Method, processor and readable storage medium for engineering equipment fault diagnosis

Abstract

The application relates to the technical field of artificial intelligence, and discloses a method for engineering equipment fault diagnosis, a processor and a readable storage medium. The method comprises the following steps: performing entity extraction on unstructured diagnosis knowledge data based on an entity recognition model, obtaining entities contained in the diagnosis knowledge data, matching the entities with known entities corresponding to the structured diagnosis knowledge data, obtaining entity triples, and importing the entity triples into a preset graph database to generate a fault knowledge graph; obtaining an input query request, determining a query fault category corresponding to the query request based on a text classification model, and determining a query result according to the query fault category and the fault knowledge graph. The fault knowledge graph is automatically constructed through the entity recognition model, which greatly reduces the workload of manually constructing a knowledge base. Moreover, the query fault category is accurately identified based on the text classification model, the query result is quickly determined according to the fault knowledge graph, and the query efficiency of fault diagnosis knowledge is greatly improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to a method, processor, and readable storage medium for fault diagnosis of engineering equipment. Background Technology

[0002] With the continuous development of science and technology, the number of brands in the construction machinery industry is increasing, the types of equipment are becoming more diverse, and the types of industry knowledge are also becoming more varied. Among these, knowledge and experience related to the diagnosis of construction machinery faults have been continuously accumulated. Reasonably and effectively utilizing this experience to resolve mechanical faults is beneficial for quickly restoring machinery operation and improving operational efficiency. Currently, some methods involve establishing an expert diagnostic system to query relevant knowledge. Specifically, this involves storing manually compiled fault cases in a relational database to build a knowledge base. Then, a user-facing query page is built, and template or text classification retrieval algorithms are used to categorize user-input questions. The categorized results are matched with the content stored in the constructed knowledge base. After matching, a fault handling method is returned to the user. While this method can automate the querying of construction machinery faults, the knowledge base needs to be manually built, resulting in high labor costs. Furthermore, this method is prone to providing limited and inaccurate query results, failing to provide users with comprehensive solutions. Summary of the Invention

[0003] To address the aforementioned shortcomings in the prior art, the purpose of this invention is to provide a method, processor, and readable storage medium for fault diagnosis of engineering equipment.

[0004] To achieve the above objectives, a first aspect of the present invention provides a method for fault diagnosis of engineering equipment, comprising:

[0005] Entity extraction is performed on unstructured diagnostic knowledge data based on an entity recognition model to obtain the entities contained in the diagnostic knowledge data;

[0006] The entities are matched with the known entities corresponding to the structured diagnostic knowledge data to obtain entity triples;

[0007] Import entity triples into a pre-defined graph database to generate a fault knowledge graph;

[0008] Get the input query request;

[0009] Determine the query failure category corresponding to the query request based on a text classification model;

[0010] The query results are determined based on the fault category and the fault knowledge graph.

[0011] In this embodiment of the invention, determining the query result based on the query fault category and fault knowledge graph includes:

[0012] The first query result is determined based on the query fault category and fault knowledge graph;

[0013] Determine the abnormal operating condition thresholds corresponding to the fault category entities in the fault knowledge graph.

[0014] Obtain operating condition data of preset associated devices;

[0015] Determine the target abnormal operating condition threshold that matches the operating condition data;

[0016] The fault category entity corresponding to the target abnormal operating condition threshold is taken as the target fault category entity;

[0017] The second query result is determined based on the target fault category entity.

[0018] In this embodiment of the invention, determining the query result based on the query fault category and fault knowledge graph further includes:

[0019] If no target abnormal operating condition threshold matching the operating condition data exists, retrieve historical query requests;

[0020] If a query request matches a historical query request, the first historical query result corresponding to the historical query request is determined.

[0021] Use the first historical query result as the second query result.

[0022] In this embodiment of the invention, determining the query result based on the query fault category and fault knowledge graph further includes:

[0023] If the query request does not match the historical query requests, identify the target historical query request that appears most frequently among all historical query requests.

[0024] Determine the second historical query result corresponding to the target historical query request;

[0025] Use the second historical query result as the second query result.

[0026] In this embodiment of the invention, entities are matched with known entities corresponding to structured diagnostic knowledge data to obtain entity triples, including:

[0027] Identify the known entities corresponding to the structured diagnostic knowledge data;

[0028] Combine the entity with known entities to obtain the target entity;

[0029] Determine the entity relationships between the target entities;

[0030] Determine entity triples based on the target entity and entity relationships.

[0031] In this embodiment of the invention, determining the query result based on the query fault category and fault knowledge graph includes:

[0032] Identify the first fault type entity in the target entity corresponding to the query fault category in the fault knowledge graph;

[0033] Identify the second fault type entity associated with the first fault type entity;

[0034] The query results are determined based on the first fault type entity and the second fault type entity.

[0035] In this embodiment of the invention, determining the query result based on the first fault type entity and the second fault type entity includes:

[0036] Determine the first relationship weight between the first fault solution entity and the first fault type entity that match the first fault type entity; and

[0037] Determine the second relationship weight between the second fault solution entity that matches the second fault type entity and the second fault type;

[0038] The graph link analysis algorithm uses the first relation weight and the second relation weight to determine the display order of the first fault solution and the second fault solution;

[0039] The first fault solution entity and the first component entity associated with the first fault solution entity, the second fault solution entity and the second component entity associated with the second fault solution entity are arranged in the order of display to generate query results.

[0040] In this embodiment of the invention, it further includes:

[0041] Obtain the input entity label training data and type label training data;

[0042] Determine the entity recognition model and the text classification model to be trained;

[0043] Input the entity labeling training data into the entity recognition model to be trained to generate the entity recognition model;

[0044] Type-labeled training data is input into the text classification model to be trained to generate the text classification model.

[0045] A second aspect of the present invention provides a processor configured to perform the steps of the method for diagnosing faults in engineering equipment as described above.

[0046] A third aspect of the present invention provides a readable storage medium storing instructions that, when executed by a processor, cause the processor to perform the method for diagnosing faults in engineering equipment as described above.

[0047] The above technical solution extracts entities from unstructured diagnostic knowledge data using an entity recognition model to obtain the entities contained within the diagnostic knowledge data. These entities are then matched with known entities in the structured diagnostic knowledge data to obtain entity triples. These entity triples are then imported into a pre-defined graph database to generate a fault knowledge graph. The system also acquires input query requests and determines the corresponding fault category based on a text classification model. The query result is then determined based on the fault category and the fault knowledge graph. The automated construction of the fault knowledge graph using the entity recognition model significantly reduces the workload of manually constructing the knowledge base. Simultaneously, the accurate identification of the query fault category using the text classification model allows for rapid determination of query results based on the fault knowledge graph, greatly improving the efficiency of fault diagnosis knowledge retrieval.

[0048] Other features and advantages of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0049] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the following detailed description to explain the invention, but do not constitute a limitation thereof. In the drawings:

[0050] Figure 1 This is a flowchart illustrating a method for diagnosing faults in engineering equipment according to an embodiment of the present invention.

[0051] Figure 2 This is a schematic diagram of an application process according to an embodiment of the present invention. Detailed Implementation

[0052] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0053] Figure 1 This is a schematic flowchart of a method for fault diagnosis of engineering equipment according to an embodiment of the present invention. Figure 1 As shown, in this embodiment of the invention, a method for fault diagnosis of engineering equipment is provided. Taking the application of this method to a processor as an example, the method may include the following steps:

[0054] Step S100: Extract entities from unstructured diagnostic knowledge data based on the entity recognition model to obtain the entities contained in the diagnostic knowledge data;

[0055] In this embodiment, it should be noted that unstructured data includes fully unstructured data and semi-structured data. Unstructured diagnostic knowledge data includes unorganized text, videos, images, web pages, and other data uploaded by engineers. Entity extraction involves extracting atomic information elements from the data, finding named entities, and classifying them. The entity recognition model refers to a pre-trained entity extraction model capable of extracting entities from unstructured diagnostic knowledge data.

[0056] Specifically, the processor extracts entities from unstructured diagnostic knowledge data using an entity recognition model, thereby obtaining the entities contained in the diagnostic knowledge data.

[0057] Step S200: Match the entities with the known entities corresponding to the structured diagnostic knowledge data to obtain entity triples;

[0058] In this embodiment, it should be noted that the structured diagnostic knowledge data includes relevant knowledge data already organized in the database, such as organized expert knowledge on engineering machinery faults. Known entities refer to the entities contained in the structured diagnostic knowledge data; entity triples include an entity-relation-entity triple framework. When unstructured data is obtained, it needs to be associated and combined with structured data to improve the knowledge database. In this embodiment, entity triples are obtained by matching the entities of the unstructured diagnostic knowledge data with the known entities of the structured diagnostic knowledge data. This allows the organization of unstructured diagnostic knowledge data to be associated with structured data, thereby improving the comprehensiveness of subsequent responses to input query requests.

[0059] Specifically, entities are matched with known entities corresponding to structured diagnostic knowledge data to obtain entity triples, including:

[0060] Step a: Identify the known entities corresponding to the structured diagnostic knowledge data;

[0061] Step b: Combine the entity with the known entities to obtain the target entity;

[0062] Step c: Determine the entity relationships between the target entities;

[0063] Step d: Determine the entity triples based on the target entity and entity relationships.

[0064] In this embodiment, it should be noted that the target entity refers to the entity obtained by combining the known entities corresponding to the structured diagnostic knowledge data with the entities contained in the unstructured diagnostic knowledge data. Combination includes pairing and matching, and the combination method is to combine them according to a logical order. Entity relationships include corresponding relationships such as sequence relationships, association relationships, and hierarchical relationships between entities.

[0065] In one embodiment, diagnostic knowledge data can extract entities including construction machinery component entities, fault category entities, and fault solution entities. An entity recognition model is used to extract entities from the semi-structured diagnostic knowledge data. The extracted results are construction machinery component entities and fault category entities arranged in the text order. The construction machinery component entities and fault category entities included in the structured diagnostic knowledge data are combined with the construction machinery component entities and fault category entities extracted from the semi-structured diagnostic knowledge data in a logical order. Simultaneously, the fault category entities in the combined target entities are matched with the fault solution entities included in the structured diagnostic knowledge data to determine the entity relationships between the target entities, resulting in the following four types of entity triples: Construction machinery component entity - superior / subordinate (including weight) - construction machinery component entity; Construction machinery component entity - related operating condition type (including weight) - fault category entity; Fault category entity - operating condition threshold (including weight) - fault solution entity; Fault solution entity - related components - construction machinery component entity. In practical applications, taking a concrete pump truck as an example, the above four three-element groups can be: boom - belongs to (weight 0.1) - boom pump; boom pump - boom current (weight 0.3) - boom no movement; boom no movement - boom current (greater than 0.3, less than 0.7, weight 0.9) - boom no movement handling solution; boom no movement handling solution - related components - boom pump.

[0066] Specifically, after the processor determines the known entities corresponding to the structured diagnostic knowledge data, it combines the entities with the known entities to obtain the target entities, and determines the entity relationships between the target entities, so as to determine the entity triples based on the target entities and entity relationships.

[0067] Step S300: Import the entity triples into the preset graph database to generate a fault knowledge graph;

[0068] In this embodiment, it should be noted that the graph database is a data management system designed for efficient storage and retrieval of graph data, using nodes and edges as basic storage units. "Nodes" represent entities, and "edges" represent relationships between entities. Importing determined entity triples into a preset graph database can generate a corresponding fault knowledge graph. After importing the entity triples into the preset graph database, all related entities need to be linked to achieve combinations between all entity triples. The preset graph database may include Neo4j.

[0069] Specifically, the processor imports entity triples into a preset graph database and checks whether all entities in the preset graph database have been linked. Once it is confirmed that all entities have been linked, a fault knowledge graph is generated.

[0070] Step S400: Obtain the input query request;

[0071] In this embodiment, it should be noted that the query request is a request input by the user. The query request may include descriptive information such as the description of the fault, the description of the mechanical parts, the description of the operating condition values, and the type of operating condition, so as to locate the corresponding diagnostic knowledge in the preset graph database.

[0072] Step S500: Determine the query fault category corresponding to the query request based on the text classification model;

[0073] In this embodiment, it should be noted that the text classification model refers to a pre-trained language representation model that can identify and classify the acquired text. The text classification model can identify the query request, obtain the fault category corresponding to the fault description in the query request, and use the fault category corresponding to the query request determined by the text classification model as the query fault category.

[0074] Step S600: Determine the query results based on the query fault category and fault knowledge graph.

[0075] In this embodiment, it should be noted that the fault knowledge graph includes target entities and the relationships between target entities. The target entities are entities extracted from diagnostic knowledge data, including fault category entities involved in the diagnostic knowledge data. After determining the query fault category, the processor can perform a query in the fault knowledge graph based on the query fault category to determine the fault category entity that matches the query fault category, and then determine the final query result based on the fault category entity.

[0076] The aforementioned method for fault diagnosis of engineering equipment extracts entities from unstructured diagnostic knowledge data using an entity recognition model. This extracts the entities contained within the diagnostic knowledge data, matches them with known entities in the structured diagnostic knowledge data to obtain entity triples, and imports these triples into a pre-defined graph database to generate a fault knowledge graph. The method also acquires input query requests and determines the corresponding fault category based on a text classification model. Finally, it determines the query result based on the fault category and the fault knowledge graph. The automated construction of the fault knowledge graph based on the entity recognition model significantly reduces the workload of manually constructing the knowledge base. Simultaneously, the accurate identification of the query fault category using the text classification model allows for rapid determination of query results based on the fault knowledge graph, greatly improving the efficiency of fault diagnosis knowledge retrieval.

[0077] In one embodiment, determining the query results based on the query fault category and fault knowledge graph includes:

[0078] Step e: Determine the first query result based on the query fault category and fault knowledge graph;

[0079] Step f: Determine the abnormal operating condition threshold corresponding to the fault category entity in the fault knowledge graph;

[0080] Step g: Obtain the operating condition data of the preset associated equipment;

[0081] Step h: Determine the target abnormal operating condition threshold that matches the operating condition data;

[0082] Step i: Take the fault category entity corresponding to the target abnormal operating condition threshold as the target fault category entity;

[0083] Step j: Determine the second query result based on the target fault category entity.

[0084] In this embodiment, it should be noted that the query results include diagnostic knowledge data such as fault solutions and associated components corresponding to the fault category of the query request in the graph database, and the query results include a first query result and a second query result.

[0085] The first query result is the response to the query request. Based on the queried fault category, a query is performed in a preset graph database to determine the fault category entity that matches the query category. Then, the fault solution entity associated with the fault category entity is obtained, and the engineering machinery component entity associated with the fault solution entity is determined. The fault solution entity and the associated engineering machinery component entity are the precise matching results corresponding to the input query request. In one embodiment, a fuzzy matching result corresponding to the query request can also be determined to improve the comprehensiveness of the query results. Using fault category entities in the fault knowledge graph similar to the queried fault category corresponding to the query request, the corresponding fault solution entity and the engineering machinery component entity associated with the fault solution entity are determined. The fault solution entity and the engineering machinery component entity associated with the fault category entity similar to the queried fault category are used as the fuzzy matching result. The precise matching result and the fuzzy matching result are output as the first query result.

[0086] The second query result is determined based on relevant information such as equipment operating conditions and historical query records. The entities stored in the fault knowledge graph include entities related to diagnostic knowledge data such as engineering machinery parts entities, fault category entities, and fault solution entities. Abnormal operating condition thresholds are operating condition thresholds associated with fault category entities in the fault knowledge graph, including the possible operating condition values ​​that may occur when the fault corresponding to that fault category entity occurs. Preset associated equipment includes engineering equipment bound by the user before inputting the query request. After receiving the user's query request, the operating condition data of the preset associated equipment is judged, determining the recent operating condition data of the preset associated equipment. This operating condition data is matched with all existing abnormal operating condition thresholds to determine if there is an abnormal operating condition threshold that matches the operating condition data. The abnormal operating condition threshold that matches the operating condition data is taken as the target abnormal operating condition threshold. At this point, it is determined that the preset associated equipment has a fault corresponding to a fault category entity associated with the target abnormal operating condition threshold. The fault category entity associated with the target abnormal operating condition threshold is taken as the target fault category entity, and the solution entity associated with the target fault category entity and the engineering machinery parts entity associated with the solution entity are output as the second query result. In this embodiment, when outputting the query results, the second query result is output after the first query result is output. Both the first and second query results can include multiple fault solution entities and component entities associated with the fault solution entities. Before the query results are output, all fault solution entities and their associated component entities in the first and second query results can be sorted respectively, so as to determine the display order of fault solution entities and their associated component entities in the first and second query results according to the sorting results.

[0087] Specifically, the processor determines the first query result based on the query fault category and fault knowledge graph, and determines the abnormal operating condition threshold corresponding to the fault category entity in the fault knowledge graph. After obtaining the operating condition data of the preset associated equipment, it determines the target abnormal operating condition threshold that matches the operating condition data, so as to take the fault category entity corresponding to the target abnormal operating condition threshold as the target fault category entity, and then determines the second query result based on the target fault category entity.

[0088] In one embodiment, determining the query results based on the query fault category and the fault knowledge graph further includes:

[0089] Step k: If there is no target abnormal operating condition threshold that matches the operating condition data, then retrieve the historical query request;

[0090] Step 1: If the query request matches the historical query request, determine the first historical query result corresponding to the historical query request;

[0091] Step m: Use the first historical query result as the second query result.

[0092] In this embodiment, it should be noted that historical query requests include query requests entered by the user before the current query request, referring to the historical record of fault queries entered by the user. When a user-inputted query request is obtained, and the operating condition data of the preset associated devices is judged, if it is determined that there is no abnormal operating condition threshold matching the operating condition data in the fault knowledge graph, then the user's historical query requests are obtained. The first historical query result is the query result output when the historical query request is entered. The historical query request matching the user's current query request refers to the historical query request whose description is similar to the current query request among all historical query requests. If there is no abnormal operating condition threshold matching the operating condition data in the fault knowledge graph, then the first historical query result is output as the second query result.

[0093] Specifically, after determining that there is no target abnormal operating condition threshold that matches the operating condition data, the processor obtains historical query requests. If the query request matches the historical query request, the processor determines the first historical query result corresponding to the historical query request and uses the first historical query result as the second query result.

[0094] In one embodiment, determining the query results based on the query fault category and the fault knowledge graph further includes:

[0095] Step n: If the query request does not match the historical query requests, identify the target historical query request that appears most frequently among all historical query requests.

[0096] Step o: Determine the second historical query result corresponding to the target historical query request;

[0097] Step p: Use the second historical query result as the second query result.

[0098] In this embodiment, it should be noted that when matching the user-input query request with historical query requests, if it is determined that there is no historical query request matching the input query request among all historical query requests, the historical query request with the highest frequency will be determined as the target historical query request, and the second historical query result will be the query result output when the target historical query request is input. If multiple historical query requests with the same highest frequency appear, the historical query request most recent to the current query request time among these multiple historical query requests with the same highest frequency will be determined as the target historical query request. The second historical query result corresponding to the target historical query request will be the second query result output.

[0099] Specifically, when a query request does not match a historical query request, the processor determines the target historical query request that appears most frequently among all historical query requests, and determines the second historical query result corresponding to the target historical query request, so as to use the second historical query result as the second query result.

[0100] In one embodiment, determining the query results based on the query fault category and fault knowledge graph includes:

[0101] Step q: Determine the first fault type entity in the target entity corresponding to the fault knowledge graph that matches the queried fault category;

[0102] Step r: Determine the second fault type entity associated with the first fault type entity;

[0103] Step s: Determine the query result based on the first fault type entity and the second fault type entity.

[0104] In this embodiment, it should be noted that the target entity refers to the entity formed by combining the known entities corresponding to the structured diagnostic knowledge data and the entities contained in the unstructured diagnostic knowledge data during the construction of the fault knowledge graph, including the combined fault type entity. The first fault type entity includes fault type entities in the fault knowledge graph that match the queried fault category, wherein the fault type entities matching the queried fault category include entities describing fault types that are the same as or similar to the queried category. The second fault type entity associated with the first fault type entity includes fault type entities whose fault types are similar to the first fault type. When determining the query result, the query result is determined based on the first fault type entity that matches the queried fault type in the query request and the second fault type entity that is similar to the first fault type entity. This allows for precise matching of fault solutions through the first fault type entity and fuzzy matching of fault solutions through the second fault type entity, enabling users to obtain more accurate and broader fault solutions.

[0105] Specifically, the query results are determined based on the first fault type entity and the second fault type entity, including:

[0106] Step s1, determine the first relationship weight between the first fault solution entity and the first fault type entity that match the first fault type entity; and

[0107] Step s2: Determine the second relationship weight between the second fault solution entity that matches the second fault type entity and the second fault type;

[0108] Step s3: Based on the graph link analysis algorithm, the display order of the first fault solution and the second fault solution is determined using the first relation weight and the second relation weight;

[0109] Step s4: Arrange the first fault solution entity and the first component entity associated with the first fault solution entity, the second fault solution entity and the second component entity associated with the second fault solution entity in the order of display to generate query results.

[0110] In this embodiment, it should be noted that the target entities in the fault knowledge graph stored in the preset graph database are associated through entity relationships. Each entity relationship has a weight, representing the tightness of the relationship between the two connected entities; that is, the higher the weight of the entity relationship, the tighter the connection between the connected entities. The output order of the query results is determined based on the weights corresponding to the entity relationships. This order highlights the relevance to the query request, allowing users to focus on the fault solutions ranked higher, thus quickly identifying solutions for the user.

[0111] When determining the query results, the query results are determined based on a first fault type entity that matches the query fault type in the query request and a second fault type entity that is similar to the first fault type entity. The first fault solution entity is a fault solution entity associated with the first fault type entity; the second fault solution entity is a fault solution entity associated with the second fault type entity; the first relationship weight includes the weight attached to the entity relationship between the first fault solution entity and the first fault type entity; the second relationship weight includes the weight attached to the entity relationship between the second fault solution entity and the second fault type entity; the graph link analysis algorithm can determine the degree of entity association between all fault solution entities and fault type entities through the weight of entity relationships. In this embodiment, the graph link analysis algorithm may include the PageRank algorithm. After determining the first and second relationship weights, a graph link analysis algorithm is used to determine the entity association degree. The display order is then determined based on the descending order of entity association degree. Specifically, the first relationship weight determines the display order of all first fault solution entities matching the first fault type entity, and the second relationship weight determines the display order of all second fault solution entities matching the second fault type entity. In one embodiment, the display order between the first and second fault solution entities can be either prioritizing the first fault solution over the second, or there can be no priority between the first and second fault solutions, with all fault solutions displayed in order of entity association degree. The first component entity is the engineering machinery component entity associated with the first fault solution entity, which is output together with the first fault solution entity; the second component entity associated with the second fault solution entity is output together with the second fault solution entity. After determining the display order, the first fault solution entity and the first component entity, as well as the second fault solution entity and the second component entity, are arranged and output according to the display order.

[0112] Specifically, the processor determines the first relationship weight between the first fault solution entity matching the first fault type entity and the first fault type entity, and determines the second relationship weight between the second fault solution entity matching the second fault type entity and the second fault type. Based on the graph link analysis algorithm, the processor uses the first relationship weight and the second relationship weight to determine the display order of the first fault solution and the second fault solution. This allows the processor to arrange the first fault solution entity and the first component entity associated with the first fault solution entity, and the second fault solution entity and the second component entity associated with the second fault solution entity, according to the display order, in order to generate query results.

[0113] In one embodiment, it also includes:

[0114] Step t: Obtain the input entity label training data and type label training data;

[0115] Step u: Determine the entity recognition model to be trained and the text classification model to be trained;

[0116] Step v: Input the entity labeling training data into the entity recognition model to be trained to generate the entity recognition model;

[0117] Step w involves inputting the type-labeled training data into the text classification model to be trained, in order to generate the text classification model.

[0118] In this embodiment, it should be noted that the entity labeling training data and type labeling training data are pre-labeled training data, including labeled entities determined based on experience and logic. For example, a boom is labeled as an engineering machinery component entity, and a boom without movement is labeled as a fault type entity. The entity recognition model and the text classification model to be trained are initial models whose training parameters have not yet been determined. After training the initial model with training data, the corresponding training parameters are determined to generate the entity recognition model and the text classification model. The entity recognition model can extract entities from diagnostic knowledge data, for example, BERT (Bidirectional Encoder Representation from Transformers, pre-trained language representation model) + BiLSTM (Bi-directional Long Short-Term Memory) + CRF (Conditional Random Field) model; the text classification model can recognize and classify text content, for example, the BERT model.

[0119] refer to Figure 2In one application scenario, entity extraction is performed on unstructured data based on an entity recognition model. The extracted entities are then combined with known entities in structured data to construct entity triples, which determine the fault knowledge graph to be stored in a preset graph database. A text classification model is used to identify the user's input query request, determining the corresponding query fault category. This query fault category is used as the first input and fed into the graph database to traverse the organized structured data. A graph link analysis algorithm is used to determine the first query result. Upon receiving the user's query request, the system triggers a judgment on the operating condition data of preset associated equipment. If it is determined that there is no abnormal operating condition threshold matching the operating condition data in the fault knowledge graph, the user's historical query requests are obtained and used as the second input and fed into the graph database to determine the second query result corresponding to the historical query request. If it is determined that there is a target abnormal operating condition threshold matching the operating condition data in the fault knowledge graph, this target abnormal operating condition threshold is used as the second input and fed into the graph database to determine the matching target fault category entity, thus determining the output second query result. The first and second query results are then output to the user.

[0120] In existing technologies, fault knowledge cannot be correlated with other knowledge, resulting in limited and inaccurate query results. This fails to provide users with comprehensive solutions, and the returned content cannot be effectively ranked based on knowledge relevance or recommended based on the characteristics of the user's current device. This embodiment utilizes a deep learning algorithm model and entity recognition model to automatically construct a fault knowledge graph from structured, semi-structured, and unstructured data, achieving effective correlation between fault diagnosis knowledge. When determining the user's query results, a graph link analysis algorithm is used to rank entity nodes, returning corresponding fault solutions based on their relevance to the input fault phenomenon. This allows users to focus on the top-ranked solutions, quickly identifying the best solution. Furthermore, abnormal operating condition data of the user's pre-associated devices are matched with corresponding and related entity nodes within the fault knowledge graph. Simultaneously, personalized knowledge recommendations are provided based on the user's historical search requests, further expanding fault handling options and enabling users to obtain broader and more accurate relevant solutions, improving the effectiveness and comprehensiveness of fault queries.

[0121] This invention provides a processor configured to execute the steps of the method for diagnosing faults in engineering equipment as described above.

[0122] This invention provides a readable storage medium storing instructions that, when executed by a processor, cause the processor to perform the method for diagnosing faults in engineering equipment as described above.

[0123] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can 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, etc.) containing computer-usable program code.

[0124] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0125] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0126] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0127] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0128] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0129] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0130] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0131] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for fault diagnosis of engineering equipment, characterized in that, include: Entity extraction is performed on unstructured diagnostic knowledge data based on an entity recognition model to obtain the entities contained in the diagnostic knowledge data; The entities are matched with the known entities corresponding to the structured diagnostic knowledge data to obtain entity triples; The entity triples are imported into a preset graph database to generate a fault knowledge graph; Get the input query request; The query failure category corresponding to the query request is determined based on a text classification model. The query results are determined based on the query fault category and the fault knowledge graph. The step of determining the query result based on the query fault category and the fault knowledge graph includes: The first query result is determined based on the query fault category and the fault knowledge graph; Determine the abnormal operating condition threshold corresponding to the fault category entity of the fault knowledge graph; Obtain operating condition data of preset associated devices; Determine the target abnormal operating condition threshold that matches the operating condition data; The fault category entity corresponding to the target abnormal operating condition threshold is taken as the target fault category entity; The second query result is determined based on the target fault category entity.

2. The method according to claim 1, characterized in that, The step of determining the query result based on the query fault category and the fault knowledge graph also includes: If no target abnormal operating condition threshold matching the operating condition data exists, retrieve historical query requests; If the query request matches the historical query request, determine the first historical query result corresponding to the historical query request; Use the first historical query result as the second query result.

3. The method according to claim 2, characterized in that, The step of determining the query result based on the query fault category and the fault knowledge graph also includes: If the query request does not match the historical query requests, determine the target historical query request that appears most frequently among all historical query requests; Determine the second historical query result corresponding to the target historical query request; Use the second historical query result as the second query result.

4. The method according to claim 1, characterized in that, The step of matching the entity with known entities corresponding to structured diagnostic knowledge data to obtain entity triples includes: Identify the known entities corresponding to the structured diagnostic knowledge data; The entity is combined with the known entity to obtain the target entity; Determine the entity relationships between the target entities; Entity triples are determined based on the target entity and the entity relationship.

5. The method according to claim 1, characterized in that, The step of determining the query results based on the query fault category and the fault knowledge graph includes: Identify the first fault type entity in the target entity corresponding to the fault knowledge graph that matches the queried fault category; Determine the second fault type entity associated with the first fault type entity; The query results are determined based on the first fault type entity and the second fault type entity.

6. The method according to claim 5, characterized in that, The step of determining the query result based on the first fault type entity and the second fault type entity includes: Determine the first relationship weight between the first fault solution entity that matches the first fault type entity and the first fault type entity; and Determine the second relationship weight between the second fault solution entity that matches the second fault type entity and the second fault type; The graph link analysis algorithm uses the first relation weight and the second relation weight to determine the display order of the first fault solution and the second fault solution; The first fault solution entity and the first component entity associated with the first fault solution entity, the second fault solution entity and the second component entity associated with the second fault solution entity are arranged in the display order to generate query results.

7. The method according to claim 1, characterized in that, Also includes: Obtain the input entity label training data and type label training data; Determine the entity recognition model and the text classification model to be trained; The entity labeling training data is input into the entity recognition model to be trained to generate the entity recognition model; The type-labeled training data is input into the text classification model to be trained to generate a text classification model.

8. A processor, characterized in that, It is configured to perform the method for fault diagnosis of engineering equipment as described in any one of claims 1 to 7.

9. A readable storage medium storing instructions, characterized in that, When the instruction is executed by the processor, it causes the processor to perform the method for fault diagnosis of engineering equipment according to any one of claims 1 to 7.

Images