Automobile fault diagnosis method and system

Abstract

The invention discloses an automobile fault diagnosis method and system, and belongs to the technical field of fault diagnosis, and the system comprises an automobile fault identification model, an automobile fault knowledge graph model and a reasoning diagnosis integration model. S100, acquiring a natural language description text of the automobile fault, preprocessing the natural language description text of the automobile fault, and extracting a description entity related to the automobile fault; s200, constructing an automobile fault identification model, inputting the description entity related to the automobile fault into the automobile fault identification model, and outputting a fault type; s300, constructing an automobile fault knowledge graph model, inputting a description entity and a fault type related to the automobile fault into the automobile fault knowledge graph model in combination with Cypher to perform semantic similarity matching, and querying a matched answer; and S400, constructing a reasoning and diagnosis integrated model, inputting the description entity and the answer related to the automobile fault into the reasoning and diagnosis integrated model in combination with prompt, and outputting a fault analysis result and a visual correlation analysis result.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to an automobile fault diagnosis method and a system thereof. Background Art

[0002] At present, with the continuous growth of the number of cars in modern society, cars have become an indispensable means of transportation in people's daily life and economic activities. However, with the accumulation of car mileage and the increase of usage time, various mechanical parts and electronic devices may wear, age or even fail. These failures will not only affect the normal use of the car, but also bring potential safety hazards and endanger the lives of drivers and passengers. Therefore, it is particularly important to diagnose car failures in a timely and accurate manner.

[0003] Related technologies can refer to traditional automobile fault diagnosis, which mainly relies on the experience of technicians, the reading results of detection equipment and the fault codes that come with the vehicle to identify the problem.

[0004] With respect to the above-mentioned related technologies, the applicant has found that these methods have certain limitations, and therefore there is an urgent need for a method to solve the problem of low efficiency and insufficient accuracy of traditional diagnostic methods when processing complex data. Summary of the invention

[0005] The present invention aims at the deficiencies in the prior art and provides an automobile fault diagnosis method and system thereof.

[0006] The technical solution of the present invention to solve the above technical problems is as follows:

[0007] An automobile fault diagnosis method, comprising an automobile fault identification model, an automobile fault knowledge graph model and a reasoning diagnosis integrated model;

[0008] S100: obtaining a natural language description text of the automobile failure, preprocessing the natural language description text of the automobile failure, and extracting description entities related to the automobile failure;

[0009] S200: constructing an automobile fault identification model, inputting description entities related to automobile faults into the automobile fault identification model, and outputting the fault type;

[0010] S300: Build a knowledge graph model for automobile faults, combine the description entities and fault types related to automobile faults with Cypher and input them into the knowledge graph model for semantic similarity matching, and query matching answers;

[0011] S400: Build a reasoning diagnosis integration model, combine the description entities and answers related to the automobile fault with prompt input into the reasoning diagnosis integration model, and output the fault analysis results and visual association analysis results.

[0012] Further, in S100, the natural language description text of the car failure is information about the vehicle problem expressed in the user's daily language, and the natural language description text of the car failure includes the symptoms, specific manifestations and scenarios of the failure.

[0013] Furthermore, preprocessing includes data cleaning, data format conversion, special tag addition and data encoding, and the NER model is used to extract the descriptive entities related to automobile failures.

[0014] Furthermore, in S200, a deep learning model is used to construct a vehicle fault identification model, and the description entities related to the vehicle fault are input into the vehicle fault identification model, and combined with historical fault data, the possible fault types of the current vehicle are predicted.

[0015] Furthermore, the automobile fault identification model uses a supervised learning method to train the model using a large number of labeled automobile fault-related descriptive entities to ensure that the model can accurately identify the fault type from the automobile fault-related descriptive entities. The automobile fault identification model outputs the fault type and its corresponding credibility rate.

[0016] Furthermore, in S300, based on the Neo4j database, an automobile fault knowledge graph model is constructed through an entity relationship diagram, in which the nodes in the graph represent various automobile parts, fault types and related technical parameter entities, and the edges represent the relationships between entities.

[0017] Furthermore, the automobile fault knowledge graph model query language Cypher is used to input the fault description entity and fault type for semantic similarity matching, and the most similar fault solution is queried through the automobile fault knowledge graph model. The automobile fault knowledge graph model queries the matching results that are closest to the user description, including possible causes of the fault and solution results.

[0018] Furthermore, in S400, the automobile fault description entity, the automobile fault knowledge graph model query result and the prompt are used as input, and a reasoning diagnosis integration model is constructed based on the generative language model, and the reasoning diagnosis integration model generates a more detailed fault analysis result.

[0019] Furthermore, after the reasoning diagnosis integrated model outputs the fault analysis results, it will visualize the analysis results and use charts to show the relationship between the faulty components.

[0020] An automobile fault diagnosis system comprises an information retrieval module, a fault reasoning diagnosis module, a data integration module and a human-computer interaction module;

[0021] The information retrieval module is used to process the natural language description text of the automobile fault input by the user, quickly and accurately retrieve the information related to the automobile fault, and output the fault type, fault cause analysis, troubleshooting strategy, maintenance suggestions and daily knowledge;

[0022] The fault reasoning diagnosis module combines the information output by the information retrieval module to generate detailed diagnostic conclusions and maintenance suggestions;

[0023] Data integration module: The data integration module is responsible for the secure storage of fault data, historical record management, and integration of user feedback information, providing data support for the continuous optimization of the diagnosis model;

[0024] The human-computer interaction module is responsible for providing an intuitive interactive interface, presenting diagnostic results and related data to users in an easy-to-understand visual form to enhance user experience.

[0025] In summary, compared with the prior art, the above technical solution has the following beneficial effects:

[0026] (1) As an emerging technical means, data retrieval methods can accurately locate the cause of faults by quickly retrieving and analyzing vehicle operation data, historical maintenance records and fault information databases. Compared with traditional fault diagnosis methods, data retrieval methods have higher processing efficiency and intelligence level, and can automatically analyze a large amount of complex fault data and provide more accurate fault diagnosis results. At the same time, data retrieval can be combined with big data and artificial intelligence technology to improve the efficiency, accuracy and reliability of diagnosis through mining and modeling of massive data;

[0027] (2) By adopting data retrieval technology, the cause of the fault can be located more quickly and accurately, improving the efficiency of automobile maintenance, thereby providing users with a more convenient service experience;

[0028] (3) Obtain and preprocess the automobile fault description text provided by the user through natural language processing, and use the NER model to extract the description entities related to the fault, build an automobile fault identification model, use deep learning technology to input the extracted description entities into the model, combine historical fault data to predict the possible fault type and its credibility of the vehicle, build an automobile fault knowledge graph model, use Cypher language to match the semantic similarity of the description entities and fault types, query the most similar fault solutions, build an inference diagnosis integration model, further analyze the fault through the generative language model, and visualize the results;

[0029] (4) Effective fault diagnosis can not only avoid serious faults and reduce maintenance costs, but also improve the safety and reliability of the vehicle and extend its service life. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A schematic diagram of the steps of a method for diagnosing automobile faults according to the present invention;

[0031] Figure 2 A schematic diagram of a flow chart of a method for diagnosing automobile faults according to the present invention;

[0032] Figure 3 A schematic diagram of an automobile fault identification model of an automobile fault diagnosis method of the present invention;

[0033] Figure 4 A schematic diagram of an automobile fault knowledge graph model of an automobile fault diagnosis method of the present invention;

[0034] Figure 5 A schematic diagram of a fault tree knowledge base of a method for diagnosing automobile faults according to the present invention;

[0035] Figure 6 The figure is a schematic diagram of an automobile fault diagnosis system according to the present invention. DETAILED DESCRIPTION

[0036] The principles and features of the present invention are described below in conjunction with all the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0037] In traditional automobile fault diagnosis, the identification of problems mainly relies on the experience of technicians, the reading results of detection equipment, and the fault codes that come with the vehicle. However, these methods have certain limitations, such as incomplete fault code coverage, diagnostic results relying on the experience level of technicians, and low diagnostic efficiency. With the complexity of automotive electronic systems and the increase in the number of sensors, the sources of fault information are becoming more diverse, and traditional methods are difficult to fully and efficiently process and analyze a large amount of fault data.

[0038] As an emerging technical means, data retrieval methods can accurately locate the cause of faults by quickly retrieving and analyzing vehicle operation data, historical maintenance records and fault information databases. Compared with traditional fault diagnosis methods, data retrieval methods have higher processing efficiency and intelligence level, and can automatically analyze a large amount of complex fault data to provide more accurate fault diagnosis results. At the same time, data retrieval can be combined with big data and artificial intelligence technology to improve the accuracy and reliability of diagnosis through mining and modeling of massive data, thus making up for the shortcomings of traditional methods.

[0039] The embodiment of the present invention discloses a method for diagnosing automobile faults. By adopting data retrieval technology, the cause of the fault can be located more quickly and accurately, and the efficiency of automobile maintenance can be improved, thereby providing users with a more convenient service experience.

[0040] Reference Figure 1 and Figure 2 , an automobile fault diagnosis method, including an automobile fault identification model, an automobile fault knowledge graph model and a reasoning diagnosis integrated model;

[0041] S100: obtaining a natural language description text of the automobile failure, preprocessing the natural language description text of the automobile failure, and extracting description entities related to the automobile failure;

[0042] In S100, the natural language description text of the vehicle failure is information about the vehicle problem expressed in the user's daily language, and the natural language description text of the vehicle failure includes the symptoms, specific manifestations and scenarios of the failure.

[0043] Preprocessing includes data cleaning, data format conversion, special tag addition and data encoding, and the NER model is used to extract descriptive entities related to automobile failures.

[0044] S200: constructing an automobile fault identification model, inputting description entities related to automobile faults into the automobile fault identification model, and outputting the fault type;

[0045] In S200, a deep learning model is used to build a vehicle fault identification model, and the description entities related to the vehicle fault are input into the vehicle fault identification model. Combined with historical fault data, the possible fault types of the current vehicle are predicted.

[0046] The automobile fault identification model uses a supervised learning method and uses a large number of labeled automobile fault-related descriptive entities to train the model, ensuring that the model can accurately identify the fault type from the automobile fault-related descriptive entities. The automobile fault identification model outputs the fault type and its corresponding credibility rate.

[0047] S300: Build a knowledge graph model for automobile faults, combine the description entities and fault types related to automobile faults with Cypher and input them into the knowledge graph model for semantic similarity matching, and query matching answers;

[0048] In S300, based on the Neo4j database, an automobile fault knowledge graph model is constructed through an entity relationship diagram. The nodes in the graph represent various automobile parts, fault types and related technical parameter entities, and the edges represent the relationships between entities.

[0049] Using the automobile fault knowledge graph model query language Cypher, the fault description entity and fault type are input for semantic similarity matching, and the most similar fault solution is queried through the automobile fault knowledge graph model. The automobile fault knowledge graph model queries the matching results that are closest to the user description, including possible causes of the fault, solutions, and other results.

[0050] S400: Build a reasoning diagnosis integration model, combine the description entities and answers related to the automobile fault with prompt input into the reasoning diagnosis integration model, and output the fault analysis results and visual association analysis results.

[0051] In S400, the automobile fault description entity, the automobile fault knowledge graph model query result and prompt are used as input, and a reasoning diagnosis integration model is constructed based on the generative language model. The reasoning diagnosis integration model generates more detailed fault analysis results through prompt.

[0052] After the reasoning diagnosis integrated model outputs the fault analysis results, it will visualize the analysis results and use charts to show the relationship between faulty components.

[0053] In this embodiment, the method for processing the natural language description text of the automobile fault and extracting the fault description entity is as follows:

[0054] The natural language description text of the car failure is the information about the vehicle problem expressed in the user's daily language; the natural language description text of the car failure includes but is not limited to the symptoms, specific manifestations and scenarios of the failure; the example input is: "My hybrid car model has an EV function limitation problem, the vehicle loses power, the throttle does not respond, and the restart does not respond. What is the problem?";

[0055] Preprocessing includes but is not limited to data cleaning, data format conversion, special mark addition and data encoding;

[0056] Data cleaning includes but is not limited to removing special marks, punctuation marks, HTML tags, etc. in the natural language description text of the car fault to ensure the cleanliness and consistency of the data; Data cleaning example: "My hybrid model has a problem with limited EV function. The vehicle loses power and the throttle does not respond to restarting. What is the problem?"

[0057] Data format conversion is to convert the cleaned data into json format; data conversion format example: "{"Questions": ["role": "user", "content": "My hybrid car has EV function limitation problem. The vehicle loses power and the throttle does not respond. The restart does not respond. What is the problem?"]}";

[0058] Special tag addition is to add corresponding tags at the beginning and end of the input natural language description text of the automobile fault and the input special nodes, including but not limited to special token tags such as [gMASK], sop, <|system|>, <|assistant|>, [CLS] and [SEP]. Examples of special tag addition: "[gMASK] My hybrid car model has a problem with limited EV function. The vehicle loses power, the throttle does not respond, and the restart does not respond. What is the problem? [gMASK]", "<|system|> My hybrid car model has a problem with limited EV function. The vehicle loses power, the throttle does not respond, and the restart does not respond. What is the problem? <|assistant|>" and "[CLS] My hybrid car model has a problem with limited EV function. The vehicle loses power, the throttle does not respond, and the restart does not respond. What is the problem? [SEP];

[0059] Data encoding is to convert the natural language description text of the car failure after adding special tags into an encoding mode, convert the text into a word or subword sequence through word segmentation operation, and then use the word embedding vector representation operation to convert the word or subword sequence into a vector representation, and combine it with position encoding for operation; word segmentation operation example: "[gMASK], my, certain, hybrid model, has, EV, function, limited, problem, vehicle, lost, power, throttle, no, response, restart, no, response, what, is the problem, [gMASK]", "<|system|>, my, certain, hybrid model, has, EV, function, limited, problem , vehicle, lost, power, throttle, no, response, restart, no, response, what, is, the, problem, <|assistant|>” and “[CLS], my, certain, hybrid model, had, EV, function, limited, problem, vehicle, lost, power, throttle, no, response, restart, no, response, what, is, the, problem, [SEP]”; word embedding vector represents operation example: “[v1_gMASK], [v2_my], [v3_certain], [v4_hybrid model], [v5_appeared], [v6_EV], [v7_function], [v8_limited], [v9_problem], [v 10 _vehicle], [v 11 _lost], [v 12 _Power], [v 13 _throttle], [v 14 _none], [v 15 _reaction], [v 16 _Restart], [v 17 _none], [v 18 _reaction], [v 19 _This], [v 20 _ is], [v 21 _what], [v 22_problem], [v 23 _gMASK]", "[v1_system], [v2_my], [v3_some], [v4_hybrid model], [v5_appear], [v6_EV], [v7_function], [v8_restricted], [v9_problem], [v 10 _vehicle], [v 11 _lost], [v 12 _Power], [v 13 _throttle], [v 14 _none], [v 15 _reaction], [v 16 _Restart], [v 17 _none], [v 18 _reaction], [v 19 _This], [v 20 _ is], [v 21 _what], [v 22 _problem], [v 23 _system]" and "[v1_CLS], [v2_my], [v3_some], [v4_hybrid], [v5_appear], [v6_EV], [v7_function], [v8_restricted], [v9_problem], [v 10 _vehicle], [v 11 _lost], [v 12 _Power], [v 13 _throttle], [v 14 _none], [v 15 _reaction], [v 16 _Restart], [v 17 _none], [v 18 _reaction], [v 19 _This], [v 20 _ is], [v 21 _what], [v 22 _problem], [v 23 _SEP]"; where v i represents the embedding vector;

[0060] Use the NER model to extract description entities related to automobile failures;

[0061] Collect more natural language description texts of automobile failures, manually annotate description entities, and build a description entity training dataset; build a pre-trained description entity extraction model based on existing NER models (including but not limited to BERT, SpaCy, or deep learning models trained based on data with custom labels); use the entity training dataset to train the pre-trained description entity extraction model to obtain a customized description entity extraction model to better extract relevant information;

[0062] Descriptive entity extraction example: Input the natural language description text of the car failure after data cleaning to the descriptive entity extraction model: "My hybrid car model has a problem with limited EV function. The vehicle has lost power, the throttle has no response, and the restart has no response. What is the problem?"; Use the descriptive entity extraction model to recognize the natural language description text of the car failure after data cleaning, and output example: {″entities″: [{″text″:″A hybrid car model″}, {″text″:″limited EV function″}, {″text″:″loss of power″}, {″text″:″no throttle response″}, {″text″:″no restart response″}]}.

[0063] In this embodiment, the automobile fault identification model is constructed using deep learning technology as follows:

[0064] Use deep learning models to build a car fault identification model; input the natural language description text of the car fault after the description entity and data encoding related to the car fault into the car fault identification model, and combine it with historical fault data to predict the possible fault type of the current vehicle;

[0065] A pre-trained automobile fault identification model is constructed based on the BERT model. The pre-trained automobile fault identification model is trained through a supervised learning method using a large number of labeled automobile fault-related description entities and data-encoded natural language description texts of automobile faults to ensure that the model can accurately identify the fault type from the automobile fault-related description entities and data-encoded natural language description texts of automobile faults. The pre-trained model is adjusted through supervised fine-tuning operations to make it more suitable for automobile fault identification tasks.

[0066] In the classification output stage of the pre-trained automobile fault identification model, a manual feedback mechanism is introduced to further adjust the accuracy of the model; automobile fault identification experts provide feedback based on the model output results to guide the model to adjust; automobile fault identification experts annotate the identification results of the pre-trained automobile fault identification model and point out incorrect or inaccurate classifications; the manual feedback data is integrated into the model training for further fine-tuning, and the data set is updated to correct the model's deviation on specific fault types to optimize the model performance; the model adjusts its strategy based on manual feedback to strengthen its classification performance in different scenarios;

[0067] The BERT model consists of three main components: input layer, encoding layer and output layer. The input layer is obtained by adding three feature embedding vector representations, including token embedding, paragraph embedding and position embedding. The token embedding vector representation of each word or subword sequence in the input sequence is obtained by using Tokenizer, and segment encoding vector representation Segment Embedding operation and position encoding vector representation Position Embedding operation are performed. The encoding layer is built based on the Transformer encoder. The encoding layer can retain long-distance information similar to RNN, and can also perform parallel calculations similar to CNN. The output layer is built based on the mask language model. The output layer will mask about 15% of the words and predict these masked words, and learn text knowledge through training in a form similar to cloze.

[0068] Inputting the description entities related to the automobile fault and the natural language description text of the automobile fault after data encoding into the automobile fault identification model; performing association processing on the description entities related to the automobile fault and the natural language description text of the automobile fault after data encoding to obtain the final input;

[0069] Input example:

[0070] An example of a natural language description of a car fault after data encoding: "[v1_CLS], [v2_my], [v3_some], [v4_hybrid model], [v5_appeared], [v6_EV], [v7_function], [v8_restricted], [v9_problem], [v 10 _vehicle], [v 11 _lost], [v 12 _Power], [v 13 _throttle], [v 14 _none], [v 15 _reaction], [v 16 _Restart], [v 17 _none], [v 18 _reaction], [v 19 _This], [v 20 _ is], [v 21 _what], [v 22 _problem], [v 23 _SEP]";

[0071] Examples of description entities related to automobile failures: {″entities″: [{″text″:″A hybrid model″}, {″text″:″EV function is limited″}, {″text″:″Loss of power″}, {″text″:″No response to throttle″}, {″text″:″No response to restart″}]};

[0072] The final input example obtained after the association process: [v1_CLS], [v3_certain], [v4_hybrid vehicle], [v6_EV], [v7_function], [v8_restricted], [v 11 _lost], [v 12 _Power], [v 13 _throttle], [v 14 _none], [v 15 _reaction], [v 16 _Restart], [v 17 _none], [v 18 _reaction], [v 23 _SEP];

[0073] Reference Figure 3 The embodiment of the present invention provides a fault classification model for automobile fault diagnosis, and the classification structure is as follows:

[0074] [CLS] is an aggregated representation of the input. After being processed by a multi-layer Transformer encoder, the final vector of the [CLS] tag captures the overall semantic information of the input. On top of the BERT model, an additional composite layer consisting of a linear layer, an activation layer, and a Dropout layer is designed, and the output dimension of this layer is exactly equal to the number of categories. The composite layer uses the final vector of the [CLS] tag as input and outputs the fault type and its corresponding credibility rate.

[0075] This embodiment introduces the construction and application of the automobile fault knowledge graph model based on the Neo4j database as follows:

[0076] Based on the Neo4j database, the automobile fault knowledge graph model is constructed through the entity relationship graph. The nodes in the graph represent various automobile parts, fault types and related technical parameters, and the edges represent the relationships between entities.

[0077] Reference Figure 4 The embodiment of the present invention provides an automobile fault knowledge graph model structure for automobile fault diagnosis, including:

[0078] Extract automobile failure-related terms, ontology concepts, ontology relationships and rule definitions from structured data, semi-structured data and unstructured data such as failure knowledge in the field of automobile failure; construct automobile failure knowledge graph ontology based on automobile failure-related terms, ontology concepts, ontology relationships and rule definitions; perform entity learning operations such as entity filling, entity disambiguation and coreference resolution on the automobile failure knowledge graph ontology, and finally obtain an automobile failure knowledge graph model;

[0079] Entity filling: Fill entities in the field of automobile failure, including but not limited to automobile parts and failure types, into the automobile failure knowledge graph ontology; entity disambiguation: handle different expressions of the same entity to ensure the uniqueness of entities in the automobile failure knowledge graph ontology; co-reference resolution: resolve different expressions of the same concept in different texts to ensure knowledge consistency;

[0080] Specifically:

[0081] Fault knowledge in the field of automobile failure includes but is not limited to structured data, semi-structured data and unstructured data such as web page knowledge, design knowledge, principle knowledge and experience knowledge; extract structured data including but not limited to brand car series knowledge, fault cause knowledge, fault maintenance knowledge, fault suggestion knowledge and expert knowledge from the fault knowledge in the field of automobile failure of structured data, semi-structured data and unstructured data; perform knowledge modeling based on Neo4j database to build but not limited to automobile type knowledge base, fault tree knowledge base, fault maintenance knowledge base and expert suggestion knowledge base;

[0082] Reference Figure 5 , the embodiment of the present invention provides a fault tree knowledge base structure for automobile fault diagnosis;

[0083] The car type knowledge base is the first knowledge base, covering all car types on the market;

[0084] The fault tree knowledge base is the second knowledge base; the fault tree knowledge base uses the car type in the car type knowledge base as the first search subject and the fault category as the second search subject to conduct a branch-and-leaf search for the cause of the fault; the fault category includes but is not limited to power battery and power management faults, suspension system faults and steering system faults; under the fault category, there are small fault categories scattered in a branch-and-leaf manner, including but not limited to power battery pack faults, power failure during driving, suspension deformation, suspension breakage, suspension rust, steering system shaking, steering system abnormal noise and steering system deviation;

[0085] The fault maintenance knowledge base is the third knowledge base, which provides maintenance suggestions for various faults; the expert suggestion knowledge base is the fourth knowledge base, which collects the suggestions and experience of experts in the field of automobile faults; the faults determined by the fault tree knowledge base are used as the search subject, combined with the fault maintenance knowledge base and the expert suggestion knowledge base, to obtain cause analysis and maintenance suggestions;

[0086] Using Cypher, a query language for automobile fault knowledge graph models, input fault description entities and fault types for semantic similarity matching, and query the most similar fault solutions through the automobile fault knowledge graph model; achieving effective application of automobile fault knowledge, especially in fault cause analysis, troubleshooting strategies, maintenance suggestions, and provision of daily knowledge;

[0087] The automobile fault knowledge graph model queries the closest matching result to the user's description, including possible causes of the fault, output results, etc.

[0088] This embodiment introduces the construction and application of a reasoning diagnosis integrated model based on a generative language model;

[0089] Build an integrated reasoning diagnosis model based on a generative language model; take the automobile fault description entity, automobile fault type, automobile fault knowledge graph model query results and prompt as input, and learn the ability to generate fault analysis reports from the input through pre-training and fine-tuning; use prompt as a guide to generate structured and detailed fault analysis content through the prompt model; the design of prompt must fully combine the input data characteristics to ensure the accuracy and readability of the output results;

[0090] After generating detailed fault analysis results, the reasoning diagnosis integrated model uses data visualization technology to display the fault analysis results in the form of a chart, making the complex fault relationship intuitive and clear; extract key fault components and the relationship information between them from the analysis results; based on the causal relationship between the fault components, construct the data structure required for the chart, including nodes (components) and connections (relationships);

[0091] Use frameworks such as Flask and echarts to visualize complex relationship graphs.

[0092] Reference Figure 6 , an automobile fault diagnosis system, including an information retrieval module, a fault reasoning diagnosis module, a data integration module and a human-computer interaction module;

[0093] The information retrieval module is used to process the natural language description text of the automobile fault input by the user, quickly and accurately retrieve the information related to the automobile fault, and output the fault type, fault cause analysis, troubleshooting strategy, maintenance suggestions and daily knowledge;

[0094] The fault reasoning diagnosis module combines the information output by the information retrieval module to generate detailed diagnostic conclusions and maintenance suggestions;

[0095] Data integration module: The data integration module is responsible for the secure storage of fault data, historical record management, and integration of user feedback information, providing data support for the continuous optimization of the diagnosis model;

[0096] The human-computer interaction module is responsible for providing an intuitive interactive interface, presenting diagnostic results and related data to users in an easy-to-understand visual form to enhance user experience.

[0097] The implementation principle of an automobile fault diagnosis method and system according to an embodiment of the present invention is as follows:

[0098] Through natural language processing, the automobile fault description text provided by the user is obtained and preprocessed, and the NER model is used to extract the fault-related descriptive entities, and an automobile fault identification model is built. The extracted descriptive entities are input into the model using deep learning technology, and the possible fault types and their credibility of the vehicle are predicted in combination with historical fault data. A knowledge graph model of automobile faults is established, and the Cypher language is used to match the semantic similarity of the descriptive entities and the fault types, and the most similar fault solutions are queried; an inference diagnosis integrated model is built, and the fault is further analyzed through a generative language model, and the results are visualized.

[0099] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for diagnosing automobile faults, characterized in that: Including automobile fault identification model, automobile fault knowledge graph model and reasoning diagnosis integrated model; S100: obtaining a natural language description text of the automobile failure, preprocessing the natural language description text of the automobile failure, and extracting description entities related to the automobile failure; S200: constructing an automobile fault identification model, inputting description entities related to automobile faults into the automobile fault identification model, and outputting the fault type; S300: Build a knowledge graph model for automobile faults, combine the description entities and fault types related to automobile faults with Cypher and input them into the knowledge graph model for semantic similarity matching, and query matching answers; S400: Build a reasoning diagnosis integration model, combine the description entities and answers related to the automobile fault with prompt input into the reasoning diagnosis integration model, and output the fault analysis results and visual association analysis results.

2. The method for diagnosing automobile faults according to claim 1, characterized in that: In S100, the natural language description text of the vehicle failure is information about the vehicle problem expressed in the user's daily language, and the natural language description text of the vehicle failure includes the symptoms, specific manifestations and scenarios of the failure.

3. A method for diagnosing automobile faults according to claim 2, characterized in that: Preprocessing includes data cleaning, data format conversion, special tag addition and data encoding, and the NER model is used to extract descriptive entities related to automobile failures.

4. The method for diagnosing automobile faults according to claim 1, characterized in that: In S200, a deep learning model is used to build a vehicle fault identification model, and the description entities related to the vehicle fault are input into the vehicle fault identification model. Combined with historical fault data, the possible fault types of the current vehicle are predicted.

5. The method for diagnosing automobile faults according to claim 4, characterized in that: The automobile fault identification model uses a supervised learning method and uses a large number of labeled automobile fault-related descriptive entities to train the model, ensuring that the model can accurately identify the fault type from the automobile fault-related descriptive entities. The automobile fault identification model outputs the fault type and its corresponding credibility rate.

6. The automobile fault diagnosis method according to claim 1, characterized in that: In S300, based on the Neo4j database, an automobile fault knowledge graph model is constructed through an entity relationship diagram. The nodes in the graph represent various automobile parts, fault types and related technical parameter entities, and the edges represent the relationships between entities.

7. The automobile fault diagnosis method according to claim 6, characterized in that: Using the automobile fault knowledge graph model query language Cypher, the fault description entity and fault type are input for semantic similarity matching, and the most similar fault solution is queried through the automobile fault knowledge graph model. The automobile fault knowledge graph model queries the matching results closest to the user description, including the possible cause of the fault and the solution result.

8. The automobile fault diagnosis method and system according to claim 1, characterized in that: In S400, the automobile fault description entity, the automobile fault knowledge graph model query result and prompt are used as input, and a reasoning diagnosis integration model is constructed based on the generative language model. The reasoning diagnosis integration model generates more detailed fault analysis results.

9. The automobile fault diagnosis method and system according to claim 8, characterized in that: After the reasoning diagnosis integrated model outputs the fault analysis results, it will visualize the analysis results and use charts to show the relationship between faulty components.

10. An automobile fault diagnosis system, characterized in that: It includes information retrieval module, fault reasoning diagnosis module, data integration module and human-computer interaction module; The information retrieval module is used to process the natural language description text of the automobile fault input by the user, quickly and accurately retrieve the information related to the automobile fault, and output the fault type, fault cause analysis, troubleshooting strategy, maintenance suggestions and daily knowledge; The fault reasoning diagnosis module combines the information output by the information retrieval module to generate detailed diagnostic conclusions and maintenance suggestions; Data integration module: The data integration module is responsible for the secure storage of fault data, historical record management, and integration of user feedback information, providing data support for the continuous optimization of the diagnosis model; The human-computer interaction module is responsible for providing an intuitive interactive interface, presenting diagnostic results and related data to users in an easy-to-understand visual form to enhance user experience.

Images