A vehicle troubleshooting method, device, equipment and storage medium

Abstract

This application provides a vehicle fault diagnosis method, apparatus, device, and storage medium. The method includes: in response to a first fault phenomenon occurring in a vehicle, querying a fault knowledge graph to obtain multiple target entities related to the first fault phenomenon, entity relationships between the multiple target entities, and probability attributes associated with the target entities; at least some of the target entities are failure modes; constructing a target Bayesian network model based on the multiple target entities, entity relationships between the multiple target entities, and probability attributes associated with the target entities; and determining the target failure mode causing the first fault phenomenon from the failure modes based on the target Bayesian network model. This application improves the accuracy of vehicle fault diagnosis by constructing a target Bayesian network model based on entity information related to the fault phenomenon queried from a fault knowledge graph, and by determining the target failure mode causing the fault phenomenon based on the target Bayesian network model.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, specifically to a vehicle fault diagnosis method, apparatus, equipment, and storage medium. Background Technology

[0002] As the automotive industry accelerates its transformation towards the "four modernizations" of electrification, intelligence, connectivity, and sharing, the integration and electronic sophistication of vehicle systems have significantly improved. The introduction of new components such as power batteries, autonomous driving controllers, and domain controllers has led to complex vehicle faults exhibiting multi-dimensional coupling and uncertain causal relationships. Therefore, the accuracy of fault diagnosis directly impacts driving safety and maintenance costs. Currently, vehicle fault diagnosis relies heavily on the manual experience of repair technicians, which is highly susceptible to subjective factors and cannot guarantee accuracy. Summary of the Invention

[0003] The main objective of this application is to provide a vehicle fault diagnosis method, apparatus, device, and storage medium, aiming to improve the accuracy of vehicle fault diagnosis.

[0004] This application provides a vehicle fault diagnosis method, including: In response to a first vehicle malfunction, multiple target entities, entity relationships between the target entities, and probability attributes associated with the target entities are retrieved from a fault knowledge graph; each target entity is an entity related to the first malfunction; at least some of the target entities are failure modes; and there are multiple failure modes. Based on the multiple target entities, the entity relationships between the multiple target entities, and the probability attributes associated with the target entities, a target Bayesian network model corresponding to the first fault phenomenon is constructed. The directed acyclic graph structure of the target Bayesian network model includes a virtual fault phenomenon total node, and the nodes of multiple failure modes are respectively directed to the fault phenomenon total node with directed edges. The conditional probability table of the target Bayesian network model is determined based on the single-point failure prior hypothesis, which is used to characterize that the fault phenomenon corresponding to the fault phenomenon total node occurs only when any one of the failure modes occurs alone. Based on the target Bayesian network model, a target failure mode is determined in the failure modes; the target failure mode is used to indicate the cause of the failure that leads to the first failure phenomenon.

[0005] In one embodiment, the probability attribute includes the prior probability of the target entity and the conditional probability of the entity relationship associated with the target entity; The step of constructing a target Bayesian network model corresponding to the first fault phenomenon based on the multiple target entities, the entity relationships between the multiple target entities, and the probability attributes associated with the target entities includes: Based on the multiple target entities and the entity relationships between the multiple target entities, a first directed acyclic graph structure is constructed; the first directed acyclic graph structure includes multiple nodes and directed edges between the multiple nodes; the nodes are used to represent the target entities; the directed edges are used to represent the entity relationships; A first conditional probability table is determined based on the prior probability of the target entity and the conditional probability of the entity relationship associated with the target entity. The target Bayesian network model is generated based on the first directed acyclic graph structure and the first conditional probability table.

[0006] In one embodiment, the number of failure modes is multiple; determining the target failure mode among the failure modes based on the target Bayesian network model includes: For each failure mode, based on the target Bayesian network model, the posterior probability of the failure mode occurring when the first failure phenomenon occurs is inferred; Based on the posterior probability of each failure mode, the failure mode with the highest posterior probability is selected from multiple failure modes. The failure mode with the highest posterior probability is determined as the target failure mode.

[0007] In one embodiment, the plurality of target entities further includes a detection means and a second fault phenomenon directly related to the detection means in the entity relationship; The step of inferring the posterior probability of the failure mode occurring under the condition that the first fault phenomenon occurs, based on the target Bayesian network model, includes: Based on the target detection result corresponding to the detection method, a first evidence variable value is determined; the first evidence variable value is used to characterize the target detection result indicating that the second fault phenomenon has occurred. Based on the first evidence variable value and the second evidence variable value, the target evidence variable value is determined; the second evidence variable value is used to characterize the occurrence of the first fault phenomenon. Based on the target Bayesian network model, infer the posterior probability of the failure mode occurring given that the target evidence variable value is true.

[0008] In one embodiment, determining the target evidence variable value based on the first evidence variable value and the second evidence variable value includes: Based on the target Bayesian network model, infer the posterior probability of the second fault phenomenon occurring when the value of the first evidence variable is true; In response to the posterior probability of the second fault phenomenon occurring being within a preset probability range, the target evidence variable value is determined based on the first evidence variable value, the second evidence variable value, and the third evidence variable value; the third evidence variable value is used to characterize the occurrence of the second fault phenomenon.

[0009] In one embodiment, after determining the target failure mode among the failure modes based on the target Bayesian network model, the method further includes: Obtain the target variable values ​​of the plurality of target entities; the target variable values ​​are the variable values ​​observed and / or inferred during the process of determining the target failure mode based on the target Bayesian network model. Update the fault diagnosis case information based on the target variable values ​​of the multiple target entities; Based on the fault diagnosis case information, update the first data of the target Bayesian network model; Based on the updated first data, update the second data in the fault knowledge graph that corresponds to the first data.

[0010] In one embodiment, the fault investigation case information contains latent variables; the latent variables are used to characterize entities where the target variable value is missing. The step of updating the first data of the target Bayesian network model based on the fault diagnosis case information includes: Based on the directed acyclic graph structure and conditional probability table of the target Bayesian network model, the expected value of the posterior probability of the latent variable is determined. Based on the expected value of the posterior probability of the latent variable, update at least some of the probability values ​​in the conditional probability table of the target Bayesian network model. The first data includes at least a portion of the probability values.

[0011] In one embodiment, the first data includes at least a portion of the directed edges in the directed acyclic graph structure of the target Bayesian network model; The step of updating the first data of the target Bayesian network model based on the fault diagnosis case information includes: Based on the fault diagnosis case information, conditional independence is tested for each pair of nodes in the directed acyclic graph structure to obtain the conditional independence test results; each pair of nodes consists of any two nodes in the directed acyclic graph structure. Based on the conditional independence test results, at least some of the directed edges in the directed acyclic graph structure are updated.

[0012] This application embodiment also provides a vehicle fault diagnosis device, the device including an entity query module, a model building module and a failure mode determination module; The entity query module is used to respond to a first vehicle malfunction by querying a fault knowledge graph to obtain multiple target entities, entity relationships between the multiple target entities, and probability attributes associated with the target entities; each target entity is an entity related to the first malfunction; at least some of the target entities are failure modes; and the number of failure modes is multiple. The model building module is used to construct a target Bayesian network model corresponding to the first fault phenomenon based on the multiple target entities, the entity relationships between the multiple target entities, and the probability attributes associated with the target entities. The directed acyclic graph structure of the target Bayesian network model includes a virtual fault phenomenon total node, and the nodes of multiple failure modes are respectively directed to the fault phenomenon total node with directed edges. The conditional probability table of the target Bayesian network model is determined based on the single-point failure prior hypothesis, which is used to characterize that the fault phenomenon corresponding to the fault phenomenon total node occurs only when any one of the failure modes occurs alone. The failure mode determination module is used to determine a target failure mode in the failure modes based on the target Bayesian network model; the target failure mode is used to indicate the cause of the failure that leads to the first failure phenomenon.

[0013] This application also provides a vehicle fault diagnosis device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the vehicle fault diagnosis method described above.

[0014] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described vehicle fault diagnosis method.

[0015] This application provides a vehicle fault diagnosis method, apparatus, device, and storage medium. By constructing a target Bayesian network model based on entity information related to the fault phenomenon retrieved from a fault knowledge graph, and determining the target failure mode that causes the fault phenomenon based on the target Bayesian network model, the accuracy of vehicle fault diagnosis can be improved. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the vehicle fault diagnosis method provided in the embodiments of this application.

[0017] Figure 2 This is a schematic diagram of the structure of the fault knowledge graph provided in the embodiments of this application. Figure 1 .

[0018] Figure 3 This is a schematic diagram of the structure of the fault knowledge graph provided in the embodiments of this application. Figure 2 .

[0019] Figure 4 This is a schematic diagram of the structure of the Bayesian network model provided in the embodiments of this application.

[0020] Figure 5 This is a schematic diagram illustrating the specific process of the vehicle fault diagnosis method provided in the embodiments of this application.

[0021] Figure 6 This is a schematic diagram of the vehicle fault diagnosis device provided in the embodiments of this application.

[0022] Figure 7 This is a schematic diagram of the vehicle fault diagnosis equipment provided in the embodiments of this application. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0024] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the digit " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0025] The vehicle fault diagnosis method provided in this application can be applied to vehicle fault diagnosis equipment or the software of such equipment. The vehicle fault diagnosis equipment can be a vehicle, a terminal, or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, or as a server cluster or distributed system composed of multiple physical servers. The software can be an application that implements the vehicle fault diagnosis method, but is not limited to the above forms.

[0026] The vehicle fault diagnosis method provided in this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0027] Please see Figure 1 The vehicle fault diagnosis method provided in this application embodiment may include: Step S101: In response to the occurrence of the first fault phenomenon in the vehicle, multiple target entities, entity relationships between multiple target entities, and probability attributes associated with the target entities are obtained from the fault knowledge graph; each target entity is an entity related to the first fault phenomenon; at least some of the target entities are failure modes; In practice, the first fault phenomenon of a vehicle can be determined by at least one of the following methods: obtaining customer complaint information related to the first fault phenomenon, reading the fault code corresponding to the first fault phenomenon, receiving the warning signal corresponding to the first fault phenomenon output by the vehicle-mounted or cloud-based warning model, or detecting the fault symptoms corresponding to the first fault phenomenon.

[0028] Optionally, the fault knowledge graph can be constructed based on materials such as vehicle Failure Mode and Effects Analysis (FMEA), Fault Tree Analysis (FTA), and a failure case library. The fault knowledge graph can include multiple entities, entity relationships between multiple entities, and probabilistic attributes associated with entities.

[0029] In practical implementation, the fault knowledge graph can be stored in a Neo4J graph database or a MySQL relational database. Then, a query statement can be written using the first fault phenomenon and / or its corresponding fault code. This query statement can then be used to retrieve multiple target entities related to the first fault phenomenon, the entity relationships between these target entities, and the probabilistic attributes associated with each target entity from the fault knowledge graph. Optionally, the retrieved target entities may include the first fault phenomenon, multiple failure modes that may lead to the first fault phenomenon, and detection methods that can be used to prove the occurrence of the first fault phenomenon.

[0030] Step S102: Based on multiple target entities, the entity relationships between multiple target entities, and the probability attributes associated with the target entities, construct a target Bayesian network model corresponding to the first fault phenomenon; Optionally, the probabilistic attributes associated with the target entity may include the prior probability of the target entity and the conditional probability of the entity relationship associated with the target entity. Optionally, the prior probability may refer to the initial estimate of the likelihood of the target entity occurring based on historical statistics or expert experience in the absence of any observational evidence.

[0031] Optionally, entity relationships can include cause, prove, and contribute to. Conditional probability can be used to characterize the strength of entity relationships between target entities, such as the probability that target entity A causes target entity B to occur, the probability that target entity C proves target entity B to occur, and the probability that target entity D contributes to target entity B to occur.

[0032] In practice, based on multiple target entities, the entity relationships between multiple target entities, and the probability attributes associated with the target entities, the data mapping relationship between the fault knowledge graph and the Bayesian network model can be queried to obtain the target data of the Bayesian network model; then, the target Bayesian network model corresponding to the first fault phenomenon can be constructed based on the target data.

[0033] Step S103: Based on the target Bayesian network model, determine the target failure mode in the failure modes; the target failure mode is used to indicate the cause of failure that leads to the occurrence of the first failure phenomenon.

[0034] In practice, for each failure mode, the posterior probability of the failure mode occurring when the first failure phenomenon occurs can be inferred based on the target Bayesian network model; then, based on the posterior probability of each failure mode, the target failure mode that causes the first failure phenomenon can be determined among multiple failure modes.

[0035] In practice, the failure mode with the highest posterior probability among multiple failure modes can be identified as the target failure mode; alternatively, the failure mode with a posterior probability greater than a probability threshold among multiple failure modes can be identified as the target failure mode; the probability threshold can be obtained through calibration.

[0036] This application embodiment improves the accuracy of vehicle fault diagnosis by constructing a target Bayesian network model based on entity information related to the fault phenomenon retrieved from a fault knowledge graph, and determining the target failure mode that causes the fault phenomenon based on the target Bayesian network model.

[0037] In practice, information such as fault phenomena, failure modes, faulty components, fault codes, detection methods, on-site detection methods, and signal inspection logic can be extracted from materials such as FMEA, FTA, and failure case libraries, and summarized into corresponding entity instances in a fault knowledge graph to construct a system like... Figure 2 The fault knowledge graph shown. For example... Figure 2As shown, a fault knowledge graph can include entities such as fault phenomena, detection methods, failure modes, faulty components, and solutions, as well as entity relationships such as "cause," "prove," "happen in," and "solve." The failure mode entity can include a priori probability attributes of the failure mode; the cause and prove entity relationships can include conditional probability attributes, representing the probability of the fault phenomenon occurring given the failure mode and the probability that the fault phenomenon has actually occurred given abnormal detection results, respectively.

[0038] Optionally, the detection methods may include at least one of signal inspection and on-site detection. To represent the AND logic in the FTA fault tree, a logical virtual entity, "and" (also called an AND gate), can be introduced into the fault knowledge graph. For example... Figure 3 As shown, the two detection methods, "Signal Check 1" and "Signal Check 2," are connected by the logical virtual entity AND to prove the existence of a fault. In this way, the fault knowledge graph can carry the AND logic gate of the FTA fault tree: only when both signal checks are simultaneously abnormal can it be determined that a fault has indeed occurred.

[0039] In practice, based on the first observed fault phenomenon in the fault troubleshooting scenario, multiple target entities related to the first fault phenomenon, the entity relationships between multiple target entities, and the probability attributes associated with the target entities can be queried from the fault knowledge graph. Then, a target Bayesian network model corresponding to the first fault phenomenon can be constructed based on the queried entity information. Based on the target Bayesian network model, the target failure mode that caused the first fault phenomenon can be determined. For specific implementation details, please refer to the relevant description below.

[0040] In one embodiment, step S102 above: constructing a target Bayesian network model corresponding to the first fault phenomenon based on multiple target entities, the entity relationships between the multiple target entities, and the probability attributes associated with the target entities, includes: Based on multiple target entities and the entity relationships between them, a first directed acyclic graph structure is constructed. The first directed acyclic graph structure includes multiple nodes and directed edges between the nodes. Nodes are used to represent target entities, and directed edges are used to represent entity relationships. The first conditional probability table is determined based on the prior probability of the target entity and the conditional probability of the entity relationship associated with the target entity. Based on the first directed acyclic graph structure and the first conditional probability table, a target Bayesian network model is generated.

[0041] In practice, we can first construct random variable nodes (variable nodes such as fault phenomena, failure modes, detection methods, etc.) of a Bayesian network based on multiple target entities, and then construct directed edges between nodes based on the entity relationships between multiple target entities to obtain the first directed acyclic graph structure. Next, a first conditional probability table of the Bayesian network can be generated based on the probability attributes associated with the target entity. This conditional probability table may include, but is not limited to, the prior probability of the failure mode, the conditional probability of the cause relationship between the failure mode and the fault phenomenon, and the conditional probability of the prove relationship between the detection method and the fault phenomenon.

[0042] Then, the first directed acyclic graph structure can be used as the skeleton of the Bayesian network, and the first conditional probability table can be used as the quantitative parameters of the Bayesian network to form a complete Bayesian network model, namely the target Bayesian network model.

[0043] Since each node in a directed acyclic graph (DAG) structure represents a random variable, when multiple target entities, including logical virtual entities, are queried, they cannot be represented in the DAG structure. Therefore, when constructing the target Bayesian network model, logical virtual entities need to be eliminated so that they are not represented in the DAG structure, but are instead converted into a specific probability distribution in a conditional probability table. For specific implementation details, please refer to the relevant description below.

[0044] The structure of the first directed acyclic graph obtained can be as follows: Figure 4 As shown, Figure 4 In the diagram, a, b, and c represent discrete random variables of different target entities related to the first fault phenomenon. The nodes corresponding to discrete random variable a and discrete random variable b are connected by directed edges to the node corresponding to discrete random variable c. Discrete random variables can characterize the mathematical representation of their corresponding target entities in a Bayesian network, typically taking binary discrete values: a value of 1 indicates that the corresponding target entity has occurred / exists, and a value of 0 indicates that the corresponding target entity has not occurred / does not exist. In this way, qualitative concepts in the knowledge graph can be transformed into mathematical variables that can be used for probability calculations.

[0045] Furthermore, based on the influence patterns of discrete random variables a and b on discrete random variable c, the probability values ​​associated with discrete random variable c in the first conditional probability table can be determined. Optionally, when discrete random variables a and b independently influence discrete random variable c, the probability values ​​associated with discrete random variable c can be as shown in Table 1: Table 1. Probability values ​​related to discrete random variable c under the independent influence model.

[0046] Wherein, P(c=1|b=1) represents the probability of the occurrence of the target entity corresponding to discrete random variable c given that the target entity corresponding to discrete random variable b has occurred, and P(c=1|a=1) represents the probability of the occurrence of the target entity corresponding to discrete random variable c given that the target entity corresponding to discrete random variable a has occurred. Both can be obtained by querying the fault knowledge graph.

[0047] Optionally, when discrete random variables a and b affect discrete random variable c through the logical virtual entity AND, the probability values ​​associated with discrete random variable c can be as shown in Table 2: Table 2. Probability values ​​related to discrete random variable c under the AND logic influence mode.

[0048] Here, P(c=1|a=1, b=1) represents the probability that the target entity corresponding to discrete random variable c occurs given that the target entities corresponding to discrete random variables a and b both occur. This probability can be obtained by querying the fault knowledge graph. This eliminates the need to represent logical virtual entities in the directed acyclic graph structure while retaining the meaning of logical virtual entities: discrete random variable c only occurs when discrete random variables a and b both occur.

[0049] This application embodiment constructs a first directed acyclic graph structure based on multiple target entities and the entity relationships between them, and determines a first conditional probability table based on the prior probability of the target entities and the conditional probability of the entity relationships associated with the target entities, thereby obtaining a target Bayesian network model. This can improve the accuracy of establishing a target Bayesian network corresponding to the first fault phenomenon, thereby enabling the target Bayesian network to accurately determine the target failure mode that causes the first fault phenomenon.

[0050] Optionally, the number of the aforementioned failure modes is multiple. In one embodiment, the generation of the target Bayesian network model based on the first directed acyclic graph structure and the first conditional probability table includes: A virtual fault phenomenon node is added to the first directed acyclic graph structure to obtain a second directed acyclic graph structure; the nodes of multiple failure modes are respectively directed to the fault phenomenon node with directed edges. Based on the first conditional probability table and the single-point failure prior hypothesis, the second conditional probability table is determined. The single-point failure prior hypothesis is used to characterize that the failure phenomenon corresponding to the total node of the failure phenomenon occurs only when any failure mode occurs alone. Based on the second directed acyclic graph structure and the second conditional probability table, a target Bayesian network model is generated.

[0051] Considering that about 90% of failures in engineering practice are single-point failures: when a failure occurs, only one failure mode / component fails, in actual implementation, a virtual fault phenomenon total node can be added to the first directed acyclic graph structure, and all failure mode nodes directly point to this total node, thus obtaining the second directed acyclic graph structure.

[0052] Meanwhile, the conditional probability of the total node can be set in the first conditional probability table to be 1 only when a single failure mode occurs, and 0 otherwise, thus obtaining the second conditional probability table. In this way, the single-point failure prior hypothesis can be integrated into the target Bayesian network model.

[0053] Subsequently, the second directed acyclic graph structure can be used as the skeleton of the Bayesian network, and the second conditional probability table can be used as the quantitative parameters of the Bayesian network to form a complete Bayesian network model, namely the target Bayesian network model.

[0054] This application's embodiments incorporate the single-point failure prior assumption into the target Bayesian network model, making the target Bayesian network model tend to identify a single failure mode as the cause of the failure during the reasoning process, rather than multiple failure modes occurring simultaneously. This can improve the efficiency of vehicle fault diagnosis while ensuring the accuracy of vehicle fault diagnosis.

[0055] Optionally, the number of failure modes is multiple. In one embodiment, the step S103 above, which determines the target failure mode based on the target Bayesian network model, includes: For each failure mode, based on the target Bayesian network model, the posterior probability of the failure mode occurring when the first failure phenomenon occurs is inferred; Based on the posterior probability of each failure mode, the failure mode with the highest posterior probability is selected from multiple failure modes. The failure mode with the highest posterior probability is identified as the target failure mode.

[0056] In practical implementation, for each failure mode, a target Bayesian network model can be used to infer the posterior probability of the failure mode occurring given the conditional probability of the target evidence variable value being true when the failure mode occurs, based on the conditional probability of the failure mode value being true and the prior probability of the failure mode. This yields the posterior probability of the failure mode occurring given the first failure phenomenon. Optionally, the target evidence variable value can be used to characterize the occurrence of the first failure phenomenon.

[0057] Alternatively, the posterior probability of the failure mode occurring given that the target evidence variable value is true can be calculated using formula (1): P(FM=1|E) = P(E|FM=1) × P(FM=1) / P(E) (1) Where E represents the value of the target evidence variable, P(FM=1|E) represents the posterior probability of the failure mode occurring given that the value of the target evidence variable is true, P(FM=1) represents the prior probability of the failure mode, and P(E) is the prior probability of the value of the target evidence variable.

[0058] Optionally, P(E|FM=1) can be obtained by chain calculation based on the Conditional Probability Table (CPT) of the target Bayesian network model, along the causal chain from the failure mode to the first failure phenomenon. The conditional probability table can be either the first conditional probability table or the second conditional probability table mentioned above.

[0059] This application embodiment improves the accuracy of vehicle fault diagnosis by inferring the posterior probability of a failure mode occurring when the first fault phenomenon occurs based on a target Bayesian network model for each failure mode, and by determining the failure mode with the highest posterior probability as the target failure mode that causes the first fault phenomenon.

[0060] Optionally, the multiple target entities also include detection methods and a second fault phenomenon directly related to the detection methods in the entity relationship. In one embodiment, the above-mentioned reasoning of the posterior probability of the failure mode occurring when the first fault phenomenon occurs based on the target Bayesian network model includes: Based on the target detection results corresponding to the detection method, the value of the first evidence variable is determined; the value of the first evidence variable is used to characterize the target detection results indicating the occurrence of the second fault phenomenon. Based on the values ​​of the first and second evidence variables, the target evidence variable value is determined; the second evidence variable value is used to characterize the occurrence of the first fault phenomenon. Based on the target Bayesian network model, the posterior probability of the failure mode occurring is obtained when the target evidence variable value is true, thus yielding the posterior probability of the failure mode occurring when the first failure phenomenon occurs.

[0061] Optionally, the entity relationship between the detection means, the first fault phenomenon, and the second fault phenomenon can be: the detection means is used to prove whether the second fault phenomenon has occurred, and the occurrence of the second fault phenomenon will cause the occurrence of the first fault phenomenon. In this case, the detection result of the detection means can indirectly prove whether the first fault phenomenon has occurred.

[0062] In practical implementation, the detection result corresponding to the detection method can be set as the first evidence variable. When the detection result of the detection method is the target detection result, the value of the first evidence variable can be set to 1. Optionally, the target detection result can be a detection result indicating the occurrence of a second fault phenomenon, such as an anomaly; setting the value of the first evidence variable to 1 can characterize that the target detection result indicates the occurrence of a second fault phenomenon.

[0063] Optionally, if the occurrence of the first fault phenomenon is observed, the value of the second evidence variable can be set to 1. The target evidence variable value can be directly determined based on the values ​​of the first and second evidence variables, such as: the target evidence variable value can be: first evidence variable value = 1, second evidence variable value = 1. Alternatively, it can be determined based on the posterior probability of the occurrence of the second fault phenomenon given the first evidence variable value, and the value of the second evidence variable. For specific implementation details, please refer to the relevant descriptions below, which will not be described here.

[0064] Next, the posterior probability of the failure mode occurring under the condition that the first fault phenomenon occurs can be obtained using the above formula (1).

[0065] This application embodiment determines the first evidence variable value based on the target detection result corresponding to the detection method; determines the target evidence variable value based on the first evidence variable value and the second evidence variable value; and infers the posterior probability of the failure mode occurring when the target evidence variable value is true based on the target Bayesian network model, taking into account the impact of the occurrence of the second fault phenomenon on the occurrence of the first fault phenomenon, which can improve the accuracy of vehicle fault diagnosis.

[0066] In one embodiment, determining the target evidence variable value based on the first evidence variable value and the second evidence variable value includes: Based on the target Bayesian network model, infer the posterior probability of the occurrence of the second fault phenomenon given that the value of the first evidence variable is true; In response to the posterior probability of the second fault occurring being within a preset probability range, a target evidence variable value is determined based on the values ​​of the first evidence variable, the second evidence variable, and the third evidence variable; the third evidence variable value is used to characterize the occurrence of the second fault.

[0067] In practical implementation, the target Bayesian network model can be used, referring to the above formula (1), to infer the posterior probability of the second fault phenomenon occurring when the value of the first evidence variable is true; then, the second fault phenomenon with a posterior probability within the preset probability range can be marked as observed, and its corresponding third evidence variable value can be determined to be 1; then, the target evidence variable value can be determined as: first evidence variable value = 1, second evidence variable value = 1, and third evidence variable value = 1. Optionally, the preset probability range can be obtained through calibration, and the preset probability range can be greater than or equal to 50%.

[0068] This application embodiment determines the observed second fault phenomenon based on the posterior probability of the second fault phenomenon occurring when the first evidence variable value is true, and determines the target evidence variable value based on the observed second fault phenomenon. This can improve the accuracy of inferring the posterior probability of the failure mode occurring when the target evidence variable value is true, thereby improving the accuracy of vehicle fault diagnosis.

[0069] In one embodiment, after determining the target failure mode based on the target Bayesian network model in step S103 above, the vehicle fault diagnosis method provided in this application further includes: Obtain the target variable values ​​of multiple target entities; the target variable values ​​are the variable values ​​observed and / or inferred during the process of determining the target failure mode based on the target Bayesian network model. Update the fault diagnosis case information based on the target variable values ​​of multiple target entities; Based on the fault diagnosis case information, update the first data of the target Bayesian network model; Based on the updated first data, update the second data in the fault knowledge graph that corresponds to the first data.

[0070] Optionally, the target variable values ​​for multiple target entities may include: Sym1=1, Sym2=1, FM1=1, FM2=0, and TR=1. Sym1=1 and Sym2=1 can respectively represent that the first fault phenomenon and the second fault phenomenon are observed; FM1=1 and FM2=0 can represent that the failure mode leading to the first fault phenomenon, as inferred, is failure mode 1, not failure mode 2; and TR=1 can represent that the detection result of the observed detection method indicates that the second fault phenomenon has occurred.

[0071] In practice, each historical fault investigation case can be organized into variable values ​​corresponding to entities such as failure mode, fault phenomenon, and detection method, thus obtaining fault investigation case information. When a new fault investigation is completed, the target variable values ​​of multiple target entities in this fault investigation can be updated to the fault investigation case information.

[0072] Optionally, the first data may include at least some probability values ​​in the conditional probability table of the target Bayesian network model and / or at least some directed edges in the directed acyclic graph structure. In actual implementation, at least some probability values ​​in the conditional probability table of the target Bayesian network model and / or at least some directed edges in the directed acyclic graph structure can be updated based on fault diagnosis case information. For specific implementation details, please refer to the relevant description below. Optionally, the conditional probability table can be either the first conditional probability table or the second conditional probability table described above; the directed acyclic graph structure can be either the first directed acyclic graph structure described above or the second directed acyclic graph structure described above.

[0073] In practice, if at least some probability values ​​in the target Bayesian network model CPT are updated, the updated probability values ​​in CPT can be written back to the corresponding probability attributes in the fault knowledge graph. For example, if P(Sym1|FM1) in CPT is updated from 0.7 to 0.85, the probability attribute corresponding to the cause relationship of FM1→Sym1 in the fault knowledge graph can be updated to 0.85.

[0074] If at least some of the directed edges in the target Bayesian network model are updated, the corresponding entity relations in the fault knowledge graph can be updated. For example, if a new directed edge is generated in the target Bayesian network model, an entity relation corresponding to that directed edge can be created in the fault knowledge graph; if a directed edge is deleted in the target Bayesian network model, the entity relation corresponding to that directed edge can be deleted in the fault knowledge graph, or the entity relation can be marked as invalid.

[0075] This application embodiment updates fault diagnosis case information based on target variable values ​​of multiple target entities; updates the first data of the target Bayesian network model based on the fault diagnosis case information; and updates the second data corresponding to the first data in the fault knowledge graph based on the updated first data. This not only enables dynamic updating of the fault knowledge graph but also ensures the accuracy of the fault knowledge graph update, thereby improving the accuracy of vehicle fault diagnosis using the fault knowledge graph.

[0076] Optionally, if there are no latent variables in the fault investigation case information, maximum likelihood estimation can be used to directly update at least some of the probability values ​​in the conditional probability table of the target Bayesian network model based on the statistical frequency of the variable values ​​corresponding to each entity; optionally, latent variables can be used to characterize entities with missing target variable values.

[0077] In one implementation, when latent variables exist in the fault diagnosis case information, the first data for updating the target Bayesian network model based on the fault diagnosis case information includes: Based on the directed acyclic graph structure and conditional probability table of the target Bayesian network model, the expected value of the posterior probability of the latent variable is determined. Update at least some of the probability values ​​in the conditional probability table of the target Bayesian network model based on the expected value of the posterior probability of the latent variables.

[0078] In practice, based on the given directed acyclic graph structure and conditional probability table of the target Bayesian network model, the expected value of the posterior probability of the latent variables can be calculated using Bayesian inference rules. This expected value reflects the probability distribution of the latent variables taking different states under the current target Bayesian network model. Then, based on the calculated expected value, at least some probability values ​​in the conditional probability table can be updated using the maximum likelihood estimation criterion to maximize the likelihood function. Subsequently, the above expected value calculation and parameter update steps can be repeated until the change in probability values ​​in the conditional probability table is less than a preset threshold or the preset maximum number of iterations is reached, thereby obtaining a converged conditional probability table.

[0079] This application embodiment determines the expected value of the posterior probability of the latent variable based on the directed acyclic graph structure and conditional probability table of the target Bayesian network model, and updates at least some probability values ​​in the conditional probability table of the target Bayesian network model based on the expected value of the posterior probability of the latent variable. This can ensure the accuracy of the update of the conditional probability table in the Bayesian network model in the absence of complete observation data.

[0080] In one embodiment, the first data for updating the target Bayesian network model based on fault diagnosis case information includes: Based on the fault diagnosis case information, conditional independence tests are performed on each pair of nodes in the directed acyclic graph structure, and the conditional independence test results are obtained; each pair of nodes consists of any two nodes in the directed acyclic graph structure. Based on the conditional independence test results, update at least some of the directed edges in the directed acyclic graph structure.

[0081] In practice, we can start with the current directed acyclic graph structure of the target Bayesian network model. Based on the fault investigation case information, we can perform conditional independence tests (such as chi-square tests or G-tests) on any two variable nodes. If the two variables are independent given other variables, we can delete the directed edge between them to eliminate redundant entity relationships in the fault knowledge graph. Conversely, if the two variables are not independent given other variables and there is no directed edge in the current directed acyclic graph structure, we can add the corresponding directed edge to supplement the entity relationships not recorded in the knowledge graph.

[0082] If the conditional independence test result is inconsistent with the entity relationship indicated by the directed edge, the direction of the directed edge can be redefined according to the v-structure (collider) criterion and domain knowledge, so that the directed edge after adjustment meets the conditional independence requirement.

[0083] This application embodiment performs conditional independence checks on each pair of nodes in the directed acyclic graph structure based on fault diagnosis case information, obtains the conditional independence check results, and updates at least some of the directed edges in the directed acyclic graph structure based on the conditional independence check results. This not only enables dynamic updating of the directed acyclic graph structure in the Bayesian network model, but also ensures the accuracy of the directed acyclic graph structure update.

[0084] Please see Figure 5 The vehicle fault diagnosis method provided in this application embodiment may include, but is not limited to, the following stages: Phase 1: Construction of the Fault Knowledge Graph; In practice, a fault knowledge graph can be built based on materials such as vehicle FMEA, FTA, and failure case library.

[0085] Phase Two: Fault Knowledge Graph Storage and Retrieval; In practice, the fault knowledge graph can be stored in a Neo4J graph database or a MySQL relational database. Then, based on the first fault phenomenon of the vehicle, corresponding query statements can be written to retrieve all target entities, entity relationships, and probability attributes that can be linked to the first fault phenomenon from the fault knowledge graph.

[0086] Phase 3: Fault diagnosis based on Bayesian network model; In practice, based on the entity information related to the first fault phenomenon obtained from the query, a target Bayesian network model corresponding to the first fault phenomenon can be constructed; then, based on the target Bayesian network model, the target failure mode that caused the first fault phenomenon can be determined.

[0087] Phase Four: Fault Knowledge Graph Update; In practice, fault diagnosis case information can be collected. Then, based on this information, the first data of the target Bayesian network model can be updated. Subsequently, based on the updated first data in the target Bayesian network model, the second data in the fault knowledge graph stored in stage two can be updated accordingly. For subsequent vehicle malfunctions, the latest fault knowledge graph can be used to query related entity information, thus further improving the accuracy of vehicle fault diagnosis.

[0088] The specific implementation process of this embodiment can be found in the relevant descriptions in the above embodiments, and will not be repeated here.

[0089] The vehicle fault diagnosis method provided in this application constructs a structured fault knowledge graph, enabling the visualization and efficient retrieval of vehicle fault knowledge, thus providing a clear knowledge foundation for fault diagnosis. Based on this, in response to specific fault phenomena, relevant entities, entity relationships, and probability attributes are extracted from the fault knowledge graph, and a corresponding target Bayesian network model is automatically constructed. This integrates the structured advantages of the knowledge graph with the quantitative reasoning capabilities of the Bayesian network. Using this Bayesian network model, intelligent fault diagnosis and analysis can be performed, accurately identifying the target failure modes that cause the fault phenomena. Furthermore, the automatic update mechanism of the model data enables the synchronous and dynamic maintenance of data in the fault knowledge graph. Therefore, this application not only effectively overcomes the shortcomings of pure knowledge graphs in quantifying the probability of fault occurrence and supporting probabilistic analysis scenarios, but also solves the problem that Bayesian networks are difficult to directly construct in the vehicle domain due to the scarcity of fault data and the difficulty in obtaining probability data.

[0090] Please see Figure 6 This application also provides a vehicle fault diagnosis device 600, which can implement the above-mentioned vehicle fault diagnosis method. The device 600 may include an entity query module 601, a model construction module 602, and a failure mode determination module 603.

[0091] The entity query module 601 can be used to query multiple target entities, the entity relationships between multiple target entities, and the probability attributes associated with the target entities in the fault knowledge graph in response to the occurrence of the first fault phenomenon in the vehicle; each target entity is an entity related to the first fault phenomenon; at least some of the target entities are failure modes. The model building module 602 can be used to build a target Bayesian network model corresponding to the first fault phenomenon based on multiple target entities, the entity relationships between multiple target entities, and the probability attributes associated with the target entities. The failure mode determination module 603 can be used to determine the target failure mode based on the target Bayesian network model; the target failure mode is used to indicate the cause of failure that leads to the occurrence of the first failure phenomenon.

[0092] The vehicle fault diagnosis device provided in this application embodiment can implement all the steps of the above-described vehicle fault diagnosis method embodiment and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0093] This application also provides a vehicle fault diagnosis device, including a processor and a memory. The memory stores a program or instructions that can be run on the processor. When the program or instructions are executed by the processor, they implement the various steps of the above-described vehicle fault diagnosis method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0094] Figure 7 To illustrate the hardware structure of the vehicle fault diagnosis device according to this application embodiment, the vehicle fault diagnosis device includes: The processor 701 can be implemented using a general-purpose central processing unit (CPU), microprocessor, application specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 702 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 702 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 702 and is called and executed by the processor 701 using the vehicle fault diagnosis method of the embodiments of this application. The input / output interface 703 is used to implement information input and output; The communication interface 704 is used to enable communication and interaction between this vehicle fault diagnosis equipment and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 705 transmits information between various components of the vehicle troubleshooting equipment (such as processor 701, memory 702, input / output interface 703, and communication interface 704); The processor 701, memory 702, input / output interface 703, and communication interface 704 are connected to each other via bus 705 within the vehicle fault diagnosis equipment.

[0095] The vehicle fault diagnosis device provided in this application embodiment can implement all the steps of the above-described vehicle fault diagnosis method embodiment and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0096] This application also provides a computer-readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various steps of the above-described vehicle fault diagnosis method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0097] The processor is the same as the processor in the vehicle fault diagnosis device described in the above embodiment. The computer-readable storage medium can be implemented in the form of a computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc.

[0098] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled. The processor is used to run programs or instructions to implement the various steps of the above-described vehicle fault diagnosis method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0099] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0100] This application provides a computer program product, which is stored in a storage medium and executed by at least one processor to implement the various steps of the vehicle fault diagnosis method embodiment described above, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0101] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not delete other identical elements present in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0102] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0103] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A vehicle troubleshooting method characterized by, include: In response to the occurrence of the first fault phenomenon in the vehicle, multiple target entities, entity relationships between the multiple target entities, and probability attributes associated with the target entities are obtained by querying the fault knowledge graph. Each of the target entities is an entity related to the first fault phenomenon; at least some of the target entities are failure modes; The number of failure modes is multiple; Based on the multiple target entities, the entity relationships between the multiple target entities, and the probability attributes associated with the target entities, a target Bayesian network model corresponding to the first fault phenomenon is constructed. The directed acyclic graph structure of the target Bayesian network model includes a virtual fault phenomenon total node, and the nodes of multiple failure modes are respectively directed to the fault phenomenon total node with directed edges. The conditional probability table of the target Bayesian network model is determined based on the single-point failure prior hypothesis, which is used to characterize that the fault phenomenon corresponding to the fault phenomenon total node occurs only when any one of the failure modes occurs alone. Based on the target Bayesian network model, the target failure mode is determined in the failure modes; The target failure mode is used to indicate the cause of the failure that leads to the first failure phenomenon.

2. The method of claim 1, wherein, The probability attributes include the prior probability of the target entity and the conditional probability of the entity relationship associated with the target entity; The step of constructing a target Bayesian network model corresponding to the first fault phenomenon based on the multiple target entities, the entity relationships between the multiple target entities, and the probability attributes associated with the target entities includes: Based on the multiple target entities and the entity relationships between the multiple target entities, a first directed acyclic graph structure is constructed; the first directed acyclic graph structure includes multiple nodes and directed edges between the multiple nodes; the nodes are used to represent the target entities; the directed edges are used to represent the entity relationships; A first conditional probability table is determined based on the prior probability of the target entity and the conditional probability of the entity relationship associated with the target entity. The target Bayesian network model is generated based on the first directed acyclic graph structure and the first conditional probability table.

3. The method of claim 1, wherein, The number of failure modes is multiple; the step of determining the target failure mode among the failure modes based on the target Bayesian network model includes: For each failure mode, based on the target Bayesian network model, the posterior probability of the failure mode occurring when the first failure phenomenon occurs is inferred; Based on the posterior probability of each failure mode, the failure mode with the highest posterior probability is selected from multiple failure modes. The failure mode with the highest posterior probability is determined as the target failure mode.

4. The method of claim 3, wherein, The plurality of target entities also include detection means, and a second fault phenomenon that is directly related to the detection means in the entity relationship; The step of inferring the posterior probability of the failure mode occurring under the condition that the first fault phenomenon occurs, based on the target Bayesian network model, includes: Based on the target detection result corresponding to the detection method, a first evidence variable value is determined; the first evidence variable value is used to characterize the target detection result indicating that the second fault phenomenon has occurred. Based on the first evidence variable value and the second evidence variable value, the target evidence variable value is determined; the second evidence variable value is used to characterize the occurrence of the first fault phenomenon. Based on the target Bayesian network model, infer the posterior probability of the failure mode occurring given that the target evidence variable value is true.

5. The method of claim 4, wherein, The step of determining the target evidence variable value based on the first evidence variable value and the second evidence variable value includes: Based on the target Bayesian network model, infer the posterior probability of the second fault phenomenon occurring when the value of the first evidence variable is true; In response to the posterior probability of the second fault phenomenon occurring being within a preset probability range, the target evidence variable value is determined based on the first evidence variable value, the second evidence variable value, and the third evidence variable value; the third evidence variable value is used to characterize the occurrence of the second fault phenomenon.

6. The method of claim 1, wherein, After determining the target failure mode from the failure modes based on the target Bayesian network model, the method further includes: Obtain the target variable values ​​of the plurality of target entities; the target variable values ​​are the variable values ​​observed and / or inferred during the process of determining the target failure mode based on the target Bayesian network model. Update the fault diagnosis case information based on the target variable values ​​of the multiple target entities; Based on the fault diagnosis case information, update the first data of the target Bayesian network model; Based on the updated first data, update the second data in the fault knowledge graph that corresponds to the first data.

7. The method of claim 6, wherein, The fault diagnosis case information contains hidden variables; these hidden variables are used to characterize entities whose target variable values ​​are missing. The step of updating the first data of the target Bayesian network model based on the fault diagnosis case information includes: Based on the directed acyclic graph structure and conditional probability table of the target Bayesian network model, the expected value of the posterior probability of the latent variable is determined. Based on the expected value of the posterior probability of the latent variable, update at least some of the probability values ​​in the conditional probability table of the target Bayesian network model. The first data includes at least a portion of the probability values.

8. The method of claim 6, wherein, The first data includes at least some of the directed edges in the directed acyclic graph structure of the target Bayesian network model; The step of updating the first data of the target Bayesian network model based on the fault diagnosis case information includes: Based on the fault diagnosis case information, conditional independence is tested for each pair of nodes in the directed acyclic graph structure to obtain the conditional independence test results; each pair of nodes consists of any two nodes in the directed acyclic graph structure. Based on the conditional independence test results, at least some of the directed edges in the directed acyclic graph structure are updated.

9. A vehicle troubleshooting device, characterized by, The device includes an entity query module, a model building module, and a failure mode determination module; The entity query module is used to respond to the first fault phenomenon of the vehicle by querying the fault knowledge graph to obtain multiple target entities, the entity relationships between the multiple target entities, and the probability attributes associated with the target entities. Each of the target entities is an entity related to the first fault phenomenon; at least some of the target entities are failure modes; The number of failure modes is multiple; The model building module is used to construct a target Bayesian network model corresponding to the first fault phenomenon based on the multiple target entities, the entity relationships between the multiple target entities, and the probability attributes associated with the target entities. The directed acyclic graph structure of the target Bayesian network model includes a virtual fault phenomenon total node, and the nodes of multiple failure modes are respectively directed to the fault phenomenon total node with directed edges. The conditional probability table of the target Bayesian network model is determined based on the single-point failure prior hypothesis, which is used to characterize that the fault phenomenon corresponding to the fault phenomenon total node occurs only when any one of the failure modes occurs alone. The failure mode determination module is used to determine the target failure mode based on the target Bayesian network model; The target failure mode is used to indicate the cause of the failure that leads to the first failure phenomenon.

10. A vehicle troubleshooting device, characterized by, The device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the vehicle fault diagnosis method as described in any one of claims 1 to 8.

11. A computer-readable storage medium storing a computer program, wherein the computer program comprises the following steps of: receiving a request for a resource from a client; determining whether the client is authorized to access the resource; and if the client is authorized to access the resource, providing the resource to the client. When the computer program is executed by the processor, it implements the vehicle fault diagnosis method as described in any one of claims 1 to 8.

Images