Maintenance data classification method, electronic equipment and computer readable storage medium
By using large language models to automatically classify the maintenance data of electronic equipment, the problems of low efficiency and high cost of manual classification are solved, and efficient and accurate classification of fault types are achieved.
Patent Information
- Application Number
- CN202411726819.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-27
- Publication Date
- 2025-04-11
AI Technical Summary
The prior art relies on manual classified repair data when repairing electronic equipment failures, which is inefficient and cost-effective.
The large language model is used to perform structured fault type classification for unstructured maintenance data. By obtaining the repair data and a set of fault type, a prompt word is generated and a large language model is input to obtain fault type classification.
Improve the classification efficiency of maintenance data, reduce labor costs, and improve the accuracy and efficiency of classification through automated processes.
Smart Images
Figure CN119807804A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of data classification, and in particular to a classification method for maintenance data, an electronic device, and a computer-readable storage medium. Background Art
[0002] When an electronic device fails, it needs to be repaired, and during the repair, it is also necessary to record the corresponding repair data for subsequent fault classification. The relevant technology adopts a manual classification method. Summary of the invention
[0003] The maintenance data classification method, electronic device, and computer-readable storage medium provided in the present application can improve the classification efficiency of maintenance data.
[0004] In a first aspect, the present application provides a maintenance data classification method, which includes: obtaining maintenance data of a target device and obtaining a set of fault types corresponding to the target device; wherein the maintenance data includes at least one of a fault phenomenon description and a maintenance description; generating prompt words based on the maintenance data and the set of fault types; inputting the prompt words into a large language model to obtain a fault type classification corresponding to the maintenance data.
[0005] Generating prompt words based on maintenance data and a set of fault types includes: generating a task description based on the maintenance data and a set of fault types; wherein the task description is used to guide the reasoning logic of a large language model; and generating prompt words based on the maintenance data, the set of fault types and the task description.
[0006] Among them, generating a task description based on maintenance data and a set of fault types includes: generating a first reasoning logic for maintenance data that only includes a description of a fault phenomenon; generating a second reasoning logic for maintenance data that only includes a maintenance description; generating a third reasoning logic for maintenance data that includes a description of a fault phenomenon and a maintenance description; and generating a task description based on the first reasoning logic, the second reasoning logic, the third reasoning logic, the maintenance data, and a set of fault types.
[0007] Among them, the maintenance data also includes maintenance parts, and a task description is generated based on the maintenance data and a set of fault types, including: generating a fourth reasoning logic for the maintenance data including maintenance parts; generating a task description based on the first reasoning logic, the second reasoning logic, the third reasoning logic, the maintenance data and all fault types, including: generating a task description based on the first reasoning logic, the second reasoning logic, the third reasoning logic, the fourth reasoning logic, the maintenance data and a set of fault types.
[0008] The generating of prompt words based on the maintenance data, the fault type set and the task description includes: splicing the task description, the maintenance data and the fault type set in order to generate the prompt words.
[0009] The fault type set includes a number of first-level fault types; each first-level fault type includes a number of second-level fault types, and the second-level fault type is a subdivision of the first-level fault type.
[0010] The generating of prompt words based on the maintenance data and the set of fault types includes: performing dimension reduction or dimension increase processing on the maintenance data to obtain target maintenance data; and generating prompt words based on the target maintenance data and the set of fault types.
[0011] Among them, after the prompt word is input into the large language model to obtain the fault type classification corresponding to the maintenance data, it includes: in response to the fault type classification indicating that the classification is not completed, the maintenance data is sent to the manual terminal; and the manual terminal feedbacks the fault type classification corresponding to the maintenance data.
[0012] In a second aspect, the present application provides an electronic device, comprising a processor and a memory coupled to the processor; wherein the memory is used to store a computer program, and the processor is used to execute the computer program to implement the classification method provided in the first aspect.
[0013] In a third aspect, the present application provides a computer-readable storage medium for storing a computer program, which, when executed by a processor, is used to implement the classification method provided in the first aspect.
[0014] The beneficial effects of the present application are as follows: Different from the prior art, the maintenance data classification method, electronic device, and computer-readable storage medium provided by the present application utilize the generative characteristics of a large language model to perform structured fault type classification on unstructured maintenance data, thereby improving the classification efficiency of maintenance data and reducing labor costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:
[0016] Figure 1 It is a flowchart of an embodiment of a maintenance data classification method provided by the present application;
[0017] Figure 2 It is a schematic diagram of the network structure of the large language model provided by this application;
[0018] Figure 3 is a flow chart of another embodiment of the maintenance data classification method provided by the present application;
[0019] Figure 4 It is a structural schematic diagram of an embodiment of an electronic device provided by the present application;
[0020] Figure 5 It is a structural schematic diagram of an embodiment of a computer-readable storage medium provided by the present application. DETAILED DESCRIPTION
[0021] 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 will be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only some but not all structures related to the present application are shown in the drawings. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.
[0022] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0023] When an electronic device fails, it needs to be repaired, and during the repair, the corresponding repair data needs to be recorded for subsequent fault classification, and the subsequent production of the equipment can be improved according to the classified faults. The relevant technology adopts the method of manual classification.
[0024] Based on this, the present application proposes to utilize the generative characteristics of a large language model to perform structured fault type classification on unstructured maintenance data, thereby improving the classification efficiency of maintenance data and reducing labor costs. For details, please refer to the technical solution of any of the following embodiments.
[0025] See also Figure 1 , Figure 1 1 is a flow chart of an embodiment of a maintenance data classification method provided by the present application. The classification method includes:
[0026] Step 11: Obtain maintenance data of the target device and obtain a set of fault types corresponding to the target device.
[0027] In some embodiments, the target device may be a home appliance, such as a dishwasher, a microwave oven, a washing machine, a television, an induction cooker, a refrigerator, an air conditioner, and a series of other home appliances.
[0028] In some embodiments, the target device may also be other types of devices, such as mobile phones, computers, tablets, and other devices.
[0029] In some embodiments, the maintenance data includes at least one of a fault phenomenon description and a maintenance description.
[0030] In some embodiments, the repair data includes a fault symptom description.
[0031] In some embodiments, the repair data includes a repair description.
[0032] In some embodiments, the maintenance data includes at least a fault phenomenon description and a maintenance description.
[0033] In some embodiments, the fault phenomenon description may be provided by a user of the target device, and may be in the form of text, voice, and / or image.
[0034] In some embodiments, the maintenance description may be provided by a maintainer of the target device, and may be in the form of text, voice, and / or image.
[0035] In some embodiments, the fault type set may include several fault types, which may be defined by summarizing historical maintenance data.
[0036] In some embodiments, different types of target devices have different sets of failure types.
[0037] Step 12: Generate prompt words based on the maintenance data and the fault type set.
[0038] Since a large language model is required in this embodiment, the large language model is a generative large language model. A prompt word can be input into the large language model to guide the reasoning of the large language model.
[0039] Prompt words are the actual input data provided by the user to the large language model, which are used to guide the large language model to generate corresponding outputs. For example, this application can first process the maintenance data, the fault type set, and some output precautions into prompt words, and optimize the prompt words in the early stage of using the large language model to make the output of the large language model more standardized and closer to the judgment logic of the actual business. After the use of the large language model tends to be stable, the prompt words, except for the input maintenance data, the fault type set can remain unchanged.
[0040] In some embodiments, the maintenance data is processed by dimensionality reduction or dimensionality increase to obtain target maintenance data; prompt words are generated based on the target maintenance data and the set of fault types. For example, if the description of the fault phenomenon in the maintenance data contains too much content, it can be processed by dimensionality reduction to retain the key content in the description of the fault phenomenon. For example, if the description of the fault phenomenon in the maintenance data contains too little content, it can be processed by dimensionality increase to increase the content of the description of the fault phenomenon for the convenience of subsequent classification. For example, if the description of the maintenance data contains too much content, it can be processed by dimensionality reduction to retain the key content in the description of the maintenance. For example, if the description of the maintenance data contains too little content, it can be processed by dimensionality increase to increase the content of the description of the maintenance for the convenience of subsequent classification.
[0041] Step 13: Input the prompt word into the large language model to obtain the fault type classification corresponding to the maintenance data.
[0042] The prompt words are input into the large language model. The large language model infers the content of the prompt words based on its internal network structure to obtain the fault type classification corresponding to the maintenance data.
[0043] In this embodiment, the large language model can be a GPT model. GPTGPT is a generative pre-training model based on the Transformer architecture, and its core components include: Self-Attention Mechanism, Multi-Head Attention, Feed-Forward Neural Network, and Positional Encoding.
[0044] Self-Attention Mechanism: The self-attention mechanism allows the model to focus on all other words in the input sequence when processing each word. This enables the model to capture long-range dependencies and contextual information.
[0045] Multi-Head Attention: By using multiple attention heads in parallel, the model can extract information from different subspaces, improving the expressiveness of the model.
[0046] Feed-Forward Neural Network: Each attention layer is followed by a feed-forward neural network for further processing and transformation of information.
[0047] Positional Encoding: Since the Transformer architecture does not contain loops or convolutional structures, positional encoding is used to add position information to each word in the input sequence, allowing the model to recognize the order of words.
[0048] In the classification processing of maintenance data, these core components enable the GPT model to perform structured processing on non-professional data (maintenance data) based on the input context and word order, and the supervised fine-tuning process ensures the consistency of the GPT model output.
[0049] In some embodiments, see Figure 2 , the construction of the large language model is as follows:
[0050] When the prompt word is input into the large language model, the text and position embedding is first performed in the text embedding layer of the large language model to obtain the corresponding position encoding data. Then the position encoding data is input into the masked multi-head self-attention layer to obtain the first encoding feature. Then the first encoding feature and the position encoding data are fused to obtain the second encoding feature. The second encoding feature is input into the first normalization layer for layer normalization to obtain the third encoding feature. Then the third encoding feature is input into the feedforward neural network to obtain the fourth encoding feature. The fourth encoding feature and the third encoding feature are fused to obtain the fifth encoding feature. Then the fifth encoding feature is input into the second normalization layer for layer normalization to obtain the sixth encoding feature. Then the sixth encoding feature is input into the text prediction and task classification layer for text prediction and task classification to obtain the fault type classification corresponding to the maintenance data.
[0051] In this embodiment, the generative characteristics of the large language model are used to perform structured fault type classification on unstructured maintenance data, thereby improving the classification efficiency of maintenance data and reducing labor costs.
[0052] See also Figure 3 , Figure 3 1 is a flow chart of another embodiment of the maintenance data classification method provided by the present application. The classification method includes:
[0053] Step 31: Obtain maintenance data of the target device and obtain a set of fault types corresponding to the target device.
[0054] The maintenance data includes at least one of a fault phenomenon description and a maintenance description.
[0055] Step 32: Generate a task description based on the maintenance data and the fault type set; wherein the task description is used to guide the reasoning logic of the large language model.
[0056] In some embodiments, it is considered that the maintenance data is prone to the following phenomena:
[0057] The first phenomenon is that the maintenance data only includes the description of the fault phenomenon.
[0058] The second phenomenon is that the maintenance data only includes the maintenance description.
[0059] The third phenomenon is that the maintenance data includes the fault phenomenon description and the maintenance description.
[0060] Therefore, a large language model is required to be able to classify the maintenance data when any of the above three phenomena occur in the maintenance data.
[0061] Based on this, when generating a task description, a first reasoning logic is generated for maintenance data that only includes a fault phenomenon description; a second reasoning logic is generated for maintenance data that only includes a maintenance description; a third reasoning logic is generated for maintenance data that includes a fault phenomenon description and a maintenance description; and a task description is generated based on the first reasoning logic, the second reasoning logic, the third reasoning logic, the maintenance data, and the fault type set.
[0062] For example, when the maintenance data only includes a fault phenomenon description, the corresponding first reasoning logic is: if there is no maintenance description in the maintenance data, then the fault type classification corresponding to the maintenance data is determined based on the fault phenomenon description and the fault type set.
[0063] For example, when the maintenance data only includes a maintenance description, the corresponding second reasoning logic is: if there is no fault phenomenon description in the maintenance data, then the fault type classification corresponding to the maintenance data is determined based on the maintenance description and the fault type set.
[0064] For example, when the maintenance data includes a fault phenomenon description and a maintenance description, the corresponding third reasoning logic is: if the maintenance data includes a fault phenomenon description and a maintenance description, then the fault type classification corresponding to the maintenance data is determined based on the fault phenomenon description, maintenance description and fault type set.
[0065] In some embodiments, the above-mentioned reasoning logic can be generated using a large language model.
[0066] In some embodiments, the above-mentioned reasoning logic can be generated using a large language model, and after being generated, it can be manually adjusted to form the final reasoning logic to improve the accuracy of the reasoning logic.
[0067] Step 33: Generate prompt words based on the maintenance data, the fault type set and the task description.
[0068] In some embodiments, the task description, maintenance data, and fault type set are spliced in order to generate prompt words.
[0069] Step 34: Input the prompt word into the large language model to obtain the fault type classification corresponding to the maintenance data.
[0070] In some embodiments, the maintenance data includes at least one of the fault phenomenon description and the maintenance description. The maintenance data also includes maintenance parts. Based on this, the maintenance data is prone to the following phenomena:
[0071] The first phenomenon is that the maintenance data only includes the description of the fault phenomenon.
[0072] The second phenomenon is that the maintenance data only includes the maintenance description.
[0073] The third phenomenon is that the maintenance data includes the fault phenomenon description and the maintenance description.
[0074] The fourth phenomenon is that the maintenance data includes maintenance parts and fault phenomenon description and / or maintenance description.
[0075] Therefore, a large language model is required to be able to classify the maintenance data when any of the above four phenomena occur in the maintenance data.
[0076] Based on this, when generating a task description, a first reasoning logic is generated for maintenance data that only includes a description of a fault phenomenon; a second reasoning logic is generated for maintenance data that only includes a maintenance description; a third reasoning logic is generated for maintenance data that includes a description of a fault phenomenon and a maintenance description; a fourth reasoning logic is generated for maintenance data that includes maintenance parts; and a task description basis is generated based on the first reasoning logic, the second reasoning logic, the third reasoning logic, the fourth reasoning logic, the maintenance data, and the fault type set.
[0077] For example, when the maintenance data includes maintenance parts and a description of a fault phenomenon and / or a maintenance description, the corresponding fourth reasoning logic is: if there is no maintenance description in the maintenance data, but there are maintenance parts and a description of a fault phenomenon, then the fault type classification corresponding to the maintenance data is determined by the maintenance parts, the description of the fault phenomenon and the set of fault types. If there is no description of a fault phenomenon in the maintenance data, but there are maintenance parts and a maintenance description, then the fault type classification corresponding to the maintenance data is determined by the maintenance parts, the maintenance description and the set of fault types. If there is a description of a fault phenomenon, maintenance parts and a maintenance description in the maintenance data, then the fault type classification corresponding to the maintenance data is determined by the description of the fault phenomenon, maintenance parts, the maintenance description and the set of fault types.
[0078] Then, prompt words are generated based on the maintenance data, the set of fault types and the task description. The prompt words are input into the large language model to obtain the fault type classification corresponding to the maintenance data.
[0079] In some embodiments, the fault type set includes a number of first-level fault types; wherein each first-level fault type includes a number of second-level fault types, and the second-level fault types are subdivisions of the first-level fault types.
[0080] In some embodiments, the fault phenomenon of some maintenance data is newly appeared and does not appear in the fault type set, or the large language model cannot classify it. Therefore, in response to the fault type classification indicating that the classification is not completed, the maintenance data is sent to the manual terminal; the fault type classification corresponding to the maintenance data fed back by the manual terminal is received. That is, the large language model is assisted to complete by means of the manual terminal, and if the fault type classification corresponding to the maintenance data fed back by the manual terminal does not exist in the fault type set, the fault type classification is added to the fault type set to increase the fault types in the fault type set and improve the subsequent classification accuracy.
[0081] In an application scenario, take the target device as a dishwasher as an example:
[0082] Get the maintenance data corresponding to the dishwasher. The maintenance data includes the fault phenomenon description, maintenance description and maintenance parts. The fault phenomenon description is: the power button cannot be pressed and the touch is invalid. The maintenance description is: power failure / replace the panel PCB.
[0083] Obtain a set of fault types corresponding to the dishwasher, which includes a first fault type: machine water leakage, abnormal heating, abnormal water inflow, and defective electric control board.
[0084] Among them, the second-level fault types corresponding to machine leakage may include LC alarm, washing pump, connecting pipe, drain pipe, drain pipe assembly, thermistor, respirator assembly, spray arm assembly, water leakage in front of the door, door assembly, water inlet pipe reconnection, water inlet pipe assembly, water inlet valve, bottom leakage, distributor, front leakage, outer door assembly, rear leakage, ground leakage, distributor assembly, safety water inlet pipe, water cup assembly, turbidity sensor, water diversion valve assembly, inner door assembly, pump, sealing strip, hot air drying system, main control board, overflow device, and unclear description.
[0085] The second-level fault types corresponding to heating abnormalities may include HC alarm, thermistor, TC alarm, washing pump, machine tripping, electric control board, drain pump, no heating, HC2 alarm, and unclear description.
[0086] The second-level fault types corresponding to abnormal water inlet may include 4C alarm, faulty electronic control board, drain pump, foreign matter cleaning, water inlet valve, water inlet pipe, water inlet pipe reconnection, display panel, safety water inlet pipe, water inlet pipe assembly, drain pipe, respirator assembly, safety water inlet pipe, 1C alarm, water inflow all the time, and unclear description.
[0087] The second-level fault types corresponding to a bad electronic control board may include power failure, bad main control board, bad buttons, bad display, bad touch, moisture penetration, small function board, washing pump, no sound, bad keypad, no sound from keys, and unclear description.
[0088] Generate prompt words based on maintenance data and fault type sets, and the prompt words contain corresponding task descriptions to guide the output of the large language model.
[0089] For example, the prompt word can be: I now need you to read two descriptions about maintenance. After reading, based on the description of the fault phenomenon, the maintenance description, the maintenance parts, and the professional knowledge I provide (a set of fault types), first determine which of the first fault types in the fault phenomenon description, the maintenance description, the maintenance parts, and the professional knowledge is related, and then determine which of the second fault types under the first fault type corresponds. Then output the determined first fault type and second fault type. With a blank interval. Note: The second fault type must be a subdivision item under the first fault type. If no clear subdivision item can be found, you can choose unclear description. The output format is as follows: "LC alarm for machine water leakage". If the first description below is empty, the judgment only refers to the second description and professional knowledge. If the second description has no content, the judgment only refers to the first description and professional knowledge. If the maintenance parts are empty, the judgment only refers to the fault phenomenon description and maintenance description in the first and second descriptions, as well as professional knowledge. If there is no corresponding subdivision item after determining the first fault type, the output corresponding to the subdivision item is the first fault type. Because a single piece of maintenance data comes from the same maintenance event, it is sufficient to output the corresponding first fault type and second fault type, and there is no need to output others. The two descriptions are: The first paragraph is: The power button cannot be pressed and the touch is invalid. The second paragraph is: Power failure / replace panel PCB. The repair part is the electric control board. Professional knowledge is: The second-level fault type corresponding to the first fault type machine leakage may include LC alarm, washing pump, connecting pipe, drain pipe, drain pipe assembly, thermistor, respirator assembly, spray arm assembly, front door leakage, door assembly, water inlet pipe reconnection, water inlet pipe assembly, water inlet valve, bottom leakage, distributor, front leakage, outer door assembly, rear leakage, ground leakage, distributor assembly, safety water inlet pipe, water cup assembly, turbidity sensor, water diversion valve assembly, inner door assembly, pump, sealing strip, hot air drying system, main control board, overflow device, unclear description. The second-level fault type corresponding to the first fault type heating abnormality may include HC alarm, thermistor, TC alarm, washing pump, machine tripping, electric control board, drain pump, no heating, HC2 alarm, unclear description. The second level fault types corresponding to the first fault type of abnormal water inflow may include 4C alarm, faulty electric control board, drain pump, foreign matter cleaning, water inlet valve, water inlet pipe, water inlet pipe reconnection, display board, safety water inlet pipe, water inlet pipe assembly, drain pipe, respirator assembly, safety water inlet pipe, 1C alarm, water inflow all the time, and unclear description. The second level fault types corresponding to the first fault type of faulty electric control board may include power failure, faulty main control board, faulty buttons, poor display, poor touch, moisture penetration, small function board, washing pump, no sound, faulty keypad, no sound from keys, and unclear description.
[0090] Input this prompt word into the large language model to get the output of the large language model. For example, the maintenance data is a description of the fault phenomenon: the power button cannot be pressed and the touch is invalid; the maintenance description is: power failure / replace the panel PCB. The maintenance part is the electric control board. The corresponding output classification of the large language model is the bad power failure of the electric control board.
[0091] In one application scenario, since most of the content in the prompt words is unified, a corresponding prompt word template can be set to leave the content that needs to be changed blank. When it is necessary to classify the maintenance data, the content of the maintenance data is filled into the blank content to form the prompt words corresponding to the maintenance data, so as to speed up the generation efficiency of the prompt words and increase the efficiency of fault type classification as a whole. For example, the fault type set and task description constitute a prompt word template, and a blank space for maintenance data is reserved in the prompt word template. When classifying the maintenance data into fault types, fill the blank space with maintenance data to obtain the corresponding prompt words. When a large amount of maintenance data needs to be classified for faults, prompt words can be automatically generated based on the maintenance data and the prompt word template, and input into the large language model to obtain the corresponding fault type classification.
[0092] In an application scenario, corresponding prompt word templates can be set for different types of equipment. When classifying the fault type of each type of equipment, the corresponding prompt word template is selected according to the type. The content of the maintenance data corresponding to the type of equipment is filled into the blank content to form the prompt word corresponding to the maintenance data, thereby speeding up the efficiency of prompt word generation and thus increasing the overall efficiency of fault type classification.
[0093] In this embodiment, the generative characteristics of the large language model are used to input prompt words into the large language model to achieve structured fault type classification of unstructured maintenance data, thereby improving the classification efficiency of maintenance data and reducing labor costs.
[0094] In an application scenario, the fault phenomenon description and / or the maintenance description may be in the form of speech. When generating the prompt word, speech recognition is performed on the fault phenomenon description and / or the maintenance description to obtain corresponding text content, and then generate the prompt word.
[0095] In one application scenario, a large language model with speech recognition can be used, that is, the fault phenomenon description and / or maintenance description in speech form can be directly used to generate prompt words.
[0096] In an application scenario, the fault phenomenon description and / or the maintenance description may be in the form of an image. A large language model with image recognition may be used. That is, the fault phenomenon description and / or the maintenance description in the form of an image may be directly used to generate prompt words.
[0097] In some embodiments, the large language model may have text, voice, and image recognition functions simultaneously, so as to achieve fault type classification of maintenance data in multiple formats.
[0098] In one application scenario, the equipment manufacturer can collect statistics on the fault type classification corresponding to the maintenance data within a preset period, and adjust the production process, process, materials, etc. of the subsequent equipment based on these fault type classifications to reduce the corresponding equipment failures.
[0099] See also Figure 4 , Figure 4 1 is a schematic diagram of the structure of an embodiment of an electronic device provided by the present application. The electronic device 40 includes a processor 41 and a memory 42 coupled to the processor 41; wherein the memory 42 is used to store a computer program, and the processor 41 is used to execute the computer program to implement the following method:
[0100] Obtain maintenance data of a target device and a set of fault types corresponding to the target device; wherein the maintenance data includes at least one of a fault phenomenon description and a maintenance description; generate prompt words based on the maintenance data and the set of fault types; input the prompt words into a large language model to obtain a fault type classification corresponding to the maintenance data.
[0101] In some embodiments, the processor 41 is further configured to execute a computer program to implement the following method:
[0102] A task description is generated based on the maintenance data and the fault type set; wherein the task description is used to guide the reasoning logic of the large language model; and a prompt word is generated based on the maintenance data, the fault type set and the task description.
[0103] In some embodiments, the processor 41 is further configured to execute a computer program to implement the following method:
[0104] Generate a first reasoning logic for maintenance data that only includes a description of a fault phenomenon; generate a second reasoning logic for maintenance data that only includes a maintenance description; generate a third reasoning logic for maintenance data that includes a description of a fault phenomenon and a maintenance description; generate a task description based on the first reasoning logic, the second reasoning logic, the third reasoning logic, the maintenance data, and a set of fault types.
[0105] In some embodiments, the maintenance data also includes maintenance parts, and the processor 41 is further configured to execute a computer program to implement the following method:
[0106] Generate a fourth reasoning logic for maintenance data including maintenance parts; generate a task description based on the first reasoning logic, the second reasoning logic, the third reasoning logic, the maintenance data and all fault types, including: generate a task description based on the first reasoning logic, the second reasoning logic, the third reasoning logic, the fourth reasoning logic, the maintenance data and a set of fault types.
[0107] In some embodiments, the processor 41 is further configured to execute a computer program to implement the following method:
[0108] The task description, maintenance data and fault type set are spliced in order to generate prompt words.
[0109] In some embodiments, the fault type set includes a number of first-level fault types; wherein each first-level fault type includes a number of second-level fault types, and the second-level fault types are subdivisions of the first-level fault types.
[0110] In some embodiments, the processor 41 is further configured to execute a computer program to implement the following method:
[0111] The maintenance data is processed by dimensionality reduction or dimensionality increase to obtain target maintenance data; prompt words are generated based on the target maintenance data and the fault type set.
[0112] In some embodiments, after the prompt word is input into the large language model to obtain the fault type classification corresponding to the maintenance data, the processor 41 is further used to execute a computer program to implement the following method:
[0113] In response to the fault type classification indicating that the classification is not completed, the maintenance data is sent to the manual terminal; and the manual terminal feedbacks the fault type classification corresponding to the maintenance data.
[0114] In some embodiments, the processor 41 is further configured to execute a computer program to implement the method of any of the above embodiments.
[0115] See also Figure 5 , Figure 5 1 is a schematic diagram of a computer readable storage medium according to an embodiment of the present application. The computer readable storage medium 50 is used to store a computer program 51. When the computer program 51 is executed by a processor, it is used to implement the following method:
[0116] Obtain maintenance data of a target device and a set of fault types corresponding to the target device; wherein the maintenance data includes at least one of a fault phenomenon description and a maintenance description; generate prompt words based on the maintenance data and the set of fault types; input the prompt words into a large language model to obtain a fault type classification corresponding to the maintenance data.
[0117] In some embodiments, when the computer program 51 is executed by a processor, it is also used to implement the method of any of the above embodiments.
[0118] In summary, the maintenance data classification method, electronic device, and computer-readable storage medium provided in the present application utilize the generative characteristics of a large language model to perform structured fault type classification on unstructured maintenance data, thereby improving the classification efficiency of maintenance data and reducing labor costs.
[0119] In the several embodiments provided in this application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device implementation described above is only illustrative, for example, the division of the modules or units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.
[0120] If the integrated units in the above other embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processing circuit component (processor) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, RandomAccess Memory), disk or optical disk and other media that can store program code.
[0121] The above description is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.
Claims
1. A maintenance data classification method, characterized in that: The classification method includes: Acquire maintenance data of a target device and acquire a set of fault types corresponding to the target device; wherein the maintenance data includes at least one of a fault phenomenon description and a maintenance description; Generate a prompt word based on the maintenance data and the fault type set; The prompt word is input into a large language model to obtain a fault type classification corresponding to the maintenance data.
2. The classification method according to claim 1, characterized in that: The generating of prompt words based on the maintenance data and the fault type set includes: Generate a task description based on the maintenance data and the set of fault types; wherein the task description is used to guide the reasoning logic of the large language model; The prompt word is generated based on the maintenance data, the fault type set and the task description.
3. The classification method according to claim 2, characterized in that: The generating a task description based on the maintenance data and the set of fault types includes: generating a first reasoning logic for the maintenance data including only the description of the fault phenomenon; generating a second reasoning logic for the maintenance data including only the maintenance description; generating a third reasoning logic for the maintenance data including the fault phenomenon description and the maintenance description; The task description is generated according to the first reasoning logic, the second reasoning logic, the third reasoning logic, the maintenance data and the fault type set.
4. The classification method according to claim 3, characterized in that: The maintenance data also includes maintenance components, and the generating of a task description based on the maintenance data and the set of fault types includes: generating fourth reasoning logic for the maintenance data including the maintenance component; The generating the task description according to the first reasoning logic, the second reasoning logic, the third reasoning logic, the maintenance data and all the fault types includes: The task description is generated according to the first reasoning logic, the second reasoning logic, the third reasoning logic, the fourth reasoning logic, the maintenance data, and the fault type set.
5. The classification method according to claim 2, characterized in that: The generating the prompt word based on the maintenance data, the fault type set and the task description includes: The task description, the maintenance data and the fault type set are spliced in order to generate the prompt word.
6. The classification method according to claim 1, characterized in that: The fault type set includes a plurality of first-level fault types; wherein each of the first-level fault types includes a plurality of second-level fault types, and the second-level fault types are subdivisions of the first-level fault types.
7. The classification method according to claim 1, characterized in that: The generating of prompt words based on the maintenance data and the fault type set includes: Performing dimension reduction or dimension increase processing on the maintenance data to obtain target maintenance data; The prompt word is generated based on the target maintenance data and the fault type set.
8. The classification method according to claim 1, characterized in that: After the prompt word is input into the large language model to obtain the fault type classification corresponding to the maintenance data, the method includes: In response to the fault type classification indicating that classification is not completed, sending the maintenance data to a manual terminal; Receive the fault type classification corresponding to the maintenance data fed back by the manual terminal.
9. An electronic device, characterized in that: The electronic device includes a processor and a memory coupled to the processor; The memory is used to store a computer program, and the processor is used to execute the computer program to implement the classification method according to any one of claims 1 to 8.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, it is used to implement the classification method according to any one of claims 1 to 8.