Aircraft maintenance activity failure impact knowledge base system for aircraft typical systems and establishment method

Abstract

This invention discloses a knowledge base system and method for establishing failure effects of aircraft maintenance activities for typical aircraft systems, belonging to the field of failure mode analysis (FMEA) technology for aircraft maintenance activities. It aims to solve problems such as traditional failure effect analysis methods not fully considering the associated effects of maintenance objects and the fact that maintenance activities are mostly stored in text format, which is not conducive to computer storage and reuse. The method includes: using FMEA to conduct failure effect analysis on typical aircraft systems and components; decomposing key operating state parameters for typical systems and components; constructing a knowledge base for the failure effects of maintenance objects; classifying maintenance activity steps and determining the maintenance objects and corresponding operating state parameters according to category correspondences; and associating failure modes and effects through operating state parameters. This improves the efficiency and accuracy of finding failure effects in maintenance activities, providing important data references for failure mode severity analysis and maintenance procedure improvement.

Description

Technical Field

[0001] This invention belongs to the field of failure mode analysis technology for aircraft maintenance activities, specifically involving a knowledge base system and method for establishing failure impacts of aircraft maintenance activities for typical aircraft systems. Background Technology

[0002] Aircraft maintenance is a crucial link in ensuring the safe operation of test aircraft. For test aircraft maintenance centered on research prototypes, the challenges are multifaceted and unpredictable. These challenges include numerous initial design flaws, immature onboard components, the first application of new technologies, significant differences in technical condition, and dynamic adjustments to maintenance support plans. Relying solely on management systems and personnel experience to control unsafe factors is too passive. Therefore, it is urgent to develop proactive safety strategies that focus on maintenance activities, starting with failure modes, analyzing the causes and impacts of maintenance failures, identifying potential safety risks, and formulating corresponding control measures.

[0003] In analyzing the potential safety risks of maintenance activities, determining the potential impact of failure modes is a crucial step. Analysis of actual operating conditions reveals that some safety risks may lead to testing machine malfunctions, causing safety hazards and, in severe cases, personal injury and property damage. If the failure impact is not quickly and accurately determined during the analysis, some serious failure modes may not be identified and controlled, resulting in even greater harm. However, traditional failure effect analysis methods rely on expert experience and primarily focus on the maintenance activity itself, failing to fully consider new failure modes arising from differences in the functional principles of the maintenance object. Furthermore, as testing machine models become increasingly complex, the number of systems increases, and their internal components become more intricate, the impact of maintenance activity failures may be overlooked, severely affecting subsequent hazard assessments and the selection of critical failure modes. In addition, maintenance activities are often stored in the form of text manuals, which, due to inconsistent writing standards, makes it difficult to retrieve maintenance activities, failure modes, and failure impacts, hindering computer storage and reuse.

[0004] Therefore, it is necessary to improve existing failure impact analysis methods for aircraft maintenance activities, enabling maintenance personnel to accurately pinpoint potential failure impacts when conducting safety risk analysis. Furthermore, considering the digitalization and intelligentization requirements of testing machine maintenance, converting maintenance activities into an intelligent knowledge database and transforming failure impacts into attributes of maintenance activity failure modes could significantly improve the efficiency of safety risk analysis and enable the secondary reuse of analysis results.

[0005] Based on this, the present invention proposes a knowledge base system and method for establishing a failure impact knowledge base for typical aircraft maintenance activities, in order to solve the problems existing in the prior art. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a knowledge base system and method for establishing a failure impact analysis system for typical aircraft maintenance activities. This system solves the problems of traditional failure impact analysis methods not fully considering the associated impacts on maintenance objects and the fact that maintenance activities are mostly stored in text format, which is not conducive to computer storage and reuse.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] This invention provides a first solution: a method for establishing a knowledge base system for the failure impacts of aircraft maintenance activities oriented towards typical aircraft systems, comprising:

[0009] Step S1: Determine typical aircraft systems, extract typical component units contained in each system based on the preset aircraft system classification criteria, and define them as the main maintenance object nodes in the graph model;

[0010] Step S2: Conduct FMEA analysis on typical components and typical systems to determine the possible failure modes, failure effects, failure causes, and severity scores corresponding to the failure modes.

[0011] Step S3: Use multiple key operating state parameters to jointly characterize the executable functions of a typical component unit, and then convert the different failure modes of the typical component unit into the intersection of events where one or more key operating state parameters do not meet the requirements.

[0012] Step S4: Utilize ontology models and knowledge graphs to construct a basic failure impact knowledge base for maintenance activities, storing failure modes, failure impacts, severity scores, and key operating status parameter information of maintenance objects;

[0013] Step S5: Classify maintenance activities according to their actions, and then construct a mapping rule table between maintenance activity classifications and key work status parameters based on the characteristics of the maintenance activities.

[0014] Step S6: Import the description text of maintenance activity steps and maintenance activity failure modes through natural language processing technology, preprocess the description text using preset regular expressions and word segmentation algorithms, then extract maintenance object information and main verbs of maintenance activities from the preprocessed maintenance activity text, and classify the information according to the verbs;

[0015] Step S7: Using the mapping rule table between maintenance activities and key working status parameters, the failure mode of maintenance activities is converted into a change in the key working status parameters of a certain maintenance object, and the corresponding failure impact of the maintenance object is obtained as the failure impact corresponding to the failure mode of maintenance activities.

[0016] Step S8: Add the associated results to the basic failure impact knowledge base and update the knowledge base to obtain a maintenance activity failure impact knowledge base for typical aircraft systems and typical component units.

[0017] In a preferred embodiment of the present invention, step S2, which involves conducting FMEA analysis on typical component units and typical systems, includes:

[0018] Step S21: Divide the FMEA analysis levels;

[0019] The lowest level of analysis is selected from aircraft maintenance actions and typical component units; the intermediate level of analysis is selected from aircraft subsystems; and the final level of analysis is selected from typical aircraft systems and the whole aircraft.

[0020] Step S22: Conduct functional analysis of typical component units and typical systems, provide functional descriptions for each task stage, and draw a structure-functional hierarchy diagram of the analysis object.

[0021] Step S23: Conduct failure analysis to determine the failure mode, failure cause, and failure effect of the object under analysis;

[0022] Among them: in the process of determining the cause of failure, only failure of the analysis object caused by erroneous maintenance activities is considered;

[0023] Step S24: Conduct risk analysis on the obtained failure modes to obtain a severity score for a particular failure mode.

[0024] In a preferred embodiment of the present invention, in step S3, during the process of using multiple key operating state parameters to jointly characterize the executable functions of a typical component unit, and then converting the different failure modes of the typical component unit into the intersection of events where one or more key operating state parameters do not meet the requirements, it is necessary to further decompose the obtained functions and failure modes of the maintenance object.

[0025] In a preferred embodiment of the present invention, step S4, which involves constructing a basic failure impact knowledge base for maintenance activities, includes:

[0026] Step S41: Construct an ontology model of the failure impact of maintenance activities, use extended entity triples to represent the failure impact information of maintenance activities, and store it in RDF format;

[0027] Step S42: Use the RDF import component of the graph database to convert the RDF format failure impact information extended triples into nodes and relationships in the knowledge graph database, forming the basic failure impact knowledge base for maintenance activities.

[0028] In a preferred embodiment of the present invention, step S5 divides maintenance activities into inspection actions, testing and functional verification actions, refilling actions, venting actions, applying actions, cleaning actions, adjustment actions, disassembly actions, installation actions, connection actions, plugging and unplugging actions, reading actions, recording actions, refurbishment actions, repair actions, and upgrade actions based on the characteristics of the maintenance activities.

[0029] In a preferred embodiment of the present invention, the mapping rule table in step S5 includes maintenance object information, maintenance action classification information, key working status parameters, key working status types, and typical physical quantities / units.

[0030] In a preferred embodiment of the present invention, step S6, which involves preprocessing the descriptive text, includes:

[0031] Step S611: Import the maintenance activity description text, use regular expressions to remove useless information from the maintenance activity description text, unify the Chinese and English punctuation information, and process the expressions that indicate separation in the text as parallel information;

[0032] Step S612: Search for possible physical units or logical symbols in the maintenance activity description text, and standardize the units and symbols in the maintenance activity text according to the standard units and symbols given in the key working status parameter information.

[0033] Step S613: Construct a synonym expression table for maintenance activities and a synonym expression table for maintenance objects to unify the expression of maintenance activities;

[0034] Step S614: Mark any light verbs that may appear in the maintenance activity description text.

[0035] In a preferred embodiment of the present invention, step S6, which involves extracting maintenance object information and key verbs of maintenance activities from the preprocessed maintenance activity text and classifying the information based on the verbs, includes:

[0036] Step S621: Identify the words in the preprocessed text, determine and label the part of speech of each word;

[0037] Step S622: Extract the maintenance object information from the maintenance activity text;

[0038] Based on the descriptions of typical components and typical systems in the existing knowledge base of failure effects of maintenance activities, the longest matching rule is used to find maintenance object information, the NER method is used to help identify the missing maintenance object information, and the possible location information of the maintenance object is determined and stored in the attributes of the maintenance object.

[0039] Step S623: Extract maintenance action information from the maintenance activity text;

[0040] The NER method is used to identify verb information contained in maintenance activity text. The input of NER is preprocessed text, and the output is labeled named entities such as verbs. The verb information is extracted using TF-IDF weights and rules. The input of TF-IDF is a set of text, and the output is the weight value of each word.

[0041] Step S624: Use a two-level mapping method to map the action information to the maintenance activity classification.

[0042] This invention provides a second solution: a knowledge base system for the failure impact of aircraft maintenance activities oriented towards typical aircraft systems, established by a method for establishing such a system.

[0043] In a preferred embodiment of the present invention, a knowledge base system for the failure impacts of aircraft maintenance activities oriented towards typical aircraft systems includes:

[0044] The typical component input module is used to input the typical aircraft system to be analyzed and the typical component units contained in the typical system;

[0045] The Functional FMEA Analysis module is used to perform functional FMEA analysis on specified typical component units and typical systems.

[0046] The typical component unit working state parameter association module is used to decompose the function of a typical component unit into one or more key working state parameters, and convert the failure mode corresponding to the function into the intersection of the working state parameters that do not meet the requirements.

[0047] The module for establishing mapping rules for maintenance activity working status parameters is used to classify maintenance activities and build mapping rules between different maintenance activity categories and key working status parameters.

[0048] The Failure Impact Knowledge Base Construction Module is used to convert the obtained failure impacts of the maintenance objects into storable knowledge data and build a basic failure impact knowledge base.

[0049] The maintenance activity import and action classification module is used to extract maintenance activity description text from maintenance activity data, preprocess the description text, extract maintenance objects and maintenance action description words from the maintenance activity text, and classify information according to maintenance actions;

[0050] The module for associating key operating status parameters with failure impacts is used to build the association between extracted maintenance action information and key operating status parameters. It uses existing classification information and maintenance object information to search for the corresponding key operating status parameters in the key operating status parameter mapping rule table and adds them to the corresponding maintenance action as attributes. Then, it associates the failure mode of maintenance activities with the failure impact of maintenance objects through key operating status parameters.

[0051] The maintenance activity failure impact database update module is used to update the associated maintenance activities and failure impact information into the constructed basic maintenance activity failure impact database, forming a complete maintenance activity failure impact database.

[0052] Compared with existing technologies, this invention provides a knowledge base system and method for establishing failure impacts of aircraft maintenance activities for typical aircraft systems, which has the following beneficial effects:

[0053] This invention constructs a mapping rule table between maintenance activities, the function of the maintenance object, and key operating status parameters. It characterizes maintenance activities as changes in the key operating status parameters of the maintenance object, standardizes the relationship between maintenance activities and the maintenance object, and lays a theoretical foundation for subsequent failure impact matching of maintenance activities.

[0054] This invention correlates the possible failure modes of each action in maintenance activities with the failure impact of the maintenance object by linking the changes in key operating state parameters, thereby quantifying the serious consequences that the potential safety risks of different maintenance actions may cause.

[0055] This invention designs an automated extraction and correlation method and system, which improves the efficiency and accuracy of analyzing the impact of maintenance failures, and provides a severity data foundation for subsequent analysis of potential safety risks and hazards in maintenance activities. Attached Figure Description

[0056] Figure 1 This invention provides a flowchart illustrating the method for establishing a knowledge base system for the failure impacts of aircraft maintenance activities for typical aircraft systems.

[0057] Figure 2 The diagram shows the ontological model of the failure impact of maintenance activities established for this invention.

[0058] Figure 3 This is a schematic diagram of the original document of the ARJ21 maintenance activity manual provided by the present invention.

[0059] Figure 4 The pre-processed maintenance activity manual description text diagram provided by the present invention.

[0060] Figure 5 This is a schematic diagram of the maintenance manual text processed by the dependency algorithm provided by the present invention.

[0061] Figure 6 The system structure diagram of the knowledge base system for failure impacts of aircraft maintenance activities for typical aircraft systems provided by this invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] As per the instruction manual Figure 1 - Appendix Figure 6 As shown, this specification first proposes a method for establishing a knowledge base system for the failure impacts of aircraft maintenance activities oriented towards typical aircraft systems, including the following steps:

[0064] Step S1: Analyze the aircraft, identify typical aircraft systems, and extract typical component units within each system based on preset aircraft system classification criteria, defining them as the main maintenance target nodes in the atlas model; specifically including the following steps:

[0065] Step S11: Based on the aircraft's primary purpose and the types of maintenance activities that may be carried out, aircraft are divided into four main categories: fighter jets, transport aircraft, helicopters, and unmanned aerial vehicles (UAVs). Each category can be further subdivided into 54 common system types. These 54 common system types are selected as the main objects of maintenance activities and identified as the typical systems of the aircraft.

[0066] Specifically, based on the classification of aircraft systems / subsystems and components in the national military standard GJB 4855A-2022 "Requirements for Aircraft System / Subsystem Classification", the aircraft is disassembled into 54 typical systems. The specific classification standards and system codes are shown in Table 1.

[0067] Table 1: Typical Aircraft Systems and Coding Examples

[0068]

[0069] Step S12: Further, analyze the main functions and possible components of the typical system, and in conjunction with the definition of the unit in the typical system in GJB 4833, give the typical component units contained in each system and identify them as the main maintenance objects.

[0070] Step S2: Conduct FMEA analysis on typical component units and typical systems to determine the possible failure modes, failure effects, failure causes, and severity scores corresponding to the failure modes. This process is based on the Functional FMEA analysis method and includes the following steps:

[0071] Step S21: Divide the FMEA analysis levels;

[0072] The lowest level of analysis is selected from aircraft maintenance actions and typical component units; the intermediate level of analysis is selected from aircraft subsystems; and the final level of analysis is selected from typical aircraft systems and the whole aircraft.

[0073] Step S22: Conduct functional analysis of typical component units and typical systems, provide functional descriptions for each task stage, and draw a structure-functional hierarchy diagram of the analysis object.

[0074] Step S23: Conduct failure analysis to determine the failure mode, failure cause, and failure effect of the analyzed object. In determining the failure cause, only failures caused by erroneous maintenance activities are considered to facilitate the correlation between failure modes and failure effects in subsequent maintenance activities.

[0075] Specifically, when conducting failure analysis, the first step is to determine the product's failure criteria based on its functions, output performance parameters, and permissible limits. Then, using an existing failure mode database, the failure modes for each function of the element of interest are identified, along with the impact of each failure mode on the next higher level of analysis and the potential causes of failure. Failure modes for commonly used typical component units can be obtained from domestic and international standards and manuals such as GJB / Z 299C and HB / Z 281-95, or determined by referring to a common failure mode sample table.

[0076] Step S24: Conduct risk analysis on the obtained failure modes to obtain a severity score for a particular failure mode.

[0077] Step S3: Combining the working principles of typical component units, using multiple key operating state parameters to jointly characterize the executable functions of a typical component unit, and then converting the different failure modes of the typical component unit into the intersection of events where one or more key operating state parameters do not meet the requirements; specifically including the following steps:

[0078] This step requires further decomposition of the functions and failure modes of the obtained maintenance objects: first, a database of working status parameters for aircraft components is constructed.

[0079] For a certain aircraft component, assuming it contains a set of operating state parameter vectors, then:

[0080] ;

[0081] in: This indicates that the aircraft component is in its operational state. The quantity is determined by the nature of the component's operation.

[0082] Specifically, taking the landing gear locking linkage of a certain type of aircraft as an example, this paper illustrates the method of decomposing the working state parameter vector. The landing gear locking linkage is installed on the aircraft landing gear and mainly functions to lock the landing gear struts, preventing the landing gear from retracting unexpectedly. Based on the function of the landing gear locking linkage, the state parameter vector of the landing gear locking linkage can be obtained. for:

[0083] ;

[0084] in: Indicates the insertion depth of the locking pin. This indicates the preload of the connecting bolts. Indicates hinge clearance. Indicates the thickness of the lubricating film or the lubrication condition. This indicates the deviation between travel distance and extreme position.

[0085] Based on the working characteristics of various systems and components of the aircraft, some common state parameters are given, as shown in Table 2.

[0086] Table 2: Typical Component Condition Parameters

[0087]

[0088] Step S4: Utilize ontology models and knowledge graphs to construct a basic failure impact knowledge base for maintenance activities, storing failure modes, failure effects, severity scores, and key operating status parameters of the maintenance objects. This includes the following steps:

[0089] Step S41: Construct an ontology model of the failure impact of maintenance activities, use extended entity triples to represent the failure impact information of maintenance activities, and store it in RDF format;

[0090] Wherein: the ontology model of the failure impact of the maintenance activity includes classes, attributes, and relationships;

[0091] Step S42: Use the n10s plugin (full name Neosemantics, an official plugin released by Neo4j Labs) to convert the RDF format failure impact information extended triples into nodes and relationships in a knowledge graph database, forming a basic failure impact knowledge base for maintenance activities;

[0092] In this system, classes and attributes serve as nodes in the graph database, while relationships act as edges connecting different nodes. The specific transformation process is as follows: First, install the n10s plugin and configure the environment in Neo4j (an open-source native graph database). Then, use the plugin's import command to load the RDF file. The input is RDF triple data, and the output is automatically generated nodes (such as classes and attributes) and edges (such as relationships), thereby constructing a complete knowledge graph.

[0093] like Figure 2 As shown, when constructing the ontology model of the failure impact of maintenance activities: First, the core concepts contained in the ontology model are defined, mainly including: "Product Composition Information", "LRU", "System", "Complete Machine", "Element of Concern", "Higher Analysis Level", "Lower Analysis Level", "Function", "Failure Mode", "Failure Cause", "Failure Impact", "Severity Score" and "State Parameter"; then, the relationships between the core concepts are sorted out, and the core concepts are transformed into classes in the ontology model to construct the ontology model of the failure impact of maintenance activities.

[0094] The constructed maintenance activity failure impact ontology model consists of three types of data: classes, relationships, and attributes. Classes include maintenance object composition information, typical component units, typical systems, the entire aircraft, the level of concern, the next lower analysis level, the previous higher analysis level, the function of the previous analysis level, the function of the concern level, the function of the next analysis level, failure modes, failure causes, failure effects, severity scores, and critical operating states. Relationships include composition relationships, designation relationships, realization relationships, inclusion relationships, and ownership relationships. Attributes include the aircraft name, aircraft serial number, typical system name, typical system serial number, typical component unit name, typical component unit serial number, description of the previous analysis level, description of the concern level, description of the next analysis level, description of the function of the previous analysis level, description of the function of the concern level, description of the function of the next analysis level, description of the failure mode, failure mode number, description of the failure effect, failure effect number, description of the failure cause, state parameter name, state parameter number, physical quantity unit, and severity score.

[0095] When constructing the knowledge graph database: First, the n10s plugin is used to convert the constructed maintenance activity failure impact ontology model into RDF triples. The classes, relations, and attributes in the maintenance activity failure impact ontology model are read and converted into nodes of the graph structure in the Neo4j knowledge graph. Then, knowledge is extracted from the semi-structured FMEA analysis content and state parameters, and Pandas in Python is used for data cleaning and mapping to populate the failure impact knowledge graph, resulting in the failure impact knowledge graph database model.

[0096] Step S5: Classify maintenance activities according to their actions, and then construct a mapping rule table between maintenance activity classifications and key working status parameters based on the characteristics of the maintenance activities; specifically including the following steps:

[0097] Based on the characteristics of maintenance activities, maintenance activities are divided into 16 categories, including inspection activities, testing and functional verification activities, adding activities, releasing activities, applying activities, cleaning activities, adjusting activities, disassembling activities, installing activities, connecting activities, plugging and unplugging activities, reading activities, recording activities, refurbishing activities, repair activities, and upgrading activities.

[0098] The process of constructing a mapping rule table between maintenance activity categories and key operating status parameters includes the following steps:

[0099] Based on the association information between maintenance objects and key operating states, a mapping rule table is constructed. The mapping rule table includes maintenance object information, maintenance action classification information, key operating state parameters, key operating state types, and typical physical quantities / units.

[0100] Key operating condition types include mechanical condition, lubrication condition, assembly condition, sealing condition, electrical condition, calibration condition, motion condition, thermal condition, vibration condition, and mode condition.

[0101] For maintenance activities and key operational status parameters, the normal execution of maintenance activities can be viewed as the process of adjusting a certain status parameter of the maintenance object. Adjust back to the design-allowed area ,Right now:

[0102] ;

[0103] The erroneous execution of a maintenance activity, i.e., the failure mode of a maintenance activity, can be equated to the state parameters of the maintenance object after a maintenance activity is executed for some reason. It landed in an abnormal area. For example, a maintenance failure mode: insufficient tightening torque during installation, can be equivalent to: , This indicates the minimum torque required to tighten.

[0104] Similarly, the failure modes of a maintenance object can be broken down into situations where the values ​​of one or more key operating parameters do not meet requirements. It can be represented as:

[0105] ;

[0106] Taking the failure mode "locking link not locked" of the "landing gear locking link" as an example, this paper explains the disassembly of the failure mode and the correlation of failure modes in maintenance activities. Analysis of the causes of the locking link not locking failure mode reveals three main reasons: insufficient locking pin insertion depth, insufficient locking stroke, and insufficient preload. Therefore, the locking link not locking failure mode can be represented as:

[0107] ;

[0108] in: This indicates the displacement stroke of the safety pin when it is fully inserted. Indicates the maximum travel of the interlock. This indicates the minimum preload of the bolt.

[0109] The process of correlating key operating condition parameters with maintenance activity failure modes includes: searching maintenance activity description texts reveals that the displacement of the safety pin is related to the installation and removal of the safety pin on the landing gear locking link; the locking stroke is related to the installation and removal of the landing gear locking link, and maintenance activities such as landing gear locking link displacement adjustment; the bolt preload is related to the installation and disassembly of the landing gear locking link, landing gear locking link stroke adjustment, and replacement of typical components of the landing gear locking link. Correlating changes in key operating condition parameters with maintenance activity failure modes yields mapping rules between maintenance activities and key operating condition parameters.

[0110] The mapping rules between maintenance activities and state parameters are as follows: For torque-related activities, if the maintenance steps include verbs such as "tighten" or "fasten," they are associated with the preload and torque in the state parameters. For clearance-related activities, if the maintenance steps include verbs such as "inspect," "measure," or "adjust," and have attributes such as "clearance" or "split," they are associated with the clearance parameter in the state parameters. For position-related activities, if the maintenance steps include verbs such as "inspect," "measure," or "confirm," and have attributes such as "in position," "locked position," "limit position," "stroke," or "deviation," they are associated with the position and stroke deviation in the state parameters. For lubrication-related activities, if the maintenance steps include verbs such as "lubricate," "apply," "coat," "add grease," or "add lubricant," they are associated with the lubrication status in the state parameters. For assembly / disassembly-related activities, if the maintenance steps include verbs such as "install," "disassemble," "install," "remove," "connect," or "disconnect," they are associated with the in-situ status in the state parameters.

[0111] It is important to note that, unlike the state parameters mentioned earlier, the in-situ state parameter is considered to be a discrete parameter, with only two states: in-situ and removed.

[0112] Step S6: Import the descriptive text of maintenance activity steps and failure modes using natural language processing technology. Preprocess the descriptive text using preset regular expressions and word segmentation algorithms. Then, extract maintenance object information and key verbs of maintenance activities from the preprocessed maintenance activity text, and classify the information based on the verbs. Specifically, this includes the following steps:

[0113] Step S61: Preprocess the description text, specifically including the following steps:

[0114] Step S611: Import the maintenance activity description text, use regularization to remove useless information such as numbers, parentheses, and redundant spaces from the maintenance activity description text, unify the Chinese and English punctuation information, and process any possible separators such as " / " in the text as parallel information.

[0115] Step S612: Search for possible physical units or logical symbols in the maintenance activity description text, and standardize the units and symbols in the maintenance activity text according to the standard units and symbols given in the key working status parameter information.

[0116] Step S613: Construct a synonym table for maintenance activities and a synonym table for maintenance objects. Based on the content of the synonym tables, normalize the synonyms that may exist in the maintenance activity description text;

[0117] Step S614: Mark the light verbs such as "conduct / implement / complete / launch" that may appear in the maintenance activity description text to facilitate the subsequent extraction of maintenance activity verbs.

[0118] Specifically, taking the ARJ21 aircraft maintenance manual as an example, the maintenance activity description text is imported, and the technical data is preprocessed, removing redundant, blank, and other useless information, and standardizing any possible physical units. The unprocessed maintenance activity description text is as follows: Figure 3 As shown, some of the preprocessed text information is as follows: Figure 4 As shown.

[0119] Step S62: Extract maintenance object information and key verbs of maintenance activities from the preprocessed maintenance activity text, and classify the information according to the verbs; specifically, this includes the following steps:

[0120] Step S621: Identify the words in the preprocessed text, determine and label the part of speech of each word. Dependency parsing is used to analyze and segment the processed text. Dependency parsing is a syntactic parsing method that analyzes the dependency relationships between sentence components. Its input is the preprocessed maintenance activity text, and its output is multiple segmented action clauses. The specific process includes constructing a dependency tree, identifying the subject-verb-object structure, and segmenting sentences according to the core verb to ensure that each action clause contains only one core action.

[0121] Step S622: Extract the maintenance object information from the maintenance activity text;

[0122] Based on the descriptions of typical component units and typical systems in the existing knowledge base of failure impacts of maintenance activities, the longest matching rule is used to find maintenance object information, and named entity recognition (NER) technology is used to help identify the missing maintenance object information and determine the possible location information of the maintenance object, which is stored in the attributes of the maintenance object.

[0123] Step S623: Extract maintenance action information from the maintenance activity text;

[0124] The Neural Error Analysis (NER) method is used to identify verb information in maintenance activity texts. NER takes preprocessed text as input and outputs labeled named entities such as verbs. TF-IDF weights and rules are then used to extract verb information. TF-IDF takes a text set as input and outputs the weight value for each word. The specific calculation formula is TF-IDF = TF × IDF, where TF is the term frequency (the number of times a word appears in a document divided by the total number of words in the document), and IDF is the inverse document frequency (log(total number of documents divided by the number of documents containing the word plus 1)). For potentially weak verbs, an extraction template is designed to filter out the true action information. The extracted verb results are normalized by mapping verbs to standard actions using an action thesaurus. If there are parallel verbs, they are retained as two separate action information entries.

[0125] Step S624: A two-level mapping method is used to map action information to 16 maintenance activity categories. The first-level mapping is the mapping of action information to maintenance activity categories, and the second-level mapping is the disambiguation of categories. Categories that cannot be determined are manually identified and entered into the maintenance activity failure impact database.

[0126] Specifically, the preprocessed text is first segmented and parsed syntactically. Dependency parsing is then used to determine the dependency relationships between verbs and objects, instruments, conditional adverbs, etc. For example, in the sentence "Remove the front landing gear pins and check the pin holes for damage," syntactic analysis can identify structures such as "remove -> pins (object)," "check -> pin holes (patient / object)," and "damage (object complement)." Through syntactic tree analysis, the maintenance actions and their corresponding maintenance objects and status characteristics are initially identified. Taking the ARJ21 maintenance activity manual as an example, the segmented text description is as follows: Figure 5 As shown.

[0127] Then, using the NER (Named Entity Recognition) method combined with deep learning technology, information on maintenance activities and maintenance objects is extracted. Specifically, a BiLSTM-CRF model is designed: first, a bidirectional LSTM layer is constructed with a hidden layer dimension of 128 to extract contextual features; then, a CRF layer is followed to ensure the globally optimal matching of the labeled sequence. The model takes a preprocessed text sequence as input at the word level and outputs a BIO tag sequence. The training process uses a labeled maintenance manual dataset with approximately 5000 sentences, employs the Adam optimizer, a learning rate of 0.001, and trains for 10 epochs, accurately labeling the key entity categories in the steps. For example, the BiLSTM-CRF model can label the "front landing gear locking linkage" as a <part> and the "torque wrench" as a <tool>.

[0128] Finally, the extracted action content is mapped to maintenance activity action categories. This mainly relies on the constructed thesaurus for verb keyword matching, selecting the appropriate action category to correspond to. For example, the maintenance action "remove" corresponds to "disassembly actions", and the maintenance action "inspect" corresponds to "inspection actions", etc.

[0129] Step S7: Using the mapping rule table between maintenance activities and key operating status parameters, the failure mode of maintenance activities is converted into a change in the key operating status parameter of a certain maintenance object, and the corresponding failure impact of the maintenance object is obtained as the failure impact corresponding to the failure mode of maintenance activities.

[0130] Step S8: Add the associated results to the basic failure impact knowledge base and update the knowledge base to obtain the failure impact knowledge base results for maintenance activities of typical aircraft systems and typical component units.

[0131] On the other hand, this specification provides a knowledge base system for the failure impacts of aircraft maintenance activities oriented towards typical aircraft systems, such as... Figure 6As shown, the aircraft maintenance activity failure impact knowledge base is used to implement steps S1-S8 of the method described above. Specifically, it includes a typical component input module, a functional FMEA analysis module, a typical component working state parameter association module, a maintenance activity working state parameter mapping rule establishment module, a failure impact knowledge base construction module, a maintenance activity import and action classification module, and a maintenance activity failure impact database update module; wherein:

[0132] The typical component input module is used to input the typical aircraft system to be analyzed and the typical component units contained in the typical system.

[0133] The functional FMEA analysis module is used to perform functional FMEA analysis on specified typical components and typical systems, determine different failure modes of typical components and systems, their impact on system and whole machine failure, and give severity scores corresponding to different failure modes.

[0134] The typical component unit working state parameter association module is used to decompose the function of the typical component unit into one or more key working state parameters, and convert the failure mode corresponding to the function into the intersection of the working state parameters not meeting the requirements.

[0135] The maintenance activity working status parameter mapping rule establishment module is used to classify maintenance activities and construct mapping rules between different maintenance activity categories and key working status parameters; the maintenance activity classification includes inspection actions, testing and functional verification actions, filling actions, venting actions, applying actions, cleaning actions, adjustment actions, disassembly actions, installation actions, connection actions, plugging and unplugging actions, reading actions, recording actions, refurbishment actions, repair actions, and upgrade actions;

[0136] The failure impact knowledge base construction module is used to convert the obtained failure impacts of the maintenance object into storable knowledge data and construct a basic failure impact knowledge base, storing the failure mode, failure impact, severity score, and key operating status parameter information of the maintenance object. The maintenance object failure impact conversion method includes converting the obtained failure mode, failure impact, severity score, and key operating status into triples to construct a failure impact ontology model of the maintenance object. Further, the n10s plugin is used to convert the ontology model into a knowledge graph, constructing the basic framework of the basic failure impact knowledge base.

[0137] The maintenance activity import and action classification module is used to extract maintenance activity description text from maintenance activity data, preprocess the description text, extract maintenance object and maintenance action description words from the maintenance activity text, and classify information according to maintenance actions. The preprocessing of the description text includes removing useless information from the text, standardizing possible physical units or logical symbols in the text, converting maintenance action description words and maintenance object description words in the description text into standard expressions, and marking verbs in maintenance activities. The maintenance action classification method includes segmenting the preprocessed text, identifying and marking the part-of-speech tags of words in the processed text, identifying maintenance object information of maintenance activities using the longest matching rule and named entity recognition technology, identifying and extracting maintenance action word information using the NER method and TF-IDF weights and rules, and mapping the action information to 16 maintenance activity categories using a two-level mapping method.

[0138] The critical operating status parameter and failure impact association module is used to construct the association between the extracted maintenance action information and critical operating status parameters. Utilizing existing classification information and maintenance object information, it searches for corresponding critical operating status parameters in the critical operating status parameter mapping rule table and adds them as attributes to the corresponding maintenance actions. Based on the critical operating status parameters and maintenance object information, it identifies the corresponding failure modes and failure impacts and associates them.

[0139] The maintenance activity failure impact database update module is used to update the associated maintenance activities and failure impact information into the constructed basic maintenance activity failure impact database, forming a complete maintenance activity failure impact database. This database stores the maintenance objects, key operating status parameters, failure modes, failure impacts, and severity scores for each maintenance activity. It also provides an output interface to output failure impact information and severity scores to the maintenance activity failure mode severity assessment system.

[0140] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for establishing a knowledge base system for the failure impacts of aircraft maintenance activities oriented towards typical aircraft systems, characterized in that, include: Step S1: Determine typical aircraft systems, extract typical component units contained in each system based on the preset aircraft system classification criteria, and define them as the main maintenance object nodes in the graph model; Step S2: Conduct FMEA analysis on typical components and typical systems to determine the possible failure modes, failure effects, failure causes, and severity scores corresponding to the failure modes. Step S3: Use multiple key operating state parameters to jointly characterize the executable functions of a typical component unit, and then convert the different failure modes of the typical component unit into the intersection of events where one or more key operating state parameters do not meet the requirements. Step S4: Utilize ontology models and knowledge graphs to construct a basic failure impact knowledge base for maintenance activities, storing failure modes, failure impacts, severity scores, and key operating status parameter information of maintenance objects; Step S5: Classify maintenance activities according to their actions, and then construct a mapping rule table between maintenance activity classifications and key work status parameters based on the characteristics of the maintenance activities. Step S5 categorizes maintenance activities into inspection, testing and functional verification, refilling, venting, applying, cleaning, adjustment, disassembly, installation, connection, plugging / unplugging, reading, recording, refurbishment, repair, and upgrade activities based on their characteristics. Step S5 Mapping rule table includes maintenance object information, maintenance action classification information, key working status parameters, key working status types and typical physical quantities / units; Step S6: Import the description text of maintenance activity steps and maintenance activity failure modes through natural language processing technology, preprocess the description text using preset regular expressions and word segmentation algorithms, then extract maintenance object information and main verbs of maintenance activities from the preprocessed maintenance activity text, and classify the information according to the verbs; Step S7: Using the mapping rule table between maintenance activities and key working status parameters, the failure mode of maintenance activities is converted into a change in the key working status parameters of a certain maintenance object, and the corresponding failure impact of the maintenance object is obtained as the failure impact corresponding to the failure mode of maintenance activities. Step S8: Add the associated results to the basic failure impact knowledge base and update the knowledge base to obtain a maintenance activity failure impact knowledge base for typical aircraft systems and typical component units.

2. The method for establishing a knowledge base system for the failure impacts of aircraft maintenance activities oriented towards typical aircraft systems as described in claim 1, characterized in that, Step S2 involves conducting FMEA analysis on typical component units and typical systems, including: Step S21: Divide the FMEA analysis levels; The lowest level of analysis is selected from aircraft maintenance actions and typical component units; the intermediate level of analysis is selected from aircraft subsystems; and the final level of analysis is selected from typical aircraft systems and the whole aircraft. Step S22: Conduct functional analysis of typical component units and typical systems, provide functional descriptions for each task stage, and draw a structure-functional hierarchy diagram of the analysis object. Step S23: Conduct failure analysis to determine the failure mode, failure cause, and failure effect of the object under analysis; Among them: in the process of determining the cause of failure, only failure of the analysis object caused by erroneous maintenance activities is considered; Step S24: Conduct risk analysis on the obtained failure modes to obtain a severity score for a particular failure mode.

3. The method for establishing a knowledge base system for the failure impact of aircraft maintenance activities oriented towards typical aircraft systems as described in claim 1, characterized in that, In step S3, during the process of using multiple key operating state parameters to jointly characterize the executable functions of a typical component unit, and then converting the different failure modes of the typical component unit into the intersection of events where one or more key operating state parameters do not meet the requirements, it is necessary to further decompose the obtained functions and failure modes of the maintenance object.

4. The method for establishing a knowledge base system for the failure impacts of aircraft maintenance activities oriented towards typical aircraft systems as described in claim 1, characterized in that, Step S4, the process of building a basic failure impact knowledge base for maintenance activities, includes: Step S41: Construct an ontology model of the failure impact of maintenance activities, use extended entity triples to represent the failure impact information of maintenance activities, and store it in RDF format; Step S42: Use the RDF import component of the graph database to convert the RDF format failure impact information extended triples into nodes and relationships in the knowledge graph database, forming the basic failure impact knowledge base for maintenance activities.

5. The method for establishing a knowledge base system for the failure impacts of aircraft maintenance activities oriented towards typical aircraft systems as described in claim 1, characterized in that, Step S6, the preprocessing of the description text, includes: Step S611: Import the maintenance activity description text, use regular expressions to remove useless information from the maintenance activity description text, unify the Chinese and English punctuation information, and process the expressions that indicate separation in the text as parallel information; Step S612: Search for possible physical units or logical symbols in the maintenance activity description text, and standardize the units and symbols in the maintenance activity text according to the standard units and symbols given in the key working status parameter information. Step S613: Construct a synonym expression table for maintenance activities and a synonym expression table for maintenance objects to unify the expression of maintenance activities; Step S614: Mark any light verbs that may appear in the maintenance activity description text.

6. The method for establishing a knowledge base system for the failure impacts of aircraft maintenance activities oriented towards typical aircraft systems as described in claim 1, characterized in that, Step S6 involves extracting maintenance object information and key verbs of maintenance activities from the preprocessed maintenance activity text, and classifying the information based on the verbs. Step S621: Identify the words in the preprocessed text, determine and label the part of speech of each word; Step S622: Extract the maintenance object information from the maintenance activity text; Based on the descriptions of typical components and typical systems in the existing knowledge base of failure effects of maintenance activities, the longest matching rule is used to find maintenance object information, the NER method is used to help identify the missing maintenance object information, and the possible location information of the maintenance object is determined and stored in the attributes of the maintenance object. Step S623: Extract maintenance action information from the maintenance activity text; The NER method is used to identify verb information contained in maintenance activity text. The input of NER is preprocessed text, and the output is labeled named entities. Verb information is extracted using TF-IDF weights and rules. The input of TF-IDF is a set of text, and the output is the weight value of each word. Step S624: Use a two-level mapping method to map the action information to the maintenance activity classification.

7. A knowledge base system for the failure impacts of aircraft maintenance activities on typical aircraft systems, established based on the method described in any one of claims 1-6, characterized in that, include: The typical component input module is used to input the typical aircraft system to be analyzed and the typical component units contained in the typical system; The Functional FMEA Analysis module is used to perform functional FMEA analysis on specified typical component units and typical systems. The typical component unit working state parameter association module is used to decompose the function of a typical component unit into one or more key working state parameters, and convert the failure mode corresponding to the function into the intersection of the working state parameters that do not meet the requirements. The module for establishing mapping rules for maintenance activity working status parameters is used to classify maintenance activities and build mapping rules between different maintenance activity categories and key working status parameters. The Failure Impact Knowledge Base Construction Module is used to convert the obtained failure impacts of the maintenance objects into storable knowledge data and build a basic failure impact knowledge base. The maintenance activity import and action classification module is used to extract maintenance activity description text from maintenance activity data, preprocess the description text, extract maintenance objects and maintenance action description words from the maintenance activity text, and classify information according to maintenance actions; The module for associating key operating status parameters with failure impacts is used to build the association between extracted maintenance action information and key operating status parameters. It uses existing classification information and maintenance object information to search for the corresponding key operating status parameters in the key operating status parameter mapping rule table and adds them to the corresponding maintenance action as attributes. Then, it associates the failure mode of maintenance activities with the failure impact of maintenance objects through key operating status parameters. The maintenance activity failure impact database update module is used to update the associated maintenance activities and failure impact information into the constructed basic maintenance activity failure impact database, forming a complete maintenance activity failure impact database.

Images