CN-122022259-A - Airplane pilot stage troubleshooting personnel recommendation method based on crew capability image

CN122022259ACN 122022259 ACN122022259 ACN 122022259ACN-122022259-A

Abstract

The invention relates to the technical field of airplane fault diagnosis, in particular to an airplane test flight stage troubleshooting personnel recommendation method based on crew capability images, which comprises the following steps of S1, constructing maintenance capability images of different crew based on multi-dimensional personnel information; the method comprises the steps of S2, dividing personnel portrait contents into a semantic association layer and a capability assessment layer, respectively calculating scores of the personnel in two aspects, S3, vectorizing text data related to the semantic association layer, vectorizing index scores of the capability assessment layer, introducing a multi-head attention mechanism, splicing the text vectors and the quantitative assessment vectors to serve as the input of the multi-head attention mechanism, finally outputting weights of the semantic association layer and the capability assessment layer, multiplying the weights by the scores of the layers respectively, and finally obtaining a comprehensive score of the maintenance capability of the personnel, S4, arranging the personnel scores in descending order, and taking TOPN as the optimal troubleshooting recommended personnel.

Inventors

JIANG MENGFAN
WANG XI
HE YE
ZHANG HANTAO
YANG CHAODONG

Assignees

中国飞行试验研究院

Dates

Publication Date: 20260512
Application Date: 20251227

Claims (8)

1. An aircraft pilot stage troubleshooting personnel recommendation method based on crew capability images is characterized by comprising the following steps: constructing a crew capability portrait based on the multidimensional personnel information; calculating a staff capacity score based on the representation content; TOP N was used as a person recommendation based on the person ability score.
2. The method of claim 1, wherein prior to constructing the crew capability representation based on the multidimensional personnel information, further comprising: and acquiring historical fault information, crew basic information, scientific research information and skill information.
3. The method of claim 2, wherein the obtaining historical fault information and crew base information, scientific information and skill information comprises: extracting structured data through a system interface; Unstructured data, including fault tickets, are extracted using OCR recognition techniques.
4. A method according to claim 3, wherein the structured data comprises fault records of fault management modules in a tester technical state management system, employee basic information in personnel data, paper information in a scientific and technological achievement management library, patent information and technical report information.
5. The method of claim 4, wherein after the obtaining the historical fault information and the crew base information, the scientific information and the skill information, further comprising: And performing invalid and redundant data cleaning, unifying data formats, and integrating a plurality of source data by name disambiguation and unified fields.
6. The method of claim 5, wherein the constructing the crew capability representation comprises: Defining a portrait construction flow, determining portrait construction dimension and realizing portrait model representation.
7. The method of claim 6, wherein calculating a staff capacity score based on the portrait content includes: Dividing the portrait index into a semantic association layer and a capability assessment layer, wherein the semantic association layer comprises professional information in basic attributes, a subject of a scientific research result in scientific research attributes and troubleshooting experiences in skill attributes; calculating fault information vector using cosine similarity Similarity to each text vector: The ability evaluation layer comprises education years, working years, skill levels and scientific research output: normalization processing is carried out on education age Eu and work age Wu, and then weighted summation is carried out to form ScoreB, namely Wherein ax(E)、 in(E)、 ax(W)、 In (W) is the minimum and maximum of the educational years and working years, respectively, in the dataset; weighting calculation and normalization are carried out on the professional skill level Su, honor name Au and service model type Mu to obtain ScoreS; Specifically, skill levels are classified into primary, medium, high and expert grades, respectively, scores of 1-4 are given, grading is carried out according to influence and authority of honor and title numbers, national honor is given 5 points, provincial honor is given 4 points, municipal honor is given 3 points, enterprise honor is given 2 points, no honor and title number is given 1 point, the number of machine types served by users is counted, a linear scoring mode is adopted, the number of machine types served by users is assumed to be n, each machine type is set to be added with 1 point, and basic score is 1 point Making ScoreS in the interval of [0,1 ]; normalizing the paper quantity Pu and the patent quantity Nu to obtain ScoreR; and carrying out weighted summation on the three scores to obtain a final quantized attribute comprehensive score: 。
8. The method of claim 7, wherein the computing a personal ability score based on the representation content further comprises: The multi-head attention mechanism is adopted to achieve fusion topic similarity and quantization attribute scoring, and comprehensive recommendation scores of all people are obtained; from multiple head attention weights Extract of Chinese and text vectors Corresponding weight parts are obtained by normalization after summation Simultaneously, extracting weight parts corresponding to the quantized vectors Squantitative, and summing and normalizing to obtain Finally, personnel recommended scores: 。

Description

Airplane pilot stage troubleshooting personnel recommendation method based on crew capability image Technical Field The application belongs to the technical field of airplane fault diagnosis, and particularly relates to an airplane pilot flight stage troubleshooting personnel recommendation method based on crew capability images. Background The crew is a key ring in the maintenance task of the testing machine, directly influences the test and can finish the test flight task on time. At present, the selection of operators in the rank of the testing machine mainly has two modes, namely, the first mode follows the progressive principle of hierarchy, namely, the mode is reported step by step from low to high according to the technical hierarchy sequence, such as a mechanic, a middle captain and the like, the linear response mode excessively depends on the organization hierarchy and ignores the specificity of professional skills, so that the personal technical expertise is erroneously matched with the fault type, and the second mode is highly dependent on the personal experience and social interaction of an organization manager in the decision making process, and the staff is scheduled in the social network through the impressing and subjective judgment of the manager, so that the systematic competence assessment is lacked. Disclosure of Invention The invention aims to solve the problem of low troubleshooting efficiency caused by unreasonable personnel selection in an aircraft troubleshooting scene, and provides an aircraft pilot flight stage troubleshooting personnel recommendation method based on crew capability images. The application provides an aircraft pilot stage troubleshooting personnel recommendation method based on crew capability images, which comprises the following steps: constructing a crew capability portrait based on the multidimensional personnel information; calculating a staff capacity score based on the representation content; TOP N was used as a person recommendation based on the person ability score. Preferably, before the crew capability portrait is constructed based on the multidimensional personnel information, the method further comprises: and acquiring historical fault information, crew basic information, scientific research information and skill information. Preferably, the acquiring the historical fault information, the crew base information, the scientific research information and the skill information includes: extracting structured data through a system interface; Unstructured data, including fault tickets, are extracted using OCR recognition techniques. Preferably, the structured data includes fault records of fault management modules in the technical state management system of the testing machine, employee basic information in personnel data, paper information in the scientific and technological achievement management library, patent information and technical report information. Preferably, after the historical fault information, the crew basic information, the scientific research information and the skill information are obtained, the method further comprises: And performing invalid and redundant data cleaning, unifying data formats, and integrating a plurality of source data by name disambiguation and unified fields. Preferably, the constructing the crew capability representation includes: Defining a portrait construction flow, determining portrait construction dimension and realizing portrait model representation. Preferably, the calculating the personnel ability score based on the portrait content includes: Dividing the portrait index into a semantic association layer and a capability assessment layer, wherein the semantic association layer comprises professional information in basic attributes, a subject of a scientific research result in scientific research attributes and troubleshooting experiences in skill attributes; calculating fault information vector using cosine similarity Similarity to each text vector: The ability evaluation layer comprises education years, working years, skill levels and scientific research output: normalization processing is carried out on education age Eu and work age Wu, and then weighted summation is carried out to form ScoreB, namely Whereinax(E)、 in(E)、ax(W)、 In (W) is the minimum and maximum of the educational years and working years, respectively, in the dataset; weighting calculation and normalization are carried out on the professional skill level Su, honor name Au and service model type Mu to obtain ScoreS; Specifically, skill levels are classified into primary, medium, high and expert grades, respectively, scores of 1-4 are given, grading is carried out according to influence and authority of honor and title numbers, national honor is given 5 points, provincial honor is given 4 points, municipal honor is given 3 points, enterprise honor is given 2 points, no honor and title number is given 1 point, the number of machine types served