Identification large model training method and device, fault classification method and device, equipment and medium

Abstract

The invention discloses an identification large model training method, a fault classification method and device, equipment and a medium, and relates to the technical field of fault identification. The identification large model training method comprises the following steps: generating a model cue word based on preset fault classification information and preset manual annotation information; inputting the model cue word into the initial fault recognition large model to obtain a fault classification result; under the condition that the fault classification result does not meet the preset accurate condition, outputting a manual annotation supplement prompt; and when supplementary annotation information fed back based on the manual annotation supplementary prompt is received, updating the model prompt word based on preset fault classification information and the supplementary annotation information until a fault classification result output by the initial fault recognition large model meets a preset accurate condition, and obtaining the fault recognition large model. In combination with automation of machine learning and manual intelligent judgment, the efficiency and accuracy of fault recognition are improved, and powerful support is provided for equipment maintenance and troubleshooting.

Description

Technical Field

[0001] This application relates to the technical field of fault identification, and specifically to a method for identifying large model training, a fault classification method, a device, a device and a medium. Background Art

[0002] Fault classification is key information in the field of engineering maintenance. The service department and the quality department must conduct component quality analysis based on this information. When there are differences in principle between the user's reported repair information and the service engineer's repair measures, it often requires the service department and the quality department to conduct secondary manual classification, which is costly and inefficient. Currently, in some applications, keyword retrieval is also used for fault classification. Keyword retrieval is based on the logic of fuzzy retrieval. When finding matching items between the fault input and the system content, it does have its unique advantages, such as being able to handle a certain degree of spelling mistakes or synonym matching, thereby improving the flexibility of retrieval classification. However, it has obvious disadvantages in dealing with semantic understanding, near-synonym matching, context dependence, limitations of fuzzy matching, and complex queries. In the field of engineering maintenance, the mechanical principle is complex and there are many industry terms. The same component may also have different names, and keyword retrieval cannot achieve effective application results. Summary of the Invention

[0003] In view of the above deficiencies in the prior art, the purpose of the embodiments of this application is to provide a method for identifying large model training, a fault classification method, a device, a device and a medium.

[0004] To achieve the above purpose, the first aspect of this application provides a method for training a large model for fault identification, including:

[0005] Generating model prompt words based on preset fault classification information and preset manual annotation information, where the manual annotation information includes fault category information, system interpretation information, and component / action example information;

[0006] Inputting the model prompt words into the initial large model for fault identification to obtain a fault classification result;

[0007] In the case where the fault classification result does not meet the preset accuracy condition, outputting a manual annotation supplementary prompt;

[0008] In the case of receiving supplementary annotation information feedback based on the manual annotation supplementary prompt, updating the model prompt words based on the preset fault classification information and the supplementary annotation information until the fault classification result output by the initial large model for fault identification meets the preset accuracy condition, and obtaining the large model for fault identification.

[0009] In the embodiments of the present application, the preset fault classification information includes system information, component / action information, and fault mode information. In the case where the fault classification result does not meet the preset accuracy condition, an artificial annotation supplement prompt is output, including:

[0010] In the case where the fault classification result does not meet the preset accuracy condition, determine the classification error type based on the fault classification result, where the classification error type includes system classification error, component / action classification error, fault mode classification error, and uncompleted classification error;

[0011] Determine and output an artificial annotation supplement prompt based on the classification error type.

[0012] In the embodiments of the present application, the supplementary annotation information includes fault category supplementary information, system interpretation supplementary information, fault mode example supplementary information, and component / action example supplementary information. Update the model prompt words based on the fault category information and the supplementary annotation information, including:

[0013] Based on the fault category supplementary information, or the system interpretation supplementary information, or the fault mode example, or the component / action example supplementary information, add a fault category, or a system interpretation, or a fault mode example, or a component / action example in the preset fault classification information;

[0014] Update the model prompt words based on the updated preset fault classification information.

[0015] In the embodiments of the present application, input the model prompt words into the initial fault recognition large model to obtain a fault classification result, including:

[0016] Input the model prompt words into the initial fault recognition large model, and vectorize the model prompt words through the initial fault recognition large model to obtain a prompt word vector;

[0017] Determine the vector distance between the prompt word vector and each preset fault vector respectively, where the preset fault vector is the vector information corresponding to the preset fault classification information;

[0018] Determine the fault classification result based on the preset fault vector corresponding to the minimum vector distance.

[0019] In the embodiments of the present application, the fault recognition large model training method further includes:

[0020] Determine historical faults and their corresponding historical fault classifications based on historical fault classification information;

[0021] Determine the preset artificial annotation information based on the historical faults and their corresponding historical fault classifications.

[0022] In the second aspect of the present application, a construction machinery fault classification method is provided, including:

[0023] Obtain the fault description information of construction machinery;

[0024] Input the fault description information of construction machinery into the fault recognition large model to obtain the fault classification of construction machinery, where the fault recognition large model is trained by the fault recognition large model training method described in the above embodiment.

[0025] The third aspect of this application provides a fault recognition large model training device, including:

[0026] A prompt word generation module, configured to generate model prompt words based on preset fault classification information and preset manual annotation information;

[0027] A fault classification module, configured to input the model prompt words into the initial fault recognition large model to obtain a fault classification result;

[0028] An annotation supplement module, configured to output a manual annotation supplement prompt when the fault classification result does not meet the preset accuracy condition;

[0029] A prompt word update module, configured to, when receiving the supplementary annotation information fed back based on the manual annotation supplement prompt, update the model prompt words based on the preset fault classification information and the supplementary annotation information until the fault classification result output by the initial fault recognition large model meets the preset accuracy condition, thereby obtaining the fault recognition large model.

[0030] The fourth aspect of this application provides a construction machinery fault classification device, including:

[0031] An information acquisition module, configured to acquire the fault description information of construction machinery;

[0032] A fault recognition module, which inputs the fault description information of construction machinery into the fault recognition large model to obtain the fault classification of construction machinery, where the fault recognition large model is trained by the fault recognition large model training method described in the above embodiment.

[0033] The fifth aspect of this application provides a computing device, including:

[0034] A memory, configured to store instructions;

[0035] A processor, configured to call instructions from the memory and be able to implement the fault recognition large model training method described in the above embodiment or the fault classification method described in the above embodiment when executing the instructions.

[0036] The sixth aspect of this application provides a machine-readable storage medium, on which instructions are stored, and these instructions are used to make the machine execute the fault recognition large model training method described in the above embodiment or the fault classification method described in the above embodiment.

[0037] Through the above technical solution, model prompting words are generated based on preset fault classification information and preset manual annotation information, where the manual annotation information includes fault category information, system interpretation information, and component / action example information; the model prompting words are input into the initial fault recognition large model to obtain a fault classification result; in the case where the fault classification result does not meet the preset accuracy condition, a manual annotation supplement prompt is output; in the case of receiving supplementary annotation information feedback based on the manual annotation supplement prompt, the model prompting words are updated based on the preset fault classification information and the supplementary annotation information until the fault classification result output by the initial fault recognition large model meets the preset accuracy condition, and a fault recognition large model is obtained. It combines the automation of machine learning and the intelligent judgment of humans, and improves the classification accuracy of the fault recognition large model through continuous iteration and optimization. Thereby improving the efficiency and accuracy of fault recognition, and providing strong support for equipment maintenance and fault troubleshooting.

[0038] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation part. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the following specific implementation, they are used to explain the embodiments of the present application, but do not constitute a limitation to the embodiments of the present application. In the drawings:

[0040] Figure 1 Schematically shows a flowchart of a method for training a large fault recognition model according to an embodiment of the present application;

[0041] Figure 2 Schematically shows a flowchart of a method for classifying construction machinery faults according to an embodiment of the present application;

[0042] Figure 3 Schematically shows a structural block diagram of a device for training a large fault recognition model according to an embodiment of the present application;

[0043] Figure 4 Schematically shows a structural block diagram of a device for training a large fault recognition model according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiments of the present application, and is not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0045] It should be noted that the acquisition, transmission, storage, use, processing, etc. of data in the technical solution of this application all comply with the relevant provisions of national laws and regulations. In the embodiments of this application, some existing solutions in the industry such as certain software, components, models, etc. may be mentioned. They should be regarded as exemplary, and their purpose is only to illustrate the feasibility in the implementation of the technical solution of this application, but it does not mean that the applicant has already or necessarily used this solution.

[0046] It should be noted that if there are directional indications (such as up, down, left, right, front, back...) involved in the embodiments of this application, then this directional indication is only used to explain the relative positional relationship, movement conditions, etc. between components in a certain specific posture (as shown in the drawings). If this specific posture changes, then this directional indication also changes accordingly.

[0047] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of this application, then the descriptions of "first", "second", etc. are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first", "second" may explicitly or implicitly include at least one such feature. In addition, the technical solutions between various embodiments can be combined with each other, but it must be based on the ability of those of ordinary skill in the art to implement. When the combination of technical solutions results in contradictions or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0048] Figure 1 Schematically shows a flowchart of a method for training a large fault recognition model according to an embodiment of the present application. As Figure 1 shown, the embodiments of the present application provide a method for training a large fault recognition model, and this method may include the following steps:

[0049] Step 100, generating model prompt words based on preset fault classification information and preset manual annotation information, where the manual annotation information includes fault category information, system interpretation information, and component / action example information;

[0050] In this embodiment, it should be noted that the model prompt is the key to guiding the model to perform tasks, including task description, context information, and expected output. Accurate prompts help the model better understand the tasks and improve classification performance. Among them, the task description is the part of the prompt that clearly tells the model what task needs to be performed. The task description will directly point out the specific task type that the model needs to complete. For example, in the fault identification task, the task description can be: "Please analyze the following text and determine the fault type it belongs to." Thus, the model can understand that the task to be performed is to identify the fault type. The context information is the key information provided to the model in the prompt to assist it in understanding and completing the task. In the fault identification task, the context information may include a detailed description of the fault phenomenon, equipment type, operating conditions, historical fault records, etc. These information help the model more accurately understand the fault situation and make judgments accordingly. For example: "This equipment is of XX model and has recently shown the following phenomena: XXX" Such a description provides the necessary background information for the model. The expected output result is the part of the prompt that specifies how the model should present its answer or result. In the fault identification task, the expected output result may be the name, number, or code of the fault type. For example: "Please output the corresponding fault type name." Such a description makes the model know that it should return the specific name of the fault type.

[0051] The preset fault classification information and the manual annotation information together constitute the model prompt, providing the model with an initial learning direction and basis. Specifically, before training the model, it is necessary to pre-construct the preset fault classification information. The preset fault classification information is a collection of faults based on the fault phenomenon category directory, including information such as the part names of subsystems, mechanical working principles, and industry terms, which is used to improve the accuracy of fault category word vectors. Taking the luffing system of an engineering crane as an example, the parts include luffing cylinders, luffing balance valves, etc.; the mechanical working principle is the basic hydraulic principle involved in the parts, and the industry terms include keywords such as arm lowering, arm raising, and luffing subsidence. The construction basis of the fault phenomenon category directory is the hierarchical structure of "system - part / action - fault mode". Specifically, refer to Table 1.

[0052] Table 1

[0053]

[0054] Adopting the hierarchical structure of "system - part / action - fault mode" helps to systematically organize and classify fault phenomena. The preset fault classification information is the core of the fault information knowledge base, providing a basic framework for the model's classification.

[0055] It should be noted that the manually marked information includes fault category information, system interpretation information, and component / action example information. Among them, the fault category information includes specific examples corresponding to the preset fault categories, including the manually selected fault sample set; the system interpretation information includes a detailed explanation of the system category, such as "the prone arm is classified into the telescopic system"; the component / action example information includes specific examples of components or actions involved when a fault occurs, such as "the prone arm is a luffing down action", "the intercooler is a component of the power system", etc.

[0056] In one embodiment, the method for training the fault recognition large model further includes:

[0057] Determining historical faults and their corresponding historical fault classifications based on historical fault classification information;

[0058] Determining preset manually marked information based on the historical faults and their corresponding historical fault classifications.

[0059] It can be understood that the preset manually marked information includes relevant fault classification information obtained by collecting and analyzing historical fault phenomena. Specifically, historical fault data is collected from databases, log files, or maintenance records. The fault data includes detailed information such as the description of the fault, the occurrence time, and the scope of influence. According to the existing fault classification system, the historical faults are classified. Among them, the existing fault classification system is the hierarchical structure included in the preset fault classification information. Each historical fault is matched with its corresponding classification. The historical fault classifications are analyzed in depth to understand the common characteristics, occurrence frequencies, and possible causes of various faults. The historical fault data is manually marked using the formulated marking rules. The verified and adjusted marking information is organized into a library to obtain the preset manually marked information. The preset manually marked information can be continuously updated and improved to adapt to new fault types and classification requirements.

[0060] Step 200, inputting the model prompt words into the initial fault recognition large model to obtain a fault classification result;

[0061] In this embodiment, it should be noted that for the large model semantic recognition model, the word vector operation logic is adopted. Through vectorization processing, the system content is converted into a vector space, and then the distance between vectors is used to judge the semantic similarity between the user input and the system content. This method is not limited to the exact match of keywords and can more accurately capture the semantic relationship between the user intention and the system content.

[0062] By constructing preset fault classification information and preset manual annotation information that include rich subsystem component names, mechanical working principles, and industry terms, etc., the accuracy of fault category word vectors can be improved. Based on the preset fault classification information and preset manual annotation information, model prompt words are generated. By vectorizing the words and phrases in the model prompt words, a more accurate and rich word vector space can be generated, thereby further improving the accuracy and relevance of semantic retrieval of the initial fault recognition large model.

[0063] In one embodiment, the model prompt words are input into the initial fault recognition large model to obtain a fault classification result, including:

[0064] The model prompt words are input into the initial fault recognition large model, and the model prompt words are vectorized by the initial fault recognition large model to obtain prompt word vectors;

[0065] The vector distances between the prompt word vectors and each preset fault vector are determined respectively, where the preset fault vector is the vector information corresponding to the preset fault classification information;

[0066] The fault classification result is determined based on the preset fault vector corresponding to the minimum vector distance.

[0067] In this embodiment, it should be noted that vectorization is the process of converting data into vector form. In the initial fault recognition large model, there is an embedding layer inside the initial fault recognition large model, which can convert the input prompt words into vectors in a high-dimensional vector space, that is, prompt word vectors. This process is usually achieved through a learned embedding matrix. After vectorization processing, a vector representing the model prompt words can be obtained, and this vector contains the position information of the prompt words in the vector space. The preset fault vectors are predefined vector information corresponding to the preset fault classification information. These vectors are learned through the training process and can represent the characteristics of different fault categories. Using a distance metric method, the distance between the prompt word vector and each preset fault vector is calculated. The smaller the distance, the more similar the prompt word vector is to the preset fault vector. The distance metric method can include Euclidean distance, Manhattan distance, Chebyshev distance, and cosine similarity, etc. Among all the calculated vector distances, find the minimum distance value, and the fault category represented by the preset fault vector corresponding to the minimum distance is the fault classification result to be determined.

[0068] Step 300, when the fault classification result does not meet the preset accuracy condition, output a manual annotation supplement prompt;

[0069] It should be noted that the preset accurate conditions may include pre-set conditions such as classification accuracy and recall rate. The classification result is judged whether it meets the requirements through the preset accurate conditions. If the preset accurate conditions are not met, it indicates that the classification ability of the initial fault recognition large model still needs to be improved. At this time, manual annotation supplementary prompts need to be output. These manual annotation supplementary prompts are based on the deficiencies and errors in the current classification result to guide manual annotation more precisely.

[0070] Step 400, in the case of receiving the supplementary annotation information fed back based on the manual annotation supplementary prompts, update the model prompt words based on the preset fault classification information and the supplementary annotation information until the fault classification result output by the initial fault recognition large model meets the preset accurate conditions, and obtain the fault recognition large model.

[0071] It should be noted that after receiving the supplementary annotation information fed back based on the manual annotation supplementary prompts, this information is used to update the model prompt words. The update process may include adding new fault category information, refining system interpretation information, or adding more component / action example information, etc. The updated model prompt words are input into the model again for a new round of classification and judgment until the classification result of the model meets the preset accurate conditions. After multiple iterations and optimizations, when the initial fault recognition large model can stably output a fault classification result that meets the preset accurate conditions, the model is regarded as trained and becomes an available fault recognition large model.

[0072] In this embodiment, model prompt words are generated based on the preset fault classification information and the preset manual annotation information, where the manual annotation information includes fault category information, system interpretation information, and component / action example information; the model prompt words are input into the initial fault recognition large model to obtain a fault classification result; in the case where the fault classification result does not meet the preset accurate conditions, manual annotation supplementary prompts are output; in the case of receiving the supplementary annotation information fed back based on the manual annotation supplementary prompts, update the model prompt words based on the preset fault classification information and the supplementary annotation information until the fault classification result output by the initial fault recognition large model meets the preset accurate conditions, and obtain the fault recognition large model. Combining the automation of machine learning and the intelligent judgment of humans, the classification accuracy of the fault recognition large model is improved through continuous iteration and optimization. Thus, the efficiency and accuracy of fault recognition are improved, providing strong support for equipment maintenance and fault troubleshooting.

[0073] In one embodiment, the preset fault classification information includes system information, component / action information, and fault mode information. In the case where the fault classification result does not meet the preset accurate conditions, the manual annotation supplementary prompts output include:

[0074] In the case where the fault classification result does not meet the preset accuracy condition, determine the classification error type based on the fault classification result, where the classification error type includes system classification error, component / action classification error, fault mode classification error, and uncompleted classification error;

[0075] Determine and output a manual annotation supplement prompt based on the classification error type.

[0076] In this embodiment, it should be noted that in the fault classification task, when the classification result of the initial fault recognition large model fails to meet the preset accuracy condition, it is necessary to conduct a detailed analysis of the classification error and provide a manual annotation supplement prompt accordingly to guide subsequent annotation and model optimization work. When a fault is wrongly classified into a category that does not match its actual system, a system classification error occurs. For example, a fault in an electrical system is wrongly classified as a fault in a mechanical system, or the boom retraction is classified into the telescoping system, etc., which are systematic classification errors. If the components or actions involved in the fault are wrongly identified, a component / action classification error is generated. For example, a fault related to the engine control module is wrongly associated with the transmission component. The fault mode classification error refers to the wrong classification of the specific manifestation form of the fault. For example, the action of not being able to stop is classified as being too fast, or a fault caused by overheating is wrongly classified as a short circuit fault, etc. When the model fails to classify the fault into any preset fault category, an uncompleted classification error occurs. This usually means that the coverage of the model is limited or the fault description is not clear enough.

[0077] When different classification error types occur, corresponding manual annotation supplement prompts will be output. For example, for the supplement prompt for system classification error, it prompts the annotator to check and confirm the system to which the fault belongs to ensure the accuracy of classification, and more detailed system descriptions or classification guidelines need to be provided. For the supplement prompt for component / action classification error, it instructs the annotator to carefully review the components or actions involved in the fault description and classify them accurately according to the preset component / action list, and detailed definitions or examples of components / actions need to be provided. For the supplement prompt for fault mode classification error, it guides the annotator to analyze the specific manifestation form of the fault and classify it accurately according to the preset fault mode list, and detailed explanations or differentiation criteria of fault modes need to be provided. For the supplement prompt for uncompleted classification error, it suggests that the annotator consider whether it is necessary to expand or modify the preset fault categories to cover the current fault. At the same time, providing additional fault descriptions or context information may help to complete the classification.

[0078] In this embodiment, by effectively identifying the classification error type and providing targeted manual annotation supplement prompts, the accuracy and efficiency of fault classification are improved.

[0079] In one embodiment, the supplementary annotation information includes fault category supplementary information, system interpretation supplementary information, fault mode example supplementary information, and component / action example supplementary information. Updating the model prompt words based on the preset fault classification information and the supplementary annotation information includes:

[0080] Based on the fault category supplementary information, or the system interpretation supplementary information, or the fault mode example, or the component / action example supplementary information, adding a fault category, or a system interpretation, or a fault mode example, or a component / action example in the preset fault classification information;

[0081] Updating the model prompt words based on the updated preset fault classification information.

[0082] In this embodiment, it should be noted that the fault category supplementary information includes newly added fault categories or refined descriptions of existing fault categories. For the fault categories that need to be subdivided based on the fault category supplementary information, necessary splitting is performed; for the categories that need to be merged, merging processing is performed. The fault category supplementary information helps the fault recognition large model to more accurately identify and understand different types of faults. The system interpretation supplementary information is a detailed explanation of the background, cause, or impact of the fault occurrence. Supplementing this part of information can help the fault recognition large model better understand the relationship between the fault and the system and improve the accuracy of classification. The fault mode example supplementary information covers different types of fault modes so that the fault recognition large model can learn the specific manifestation forms of the faults. The component / action example supplementary information integrates examples of components or actions related to the fault to ensure that these examples can accurately reflect the association between the fault and the components / actions.

[0083] Specifically, adding the newly added fault categories in the preset fault classification information; updating the descriptions of the existing fault categories to include more detailed information or more accurate definitions; adding or updating the system interpretation to reflect the new understanding between the fault and the system; introducing fault mode examples and component / action examples to enhance the model's ability to identify the specific manifestation forms of the faults. Based on the updated preset fault classification information, updating the model prompt words.

[0084] In this embodiment, the model prompt words are updated through the supplementary annotation information, thereby improving the accuracy and generalization ability of the fault recognition large model.

[0085] Figure 2 Schematically shows a flowchart of a construction machinery fault classification method according to an embodiment of the present application. As Figure 2 shown, the embodiment of the present application also provides a construction machinery fault classification method, including the following steps:

[0086] Step 500, obtaining the construction machinery fault description information;

[0087] Step 600: Input the construction machinery fault description information into the fault recognition large model to obtain the construction machinery fault classification, where the fault recognition large model is trained by the fault recognition large model training method described in the above embodiments.

[0088] In this embodiment, it should be noted that the construction machinery fault description information includes the time and location of the fault occurrence, specific phenomena, model and specifications of the construction machinery, and any relevant operation or maintenance history, etc. In one embodiment, before inputting the construction machinery fault description information into the fault recognition large model, preprocessing such as text cleaning, word segmentation, and stop word removal is also performed to ensure the accuracy and consistency of the input information.

[0089] It can be understood that with the continuous development of construction machinery technology and the emergence of new fault types, the fault recognition large model will also be updated or retrained regularly to ensure that it can continuously and accurately identify new fault types.

[0090] In this embodiment, by using the fault recognition large model to classify construction machinery faults, the maintenance efficiency is improved and the maintenance cost is reduced.

[0091] Figure 3 Schematically shows a structural block diagram of a fault recognition large model training device according to an embodiment of the present application. As Figure 3 shown, an embodiment of the present application provides a fault recognition large model training device 1000, which may include:

[0092] A prompt word generation module 1001, configured to generate model prompt words based on preset fault classification information and preset manual annotation information;

[0093] A fault classification module 1002, configured to input the model prompt words into the initial fault recognition large model to obtain a fault classification result;

[0094] An annotation supplement module 1003, configured to output a manual annotation supplement prompt when the fault classification result does not meet the preset accuracy condition;

[0095] A prompt word update module 1004, configured to update the model prompt words based on the preset fault classification information and the supplementary annotation information when receiving the supplementary annotation information fed back based on the manual annotation supplement prompt, until the fault classification result output by the initial fault recognition large model meets the preset accuracy condition, so as to obtain the fault recognition large model.

[0096] Figure 4 Schematically shows a structural block diagram of a fault recognition large model training device according to an embodiment of the present application. As Figure 4 shown, an embodiment of the present application provides a construction machinery fault classification device 2000, which may include:

[0097] An information acquisition module 2001, configured to acquire construction machinery fault description information;

[0098] A fault identification module 2002 inputs the construction machinery fault description information into a fault identification large model to obtain a classification of construction machinery faults, where the fault identification large model is trained by the fault identification large model training method described in the above embodiments.

[0099] An embodiment of the present application further provides a computing device, including:

[0100] A memory, configured to store instructions;

[0101] A processor, configured to call instructions from the memory and, when executing the instructions, be able to implement the fault identification large model training method described in the above embodiments or the fault classification method described in the above embodiments.

[0102] An embodiment of the present application further provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to cause a machine to execute the fault identification large model training method described in the above embodiments or the fault classification method described in the above embodiments.

[0103] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0104] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the specified functions in one Figure 1 one or more flows and / or multiple flows and / or blocks Figure 1 one or more blocks or multiple blocks.

[0105] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means that implement the functions specified in the flowchart Figure 1 one or more flowcharts and / or boxes Figure 1 specified in the box or boxes.

[0106] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart Figure 1 one or more flowcharts and / or boxes Figure 1 specified in the box or boxes.

[0107] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0108] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM) and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

[0109] Computer-readable media includes both permanent and non-permanent, removable and non-removable media implemented by any method or technology for storing information. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile discs (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0110] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, commodity or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, commodity or device comprising the element.

[0111] The above are only embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A large fault identification model training method, characterized in that: include: Generate model prompt words based on preset fault classification information and preset manual annotation information, wherein the manual annotation information includes fault category information, system interpretation information, and component / action example information; Inputting the model prompt words into the initial fault identification model to obtain a fault classification result; When the fault classification result does not meet the preset accuracy condition, outputting a supplementary manual annotation prompt; When supplementary annotation information based on the manual annotation supplementary prompt feedback is received, the model prompt word is updated based on the preset fault classification information and the supplementary annotation information until the fault classification result output by the initial fault identification large model meets the preset accuracy condition, thereby obtaining the fault identification large model.

2. The fault identification large model training method according to claim 1, characterized in that: The preset fault classification information includes system information, component / action information and fault mode information. When the fault classification result does not meet the preset accuracy condition, outputting a manual annotation supplement prompt includes: In the case where the fault classification result does not meet the preset accuracy condition, determining the classification error type based on the fault classification result, wherein the classification error type includes system classification error, component / action classification error, fault mode classification error and incomplete classification error; Determine and output a supplementary prompt for manual annotation based on the classification error type.

3. The fault identification large model training method according to claim 2 is characterized in that: The supplementary annotation information includes supplementary information on fault categories, supplementary information on system definitions, supplementary information on fault mode examples, and supplementary information on component / action examples. The updating of the model prompt words based on the preset fault classification information and the supplementary annotation information includes: Based on the fault category supplementary information or the system interpretation supplementary information or the fault mode example supplementary information or the component / action example supplementary information, the fault category or system interpretation or the fault mode example or component / action example in the preset fault classification information is added; The model prompt word is updated based on the newly added preset fault classification information.

4. The fault identification large model training method according to claim 1 is characterized in that: The step of inputting the model prompt words into the initial fault identification model to obtain a fault classification result includes: Inputting the model prompt word into the initial fault identification large model, vectorizing the model prompt word through the initial fault identification large model, and obtaining a prompt word vector; respectively determining the vector distances between the prompt word vector and each preset fault vector, wherein the preset fault vector is the vector information corresponding to the preset fault classification information; A fault classification result is determined based on a preset fault vector corresponding to the minimum vector distance.

5. The fault identification large model training method according to claim 1, characterized in that: Also includes: Determine historical faults and corresponding historical fault classifications based on historical fault classification information; The preset manual labeling information is determined based on the historical faults and the corresponding historical fault classifications.

6. A method for classifying engineering machinery faults, characterized in that: include: Obtain engineering machinery fault description information; The engineering machinery fault description information is input into a fault identification large model to obtain an engineering machinery fault classification, wherein the fault identification large model is trained according to the fault identification large model training method according to any one of claims 1 to 5.

7. A large model training device for fault identification, characterized in that: include: A prompt word generation module, used to generate model prompt words based on preset fault classification information and preset manual annotation information; A fault classification module, used for inputting the model prompt words into the initial fault identification large model to obtain a fault classification result; A labeling supplement module, used for outputting a manual labeling supplement prompt when the fault classification result does not meet the preset accuracy condition; A prompt word updating module is used to update the model prompt word based on the preset fault classification information and the supplementary annotation information when receiving the supplementary annotation information based on the manual annotation supplementary prompt feedback, until the fault classification result output by the initial fault identification large model meets the preset accuracy condition, thereby obtaining the fault identification large model.

8. A device for classifying engineering machinery faults, characterized in that: include: An information acquisition module is used to obtain engineering machinery fault description information; The fault identification module inputs the engineering machinery fault description information into a fault identification large model to obtain an engineering machinery fault classification, wherein the fault identification large model is trained according to the fault identification large model training method according to any one of claims 1 to 5.

9. A computing device, characterized in that include: a memory configured to store instructions; A processor is configured to call the instructions from the memory and to implement the fault identification large model training method according to any one of claims 1 to 5 or the fault classification method according to claim 6 when executing the instructions.

10. A machine-readable storage medium, characterized in that: The machine-readable storage medium stores instructions for enabling a machine to execute the fault identification large model training method according to any one of claims 1 to 5 or the fault classification method according to claim 6.

Images