Aircraft cabin door state monitoring method and device and electronic equipment

Abstract

The invention provides an aircraft cabin door state monitoring method and device and electronic equipment. The method comprises the following steps: acquiring target pressure change data of an aircraft cabin door; obtaining a target triple according to the target voltage change data; wherein the target triad comprises an entity, a relationship and an attribute; performing matching in a preset cabin door fault knowledge base by utilizing the target triad to obtain a matched hit triad; wherein the preset cabin door fault knowledge base comprises a plurality of historical triples; the historical triple is obtained based on historical pressure change data of the aircraft cabin door; the historical triples comprise hit triples; and according to the relationship corresponding to the hit triple, generating target alarm information of the aircraft cabin door. Based on the technical scheme, monitoring and early warning of the state of the cabin door of the aircraft can be realized, and the safety of the aircraft is improved.

Description

Technical Field

[0001] This application relates to the field of aircraft control technology, and in particular to a method, device, and electronic equipment for monitoring the status of aircraft doors. Background Technology

[0002] As frequently used moving functional components, aircraft doors directly impact flight safety in terms of function, lifespan, safety, and reliability. With repeated opening and closing of doors, and the mechanical loads and cabin pressurization loads they endure during flight, the health of doors faces potential risks. Minor deformations and cracks may occur in the door structure, difficult to detect manually, leading to door opening and closing difficulties, or even airtight defects causing cabin depressurization or door detachment—serious flight accidents. Therefore, the need for timely, sensitive, and comprehensive monitoring of door health is increasingly urgent.

[0003] In related technologies, model-based methods require the establishment of an accurate mathematical model of the monitored location (hatch), and the state of the monitored location is determined by comparing the model with actual data, which has high computational complexity. Summary of the Invention

[0004] This application proposes a method, device, and electronic equipment for monitoring the status of aircraft cabin doors to solve the technical problem of high complexity in calculating early warning information.

[0005] To achieve the above objectives, according to a first aspect of this application, a method for monitoring the status of an aircraft door is provided, comprising:

[0006] Acquire target pressure change data for the aircraft hatch;

[0007] Based on the target pressure transformer data, a target triple is obtained; wherein, the target triple includes: entity, relation, and attribute;

[0008] The target triplet is matched against a pre-defined door fault knowledge base to obtain a matching hit triplet; wherein, the pre-defined door fault knowledge base includes multiple historical triplets; the historical triplets are obtained based on the historical pressure transformer data of the aircraft door; the historical triplets include the hit triplet.

[0009] Based on the relationship between the hit triplet, target alarm information for the aircraft door is generated.

[0010] In some embodiments, acquiring target pressure change data of the aircraft door includes:

[0011] Acquire the initial pressure-transformation data of the aircraft cabin door;

[0012] The initial voltage transformer data is aligned and standardized to obtain the first voltage transformer data;

[0013] Change point detection is performed on the first voltage transformer data to determine the time nodes of state change in the first voltage transformer data;

[0014] A window is constructed based on the aforementioned state change time points;

[0015] The first voltage transformer data outside the window is compressed to obtain the target voltage transformer data.

[0016] In some embodiments, the entity includes: a fault ontology and a data ontology;

[0017] The relationships include: the relationships between the fault entities, the relationships between the data entities, and the relationships between the fault entities and the data entities.

[0018] In some embodiments, obtaining a target triplet based on the target voltage transformer data includes:

[0019] The fault entity is extracted from the door text data, and the relationships between the fault entities and the attributes of the fault entities are determined.

[0020] The data ontology is extracted from the target pressure transformer data, and the attributes of the data ontology and the relationships between the data ontology are determined. The attributes of the data ontology are obtained based on the data change characteristics of the data ontology. The relationships between the data ontology include proportional relationships and sequential relationships. The proportional relationships are obtained based on the correlation coefficient of the data ontology. The sequential relationships are obtained based on the timestamp of the data ontology.

[0021] The data ontology and the fault ontology are input into a neural network model to obtain the relationship between the fault ontology and the data ontology.

[0022] Construct a target triple based on the data ontology, the fault ontology, the attributes of the data ontology, the attributes of the fault ontology, the relationships between the fault ontologs, the relationships between the data ontologs, and the relationships between the fault ontology and the data ontology.

[0023] In some embodiments, the method further includes:

[0024] Calculate the first similarity between every two of the data ontologies;

[0025] If the first similarity is greater than a first preset value, the two data ontologies are merged to obtain a merged data ontology; the merged data ontology is used to construct the target triplet.

[0026] Calculate the second similarity between every two of the fault entities;

[0027] If the second similarity is greater than the second preset value, the two fault entities are fused to obtain a fused fault entity; the fused fault entity is used to construct the target triplet.

[0028] In some embodiments, the target triple is matched against a preset door fault knowledge base to obtain a matching triple, including:

[0029] Using the attributes of the target triple, a matching is performed in a preset door fault knowledge base, and the overlap between the attributes of the target triple and the attributes of each historical triple is calculated.

[0030] Based on the overlap, a hit triple is determined; the overlap between the attributes of the hit triple and the attributes of the target triple is greater than a third preset value.

[0031] In some embodiments, the method further includes:

[0032] If the target triplet does not exist in the preset door fault knowledge base, the target triplet is added to the preset door fault knowledge base.

[0033] In some embodiments, target alarm information for the aircraft door is generated based on the relationship between the hit triples, including:

[0034] Based on the relationship corresponding to the hit triples, a query is performed in a preset door fault knowledge base to obtain the target entity; the target entity is associated with the entity in the hit triples.

[0035] Based on the relationship corresponding to the target entity, a query is performed in the preset door fault knowledge base to obtain the target alarm information.

[0036] According to a second aspect of this application, an aircraft door status monitoring device is provided, comprising:

[0037] The acquisition module is used to acquire target pressure change data of the aircraft hatch;

[0038] An extraction module is used to obtain target triples based on the target pressure transformer data; wherein the target triples include: entity, relation, and attribute;

[0039] The matching module is used to match the target triplet in a preset door fault knowledge base to obtain a matching hit triplet; wherein, the preset door fault knowledge base includes multiple historical triplets; the historical triplets are obtained based on the historical pressure transformer data of the aircraft door; the historical triplets include the hit triplet.

[0040] The generation module is used to generate target alarm information for the aircraft door based on the relationship between the hit triples.

[0041] The apparatus of this application may be divided into more or fewer modules, or may adopt other module layouts, without limitation. For example, the acquisition module, extraction module, matching module, and generation module may be integrated into a processing module (or processing unit).

[0042] According to a third aspect of this application, an electronic device is provided, the electronic device including a processor and a memory for storing processor-executable instructions, wherein the processor executes the instructions to implement the steps of the aircraft door status monitoring method described in any of the above embodiments.

[0043] According to a fourth aspect of this application, a storage medium is provided that stores computer instructions thereon, which, when executed by a processor, implement the steps of the aircraft door status monitoring method described in any of the above embodiments.

[0044] According to a fifth aspect of this application, a computer program product is provided, comprising a computer program / instructions that, when executed by a processor, implement the steps of the aircraft door status monitoring method described in any of the above embodiments.

[0045] The technical solution of this application can achieve the following beneficial effects: This application provides a method for monitoring the status of aircraft doors. A preset door fault knowledge base stores historical triples obtained from a large amount of historical fault data. The target triples are matched against the preset door fault knowledge base to obtain matching hit triples. Target alarm information is generated based on the hit triples. It does not require the establishment of a complex status monitoring model, thus reducing computational complexity. Since the target alarm information is also based on historical triples, it can contain comprehensive and rich potential fault information, realizing all-round monitoring and early warning of aircraft door status, and improving aircraft safety.

[0046] Other features and advantages of this application will be described in detail in the following detailed description section. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] To gain a more complete understanding of this application and its beneficial effects, the following description will be provided in conjunction with the accompanying drawings, wherein the same reference numerals in the following description denote the same parts.

[0049] Figure 1 A flowchart illustrating a method for monitoring the status of an aircraft door, provided as an embodiment of this application;

[0050] Figure 2 A schematic diagram of the structure of a door failure knowledge system provided in this application embodiment;

[0051] Figure 3 This is a schematic diagram of the structure of a data processing system provided in an embodiment of this application;

[0052] Figure 4 A flowchart illustrating a method for monitoring the status of an aircraft door, provided as an embodiment of this application;

[0053] Figure 5 A flowchart illustrating a method for monitoring the status of an aircraft door, provided as an embodiment of this application;

[0054] Figure 6 A schematic diagram of a knowledge extraction process provided for an embodiment of this application;

[0055] Figure 7 This is a schematic diagram of the structure of an aircraft door status monitoring device provided in an embodiment of this application. Detailed Implementation

[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0057] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0058] "A and / or B" includes the following three combinations: A only, B only, and a combination of A and B.

[0059] The use of "applies to" or "configured to" in this application implies open and inclusive language, which does not exclude the applicability to or configuration to devices performing additional tasks or steps. Additionally, the use of "based on" implies openness and inclusivity, because processes, steps, calculations, or other actions "based on" one or more of the stated conditions or values ​​may in practice be based on additional conditions or values ​​beyond those stated.

[0060] In this application, the term "exemplary" is used to mean "used as an example, illustration, or description." Any embodiment described as "exemplary" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes are not described in detail to avoid obscuring the description of this application with unnecessary detail. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0061] In related technologies, model-based methods require establishing an accurate mathematical model of the monitored location (door), and then comparing the model with actual data to determine the status of the monitored location. The accuracy of status monitoring using model-based methods is strongly correlated with the accuracy of the model; however, due to the complexity of aircraft systems, models are often difficult to establish and computationally complex. Knowledge-based methods utilize expert experience and knowledge to determine the status of the monitored location, and then achieve automated fault diagnosis and status monitoring by simulating the expert reasoning process. Knowledge-based methods rely on accurate expert knowledge for judgment, requiring continuous maintenance and updating of the knowledge base, and involve a high degree of human intervention. Data-driven methods directly extract fault features from monitoring data, using machine learning algorithms or deep learning models to process and analyze the data to determine whether a fault exists, thus achieving status monitoring. Data-driven methods require a large amount of data to train the deep learning model in the early stages, placing high demands on data acquisition. Furthermore, for systems with many variables in the monitoring process, the complexity of the model and the operating cost increase accordingly.

[0062] On one hand, embodiments of this application provide a method for monitoring the status of an aircraft hatch, which can be executed by the aircraft or related components (such as chips). Figure 1 As shown, the aircraft hatch status monitoring method includes the following steps:

[0063] S101: Acquire target pressure change data of the aircraft hatch.

[0064] S102: Based on the target pressure change data, obtain the target triplet; where the target triplet includes: entity, relation, and attribute.

[0065] S103: Match the target triplet in a preset door fault knowledge base to obtain a matching hit triplet; wherein, the preset door fault knowledge base includes multiple historical triplets; the historical triplets are obtained based on the historical pressure change data of the aircraft door; the historical triplets include the hit triplet.

[0066] S104: Generate target alarm information for the aircraft door based on the relationship of the hit triplet.

[0067] The following is a detailed description of a method for monitoring the status of an aircraft cabin door provided in an embodiment of this application.

[0068] S101: Acquire target pressure change data of the aircraft hatch.

[0069] In some embodiments, the aircraft includes an airplane.

[0070] In some embodiments, obtaining the target pressure change data of the aircraft door specifically includes: obtaining the target pressure change data of the aircraft door at the current moment.

[0071] In some embodiments, acquiring target pressure change data of the aircraft door specifically includes:

[0072] S1: Acquire the initial pressure change data of the aircraft hatch;

[0073] S2: Perform data alignment and standardization on the initial voltage transformer data to obtain the first voltage transformer data;

[0074] S3: Perform change point detection on the first voltage transformer data to determine the time nodes of state change in the first voltage transformer data;

[0075] S4: Construct a window based on the state change time nodes;

[0076] S5: Compress the first transformer data outside the window to obtain the target transformer data.

[0077] In some embodiments, pressure transducers deployed on the hatch can be used to collect initial pressure transducer data. This initial pressure transducer data is pressure data arranged in chronological order, reflecting the stress state of the hatch. The stress state of the hatch reflects its sealing performance, deformation, and overall health. A pressure transducer is a device capable of sensing pressure and converting it into an electrical signal output, specifically including a resistive strain gauge. The initial pressure transducer data includes multiple pressure data points arranged according to their acquisition time. Each pressure data point has its corresponding acquisition time, also known as a timestamp.

[0078] In some embodiments, the number of pressure transformer sensors is set to x, where x ≥ 2. The initial pressure transformer data collected by the x sensors are aligned according to their timestamps, and then standardized to eliminate standard deviation, resulting in the first pressure transformer data. It can be understood that the first pressure transformer data includes multiple pressure data points arranged according to their data acquisition time, with each pressure data point having its corresponding timestamp.

[0079] In some embodiments, the cumulative sum (CUSUM) algorithm is used to detect state change points (change point detection) for each first pressure data point. A threshold θ and an offset parameter k are set. For each pressure data point, the deviation is calculated recursively, and the accumulated deviation is compared with the threshold. When the accumulated deviation is greater than the threshold, this timestamp is the state change time node (change point). The specific CUSUM algorithm flow is as follows:

[0080] When t=1, the first voltage transformer data is initialized:

[0081] The mean value of the first voltage transformer data μ t =data t =data1, positive deviation negative bias t represents the timestamp sequence number, data t This represents the t-th pressure data point. The first pressure data point consists of multiple data points. t A data sequence formed by arranging data in ascending order of timestamps.

[0082] When t>1, μ is calculated recursively. t and standard deviation σ t :

[0083]

[0084]

[0085] The bias was calculated using the Z-standardized form of the pressure data as follows:

[0086]

[0087] in, This represents the positive deviation corresponding to t. This represents the reverse deviation corresponding to t. This represents the positive deviation corresponding to t-1. This represents the reverse deviation corresponding to t-1.

[0088] After each calculation, it is compared with the threshold θ to determine the time point of state change: if or Then the t-th timestamp is the state change time node, and the state change time node is recorded as p. n And reset the deviation, i.e.

[0089] Suppose that in the a1th first voltage transformer data point, the b1th timestamp is the state change time node; in the a2th first voltage transformer data point, the b2th timestamp is the state change time node; and in the a3th first voltage transformer data point, the b3rd timestamp is the state change time node. Then, to fully reflect the information implied by the state change time nodes, for each first voltage transformer data point, the b1th, b2th, and b3th timestamps are all state change time nodes.

[0090] In some embodiments, at the state change time node p n Next, n represents the sequence number of the state change time node. The window length is set to L, with state change time node p... n Taking the center, take the following directions to the left and right respectively. The length of the window is obtained. A window is a range. Reserve the window. The internal pressure data is not compressed.

[0091] For pressure data outside the window, time-series compression techniques are used to compress the data and condense the effective information. Run-length encoding (RLE) is a common data compression algorithm. RLE averages the data between two windows before saving it. The specific steps are as follows:

[0092] Calculate the length S from the end of the previous window to the front of the current window. The previous window is represented as... The current window is represented as Then S is represented as:

[0093]

[0094] The interval outside the window is denoted as non-window, and S represents the length of the non-window interval. For non-window stress data, data compression is performed by averaging, and the calculation formula is as follows:

[0095]

[0096] N represents the compressed pressure data. The non-window pressure data is denoted as (S,N), meaning there are S pressure data with a value of N, thus completing the compression of redundant data (non-window pressure data).

[0097] Finally, the pressure data before compression was replaced with the compressed pressure data to obtain the target pressure change data.

[0098] S102: Based on the target pressure change data, obtain the target triplet; where the target triplet includes: entity, relation, and attribute.

[0099] In some embodiments, the target triple is an "entity-attribute-relationship" triple.

[0100] In some embodiments, the entity includes a fault ontology and a data ontology. The fault ontology may include fault analysis, and may also include fault location and fault mode. The data ontology is the target transformer data.

[0101] In some embodiments, the relationships include: relationships between fault entities, relationships between data entities, and relationships between fault entities and data entities.

[0102] In some embodiments, the attributes include: attributes of the fault ontology and attributes of the data ontology.

[0103] In some embodiments, the target triplet is obtained based on the target transformer data, specifically including:

[0104] S1: Extract the fault entity from the hatch text data, determine the relationship between the fault entities and the attributes of the fault entities;

[0105] S2: Extract the data ontology from the target pressure transformer data, determine the attributes of the data ontology and the relationships between the data ontology; the attributes of the data ontology are obtained based on the data change characteristics of the data ontology; the relationships between the data ontology include proportional relationships and sequential relationships; proportional relationships are obtained based on the correlation coefficient of the data ontology; sequential relationships are obtained based on the timestamp of the data ontology.

[0106] S3: Input the data ontology and the fault ontology into the neural network model to obtain the relationship between the fault ontology and the data ontology;

[0107] S4: Construct the target triple based on the data ontology, the fault ontology, the attributes of the data ontology, the attributes of the fault ontology, the relationships between the fault ontology, the relationships between the data ontology, and the relationships between the fault ontology and the data ontology.

[0108] In some embodiments, the hatch text data includes: hatch design report (i.e., the hatch status under normal conditions, including the hatch's mechanical parameters, dimensional data, materials, load data, etc.) and hatch historical failure report (including the hatch's historical failure status).

[0109] In some embodiments, historical pressure transformer data of the aircraft hatch is acquired. This historical data refers to pressure transformer data collected at historical moments that exhibit faults. The structure of the historical data is identical to that of the target data. Based on this historical data, and combined with expert knowledge, design reports, etc., concepts related to hatch faults are summarized, and relationships between these summarized concepts are established to construct a hatch fault knowledge system, forming a hatch fault knowledge structure. (See also...) Figure 2 As shown, the knowledge system for door failures decomposes door failure analysis into: failure location, failure analysis, and failure mode. Failure location refers to the door component where the failure occurs, such as the stop block, hinge, fastener, etc. Failure analysis includes failure type, failure name, failure description, failure cause, and solution. Failure mode includes the performance of failure data (target transformer data with failure) when the failure occurs, which is the data characteristic.

[0110] Based on the knowledge system of hatch failure, entities, relationships, and attributes in hatch status monitoring are defined, as shown in Table 1. Table 1 represents the construction rules of "entity-attribute-relationship" triples. Based on Table 1, knowledge extraction can be completed subsequently through a joint extraction model, that is, constructing "entity-attribute-relationship" triples (target triples).

[0111] Table 1

[0112]

[0113]

[0114] As can be seen from Table 1, the entities include: the fault entity and the data entity.

[0115] Relationships include: relationships between fault entities, relationships between data entities, and relationships between fault entities and data entities.

[0116] The attributes include: attributes of the fault ontology and attributes of the data ontology.

[0117] The attributes of the fault entity include: fault type, fault name, fault description, fault cause, and solution.

[0118] The attributes of the data ontology include: timestamp, change status, number of changes, and whether it has been restored.

[0119] The relationships between fault entities include: inheritance relationships, class relationships, and action class relationships.

[0120] The relationship between the fault ontology and the data ontology includes: complete correspondence and partial correspondence.

[0121] The relationships between data ontology include: proportional relationships and sequential relationships.

[0122] Inheritance classes include: attribute relationships, parent-child inheritance relationships, and whole-part relationships.

[0123] The class includes: contains.

[0124] Action categories include: causing, resolving, and expressing.

[0125] In some embodiments, the target voltage transformer data is used as the data body, and the "change status" of the data body is determined by calculating the difference; the number of state change time nodes contained in the target voltage transformer data is counted as the "change quantity"; and the first and last values ​​of the target voltage transformer data are compared to determine "whether it has been restored".

[0126] Specifically, a recovery threshold can be set. The target pressure transformer data is a data sequence formed by arranging multiple pressure data in ascending order of timestamps. The first and last data in the data sequence are extracted, and the absolute value of the difference between the two is calculated. If the absolute value is less than or equal to the recovery threshold, it is determined that "recovery" has been achieved. If the absolute value is greater than the recovery threshold, it is determined that "no recovery has been achieved".

[0127] Based on the above embodiments, the attributes of the data ontology can be obtained.

[0128] In some embodiments, sequential class relationships are obtained by sorting timestamps.

[0129] In some embodiments, the proportional relationship is obtained based on the correlation coefficient of the data ontology; the correlation coefficient is specifically the Pearson coefficient.

[0130] In some embodiments, the Pearson coefficient between every two target pressure transformer data points can be calculated in the following manner.

[0131] Set two different target pressure transformer data points, A1 and B1, with equal lengths (i.e., containing the same number of pressure data points). Calculate the mean value μ of A1. A :

[0132]

[0133] Where n1 represents the number of pressure data in A1, A i This represents the i-th pressure data in A1.

[0134] Calculate the mean μ of B1 B :

[0135]

[0136] Among them, B i This represents the i-th pressure data in B1.

[0137] The Pearson coefficient rho(A1,B1) between A1 and B1 is calculated as follows:

[0138]

[0139] The Pearson coefficient ranges from -1 to +1. -1 indicates a perfect negative correlation between the two data sequences, while +1 indicates a perfect positive correlation. The Pearson coefficient is used to represent a "proportional relationship".

[0140] Based on the above embodiments, the relationships between data ontologies can be obtained.

[0141] In some embodiments, firstly, hatch text data is collected using a character embedding layer, and the hatch states in the hatch text data are labeled according to the definition in Table 1. Then, deep contextual features of the labeled data are obtained through a long short-term memory (LSTM) network layer, and a fault ontology is output using a conditional random field (CRF). The BERT model is used to extract the relationships between the fault ontology. Algorithms such as Word2Vector are used for pre-training as initial values. The fault ontology is input into the BERT model, and the output is a vector sequence of each character / word fused with the semantic information of the entire text. Subsequently, the vector sequence output by the BERT model is input into a bidirectional long short-term memory neural network to obtain the last hidden layer state in each direction, and relationships are extracted to obtain the relationships between the fault ontology.

[0142] In some embodiments, the attributes of the faulty entity can be extracted using a gated recurrent unit (GRU).

[0143] Based on the above embodiments, the fault entity, the attributes of the fault entity, and the relationship between the fault entities can be obtained.

[0144] In some embodiments, the fault ontology and data ontology can be input into a trained neural network model to extract the "relationship between the fault ontology and the data ontology". The neural network model includes: a recurrent neural network model and a joint extraction model.

[0145] In some embodiments, the method further includes:

[0146] S1: Calculate the first similarity between every two data ontologies;

[0147] S2: If the first similarity is greater than the first preset value, merge the two data ontology to obtain the merged data ontology; the merged data ontology is used to construct the target triplet.

[0148] S3: Calculate the second similarity between every two faulty entities;

[0149] S4: If the second similarity is greater than the second preset value, merge the two fault entities to obtain the merged fault entity; the merged fault entity is used to construct the target triplet.

[0150] In some embodiments, for a data ontology, a cosine similarity between two data ontologs can be calculated as a first similarity.

[0151] Let two different data bodies be A2 and B2, with a first preset value of α1 and a fifth preset value of α2, where α1 > α2. Cosine Similarity(A2, B2) is the cosine similarity between A2 and B2. When A2 and B2 are deemed to have high similarity, it is considered that there are duplicate, conflicting, or inconsistent concepts, and data fusion is required: data ontology with the same concept or attribute is merged and the name is unified; conflicting concepts are selectively retained to eliminate ambiguity; and inconsistent concepts are identified and then merged or deleted based on the results.

[0152] when At that time, manual annotation and judgment are performed to identify conflicts or inconsistencies and then merge them.

[0153] In some embodiments, for a faulty entity, a cosine similarity between two faulty entities can be calculated as a second similarity.

[0154] Let two distinct fault entities be A3 and B3, a second preset value be α3, and a fourth preset value be α4, where α3 > α4. Cosine Similarity(A3, B3) represents the cosine similarity between A3 and B3. When A3 and B3 are deemed to have a high degree of similarity, it is considered that there are duplicate, conflicting, or inconsistent concepts, and data fusion is required: fused fault ontology with the same concept or the same attribute, unified name; selectively retain conflicting concepts to eliminate ambiguity; and identify inconsistent concepts and then choose to fuse or delete them based on the results.

[0155] when At that time, manual annotation and judgment are performed to identify conflicts or inconsistencies and then merge them.

[0156] Replace the two original data ontologies with the merged data ontology. Replace the two original fault ontologies with the merged fault ontology. Construct the target triple using the merged data ontology.

[0157] Data fusion reduces the total amount of data, improves the computational efficiency of subsequent data matching, eliminates redundant data, and enhances data quality. The fused fault ontology is then used to construct the target triplet.

[0158] Based on the above embodiments, an "entity-attribute-relationship" triple was constructed as the target triple.

[0159] S103: Match the target triplet in a preset door fault knowledge base to obtain a matching hit triplet; wherein, the preset door fault knowledge base includes multiple historical triplets; the historical triplets are obtained based on the historical pressure change data of the aircraft door; the historical triplets include the hit triplet.

[0160] In some embodiments, a knowledge base for door failures is pre-built based on historical pressure change data of the aircraft door.

[0161] Specifically, historical pressure-changing data of the aircraft hatch is acquired. This historical pressure-changing data refers to pressure-changing data collected at historical moments that exhibit faults. The structure of the historical pressure-changing data is the same as that of the target pressure-changing data. Historical triples are obtained from the historical pressure-changing data, and a hatch fault knowledge base is constructed using these historical triples. For a detailed implementation of obtaining historical triples from historical pressure-changing data, please refer to the detailed implementation of obtaining target triples from target pressure-changing data; it will not be elaborated here. The structure of the historical triples is the same as that of the target triples, and the historical triples also include: entity, relation, and attribute, forming an "entity-attribute-relationship" triple.

[0162] Both historical triples and target triples are forms of knowledge graphs. Knowledge graphs can visualize information and provide convenient query and matching functions, providing basic knowledge for subsequent fault diagnosis.

[0163] In some embodiments, the preset door fault knowledge base is a Neoj4 graph database.

[0164] In some embodiments, the target triples are matched against a preset door fault knowledge base to obtain matching hit triples, specifically including:

[0165] S1: Using the attributes of the target triple, match it in the preset door fault knowledge base, and calculate the overlap between the attributes of the target triple and the attributes of each historical triple.

[0166] S2: Determine the hit triple based on the overlap; the overlap between the attributes of the hit triple and the attributes of the target triple is greater than the third preset value.

[0167] In some embodiments, attributes of the data ontology are extracted from the target triple, and the Cypher statement is used to query and match them sequentially in a preset door fault knowledge base. A third preset value is set to 80%. If the overlap between the attributes of the data ontology of a historical triple and the attributes of the data ontology extracted from the target triple is greater than the third preset value (80%), then the historical triple is determined to be a hit triple. The number of hit triples can be 0, 1 or more.

[0168] In some embodiments, if the target triplet is not found in the preset door fault knowledge base, the target triplet is added to the preset door fault knowledge base.

[0169] In some embodiments, if the number of matched triples is zero, meaning there is no data body with a data body whose attributes overlap with the target triple's data body attributes with a third preset value, it is considered that no matched triple exists in the preset door fault knowledge base, and the match fails. In this case, the target triple is added to the preset door fault knowledge base to expand and update it. At this point, a "match failed" alarm message can be generated and fed back to the aircraft's onboard alarm system.

[0170] In some embodiments, if there is no matching triplet in the preset door fault knowledge base, the probability that the target triplet belongs to fault data can be further analyzed by a neural network model. If the probability that the target triplet belongs to fault data exceeds a sixth preset value, the target triplet is added to the preset door fault knowledge base.

[0171] S104: Generate target alarm information for the aircraft door based on the relationship of the hit triplet.

[0172] In some embodiments, target alarm information for the aircraft door is generated based on the relationship between the hit triples, specifically including:

[0173] S1: Based on the relationship corresponding to the hit triples, query the preset door fault knowledge base to obtain the target entity; the target entity is associated with the entity in the hit triples; the entity in the hit triples refers to the "data ontology";

[0174] S2: Based on the relationship corresponding to the target entity, query the preset door fault knowledge base to obtain the target alarm information.

[0175] Specifically, the data ontology in the hit triplet is identified as D. Further, a query is performed in the pre-defined hatch fault knowledge base to find a fault ontology that has a "complete correspondence" with data ontology D, denoted as G1. Then, the attributes corresponding to fault ontology G1 are queried, and fault ontology G1 and its corresponding attributes are used as the first alarm information.

[0176] Assuming that data ontology D does not have a fault ontology G1 with a "complete correspondence" relationship, then query out the fault ontology with a "partial correspondence" relationship with data ontology D, denoted as G2, and use fault ontology G2 as the first alarm information, or use fault ontology G2 and the attributes corresponding to fault ontology G2 as the first alarm information.

[0177] Both the fault ontology G1, which has a "complete correspondence" with the data ontology D, and the fault ontology G2, which has a "partial correspondence" with the data ontology D, are "target entities".

[0178] If a fault entity G1 exists, a query is performed in the preset door fault knowledge base to find fault entities that have a "class" relationship, an "action" relationship, and an "inheritance" relationship with fault entity G1, denoted as Gs. The attributes of fault entity Gs are then obtained, and the attributes of fault entity Gs are used as auxiliary information. Fault entity Gs and its attributes are used as second alarm information.

[0179] In the absence of a faulty entity G1, query the faulty entity that has a "class" relationship, an "action" relationship, and an "inheritance" relationship with the faulty entity G2, and denote it as Gs. Further obtain the attributes of the faulty entity Gs, use the attributes of the faulty entity Gs as auxiliary information, and use the faulty entity Gs and its attributes as the second alarm information.

[0180] A query is performed in the preset door fault knowledge base to find the data ontology that has a "sequential" relationship and a "proportional" relationship with the data ontology D, denoted as Ds. The attribute information of the data ontology Ds is then queried, and the data ontology Ds and the data ontology Ds are used as the second alarm information.

[0181] The target alarm information is obtained by combining the first alarm information and the second alarm information.

[0182] In some embodiments, the first alarm message has a higher feedback priority and should be fed back first; the second alarm message has a lower feedback priority and should be fed back after the first alarm message is sent. For example, if the aircraft is at high altitude and signal transmission is blocked, the amount of data that can be received in a short time is limited, so the first alarm message is sent to the aircraft's onboard alarm system first, and the second alarm message is sent to the aircraft's onboard alarm system after the first alarm message has been sent.

[0183] In some embodiments, an emergency tag can be added to some data in the target alarm information to indicate that the data is of a high degree of urgency. The fault types involved in the data that need to be tagged with emergency include: mechanical failure, human error, mechanical jamming, mechanical deformation, etc.; the fault names involved include: hatch seal failure, locking mechanism failure, minor deformation of the door frame structure, wear of the stop block, etc.

[0184] Based on the above embodiments, the preset door fault knowledge base stores historical triples obtained from a large amount of historical fault data. The target triple is matched against the preset door fault knowledge base to obtain a matching hit triple. Target alarm information is generated based on the hit triple. There is no need to build a complex state monitoring model, which reduces computational complexity. Since the target alarm information is also based on historical triples, it can contain comprehensive and rich potential fault information, realizing all-round monitoring and early warning of the aircraft door status and improving the safety of the aircraft.

[0185] In some embodiments, see Figure 3 As shown, the aircraft is specifically an airplane, which has an onboard section including: a cabin door (equipped with a pressure transducer sensor), an onboard alarm system, and a data acquisition and processing subsystem. The pressure transducer sensor, in conjunction with the data acquisition and processing subsystem, is used to collect the initial pressure transducer data from the cabin door. The 5G base station provides a 5G ATG (air-to-ground) air-to-ground communication link. Initial pressure-changing data is transmitted to the ground analysis system via this link. The ground analysis system is used to: acquire the initial pressure-changing data of the aircraft door; perform data alignment and standardization on the initial pressure-changing data to obtain the first pressure-changing data; perform change point detection on the first pressure-changing data to determine the state change time nodes; construct a window based on the state change time nodes; compress the first pressure-changing data outside the window to obtain the target pressure-changing data; obtain the target triplet based on the target pressure-changing data; the target triplet includes an entity, a relation, and an attribute; match the target triplet against a pre-set door fault knowledge base to obtain the matching hit triplet; and generate target alarm information for the aircraft door based on the relation corresponding to the hit triplet. The ground-based status monitoring system is used to: send the target alarm information to the onboard alarm system according to its priority, allowing the flight crew to receive the target alarm information and take corresponding control measures.

[0186] In some embodiments, see Figure 4 As shown, a method for monitoring the status of an aircraft cabin door specifically includes the following steps S1 to S5.

[0187] S1 data acquisition.

[0188] S1 data acquisition specifically includes: acquiring the initial pressure change data of the aircraft cabin door.

[0189] S2 data transmission.

[0190] S2 data transmission specifically includes transmitting initial voltage transformer data to the ground analysis system via a 5G ATG air-to-ground communication link.

[0191] S3 data preprocessing.

[0192] S3 data preprocessing specifically includes: acquiring initial pressure-transformation data of the aircraft door; performing data alignment and standardization on the initial pressure-transformation data to obtain first pressure-transformation data; performing change point detection on the first pressure-transformation data to determine the state change time nodes in the first pressure-transformation data; constructing a window based on the state change time nodes; and compressing the first pressure-transformation data outside the window to obtain target pressure-transformation data.

[0193] S4 builds a fault knowledge base.

[0194] The S4 fault knowledge base includes the following components.

[0195] S41: Based on historical transformer data, combined with expert knowledge, design reports, etc., the concepts in the field of hatch failure are summarized, the relationships between the summarized concepts are established, a knowledge system of hatch failure is constructed, a hierarchical knowledge classification system is formed, a knowledge structure of hatch failure is formed, and the knowledge ontology of hatch failure is structured.

[0196] S42: Based on the fault detection knowledge system and combined with expert knowledge, design a set of knowledge extraction rules for target transformer data and historical transformer data, and determine the information identification template and matching rules in accordance with the requirements of clarity, completeness and accuracy.

[0197] S43: Perform fault knowledge fusion and deambiguity elimination on historical pressure transformer data of the cabin door, check for inconsistencies, conflicts, and duplications in information, match entities and merge information to improve the conceptual system of aircraft cabin door fault knowledge.

[0198] S44: Based on the knowledge extraction rule set and fault detection knowledge system, a joint extraction model is used to extract knowledge from historical transformer data, resulting in structured "entity-relationship-attribute" triples, which are the historical triples. Similarly, the joint extraction model is used to extract knowledge from target transformer data, resulting in structured "entity-relationship-attribute" triples, which are the target triples. The joint extraction model includes rule-based entity and attribute extraction, and pattern matching and semi-supervised relation extraction. Entity and attribute extraction primarily relies on extraction rules defined by expert knowledge; pattern matching is used to extract relationships between "faults" and between "data," while semi-supervised methods are used to extract relationships between "faults and data." "Fault" refers to a fault entity, and "data" refers to a data entity. After knowledge extraction, concept fusion is performed on the historical triples. Concept fusion includes: using a semi-supervised entity alignment model based on entity relation clustering to achieve the fusion of entities and entity relations; using the similarity calculation of parallel entity relations to filter out different fault entity names pointing to the same attribute; merging the above-labeled entity names pointing to the same fault through entity relation clustering; and finally using semi-supervised learning entity alignment to fuse and align the merged entity relations, thereby achieving fault knowledge fusion for each triple.

[0199] S45: Uses Neo4j to store historical triples to establish a complete knowledge base for door failures.

[0200] S5 door status monitoring.

[0201] The S5 door status monitoring specifically includes: inputting the target triplet into the door fault knowledge base for querying, monitoring the aircraft door status in real time, and determining whether the door has malfunctioned or has potential malfunctions; if there is a major malfunction that may affect flight, it is transmitted to the onboard alarm system via the 5G ATG air-to-ground transmission link to notify the crew.

[0202] Based on the above embodiments, the data processing flow is mainly carried out in the ground analysis system and the status monitoring system. Air-to-ground transmission via 5G ATG reduces the onboard computational load. Fault analysis and judgment performed in the ground analysis system are more accurate, and fault information (target alarm information) is only transmitted to the onboard alarm system when necessary, ensuring flight safety without excessively consuming communication resources. Real-time comparison and analysis of door status and faults are also performed. The constructed door fault knowledge base helps achieve full-cycle management of door health, and can be continuously expanded and updated, contributing to long-term maintenance efficiency improvements.

[0203] In some embodiments, see Figure 5 As shown, a method for monitoring the status of an aircraft cabin door specifically includes the following steps 201 to 205.

[0204] Step 201: Preprocess the hatch strain data (initial pressure transformer data) by synchronizing timestamps to remove noise, filter outliers, and remove redundant information.

[0205] Step 202: Construct the ontology using the skeleton method, combine expert knowledge, historical fault database design reports, etc. to summarize concepts, define and evaluate the domain ontology.

[0206] Step 203: Based on the inductive concept system and the defined knowledge extraction rules, obtain the triples of entity, relation, and attribute through the joint extraction model.

[0207] Step 204: Store entities and relationships in the Neoj4 graph database system to establish a complete knowledge graph of hatch states.

[0208] Step 205: Query the target triplet collected in real time on the aircraft in the door fault knowledge graph to determine whether the door is faulty and related information.

[0209] In some embodiments, see Figure 6 As shown, Figure 6 This document outlines the specific process for knowledge extraction using a joint extraction model. After S3 data preprocessing, attributes are extracted through statistical and numerical processing, and relationships between data are extracted through numerical processing. Combining design documents, expert knowledge, and operational reports, attributes are extracted using an artificial neural network, and relationships between faults are extracted using a BERT model. Finally, a neural network model is used to extract the relationships between faults and data.

[0210] On the other hand, see Figure 7 As shown, this application embodiment provides an aircraft door status monitoring device, including:

[0211] Acquisition module 701 is used to acquire target pressure change data of the aircraft hatch;

[0212] Extraction module 702 is used to obtain target triples based on target compression transformer data; wherein, target triples include: entity, relation and attribute;

[0213] The matching module 702 is used to match the target triplet in a preset door fault knowledge base to obtain the matched hit triplet; wherein, the preset door fault knowledge base includes multiple historical triplets; the historical triplets are obtained based on the historical pressure transformer data of the aircraft door; the historical triplets include the hit triplet.

[0214] The generation module 704 is used to generate target alarm information for the aircraft door based on the relationship between the hit triplet.

[0215] In some embodiments, the acquisition module 701 is specifically used to: acquire initial pressure-transformation data of the aircraft door; perform data alignment and standardization on the initial pressure-transformation data to obtain first pressure-transformation data; perform change point detection on the first pressure-transformation data to determine the state change time nodes in the first pressure-transformation data; construct a window based on the state change time nodes; and compress the first pressure-transformation data outside the window to obtain target pressure-transformation data.

[0216] In some embodiments, the extraction module 702 is specifically used for: extracting fault entities from target transformer data, determining the relationships between fault entities and their attributes; extracting data entities from target transformer data, determining the attributes of data entities and the relationships between them; the attributes of the data entities are obtained based on the data change characteristics of the data entities; the relationships between the data entities include proportional relationships and sequential relationships; proportional relationships are obtained based on the correlation coefficients of the data entities; sequential relationships are obtained based on the timestamps of the data entities; inputting the data entities and fault entities into a neural network model to obtain the relationships between the fault entities and the data entities; and constructing target triples based on the data entities, fault entities, attributes of the data entities, attributes of the fault entities, relationships between fault entities, relationships between data entities, and relationships between fault entities and data entities.

[0217] In some embodiments, the matching module 703 is specifically used to: use the attributes of the target triple to perform matching in a preset door fault knowledge base, calculate the overlap between the attributes of the target triple and the attributes of each historical triple, determine the hit triple based on the overlap, and the overlap between the attributes of the hit triple and the attributes of the target triple is greater than a third preset value.

[0218] In some embodiments, the generation module 704 is specifically used to: query a preset door fault knowledge base according to the relationship corresponding to the hit triples to obtain the target entity; associate the target entity with the entity in the hit triples; and query the preset door fault knowledge base according to the relationship corresponding to the target entity to obtain the target alarm information.

[0219] This application provides an electronic device, which includes a processor and a memory for storing processor-executable instructions. When the processor executes the instructions, it implements the steps described in any of the above-described aircraft door status monitoring methods.

[0220] This application provides a storage medium storing computer instructions, which, when executed by a processor, implement the steps described in any of the above-described aircraft door status monitoring methods.

[0221] In the embodiments of this application, the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0222] This application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps described in any of the above-described aircraft door status monitoring methods.

[0223] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0224] The above provides a detailed description of the aircraft door status monitoring method, device, and electronic equipment provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for monitoring the status of an aircraft hatch, characterized in that, include: Acquire target pressure change data for the aircraft hatch; Based on the target pressure transformer data, a target triple is obtained; wherein, the target triple includes: entity, relation, and attribute; The target triplet is matched against a pre-defined door fault knowledge base to obtain a matching hit triplet; wherein, the pre-defined door fault knowledge base includes multiple historical triplets; the historical triplets are obtained based on the historical pressure transformer data of the aircraft door; the historical triplets include the hit triplet. Based on the relationship between the hit triplet, target alarm information for the aircraft door is generated.

2. The method according to claim 1, characterized in that, Acquire target pressure change data for the aircraft hatch, including: Acquire the initial pressure-transformation data of the aircraft cabin door; The initial voltage transformer data is aligned and standardized to obtain the first voltage transformer data; Change point detection is performed on the first voltage transformer data to determine the time nodes of state change in the first voltage transformer data; A window is constructed based on the aforementioned state change time points; The first voltage transformer data outside the window is compressed to obtain the target voltage transformer data.

3. The method according to claim 1 or 2, characterized in that, The entities include: the fault ontology and the data ontology; The relationships include: the relationships between the fault entities, the relationships between the data entities, and the relationships between the fault entities and the data entities.

4. The method according to claim 3, characterized in that, Based on the target pressure transformer data, the target triplet is obtained, including: The fault entity is extracted from the door text data, and the relationships between the fault entities and the attributes of the fault entities are determined. The data ontology is extracted from the target pressure transformer data, and the attributes of the data ontology and the relationships between the data ontology are determined. The attributes of the data ontology are obtained based on the data change characteristics of the data ontology. The relationships between the data ontology include proportional relationships and sequential relationships. The proportional relationships are obtained based on the correlation coefficient of the data ontology. The sequential relationships are obtained based on the timestamp of the data ontology. The data ontology and the fault ontology are input into a neural network model to obtain the relationship between the fault ontology and the data ontology. Construct a target triple based on the data ontology, the fault ontology, the attributes of the data ontology, the attributes of the fault ontology, the relationships between the fault ontologs, the relationships between the data ontologs, and the relationships between the fault ontology and the data ontology.

5. The method according to claim 4, characterized in that, The method further includes: Calculate the first similarity between every two of the data ontologies; If the first similarity is greater than a first preset value, the two data ontologies are merged to obtain a merged data ontology; the merged data ontology is used to construct the target triplet. Calculate the second similarity between every two of the fault entities; If the second similarity is greater than the second preset value, the two fault entities are fused to obtain a fused fault entity; the fused fault entity is used to construct the target triplet.

6. The method according to claim 1, characterized in that, The target triple is matched against a pre-defined hatch fault knowledge base to obtain matching hit triples, including: Using the attributes of the target triple, a matching is performed in a preset door fault knowledge base, and the overlap between the attributes of the target triple and the attributes of each historical triple is calculated. Based on the overlap, a hit triple is determined; the overlap between the attributes of the hit triple and the attributes of the target triple is greater than a third preset value.

7. The method according to claim 6, characterized in that, The method further includes: If the target triplet does not exist in the preset door fault knowledge base, the target triplet is added to the preset door fault knowledge base.

8. The method according to claim 1, characterized in that, Based on the relationship between the hit triplet, target alarm information for the aircraft door is generated, including: Based on the relationship corresponding to the hit triples, a query is performed in a preset door fault knowledge base to obtain the target entity; the target entity is associated with the entity in the hit triples; Based on the relationship corresponding to the target entity, a query is performed in the preset door fault knowledge base to obtain the target alarm information.

9. A device for monitoring the status of an aircraft cabin door, characterized in that, include: The acquisition module is used to acquire target pressure change data of the aircraft hatch; An extraction module is used to obtain target triples based on the target pressure transformer data; wherein the target triples include: entity, relation, and attribute; The matching module is used to match the target triplet in a preset door fault knowledge base to obtain a matching hit triplet; wherein, the preset door fault knowledge base includes multiple historical triplets; the historical triplets are obtained based on the historical pressure transformer data of the aircraft door; the historical triplets include the hit triplet. The generation module is used to generate target alarm information for the aircraft door based on the relationship between the hit triples.

10. An electronic device, characterized in that, The electronic device includes a processor and a memory for storing processor-executable instructions, wherein the processor executes the instructions to implement the steps of the method according to any one of claims 1 to 8.

11. A storage medium, characterized in that, It stores computer instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 8.

12. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method as described in any one of claims 1 to 8.

Images