Fault root cause analysis method and device, electronic equipment and storage medium

Abstract

The invention provides a fault root cause analysis method and device, electronic equipment and a storage medium, and can be applied to the technical field of computers. The method comprises the following steps: converting a fault description text input by a user into vector representation to obtain a query vector; performing query search in a preset operation and maintenance knowledge vector library according to the query vector, and obtaining an operation and maintenance knowledge vector meeting a similarity condition with the query vector in the operation and maintenance knowledge vector library; wherein the operation and maintenance knowledge vector is a vector representation of the operation and maintenance knowledge text; based on the operation and maintenance knowledge text corresponding to the operation and maintenance knowledge vector and the fault description text, obtaining fault prompt information; and inputting the fault prompt information into the large language model to obtain a fault root cause analysis result. According to the method, by integrating the large language model and the retrieval enhancement technology, fault root cause analysis can be automatically carried out, and the analysis efficiency of fault root causes is effectively improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a fault root cause analysis method, device, electronic device and storage medium. Background Art

[0002] Data Center is a key infrastructure of modern information technology, used to realize centralized processing, storage, transmission, exchange and management of data information.

[0003] As a key business support platform, the stability and reliability of the data center of a financial institution are of vital importance. Any infrastructure failure may lead to business interruption, affect the safe production and operation of the data center, and cause economic losses to the financial institution.

[0004] With the development of technology, the scale and complexity of data centers continue to increase, involving more and more equipment and systems, making fault diagnosis more difficult. When a data center fails, operation and maintenance personnel need to locate the fault location and diagnose the fault type from the massive amount of monitoring data and log information generated by the data center. However, traditional manual analysis methods are difficult to cope with such a large amount of data, resulting in the inability to quickly and accurately locate the fault location and diagnose the fault type, and there is a problem of low efficiency in fault root cause analysis. Summary of the invention

[0005] The present application provides a fault root cause analysis method, device, electronic device and storage medium to solve the technical problem of low efficiency of fault root cause analysis in existing fault root cause analysis methods.

[0006] According to the first aspect disclosed in the present application, the present application provides a fault root cause analysis method, comprising:

[0007] Convert the fault description text entered by the user into a vector representation to obtain a query vector;

[0008] Performing a query search in a preset operation and maintenance knowledge vector library according to the query vector, and obtaining an operation and maintenance knowledge vector in the operation and maintenance knowledge vector library that satisfies a similarity condition with the query vector; wherein the operation and maintenance knowledge vector is a vector representation of an operation and maintenance knowledge text;

[0009] Obtaining fault prompt information based on the operation and maintenance knowledge text corresponding to the operation and maintenance knowledge vector and the fault description text;

[0010] The fault prompt information is input into a large language model to obtain a fault root cause analysis result.

[0011] In a feasible implementation manner, the method further includes:

[0012] Obtaining operation and maintenance knowledge text in a preset operation and maintenance knowledge database;

[0013] Converting the operation and maintenance knowledge text into a vector representation to obtain the operation and maintenance knowledge vector;

[0014] The operation and maintenance knowledge vector is stored in an operation and maintenance knowledge vector library.

[0015] In a feasible implementation manner, the method further includes:

[0016] Performing word segmentation on the target text and removing stop words to obtain multiple vocabulary units; wherein the target text is the fault description text or the operation and maintenance knowledge text;

[0017] Input each vocabulary unit into the BERT model and convert it into a vector representation to obtain the vocabulary vector of each vocabulary unit;

[0018] Based on the weighted sum or average value of the vocabulary vectors of each vocabulary unit, the vector representation of the target text is obtained.

[0019] In a feasible implementation manner, the method further includes:

[0020] Collect data documents and perform data cleaning on the data documents; wherein the data documents include data center fault handling records, system technical configuration manuals, operation and maintenance manuals, and historical fault cases;

[0021] Performing text segmentation on the data document based on paragraphs to obtain operation and maintenance knowledge text;

[0022] The operation and maintenance knowledge text is stored in the operation and maintenance knowledge database.

[0023] In a feasible implementation manner, obtaining fault prompt information based on the operation and maintenance knowledge text corresponding to the operation and maintenance knowledge vector and the fault description text includes:

[0024] Creating a prompt template; wherein the prompt template includes a task description, background knowledge, and user input;

[0025] The operation and maintenance knowledge text corresponding to the operation and maintenance knowledge vector and the fault description text are input into the prompt template to obtain fault prompt information.

[0026] In a feasible implementation manner, the method further includes:

[0027] Based on a preset nearest neighbor search algorithm, an operation and maintenance knowledge vector in the operation and maintenance knowledge vector library that meets a similarity condition with the query vector is obtained.

[0028] In a feasible implementation manner, the method further includes:

[0029] Collect multiple fault analysis corpora to construct a fault analysis corpus set; wherein the fault analysis corpus includes fault description data and corresponding fault root cause analysis data;

[0030] The pre-trained large language model is subjected to parameter fine-tuning training based on the fault analysis corpus; wherein the pre-training of the large language model is completed based on a general corpus.

[0031] According to a second aspect disclosed in the present application, the present application provides a fault root cause analysis device, comprising:

[0032] A vector conversion module is used to convert the fault description text input by the user into a vector representation to obtain a query vector;

[0033] A vector query module, used to perform a query search in a preset operation and maintenance knowledge vector library according to the query vector, and obtain an operation and maintenance knowledge vector in the operation and maintenance knowledge vector library that meets a similarity condition with the query vector; wherein the operation and maintenance knowledge vector is a vector representation of a corresponding operation and maintenance knowledge text;

[0034] A prompt generating module, used for obtaining fault prompt information based on the operation and maintenance knowledge text corresponding to the operation and maintenance knowledge vector and the fault description text;

[0035] The fault analysis module is used to input the fault prompt information into the large language model to obtain the fault root cause analysis result.

[0036] According to a third aspect disclosed in the present application, the present application provides an electronic device, including a processor, and a memory communicatively connected to the processor;

[0037] The memory stores computer-executable instructions;

[0038] The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of the first aspects.

[0039] According to the fourth aspect disclosed in the present application, the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed, they are used to implement any method in the first aspect.

[0040] According to a fifth aspect disclosed in the present application, the present application provides a computer program product, including a computer program, and when the computer program is executed, it is used to implement any one of the methods in the first aspect.

[0041] Compared with the prior art, this application has the following beneficial effects:

[0042] The present application provides a method, device, electronic device and storage medium for fault root cause analysis. By integrating a large language model (LLM) and retrieval enhancement technology (RAG), the method effectively combines the powerful text processing capabilities of the large language model and the rich operation and maintenance knowledge provided by the enhanced retrieval technology to automatically process and analyze the massive logs and monitoring data generated by the data center, and can more accurately identify the fault mode and root cause from a large amount of data, thereby improving the accuracy of fault root cause analysis, improving the efficiency of fault root cause analysis, and reducing the burden on operation and maintenance personnel. It can also better adapt to changing monitoring scenarios and needs, and accurate fault root cause analysis also provides strong decision support for operation and maintenance personnel, helping them to formulate more effective fault response strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0044] Figure 1 A flowchart of a fault root cause analysis method provided in an embodiment of the present application;

[0045] Figure 2 A flowchart of another fault root cause analysis method provided in an embodiment of the present application;

[0046] Figure 3 A schematic diagram of the structure of a fault root cause analysis device provided in an embodiment of the present application;

[0047] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0048] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0049] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0050] Data Center is a key infrastructure of modern information technology, used to realize the centralized processing, storage, transmission, exchange and management of data information. It includes servers, storage devices, switches, routers, firewalls and other equipment, as well as supporting networks, redundant data communication connections, environmental control, monitoring and safety devices. The main function of the data center is to provide a stable server and network environment, ensure the security and reliability of data, and meet the user's data storage, computing, transmission and management needs.

[0051] As a key business support platform, the stability and reliability of the data center of a financial institution are of vital importance. Any infrastructure failure may lead to business interruption, affect the safe production and operation of the data center, and cause economic losses to the financial institution.

[0052] With the development of technology, the scale and complexity of data centers continue to increase, involving more and more equipment and systems, making fault diagnosis more difficult. When a data center fails, operation and maintenance personnel need to locate the fault location and diagnose the fault type from the massive amount of monitoring data and log information generated by the data center. However, traditional manual analysis methods are difficult to cope with such a large amount of data, resulting in the inability to quickly and accurately locate the fault location and diagnose the fault type, and there is a problem of low efficiency in fault root cause analysis.

[0053] In response to the above technical problems, this application proposes a fault root cause analysis method, which can automatically perform fault root cause analysis by integrating a large language model and retrieval enhancement technology, effectively improving the efficiency of fault root cause analysis. A more intelligent, efficient and automated fault root cause analysis system is created to cope with the complexity and dynamics of data center operation and maintenance.

[0054] The technical solution of the fault root cause analysis method provided by the present application is described in detail below through specific embodiments. It should be noted that the following embodiments may exist alone or in combination with each other, and the same or similar contents may not be described repeatedly in different embodiments.

[0055] Figure 1 A schematic diagram of a method for analyzing the root cause of a fault provided by an embodiment of the present application is shown in FIG. Figure 1 In some embodiments, the process of the fault root cause analysis method includes the following steps:

[0056] S101, converting the fault description text input by the user into a vector representation to obtain a query vector.

[0057] Among them, the advantage of converting the input text data into vector representation is that it can quickly retrieve the data most relevant to the query conditions by calculating the similarity between vectors, which is especially useful when large-scale unstructured data is involved. Therefore, by converting the fault description text entered by the user into a vector representation, the query vector can be used to quickly query the operation and maintenance knowledge text related to the fault description text in the subsequent process.

[0058] S102, performing a query search in a preset operation and maintenance knowledge vector library according to the query vector, and obtaining an operation and maintenance knowledge vector in the operation and maintenance knowledge vector library that meets a similarity condition with the query vector; wherein the operation and maintenance knowledge vector is a vector representation of the operation and maintenance knowledge text.

[0059] Among them, according to the query vector, the operation and maintenance knowledge vector that meets the similarity condition with the query vector is quickly searched in the operation and maintenance knowledge vector library, and the corresponding operation and maintenance knowledge text is obtained according to the queried operation and maintenance knowledge vector. Since the text itself is a kind of unstructured data, it is difficult to directly compare and analyze. By converting text into vectors, unstructured text data can be converted into structured numerical data, which makes it easier to judge whether the texts are similar.

[0060] Specifically, a vector database is a database system specifically used to store, retrieve, and analyze vector data. These vectors are usually high-dimensional numerical vectors converted from raw data (such as text, images, audio, and other unstructured data) through embedding technology.

[0061] S103, obtaining fault prompt information based on the operation and maintenance knowledge text and the fault description text corresponding to the operation and maintenance knowledge vector.

[0062] The queried operation and maintenance knowledge text and the fault description text entered by the user are combined to form complete fault prompt information. The operation and maintenance knowledge text obtained by querying the retrieval-augmented generation (RAG) technology is used to expand the fault description text, so as to provide richer context information for the subsequent large language model fault root cause analysis, thereby improving the accuracy of fault root cause analysis.

[0063] Specifically, Retrieval-Augmented Generation (RAG) technology is a natural language processing technology that combines information retrieval (Retrieval) and text generation (Generation). The core idea of ​​RAG is to improve the accuracy and relevance of generated text by retrieving information related to the current task from a large amount of document data and incorporating it into the generation process of the language model. This method is particularly suitable for scenarios that require reference to external knowledge or data.

[0064] S104, inputting the fault prompt information into the large language model to obtain the fault root cause analysis result.

[0065] The generated fault prompt information is input into the large language model, and the fault root cause analysis is performed through the large language model to obtain the fault root cause analysis result.

[0066] Specifically, the root cause analysis results include fault cause analysis, fault type analysis, fault location analysis, fault operation and maintenance suggestions, etc. Through the root cause analysis results, operation and maintenance personnel can quickly understand the cause of the data center failure, locate the fault location in time, and achieve the purpose of rapid maintenance.

[0067] Specifically, Large Language Model (LLM) is a class of machine learning models designed to handle natural language understanding and generation tasks. These models usually have a large number of parameters (usually reaching billions or even tens of billions), so they are called "large" models. The working principle of LLM is usually based on the Transformer architecture, which is a deep learning model that is particularly suitable for processing sequence data. The Transformer architecture uses the Self-Attention Mechanism to process each element in the sequence, allowing the model to take into account other elements in the sequence when processing the current element, which makes LLM perform well in capturing the contextual information of the text.

[0068] In this embodiment, by integrating the large language model (LLM) and the retrieval enhancement technology (RAG), the powerful text processing capability of the large language model and the rich operation and maintenance knowledge provided by the enhanced retrieval technology are effectively combined to automatically process and analyze the massive logs and monitoring data generated by the data center, and the fault mode and root cause can be more accurately identified from a large amount of data, thereby improving the accuracy of the root cause analysis of the fault, improving the efficiency of the root cause analysis of the fault, and reducing the burden on the operation and maintenance personnel. It can also better adapt to the changing monitoring scenarios and needs, and the accurate root cause analysis of the fault also provides strong decision-making support for the operation and maintenance personnel, helping them to formulate more effective fault response strategies.

[0069] exist Figure 1 Based on the embodiment shown, the following Figure 2 , the technical solution of the above-mentioned fault root cause analysis method is further introduced.

[0070] Figure 2 A flowchart of another fault root cause analysis method provided in an embodiment of the present application is shown in FIG. Figure 2 In some embodiments, the process of the fault root cause analysis method includes the following steps:

[0071] S201, collecting data documents and performing data cleaning on the data documents; wherein the data documents include the data center's fault handling records, system technical configuration manuals, operation and maintenance manuals, and historical fault cases.

[0072] Among them, data cleaning is used to identify, correct and remove erroneous, invalid, redundant, incomplete, inconsistent or repeated information in data documents to improve the accuracy, consistency and completeness of the data, ensure that the data can truly reflect the business situation, and provide a reliable data foundation for subsequent data analysis and decision-making.

[0073] Specifically, fault handling records, system technical configuration manuals, operation and maintenance manuals, and historical fault cases all contain operation and maintenance knowledge data related to fault root cause analysis.

[0074] S202, performing text segmentation on the data document based on paragraphs to obtain operation and maintenance knowledge text.

[0075] Among them, for a data document, one paragraph generally includes a complete piece of data information. Therefore, the data document is segmented based on the paragraphs and the data document is segmented into text blocks to obtain the operation and maintenance knowledge text.

[0076] S203, storing the operation and maintenance knowledge text into the operation and maintenance knowledge database.

[0077] The operation and maintenance knowledge text is stored in the operation and maintenance knowledge database as text content for subsequent expansion of the fault description text.

[0078] S204, obtaining operation and maintenance knowledge text in a preset operation and maintenance knowledge database.

[0079] S205, converting the operation and maintenance knowledge text into a vector representation to obtain an operation and maintenance knowledge vector.

[0080] Among them, the operation and maintenance knowledge text is converted into the corresponding operation and maintenance knowledge vector, so as to facilitate the subsequent use of the query vector for data query.

[0081] S206, storing the operation and maintenance knowledge vector into an operation and maintenance knowledge vector library.

[0082] S207, converting the fault description text input by the user into a vector representation to obtain a query vector.

[0083] Preferably, when searching for operation and maintenance knowledge text related to the fault description text, it is necessary to convert the fault description text or the operation and maintenance knowledge text into a vector representation, specifically including:

[0084] Step 1: segment the target text and remove stop words to obtain multiple vocabulary units; wherein the target text is a fault description text or an operation and maintenance knowledge text;

[0085] Among them, word segmentation is the process of splitting text into individual words or phrases, and removing stop words is to remove words that frequently appear in the text but contribute little to the meaning of the text, such as "de", "le", "zai", etc. Removing these words can reduce data sparsity and improve the efficiency of subsequent processing.

[0086] Step 2: Input each lexical unit into the BERT model to convert it into a vector representation, and obtain the lexical vectors of each lexical unit.

[0087] Among them, the BERT model, whose full name is Bidirectional Encoder Representations from Transformers, is an open-source machine learning framework designed for the field of natural language processing (NLP). The BERT model adopts a bidirectional Transformer architecture and can consider the context information before and after words simultaneously. This bidirectional encoding ability enables BERT to better understand the meaning and association of words in a sentence and generate more semantically representative word vectors. Therefore, the word vector representation obtained using BERT can more accurately reflect the meaning of a word in a specific context.

[0088] Step 3: Based on the weighted sum result or average result of the lexical vectors of each lexical unit, obtain the vector representation of the target text.

[0089] Among them, taking the weighted sum or average of the lexical vectors of each lexical unit can handle texts of different lengths because no matter how many words the text contains, a fixed-length vector representation can ultimately be obtained. Once the text is converted into a vector representation, various distance metrics (such as cosine similarity) can be used to calculate the similarity between texts, which is beneficial for subsequent information retrieval tasks.

[0090] Specifically, for the weighted sum, by assigning a weight to each lexical vector, the importance of certain words to the overall meaning of the text can be emphasized. Specifically, the weight can be determined based on the frequency of the word, TF-IDF value, the norm of the word embedding, or other features.

[0091] Specifically, the average is to calculate the average of all lexical vectors, which assumes that the contribution of all words to the overall meaning of the text is equal. Although this method may ignore the influence of some important words, it can provide a good enough text representation in many cases and has the advantages of simplicity and effectiveness.

[0092] S208: Query and search in the preset operation and maintenance knowledge vector library according to the query vector, and obtain the operation and maintenance knowledge vectors in the operation and maintenance knowledge vector library that meet the similarity condition with the query vector; among them, the operation and maintenance knowledge vector is the vector representation of the operation and maintenance knowledge text.

[0093] Preferably, based on a preset nearest neighbor search algorithm, an operation and maintenance knowledge vector in the operation and maintenance knowledge vector library that satisfies a similarity condition with the query vector is obtained.

[0094] Among them, the Approximate Nearest Neighbor (ANN) algorithm is a method for finding a data point in a data set that is very close to a given query point. The core idea is to quickly find a data point that is close enough to a given query vector in a large-scale data set without an exhaustive search. This method allows the returned result to not be the absolute closest, but is usually within an acceptable error range. In this way, the ANN algorithm can significantly increase the search speed while reducing the demand for computing resources.

[0095] Specifically, the similarity between vectors can be determined by the distance between vectors. Specifically, it includes:

[0096] Step 1: Obtain the similarity between the operation and maintenance knowledge vector and the query vector.

[0097] Specifically, the similarity between the operation and maintenance knowledge vector and the query vector can be obtained by using algorithms such as the cosine similarity algorithm, the Euclidean distance algorithm, and the Manhattan distance algorithm.

[0098] Step 2: If the similarity is greater than a preset threshold, the operation and maintenance knowledge vector is returned as the search result.

[0099] Among them, by presetting a similarity threshold, when the similarity between the operation and maintenance knowledge vector and the query vector is greater than the threshold, it is determined that the operation and maintenance knowledge vector and the query vector meet the similarity condition, and the operation and maintenance knowledge vector is returned as the search result.

[0100] Preferably, in order to avoid too many search results in the range, a quantity limit value can also be set, and the operation and maintenance knowledge vectors that meet the preset threshold are sorted by similarity, and multiple operation and maintenance knowledge vectors with greater similarity are selected based on the quantity limit value as search results to be returned.

[0101] S209, creating a prompt template; wherein the prompt template includes task description, background knowledge and user input.

[0102] Among them, using the structured format contained in the prompt template helps the model better understand the task requirements and guide the large language model to generate more accurate and relevant content.

[0103] Specifically, the prompt template includes task description, background knowledge, and user input.

[0104] Task description: Please generate the root cause analysis results based on the following retrieved information and user input text.

[0105] Background knowledge: Fill in the relevant operation and maintenance knowledge text information retrieved from the operation and maintenance knowledge database.

[0106] User input: Fill in the user's original or rewritten fault description text.

[0107] S210, inputting the operation and maintenance knowledge text and the fault description text corresponding to the operation and maintenance knowledge vector into the prompt template to obtain fault prompt information.

[0108] Among them, the operation and maintenance knowledge text corresponding to the operation and maintenance knowledge vector returned by the query is obtained from the operation and maintenance knowledge database, and the queried operation and maintenance knowledge text is combined with the fault description text input by the user and input into the preset prompt template to obtain the fault prompt information.

[0109] S211, input the fault prompt information into the large language model to obtain the fault root cause analysis result.

[0110] Preferably, in order to improve the fault root cause analysis capability of the large language model, the large language model is fine-tuned using a dedicated corpus, specifically including:

[0111] Step 1: collect multiple fault analysis corpora and construct a fault analysis corpus set; wherein the fault analysis corpus includes fault description data and corresponding fault root cause analysis data.

[0112] Herein, a corpus with fault description data and corresponding fault root cause analysis data is collected to adaptively adjust the large language model.

[0113] Step 2: fine-tune the parameters of the pre-trained large language model based on the fault analysis corpus; wherein the large language model is pre-trained based on a general corpus.

[0114] Fine-tuning is the process of further training a large language model (LLM) on a specific task or domain, aiming to improve the performance of the LLM on a specific task or domain. Fine-tuning can significantly improve the performance of the model on a specific task, making it better adapted to the specific requirements of the target task or data distribution. In addition, fine-tuning can also make the LLM more interpretable, which helps to debug and understand the behavior of the model.

[0115] In addition, by learning from historical failure data, LLM and RAG technologies can also help predict potential failures and achieve preventive maintenance.

[0116] Specifically, the fine-tuning algorithms for large language models include Prefix Tuning, Prompt Tuning, and Supervised Fine Tuning.

[0117] Specifically, a general corpus is a set of text data used to train and evaluate models. These data are usually sourced from a variety of sources to ensure the wide applicability and accuracy of the model. Take large-scale corpora such as C4 and Pile as examples. They crawl large amounts of text from the Internet and apply filters to clean and process them, ultimately forming high-quality corpora. These corpora play an important role in the training of large language models, improving the performance and generalization capabilities of the models.

[0118] In this embodiment, by integrating a large language model and retrieval enhancement technology, fault root cause analysis can be performed automatically, effectively improving the efficiency of fault root cause analysis.

[0119] Figure 3 is a schematic diagram of a fault root cause analysis device provided in an embodiment of the present application, see Figure 3 The fault root cause analysis device includes various functional modules for implementing the aforementioned fault root cause analysis method, and any functional module can be implemented by software and / or hardware.

[0120] In some embodiments, the fault root cause analysis device 300 includes a vector conversion module 301, a vector query module 302, a prompt generation module 303 and a fault analysis module 304. Among them:

[0121] The vector conversion module 301 is used to convert the fault description text input by the user into a vector representation to obtain a query vector;

[0122] The vector query module 302 is used to perform a query search in a preset operation and maintenance knowledge vector library according to the query vector, and obtain an operation and maintenance knowledge vector in the operation and maintenance knowledge vector library that meets the similarity condition with the query vector; wherein the operation and maintenance knowledge vector is a vector representation of the corresponding operation and maintenance knowledge text;

[0123] The prompt generation module 303 is used to obtain fault prompt information based on the operation and maintenance knowledge text and the fault description text corresponding to the operation and maintenance knowledge vector;

[0124] The fault analysis module 304 is used to input the fault prompt information into the large language model to obtain the fault root cause analysis result.

[0125] In some embodiments, the fault root cause analysis device 300 further includes a vector library construction module 305, and the vector library construction module 305 is specifically used to:

[0126] Obtaining operation and maintenance knowledge text in a preset operation and maintenance knowledge database;

[0127] Convert the operation and maintenance knowledge text into a vector representation to obtain the operation and maintenance knowledge vector;

[0128] The operation and maintenance knowledge vector is stored in the operation and maintenance knowledge vector library.

[0129] In some embodiments, the vector conversion module 301 is specifically used for:

[0130] Perform word segmentation on the target text and remove stop words to obtain multiple vocabulary units; wherein the target text is a fault description text or an operation and maintenance knowledge text;

[0131] Input each vocabulary unit into the BERT model and convert it into a vector representation to obtain the vocabulary vector of each vocabulary unit;

[0132] Based on the weighted sum or average of the vocabulary vectors of each vocabulary unit, the vector representation of the target text is obtained.

[0133] In some embodiments, the fault root cause analysis device 300 further includes a database construction module 306, and the database construction module 306 is specifically used to:

[0134] Collect data documents and clean them; data documents include data center fault handling records, system technical configuration manuals, operation and maintenance manuals, and historical fault cases;

[0135] Perform text segmentation on data documents based on paragraphs to obtain operation and maintenance knowledge text;

[0136] The operation and maintenance knowledge text is stored in the operation and maintenance knowledge database.

[0137] In some embodiments, the prompt generation module 303 is specifically used to:

[0138] Creating a prompt template; wherein the prompt template includes a task description, background knowledge, and user input;

[0139] The operation and maintenance knowledge text and the fault description text corresponding to the operation and maintenance knowledge vector are input into the prompt template to obtain the fault prompt information.

[0140] In some embodiments, the vector query module 302 is further configured to:

[0141] Based on the preset nearest neighbor search algorithm, the operation and maintenance knowledge vector in the operation and maintenance knowledge vector library that meets the similarity condition with the query vector is obtained.

[0142] In some embodiments, the fault root cause analysis device 300 further includes a model fine-tuning module 307, and the model fine-tuning module 307 is specifically used to:

[0143] Collect multiple fault analysis corpora and construct a fault analysis corpus set; wherein the fault analysis corpus includes fault description data and corresponding fault root cause analysis data;

[0144] The pre-trained large language model is fine-tuned based on the fault analysis corpus. The large language model is pre-trained based on a general corpus.

[0145] The fault root cause analysis device 300 provided in the embodiment of the present application is used to execute the technical solution provided in the aforementioned fault root cause analysis method embodiment. Its implementation principle and technical effect are similar to those in the aforementioned method embodiment and will not be repeated here.

[0146] It should be noted that it should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements, or all in the form of hardware, or some modules can be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the vector conversion module 301 can be a separately established processing element, or it can be integrated in a certain chip of the above device for implementation. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a certain processing element of the above device. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0147] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application, see Figure 4 The electronic device 400 includes a processor 401 and a memory 02 communicatively connected to the processor 401;

[0148] Memory 402 stores computer executable instructions;

[0149] The processor 401 executes the computer-executable instructions stored in the memory 402 to implement the technical solution of the aforementioned fault root cause analysis method.

[0150] In the above-mentioned electronic device 400, the memory 402 and the processor 401 are electrically connected directly or indirectly to realize data transmission or interaction. For example, these elements can be electrically connected to each other through one or more communication buses or signal lines, such as through a bus. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc., but it does not mean that there is only one bus or one type of bus. The memory 402 stores computer execution instructions for implementing the above-mentioned root cause analysis method, including at least one software function module that can be stored in the memory 402 in the form of software or firmware. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402.

[0151] The memory 402 includes at least one type of readable storage medium, including but not limited to random access memory (RAM), read only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), and electrically erasable read-only memory (EEPROM). The memory 402 is used to store programs, and the processor 401 executes the programs after receiving the execution instruction. Furthermore, the software programs and modules in the above-mentioned memory 402 may also include an operating system, which may include various software components and / or drivers for managing system tasks (such as memory management, storage device control, power management, etc.), and may communicate with various hardware or software components to provide an operating environment for other software components.

[0152] The processor 401 may be an integrated circuit chip having the ability to process signals. The above-mentioned processor 401 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), etc. The disclosed methods, steps and logic block diagrams in the embodiments of the present application may be implemented or executed. The general-purpose processor may be a microprocessor, or the processor 401 may be any conventional processor, etc.

[0153] The electronic device 400 is used to execute the technical solution provided by the aforementioned fault root cause analysis method embodiment. Its implementation principle and technical effect are similar to those in the aforementioned method embodiment and will not be repeated here.

[0154] An embodiment of the present application further provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed, they are used to implement the technical solution of the aforementioned fault root cause analysis method.

[0155] The computer-readable storage medium mentioned above can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The computer-readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0156] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist as discrete components in the control device of the fault root cause analysis device.

[0157] The embodiment of the present application also provides a computer program product, including a computer program, which, when executed, is used to implement the technical solution of the aforementioned fault root cause analysis method.

[0158] In the above embodiments, those skilled in the art will appreciate that the implementation of the above method embodiments can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless network, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a DVD), or a semiconductor medium (eg, a solid state disk (SSD)).

[0159] In the above embodiments, the description of each embodiment has its own emphasis. For the part not described in detail in a certain embodiment, please refer to the relevant description of other embodiments. The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not 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.

[0160] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the appended claims.

[0161] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for root cause analysis of a fault, characterized in that: include: Convert the fault description text entered by the user into a vector representation to obtain a query vector; Performing a query search in a preset operation and maintenance knowledge vector library according to the query vector, and obtaining an operation and maintenance knowledge vector in the operation and maintenance knowledge vector library that satisfies a similarity condition with the query vector; wherein the operation and maintenance knowledge vector is a vector representation of an operation and maintenance knowledge text; Obtaining fault prompt information based on the operation and maintenance knowledge text corresponding to the operation and maintenance knowledge vector and the fault description text; The fault prompt information is input into a large language model to obtain a fault root cause analysis result.

2. The method according to claim 1, characterized in that The method further comprises: Obtaining operation and maintenance knowledge text in a preset operation and maintenance knowledge database; Converting the operation and maintenance knowledge text into a vector representation to obtain the operation and maintenance knowledge vector; The operation and maintenance knowledge vector is stored in an operation and maintenance knowledge vector library.

3. The method according to claim 2, characterized in that The method further comprises: Performing word segmentation on the target text and removing stop words to obtain multiple vocabulary units; wherein the target text is the fault description text or the operation and maintenance knowledge text; Input each vocabulary unit into the BERT model and convert it into a vector representation to obtain the vocabulary vector of each vocabulary unit; Based on the weighted sum or average value of the vocabulary vectors of each vocabulary unit, the vector representation of the target text is obtained.

4. The method according to claim 2, characterized in that: The method further comprises: Collect data documents and perform data cleaning on the data documents; wherein the data documents include data center fault handling records, system technical configuration manuals, operation and maintenance manuals, and historical fault cases; Performing text segmentation on the data document based on paragraphs to obtain operation and maintenance knowledge text; The operation and maintenance knowledge text is stored in the operation and maintenance knowledge database.

5. The method according to any one of claims 1 to 4, characterized in that: Obtaining fault prompt information based on the operation and maintenance knowledge text corresponding to the operation and maintenance knowledge vector and the fault description text, including: Creating a prompt template; wherein the prompt template includes a task description, background knowledge, and user input; The operation and maintenance knowledge text corresponding to the operation and maintenance knowledge vector and the fault description text are input into the prompt template to obtain fault prompt information.

6. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: Based on a preset nearest neighbor search algorithm, an operation and maintenance knowledge vector in the operation and maintenance knowledge vector library that meets a similarity condition with the query vector is obtained.

7. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: Collect multiple fault analysis corpora to construct a fault analysis corpus set; wherein the fault analysis corpus includes fault description data and corresponding fault root cause analysis data; The pre-trained large language model is subjected to parameter fine-tuning training based on the fault analysis corpus; wherein the pre-training of the large language model is completed based on a general corpus.

8. A fault root cause analysis device, characterized in that: include: A vector conversion module is used to convert the fault description text input by the user into a vector representation to obtain a query vector; A vector query module, used to perform a query search in a preset operation and maintenance knowledge vector library according to the query vector, and obtain an operation and maintenance knowledge vector in the operation and maintenance knowledge vector library that meets a similarity condition with the query vector; wherein the operation and maintenance knowledge vector is a vector representation of a corresponding operation and maintenance knowledge text; A prompt generating module, used for obtaining fault prompt information based on the operation and maintenance knowledge text corresponding to the operation and maintenance knowledge vector and the fault description text; The fault analysis module is used to input the fault prompt information into the large language model to obtain the fault root cause analysis result.

9. An electronic device, characterized in that: comprising a processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed, they are used to implement the method according to any one of claims 1 to 7.

Images