Construction method, device and equipment of aero-engine fault data extraction model

Abstract

The invention relates to a construction method, device and equipment of an aero-engine fault data extraction model. According to the method, fault data generated by the aero-engine in a fault state can be obtained, meanwhile, a fault data ontology concept corresponding to the aero-engine can be constructed, a labeling rule is formulated according to the fault data ontology concept to label the fault data, and then a preset data extraction model is trained by utilizing the labeled fault data; according to the method, the target data extraction model is obtained, finally, accurate extraction of the aero-engine fault data can be realized by using the target data extraction model, and fault information corresponding to different fault data is obtained, so that construction of a fault knowledge base is facilitated. Meanwhile, the technical development in the aero-engine troubleshooting task can be promoted, the intelligent level is improved, more reliable technical support is provided for safety management and maintenance of the aero-engine, and reference is provided for information extraction tasks with small sample data and high specialty.

Description

Technical Field

[0001] The present disclosure relates to the technical field of information processing technology, and in particular to a method, device and equipment for constructing an aircraft engine fault data extraction model. Background Art

[0002] A fault knowledge base may be constructed for different faults, and the fault knowledge base may include fault information corresponding to different faults.

[0003] In the related art, fault information is mostly edited manually, that is, fault information is obtained by manually sorting fault data, but manual editing is inefficient and long-term editing may cause workers to become tired, resulting in poor editing accuracy of the fault knowledge base. Summary of the invention

[0004] In view of this, the embodiments of the present disclosure provide a method, device and equipment for constructing an aircraft engine fault data extraction model to solve the problems existing in the related technology.

[0005] According to a first aspect of an embodiment of the present disclosure, a method for constructing an aircraft engine fault data extraction model is provided, the method comprising: constructing a fault data ontology concept, and determining a labeling rule based on the fault data ontology concept, wherein the fault data ontology concept is used to describe entities, relationships between entities, and attributes corresponding to entities; obtaining multiple fault data generated by the aircraft engine in a fault state, and labeling the multiple fault data respectively according to the labeling rule; training a preset data extraction model according to the labeled fault data to obtain a target data extraction model.

[0006] According to a second aspect of an embodiment of the present disclosure, there is provided a device for constructing an aircraft engine fault data extraction model, which is applied to the method for constructing an aircraft engine fault data extraction model as in the first aspect. The device comprises: a construction module, which is used to construct a fault data ontology concept and determine a labeling rule based on the fault data ontology concept, wherein the fault data ontology concept is used to describe entities, relationships between entities and attributes corresponding to entities; an acquisition module, which is used to acquire multiple fault data generated by an aircraft engine in a fault state and label the multiple fault data respectively according to the labeling rules; and a training module, which is used to train a preset data extraction model according to the labeled fault data to obtain a target data extraction model.

[0007] A third aspect of an embodiment of the present disclosure provides a computer device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method for constructing the above-mentioned aircraft engine fault data extraction model.

[0008] A fourth aspect of the embodiments of the present disclosure provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the steps of the method for constructing the above-mentioned aircraft engine fault data extraction model.

[0009] According to a fifth aspect of the present disclosure, a computer program product is provided. When the computer program / instructions are executed by a processor, the steps of the method for constructing the above-mentioned aircraft engine fault data extraction model are implemented.

[0010] At least one of the above technical solutions adopted in the embodiment of the present disclosure can achieve the following beneficial effects: by constructing the concept of fault data ontology, and determining the labeling rules based on the concept of fault data ontology, wherein the concept of fault data ontology is used to describe entities, the relationship between entities and the attributes corresponding to entities; obtaining multiple fault data generated by the aircraft engine in a fault state, and labeling the multiple fault data respectively according to the labeling rules; training the preset data extraction model according to the labeled fault data to obtain the target data extraction model. Based on this, the fault data generated by the aircraft engine in a fault state can be obtained, and at the same time, the fault data ontology concept corresponding to the aircraft engine can be constructed, and the labeling rules can be determined according to the fault data ontology concept, and then the fault data can be labeled, and then the preset data extraction model can be trained using the labeled fault data to obtain the target data extraction model, and finally the target data extraction model can be used to realize the accurate extraction of the fault data of the aircraft engine, and the fault information corresponding to different fault data can be obtained, so as to facilitate the construction of the fault knowledge base. At the same time, it can promote the technical development in the task of troubleshooting aircraft engines, improve the level of intelligence, provide more reliable technical support for the safety management and maintenance of aircraft engines, and provide a reference for small sample data and highly professional information extraction tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other purposes, features and advantages of the present disclosure will become more apparent by describing the embodiments of the present disclosure in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation of the present disclosure. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0012] Figure 1 A schematic flow chart of a method for constructing an aircraft engine fault data extraction model provided by an exemplary embodiment of the present disclosure; Figure 2 A flowchart of an evaluation iteration method provided for an exemplary embodiment of the present disclosure; Figure 3A schematic diagram of the structure of an aircraft engine fault information ontology model layer provided by an exemplary embodiment of the present disclosure; Figure 4 A schematic diagram of the structure of a fault data extraction framework provided by an exemplary embodiment of the present disclosure; Figure 5 A schematic diagram of the internal architecture of a fault data extraction method provided by an exemplary embodiment of the present disclosure; Figure 6 A schematic flow chart of another method for constructing an aircraft engine fault data extraction model provided by an exemplary embodiment of the present disclosure; Figure 7 A schematic diagram of the structure of a device for constructing an aircraft engine fault data extraction model provided by an exemplary embodiment of the present disclosure; Figure 8 A schematic diagram of the structure of an electronic device provided by an exemplary embodiment of the present disclosure; Fig. 9 A schematic diagram of the structure of a computer system provided for an exemplary embodiment of the present disclosure; Fig.10 A schematic diagram of a computer program product provided for an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0013] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0014] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0015] The term "including" and its variations used in this document are open inclusions, that is, "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments". Relevant definitions of other terms will be given in the description below. It should be noted that the concepts of "first", "second", etc. mentioned in this disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0016] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0017] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0018] In practical applications, aircraft engines are high-end mechanical equipment with complex structures and high processing precision. Their operating status will directly affect the safety and reliability of aviation equipment. Various faults may occur during the operation of aircraft engines. Usually, when a fault occurs, relevant technicians are required to troubleshoot the aircraft engine, determine the cause of the fault, locate the fault, and take effective measures to repair the fault in a timely manner. In this process, how to use the fault data corresponding to the aircraft engine to generate valuable troubleshooting knowledge, and then assist relevant technicians to quickly complete the troubleshooting task is a difficult problem to be solved. At present, the fault knowledge base involved in related technologies is mostly manually edited, and there is little automatic extraction of fault data. The traditional manual knowledge extraction method is not only inefficient but also prone to errors.

[0019] In the existing technology, with the development of artificial intelligence technology, natural language processing (NLP) technology is also constantly improving, providing new solutions for the automatic extraction of aircraft engine fault data. Among them, data extraction is a very important content in natural language processing tasks, and it is the technical basis for application fields such as knowledge graphs, recommendation systems, and information retrieval. Its purpose is to extract structured data from unstructured or semi-structured text data, such as entities, relationships, attributes, and events. However, due to the high professionalism of the fault data corresponding to aircraft engines and the difficulty in obtaining fault data, the amount of fault data is insufficient and there is a lack of a large amount of labeled training data, which in turn affects the sufficient training of the commonly used basic models based on neural networks.

[0020] In view of the above problems, the disclosed embodiment provides a method for constructing an aircraft engine fault data extraction model, which can obtain fault data generated by an aircraft engine in a fault state, and at the same time, can construct a fault data ontology concept corresponding to the aircraft engine, and formulate labeling rules according to the fault data ontology concept to label the fault data, and then divide the labeled fault data into a training set, a validation set and a test set, and then can train the preset data extraction model through the training set, the validation set and the test set to obtain the target data extraction model, and finally can use the target data extraction model to achieve accurate extraction of aircraft engine fault data, thereby improving the accuracy and efficiency of aircraft engine fault diagnosis. At the same time, it can promote the technical development in the task of troubleshooting aircraft engines, improve the level of intelligence, provide more reliable technical support for the safe management and maintenance of aircraft engines, and provide a reference for small sample data and highly professional information extraction tasks. It should be understood that the Universal Information Extraction (UIE) model is an advanced information extraction technology that can achieve entity extraction, relationship extraction, event extraction, and sentiment analysis through a unified framework. The unified modeling of multiple tasks enables good migration and generalization capabilities between different tasks. Therefore, even when the sample size is small, the model can complete the extraction task well, thereby greatly reducing the cost and threshold of domain migration.

[0021] Figure 1 The following is a flow chart of a method for constructing an aircraft engine fault data extraction model provided by an exemplary embodiment of the present disclosure. Figure 1 As shown, specifically including: S101, acquiring fault data generated by the aircraft engine in a fault state, and storing and preprocessing the fault data. Specifically, the following steps are included: S1011, obtaining initial data generated during the operation of the aircraft engine.

[0022] In some embodiments, various faults may occur in the aircraft engine during operation. In order to effectively solve these faults, the initial data generated by the aircraft engine during the corresponding operation cycle can be collected, wherein the initial data can be all the data of the aircraft engine during the entire life cycle, or the initial data can also be the data generated by the aircraft engine during a certain working period. These initial data can then be analyzed to obtain fault data, based on which the corresponding fault resolution strategy can be obtained through the analysis of the fault data.

[0023] S1012, performing data cleaning on the initial data to obtain fault data.

[0024] In some embodiments, the collected initial data may come from standards, technical documents, analysis reports, engine field maintenance records, intelligence documents and other materials. Since there is no limitation on the type of data collected when collecting data, the initial data may include not only data generated under fault conditions, but also data generated under non-fault conditions. Therefore, after obtaining the initial data, it is necessary to desensitize and filter the initial data, so as to obtain the fault data generated by the aircraft engine under fault conditions.

[0025] Specifically, when the aircraft engine is just out of the factory, that is, at the beginning of its life cycle, since the parts of each part are relatively new, the failure rate is low, and the initial data collected may mostly be data in a non-fault state, that is, data generated when kinetic energy is normally provided; and when the aircraft engine has been used for a long time, that is, the life cycle enters the countdown, the parts of each part may age due to long-term use, at this time, the failure rate will increase, and the initial data collected may mostly be data in a fault state. Based on this, the disclosed embodiment can also perform data desensitization and screening and sorting on the collected initial data, that is, it can perform deformation processing on sensitive data, so that the processed data no longer contains sensitive information on the basis of retaining the original data format and certain statistical characteristics, thereby achieving the purpose of protecting data privacy, and at the same time filtering out the corresponding fault data in the fault state from the initial data, thereby reducing irrelevant workload and accelerating analysis efficiency, while improving the accuracy and reliability of subsequent fault data analysis.

[0026] For example, the files corresponding to the initial data can be converted into a unified text file format to ensure the consistency of the data format, which is convenient for subsequent neural network model training. After conversion to a unified text file format, the data irrelevant to aircraft engine failures in these files can be removed, and irrelevant rows, duplicate values, missing values, and redundant symbols that are obviously irrelevant to aircraft engine failure research can be deleted, and only relevant data content can be retained to obtain fault data under fault conditions.

[0027] S1013, pre-processing the fault data.

[0028] In some embodiments, there is generally a character count requirement for the input of the neural network model, so paragraphs or long sentences may not be used for neural network model training. Based on this, the disclosed embodiments can also perform word segmentation or sentence segmentation on paragraphs, that is, the full stop can be used as a cutoff point for sentence segmentation, and the long text can be split into independent sentences, which can facilitate the training of the neural network model and improve the accuracy of subsequent fault data extraction.

[0029] S102, formulate labeling rules according to the fault data ontology concept, and divide the fault data into training set, verification set and test set data according to the labeling rules.

[0030] In some embodiments, a fault data ontology concept corresponding to the aircraft engine fault data can be constructed, and labeling rules can be formulated based on the fault data ontology concept. Then, the fault data can be labeled according to the labeling rules. Finally, the labeled fault data can be segmented to obtain training set, verification set and test set data.

[0031] Specifically, the fault data ontology concept corresponding to aircraft engine fault data is a formalized, clear and normative description of the concepts and the relationships between concepts in the field of aircraft engines. It is like a "dictionary" and "map" of domain knowledge, providing an accurate description of various concepts, their attributes and their relationships in a specific field.

[0032] The above ontology is a rule for conceptual modeling, an abstract description of the objective world, and a formalized definition of concepts and the connections between concepts. Among them, ontology construction requires the definition of entities, attributes, and relationships respectively. Entity definition refers to the descriptive definition of a specific entity, attribute definition describes the characteristics and features of an entity, and relationship definition represents the semantic connection between entities.

[0033] For example, an entity may be a fault event or a fault detection method; a relationship may be a relationship between a fault time and a fault detection method, and the attributes corresponding to the fault event may include the fault occurrence time and the fault location, etc.

[0034] Based on this, the processed fault data can be manually labeled using this standard, and then the labeled data can be processed and split to generate training set, validation set and test set data.

[0035] Specifically, step S102 may include the following steps: S1021, constructing a fault data ontology concept corresponding to the aircraft engine fault data.

[0036] In some embodiments, when constructing the fault data ontology concept corresponding to the aircraft engine, the following principles may be included: First, the concept of fault data ontology must meet the needs of aircraft engine troubleshooting. To achieve this, during the construction of the fault data ontology concept, sufficient consultations can be conducted with relevant technical personnel and front-line aircraft maintenance personnel to fully understand the knowledge elements required for troubleshooting. Then, based on the knowledge elements required for troubleshooting, a fault data ontology concept for aircraft engine fault data that is recognized by industry professionals can be created. Then, based on the fault data ontology concept, a pattern layer can be obtained, which can provide a clear structural framework for fault data, so that fault data can be stored according to certain rules. It should be understood that the fault data ontology concept here can provide a semantically rich knowledge structure foundation for the pattern layer, and the concepts, relationships, and attributes in the fault data ontology concept can be mapped to the pattern layer to build a more semantic data model.

[0037] Secondly, continuous improvement based on application feedback and technical challenges is necessary. In the initial stage of the conceptual design of the fault data ontology, multiple entity types at different levels can be defined. For example, a system can include multiple modules, each module has its own corresponding components, and each component can be formed by splicing corresponding parts, that is, in terms of level, it can be system-module-component-part.

[0038] However, in the process of data extraction, all data are often extracted together. Thus, when determining the data corresponding to each level, it is necessary to search from multiple data, which is time-consuming and inefficient. Therefore, it becomes very challenging to determine the level of specific data only by context. Therefore, the disclosed embodiment can facilitate data search through the design of the pattern layer.

[0039] At the same time, when designing the model layer, the model layer should be as simple and effective as possible. In practical applications, as the number of entity types increases, the complexity of the model layer will increase. Therefore, a large amount of corpus labeling costs are required, and the workload required to extract the data corresponding to the entity and the relationship between the entities from the fault data will also increase. Therefore, while ensuring that the model layer meets the troubleshooting requirements, the model layer should be simplified as much as possible to avoid unacceptable manual labeling costs.

[0040] Finally, when designing the model layer, it is considered that fault data may contain both structured data and unstructured data. Therefore, structured data, such as Excel data and data in standardized XML format, contain rich information and can be directly converted into triples. Ideally, the model layer should include all possible structured data information. On the other hand, for unstructured data, mainly including Word and PDF documents, data extraction can be performed through deep learning or other technologies.

[0041] Figure 2The following is a flow chart of an iterative evaluation method provided by an exemplary embodiment of the present disclosure. Figure 2 As shown, specifically including: S201, determining the elements and scope of fault data.

[0042] In some embodiments, the time range of the fault occurrence can be determined, and data collection can be performed on the aircraft engine in the fault state within the time range. At the same time, data corresponding to designated elements can also be collected.

[0043] For example, if the aircraft engine starts to fail at 12:00 and stops working at 12:30, and if the compressor fails, the data corresponding to the compressor from 12:00 to 12:30 can be directly retrieved. It should be understood that an element refers to a type of data that exists in electronic form, participates in production and operation activities through calculations, and plays an important role.

[0044] S202, design an aircraft engine fault data extraction model.

[0045] In some embodiments, entity types related to fault data can be defined, and attributes corresponding to the entities can be defined, as well as relationships between different entities. Based on this, a fault data ontology concept for aircraft engine fault data recognized by industry professionals can be designed, and then an aircraft engine fault data extraction mode can be designed based on the fault data ontology concept.

[0046] S203, extracting fault data using the designed aircraft engine fault data extraction mode.

[0047] In some embodiments, data in the fault data can be extracted according to the data extraction rules specified by the aircraft engine fault data extraction pattern. Since the construction of the pattern layer is often based on the results obtained from the extraction pattern, the pattern layer can also be determined based on the extracted data.

[0048] S204, evaluating the fault data extraction mode.

[0049] In some embodiments, the maintenance engineers and domain experts within the engine life cycle can evaluate the pattern layer obtained by the fault data extraction pattern to obtain an evaluation result, and S202 can be executed based on the evaluation result, that is, the relevant definition content in the fault data extraction pattern is adjusted based on the evaluation result.

[0050] Following the above principles, the “evaluation iteration method” can be used at the model layer to construct an extraction model for aircraft engine fault data, and it can be continuously iterated and improved based on the evaluation of maintenance personnel and domain experts.

[0051] S1022, define entities, attributes and relationships between different entities corresponding to aircraft engine failure data.

[0052] Figure 3 A schematic diagram of the structure of an aircraft engine fault information ontology model layer provided for an exemplary embodiment of the present disclosure. The fault data ontology concept can define entities, attributes, and relationships respectively, where entity definition refers to a descriptive definition of a specific entity, attribute definition describes the characteristics and features of an entity, and relationship definition represents the semantic connection between entities.

[0053] like Figure 3 As shown in the figure, the rectangular box is the attribute, the circular box is the entity, the arrowed line indicates the relationship between them, and the line without arrows is the attribute corresponding to the entity. Specifically, the entity may include: unit, signal, fault, fault case and detection method, etc.; and the relationship may include: cause, detection, is part of, connected to, belongs to, location, indication and related, etc. For specific explanations, please refer to Table 1 and Table 2.

[0054] Table 1 Entity type definition

[0055] Table 2 Relationship type definition

[0056] Step S1023, formulating manual labeling rules.

[0057] In the related technology, there is no accurately labeled training data related to aircraft engine fault data. Therefore, manual labeling is required, and the labeled fault data is used for model training.

[0058] Specifically, the data obtained in S1013, that is, the pre-processed fault data, can be manually labeled using the open source software Wizard Labeling Assistant software based on the relationship between the entity and different entities defined in step S1022. After completing the labeling task, the data in json format is output and adjusted to a data format that can be input by the UIE model to prepare for subsequent model training.

[0059] Table 3 Data annotation

[0060] S1024, dividing the manually labeled fault data.

[0061] In some embodiments, the manually annotated fault data may be divided into a training set, a validation set, and a test set in a ratio of 8:1:1 for subsequent model training and evaluation.

[0062] S103, training a preset data extraction model using a training set, a validation set, and a test set to obtain a target data extraction model.

[0063] S1031, determine a preset data extraction model.

[0064] In some embodiments, the preset data extraction model can be a UIE model. The UIE model designs a structural schema instructor (SSI), which is mainly used to control the data extraction, data association and data generation of the UIE model. The structured extraction language (SEL) is designed to uniformly represent text data of different structures under different information extraction tasks. In addition, text data of different structures can also be uniformly encoded, and the target object can be adaptively extracted through a pattern-based prompt mechanism to achieve the extraction of large-scale text data into a specified structure. It should be understood that the UIE model can realize the joint extraction of entities and relationships in this task.

[0065] Specifically, the UIE model can use a structural pattern prompter SSI based on the schema prompt mechanism to adaptively generate the expected information during the extraction process. It should be understood that the schema is a structure used to organize and manage database objects. It defines the properties, relationships and constraints of various objects in the database, and the prompt refers to the text or instructions that provide input to the model to guide it to generate specific outputs. It is a text paragraph provided by the user when interacting with the model, which is used to describe the information, answers, text, etc. that the user wants to obtain from the model.

[0066] In some embodiments, the structural mode indicator SSI may include the following three types of segments: (1) SPOTNAME: extracts the target data fragments required in the task, such as entities; (2) ASSONAME: the association between two target data fragments, such as a relationship; (3) Token: The smallest unit in the text processing process, including [spot], [asso], and [text]. These symbols are used to distinguish which part the content in the input fault data belongs to.

[0067] For example, the UIE model can use the structured pattern prompter SSI based on the schema prompt mechanism to adaptively generate the expected data during the extraction process. The UIE model takes the defined SSI and fault data as input to generate a linearized SEL output result, whose expression is shown in Formula 1.

[0068] Formula 1 in, is the input fault data, s is the structural mode indicator, To extract and generate structured results, Indicates a connection symbol.

[0069] The entire input form of the UIE model is shown in Formula 2: Formula 2 in, Extract the elements of the pattern for the first structure, Extract the elements of the pattern for the second structure, is the element of the s-th structure extraction pattern, The specific content of the first fault data. The specific content of the second fault data is: is the specific content of the xth fault data. After expanding the content of Formula 2, the final model input result is shown in Formula 3: Formula 3 in, are the symbols corresponding to different entity categories, A symbol of relationship. For later access to data content.

[0070] Figure 4 FIG. 1 is a schematic diagram of a fault data extraction framework provided by an exemplary embodiment of the present disclosure. Figure 4 As shown in the figure, "Post-flight visual inspection found stains in the engine system, disassembly inspection / detection found that the main lubricating oil pump seal was damaged, and the problem was solved after replacing the main lubricating oil pump seal." can be input into the UIE model, and then "Fault: Main lubricating oil pump seal is damaged, detection method / detection: disassembly inspection" can be used as an example for architecture analysis, including: Figure 5 The internal architecture diagram of a fault data extraction method provided by an exemplary embodiment of the present disclosure is shown in FIG. Figure 5As shown, the tag (Class, CLS) represents a special symbol added to the beginning of the input text sequence, Encoder represents the encoder, Prompt is a text or instruction used to guide the model to generate a specific output, the boundary marker (Separator, SEP) is used to distinguish different parts of the text input, and Content represents the text content that actually needs to extract data. Based on this, in order to realize the linearized SEL, the encoder-decoder (Encoder-Decoder) architecture is used. When given an SSI(s) and fault data x as input, the UIE model can calculate the hidden representation of each token and encode it. Its expression is shown in Formula 4: Formula 4 in, is the hidden representation of token, is a Transformer encoder.

[0071] Furthermore, the UIE model can decode the input SSI(s) and fault data (x) into a linearized SEL in an autoregressive manner. In the i-th step of decoding, UIE will generate the result corresponding to the i-th token in the SEL sequence , and decode the corresponding state , its decoding expression is shown in Formula 5: Formula 5 in, is a Transformer decoder, and the prediction condition is In the first step of decoding, UIE will generate the result corresponding to the first token in the SEL sequence. , and decode the corresponding status result , decoded Step 1, UIE will generate the first Token corresponding to , and decode the corresponding status result , i represents the serial number corresponding to token.

[0072] In summary, a preset data extraction model can be determined.

[0073] Step S1032: pre-training the preset data extraction model, specifically including: (1) Data to structure pre-training.

[0074] Through pre-training, the UIE model can learn the mapping ability from data to structure, and generate the data structure to be extracted according to the schema and fault data. Its loss function As shown in Formula 6: Formula 6 in, The data pair is specifically represented as , l is the prompt before the input sentence, for each ,in is the fault data, is the data structure to be extracted, Represents the construction of negative samples, are the encoder parameters, is the decoder parameter, and p represents the probability.

[0075] (2) Structured generative pre-training.

[0076] In order to pre-train the UIE model to acquire the ability to generate structures that conform to SEL and schema, the input is only the front part of the structured data record and only the decoding part of UIE is trained to generate the remaining part, so that it can learn the SEL generation structure. Its autoregressive loss function is As shown in Formula 7: Formula 7 in, Represents a data record, specifically represented as a triple ,in, Indicates the data to be filled. is the conditional transition probability of the token.

[0077] (3) Improve semantic representation.

[0078] The training strategy (Masked Language Modeling, MLM) is used in the fault data to train the UIE model and improve its semantic representation ability. The semantic representation loss function As shown in Formula 8: Formula 8 in, The text dataset is specifically represented as , is the corrupted source fault data, The target span for destruction.

[0079] (4) Formula 9: pre-training loss function.

[0080] The final training task loss function of the UIE model is the final value of the above three tasks combined, and its loss function L is shown in Formula 9: Formula 9 S1033, UIE model fine-tuning.

[0081] In some embodiments, based on a given pre-trained UIE model, the model can be fine-tuned using aircraft engine fault data to achieve the task of extracting aircraft engine fault information. The training set, validation set, and test set divided in S102 are fine-tuned using the teacher-forcing cross loss function, as shown in Formula 10: Formula 10 in, represents the cross loss function, ={(l,m,n)} represents a given labeled corpus. Based on this, the fine-tuning function of the UIE model can be used to fine-tune the labeled aircraft engine fault data, and the aircraft engine fault entity and relationship extraction tasks can be achieved at the same time.

[0082] In summary, the target data extraction model can be obtained.

[0083] S104. Use NER evaluation index to evaluate the model test effect.

[0084] In some embodiments, the test result evaluation of the target data extraction model can use NER evaluation indicators, including precision P (Precision), recall R (Recall) and F1 value, and compare the UIE model with other knowledge extraction models under the same experimental configuration environment. Specifically, it includes: S1041, selection of evaluation indicators.

[0085] The model test results are evaluated using the named entity recognition model (NER), including precision P (Precision), recall R (Recall) and comprehensive metric F1 value to evaluate the UIE model and other model experimental results. The specific calculation formula is shown in Formula 11-13: Formula 11 Formula 12 Formula 13 Among them, TP is the number of correct identifications by the model, FP is the number of incorrect identifications by the model, and FN is the number of failures to identify by the model.

[0086] S1042. Comparative test.

[0087] In some embodiments, the UIE model can be compared with other mainstream knowledge extraction models, such as language models (Bidirectional Encoder Representations from Transformers, BERT), under the same experimental configuration to verify the superiority of the UIE model in the task of aircraft engine fault data extraction.

[0088] The method for constructing an aircraft engine fault data extraction model provided in the embodiment of the present disclosure may be executed by a terminal or by a chip applied to the terminal.

[0089] Exemplarily, the above-mentioned terminals may include one or more of mobile phones, tablet computers, wearable devices, vehicle-mounted devices, laptop computers, ultra-mobile personal computers (UMPCs), netbooks, personal digital assistants (PDAs), and wearable devices based on augmented reality (AR) and / or virtual reality (VR) technology, and may also include but not be limited to remote control devices, wearable devices, street lamps, home appliances and other smart terminals, which are not specifically limited in the embodiments of the present disclosure.

[0090] Figure 6 A flowchart of another method for constructing an aircraft engine fault data extraction model provided by an exemplary embodiment of the present disclosure. Figure 6 As shown, specifically including: S601, constructing a fault data ontology concept, and determining a labeling rule based on the fault data ontology concept, wherein the fault data ontology concept is used to describe entities, relationships between entities, and attributes corresponding to entities.

[0091] In some embodiments, a fault data ontology concept corresponding to the aircraft engine fault data may be constructed, and labeling rules may be formulated based on the fault data ontology concept. The fault data may then be labeled based on the labeling rules, thereby facilitating the processing of the fault data.

[0092] Specifically, the fault data ontology concept corresponding to aircraft engine fault data is a formalized, clear and normative description of the concepts and the relationships between concepts in the field of aircraft engines. It is like a "dictionary" and "map" of domain knowledge, providing an accurate description of various concepts, their attributes and their relationships in a specific field.

[0093] The above ontology is a rule for conceptual modeling, an abstract description of the objective world, and a formalized definition of concepts and the connections between concepts. Among them, ontology construction requires the definition of entities, attributes, and relationships respectively. Entity definition refers to the descriptive definition of a specific entity, attribute definition describes the characteristics and features of an entity, and relationship definition represents the semantic connection between entities.

[0094] For example, an entity may be a fault event or a fault detection method; a relationship may be the relationship between fault time and fault detection method, and the attributes corresponding to the fault event may include the fault occurrence time and fault location, etc. Based on this, the annotation rules can be determined based on the fault data ontology concept.

[0095] S602, obtaining a plurality of fault data generated by the aircraft engine in a fault state, and marking the plurality of fault data respectively according to a marking rule.

[0096] In some embodiments, data of the aircraft engine in a working state can be directly obtained through data acquisition equipment and other devices. Since faults may occur during the operation process, the collected data may include not only fault data in a fault state, but also data in a non-fault state. Based on this, the fault data can be first filtered out from the collected data. Specifically, the data generated during the fault period can be obtained starting from the time of the fault occurrence, that is, the data generated by the aircraft engine during the fault period can be obtained. Furthermore, when it is determined that a specific component of the aircraft engine has a fault, only the fault data corresponding to that part can be filtered out.

[0097] During the data collection process, multiple fault data may be obtained. Therefore, after obtaining the fault data, multiple fault data can be labeled separately according to the labeling rules. Specifically, manual labeling can be used. The open source software Wizard Labeling Assistant can be used for manual labeling. After completing the labeling task, the json format data is output and adjusted to the data format that can be input by the UIE model to prepare for subsequent model training.

[0098] S603, training a preset data extraction model according to the annotated fault data, and adjusting the trained preset extraction model using a cross loss function to obtain a target data extraction model.

[0099] Based on this, the fault data generated by the aircraft engine in the fault state can be obtained. At the same time, the fault data ontology concept corresponding to the aircraft engine can be constructed, and the annotation rules can be formulated according to the fault data ontology concept to annotate the fault data. Then, the preset data extraction model is trained using the annotated fault data, and the trained preset extraction model is adjusted using the cross loss function to obtain the target data extraction model. Finally, the target data extraction model can be used to accurately extract the fault data of the aircraft engine, thereby improving the accuracy and efficiency of the fault diagnosis of the aircraft engine. At the same time, it can promote the technical development in the task of troubleshooting aircraft engines, improve the level of intelligence, provide more reliable technical support for the safe management and maintenance of aircraft engines, and provide a reference for small sample data and highly professional information extraction tasks.

[0100] In some embodiments, determining the labeling rules based on the fault data ontology concept includes: determining a pattern layer according to the fault data ontology concept, and determining the labeling rules according to the pattern layer.

[0101] Specifically, the fault data ontology concept can accurately define the various entities, attributes and constraint relationships in aircraft engines, and the construction of the pattern layer often relies on the ontology concept to define its semantic basis. Based on this, the pattern layer can be determined according to the fault data ontology concept, thereby providing a clear structural framework for the fault data, so that the fault data can be stored according to certain rules.

[0102] In practical applications, the pattern layer can provide a systematic rule framework for data annotation. For example, in the information extraction task of natural language processing, the pattern layer will specify the annotation method of entities and relationships. It can clarify which data should be labeled as entities and which data should be labeled as relationships. Based on this, the annotation rules can be determined according to the pattern layer.

[0103] Specifically, in the field of aircraft engines, entities may include systems, components, and parts on aircraft engines, as well as fault cases and detection methods corresponding to each system, while relationships may include the connection relationship between systems and components, the connection relationship between components and parts, and the cause of the fault.

[0104] In some embodiments, a plurality of fault data generated by an aircraft engine in a fault state is obtained, including: obtaining initial data generated by the aircraft engine during operation; screening the initial data, and cleaning the initial data according to preset data cleaning rules to obtain fault data.

[0105] Specifically, there may be various faults in the operation of an aircraft engine. In order to effectively solve these faults, the initial data generated by the aircraft engine in the corresponding operation cycle can be collected, wherein the initial data can be all the data in the entire life cycle of the aircraft engine, or the initial data can also be the data generated by the aircraft engine in a certain working period, and then these initial data can be analyzed to obtain fault data. Based on this, the corresponding fault resolution strategy can be obtained by analyzing the fault data.

[0106] In some embodiments, the collected initial data may come from standards, technical documents, analysis reports, engine field maintenance records, intelligence documents and other materials. Since there is no limitation on the type of data collected when collecting data, the initial data may include not only data generated under fault conditions, but also data generated under non-fault conditions. Therefore, after obtaining the initial data, it is necessary to desensitize and filter the initial data, so as to obtain the fault data generated by the aircraft engine under fault conditions.

[0107] Specifically, when the aircraft engine is just out of the factory, that is, at the beginning of its life cycle, since the parts of each part are relatively new, the failure rate is low, and the initial data collected may mostly be data in a non-fault state, that is, data generated when kinetic energy is normally provided; and when the aircraft engine has been used for a long time, that is, the life cycle enters the countdown, the parts of each part may age due to long-term use, at this time, the failure rate will increase, and the initial data collected may mostly be data in a fault state. Based on this, the disclosed embodiment can also perform data desensitization and screening and sorting on the collected initial data, that is, it can perform deformation processing on sensitive data, so that the processed data no longer contains sensitive information on the basis of retaining the original data format and certain statistical characteristics, thereby achieving the purpose of protecting data privacy, and at the same time filtering out the corresponding fault data in the fault state from the initial data, thereby reducing irrelevant workload and accelerating analysis efficiency, while improving the accuracy and reliability of subsequent fault data analysis.

[0108] For example, the files corresponding to the initial data can be converted into a unified text file format to ensure the consistency of the data format, which is convenient for subsequent neural network model training. After conversion to a unified text file format, the data irrelevant to aircraft engine failures in these files can be removed, and irrelevant rows, duplicate values, missing values, and redundant symbols that are obviously irrelevant to aircraft engine failure research can be deleted, and only relevant data information is retained to obtain fault data under fault conditions.

[0109] In practical applications, the input of the neural network model generally has a character count requirement, so paragraphs or long sentences may not be used for neural network model training. Based on this, the disclosed embodiment can also perform word segmentation or sentence segmentation on the paragraph, that is, the full stop can be used as the cutoff point for sentence segmentation, and the long text can be split into independent sentences, which can facilitate the training of the neural network model and improve the accuracy of subsequent fault data extraction.

[0110] In some embodiments, a preset data extraction model is trained based on the labeled fault data, and the trained preset extraction model is adjusted using a cross-loss function to obtain a target data extraction model, including: dividing the labeled fault data according to a preset ratio to obtain a training set, a validation set, and a test set; training the preset data extraction model using the training set, the validation set, and the test set, and adjusting the trained preset extraction model using a cross-loss function to obtain a target data extraction model.

[0111] Specifically, the labeled fault data may be divided into a training set, a validation set, and a test set in a ratio of 8:1:1 for training the preset data extraction model.

[0112] In some embodiments, the fault data to be extracted corresponding to the aircraft engine to be processed can also be obtained; the target data extraction model is used to extract the fault data to be extracted to obtain a data extraction result.

[0113] Specifically, for the constructed target data extraction model, when in use, the fault data to be extracted can be directly input into the target data extraction model. After analysis, the target data extraction model can output a data extraction result. The data extraction result can be a structured information or a complete text fragment.

[0114] For example, fault data can be text data. When the input fault data content is "the temperature of the gas engine is too high at 12:00", the structured information can be "entity: gas engine, temperature is too high. 12:00; relationship: the temperature is too high on the gas engine, the fault occurs at 12:00". When the input fault data content is "gas engine, 12:00, temperature is too high", a complete text segment can be "the gas engine fails at 12:00". Based on this, valid fault information can be output for the input fault data. This can improve the efficiency and accuracy of aircraft engine fault information extraction and provide strong support for aircraft engine safety management and maintenance.

[0115] In some embodiments, the target data extraction model includes an encoder and a decoder, and the target data extraction model is used to extract data from multiple fault data, including: using the encoder to extract features of the fault data to be extracted; using the decoder to determine the data extraction results corresponding to the fault data to be extracted based on the features extracted by the encoding.

[0116] Specifically, when the fault data is text data, the encoder can extract features from the input text data. It converts information such as vocabulary and grammatical structure in the text data into an intermediate representation that the machine can understand; at the same time, the encoder can understand the semantic content of the text data. It analyzes factors such as the structure of sentences and the order of words, and encodes the semantic information of the text data into a vector space. The vectors in this vector space can represent different semantic features of the text data; it can capture contextual information in the text data. For example, in a long text paragraph, the meaning of a word is often affected by its context. At this time, the encoder will consider this contextual relationship so that the extracted features contain complete semantic information.

[0117] The decoder can decode the intermediate feature vectors extracted by the encoder and output structured information that meets the output requirements; it can also generate specific information based on the task objectives. For example, if it is used for event extraction, the decoder will generate key information about the event based on the encoded features of the input text data and then output it.

[0118] The above mainly introduces the solution provided by the embodiment of the present disclosure from the perspective of the server. It is understandable that, in order to implement the above functions, the server includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present disclosure can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present disclosure.

[0119] The embodiments of the present disclosure may divide the server into functional units according to the above method examples. For example, each functional module may be divided corresponding to each function, or two or more functions may be integrated into one management module. The above integrated modules may be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiments of the present disclosure is schematic and is only a logical functional division. There may be other division methods in actual implementation.

[0120] In the case of dividing each functional module according to each function, an exemplary embodiment of the present disclosure provides an apparatus for extracting fault data of an aircraft engine, and the apparatus for extracting fault data of an aircraft engine may be a server or a chip applied to a server. Figure 7 The present invention provides a schematic diagram of a device for constructing an aircraft engine fault data extraction model according to an exemplary embodiment of the present invention. Figure 7 As shown, the construction device 700 of the aircraft engine fault data extraction model includes: A construction module 701 is used to construct a fault data ontology concept and determine a labeling rule based on the fault data ontology concept, wherein the fault data ontology concept is used to describe entities, relationships between entities, and attributes corresponding to the entities; The acquisition module 702 is used to acquire a plurality of fault data generated by the aircraft engine in a fault state, and mark the plurality of fault data respectively according to the marking rule; The training module 703 is used to train the preset data extraction model according to the labeled fault data, and adjust the trained preset extraction model using a cross loss function to obtain a target data extraction model.

[0121] In an optional manner, the construction module 701 is further configured to determine a pattern layer according to the fault data ontology concept, and determine the annotation rule according to the pattern layer.

[0122] In an optional manner, the acquisition module 702 is also used to acquire initial data generated by the aircraft engine during operation; perform data screening on the initial data, and clean the initial data according to preset data cleaning rules to obtain the fault data.

[0123] In an optional manner, the training module 703 is also used to divide the labeled fault data according to a preset ratio to obtain a training set, a validation set and a test set; and use the training set, the validation set and the test set to train the preset data extraction model to obtain the target data extraction model.

[0124] In an optional manner, the fault data to be extracted corresponding to the aircraft engine to be processed is obtained; and the target data extraction model is used to extract the fault data to be extracted to obtain a data extraction result.

[0125] In an optional manner, the target data extraction model includes an encoder and a decoder, and the use of the target data extraction model to extract data from multiple fault data includes: using the encoder to extract features of the fault data to be extracted; and using the decoder to determine the data extraction result corresponding to the fault data to be extracted based on the features extracted by the encoding.

[0126] An embodiment of the present disclosure also provides an electronic device, comprising: at least one processor; a memory for storing instructions executable by at least one processor; wherein the at least one processor is used to execute instructions to implement the steps of the above method disclosed in the embodiment of the present disclosure.

[0127] Figure 8 The structure diagram of an electronic device provided by an exemplary embodiment of the present disclosure is shown in FIG. Figure 8 As shown, the electronic device 800 includes at least one processor 801 and a memory 802 coupled to the processor 801 , and the processor 801 can execute corresponding steps in the above method disclosed in the embodiment of the present disclosure.

[0128] The processor 801 may also be referred to as a central processing unit (CPU), which may be an integrated circuit chip having signal processing capabilities. Each step in the method disclosed in the embodiment of the present disclosure may be completed by an integrated logic circuit of hardware in the processor 801 or by instructions in the form of software. The processor 801 may be a general-purpose processor, a digital signal processor (DSP), an ASIC, a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in the embodiment of the present disclosure may be directly embodied as being executed by a hardware decoding processor, or may be executed by a combination of hardware and software modules in a decoding processor. The software module may be located in a memory 802, such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, or other mature storage media in the art. The processor 801 reads the information in the memory 802 and completes the steps of the method in combination with its hardware.

[0129] In addition, when various operations / processes according to the present disclosure are implemented by software and / or firmware, they can be transmitted from a storage medium or a network to a computer system having a dedicated hardware structure, for example, Fig. 9 The computer system 900 shown installs the programs constituting the software, and when the various programs are installed, the computer system can execute various functions, including the functions described above. Fig. 9 A schematic diagram of the structure of a computer system provided by an exemplary embodiment of the present disclosure.

[0130] Computer system 900 is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0131] like Fig. 9 As shown, the computer system 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 to a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the computer system 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0132] A plurality of components in the computer system 900 are connected to the I / O interface 905, including: an input unit 906, an output unit 907, a storage unit 908, and a communication unit 909. The input unit 906 may be any type of device capable of inputting information to the computer system 900, and the input unit 906 may receive input digital or character information, and generate key signal inputs related to user settings and / or function control of the electronic device. The output unit 907 may be any type of device capable of presenting information, and may include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 908 may include, but is not limited to, a disk, an optical disk. The communication unit 909 allows the computer system 900 to exchange information / data with other devices over a network such as the Internet, and may include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, for example, a Bluetooth™ device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0133] The computing unit 901 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 901 performs the various methods and processes described above. For example, in some embodiments, the above methods disclosed in the embodiments of the present disclosure may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on an electronic device via a ROM 902 and / or a communication unit 909. In some embodiments, the computing unit 901 may be configured to perform the above methods disclosed in the embodiments of the present disclosure in any other appropriate manner (e.g., by means of firmware).

[0134] The embodiment of the present disclosure also provides a computer-readable storage medium, wherein when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the above method disclosed in the embodiment of the present disclosure.

[0135] The computer-readable storage medium in the embodiments of the present disclosure may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. The computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the above. More specifically, the computer-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0136] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0137] Fig.10 Schematic diagram of a computer program product provided by an exemplary embodiment of the present disclosure. Fig.10 As shown, the computer program product 1000 includes a computer program 1001, wherein the computer program 1001 implements the above method disclosed in the embodiment of the present disclosure when executed by a processor.

[0138] In an embodiment of the present disclosure, computer program code for performing the operations of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including but not limited to object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer.

[0139] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0140] The modules, components or units involved in the embodiments described in the present disclosure may be implemented by software or hardware, wherein the names of the modules, components or units do not, in some cases, limit the modules, components or units themselves.

[0141] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, exemplary hardware logic components that may be used include, without limitation, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0142] The above descriptions are only some embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

[0143] Although some specific embodiments of the present disclosure have been described in detail by way of example, it should be understood by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present disclosure. It should be understood by those skilled in the art that the above embodiments may be modified without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A method for constructing an aircraft engine fault data extraction model, characterized in that: include: Constructing a fault data ontology concept and determining a labeling rule based on the fault data ontology concept, wherein the fault data ontology concept is used to describe entities, relationships between entities, and attributes corresponding to the entities; Acquire a plurality of fault data generated by the aircraft engine in a fault state, and respectively mark the plurality of fault data according to the marking rule; The preset data extraction model is trained according to the labeled fault data, and the trained preset extraction model is adjusted using a cross loss function to obtain a target data extraction model.

2. The method according to claim 1, characterized in that The determining of the labeling rule based on the fault data ontology concept includes: A pattern layer is determined according to the fault data ontology concept, and the labeling rule is determined according to the pattern layer.

3. The method according to claim 1, characterized in that The obtaining of a plurality of fault data generated by the aircraft engine in a fault state includes: Acquiring initial data generated by the aircraft engine during operation; The initial data is screened and cleaned according to preset data cleaning rules to obtain the fault data.

4. The method according to claim 1, characterized in that The method of training a preset data extraction model according to the labeled fault data and adjusting the trained preset extraction model using a cross loss function to obtain a target data extraction model includes: Dividing the annotated fault data according to a preset ratio to obtain a training set, a validation set, and a test set; The preset data extraction model is trained using the training set, the validation set and the test set, and the trained preset extraction model is adjusted using a cross loss function to obtain the target data extraction model.

5. The method according to claim 1, characterized in that The method further comprises: Obtaining fault data to be extracted corresponding to the aircraft engine to be processed; The target data extraction model is used to extract the fault data to be extracted to obtain a data extraction result.

6. The method according to claim 5, characterized in that The target data extraction model includes an encoder and a decoder, and the method of extracting data from a plurality of fault data using the target data extraction model includes: Using the encoder to perform feature extraction on the fault data to be extracted; The decoder is used to determine the data extraction result corresponding to the fault data to be extracted according to the features extracted by the encoding.

7. A device for constructing an aircraft engine fault data extraction model, characterized in that: include: A construction module, used to construct a fault data ontology concept and determine a labeling rule based on the fault data ontology concept, wherein the fault data ontology concept is used to describe entities, relationships between entities, and attributes corresponding to the entities; An acquisition module, used for acquiring a plurality of fault data generated by the aircraft engine in a fault state, and marking the plurality of fault data respectively according to the marking rule; The training module is used to train a preset data extraction model according to the labeled fault data to obtain a target data extraction model.

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

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method described in any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method described in any one of claims 1 to 6 are implemented.

Images