A method and device for updating automobile fault detection model and predicting fault

By extracting the features of the original data of the car and training the Bayesian detection network, the problems of fault detection misjudgment and misjudgment in the Internet of Vehicles data training model are solved, and the accuracy and comprehensiveness of vehicle fault detection are improved.

CN118034733BActive Publication Date: 2024-11-15PISTON INTELLIGENCE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410139247.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-01-31
Publication Date
2024-11-15
Estimated Expiration
2044-01-31

AI Technical Summary

Technical Problem

Automobile fault detection models trained based on Internet of Vehicles data are prone to misjudgment and misjudgment, resulting in reduced detection accuracy and stability.

Method used

By obtaining the original data of the car, extracting characteristic data, including vehicle age data, Internet of Vehicles data and expert experience, using survival analysis models and classification algorithms to obtain the survival functions and impact variables of the fault, training Bayesian detection network, and updating the model to reduce errors The probability of judgment and missed judgment is improved to improve the accuracy of fault detection.

Benefits of technology

Real-time update of the car fault detection model is achieved, which improves the accuracy and comprehensiveness of detection, reduces the probability of misjudgment and misjudgment, and improves the accuracy of fault type identification.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118034733B_ABST
    Figure CN118034733B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and device for updating a vehicle fault detection model and predicting faults. Extract the first fault influencing variable that affects the fault from the original data of the vehicle and the survival function corresponding to each fault; when the original data of the vehicle exceeds a preset ratio, obtain the second fault influencing variable that affects the fault and the last layer of neurons of the second neural network by training the first neural network and the second neural network respectively; update the Bayesian detection network for the major fault category and several Bayesian detection networks for specific faults in the fault detection model according to the first fault influencing variable, the second fault influencing variable, the last layer of neurons of the second neural network, the survival function and the vehicle age data in the original data of the vehicle, and obtain the updated vehicle fault detection model. Through this application, the misjudgment rate and missed judgment rate of the vehicle fault detection model can be reduced, and the accuracy of vehicle fault detection can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method and device for updating an automobile fault detection model and predicting faults. Background Art

[0002] With the popularization of intelligent vehicle networking, more and more cars have Internet of Vehicles data, which makes it possible to train models based on Internet of Vehicles data and use it to predict or diagnose vehicle failures. Using Internet of Vehicles data to diagnose or predict failures has positive significance in ensuring vehicle safety, reducing maintenance costs, and improving user experience.

[0003] However, there are two main problems in predicting and diagnosing automobile failures based on the training model of Internet of Vehicles data: 1. Using only the Internet of Vehicles data of the target component to predict or diagnose the occurrence of a fault can easily lead to a high rate of misjudgment; 2. Since some automobile failures do not directly affect the use, they will not be discovered immediately when they occur or will only be discovered when other components have problems, resulting in missed faults, which in turn causes label errors in the training set, and ultimately reduces the accuracy and stability of the model. Summary of the invention

[0004] The present invention provides a method and device for updating an automobile fault detection model and predicting faults, so as to solve the technical problem of how to reduce the misjudgment and missed judgment of automobile faults by the Internet of Vehicles data training model, thereby improving the accuracy of automobile fault detection and prediction.

[0005] In order to solve the above technical problems, in a first aspect, an embodiment of the present invention provides a method for updating a vehicle fault detection model, comprising:

[0006] Obtaining original vehicle data and extracting characteristic data of the original vehicle data; wherein the original vehicle data includes: vehicle age data, vehicle networking data, vehicle fault information and expert experience; the characteristic data includes: a survival function of each vehicle fault obtained by a survival analysis model algorithm and a first fault influencing variable of each vehicle fault extracted by a classification algorithm;

[0007] When the acquired original automobile data exceeds a preset ratio, the Internet of Vehicles data is used as input, and the first neural network corresponding to each Bayesian detection network in the automobile fault detection model is trained in sequence according to the first preset data set, and the second fault influencing variable corresponding to the Bayesian detection network is screened according to each of the first neural networks; the automobile fault detection model includes: a fault category Bayesian detection network for detecting a fault category and a plurality of specific fault Bayesian detection networks corresponding to each fault category;

[0008] According to the second fault influencing variable and the second preset data set of each Bayesian detection network, train the second neural network corresponding to each Bayesian detection network in turn, and obtain the last layer of neurons of the second neural network;

[0009] According to the survival function, the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network and the second fault influencing variable corresponding to each Bayesian detection network, each Bayesian detection network is updated to output an updated vehicle fault detection model.

[0010] Compared with the prior art, the above embodiment has the following beneficial effects: by determining whether to update the automobile fault detection model through the Internet of Vehicles data obtained in real time, the model used to detect automobile faults can be updated in time, thereby improving the accuracy of the model; in addition, by obtaining important variables that affect the output of each model when updating the model, the accuracy of judging whether the automobile will fail in the future is improved according to the predicted value of the important variable; finally, by updating multiple detection networks Bayesian detection networks, the type of automobile fault can be identified more finely, and at the same time, the Bayesian network can associate the data of multiple related components, and use the mutual influence between the operating states of different related components to improve the accuracy of the final fault judgment, and introduce the survival function as prior knowledge into the Bayesian network, and substitute the historical fault law into the fault judgment process, thereby reducing the probability of missed and misjudgment of automobile faults, and improving the accuracy of identifying and predicting the type of automobile faults.

[0011] In an embodiment of the first aspect, updating each Bayesian detection network according to the survival function, the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network corresponding to each Bayesian detection network and the second fault influencing variable in sequence includes:

[0012] updating the graph structure of each Bayesian detection network according to the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network and the second fault influencing variable corresponding to each Bayesian detection network;

[0013] The conditional probability distribution of each node in each Bayesian detection network is updated according to the survival function, expert experience and the uniform distribution of the second fault influencing variable corresponding to each Bayesian detection network in a value range.

[0014] Compared with the prior art, the above embodiment has the following beneficial effects: by updating the graph structure of the fault detection model according to all important variables, the model can better reflect the relationship between different variables and components and the ultimate impact on the vehicle failure, thereby improving the comprehensiveness and accuracy of the fault detection model; in addition, the survival function, expert experience and the uniform distribution of variables in the value domain are used as the probability distribution of the model nodes, which effectively evaluates and reflects the relationship between each variable and the fault, wherein the survival function also introduces the effective information contained in the massive historical data into the detection model, thereby improving the comprehensiveness and accuracy of the fault detection model.

[0015] In an embodiment of the first aspect, obtaining original vehicle data and extracting feature data of the original vehicle data includes:

[0016] Preprocessing the original automobile data, performing feature engineering processing on the preprocessed original automobile data, and obtaining preprocessing variables;

[0017] Extracting the first fault influencing variables corresponding to each of the automobile faults from the preprocessing variables by a classification algorithm, and cross-grouping all the first fault influencing variables of the same automobile fault to obtain a first fault influencing variable group set corresponding to each of the automobile faults;

[0018] By using a survival analysis algorithm, a functional relationship between the occurrence probability of each of the automobile failures and the vehicle age data and the first failure influencing variable group set is established to obtain the survival function of the automobile failure on each group set.

[0019] Compared with the prior art, the above embodiment has the following beneficial effects: by preprocessing the data, the influence of noise is reduced, which helps to extract a useful and important first variable; in addition, by cross-grouping the first fault influencing variables, it helps to find the influence of different variable value combinations on the fault probability; finally, by establishing a survival function, the problem of inaccurate fault probability judgment caused by data sparsity is solved, and the accuracy of judging the probability of fault occurrence is improved.

[0020] In an embodiment of the first aspect, according to each of the first neural networks, screening a second fault influencing variable corresponding to the Bayesian detection network includes:

[0021] Based on the neural network explanatory technology, calculating the feature contribution of the input variables of the first neural network;

[0022] According to the characteristic contribution of the input variables of the first neural network, the input variables of the first neural network are sorted from large to small, and a preset number of input variables of the first neural network or those that meet a threshold are selected in sequence as the second fault influencing variables of the corresponding Bayesian detection network.

[0023] Compared with the prior art, the above embodiment has the following beneficial effects: by analyzing the feature contribution of the trained neural network input variables, the variables in the input variables that have an important impact on the detection results can be more accurately reflected according to the obtained feature contribution, thereby avoiding the problem of poor robustness of the results caused by overfitting due to too many neural network variables, and avoiding the problem of too complex a structure caused by too many neurons in the last layer of the neural network used to design the Bayesian network graph structure, thereby improving the accuracy of the detection model.

[0024] In a second aspect, an embodiment of the present invention further provides a method for predicting automobile faults, comprising:

[0025] Acquire a vehicle fault detection model; wherein the vehicle fault detection model is updated according to the vehicle fault detection model updating method described in any one of the embodiments of the present application;

[0026] According to the record information of the second fault influencing variable of the Bayesian detection network for major fault categories in the automobile fault detection model, obtaining a next time period prediction value of the second fault influencing variable of the Bayesian detection network for major fault categories;

[0027] According to the next time period prediction value of the second fault influencing variable of the fault major category Bayesian detection network, first prediction data is obtained through the fault major category Bayesian detection network; the first prediction data includes: whether a fault occurs in the next time period and the major fault category when the fault occurs;

[0028] When it is predicted that a fault will occur in the next time period, according to the major fault category, a corresponding specific fault Bayesian detection network in the automobile fault detection model is selected, and according to the record information of the second fault influencing variable of the selected specific fault Bayesian detection network, a next time period prediction value of the second fault influencing variable of the specific fault Bayesian detection network is obtained;

[0029] According to the next time period prediction value of the second fault influencing variable of the specific fault Bayesian detection network, the specific type and probability of the fault occurring in the next time period of the vehicle are obtained through the specific fault Bayesian detection network.

[0030] Compared with the prior art, the above embodiment has the following beneficial effects: by predicting the values ​​of important variables and combining with the first-layer fault detection model, it is determined whether a fault will occur in the future and the major type of fault that will occur; according to the major type of fault that has occurred, the corresponding second-layer fault detection model is selected to obtain the specific type of fault that will occur in the future, thereby solving the problem of instability and inaccuracy caused by too many types of faults being put in one model, and improving the accuracy of the automobile fault detection model.

[0031] In an embodiment of the second aspect, obtaining a predicted value of the second fault influencing variable according to recorded information of the second fault influencing variable includes:

[0032] According to a preset value range and a current value of the second fault influencing variable, obtaining a value label of the second fault influencing variable in a current period;

[0033] According to the regression algorithm, combining the value label of the second fault influencing variable in the current period and all values ​​of the second fault influencing variable in the previous period, obtaining the value label of the second fault influencing variable in the next period;

[0034] An average value of the values ​​of the second fault influencing variable in other time periods having the same value label as the value of the second fault influencing variable in the next time period is used as the predicted value of the second fault influencing variable in the next time period.

[0035] Compared with the prior art, the above embodiment has the following beneficial effects: using labels instead of values ​​can capture nonlinear effects, and by determining the current labels of important variables that affect fault detection results and combining all previous values ​​of all important variables, the accuracy of predicting the variable values ​​in the next time period is improved, thereby effectively reducing the volatility of the prediction and improving the accuracy of the prediction.

[0036] In a third aspect, an embodiment of the present invention further provides a vehicle fault detection model updating device, comprising: a feature data extraction module, a first neural network training module, a second neural network training module and a Bayesian detection network updating module;

[0037] The feature data extraction module is used to obtain original vehicle data and extract feature data of the original vehicle data; the original vehicle data includes: vehicle age data, vehicle networking data, vehicle fault information and expert experience; the feature data includes: the survival function of each vehicle fault obtained by the survival analysis model algorithm and the first fault influencing variable of each vehicle fault extracted by the classification algorithm;

[0038] The first neural network training module is used to take the Internet of Vehicles data as input when the acquired original automobile data exceeds a preset ratio, and train the first neural network corresponding to each Bayesian detection network in the automobile fault detection model in sequence according to the first preset data set, and screen the second fault influencing variable corresponding to the Bayesian detection network according to each first neural network; the automobile fault detection model includes: a Bayesian detection network for a large fault category and a plurality of Bayesian detection networks for specific faults;

[0039] The second neural network training module is used to train the second neural network corresponding to each Bayesian detection network according to the second fault influencing variable and the second preset data set of each Bayesian detection network in turn, and obtain the last layer of neurons of the second neural network;

[0040] The Bayesian detection network update module is used to update each Bayesian detection network according to the survival function, the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network of each Bayesian detection network and the second fault influencing variable, and output an updated automobile fault detection model.

[0041] In an embodiment of the third aspect, the Bayesian detection network update model includes: a network structure update unit and a probability distribution update unit;

[0042] The network structure updating unit is used to update the graph structure of each Bayesian detection network according to the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network corresponding to each Bayesian detection network and the second fault influencing variable in sequence;

[0043] The probability distribution updating unit is used to update the conditional probability distribution of each node in each Bayesian detection network according to the survival function, expert experience and the uniform distribution of the second fault influencing variable corresponding to each Bayesian detection network in a value domain.

[0044] In an embodiment of the third aspect, the feature extraction module includes: a data preprocessing module unit, a first fault influencing variable extraction unit and a survival function acquisition unit;

[0045] The data preprocessing module is used to preprocess the original automobile data, perform feature engineering on the preprocessed original automobile data, and obtain preprocessing variables;

[0046] The first fault influencing variable extraction unit is used to extract the first fault influencing variable corresponding to each of the vehicle faults from the preprocessing variables through a classification algorithm, and cross-group all the first fault influencing variables of the same vehicle fault to obtain a first fault influencing variable group set corresponding to each of the vehicle faults;

[0047] The survival function acquisition unit is used to establish a functional relationship between the occurrence probability of each of the vehicle failures and the vehicle age data and the first failure influencing variable group set through a survival analysis algorithm, and to obtain the survival function of the vehicle failure on each group set.

[0048] In a fourth aspect, an embodiment of the present invention further provides an automobile fault prediction device, comprising: an automobile fault detection model acquisition module, a second fault influencing variable prediction module, a fault category detection module, and a specific fault detection module;

[0049] The vehicle fault detection model acquisition module is used to acquire a vehicle fault detection model; wherein the vehicle fault detection model is updated and acquired by the vehicle fault detection model updating device according to the embodiment of the present invention;

[0050] The second fault influencing variable prediction module is used to obtain the next time period prediction value of the second fault influencing variable of the fault major category Bayesian detection network according to the record information of the second fault influencing variable of the fault major category Bayesian detection network;

[0051] The fault category detection module is used to obtain first prediction data through the fault category Bayesian detection network according to the next time period prediction value of the second fault influencing variable of the fault category Bayesian detection network in the automobile fault detection model; the first prediction data includes: whether a fault will occur in the next time period and the fault category when the fault occurs;

[0052] The second fault influencing variable prediction module is further used to select a corresponding specific fault Bayesian detection network in the automobile fault detection model according to the major fault category when predicting that a fault will occur in the next time period, and obtain a next time period prediction value of the second fault influencing variable of the specific fault Bayesian detection network according to the record information of the second fault influencing variable of the selected specific fault Bayesian detection network;

[0053] The specific fault detection module is used to obtain the specific type and probability of the vehicle's fault occurring in the next time period through the specific fault Bayesian detection network according to the next time period prediction value of the second fault influencing variable of the specific fault Bayesian detection network.

[0054] In the fifth aspect, an embodiment of the present invention further provides a terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the above-mentioned method for updating a vehicle fault detection model, or the above-mentioned method for predicting a vehicle fault.

[0055] In the sixth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned method for updating a vehicle fault detection model, or the above-mentioned method for predicting a vehicle fault. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A schematic flow chart of an embodiment of a method for updating a vehicle fault detection model provided by the present invention;

[0057] Figure 2 A schematic diagram of a flow chart of an embodiment of a method for predicting automobile faults provided by the present invention;

[0058] Figure 3 A schematic diagram of the structure of an embodiment of a vehicle fault detection model updating device provided by the present invention;

[0059] Figure 4 A schematic structural diagram of an embodiment of a vehicle fault prediction device provided by the present invention. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] Embodiment 1

[0062] Please refer to Figure 1 , a method for updating a vehicle fault detection model provided by an embodiment of the present invention, includes S101 to S104, specifically:

[0063] S101: Acquire original vehicle data and extract characteristic data of the original vehicle data; wherein the original vehicle data includes: vehicle age data, vehicle networking data, vehicle fault information and expert experience; the characteristic data includes: a survival function of each vehicle fault obtained by a survival analysis model algorithm and a first fault influencing variable of each vehicle fault extracted by a classification algorithm;

[0064] In one embodiment, the step of extracting the characteristic data of the original automobile data specifically includes the following steps:

[0065] S1011: preprocessing the original automobile data, performing feature engineering processing on the preprocessed original automobile data, and obtaining preprocessing variables;

[0066] S1012: extracting the first fault influencing variables corresponding to each of the automobile faults from the preprocessing variables by a classification algorithm, and cross-grouping all the first fault influencing variables of the same automobile fault to obtain a first fault influencing variable group set corresponding to each of the automobile faults;

[0067] S1013: By using a survival analysis algorithm, a functional relationship between the occurrence probability of each of the vehicle failures and the vehicle age data and the first failure influencing variable group set is established to obtain the survival function of the vehicle failure on each group set.

[0068] Specifically, the feature engineering process in step S1011 may adopt the following method:

[0069] (1) Aggregate the original fields using different methods such as average, variance, median, and range in different time periods, and then transform them using moving average, difference, delay, and other methods in the time dimension;

[0070] (2) Use new car sales to estimate the market ownership in each region, and use the market ownership to standardize the parts failure data:

[0071] (a) Calculate the market holdings:

[0072]

[0073] Among them, t represents a certain period of time, age represents the age of the vehicle, and p i represents the scrapping rate of vehicles with age i;

[0074] (b) Use the market inventory to convert the number of failures into a failure rate:

[0075]

[0076] Among them, the number of faults j t,age Represents the number of times fault j occurs on a vehicle with age age at time t.

[0077] (3) Multiple fields combined to generate new variables

[0078]

[0079] Where t represents a time period, v represents a vehicle, and K represents the number of trips that exceed a certain length (such as 1 kilometer) within the time period.

[0080] Specifically, the method for extracting and grouping the first fault influencing variables in step S1012 may be:

[0081] (1) Segment the continuous variable so that the probability of component failure is different in different segments; for example, a decision tree is established for battery failure, identifying the impact of temperature on battery failure and dividing it into three temperature segments: low (<5 degrees), medium, and high (>35 degrees);

[0082] (2) Clustering discrete variables. Different categories have different impacts on part failures. For example, provinces are discrete variables and are divided into different regions such as Northeast and Northwest by the algorithm.

[0083] (3) All variables that affect the failure of the same part are crossed and grouped according to segments and categories; for example, the groups obtained in (2) are low temperature in the northeast and high temperature in the northwest. In order to prevent data sparseness due to too many groups, groups with little difference in part failure can also be merged to obtain the final grouping to ensure that the difference in part failure in each group is significant.

[0084] Specifically, the method for obtaining the survival function in step S1013 may be: according to any survival analysis model algorithm, obtaining the functional relationship between the probability of failure occurrence and the vehicle age and the first failure influencing variable group.

[0085] In order to fully explain the method of obtaining the survival function based on the survival analysis model algorithm, the following explanation is given using the Cox model as an example:

[0086] f j (age|Z j )=f0 j (age)*exp(β j ′ Z j )

[0087] Where j represents fault j, f0 j (age) is the original survival function, Z j represents the group of factors affecting fault j, β j ′ is the influence coefficient of each group; the original survival function can be calculated by Kaplan-Meier method, β j ′ It can be estimated by the maximum likelihood function.

[0088] Compared with the prior art, the above embodiment has the following beneficial effects: by preprocessing the data, the influence of noise is reduced, which helps to extract important variables that affect the fault; in addition, by cross-grouping the first fault influencing variables, it helps to find the influence of different variable value combinations on the fault probability; finally, by establishing a survival function, the problem of inaccurate fault probability judgment caused by data sparsity is solved, and the accuracy of judging the probability of fault occurrence is improved.

[0089] S102: when the acquired original data of the automobile exceeds a preset ratio, the Internet of Vehicles data is used as input, and according to a first preset data set, the first neural network corresponding to each Bayesian detection network in the automobile fault detection model is trained in sequence, and according to each of the first neural networks, the second fault influencing variable corresponding to the Bayesian detection network is screened; the automobile fault detection model includes: a fault category Bayesian detection network for detecting a fault category and a plurality of specific fault Bayesian detection networks corresponding to each fault category;

[0090] In one embodiment, the second fault influencing variable of the corresponding Bayesian detection network is screened according to each of the first neural networks, specifically: based on the neural network explanatory technology, the feature contribution of the input variables of the first neural network is calculated; according to the feature contribution of the input variables of the first neural network, the input variables of the first neural network are sorted from large to small, and a preset number of input variables of the first neural network or those that meet a threshold are selected in order as the second fault influencing variables of the corresponding Bayesian detection network.

[0091] Compared with the prior art, the above embodiment has the following beneficial effects: by analyzing the feature contribution of the trained neural network input variables, the variables in the input variables that have an important impact on the detection results can be more accurately reflected according to the obtained feature contribution, thereby avoiding the problem of poor robustness of the results caused by overfitting due to too many neural network variables, and avoiding the problem of too complex a structure caused by too many neurons in the last layer of the neural network used to design the Bayesian network graph structure, thereby improving the accuracy of the detection model.

[0092] In one embodiment, the data label of the preset data set is set according to the type of Bayesian detection network that needs to be trained, specifically including: dividing all faults into several major fault categories through the expert experience and the vehicle fault information; when training the first neural network corresponding to the major fault category Bayesian detection network, setting the data label of the data set according to the major fault category to which the fault belongs; when training the first neural network corresponding to the specific fault Bayesian detection network, setting the data label of the data set according to the specific fault that occurs.

[0093] Compared with the prior art, the above embodiment has the following beneficial effects: when training the fault detection model, the accuracy of the fault detection model training result is improved by introducing expert experience.

[0094] S103: training the second neural network corresponding to each of the Bayesian detection networks according to the second fault influencing variables and the second preset data set of each of the Bayesian detection networks in turn, and obtaining the last layer of neurons of the second neural network;

[0095] Specifically, in order to fully explain the training process of the neural network described in S102 and S103, the following scheme is used as an example for explanation:

[0096] (1) Based on expert experience or by experts, classify specific automobile faults into several major categories;

[0097] (2) Divide the dataset into training set, validation set, and test set according to the ratio of 6:2:2;

[0098] (3) Taking the IoV data of each vehicle at a certain time granularity and the probability of occurrence of each type of fault obtained according to the survival function as input;

[0099] (4) When the Bayesian detection network corresponding to the neural network is a fault category Bayesian detection network, the fault category that occurred is used as the label of the data set data; when the Bayesian detection network corresponding to the neural network is a specific fault Bayesian detection network, the specific fault that occurred is used as the label of the data set data;

[0100] (5) training a first neural network, analyzing the feature contribution of input variables based on feature importance ranking or SHAP value, and selecting a first fault influencing variable;

[0101] (6) Based on the selected first fault influencing variable, train the second neural network to obtain the last layer of neurons of the second neural network.

[0102] S104: Update each Bayesian detection network according to the survival function, the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network and the second fault influencing variable corresponding to each Bayesian detection network, and output an updated vehicle fault detection model.

[0103] In one embodiment, the Bayesian detection network updating method can be implemented by the following steps:

[0104] S1041: updating the graph structure of each of the fault Bayesian detection networks according to the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network corresponding to each of the Bayesian detection networks and the second fault influencing variable in sequence;

[0105] S1042: updating the conditional probability distribution of each node in each Bayesian detection network according to the survival function, expert experience, and the uniform distribution of the second fault influencing variable corresponding to each Bayesian detection network in a value range.

[0106] Specifically, in S1041, the graph structure updating step may be:

[0107] (1) updating the nodes in the Bayesian detection network based on the second fault influencing variable obtained by the neural network, the first fault influencing variable extracted by feature engineering, the neurons in the last layer of the second neural network, and the vehicle age;

[0108] (2) Use the Constraint-Based algorithm to add edges to the updated nodes;

[0109] (3) adding or deleting nodes in the Bayesian detection network, and adding or deleting edges between nodes based on expert experience or by experts;

[0110] (4) Finally, the structure of the Bayesian detection network is optimized based on the Bayesian Information Criterion (BIC) or Likelihood Estimate (LL).

[0111] Specifically, in S1042, the step of updating the conditional probability distribution of the Bayesian detection network node may be:

[0112] (1) Node prior distribution update: For the first fault influencing variable and vehicle age node extracted by feature engineering, the survival function obtained by survival analysis is used as the prior distribution; expert experience or uniform distribution within the variable value range is used as the prior distribution of other variables;

[0113] (2) Update the posterior distribution of the node: Based on the Bayesian formula, multiply the prior distribution by the likelihood function to obtain the conditional probability distribution under the given parent node condition.

[0114] Specifically, the vehicle fault detection model updating method from S101 to S104 can also be used to construct the vehicle fault detection model.

[0115] Embodiment 2

[0116] Please refer to Figure 2 , a method for predicting automobile faults provided by an embodiment of the present invention, includes S201 to S205, specifically:

[0117] S201: Acquire a vehicle fault detection model; wherein the vehicle fault detection model is updated according to any one of the vehicle fault detection model updating methods in the embodiments of the present invention;

[0118] S202: acquiring a next time period prediction value of the second fault influencing variable corresponding to the fault category Bayesian detection network in the automobile fault detection model according to the record information of the second fault influencing variable of the fault category Bayesian detection network;

[0119] S203: acquiring first prediction data through the fault category Bayesian detection network according to the next time period prediction value of the second fault influencing variable of the fault category Bayesian detection network; the first prediction data includes: whether a fault will occur in the next time period and the fault category when the fault occurs;

[0120] S204: when it is predicted that a fault will occur in the next time period, selecting a corresponding specific fault Bayesian detection network in the automobile fault detection model according to the major fault category, and obtaining a next time period prediction value of the second fault influencing variable of the specific fault Bayesian detection network according to the record information of the second fault influencing variable of the selected specific fault Bayesian detection network;

[0121] S205: According to the next time period prediction value of the second fault influencing variable of the specific fault Bayesian detection network, the specific type and probability of the fault occurring in the next time period of the vehicle are obtained through the specific fault Bayesian detection network.

[0122] Compared with the existing basis, the above embodiment has the following beneficial effects: by predicting the values ​​of important variables and combining with the first-level fault detection model, it is determined whether a fault will occur in the future and the major type of fault that will occur; according to the major type of fault that has occurred, the corresponding second fault detection model is selected to obtain the specific type of fault that will occur in the future, thereby solving the problem of instability and inaccuracy caused by too many types of faults being put in one model, and improving the accuracy of the automobile fault detection model.

[0123] In one embodiment, in S202 and S204, the predicted value of the second fault influencing variable is obtained based on the recorded information of the second fault influencing variable, specifically including: obtaining the value label of the second fault influencing variable in the current time period based on a preset value range and the current value of the second fault influencing variable; obtaining the value label of the second fault influencing variable in the next time period based on a regression algorithm, combined with the value label of the second fault influencing variable in the current time period and all values ​​of the second fault influencing variable in previous time periods; and taking the average value of the second fault influencing variable in other time periods with the same value label as the second fault influencing variable in the next time period as the predicted value of the second fault influencing variable in the next time period.

[0124] Specifically, in order to further illustrate the specific process of the above embodiment, the following scheme is used as an example:

[0125] (1) Based on the correlation between the value of the influencing variable and the fault, and combined with expert experience, the value of the second fault influencing variable is labeled as normal, slightly abnormal, or severely abnormal; where normal refers to the normal value range, slightly abnormal refers to the value that has a certain predictive effect on the fault, and severely abnormal refers to the value that has a significant predictive effect on the fault.

[0126] (2) Set t1, t2, …, t n-1 The value of the second fault affecting the variable in the time period and t n The labels of the time periods are used as training data sets;

[0127] (3) Based on multi-class logistic regression or Softmax regression with autoregressive structure, predict the n+1 What are the probabilities of the variable being normal, slightly abnormal, and severely abnormal?

[0128] (4) When the probability of mild abnormality or severe abnormality is greater than the normal probability and exceeds the threshold, the prediction t n+1 The time period value is the average of all time period values ​​of its corresponding label.

[0129] Compared with the prior art, the above embodiment has the following beneficial effects: using labels instead of values ​​can capture nonlinear effects, and by determining the current labels of important variables that affect fault detection results and combining all previous values ​​of all important variables, the accuracy of predicting the variable values ​​in the next time period is improved, thereby effectively reducing the volatility of the prediction and improving the accuracy of the prediction.

[0130] Specifically, the automobile fault detection model can also be used to detect whether the automobile has a fault in the current time period and the specific type of the fault. The specific method is: input the value of the second fault influencing variable in the current time period into the fault category detection model to determine whether the automobile has a fault in the current time period. If the automobile has a fault in the current time period, select the corresponding specific fault Bayesian detection network according to the major fault category, and input the value of the second fault influencing variable in the current time period of the corresponding specific fault detection Bayesian network to obtain the specific fault type of the automobile in the current time period.

[0131] Specifically, the automobile fault detection model can also be used to simulate and analyze the relationship between the probability of fault occurrence and variables. The specific method is: experts input the numerical values ​​of the variables to be analyzed into the fault detection model, and based on the results output by the automobile fault detection model, evaluate the relationship between the probability of fault occurrence and the variables.

[0132] Embodiment 3

[0133] See also Figure 3 , Figure 3A schematic diagram of the structure of an embodiment of the vehicle fault detection model updating device provided by the present invention is shown in FIG. Figure 3 As shown, the device includes a feature data extraction module 301, a first neural network training module 302, a second neural network training module 303 and a Bayesian detection network update module 304; wherein the feature data extraction module 301 also includes a data preprocessing module 3011, a first fault influencing variable extraction unit 3012 and a survival function acquisition unit 3013; the Bayesian detection network update module 304 also includes a network structure update unit 3041 and a probability distribution update unit 3042, specifically:

[0134] In one embodiment, the feature data extraction module 301 is used to obtain original automobile data and extract feature data of the original automobile data; wherein the original automobile data includes: vehicle age data, Internet of Vehicles data, vehicle fault information and expert experience; the feature data includes: the first fault influencing variable of each automobile fault extracted by the classification algorithm and the survival function of each automobile fault.

[0135] In one embodiment, the data preprocessing module 3011 is used to preprocess the original automobile data, perform feature engineering processing on the preprocessed original automobile data, and obtain preprocessing variables.

[0136] In one embodiment, the first fault influencing variable extraction unit 3012 is used to extract the first fault influencing variable corresponding to each of the vehicle faults in the preprocessing variables through a classification algorithm, and cross-group all the first fault influencing variables of the same type of vehicle fault to obtain a set of first fault influencing variable groups corresponding to each of the vehicle faults.

[0137] In one embodiment, the survival function acquisition unit 3013 is used to establish a model for each of the automobile faults through a survival analysis algorithm, fit the functional relationship between its occurrence probability and the vehicle age data and its first fault influencing variable group set, and obtain the survival function of the automobile fault in each group set.

[0138] In one embodiment, the first neural network training module 302 is used to use the Internet of Vehicles data as input when the acquired original data of the vehicle exceeds a preset ratio, and train the first neural network corresponding to each Bayesian detection network in the vehicle fault detection model in sequence according to the first preset data set, and screen the second fault influencing variable of its Bayesian detection network according to each first neural network; the vehicle fault detection model includes: a Bayesian detection network for a large fault category and a number of Bayesian detection networks for specific faults.

[0139] In one embodiment, the second neural network training module 303 is used to train the second neural network corresponding to each Bayesian detection network in turn according to the second fault influencing variable and the second preset data set of each Bayesian detection network, and obtain the last layer of neurons of the second neural network.

[0140] In one embodiment, the first neural network training module 302 is further used to set the data label of the preset data set according to the type of Bayesian detection network to be trained, specifically: all faults are divided into several major fault categories through the expert experience and the vehicle fault information; when training the first neural network corresponding to the major fault category Bayesian detection network, the data label of the data set is set according to the major fault category to which the fault belongs; when training the first neural network corresponding to the specific fault Bayesian detection network, the data label of the data set is set according to the specific fault that occurs.

[0141] In one embodiment, the first neural network training module 302 is used to screen the second fault influencing variables of the corresponding Bayesian detection network according to each of the first neural networks, specifically: based on the neural network explanatory technology, calculate the feature contribution of the input variables of the first neural network; according to the feature contribution of the input variables of the first neural network, sort the input variables of the first neural network from large to small, and select a preset number or input variables of the first neural network that meet the threshold in order as the second fault influencing variables of the corresponding Bayesian detection network.

[0142] In one embodiment, the Bayesian detection network update module 304 is used to update each Bayesian detection network in turn according to the survival function, the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network corresponding to each Bayesian detection network and the second fault influencing variable, and output an updated vehicle fault detection model.

[0143] In one embodiment, the network structure updating unit 3041 is used to update the graph structure of each Bayesian detection network based on the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network of each Bayesian detection network and the second fault influencing variable.

[0144] In one embodiment, the probability distribution updating unit 3042 is used to update the conditional probability distribution of each node in each Bayesian detection network according to the survival function, expert experience and the uniform distribution of the second fault influencing variable of each Bayesian detection network within the value domain.

[0145] Embodiment 4

[0146] See also Figure 4 , Figure 4 A schematic diagram of the structure of an embodiment of the vehicle fault prediction device provided by the present invention is shown as follows: Figure 4 As shown, the automobile fault prediction device includes an automobile fault detection model acquisition module 401, a second fault influencing variable prediction module 402, a fault category detection module 403 and a specific fault detection module 404, which are specifically:

[0147] In one embodiment, the automobile fault detection model acquisition module 401 is used to acquire an automobile fault detection model; wherein the automobile fault detection model is updated and obtained according to the automobile fault detection model updating device described in any one of the above embodiments.

[0148] In one embodiment, the second fault influencing variable prediction module 402 is used to obtain the next time period prediction value of the second fault influencing variable of the fault major category Bayesian detection network according to the record information of the second fault influencing variable of the fault major category Bayesian detection network.

[0149] In one embodiment, the fault category detection module 403 is used to obtain first prediction data through the fault category Bayesian detection network according to the next time period prediction value of the second fault influencing variable of the fault category Bayesian detection network in the automobile fault detection model; the first prediction data includes: whether a fault will occur in the next time period and the fault category when the fault occurs.

[0150] In one embodiment, the second fault influencing variable prediction module 402 is also used to select a corresponding specific fault Bayesian detection network in the automobile fault detection model according to the major fault category when predicting that a fault will occur in the next time period, and obtain the next time period prediction value of the second fault influencing variable of the specific fault Bayesian detection network according to the record information of the second fault influencing variable of the selected specific fault Bayesian detection network.

[0151] In one embodiment, the specific fault detection module 404 is used to obtain the specific type and probability of the vehicle fault occurring in the next time period through the specific fault Bayesian detection network according to the next time period prediction value of the second fault influencing variable of the specific fault Bayesian detection network.

[0152] In one embodiment, the second fault influencing variable prediction module 402 is used to obtain the predicted value of the second fault influencing variable based on the recorded information of the second fault influencing variable, specifically: according to a preset value range and the current value of the second fault influencing variable, obtain the value label of the second fault influencing variable in the current time period; according to the regression algorithm, combined with the value label of the second fault influencing variable in the current time period and all values ​​of the second fault influencing variable in previous time periods, obtain the value label of the second fault influencing variable in the next time period; and use the average value of the second fault influencing variable in other time periods with the same value label as the second fault influencing variable in the next time period as the predicted value of the second fault influencing variable in the next time period.

[0153] Embodiment 5

[0154] Based on the above-mentioned embodiments of the automobile fault detection model updating method and the automobile fault prediction method, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the automobile fault detection model updating method or the automobile fault prediction method of any embodiment of the present invention is implemented.

[0155] Exemplarily, in this embodiment, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the terminal device.

[0156] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0157] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device. The memory may be used to store the computer program and / or module, and the processor implements various functions of the terminal device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store data created according to the use of the mobile phone, etc. In addition, the memory may include a high-speed random access memory and may also include a non-volatile memory, such as a hard disk, an internal memory, a plug-in hard disk, a smart memory card (SMC), a secure digital (SD) card, a flash memory card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0158] Embodiment 6

[0159] Based on the above-mentioned embodiments of the automobile fault detection model updating method and the automobile fault prediction method, another embodiment of the present invention provides a storage medium, wherein the storage medium includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute any one of the automobile fault detection model updating method or automobile fault prediction method of the embodiments of the present invention.

[0160] In this embodiment, the storage medium is a computer-readable storage medium, and the computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0161] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for updating a vehicle fault detection model, It is characterized in that include: Obtaining original vehicle data and extracting characteristic data of the original vehicle data; wherein the original vehicle data includes: vehicle age data, vehicle networking data, vehicle fault information and expert experience; the characteristic data includes: a survival function of each vehicle fault obtained by a survival analysis model algorithm and a first fault influencing variable of each vehicle fault extracted by a classification algorithm; When the acquired original automobile data exceeds a preset ratio, the Internet of Vehicles data is used as input, and the first neural network corresponding to each Bayesian detection network in the automobile fault detection model is trained in sequence according to the first preset data set, and the second fault influencing variable corresponding to the Bayesian detection network is screened according to each of the first neural networks; the automobile fault detection model includes: a fault category Bayesian detection network for detecting a fault category and a plurality of specific fault Bayesian detection networks corresponding to each fault category; According to the second fault influencing variable and the second preset data set of each Bayesian detection network, train the second neural network corresponding to each Bayesian detection network in turn, and obtain the last layer of neurons of the second neural network; According to the survival function, the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network and the second fault influencing variable corresponding to each Bayesian detection network, the graph structure of each Bayesian detection network and the conditional probability distribution of each node are updated, and an updated vehicle fault detection model is output; The method for updating the graph structure of the Bayesian detection network includes: updating the nodes of the Bayesian detection network according to the first fault influencing variable, the vehicle age data, the last layer of neurons of the second neural network corresponding to each Bayesian detection network, and the second fault influencing variable; By using a Constraint-Based algorithm, adding edges to the nodes of the updated Bayesian detection network, and adding or deleting nodes in the Bayesian detection network, and adding or deleting edges between nodes according to the expert experience; Optimizing the graph structure of the Bayesian detection network based on the Bayesian information criterion or likelihood estimation; The updating method of the conditional probability distribution comprises: For the first fault influencing variable and the node corresponding to the vehicle age data, the survival function is used as a prior distribution, and the expert experience or the uniform distribution within the variable value range is used as the prior distribution of other variables; Based on the Bayesian formula, the prior distribution is multiplied by the likelihood function to obtain the conditional probability distribution under the given parent node condition.

2. A method for updating a vehicle fault detection model as claimed in claim 1, It is characterized in that The updating of each Bayesian detection network according to the survival function, the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network corresponding to each Bayesian detection network and the second fault influencing variable in sequence includes: updating the graph structure of each Bayesian detection network according to the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network and the second fault influencing variable corresponding to each Bayesian detection network; The conditional probability distribution of each node in each Bayesian detection network is updated according to the survival function, expert experience and the uniform distribution of the second fault influencing variable corresponding to each Bayesian detection network in a value range.

3. A method for updating a vehicle fault detection model as claimed in claim 1, It is characterized in that The step of acquiring original automobile data and extracting characteristic data of the original automobile data includes: Preprocessing the original automobile data, performing feature engineering processing on the preprocessed original automobile data, and obtaining preprocessing variables; Extracting the first fault influencing variables corresponding to each of the automobile faults from the preprocessing variables by a classification algorithm, and cross-grouping all the first fault influencing variables of the same automobile fault to obtain a first fault influencing variable group set corresponding to each of the automobile faults; By using a survival analysis algorithm, a functional relationship between the occurrence probability of each of the automobile failures and the vehicle age data and the first failure influencing variable group set is established to obtain the survival function of the automobile failure on each group set.

4. A method for updating a vehicle fault detection model as claimed in claim 1, It is characterized in that The step of screening the second fault influencing variable of the corresponding Bayesian detection network according to each of the first neural networks includes: Based on the neural network explanatory technology, calculating the feature contribution of the input variables of the first neural network; According to the characteristic contribution of the input variables of the first neural network, the input variables of the first neural network are sorted from large to small, and a preset number of input variables of the first neural network or those that meet a threshold are selected in sequence as the second fault influencing variables of the corresponding Bayesian detection network.

5. A method for predicting automobile failures, It is characterized in that include: Acquire a vehicle fault detection model; wherein the vehicle fault detection model is updated according to the vehicle fault detection model updating method according to any one of claims 1 to 4; According to the record information of the second fault influencing variable of the Bayesian detection network for major fault categories in the automobile fault detection model, obtaining a next time period prediction value of the second fault influencing variable of the Bayesian detection network for major fault categories; According to the next time period prediction value of the second fault influencing variable of the fault major category Bayesian detection network, first prediction data is obtained through the fault major category Bayesian detection network; the first prediction data includes: whether a fault occurs in the next time period and the major fault category when the fault occurs; When it is predicted that a fault will occur in the next time period, according to the major fault category, a corresponding specific fault Bayesian detection network in the automobile fault detection model is selected, and according to the record information of the second fault influencing variable of the selected specific fault Bayesian detection network, a next time period prediction value of the second fault influencing variable of the specific fault Bayesian detection network is obtained; According to the next time period prediction value of the second fault influencing variable of the specific fault Bayesian detection network, the specific type and probability of the fault occurring in the next time period of the vehicle are obtained through the specific fault Bayesian detection network.

6. A method for predicting automobile failures as claimed in claim 5, It is characterized in that Acquiring a predicted value of the second fault influencing variable according to the recorded information of the second fault influencing variable includes: According to a preset value range and a current value of the second fault influencing variable, obtaining a value label of the second fault influencing variable in a current period; According to the regression algorithm, combining the value label of the second fault influencing variable in the current period and all values ​​of the second fault influencing variable in the previous period, obtaining the value label of the second fault influencing variable in the next period; An average value of the values ​​of the second fault influencing variable in other time periods having the same value label as the value of the second fault influencing variable in the next time period is used as the predicted value of the second fault influencing variable in the next time period.

7. A vehicle fault detection model updating device, It is characterized in that The automobile fault detection model updating device comprises: a feature data extraction module, a first neural network training module, a second neural network training module and a detection network Bayesian detection network updating module; The feature data extraction module is used to obtain the original data of the vehicle and extract the feature data of the original data of the vehicle; wherein the original data of the vehicle includes: vehicle age data, vehicle networking data, vehicle fault information and expert experience; the feature data includes: the survival function of each vehicle fault obtained by the survival analysis model algorithm and the first fault influencing variable of each vehicle fault extracted by the classification algorithm; The first neural network training module is used to take the Internet of Vehicles data as input when the acquired original automobile data exceeds a preset ratio, and train the first neural network corresponding to each Bayesian detection network in the automobile fault detection model in sequence according to the first preset data set, and screen the second fault influencing variable corresponding to the Bayesian detection network according to each of the first neural networks; the automobile fault detection model includes: a fault category Bayesian detection network for detecting a fault category and a plurality of specific fault Bayesian detection networks corresponding to each fault category; The second neural network training module is used to train the second neural network corresponding to each Bayesian detection network according to the second fault influencing variable and the second preset data set of each Bayesian detection network in turn, and obtain the last layer of neurons of the second neural network; The Bayesian detection network update module is used to update the graph structure of each Bayesian detection network and the conditional probability distribution of each node according to the survival function, the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network corresponding to each Bayesian detection network and the second fault influencing variable, and output an updated vehicle fault detection model; The method for updating the graph structure of the Bayesian detection network includes: updating the nodes of the Bayesian detection network according to the first fault influencing variable, the vehicle age data, the last layer of neurons of the second neural network corresponding to each Bayesian detection network, and the second fault influencing variable; By using a Constraint-Based algorithm, adding edges to the nodes of the updated Bayesian detection network, and adding or deleting nodes in the Bayesian detection network, and adding or deleting edges between nodes according to the expert experience; Optimizing the graph structure of the Bayesian detection network based on the Bayesian information criterion or likelihood estimation; The updating method of the conditional probability distribution comprises: For the first fault influencing variable and the node corresponding to the vehicle age data, the survival function is used as a prior distribution, and the expert experience or the uniform distribution within the variable value range is used as the prior distribution of other variables; Based on the Bayesian formula, the prior distribution is multiplied by the likelihood function to obtain the conditional probability distribution under the given parent node condition.

8. A vehicle fault detection model updating device as claimed in claim 7, It is characterized in that The Bayesian detection network update model includes: a network structure update unit and a probability distribution update unit; The network structure updating unit is used to update the graph structure of each Bayesian detection network according to the expert experience, the first fault influencing variable, the vehicle age data, and the last layer of neurons of the second neural network corresponding to each Bayesian detection network and the second fault influencing variable in sequence; The probability distribution updating unit is used to update the conditional probability distribution of each node in each Bayesian detection network according to the survival function, expert experience and the uniform distribution of the second fault influencing variable corresponding to each Bayesian detection network in a value domain.

9. The vehicle fault detection model updating device according to claim 7, It is characterized in that The characteristic data extraction module includes: a data preprocessing module unit, a first fault influencing variable extraction unit and a survival function acquisition unit; The data preprocessing module is used to preprocess the original automobile data, perform feature engineering on the preprocessed original automobile data, and obtain preprocessing variables; The first fault influencing variable extraction unit is used to extract the first fault influencing variable corresponding to each of the vehicle faults from the preprocessing variables through a classification algorithm, and cross-group all the first fault influencing variables of the same vehicle fault to obtain a first fault influencing variable group set corresponding to each of the vehicle faults; The survival function acquisition unit is used to establish a functional relationship between the occurrence probability of each of the vehicle failures and the vehicle age data and the first failure influencing variable group set through a survival analysis algorithm, and to obtain the survival function of the vehicle failure on each group set.

10. A vehicle fault prediction device, It is characterized in that The automobile fault prediction device comprises: an automobile fault detection model acquisition module, a second fault influencing variable prediction module, a fault category detection module and a specific fault detection module; The vehicle fault detection model acquisition module is used to acquire a vehicle fault detection model; wherein the vehicle fault detection model is updated and obtained according to the vehicle fault detection model updating device according to any one of claims 7 to 9; The second fault influencing variable prediction module is used to obtain the next time period prediction value of the second fault influencing variable of the fault major category Bayesian detection network according to the record information of the second fault influencing variable of the fault major category Bayesian detection network; The fault category detection module is used to obtain first prediction data through the fault category Bayesian detection network according to the next time period prediction value of the second fault influencing variable of the fault category Bayesian detection network in the automobile fault detection model; the first prediction data includes: whether a fault will occur in the next time period and the fault category when the fault occurs; The second fault influencing variable prediction module is further used to select a corresponding specific fault Bayesian detection network in the automobile fault detection model according to the major fault category when predicting that a fault will occur in the next time period, and obtain a next time period prediction value of the second fault influencing variable of the specific fault Bayesian detection network according to the record information of the second fault influencing variable of the selected specific fault Bayesian detection network; The specific fault detection module is used to obtain the specific type and probability of the vehicle's fault occurring in the next time period through the specific fault Bayesian detection network according to the next time period prediction value of the second fault influencing variable of the specific fault Bayesian detection network.