design records and production batches When associating a design record (such as a BOM version) with a production batch, determine whether the production time of that batch falls within the validity period of that BOM version.

[0050] If the production time of the batch falls between the start and end dates of the BOM validity period, then Sim time (c)=1, otherwise Sim time (c)=0; Because BOM versions typically have effective and expiration dates, only batches produced within that validity period can use that BOM.

[0051] (2) Between the running window and the service order When associating a running window (such as an error window) with a service order, the interval between the end time of the error window and the time of the repair request should be considered.

[0052] The time correlation degree Sim is calculated using the following formula. time (c): Sim time (c)=max(0,1-G / T service ); Where G is the interval between the end time of the exception window and the time of service order reporting; T service This is a preset service response time constant (e.g., 3 days) used to measure the length of the interval.

[0053] (3) Between production batches and service work orders When associating production batches with service orders, consider whether the equipment's usage time from production to failure / repair is within a reasonable lifespan.

[0054] The service time minus the production time falls within the preset lifespan range [L] min ,L max The value is taken as 1 if it is within the range, and 0 if it is outside the range.

[0055] The attribute similarity Sim is calculated using the following formula. attr (c): Sim attr (c)=Σ[a k ×Match k (c)] / Σa k ; Where k is the index of the attribute field being compared (such as material grade, foaming equipment, process formula, capacity range, fault phenomenon, replacement part code, etc.); a k The weight of attribute field k, such as material grade, foaming equipment, process formula, capacity range, and fault phenomenon or replacement part code. k Use values ​​of 0.25, 0.25, 0.25, 0.15, and 0.10 respectively; Match k (c) is the matching indicator function, which takes the value 1 when the target matches the candidate object in the field and 0 when they do not. If one of the targets is unknown (missing) in the field, then the field is not included in the numerator and denominator.

[0056] Furthermore, after obtaining the model similarity Sim model Time correlation Sim time and attribute consistency Sim attr Then, the candidate association edge score for the candidate association edge corresponding to the candidate object is calculated according to the following formula. link : Score link =W1×Sim model + W2×Sim time + W3×Sim attr ; Among them, W1, W2, and W3 are the weights of model similarity, time correlation, and attribute consistency, respectively, and can be set to W1=0.45, W2=0.30, and W3=0.25.

[0057] When a certain score category is not computable, the remaining W values ​​are normalized, and the score is calculated only if there are at least two computable score categories and no fewer than three valid fields. link Otherwise, no candidate edge will be created and "Field to be filled in" will be output.

[0058] Finally, in Score link Greater than the preset candidate score threshold θ link When the value is 0.8, candidate related edges are established.

[0059] In addition, the threshold for distinguishing the highest and second-highest scores can be set to 0.10; if the difference between the highest and second-highest scores does not reach the threshold, multiple candidate related edges are retained and marked as conflict.

[0060] In some feasible implementations, after constructing the candidate evidence chain for the faulty device based on the determined associated edge and the candidate associated edge, the method further includes: obtaining the determined associated edge score of the determined associated edge; and calculating the second association reliability score of the candidate evidence chain based on the determined associated edge score and the candidate associated edge score.

[0061] In this embodiment, after constructing a candidate evidence chain for the faulty device based on the determined associated edges and candidate associated edges, the determined associated edge score is further obtained. For example, the score of the determined associated edge is fixed at 1, while the reliability score of the candidate associated edge is the candidate associated edge score at the time of its establishment. Then, based on the determined associated edge score and the candidate associated edge score, a second association reliability score for the candidate evidence chain is calculated. This second association reliability score is used to determine whether the candidate evidence chain enters physical verification and root cause confirmation, thereby reducing the risk of false associations through the reliability score.

[0062] As an example, determining the reliability score of associated edges (i.e., determining the associated edge score) Rel link =1, Reliability score of candidate associated edges (i.e., candidate associated edge score) Rel link = Score link ; The reliability score (second association reliability score) of the candidate chain of evidence is calculated using the following formula: Rel chain : Rel chain =[Π(e∈E req Rel link (e)^ w e ]^(1 / Σw e )×C ver ×C time ; Among them, C ver The version compatibility coefficient is determined according to the version compatibility mapping table. For example, if the model series, insulation structure configuration, and BOM main version are consistent, they are directly compatible; C time This is a time sequence coefficient, which is set to 1 when the production event precedes the running window, or when the running window precedes the fault or service event; otherwise, it is set to 0.

[0063] If the alternative chain of evidence lacks a necessary edge, the second correlation reliability score Rel chain Below the preset correlation threshold θchain (e.g., 0.85), C ver 0 or C time If the value is 0, the candidate evidence chain will not proceed to subsequent physical verification and root cause confirmation.

[0064] Among some feasible implementation methods, refer to Figure 2 The specific implementation process of constructing a time-series knowledge graph is as follows: The runtime data in the entire lifecycle data is transformed into time-series state nodes. Specifically, the runtime records of the same device serial number SN are segmented according to a preset window length L and step size S. Taking the first valid sampling time t0 as the baseline, the m-th window is: W m =[t0+m×S, t0+m×S+L]; Each window is aggregated into a status node, and each node contains at least the window start and end times, statistical characteristics of the running data, data quality tags, and parameter versions. For example: for the chamber temperature T... box Record the mean, standard deviation, and slope of change of the last segment relative to the first segment; for ambient temperature T amb Record the mean and slope of change; for compressor current I comp Record the average and peak current during operation; for the operating rate D comp The startup rate and number of starts and stops are obtained by dividing the cumulative compressor startup time in the record window by L.

[0065] The status nodes are labeled with data quality and parameter version. For example: windows with an effective sampling rate of not less than 90%, a maximum consecutive missing time of no more than 5 minutes, and no range anomalies are labeled A; windows with an effective sampling rate of 80% to 90% are labeled B and can only be used for auxiliary association; the remaining windows are labeled C and excluded. The parameter version record is V={model, BOM version, material calibration version, control parameter version}. Only nodes with the same version or those confirmed to be compatible through version mapping can participate in the retrieval of the same causal path.

[0066] Accordingly, in addition to status nodes, other types of nodes such as production events, material batches, service orders, test results, and fault events will be constructed based on the full lifecycle data to achieve a time-series knowledge graph covering the entire lifecycle.

[0067] Furthermore, we construct multi-type edges in the temporal knowledge graph. Specifically, we establish temporal sequence edges between nodes based on their chronological order; we establish physical causal edges between nodes based on their physical mechanisms; and we establish definitive and candidate association edges between nodes based on the entire lifecycle evidence chain. For example, a definitive association edge is established between a production event and a status node through the equipment serial number, provided that the production time is earlier than the start time of the status window; a temporal sequence edge is established between a status node and a fault event; and a physical causal edge is established between a foaming density deviation node and a candidate thermal resistance decrease node.

[0068] Finally, the nodes, edges, temporal sequence relationships, and physical causal relationships are organized to form a temporal knowledge graph with temporal causal relationships.

[0069] For example, refer to Figure 3 Based on the time-series knowledge graph, for the fault description information "poor cooling effect", a causal path can be formed as "foaming density deviation - candidate thermal resistance decrease - heat load increase or compressor compensation operation - poor cooling effect". At the same time, there are also causal paths composed of competing causes such as environmental load, door seal cold leakage, refrigerant abnormality and compressor performance degradation, for example: "door seal cold leakage - additional heat leakage load increase - heat load increase - poor cooling effect" and "refrigerant abnormality - insufficient cooling capacity - increased cabinet temperature - poor cooling effect".

[0070] In some feasible implementations, the physical verification of the initial causal path includes: obtaining a physical model for the initial causal path; using the physical model to perform physical verification on the initial causal path, and obtaining the physical verification result of the initial causal path.

[0071] In this embodiment of the application, for the multiple initial causal paths extracted from the time-series knowledge graph, that is, multiple candidate causal paths, physical laws will be used to verify whether the candidate causal paths can really explain the corresponding fault phenomena.

[0072] In the specific implementation process, for each initial causal path, a corresponding physical model is obtained. Examples include: a thermal resistance and heat leakage model for candidate root causes of insulation layer problems; an additional heat leakage load model for door seal cold leakage; a cooling capacity balance model for refrigerant anomalies; and a performance model for compressor performance degradation. The physical model is then used to physically verify the initial causal path, determining whether it passes the physical verification. If the physical verification fails, the initial causal path is excluded and not used for subsequent root cause localization. The physical verification results can include whether the physical verification passes or fails, as well as the amount of improvement in physical residuals, the direction of observed change, the number of direction-consistent windows, the number of effective verification windows, and the total number of windows. This embodiment of the application uses a physical model to physically verify causal paths, filtering out causal paths that do not meet physical consistency requirements, ensuring that the finally determined root cause of the fault corresponds to the fault phenomenon in terms of physical mechanism, effectively improving the accuracy of fault root cause localization.

[0073] In some feasible implementations, the step of using the physical model to perform physical verification on the initial causal path and obtaining the physical verification result of the initial causal path includes: analyzing the initial causal path using the physical model to obtain the physical residual improvement amount and observed change direction of the initial causal path; if the physical residual improvement amount is greater than a preset residual threshold, and the observed change direction is consistent with the predicted change direction of the physical model, then the physical verification of the initial causal path passes; if the physical residual improvement amount is less than or equal to the preset residual threshold, or if the observed change direction is inconsistent with the predicted change direction of the physical model, then the physical verification of the initial causal path fails.

[0074] In this embodiment, a physical model is used to analyze the initial causal path to obtain the physical residual improvement and the observed change direction. The physical residual improvement refers to the difference between the physical residual calculated by the physical model with the candidate root and the actual observation on the operating side, and the baseline residual without the candidate root. The observed change direction refers to the trend of the actual operating data within the window, such as whether the chamber temperature increases or decreases, whether the compressor current increases or decreases, or whether the operating rate increases or decreases. The physical model also provides a predicted change direction.

[0075] Accordingly, if the improvement in physical residuals is greater than a preset residual threshold, and the observed change direction is consistent with the predicted change direction, then the physical verification of the initial causal path is deemed successful; if the improvement in physical residuals is less than or equal to the preset residual threshold, or the observed change direction is inconsistent with the predicted change direction, then the physical verification of the initial causal path is deemed unsuccessful. The preset residual threshold can be determined based on calibration samples of the same model and parameter version that have been confirmed to have no candidate root cause.

[0076] As an example, thermal resistance and heat leakage are calculated using a physical model of candidate root causes of insulation: R ins =d / (k(ρ)×A); Q loss (t)= (T amb (t)- T box (t) ) / (R in +R ins +R out ); Where d is the thickness of the insulation layer; k(ρ) is the effective thermal conductivity of the insulation material, a function of the foaming density ρ, provided by the material calibration table; A is the heat exchange area; R ins It is the thermal resistance of the insulation layer; R in It is the convective thermal resistance between the air inside the box and the inner liner, R in =1 / (h in ×A), h in It is the convective heat transfer coefficient inside the box; R out It is the convective thermal resistance between the outer surface of the enclosure and the ambient air, R out =1 / (h out ×A), h out Q is the convective heat transfer coefficient on the outside of the box; loss (t) is the heat loss that is transferred from the environment into the box through the insulation layer at time t.

[0077] For a candidate root cause v, within the same sliding time window W, the parameter state x corresponding to that candidate root cause is determined. v Based on the baseline state x0, generate the predicted observation Y. v (t), and obtain the actual observation Y on the operating side. obs (t). Wherein, for the parameter state x of the candidate root cause v. v This refers to the set of root cause-related parameters substituted into the corresponding physical model. For example, the candidate for insulation layer is x. v ={d,ρ,k(ρ),A}, candidate for door seal leakage is x v ={Door seal leakage coefficient, open / closed state}, refrigerant anomaly candidate is x v ={Evaporation temperature, Condensation temperature, Charge state}. The baseline state x0 is taken from the median parameter of a calibration sample of the same model, BOM version, and material calibration version, and confirmed by maintenance to have no corresponding fault. Both the baseline model and candidate models use the same thermal equilibrium or refrigeration cycle model. Calculations are performed according to the following formula: e phy (v)=mean{ |Y obs (t)- Y v (t)| / (|Y obs(t)|+ε)},t∈W; Δe(v)=e base - e phy (v); Among them, e phy (v) represents the physical residual, and ε is a preset positive constant used to avoid a denominator of zero; e base Δe(v) represents the baseline residual; Δe(v) represents the improvement in the physical residual.

[0078] The residual threshold (i.e., the preset residual threshold) is determined based on calibration samples of the same model, parameter version, and confirmed to have no candidate root cause: the distribution of Δe is calculated for each calibration window, and its 95th percentile is taken as the residual threshold θ. e Candidate root causes only exist if Δe(v) > θ e Conditions can be improved by using residuals.

[0079] When the thermal resistance of the insulation layer decreases, the expected T box I comp Or D comp At least two valid channels must exhibit an upward trend consistent with the prediction; if there are fewer than two valid channels, all valid channels must be in the same direction. Only when θ is satisfied... e Candidate root factors that meet the threshold and direction consistency conditions are marked as "physical verification passed".

[0080] For example, referring to Table 2, if the candidate root cause is "low foam density", and the thermal balance residual decreases after introducing the candidate root cause, and the predicted direction of the compressor compensation operation is consistent with the observed direction, then the causal path of "foam density deviation - decrease in thermal resistance of insulation layer - increase in heat load - poor cooling effect" is retained; otherwise, even if there is statistical correlation between batches or work orders, it is not taken as a priority root cause of failure.

[0081] Door seal cold leakage is addressed by establishing an additional heat loss load model based on door seal status, door opening and closing records, or seal test results; refrigerant anomalies are addressed by establishing a cooling capacity balance model based on evaporation temperature, condensation temperature, compressor current, and operating time; compressor performance degradation is addressed by establishing a performance model based on operating current, start-stop frequency, and time required to reach the set temperature. The physical residuals of the initial causal paths corresponding to each candidate root cause are calculated within the same window. Paths that fail the corresponding model validation are marked as pending verification or excluded.

[0082]

[0083] Table 2 In some feasible implementations, the physical verification result further includes the number of directional consistency windows, the number of valid verification windows, and the total number of windows in the initial causal path; the step of selecting target fault root causes that satisfy the full lifecycle evidence chain from the initial causal path based on the physical verification result includes: when the physical verification of the initial causal path passes, obtaining the target residual improvement amount corresponding to the faulty device, and calculating the physical verification score of the initial causal path based on the target residual improvement amount, the physical residual improvement amount, the number of directional consistency windows, the number of valid verification windows, and the total number of windows; calculating the association reliability score of the full lifecycle evidence chain based on the full lifecycle evidence chain; calculating the root cause score of the initial causal path based on the physical verification score and the association reliability score; and selecting target fault root causes that satisfy the full lifecycle evidence chain from the initial causal path based on the root cause score.

[0084] In this embodiment, the physical verification result further includes the number of direction-consistent windows, the number of valid verification windows, and the total number of windows. The number of direction-consistent windows is the number of valid verification windows where the predicted direction of change matches the actual observed direction of change. The number of valid verification windows is the number of windows that simultaneously meet the following criteria: data quality is A, parameter version compatibility, complete input, not under exclusion conditions (exclusion conditions include defrosting, prolonged door opening, etc.), and have completed residual and direction calculations. The total number of windows is the total number of available windows for the candidate root cause.

[0085] When the initial causal path passes physical verification, the target residual improvement amount corresponding to the faulty device is obtained. The target residual improvement amount is the target residual improvement amount calibrated for the same model. Based on the target residual improvement amount, physical residual improvement amount, number of directional consistency windows, number of effective verification windows, and total number of windows, the physical verification score of the initial causal path is calculated. Specifically, the physical verification score can be obtained by weighting the residual improvement degree, the proportion of directional consistency windows, and the coverage of effective windows. Optionally, for initial causal paths that pass physical verification, scoring is only allowed when the number of effective verification windows is greater than a preset number (e.g., more than 3 effective verification windows) to avoid randomness caused by too few windows.

[0086] Simultaneously, a correlation reliability score is calculated based on the entire lifecycle evidence chain. If the entire lifecycle evidence chain is a deterministic evidence chain, the correlation reliability score is typically 1; if the entire lifecycle evidence chain is a candidate evidence chain, the second correlation reliability score mentioned above applies. Furthermore, the physical verification score and the correlation reliability score are fused to calculate the root cause score of the initial causal path. Finally, based on the root cause score, target failure root causes that satisfy the entire lifecycle evidence chain are selected from the initial causal path.

[0087] As an example, refer to Figure 4 First, for the candidate root cause v that passes physical verification, the reliability score Rel(v) of the entire lifecycle evidence chain upon which the candidate root cause depends is calculated according to the following formula. Rel(v) can be the minimum value or the weighted geometric mean of the reliability of the necessary associated edges in the causal path: Rel(v)=[Π(e∈E req Rel link (e)^ w e ] ^(1 / Σw e );or, Rel(v) = min(Rel) link (e) | (e∈E req )); Secondly, the physical verification score E is obtained by weighting the residual improvement degree, the proportion of windows with consistent orientation, and the effective window coverage according to the following formula. phy (v): E phy (v) = 0.50 × min[Δe(v) / Δe] ref ,1]+0.30×(N dir / N eff )+0.20×(N eff / N all ); Where, N dir N represents the number of windows with consistent orientation. eff N is the number of valid verification windows. all Δe(v) represents the total number of windows; Δe(v) represents the improvement in physical residuals; Δe ref This represents the target residual improvement amount calibrated for the same model; Δe(v) / Δe ref The residual improvement level measures the ratio of the actual improvement to the target improvement; a value of 1 is used when the ratio exceeds 1, and the weight can be set to 0.5; (N dir / N eff ) represents the proportion of windows with consistent orientation within the effective window; its weight can be set to 0.3. eff / N all Effective window coverage represents the proportion of valid verification windows among all available windows, and the weight can be set to 0.20.

[0088] Furthermore, in the temporal knowledge graph, each causal path consists of several edges, and the path weight α(i,j) of each edge (i,j) reflects the credibility of that edge in the reasoning. The path weight α(i,j) is calculated according to the following formula: α(i,j)=α type ×α time ×αver ; Where, α type Weights for edge types, including verified physically causal edges (α) type 1) Determine the associated edges (α) type (take 0.95) and candidate associated edges (α) type (Take 0.7); α time This is used to measure whether the temporal order of the nodes at both ends of an edge meets the preset time delay requirement. If the temporal order meets the preset time delay, then α... time Take 1, only when it falls within the allowable time delay α time Take 0.80, otherwise α time Take 0; α ver Used to measure parameter version consistency; when parameter versions are consistent, α... ver If the value is 1, then α is not consistent. ver Take 0.

[0089] Finally, the root cause score RCS(v) is calculated using the following formula, based on the physical verification score, the association reliability score, and the path weight: RCS(v) = Normalize{Rel(v) × E phy (v)×Sum[p∈P(v,f)]Product[α(i,j)]}; Normalize means normalization.

[0090] In some feasible implementations, calculating the root cause score of the initial causal path based on the physical verification score and the correlation reliability score includes: when the entire lifecycle evidence chain is a deterministic evidence chain, calculating the first correlation reliability score of the deterministic evidence chain, and calculating the root cause score of the initial causal path based on the first correlation reliability score corresponding to the deterministic evidence chain and the physical verification score; when the entire lifecycle evidence chain is a candidate evidence chain, if the second correlation reliability score corresponding to the candidate evidence chain is greater than or equal to a preset correlation threshold, then calculating the root cause score of the initial causal path based on the physical verification score and the second correlation reliability score.

[0091] In this embodiment, if the second correlation reliability score of the candidate evidence chain is lower than a preset correlation threshold, the candidate evidence chain will not proceed to subsequent physical verification and root cause confirmation. Therefore, in calculating the root cause score, if the entire lifecycle evidence chain is a deterministic evidence chain, the root cause score of the initial causal path is calculated based on the first correlation reliability score and physical verification score corresponding to the deterministic evidence chain; if the entire lifecycle evidence chain is a candidate evidence chain, then a candidate evidence chain with a second correlation reliability score greater than the preset correlation threshold is used, and the root cause score of the initial causal path is calculated based on the physical verification score and the second correlation reliability score. This embodiment, by setting admission conditions for candidate evidence chains, ensures that only candidate evidence chains with sufficiently high reliability are allowed to participate in root cause scoring, thus achieving both the utilization of candidate evidence chains to supplement information and the avoidance of uncertainties introduced by candidate evidence chains, further improving the accuracy of fault root cause location.

[0092] In some feasible implementations, the step of selecting the target failure root cause that satisfies the full life cycle evidence chain from the initial causal paths based on the root cause score includes: selecting the target causal path and the target failure root cause corresponding to the target causal path from all the initial causal paths based on the root cause score.

[0093] In this embodiment, after calculating the root cause score of each initial causal path, all initial causal paths are sorted from highest to lowest root cause score. The target causal path with the highest root cause score or meeting a preset score condition is selected from all initial causal paths, and the initial root cause node corresponding to this target causal path is determined as the target fault root cause. This embodiment, by using the root cause score calculated based on the physical verification score and the reliability of the entire lifecycle evidence chain association, accurately determines the target causal path and the target fault root cause from multiple candidate causal paths. The target fault root cause simultaneously satisfies physical consistency and evidence chain traceability, effectively improving the accuracy of fault root cause location.

[0094] Optionally, when the target causal path is the corresponding candidate evidence chain, the output is "to be verified" to reduce the risk of misjudgment.

[0095] In some feasible implementations, after outputting the target root cause of the failure, the target root cause will be verified and confirmed. This can be done through maintenance testing, batch sampling, or rework disassembly to confirm whether the target root cause is the actual root cause. If it is the actual root cause, the corresponding full lifecycle evidence chain, physical verification results, and causal path will be marked as verified knowledge, and the applicable model, batch range, and calibration parameter version will be recorded. Unconfirmed candidate inferences will only be saved as records pending verification.

[0096] Among them, the alternative evidence chains, paths that fail physical verification, and unconfirmed high-scoring candidate root causes and their paths are not used to update relation weights, material calibration parameters, or subsequent models.

[0097] In addition, when process parameters, insulation materials, or BOM versions change, the corresponding calibration table is re-linked and recalculated; if the versions do not match, "Parameters to be calibrated" is output instead of the verified root cause.

[0098] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of this application.

[0099] This application also provides an equipment fault analysis device 50, please refer to... Figure 5 ,include: Data acquisition module 501 is used to acquire full lifecycle data and fault description information corresponding to the faulty equipment; The evidence chain determination module 502 is used to determine the full lifecycle evidence chain corresponding to the full lifecycle data; The graph construction module 503 is used to construct a temporal knowledge graph for the faulty device based on the full life cycle data and the full life cycle evidence chain. The temporal knowledge graph includes root cause nodes and temporal causal relationships between the root cause nodes. The root cause path determination module 504 is used to match the fault description information with the time series knowledge graph, and extract the initial root cause node and the initial causal path corresponding to the initial root cause node from the time series knowledge graph based on the matching result. The root cause localization module 505 is used to perform physical verification on the initial causal path and, based on the physical verification results, select the target fault root cause that satisfies the full life cycle evidence chain from the initial causal path.

[0100] In some feasible implementations, the full lifecycle evidence chain includes a deterministic evidence chain and a candidate evidence chain; the evidence chain determination module 502 is specifically used for: Obtain the association primary key used to associate the entire lifecycle data; The associated primary key is compared with the full lifecycle data; If the full lifecycle data has the associated primary key, then a deterministic chain of evidence for the faulty device is constructed based on the associated primary key; If at least one of the associated primary keys is missing from the full lifecycle data, then candidate data of the candidate object is obtained, and a candidate evidence chain for the faulty device is constructed based on the candidate data.

[0101] In some feasible implementations, the evidence chain determination module 502 is specifically used for: The full lifecycle data is compared with the candidate data to obtain the model consistency, time correlation, and attribute consistency between the faulty device and the candidate object; Based on the model consistency, the time correlation degree, and the attribute consistency, the candidate association edge score of the candidate object is calculated; If the candidate association edge score is greater than the preset candidate score threshold, then based on the candidate data, a candidate association edge for the faulty device with the missing primary key is constructed. Obtain the defined associated edges of the faulty device for the associated primary key; Based on the determined associated edges and the candidate associated edges, a candidate evidence chain for the faulty device is constructed.

[0102] In some feasible implementations, the apparatus further includes: The reliable correlation calculation module is used to obtain the score of the determined correlation edge; and to calculate the second correlation reliability score of the candidate evidence chain based on the determined correlation edge score and the candidate correlation edge score. In some feasible implementations, the root cause localization module 505 includes: The physical verification submodule is used to obtain a physical model for the initial causal path; and to perform physical verification on the initial causal path using the physical model to obtain the physical verification result of the initial causal path.

[0103] In some feasible implementations, the physical verification submodule is specifically used for: The physical model is used to analyze the initial causal path to obtain the physical residual improvement and the direction of observed change of the initial causal path; If the improvement in the physical residual is greater than the preset residual threshold, and the observed change direction is consistent with the predicted change direction of the physical model, then the physical verification of the initial causal path passes. If the improvement in the physical residual is less than or equal to a preset residual threshold, or if the observed change direction is inconsistent with the predicted change direction of the physical model, then the physical verification of the initial causal path fails.

[0104] In some feasible implementations, the physical verification result also includes the number of directional consistent windows, the number of valid verification windows, and the total number of windows in the initial causal path; the root cause localization module 505 includes: The verification score calculation submodule is used to obtain the target residual improvement amount corresponding to the faulty device when the physical verification of the initial causal path passes, and calculate the physical verification score of the initial causal path based on the target residual improvement amount, the physical residual improvement amount, the number of direction-consistent windows, the number of effective verification windows, and the total number of windows. The reliability score calculation module is used to calculate the correlation reliability score of the full life cycle evidence chain based on the full life cycle evidence chain. The root cause score calculation submodule is used to calculate the root cause score of the initial causal path based on the physical verification score and the correlation reliability score. The root cause localization submodule is used to filter out target fault root causes that satisfy the full life cycle evidence chain from the initial causal path based on the root cause score.

[0105] In some feasible implementations, the root cause score calculation submodule is specifically used for: When the whole lifecycle evidence chain is a deterministic evidence chain, the first association reliability score of the deterministic evidence chain is calculated, and the root cause score of the initial causal path is calculated based on the first association reliability score corresponding to the deterministic evidence chain and the physical verification score. When the full lifecycle evidence chain is a candidate evidence chain, if the second association reliability score corresponding to the candidate evidence chain is greater than or equal to the preset association threshold, the root cause score of the initial causal path is calculated based on the physical verification score and the second association reliability score.

[0106] In some feasible implementations, the root cause localization submodule is specifically used for: Based on the root cause score, a target causal path and the target fault root cause corresponding to the target causal path are selected from all the initial causal paths.

[0107] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0108] This application also provides an electronic device, including: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the various processes of the above method embodiments and achieves the same technical effects. To avoid repetition, it will not be described again here.

[0109] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods described in any embodiment of this application.

[0110] Figure 6 A schematic diagram of the hardware structure of an electronic device for implementing the various embodiments of this application.

[0111] The electronic device 600 includes, but is not limited to, components such as: a radio frequency unit 601, a network module 602, an audio output unit 603, an input unit 604, a sensor 605, a display unit 606, a user input unit 607, an interface unit 608, a memory 609, a processor 610, and a power supply 611. Those skilled in the art will understand that the electronic device structure described in the embodiments of this application does not constitute a limitation on the electronic device. An electronic device may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. In the embodiments of this application, the electronic device includes, but is not limited to, mobile phones, tablet computers, laptop computers, PDAs, in-vehicle terminals, wearable devices, and pedometers.

[0112] It should be understood that, in this embodiment, the radio frequency unit 601 can be used for receiving and transmitting signals during information transmission or calls. Specifically, it receives downlink data from the base station and processes it with the processor 610; additionally, it transmits uplink data to the base station. Typically, the radio frequency unit 601 includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low-noise amplifier, and a duplexer. Furthermore, the radio frequency unit 601 can also communicate with networks and other devices via a wireless communication system.

[0113] The electronic device provides users with wireless broadband internet access through the network module 602, such as helping users send and receive emails, browse web pages, and access streaming media.

[0114] The audio output unit 603 can convert audio data received by the radio frequency unit 601 or the network module 602 or stored in the memory 609 into audio signals and output them as sound. Furthermore, the audio output unit 603 can also provide audio output related to specific functions performed by the electronic device 600 (e.g., call signal reception sound, message reception sound, etc.). The audio output unit 603 includes a speaker, a buzzer, and a receiver, etc.

[0115] Input unit 604 is used to receive audio or video signals. Input unit 604 may include a graphics processing unit (GPU) 6041 and a microphone 6042. The GPU 6041 processes image data of still images or videos acquired by an image capture device (such as a camera) in video capture mode or image capture mode. The processed image frames can be displayed on display unit 606. The image frames processed by GPU 6041 can be stored in memory 609 (or other storage media) or transmitted via radio frequency unit 601 or network module 602. Microphone 6042 can receive sound and process such sound into audio data. The processed audio data can be converted into a format that can be transmitted to a mobile communication base station via radio frequency unit 601 in telephone call mode.

[0116] The electronic device 600 also includes at least one sensor 605, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor includes an ambient light sensor and a proximity sensor. The ambient light sensor can adjust the brightness of the display panel 6061 according to the ambient light level, and the proximity sensor can turn off the display panel 6061 and / or backlight when the electronic device 600 is moved to the ear. As a type of motion sensor, an accelerometer sensor can detect the magnitude of acceleration in various directions (generally three axes). When stationary, it can detect the magnitude and direction of gravity and can be used to identify the posture of the electronic device (such as landscape / portrait switching, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer, tapping), etc. The sensor 605 may also include a fingerprint sensor, pressure sensor, iris sensor, molecular sensor, gyroscope, barometer, hygrometer, thermometer, infrared sensor, etc., which will not be described in detail here.

[0117] The display unit 606 is used to display information input by the user or information provided to the user. The display unit 606 may include a display panel 6061, which may be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED), or the like.

[0118] User input unit 607 can be used to receive input numerical or character information, and to generate key signal inputs related to user settings and function control of electronic devices. Specifically, user input unit 607 includes a touch panel 6071 and other input devices 6072. Touch panel 6071, also known as a touch screen, can collect touch operations performed by the user on or near it (such as operations performed by the user using a finger, stylus, or any suitable object or accessory on or near touch panel 6071). Touch panel 6071 may include two parts: a touch detection device and a touch controller. The touch detection device detects the user's touch position and the signal generated by the touch operation, and transmits the signal to the touch controller; the touch controller receives touch information from the touch detection device, converts it into touch point coordinates, and sends it to processor 610, which receives and executes commands from processor 610. In addition, touch panel 6071 can be implemented using various types such as resistive, capacitive, infrared, and surface acoustic wave. Besides touch panel 6071, user input unit 607 may also include other input devices 6072. Specifically, other input devices 6072 may include, but are not limited to, physical keyboards, function keys (such as volume control buttons, power buttons, etc.), trackballs, mice, joysticks, etc., which will not be described in detail here.

[0119] Furthermore, the touch panel 6071 can cover the display panel 6061. When the touch panel 6071 detects a touch operation on or near it, it transmits the information to the processor 610 to determine the type of touch event. Subsequently, the processor 610 provides corresponding visual output on the display panel 6061 according to the type of touch event. It is understood that in one embodiment, the touch panel 6071 and the display panel 6061 are implemented as two independent components to realize the input and output functions of the electronic device. However, in some embodiments, the touch panel 6071 and the display panel 6061 can be integrated to realize the input and output functions of the electronic device. The specific implementation is not limited here.

[0120] Interface unit 608 serves as an interface for connecting external devices to electronic device 600. For example, external devices may include a wired or wireless headphone port, an external power supply (or battery charger) port, a wired or wireless data port, a memory card port, a port for connecting a device with an identification module, an audio input / output (I / O) port, a video I / O port, a headphone port, and so on. Interface unit 608 can be used to receive input from external devices (e.g., data, power, etc.) and transmit the received input to one or more components within electronic device 600, or it can be used to transmit data between electronic device 600 and external devices.

[0121] The memory 609 can be used to store software programs and various data. The memory 609 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function (such as sound playback, image playback, etc.), etc.; the data storage area may store data created based on the use of the mobile phone (such as audio data, phonebook, etc.). Furthermore, the memory 609 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0122] The processor 610 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines. By running or executing software programs and / or modules stored in the memory 609, and by calling data stored in the memory 609, it performs various functions and processes data, thereby providing overall monitoring of the electronic device. The processor 610 may include one or more processing units; preferably, the processor 610 may integrate an application processor and a modem processor. The application processor mainly handles the operating system, user interface, and applications, while the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 610.

[0123] The electronic device 600 may also include a power supply 611 (such as a battery) that supplies power to various components. Preferably, the power supply 611 can be logically connected to the processor 610 through a power management system, thereby enabling functions such as managing charging, discharging, and power consumption through the power management system.

[0124] In addition, the electronic device 600 includes some functional modules not shown, which will not be described in detail here.

[0125] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0126] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0127] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

[0128] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0129] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0130] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0131] 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.

[0132] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0133] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they 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 a portion 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, ROM, RAM, magnetic disks, or optical disks.

[0134] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for analyzing equipment failures, characterized in that, The method includes: Obtain full lifecycle data and fault description information for the faulty device; Determine the full lifecycle evidence chain corresponding to the full lifecycle data; Based on the full lifecycle data and the full lifecycle evidence chain, a temporal knowledge graph for the faulty device is constructed. The temporal knowledge graph includes root cause nodes and temporal causal relationships between the root cause nodes. The fault description information is matched with the temporal knowledge graph, and the initial root cause node and the initial causal path corresponding to the initial root cause node are extracted from the temporal knowledge graph based on the matching result and the temporal causal relationship. The initial causal path is physically verified, and the target failure root cause that satisfies the full life cycle evidence chain is selected from the initial causal path based on the physical verification results.

2. The method according to claim 1, characterized in that, The full lifecycle evidence chain includes a deterministic evidence chain and a candidate evidence chain; determining the full lifecycle evidence chain corresponding to the full lifecycle data includes: Obtain the association primary key used to associate the entire lifecycle data; Compare the associated primary key with the full lifecycle data; If the full lifecycle data has the associated primary key, then a deterministic chain of evidence for the faulty device is constructed based on the associated primary key; If at least one of the associated primary keys is missing from the full lifecycle data, then candidate data of the candidate object is obtained, and a candidate evidence chain for the faulty device is constructed based on the candidate data.

3. The method according to claim 2, characterized in that, The step of constructing a candidate evidence chain for the faulty device based on the candidate data includes: The full lifecycle data is compared with the candidate data to obtain the model consistency, time correlation, and attribute consistency between the faulty device and the candidate object; Based on the model consistency, the time correlation degree, and the attribute consistency, the candidate association edge score of the candidate object is calculated; If the candidate association edge score is greater than the preset candidate score threshold, then based on the candidate data, a candidate association edge for the faulty device with the missing primary key is constructed. Obtain the defined associated edges of the faulty device for the associated primary key; Based on the determined associated edges and the candidate associated edges, a candidate evidence chain for the faulty device is constructed.

4. The method according to claim 3, characterized in that, After constructing the candidate evidence chain for the faulty device based on the determined associated edges and the candidate associated edges, the method further includes: Obtain the score of the determined associated edge; Based on the determined associated edge score and the candidate associated edge score, the second association reliability score of the candidate evidence chain is calculated.

5. The method according to claim 1, characterized in that, The physical verification of the initial causal path includes: Obtain a physical model for the initial causal path; The physical model is used to perform physical verification on the initial causal path, and the physical verification result of the initial causal path is obtained.

6. The method according to claim 5, characterized in that, The step of using the physical model to perform physical verification on the initial causal path and obtaining the physical verification result of the initial causal path includes: The physical model is used to analyze the initial causal path to obtain the physical residual improvement and the direction of observed change of the initial causal path; If the improvement in the physical residual is greater than the preset residual threshold, and the observed change direction is consistent with the predicted change direction of the physical model, then the physical verification of the initial causal path passes. If the improvement in the physical residual is less than or equal to a preset residual threshold, or if the observed change direction is inconsistent with the predicted change direction of the physical model, then the physical verification of the initial causal path fails.

7. The method according to claim 6, characterized in that, The physical verification result also includes the number of directional consistency windows, the number of valid verification windows, and the total number of windows in the initial causal path; the step of selecting the target failure root cause that satisfies the full lifecycle evidence chain from the initial causal path based on the physical verification result includes: When the physical verification of the initial causal path passes, the target residual improvement amount corresponding to the faulty device is obtained, and the physical verification score of the initial causal path is calculated based on the target residual improvement amount, the physical residual improvement amount, the number of directional consistency windows, the number of effective verification windows, and the total number of windows. The association reliability score of the full lifecycle evidence chain is obtained by calculation based on the full lifecycle evidence chain. Based on the physical verification score and the correlation reliability score, the root cause score of the initial causal path is calculated. Based on the root cause score, target failure root causes that satisfy the full life cycle evidence chain are selected from the initial causal path.

8. The method according to claim 7, characterized in that, The root cause score of the initial causal path is calculated based on the physical verification score and the correlation reliability score, including: When the whole lifecycle evidence chain is a deterministic evidence chain, the first association reliability score of the deterministic evidence chain is calculated, and the root cause score of the initial causal path is calculated based on the first association reliability score corresponding to the deterministic evidence chain and the physical verification score. When the full lifecycle evidence chain is a candidate evidence chain, if the second association reliability score corresponding to the candidate evidence chain is greater than or equal to the preset association threshold, the root cause score of the initial causal path is calculated based on the physical verification score and the second association reliability score.

9. The method according to claim 7, characterized in that, The step of filtering target failure root causes that satisfy the full lifecycle evidence chain from the initial causal path based on the root cause score includes: Based on the root cause score, a target causal path and the target fault root cause corresponding to the target causal path are selected from all the initial causal paths.

10. A device for analyzing equipment failures, characterized in that, The device includes: The data acquisition module is used to acquire full lifecycle data and fault description information corresponding to the faulty equipment; The evidence chain determination module is used to determine the full lifecycle evidence chain corresponding to the full lifecycle data. The graph construction module is used to construct a temporal knowledge graph for the faulty device based on the full lifecycle data and the full lifecycle evidence chain. The temporal knowledge graph includes root cause nodes and temporal causal relationships between the root cause nodes. The root cause path determination module is used to match the fault description information with the time series knowledge graph, and extract the initial root cause node and the initial causal path corresponding to the initial root cause node from the time series knowledge graph based on the matching result. The root cause localization module is used to physically verify the initial causal path and, based on the physical verification results, select the target fault root cause that satisfies the full lifecycle evidence chain from the initial causal path.

11. An electronic device, characterized in that, Including processor and memory, among which, Memory, used to store computer programs; A processor for executing a program stored in memory to implement the method described in any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in any one of claims 1-9.

Images