Knowledge model-driven avionics fault reasoning method, medium, and device

Abstract

Provided in the present invention are a knowledge model-driven avionics fault reasoning method, a medium, and a device. The method comprises: building an entity extraction model, constructing an entity extraction data set, and using the entity extraction data set to train the entity extraction model; building a semantic similarity calculation model, constructing a semantic similarity calculation data set, and using the semantic similarity calculation data set to train the semantic similarity calculation model; inputting a fault description text into the entity extraction model and extracting entity information; processing the extracted entity information to generate a new fault description text; and using the semantic similarity calculation model to perform similarity calculation on the fault description text and a plurality of standard fault modes in a corresponding faulty module so as to obtain a fault reasoning result. The present invention improves the maintenance success rate and efficiency of maintenance personnel, reduces the factory return rate of modules, improves the response efficiency of battlefield equipment, and has great significance and value.

Description

A knowledge model driven avionics fault reasoning method, medium and device Technical Field

[0001] The present invention relates to the technical field of avionics equipment, and in particular to a knowledge model-driven avionics fault reasoning method, medium, and device. Background Art

[0002] Avionics and information equipment exhibits the technical characteristics of deeply integrated design of multiple systems such as communication, navigation, identification, reconnaissance, and countermeasures. The experience-based troubleshooting and location rate of field support personnel is very low, while a large amount of textual fault description information is generated in the flight records of pilots and during the daily maintenance and repair of aircraft by support personnel. By mining the inherent connections of these textual information and inputting the textual description of the fault phenomenon, the system can intelligently match or recommend relevant standard fault modes, which can greatly improve the maintenance efficiency and success rate of maintenance personnel and reduce the return rate of modules.

[0003] However, there are currently only a large number of non-standard maintenance record texts in the aviation field, and the accuracy of traditional text matching and database fuzzy retrieval methods is difficult to meet the actual needs of fault location.

[0004] Summary of the Invention

[0005] The present invention aims to provide a knowledge model-driven avionics fault reasoning method, medium and device to solve the problem that the accuracy of traditional text matching and database fuzzy retrieval methods cannot meet the actual needs of fault location.

[0006] The present invention provides a knowledge model-driven avionics fault reasoning method, comprising the following steps:

[0007] S1: Build an entity extraction model, construct an entity extraction dataset, and use the entity extraction dataset to train the entity extraction model;

[0008] S2: Build a semantic similarity calculation model, construct a semantic similarity calculation dataset, and use the semantic similarity calculation dataset to train the semantic similarity calculation model;

[0009] S3, input the fault phenomenon description text into the entity extraction model to extract entity information;

[0010] S4, processing the extracted entity information to generate a new fault phenomenon description text;

[0011] S5, using a semantic similarity calculation model to perform similarity calculation on the fault phenomenon description text and a number of standard fault modes in the corresponding fault module to obtain a fault inference result.

[0012] Furthermore, step S1 includes:

[0013] Constructing an entity extraction dataset: Entities in text data are annotated. The entity extraction model is treated as a word-level annotation task, using the BIO sequence annotation rule, where "B" indicates the beginning of an entity, "I" indicates the middle of an entity, and "O" indicates a non-entity word. The input of the entity extraction model includes word vectors, position vectors, and segmentation vectors, and the output is the annotation label for each word.

[0014] Build an entity extraction model: The entity extraction model consists of an input layer, an encoding layer, a BERT layer, and a CRF layer. First, the encoding layer adds positional encoding before data preprocessing and sums it with the input vector. Then, the BERT layer extracts the text feature sequence. The CRF layer uses a log-linear model of conditional random fields (CRF) to represent the joint probability of the entire text feature sequence, thereby better predicting the annotation labels in the text feature sequence.

[0015] The entity extraction model is continuously optimized according to the total score of the predicted annotation label sequence until the number of training times reaches the set value.

[0016] Furthermore, assuming that the sentence length is n, the extracted text feature sequence is X = (x1, x2, ..., x n ), the corresponding predicted label sequence is Y=(y1,y2,...,y n ), the total score of the predicted annotation label sequence is:

[0017] Among them, T represents the transfer score between labels, and P represents the value of each word corresponding to y i The score of the label.

[0018] Furthermore, step S2 includes:

[0019] Constructing a semantic similarity calculation dataset: First, the input fault phenomenon description text and the standard fault mode in the standard fault library are combined into sentence pairs to construct a semantic similarity calculation dataset containing sentence pairs and labels;

[0020] Build a semantic similarity calculation model: Input sentence pairs into the Tokenization layer for word segmentation and concatenation, and then input the processed sentences into the Embedding layer for encoding. During the encoding process, due to the bidirectional Attention mechanism of the BERT model, the two sentences in each other's context will influence each other, ultimately learning the similarity between the two sentences. After calculation using the BERT model, the final encoding of the concatenated sentences is obtained in the output layer and input into the Dropout layer to prevent overfitting of the algorithm. After the Dropout layer, connect a fully connected layer with an output dimension of 2, and then use softmax to calculate the probabilities of similarity and dissimilarity.

[0021] The semantic similarity calculation model is continuously optimized according to the similarity score until the number of training times reaches the set value.

[0022] Furthermore, the similarity calculation is expressed as: P = softmax(cW)

[0023] Among them, P represents the similarity, c represents the final code, and W represents the weight of the fully connected layer.

[0024] Furthermore, in step S3, the entity information includes component units and fault phenomenon descriptions.

[0025] Furthermore, in step S4, processing the extracted entity information includes:

[0026] Disambiguate entity information and find standard names in the standard fault library; combine the extracted modules, keywords and related descriptive words into a new and more concise sentence to generate a new fault phenomenon description text.

[0027] Furthermore, step S5 includes:

[0028] According to the standard name of the component unit, the standard fault library of the component unit is found, and the semantic similarity calculation model is used to calculate the similarity between the fault phenomenon description text and several standard fault modes in the corresponding fault module. The standard fault mode with a similarity higher than the set value is taken as the fault inference result.

[0029] The present invention also provides a computer terminal storage medium storing computer terminal executable instructions, wherein the computer terminal executable instructions are used to execute the above-mentioned knowledge model-driven avionics fault reasoning method.

[0030] The present invention further provides a computing device, comprising:

[0031] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the above-mentioned knowledge model-driven avionics fault reasoning method.

[0032] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0033] The knowledge model-driven avionics fault reasoning method of the present invention can be applied in multiple fields. First, by extracting and analyzing text data, an entity extraction model is constructed and trained to extract key information from the text, such as component units and fault status. A semantic similarity calculation model is then constructed and trained to establish correlations with a standard fault database. During use, the fault phenomenon description text is input into the entity extraction model to extract entity information, such as component units and fault phenomenon descriptions. The extracted entity information data is then processed to generate a new fault phenomenon description text to meet the needs of subsequent algorithms. Finally, the semantic similarity calculation model is used to calculate similarity between the fault phenomenon description text and several fault modes in the corresponding fault module. Once the similarity score exceeds a certain set value, the result is used as the output of the fault reasoning system. By locating inaccurate and non-standard fault phenomenon description information to standard fault modes, it can provide assistance and support for fault diagnosis and maintenance decision-making for avionics equipment maintenance personnel. Therefore, the knowledge model-driven electronic equipment fault reasoning method is of great significance and value in improving maintenance personnel's repair success rate and efficiency, reducing module return rates, and improving the efficiency of battlefield equipment response. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings in the embodiments will be briefly introduced below. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0035] FIG1 is a flow chart of a knowledge model-driven avionics fault reasoning method according to an embodiment of the present invention.

[0036] FIG2 is a schematic diagram of an entity extraction model in an embodiment of the present invention.

[0037] FIG3 is a schematic diagram of a semantic similarity calculation model in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0039] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are intended to fall within the scope of protection of the present invention.

[0040] Example

[0041] As shown in FIG1 , this embodiment proposes a knowledge model-driven avionics fault reasoning method, including the following steps:

[0042] S1: Build an entity extraction model, construct an entity extraction dataset, and use the entity extraction dataset to train the entity extraction model; specifically, it includes:

[0043] Constructing an entity extraction dataset: Entities in text data are annotated. The entity extraction model is treated as a word-level annotation task, using the BIO sequence annotation rule, where "B" indicates the beginning of an entity, "I" indicates the middle of an entity, and "O" indicates a non-entity word. The input of the entity extraction model includes word vectors, position vectors, and segmentation vectors, and the output is the annotation label for each word.

[0044] Building an entity extraction model: As shown in Figure 2, the entity extraction model consists of an input layer, an encoding layer, a BERT layer, and a CRF layer. First, the encoding layer adds positional encoding before data preprocessing and sums it with the input vector. Then, the BERT layer extracts the text feature sequence. The CRF layer uses a log-linear model of conditional random fields (CRF) to represent the joint probability of the entire text feature sequence, thereby better predicting the annotation labels in the text feature sequence.

[0045] Assume that the sentence length is n, and the extracted text feature sequence is X=(x1,x2,...,x n ), the corresponding predicted label sequence is Y=(y1,y2,...,y n ), the total score of the predicted annotation label sequence is:

[0046] Among them, T represents the transfer score between labels, and P represents the value of each word corresponding to y i The score of the label;

[0047] The entity extraction model is continuously optimized according to the total score of the predicted annotation label sequence until the number of training times reaches the set value.

[0048] S2: Build a semantic similarity calculation model, construct a semantic similarity calculation dataset, and use the semantic similarity calculation dataset to train the semantic similarity calculation model; specifically, it includes:

[0049] Constructing a semantic similarity calculation dataset: First, the input fault phenomenon description text and the standard fault mode in the standard fault library are combined into sentence pairs to construct a semantic similarity calculation dataset containing sentence pairs and labels;

[0050] Building a semantic similarity calculation model: As shown in Figure 3, sentence pairs are input into the Tokenization layer for word segmentation and concatenation. The processed sentences are then input into the Embedding layer for encoding. During the encoding process, due to the bidirectional Attention mechanism of the BERT model, the two sentences in each other's context will influence each other, ultimately learning the degree of similarity between the two sentences. After calculation using the BERT model, the final encoding of the concatenated sentences is obtained at the output layer and input into the Dropout layer to prevent overfitting of the algorithm. A fully connected layer with an output dimension of 2 is connected after the Dropout layer, and softmax is used to calculate the probabilities of similarity and dissimilarity.

[0051] Similarity calculation is expressed as: P = softmax (cW)

[0052] Among them, P represents the similarity, c represents the final code, and W represents the weight of the fully connected layer;

[0053] The semantic similarity calculation model is continuously optimized according to the similarity score until the number of training times reaches the set value.

[0054] S3, inputting the fault phenomenon description text into the entity extraction model to extract entity information; the entity information includes component units and fault phenomenon description, etc.;

[0055] S4, processing the extracted entity information: disambiguating the entity information and finding the standard name in the standard fault database; combining the extracted modules, keywords, and related descriptive words into a new, more concise sentence, thus generating a new fault phenomenon description text;

[0056] S5, find the standard fault library of the component unit according to the standard name of the component unit, use the semantic similarity calculation model to calculate the similarity between the fault phenomenon description text and several standard fault modes in the corresponding fault module, and use the standard fault mode with a similarity higher than the set value as the fault inference result.

[0057] The above-described knowledge model-driven avionics fault reasoning method can be applied in multiple fields. First, by extracting and analyzing features from text data, an entity extraction model is constructed and trained to extract key information from the text, such as component units and fault states. A semantic similarity calculation model is then constructed and trained to establish correlations with a standard fault database. During use, the fault phenomenon description text is input into the entity extraction model to extract entity information, such as component units and fault phenomenon descriptions. The extracted entity information data is then processed to generate a new fault phenomenon description text to meet the needs of subsequent algorithms. Finally, the semantic similarity calculation model is used to calculate the similarity between the fault phenomenon description text and several fault modes in the corresponding fault module. Once the similarity score exceeds a certain set value, the result is used as the output of the fault reasoning system. By locating inaccurate and non-standard fault phenomenon descriptions to standard fault modes, it can provide assistance and support to avionics equipment maintenance personnel in fault diagnosis and maintenance decision-making. Therefore, the knowledge model-driven electronic equipment fault reasoning method is of great significance and value in improving maintenance personnel's repair success rate and efficiency, reducing module return rates, and improving the efficiency of battlefield equipment response.

[0058] Furthermore, in some embodiments, a computer terminal storage medium is provided, storing computer terminal executable instructions for executing the knowledge model-driven avionics fault reasoning method described in the preceding embodiments. Examples of computer storage media include magnetic storage media (e.g., floppy disks, hard disks, etc.), optical recording media (e.g., CD-ROMs, DVDs, etc.), or memory devices such as memory cards, ROM, or RAM. Computer storage media can also be distributed across networked computer systems, such as in an application store.

[0059] Furthermore, in some embodiments, a computing device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the knowledge model-driven avionics fault reasoning method described in the above embodiments. Examples of computing devices include PCs, tablet computers, smartphones, and PDAs.

[0060] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A knowledge model driven avionics fault reasoning method, characterized in that: The following steps are involved: S1, build an entity extraction model, construct an entity extraction dataset, and use the entity extraction dataset to train the entity extraction model; S2, building a semantic similarity calculation model, constructing a semantic similarity calculation dataset, and using the semantic similarity calculation dataset to train the semantic similarity calculation model; S3, input the fault phenomenon description text into the entity extraction model to extract entity information; S4, processing the extracted entity information to generate a new fault phenomenon description text; S5, using a semantic similarity calculation model to perform similarity calculation on the fault phenomenon description text and a number of standard fault modes in the corresponding fault module to obtain a fault inference result.

2. The knowledge model-driven avionics fault reasoning method according to claim 1, characterized in that: Step S1 includes: Constructing an entity extraction dataset: annotating the entities in the text data, treating the entity extraction model as a word-level annotation task, and using the BIO sequence annotation rule, where "B" indicates the beginning of the entity, "I" indicates the middle of the entity, and "O" indicates a non-entity word; the input of the entity extraction model includes word vectors, position vectors, and segmentation vectors, and the output is the annotation label for each word; Build an entity extraction model: The entity extraction model includes an input layer, an encoding layer, a BERT layer, and a CRF layer. First, the encoding layer adds position encoding before data preprocessing and sums it with the input vector. Then, the BERT layer extracts the text feature sequence, and the CRF layer uses the log-linear model of the conditional random field (CRF) to represent the joint probability of the entire text feature sequence, so as to better predict the annotation labels in the text feature sequence. The entity extraction model is continuously optimized according to the total score of the predicted annotation label sequence until the number of training times reaches the set value.

3. The knowledge model driven avionics fault reasoning method according to claim 2 is characterized in that, assuming that the sentence length is n, the extracted text feature sequence is X = (x1, x2, ..., x n ), the corresponding predicted label sequence is Y = (y1, y2, ..., y n ), the total score of the predicted annotation label sequence is: in, T represents the transfer score between labels, and P represents the correspondence between each word and y i The score of the label.

4. The knowledge model-driven avionics fault reasoning method according to claim 1, characterized in that: Step S2 includes: Constructing a semantic similarity calculation dataset: First, the input fault phenomenon description text and the standard fault mode in the standard fault library are combined into sentence pairs to construct a semantic similarity calculation dataset containing sentence pairs and labels; Build a semantic similarity calculation model: Input the sentence pairs into the Tokenization layer for word segmentation and concatenation, and then input the processed sentences into the Embedding layer for encoding. During the encoding process, due to the bidirectional Attention mechanism of the BERT model, the two sentences in each other's context will affect each other, and finally learn the similarity between the two sentences; after calculation through the Bert model, the final encoding of the concatenated sentence is obtained in the output layer, and the final encoding is input into the Dropout layer to suppress the overfitting of the algorithm; after the Dropout layer, connect a fully connected layer with an output dimension of 2, and then use softmax to calculate the probabilities of similarity and dissimilarity; The semantic similarity calculation model is continuously optimized according to the similarity score until the number of training times reaches the set value.

5. The knowledge model-driven avionics fault reasoning method according to claim 4, characterized in that: The similarity calculation is expressed as: P = softmax (cW) Among them, P represents the similarity, c represents the final encoding, and W represents the weight of the fully connected layer.

6. The knowledge model-driven avionics fault reasoning method according to claim 1, characterized in that: In step S3, the entity information includes component units and fault phenomenon description.

7. The knowledge model-driven avionics fault reasoning method according to claim 6, characterized in that: In step S4, processing the extracted entity information includes: Disambiguate entity information and find standard names in the standard fault library; combine the extracted modules, keywords and related descriptive words into a new and more concise sentence to generate a new fault phenomenon description text.

8. The knowledge model-driven avionics fault reasoning method according to claim 7, characterized in that: Step S5 includes: The standard fault library of the component unit is found according to the standard name of the component unit, and the similarity between the fault phenomenon description text and several standard fault modes in the corresponding fault module is calculated using the semantic similarity calculation model. The standard fault mode with a similarity higher than the set value is taken as the fault inference result.

9. A computer terminal storage medium storing computer terminal executable instructions, characterized in that: The computer terminal executable instructions are used to execute the knowledge model-driven avionics fault reasoning method as claimed in any one of claims 1 to 8.

10. A computing device, characterized in that: include: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the knowledge model-driven avionics fault reasoning method as described in any one of claims 1 to 8.

Images