CN117290484B

CN117290484B - Intelligent question-answering system and method for automobile fault positioning and maintenance suggestion

Info

Publication number: CN117290484B
Application number: CN202311308336.XA
Authority: CN
Inventors: 司徒俊豪; 张世恒; 张江波; 孙涛; 蔡鸿平; 吴浩驰
Original assignee: Shenzhen Mingrui Data Technology Co ltd
Current assignee: Shenzhen Mingrui Data Technology Co ltd
Priority date: 2023-10-10
Filing date: 2023-10-10
Publication date: 2024-06-18
Anticipated expiration: 2043-10-10
Also published as: CN117290484A

Abstract

The invention provides an intelligent question-answering system and method for automobile fault location and maintenance suggestion, wherein the system comprises the following components: the device comprises an information input module, a data searching module and a result output module; the method comprises the following steps: acquiring automobile identity information and fault description content input by a user in real time, and screening target characteristic information required by a model; searching data corresponding to the target feature information in the database, splicing the historical dialogue, the target feature information and a storage catalog of the database, and inputting the spliced data into a large language model; and the large language model processes the spliced content to obtain information such as fault reasons, maintenance schemes, maintenance time, cost and the like corresponding to the fault description content of the information input module. The invention realizes an intelligent fault diagnosis system by integrating the user input, the database and the large language model, provides a convenient, accurate and time-and cost-saving automobile fault solution, and has important significance for users and automobile industries.

Description

Intelligent question-answering system and method for automobile fault positioning and maintenance suggestion

Technical Field

The invention relates to the technical field of intelligent man-machine interaction, in particular to an intelligent question-answering system and method for automobile fault positioning and maintenance suggestion.

Background

A large language model (Large Language Model, abbreviated LLM), also known as a large language model, is an artificial intelligence model intended to understand and generate human language; training on a large amount of text data can perform a wide range of tasks including text summarization, translation, emotion analysis, and the like. Large language models are characterized by a large scale, containing billions of parameters, which help them learn complex patterns in language data. These models are typically based on deep learning architectures, such as translators, which help them to perform impressively on the task of various natural language processing. In the automotive industry, with the continuous development of technology and the increasing complexity of vehicles, there is an increasing demand for automobile repair, maintenance, and troubleshooting. When a user is faced with an automotive problem, the user often needs to review a large number of documents, manuals, or consult with a professional to obtain an accurate solution. Then, through network searching or inquiring to professional telephone, often, the obtained result is poor in accuracy due to inaccurate description, and an answer with higher accuracy cannot be provided for the user.

First, application number: CN202011093108.1 discloses a big data intelligent system for accurately diagnosing automobile faults, which comprises a fault point identification module and an automobile accessory determination module, and solves the problems that the existing diagnostic equipment technology depends on individual experience excessively by means of a coding mechanism, and lacks the capability of intelligent diagnosis and guiding supported by big data technology, so that the fault point and accessories are accurately hit while the correct judgment of machine automation on faults is difficult to realize, although the intelligent, prepositioned, online and remote data link service modes of automobile aftermarket service are realized; but is not suitable for the personalized service of the user, and the personalized user cannot be provided with the targeted service, so that the experience of the user is poor.

Second prior art, application number: CN201810743122.8 discloses a word vector based intelligent auxiliary diagnosis method and system for automobile faults, comprising: establishing a fault-fault correlation factor dictionary; acquiring fault description; establishing a replacement word list by using a word vector model, and carrying out keyword matching on fault description; obtaining a relevance score of the fault correlation factor corresponding to the fault by combining the fault-fault correlation factor dictionary, and calculating the score of the fault; and sequencing the scores of the faults to determine the faults. Although the word vector embedding technology is used, the processing capacity of natural language is greatly enhanced, the effective utilization of the past automobile maintenance communication records is realized, and fragmented natural language data is changed into available numerical data; however, the operation is performed by a professional in a 4S store, and the autonomous inquiry of the fault by the user cannot be realized, so that the application range is narrow, and the learning ability of the model is not fully exerted.

Third, application number: CN202211687840.0 discloses an intelligent question-answering system of automobile based on knowledge base, comprising: the knowledge base module stores local information aiming at automobile knowledge; the problem acquisition module acquires user voice to obtain user problem voice or user correction voice; the voice recognition module performs voice recognition on the user problem voice or the user correction voice to obtain user problem information or user correction information; the confirmation correction module feeds the user problem information back to the user for confirmation, and corrects the information according to the user correction information to obtain problem request information; the matching retrieval module matches in the knowledge base module according to the question request information to obtain answer information; the answer feedback module feeds answer information back to the user. Although the correction module is calibrated aiming at the problem information, the accuracy of the intelligent question-answering system of the automobile is improved, the satisfaction degree of the user is improved, the influence on the emotion of the user is avoided, and meanwhile, the driving danger caused by emotional driving is also avoided; however, the knowledge base module lacks the content of automobile fault positioning and maintenance suggestions, so that the results of question and answer are incomplete and the degree of intellectualization is low.

The system for detecting the automobile faults in the first, second and third prior art can not provide personalized service for users, and needs professional personnel to operate, so that the operability of the system is poor, and the user can not be helped to give out the problems of fault positioning and maintenance schemes according to fault descriptions.

Disclosure of Invention

In order to solve the technical problems, the invention provides an intelligent question-answering system for automobile fault location and maintenance suggestion, comprising:

the information input module is used for acquiring the automobile identity information and the fault description content input by a user in real time and screening target characteristic information required by the model; wherein, the car identity information contains: branding, train, purchase date and maintenance period, the fault description content comprises: failure phenomena, duration, and frequency of occurrence;

The data searching module is responsible for searching data corresponding to the target characteristic information in the database, splicing the history dialogue of the information input module, the target characteristic information and the storage catalog of the database, and inputting the information into the large language model; the database comprises knowledge maps of various automobile identity information, wherein the knowledge maps comprise specific faults, fault reasons, maintenance schemes, maintenance time and cost corresponding to the automobile identity information;

And the result output module is responsible for processing the spliced content by the large language model to obtain fault reasons, maintenance schemes, maintenance time and cost information corresponding to the fault description content of the information input module.

Optionally, the information input module includes:

The first information input sub-module is in charge of receiving the automobile identity information input by a user, screening out target characteristic information in the automobile identity information according to the requirement of a large language model, searching a corresponding database according to the screened out target characteristic information in the automobile identity information by the large language model, and calling a knowledge graph cache stored in the database;

The second information input sub-module is in charge of receiving fault description contents input by a user, screening out target feature information in the fault description contents according to the requirements of the large language model, and calling out a cached knowledge graph according to the target feature information in the screened out fault description contents by the large language model;

And the third information input sub-module is responsible for receiving new fault description contents input by a user, searching the database by the large language model according to the new fault description contents, and giving out prompts of the fault description contents by the large language model when the fault description contents of the user data cannot be identified by the large language model, and inputting by the user according to the prompts.

Optionally, the third information input sub-module includes:

The input detection unit is used for detecting whether a user has input of new fault description content, if so, triggering the large language model to execute a new search action program, and if not, continuously detecting whether the user has input of the new fault description content until the user has input of the new fault description content;

The mode establishing unit is in charge of establishing a new question-answer mode according to the new fault description content, and taking the new fault description content as feedback information expected by a characterization user as a starting point of a question-answer under the new question-answer mode; searching matched data in a database according to a key sub-of the new fault description content and acquiring an instruction for retrieving the data in the database according to the request of the new fault description content;

And the condition judging unit is in charge of judging whether the feedback information meets the preset condition according to the number of the matched data in the database, and outputting N pieces of feedback information aiming at the new fault description content M, wherein M and N are positive integers larger than 1, and N is smaller than or equal to M.

Optionally, the mode establishing unit includes:

The information updating subunit is responsible for establishing a new question-answer mode according to the request of the new fault description content and updating the new question-answer mode and the statistical information searched by the corresponding database;

The threshold comparison subunit is in charge of calculating the matching degree of the matched data in the database and the new fault description content, comparing the matching degree with a preset threshold, creating another new question-answer mode when the matching degree is smaller than the preset threshold, prompting a user to further describe the new fault description content, and feeding back information in the current new question-answer mode when the matching degree is larger than the preset threshold;

the threshold updating subunit is responsible for continuously establishing X new question-answer modes in a new fault description content, and updating a preset threshold according to the average value of the matching degree of X times; x is not less than 2.

Optionally, the threshold updating subunit includes:

the data receiving component is in charge of receiving the matching degree of X times and a preset threshold value, wherein the matching degree of X times comprises the matching degree of the matching data in the database each time and the new fault description content;

a candidate threshold component responsible for determining a candidate preset threshold representing a candidate for a possible match between the first through the X-th question-answer patterns; providing a matching compensation value for each matching degree according to the distance from each matching degree to the average value of the X times of matching degrees, establishing a matching compensation function for calculating a candidate preset threshold value, inputting the matching compensation value into the matching compensation function, and calculating to obtain a preset threshold value;

the threshold output component is in charge of receiving a calculated preset threshold and outputting an updated preset threshold.

Optionally, the data searching module includes:

the incidence relation establishing sub-module is responsible for establishing incidence relation between the target feature information and a database storage catalog, obtaining a corresponding relation table of the target feature information and a data source in the database, generating a query path according to the corresponding relation table, determining keywords for querying the target feature information according to the query path, searching data matched with the keywords from the corresponding relation table, and displaying a list of the searched data;

The target network generation sub-module is responsible for collecting the history dialogue, the target feature information and a storage catalog of the database, generating a target network according to the history dialogue, the target feature information and the storage catalog, establishing a mapping relation between a subarea of the history dialogue, the target feature information and the database and the target network, reading the content in the subarea into a memory, and establishing a data set with a target grid;

and the connection relation establishing sub-module is in charge of receiving the data set, sending out a configuration instruction of the large predictive model when the large language model detects the data set, realizing initialization processing, and establishing connection between the data set and an input port of the large language model.

Optionally, the result output module includes:

The model training sub-module is in charge of selecting sample data from the database, and forming a training data set by the sample data, wherein the sample data is data which is used for users to consult for more than a threshold number of times, and the training data set is used for training a large language model;

The content cleaning sub-module is in charge of cleaning spliced content to obtain cleaned spliced content, inputting the cleaned spliced content into the large language model for identification, and searching fault reasons, maintenance schemes, maintenance time and cost information corresponding to the fault description content in the database;

And the question-answer output sub-module is responsible for taking the fault description content as a title of the result output, and the fault reason, the maintenance scheme, the maintenance time and the cost information as contents of the result output, establishing a connection between the title and the content, generating a question-answer result corresponding to the fault description content, and forming a text result containing automobile fault positioning and maintenance suggestions.

Optionally, the model training sub-module includes:

The information screening unit is responsible for determining that the requirement of the large language model is to generate natural language text, adding a mark and a keyword which are related to or need to be known into the input information according to the requirement of the large language model, indicating specific information focused by the large language model, and screening out target feature information in the input information;

The information fusion unit is responsible for converting a text into a vector representation by using an encoder, converting input information into a representation form understood by a model, and fusing the input information with previous context information;

and the result output unit is responsible for converting the fused input information into a generated text by using a decoder, generating a word or character sequence, and converting the generated word or character into a final text result according to the output of the large language model.

The invention provides an intelligent question-answering method for automobile fault positioning and maintenance suggestion, which comprises the following steps:

Acquiring automobile identity information and fault description content input by a user in real time, and screening target characteristic information required by a model; wherein, the car identity information contains: branding, train, purchase date and maintenance period, the fault description content comprises: failure phenomena, duration, and frequency of occurrence;

Searching data corresponding to the target characteristic information in the database, splicing the history dialogue of the information input module, the target characteristic information and a storage catalog of the database, and inputting the information into the large language model; the database comprises knowledge maps of various automobile identity information, wherein the knowledge maps comprise specific faults, fault reasons, maintenance schemes, maintenance time and cost corresponding to the automobile identity information;

And the large language model processes the spliced content to obtain fault cause, maintenance scheme, maintenance time and cost information corresponding to the fault description content of the information input module.

Optionally, the process of acquiring the automobile identity information and the fault description content input by the user in real time comprises the following steps:

Receiving automobile identity information input by a user, screening out target feature information in the automobile identity information according to the requirement of a large language model, searching a corresponding database according to the screened out target feature information in the automobile identity information by the large language model, and calling a knowledge graph cache stored in the database;

Receiving fault description contents input by a user, screening out target feature information in the fault description contents according to the requirements of a large language model, and calling out a cached knowledge graph by the large language model according to the screened out target feature information in the fault description contents;

And receiving new fault description contents input by the user, searching the database by the large language model according to the new fault description contents, and giving out prompts of the fault description contents by the large language model when the fault description contents model of the user data cannot be identified, and inputting by the user according to the prompts.

The information input module acquires the automobile identity information and the fault description content input by a user in real time, and screens out target characteristic information required by a model; wherein, the car identity information contains: the fault description contents include brands, train systems, purchase dates, maintenance periods and the like: failure phenomena, duration, frequency of occurrence, etc.; the data searching module searches data corresponding to the target characteristic information in the database, splices the history dialogue of the information input module, the target characteristic information and a storage catalog of the database, and inputs the spliced data into the large language model; the database comprises knowledge maps of various automobile identity information, wherein the knowledge maps comprise specific faults, fault reasons, maintenance schemes, maintenance time, cost and the like corresponding to the automobile identity information; the result output module is used for processing the spliced content by the large language model to obtain information such as fault reasons, maintenance schemes, maintenance time, cost and the like corresponding to the fault description content of the information input module; the scheme improves the user experience: the user can quickly acquire the related information of the automobile fault through simple input description, professional automobile knowledge is not needed, and the operation convenience and satisfaction of the user are improved; providing accurate fault diagnosis results: the system can give accurate information such as fault reasons, maintenance schemes, maintenance time, cost and the like by combining the information provided by the user through the knowledge graph in the database, so that the user is helped to better know and solve the problem of the automobile fault; time and cost are saved: the system can rapidly locate and diagnose faults, give corresponding maintenance schemes and cost, and save time and cost for users to find professional technicians and consultation; support knowledge management and sharing: the database in the system is a knowledge graph of the automobile identity information, can be updated and expanded continuously, improves the knowledge management and sharing capacity of the database, and has reference and learning values for professional technicians and related personnel in the automobile maintenance industry. The intelligent fault diagnosis system is realized by integrating the user input, the database and the large language model, a convenient, accurate and time-and cost-saving automobile fault solution is provided, and the intelligent fault diagnosis system has important significance for users and automobile industries.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.

The technical scheme of the invention is further described in detail through the drawings and the embodiments.

Drawings

The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention. In the drawings:

FIG. 1 is a block diagram of an intelligent question-answering system for automobile fault location and repair advice in accordance with embodiment 1 of the present invention;

FIG. 2 is a block diagram of an information input module in embodiment 2 of the present invention;

FIG. 3 is a block diagram of a third information input sub-module in embodiment 3 of the present invention;

FIG. 4 is a block diagram of a mode establishing unit in embodiment 4 of the present invention;

FIG. 5 is a block diagram of a threshold updating subunit of embodiment 5 of the present invention;

FIG. 6 is a block diagram of a data search module in embodiment 6 of the present invention;

FIG. 7 is a block diagram of a result output module in embodiment 7 of the present invention;

FIG. 8 is a block diagram of a model training submodule in embodiment 8 of the present invention;

FIG. 9 is a flow chart of the intelligent question-answering method for automobile fault location and maintenance advice in embodiment 9 of the invention;

FIG. 10 is a process diagram of acquiring real-time input of automobile identity information and fault description content by a user according to embodiment 10 of the present invention;

FIG. 11 is a process diagram of searching for data corresponding to target feature information in a database according to embodiment 11 of the present invention;

FIG. 12 is a process diagram of the large language model processing spliced content in embodiment 12 of the present invention.

Detailed Description

The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the preferred embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention.

The terminology used in the embodiments of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of embodiments of the application. As used in this application and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any or all possible combinations of one or more of the associated listed items.

When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the application as detailed in the accompanying claims. In the description of the present application, it should be understood that the terms "first," "second," "third," and the like are used merely to distinguish between similar objects and are not necessarily used to describe a particular order or sequence, nor should they be construed to indicate or imply relative importance. The specific meaning of the above terms in the present application can be understood by those of ordinary skill in the art according to the specific circumstances.

Example 1: as shown in fig. 1, an embodiment of the present invention provides an intelligent question-answering system for automobile fault location and maintenance suggestion, including:

The information input module is used for acquiring the automobile identity information and the fault description content input by a user in real time and screening target characteristic information required by the model; wherein, the car identity information contains: the fault description contents include brands, train systems, purchase dates, maintenance periods and the like: failure phenomena, duration, frequency of occurrence, etc.;

The data searching module is responsible for searching data corresponding to the target characteristic information in the database, splicing the history dialogue of the information input module, the target characteristic information and the storage catalog of the database, and inputting the information into the large language model; the database comprises knowledge maps of various automobile identity information, wherein the knowledge maps comprise specific faults, fault reasons, maintenance schemes, maintenance time, cost and the like corresponding to the automobile identity information;

The result output module is in charge of processing spliced contents by the large language model to obtain information such as fault reasons, maintenance schemes, maintenance time, cost and the like corresponding to the fault description contents of the information input module;

The working principle and beneficial effects of the technical scheme are as follows: the information input module of the embodiment obtains the automobile identity information and the fault description content input by a user in real time, and screens out target characteristic information required by the model; wherein, the car identity information contains: the fault description contents include brands, train systems, purchase dates, maintenance periods and the like: failure phenomena, duration, frequency of occurrence, etc.; the data searching module searches the data corresponding to the target feature information in the database, splices the history dialogue of the information input module, the target feature information and the storage catalog of the database, and inputs the spliced data into the large language model (in the embodiment, the LLM of the large language model preferentially selects the ChatGLM-6B model with the bloom open source as a base of the LLM, and other models can be selected according to actual demands); the database comprises knowledge maps of various automobile identity information, wherein the knowledge maps comprise specific faults, fault reasons, maintenance schemes, maintenance time, cost and the like corresponding to the automobile identity information; the result output module is used for processing the spliced content by the large language model to obtain information such as fault reasons, maintenance schemes, maintenance time, cost and the like corresponding to the fault description content of the information input module; the scheme improves the user experience: the user can quickly acquire the related information of the automobile fault through simple input description, professional automobile knowledge is not needed, and the operation convenience and satisfaction of the user are improved; providing accurate fault diagnosis results: the system can give accurate information such as fault reasons, maintenance schemes, maintenance time, cost and the like by combining the information provided by the user through the knowledge graph in the database, so that the user is helped to better know and solve the problem of the automobile fault; time and cost are saved: the system can rapidly locate and diagnose faults, give corresponding maintenance schemes and cost, and save time and cost for users to find professional technicians and consultation; support knowledge management and sharing: the database in the system is a knowledge graph of the automobile identity information, can be updated and expanded continuously, improves the knowledge management and sharing capacity of the database, and has reference and learning values for professional technicians and related personnel in the automobile maintenance industry. The intelligent fault diagnosis system is realized by integrating the user input, the database and the large language model, a convenient, accurate and time-and cost-saving automobile fault solution is provided, and the intelligent fault diagnosis system has important significance for users and automobile industries.

Example 2: as shown in fig. 2, on the basis of embodiment 1, an information input module provided in an embodiment of the present invention includes:

The third information input sub-module is in charge of receiving new fault description contents input by a user, the large language model searches in the database according to the new fault description contents, and when the fault description contents model of the user data cannot be identified, the large language model gives out prompts of the fault description contents, and the user inputs according to the prompts;

the working principle and beneficial effects of the technical scheme are as follows: the first information input submodule of the embodiment receives the automobile identity information input by the user, screens out target characteristic information in the automobile identity information according to the requirement of a large language model, searches a corresponding database according to the screened out target characteristic information in the automobile identity information, and retrieves a knowledge graph cache stored in the database; the second information input sub-module receives fault description contents input by a user, screens out target feature information in the fault description contents according to the requirements of a large language model, and the large language model invokes a cached knowledge graph according to the target feature information in the screened out fault description contents; the third information input submodule receives new fault description contents input by a user, the large language model searches in the database according to the new fault description contents, and when the fault description contents of the user data cannot be identified by the large language model, the large language model gives out prompts of the fault description contents, and the user inputs according to the prompts; the scheme improves the accuracy of fault diagnosis: by screening out target characteristic information and calling a knowledge graph, the system can provide more accurate and precise fault diagnosis results, and helps users to better solve the problems; and the user experience is improved: the system can give a prompt according to the fault description content input by the user, so that the user can better describe the fault condition, more personalized and customized service is provided, and the user satisfaction is improved; the efficiency of fault diagnosis is improved: the system can automatically screen out target characteristic information and call a knowledge graph, reduces the workload of manual intervention, improves the efficiency of fault diagnosis, and saves time and resource cost; extensibility and maintainability: the system adopts a modularized design, and each sub-module can be independently developed and maintained, so that the expansion and upgrading of the system are facilitated, and the maintainability and expandability of the system are improved. The embodiment provides a high-efficiency, accurate and personalized fault diagnosis system, realizes a one-to-one interaction mode of users, improves the experience of books on one hand, guides the professional description of the users on the other hand, improves the working efficiency of a model, helps the users solve the problem of automobile faults, and improves the user experience and satisfaction.

Example 3: as shown in fig. 3, on the basis of embodiment 2, a third information input sub-module provided in the embodiment of the present invention includes:

The condition judging unit is in charge of judging whether the feedback information meets a preset condition according to the number of the matched data in the database, and outputting N pieces of feedback information aiming at new fault description content M, wherein M and N are positive integers larger than 1, and N is smaller than or equal to M;

The working principle and beneficial effects of the technical scheme are as follows: the input detection unit of the embodiment detects whether a user has input of new fault description content, if so, the large language model is triggered to execute a new search action program, and if not, whether the user has input of the new fault description content is continuously detected until the user has input of the new fault description content; the mode establishing unit establishes a new question-answer mode according to the new fault description content, and takes the new fault description content as feedback information expected by a characterization user as a starting point of a question-answer under the new question-answer mode; searching matched data in a database according to a key sub-of the new fault description content and acquiring an instruction for retrieving the data in the database according to the request of the new fault description content; the condition judging unit judges whether the feedback information meets a preset condition according to the number of the matched data in the database, and outputs N pieces of feedback information aiming at new fault description content M, wherein M and N are positive integers larger than 1, and N is smaller than or equal to M; the scheme improves the user experience: by monitoring whether a user has input of new fault description content and establishing a new question-answer mode according to the new description, the user is ensured to obtain corresponding feedback information, and satisfaction and experience of the user are improved; accuracy is improved: the data in the database is matched through the keywords, so that instructions related to fault description contents are obtained, and accuracy of feedback information is improved; efficiency is improved: the condition judgment unit is used for outputting feedback information meeting the requirements according to preset conditions, so that excessive feedback information is prevented from being output, and the processing efficiency and the information acquisition efficiency of a user are improved; scalability: new fault description contents and corresponding question-answering modes are added at any time according to the needs so as to adapt to the continuously changing user requirements and fault conditions. The embodiment improves the user experience, accuracy and efficiency, has certain expandability and can better meet the requirements of users.

Example 4: as shown in fig. 4, on the basis of embodiment 3, a mode establishing unit provided in an embodiment of the present invention includes:

The threshold updating subunit is responsible for continuously establishing X new question-answer modes in a new fault description content, and updating a preset threshold according to the average value of the matching degree of X times; x is not less than 2;

The working principle and beneficial effects of the technical scheme are as follows: the information updating subunit of the embodiment establishes a new question-answer mode according to the request of the new fault description content, and updates the new question-answer mode and the statistical information searched by the corresponding database; the threshold comparison subunit calculates the matching degree of the matched data in the database and the new fault description content, compares the matching degree with a preset threshold, creates another new question-answer mode if the matching degree is smaller than the preset threshold, prompts the user to further describe the new fault description content, and feeds back information in the current new question-answer mode if the matching degree is larger than the preset threshold; the threshold updating subunit continuously establishes X new question-answer modes in a new fault description content, and updates a preset threshold according to the average value of the matching degree of X times; the scheme improves the user experience: by creating a new question-answering mode according to the fault description content, the questions of the user are answered more accurately, and related help and solutions are provided, so that the satisfaction degree and the use experience of the user are improved; improving the problem solving efficiency: and (3) rapidly judging whether the matching degree exists enough to provide an accurate answer by calculating the matching degree of the fault description content and the question-answer mode in the database and comparing the matching degree with a preset threshold value. If the degree of matching is less than a preset threshold, prompting the user to provide a more detailed description so as to better understand the problem; if the matching degree is greater than a preset threshold, corresponding answers can be directly provided, and the problem solving efficiency is improved; dynamically updating a threshold value: and updating a preset threshold according to the average value of the matching degree for X times continuously, so that the system can adapt to the change of the user question and the update of the question-answer mode in the database. The threshold value is dynamically updated, so that the matching degree requirements under different scenes are better met, and the intelligent and self-adaptive capacity of the system is improved. According to the method and the system, the understanding capability of the system to the user fault description and the problem solving efficiency are improved through continuously updating the question-answering mode, the matching degree threshold value and the statistical information searched by the database, so that the user experience and the intelligent degree of the system are improved.

Example 5: as shown in fig. 5, on the basis of embodiment 4, the threshold updating subunit provided in the embodiment of the present invention includes:

The threshold output component is in charge of receiving a preset threshold obtained by calculation and outputting an updated preset threshold;

The working principle and beneficial effects of the technical scheme are as follows: the data receiving component of the embodiment receives the matching degree of X times and a preset threshold value, wherein the matching degree of X times comprises the matching degree of the matching data in the database of each time and the new fault description content; the candidate threshold component determines a candidate preset threshold, wherein the candidate preset threshold represents a candidate which is possibly matched between the first question-answer mode and the X-th question-answer mode; providing a matching compensation value for each matching degree according to the distance from each matching degree to the average value of the X times of matching degrees, establishing a matching compensation function for calculating a candidate preset threshold value, inputting the matching compensation value into the matching compensation function, and calculating to obtain a preset threshold value; the threshold output is responsible for receiving a calculated preset threshold and outputting an updated preset threshold; the scheme improves the accuracy of the matching degree: the matching degree of X times is received, the matching degree of the matching data in the database and the new fault description content is included, comprehensive information of multiple matching results is obtained, a candidate threshold component determines a candidate preset threshold according to the matching degrees, one possible matching candidate between the first time question-answer mode and the X time question-answer mode is provided for subsequent processing, accuracy of the matching degree is improved, and user fault description is better understood; consider matching compensation: in order to further improve the accuracy of the matching degree, a matching compensation function for calculating a candidate preset threshold value is introduced. Providing a matching compensation value for each matching degree according to the distance from each matching degree to the average value of the X times of matching degrees, more accurately calculating a candidate preset threshold value, correcting the matching degree according to actual conditions, and improving the accuracy and reliability of the matching degree; dynamically updating a preset threshold value: and outputting the updated preset threshold value to realize dynamic adjustment and updating of the threshold value. According to the calculated preset threshold value, whether the matching degree meets the requirement or not is judged better in the subsequent matching process, so that the system can adaptively adjust the matching requirement, and the intellectualization and adaptability of the system are improved. According to the embodiment, through the cooperative work of the data receiving component and the candidate threshold component and the combination of the matching compensation function, the dynamic updating of the preset threshold is realized, the accuracy of the matching degree and the self-adaptive capacity of the system are improved, and therefore the user experience and the intelligent degree of the system are improved.

Example 6: as shown in fig. 6, on the basis of embodiment 1, a data searching module provided in an embodiment of the present invention includes:

the connection relation establishing sub-module is in charge of receiving the data set, sending out a configuration instruction of the large predictive model when the large language model detects the data set, realizing initialization processing, and establishing connection between the data set and an input port of the large language model;

The working principle and beneficial effects of the technical scheme are as follows: the association relation establishing submodule of the embodiment establishes an association relation between the target feature information and a database storage catalog to obtain a corresponding relation table of the target feature information and a data source in the database, generates a query path according to the corresponding relation table, determines keywords for querying the target feature information according to the query path, searches data matched with the keywords from the corresponding relation table, and displays a list of the searched data; the target network generation submodule collects the history dialogue, the target feature information and a storage catalog of the database, generates a target network according to the history dialogue, the target feature information and the storage catalog, establishes a mapping relation between a subarea of the history dialogue, the target feature information and the database and the target network, reads the content in the subarea into a memory, and establishes a data set with a target grid; the connection relation establishment submodule receives the data set, when the large language model detects the data set, a configuration instruction of the large language model is sent out, initialization processing is realized, and connection between the data set and an input port of the large language model is established; the scheme improves the information retrieval efficiency: by establishing a corresponding relation table and a query path of the target feature information and the database, the target feature information is quickly searched and displayed, and the information retrieval efficiency is improved; an analysis platform providing historical dialog and target feature information: the historical dialogue and the target characteristic information are collected through the target network generation sub-module and mapped into the target network, so that convenience is brought to further analysis; the connection of the data set and the large language model is realized: the data set and the large language model can be connected through the connection relation establishment sub-module, so that the initialization processing of the data and the connection of the input port are realized, and a foundation is provided for the subsequent language model processing; intelligent level of the lifting system: through establishing association relation, collecting historical dialogue and target characteristic information and connecting a large language model, the intelligent level of the system is improved, and the user requirements can be better understood and processed.

Example 7: as shown in fig. 7, on the basis of embodiment 1, the result output module provided in the embodiment of the present invention includes:

The content cleaning sub-module is in charge of cleaning spliced content to obtain cleaned spliced content, inputting the cleaned spliced content into the large language model for identification, and searching information such as fault reasons, maintenance schemes, maintenance time and cost corresponding to the fault description content in the database;

the question-answer output sub-module is responsible for taking the fault description content as the content of the output result, taking the information such as fault cause, maintenance scheme, maintenance time and cost as the content of the output result, establishing a connection between the title and the content, generating a question-answer result corresponding to the fault description content, and forming a text result containing automobile fault positioning and maintenance suggestions;

The working principle and beneficial effects of the technical scheme are as follows: the model training submodule of the embodiment selects sample data from the database, the sample data form a training data set, the sample data is data which is used for users to consult for more than a threshold number of times, and the training data set is used for training a large language model; the content cleaning sub-module cleans the spliced content to obtain the cleaned spliced content, inputs the spliced content into the large language model for identification, and searches information such as fault reasons, maintenance schemes, maintenance time and cost corresponding to the fault description content in the database; the question-answer output sub-module takes the fault description content as the title of the result output, and the information such as fault reason, maintenance scheme, maintenance time and cost as the content of the result output, and the title and the content establish a connection to generate a question-answer result corresponding to the fault description content, so as to form a text result containing automobile fault positioning and maintenance suggestions; the scheme provides an automatic mode to identify the automobile fault and give the maintenance suggestion, so that the workload and time cost of manual processing are reduced; by using a large language model, the system can process more complex fault descriptions, and the accuracy and reliability of positioning and maintenance suggestions are improved; by cleaning and processing the input content, the system quickly and accurately finds maintenance information corresponding to the fault description from a large amount of database information, so that the query efficiency is improved; the generated question and answer results can be directly presented to the user, so that the user can be helped to quickly know the fault cause and the maintenance scheme, and convenient service experience is provided; the method can be applied to the automobile maintenance industry, helps technicians to accurately diagnose and solve faults, and improves maintenance efficiency and quality. The embodiment utilizes a large language model and a database query technology to realize an automatic automobile fault positioning and maintenance suggestion system, provides convenient, accurate and efficient service, and helps a user to solve the automobile fault problem.

Example 8: as shown in fig. 8, on the basis of embodiment 7, the model training submodule provided in the embodiment of the present invention includes:

the result output unit is responsible for converting the fused input information into a generated text by using a decoder, generating a word or character sequence, and converting the generated word or character into a final text result according to the output of the large language model;

the working principle and beneficial effects of the technical scheme are as follows: the information screening unit of the embodiment determines that the requirement of the large language model is to generate a natural language text, adds a mark and a keyword to the input information according to the requirement of the large language model, indicates specific information focused by the large language model, and screens out target feature information in the input information; the information fusion unit converts the text into vector representation by using an encoder, converts the input information into a representation form understood by a model, and fuses the input information with the previous context information; the result output unit converts the fused input information into a generated text by using a decoder, generates a word or character sequence, and converts the generated word or character into a final text result according to the output of the large language model; the large language model of the scheme provides a method for screening, fusing and outputting information so as to meet the requirement of generating natural language text, and the key mark and the key word are added in the input information to indicate that the model pays attention to specific information, so that more definite and accurate task guidance is provided; the encoder of the information fusion unit is used for converting the text into vector representation, and fusing the input information with the context information, so that the model is helpful for understanding the semantics and the context background of the input information, and can better understand the input information and generate more reasonable and accurate text; the result output unit converts the fused input information into a generated text sequence using a decoder and converts it into a final text result. The output mechanism can ensure that the generated text is consistent with the input information and the context information, and the consistency and consistency of the generated text are improved. The embodiment provides a systematic method, so that a large language model can better understand and generate natural language texts, and the quality and the relevance of the generated results are improved. By screening, fusing and outputting information, the model can more accurately interpret task demands and generate text results more in line with expectations.

Example 9: as shown in fig. 9, the embodiment of the invention provides an intelligent question-answering method for automobile fault location and maintenance suggestion, which comprises the following steps:

S100: acquiring automobile identity information and fault description content input by a user in real time, and screening target characteristic information required by a model; wherein, the car identity information contains: the fault description contents include brands, train systems, purchase dates, maintenance periods and the like: failure phenomena, duration, frequency of occurrence, etc.;

S200: searching data corresponding to the target characteristic information in the database, splicing the history dialogue of the information input module, the target characteristic information and a storage catalog of the database, and inputting the information into the large language model; the database comprises knowledge maps of various automobile identity information, wherein the knowledge maps comprise specific faults, fault reasons, maintenance schemes, maintenance time, cost and the like corresponding to the automobile identity information;

S300: the large language model processes the spliced content to obtain information such as fault reasons, maintenance schemes, maintenance time and cost corresponding to the fault description content of the information input module;

The working principle and beneficial effects of the technical scheme are as follows: firstly, acquiring automobile identity information and fault description content input by a user in real time, and screening target characteristic information required by a model; wherein, the car identity information contains: the fault description contents include brands, train systems, purchase dates, maintenance periods and the like: failure phenomena, duration, frequency of occurrence, etc.; secondly, searching data corresponding to the target feature information in a database, splicing the history dialogue of the information input module, the target feature information and a storage catalog of the database, and inputting the spliced data into a large language model; the database comprises knowledge maps of various automobile identity information, wherein the knowledge maps comprise specific faults, fault reasons, maintenance schemes, maintenance time, cost and the like corresponding to the automobile identity information; finally, the large language model is realized to process the spliced content, and information such as fault reasons, maintenance schemes, maintenance time and cost corresponding to the fault description content of the information input module is obtained; the scheme improves the user experience: the user can quickly acquire the related information of the automobile fault through simple input description, professional automobile knowledge is not needed, and the operation convenience and satisfaction of the user are improved; providing accurate fault diagnosis results: the system can give accurate information such as fault reasons, maintenance schemes, maintenance time, cost and the like by combining the information provided by the user through the knowledge graph in the database, so that the user is helped to better know and solve the problem of the automobile fault; time and cost are saved: the system can rapidly locate and diagnose faults, give corresponding maintenance schemes and cost, and save time and cost for users to find professional technicians and consultation; support knowledge management and sharing: the database in the system is a knowledge graph of the automobile identity information, can be updated and expanded continuously, improves the knowledge management and sharing capacity of the database, and has reference and learning values for professional technicians and related personnel in the automobile maintenance industry. The intelligent fault diagnosis system is realized by integrating the user input, the database and the large language model, a convenient, accurate and time-and cost-saving automobile fault solution is provided, and the intelligent fault diagnosis system has important significance for users and automobile industries.

Example 10: as shown in fig. 10, on the basis of embodiment 9, the process for obtaining the real-time input of the identity information and the fault description content of the automobile provided by the embodiment of the invention includes the following steps:

S101: receiving automobile identity information input by a user, screening out target feature information in the automobile identity information according to the requirement of a large language model, searching a corresponding database according to the screened out target feature information in the automobile identity information by the large language model, and calling a knowledge graph cache stored in the database;

S102: receiving fault description contents input by a user, screening out target feature information in the fault description contents according to the requirements of a large language model, and calling out a cached knowledge graph by the large language model according to the screened out target feature information in the fault description contents;

S103: receiving new fault description contents input by a user, searching a database according to the new fault description contents by a large language model, giving out prompts of the fault description contents by the large language model when the fault description contents model of user data cannot be identified, and inputting by the user according to the prompts;

The working principle and beneficial effects of the technical scheme are as follows: according to the embodiment, firstly, automobile identity information input by a user is received, target feature information in the automobile identity information is screened out according to the requirement of a large language model, the large language model searches a corresponding database according to the screened out target feature information in the automobile identity information, and a knowledge graph cache stored in the database is called out; secondly, receiving fault description contents input by a user, screening target feature information in the fault description contents according to the requirements of a large language model, and calling a cached knowledge graph by the large language model according to the screened target feature information in the fault description contents; finally, receiving new fault description contents input by a user, searching a database by the large language model according to the new fault description contents, and giving out prompts of the fault description contents by the large language model when the fault description contents model of the user data cannot be identified, and inputting by the user according to the prompts; the scheme improves the accuracy of fault diagnosis: by screening out target characteristic information and calling a knowledge graph, the system can provide more accurate and precise fault diagnosis results, and helps users to better solve the problems; and the user experience is improved: the system can give a prompt according to the fault description content input by the user, so that the user can better describe the fault condition, more personalized and customized service is provided, and the user satisfaction is improved; the efficiency of fault diagnosis is improved: the system can automatically screen out target characteristic information and call a knowledge graph, reduces the workload of manual intervention, improves the efficiency of fault diagnosis, and saves time and resource cost; extensibility and maintainability: the system adopts a modularized design, and each sub-module can be independently developed and maintained, so that the expansion and upgrading of the system are facilitated, and the maintainability and expandability of the system are improved. The embodiment provides a high-efficiency, accurate and personalized fault diagnosis system, realizes a one-to-one interaction mode of users, improves the experience of books on one hand, guides the professional description of the users on the other hand, improves the working efficiency of a model, helps the users solve the problem of automobile faults, and improves the user experience and satisfaction.

Example 11: as shown in fig. 11, on the basis of embodiment 9, the process of searching the database for the data corresponding to the target feature information provided in the embodiment of the present invention includes the following steps:

S201: establishing an association relation between the target feature information and a database storage directory, obtaining a corresponding relation table of the target feature information and a data source in the database, generating a query path according to the corresponding relation table, determining keywords for querying the target feature information according to the query path, searching data matched with the keywords from the corresponding relation table, and displaying a list of the searched data;

S202: collecting historical dialogue, target feature information and a storage catalog of a database, generating a target network according to the historical dialogue, the target feature information and the storage catalog, establishing a mapping relation between a subarea of the historical dialogue, the target feature information and the database and the target network, reading contents in the subarea into a memory, and establishing a data set with a target grid;

S203: receiving a data set, when the large language model detects the data set, sending out a configuration instruction of the large language model, realizing initialization processing, and establishing connection between the data set and an input port of the large language model;

The working principle and beneficial effects of the technical scheme are as follows: the method comprises the steps of firstly establishing an association relation between target feature information and a database storage catalog to obtain a corresponding relation table of the target feature information and a data source in the database, generating a query path according to the corresponding relation table, determining keywords for querying the target feature information according to the query path, searching data matched with the keywords from the corresponding relation table, and displaying a list of the searched data; secondly, collecting historical dialogue, target feature information and a storage catalog of a database, generating a target network according to the historical dialogue, the target feature information and the storage catalog, establishing a mapping relation between a subarea of the historical dialogue, the target feature information and the database and the target network, reading contents in the subarea into a memory, and establishing a data set with a target grid; finally, receiving a data set, and when the large language model detects the data set, sending out a configuration instruction of the large language model, realizing initialization processing, and establishing connection between the data set and an input port of the large language model; the scheme improves the information retrieval efficiency: by establishing a corresponding relation table and a query path of the target feature information and the database, the target feature information is quickly searched and displayed, and the information retrieval efficiency is improved; an analysis platform providing historical dialog and target feature information: the historical dialogue and the target characteristic information are collected through the target network generation sub-module and mapped into the target network, so that convenience is brought to further analysis; the connection of the data set and the large language model is realized: the data set and the large language model can be connected through the connection relation establishment sub-module, so that the initialization processing of the data and the connection of the input port are realized, and a foundation is provided for the subsequent language model processing; intelligent level of the lifting system: through establishing association relation, collecting historical dialogue and target characteristic information and connecting a large language model, the intelligent level of the system is improved, and the user requirements can be better understood and processed.

Example 12: as shown in fig. 12, on the basis of embodiment 9, the process of processing the spliced content by the large language model provided by the embodiment of the present invention includes the following steps:

s301: sample data are selected from the database, a training data set is formed by the sample data, the sample data are data which are used for users to consult for more than a threshold number of times, and the training data set is used for training a large language model;

S302: cleaning spliced contents to obtain cleaned spliced contents, inputting the cleaned spliced contents into a large language model for identification, and searching information such as fault reasons, maintenance schemes, maintenance time, cost and the like corresponding to the fault description contents in a database;

s303: the method comprises the steps of taking fault description contents as titles, fault reasons, maintenance schemes, maintenance time, cost and other information output by results, establishing a connection between the titles and the contents, generating question-answer results corresponding to the fault description contents, and forming text results containing automobile fault positioning and maintenance suggestions;

The working principle and beneficial effects of the technical scheme are as follows: firstly, sample data are selected from a database, the sample data form a training data set, the sample data are data which are used for a user to consult for more than a threshold number of times, the training data set is used for training a large language model, the requirement of the large language model is determined to be that natural language text is generated, according to the requirement of the large language model, marks and keywords are added into input information about or required to be known, specific information focused by the large language model is indicated, and target feature information in the input information is screened out; converting the text into a vector representation using an encoder, converting the input information into a representation of model understanding, fusing the input information with previous context information; converting the fused input information into a generated text by using a decoder, generating a word or character sequence, and converting the generated word or character into a final text result according to the output of the large language model; secondly, cleaning spliced contents to obtain cleaned spliced contents, inputting the cleaned spliced contents into a large language model for identification, and searching information such as fault reasons, maintenance schemes, maintenance time and cost corresponding to the fault description contents in a database; finally, the fault description content is used as a title of the result output, information such as fault reasons, maintenance schemes, maintenance time and cost are used as the result output content, the title and the content are connected, a question-answer result corresponding to the fault description content is generated, and a text result containing automobile fault positioning and maintenance suggestions is formed; the scheme provides an automatic mode to identify the automobile fault and give the maintenance suggestion, so that the workload and time cost of manual processing are reduced; by using a large language model, the system can process more complex fault descriptions, and the accuracy and reliability of positioning and maintenance suggestions are improved; by cleaning and processing the input content, the system quickly and accurately finds maintenance information corresponding to the fault description from a large amount of database information, so that the query efficiency is improved; the generated question and answer results can be directly presented to the user, so that the user can be helped to quickly know the fault cause and the maintenance scheme, and convenient service experience is provided; the method can be applied to the automobile maintenance industry, helps technicians to accurately diagnose and solve faults, and improves maintenance efficiency and quality. The embodiment utilizes a large language model and a database query technology to realize an automatic automobile fault positioning and maintenance suggestion system, provides convenient, accurate and efficient service, and helps a user to solve the automobile fault problem.

It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.

Claims

1. An intelligent question-answering system for automobile fault location and repair advice, comprising:

the result output module is in charge of processing spliced contents by the large language model to obtain fault reasons, maintenance schemes, maintenance time and cost information corresponding to the fault description contents of the information input module;

an information input module comprising:

a third information input sub-module comprising:

a mode establishing unit comprising:

2. The intelligent question-answering system for vehicle fault locating and repair advice of claim 1, wherein the threshold updating subunit comprises:

3. The intelligent question-answering system for vehicle fault locating and repair advice of claim 1, wherein the data lookup module comprises:

4. The intelligent question-answering system for vehicle fault locating and repair advice of claim 1, wherein the results output module comprises:

5. The intelligent question-answering system for vehicle fault locating and repair advice of claim 4, wherein the model training sub-module comprises:

6. An intelligent question-answering method for automobile fault positioning and maintenance advice is characterized by comprising the following steps:

The large language model processes the spliced content to obtain fault cause, maintenance scheme, maintenance time and cost information corresponding to the fault description content of the information input module;

The process for acquiring the automobile identity information and the fault description content input by the user in real time comprises the following steps:

Receiving new fault description contents input by a user, searching a database according to the new fault description contents by a large language model, giving out prompts of the fault description contents by the large language model when the fault description contents model of user data cannot be identified, and inputting by the user according to the prompts; the method specifically comprises the following steps:

The input detection unit detects whether a user has input of new fault description content, if so, the large language model is triggered to execute a new search action program, and if not, whether the user has input of the new fault description content is continuously detected until the user has input of the new fault description content;

The mode establishing unit establishes a new question-answer mode according to the new fault description content, and takes the new fault description content as feedback information expected by a characterization user as a starting point of a question-answer under the new question-answer mode; searching matched data in a database according to a key sub-of the new fault description content and acquiring an instruction for retrieving the data in the database according to the request of the new fault description content; comprising: the information updating subunit establishes a new question-answer mode according to the request of the new fault description content, and updates the new question-answer mode and the statistical information searched by the corresponding database; the threshold comparison subunit calculates the matching degree of the matched data in the database and the new fault description content, compares the matching degree with a preset threshold, creates another new question-answer mode if the matching degree is smaller than the preset threshold, prompts the user to further describe the new fault description content, and feeds back information in the current new question-answer mode if the matching degree is larger than the preset threshold; the threshold updating subunit continuously establishes X new question-answer modes in a new fault description content, and updates a preset threshold according to the average value of the matching degree of X times; x is not less than 2;

The condition judging unit judges whether the feedback information meets the preset condition according to the number of the matched data in the database, and outputs N pieces of feedback information aiming at the new fault description content M, wherein M and N are positive integers larger than 1, and N is smaller than or equal to M.