CN119691130A - Equipment fault prediction method and device, storage medium and electronic equipment - Google Patents

Abstract

The present application provides a method, device, storage medium and electronic device for predicting equipment failure, which obtains a text feature vector describing the problem of the abnormal condition of the equipment; matches the text feature vector from a preset question-and-answer feature vector database to determine whether the problem description is a problem description related to the equipment failure; if it is determined to be a problem description related to the failure, obtain the system failure prediction results and system solutions, network failure prediction results and network solutions, and graph prediction failure results and graph solutions from the question-and-answer knowledge base respectively; use a hybrid expert model to obtain the system failure prediction results and system solutions, network failure prediction results and network solutions, and graph prediction failure results and graph solutions from the question-and-answer knowledge base, and obtain the optimal failure prediction results and the corresponding optimal solutions. Thus, the failure prediction results and failure solutions are determined efficiently and accurately, thereby improving the operation and maintenance efficiency of the equipment.

Description

Equipment fault prediction method and device, storage medium and electronic equipment

Technical Field

The present application relates to the field of large language models, and in particular, to a method and apparatus for predicting an equipment failure, a storage medium, and an electronic device.

Background

In the daily operation and maintenance of certification equipment, fault diagnosis is a core task of ensuring system stability and business continuity. Operation and maintenance work has been a complex and difficult task in the IT industry. As the scale of systems increases and the complexity increases, so does the challenges facing the operators. In operation and maintenance work, the log is a main basis for troubleshooting. However, as the complexity of systems increases, the amount of logs increases dramatically, and how to quickly find the point of failure therefrom becomes a significant challenge. Traditional log analysis methods often rely on manual screening, which is time consuming and prone to missing important information. The complexity of log information tends to be difficult to understand even if faults can be located. The operation and maintenance personnel need to interpret a large number of technical terms and error information, which is time-consuming and also easily causes erroneous interpretation. The solution of equipment operation faults is generally in an enterprise knowledge base, and needs to be periodically synchronized to operation and maintenance engineers in each province and city, so that knowledge synchronization is not timely. Therefore, how to provide a method capable of rapidly detecting equipment faults or anomalies, which can efficiently and accurately assist operation and maintenance personnel to learn equipment problems and provide a solution in a targeted manner, is convenient for the operation and maintenance personnel to improve operation and maintenance efficiency, and becomes a technical problem to be solved in industry.

Disclosure of Invention

In view of the above, embodiments of the present application provide a method, an apparatus, a storage medium, and an electronic device for predicting a device failure, so as to at least or partially solve the above problems.

In a first aspect, an embodiment of the present application provides a method for predicting an equipment failure, including:

acquiring a text feature vector for describing the problem of the abnormal condition of the equipment;

Matching the text feature vector from a preset question-answer feature vector database to determine whether the problem description is a problem description related to equipment failure;

If the text feature is determined to be the problem description related to the fault, performing feature matching on the text feature in a fault knowledge question-answering library by using a retrieval enhancement mode to obtain a system fault prediction result and a system solution of the question-answering knowledge library;

And obtaining a fault optimal prediction result and a corresponding optimal solution based on the system fault prediction result and the system solution, the networking fault prediction result and the networking solution, and the map prediction fault result and the map solution by using a mixed expert model.

Optionally, in one embodiment of the present application, the method further comprises obtaining a text feature vector describing a problem of an abnormal condition of the device;

Matching the text feature vector from a preset question-answer feature vector database to determine whether the problem description is a problem description related to equipment failure;

If the text feature is determined to be the problem description related to the fault, performing feature matching on the text feature in a fault knowledge question-answering library by using a retrieval enhancement mode to obtain a system fault prediction result and a system solution of the question-answering knowledge library;

And obtaining a fault optimal prediction result and a corresponding optimal solution based on the system fault prediction result and the system solution, the networking fault prediction result and the networking solution, and the map prediction fault result and the map solution by using a mixed expert model.

Optionally, in an embodiment of the present application, the feature matching is performed on the text feature in the fault knowledge question-answering library by using a search enhancement manner, to obtain a system fault prediction result and a system solution of the question-answering knowledge library, including:

Determining the fault knowledge question-answering library, wherein the fault knowledge question-answering library at least comprises one of a local fault knowledge question-answering library or an external fault knowledge question-answering library;

generating a query request according to the text characteristics by adopting a high-efficiency search algorithm to select related documents from the selected fault knowledge question-answering library;

extracting information from the related document, and extracting key information of the related document, wherein the key information comprises one or more of entity information, entity relation information and context information;

Fusing the key information by using a large language model to generate fault prediction information and prediction fault resolution information;

And performing at least one text input process of illusion suppression, noise information processing, document screening and error correction on the fault prediction information and the prediction fault solution information to generate a system fault prediction result and a system solution of a question-answer knowledge base.

Optionally, in an embodiment of the present application, before the obtaining the graph prediction fault result and the graph solution according to the text feature vector in combination with the constructed knowledge graph for the equipment fault, the method further includes constructing a knowledge graph of the equipment fault;

determining equipment operation and maintenance information consisting of fault description, solution and fault cause of the equipment from open world data;

converting the operation and maintenance information into structural semantic block data in a predefined format;

Performing incremental entity extraction and incremental relation extraction on the structured semantic block data to obtain unique entity information in each semantic block, and determining relation information among the entities;

Integrating the acquired unique entity information and the acquired relationship information into a preset graphic database, and generating a specific fault knowledge graph and a specific equipment knowledge graph by combining equipment related knowledge graph subgraphs retrieved from the existing knowledge graph, wherein the specific equipment knowledge graph has sequence data during the operation of the specific equipment;

Performing node and edge clustering on the generated specific fault knowledge graph, and creating a fault aggregation knowledge graph;

And combining the related specific fault pattern and time information about sequential data in the specific knowledge pattern, generating a personalized knowledge pattern of the equipment, and determining the personalized knowledge pattern as the knowledge pattern of the equipment fault.

Optionally, in an embodiment of the present application, the obtaining the graph prediction fault result and the graph solution according to the text feature vector in combination with the constructed knowledge graph for the equipment fault includes performing fault prediction on the equipment based on the personalized knowledge graph by using dual attention enhancement and a graph neural network according to the text feature vector to obtain the graph prediction fault result and the graph solution.

Optionally, in one embodiment of the present application, the method further comprises periodically acquiring a device operational knowledge base;

and extracting key information in the equipment operation knowledge base by using the trained large language model, wherein the key information at least comprises one of equipment problem description, fault reason description and fault solution.

And according to the extracted key information, updating data of a knowledge graph or a fault knowledge base of the equipment fault.

Optionally, in an embodiment of the present application, after the obtaining the fault optimal prediction result and the corresponding optimal solution, the method further includes:

Performing abnormality detection on the equipment according to the optimal fault prediction result;

And carrying out early warning prompt on the equipment according to the abnormal detection result.

Optionally, in an embodiment of the present application, the detecting the abnormality of the device according to the optimal prediction result of the fault includes:

creating a digital twin analog device for the device;

according to simulation operation data obtained by the simulation equipment through working state simulation, actual monitoring data obtained by monitoring the working state of the equipment, and statistical data obtained by statistics of the working state of the equipment of the same type, carrying out data fusion to obtain fusion data;

Constructing a health index and equipment residual life prediction neural network model aiming at the equipment according to the fusion data;

and performing performance evaluation on the digital twin simulation equipment by using a health index and equipment residual life prediction neural network model and combining the fault optimal prediction result so as to determine the abnormal detection result according to the performance evaluation result.

Optionally, in an embodiment of the present application, the creating, at the server, a digital twin simulation device for the device includes:

Collecting multi-source data of the device;

Based on the multi-source data, a digital twin simulation device of the device is created using a preset geometric model, a data analysis model and a physical model.

In a second aspect, based on the method for predicting an equipment failure according to the first aspect of the present application, an embodiment of the present application further provides an apparatus for predicting an equipment failure, including:

The extraction module is used for obtaining text feature vectors for describing problems of abnormal conditions of equipment;

The matching module is used for matching the text feature vector from a preset question-answer feature vector database so as to determine whether the problem description is related to equipment faults;

The system comprises a search module, a graph prediction result and a graph solution, wherein the search module is used for carrying out feature matching on the text features in a fault knowledge question-answer library by using a search enhancement mode if the text features are determined to be related to the faults;

And the generation module is used for obtaining a fault optimal prediction result and a corresponding optimal solution based on the system fault prediction result and the system solution, the networking fault prediction result and the networking solution and the map prediction fault result and the map solution by utilizing a mixed expert model.

In a third aspect, an embodiment of the present application further provides a computer storage medium, where computer executable instructions are stored, where the computer executable instructions, when executed, perform a method for predicting any device failure according to the first aspect of the embodiment of the present application.

In a fourth aspect, the embodiment of the application also provides an electronic device, which comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;

the memory is configured to store at least one executable instruction that causes the processor to perform any one of the method for predicting a device failure according to the first aspect of the embodiment of the present application.

The application provides a prediction method, a device, a storage medium and electronic equipment for equipment faults, which are used for acquiring text feature vectors for carrying out problem description on abnormal conditions of the equipment, matching the text feature vectors from a preset question feature vector database to determine whether the problem description is related to equipment faults or not, carrying out feature matching on the text features in a fault knowledge question-answer library by using a retrieval enhancement mode if the problem description is determined to be related to the faults, acquiring a system fault prediction result and a system solution of a question-answer knowledge library, acquiring a networking fault prediction result and a networking solution by a networking query mode according to the text feature vectors, acquiring a map prediction fault result and a map solution according to the text feature vectors in combination with a constructed knowledge map for the equipment faults, and acquiring a fault optimal prediction result and a corresponding optimal solution by utilizing a mixed expert model based on the system fault prediction result and the system solution, the fault prediction result and the networking solution and the map prediction fault result and the map solution. The equipment fault prediction mode provided by the embodiment of the application can efficiently provide a relatively accurate fault prediction result for a user, and provides an optimal solution in a targeted manner, so that the experience of the user is obviously improved, and the operation and maintenance efficiency of the equipment is effectively improved.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments described in the embodiments of the present application, and other drawings may be obtained according to these drawings for a person having ordinary skill in the art.

FIG. 1 is a schematic workflow diagram of a method for predicting equipment failure according to an embodiment of the present application;

fig. 2 is a schematic structural diagram of a device for predicting equipment failure according to an embodiment of the present application;

fig. 3 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

In order to better understand the technical solutions in the embodiments of the present application, the following description will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application, and it is obvious that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which are derived by a person skilled in the art based on the embodiments of the present application, shall fall within the scope of protection of the embodiments of the present application.

It should be understood that the various steps recited in the method embodiments of the present application may be performed in a different order and/or performed in parallel. Furthermore, method embodiments may include additional steps and/or omit performing the illustrated steps. The scope of the application is not limited in this respect.

Embodiment 1,

An embodiment of the present application provides a method for predicting an equipment failure, as shown in fig. 1, and fig. 1 is a schematic workflow diagram of the method for predicting an equipment failure provided in the embodiment of the present application, including:

step S101, obtaining a text feature vector for describing the problem of the abnormal condition of the equipment. In the embodiment of the application, the problem is described as text content of a question which is asked to the system by a user according to the actual abnormal condition of the equipment.

Optionally, in one implementation manner of the embodiment of the application, acquiring the text feature vector for carrying out problem description on the abnormal condition of the equipment comprises the steps of preprocessing text content of the problem description, acquiring a preprocessed text, and extracting the text feature vector from the preprocessed text to acquire the text feature vector, wherein the preprocessing comprises the step of converting the problem description into text data in a structured preset format conforming to system identification. So that the system can efficiently and accurately identify the problem description content, thereby accurately extracting the text feature vector.

And step S102, matching the text feature vector from a preset question and answer feature vector database to determine whether the problem description is related to equipment faults. In the embodiment of the application, whether the current system question of the user is the question description related to the equipment fault is determined by the characteristic matching mode of the question description, the process is simple and easy to implement, and the judgment accuracy is high.

Step S103, if the description of the fault related problems is determined, performing feature matching on the text features in a fault knowledge question-answering library by using a retrieval enhancement mode to obtain a system fault prediction result and a system solution of the question-answering knowledge library, obtaining a networking fault prediction result and a networking solution by a networking query mode according to the text feature vector, and obtaining the map prediction fault result and the map solution according to the text feature vector and a constructed knowledge map aiming at the equipment fault.

Optionally, in one implementation manner of the embodiment of the application, a retrieval enhancement manner is used for carrying out feature matching on the text features in a fault knowledge question-answering library to obtain a system fault prediction result and a system solution of the question-answering knowledge library, wherein the method comprises the steps of determining the fault knowledge question-answering library, wherein the fault knowledge question-answering library at least comprises one of a local fault knowledge question-answering library or an external fault knowledge question-answering library; the method comprises the steps of generating a query request according to text characteristics, selecting related documents from a selected fault knowledge question-answer library, extracting information from the related documents, extracting key information of the related documents, wherein the key information comprises one or more of entity information (namely information such as fault description, solution or fault cause), entity relation information (namely fault description-fault cause relation, fault cause-solution relation) and context information, fusing the key information by using a large language model to generate fault prediction information and predictive fault solution information, and performing at least one text processing mode of illusion suppression and noise information processing on the fault prediction information and the predictive fault solution information to generate a system fault prediction result and a system solution of the question-answer knowledge library. The embodiment of the application firstly selects various knowledge which can cover the operation field of the equipment more comprehensively, such as fault problem description of the equipment, fault solution or knowledge base of content such as upgrade information of the equipment, so that the process for supporting matching of the problem description proposed by the user is more reliable and comprehensive, thereby ensuring the generation of the accurate and comprehensive text content for feeding back the problem to the user and improving the use satisfaction of the user. Further, in this stage, when checking and inquiring are performed again, efficient searching algorithm is adopted, such as

And the BM25 (Best Matching) or vector search algorithm is used for carrying out associated information search query on the selected knowledge question-answer library by the generated query request, so that the accuracy and efficiency of the search query can be improved. And more accurate and comprehensive generated fault prediction information and corresponding solutions can be obtained through information fusion. In the embodiment of the present application, the illusion suppression refers to checking whether the generated text is consistent with the known facts, for example, by comparing with the information in the knowledge base, or determining the accuracy of the generated text content by using a special fact checking model, and meanwhile, the generated text may be configured to be inconsistent with the facts by using constraint conditions, where the constraint conditions may be constructed based on language rules, domain knowledge or priori text content, and the embodiment of the present application is not limited in this regard. The noise processing is to remove the content of the document with low correlation or pseudo correlation in the retrieved document mainly through two text processing mechanisms, namely document screening and error correction. Meanwhile, a corresponding text correction processing algorithm is designed by using a text content analysis mode, and documents containing error information are corrected by using other reliable data through verification or an interactive supplement mode according to the extracted context information. Thereby improving the accuracy of the feedback document.

Optionally, in one implementation manner of the embodiment of the application, when generating the system fault prediction result and the system solution of the question-answer knowledge base, the method further comprises the step of performing text optimization on the system fault prediction result and the system solution, wherein the text optimization process performs text filtering, text compression and reordering on the data processing manner, so that the text content of the finally generated fault prediction result and the system solution is determined with better accuracy.

And step S104, obtaining a fault optimal prediction result and a corresponding optimal solution based on the system fault prediction result and the system solution, the networking fault prediction result and the networking solution, and the map prediction fault result and the map solution by utilizing a hybrid expert model. The hybrid expert model (MoE, holo Mixture ofExperts) is a deep learning technique that allows better predictive performance by combining multiple models (these models are called "experts") directly together to speed up model convergence. This model design strategy is particularly important in large models, which can solve some of the problems faced by large models when they are trained. In the embodiment of the application, the mixed expert model is used for combining the multiple failure sister results and the corresponding solutions to obtain the failure optimal prediction result and the corresponding optimal solution which are finally used for feedback, so that the efficiency is high and the accuracy is good.

Optionally, in an embodiment of the present application, after obtaining the optimal failure prediction result and the optimal solution corresponding to the optimal failure prediction result, the method further includes determining an abnormality detection result of the device based on the optimal failure prediction result in combination with operation data of the device, and performing early warning prompt according to the abnormality detection result. Specifically, by means of the method, operation and maintenance personnel can find faults of equipment in time so as to adjust and optimize equipment operation cost parameters, for example, for printing equipment, the equipment operation parameters can be optimized by generating targeted early warning prompts, and according to the environment where the equipment is deployed and the types of certificates, anti-counterfeiting materials and consumables, personalized parameter setting schemes aiming at the early warning prompts are generated according to processing and production processes such as thermal transfer printing, laser etching, color UV and the like, and the contents of relevant parameter adjustment such as the working temperature of the printing equipment and the position of a sensor are included, and the working temperature, the laminating speed and the speed of a conveying station of a laminating machine are set. The embodiments of the present application are not described herein. In the embodiment of the application, the content feedback about the problem description can be performed pertinently for the user, the abnormal detection can be further performed by combining the fault prediction result in the mode, and the early warning is performed on the abnormal detection result, so that the system can timely acquire the possible abnormal state and running risk of the equipment while feeding back the determined fault and solution to the user, the stable running of the equipment is further ensured, and the satisfaction degree of the user is improved.

Optionally, in an implementation manner of the embodiment of the present application, before obtaining the graph prediction fault result and the graph solution according to the text feature vector in combination with the constructed knowledge graph for the equipment fault, the method further includes determining equipment operation and maintenance information composed of the fault description, the solution and the fault cause of the equipment from open world data, converting the operation and maintenance information into structured semantic block data in a predefined format, performing incremental entity extraction and incremental relation extraction on the structured semantic block data to obtain unique entity information in each semantic block and determine relation information between the entities, integrating the obtained unique entity information and the relation information into a preset graph database, combining equipment related knowledge graph retrieved from existing knowledge graphs, generating a specific fault knowledge graph and a specific equipment knowledge graph, wherein the specific equipment knowledge graph has sequence data during operation of the specific equipment, performing node and edge clustering on the generated specific fault knowledge graph, combining the specific related fault and the specific equipment related knowledge graph to be the information about the time of the personalized knowledge graph, and generating the personalized knowledge graph. In the embodiment of the application, the open world data comprises system local data, internet data and the like so as to ensure the multisource and the comprehensiveness of the acquired data. The embodiment of the application can effectively ensure the abundance of the information covered by the constructed knowledge graph of the equipment fault, avoid the information containing semantic mixture and conflicting information, and facilitate the user to intuitively understand the relationship of complex equipment operation and maintenance knowledge points.

Optionally, in one embodiment of the present application, the obtaining the graph prediction fault result and the graph solution according to the text feature vector in combination with the constructed knowledge graph for the equipment fault includes performing fault prediction on the equipment based on the personalized knowledge graph by using dual attention enhancement (BAT) and Graph Neural Network (GNN) according to the text feature vector to obtain the graph prediction fault result and the graph solution, so as to improve accuracy of the obtained graph prediction fault result and the graph solution.

Specifically, in the implementation manner of the embodiment of the application, the specific process of constructing the knowledge graph can be realized through 4 functional modules, namely a document analysis module, an increment entity extraction module, an increment relation extraction module and a graph integration module. The constructed knowledge graph may be expressed as g= { E, R, F, T }, where E represents the entity set of the extracted device, R is the extracted entity relationship set, F represents the corresponding set of fault solutions, T represents the time at which the event occurred, and is combined by different types of relationships between the entities.

In the process of constructing the knowledge graph, the document analysis module is used for taking documents such as a solution of equipment operation faults, operation log data, an equipment operation and maintenance manual, a user manual and the like related to equipment use description as input, extracting semantic block entity information from the input documents and text feature vectors corresponding to the input documents by using a large language model to acquire an entity information set containing 4-tuple (e, r, f, t), wherein e represents equipment entities, the entities at least comprise equipment names, fault names and actual fault phenomena, r represents the relation among the entities, f represents related fault solutions, t represents the time of occurrence of fault events, and t _i<t_j is when i < j. The knowledge graph G _t of the constructed device at time t can be expressed as:

G_t＝{(e_i,r_i,f_i,t_i):e_i∈E,r_i∈R,f_i∈F,t_i<t}

When the knowledge graph is constructed, the document analysis module can also carry out quality filtering on the extracted semantic block data in advance to remove useless information in the document, so as to ensure that the text characteristic data of the constructed knowledge graph accords with the format and the content accords with the requirement of constructing the knowledge graph.

The increment entity extraction module is used for iteratively extracting the obtained semantic blocks by using a preset increment entity matcher and extracting global document entity information. Specifically, in the first step, a large language model is used to extract entities from the first semantic block D to form a global entity set E. According to constraint C, the large language model is constrained to extract entities representing a unique concept to avoid extracting semantic hybrid entities. For the semantic block data input subsequently, the entity extracted from the semantic block data is taken as a local entity E _d. And secondly, matching the extracted local entity E _d with the global entities in the local global entity set E. If a corresponding local entity E _i is matched in E, it is added to the matched set E _{d Matching} , which E _{d Matching} represents a set of incremental entities that are temporarily created when entity extraction is performed on these subsequently entered semantic block data. If not, a cosine similarity measure with a predefined threshold is used to search for similar entities in E. If no match is found, the local entity is added to E _{d Matching} , otherwise it is added to the best matching global entity E _i' based on the maximum similarity algorithm. And thirdly, unifying E and E _{d Matching} to update the global entity set E. This process is repeated for each semantic block, thereby generating a more comprehensive global entity set E.

The increment relation extraction module is used for extracting the global document relation by using the increment relation matcher to take the global document entity E as a context and each semantic block. The method is the same as that of the entity matcher, when the global entity is provided as a context, extracting the relation between semantic blocks by using a large language model for the entity which does not definitely exist in the semantic blocks, and performing incremental update on the entity relation obtained in the previous step to obtain a more comprehensive global entity relation;

The map integration module stores the global semantic entity and the global semantic relation obtained in the previous step into a map database, and constructs a knowledge map according to the map database. Preferably, in this embodiment, a Graph Neural Network (GNN) model based on a dual-attention enhancement mechanism is used to construct a knowledge graph, and the specific process is as follows:

The attention calculating method is characterized in that the representation of the map is { E, R, F, T }, at the first layer, an embedded representation Y ^l of the entity is generated by using a neural network model, and the initial input is that Y ⁰.D_l is the characteristic dimension of the first layer, and N= |E|.

The update function for each layer is:

Y^l(f)＝A(∑_e∈EAtn^l(e,r,f)·Msg^l(e,r,f))

A is an activation function, e.g. Leaky-ReLU

Atn ^l (e, r, f) attention weight for indicating the degree of influence of the equipment, the failure event entity e, on the solution f;

Msg ^l (e, r, f) message vector containing information of the device entity;

Wherein the dual attention enhancement mechanism of the neural network model represents the input features of each layer as a combination of subspaces, denoted Y ^l＝[Y^l,1,...,Y^l,H]∈R^N×Dl, where H is the number of attention heads, herein h=2, D ^l is the feature dimension of the first layer, D ^l is the feature dimension of the first layer, n= |e|.

The following formula is used in calculating the attention score S for each attention head:

W ^l[e]^T represents the key vector of the entity, W ^l represents a relationship-specific trainable weight matrix, Q ^l [ f ] is the query vector of the solution; is a matrix of relationships between entities and solutions;

In order to better perform time-based modeling to facilitate equipment failure prediction and residual life assessment, embodiments of the present application introduce time information into the model at this stage. In particular, a recurrent neural network (Recurrent Neural Network, RNN) is used for time-varying modeling Wherein, Time embedding, which is time t;

To capture the change of knowledge-graph over time, this stage of the embodiment of the application uses a recurrent neural network (Recurrent Neural Network, RNN) to update the temporal embedding (Temporal Embeddings) and structural embedding (Structural Embeddings) of entities and relationships.

Furthermore, the embodiment of the application can also update the full amount of the knowledge graph through probability modeling and learning of the dynamic knowledge graph. Specifically, probabilistic modeling is performed by defining a probabilistic model of the occurrence of an event:

wherein, Representing a new set of events that occur at time t. The conditional probability is decomposed into two parts, namely a structure and time:

And then carrying out iterative updating on the constructed knowledge graph by utilizing the probability modeling, and setting a loss function definition at least comprising time and structural information:

mu ₁ and mu ₂ are preset adjustable parameters;

According to the embodiment of the application, the final knowledge graph is constructed in the mode, so that the final constructed knowledge graph can be better ensured to have more comprehensive equipment entity information and relation information, and the accuracy of the output graph prediction fault result and the accuracy of the graph solution are further improved.

Optionally, in one embodiment of the application, the method further comprises the steps of periodically acquiring a device operation knowledge base, and extracting key information in the device operation knowledge base by using a trained large language model, wherein the key information at least comprises one of device problem description, fault reason description and fault solution. And according to the extracted key information, updating the data of the knowledge graph or the fault knowledge base of the equipment fault. To improve the novelty and accuracy of the system with respect to equipment failure feedback or solutions.

Specifically, in an optional implementation manner of the embodiment of the application, the device is subjected to abnormality detection according to the optimal failure prediction result, and the method comprises the steps of creating digital twin simulation equipment aiming at the device, carrying out data fusion according to simulation operation data obtained through working state simulation of the simulation equipment, actual monitoring data obtained through monitoring the working state of the device and statistical data obtained through statistics of the working state of the same type of device, constructing a health index and device residual life prediction neural network model according to the fusion data, carrying out performance evaluation on the digital twin simulation equipment by combining the optimal failure prediction result by using the health index and the device residual life prediction neural network model, and determining the abnormality detection result according to the performance evaluation result. According to the embodiment of the application, the digital twin simulation equipment is used as a monitoring target for monitoring equipment abnormality, and the system is not required to be connected with actual equipment, so that the effect of monitoring the health condition of the equipment with low cost is realized, complicated processes such as shutdown detection and the like are not required to be carried out on the actual equipment in the monitoring effect, and the equipment performance is accurately assessed in health when the equipment operation efficiency is ensured.

Optionally, in an actual application scene of the embodiment of the application, the digital twin simulation equipment for the equipment is created, and the method comprises the steps of collecting multi-source data of the equipment, and creating the digital twin simulation equipment for the equipment by using a preset geometric structure model, a data analysis model and a physical model according to the multi-source data. In the embodiment of the application, the geometric model can more accurately construct the appearance of each part of the digital twin model corresponding to the equipment according to the acquired equipment shape data such as the point cloud data of the equipment, and dynamically update the corresponding appearance of the created digital twin simulation equipment according to the appearance change of each part caused by the running condition of the equipment. The data analysis model may be calibrated and updated for the equipment operation conditions by incremental learning or continuous learning. Meanwhile, the physical model can be used for discovering new relevant physical knowledge data which has an influence on the running state of the equipment and updating the running environment of the equipment or a new fault source. The embodiment of the application at least uses the three models to create the digital twin simulation equipment model so as to realize high-assurance mapping of real world assets and simulation, and more accurately predicts the actual abnormal condition of the equipment through monitoring the abnormal condition of the digital twin simulation equipment model.

The application provides a prediction method of equipment faults, which comprises the steps of obtaining text feature vectors for carrying out problem description on abnormal conditions of the equipment, matching the text feature vectors from a preset question-answer feature vector database to determine whether the problem description is related to the equipment faults, carrying out feature matching on the text features in a fault knowledge question-answer library by using a retrieval enhancement mode if the problem description is determined to be related to the faults, obtaining a system fault prediction result and a system solution of the question-answer knowledge library, obtaining a networking fault prediction result and the networking solution by a networking query mode according to the text feature vectors, obtaining a map prediction fault result and a map solution by combining a constructed knowledge map for the equipment faults according to the text feature vectors, and obtaining a fault optimal prediction result and a corresponding optimal solution by utilizing a mixed expert model based on the system fault prediction result and the system solution, the networking fault prediction result and the networking solution and the map solution. The equipment fault prediction mode provided by the embodiment of the application can efficiently provide a relatively accurate fault prediction result for a user, and provides an optimal solution in a targeted manner, so that the experience of the user is obviously improved, and the operation and maintenance efficiency of the equipment is effectively improved.

Embodiment II,

Based on the method for predicting an equipment failure provided in the first embodiment of the present application, the embodiment of the present application further provides a device for predicting an equipment failure, as shown in fig. 2, fig. 2 is a schematic structural diagram of a device for predicting an equipment failure 20 provided in the first embodiment of the present application, where the device for predicting an equipment failure 20 includes:

an extraction module 201, configured to obtain a text feature vector for describing a problem of an abnormal condition of a device;

A matching module 202, configured to match the text feature vector from a preset question-answer feature vector database, so as to determine whether the problem description is a problem description related to a device fault;

the retrieval module 203 is configured to, if it is determined that the description of the problem is related to the fault, perform feature matching on the text feature in a fault knowledge question-answering library by using a retrieval enhancement manner, obtain a system fault prediction result and a system solution of the question-answering knowledge library;

And the generating module 204 is configured to obtain a fault optimal prediction result and a corresponding optimal solution based on the system fault prediction result and the system solution, the networking fault prediction result and the networking solution, and the map prediction fault result and the map solution by using a hybrid expert model.

Optionally, in an implementation manner of the embodiment of the present application, the retrieving module 203 is further configured to determine the fault knowledge question-and-answer library, where the fault knowledge question-and-answer library includes at least one of a local fault knowledge question-and-answer library and an external fault knowledge question-and-answer library, generate a query request according to the text feature by adopting an efficient retrieving algorithm to select a relevant document from the selected fault knowledge question-and-answer library, extract information from the relevant document, extract key information of the relevant document, where the key information includes one or more of entity information, entity relationship information and context information, fuse the key information by using a large language model to generate fault prediction information and predicted fault resolution information, and perform at least one of hallucination suppression, noise information processing, document screening and text input processing of the fault prediction information and the predicted fault resolution information to generate a system fault prediction result and a system solution of the question-and-answer knowledge library.

Optionally, in an implementation manner of the embodiment of the present application, the apparatus 20 further includes a construction module (not shown in the drawing), where the construction module is configured to determine, from open world data, equipment operation and maintenance information composed of a fault description, a solution and a fault cause of the equipment before the constructed knowledge graph for the equipment fault is combined according to the text feature vector to obtain the graph prediction fault result and the graph solution, convert the operation and maintenance information into structured semantic block data in a predefined format, perform incremental entity extraction and incremental relation extraction on the structured semantic block data to obtain unique entity information in each semantic block, determine relationship information between the entities, integrate the obtained unique entity information and the relationship information into a preset graph database, and combine the equipment related knowledge graph retrieved from the existing knowledge graph to generate a specific fault knowledge graph and a specific equipment knowledge graph, where the specific equipment knowledge has order data during operation of the specific equipment, aggregate node and edge of the generated specific fault graph, and cluster the specific equipment related knowledge graph is the specific knowledge graph, and the specific equipment related knowledge graph is integrated into the personalized knowledge graph.

Optionally, in an implementation manner of the embodiment of the present application, the retrieving module 203 is further configured to perform, according to the text feature vector, fault prediction on the device based on the personalized knowledge graph, using a dual-attention enhancement and graph neural network, so as to obtain the graph prediction fault result and the graph solution.

Optionally, in an implementation manner of the embodiment of the present application, the apparatus 20 further includes an updating module (not shown in the drawing), where the updating module is configured to periodically obtain a device operation knowledge base, extract key information in the device operation knowledge base by using a trained large language model, where the key information includes at least one of a device problem description, a fault reason description, and a fault solution, and update the personalized knowledge graph or the fault knowledge base according to the extracted key information.

Optionally, in an implementation manner of the embodiment of the present application, the apparatus 20 further includes a detection module (not shown in the drawing), where the detection module is configured to perform anomaly detection on the device according to the failure optimal prediction result after the failure optimal prediction result and the corresponding optimal solution are obtained, and perform early warning prompt on the device according to the result of the anomaly detection.

Optionally, in an implementation manner of the embodiment of the present application, the detection module specifically further uses:

creating a digital twin analog device for the device;

according to simulation operation data obtained by the simulation equipment through working state simulation, actual monitoring data obtained by monitoring the working state of the equipment, and statistical data obtained by statistics of the working state of the equipment of the same type, carrying out data fusion to obtain fusion data;

Constructing a health index and equipment residual life prediction neural network model aiming at the equipment according to the fusion data;

and performing performance evaluation on the digital twin simulation equipment by using a health index and equipment residual life prediction neural network model and combining the fault optimal prediction result so as to determine the abnormal detection result according to the performance evaluation result.

Optionally, in an implementation manner of the embodiment of the present application, the detection module specifically further collects multi-source data of the device;

Based on the multi-source data, a digital twin simulation device of the device is created using a preset geometric model, a data analysis model and a physical model.

The application provides a device for predicting equipment faults, which comprises a setting extraction module for obtaining text feature vectors for carrying out problem description on abnormal conditions of the equipment, a setting matching module for matching the text feature vectors from a preset question-answer feature vector database to determine whether the problem description is related to equipment faults, a setting retrieval module for carrying out feature matching on the text features in a fault knowledge question-answer library by using a retrieval enhancement mode if the problem description is determined to be related to faults, obtaining a system fault prediction result and a system solution of a question-answer knowledge library, obtaining a networking fault prediction result and a networking solution by a networking query mode according to the text feature vectors, obtaining the map prediction fault result and the map solution by combining a constructed knowledge map for the equipment faults according to the text feature vectors, and a setting generation module for obtaining a fault optimal prediction result and a corresponding optimal solution by utilizing a mixed expert model based on the system fault prediction result and the system solution, the networking fault prediction result and the networking solution and the map prediction fault result and the solution. The device for predicting the equipment faults can provide accurate fault prediction results for users with high efficiency, provides an optimal solution in a targeted manner, remarkably improves the experience of the users, effectively improves the operation and maintenance efficiency of the equipment, and is simple in structure and easy to realize.

Third embodiment,

The embodiment of the application also provides a storage medium, on which a computer program is stored, which when executed by a processor, implements the method for predicting any equipment failure according to the embodiment of the application.

Fourth embodiment,

An embodiment of the present application further provides an electronic device, as shown in fig. 3, and fig. 3 is a schematic structural diagram of an electronic device 30 according to an embodiment of the present application, where the electronic device 30 includes:

One or more processors 301, communication interfaces 302, memory 303, and communication buses 304, the processors 301, memory 303, and communication interfaces 302 completing communication among each other through the communication buses 304;

A memory 303 for storing one or more programs;

When the one or more programs are executed by the one or more processors 301, the one or more processors 301 implement any one of the device failure prediction methods described in the first embodiment of the present application.

Thus, the present application has been described with respect to particular embodiments of the present subject matter. In some cases, the actions recited in the claims can be performed in a different order and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may be advantageous.

In the 90 s of the 20 th century, improvements to one technology could clearly be distinguished as improvements in hardware (e.g., improvements to circuit structures such as diodes, transistors, switches, etc.) or software (improvements to the process flow). However, with the development of technology, many improvements of the current method flows can be regarded as direct improvements of hardware circuit structures. Designers almost always obtain corresponding hardware circuit structures by programming improved method flows into hardware circuits. Therefore, an improvement of a method flow cannot be said to be realized by a hardware entity module. For example, a programmable logic device (Programmable Logic Device, PLD) (e.g., field programmable gate array (Field Programmable GATE ARRAY, FPGA)) is an integrated circuit whose logic functions are determined by user programming of the device. A designer programs to "integrate" a digital system layer onto a PLD without having to ask the chip manufacturer to design and fabricate application specific integrated circuit chips. Moreover, nowadays, instead of manually manufacturing integrated circuit chips, such programming is mostly implemented with "logic compiler (logic compiler)" software, which is similar to the software compiler used in program development and writing, and the original code before being compiled is also written in a specific programming language, which is called hardware description language (Hardware Description Language, HDL), but HDL is not just one, but a plurality of kinds, such as ABEL(Advanced Boolean Expression Language)、AHDL(Altera Hardware Description Language)、Confluence、CUPL(Cornell University Programming Language)、HDCal、JHDL(Java Hardware Description Language)、Lava、Lola、MyHDL、PALASM、RHDL(Ruby Hardware Description Language), and VHDL (Very-High-SPEED INTEGRATED Circuit Hardware Description Language) and Verilog are currently most commonly used. It will also be apparent to those skilled in the art that a hardware circuit implementing the logic method flow can be readily obtained by merely slightly programming the method flow into an integrated circuit using several of the hardware description languages described above.

The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer readable medium storing computer readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, application SPECIFIC INTEGRATED Circuits (ASICs), programmable logic controllers, and embedded microcontrollers, examples of which include, but are not limited to, ARC 625D, atmel AT91SAM, microchip PIC18F26K20, and Silicone Labs C8051F320, and the memory controller may also be implemented as part of the control logic of the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller in a pure computer readable program code, it is well possible to implement the same functionality by logically programming the method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Such a controller may thus be regarded as a kind of hardware component, and means for performing various functions included therein may also be regarded as structures within the hardware component. Or even means for achieving the various functions may be regarded as either software modules implementing the methods or structures within hardware components.

The system layers, apparatuses, modules or units set forth in the above embodiments may be implemented in particular by a computer chip or entity, or by a product having a certain function. One typical implementation is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

For convenience of description, the above devices are described as being functionally divided into various units, respectively. Of course, the functions of each element may be implemented in the same piece or pieces of software and/or hardware when implementing the present application.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises an element.

It will be appreciated by those skilled in the art that embodiments of the application may be provided as a method, system layer, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular transactions or implement particular abstract data types. The application may also be practiced in distributed computing environments where transactions are performed by remote processing devices that are connected through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system layer embodiments, the description is relatively simple, as it is substantially similar to method embodiments, with reference to the section of the method embodiments being relevant.

The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and variations of the present application will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the application are to be included in the scope of the claims of the present application.

Claims (10)

1. A method for predicting equipment failure, comprising:

acquiring a text feature vector for describing the problem of the abnormal condition of the equipment;

Matching the text feature vector from a preset question-answer feature vector database to determine whether the problem description is a problem description related to equipment failure;

If the text feature is determined to be the problem description related to the fault, performing feature matching on the text feature in a fault knowledge question-answering library by using a retrieval enhancement mode to obtain a system fault prediction result and a system solution of the question-answering knowledge library;

And obtaining a fault optimal prediction result and a corresponding optimal solution based on the system fault prediction result and the system solution, the networking fault prediction result and the networking solution, and the map prediction fault result and the map solution by using a mixed expert model.

2. The method for predicting equipment failure according to claim 1, wherein the feature matching is performed on the text features in a failure knowledge question-and-answer library by using a search enhancement method, so as to obtain a system failure prediction result and a system solution of the question-and-answer knowledge library, and the method comprises the following steps:

Determining the fault knowledge question-answering library, wherein the fault knowledge question-answering library at least comprises one of a local fault knowledge question-answering library or an external fault knowledge question-answering library;

generating a query request according to the text characteristics by adopting a high-efficiency search algorithm to select related documents from the selected fault knowledge question-answering library;

extracting information from the related document, and extracting key information of the related document, wherein the key information comprises one or more of entity information, entity relation information and context information;

Fusing the key information by using a large language model to generate fault prediction information and prediction fault resolution information;

And performing at least one text input process of illusion suppression, noise information processing, document screening and error correction on the fault prediction information and the prediction fault solution information to generate a system fault prediction result and a system solution of a question-answer knowledge base [ HZTC-HXJKJD241005] [ HS2411223CCN ].

3. The method for predicting an equipment failure according to claim 1, wherein before the knowledge graph for the equipment failure constructed according to the text feature vector is combined, the method further comprises:

determining equipment operation and maintenance information consisting of fault description, solution and fault cause of the equipment from open world data;

converting the operation and maintenance information into structural semantic block data in a predefined format;

Performing incremental entity extraction and incremental relation extraction on the structured semantic block data to obtain unique entity information in each semantic block, and determining relation information among the entities;

Integrating the acquired unique entity information and the acquired relationship information into a preset graphic database, and generating a specific fault knowledge graph and a specific equipment knowledge graph by combining equipment related knowledge graph subgraphs retrieved from the existing knowledge graph, wherein the specific equipment knowledge graph has sequence data during the operation of the specific equipment;

Performing node and edge clustering on the generated specific fault knowledge graph, and creating a fault aggregation knowledge graph;

And combining the related specific fault pattern and time information about sequential data in the specific knowledge pattern, generating a personalized knowledge pattern of the equipment, and determining the personalized knowledge pattern as the knowledge pattern of the equipment fault.

4. The method for predicting an equipment failure according to claim 3, wherein the obtaining the graph prediction failure result and the graph solution according to the text feature vector and the constructed knowledge graph for the equipment failure includes:

And according to the text feature vector, performing fault prediction on the equipment based on the personalized knowledge graph by adopting a dual-attention enhancement and graph neural network so as to acquire a graph prediction fault result and a graph solution.

5. The method for predicting equipment failure of claim 1, further comprising periodically acquiring a new equipment operation knowledge base;

[ HZTC-HXJKJD241005] [ HS2411223CCN ] uses the trained large language model to extract key information in the equipment operation knowledge base, wherein the key information at least comprises one of equipment problem description, fault reason description and fault solution.

And according to the extracted key information, updating data of a knowledge graph or a fault knowledge base of the equipment fault.

6. The method for predicting an equipment failure according to claim 1, wherein after the obtaining of the failure optimal prediction result and the corresponding optimal solution, the method further comprises:

Performing abnormality detection on the equipment according to the optimal fault prediction result;

And carrying out early warning prompt on the equipment according to the abnormal detection result.

7. The method for predicting equipment failure according to claim 6, wherein the performing anomaly detection on the equipment according to the failure optimal prediction result comprises:

creating a digital twin analog device for the device;

according to simulation operation data obtained by the simulation equipment through working state simulation, actual monitoring data obtained by monitoring the working state of the equipment, and statistical data obtained by statistics of the working state of the equipment of the same type, carrying out data fusion to obtain fusion data;

Constructing a health index and equipment residual life prediction neural network model aiming at the equipment according to the fusion data;

and performing performance evaluation on the digital twin simulation equipment by using a health index and equipment residual life prediction neural network model and combining the fault optimal prediction result so as to determine the abnormal detection result according to the performance evaluation result.

8. The method of claim 7, wherein creating a digital twin simulation device for the device at a server comprises:

Collecting multi-source data of the device;

Based on the multi-source data, a digital twin simulation device of the device is created using a preset geometric model, a data analysis model and a physical model.

9. A device for predicting equipment failure, comprising:

The [ HZTC-HXJKJD241005] [ HS2411223CCN ] extraction module is used for acquiring text feature vectors for describing the problem of the abnormal condition of the equipment;

The matching module is used for matching the text feature vector from a preset question-answer feature vector database so as to determine whether the problem description is related to equipment faults;

The system comprises a search module, a graph prediction result and a graph solution, wherein the search module is used for carrying out feature matching on the text features in a fault knowledge question-answer library by using a search enhancement mode if the text features are determined to be related to the faults;

And the generation module is used for obtaining a fault optimal prediction result and a corresponding optimal solution based on the system fault prediction result and the system solution, the networking fault prediction result and the networking solution and the map prediction fault result and the map solution by utilizing a mixed expert model.

10. A computer storage medium having stored thereon computer executable instructions which when executed perform the method of predicting a device failure according to any one of claims 1-8.