New energy vehicle supply chain problem traceability diagnosis method based on knowledge graph

Abstract

The application discloses a new energy automobile supply chain problem tracing diagnosis method based on a knowledge graph, relates to the technical field of new energy automobile supply chain data management, and forms a new energy automobile supply chain problem tracing knowledge graph through supply chain tracing data; the new energy automobile supply chain problem tracing knowledge graph is constructed by taking a part version identifier as an independent entity; a version evolution edge is additionally arranged to represent the version iteration relationship of the same coded part, the design realizes modeling and fault correlation analysis of each part version under the same part code, and can effectively determine whether the fault occurs in a certain version; version fault correlation analysis is carried out by generating a same-family comparable version node set, the abnormal correlation part version node is determined by combining the same-family correlation rate benchmark value and the box plot anomaly detection, and finally, a problem tracing path is generated. The target part version and the corresponding supplier causing a specific fault can be more accurately identified, and the tracing range can be effectively reduced.

Description

Technical Field

[0001] This invention relates to the field of new energy vehicle supply chain data management technology, and in particular to a knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain. Background Technology

[0002] With the development of the new energy vehicle industry, the types of new energy vehicle components, the number of suppliers, and the scale of vehicle production are constantly increasing. To achieve quality traceability during the vehicle production process and after-sales fault processes, existing technologies typically record the assembly relationship between vehicles and components through manufacturing systems, supply chain management systems, or quality management systems, and record data such as supplier information and vehicle fault information for the corresponding components. When a vehicle malfunctions, the components assembled in the vehicle can be queried based on the vehicle identification, and further traceable based on the component codes to obtain the corresponding component and supplier information. Some existing technologies also use knowledge graphs and other methods to establish relationships between vehicles, components, suppliers, and vehicle faults, thereby assisting in the traceability analysis of new energy vehicle production quality problems or vehicle fault problems.

[0003] However, during the production and lifecycle of new energy vehicles, the same part code may correspond to multiple different part versions due to design optimization, process adjustments, material changes, or changes in supply relationships. These different part versions may be used in different batches or different vehicles. Existing traceability methods typically establish the association between vehicles and parts based primarily on part codes. Queries can only determine "which vehicles used this part code," but cannot further determine whether the faults are concentrated in a particular version. Because it is difficult to accurately characterize the version evolution relationship between different part versions under the same part code, and also difficult to distinguish the specific part versions actually assembled in the vehicle and the differences in the association between different versions and vehicle faults, when a vehicle experiences a specific type of fault, it is easy to analyze multiple historical versions corresponding to the same part code as the same object. This makes it difficult to determine whether the fault has an abnormal association with a specific part version, which may lead to an excessively broad traceability scope for the problematic part version and its corresponding supplier, affecting the efficiency and accuracy of locating quality problems in the new energy vehicle supply chain. Summary of the Invention

[0004] To address the technical problems existing in the background art, this invention proposes a knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain.

[0005] The knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain proposed in this invention includes the following steps: S1. Obtain supply chain traceability data in the new energy vehicle supply chain; S2. Generate a knowledge graph for tracing issues in the new energy vehicle supply chain based on supply chain traceability data; S3. Extract a vehicle fault data to be traced from multiple vehicle fault data sets, and use it as the vehicle fault data to be traced; generate a set of candidate component version nodes corresponding to the vehicle fault data to be traced based on the knowledge graph of new energy vehicle supply chain problem traceability. S4. For each candidate component version node in the candidate component version node set, generate a set of comparable version nodes of the same family corresponding to each candidate component version node. S5. Obtain the set of comparable version nodes in the same family corresponding to all candidate component version nodes; for any candidate component version node corresponding to the set of comparable version nodes in the same family, perform version fault correlation analysis on each comparable component version node in the set of comparable version nodes in the same family to obtain the version fault correlation parameters corresponding to each comparable component version node. S6. Generate abnormal associated component version nodes corresponding to the fault data of the vehicle to be traced based on the version fault association parameters corresponding to each comparable component version node. S7. Obtain the version nodes of abnormal related components corresponding to all vehicle fault data to be traced, and generate the problem tracing path corresponding to the vehicle fault data to be traced based on the knowledge graph of new energy vehicle supply chain problem tracing.

[0006] Preferably, in S1, the supply chain traceability data includes multiple component version data, multiple vehicle assembly data, and multiple vehicle fault data; in, Each component version data includes the component code, several component version identifiers corresponding to the component code, the effective time of the version corresponding to the component version identifier, and the supplier information corresponding to the component version identifier; Each vehicle assembly data includes the vehicle identification number, the part code installed on the vehicle corresponding to the vehicle identification number, the version identification number of the part installed on the vehicle corresponding to the vehicle identification number, the assembly time, and the service period of the version. Version service period is used to characterize the actual assembly and usage time of a component version on a vehicle; For any vehicle assembly data version service period, the start time of the version service period is the assembly time of that vehicle assembly data. Vehicle assembly data with the same vehicle identifier and the same part code as the vehicle assembly data are obtained from multiple vehicle assembly data, and arranged in chronological order of assembly time. If there is a next vehicle assembly data after the vehicle assembly data, the assembly time in the next vehicle assembly data is taken as the end time of the service period of the version corresponding to the vehicle assembly data; otherwise, the service period of the version of the vehicle assembly data is the continuous period starting from the assembly time of the vehicle assembly data. As an explanation, the part version identifier on the vehicle corresponding to the vehicle identification corresponds to the part code on the vehicle corresponding to the vehicle identification. In other words, the part version identifier is one of several part version identifiers corresponding to the part code.

[0007] Each vehicle fault data includes vehicle identification, fault occurrence time, and fault type information.

[0008] Preferably, in step S2, a knowledge graph for tracing issues in the new energy vehicle supply chain is generated based on supply chain traceability data, including the following steps: S21. Using the distinct component codes appearing in the component version data as component entities, generate component nodes corresponding to each component code. Each distinct component version identifier appearing in the component version data is used as a component version entity. A component version node corresponding to each component version identifier is generated, and the effective time of the version corresponding to the component version identifier is used as the node attribute of the component version node. Using the distinct supplier information appearing in the component version data as supplier entities, a supplier node corresponding to each supplier information is generated. Using the distinct vehicle identifiers appearing in the vehicle assembly data as vehicle entities, vehicle nodes corresponding to each vehicle identifier are generated. Each vehicle fault data is used as a fault event entity, and a fault event node corresponding to each vehicle fault data is generated. The fault occurrence time and fault type information in each vehicle fault data are used as the node attributes of the fault event node. S22. Generate a version attribution edge between component nodes and component version nodes corresponding to the same component code; wherein, for each component version node, obtain the component code corresponding to the component version node, and generate a version attribution edge between the component version node and component nodes with the same component code. For multiple component version nodes corresponding to the same component code, sort the multiple component version nodes according to the version effective time of each component version node, and generate a version evolution edge between two adjacent component version nodes. The version evolution edge points from the component version node with the earlier version effective time to the component version node with the later version effective time. For each component version identifier, a supply edge is generated between the component version node and the corresponding supplier node based on the supplier information corresponding to the component version identifier. For each vehicle assembly data, obtain the corresponding vehicle node based on the vehicle identifier in the vehicle assembly data, and obtain the corresponding component version node based on the component version identifier in the vehicle assembly data. Generate an assembly edge between the vehicle node and the component version node, and use the assembly time and version service period in the vehicle assembly data as the edge attribute of the assembly edge. For each vehicle fault data, obtain the corresponding vehicle node based on the vehicle identifier in the vehicle fault data, and generate a fault edge between the vehicle node and the fault event node based on the fault event node corresponding to the vehicle fault data. S23. Combine the component nodes, component version nodes, supplier nodes, vehicle nodes, and fault event nodes generated in step S21, as well as the version attribution edge, version evolution edge, supply edge, assembly edge, and fault edge generated in step S22, to form a knowledge graph for tracing issues in the new energy vehicle supply chain.

[0009] Preferably, in S3, based on the knowledge graph for tracing issues in the new energy vehicle supply chain, a set of candidate component version nodes corresponding to the fault data of the vehicle to be traced is generated, including the following steps: S31. Obtain the fault event nodes corresponding to the fault data of the vehicle to be traced from the knowledge graph of the new energy vehicle supply chain problem tracing, and use them as target fault event nodes. S32. Take the fault occurrence time in the fault data of the vehicle to be traced as the target fault occurrence time, and take the fault type information in the fault data of the vehicle to be traced as the target fault type information. S33. Take the vehicle node that has a fault edge with the target fault event node as the target vehicle node; S34. Obtain the version nodes of each component that have an assembly edge with the target vehicle node, and obtain the version effective time of each component version node that has an assembly edge with the target vehicle node. For each component version node that has an assembly edge with the target vehicle node, obtain the assembly time of the corresponding assembly edge. S35. Select the component version nodes whose version effective time is no later than the target failure time, whose assembly time of the corresponding assembly edge is no later than the target failure time, and whose target failure time falls within the service period of the version of the corresponding assembly edge as candidate component version nodes, and form a set of candidate component version nodes from each candidate component version node.

[0010] Preferably, in step S4, for each candidate component version node in the candidate component version node set, a set of comparable version nodes of the same family corresponding to each candidate component version node is generated, including the following steps: S41. For any candidate component version node in the candidate component version node set, obtain the component node that has a version ownership edge with the candidate component version node, and use it as a candidate component node. S42. In the knowledge graph for tracing issues in the new energy vehicle supply chain, starting from the candidate component node, obtain all component version nodes that have version attribution edges with the candidate component node; S43. Based on the version evolution edges between each component version node, arrange each component version node in sequence to form the version evolution sequence corresponding to the candidate component node. S44. The component version node in the version evolution sequence whose version effective time is no later than the target failure time is taken as the comparable component version node corresponding to the candidate component version node. S45. The candidate component version node and its corresponding comparable component version node form a set of comparable version nodes of the same family.

[0011] Preferably, in S5, for any candidate component version node corresponding to a set of comparable version nodes of the same family, version fault correlation analysis is performed on each comparable component version node in the set of comparable version nodes of the same family to obtain the version fault correlation parameters corresponding to each comparable component version node, including the following steps: S51. For any candidate component version node, the set of comparable version nodes in the same family: S52. For any comparable component version node in the set of comparable version nodes of the same family, obtain the vehicle nodes that have an assembly edge with the comparable component version node, and form the set of associated vehicle nodes corresponding to the comparable component version node. S53. Count the number of vehicle nodes in the set of associated vehicle nodes to obtain the number of associated vehicle nodes corresponding to the comparable component version node. S54. For each vehicle node in the set of associated vehicle nodes, obtain the fault event nodes that have a fault edge with the vehicle node, and filter out the fault event nodes of the same type from the fault event nodes. Among them, the same type of fault event node is a fault event node whose fault type information is the same as the target fault type information, whose fault occurrence time is no later than the target fault occurrence time, and whose fault occurrence time falls within the service period of the version of the assembly edge between the comparable component version node and the corresponding vehicle node. S55. Take the vehicle node with at least one similar fault event node as the similar fault vehicle node, and count the number of similar fault vehicle nodes to obtain the number of similar fault vehicle nodes corresponding to the comparable component version node. S56. When the number of associated vehicle nodes is greater than zero, the ratio of the number of similar fault vehicle nodes corresponding to the comparable component version node to the number of associated vehicle nodes shall be used as the version fault association rate corresponding to the comparable component version node. S57. The number of associated vehicle nodes, the number of similar faulty vehicle nodes, and the version fault association rate are used as the version fault association parameters corresponding to the version node of the comparable component.

[0012] Preferably, in step S6, generating abnormal associated component version nodes corresponding to the vehicle fault data to be traced, based on the version fault association parameters corresponding to each comparable component version node, includes the following steps: S61. For each candidate component version node, obtain the version fault association parameters corresponding to each comparable component version node from the set of comparable version nodes in the same family corresponding to the candidate component version node. S62. Based on the version fault association parameters corresponding to each candidate component version node, determine the abnormal associated component version node, specifically including: S63. Determine the family association rate benchmark value corresponding to the family of comparable version nodes based on the version fault association rate in the version fault association parameters corresponding to each comparable component version node. S64. Obtain the comparable component version node that is the same as the candidate component version node from the set of comparable version nodes of the same family, and use the version fault association rate in the version fault association parameter corresponding to the comparable component version node as the candidate version fault association rate corresponding to the candidate component version node. S65. Obtain the difference between the candidate version fault association rate and the family association rate benchmark value, and use it as the version difference association value corresponding to the candidate component version node. As an explanation, the difference between the candidate version failure correlation rate and the baseline value of the family correlation rate is: candidate version failure correlation rate − family correlation rate baseline value = version difference correlation value. S66. Obtain the version difference correlation value corresponding to each candidate component version node, and use the box plot method to perform anomaly detection on the version difference correlation value corresponding to each candidate component version node. The candidate component version node that meets the anomaly judgment condition of the box plot method is determined as the abnormally correlated component version node.

[0013] Preferably, in step S7, based on the knowledge graph of new energy vehicle supply chain problem tracing, a problem tracing path corresponding to the fault data of the vehicle to be traced is generated, including the following steps: S71. Based on all abnormal associated component version nodes, generate the problem tracing path corresponding to the vehicle fault data to be traced in the new energy vehicle supply chain problem tracing knowledge graph. S72. The problem tracing path includes the path from the target fault event node to the target vehicle node via the fault edge, from the target vehicle node to the abnormal associated component version node via the assembly edge, from the abnormal associated component version node to the corresponding component node via the version ownership edge, and from the abnormal associated component version node to the corresponding supplier node via the supply edge.

[0014] As an explanation, the problem tracing path is based on the knowledge graph of problem tracing in the new energy vehicle supply chain. Starting from the target fault event node, the path traces along the fault edge to the target vehicle node, from the target vehicle node to the assembly edge to the abnormal associated component version node, from the abnormal associated component version node to the version affiliation edge to the corresponding component node, and from the abnormal associated component version node to the supply edge to the corresponding supplier node. Then, the above paths are connected to generate the problem tracing path corresponding to the fault data of the vehicle to be traced.

[0015] An electronic device, characterized in that the electronic device comprises: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the aforementioned knowledge graph-based new energy vehicle supply chain problem tracing and diagnosis method.

[0016] A computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the above-described knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain.

[0017] The knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain proposed in this invention has the following beneficial technical effects: This application, when forming a knowledge graph for tracing issues in the new energy vehicle supply chain using supply chain traceability data, constructs the knowledge graph by treating component version identifiers as independent entities. It adds version evolution edges to represent the version iteration relationship of components with the same code. This design enables modeling and fault correlation analysis of various component versions under the same component code, effectively determining whether faults are concentrated in a particular version. Then, based on the fault occurrence time and fault type information of the vehicle fault data to be traced, a set of candidate component version nodes is obtained. Version fault correlation analysis is conducted by generating a set of comparable version nodes within the same family. Combining the family correlation rate benchmark value and box plot anomaly detection, abnormally associated component version nodes are identified. Finally, a problem tracing path corresponding to the vehicle fault data to be traced is generated in the new energy vehicle supply chain problem tracing knowledge graph. This application, by overcoming the limitation of traditional traceability that only anchors to component codes, can more accurately identify the target component version and corresponding supplier causing a specific fault, effectively narrowing the traceability scope and significantly improving the efficiency and accuracy of locating and tracing quality problems in the new energy vehicle supply chain. Attached Figure Description

[0018] Figure 1 This is a flowchart of the knowledge graph-based new energy vehicle supply chain problem tracing and diagnosis method of the present invention. Figure 2 This is a flowchart of step S4 of the knowledge graph-based new energy vehicle supply chain problem tracing and diagnosis method of the present invention; Figure 3 This is a logic diagram for determining the version nodes of abnormal related components in the knowledge graph-based new energy vehicle supply chain problem tracing and diagnosis method of the present invention. Detailed Implementation

[0019] Embodiments of the present invention are described in detail below. Examples of these embodiments are illustrated in the accompanying drawings, wherein the same or similar symbols denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0020] like Figure 1-3The knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain, as presented in this application, achieves modeling and fault association analysis for different versions of components under the same component code. This effectively determines whether faults are concentrated in a particular version, thus more accurately identifying whether a fault has an abnormal association with a specific component version. Based on this, a problem tracing path is generated. This alleviates the limitations of existing technologies, which typically rely primarily on component codes to establish the association between vehicles and components. These technologies often fail to further determine whether faults are concentrated in a particular version, accurately characterize the version evolution relationship between different component versions under the same component code, and struggle to distinguish the specific component versions actually assembled in the vehicle and the differences in association between different versions and vehicle faults. Furthermore, when a vehicle experiences a specific type of fault, multiple historical versions corresponding to the same component code are easily analyzed as the same object, making it difficult to determine whether the fault has an abnormal association with a specific component version. This can lead to an excessively large tracing scope for problematic component versions and their corresponding suppliers, impacting the efficiency and accuracy of locating and tracing quality problems in the new energy vehicle supply chain.

[0021] Includes the following steps: S1. Obtain supply chain traceability data in the new energy vehicle supply chain; In an optional embodiment, in S1, the supply chain traceability data includes multiple component version data, multiple vehicle assembly data, and multiple vehicle fault data. in, Each component version data includes the component code, several component version identifiers corresponding to the component code, the effective time of the version corresponding to the component version identifier, and the supplier information corresponding to the component version identifier; Each vehicle assembly data includes the vehicle identification number, the part code installed on the vehicle corresponding to the vehicle identification number, the version identification number of the part installed on the vehicle corresponding to the vehicle identification number, the assembly time, and the service period of the version. Version service period is used to characterize the actual assembly and usage time of a component version on a vehicle; For any vehicle assembly data version service period, the start time of the version service period is the assembly time of that vehicle assembly data. Vehicle assembly data with the same vehicle identifier and the same part code as the vehicle assembly data are obtained from multiple vehicle assembly data, and arranged in chronological order of assembly time. If there is a next vehicle assembly data after the vehicle assembly data, the assembly time in the next vehicle assembly data is taken as the end time of the service period of the version corresponding to the vehicle assembly data; otherwise, the service period of the version of the vehicle assembly data is the continuous period starting from the assembly time of the vehicle assembly data. As an explanation, the component version identifier on the vehicle corresponding to the vehicle identifier corresponds to the component code on the vehicle corresponding to the vehicle identifier. That is to say, the component version identifier is one of several component version identifiers corresponding to the component code. The assembly time is used to characterize the time when the component version corresponding to the component version identifier is assembled to the vehicle corresponding to the vehicle identifier. When the component corresponding to the same component code in the vehicle is replaced, the time when the replaced component version is assembled to the vehicle is used as the assembly time in the corresponding vehicle assembly data, and new vehicle assembly data is formed.

[0022] Each vehicle fault data includes vehicle identification, fault occurrence time, and fault type information; As further clarification, step S1 primarily establishes the data foundation required for subsequent problem tracing, enabling business data originally scattered across the parts management system, supplier management system, vehicle assembly system, and vehicle fault management system to be organized according to a unified data association relationship. By acquiring data across three dimensions—part version, vehicle assembly, and vehicle fault—the objective state of the new energy vehicle supply chain can be described from three aspects: "the evolution of the parts themselves," "the actual usage of the parts in the vehicle," and "the faults that occur during vehicle use." This provides data support for subsequently establishing the association relationship between parts, vehicles, suppliers, and faults.

[0023] The component code identifies the product category or component family to which a component belongs, and the same component code can correspond to multiple different component versions. In actual production, components may undergo version iterations due to design changes, raw material adjustments, process optimizations, supplier switching, quality improvements, etc. Therefore, relying solely on component codes for fault analysis can easily mix the quality performance of different versions, making it impossible to determine whether faults are concentrated in a specific version. To address this, this implementation further utilizes component version identifiers to distinguish different technical states of the same component.

[0024] The version effective time reflects the time when the corresponding component version begins to enter an effective usage state. This time not only represents the sequential evolution relationship between different versions, but can also be used for subsequent time constraints to ensure that component versions that become effective after a failure occur are not included in the analysis of that failure, thereby avoiding misjudgments due to unreasonable time logic.

[0025] Supplier information is used to determine the source of supply for corresponding component versions. In the actual supply chain, different versions of the same component may be supplied by different suppliers, and suppliers may change during version upgrades. Therefore, associating supplier information with specific component versions, rather than just component codes, can further improve the granularity of supply chain traceability. When it is subsequently determined that a certain component version has an abnormal association with the target fault, it is possible to further trace back to the supplier corresponding to that version.

[0026] Vehicle assembly data is used to establish the actual assembly relationship between the complete vehicle and specific component versions. In other words, this data indicates which version of a component was actually used during the production and assembly of a particular vehicle. Since different vehicles may install different versions of components due to factors such as production batch, production time, and model configuration, the actual component composition of a faulty vehicle can be accurately determined by the correspondence between vehicle identification and component version identification, without the need for indirect inference based on production time or material planning.

[0027] Vehicle fault data is used to record fault events that occur during vehicle production, testing, after-sales service, or actual operation. Vehicle identifiers are used to establish a correspondence between fault events and specific vehicles, the fault occurrence time is used to indicate the location of the fault event in the time dimension, and fault type information is used to distinguish different types of faults. Therefore, it is possible not only to determine whether a vehicle has a fault, but also to further determine what type of fault it is and when it occurred.

[0028] Therefore, this step actually completes the preparation of the basic entity data and association basis required for subsequent knowledge graph construction, so that the data conditions between component codes, component versions, suppliers, vehicles and vehicle faults can be further structured and expressed.

[0029] S2. Generate a knowledge graph for tracing issues in the new energy vehicle supply chain based on supply chain traceability data; In an optional embodiment, in S2, a knowledge graph for tracing issues in the new energy vehicle supply chain is generated based on supply chain traceability data, including the following steps: S21. Using the distinct component codes appearing in the component version data as component entities, generate component nodes corresponding to each component code; and use the component code as the node attribute of the corresponding component node. Each distinct component version identifier appearing in the component version data is used as a component version entity. A component version node corresponding to each component version identifier is generated, and the effective time of the version corresponding to the component version identifier is used as the node attribute of the component version node. Each distinct supplier information appearing in the component version data is used as a supplier entity, and a supplier node corresponding to each supplier information is generated; the supplier information is used as the node attribute of the corresponding supplier node. Using the distinct vehicle identifiers appearing in the vehicle assembly data as vehicle entities, vehicle nodes corresponding to each vehicle identifier are generated; the vehicle identifiers are used as node attributes of the corresponding vehicle nodes. Each vehicle fault data is used as a fault event entity, and a fault event node corresponding to each vehicle fault data is generated. The fault occurrence time and fault type information in each vehicle fault data are used as the node attributes of the fault event node. S22. Generate a version affiliation edge between component nodes and component version nodes corresponding to the same component code; wherein, for each component version node, obtain the component code corresponding to the component version node, and generate a version evolution edge between the component version node and component nodes with the same component code. For multiple component version nodes corresponding to the same component code, sort the multiple component version nodes according to the version effective time of each component version node, and generate a version evolution edge between two adjacent component version nodes. The version evolution edge points from the component version node with the earlier version effective time to the component version node with the later version effective time. For each component version identifier, a supply edge is generated between the component version node and the corresponding supplier node based on the supplier information corresponding to the component version identifier. For each vehicle assembly data, obtain the corresponding vehicle node based on the vehicle identifier in the vehicle assembly data, and obtain the corresponding component version node based on the component version identifier in the vehicle assembly data. Generate an assembly edge between the vehicle node and the component version node, and use the assembly time and version service period in the vehicle assembly data as the edge attribute of the assembly edge. For each vehicle fault data, obtain the corresponding vehicle node based on the vehicle identifier in the vehicle fault data, and generate a fault edge between the vehicle node and the fault event node based on the fault event node corresponding to the vehicle fault data. S23. Combine the component nodes, component version nodes, supplier nodes, vehicle nodes, and fault event nodes generated in step S21, as well as the version attribution edge, version evolution edge, supply edge, assembly edge, and fault edge generated in step S22, to form a knowledge graph for tracing issues in the new energy vehicle supply chain.

[0030] As further explained, step S2 is mainly used to convert the supply chain traceability data obtained in step S1 from traditional tabular and field-based data into a knowledge graph with an entity relationship structure, so that the evolution relationship of component versions, supply relationship, assembly relationship and vehicle fault relationship can be associated and traced in a unified data structure.

[0031] When constructing a knowledge graph, treating components, component versions, suppliers, vehicles, and failure events as different types of entity nodes can avoid semantic confusion between different types of data. For example, a component node represents a type of component itself, while a component version node represents the specific version status of that component at a certain stage. By setting them as independent entities, both component category information and version iteration information can be preserved simultaneously.

[0032] Setting the version effective time as a node attribute of the component version node allows for direct time-based judgment when performing version filtering or version evolution analysis in the knowledge graph. Similarly, setting the fault occurrence time and fault type information as node attributes of the fault event node enables the fault event to both establish a connection with the vehicle through the graph structure and perform fault type and time-based filtering using its own attributes.

[0033] Version attribution edges are used to express the relationship of which component a given component version belongs to. After setting this relationship, it's possible to find the component to which a given component version belongs, or to retrieve all corresponding versions from a given component. This relationship is a crucial foundation for subsequently constructing a set of comparable version nodes within the same family.

[0034] Version evolution edges are used to describe the temporal evolution relationship between different versions of the same component. By sorting them according to their effective date, multiple originally independent component versions can be formed into a sequential version evolution chain. This setting can reflect the process of a component gradually evolving from an older version to a newer version, facilitating subsequent analysis of whether a fault correlation changes significantly with version upgrades. For example, when the fault correlation rate of a certain version is significantly higher than its predecessor and subsequent versions, the quality anomalies that version changes may bring can be more intuitively identified.

[0035] The supply side is used to express the supply relationship between component versions and suppliers. Setting the supply relationship at the component version level can accommodate the actual situation where the same component is supplied by different suppliers at different version stages, thereby further determining the supply source of the abnormal version after identifying the abnormal version.

[0036] Assembly edges represent the actual versions of a component installed on a vehicle. This relationship allows for the quick retrieval of all actually installed component versions from a vehicle node, and also enables the reverse lookup of all vehicles equipped with a specific component version. Therefore, it supports both single-vehicle issue tracing and group vehicle fault statistics.

[0037] Fault edges are used to represent the association between a vehicle and its corresponding fault event. Since the same vehicle may have multiple fault events at different times, establishing each vehicle fault data as a fault event node can completely preserve the vehicle's fault history and prevent information from being overwritten between different fault events due to simply setting the "fault attributes" of vehicle nodes.

[0038] After the above processing, the originally relatively independent data is organized into a relational network that allows for multi-hop queries. For example, a fault event can be used to sequentially query the corresponding vehicle, the version of the parts installed in the vehicle, the parts to which the parts belong, and the corresponding supplier. This provides a unified data structure foundation for subsequent processes such as candidate version screening, version comparison analysis, anomaly detection, and supplier tracing.

[0039] S3. Extract the vehicle fault data to be traced from multiple vehicle fault data sets as the vehicle fault data to be traced; generate a set of candidate component version nodes corresponding to the vehicle fault data to be traced based on the knowledge graph of new energy vehicle supply chain problem traceability. In an optional embodiment, in S3, a set of candidate component version nodes corresponding to the fault data of the vehicle to be traced is generated based on the knowledge graph of the new energy vehicle supply chain problem tracing, including the following steps: S31. Obtain the fault event nodes corresponding to the fault data of the vehicle to be traced from the knowledge graph of the new energy vehicle supply chain problem tracing, and use them as target fault event nodes. S32. Take the fault occurrence time in the fault data of the vehicle to be traced as the target fault occurrence time, and take the fault type information in the fault data of the vehicle to be traced as the target fault type information. S33. Take the vehicle node that has a fault edge with the target fault event node as the target vehicle node; S34. Obtain the version nodes of each component that have an assembly edge with the target vehicle node, and obtain the version effective time of each component version node that has an assembly edge with the target vehicle node. For each component version node that has an assembly edge with the target vehicle node, obtain the assembly time of the corresponding assembly edge. S35. Select the component version nodes whose version effective time is no later than the target failure time, whose assembly time of the corresponding assembly edge is no later than the target failure time, and whose target failure time falls within the service period of the version of the corresponding assembly edge as candidate component version nodes, and form a set of candidate component version nodes from each candidate component version node.

[0040] As further explanation, step S3 primarily focuses on identifying component versions that may be associated with the specific fault event requiring diagnosis, thereby narrowing down the scope of subsequent analysis. Since a new energy vehicle typically comprises a large number of components, directly performing fault correlation analysis on all component versions in the supply chain would not only be computationally intensive but would also introduce a significant amount of data that has no actual assembly relationship with the target vehicle. Therefore, this step first utilizes vehicle assembly relationships within a knowledge graph to limit the analysis object to the component versions actually installed in the target vehicle.

[0041] Specifically, by using the fault edge between the target fault event node and the vehicle node, the specific vehicle where the current traceable fault occurred can be determined. Then, by using the assembly edge between the vehicle node and the component version node, the actual component versions assembled in the target vehicle can be obtained. This process ensures that all component versions entering the candidate range have a direct assembly relationship with the target vehicle, thus possessing a real possibility of becoming fault-related objects.

[0042] Furthermore, setting version activation time constraints for the acquired component versions aims to ensure the logical consistency of candidate component versions in terms of time. Fault tracing is a process of cause analysis based on historical states. If a component version only becomes effective after the target fault occurs, that version should not, in principle, be used to explain faults that have already occurred. Therefore, by comparing the version activation time with the target fault occurrence time, versions that were not yet effective when the fault occurred can be excluded.

[0043] The time-based filtering here also helps avoid data interference caused by the continuous updating of the knowledge graph. For example, the knowledge graph may already contain newly released component versions after the target failure occurred. Without time constraints, these subsequent versions might be included in the comparison analysis of later versions, thus affecting the statistical benchmark. Therefore, using the failure occurrence time to form an analytical time section ensures that subsequent analyses are based as much as possible on the supply chain state that already existed at the time of the failure.

[0044] The final set of candidate component version nodes can be understood as the initial investigation scope for the current target fault. This set does not directly indicate that each component version is the cause of the fault, but only that these component versions have the conditions for further anomaly correlation analysis in terms of assembly and time relationships, thus providing an object basis for subsequently evaluating the fault correlation degree of each candidate component version.

[0045] S4. For each candidate component version node in the candidate component version node set, generate a set of comparable version nodes of the same family corresponding to each candidate component version node. In an optional embodiment, such as Figure 2 As shown, in S4, for each candidate component version node in the candidate component version node set, a set of comparable version nodes of the same family corresponding to each candidate component version node is generated, including the following steps: S41. For any candidate component version node in the candidate component version node set, obtain the component node that has a version ownership edge with the candidate component version node, and use it as a candidate component node. S42. In the knowledge graph for tracing issues in the new energy vehicle supply chain, starting from the candidate component node, obtain all component version nodes that have version attribution edges with the candidate component node; S43. Based on the version evolution edges between each component version node, arrange each component version node in sequence to form the version evolution sequence corresponding to the candidate component node. S44. The component version node in the version evolution sequence whose version effective time is no later than the target failure time is taken as the comparable component version node corresponding to the candidate component version node. As an explanation, S3 filters "the version nodes of the components actually assembled in the target vehicle"; S4 filters "all version nodes belonging to the same component as the candidate component version node". Therefore, S3 can only guarantee that the version effective time of the candidate component version node itself is no later than the time of the target failure, but cannot guarantee that other version nodes of the same family obtained by S4 also meet this time condition.

[0046] S45. The candidate component version node and its corresponding comparable component version node form a set of comparable version nodes of the same family.

[0047] As further explained, step S4 is mainly used to establish a comparable set of reference versions for each candidate component version, so that subsequent fault correlation analysis will not only evaluate the fault occurrence rate of a certain version itself, but will be able to make a horizontal comparison between the version and other historical versions of the same component.

[0048] In practical applications, the basic failure probabilities of different types of components may vary significantly. For example, the proportion of certain high-frequency electronic devices experiencing a particular type of failure may generally be higher than that of structural components. If an abnormality is judged solely based on the absolute failure rate of a particular component version, it is easily influenced by the inherent failure level of the component type itself. Therefore, this implementation method uses "different versions of the same component" as the main comparison object, ensuring that the comparison objects maintain a high degree of consistency in component usage, function, and basic structure, thereby improving the rationality of the comparison results.

[0049] For any candidate component version, its corresponding component node is first determined through version attribution. Then, all versions belonging to that component node are further obtained, thus establishing the "family" range of the candidate version. Here, "family" mainly refers to different component versions with the same component code.

[0050] After obtaining multiple versions of the same family, the version evolution relationship can be further utilized to determine the chronological order of each version, thus forming a version evolution sequence for the corresponding components. This version evolution sequence not only displays the chronological order of different versions but also provides a structural basis for subsequent analysis of fault correlation trends. For example, it can be determined whether a certain anomaly began to appear in a certain version, or whether a certain anomaly was improved in subsequent versions.

[0051] Meanwhile, this step still uses the time of the target failure as the time boundary to further restrict versions within the same family. This is because versions that take effect after the target failure occurs are not historical versions observable at the time of the failure. Using these future versions to build a comparison benchmark for the current failure could lead to information overflow. By retaining only versions that were already in effect when the target failure occurred, subsequent comparisons can conform to the temporal causal logic of failure tracing.

[0052] The resulting set of comparable version nodes is equivalent to configuring a reference group with common component attributes and time conditions for each candidate component version. Subsequently, the fault correlation performance of the candidate version can be compared with other versions within this reference group to determine whether the candidate version exhibits an abnormally high degree of fault correlation relative to similar historical versions, rather than simply making an isolated judgment based on its own failure rate.

[0053] S5. Obtain the set of comparable version nodes of the same family corresponding to all candidate component version nodes; For any candidate component version node, the set of comparable version nodes in the same family is analyzed for version fault correlation, and the version fault correlation parameters corresponding to each comparable component version node are obtained. In an optional embodiment, in S5, for any candidate component version node corresponding to a set of comparable version nodes in the same family, version fault correlation analysis is performed on each comparable component version node in the set of comparable version nodes in the same family to obtain the version fault correlation parameters corresponding to each comparable component version node, including the following steps: S51. For any candidate component version node, the set of comparable version nodes in the same family: S52. For any comparable component version node in the set of comparable version nodes of the same family, obtain the vehicle nodes that have an assembly edge with the comparable component version node, and form the set of associated vehicle nodes corresponding to the comparable component version node. S53. Count the number of vehicle nodes in the set of associated vehicle nodes to obtain the number of associated vehicle nodes corresponding to the comparable component version node. S54. For each vehicle node in the set of associated vehicle nodes, obtain the fault event nodes that have a fault edge with the vehicle node, and filter out the fault event nodes of the same type from the fault event nodes. Among them, the same type of fault event node is a fault event node whose fault type information is the same as the target fault type information, whose fault occurrence time is no later than the target fault occurrence time, and whose fault occurrence time falls within the service period of the version of the assembly edge between the comparable component version node and the corresponding vehicle node. S55. Take the vehicle node with at least one similar fault event node as the similar fault vehicle node, and count the number of similar fault vehicle nodes to obtain the number of similar fault vehicle nodes corresponding to the comparable component version node. S56. When the number of associated vehicle nodes is greater than zero, the ratio of the number of similar fault vehicle nodes corresponding to the comparable component version node to the number of associated vehicle nodes shall be used as the version fault association rate corresponding to the comparable component version node. S57. The number of associated vehicle nodes, the number of similar faulty vehicle nodes, and the version fault association rate are used as the version fault association parameters corresponding to the version node of the comparable component.

[0054] As further explained, step S5 is mainly used to quantify the statistical correlation between each component version and the target fault type, providing a quantitative basis for subsequent judgment on whether the candidate component version belongs to the abnormal version.

[0055] For any comparable component version, the assembly edges in the knowledge graph can be used to query all vehicles that have actually installed that version. These vehicles collectively form the actual usage sample of that version. By counting the number of vehicles in this sample, the number of associated vehicle nodes corresponding to that version can be obtained, which reflects the sample size of that component version within the current data range.

[0056] Using the number of vehicles instead of simply counting the number of assembly records avoids duplicate counting when the same vehicle has duplicate records. Similarly, when counting faulty vehicles, this implementation focuses on the "number of vehicles that have experienced at least one target type of fault" rather than directly counting all fault events. Therefore, it avoids overestimating the statistical results due to a vehicle repeatedly experiencing the same type of fault.

[0057] For each vehicle equipped with this version, the corresponding fault events are further queried through the fault edge, and filtered by combining fault type information and fault occurrence time. Maintaining consistency between the fault type and the target fault type ensures that the statistically analyzed fault events belong to the same fault category as the current problem to be diagnosed, thereby avoiding interference from other irrelevant faults in the analysis results.

[0058] By limiting the fault occurrence time to no later than the target fault occurrence time, a unified historical statistical window is created. In other words, when tracing a target fault, only historical faults that occurred up to the time the target fault occurred are used for statistics, not data that appeared after the target fault occurred. This prevents newly added fault information from negatively impacting the diagnostic results of historical faults, and also ensures that statistical data across different versions have a unified time reference.

[0059] After obtaining the total number of associated vehicles and the number of vehicles experiencing the target type of fault, the ratio of these two numbers forms the version fault correlation rate. This correlation rate can be understood as the proportion of vehicles actually equipped with this version of the component that experienced the target type of fault up to the time the target fault occurred. Therefore, this indicator can normalize the fault performance of different versions, making versions with a large number of installations comparable to those with a small number of installations.

[0060] For example, if two versions are installed on different numbers of vehicles, directly comparing the absolute number of faulty vehicles may not accurately reflect the quality performance of the versions; however, by calculating the proportion of faulty vehicles to the total number of related vehicles, the impact of differences in the scale of installation can be reduced.

[0061] Simultaneously retaining the number of associated vehicle nodes, the number of similar faulty vehicle nodes, and the version fault association rate allows subsequent analysis to obtain not only a proportional indicator but also the sample size information corresponding to that proportion. In practice, the sample size can be further considered to assist in judging the analysis results. For example, when a certain version corresponds to only a very small number of vehicles, its fault association rate may have a high degree of randomness; therefore, retaining the number of associated vehicles is beneficial for further evaluating the statistical reliability of the results.

[0062] This step allows for the establishment of version fault association parameters with a unified structure for each comparable version within the same family, further transforming the original knowledge graph-based relational query results into quantitative indicators capable of mathematical comparison and anomaly analysis.

[0063] S6. Generate abnormal associated component version nodes corresponding to the fault data of the vehicle to be traced based on the version fault association parameters corresponding to each comparable component version node. In an optional embodiment, in S6, based on the version fault association parameters corresponding to each comparable component version node, an abnormal associated component version node corresponding to the vehicle fault data to be traced is generated, such as... Figure 3 As shown, it includes the following steps: S61. For each candidate component version node, obtain the version fault association parameters corresponding to each comparable component version node from the set of comparable version nodes in the same family corresponding to the candidate component version node. S62. Based on the version fault association parameters corresponding to each candidate component version node, determine the abnormal associated component version node, specifically including: S63. Determine the family association rate benchmark value corresponding to the family of comparable version nodes based on the version fault association rate in the version fault association parameters corresponding to each comparable component version node. In an optional embodiment, a family association rate benchmark value corresponding to the family of comparable version nodes is determined based on the version fault association rate in the version fault association parameters corresponding to each comparable component version node; the average value of the version fault association rates corresponding to each comparable component version node in the family of comparable version nodes can be calculated as the family association rate benchmark value; or, from the family of comparable version nodes corresponding to the candidate component version node, comparable component version nodes that are the same as the candidate component version node are excluded to obtain a benchmark version node set; based on the version fault association rate in the version fault association parameters corresponding to each comparable component version node in the benchmark version node set, the average value of the version fault association rates of each comparable component version node in the benchmark version node set is calculated as the family association rate benchmark value corresponding to the candidate component version node.

[0064] S64. Obtain the comparable component version node that is the same as the candidate component version node from the set of comparable version nodes of the same family, and use the version fault association rate in the version fault association parameter corresponding to the comparable component version node as the candidate version fault association rate corresponding to the candidate component version node. S65. Obtain the difference between the candidate version fault association rate and the family association rate benchmark value, and use it as the version difference association value corresponding to the candidate component version node. As an explanation, the difference between the candidate version failure correlation rate and the baseline value of the family correlation rate is: candidate version failure correlation rate − family correlation rate baseline value = version difference correlation value. S66. Obtain the version difference correlation value corresponding to each candidate component version node, and use the box plot method to perform anomaly detection on the version difference correlation value corresponding to each candidate component version node. The candidate component version node that meets the anomaly judgment condition of the box plot method is determined as the abnormally correlated component version node.

[0065] As further explained, step S6 is mainly used to determine the relative anomaly of candidate component versions based on the version fault association parameters obtained in step S5, so as to identify component versions whose fault association degree is significantly higher than the normal level of the same family.

[0066] Since the normal failure correlation levels of different components are not entirely the same, this implementation method does not use a uniform fixed failure rate threshold to judge all component versions. Instead, for each candidate component version, a corresponding correlation rate benchmark is established within its corresponding comparable version family. This approach can, to some extent, eliminate the impact of differences in component types.

[0067] The family association rate benchmark is used to characterize the general association level of multiple versions of the same component on the target fault type. The benchmark can be calculated by combining the version fault association rates of all versions in the same family, or it can be calculated using only other versions in the same family after excluding the current candidate version.

[0068] When calculating the average value using all versions, the overall fault correlation level of the entire version family can be obtained relatively directly, and the calculation process is relatively simple. When calculating the baseline value by excluding the current candidate version, the influence of the candidate version's own data on the baseline value can be reduced. Especially when the fault correlation rate of the candidate version is obviously abnormal, it can prevent the outlier from raising the baseline of the same family, thereby improving the sensitivity of anomaly identification.

[0069] The candidate version failure correlation rate represents the failure correlation level of the candidate component version currently being evaluated. The version difference correlation value is obtained by subtracting the family correlation rate benchmark from the candidate version failure correlation rate. This difference value reflects the degree of deviation of the candidate version from the general level of its family.

[0070] When the version difference correlation value is close to zero, it indicates that the fault correlation performance of the candidate version is relatively close to the overall level of the same family; when the version difference correlation value is significantly positive, it indicates that the correlation rate of the target fault type of the candidate version is higher than the reference level of the same family; and the larger the difference value, the higher its degree of abnormality may be relative to the same family version.

[0071] After obtaining the version difference correlation values ​​of multiple candidate component versions corresponding to the current target vehicle, the box plot method is used for anomaly detection, which can further identify versions that are statistically significantly deviated from the overall distribution from multiple candidate versions.

[0072] Box plots are a method for anomaly identification based on data distribution. They do not require a pre-defined fixed failure rate threshold and can adaptively form anomaly judgment boundaries based on the actual distribution of the correlation values ​​between candidate versions. For example, in one implementation, the anomaly detection range can be determined based on the first quartile, the third quartile, and the interquartile range. When the correlation value between the version differences of a candidate component version exceeds the corresponding upper anomaly boundary, it can be considered that this version exhibits an abnormally high degree of fault correlation compared to other candidate component versions.

[0073] Since this implementation method primarily focuses on highly correlated versions that may cause target malfunctions, in actual implementation, it can prioritize identifying candidate component versions whose version difference correlation values ​​fall within the upper abnormal region. This allows for the combination of multiple factors, such as "actual installation on the target vehicle," "abnormally high failure rate of the same family of versions," and "abnormal values ​​relative to other candidate components in the current vehicle," to further narrow down the scope of fault investigation.

[0074] It should be noted that the abnormally related component version nodes identified in this step reflect a significant statistical anomaly correlation between the component version and the target failure, but do not necessarily equate to confirming that the component version is the sole direct cause of the failure. In practical applications, this result can serve as a priority for quality engineers, R&D personnel, and supply chain managers when conducting further physical testing, failure analysis, or supplier quality investigations.

[0075] S7. Obtain the version nodes of abnormal related components corresponding to all vehicle fault data to be traced, and generate the problem tracing path corresponding to the vehicle fault data to be traced based on the knowledge graph of new energy vehicle supply chain problem tracing.

[0076] In an optional embodiment, in S7, based on the knowledge graph of new energy vehicle supply chain problem tracing, a problem tracing path corresponding to the fault data of the vehicle to be traced is generated, including the following steps: S71. Based on all abnormal associated component version nodes, generate the problem tracing path corresponding to the vehicle fault data to be traced in the new energy vehicle supply chain problem tracing knowledge graph. S72. The problem tracing path includes the path from the target fault event node to the target vehicle node via the fault edge, from the target vehicle node to the abnormal associated component version node via the assembly edge, from the abnormal associated component version node to the corresponding component node via the version ownership edge, and from the abnormal associated component version node to the corresponding supplier node via the supply edge.

[0077] As an explanation, the problem tracing path is based on the knowledge graph of problem tracing in the new energy vehicle supply chain. Starting from the target fault event node, the path traces along the fault edge to the target vehicle node, from the target vehicle node to the assembly edge to the abnormal associated component version node, from the abnormal associated component version node to the version affiliation edge to the corresponding component node, and from the abnormal associated component version node to the supply edge to the corresponding supplier node. Then, the above paths are connected to generate the problem tracing path corresponding to the fault data of the vehicle to be traced.

[0078] As a further explanation, step S7 is mainly used to remap the results obtained from the aforementioned anomaly detection to the knowledge graph of the new energy vehicle supply chain problem tracing, so that the simple abnormal part version identification results can be further expanded into a problem tracing path that can reflect the complete relationship between "fault - vehicle - part version - part - supplier".

[0079] The aforementioned steps have been sufficient to identify which component versions actually installed on the target vehicle exhibit abnormal fault correlations compared to their counterparts. However, for actual supply chain quality management, simply obtaining the abnormal version is often insufficient for direct follow-up processing. Quality personnel typically need to further clarify which vehicle the abnormal version is installed on, which component category it belongs to, and which supplier provided it. Therefore, this step utilizes the various relationship edges already established in the knowledge graph to connect the above information.

[0080] First, by tracing the fault relationship from the target fault event node, the target vehicle where the fault occurred can be identified; this part is used to clarify the object of the problem. Subsequently, the target vehicle is located to the abnormal associated component version identified in step S6 through the assembly relationship, thereby forming direct assembly evidence between the faulty vehicle and the suspected abnormal component version.

[0081] After identifying the versions of the abnormally associated components, their corresponding components can be further determined through version attribution, thus clarifying which component the abnormal version belongs to. This process helps R&D or quality personnel to conduct further technical analysis from the perspectives of component function, structure, and version changes.

[0082] Meanwhile, by identifying the supply relationship between the abnormal component version and the supplier, the supplier corresponding to that specific version can be determined, allowing the fault analysis results to be further extended to the supply chain source. Compared to determining the supplier solely based on the component code, this implementation establishes a supply traceability relationship at the component version level. Therefore, even if the same component is supplied by different suppliers at different version stages, the actual supply source corresponding to the abnormal version can be located more accurately.

[0083] By combining the above relationships, a complete problem tracing chain can be formed around the target fault event. This chain can reflect not only the vehicle object where the problem occurred, but also the specific component version that may be associated with the fault anomaly. Furthermore, it can further display the component to which the version belongs and its supplier. Therefore, fault problems discovered at the vehicle operation end or after-sales end can be traced step by step back to the production and assembly stage, the component version stage, and the supplier stage.

[0084] When step S6 identifies multiple abnormally related component versions, corresponding traceability branches can be established for each abnormally related component version, thus forming a traceability result for a target failure event corresponding to multiple potentially abnormal component versions and their supply sources. Relevant personnel can determine the priority of further investigation based on the version difference correlation value, version failure correlation rate, vehicle installation scale, and specific supplier information for each abnormal version.

[0085] This step transforms the aforementioned anomaly identification results obtained from statistical analysis into supply chain traceability results with clear business semantics. This avoids the problem of manually querying vehicle production data, component version data, and supplier data layer by layer in traditional fault analysis, thereby improving the efficiency of fault location and supply chain quality problem tracing in the new energy vehicle supply chain. It also provides data support for subsequent determination of component recall scope, supplier quality rectification, version change assessment, and further verification of fault causes.

[0086] This application, when forming a knowledge graph for tracing issues in the new energy vehicle supply chain using supply chain traceability data, constructs the knowledge graph by treating component version identifiers as independent entities. It adds version evolution edges to represent the version iteration relationship of components with the same code. This design enables modeling and fault correlation analysis of various component versions under the same component code, effectively determining whether faults are concentrated in a particular version. Then, based on the fault occurrence time and fault type information of the vehicle fault data to be traced, a set of candidate component version nodes is obtained. Version fault correlation analysis is conducted by generating a set of comparable version nodes within the same family. Combining the family correlation rate benchmark value and box plot anomaly detection, abnormally associated component version nodes are identified. Finally, a problem tracing path corresponding to the vehicle fault data to be traced is generated in the new energy vehicle supply chain problem tracing knowledge graph. This application, by overcoming the limitation of traditional traceability that only anchors to component codes, can more accurately identify the target component version and corresponding supplier causing a specific fault, effectively narrowing the traceability scope and significantly improving the efficiency and accuracy of locating and tracing quality problems in the new energy vehicle supply chain.

[0087] For clarification, "acquisition" in this application refers to obtaining the required content or data using existing technical means.

[0088] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0089] In the embodiments provided by this invention, it should be understood that the disclosed system or method can be implemented in other ways. For example, the embodiments of the invention described above are merely illustrative; for instance, the division of modules is only a logical functional division, and there may be other division methods in actual implementation.

[0090] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0091] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or in the form of hardware plus software functional modules.

[0092] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the basic characteristics of the present invention.

[0093] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain, characterized in that: Includes the following steps: S1. Obtain supply chain traceability data in the new energy vehicle supply chain; S2. Generate a knowledge graph for tracing issues in the new energy vehicle supply chain based on supply chain traceability data; S3. Extract a vehicle fault data to be traced from multiple vehicle fault data and use it as the vehicle fault data to be traced. Based on the knowledge graph of new energy vehicle supply chain problem tracing, a set of candidate component version nodes corresponding to the vehicle fault data to be traced is generated. S4. For each candidate component version node in the candidate component version node set, generate a set of comparable version nodes of the same family corresponding to each candidate component version node; the comparable component version node is the component version node that belongs to the same component as the corresponding candidate component version node and is used for version fault association comparison. S5. Obtain the set of comparable version nodes in the same family corresponding to all candidate component version nodes; for any candidate component version node corresponding to the set of comparable version nodes in the same family, perform version fault correlation analysis on each comparable component version node in the set of comparable version nodes in the same family to obtain the version fault correlation parameters corresponding to each comparable component version node. S6. Generate abnormal associated component version nodes corresponding to the fault data of the vehicle to be traced based on the version fault association parameters corresponding to each comparable component version node. S7. Obtain the version nodes of abnormal related components corresponding to all vehicle fault data to be traced, and generate the problem tracing path corresponding to the vehicle fault data to be traced based on the knowledge graph of new energy vehicle supply chain problem tracing.

2. The knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain according to claim 1, characterized in that, In S1, supply chain traceability data includes multiple component version data, multiple vehicle assembly data, and multiple vehicle fault data. in, Each component version data includes the component code, several component version identifiers corresponding to the component code, the effective time of the version corresponding to the component version identifier, and the supplier information corresponding to the component version identifier; Each vehicle assembly data includes the vehicle identification number, the part code installed on the vehicle corresponding to the vehicle identification number, the version identification number of the part installed on the vehicle corresponding to the vehicle identification number, the assembly time, and the service period of the version. Version service period is used to characterize the actual assembly and usage time of a component version on a vehicle; For any vehicle assembly data version service period, the start time of the version service period is the assembly time of that vehicle assembly data. Vehicle assembly data with the same vehicle identifier and the same part code as the vehicle assembly data are obtained from multiple vehicle assembly data, and arranged in chronological order of assembly time. If there is a next vehicle assembly data after the vehicle assembly data, the assembly time in the next vehicle assembly data is taken as the end time of the service period of the version corresponding to the vehicle assembly data; otherwise, the service period of the version of the vehicle assembly data is the continuous period starting from the assembly time of the vehicle assembly data. Each vehicle fault data includes vehicle identification, fault occurrence time, and fault type information.

3. The knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain according to claim 2, characterized in that, In S2, a knowledge graph for tracing issues in the new energy vehicle supply chain is generated based on supply chain traceability data, including the following steps: S21. Using the distinct component codes appearing in the component version data as component entities, generate component nodes corresponding to each component code. Each distinct component version identifier appearing in the component version data is used as a component version entity. A component version node corresponding to each component version identifier is generated, and the effective time of the version corresponding to the component version identifier is used as the node attribute of the component version node. Using the distinct supplier information appearing in the component version data as supplier entities, a supplier node corresponding to each supplier information is generated. Using the distinct vehicle identifiers appearing in the vehicle assembly data as vehicle entities, vehicle nodes corresponding to each vehicle identifier are generated. Each vehicle fault data is used as a fault event entity, and a fault event node corresponding to each vehicle fault data is generated. The fault occurrence time and fault type information in each vehicle fault data are used as the node attributes of the fault event node. S22. Generate a version affiliation edge between component nodes and component version nodes corresponding to the same component code; wherein, for each component version node, obtain the component code corresponding to the component version node, and generate a version evolution edge between the component version node and component nodes with the same component code. For multiple component version nodes corresponding to the same component code, sort the multiple component version nodes according to the version effective time of each component version node, and generate a version evolution edge between two adjacent component version nodes. The version evolution edge points from the component version node with the earlier version effective time to the component version node with the later version effective time. For each component version identifier, a supply edge is generated between the component version node and the corresponding supplier node based on the supplier information corresponding to the component version identifier. For each vehicle assembly data, obtain the corresponding vehicle node based on the vehicle identifier in the vehicle assembly data, and obtain the corresponding component version node based on the component version identifier in the vehicle assembly data. Generate an assembly edge between the vehicle node and the component version node, and use the assembly time and version service period in the vehicle assembly data as the edge attribute of the assembly edge. For each vehicle fault data, obtain the corresponding vehicle node based on the vehicle identifier in the vehicle fault data, and generate a fault edge between the vehicle node and the fault event node based on the fault event node corresponding to the vehicle fault data. S23. Combine the component nodes, component version nodes, supplier nodes, vehicle nodes, and fault event nodes generated in step S21, as well as the version attribution edge, version evolution edge, supply edge, assembly edge, and fault edge generated in step S22, to form a knowledge graph for tracing issues in the new energy vehicle supply chain.

4. The knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain according to claim 3, characterized in that, In S3, based on the knowledge graph for tracing issues in the new energy vehicle supply chain, a set of candidate component version nodes corresponding to the fault data of the vehicle to be traced is generated, including the following steps: S31. Obtain the fault event nodes corresponding to the fault data of the vehicle to be traced from the knowledge graph of the new energy vehicle supply chain problem tracing, and use them as target fault event nodes. S32. Take the fault occurrence time in the fault data of the vehicle to be traced as the target fault occurrence time, and take the fault type information in the fault data of the vehicle to be traced as the target fault type information. S33. Take the vehicle node that has a fault edge with the target fault event node as the target vehicle node; S34. Obtain the version nodes of each component that have an assembly edge with the target vehicle node, and obtain the version effective time of each component version node that has an assembly edge with the target vehicle node. For each component version node that has an assembly edge with the target vehicle node, obtain the assembly time of the corresponding assembly edge. S35. Select the component version nodes whose version effective time is no later than the target failure time, whose assembly time of the corresponding assembly edge is no later than the target failure time, and whose target failure time falls within the service period of the version of the corresponding assembly edge as candidate component version nodes, and form a set of candidate component version nodes from each candidate component version node.

5. The knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain according to claim 4, characterized in that, In S4, for each candidate component version node in the candidate component version node set, a set of comparable version nodes of the same family corresponding to each candidate component version node is generated, including the following steps: S41. For any candidate component version node in the candidate component version node set, obtain the component node that has a version ownership edge with the candidate component version node, and use it as a candidate component node. S42. In the knowledge graph for tracing issues in the new energy vehicle supply chain, starting from the candidate component node, obtain all component version nodes that have version attribution edges with the candidate component node; S43. Based on the version evolution edges between each component version node, arrange each component version node in sequence to form the version evolution sequence corresponding to the candidate component node. S44. The component version node in the version evolution sequence whose version effective time is no later than the target failure time is taken as the comparable component version node corresponding to the candidate component version node. S45. The candidate component version node and its corresponding comparable component version node form a set of comparable version nodes of the same family.

6. The knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain according to claim 5, characterized in that, In S5, for any candidate component version node corresponding to a set of comparable version nodes in the same family, version fault correlation analysis is performed on each comparable component version node in the set of comparable version nodes in the same family to obtain the version fault correlation parameters corresponding to each comparable component version node. This includes the following steps: S51. For any candidate component version node, the set of comparable version nodes in the same family: S52. For any comparable component version node in the set of comparable version nodes of the same family, obtain the vehicle nodes that have an assembly edge with the comparable component version node, and form the set of associated vehicle nodes corresponding to the comparable component version node. S53. Count the number of vehicle nodes in the set of associated vehicle nodes to obtain the number of associated vehicle nodes corresponding to the comparable component version node. S54. For each vehicle node in the set of associated vehicle nodes, obtain the fault event nodes that have a fault edge with the vehicle node, and filter out the fault event nodes of the same type from the fault event nodes. Among them, the same type of fault event node is a fault event node whose fault type information is the same as the target fault type information, whose fault occurrence time is no later than the target fault occurrence time, and whose fault occurrence time falls within the service period of the version of the assembly edge between the comparable component version node and the corresponding vehicle node. S55. Take the vehicle node with at least one similar fault event node as the similar fault vehicle node, and count the number of similar fault vehicle nodes to obtain the number of similar fault vehicle nodes corresponding to the comparable component version node. S56. When the number of associated vehicle nodes is greater than zero, the ratio of the number of similar fault vehicle nodes corresponding to the comparable component version node to the number of associated vehicle nodes shall be used as the version fault association rate corresponding to the comparable component version node. S57. The number of associated vehicle nodes, the number of similar faulty vehicle nodes, and the version fault association rate are used as the version fault association parameters corresponding to the version node of the comparable component.

7. The knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain according to claim 6, characterized in that, In S6, based on the version fault association parameters corresponding to each comparable component version node, abnormal associated component version nodes are generated corresponding to the fault data of the vehicle to be traced, including the following steps: S61. For each candidate component version node, obtain the version fault association parameters corresponding to each comparable component version node from the set of comparable version nodes in the same family corresponding to the candidate component version node. S62. Based on the version fault association parameters corresponding to each candidate component version node, determine the abnormal associated component version node, specifically including: S63. Determine the family association rate benchmark value corresponding to the family of comparable version nodes based on the version fault association rate in the version fault association parameters corresponding to each comparable component version node. S64. Obtain the comparable component version node that is the same as the candidate component version node from the set of comparable version nodes of the same family, and use the version fault association rate in the version fault association parameter corresponding to the comparable component version node as the candidate version fault association rate corresponding to the candidate component version node. S65. Obtain the difference between the candidate version fault association rate and the family association rate benchmark value, and use it as the version difference association value corresponding to the candidate component version node. S66. Obtain the version difference correlation value corresponding to each candidate component version node, and use the box plot method to perform anomaly detection on the version difference correlation value corresponding to each candidate component version node. The candidate component version node that meets the anomaly judgment condition of the box plot method is determined as the abnormally correlated component version node.

8. The knowledge graph-based method for tracing and diagnosing problems in the new energy vehicle supply chain according to claim 7, characterized in that, In S7, based on the knowledge graph of new energy vehicle supply chain problem tracing, a problem tracing path corresponding to the fault data of the vehicle to be traced is generated, including the following steps: S71. Based on all abnormal associated component version nodes, generate the problem tracing path corresponding to the vehicle fault data to be traced in the new energy vehicle supply chain problem tracing knowledge graph. S72. The problem tracing path includes the path from the target fault event node to the target vehicle node via the fault edge, from the target vehicle node to the abnormal associated component version node via the assembly edge, from the abnormal associated component version node to the corresponding component node via the version ownership edge, and from the abnormal associated component version node to the corresponding supplier node via the supply edge.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the knowledge graph-based new energy vehicle supply chain problem tracing and diagnosis method as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the knowledge graph-based new energy vehicle supply chain problem tracing and diagnosis method as described in any one of claims 1 to 8.

Images