A method and system for flight training saturation time evaluation

Abstract

This invention relates to the field of flight training assessment technology, specifically a method and system for assessing flight training saturation time. The method first constructs a standard library of flight maneuvers based on flight data from excellent instructors. After collecting and preprocessing student training flight data, it uses an improved Dynamic Time Warping (MDTW) algorithm to calculate the similarity score between the student's maneuvers and the standard maneuvers. Then, it determines the saturation state of individual maneuvers based on the score trends. When all maneuvers are saturated, it calculates the training saturation time for each student. Finally, it estimates the training saturation time range for the current aircraft type using a t-distribution interval. This invention enables precise matching between student capability development and training cycles, breaking through the experience-based extensive training model, optimizing training resource allocation, and effectively improving the overall efficiency of student training.

Description

Technical Field

[0001] This invention relates to the field of flight training assessment technology, specifically a method and system for assessing flight training saturation time. Background Technology

[0002] Flight training is characterized by high resource consumption, long training cycles, and high costs. For a long time, flight training durations have typically been based on experience, lacking a scientific basis for evaluating the individual pilot's skill development process, making it difficult to achieve rational allocation of training resources and maximize training effectiveness. With the rapid development of modern aviation technology and the continuous increase in the complexity of flight missions, the requirements for precision and scientific rigor in pilot training are becoming increasingly urgent. There is a pressing need to achieve a precise match between training cycles and pilot skill development. How to accurately assess the "saturation time" in flight training—that is, the point at which trainees enter a plateau phase after reaching peak skill internalization efficiency—has become a key technical challenge that urgently needs to be addressed to optimize flight training effectiveness.

[0003] In current technologies, flight training generally adopts an experience-based and extensive management model. Training plans are mainly formulated based on the instructors' subjective experience, making it impossible to personalize adjustments to address the differences in the ability development of individual trainees. This can easily lead to problems such as wasted training resources or overtraining. Furthermore, existing technologies lack quantitative methods for assessing pilot skills, making it difficult to accurately capture the inflection point in trainee skill development and determine the specific time point when trainees reach capability saturation. This results in a severe disconnect between the training cycle and the capability development process, affecting both the quality of trainee training and increasing the overall cost of flight training. Therefore, there is an urgent need to develop a method and system for assessing flight training saturation time to overcome the shortcomings in current practical applications. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for evaluating flight training saturation time, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for assessing flight training saturation time includes the following steps:

[0007] S1. Construct a standard library of flight maneuver entries based on flight data from outstanding instructors;

[0008] S2. Collect and preprocess student training flight data, and use the improved dynamic time warping (MDTW) algorithm to calculate the similarity score between the student's actions and the corresponding standard actions in the standard library.

[0009] S3. Determine the saturation state of a single flight maneuver based on the changing trend of the similarity score. When all flight maneuvers have reached saturation and are qualified, calculate the training saturation time of a single trainee.

[0010] S4. Collect training saturation time data from multiple trainees and estimate the training saturation time range for the current model stage using the t-distribution interval.

[0011] As a further aspect of the present invention: step S1 specifically includes:

[0012] S11. Determine the scope of flight maneuver entries;

[0013] S12. Select outstanding instructors who meet the preset conditions as the data collection objects for the standard library;

[0014] S13. Collect the instructor's flight data under standard weather conditions, and divide it into independent data segments for each flight action item according to the preset start and end rules.

[0015] S14. Perform outlier removal, missing value imputation, and data standardization preprocessing on the segmented flight data;

[0016] S15. Store the preprocessed flight maneuver data fragments according to maneuver entries to form a standard library of flight maneuver entries.

[0017] As a further aspect of the present invention: In step S2, the specific steps for calculating the similarity score using the MDTW algorithm include:

[0018] S21. Construct a local cost matrix for student motion data and standard motion data, where the local cost is the weighted Euclidean distance between the two sets of flight data. The formula for calculating the weighted Euclidean distance is:

[0019]

[0020] Where W represents the number of parameters in the flight data. The weight coefficient for the Wth parameter. and These are the flight parameter vectors for the trainee and the standard maneuver at the corresponding moments;

[0021] S22. Under the slope constraints that limit the maximum slope of the matching path to 2 and the minimum slope to 1 / 2, the cumulative distance matrix is ​​calculated by dynamic programming to obtain the MDTW distance between the trainee's action and the standard action.

[0022] S23. Calculate the similarity metric based on the MDTW distance. The calculation formula is as follows:

[0023]

[0024] In the formula, D(m,n) is the final element value of the cumulative distance matrix, and m and n are the lengths of the student's motion data and the standard motion data, respectively;

[0025] S24. Take the maximum similarity metric between the student's action and all corresponding standard actions in the standard library, and use it as the score for that action item for the student.

[0026] As a further aspect of the present invention: In step S3, the specific steps for determining the saturation state of a single flight maneuver and calculating the training saturation time of a single trainee include:

[0027] S31. Perform exponential smoothing on the historical score sequence of each flight maneuver item for trainees to eliminate random fluctuations;

[0028] S32. Calculate the growth rate sequence of the smoothed score;

[0029] S33. Use a sliding window to calculate the mean and standard deviation of the score growth rate to determine the saturation judgment threshold.

[0030] S34. When the score growth rate of a certain flight action item is lower than the threshold for a preset number of consecutive times, it is determined that the action has reached saturation.

[0031] S35. Verify whether the final score of the action under saturation reaches the preset qualified level. If it does not reach the preset qualified level, the action is judged to be unqualified.

[0032] S36. When all flight maneuver items of a trainee have reached saturation and the score is qualified, the total training time before the trainee reaches saturation is accumulated to obtain the trainee's training saturation time.

[0033] As a further aspect of the present invention: in step S4, the calculation formula for the t-distribution interval estimation is as follows:

[0034]

[0035] in, The average training saturation time for trainees. Let γ be the standard deviation of the trainee's training saturation time, and γ be the confidence level. By consulting the t-distribution table, we can find that N is the total number of all flight trainees participating in the evaluation;

[0036] Choose the appropriate confidence level based on training needs; high confidence level corresponds to low attrition rate requirements, and low confidence level corresponds to low resource consumption requirements.

[0037] A flight training saturation time assessment system, comprising:

[0038] The standard library construction module is used to build a standard library of flight maneuver entries based on flight data from excellent instructors;

[0039] The data acquisition and preprocessing module is used to collect and preprocess student training flight data.

[0040] The motion scoring module has a built-in MDTW algorithm to calculate the similarity score between the student's motion and the corresponding standard motion in the standard library.

[0041] The saturation time calculation module is used to determine the saturation state of a single flight maneuver based on the changing trend of the similarity score. When all flight maneuvers have reached saturation and are qualified, the training saturation time of a single trainee is calculated.

[0042] The group statistics module is used to collect training saturation time data from multiple trainees and estimate the training saturation time range for the current model stage through t-distribution interval estimation.

[0043] As a further aspect of the present invention: the standard library construction module is specifically used for:

[0044] Define the scope of flight maneuver entries;

[0045] Select outstanding instructors who meet the preset criteria as the data collection objects for the standard library;

[0046] Flight data of instructors is collected under standard weather conditions and segmented into independent data segments for each flight maneuver according to preset start and end rules.

[0047] The segmented flight data is preprocessed by outlier removal, missing value imputation, and data standardization.

[0048] The preprocessed flight maneuver data fragments are categorized and stored according to maneuver entries, forming a standard library of flight maneuver entries.

[0049] As a further aspect of the present invention: the action scoring module is specifically used for:

[0050] Construct a local cost matrix between trainee motion data and standard motion data, where the local cost is the weighted Euclidean distance between the two sets of flight data;

[0051] Under the slope constraints that limit the maximum slope of the matching path to 2 and the minimum slope to 1 / 2, the cumulative distance matrix is ​​calculated by dynamic programming to obtain the MDTW distance between the trainee's action and the standard action.

[0052] Calculate similarity metrics based on MDTW distance;

[0053] The maximum similarity metric between the student's action and all corresponding standard actions in the standard library is taken as the score for that student's action item.

[0054] As a further aspect of the present invention: the saturation time calculation module is specifically used for:

[0055] The historical score sequence of each flight maneuver item for trainees is exponentially smoothed to eliminate random fluctuations;

[0056] Calculate the growth rate sequence of the smoothed score;

[0057] The mean and standard deviation of the score growth rate are statistically analyzed using a sliding window to determine the saturation threshold.

[0058] When the score growth rate of a certain flight action item is lower than the threshold for a preset number of consecutive times, the action is determined to have reached saturation.

[0059] Verify whether the final score of the action under saturation condition reaches the preset qualified level. If it does not reach the preset qualified level, the action is judged to be unqualified.

[0060] When a student has reached saturation and scored the required points for all flight maneuvers, the total training time before reaching saturation is accumulated to obtain the student's training saturation time.

[0061] As a further aspect of the present invention, it also includes a result display module for visually displaying the action score curve, saturation state, and training saturation time range of a single trainee at the current model stage.

[0062] Compared with the prior art, the beneficial effects of the present invention are:

[0063] This invention provides a method and system for assessing flight training saturation time, which can break through the previous experience-based and extensive training model and achieve a precise combination of flight trainee capability generation and training cycle.

[0064] By constructing a standardized library of flight maneuver items and using the improved dynamic time warping (MDTW) algorithm to objectively and quantitatively score trainees' maneuvers, the trainees' skill mastery can be accurately reflected.

[0065] By analyzing the changing trends of trainees' scores, we can accurately determine the saturation state of individual movements and overall training, and precisely calculate the training saturation time of individual trainees.

[0066] By statistically estimating the saturation time of multiple trainees, the general training saturation time range for the current aircraft type can be obtained, providing a scientific theoretical and methodological support for optimizing flight training program design, rationally allocating training resources, and improving the overall effectiveness of flight trainee training. Attached Figure Description

[0067] Figure 1 This is a flowchart of the flight training saturation time assessment method in an embodiment of the present invention.

[0068] Figure 2 This is a schematic diagram illustrating the calculation and evaluation process of flight training saturation time in an embodiment of the present invention. Detailed Implementation

[0069] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0070] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0071] Please see Figure 1 and Figure 2 The present invention provides a method and system for evaluating flight training saturation time, which can solve the technical problems in traditional flight training where training duration is based on experience and the training cycle does not match the development of pilot capabilities, resulting in wasted training resources or overtraining.

[0072] First, a standardized flight maneuver standard library is constructed based on flight data from excellent instructors. Second, flight data during student training is collected, and the similarity score between student maneuvers and standard maneuvers is calculated using an improved Dynamic Time Warping (MDTW) algorithm. Then, the saturation state of individual maneuvers is determined based on the trend of score changes. When all maneuvers reach saturation, the cumulative training time is the training saturation time for a single student. Finally, the general training saturation time range for the current aircraft type is obtained through statistical interval estimation, providing a quantitative basis for optimizing the training program.

[0073] In one embodiment of the present invention, the steps for establishing the standard library of flight action entries are as follows:

[0074] Step S101: Determine the scope of flight maneuvers. In this embodiment, the flight maneuvers are all the maneuvers that need to be mastered during the pilot's basic training phase, including but not limited to: takeoff and departure, approach and landing, continuous takeoff, go-around, level flight, constant altitude turn, constant bank turn, constant speed climb, constant altitude climb, constant speed descent, constant altitude descent, low speed flight, sideslip flight, stall detection and recovery, high bank turn, point turn, point hover, and low-altitude flyby.

[0075] Step S102: Screening data collection targets for the standard library. In this embodiment, an excellent instructor is defined as a pilot with over 1000 hours of flight experience on this aircraft type, over 500 hours of teaching experience, and an excellent annual flight assessment rating with no safety incidents in the past three years. At least five instructors meeting the above criteria are invited to participate in the standard library data collection to ensure the representativeness and authority of the standard library.

[0076] Step S103: Flight Maneuver Data Acquisition and Segmentation. Flight instructors complete training for all the above flight maneuvers under standard weather conditions (visibility ≥ 10 km, wind speed ≤ 5 m / s, no precipitation). The aircraft automatically records flight data at a frequency of no less than 5 Hz. Flight data includes two categories: continuous parameters and discrete parameters. Continuous parameters include altitude, airspeed, G-force, pitch angle, roll angle, heading angle, and rate of climb / descendancy. Discrete parameters include landing gear retraction / extension and flap retraction / extension.

[0077] During the flight, the instructors marked the start and end times of each flight maneuver according to standardized rules:

[0078] Takeoff and departure: from the moment the pilot releases the brakes until the moment the aircraft climbs to an altitude of 1,000 feet and maintains level flight for 10 seconds;

[0079] Approach and landing: from the moment the aircraft intercepts the Instrument Landing System (ILS) glide path until the moment the post-touchdown speed drops below 30 knots;

[0080] Constant speed climb: This begins when the engine thrust is adjusted to the climb thrust and the airspeed is stabilized at the target climb speed, and ends when the target altitude is reached.

[0081] The remaining action entries follow the same logic, with the start point being the moment the action instruction is triggered and the end point being the moment the action is completed and the state returns to stability.

[0082] The flight data is segmented according to the start and end times of the markings to obtain independent data fragments for each flight maneuver item. At least 10 valid data fragments from different instructors are collected for each maneuver item to construct the original dataset for the standard library. This step, through standardized maneuver segmentation rules, eliminates data differences caused by different instructors' marking habits, ensuring the consistency of the baseline for subsequent similarity calculations.

[0083] Step S104: Flight data preprocessing. Outlier removal, missing value imputation, and data standardization are performed on the segmented continuous parameter data; discrete parameter data does not require preprocessing.

[0084] 1. Outlier Removal: Given the time series data of a certain parameter in the flight data, L is the data length, and the detection threshold is calculated. :

[0085]

[0086] By comparing each data point in the dataset, if the data at time t satisfies the following condition:

[0087]

[0088] and:

[0089]

[0090] This data value Data outliers are set to null values. If a data segment contains more than 5 consecutive outliers or the outlier rate exceeds 20%, the data segment is discarded and re-acquired. This step effectively removes abnormal data caused by sudden sensor noise, electromagnetic interference, etc., improving the quality of the standard library data.

[0091] 2. Missing Value Imputation: Missing values ​​are imputed using linear interpolation. If the data at time t is empty, the nearest non-empty data point is found, and their time scales are denoted as follows: and The actual values ​​are respectively and Then the data filling value at time t is:

[0092]

[0093] Linear interpolation can restore the true trend of flight parameters to the greatest extent possible while ensuring data continuity.

[0094] 3. Data standardization: The Z-score method is used to standardize the data of each parameter to eliminate differences in units and differences in training conditions. and Let the mean and standard deviation of this parameter be respectively, then the data value at time t is standardized as follows:

[0095]

[0096] After standardization, the value range of all parameters is unified, avoiding the excessive influence of high-level parameters on the similarity calculation results.

[0097] Step S105: Construct a standard flight maneuver library. All preprocessed flight maneuver data fragments are categorized and stored according to maneuver entries, forming a standard flight maneuver entry library. Each maneuver entry in the library corresponds to multiple standardized flight data sequences, serving as the benchmark for subsequent trainee maneuver scoring.

[0098] In one embodiment of the present invention, the pilot action item scoring steps are as follows:

[0099] Step S201: Trainee Training Data Collection and Preprocessing. Flight data is collected from each training flight, with the collection frequency consistent with the standard library. The flight instructor marks the start and end times of each flight maneuver item according to the same start and end marking rules as in Step S103, thus segmenting the trainee's flight maneuver item data. Following the same preprocessing method as in Step S104, outlier removal, missing value imputation, and data standardization are performed on the trainee's continuous parameter data.

[0100] Step S202: Similarity measurement based on MDTW algorithm. The preprocessed student action data is compared with all data sequences of the corresponding action items in the standard library for similarity measurement. This invention uses the improved Dynamic Time Warping (MDTW) method to calculate the similarity score.

[0101] In this embodiment, the improvement of the MDTW algorithm over the conventional DTW algorithm lies in the addition of slope constraints during the dynamic programming path calculation process. This limits the maximum slope of the matching path to 2 and the minimum slope to 1 / 2, preventing excessively distorted matching paths and eliminating distance deviations between sequences of different lengths. The specific calculation steps are as follows:

[0102] 1. Construct a local cost matrix: Let the data for a certain action item in pilot training be... The data in the standard library is ,in and All are data vectors, representing the set of flight parameters collected at a certain moment, with subscripts... These are the sequence numbers sorted by time. Let C be the length of the two sets of sequence data. Construct the local cost matrix C:

[0103]

[0104] in, The weighted Euclidean distance between the two sets of flight data:

[0105]

[0106] Where W represents the number of parameters in the flight data. This represents the weighting coefficient for the Wth parameter. In this embodiment, the weighting coefficients for pitch angle, roll angle, and G-force are set to 2; the weighting coefficients for airspeed and rate of climb are set to 1; and the weighting coefficients for altitude and heading angle are set to 0.5. These weighting coefficients are determined using the analytic hierarchy process (AHP) combined with scores from 10 senior instructors. The weighted Euclidean distance reflects the varying degrees of influence of different flight parameters on maneuver quality, improving the accuracy of the scoring.

[0107] 2. Dynamic programming to calculate the cumulative distance matrix: Define the cumulative distance matrix. , Indicates starting from the origin arrive The minimum cumulative distance.

[0108] Initialize the first row and first column of the matrix:

[0109]

[0110]

[0111]

[0112] The entire matrix is ​​calculated recursively, with slope constraints applied:

[0113]

[0114] Among them, subscript These are the sequence numbers of the training data and the data in the standard library, sorted by time. , If the slope of a path is greater than 2 or less than 1 / 2, the path is prohibited, and the corresponding cumulative distance is set to infinity. Matrix The last item in the last column That is and The MDTW distance (Modified Dynamic Time Warping).

[0115] 3. Similarity metric calculation: Based on the MDTW distance, the similarity metric is calculated as follows:

[0116]

[0117] The similarity metric ranges from (0,1), with a value closer to 1 indicating a higher similarity between the student's movement and the standard movement.

[0118] Step S203: Determine the score of the action item. Assume there are p data items of this type in the standard library. Then the student's score for action item X is:

[0119]

[0120] Taking the maximum similarity metric as the actual score for that action item can reflect the trainee's best performance level for that action.

[0121] In one embodiment of the present invention, the steps for calculating the training saturation time of a single trainee are as follows:

[0122] Step S301: Smoothing of motion score data. Assume a trainee's score for a certain flight motion item, in order of training, is as follows: There are a total of l data points, representing the scores from l training iterations.

[0123] First, smooth the score S to obtain the smoothed score data S':

[0124]

[0125]

[0126] Here, α represents the smoothing factor, with a value between 0 and 1. In this embodiment, α is set to 0.2. For stunt maneuvers with high complexity (such as stall detection and recovery, and steep turns), α can be adjusted to 0.3-0.4 to balance the smoothness and sensitivity of the score. Let represent the k-th training score before and after smoothing, respectively. Exponential smoothing can eliminate random fluctuations in individual training scores, more accurately reflecting the true growth trend of learners' skills.

[0127] Step S302: Calculate the score growth rate. Calculate the growth rate of the smoothed score:

[0128]

[0129] The growth rate reflects the extent to which trainees improve their skill level in that movement after each training session.

[0130] Step S303: Sliding window statistics and threshold calculation. Take a sliding window sw; in this embodiment, sw is 5, corresponding to the average frequency of training this action per flight by the trainee. Calculate the mean and standard deviation of the score growth rate based on the sliding window:

[0131]

[0132]

[0133] in, and Let represent the mean and standard deviation of the score growth rate within the k-th sliding window, respectively.

[0134] The threshold is set as the mean minus the standard deviation within the sliding window, and the threshold should not be less than 0. When the score growth rate is less than the threshold:

[0135]

[0136] Step S304: Determining the Saturation State of a Single Action. When the score growth rate of a training session is less than a threshold, it indicates that the skill level of that action has not improved significantly during this training. If the score growth rate of the action item is lower than the threshold for three consecutive training sessions, the action item is determined to have reached saturation. Here, "three consecutive times" refers to three consecutive training sessions of the action. If the action is not trained in a particular flight, the consecutive number of sessions is extended without interruption.

[0137] Step S305: Maneuver Compliance Verification. When a maneuver item reaches saturation, it is necessary to verify whether its final score meets the qualification level. In this embodiment, the qualification level for the maneuver score is set to 0.8. This threshold is determined by statistically analyzing the maneuver data of 50 pilots who have passed this stage of assessment and taking 90% of their average score. For high-risk maneuvers such as stall perception and recovery, and steep turns, the qualification threshold is increased to 0.85. If the score does not reach the qualification level when saturation occurs, the trainee is deemed to have failed the maneuver and needs targeted retraining.

[0138] Step S306: Individual trainee saturation state judgment and saturation time calculation. When a trainee has reached saturation in all flight maneuvers and scored qualified, the trainee is judged to have reached the ability saturation level for this stage of training.

[0139] The total training time when a trainee's ability reaches saturation is recorded; this is the trainee's training saturation time. The specific calculation method is as follows:

[0140] Assuming the student reaches saturation during the Mth flight training session, and the duration of each training session prior to this session was as follows: The training time before this sortie reaches saturation is (Training time after reaching saturation is not included), therefore, its training saturation time is:

[0141]

[0142] This calculation method can accurately pinpoint the specific time point when trainees reach their ability saturation, avoiding the inclusion of ineffective training time in the saturation time and improving the accuracy of the evaluation results.

[0143] Example of single action saturation judgment:

[0144] Taking a student's score in a constant-speed climb exercise as an example, we will demonstrate the calculation process for saturation. Assume the student's scores for the 16 training sessions in the constant-speed climb exercise are as follows:

[0145] [0.22,0.28,0.38,0.45,0.49,0.55,0.61,0.66,0.73,0.77,0.8,0.82,0.82,0.83,0.84,0.83];

[0146] With a smoothing factor α of 0.2, the smoothed data is obtained as follows:

[0147] [0.22,0.268,0.358,0.432,0.478,0.536,0.595,0.647,0.713,0.759,0.792,0.814,0.819,0.828,0.838,0.832];

[0148] Calculate the score growth rate:

[0149] [-,0.218,0.334,0.207,0.108,0.120,0.111,0.087,0.103,0.063,0.044,0.029,0.006,0.011,0.012,-0.007];

[0150] Using a sliding window sw=5, calculate the average growth rate:

[0151] [-,-,-,-,-,0.198,0.176,0.127,0.106,0.097,0.082,0.065,0.049,0.030,0.020,0.010];

[0152] Calculate the standard deviation of the growth rate:

[0153] [-,-,-,-,-,0.081,0.087,0.041,0.011,0.020,0.025,0.027,0.033,0.021,0.014,0.012];

[0154] Calculate the judgment threshold:

[0155] [-,-,-,-,-,0.116,0.089,0.085,0.095,0.077,0.057,0.038,0.016,0.009,0.006,0];

[0156] The difference is obtained by comparing the score growth rate with the threshold:

[0157] [-,-,-,-,-,0.004,0.022,0.002,0.008,-0.014,-0.013,-0.009,-0.010,0.002,0.006,-0.007];

[0158] It can be seen that starting from the 10th training session, the score growth rate was below the threshold for three consecutive times (10th, 11th, and 12th). Therefore, it is determined that the constant speed climbing item reached saturation after the 12th training session, at which point the score was 0.82, which is considered a passing grade.

[0159] In one embodiment of the present invention, the interval estimation step for the training saturation time of the current model stage is as follows:

[0160] Step S401: Sample Data Collection. Collect the training saturation time data of N trainees who have completed this stage of training to form a sample set.

[0161] Step S402: t-distribution interval estimation. Since the trainee training saturation time follows a normal distribution, the population mean is estimated using the t-distribution interval:

[0162]

[0163] in, The average training saturation time for trainees. Let γ be the standard deviation of the trainee's training saturation time, and γ be the confidence level. By consulting the t-distribution table, we can find that N is the total number of all flight trainees participating in the evaluation;

[0164] The confidence level γ is selected based on training requirements: when a low attrition rate is required, choose 99% confidence; when low resource consumption is required, choose 95% confidence.

[0165] The saturation time obtained from the evaluation can be used to optimize the design of pilot training programs and achieve a match between pilot capability growth and training cycle.

[0166] Example of population saturation time estimation:

[0167] Assume that the training saturation time data of 30 trainees are collected as shown in Table 1.

[0168] Table 1. Statistics on training saturation time of 30 trainees (unit: minutes)

[0169] 504.14 513.73 538.82 520.36 515.66 529.54 509.41 505.01 510.81 522.40 549.22 538.81 540.81 510.90 539.39 509.61 529.65 509.49 509.97 529.87 513.10 508.57 496.03 526.21 501.37 495.68 542.85 538.41 508.33 542.02

[0170] The calculated mean saturation time is 520.34, and the standard deviation is 15.54. Using interval estimation at a 99% confidence level, the following table can be consulted: (29) = 3.038;

[0171] The estimation error was calculated as follows: =8.62;

[0172] Therefore, the confidence interval for the training saturation time in this model stage is 520.34 ± 8.62 minutes, or [511.72, 528.96] minutes.

[0173] The results of this interval estimation can be used to guide the development of training plans: when the requirements for flight progress are not high but the requirement for a low elimination rate is required, the training time can adopt the upper limit of the interval, 528.96 minutes; when the requirements for the elimination rate are not high but the requirement for low resource consumption is required, the training time can adopt the lower limit of the interval, 511.72 minutes.

[0174] In one embodiment of the present invention, a flight training saturation time assessment system for implementing the above method is also provided, comprising:

[0175] Standard Library Construction Module: Used to receive flight data from excellent instructors, complete motion segmentation and data preprocessing, and build and store a standard library of flight motion entries;

[0176] Data acquisition and preprocessing module: used to collect flight data during student training and perform motion segmentation and preprocessing according to unified rules;

[0177] Motion scoring module: Built-in MDTW algorithm, used to calculate the similarity score between the student's motion and the standard library motion;

[0178] Saturation Time Calculation Module: Used to calculate the saturation state of a single action based on the score sequence, and to calculate the training saturation time of a single trainee when all actions are saturated;

[0179] Group statistics module: used to estimate the t-distribution interval of saturation time data of multiple trainees and output the training saturation time range of the current model stage;

[0180] Results Display Module: Used to visually display the individual student's action score curve, saturation state, and group saturation time interval.

[0181] The modules mentioned above communicate with each other via a data bus to achieve real-time data transmission and processing.

[0182] It should be noted that, in this invention, although the specification describes the embodiments, not every embodiment contains only one independent technical solution. This way of describing the specification is only for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for assessing flight training saturation time, characterized in that, Includes the following steps: S1. Construct a standard library of flight maneuver entries based on flight data from outstanding instructors; S2. Collect and preprocess student training flight data, and use the improved dynamic time warping (MDTW) algorithm to calculate the similarity score between the student's actions and the corresponding standard actions in the standard library. S3. Determine the saturation state of a single flight maneuver based on the changing trend of the similarity score. When all flight maneuvers have reached saturation and are qualified, calculate the training saturation time of a single trainee. S4. Collect training saturation time data from multiple trainees and estimate the training saturation time range for the current model stage using the t-distribution interval.

2. The method for evaluating flight training saturation time according to claim 1, characterized in that, Step S1 specifically includes: S11. Determine the scope of flight maneuver entries; S12. Select outstanding instructors who meet the preset conditions as the data collection objects for the standard library; S13. Collect the instructor's flight data under standard weather conditions, and divide it into independent data segments for each flight action item according to the preset start and end rules. S14. Perform outlier removal, missing value imputation, and data standardization preprocessing on the segmented flight data; S15. Store the preprocessed flight maneuver data fragments according to maneuver entries to form a standard library of flight maneuver entries.

3. The method for evaluating flight training saturation time according to claim 1, characterized in that, In step S2, the specific steps for calculating the similarity score using the MDTW algorithm include: S21. Construct a local cost matrix for trainee motion data and standard motion data, where the local cost is the weighted Euclidean distance between the two sets of flight data. The formula for calculating the weighted Euclidean distance is: ； Where W represents the number of parameters in the flight data. The weight coefficient for the Wth parameter. and These are the flight parameter vectors for the trainee and the standard maneuver at the corresponding moments; S22. Under the slope constraints that limit the maximum slope of the matching path to 2 and the minimum slope to 1 / 2, the cumulative distance matrix is ​​calculated by dynamic programming to obtain the MDTW distance between the trainee's action and the standard action. S23. Calculate the similarity metric based on the MDTW distance. The calculation formula is as follows: ； In the formula, D(m,n) is the final element value of the cumulative distance matrix, and m and n are the lengths of the student's motion data and the standard motion data, respectively; S24. Take the maximum similarity metric between the student's action and all corresponding standard actions in the standard library, and use it as the score for that action item for the student.

4. The method for evaluating flight training saturation time according to claim 1, characterized in that, In step S3, the specific steps for determining the saturation state of a single flight maneuver and calculating the training saturation time of a single trainee include: S31. Perform exponential smoothing on the historical score sequence of each flight maneuver item for trainees to eliminate random fluctuations; S32. Calculate the growth rate sequence of the smoothed score; S33. Use a sliding window to calculate the mean and standard deviation of the score growth rate to determine the saturation judgment threshold. S34. When the score growth rate of a certain flight action item is lower than the threshold for a preset number of consecutive times, it is determined that the action has reached saturation. S35. Verify whether the final score of the action under saturation state reaches the preset qualified level. If it does not reach the preset qualified level, the action is judged to be unqualified. S36. When all flight maneuver items of a trainee have reached saturation and the score is qualified, the total training time before the trainee reaches saturation is accumulated to obtain the trainee's training saturation time.

5. The method for evaluating flight training saturation time according to claim 1, characterized in that, In step S4, the formula for calculating the t-distribution interval estimate is: ； in, The average training saturation time for trainees. Let γ be the standard deviation of the trainee's training saturation time, and γ be the confidence level. By consulting the t-distribution table, we can find that N is the total number of all flight trainees participating in the evaluation; Choose the appropriate confidence level based on training needs; high confidence level corresponds to low attrition rate requirements, and low confidence level corresponds to low resource consumption requirements.

6. A flight training saturation time assessment system, characterized in that, include: The standard library construction module is used to build a standard library of flight maneuver entries based on flight data from excellent instructors; The data acquisition and preprocessing module is used to collect and preprocess student training flight data. The motion scoring module has a built-in MDTW algorithm to calculate the similarity score between the student's motion and the corresponding standard motion in the standard library. The saturation time calculation module is used to determine the saturation state of a single flight maneuver based on the changing trend of the similarity score. When all flight maneuvers have reached saturation and are qualified, the training saturation time of a single trainee is calculated. The group statistics module is used to collect training saturation time data from multiple trainees and estimate the training saturation time range for the current model stage through t-distribution interval estimation.

7. The flight training saturation time assessment system according to claim 6, characterized in that, The standard library construction module is specifically used for: Define the scope of flight maneuver entries; Select outstanding instructors who meet the preset criteria as the data collection objects for the standard library; Flight data of instructors is collected under standard weather conditions and segmented into independent data segments for each flight maneuver according to preset start and end rules. The segmented flight data is preprocessed by outlier removal, missing value imputation, and data standardization. The preprocessed flight maneuver data fragments are categorized and stored according to maneuver entries, forming a standard library of flight maneuver entries.

8. The flight training saturation time assessment system according to claim 6, characterized in that, The action scoring module is specifically used for: Construct a local cost matrix between trainee motion data and standard motion data, where the local cost is the weighted Euclidean distance between the two sets of flight data; Under the slope constraints that limit the maximum slope of the matching path to 2 and the minimum slope to 1 / 2, the cumulative distance matrix is ​​calculated by dynamic programming to obtain the MDTW distance between the trainee's action and the standard action. Calculate similarity metrics based on MDTW distance; The maximum similarity metric between the student's action and all corresponding standard actions in the standard library is taken as the score for that student's action item.

9. The flight training saturation time assessment system according to claim 6, characterized in that, The saturation time calculation module is specifically used for: The historical score sequence of each flight maneuver item for trainees is exponentially smoothed to eliminate random fluctuations; Calculate the growth rate sequence of the smoothed score; The mean and standard deviation of the score growth rate are statistically analyzed using a sliding window to determine the saturation threshold. When the score growth rate of a certain flight action item is lower than the threshold for a preset number of consecutive times, the action is determined to have reached saturation. Verify whether the final score of the action under saturation condition reaches the preset qualified level. If it does not reach the preset qualified level, the action is judged to be unqualified. When a student has reached saturation and scored the required points for all flight maneuvers, the total training time before reaching saturation is accumulated to obtain the student's training saturation time.

10. The flight training saturation time assessment system according to claim 6, characterized in that, It also includes a results display module, which is used to visualize the action score curve, saturation status, and training saturation time range of a single trainee at the current model stage.

Images