Aircraft maintenance manual data labeling method and system oriented to deep learning
By employing a data annotation method based on deep learning, the problem of large amounts of aircraft maintenance data with diverse formats was solved, enabling intelligent processing and fast and accurate information retrieval.
Patent Information
- Application Number
- CN202510994542.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-18
- Publication Date
- 2025-11-04
AI Technical Summary
Aircraft maintenance data is voluminous and comes in various formats, making it difficult to quickly and accurately access information, and existing large language models are complex to process.
We employ a deep learning-oriented data annotation method, which includes acquiring data annotation standards, standardizing format processing, automated annotation, and manual verification. We utilize big data models and convolutional neural networks for text and image recognition, and train and validate deep learning models.
It enables intelligent processing of maintenance data, improves information retrieval and analysis efficiency, and helps users quickly and accurately find the information they need.
Smart Images
Figure CN120893404A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of aircraft maintenance technology, and specifically relates to a data annotation method and system for aircraft maintenance manuals based on deep learning. Background Technology
[0002] Aircraft maintenance involves complex technical knowledge and a vast amount of maintenance documentation, including technical manuals, maintenance records, and drawings, resulting in a massive amount of information. This information often contains highly specialized technical terminology and knowledge, making it difficult to quickly and accurately retrieve the required information. Furthermore, aircraft maintenance documentation encompasses various forms of information, such as text, images, and charts. Current technological advancements suggest applying large language modeling (MLM) technology to user technical documentation. While MLM can process and understand human language on a large scale, making it highly proficient in language processing tasks, the need for comprehensive processing of data in different formats and types increases the complexity of retrieving and accessing MLM data. Therefore, a data processing approach based on deep learning and understanding is required to facilitate the engineering application of MLM. Summary of the Invention
[0003] To address the aforementioned issues, this application provides a deep learning-based method and system for annotating aircraft maintenance manual data.
[0004] The first aspect of this application provides a deep learning-based method for annotating aircraft maintenance manual data, which mainly includes:
[0005] Step S1: Obtain the data labeling standard;
[0006] Step S2: Standardize the format of data obtained from different data sources;
[0007] Step S3: Use big data models to perform automated text tagging and automated image tagging respectively;
[0008] Step S4: Based on the automated identification and marking, perform manual verification and correction;
[0009] Step S5: Train and validate the deep learning model.
[0010] Preferably, step S1 further includes:
[0011] Determine the object to be marked, the mark category, the terminology or abbreviation, and the mark format.
[0012] Preferably, step S2 further includes:
[0013] All graphics will be converted to PNG format, and text content will be converted to txt format.
[0014] Preferably, in step S3, the automated text tagging includes:
[0015] Step S31: Identify entities in the text and add marks or labels to them using uniform terminology or abbreviations;
[0016] Step S32: Identify the relationships between entities based on the sentence structure in the text;
[0017] Step S33: Identify the specific operation steps or processes described in the text and add tags to them;
[0018] Step S34: Identify the fault phenomena, causes, and solutions described in the text;
[0019] Step S35: Identify safety precautions and warnings in the text and add specified tags to them.
[0020] Preferably, in step S3, the automated image labeling includes:
[0021] Identify objects in images and label the corresponding components, systems, regions, and fault phenomena.
[0022] The second aspect of this application provides a deep learning-based aircraft maintenance manual data annotation system, which mainly includes:
[0023] The tagging standard acquisition module is used to acquire data tagging standards;
[0024] The data preprocessing module is used to standardize the format of data obtained from different data sources;
[0025] The automated tagging module is used to perform automated text tagging and automated image tagging using big data models, respectively.
[0026] The manual verification and correction module is used to perform manual verification and correction based on the automated identification of tags;
[0027] The validation module is used for training and validation of deep learning models.
[0028] Preferably, the marking standard acquisition module includes:
[0029] The marking standard determination unit is used to determine the marking object, marking category, terminology or abbreviation, and marking format.
[0030] Preferably, the data preprocessing module includes:
[0031] The graphics preprocessing unit is used to convert various graphics into PNG format.
[0032] The text preprocessing unit is used to convert text content into a unified txt format.
[0033] Preferably, the automated marking module includes:
[0034] Entity recognition unit, used to identify entities in text and add marks or labels to them using uniform terminology or abbreviations;
[0035] The entity relationship recognition unit is used to identify the relationships between entities based on the sentence structure in the text;
[0036] The process identification unit is used to identify the specific operation steps or processes described in the text and add tags to them.
[0037] The fault identification unit is used to identify the fault phenomena, causes, and solutions described in the text.
[0038] The safety warning recognition unit is used to identify safety precautions and warning messages in text and add specified tags to them.
[0039] Preferably, the automated marking module includes:
[0040] The image recognition and labeling unit is used to identify objects in an image and label the corresponding components, systems, areas, and fault phenomena.
[0041] This application enables intelligent processing and understanding of maintenance data, improves information retrieval and analysis efficiency, and helps users quickly and accurately find the information they need. Attached Figure Description
[0042] Figure 1 This is a flowchart of a preferred embodiment of the aircraft maintenance manual data annotation method based on deep learning in this application. Detailed Implementation
[0043] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are only some, not all, of the embodiments of this application. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. The embodiments of this application will be described in detail below with reference to the accompanying drawings.
[0044] The first aspect of this application provides a deep learning-based method for annotating aircraft maintenance manual data, such as... Figure 1 As shown, it mainly includes:
[0045] Step S1: Obtain the data labeling standard;
[0046] Step S2: Standardize the format of data obtained from different data sources;
[0047] Step S3: Use big data models to perform automated text tagging and automated image tagging respectively;
[0048] Step S4: Based on the automated identification and marking, perform manual verification and correction;
[0049] Step S5: Train and validate the deep learning model.
[0050] In some alternative implementations, step S1 further includes:
[0051] Determine the object to be marked, the mark category, the terminology or abbreviation, and the mark format.
[0052] In this embodiment, the labeled objects include, for example, text content, image information, and chart data. The label category refers to the specific category that needs to be labeled, and the terminology or abbreviation refers to the standardized terminology and abbreviations used to ensure consistency in terminology during data labeling and use, avoiding ambiguity. The label format includes, for example, the position, color, size, and style of the text, ensuring the label is clearly visible.
[0053] In some alternative implementations, step S2 further includes:
[0054] All graphics will be converted to PNG format, and text content will be converted to txt format.
[0055] This embodiment collects data from various data sources in the aircraft maintenance manual and then performs format conversion of the graphic and text content according to the above requirements.
[0056] In some alternative implementations, step S3, performing automated text tagging, includes:
[0057] Step S31: Identify entities in the text and add marks or labels to them using uniform terminology or abbreviations;
[0058] Step S32: Identify the relationships between entities based on the sentence structure in the text;
[0059] Step S33: Identify the specific operation steps or processes described in the text and add tags to them;
[0060] Step S34: Identify the fault phenomena, causes, and solutions described in the text;
[0061] Step S35: Identify safety precautions and warnings in the text and add specified tags to them.
[0062] In this embodiment, for automated text tagging, a large language model can be used to recognize the text, such as using AliTranslate or Wenxin Yiyan. Step S31 is used to identify entities in the text, such as important information like component names, tool names, and operating steps, and add tags or labels to them. Additionally, it is necessary to standardize the terminology and abbreviations in the text to ensure consistency and accuracy of the tagging. Step S32 is used to extract relationships, analyze the sentence structure in the text, and identify the relationships between entities, such as the association between components and fault phenomena, which helps in understanding the text content. Step S33 is used to identify the specific operating steps or processes described in the text and add tags to them to better organize and understand the content of the maintenance manual. Step S34 is used to identify the fault phenomena, causes, and solutions described in the text to help technicians quickly locate and resolve faults. Step S35 is used to identify safety precautions and warnings in the text and add specific tags to them to ensure that personnel comply with relevant safety regulations during maintenance.
[0063] In addition, keyword extraction should be performed to extract representative and important keywords or phrases from the text to help quickly understand the key points of the text content.
[0064] In some alternative implementations, step S3, performing automated image tagging, includes:
[0065] Identify objects in images and label the corresponding components, systems, regions, and fault phenomena.
[0066] This embodiment can utilize Baidu PaddlePaddle's Convolutional Neural Network (CNN) algorithm to perform object recognition on PNG images.
[0067] Step S4 is used to verify and correct the automatically identified tags. Step S5 is used to train and validate the deep learning model. In this embodiment, deep learning technology is used to train a model on the labeled and annotated data to recognize and understand various information in the aircraft maintenance manual. The deep learning-trained model is integrated into the aircraft maintenance manual system for validation, and the tagging method is iteratively optimized.
[0068] The second aspect of this application provides a deep learning-based aircraft maintenance manual data annotation system corresponding to the above-described method, mainly comprising:
[0069] The tagging standard acquisition module is used to acquire data tagging standards;
[0070] The data preprocessing module is used to standardize the format of data obtained from different data sources;
[0071] The automated tagging module is used to perform automated text tagging and automated image tagging using big data models, respectively.
[0072] The manual verification and correction module is used to perform manual verification and correction based on the automated identification of tags;
[0073] The validation module is used for training and validation of deep learning models.
[0074] In some optional implementations, the tagging standard acquisition module includes:
[0075] The marking standard determination unit is used to determine the marking object, marking category, terminology or abbreviation, and marking format.
[0076] In some alternative implementations, the data preprocessing module includes:
[0077] The graphics preprocessing unit is used to convert various graphics into PNG format.
[0078] The text preprocessing unit is used to convert text content into a unified txt format.
[0079] In some alternative implementations, the automated tagging module includes:
[0080] Entity recognition unit, used to identify entities in text and add marks or labels to them using uniform terminology or abbreviations;
[0081] The entity relationship recognition unit is used to identify the relationships between entities based on the sentence structure in the text;
[0082] The process identification unit is used to identify the specific operation steps or processes described in the text and add tags to them.
[0083] The fault identification unit is used to identify the fault phenomena, causes, and solutions described in the text.
[0084] The safety warning recognition unit is used to identify safety precautions and warning messages in text and add specified tags to them.
[0085] In some alternative implementations, the automated tagging module includes:
[0086] The image recognition and labeling unit is used to identify objects in an image and label the corresponding components, systems, areas, and fault phenomena.
[0087] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method for annotating aircraft maintenance manual data based on deep learning, characterized in that, include: Step S1: Obtain the data labeling standard; Step S2: Standardize the format of data obtained from different data sources; Step S3: Use big data models to perform automated text tagging and automated image tagging respectively; Step S4: Based on the automated identification and marking, perform manual verification and correction; Step S5: Train and validate the deep learning model.
2. The aircraft maintenance manual data annotation method based on deep learning as described in claim 1, characterized in that, Step S1 further includes: Determine the object to be marked, the mark category, the terminology or abbreviation, and the mark format.
3. The deep learning-based aircraft maintenance manual data annotation method as described in claim 2, characterized in that, Step S2 further includes: All graphics will be converted to PNG format, and text content will be converted to txt format.
4. The deep learning-based aircraft maintenance manual data annotation method as described in claim 1, characterized in that, In step S3, the automated text tagging includes: Step S31: Identify entities in the text and add marks or labels to them using uniform terminology or abbreviations; Step S32: Identify the relationships between entities based on the sentence structure in the text; Step S33: Identify the specific operation steps or processes described in the text and add tags to them; Step S34: Identify the fault phenomena, causes, and solutions described in the text; Step S35: Identify safety precautions and warnings in the text and add specified tags to them.
5. The deep learning-based aircraft maintenance manual data annotation method as described in claim 1, characterized in that, In step S3, the automated image labeling includes: Identify objects in images and label the corresponding components, systems, regions, and fault phenomena.
6. A data annotation system for aircraft maintenance manuals based on deep learning, characterized in that, include: The tagging standard acquisition module is used to acquire data tagging standards; The data preprocessing module is used to standardize the format of data obtained from different data sources; The automated tagging module is used to perform automated text tagging and automated image tagging using big data models, respectively. The manual verification and correction module is used to perform manual verification and correction based on the automated identification of tags; The validation module is used for training and validation of deep learning models.
7. The aircraft maintenance manual data annotation system based on deep learning as described in claim 6, characterized in that, The tagging standard acquisition module includes: The marking standard determination unit is used to determine the marking object, marking category, terminology or abbreviation, and marking format.
8. The aircraft maintenance manual data annotation system based on deep learning as described in claim 7, characterized in that, The data preprocessing module includes: The graphics preprocessing unit is used to convert various graphics into PNG format. The text preprocessing unit is used to convert text content into a unified txt format.
9. The aircraft maintenance manual data annotation system based on deep learning as described in claim 6, characterized in that, The automated tagging module includes: Entity recognition unit, used to identify entities in text and add marks or labels to them using uniform terminology or abbreviations; The entity relationship recognition unit is used to identify the relationships between entities based on the sentence structure in the text; The process identification unit is used to identify the specific operation steps or processes described in the text and add tags to them. The fault identification unit is used to identify the fault phenomena, causes, and solutions described in the text. The safety warning recognition unit is used to identify safety precautions and warning messages in text and add specified tags to them.
10. The aircraft maintenance manual data annotation system based on deep learning as described in claim 6, characterized in that, The automated tagging module includes: The image recognition and labeling unit is used to identify objects in an image and label the corresponding components, systems, areas, and fault phenomena.