Equipment fault processing answer determination method and device, computer equipment, storage medium and computer program product

By obtaining the word vector of equipment failure question information and the embedded vector of the equipment maintenance knowledge base, and combining the language model, the problem of low accuracy in determining equipment failure answers is solved, achieving more efficient and accurate troubleshooting answers. .

CN118982016APending Publication Date: 2024-11-19GUANGZHOU MINO AUTOMOTIVE EQUIP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410926363.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

The prior art has low accuracy in determining the answers to equipment fault handling, which is mainly due to relying on the experience of experts and subjective factors are prone to errors.

Method used

By obtaining the equipment failure question information, determining its corresponding word vector, and comparing the maximum similarity with the preset embed vector in the pre-constructed device maintenance knowledge base, obtaining the answer from the knowledge base when the preset similarity is met. Use pre-trained language models to determine the answer when similarity is not met.

Benefits of technology

It improves the accuracy of determining the answers to equipment fault handling, reduces the dependence on expert experience, avoids errors caused by subjective factors, and improves the efficiency and accuracy of fault handling.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982016A_ABST
    Figure CN118982016A_ABST
Patent Text Reader

Abstract

The invention relates to an equipment fault processing answer determination method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring equipment fault question information associated with to-be-analyzed equipment; determining a word vector corresponding to the equipment fault question information; determining the maximum similarity between the word vector and a preset embedded vector in a pre-constructed equipment maintenance knowledge base; under the condition that the maximum similarity is greater than a preset similarity, determining a preset embedding vector corresponding to the maximum similarity as a target embedding vector, and determining answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as target answer information corresponding to the equipment fault question information; and under the condition that the maximum similarity is smaller than or equal to a preset similarity, determining target answer information corresponding to the equipment fault question information through a pre-trained language model. By adopting the method, the determination accuracy of the equipment fault processing answer can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method and apparatus for determining an answer to a device fault processing, a computer device, a computer-readable storage medium, and a computer program product. Background Art

[0002] At present, in the production process, when equipment failure occurs, it is crucial to accurately determine the answer to the equipment failure.

[0003] In traditional technology, when determining the answer to equipment fault handling, it mainly relies on the experience of experts; however, this method has subjective factors and is prone to errors, resulting in low accuracy in determining the answer to equipment fault handling. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, computer device, computer-readable storage medium and computer program product for determining an equipment fault handling answer that can improve the accuracy of determining the equipment fault handling answer in response to the above technical problems.

[0005] In a first aspect, the present application provides a method for determining an answer to a device fault processing, comprising:

[0006] Obtain equipment fault question information associated with the equipment to be analyzed;

[0007] Determine a word vector corresponding to the equipment failure question information;

[0008] Determining the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base;

[0009] In the case where the maximum similarity is greater than a preset similarity, determining a preset embedding vector corresponding to the maximum similarity as a target embedding vector, and determining answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as target answer information corresponding to the equipment fault question information;

[0010] When the maximum similarity is less than or equal to the preset similarity, target answer information corresponding to the equipment fault question information is determined through a pre-trained language model.

[0011] In one embodiment, the pre-built equipment maintenance knowledge base is built by the following method:

[0012] Acquire sample fault record information of a sample device and sample historical operation status information of the sample device;

[0013] Obtaining a constructed equipment maintenance knowledge base based on the sample fault record information and the sample historical operation status information;

[0014] According to the preset maintenance information associated with the sample device, the constructed equipment maintenance knowledge base is adjusted to obtain the pre-constructed equipment maintenance knowledge base.

[0015] In one embodiment, obtaining sample fault record information of sample devices includes:

[0016] Obtaining sample fault document information of the sample device;

[0017] Parsing the sample fault document information to obtain a parsed document;

[0018] Extracting target fault information from the parsed document;

[0019] According to the target fault information, sample fault record information of the sample device is obtained.

[0020] In one embodiment, obtaining sample fault record information of a sample device according to the target fault information includes:

[0021] Segmenting the target fault information to obtain multiple text blocks;

[0022] Performing fusion processing on each of the text blocks to obtain a fused text block;

[0023] Extracting key information from the fused text block;

[0024] Performing fusion processing on the key information to obtain fused key information;

[0025] Determining a preset embedding vector corresponding to the fusion key information;

[0026] According to the preset embedding vector and each of the text blocks, sample fault record information of the sample device is obtained.

[0027] In one embodiment, the pre-trained language model is trained by the following method:

[0028] Inputting the target fault information into the language model to be trained to obtain predicted answer information corresponding to the target fault information; the predicted answer information is used to represent the predicted recommended measures information corresponding to the target fault information;

[0029] According to the difference between the predicted answer information and the actual answer information corresponding to the target fault information, the language model to be trained is iteratively trained to obtain the pre-trained language model; the actual answer information is used to represent the actual recommended measures information corresponding to the target fault information.

[0030] In one embodiment, determining answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as target answer information corresponding to the equipment fault question information includes:

[0031] Determining a target fault identifier corresponding to the target embedding vector from the equipment maintenance knowledge base;

[0032] Determine the recommended action information corresponding to the target fault identifier as the recommended action information corresponding to the equipment fault question information, and determine the historical operation status information corresponding to the target fault identifier as the historical operation status information corresponding to the equipment fault question information;

[0033] The suggested measures information and the historical operation status information corresponding to the equipment failure question information are both used as the target answer information corresponding to the equipment failure question information.

[0034] In one embodiment, determining the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base includes:

[0035] Determining the similarity between the word vector and a plurality of preset embedding vectors in a pre-built equipment maintenance knowledge base;

[0036] From the similarities, the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base is determined.

[0037] In a second aspect, the present application also provides a device for determining an answer to a device fault processing, comprising:

[0038] A question information acquisition module is used to acquire device fault question information associated with the device to be analyzed;

[0039] An information vector determination module, used to determine a word vector corresponding to the equipment failure question information;

[0040] A maximum similarity determination module, used to determine the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base;

[0041] A first answer determination module is used to determine, when the maximum similarity is greater than a preset similarity, a preset embedding vector corresponding to the maximum similarity as a target embedding vector, and determine answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as target answer information corresponding to the equipment fault question information;

[0042] The second answer determination module is used to determine the target answer information corresponding to the equipment fault question information through a pre-trained language model when the maximum similarity is less than or equal to the preset similarity.

[0043] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0044] Obtain equipment fault question information associated with the equipment to be analyzed;

[0045] Determine a word vector corresponding to the equipment failure question information;

[0046] Determining the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base;

[0047] In the case where the maximum similarity is greater than a preset similarity, determining a preset embedding vector corresponding to the maximum similarity as a target embedding vector, and determining answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as target answer information corresponding to the equipment fault question information;

[0048] When the maximum similarity is less than or equal to the preset similarity, target answer information corresponding to the equipment fault question information is determined through a pre-trained language model.

[0049] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0050] Obtain equipment fault question information associated with the equipment to be analyzed;

[0051] Determine a word vector corresponding to the equipment failure question information;

[0052] Determining the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base;

[0053] In the case where the maximum similarity is greater than a preset similarity, determining a preset embedding vector corresponding to the maximum similarity as a target embedding vector, and determining answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as target answer information corresponding to the equipment fault question information;

[0054] When the maximum similarity is less than or equal to the preset similarity, target answer information corresponding to the equipment fault question information is determined through a pre-trained language model.

[0055] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0056] Obtain equipment fault question information associated with the equipment to be analyzed;

[0057] Determine a word vector corresponding to the equipment failure question information;

[0058] Determining the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base;

[0059] In the case where the maximum similarity is greater than a preset similarity, determining a preset embedding vector corresponding to the maximum similarity as a target embedding vector, and determining answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as target answer information corresponding to the equipment fault question information;

[0060] When the maximum similarity is less than or equal to the preset similarity, target answer information corresponding to the equipment fault question information is determined through a pre-trained language model.

[0061] The above-mentioned equipment fault handling answer determination method, device, computer device, storage medium and computer program product first obtains equipment fault question information associated with the equipment to be analyzed, then determines the word vector corresponding to the equipment fault question information, and then determines the maximum similarity between the word vector and the preset embedding vector in the pre-constructed equipment maintenance knowledge base. When the maximum similarity is greater than the preset similarity, the preset embedding vector corresponding to the maximum similarity is determined as the target embedding vector, and the answer information corresponding to the target embedding vector is determined from the equipment maintenance knowledge base as the target answer information corresponding to the equipment fault question information. When the maximum similarity is less than or equal to the preset similarity, the target answer information corresponding to the equipment fault question information is determined through a pre-trained language model. In this way, when determining the answer to equipment fault handling, by converting the equipment fault question information into a word vector and comparing it with the preset embedding vector in the equipment maintenance knowledge base, the equipment maintenance knowledge base and the language model can be used to quickly determine the answer information that matches the equipment fault question information, which is beneficial to improving the efficiency of determining the answer to equipment fault handling; moreover, the entire process does not rely entirely on the experience of experts, avoiding the defect that the method that relies solely on the experience of experts has subjective factors and is prone to errors, resulting in a low accuracy in determining the answer to equipment fault handling, thereby improving the accuracy of determining the answer to equipment fault handling. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments of the present application or the related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0063] Figure 1 A schematic diagram of a flow chart of a method for determining an answer to a device fault processing in one embodiment;

[0064] Figure 2 A schematic diagram of a flow chart of steps for constructing a pre-built equipment maintenance knowledge base in one embodiment;

[0065] Figure 3 A flowchart of steps for obtaining sample fault record information of a sample device in one embodiment;

[0066] Figure 4 A schematic diagram of part of the information in a maintenance manual for equipment failure in one embodiment;

[0067] Figure 5 A schematic diagram of part of the information in a maintenance manual for equipment failure in another embodiment;

[0068] Figure 6 A schematic flow chart of a method for determining an answer to a device failure processing in another embodiment;

[0069] Figure 7 It is a flowchart of an intelligent question-answering system and method for equipment failure maintenance based on retrieval enhancement generation in one embodiment;

[0070] Figure 8 A structural block diagram of a device for determining an answer to a device failure processing in one embodiment;

[0071] Fig. 9 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0073] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0074] In an exemplary embodiment, Figure 1 As shown, a method for determining an answer to a device fault processing is provided. This embodiment uses the method applied to a server as an example for illustration; it is understandable that the method can also be applied to a terminal, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. Among them, the terminal can be, but is not limited to, various personal computers, laptops, smart phones, and tablet computers; the server can be implemented as an independent server or a server cluster composed of multiple servers. In this embodiment, the method includes the following steps:

[0075] Step S101, obtaining device fault question information associated with the device to be analyzed.

[0076] The device to be analyzed refers to the device that needs to be analyzed for faults, such as a robot.

[0077] The equipment failure question information refers to the question information related to equipment failure maintenance.

[0078] Exemplarily, the server receives question information related to equipment failure maintenance of the device to be analyzed sent by the user through the question-answer interaction interface in the cloud platform as equipment failure question information associated with the device to be analyzed.

[0079] Step S102, determining the word vector corresponding to the equipment failure question information.

[0080] The word vector refers to the feature vector corresponding to the equipment failure question information.

[0081] Exemplarily, the server inputs the equipment failure question information into the feature extraction model, and vectorizes the equipment failure question information through the feature extraction model to obtain a feature vector corresponding to the equipment failure question information as a word vector corresponding to the equipment failure question information.

[0082] Step S103, determining the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base.

[0083] Among them, the equipment maintenance knowledge base refers to a database that stores preset embedding vectors and fault record information.

[0084] The preset embedding vector refers to a pre-set embedding vector, such as [0.12089, 0.18089, 0.4656, ...].

[0085] The maximum similarity refers to the maximum value corresponding to the similarity between the word vector and the preset embedding vector.

[0086] Exemplarily, the server determines a maximum value corresponding to the similarity between the word vector and the preset embedding vector as the maximum similarity between the word vector and the preset embedding vector in the pre-built equipment maintenance knowledge base.

[0087] Step S104, when the maximum similarity is greater than the preset similarity, determine the preset embedding vector corresponding to the maximum similarity as the target embedding vector, and determine the answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as the target answer information corresponding to the equipment fault question information.

[0088] The preset similarity refers to a pre-set similarity. It should be noted that the preset similarity may be determined according to circumstances.

[0089] Among them, the target embedding vector refers to the preset embedding vector corresponding to the maximum similarity.

[0090] The answer information corresponding to the target embedding vector refers to the answer information output by the equipment maintenance knowledge base, including the recommended measures information corresponding to the equipment failure question information and the historical operating status information of the equipment to be analyzed.

[0091] Exemplarily, when the maximum similarity is greater than a preset similarity, the server determines a preset embedding vector corresponding to the maximum similarity, and uses the preset embedding vector as the target embedding vector; then, the server determines the answer information corresponding to the target embedding vector from the equipment maintenance knowledge base, and uses the answer information as the target answer information corresponding to the equipment fault question information.

[0092] Step S105 , when the maximum similarity is less than or equal to the preset similarity, the target answer information corresponding to the equipment fault question information is determined through a pre-trained language model.

[0093] Among them, language models refer to large language models, such as ChatGPT (Chat Generative Pre-trainedTransformer), Kimi (a large language model), Wenxinyiyan, etc.

[0094] Exemplarily, when the maximum similarity is less than or equal to a preset similarity, the server inputs the equipment failure question information into a pre-trained language model, and determines answer information corresponding to the equipment failure question information through the pre-trained language model as target answer information corresponding to the equipment failure question information.

[0095] In the above-mentioned method for determining the answer to the equipment fault processing, the equipment fault question information associated with the equipment to be analyzed is first obtained, and then the word vector corresponding to the equipment fault question information is determined. Then, the maximum similarity between the word vector and the preset embedding vector in the pre-constructed equipment maintenance knowledge base is determined. When the maximum similarity is greater than the preset similarity, the preset embedding vector corresponding to the maximum similarity is determined as the target embedding vector, and the answer information corresponding to the target embedding vector is determined from the equipment maintenance knowledge base as the target answer information corresponding to the equipment fault question information. When the maximum similarity is less than or equal to the preset similarity, the target answer information corresponding to the equipment fault question information is determined by the pre-trained language model. In this way, when determining the answer to the equipment fault processing, by converting the equipment fault question information into a word vector and comparing it with the preset embedding vector in the equipment maintenance knowledge base, the answer information matching the equipment fault question information can be quickly determined through the equipment maintenance knowledge base and the language model, which is conducive to improving the efficiency of determining the answer to the equipment fault processing; moreover, the whole process does not completely rely on the experience of experts, avoiding the defects of subjective factors and easy errors in the method that relies only on the experience of experts, resulting in low accuracy in determining the answer to the equipment fault processing, thereby improving the accuracy of determining the answer to the equipment fault processing.

[0096] In an exemplary embodiment, Figure 2As shown, the method for determining the equipment fault handling answer provided by the present application also includes the step of constructing a pre-built equipment maintenance knowledge base, which specifically includes the following steps:

[0097] Step S201: acquiring sample fault record information of a sample device and sample historical operation status information of a sample device.

[0098] Step S202: Obtain a constructed equipment maintenance knowledge base based on sample fault record information and sample historical operation status information.

[0099] Step S203: adjusting the constructed equipment maintenance knowledge base according to the preset maintenance information associated with the sample equipment to obtain a pre-constructed equipment maintenance knowledge base.

[0100] The sample equipment refers to the equipment used to construct the equipment maintenance knowledge base.

[0101] The sample fault record information refers to the fault record information of the sample device, including fault code, fault description, fault explanation, recommended measures, embedded vector, etc. As shown in Table 1, it is a key information record with the fault code "71058".

[0102] Table 1 Key information record for fault code "71058"

[0103]

[0104] The sample historical operation status information refers to the historical operation status information of the sample device. In actual scenarios, the sample historical operation status information refers to the operation status data N (for example, 1) hours before the failure of the sample device. As shown in Table 2, this is the operation status data information before the failure.

[0105] Table 2 Operation status data information before the fault

[0106]

[0107] Among them, the preset maintenance information refers to the expert maintenance experience.

[0108] Exemplarily, in response to a construction instruction for an equipment maintenance knowledge base, the server obtains fault record information of a sample device from the data as sample fault record information, and obtains historical operating status information of the sample device as sample historical operating status information; then, the server stores the sample fault record information in the first table of the vector database, and stores the sample historical operating status information in the second table of the vector database; then, the server obtains a knowledge base for equipment maintenance initialization based on the first table and the second table of the vector database as a constructed equipment maintenance knowledge base; then, the server supplements and corrects the constructed equipment maintenance knowledge base based on preset maintenance information through an information entry interface, and obtains a final equipment maintenance knowledge base as a pre-constructed equipment maintenance knowledge base.

[0109] In this embodiment, by pre-building an equipment maintenance knowledge base, it is convenient in actual applications that after obtaining equipment fault question information associated with the equipment to be analyzed, the target answer information corresponding to the equipment fault question information can be output, thereby providing a comprehensive and accurate reference for equipment maintenance, which is conducive to quickly locating and solving potential problems of the equipment to be analyzed.

[0110] In an exemplary embodiment, Figure 3 As shown, the above step S201, obtaining sample fault record information of the sample device, specifically includes the following steps:

[0111] Step S301: Obtain sample fault document information of a sample device.

[0112] Step S302: parse the sample fault document information to obtain a parsed document.

[0113] Step S303: extract target fault information from the parsed document.

[0114] Step S304: obtaining sample fault record information of the sample device according to the target fault information.

[0115] The sample fault document information refers to the fault document information of the sample equipment, including equipment fault maintenance manual, equipment fault maintenance report and other documents. It should be noted that the sample fault document information can be in the form of pdf, word, ppt, excel, etc. Figure 4 The following is part of the equipment failure maintenance manual pdf. Figure 5 The following is part of the information in the equipment failure maintenance manual excel.

[0116] Among them, parsing processing refers to the corresponding processing process of parsing the sample fault document information, including the processing process of deleting information such as headers and footers.

[0117] The parsed document refers to the document information obtained after parsing the sample fault document information.

[0118] The target fault information refers to the information related to the fault in the parsed document, including fault code, fault description, cause of the fault, recommended measures for fault maintenance, fault description and other information.

[0119] Exemplarily, the server obtains fault document information of a sample device as sample fault document information; then, the server parses the sample fault document information to obtain the parsed sample fault document information as a parsed document; then, the server extracts and processes the parsed document according to a preset fault information identifier (such as fault code, fault description, cause of the fault, recommended measures for fault maintenance, and fault description) to obtain target fault information in the parsed document; finally, the server obtains sample fault record information of the sample device based on the target fault information.

[0120] For example, the server obtains the equipment fault maintenance manual, parses the manual, deletes information such as headers and footers, and extracts effective information on equipment fault maintenance (including fault code, fault description, cause of fault, recommended measures for fault maintenance, fault description, etc.), and then obtains sample fault record information of sample equipment based on the effective information on equipment fault maintenance.

[0121] In this embodiment, through automated parsing processing, key target fault information can be quickly and accurately extracted from a large number of sample fault documents, reducing the time and errors of manual processing, and facilitating the standardized organization of fault information in sample fault documents for subsequent analysis and utilization.

[0122] In an exemplary embodiment, the above step S304 obtains sample fault record information of the sample device based on the target fault information, which specifically includes the following contents: segmenting the target fault information to obtain multiple text blocks; fusing the text blocks to obtain fused text blocks; extracting key information from the fused text blocks; fusing the key information to obtain fused key information; determining a preset embedding vector corresponding to the fused key information; and obtaining the sample fault record information of the sample device based on the preset embedding vector and the text blocks.

[0123] The segmentation process refers to the process of segmenting the target fault information into multiple text blocks.

[0124] The text block is used to represent the text information in the target fault information, for example, the text block device fault code is "71058" and the text block fault description is "communication interruption with the I / O device".

[0125] The fusion process may refer to a splicing process.

[0126] The fused text block refers to a text block obtained by fusion processing of various text blocks.

[0127] The key information refers to the information of corresponding importance in the fused text block.

[0128] The fused key information refers to the fused information obtained by fusion processing the key information.

[0129] Exemplarily, the server performs segmentation processing on the target fault information to obtain text blocks corresponding to multiple target fault information; then, the server performs fusion processing on each text block to obtain a fused text block; for example, the server performs splicing processing on each text block to obtain a fused text block; then, the server determines the importance corresponding to each text information in the fused text block, and filters out text information whose corresponding importance is greater than a preset importance from each text information, and uses these text information as key information in the fused text block; then, the server performs fusion processing on the key information to obtain fused key information; for example, the server performs splicing processing on the key information to obtain fused key information; then, the server inputs the fused key information into a feature extraction model, and vectorizes the fused key information through the feature extraction model to obtain a feature vector corresponding to the fused key information, which is used as a preset embedding vector corresponding to the fused key information; finally, the server combines the preset embedding vector with each text block to obtain sample fault record information of the sample device.

[0130] For example, the server performs text segmentation on the effective information of equipment fault maintenance to obtain multiple text blocks, such as the text block equipment fault code "71058" and the text block fault description "communication interruption with I / O (input / output) device"; then, the server performs string fusion on the text blocks such as the equipment fault code and the fault description to obtain the fused string "71058 Communication interruption with I / O device"; then, the server extracts the keyword information "71058", "I / O", "device", "communication interruption" of the fused string; then, the server fuses the above keywords to obtain the fused keyword "71058 I / O device communication interruption"; then, the server performs vectorization processing on the fused keywords to obtain an embedded vector [0.12089, 0.18089, 0.4656, ...]; then, based on the embedded vector, the text block equipment fault code, the text block fault description, the text block fault cause, the text block fault maintenance recommended measures, the text block fault description, etc., the server obtains a fault record information.

[0131] In this embodiment, by performing various processes such as segmentation, fusion, and extraction on the target fault information, the sample fault record information of the sample device can be accurately obtained; moreover, the entire process does not require human intervention, thus avoiding the defect that errors easily occur in manual operation, resulting in low accuracy of the sample fault record information of the sample device.

[0132] In an exemplary embodiment, the device fault handling answer determination method provided in the present application also includes a training step of a pre-trained language model, which specifically includes the following contents: inputting the target fault information into the language model to be trained to obtain predicted answer information corresponding to the target fault information; the predicted answer information is used to represent the predicted recommended measures information corresponding to the target fault information; based on the difference between the predicted answer information corresponding to the target fault information and the actual answer information, the language model to be trained is iteratively trained to obtain a pre-trained language model; the actual answer information is used to represent the actual recommended measures information corresponding to the target fault information.

[0133] The prediction answer information refers to the prediction recommended measure information corresponding to the target fault information.

[0134] The predicted recommended action information refers to the predicted value of the recommended action information corresponding to the target fault information output by the language model to be trained.

[0135] The actual answer information refers to the actual recommended measure information corresponding to the target fault information.

[0136] The actual recommended measures information refers to the recommended measures information in a pre-built equipment maintenance knowledge base.

[0137] Exemplarily, the server inputs the target fault information into the language model to be trained, and obtains the predicted recommended measures information corresponding to the target fault information through the language model to be trained, as the predicted answer information corresponding to the target fault information; then, the server obtains the actual recommended measures information corresponding to the target fault information, as the actual answer information corresponding to the target fault information; then, the server obtains the loss value based on the difference between the predicted answer information and the actual answer information corresponding to the target fault information; then, the server adjusts the model parameters of the language model to be trained based on the loss value; then, the server re-trains the language model after the model parameters are adjusted until the loss value obtained by the trained language model is less than the loss value threshold, then stops the training, and uses the trained language model as the pre-trained language model.

[0138] In this embodiment, by pre-training the language model, it is convenient to predict the answer information corresponding to the equipment failure question information after obtaining the equipment failure question information in actual applications; moreover, the language model receives new data in each round of iteration, and performs internal improvements and optimizations on the model, so that predictions can be made more effectively, which is beneficial to improving the prediction accuracy of the language model.

[0139] In an exemplary embodiment, the above step S104, determining the answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as the target answer information corresponding to the equipment fault question information, specifically includes the following contents: determining the target fault identification corresponding to the target embedding vector from the equipment maintenance knowledge base; determining the recommended measures information corresponding to the target fault identification as the recommended measures information corresponding to the equipment fault question information, and determining the historical operating status information corresponding to the target fault identification as the historical operating status information corresponding to the equipment fault question information; using the recommended measures information and the historical operating status information corresponding to the equipment fault question information as the target answer information corresponding to the equipment fault question information.

[0140] The target fault identifier refers to the fault identifier corresponding to the target embedding vector, such as a fault code.

[0141] The recommended measures information is used to represent the recommended measures corresponding to the device fault question information, such as "(1) Ensure that the network cable is plugged into the controller; (2) Ensure that the I / O device is powered normally; (3) Ensure that the wiring to the I / O device is correctly connected."

[0142] The historical operation status information refers to the operation status data N (for example, 1) hours before the failure of the device to be analyzed occurs.

[0143] Exemplarily, the server determines the fault identification corresponding to the target embedding vector from the equipment maintenance knowledge base as the target fault identification; then, the server determines the recommended measures information corresponding to the target fault identification as the recommended measures information corresponding to the equipment fault question information; then, the server determines the historical operating status information corresponding to the target fault identification as the historical operating status information corresponding to the equipment fault question information; then, the server constructs an operating status time series data graph corresponding to the equipment fault question information based on the historical operating status information corresponding to the equipment fault question information; finally, the server uses the recommended measures information and the operating status time series data graph corresponding to the equipment fault question information as the target answer information corresponding to the equipment fault question information.

[0144] In this embodiment, by quickly determining the fault identifier corresponding to the target embedded vector from the knowledge base and obtaining the recommended measures and historical operating status information corresponding to the fault identifier, direct and effective guidance is provided for handling equipment faults, saving time for fault troubleshooting and solution formulation, and helping to improve fault handling efficiency.

[0145] In an exemplary embodiment, the above step S103, determining the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base, specifically includes the following contents: determining the similarity between the word vector and multiple preset embedding vectors in the pre-built equipment maintenance knowledge base; from each similarity, determining the maximum similarity between the word vector and the preset embedding vector in the pre-built equipment maintenance knowledge base.

[0146] Here, the similarity refers to the cosine similarity.

[0147] Exemplarily, the server obtains multiple preset embedding vectors in a pre-built equipment maintenance knowledge base, and determines the similarity between the word vector and each preset embedding vector; then, the server determines the maximum similarity from each similarity as the maximum similarity between the word vector and the preset embedding vector in the pre-built equipment maintenance knowledge base.

[0148] In this embodiment, by determining the maximum similarity, the preset embedding vector most relevant to the input word vector can be accurately found, thereby improving the accuracy of information retrieval and matching; moreover, by screening out the maximum similarity from multiple similarities, unnecessary comparisons and calculations are reduced, thereby improving processing speed and efficiency.

[0149] In an exemplary embodiment, Figure 6 As shown, another method for determining an answer to a device fault processing is provided, and the method is applied to a server as an example, and includes the following steps:

[0150] Step S601: Obtain device fault question information associated with the device to be analyzed.

[0151] Step S602: determine the word vector corresponding to the equipment failure question information.

[0152] Step S603, determining the similarity between the word vector and a plurality of preset embedding vectors in a pre-built equipment maintenance knowledge base.

[0153] Step S604: Determine the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base from among the similarities.

[0154] Step S605: When the maximum similarity is greater than a preset similarity, a preset embedding vector corresponding to the maximum similarity is determined as a target embedding vector.

[0155] Step S606: Determine the target fault identifier corresponding to the target embedded vector from the equipment maintenance knowledge base.

[0156] Step S607, determining the recommended action information corresponding to the target fault identification as the recommended action information corresponding to the equipment fault question information, and determining the historical operation status information corresponding to the target fault identification as the historical operation status information corresponding to the equipment fault question information.

[0157] Step S608: The recommended measures information and the historical operation status information corresponding to the equipment failure question information are used as target answer information corresponding to the equipment failure question information.

[0158] Step S609: when the maximum similarity is less than or equal to the preset similarity, the target answer information corresponding to the equipment fault question information is determined by using a pre-trained language model.

[0159] In the above-mentioned method for determining the answer to equipment fault handling, when determining the answer to equipment fault handling, by converting the equipment fault question information into a word vector and comparing it with the preset embedding vector in the equipment maintenance knowledge base, the equipment maintenance knowledge base and the language model can be used to quickly determine the answer information that matches the equipment fault question information, which is beneficial to improving the efficiency of determining the answer to equipment fault handling; moreover, the entire process does not rely entirely on the experience of experts, avoiding the defect that the method that relies solely on the experience of experts has subjective factors and is prone to errors, resulting in a low accuracy in determining the answer to equipment fault handling, thereby improving the accuracy of determining the answer to equipment fault handling.

[0160] In an exemplary embodiment, in order to more clearly illustrate the device fault processing answer determination method provided in the embodiment of the present application, the device fault processing answer determination method is specifically described below with a specific embodiment. In one embodiment, Figure 7 As shown, the present application also provides an intelligent question-answering system and method for equipment fault maintenance based on retrieval enhancement generation. When determining the answer to the equipment fault handling, first obtain the equipment fault question information associated with the equipment to be analyzed, then determine the word vector corresponding to the equipment fault question information, and then determine the maximum similarity between the word vector and the preset embedding vector in the pre-built equipment maintenance knowledge base. When the maximum similarity is greater than the preset similarity, determine the preset embedding vector corresponding to the maximum similarity as the target embedding vector, and determine the answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as the target answer information corresponding to the equipment fault question information. When the maximum similarity is less than or equal to the preset similarity, determine the target answer information corresponding to the equipment fault question information through the pre-trained language model. Specifically including the following contents:

[0161] S1, collect various types of maintenance documents of enterprise equipment, including equipment failure maintenance manuals, equipment failure maintenance reports and other files, involving pdf, word, ppt, etc.; collect the operating status data N hours before the equipment failure occurs.

[0162] S2, based on equipment failure maintenance manuals and report files, builds a text analysis and processing model. First, it parses various files and extracts them into text information. Then, it segments the text information into text blocks and vectorizes the text blocks.

[0163] S3, vector database design, uses the designed vector database to store the above text block vectorization information and operating status data, where the first table in the vector database stores information including fault code, fault description, fault cause, fault maintenance recommended measures, fault text block vector, etc.; the second table in the vector database stores equipment fault code, fault occurrence time, and operating status data N hours before the fault occurrence; and obtains the knowledge base for equipment maintenance initialization.

[0164] S4, experts review the initialized knowledge base, supplement and correct the equipment maintenance initialized knowledge base based on expert maintenance experience and information input interface, and obtain the final equipment maintenance knowledge base.

[0165] S5, based on the text information of equipment failure maintenance in the vector database, fine-tune the existing large language model parameters to obtain a fine-tuned large language model.

[0166] S6, through the incremental information entry interface, the new treatment measures for equipment failure maintenance are continuously updated in the form of incremental records in the knowledge base.

[0167] S7, through the user question-answering interactive interface, integrates the answers given by the vector knowledge base and the large language model, and feeds back the final equipment fault maintenance measures and fault tracing operation status data to the user.

[0168] The aforementioned N is an adjustable parameter.

[0169] Specifically, the following steps are included:

[0170] Step 1: Collect various equipment failure maintenance documents of the enterprise, including equipment failure maintenance manuals, equipment failure maintenance reports and other documents, involving pdf, word, ppt, etc., such as Figure 2 As shown; collect the operating status data of various types of faults 1 (N=1) hour before.

[0171] Step 2: Taking the equipment fault maintenance manual as an example, a text analysis and processing model is constructed: first, the manual is parsed, information such as headers and footers are deleted, and effective information about equipment fault maintenance is extracted, including fault code, fault description, cause of fault, recommended measures for fault maintenance, fault description, etc., and then the text is segmented to obtain text blocks.

[0172] Step 3: Perform string fusion on the text blocks such as the device fault code and fault description, and extract keyword information of the fused string. For example, if the text block device fault code is "71058" and the text block fault description is "communication interruption with I / O device", the fusion result is "71058 communication interruption with I / O device", and the extracted keywords are "71058", "I / O", "device", and "communication interruption".

[0173] Step 4: merge the above keywords to obtain merged keywords, such as "71058 I / O device communication interruption".

[0174] Step 5: Vectorize the fused key to obtain an embedding vector, such as [0.12089, 0.18089, 0.4656, ...].

[0175] Step 6, based on the embedded vector, text block device fault code, text block fault description, text block fault cause, text block fault maintenance recommended measures, text block fault description, etc., obtain a record information under the fault code 71058. In particular, a record information can be obtained for each fault.

[0176] Step 7: Repeat steps 3 to 6 to obtain the record information for each fault code in the entire equipment fault maintenance manual.

[0177] Step 8, store the recorded information under each fault code in step 7 into the first table of the vector database; store the actual time when the equipment fault occurs and the operating status data 1 hour before the fault occurs into the second table of the vector database; and obtain the knowledge base for equipment maintenance initialization.

[0178] Step 9: Supplement and correct the equipment maintenance initialization knowledge base based on the expert maintenance experience and the information input interface to obtain the final equipment maintenance knowledge base.

[0179] Step 10: Using the final equipment maintenance knowledge base record information as the first input corpus.

[0180] Step 11, construct prompt words with equipment fault code, fault description and fault explanation as information, call large model third-party interfaces such as ChatGPT, Kimi, Wenxinyiyan, etc., generate corresponding answers, and integrate the fault recommended measures of the final equipment maintenance knowledge base into question and answer corpus.

[0181] Step 12: Based on the above question-answer corpus, fine-tune the model parameters of the existing large language model; check whether the answers of the fine-tuned large model are reasonable. After testing, a large model dedicated to equipment fault maintenance that is suitable for the enterprise is obtained.

[0182] In step 13, the user asks questions through the question-answering interactive interface, and the equipment fault maintenance question-answering system converts the question words into a word vector (consistent with the processing method in step 5), and searches the information in the final vector database through text similarity; based on information association, the text answer given by the final equipment maintenance knowledge base is obtained; based on the equipment fault code in the answer, the fault code in the second table in the associated vector database is extracted, the operating status time series data of the most recent fault is extracted, and a time series diagram is drawn, and the text answer and the operating status time series data diagram are used as the first answer; if the final equipment maintenance knowledge base does not give a text answer, the result of the fine-tuned large language model in step 12 is called as the second answer; compared with the second answer, the first answer is more accurate because the fault code or fault description of the user's question already exists in the final equipment maintenance knowledge base, and the second answer is a rough answer obtained without obtaining the first answer.

[0183] Step 14: deploy the final equipment maintenance knowledge base and the large model dedicated to equipment failure maintenance to the cloud platform, and provide an intelligent question-and-answer service interface for equipment failure maintenance to provide services to enterprises.

[0184] In step 15, experts and equipment maintenance engineers enter the new equipment failure maintenance information into the system based on the information entry interface, and execute steps 9 to 14 to continuously enrich the equipment maintenance knowledge base and iterate the large model.

[0185] In the above embodiment, when determining the answer to the equipment fault handling, by converting the equipment fault question information into a word vector and comparing it with the preset embedding vector in the equipment maintenance knowledge base, the equipment maintenance knowledge base and the language model can quickly determine the answer information that matches the equipment fault question information, which is conducive to improving the efficiency of determining the answer to the equipment fault handling; moreover, the entire process does not rely entirely on the experience of experts, avoiding the subjective factors of the method that relies solely on the experience of experts, which is prone to errors and leads to the defect of low accuracy in determining the answer to the equipment fault handling, thereby improving the accuracy of determining the answer to the equipment fault handling. At the same time, by combining the enterprise equipment fault maintenance manual and report, expert experience, and large language fine-tuning model question and answer corpus, the suggestions for equipment fault maintenance measures are made more refined and accurate, which is conducive to improving the rationality and scientificity of equipment fault maintenance.

[0186] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0187] Based on the same inventive concept, the embodiment of the present application also provides a device for determining an answer to a device fault processing for implementing the device fault processing answer determination method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of one or more device fault processing answer determination devices provided below can refer to the limitations of the device fault processing answer determination method above, and will not be repeated here.

[0188] In an exemplary embodiment, Figure 8 As shown, a device for determining an answer to a device fault processing is provided, comprising: a question information acquisition module 801, an information vector determination module 802, a maximum similarity determination module 803, a first answer determination module 804, and a second answer determination module 805, wherein:

[0189] The question information acquisition module 801 is used to acquire device fault question information associated with the device to be analyzed.

[0190] The information vector determination module 802 is used to determine the word vector corresponding to the equipment failure question information.

[0191] The maximum similarity determination module 803 is used to determine the maximum similarity between the word vector and the preset embedding vector in the pre-built equipment maintenance knowledge base.

[0192] The first answer determination module 804 is used to determine a preset embedding vector corresponding to the maximum similarity as a target embedding vector when the maximum similarity is greater than a preset similarity, and to determine answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as target answer information corresponding to the equipment fault question information.

[0193] The second answer determination module 805 is used to determine the target answer information corresponding to the equipment fault question information by using a pre-trained language model when the maximum similarity is less than or equal to a preset similarity.

[0194] In an exemplary embodiment, the device for determining an answer to equipment fault handling also includes a knowledge base construction module, which is used to obtain sample fault record information of sample equipment and sample historical operating status information of sample equipment; obtain a constructed equipment maintenance knowledge base based on the sample fault record information and the sample historical operating status information; and adjust the constructed equipment maintenance knowledge base based on the preset maintenance information associated with the sample equipment to obtain a pre-constructed equipment maintenance knowledge base.

[0195] In an exemplary embodiment, the knowledge base construction module is also used to obtain sample fault document information of a sample device; parse the sample fault document information to obtain a parsed document; extract target fault information from the parsed document; and obtain sample fault record information of the sample device based on the target fault information.

[0196] In an exemplary embodiment, the knowledge base construction module is also used to segment the target fault information to obtain multiple text blocks; fuse the text blocks to obtain fused text blocks; extract key information from the fused text blocks; fuse the key information to obtain fused key information; determine a preset embedding vector corresponding to the fused key information; and obtain sample fault record information of the sample device based on the preset embedding vector and the text blocks.

[0197] In an exemplary embodiment, the device for determining an answer to a device fault handling also includes a language model training module, which is used to input the target fault information into the language model to be trained to obtain predicted answer information corresponding to the target fault information; the predicted answer information is used to represent the predicted recommended measures information corresponding to the target fault information; based on the difference between the predicted answer information corresponding to the target fault information and the actual answer information, the language model to be trained is iteratively trained to obtain a pre-trained language model; the actual answer information is used to represent the actual recommended measures information corresponding to the target fault information.

[0198] In an exemplary embodiment, the first answer determination module 804 is also used to determine the target fault identification corresponding to the target embedding vector from the equipment maintenance knowledge base; determine the recommended measures information corresponding to the target fault identification as the recommended measures information corresponding to the equipment fault question information, and determine the historical operating status information corresponding to the target fault identification as the historical operating status information corresponding to the equipment fault question information; and use the recommended measures information and the historical operating status information corresponding to the equipment fault question information as the target answer information corresponding to the equipment fault question information.

[0199] In an exemplary embodiment, the maximum similarity determination module 803 is also used to determine the similarity between the word vector and multiple preset embedding vectors in a pre-built equipment maintenance knowledge base; from each similarity, determine the maximum similarity between the word vector and the preset embedding vector in the pre-built equipment maintenance knowledge base.

[0200] Each module in the above-mentioned device for determining the answer to the equipment fault processing can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each module above.

[0201] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Fig. 9 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as device fault question information and answer information. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for determining an answer to device fault processing is implemented.

[0202] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0203] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0204] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0205] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0206] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0207] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0208] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for determining an answer to a device fault processing, It is characterized in that The method comprises: Obtain equipment fault question information associated with the equipment to be analyzed; Determine a word vector corresponding to the equipment failure question information; Determining the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base; In the case where the maximum similarity is greater than a preset similarity, determining a preset embedding vector corresponding to the maximum similarity as a target embedding vector, and determining answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as target answer information corresponding to the equipment fault question information; When the maximum similarity is less than or equal to the preset similarity, target answer information corresponding to the equipment fault question information is determined through a pre-trained language model.

2. The method according to claim 1, It is characterized in that The pre-built equipment maintenance knowledge base is built by the following method: Acquire sample fault record information of a sample device and sample historical operation status information of the sample device; Obtaining a constructed equipment maintenance knowledge base based on the sample fault record information and the sample historical operation status information; According to the preset maintenance information associated with the sample device, the constructed equipment maintenance knowledge base is adjusted to obtain the pre-constructed equipment maintenance knowledge base.

3. The method according to claim 2, It is characterized in that The obtaining of sample fault record information of sample devices includes: Obtaining sample fault document information of the sample device; Parsing the sample fault document information to obtain a parsed document; Extracting target fault information from the parsed document; According to the target fault information, sample fault record information of the sample device is obtained.

4. The method according to claim 3, It is characterized in that The step of obtaining sample fault record information of a sample device according to the target fault information includes: Segmenting the target fault information to obtain multiple text blocks; Performing fusion processing on each of the text blocks to obtain a fused text block; Extracting key information from the fused text block; Performing fusion processing on the key information to obtain fused key information; Determining a preset embedding vector corresponding to the fusion key information; According to the preset embedding vector and each of the text blocks, sample fault record information of the sample device is obtained.

5. The method according to claim 3, It is characterized in that The pre-trained language model is trained by the following method: Inputting the target fault information into the language model to be trained to obtain predicted answer information corresponding to the target fault information; the predicted answer information is used to represent the predicted recommended measures information corresponding to the target fault information; According to the difference between the predicted answer information and the actual answer information corresponding to the target fault information, the language model to be trained is iteratively trained to obtain the pre-trained language model; the actual answer information is used to represent the actual recommended measures information corresponding to the target fault information.

6. The method according to claim 1, It is characterized in that The step of determining answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as target answer information corresponding to the equipment fault question information includes: Determining a target fault identifier corresponding to the target embedding vector from the equipment maintenance knowledge base; Determine the recommended action information corresponding to the target fault identifier as the recommended action information corresponding to the equipment fault question information, and determine the historical operation status information corresponding to the target fault identifier as the historical operation status information corresponding to the equipment fault question information; The suggested measures information and the historical operation status information corresponding to the equipment failure question information are both used as the target answer information corresponding to the equipment failure question information.

7. The method according to any one of claims 1 to 6, It is characterized in that The determining of the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base includes: Determining the similarity between the word vector and a plurality of preset embedding vectors in a pre-built equipment maintenance knowledge base; From the similarities, the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base is determined.

8. A device for determining an answer to a device failure, It is characterized in that The device comprises: A question information acquisition module is used to acquire device fault question information associated with the device to be analyzed; An information vector determination module, used to determine a word vector corresponding to the equipment failure question information; A maximum similarity determination module, used to determine the maximum similarity between the word vector and a preset embedding vector in a pre-built equipment maintenance knowledge base; A first answer determination module is used to determine, when the maximum similarity is greater than a preset similarity, a preset embedding vector corresponding to the maximum similarity as a target embedding vector, and determine answer information corresponding to the target embedding vector from the equipment maintenance knowledge base as target answer information corresponding to the equipment fault question information; The second answer determination module is used to determine the target answer information corresponding to the equipment fault question information through a pre-trained language model when the maximum similarity is less than or equal to the preset similarity.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program. It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

11. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.