Fault data processing method and device based on large model, computer equipment, readable storage medium and program product

Abstract

The invention relates to a fault data processing method and device based on a large model, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring target fault record data in a target time period; removing irrelevant contents in the target fault record data through a preset large language model and target knowledge base data to obtain key fault information, and performing attribution analysis on the target fault record data to obtain a fault reason; and classifying the fault phenomena and the fault causes to obtain a fault phenomenon classification standard and a fault cause classification standard. By adopting the method, the original fault report data can be automatically analyzed and sorted in combination with a large language model and a knowledge base plan, so that automatic data classification and information extraction are realized, manual operation is avoided, and efficient classification and analysis of fault data are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a fault data processing method and device based on a large model, a computer device, a computer readable storage medium and a computer program product. BACKGROUND

[0002] During the operation of electric locomotives, various abnormalities and faults will inevitably occur. The personnel responsible for maintenance and processing will record rich processing records. These abnormalities and faults and their processing records contain rich knowledge about fault rules and equipment maintenance. Effective integration and analysis of data accumulated over the years can not only reduce the downtime of electric locomotives and improve maintenance efficiency, but also predict potential faults and take preventive measures in advance, thereby greatly improving the reliability and operation efficiency of electric locomotives.

[0003] However, there are many problems in the processing and mining of these data. First, the format of historical data is relatively arbitrary and does not have a unified standard, with a large number of abbreviations, aliases, and many typos. Second, electric locomotives are complex integrated devices that span electrical, mechanical, control, network and many other fields, and the faults that occur on them are therefore highly complex and uncertain. For example, a seemingly simple electrical fault may involve the coordinated work of multiple subsystems, while a mechanical fault may trigger a chain reaction and affect the operation state of the entire electric locomotive. The data analysis method in the related art is generally manual processing, and the data processing process is relatively lengthy, and the manual analysis of the association between data is also limited, resulting in low data utilization. SUMMARY

[0004] Therefore, it is necessary to provide a fault data processing method, device, computer equipment, computer readable storage medium and computer program product based on a large model, which can improve the data utilization efficiency.

[0005] In a first aspect, the present application provides a fault data processing method based on a large model, comprising:

[0006] obtaining target fault record data in a target time period;

[0007] removing irrelevant content in the target fault record data through a pre-set large language model and target knowledge base data to obtain key fault information, and performing attribution analysis on the target fault record data to obtain fault causes; the key fault information includes fault objects and fault phenomena;

[0008] classifying the fault phenomena and the fault causes to obtain fault phenomenon classification standards and fault cause classification standards.

[0009] In one embodiment, the method further comprises:

[0010] In one embodiment, the method further comprises:

[0011] In one embodiment, the method further comprises:

[0012] In one embodiment, the method further comprises:

[0013] In one embodiment, the method further comprises:

[0014] In one embodiment, the method further comprises:

[0015] In one embodiment, the method further comprises:

[0016] In one embodiment, the method further comprises:

[0017] In one embodiment, the method further comprises:

[0018] In one embodiment, the method further comprises:

[0019] In one embodiment, the method further comprises:

[0020] In one embodiment, the method further comprises:

[0021] In one embodiment, the method further comprises:

[0022] In a second aspect, the present application also provides a fault data processing device based on a large model, comprising:

[0023] In a second aspect, the present application also provides a fault data processing device based on a large model, comprising: In a second aspect, the present application also provides a fault data processing device based on a large model, comprising:

[0024] a first obtaining module, configured to obtain target fault record data in a target time period;

[0025] a first removing module, configured to remove irrelevant content in the target fault record data by using a preset large language model and target knowledge base data, to obtain key fault information, and to perform attribution analysis on the target fault record data to obtain a fault cause; the key fault information includes a fault object and a fault phenomenon;

[0026] a first classifying module, configured to classify the fault phenomenon and the fault cause to obtain a fault phenomenon classification standard and a fault cause classification standard.

[0027] In a third aspect, the present application further provides a computer device, comprising a memory and a processor, the memory stores a computer program, and the processor realizes the following steps when executing the computer program:

[0028] obtain target fault record data in a target time period;

[0029] remove irrelevant content in the target fault record data by using a preset large language model and target knowledge base data, to obtain key fault information, and perform attribution analysis on the target fault record data to obtain a fault cause; the key fault information includes a fault object and a fault phenomenon;

[0030] classify the fault phenomenon and the fault cause to obtain a fault phenomenon classification standard and a fault cause classification standard.

[0031] In a fourth aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the computer program realizes the steps in the embodiments when executed by a processor.

[0032] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, and the computer program realizes the steps in the embodiments when executed by a processor.

[0033] The aforementioned fault data processing method, apparatus, computer equipment, computer-readable storage medium, and computer program product based on a large-scale model include the following steps: acquiring target fault record data within a target time period; removing irrelevant content from the target fault record data using a pre-set large-scale language model and target knowledge base data to obtain key fault information; and performing attribution analysis on the target fault record data to obtain fault causes. Key fault information includes the fault object and fault phenomena; and classifying the fault phenomena and fault causes to obtain fault phenomenon classification criteria and fault cause classification criteria. By employing this method, combined with a large-scale language model and knowledge base plan, the original fault report data can be automatically analyzed and organized, achieving automated data classification and information extraction, avoiding manual operation, and realizing efficient classification and analysis of fault data. This significantly shortens data processing time and improves the accuracy, reliability, and timeliness of data processing. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a flowchart illustrating a fault data processing method based on a large model in one embodiment;

[0036] Figure 2 This is a flowchart illustrating the steps for obtaining knowledge base data in one embodiment;

[0037] Figure 3a This is a flowchart illustrating a fault data processing method based on a large model, as described in another embodiment.

[0038] Figure 3b This is a schematic diagram of an online usage process based on a large model in another embodiment;

[0039] Figure 4 This is a structural block diagram of a fault data processing device based on a large model in one embodiment;

[0040] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0041] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not intended to limit the present application.

[0042] It should be noted that the terms "first", "second", etc. used in the present application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "include" and "have" and any variations thereof used in the present application are intended to cover non-exclusive inclusion. The term "a plurality of" used in the present application means two or more. The term "and / or" used in the present application means one of the options or any combination of multiple options.

[0043] The method for processing fault data based on a large model provided by the embodiments of the present application is described by taking the case of applying the method to a terminal device. The method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is realized through the interaction of the terminal and the server. The terminal can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers. The server can be a stand-alone physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services; as shown in Figure 1 The method for processing fault data based on a large model provided by the embodiments of the present application is described by taking the case of applying the method to a terminal device. The method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is realized through the interaction of the terminal and the server. The terminal can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers. The server can be a stand-alone physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services; as shown in

[0044] Step 102, obtaining target fault record data in a target time period.

[0045] The target time period can be a historical time period, for example, a period of time before the current time point, and the specific value of the historical time period is not limited in the present disclosure. The target fault record data can be obtained by preprocessing the initial fault record data. The initial fault record data includes fault reports and fault handling records. The fault report can be a description of the fault situation, which can include the object of the fault and the fault description. The fault handling record can include the fault report and the fault handling strategy.

[0046] Specifically, the terminal can obtain the initial fault record data of the target time period, and preprocess the initial fault record data to obtain the target fault record data.

[0047] Step 104, removing irrelevant content in the target fault record data by using a pre-set large language model and target knowledge base data to obtain key fault information, and performing attribution analysis on the target fault record data to obtain the fault reason.

[0048] The key fault information includes a fault object and a fault phenomenon. The large language model LLM can be a large model based on a large number of model parameters, and can analyze a large amount of text data to process the data. The target knowledge base data can be knowledge data matched with the target fault record data filtered from the preset knowledge base data. The fault object can be a locomotive.

[0049] Specifically, the terminal can remove irrelevant content in the target fault record data by the preset large language model and the determined target knowledge base data, to obtain a fault object and a fault phenomenon corresponding to the target fault record data, where the fault object can be an object that has a fault. The irrelevant content can be a locomotive number, a fault location, a fault time, a driver identifier of the fault object, and the like. The obtained key fault information can avoid data redundancy and further improve the efficiency of fault data analysis. The target fault record data can be a fault report.

[0050] In this way, the terminal can also perform attribution analysis on the fault report and the fault handling strategy in the target fault record data by the preset large language model and the determined target knowledge base data, determine the fault phenomenon based on the fault report, and perform reverse reasoning based on the fault handling strategy to obtain an abnormal reason. Optionally, for example, the fault handling strategy can be firmware upgrade, and the abnormal reason obtained by the terminal after reverse reasoning can be that the firmware version is too old and has not been upgraded in time, and the like.

[0051] The target fault record data can be a fault handling record, and the fault handling record includes a fault report and a fault handling strategy.

[0052] In step 106, the fault phenomenon and the fault reason are classified to obtain a fault phenomenon classification standard and a fault reason classification standard.

[0053] The fault phenomenon classification standard is a standard for classifying the fault phenomenon, and the fault reason classification standard is a standard for classifying the fault reason.

[0054] Specifically, the terminal can divide the fault phenomenon and the fault cause corresponding to the target fault record data into multiple batches of data, and sequentially input each batch of data into the large language model to obtain an output result of the large language model, which can be a fault phenomenon classification standard and a fault cause classification standard. For example, the terminal can input the first batch of data into the large language model to obtain an initial fault phenomenon classification standard and an initial fault cause classification standard, and input the second batch of data into the large language model to obtain a classification standard updated based on the initial fault phenomenon classification standard and the initial fault cause classification standard, until all batches of data are input into the large language model to obtain a final fault phenomenon classification standard and a final fault cause classification standard. The terminal can classify the target fault record data based on the final fault phenomenon classification standard and the final fault cause classification standard to obtain a fault phenomenon type and a fault cause type of the target fault record data.

[0055] In the above method for processing fault data based on a large model, target fault record data in a target time period is obtained; irrelevant content in the target fault record data is removed by a preset large language model and target knowledge base data to obtain key fault information, and the target fault record data is subjected to attribution analysis to obtain a fault cause; the key fault information includes a fault object and a fault phenomenon; the fault phenomenon and the fault cause are classified to obtain a fault phenomenon classification standard and a fault cause classification standard. By using the method, original fault report data can be automatically analyzed and sorted in combination with a large language model and a knowledge base, automatic data classification and information extraction are realized, manual operation is avoided, efficient classification and analysis of fault data are realized, data processing time is greatly shortened, and the accuracy, reliability, and timeliness of data processing are improved.

[0056] In one embodiment, the step of "obtaining target fault record data in a target time period" can include:

[0057] Obtaining initial fault record data in a target time period, the initial fault record data including fault reports and fault handling records; preprocessing abnormal text in the initial fault record data to obtain target fault record data.

[0058] The initial fault record data includes fault reports and fault handling records. The fault report can be a description of a fault condition, which can include a fault object and a fault description. The fault handling record can include a fault report and a fault handling strategy. The fault description can be a text description, such as xx component oil leakage, xx device oil leakage, etc. The abnormal text can be a wrong character, a wrong punctuation, a continuous symbol, etc. in the initial fault record data.

[0059] Specifically, the terminal can obtain initial fault record data in a target time period, determine abnormal text in the initial fault record data, i.e., determine typos, incorrect punctuation marks, line break marks, continuous same symbols, etc. in the initial fault record data, and perform removal processing on the abnormal text to obtain target fault record data. That is, the terminal can perform cleaning processing on the abnormal text in the initial fault record data to obtain target fault record data that does not contain abnormal text.

[0060] In this embodiment, by preprocessing the initial fault record data, the data used for subsequent fault analysis and fault processing can be ensured to be clean, and a reliable data basis is provided for the subsequent analysis process.

[0061] In one embodiment, the method further comprises:

[0062] Based on the fault phenomenon classification standard and the fault cause classification standard, the target fault record data is classified to obtain the phenomenon type and the cause type corresponding to each piece of fault record data in the target fault record data.

[0063] Specifically, the terminal can input the fault phenomenon classification standard, the fault cause classification standard, and the target fault record data into a large language model, and perform phenomenon classification and cause classification through the large language model. For example, the large language model can perform format conversion on the fault phenomenon classification standard and the fault cause classification standard to obtain a classification standard feature vector, and based on the classification strategy indicated by the classification standard feature vector, classify the phenomenon and the cause of each piece of fault record data in the target fault record data to obtain the phenomenon type and the cause type corresponding to each piece of fault record data. The phenomenon type can include shutdown, power failure, oil leakage, performance degradation, etc., and the cause type can include electrical fault, mechanical fault, software error, abnormal human operation, etc.

[0064] In this embodiment, the large language model is used to automatically classify fault data, which avoids the need for experts to label fault data and further improves the accuracy and efficiency of data classification.

[0065] In one embodiment, the method further comprises:

[0066] The fault data set is used to train the to-be-trained model to obtain a fault processing model, and the fault processing model includes one or more of a fault phenomenon extraction model, a fault phenomenon classification model, and a fault cause analysis model.

[0067] The to-be-trained model can be a small model of the to-be-trained online application, for example, can be a neural network model, a deep learning model, and the like traditional small model. The fault phenomenon extraction model is used to extract the fault phenomenon in the fault record data; the fault phenomenon classification model is used to classify the fault phenomenon in the fault record data to obtain the phenomenon type of the fault record data; and the fault reason analysis model is used to classify the fault reason in the fault record data to obtain the reason type of the fault reason corresponding to the fault record data. The fault data set includes sample training data, and the sample training data at least includes the fault record data, the fault phenomenon, the fault reason, the phenomenon type of the fault phenomenon, and the reason type of the fault reason. The fault data set represents the corresponding relationship between the original fault phenomenon of the fault record data and the fault phenomenon (standardized fault phenomenon) extracted by the large language model, and also represents the corresponding relationship between the original fault processing record and the fault reason.

[0068] Specifically, the terminal can train the to-be-trained model through the fault record data and the fault phenomenon in the fault data set to obtain the fault phenomenon extraction model; the terminal can train the to-be-trained model through the fault record data, the fault phenomenon, and the phenomenon type in the fault data set to obtain the fault phenomenon classification model; the terminal can train the to-be-trained model through the fault record data and the fault reason in the fault data set to obtain the fault reason analysis model; and optionally, the terminal can also train the to-be-trained model through the fault record data, the fault reason, and the reason type in the fault data set to obtain the fault reason analysis model.

[0069] In this embodiment, the small model is trained through the fault data set, which can avoid the high computing cost of the large language model, reduce the computing cost, realize more cost-effective fault phenomenon extraction and classification, ensure the timeliness of the online application, and improve the user experience of using the online application.

[0070] In one embodiment, the fault data set at least includes sample fault record data and sample fault phenomenon corresponding to the sample fault record data; and the method further includes:

[0071] The to-be-trained model is used to process the sample fault record data to obtain a predicted fault phenomenon corresponding to the sample fault record data; and the to-be-trained model is updated based on the predicted fault phenomenon and the sample fault phenomenon to obtain a fault phenomenon extraction model meeting a preset training completion condition.

[0072] The sample fault record data can be original fault record data, i.e., original fault phenomenon and original processing record; and the sample fault phenomenon can be a standardized fault phenomenon, i.e., a fault phenomenon extracted by a large language model.

[0073] Specifically, the terminal can input the sample fault record data into the to-be-trained model, i.e., the to-be-trained fault phenomenon extraction model, to obtain a predicted fault phenomenon, and calculate a loss value corresponding to a loss function between the predicted fault phenomenon and the sample fault phenomenon. If it is determined that the loss value does not satisfy a preset training completion condition, the model parameters of the to-be-trained fault phenomenon extraction model are updated to obtain an updated model. Based on the updated model, the step of processing the sample fault record data by the to-be-trained model is re-executed until the preset training completion condition is satisfied, and a trained fault phenomenon extraction model is obtained.

[0074] Optionally, in a case where the terminal determines that the current number of training iterations has satisfied the preset threshold of the number of training iterations, the terminal can determine that the preset training completion condition has been satisfied; in another example, in a case where the terminal determines that the loss value corresponding to the loss function has satisfied a preset convergence condition, the terminal can determine that the preset training completion condition has been satisfied. The preset convergence condition can be that the loss value corresponding to the loss function has not changed, or that the loss value has reached a minimum loss value threshold, etc.

[0075] In this embodiment, the small model applied online is trained based on the fault data set, which can realize more cost-effective extraction and classification of fault phenomena, reduce the training cost and calculation cost of the model, and ensure the efficiency of online application.

[0076] In one embodiment, as shown in Figure 2 the method further includes:

[0077] In step 202, based on the target fault record data, the initial knowledge data with a semantic similarity satisfying a preset similarity condition is extracted from a preset knowledge base.

[0078] The preset knowledge base can be professional knowledge in the field of locomotives, for example, can include locomotive technical manuals, locomotive maintenance guidelines, etc. The terminal can segment the professional knowledge in the field of locomotives, and perform feature conversion on the segmented data to obtain a knowledge feature vector. The knowledge feature vector corresponding to the professional knowledge in the field of locomotives is added to a vector database to obtain the preset knowledge base.

[0079] Specifically, the terminal can calculate semantic similarities between the target fault record data and each knowledge feature vector in the preset knowledge base respectively, determine the knowledge feature vectors with the largest semantic similarities as the initial knowledge data under the preset similarity condition, or determine the knowledge feature vectors with the semantic similarities greater than a preset similarity threshold as the initial knowledge data under the preset similarity condition. In this embodiment, the specific values of the target number and the preset similarity threshold are not limited, and can be determined based on actual application scenarios by those skilled in the art. For example, the target number can be 10, and the preset similarity threshold can be 95%.

[0080] In step 204, the initial knowledge data is subjected to data reconstruction processing to obtain the knowledge base data, and / or the invalid part in the initial knowledge data is removed to obtain the knowledge base data.

[0081] Specifically, the terminal can reconstruct the professional knowledge represented by the initial knowledge data, and output the knowledge base data with different expression forms but unchanged expression meanings. The terminal can also reconstruct the professional knowledge represented by the initial knowledge data to obtain reconstructed data, remove the invalid part in the reconstructed data, and obtain the knowledge base data. The terminal can also query the invalid part in the initial knowledge data, remove the invalid part, and obtain the knowledge base data, so as to distill the knowledge base data and further improve the speed of LLM (Large Language Model) data processing.

[0082] In this embodiment, by screening in the preset knowledge base based on semantic similarity, the obtained knowledge data can be input into the LLM, which significantly improves the accuracy and reliability of the model on professional problems. By reconstructing and cutting the knowledge data, the text data input into the LLM can be reduced, the data processing speed of the LLM can be significantly improved, and the balance between the processing speed and the processing accuracy can be ensured.

[0083] In the following, the implementation process of the above-mentioned fault data processing method based on a large model is described in detail in combination with a specific embodiment.

[0084] The method for processing fault data based on a large model provided in this embodiment realizes automatic sorting and analysis of electric locomotive fault reports by combining a large language model and knowledge base technology, thereby improving the efficiency and accuracy of data processing and providing a scientific basis for the maintenance and management of electric locomotives. Mainly includes: using a large language model (LLM) and knowledge base technology, it can automatically identify and correct the non-uniform data and non-standard data in the traditional fault record data, convert the non-uniform data into a unified format, and classify, which improves the standardization of data and reduces the error rate of manual processing. Optionally, the large language model can understand the context and analyze the semantics to unify "electric control failure" and "electronic control failure" as "electrical control failure", thereby ensuring the consistency and accuracy of the data.

[0085] Secondly, by using a large language model and knowledge base technology, the method provided in this embodiment can quickly extract and sort key information, generate detailed statistical reports, shorten the data processing time, and improve the accuracy, processing efficiency and reliability of the data. Through automated tools, efficient statistics and analysis of fault data are realized. For example, the system can automatically count the number of occurrences of a certain type of fault within a certain period of time, and analyze the causes and processing methods, providing a scientific basis for subsequent fault prevention and equipment maintenance.

[0086] Further, the method provided in this embodiment ensures the consistency and accuracy of the data through standardized data processing procedures and automated tools. The system can automatically detect and correct data errors, generate standardized reports, and facilitate maintenance personnel to review and use. It also provides visual data analysis tools to make the data processing process more intuitive and convenient.

[0087] In summary, the method provided in this embodiment realizes automatic sorting and analysis of electric locomotive fault reports by combining a large language model and knowledge base technology, not only improving the efficiency and accuracy of data processing, but also providing a scientific basis for the maintenance and management of electric locomotives. This will greatly improve the reliability and operating efficiency of electric locomotives, reduce downtime, improve repair efficiency, and reduce maintenance costs.

[0088] As Figure 3a shown, it is another embodiment of the data processing flow, which realizes the above effects, and the specific process is as follows:

[0089] The historical data is processed, for example, the fault report is processed to obtain a processing record; the industry knowledge includes a maintenance manual, a locomotive architecture, and a term dictionary; the terminal can sample the processing record, process the sampled processing record, a large model summary instruction, and industry knowledge through a large model to obtain a fault summary corresponding to the fault report, train based on the fault report and the fault summary to obtain a fault summary model.

[0090] The semantic query is performed by the large model in combination with the sampled processing record, the large model attribution instruction and the industry knowledge to obtain the fault attribution corresponding to the fault report + processing record; the fault attribution model is obtained based on the fault attribution and the fault report + processing record.

[0091] Batch reasoning is performed on the full historical data and the fault attribution model to obtain the fault summary + fault attribution corresponding to the fault report + processing record.

[0092] The fault summary + fault attribution is input into the recursive classification process, and another large model is used in combination with the fault summary + fault attribution, the large model category identification prompt and the sampled fault details to obtain the fault category; the fault category corresponding to the fault details is obtained based on the fault summary + fault attribution, the sampled fault details, the large model classification prompt and the fault category in combination with another large model, and the text classification model is obtained based on the fault details and the fault category.

[0093] Batch reasoning is performed based on the full fault details data and the text classification model to output the fault category corresponding to the fault details, and the fault category corresponding to the fault details is re-input into the recursive classification process.

[0094] As shown in Figure 3b , it can be a process schematic diagram of online use of the model:

[0095] The fault report is input into the fault summary model to obtain the fault summary, and the fault summary is input into the text classification model to obtain the category of the fault summary. The category of the fault summary and the fault report are input into the fault type classification to obtain the category of the fault. The processing record is input into the fault attribution model to obtain the fault reason, and the category of the fault and the fault reason are input into the text classification model to perform fault reason classification to obtain the category of the fault reason. Each text classification model can be trained based on the output of the text classification model.

[0096] Specifically, the specific execution process of the fault data processing method of the large model can be as follows:

[0097] Step 1, data preprocessing. Data preprocessing is to clean the input data to ensure the accuracy and efficiency of the subsequent model. Specifically, common misspelling and incorrect punctuation in the input data are preprocessed to remove line breaks, continuous spaces or symbols that affect abnormal text. The data is cleaned to lay a reliable data foundation for subsequent processing, that is, the initial fault record data is cleaned to obtain the target fault record data.

[0098] Step 2: Extracting key fault information based on large language models, in order to extract the part that really involves the fault in the fault report. Specifically, using a large language model, combined with knowledge base knowledge, irrelevant content (such as fault locomotive number, fault location and time, driver, etc.) is removed, key fault information is extracted, greatly reducing data redundancy and improving the efficiency of subsequent analysis. Key fault information includes fault object and fault phenomenon.

[0099] Step 3: Automatically analyzing fault causes based on large language models, in order to preliminarily extract possible causes of the fault. Specifically, using a large language model, combined with knowledge base knowledge, potential fault causes are extracted from the processing records, providing another dimension for subsequent fault classification and improving the application value of the analysis results.

[0100] Step 4: Classifying faults based on large language models, in order to classify the extracted fault phenomena or fault causes. Specifically, based on a large language model, the input data is divided into several batches, and the model is output in sequence. The model will give and gradually improve the classification criteria and examples, and then based on this classification criteria, use the large model to classify the input data one by one. Automatically classify fault data, reduce the demand for expert annotation, and improve classification efficiency.

[0101] Among them, the method also combines knowledge base to improve the performance of large language models on professional problems. Specifically, the locomotive technical manual, maintenance guide, etc. are cut into vector databases, and when the large model processes fault records, it is queried based on semantic similarity and provided to the model as a reference. This step significantly improves the accuracy and reliability of the model on professional problems.

[0102] The method also distills the knowledge base of the large language model to speed up processing. Specifically, when querying the knowledge base, use a smaller language model to cut off irrelevant parts, or completely refine and rewrite the references found as needed, which can reduce the amount of text input to the large model and significantly improve processing speed.

[0103] Step 5: Training small models based on large language model standard results to realize online application, in order to realize online application and reduce computing cost. Specifically, considering the high computing cost of large language models, use large language models as automatic annotators to train small models to achieve more affordable fault phenomenon extraction and classification. This step not only reduces the computing cost, but also ensures the efficiency of online application.

[0104] The following describes one specific embodiment: A railway company has accumulated a large number of fault reports during the long-term operation of electric locomotives. These reports have some standard fields, such as time, location, repair person, etc., but the core fault phenomenon and handling records are in different formats, contain a large number of abbreviations, aliases and typos, and are also mixed with a large number of auxiliary descriptions other than the fault phenomenon, such as fault time, locomotive number, running state, etc. Maintenance personnel need to regularly sort and analyze these reports to promptly discover and handle faults. However, due to the large amount of data and non-uniform format, manually processing these reports not only consumes time and effort, but also is prone to errors and lacks timeliness. Therefore, the fault data processing method based on large models provided in this embodiment can be applied to a fault report management special software system to realize automatic sorting and analysis of fault reports.

[0105] S1, model offline training; specifically, in order to efficiently and accurately identify abbreviations, aliases and special expressions between different repair teams, fault reports in a historical time period are obtained for model training and learning.

[0106] After cleaning, key fault phenomenon extraction, fault phenomenon classification, automatic fault cause analysis and classification, and manual verification, the original fault phenomenon to standardized fault phenomenon and classification, and the original handling record to fault cause dataset is obtained. The dataset is used to train smaller fault phenomenon extraction models, fault phenomenon classification models and fault cause analysis models, and the effects are verified.

[0107] S2, software system integration; specifically, the trained models are organically integrated into the fault report management special software system. When filling in the fault phenomenon, the fault phenomenon extraction model and the fault phenomenon classification model are automatically called, and when filling in the maintenance record after fault repair, the fault cause is automatically analyzed.

[0108] S3, statistical analysis; specifically, based on the automatically calculated fault phenomenon, fault classification and fault cause, the newly added faults are automatically statistically analyzed to generate related reports and charts. These reports provide scientific basis for maintenance personnel, improve maintenance efficiency and reduce maintenance cost.

[0109] S4, regular retraining; specifically, the introduction of automated systems will inevitably affect users' expression habits, leading to changes in users' abbreviations, aliases, commonly used expressions, etc. To cope with this change, after the system goes online, new data is regularly fed back to the offline training stage to update the corresponding processing models, so as to continuously learn users' expression habits and language, achieving the best effect.

[0110] Through the above steps, the automatic sorting and analysis of fault reports are successfully realized, and the data processing efficiency is significantly improved. The data that originally required manual processing is now completed by the system in real time. The random and uneven fault phenomena and processing are converted into structured fault reports, and the consistency and accuracy of the data are greatly improved. Combined with the statistical analysis function of the system, the maintenance team can timely discover fault trends and deal with them in a targeted manner, reducing locomotive downtime and providing operational efficiency.

[0111] That is, the fault data processing method based on a large model provided in this embodiment realizes the automatic sorting and analysis of electric locomotive fault reports, significantly improves the efficiency and accuracy of data processing, and provides a scientific basis for the maintenance and management of electric locomotive. This will greatly improve the reliability and operational efficiency of electric locomotive, reduce downtime, improve maintenance efficiency, and reduce maintenance costs. Combined with large language model and knowledge base technology, the automatic sorting and analysis of electric locomotive fault reports are realized. Through automated data processing, classification, and information extraction, a large amount of manual operation is eliminated, efficient classification and analysis of fault data are realized, the data processing time is greatly shortened, and the accuracy, reliability, and timeliness of the data are improved.

[0112] Due to the complexity of electric locomotive, personnel with considerable professional knowledge and industry experience are required for manual statistical classification. Through automated processing, the professional ability requirements of data statisticians are reduced, efficient data processing and analysis can be performed with the help of the system, human resources are saved, and the efficiency and quality of data processing are improved.

[0113] The fault data processing method based on a large model provided in this embodiment can realize real-time analysis of large-scale data, timely discovery of potential fault patterns and trends, and in-depth mining and long-term trend analysis of historical data through automated tools. The system can automatically integrate fault data accumulated over many years to generate historical fault rule reports, helping maintenance personnel to predict and prevent potential faults in advance, thereby improving the reliability and operational efficiency of electric locomotive. Through automated processing and analysis, faults can be timely discovered and handled, reducing electric locomotive downtime and improving operational efficiency. The detailed statistical reports and analysis results generated by the system provide a scientific basis for maintenance personnel, improving maintenance efficiency and reducing maintenance costs. Through real-time analysis of data, maintenance personnel can quickly discover high-frequency faults or new fault trends and take preventive measures in advance, further improving the reliability and operational efficiency of electric locomotive.

[0114] It should be understood that although each step in the flowchart involved in each embodiment as described above is shown in sequence according to the direction of the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise explicitly stated herein, there is no strict order limitation for the execution of these steps, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or multiple stages, which 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 sequential, but can be alternately or alternately executed with at least part of other steps or steps or stages in other steps. It can be understood that the steps in different embodiments can be freely combined as needed, and various non-contradictory schemes formed by the combination are within the scope of protection of the present application.

[0115] Based on the same inventive concept, the embodiments of the present application also provide a large model-based fault data processing device for implementing the large model-based fault data processing method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more large model-based fault data processing device embodiments provided below can refer to the limitations of the large model-based fault data processing method described above, which will not be repeated here.

[0116] In one exemplary embodiment, as shown in Figure 4 a large model-based fault data processing device 400 is provided, comprising:

[0117] A first acquisition module 402 is configured to acquire target fault record data in a target time period;

[0118] A first removal module 404 is configured to remove irrelevant content in the target fault record data by using a preset large language model and target knowledge base data to obtain key fault information, and perform attribution analysis on the target fault record data to obtain a fault cause; the key fault information includes a fault object and a fault phenomenon;

[0119] A first classification module 406 is configured to classify the fault phenomenon and the fault cause to obtain a fault phenomenon classification standard and a fault cause classification standard.

[0120] In one embodiment, the first acquisition module is specifically configured to:

[0121] acquire initial fault record data in a target time period, the initial fault record data including fault reports and fault handling records;

[0122] Preprocess the abnormal text in the initial fault record data to obtain target fault record data.

[0123] In one embodiment, the apparatus further comprises:

[0124] The second classification module is configured to classify the target fault record data based on the fault phenomenon classification standard and the fault cause classification standard to obtain a phenomenon type and a cause type corresponding to the target fault record data.

[0125] In one embodiment, the apparatus further comprises:

[0126] The first training module is configured to train the to-be-trained model based on a fault data set to obtain a fault processing model, the fault processing model comprising one or more of a fault phenomenon extraction model, a fault phenomenon classification model, and a fault cause analysis model.

[0127] In one embodiment, the fault data set at least comprises sample fault record data and sample fault phenomena corresponding to the sample fault record data; and the apparatus further comprises:

[0128] The first processing module is configured to process the sample fault record data based on the to-be-trained model to obtain predicted fault phenomena corresponding to the sample fault record data.

[0129] The first updating module is configured to update the to-be-trained model based on the predicted fault phenomena and the sample fault phenomena to obtain a fault phenomenon extraction model that satisfies a preset training completion condition.

[0130] In one embodiment, the apparatus further comprises:

[0131] The first query module is configured to query the target fault record data in a preset knowledge base to extract initial knowledge data with a semantic similarity satisfying a preset similarity condition.

[0132] The knowledge base data determination module is configured to perform data reconstruction processing on the initial knowledge data to obtain knowledge base data; and / or remove invalid parts in the initial knowledge data to obtain knowledge base data.

[0133] The above-described various modules in the fault data processing apparatus based on a large model can be realized in whole or in part by software, hardware, and a combination thereof. The above-described various modules can be embedded in or independent of a processor in a computer device in hardware form, or can be stored in a memory in a computer device in software form, so as to be called and executed by a processor to perform operations corresponding to the above-described various modules.

[0134] In an example embodiment, a computer device, which can be a server, is provided, and an internal structure diagram of the computer device can be as shown in FIG. 1. Figure 5 The computer device includes a processor, a memory, an input / output interface, 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 configured 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 running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store fault data. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with terminals outside through a network connection. The computer program is executed by the processor to implement a fault data processing method based on a large model.

[0135] Those skilled in the art can understand that Figure 5 The structure shown in the above embodiment is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the diagram, or combine certain components, or have a different arrangement of components.

[0136] In an example embodiment, a computer device is provided, including a memory and a processor, and the memory stores a computer program. The processor executes the computer program to implement the steps in the embodiment.

[0137] In an example embodiment, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program. The computer program is executed by the processor to implement the steps in the embodiment.

[0138] In an example embodiment, a computer program product is provided, and the computer program product includes a computer program. The computer program is executed by the processor to implement the steps in the embodiment.

[0139] 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 for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use, and processing of the related data need to comply with relevant regulations.

[0140] 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. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. The non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, 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. The volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0141] The technical features of the above embodiments can be combined in any manner. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present application.

[0142] The above-described embodiments are merely illustrative of several embodiments of the present application, and the description is relatively specific and detailed, but should not be understood as a limitation on the scope of the patent. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. A large model-based fault data processing method, characterized in that, The method comprises: acquiring target fault record data in a target time period; removing irrelevant content in the target fault record data by a preset large language model and target knowledge base data to obtain key fault information, and performing attribution analysis on the target fault record data to obtain a fault cause; the key fault information comprises a fault object and a fault phenomenon; classifying the fault phenomenon and the fault cause to obtain a fault phenomenon classification standard and a fault cause classification standard.

2. The method of claim 1, wherein, The acquisition of the target fault record data in the target time period comprises: acquiring initial fault record data in the target time period, the initial fault record data comprising fault reports and fault handling records; preprocessing abnormal text in the initial fault record data to obtain the target fault record data.

3. The method of claim 1, wherein, The method further comprises: classifying the target fault record data based on the fault phenomenon classification standard and the fault cause classification standard to obtain a phenomenon type and a cause type corresponding to the target fault record data respectively.

4. The method of claim 1, wherein, The method further comprises: training a to-be-trained model by a fault data set to obtain a fault handling model, the fault handling model comprising one or more of a fault phenomenon extraction model, a fault phenomenon classification model, and a fault cause analysis model.

5. The method of claim 4, wherein, The fault data set at least comprises sample fault record data and sample fault phenomena corresponding to the sample fault record data; the method further comprises: processing the sample fault record data by the to-be-trained model to obtain predicted fault phenomena corresponding to the sample fault record data; updating the to-be-trained model based on the predicted fault phenomena and the sample fault phenomena to obtain a fault phenomenon extraction model meeting a preset training completion condition.

6. The method of claim 1, wherein, The method further comprises: querying the target fault record data in a preset knowledge base to extract initial knowledge data with a semantic similarity meeting a preset similarity condition; performing data reconstruction processing on the initial knowledge data to obtain knowledge base data; and / or removing invalid parts in the initial knowledge data to obtain knowledge base data. 7.A large model-based fault data processing apparatus, characterized by, The apparatus comprises: a first acquisition module configured to acquire target fault record data in a target time period; a first removal module configured to remove irrelevant content in the target fault record data by a preset large language model and target knowledge base data to obtain key fault information, and perform attribution analysis on the target fault record data to obtain a fault cause; the key fault information comprises a fault object and a fault phenomenon; a first classification module configured to classify the fault phenomenon and the fault cause to obtain a fault phenomenon classification standard and a fault cause classification standard.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the steps of the method of any one of claims 1 to 6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.

Images