A multi-dimensional quantitative evaluation method and system for the post matching degree of aviation test flight maintenance personnel

Abstract

The application belongs to the technical field of human resource management, and discloses a multi-dimensional quantitative evaluation method and system for post matching degree of aviation test flight maintenance personnel, which comprises the following steps: S1, a five-dimensional quantitative evaluation system special for the aviation test flight maintenance field is constructed, including safety psychology and stress resilience, complex system cognition and operation, team cooperation and communication, trait will and responsibility, physical endurance and environmental adaptability, and a weighted post capacity demand characteristic vector is generated based on a combination weighting method; S2, a multi-modal measurement tool including structured interviews, simulation situation tests, psychological scales and physiological monitoring is used to quantitatively form a weighted personal capacity characteristic vector; S3, a difference vector of the personal and post vectors is calculated, input into a matching degree prediction model based on a gradient boosting tree to output a matching degree score, and SHAP explainable artificial intelligence technology is introduced to quantitatively evaluate the contribution of each dimension to the matching degree score, and finally a visual diagnostic report containing the total matching degree score and "short board dimension" identification is generated. The application aims to solve the problems of dimension generalization, process subjectification and result "black box" in the existing evaluation. The application introduces explainable AI into the field of human-post matching, and realizes the scientization, objectification and transparency of high-risk and high-precision post personnel selection and training decision-making.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of aviation maintenance human resource management and artificial intelligence applications, specifically involving a multi-dimensional quantitative assessment method and corresponding assessment system for the job matching degree of aircraft test maintenance personnel. This invention is particularly aimed at the inherent high-risk, high-complexity, high-time-sensitivity, and highly uncertain characteristics of aviation maintenance tasks during the flight test phase, and aims to replace traditional job matching methods based on general competency models and subjective experience, addressing their fundamental shortcomings in terms of assessment specificity, objectivity, and result interpretability. Background Technology

[0002] Person-job fit is the core of corporate human resource management, aiming to ensure that an individual's knowledge, skills, abilities, and other characteristics (KSAOs) match the responsibilities and requirements of a specific position. In the aviation field, especially in the critical stage of aircraft development involving "flight testing," the degree of matching between maintenance personnel's capabilities and their positions directly affects flight test safety, mission efficiency, and model development progress.

[0003] Traditional methods for person-job matching and competency assessment have mainly gone through the following stages, each with significant shortcomings: First, there is the matching method based on expert subjective evaluation. A panel of senior experts conducts qualitative comparisons. While this method is simple, the evaluation criteria are inconsistent, and the results heavily rely on individual expert experience and subjective judgment, resulting in low reliability and validity. It is particularly ill-suited to handling non-standardized, unpredictable, and complex faults encountered during flight test maintenance.

[0004] Second, there are assessment methods based on general competency models. While these models have a certain degree of universality, they prove inadequate when applied to the highly specialized field of flight test and maintenance. For example, general models often emphasize "communication skills," but fail to highlight the specific requirements of flight test and maintenance, such as "precise technical communication under extreme time pressure" and "fault reasoning communication based on incomplete information." Furthermore, they completely ignore the unique psychological and physiological demands faced by flight test and maintenance engineers.

[0005] Third, methods based on psychometrics and weighted scoring. These methods quantify the assessment process to some extent, but their core flaws are: the selection of dimensions and the setting of weights are highly subjective; they usually assume that the dimensions are linearly separable and independent of each other, which does not match the complex interactions of abilities in reality; at the same time, it is still a "black box" process, and managers have no way of knowing the specific reasons for low matching and cannot provide targeted training and guidance.

[0006] Fourth, matching models based on traditional machine learning algorithms. These models improve prediction accuracy to some extent and can capture non-linear relationships. However, they face the new challenge of extremely poor model interpretability. Especially in the high-risk, high-reliability test and maintenance field, an unexplainable "black box" model is unlikely to gain the trust of managers and front-line engineers, and cannot meet the stringent compliance and review requirements of the aviation industry.

[0007] In summary, existing technologies have significant shortcomings in terms of dimension specificity, evaluation objectivity, and model interpretability. There is an urgent need for an intelligent evaluation method that can scientifically construct a dedicated dimension weight system, acquire multimodal quantitative data, and provide transparent and interpretable matching results. Summary of the Invention

[0008] The purpose of this invention is to provide a multi-dimensional quantitative evaluation method and system for the job matching degree of aircraft test maintenance personnel, so as to overcome the defects of the existing technology, such as the generalization of evaluation dimensions, the subjectivity of evaluation process, and the "black box" of matching results. It provides a scientific, quantitative, objective and evidence-based decision support system for the selection, accurate job assignment and personalized training of high-risk positions for aircraft test maintenance personnel.

[0009] The technical solution of this invention is implemented as follows: Firstly, this invention provides a multi-dimensional quantitative evaluation method for the job suitability of aircraft test maintenance personnel, including the following steps: S1. Construct an assessment system and job requirement model: Build a multi-dimensional quantitative assessment system specifically for the field of aircraft test maintenance. This system should include at least five primary dimensions: "Safety Psychology and Stress Resilience," "Complex System Cognition and Operation," "Teamwork and Communication," "Trait of Willpower and Sense of Responsibility," and "Physical Endurance and Environmental Adaptability." A weighted feature vector of job competency requirements should be generated. R w .

[0010] Specifically, step S1 further includes: S11. Through behavioral event interviews, the Delphi method, and expert focus groups, the five primary dimensions and their subordinate secondary indicators are determined.

[0011] S12, for specific positions in flight test and maintenance, uses the Analytic Hierarchy Process (AHP) to determine the subjective weight vector of each indicator. w ahp The objective weight vector of each indicator is calculated using the entropy weight method based on historical performance data. w entropy By combining weighting formulas w =αw ahp + (1- α ) w entropy obtain a final comprehensive weight vector, wherein α is a preset empirical coefficient (for example, the value is 0.6).

[0012] S13, setting a benchmark demand value for each secondary indicator R j (0-100 points) to form a post demand vector R and weighting the post demand vector by using said comprehensive weight vector w to obtain the weighted post capability demand feature vector R w .

[0013] S2, personal capability quantification and modeling: adopting a multimodal measurement tool matched with said evaluation system to collect actual capability level data of a personnel to be evaluated, and quantifying the data into a weighted personal capability feature vector P w .

[0014] Specifically, said multimodal measurement tool comprises at least two of the following or a combination thereof: a structured behavioral interview tool based on critical incident technology, a high-fidelity simulation situation operation test system, a standardized psychological scale, and a wearable physiological monitoring device. Step S2 further comprises: performing Z-score standardization processing on all collected original scores, and then performing weighted aggregation with said comprehensive weight vector w to form said weighted personal capability feature vector P w .

[0015] S3, matching degree prediction and interpretability diagnosis: calculating the difference vector between said weighted personal capability feature vector P w and said weighted post capability demand feature vector R w D = P w - R w ; inputting said difference vector D into a pre-trained matching degree prediction model based on gradient boosting decision tree, and outputting a prediction score representing post matching degree; and calling a SHAP (Shapley Additive Explanations) interpreter to calculate said difference vector D ​The SHAP value for each dimension feature in the predicted score is calculated; finally, based on the sign and absolute value of the SHAP value, an interpretable diagnostic report is generated that includes the total matching score and at least one "weakness dimension" identification.

[0016] The pre-trained gradient boosting tree-based matching prediction model is trained by: constructing a training dataset, wherein each training sample contains: a difference vector between the weighted individual ability vector of a historical personnel sample and the weighted requirement vector of the corresponding position. D k and a label y k The label y k Characterize the performance level of this historical personnel sample in subsequent work (e.g., high performance is labeled 1, low performance is labeled 0); train a gradient boosting tree model (such as XGBoost) on the training dataset to fit the difference vector. D To predict the matching score The nonlinear mapping relationship.

[0017] The specific method for generating an interpretable diagnostic report is as follows: Let the average expected value of the matching degree prediction model for all predictions be... The SHAP interpreter is the difference vector of the current input. D new The first in i The SHAP value obtained from the calculation of the features in each dimension is: and satisfy Based on the magnitude and sign of the SHAP values ​​for each dimension of features, if If the value is negative and its absolute value is greater than a preset threshold or it is ranked higher, then this value will be... The corresponding dimension is determined as the "weakness dimension". The diagnostic report includes at least an overall matching score. And the names of the indicators identified as "weakness dimensions" and their SHAP values.

[0018] As a further technical solution of the present invention, the diagnostic report also includes: based on the indicators determined to be "weakness dimensions", retrieving and outputting targeted training courses or practical task suggestions from a pre-built "capability-training comparison library".

[0019] Secondly, this invention provides a multi-dimensional quantitative evaluation system for the job matching degree of aircraft test maintenance personnel, including: The job requirement modeling module is used to build and store a multi-dimensional quantitative evaluation system specific to the field of aircraft test and maintenance, and to generate and output a weighted feature vector of job competency requirements based on this system. Rw ; wherein the evaluation system includes at least five first-level dimensions: "safety psychology and stress resilience", "complex system cognition and operation", "team collaboration and communication", "trait will and sense of responsibility", and "physical endurance and environmental adaptability".

[0020] a personal capability quantification module, configured to control and execute a multimodal measurement tool matching the evaluation system, collect actual capability level data of personnel to be evaluated, and quantify the data into a weighted personal capability feature vector P w .

[0021] a matching degree prediction and diagnosis module, which includes: a difference vector calculation unit, configured to calculate D = P w - R w ; a matching degree prediction unit, which is embedded with a pre-trained matching degree prediction model based on gradient boosting tree, and is configured to receive the difference vector D and output a prediction score representing post matching degree; a SHAP interpretability analysis unit, configured to call a SHAP interpreter to calculate the SHAP value of each dimensional feature in D the difference vector for the prediction score; a diagnosis report generation unit, configured to generate and output an interpretable diagnosis report including a total matching degree score and at least one "short board dimension" identification according to the positive-negative sign and absolute value of the SHAP value.

[0022] Compared with the prior art, the beneficial effects of the present invention are: 1. Innovation of dimensional system: The "special five-dimensional evaluation system for aviation flight test aircraft maintenance personnel" that is different from general competency models is clearly proposed and constructed for the first time. In particular, "safety psychology and stress resilience" and "physical endurance and environmental adaptability" are taken as independent first-level dimensions, which profoundly reflect the essential requirements of high risk, high pressure and high precision for flight test maintenance posts, and the evaluation has strong specificity.

[0023] 2. Innovation of weight determination method: Abandoning the practice of purely relying on subjective experience, a combined weighting method combining AHP and entropy weight method is adopted, which not only respects the prior knowledge of domain experts, but also utilizes the objective information of historical performance data, making weight setting more scientific and reasonable.

[0024] 3. Innovation of measurement tools: A multimodal measurement tool set of "interview + situational test + psychological scale + physiological monitoring" is designed, in particular, physiological data collected by wearable devices is introduced, which realizes objective quantification of abstract dimensions such as psychological resilience and physical endurance, and improves the reliability and validity of data collection.

[0025] 4. Algorithm and Technology Innovation: Gradient Boosting Tree (XGBoost) and SHAP interpretable AI technology are introduced into the field of job matching. XGBoost ensures high accuracy in matching degree prediction and can capture the nonlinear interaction effects between dimensions. The application of SHAP technology transforms the model from a "black box" to a "white box," enabling precise quantification of the marginal contribution of each evaluation dimension to the final matching score and directly locating the "weakest link dimension" that leads to mismatch.

[0026] 5. High Decision Support Value: The final output is not just a single matching score, but an interpretable diagnostic report that includes the score, strengths, weaknesses, and personalized development suggestions. This makes the evaluation results highly transparent and actionable, gaining the trust of managers and frontline engineers. It provides a solid scientific basis for personnel selection, job placement, and precise training, significantly improving the quality and safety of human resource decisions.

[0027] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating the overall method of an embodiment of the present invention. The flowchart summarizes the complete process from dimension construction and measurement implementation to intelligent matching and interpretation.

[0029] Figure 2 This is a flowchart of the person-job matching prediction and SHAP interpretability analysis sub-flowchart in an embodiment of the present invention. The diagram details the internal calculation logic from the difference vector to the final diagnostic report in step S3. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some embodiments of this invention, but not all embodiments.

[0031] The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the invention, and should not be construed as limiting the invention.

[0032] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0033] The following is in conjunction with the appendix Figure 1-2 The embodiments of the present invention will be described in detail below.

[0034] Example 1 This invention evaluates the match between an intermediate-level flight test maintenance mechanic and the "flight test maintenance mechanic position for a certain type of third-generation fighter jet".

[0035] I. Preliminary Preparations (corresponding to step S1) First, for the "flight test and maintenance mechanic position of a certain type of third-generation fighter jet", a job requirement model is constructed by implementing steps S11 to S13 of the present invention.

[0036] Assuming that after expert panel review and weighting calculations, the final comprehensive weights for some secondary indicators are... w and job baseline requirements R j As shown in the table below:

[0037] Then the weighted job demand vector R w The corresponding element is: R w,1 =88 × 0.15 = 13.2; R w,2 =80 × 0.10 = 8.0; R w,3 =90 × 0.12 = 10.8; R w,4 =75×0.08=6.0;… (Other dimensions omitted). This vector serves as the benchmark for subsequent evaluation.

[0038] II. Quantitative Assessment of Individual Competencies (corresponding to step S2) The person to be evaluated is engineer Li. Multimodal measurements were conducted for the aforementioned indicators: 1. Structured Interview: Two assessors conducted a retrospective interview with Mr. Li. Regarding "the pursuit of zero defects in work results" (D3), Mr. Li described a case of proactively identifying potential hazards. The two assessors gave him scores of 4.5 and 4.0 (out of 5), with an average of 4.25, which translates to 85 points on a percentage scale. Assume the mean score of the reference population on this indicator is... μ=75 Standard deviation σ=25 Then the standardized score P 3 = (85-75) / 25 = 0.4.

[0039] 2. Simulation Scenario Test: An engine vibration fault troubleshooting test was conducted on a simulator. The system recorded its "System Fault Logic Tracing Ability" (D1), with a raw score of 53.8 points (based on a combination of path and time consumption). Assuming the average score of the reference population... m =70, standard deviation s =20, then the standardized score P1 = (53.8 - 70) / 20 = -0.81.

[0040] 3. Physiological monitoring: Heart rate variability (HRV) under high-intensity tasks was monitored using wearable devices to characterize "resting and dynamic endurance" (D4). Mr. Li's HRV-RMSSD value was in the top 15% percentile of the reference population, and his standardized score was... P 4 = 1.2.

[0041] 4. Psychological scales: Li's "psychological resilience and ability to withstand setbacks" (D2) was assessed using the Occupational Resilience Scale. Li's standardized score... P 2 = -0.6.

[0042] The weighted individual competency vector is obtained by multiplying the standardized scores of all secondary indicators by their corresponding weights and then summing them up (example dimensions are listed only): P w =[ ,...]=[(-0.81)×0.15,(-0.6)×0.10,0.4×0.12,1.2×0.08,...]=[-0.1215,-0.06,0.048,0.096,...].

[0043] III. Match Prediction and Diagnosis (corresponding to step S3) 1. Calculate the difference vector: D 李某 = P w - R w .

[0044] D 1 = -0.1215 - 13.2 = -13.3215 D 2 = -0.06 - 8.0 = -8.06 D 3 = 0.048 - 10.8 = -10.752 D 4 = 0.096 - 6.0 = -5.904 ... In practical engineering applications, due to P w and R w The magnitudes may differ. Normalization can be performed on both before calculating the difference vector, or the difference can be calculated directly using an unweighted but standardized vector, followed by weighting. For simplicity, the above calculation logic is used here, and its physical meaning is "the degree to which the weighted capability exceeds / falls below the requirement."

[0045] 2. Model prediction: D 李某 Input a pre-trained XGBoost model. Assume the model outputs a matching score. =0.72, indicating that the predicted match between Li and the position is 72%.

[0046] 3. SHAP Interpretability Analysis: Call SHAP TreeExplainer to calculate the SHAP values ​​for each dimension. Assume the model baseline values... =0.65. The calculation result is: (Fault Origin) = -0.15 (Psychological resilience) = -0.10 (Pursuit of zero defects) = -0.08 (Static endurance) = +0.40 Total contribution from other dimensions =0.00 Verification: 0.65 + (-0.15 - 0.10 - 0.08 + 0.40) = 0.72, correct.

[0047] 4. Generate diagnostic report: The system automatically generates the following report: Overall match rate: 72% Core strength: Its strongest advantage lies in the "static and dynamic endurance" dimension (D4, SHAP=+0.40), which far exceeds the job benchmark and makes a significant positive contribution to the match.

[0048] Key weaknesses: “System failure logic tracing ability” (D1, SHAP=-0.15) and “psychological resilience and setback resistance” (D2, SHAP=-0.10) are the main factors that lower the matching degree.

[0049] Development Recommendations: We recommend that Mr. Li attend the "Advanced Training Course on Fault Tree Analysis of Complex Hydraulic Systems" to improve his ability to trace systemic causes; at the same time, we recommend arranging a "Psychological Resilience Training Workshop under High-Pressure Situations" for him to enhance his psychological stability under unexpected situations during flight tests.

[0050] Example 2 This invention evaluates the suitability of a senior avionics systems maintenance expert for a "new model UAV test flight maintenance position." This example aims to illustrate the flexibility of applying this invention in different positions.

[0051] S1 Step Review: For the "New Model UAV Test Flight Maintenance Position", the expert group increased the weight of the sub-indicator "Cross-system Fault Chain Logic Tracing" in the "Complex System Cognition" dimension (Dimension 2) (e.g., adjusted to 0.18), while significantly increasing the job benchmark requirement value of "Seeking Consensus in Disputes" in the "Team Collaboration" dimension (Dimension 3) (e.g., adjusted to 95 points), and lowered the benchmark requirement of the "Physical Endurance" dimension.

[0052] Step S2 Implementation: The person to be evaluated is Senior Engineer Wang. Evaluation results show: Simulation scenario test: Wang performed excellently in troubleshooting complex faults in a new type of UAV, achieving a high standardized score in "information integration ability". P 1 = 1.5.

[0053] Structured Interview: Wang was unable to clearly describe an effective collaboration case with designers; his standardized score for "cross-disciplinary information synchronization" ability was [not specified]. P 2 = -0.9.

[0054] All other dimensions are at a medium to high level.

[0055] Calculate its weighted individual ability vector P w .

[0056] Step S3 Implementation: After calculating the difference vector, input it into the XGBoost model and output the matching score. =0.53, which is lower than the job hiring threshold (assuming it is 0.7).

[0057] The SHAP interpretation result is: SHAP value for the high-tech capability dimension. 技术 =+0.25, SHAP value for low collaboration capability dimension 协作 = -0.32, and the sum of other dimensions is approximately 0.00. Final +0.25-0.32=0.53.

[0058] The generated diagnostic report: Overall match rate: 53% (no match) Core strength: Excellent technical problem-solving and information integration capabilities (SHAP+0.25).

[0059] The key weakness is the ability to collaborate across disciplines and resolve disputes (SHAP=-0.32), which is the decisive factor leading to the mismatch.

[0060] Recommendation: Wang is not suitable to be the frontline maintenance manager for this drone team. It is recommended that he be reassigned to a position requiring in-depth individual research, such as a "Technical Expert" or "Fault Analysis Center" role. If he must be retained, a specific behavioral improvement plan must be developed.

[0061] In summary, the method and system provided by this invention, by constructing a dedicated evaluation system, implementing multimodal objective measurement, and combining XGBoost and SHAP technologies, achieves a precise, objective, and transparent evaluation of the job matching degree of aircraft test maintenance personnel, which is significantly superior to existing technologies.

[0062] Example 3 This invention discloses a multi-dimensional quantitative assessment method for the job matching degree of aircraft test maintenance personnel, aiming to solve the problems of generalized dimensions, subjective assessment, and "black box" results in traditional personnel-job matching assessments. The core of this invention lies in: First, constructing a five-dimensional quantitative assessment system specifically for the field of aircraft test maintenance, including "safety psychology and stress resilience dimension," "complex system cognition and operation dimension," "teamwork and communication dimension," "trait will and sense of responsibility dimension," and "physical endurance and environmental adaptability dimension." The combined weights of each dimension and sub-dimension are determined based on the Analytic Hierarchy Process (AHP) and entropy weight method, forming a job competency requirement feature vector. Second, developing supporting multimodal measurement tools for the above dimensions, including structured behavioral interviews based on critical incident technology, high-fidelity simulation scenario operation tests, standardized psychological scales, and physiological indicator monitoring, to objectively obtain data on the actual competency level of maintenance personnel and quantify it into an individual competency feature vector. Finally, an innovative intelligent algorithm for person-job matching is designed. This algorithm does not simply calculate vector distance, but instead constructs a matching degree prediction model based on gradient boosting trees (such as XGBoost) and introduces SHAP (Shapley Additive Explanations) interpretable artificial intelligence technology. By quantifying the contribution of each evaluation dimension to the final matching degree score through SHAP values, it overcomes the "black box" limitation. It not only outputs a quantitative total score for person-job matching but also provides a visualized and interpretable diagnostic report of the reasons for matching / mismatch. For example, it precisely identifies the weakest dimension leading to low matching degree as "stress resistance" or "system operation accuracy." This invention provides a scientific, objective, and quantitative decision support tool for the selection, placement, and training of personnel in the high-risk, high-precision positions of aircraft test and maintenance, significantly improving the accuracy, relevance, and transparency of the assessment.

[0063] The specific technical solution is as follows: A multi-dimensional quantitative evaluation method for post matching degree of aviation flight test maintenance personnel, characterized by comprising the following steps: S1, constructing a multi-dimensional quantitative evaluation system and weights dedicated to aviation flight test maintenance personnel, and establishing a post capability demand feature vector; S2, designing and implementing a multimodal measurement tool matched with the evaluation system, acquiring actual capability level data of the personnel to be evaluated, and forming a personal capability feature vector; S3, constructing and executing a personnel-post matching intelligent evaluation algorithm based on the gradient boosting decision tree model and SHAP interpretable AI technology, and outputting a total matching score and an interpretable diagnosis report.

[0064] Further, constructing the dedicated multi-dimensional quantitative evaluation system in said step S1 specifically comprises: S11, screening five first-level dimensions different from general competencies through behavioral event interview, Delphi method and expert focus group, namely "safety psychology and stress resilience dimension", "complex system cognition and operation dimension", "teamwork and communication dimension", "trait will and sense of responsibility dimension" and "physical endurance and environmental adaptability dimension"; S12, further decomposing each first-level dimension into 2-4 measurable second-level indicators to form a complete evaluation indicator system.

[0065] Further, establishing the post capability demand feature vector in said step S1 specifically comprises: S13, for different posts of flight test maintenance (e.g., mechanic, avionics engineer), respectively adopting analytic hierarchy process (AHP) to determine the subjective weight vector of indicators at all levels w ahp ; S14, combining historical accident data and post performance data, adopting entropy weight method to calculate the objective weight vector w entropy ; S15, through a combination weighting formula w =αw ahp + (1- α ) w entropy obtaining the final comprehensive weight vector, wherein α is an empirical coefficient (e.g., 0.6); S16, setting a benchmark demand value for each second-level indicator according to post descriptions and expert consultation R j (0-100 points), then the post demand vector can be expressed as R = R 1, R 2,..., R n , and multiplying by the comprehensive weight w obtaining the weighted post demand vector R w .

[0066] Furthermore, step S2 involves designing and implementing a multimodal measurement tool, specifically including: S21, developing a structured behavioral retrospective interview outline based on key events to assess dimensions that are difficult to measure directly, such as "trait willpower and sense of responsibility," and having at least two senior assessors independently score using a Likert 5-point scale; S22, setting up typical fault scenarios to automatically record objective data such as the maintenance personnel's operation sequence, diagnostic time, and tool usage correctness, and mapping this data to a quantitative score for the "complex system cognition and operation dimension"; S23, using the revised "Aviation Personnel Stress Resilience Scale" and the "Big Five Personality Inventory (Revised Version)" to obtain quantitative data for psychological and trait dimensions; S24, monitoring physiological indicators such as heart rate variability during operations using wearable devices as supplementary objective data for the "physical endurance and environmental adaptability dimension"; and S25, after Z-score standardization of all raw scores, weighted aggregation to form an individual's actual ability feature vector. P =[ P 1, P 2,..., P n ] and its weighted form P w .

[0067] Furthermore, the construction of the intelligent assessment algorithm for person-job matching in step S3 specifically includes: S31, constructing a training dataset, where each data point contains a weighted individual ability vector of a sample. P w Corresponding weighted job requirement vector R w Difference vector D = P w - R w and tags y (The performance rating of this sample in subsequent work, e.g., high performance is labeled 1, low performance is labeled 0); S32, use a gradient boosting tree model (XGBoost) to train on the dataset and fit a non-linear mapping function. f (D) → ,in ∈[0,1] represents the predicted job match score; S33, when evaluating a new person-job pair, first calculate its difference vector. D new Then input the trained XGBoost model to obtain the matching score. S34, invoke the SHAP interpreter, targeting D new Calculate features for each dimensionD i SHAP value ,satisfy ,in S35, based on the model baseline expected value, generate a matching degree diagnostic report according to the magnitude and sign of the SHAP values ​​of each dimension. If the value is negative and the absolute value is large, then the dimension is determined to be a "weakness dimension" and targeted improvement suggestions are given.

[0068] The innovation of this invention lies in: 1. Dimensional System Innovation: For the first time, a "Five-Dimensional Assessment System for Aircraft Test and Maintenance Personnel" was clearly proposed and constructed, which is different from the general competency model. In particular, "Safety Psychology and Stress Resilience" and "Physical Endurance and Environmental Adaptability" are set as independent first-level dimensions, which profoundly reflect the essential requirements of this job.

[0069] 2. Innovative weighting method: Abandoning the practice of relying solely on subjective experience, a combined weighting method combining AHP and entropy weighting is adopted. This method respects the prior knowledge of domain experts while utilizing the objective information of historical performance data, making the weight setting more scientific and reasonable.

[0070] 3. Innovative Measurement Tools: A multimodal measurement toolset consisting of “interviews + situational tests + psychological scales + physiological monitoring” was designed. In particular, the introduction of physiological data from wearable devices enabled the objective quantification of abstract dimensions.

[0071] 4. Algorithm and Technology Innovation: Gradient Boosting Tree (XGBoost) and SHAP interpretable AI technology are introduced into the field of job matching. XGBoost ensures high accuracy in matching degree prediction, while the application of SHAP technology transforms the model from a "black box" to a "white box," enabling precise quantification of the marginal contribution of each evaluation dimension to the final matching score, thereby generating decision reports with diagnostic value.

[0072] Specific application examples: Assess the match between an intermediate-level flight test maintenance mechanic and the position of "flight test maintenance mechanic for a certain type of third-generation fighter jet".

[0073] Step S1 Review (Job Requirements Construction): Assuming the position is "Test Flight Maintenance Mechanic for a Certain Type of Third-Generation Fighter Jet", the secondary indicators are assigned using the AHP entropy weight combination weighting method. D 1. "System Fault Logic Tracing Ability" (belonging to Dimension 2), its comprehensive weight =0.15; Baseline job requirement R 1 = 88 points. Another secondary indicator. D 2. "Psychological resilience and ability to withstand setbacks" (belonging to dimension 1), weight =0.10, Job requirement valueR 2 = 80 points. The same logic applies to other dimensions. This leads to the calculation of the weighted job requirement vector. R w =[ ,...].

[0074] Step S2 implementation (Assessment subject: Engineer Li): S21 Structured Interview: Two assessors conducted a retrospective interview with Mr. Li. Regarding the "responsibility" indicator, based on his description of proactively identifying potential hazards, the two assessors gave him scores of 4.5 and 4.0 (out of 5), with an average of 4.25, converted to a percentage. x raw,1 =85 points. This was obtained through standardized calculation. P 1 = 0.4 (assuming the mean of the reference population) m =75, s =25).

[0075] S22 Simulation Simulator Test: Mr. Li troubleshoots a vibration fault in a simulated scenario. The system records a path matching accuracy of 92% and a total time consumption. T actual =18 minutes, and T std =25 minutes T opt =12 minutes. Substitute into the formula to calculate its original score for "fault logic tracing ability": Score =100×(25-18) / (25-12)=100×7 / 13≈53.8 (This indicates that although the speed is fast, the path is not optimal, resulting in a low score.) Original Score x raw,2 =53.8. This was obtained after standardization. P 2 = -0.85 (below the mean).

[0076] S23 Psychological Scale: Li scored 78 points on the NEO-FFI Responsibility Scale, slightly above the mean. After standardization... P 3 = 0.2.

[0077] S24 physiological monitoring: its HRV-RMSSD value was in the top 15% percentile of the reference population, after standardization P 4 = 1.2 (indicating that it has a strong physiological stress resistance).

[0078] Ultimately, Li's weighted personal ability vector P w =[0.4 0.15, -0.85 0.10, 0.2 0.12, 1.2 0.08,...]=[0.06,-0.085,0.024,0.096,...].

[0079] Step S3 is implemented (matching calculation): Calculate the difference vector: D 李某 = P w - R w Assuming R w If the corresponding positions are [0.10, 0.05, 0.04, 0.03, ...], then the difference vector contains... D 1 = 0.06 - 0.10 = -0.04; D 2 = -0.085 - 0.05 = -0.135; D 3 = 0.024 - 0.04 = -0.016; D 4 = 0.096 - 0.03 = 0.066; …

[0080] Model prediction: D 李某 Input a pre-trained XGBoost model, and the model outputs a matching score. =0.72. This indicates that Li's suitability for the position is 72% (slightly above average).

[0081] SHAP Interpretation: Calls the SHAP interpreter to calculate the SHAP values ​​for each dimension. Assume the model baseline values... =0.65. The calculation result is: ( D 1. Fault tracing) = -0.15 ( D 2, psychological resilience) = -0.20 ( D 3. Sense of responsibility) = -0.02 ( D 4. Compressive strength) = +0.44 Total contribution from other dimensions =0.00 Verification: 0.65 + (-0.15 - 0.20 - 0.02 + 0.44) = 0.72, correct.

[0082] Diagnostic report: Overall match rate: 72% Core strengths: Its strongest advantage is its "physiological stress resistance and endurance" (D4, SHAP=+0.44), which far exceeds the job benchmark and makes a significant positive contribution to environmental adaptability.

[0083] Key weaknesses: "Psychological resilience and setback tolerance" (D2, SHAP=-0.20) and "System failure logic tracing ability" (D1, SHAP=-0.15) are the main factors lowering the match. In particular, the psychological resilience score is far below the requirements for intermediate mechanics, which may lead to psychological exhaustion when dealing with unexpected events during test flights.

[0084] Development Recommendations: We recommend that Li participate in the "Psychological Resilience Training Workshop under High-Pressure Situations"; at the same time, we recommend that he undergo a two-week specialized simulation training in the area of ​​"Fault Diagnosis of Complex Hydraulic Systems" to improve his systematic tracing ability.

[0085] Example 4 This invention discloses a multi-dimensional quantitative assessment method for the job matching degree of aircraft test maintenance personnel, overcoming the shortcomings of existing technologies such as generalized assessment dimensions, subjective assessment process, and "black box" matching results. The invention mainly includes: First, constructing a quantitative assessment system specifically for the field of aircraft test maintenance, encompassing five specific dimensions: safety psychology, complex cognition, teamwork, trait willpower, and physical endurance. The weights of each dimension are determined through a combined weighting method, thereby forming a standardized job competency requirement feature vector. Second, designing a multimodal measurement tool integrating structured behavioral interviews, maintenance scenario testing, standardized scales, and physiological monitoring to objectively collect and quantify the actual competency level of personnel, forming an individual competency feature vector. Finally, inventing a person-job matching intelligent algorithm based on XGBoost and SHAP technologies. This algorithm calculates the difference between individual and job feature vectors and inputs them into a prediction model, ultimately outputting a matching degree score and SHAP values ​​for the contribution of each dimension, achieving intelligent and interpretable assessment results. The purpose of this invention is to provide a scientific, quantitative, objective, and evidence-based decision support system for the selection, precise job assignment, and personalized training of high-risk personnel for aircraft test and maintenance.

[0086] The specific implementation steps are as follows: S1. Construct a multi-dimensional quantitative evaluation system and weights specifically for aircraft test maintenance personnel, and establish a feature vector of job competency requirements.

[0087] This step is the basis for the invention's distinction from the general model, aiming to extract the evaluation dimensions that best represent the essence of flight test and maintenance work.

[0088] S11, Screening and Determining the Evaluation Dimension System. The Behavioral Event Interview (BEI) method was used to conduct in-depth interviews with 20 top-performing (top 20% performance) and 20 average-performing (median performance) flight test and maintenance personnel (covering mechanical, avionics, and other specialties). The core interview questions revolved around "your most memorable successful / tricky troubleshooting experience." Through transcription, coding, and thematic analysis of the interview recordings, key behavioral characteristics that frequently occurred and distinguished between top-performing and average personnel were extracted. Subsequently, a 12-member expert group, consisting of the chief flight test engineer, maintenance team leader, senior test pilot, aviation psychology expert, and human factors engineering expert, conducted three rounds of Delphi method consultation. Ultimately, the expert group unanimously confirmed and defined the following five primary dimensions, which are significantly different from general competencies: (1) Safety psychology and stress resilience dimension: refers to the ability to maintain a high level of vigilance regarding safety procedures, manage one's own emotions and stress responses, and recover quickly from failures or unexpected events under high-pressure and high-risk test flight environments. Its secondary indicators include: decision-making calmness in risk situations, internalization and execution of safety procedures, psychological resilience and ability to withstand setbacks.

[0089] (2) Complex System Cognition and Operation Dimension: This refers to the ability to deeply understand the complex electromechanical-hydraulic control systems of new aircraft models, rapidly integrate multimodal information, accurately diagnose faults, and precisely operate maintenance equipment. Its secondary indicators include: the ability to mentally rotate and reason about system schematic diagrams, the ability to logically trace fault chains across systems, and the hand-eye coordination and precision of maintenance operations.

[0090] (3) Teamwork and Communication Dimension: This refers to the ability to conduct efficient and accurate collaboration and communication with test pilots, designers, and other maintenance technicians during troubleshooting, especially under time pressure and uncertain information. Its secondary indicators include: cross-professional information synchronization and sharing, coordination ability to seek consensus in disputes, and clear issuance and confirmation of technical instructions (closed-loop communication).

[0091] (4) Trait Will and Sense of Responsibility Dimension: refers to the sense of responsibility, rigor, initiative, and pursuit of perfection necessary for long-term high-risk and high-intensity work. Its secondary indicators include: vigilance in proactively identifying potential hazards, the pursuit of zero defects in work results, and self-discipline in the absence of supervision.

[0092] (5) Physical endurance and environmental adaptability dimension: refers to the physical strength, endurance and fatigue resistance necessary to maintain efficient maintenance operations in extreme environments (such as high temperature and high noise) and special periods (such as night). Its secondary indicators include: static and dynamic endurance, hand stability under fine operation, and physiological adaptability to noise and vibration environments.

[0093] S12, Establish the feature vector of job competency requirements. After determining the dimensions, it is necessary to set the ideal competency requirement standards for the specific job (let's call it job A).

[0094] S13, Using the Analytic Hierarchy Process (AHP) to determine subjective weights. w ahp The expert panel conducted pairwise comparisons of the five primary dimensions and their subordinate secondary indicators established in S11, constructing a judgment matrix. For example, the importance of the secondary indicators "decision-making composure under risk situations" and "internalization and execution of safety procedures" relative to their parent objective, "safety psychology and stress resilience," was assessed. The eigenvector corresponding to the largest eigenvalue of the judgment matrix was calculated, and a consistency test (CR < 0.1) was performed. This yielded the final results for each secondary indicator. j Subjective weight w ahp,j , forming vectors w ahp .

[0095] S14, Determining objective weights using the entropy weight method. w entropy Extract historical performance data (such as quarterly performance appraisal scores, number of major tasks completed, and human error rate) and their historical evaluation scores on each secondary indicator from the company's human resources database for at least 50 employees in similar positions (e.g., flight test maintenance mechanics) over the past three years. This data can be retrospective scores from existing evaluation data. After normalizing the data, calculate the scores for each indicator. j Information entropy E j The higher the information entropy, the lower the discrimination of the indicator, and the smaller its weight. Objective weight. w entropy,j The calculation formula is:

[0096] S15, Determine the final overall weight. w To balance expert experience with factual data, a linear combination weighting method is adopted:

[0097] in, α This is a weighting coefficient, representing the degree of reliance on subjective experience. In high-safety areas such as flight testing, expert experience is crucial, and it is recommended to... α The value is set to 0.6. The final comprehensive weight vector is obtained. w =[ ], n This represents the total number of secondary indicators.

[0098] S16. Determine the baseline requirements for the position. Invite at least three senior managers from the department to determine a minimum expected baseline score for each secondary indicator j, based on the job description, job analysis, and future task planning, using a 100-point scale (0-100 points). R j For example, for the indicator of "fault logic tracing ability," the job requirement for an intermediate flight test maintenance engineer might be 85 points, while a senior position might require 95 points. This yields a job requirement vector. R =[ R 1,..., R n Finally, the weighted feature vector of job requirements is: R w =[ ] S2. Design and implement multimodal measurement tools to complement the assessment system, obtain data on the actual ability level of the personnel to be assessed, and form an individual ability characteristic vector.

[0099] S21, Structured Behavioral Retrospective Interview (for Dimension 4 and some aspects of Dimension 1). Design a standardized interview outline containing 6-8 key event questions, such as: "Please recall a time when you discovered a potential, but not explicitly identified, safety hazard while performing an urgent maintenance task. What were your thoughts at the time? What did you do? What was the result?" The entire interview is recorded. Two trained, non-interfering assessors score the interviewee on a 5-point Likert scale (1 = far below expectations, 5 = far above expectations) based on the interviewee's answers, referring to the *Behavioral Indicator Scoring Anchoring Manual*. The average score is then linearly converted to a percentage (score = average score × 20). This process primarily quantifies "trait willpower and responsibility" and "stress resilience."

[0100] S22, Scenario-based Maintenance Test (for Dimension 2). Several typical flight test emergency failure scenarios are designed (such as vibration of a certain type of engine at a specific altitude, intermittent avionics system errors, etc.), and subjects simulate troubleshooting operations. The system automatically records and quantifies the following indicators: (1) System diagnostic logic: Record the operation path, compare it with the optimal solution path, and calculate the path matching degree.

[0101] (2) Operational accuracy and time: Record the correctness of replacing parts or adjusting parameters, and the total time from the occurrence of the fault to the confirmation of the root cause. T total ).

[0102] (3) Information search efficiency: Eye tracking records the switching frequency and fixation duration between multiple system interfaces to evaluate information integration efficiency.

[0103] The system converts the collected raw data into a secondary indicator score ranging from 0 to 100. For example, the formula for calculating the score for "fault logic tracing capability" could be:

[0104] in T std The preset standard time (0 minutes if exceeded). T opt For optimal time consumption, T actual This represents the actual time spent.

[0105] S23, Standardized Psychological Scale Testing (for dimensions 1, 3, and 4). Commercially available psychological scales with good reliability and validity were selected and revised. For example, the Aviation Safety Attitude Questionnaire (ASQ) subscale was used to assess "internalization of safety procedures"; the Occupational Resilience Scale was used to assess psychological resilience; the Conscientiousness sub-dimension of the Big Five Personality Inventory (NEO-FFI) was used to assess "pursuit of work results"; and the Team Situational Awareness Scale was used to assess cross-disciplinary information synchronization ability. Scale scores were converted to a 0-100 scale according to their standard scoring rules.

[0106] S24, Physiological Indicator Monitoring (for Dimension 5). During the S22 simulation test, subjects were required to wear a medical-grade smart bracelet / chest strap to monitor electrocardiogram (ECG) and electrical activity of the skin (EDA). After the test, the root mean square difference (RMSSD) of heart rate variability (HRV) and the frequency of fluctuations in EDA were calculated during the high-pressure, long-duration maintenance task. Lower HRV and higher EDA fluctuation frequency are generally associated with high stress and low emotional regulation ability, indirectly reflecting physical exertion and recovery status. Objective data such as RMSSD were mapped to indicators such as "static endurance" and "hand stability," with the mapping function established based on previous regression analysis of typical samples.

[0107] S25, Data Standardization and Aggregation. All raw scores from S21-S24. x raw (Possibly different units of measurement), uniformly perform Z-score standardization to eliminate the influence of units of measurement:

[0108] in, m j and s j This is the secondary indicator. jThe mean and standard deviation in the reference population (e.g., all maintenance personnel of the same level). After standardization, the individual's actual ability vector is obtained. P =[ P 1,..., P n Finally, multiply by the comprehensive weights to obtain the weighted individual ability feature vector: P w =[ ] S3, based on the gradient boosting tree model and SHAP interpretable AI technology, constructs and executes an intelligent assessment algorithm for matching people to jobs.

[0109] S31, Construct the training dataset. Extract from the enterprise historical database. N records (e.g., N =200), each record is a sample. k ,Include: (1) Weighted individual ability vector of the sample P w,k .

[0110] (2) Weighted demand vector of the corresponding job positions in the sample R w,k (Different positions, R w different).

[0111] (3) Calculate the difference vector between the two. D k = P w,k - R w,k =[ D 1, D 2,..., D n Each element of this vector represents the degree to which the person exceeds or falls below the job requirements in that dimension.

[0112] (4) Label y k This represents the employee's performance level one year after joining the company or being assigned to a specific position. Employees in the top 30% of performance are labeled 1 (high match), and those in the bottom 30% are labeled 0 (low match). Employees in between can be selectively removed or processed in other ways. This binary label is defined as the prediction target.

[0113] S32, Train the XGBoost matching prediction model. Split the dataset into training and test sets (e.g., 80 / 20 split). Train using the XGBoost classifier, with the difference vector as input. D The output is the predicted matching probability. ∈[0,1]. XGBoost iteratively adds decision trees, with each new tree fitting the residuals predicted by the previous tree. Its objective function is:

[0114] in L It is a loss function (such as log loss). Oh This is a regularization term (penalizing model complexity). By optimizing hyperparameters (such as tree depth, learning rate, and number of weak learners) through grid search, a high-precision matching prediction model capable of handling complex nonlinear relationships is ultimately obtained. f .

[0115] S33, predict the matching degree for the new sample. For a person-job pair to be evaluated (person... M ,post J ), measure and calculate its according to step S2. P w And obtain the job title pre-generated in step S1. R w Calculate the difference vector. D new = P w -R w .Will D new Input into the trained XGBoost model f In the process, the predicted matching score is obtained:

[0116] This score directly reflects the personnel M With position J The matching probability.

[0117] S34, Introducing SHAP for interpretability analysis. This is the core innovation of this invention. It calls the SHAP interpreter (such as TreeExplainer from Python's shap library) of the pre-trained XGBoost model. SHAP, based on Shapley values ​​in game theory, fairly distributes the model's predictions to each input feature. For predictions of new samples... SHAP value i satisfy:

[0118] in, 0 is the model's average expected value (baseline value) across all predictions. i That is the firsti Each dimension (i.e., in the difference vector) D i The contribution of ) to this prediction. A positive i This indicates that the feature value of this dimension has driven the prediction result higher than the baseline value, while the opposite indicates suppression.

[0119] S35 generates diagnostic reports and decision support. The system automatically generates a person-job matching assessment report, including: (1) Overall matching degree: (For example, 0.85).

[0120] (2) Short board / long board analysis: SHAP values ​​of all dimensions i Sort by absolute value from largest to smallest. The system identifies the top 2-3 dimensions with negative SHAP values ​​and the largest absolute values ​​as "critical weakness dimensions," indicating that personnel are significantly below job requirements in these areas, constituting the main cause of the mismatch. Dimensions with positive and large SHAP values ​​are identified as "core strength dimensions."

[0121] (3) Visualization charts: Draw force plots or bar charts to visually demonstrate how each feature affects the baseline value. 0 pushes towards the final predicted value .

[0122] (4) Personalized development suggestions: Based on the identified shortcomings, targeted training courses or practical task suggestions are retrieved and matched from the pre-built "capability-training comparison library". For example: "Your 'fault logic tracing ability' has a SHAP value of -0.32, which is the main reason for the low matching degree. It is recommended to participate in the 'Advanced Training Course on Fault Tree Analysis of Complex Systems' and complete 5 simulated case training." Specific application examples: Assess the suitability of a senior avionics systems maintenance expert for the "new model UAV test flight maintenance position".

[0123] Background: The test flight and maintenance of new UAV models pose new challenges to traditional avionics engineers, relying heavily on autonomous decision-making, cross-system data fusion and analysis capabilities, and often requiring remote support, thus changing communication patterns.

[0124] Step S1 Review: For this position, the expert panel increased the weight of the sub-indicators "Cross-system Fault Chain Logical Origin" and "Multimodal Information Integration" in "Complex System Cognition" (Dimension 2). w At the same time, the baseline requirement for the role of "seeking consensus amidst disputes" in "team collaboration" (dimensional 3) was significantly increased, because ground station and field maintenance personnel need to collaborate efficiently remotely. Meanwhile, the baseline for "physical endurance" was lowered, because the ground support environment for drones is relatively better than that for manned drones.

[0125] Step S2 Implementation (Assessment Subject: Senior Engineer Wang): Simulated scenario test: Wang, in a simulated ground station control console for a new model of UAV, faced a complex fault involving flight control, data links, and payload. The system evaluation scored his "information integration capability" as high as 96 points (raw), far exceeding the average person's performance. Standardized score P 1 = 1.5.

[0126] Structured Interview: Regarding the "cross-disciplinary information synchronization" indicator, Wang was unable to clearly describe a case where he effectively resolved a dispute with designers; his answers were mostly "working on it alone." The assessor awarded him 3.0 points (out of 5, passing grade). Standardized score. P 2 = -0.9.

[0127] Other dimensions: Excellent performance in both sense of responsibility and resilience under pressure. P 30.8).

[0128] calculate P w : Weighting the weights, its P w It exhibits characteristics of "high technical capabilities and medium to low collaborative capabilities".

[0129] Step S3 is implemented as follows: Calculate the difference vector: due to job requirements R w The collaboration ability score was highly weighted and the baseline was high, but Wang's collaboration ability score was low and negative, resulting in... D 协作 It is a very large negative value (e.g., -0.25).

[0130] Model prediction: Input to XGBoost model, output matching score =0.53. This score is below the company's 70% hiring cutoff.

[0131] SHAP Explanation: Calculates the SHAP value. Assume... 0 = 0.60.

[0132] SHAP value in the high-tech capability dimension: 技术 =+0.25 (Advantage) SHAP value for low collaboration capability dimension: 协作=-0.32 (Severe weakness) The sum of other dimensions ∑=0.00 Final: 0.60 + 0.25 - 0.32 = 0.53.

[0133] Diagnostic report: Overall match rate: 53% (no match) Core strengths: Excellent technical problem-solving and information integration capabilities (SHAP+0.25), fully capable of handling the complex technical challenges of new drones.

[0134] The key weakness—cross-disciplinary collaboration and dispute resolution capabilities (SHAP=-0.32)—is the decisive factor leading to the mismatch. The results show that although Wang possesses exceptional technical skills, his aversion to communication is severely incompatible with the highly efficient "ground-on-site" collaboration requirements emphasized in the test flight maintenance of the new model.

[0135] Decision Recommendation: This result suggests to the Human Resources department and project manager that Wang is not suitable as the frontline maintenance leader for this drone team. However, given his strong technical skills, it is recommended that he be reassigned to a position such as "Technical Expert" or "Fault Analysis Center," which focuses more on in-depth individual research rather than real-time collaboration. If he must be retained, a very detailed Performance Improvement Plan (PIP) must be developed, and he should be assigned a mentor for at least 6 months of intensive collaborative communication training.

[0136] Thus, the objective of this invention has been achieved.

[0137] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-dimensional quantitative evaluation method for the job matching degree of aircraft test maintenance personnel, characterized in that, Includes the following steps: S1. Construct an assessment system and job requirement model: Build a multi-dimensional quantitative assessment system specifically for the field of aircraft test maintenance. This system should include at least five primary dimensions: "Safety Psychology and Stress Resilience," "Complex System Cognition and Operation," "Teamwork and Communication," "Trait of Willpower and Sense of Responsibility," and "Physical Endurance and Environmental Adaptability." A weighted feature vector of job competency requirements should be generated. R w ; S2, Quantification and modeling of personal ability: adopting a multimodal measurement tool matching the assessment system, collecting actual ability level data of the personnel to be assessed, and quantifying the data into a weighted personal ability feature vector P w ; S3, Matching Prediction and Interpretability Diagnosis: Calculate the weighted individual ability feature vector. P w Weighted Job Competency Requirement Feature Vector R w Difference vectors between D = P w - R w ; the difference vector D The input is fed into a pre-trained gradient boosting tree-based matching prediction model, which outputs a predicted score representing the job matching degree; and the SHAP interpreter is invoked to calculate the difference vector. D The SHAP value of each dimension feature for the predicted score is calculated; finally, based on the sign and absolute value of the SHAP value, an interpretable diagnostic report containing the total matching score and at least one "weakness dimension" is generated.

2. The method according to claim 1, characterized in that, In step S1, a multi-dimensional quantitative evaluation system specifically for the field of aircraft test and maintenance is constructed, and a weighted feature vector of job competency requirements is generated. R w Specifically, it includes: S11. Through behavioral event interviews, the Delphi method, and expert focus groups, the five primary dimensions and their multiple secondary indicators are determined. S12, for specific positions in flight test and maintenance, uses the analytic hierarchy process (AHP) to determine the subjective weight vectors of each indicator. w ahp The objective weight vector of each indicator is calculated using the entropy weight method based on historical performance data. w entropy By combining weighting formulas w =αw ahp + (1- α ) w entropy The final comprehensive weight vector is obtained, where α The preset empirical coefficient; S13, set a baseline demand value for each secondary indicator. R j To form a vector of job requirements R And using the comprehensive weight vector w The weighted feature vector of job competency requirements is obtained by weighting it. R w .

3. The method according to claim 1, characterized in that, The multimodal measurement tool in step S2 includes at least two of the following: A structured behavioral interview tool based on critical incident technology is used to assess dimensions that are difficult to quantify directly; A high-fidelity simulation scenario operation test system is used to automatically record operation sequences, diagnostic time consumption, and tool usage correctness data in simulated fault scenarios; Standardized psychological scales are used to collect data on psychological traits and states. Wearable physiological monitoring devices are used to collect physiological indicator data during the work process; Furthermore, step S2 further includes: standardizing all collected raw scores, and then combining them with the comprehensive weight vector. w Weighted aggregation to form the weighted individual ability feature vector. P w .

4. The method according to claim 1, characterized in that, The training process of the pre-trained gradient boosting tree-based matching degree prediction model includes: Construct a training dataset where each training sample contains: a difference vector between a weighted individual ability vector of a historical personnel sample and a weighted requirement vector for the corresponding job position. D k and a label y k The label y k Characterize the performance level of this historical personnel sample in subsequent work; The gradient boosting tree model is trained on the training dataset to fit the difference vector. D To predict the matching score The nonlinear mapping relationship.

5. The method according to claim 1, characterized in that, In step S3, the specific method for generating an interpretable diagnostic report is as follows: Let the average expected value of the matching degree prediction model for all predictions be _____. The SHAP interpreter is the difference vector of the current input. D new The first in i The SHAP value obtained from the calculation of the features in each dimension is: and satisfy ; Based on the magnitude and sign of the SHAP values ​​for each dimension of features, if If the value is negative and its absolute value is greater than a preset threshold or it is ranked higher, then this value will be... The corresponding dimension is determined as the "weakness dimension"; the diagnostic report includes at least the overall matching score. And the names of the indicators identified as "weakest link dimensions" and their SHAP values.

6. The method according to claim 5, characterized in that, The diagnostic report also includes: Based on the indicators identified as "weakness dimensions", targeted training courses or practical task suggestions are retrieved and output from a pre-built "capability-training comparison library".

7. A multi-dimensional quantitative evaluation system for the job matching degree of aircraft test maintenance personnel, characterized in that, include: The job requirement modeling module is used to build and store a multi-dimensional quantitative evaluation system specific to the field of aircraft test and maintenance, and to generate and output a weighted feature vector of job competency requirements based on this system. R w The assessment system mentioned above includes at least five primary dimensions: "safety psychology and stress resilience", "complex system cognition and operation", "teamwork and communication", "trait will and sense of responsibility" and "physical endurance and environmental adaptability". A personal ability quantification module, configured to control and execute a multi-modal measurement tool matching the evaluation system, collect actual ability level data of a person to be evaluated, and quantify the data into a weighted personal ability feature vector P w ; The matching degree prediction and diagnosis module includes: Difference vector calculation unit, used to calculate D = P w - R w ; The matching degree prediction unit embeds a pre-trained matching degree prediction model based on gradient boosting trees, used to receive the difference vector. D It also outputs a predicted score that represents the degree of job matching; The SHAP interpretability analysis unit is used to invoke the SHAP interpreter to calculate the difference vector. D The SHAP value of each dimension feature for the predicted score; The diagnostic report generation unit is used to generate and output an interpretable diagnostic report containing a total matching score and at least one "weakness dimension" identification based on the sign and absolute value of the SHAP value.

8. The system according to claim 7, characterized in that, The job requirement modeling module further includes: The subjective weight calculation unit is used to perform the analytic hierarchy process (AHP) to determine the subjective weight vectors of each indicator. w ahp ; The objective weight calculation unit is used to perform the entropy weight method to calculate the objective weight vector of each indicator based on historical performance data. w entropy ; Combined weighting units are used to assign weights through formulas. w =αw ahp + (1- α ) w entropy Calculate the comprehensive weight vector, where α The preset empirical coefficient; The demand vector generation unit is used to generate a demand vector based on the baseline demand value of each input secondary indicator. R j Forming a job requirement vector R And using the comprehensive weight vector w The features are weighted to generate the weighted job competency requirement feature vector. R w .

9. The system according to claim 7, characterized in that, The multimodal measurement tools controlled by the personal ability quantification module include at least two of the following: Structured behavioral interview tools, high-fidelity simulation scenario operation testing systems, standardized psychological scales, and wearable physiological monitoring devices; Furthermore, the individual ability quantification module also includes a data standardization and aggregation unit, used to standardize all collected raw scores and then combine them with the comprehensive weight vector. w Weighted aggregation to generate the weighted individual ability feature vector. P w .

10. The system according to claim 7, characterized in that, The diagnostic report generation unit in the matching degree prediction and diagnosis module is also used for: Based on the indicators identified as "weakness dimensions", targeted training courses or practical task suggestions are retrieved and output from a pre-built "capability-training comparison library" and included in the interpretability diagnostic report.

Images