Multi-dimensional data mining and analysis system based on flight simulation

Abstract

The invention relates to the technical field of data mining, in particular to a multi-dimensional data mining and analysis system based on flight simulation, which comprises a flight path characteristic analysis module, an action characteristic correlation analysis module, a key action offset prediction module and a training task distribution module. According to the method, the accuracy of track offset dynamic evaluation is improved by decomposing flight track change parameters and extracting track offset characteristic distribution, the capture capability of track change rules is remarkably enhanced, the correlation degree of control torque change and control surface offset is analyzed, and the key correlation between actions and tracks is quantified; the method lays a foundation for prediction of a high-frequency offset action distribution range, combines action characteristic parameter trend analysis, clears an offset direction and amplitude difference, optimizes control surface adjustment and control moment distribution, enables training distribution to meet requirements, realizes comprehensive coverage of flight data from law discovery to task optimization, improves decision precision and training efficiency, and improves the prediction efficiency. And meanwhile, the flight safety is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of data mining, and particularly to a multi-dimensional data mining and analysis system based on flight simulation. Background Art

[0002] The technical field of data mining involves extracting valuable knowledge and patterns from large amounts of data to assist in decision-making, predictive analysis, and pattern recognition. Through various algorithms and methods, such as classification, clustering, regression analysis, association rule discovery, etc., data mining can discover potential trends, associations, and regularities from structured or unstructured data, and is widely applied in multiple industries, such as finance, healthcare, marketing, industry, aviation, etc. When dealing with complex and large-scale data sets, it can automatically identify important information in the data to provide support for decision-making. With the improvement of computing power and the continuous growth of data scale, the development of data mining technology is also constantly deepening, promoting the implementation and application of intelligent systems.

[0003] Among them, a multi-dimensional data mining and analysis system is a system that integrates multiple data mining technologies, aiming to extract valuable information from multi-dimensional and multi-source aviation data to support real-time decision-making and optimization. This system can analyze real-time data in flight simulation, identify anomalies or optimization spaces during flight, and help pilots make effective decisions during training and actual flight. It can not only provide retrospective analysis based on historical data, but also conduct dynamic monitoring and decision-making suggestions based on real-time data, improving the efficiency and accuracy of flight training and flight safety.

[0004] The existing technology lacks refined processing in extracting flight trajectory characteristics from multi-dimensional data. The evaluation of trajectory deviation mostly relies on simple analysis of the overall trend, and fails to deeply analyze the subtle laws of trajectory changes, resulting in insufficient grasp of the dynamic characteristics of complex flight trajectories. In the analysis of the correlation between actions and trajectories, existing methods mostly focus on the static correlation evaluation of single variables, while ignoring the extraction of dynamic correlation characteristics and distribution characteristics, restricting the comprehensive analysis ability of key action characteristics and making it difficult to provide sufficient basis for the prediction of high-frequency deviation actions. In the prediction of key action deviation and task matching, the existing technology mostly stays at simple retrospective based on historical data, lacking dynamic monitoring and trend analysis of real-time data, resulting in insufficient accuracy and practicality of training task allocation, causing the disconnection between flight training and actual needs, affecting training effects, and posing potential risks to flight safety. In flight training, the distribution characteristics of high-frequency deviation actions cannot be accurately captured, resulting in insufficient correction of some key actions, and further affecting the scientificity and safety of flight decisions. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies existing in the prior art, and to propose a multi-dimensional data mining and analysis system based on flight simulation.

[0006] To achieve the above object, the present invention adopts the following technical solutions: A multi-dimensional data mining and analysis system based on flight simulation includes:

[0007] The flight trajectory characteristic analysis module is based on the flight trajectory data recorded by flight simulation, decomposes the flight trajectory change parameters, analyzes the dynamic changes of the heading angle and acceleration, evaluates the trajectory offset amplitude and change range, obtains the trajectory offset characteristic distribution result, and generates the flight trajectory deviation distribution result by comparing the flight attitude angles;

[0008] The action characteristic correlation analysis module is based on the flight trajectory deviation distribution result, performs matching of deviation distribution influencing factors, analyzes the correlation degree between the control moment change and the rudder surface offset, generates an action offset correlation matrix, extracts the distribution characteristics, analyzes the key action characteristics, and generates the key action characteristic analysis result;

[0009] The key action offset prediction module is based on the key action characteristic analysis result, identifies the distribution range of high-frequency offset actions, analyzes the change trend of action characteristic parameters, obtains the key action offset range, divides different categories according to the offset direction and amplitude, performs interval statistics and distribution analysis, and generates the key action offset prediction result;

[0010] The training task allocation module is based on the key action offset prediction result, analyzes the matching degree between the action characteristic data and the training task stage, optimizes the training distribution of the rudder surface adjustment and control moment, combines the task stage characteristics and the action correction characteristics, and generates a multi-dimensional training allocation scheme for flight simulation.

[0011] As a further solution of the present invention, the specific steps for obtaining the trajectory offset characteristic distribution result are as follows:

[0012] Based on the flight trajectory data recorded by flight simulation, extract the heading angle and acceleration data points, analyze the data at the time points, identify the heading angle offset and acceleration change, calculate the offset and change parameters of each point of the flight trajectory, and generate a flight trajectory change parameter list;

[0013] Based on the flight trajectory change parameter list, perform trajectory offset characteristic analysis, focus on the frequency distribution and change range of the trajectory offset, measure the volatility and dispersion of the offset data points, and generate the trajectory offset volatility and dispersion evaluation result;

[0014] Utilize the trajectory offset volatility and dispersion evaluation result, and perform correlation analysis with the acceleration change data, using the formula:

[0015]

[0016] Identify the influence of acceleration and trajectory deviation to obtain the distribution result of trajectory deviation characteristics;

[0017] Among them, C represents the distribution value of trajectory deviation characteristics, a i represents the acceleration value of the i-th data point, and Δθ i represents the heading angle deviation of the i-th data point, and n is the total number of data points.

[0018] As a further solution of the present invention, the steps for obtaining the flight trajectory deviation distribution result are specifically as follows:

[0019] Based on the distribution result of the trajectory deviation characteristics, extract the deviation angle data of the position points in the flight trajectory, analyze the angle change between adjacent position points, serialize the angle change in combination with the time axis, assign time marks to the trajectory points and reorder them to obtain the flight trajectory deviation angle sequence;

[0020] Based on the flight trajectory deviation angle sequence, select the flight attitude angle as a reference parameter, calculate the difference between the deviation angle and the attitude angle, analyze the mapping relationship between the trajectory points and the difference value, integrate the difference data, supplement the time axis parameters and perform classification and sorting to generate a trajectory angle deviation difference matrix;

[0021] Based on the trajectory angle deviation difference matrix, extract the difference abnormal points and the corresponding position coordinates, draw a trajectory space model in combination with the trajectory deviation angle sequence, mark the abnormal points in the trajectory model, match the relationship between the deviation difference value and the trajectory point position, and generate the flight trajectory deviation distribution result.

[0022] As a further solution of the present invention, the steps for obtaining the motion deviation correlation matrix are specifically as follows:

[0023] Based on the flight trajectory deviation distribution result, analyze the absolute change rates of the control torque change value and the rudder surface offset value, and screen the data points with change rates exceeding the threshold to obtain the control torque and rudder surface offset data groups;

[0024] Using the control torque and rudder surface offset data groups, extract the change values and analyze the difference amplitude, identify the correlation factors of the data point corresponding relationship, through the formula:

[0025]

[0026] Calculate the correlation score value of each group of data to generate a preliminary correlation score matrix;

[0027] Among them, R represents the correlation score value, α and β are the weight coefficients of the control torque and the rudder surface offset respectively, T k and S k represent the control torque and the rudder surface offset in the k-th group of data respectively, and v is the number of data groups;

[0028] Based on the preliminary correlation scoring matrix, select the data groups with scores higher than the preset threshold, perform normalization processing on the data groups, and generate an action offset correlation matrix.

[0029] As a further solution of the present invention, the steps for obtaining the analysis result of the key action characteristics are specifically as follows:

[0030] Based on the action offset correlation matrix, extract the action offset values and match them with the corresponding correlation parameters, analyze the cumulative distribution of each action offset value, classify and integrate the cumulative distribution values and the correlation parameters, and record the distribution intervals of the high-frequency correlation parameters and the action offsets to generate an action distribution characteristic data set;

[0031] Based on the action distribution characteristic data set, screen the offset intervals of the correlation parameter frequencies in the cumulative distribution, extract the action position data corresponding to the offset intervals, sort the offset characteristics of each group of actions according to the interval size based on the action positions and the offset characteristics, and mark the characteristic values of the offset weights to generate a key action characteristic grouping table;

[0032] Based on the key action characteristic grouping table, analyze the distribution of the weight characteristics in the action sequence, adjust the mapping relationship of the offset characteristic values according to the positions in the action sequence, and integrate the mapped distribution characteristics to generate the analysis result of the key action characteristics.

[0033] As a further solution of the present invention, the steps for obtaining the key action offset range are specifically as follows:

[0034] Based on the analysis result of the key action characteristics, call the initial data set of the offset action distribution range, parse the key characteristic parameters of each action, analyze the trend range of the parameter changes, determine the peak offset points in the action characteristics, and obtain the preliminary analysis data of the action characteristic offsets;

[0035] Based on the preliminary analysis data of the action characteristic offsets, calculate the distribution density of the action offset points, using the formula:

[0036]

[0037] Screen the set of offset points with distribution density exceeding the preset threshold to obtain the high-frequency offset point distribution range;

[0038] where D is the distribution density value, P j is the weight of the offset point, W j is the characteristic weight value of the action offset point, is the mean value of the offset point weights, and m is the total number of offset points;

[0039] Based on the distribution range of the high-frequency offset points and superimposing it with the action characteristic trend range, analyze the change interval of the characteristic parameters, introduce the action offset influence factor, and obtain the key action offset range.

[0040] As a further solution of the present invention, the steps for obtaining the prediction result of the key action offset are specifically as follows:

[0041] Based on the key action offset range, extract the numerical values of the action offset direction and amplitude, classify and label each action data according to the offset direction, assign the offset amplitude values to a continuous interval range, count the number of actions in each classification, and generate a key action offset classification data set;

[0042] Based on the key action offset classification data set, through each group of data in the offset direction classification, count the number of actions in the differential amplitude interval, and integrate and classify according to the offset direction, compare the quantity differences of the differential offset directions within the amplitude range, and generate a key action offset distribution model;

[0043] Based on the key action offset distribution model, extract the interval data of the offset direction and amplitude distribution ratio, sort the action data in a time series according to the interval distribution trend, predict the change range and amplitude interval of the action offset direction in the future time period, and combine the distribution trend and prediction data to generate a key action offset prediction result.

[0044] As a further solution of the present invention, the steps for obtaining the multi-dimensional training allocation scheme of the flight simulation are specifically as follows:

[0045] Based on the key action offset prediction result, extract the key action characteristic parameter set, including the action execution duration parameter, the action offset amplitude parameter, and the action execution accuracy parameter, and generate an initial set of matching key action characteristics and task phases by analyzing the characteristic weights of the parameters matching the task phases;

[0046] Based on the initial set of matching key action characteristics and task phases, by calculating the characteristic matching degree distribution value, using the formula:

[0047]

[0048] Optimize the influence degree of the matching degree value on the rudder surface adjustment distribution, and generate an optimized matching characteristic distribution matrix;

[0049] Among them, M c represents the characteristic matching degree distribution value, T a represents the action execution duration, D a represents the action offset amplitude, P a represents the action execution accuracy, Q 1 is the action offset weight parameter, Q 2is the action precision weight parameter;

[0050] Call the optimized matching characteristic distribution matrix, combine the control torque distribution and the task stage correction characteristics, analyze the rationality of optimizing the rudder surface adjustment and the control torque distribution, and generate a multi-dimensional training allocation scheme for flight simulation by screening the optimal distribution combination.

[0051] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0052] In the present invention, by decomposing the change parameters of the flight trajectory, deeply analyzing the dynamic changes of the heading angle and the acceleration, and extracting the characteristic distribution results of the trajectory deviation, the dynamic evaluation of the trajectory deviation is made more accurate, significantly improving the ability to capture the change law of the flight trajectory. By further matching the influencing factors of the deviation distribution, analyzing the correlation degree between the control torque change and the rudder surface deviation, strengthening the analysis depth of the action characteristics during the flight process, quantifying the key correlation between the action and the trajectory, refining the processing method of the action characteristic analysis, making the extraction of the action characteristic distribution and the dynamic evaluation of the key characteristics more comprehensive, laying a foundation for accurately predicting the distribution range of high-frequency offset actions. In the prediction of high-frequency action offset, comprehensively analyzing the change trend of the action characteristic parameters, further refining the key action offset range, and through interval statistics and distribution analysis, clarifying the differential categories of the offset direction and amplitude, making the dynamic capture of high-frequency offset more practical and accurate. Combining the matching analysis of the training task stage, optimizing the allocation of the rudder surface adjustment and the control torque, making the training allocation of flight simulation more in line with the actual needs, realizing the full-chain coverage of flight data from law discovery to task optimization, not only improving the accuracy of real-time decision-making, but also enhancing the scientific nature of training task allocation, and thus effectively improving flight safety and training efficiency. Brief Description of the Drawings

[0053] Figure 1 is the system flowchart of the present invention;

[0054] Figure 2 is the flowchart of the characteristic distribution result of the trajectory deviation in the present invention;

[0055] Figure 3 is the flowchart of the flight trajectory deviation distribution result in the present invention;

[0056] Figure 4 is the flowchart of the action offset correlation matrix in the present invention;

[0057] Figure 5 is the flowchart of the analysis result of the key characteristics of the action in the present invention;

[0058] Figure 6 is the flowchart of the key action offset range in the present invention;

[0059] Figure 7 It is a flowchart of the prediction result of the key action offset in the present invention;

[0060] Figure 8 It is a flowchart of the multi-dimensional training allocation scheme for flight simulation in the present invention. Specific embodiments

[0061] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0062] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0063] Please refer to Figure 1 , a multi-dimensional data mining and analysis system based on flight simulation includes:

[0064] The flight trajectory characteristic analysis module decomposes the flight trajectory change parameters based on the flight trajectory data recorded by flight simulation, analyzes the dynamic changes of the heading angle and acceleration, evaluates the trajectory offset amplitude and change range, obtains the trajectory offset characteristic distribution result, and generates the flight trajectory deviation distribution result by comparing the flight attitude angles;

[0065] The action characteristic correlation analysis module matches the influencing factors of the deviation distribution based on the flight trajectory deviation distribution result, analyzes the correlation degree between the control torque change and the rudder surface offset, generates the action offset correlation matrix, extracts the distribution characteristics, analyzes the key action characteristics, and generates the key action characteristic analysis result;

[0066] The key action offset prediction module identifies the distribution range of high-frequency offset actions based on the key action characteristic analysis result, analyzes the change trend of the action characteristic parameters, obtains the key action offset range, divides different categories according to the offset direction and amplitude, performs interval statistics and distribution analysis, and generates the key action offset prediction result;

[0067] Based on the prediction results of key action offsets, the training task allocation module analyzes the matching degree between action characteristic data and training task phases, optimizes the training distribution of rudder surface adjustment and control torque, and combines the characteristics of task phases and action correction characteristics to generate a multi-dimensional training allocation plan for flight simulation.

[0068] The trajectory offset characteristic distribution results include offset amplitude distribution, change range distribution, and dynamic characteristic distribution. The flight trajectory deviation distribution results include heading angle deviation distribution, acceleration deviation distribution, and attitude angle deviation distribution. The action offset correlation matrix includes control torque change matrix, rudder surface offset correlation matrix, and action distribution characteristic matrix. The analysis results of key action characteristics include key action distribution characteristics, changes in action characteristic parameters, and analysis of key characteristic trends. The key action offset range includes high-frequency offset action range, offset direction range, and offset amplitude range. The prediction results of key action offsets include offset interval distribution, action distribution statistics, and differential category division. The multi-dimensional training allocation plan for flight simulation includes task phase matching distribution, optimization distribution of rudder surface adjustment, and optimization distribution of control torque distribution.

[0069] Please refer to Figure 2 , and the specific steps for obtaining the trajectory offset characteristic distribution results are as follows:

[0070] Based on the flight trajectory data recorded in the flight simulation, extract the heading angle and acceleration data points, analyze the data at each time point, identify the heading angle offset and acceleration change, calculate the offset and change parameters of each point on the flight trajectory, and generate a list of flight trajectory change parameters;

[0071] Extract the heading angle and acceleration data points from the flight simulation data. By parsing the original data in the flight trajectory record, extract the change parameters of the corresponding heading angle and acceleration in the time series. Perform a difference calculation on the heading angle data to obtain the offset between each time point. Calculate the list of heading offset values at each time point in the time series through the angular amplitude change of the heading angle change value. Use the acceleration data for integration operations, accumulate the discrete acceleration values within each time period, calculate the list of speed changes within the corresponding time period, combine the list of heading offset values, match and normalize the two according to the time point, eliminate the interference caused by uneven time intervals or sampling frequency errors, combine and store the normalized acceleration change and heading angle offset values, and at the same time calculate the offset direction and relative change rate of acceleration in each set of matching data to reflect the dynamic adjustment characteristics of direction and speed in the flight trajectory, and finally generate a list of flight trajectory change parameters.

[0072] Based on the list of flight trajectory change parameters, perform trajectory offset characteristic analysis, focus on the frequency distribution and change range of trajectory offsets, measure the volatility and dispersion of offset data points, and generate an evaluation result of trajectory offset volatility and dispersion;

[0073] Analyze the distribution characteristics of the heading angle offset and acceleration change data. Divide the amplitude values of the offset into intervals, divide the data points into multiple amplitude intervals according to the size of the offset, and statistically analyze the frequency distribution characteristics of each interval. At the same time, calculate the standard deviation and coefficient of variation of the offset amplitude to measure the volatility and dispersion of the offset amplitude within each interval. By comparing the change trends of the offset amplitudes in adjacent intervals, screen out the intervals with high volatility and high dispersion, and further analyze the acceleration changes in each interval. Calculate the maximum, minimum, and average values of the acceleration data, and perform a fitting analysis on the joint distribution characteristics of acceleration and offset through the data within the interval to generate a joint distribution diagram of offset amplitude and acceleration. Combine the joint distribution diagram and organize the calculation results into an evaluation result of the volatility and dispersion of trajectory offset.

[0074] Utilize the evaluation result of the volatility and dispersion of trajectory offset, and conduct a correlation analysis with the acceleration change data. Use the formula:

[0075]

[0076] Identify the influence of acceleration and trajectory offset to obtain the distribution result of trajectory offset characteristics;

[0077] where C represents the distribution value of trajectory offset characteristics, a i represents the acceleration value of the i-th data point, Δθ i represents the heading angle offset of the i-th data point, and n is the total number of data points;

[0078] The advantage of the formula is that by comprehensively multiplying the acceleration and the heading angle offset, combined with the normalization calculation method of the sum of squares, it reduces the influence of outliers on the overall result and ensures the stability and reliability of the offset characteristic calculation;

[0079] The acceleration data extracted from the flight trajectory record is 0.8, 1.2, 1.0, 1.5, 1.3, and the heading angle offset data is 0.05, 0.04, 0.03, 0.06, 0.05. The acceleration data is calculated from the 6-degree-of-freedom motion equation, and its unit is m / s 2 , the data has been filtered to remove interference values. The heading angle offset data is calculated from the 6-degree-of-freedom motion equation, with the unit of radians, and the sampling time interval is 1 second;

[0080] The calculation of the numerator part is as follows:

[0081] The calculation of the denominator part is as follows:

[0082]

[0083] Finally calculated

[0084] The result shows that the comprehensive offset characteristic distribution value is 0.1068, representing the dynamic coupling degree of acceleration and heading angle offset, and can directly reflect the characteristic result of the comprehensive offset trend in the trajectory offset characteristic distribution.

[0085] Please refer to Figure 3 , and the steps for obtaining the flight trajectory deviation distribution result are specifically as follows:

[0086] Based on the trajectory offset characteristic distribution result, extract the offset angle data of the position points in the flight trajectory, analyze the angle changes between adjacent position points, serialize the angle changes in combination with the time axis, assign time marks to the trajectory points and reorder them to obtain the flight trajectory offset angle sequence;

[0087] By dividing the flight path into multiple sampling points with equal time intervals, analyze the offset angle of each sampling point separately, record the offset change amount between adjacent position points, based on the absolute value of the angle change, construct the angle change sequence on the time axis in combination with the time parameter, assign a unique value to the time mark of each position point according to the order of position change, ensure that all data can be strictly sorted according to the time dimension, so as to avoid data overlap or loss, rearrange all the offset angle data according to the time axis, and calculate the change rate between two consecutive time points. The change rate is obtained by dividing the angle difference by the time difference, so as to ensure that the trend of the offset angle changing with time can be captured. Through this processing, the serialization of the flight trajectory offset angle is completed, and a directional identifier is added in the analysis to distinguish different offset directions. Combine the serialized data of each segment to construct the flight trajectory offset angle sequence and output it in the form of a list to complete the structural conversion of the flight trajectory offset data.

[0088] Based on the flight trajectory offset angle sequence, select the flight attitude angle as the reference parameter, calculate the difference between the offset angle and the attitude angle, analyze the mapping relationship between the trajectory points and the difference value, integrate the difference value data, supplement the time axis parameter and perform classification and sorting to generate the trajectory angle offset difference matrix;

[0089] Extract the attitude angle data of the trajectory points and perform linear interpolation on it to ensure the continuity and accuracy of the attitude angle in the time dimension. By calculating the difference between the offset angle and the attitude angle point by point, the two sets of data are corresponded one by one. Generate a new angle difference sequence according to the time stamp, and integrate and store the difference and its corresponding trajectory point position data in matrix form. Conduct a deep analysis of the mapping relationship between the trajectory points and the differences, set the difference threshold range in combination with the data distribution characteristics, and mark the difference points that exceed this range. To further supplement the time dimension parameters of the angle difference data, add the time stamp to the difference matrix, and reorder the trajectory points according to the magnitude of the differences, thereby generating a trajectory angle offset difference matrix. In this way, the fusion processing of the angle offset characteristics and the flight attitude characteristics is completed, which can effectively reflect the differences in trajectory offset and flight attitude changes in different time periods, and form a multi-dimensional classification and sorting result among the trajectory points, time, and differences.

[0090] Based on the trajectory angle offset difference matrix, extract the difference anomaly points and the corresponding position coordinates, draw the trajectory space model in combination with the trajectory offset angle sequence, mark the anomaly points in the trajectory model, match the relationship between the offset difference and the trajectory point position, and generate the flight trajectory deviation distribution result.

[0091] Extract the difference anomaly points that exceed the threshold range in the angle deviation, and extract their corresponding trajectory point position coordinates. Combine the change trend of the trajectory offset angle sequence, and use the interpolation method to perform spatial fitting on the trajectory offset angle at the position of the anomaly points to construct a trajectory space model including the offset angle and the position coordinates. In the model, visualize all the trajectory points according to their position coordinates and offset angle distribution to ensure that the model can accurately display the overall characteristics of the angle offset. For the marked anomaly points, embed them in the model, and add the position coordinate marks and the difference range information of the offset angle for each anomaly point to distinguish the relationship between the anomaly points and the trajectory points. Calculate the distribution density between all the offset differences and the trajectory point positions, and describe the offset characteristics of the flight trajectory in space with the density change. According to the model, match the specific mapping relationship between the offset difference and the trajectory point position, and output the flight trajectory deviation distribution result to intuitively reflect the offset characteristics and risk distribution area of the flight trajectory in space.

[0092] Please refer to Figure 4 , the steps for obtaining the action offset correlation matrix are specifically as follows:

[0093] Based on the flight trajectory deviation distribution result, analyze the absolute change rate of the control moment change value and the rudder surface offset value, and screen the data points whose change rate exceeds the threshold to obtain the control moment and rudder surface offset data groups.

[0094] By calling the control moment change values and rudder surface offset values corresponding to each time node in the flight trajectory record file, two sets of data are extracted respectively and a data set is constructed. Calculate the absolute change rates of the control moment change values and the rudder surface offset values, extract the specific formulas for the change rates and calculate the absolute change rate values for each point in the data group. Based on the calculation results, gradually screen the data points with change rates greater than the specified threshold, rematch the specific corresponding relationship between the control moment change values and the rudder surface offset values from the valid data points, eliminate or replace the missing values or abnormal points in the data to avoid the influence of invalid data on subsequent calculations. By sorting and classifying the matching relationships in the remaining valid data points, ensure that each set of matching data has a clear calculation basis and validity. Finally, output the control moment and rudder surface offset data groups.

[0095] Using the control moment and rudder surface offset data groups, extract the change values and analyze the difference amplitudes, identify the correlation factors of the data point corresponding relationships, through the formula:

[0096]

[0097] Calculate the correlation scoring values for each group of data to generate a preliminary correlation scoring matrix;

[0098] Among them, R represents the correlation scoring value, α and β are the weight coefficients of the control moment and the rudder surface offset respectively, T k and S k represent the control moment and the rudder surface offset in the kth group of data respectively, and v is the number of data groups;

[0099] The advantage of the formula is that by introducing weight parameters to perform weighted processing on the control moment and rudder surface offset data, it can more carefully reflect the influence ratio of the two on the correlation scoring. At the same time, through the processing of the numerator and denominator, ensure that the final scoring value can adapt to the data distribution of different sample scales;

[0100] Calculate the absolute value part of the data: Assume that there are 5 groups of data in the data set, T 1 = 3.2, T 2 = 5.1, T 3 = 2.8, T 4 = 4.6, T 5 = 3.9, S 1 = 1.7, S 2 = 2.5, S 3 = 2.3, S 4 = 3.1, S 5 = 2.8, and calculate the weighted absolute values of each group of data respectively;

[0101] Set the weight parameters: α = 0.6, β = 0.4, corresponding to the control torque and rudder surface deflection weights respectively, and reflect the importance of the two parameters through weight distribution;

[0102] Substitute into the formula for calculation:

[0103] Calculate the internal sum:

[0104] Final result: R = 3.344;

[0105] This result indicates that the average correlation score between the control torque and the rudder surface deflection data is 3.344. Combining with the scoring matrix, it can be used to further analyze the comprehensive impact of the two parameters on the overall deviation.

[0106] Based on the preliminary correlation scoring matrix, select the data groups with scores higher than the preset threshold, perform normalization processing on the data groups, and generate the action offset correlation matrix;

[0107] By calling the specific values of the correlation scores of each group in the preliminary scoring matrix, compare the scoring matrix with the set preset threshold one by one, screen the data groups that meet the conditions according to the rule that the score value is higher than the preset threshold, call the matching data groups that meet the conditions and calculate the normalization processing results of each group of data, process the score value range, so as to uniformly reduce all score values to the standard range, prevent deviations in subsequent calculations caused by differences in data ranges, and through further grouping and integration processing of the normalized score values, reorganize the data into a standard matrix form and output the action offset correlation matrix.

[0108] Please refer to Figure 5 , and the specific steps for obtaining the analysis results of the key action characteristics are as follows:

[0109] Based on the action offset correlation matrix, extract the action offset values and match them with the corresponding correlation parameters, analyze the cumulative distribution of each action offset value, classify and integrate the cumulative distribution values and the correlation parameters, and record the distribution intervals of the high-frequency correlation parameters and the action offsets to generate the action distribution characteristic data set;

[0110] Compare each action offset value one by one according to the category of the associated parameter to form a mapping relationship. Calculate the cumulative distribution of the mapped action offset values, use the increasing cumulative result of the offset value as the distribution index to generate a cumulative distribution sequence. Integrate the cumulative distribution values with the corresponding associated parameter categories. By calculating the cumulative distribution range and frequency density under different associated parameter categories, judge the distribution characteristics of high-frequency associated parameters, and record the distribution interval ranges of high-frequency parameters and the corresponding action offset values. For each offset value within the distribution interval, reorder it using the action time series to ensure that the interval mapping relationship between the offset value and the associated parameter is accurately expressed. Based on the distribution range of the offset value, extract the offset cumulative values of each group of high-frequency associated parameters and output an action distribution characteristic data set to further characterize the distribution law and characteristic trend of action offset.

[0111] Based on the action distribution characteristic data set, screen the offset intervals of the associated parameter frequencies in the cumulative distribution, extract the action position data corresponding to the offset intervals. According to the action position and offset characteristics, sort the offset characteristics of each group of actions by the interval size and mark the characteristic values of the offset weights to generate a key action characteristic grouping table.

[0112] Mark the associated parameters with cumulative frequencies exceeding a specific interval as high-offset characteristic parameters, and further extract the action position data corresponding to the high-frequency offset intervals. By analyzing the action positions within the offset intervals, summarize the offset ranges and frequency characteristics of each action. Combining the distribution interval information already generated in the distribution characteristic data set, sort the offset characteristics of the actions with the interval size as the sorting criterion, and output the order of action offsets from high to low. For the actions within the high-frequency offset intervals, calculate the offset characteristic values, generate the corresponding offset weight values and mark them in the data table. Integrate the weight values with the interval ranges to form a grouping with action positions, offset characteristics and weight values. According to the size of the weight characteristic values, mark the key actions as actions with significant characteristics, and at the same time clearly distinguish the distribution ranges of high-offset actions and ordinary actions to generate a key action characteristic grouping table to illustrate the offset weights of each group of actions in different position intervals and their influence on the overall distribution.

[0113] Based on the key action characteristic grouping table, analyze the distribution of the weight characteristics in the action sequence, adjust the mapping relationship of the offset characteristic values according to the positions in the action sequence, and integrate the mapped distribution characteristics to generate the analysis results of the key action characteristics.

[0114] By matching the positions of the action sequences and the offset weight characteristics of the key actions, an action offset characteristic distribution map is constructed. In the distribution map, the mapping relationship of the offset characteristic values is readjusted according to the weight characteristic values of the actions, so that the offset characteristics of the high-weight actions can more accurately reflect their influence on the action sequences. During the adjustment process, the characteristics of each group of key actions are mapped to their corresponding time-axis positions to form a dynamic change curve of the offset characteristic values. Combining the adjusted distribution characteristic values, each group of actions within the action sequence is reclassified according to the characteristic distribution to ensure that the distribution map can clearly reflect the offset characteristics and weight distribution characteristics of the actions. Finally, the adjusted distribution characteristics are integrated, and the distribution of the key actions in the sequence is presented in the form of a visualization chart to generate the analysis results of the key characteristics of the actions, which are used to describe the spatial distribution of the offset characteristic values in the action sequence and their dynamic influence relationships.

[0115] Please refer to Figure 6 , and the steps for obtaining the offset range of the key actions are specifically as follows:

[0116] Based on the analysis results of the key characteristics of the actions, the initial data set of the distribution range of the offset actions is called, the key characteristic parameters of each action are parsed, the trend range of the parameter changes is analyzed, the peak offset points in the action characteristics are determined, and the preliminary analysis data of the action characteristic offset is obtained;

[0117] First, the time series information is extracted from the action sequence to obtain the continuous amplitude change data of the action at different time points. At the same time, the rate of change of the action speed is calculated. The rate of change of the speed can be calculated by dividing the displacement difference between adjacent time points by the time interval. Further, the amplitude change range of the action is extracted in combination with the amplitude data. Finally, the maximum value, minimum value, corresponding time points, and frequency characteristics of each action are statistically analyzed. By comparing the relative differences between the parameters and the mean value and the upper and lower threshold values of the characteristic range, the peak change points of each action are calculated and their change curves are extracted. According to the peak characteristics, the key characteristic parameters are selected. By judging whether the peak change points of the characteristics meet the key offset characteristic conditions of the parameters, the conditions include that the amplitude peak is greater than several times the mean value, the absolute value of the speed direction change exceeds a certain amplitude threshold, etc. Finally, the set of peak change points that meet the conditions is used as the offset points in the action characteristics, and the preliminary analysis data of the action characteristic offset is obtained.

[0118] Based on the preliminary analysis data of the action characteristic offset, calculate the distribution density of the action offset points, using the formula:

[0119]

[0120] Screen the set of offset points whose distribution density exceeds the preset threshold to obtain the distribution range of high-frequency offset points;

[0121] where D is the distribution density value, P jis the weight of the offset point, W j is the feature weight value of the action offset point is the mean value of the weights of the offset points, and m is the total number of offset points;

[0122] The advantage of the formula is that by introducing the combination of weight distribution and standard deviation to calculate the distribution density, it can effectively screen out high-frequency offset points and accurately quantify their distribution density, thereby further improving the recognition accuracy of high-frequency offset actions and the ability to capture data characteristics;

[0123] P j represents the weight of the offset point, and this value is obtained by comprehensively calculating the parameters of the three dimensions in the offset characteristics. The calculation method is P j = a 1 ·A j + a 2 ·F j + a 3 ·S j , where A j is the amplitude characteristic value, F j is the frequency characteristic value, S j is the speed characteristic value, and the weight coefficients a 1 , a 2 , a 3 are set based on the influence degrees of the three on the offset characteristics in the actual data. Specifically, the weight ratio is determined by analyzing the significance index of the offset points in the dataset, and the weight values satisfy a 1 + a 2 + a 3 = 1, W j represents the feature weight value of the action offset point, which is obtained by normalizing the frequency distribution. The formula is is the mean value of the weights of the offset points, and the formula is

[0124] Substitute the parameter values, and set m = 3. The three groups of offset point feature values are:

[0125] The first group A 1 = 5, F 1 = 3, S 1 = 4;

[0126] The second group A 2 = 6, F 2 = 4, S 2 = 3;

[0127] The third group A 3 = 7, F 3 = 5, S 3 = 2;

[0128] Weight coefficient a 1 = 0.4, a 2 = 0.3, a 3 = 0.3:

[0129] Calculate the weight P of each offset point j :

[0130] P 1 = 0.4·5 + 0.3·3 + 0.3·4 = 4.3;

[0131] P 2 = 0.4·6 + 0.3·4 + 0.3·3 = 4.9;

[0132] P 3 = 0.4·7 + 0.3·5 + 0.3·2 = 5.5;

[0133] Calculate the characteristic weight value W of the offset point j :

[0134] W 1 = 3 / (3 + 4 + 5) = 0.25;

[0135] W 2 = 4 / (3 + 4 + 5) = 0.33;

[0136] W 3 = 5 / (3 + 4 + 5) = 0.42;

[0137] Calculate the mean value

[0138] Calculate the distribution density D:

[0139]

[0140] This result shows that the distribution density value is 5.89, indicating that the distribution density of the offset points in the current data set is relatively high, which can accurately identify the high-frequency offset area and generate the distribution range of high-frequency offset points.

[0141] Based on the distribution range of high-frequency offset points, and superimposing it with the action characteristic trend range, analyze the change interval of characteristic parameters, introduce the action offset influence factor, and obtain the key action offset range;

[0142] Normalize each action characteristic parameter within the trend range through an overlay process. During the normalization process, standardize the parameters in three dimensions: amplitude characteristic, speed characteristic, and frequency characteristic. Normalize the maximum and minimum values of each parameter to the standard range through logarithmic function transformation. Perform standardization operations by mapping the parameters within the standard range to the unit interval. Use the normalization result as the reference parameter for overlay, calculate the change interval of each action characteristic trend range, and further screen the overlaid parameter set by setting the judgment criterion of whether the change interval exceeds the offset influence factor. Finally, obtain the key offset influence factor set in the parameter change interval to generate the key action offset range.

[0143] Please refer to Figure 7 , and the steps for obtaining the key action offset prediction result are specifically as follows:

[0144] Based on the key action offset range, extract the numerical values of the action offset direction and amplitude. Classify and label each action data according to the offset direction, assign the offset amplitude values to the continuous interval range, count the number of actions in each classification, and generate a key action offset classification data set.

[0145] Define the direction data of each action offset as the deviation angle with respect to the trajectory reference direction. A positive value represents a clockwise offset, and a negative value represents a counterclockwise offset. Classify and label the offset direction data of each group of actions, divide the action data into multiple classification groups according to the directionality, such as the "clockwise offset group" and the "counterclockwise offset group". Perform continuous interval division on each action offset amplitude data, allocate the amplitude values according to the interval range, and set the size of each interval as a fixed amplitude difference to ensure the continuity and comparability of all action offset data in terms of amplitude. Count the actions in the same amplitude interval in each classification group, and record the number of actions in different amplitude intervals in each direction classification. By integrating the action statistics results of all classification groups, generate a key action offset classification data set that includes the offset direction, amplitude interval, and number of actions. This data set can intuitively reflect the distribution law of action offset in terms of direction and amplitude, providing a data basis for subsequent offset distribution analysis.

[0146] Based on the key action offset classification data set, count the number of actions in the differential amplitude intervals through each group of data in the offset direction classification, integrate and classify them according to the offset direction, compare the quantity differences in the amplitude range of different offset directions, and generate a key action offset distribution model.

[0147] Precisely compare the number of amplitude interval actions in each direction classification. During the statistical process, first calculate the cumulative number of actions in each amplitude interval for each offset direction, then horizontally compare the number of actions in the same amplitude interval in different direction classifications, mark the significantly different interval ranges, integrate the amplitude intervals and offset directions to generate a direction-amplitude distribution map reflecting the difference in the number of actions. By analyzing this distribution map, the concentration and sparsity of the number of actions in each amplitude range in different differential offset directions can be clearly identified. Deeply integrate the distribution relationship between each group of offset directions and amplitude ranges, and finally generate a key action offset distribution model. This model contains the complete mapping relationship of direction, amplitude, and the number of actions, and can reflect the similarities and differences in action distribution under different directions and amplitudes, providing an important basis for subsequent trend analysis.

[0148] Based on the key action offset distribution model, extract the interval data of the offset direction and amplitude distribution ratio, sort the action data in a time series according to the interval distribution trend, predict the change range and amplitude interval of the action offset direction in the future time period, and combine the distribution trend and prediction data to generate the key action offset prediction result.

[0149] Dynamically sort the number of actions in each offset direction and amplitude interval through the time axis, continuously process the action data in a time series to ensure that the distribution ratio can reflect the law of change over time. During the processing, fit the distribution trend of the offset direction, combine the proportional change of the amplitude interval, predict the change range of the action offset direction and the dynamic distribution of the amplitude interval in the future time period. By constructing a time series trend chart of the offset direction and amplitude, clarify the trend range of each offset direction and amplitude in the future time period. Combine the prediction data to estimate the distribution density of the action offset in the future time period, output the ranges of the high-frequency interval and low-frequency interval, and integrate the distribution trend and prediction data to generate the key action offset prediction result. This result intuitively presents the direction change trend and amplitude distribution range of the action offset in the future time period, providing strong support for flight trajectory adjustment and abnormal action warning.

[0150] Please refer to Figure 8 , the specific steps for obtaining the multi-dimensional training allocation scheme for flight simulation are as follows:

[0151] Based on the key action offset prediction result, extract the key action characteristic parameter set, including the action execution duration parameter, action offset amplitude parameter, and action execution accuracy parameter. By analyzing the characteristic weights of the parameter matching with the task stage, generate the initial set of key action characteristic parameters matching the task stage.

[0152] First, by parsing the original data of motion offset prediction, the data is segmented into multiple time intervals corresponding to task phases. The changing trend of the motion execution duration within each interval is analyzed to extract the motion execution duration parameter. Subsequently, the motion offset amplitude is normalized. Combining the standard deviation interval of the maximum and minimum amplitudes, the significantly offset-fluctuating part is selected, abnormal data is removed, and the final offset amplitude is calculated. Then, through the standardization of the measurement of execution accuracy, the accuracy measurement data is converted into characteristic parameter values that conform to the model. Next, the offset amplitude, execution duration, and execution accuracy are weighted and calculated to obtain the preliminary matching relationship between the task phase and motion characteristics, and this matching relationship is expressed in matrix form for subsequent further optimization calculations. Finally, the initial set of matching key motion characteristic parameters and task phases is generated.

[0153] Based on the initial set of matching key motion characteristic parameters and task phases, by calculating the distribution value of the characteristic matching degree, using the formula:

[0154]

[0155] Optimize the influence degree of the matching degree value on the rudder surface adjustment distribution, and generate an optimized matching characteristic distribution matrix;

[0156] Among them, M c represents the distribution value of the characteristic matching degree, T a represents the motion execution duration, D a represents the motion offset amplitude, P a represents the motion execution accuracy, Q 1 is the motion offset weight parameter, Q 2 is the motion accuracy weight parameter;

[0157] The benefit of the formula is to improve the calculation accuracy of the matching degree value by introducing the motion accuracy parameter and adjusting the weight term. At the same time, by combining the sum of the squares of the motion execution duration and the offset amplitude, the balance of the distribution is ensured. The weight parameter can be dynamically adjusted according to the data characteristics, enhancing the flexibility of the calculation;

[0158] The motion execution duration parameter T a = 5.4, obtained by measuring the actual operation time in the flight simulation task point by point and calculating the mean value;

[0159] The motion offset amplitude parameter D a = 3.2, obtained by calculating the difference in the motion position offset in the simulation data and taking the mean value of its fluctuation range;

[0160] The motion execution accuracy parameter P a = 0.85, obtained by standardizing the motion execution data and calculating the mean value after taking the deviation ratio of each motion execution;

[0161] Offset weight parameter Q 1 = 1.2, which is set according to the weight distribution curve after analyzing the influence of the historical action offset amplitude in different task stages;

[0162] Precision weight parameter Q 2 = 0.8, which is obtained by analyzing the historical data distribution relationship of precision in task characteristic matching and normalizing;

[0163] Calculate the numerator part:

[0164] T a ·P a = 5.4·0.85 = 4.59, D a ·Q 1 = 3.2·1.2 = 3.84, |4.59 - 3.84| = 0.75;

[0165] Calculate the denominator part:

[0166]

[0167] Calculate the characteristic matching degree distribution value:

[0168] The result shows that the characteristic matching degree distribution value M c = 0.1186 can quantify and optimize the weight influence in the matching characteristic distribution, and be used as the reference basis for the subsequent rudder surface adjustment distribution.

[0169] Call the optimized matching characteristic distribution matrix, combine the control moment distribution and the task stage correction characteristics, analyze the rationality of the optimized rudder surface adjustment and the control moment distribution, and generate a multi-dimensional training allocation scheme for flight simulation by screening the optimal distribution combination;

[0170] First, extract the parameter subset most relevant to the current task phase objective from the optimized feature distribution matrix. Screen the feature weights corresponding to each parameter combination row by row in the matrix, and set a feature threshold to eliminate the non-conforming subsets. Then, cross-map the selected subsets with the matrix in the control moment distribution model. Through the linear adjustment of the control moment distribution matrix, calculate the corrective influence of each parameter on the rudder surface adjustment. Next, dynamically correct the selected feature weights in sequence, and gradually update the weight values and matrix content of each group of data. Finally, accumulate and classify the results that meet the feature combination in the matrix by combining the task phase corrected feature weights, so as to generate a multi-dimensional training allocation plan for flight simulation. The multi-dimensional training allocation plan mainly uses the data collected from flight simulation to optimize the flight training task through multi-dimensional data mining technology. Specifically, this plan relies on in-depth analysis of flight trajectories, assessment of the correlation between actions and trajectories, and offset prediction based on action characteristics. Through these complex analyses and predictions, a more accurate training task allocation can be formulated, which not only improves the efficiency and safety of flight training, but also enables pilots to conduct more targeted training for specific flight actions and trajectory offsets. In practice, the multi-dimensional training allocation plan specifically includes the adjustment of training subjects. For example, according to the results of flight data analysis, the training allocation of rudder surface adjustment and control moment can be adjusted to ensure that each training targets the weaknesses of the pilot and the upcoming flight challenges. Such a system not only improves the adaptability of individual training, but also improves the overall training quality, making the training more in line with various situations that may be encountered in actual flight.

[0171] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A multidimensional data mining and analysis system based on flight simulation, characterized in that: The system comprises: The flight trajectory characteristic analysis module decomposes the flight trajectory change parameters based on the flight trajectory data recorded by the flight simulation, analyzes the dynamic changes of the heading angle and acceleration, evaluates the trajectory deviation amplitude and change range, obtains the trajectory deviation characteristic distribution results, and generates the flight trajectory deviation distribution results by comparing the flight attitude angles; The action characteristic correlation analysis module matches the deviation distribution influencing factors based on the flight trajectory deviation distribution result, analyzes the correlation between the control torque change and the rudder surface deviation, generates the action deviation correlation matrix, extracts the distribution characteristics, analyzes the key action characteristics, and generates the action key characteristic analysis results; The key action offset prediction module identifies the distribution range of high-frequency offset actions based on the key characteristic analysis results of the actions, analyzes the change trend of action characteristic parameters, obtains the key action offset range, divides the differential categories according to the offset direction and amplitude, performs interval statistics and distribution analysis, and generates key action offset prediction results; The training task allocation module analyzes the matching degree between the action characteristic data and the training task stage based on the key action offset prediction results, optimizes the training distribution of the rudder adjustment and the control torque, combines the task stage characteristics with the action correction characteristics, and generates a multi-dimensional training allocation plan for flight simulation.

2. The multidimensional data mining and analysis system based on flight simulation according to claim 1, characterized in that: The steps for obtaining the trajectory deviation characteristic distribution result are specifically as follows: Based on the flight trajectory data recorded by the flight simulation, the heading angle and acceleration data points are extracted, the data at the time point is analyzed, the heading angle offset and acceleration change are identified, the offset and change parameters of each point of the flight trajectory are calculated, and a list of flight trajectory change parameters is generated; Based on the flight trajectory change parameter list, performing trajectory deviation characteristic analysis, focusing on the frequency distribution and variation range of trajectory deviation, measuring the volatility and dispersion of deviation data points, and generating trajectory deviation volatility and dispersion evaluation results; The trajectory deviation volatility and dispersion evaluation results are used to perform correlation analysis with the acceleration change data, using the formula: Identify the influence of acceleration and trajectory deviation and obtain the trajectory deviation characteristic distribution result; Where C represents the distribution value of the trajectory deviation characteristic, a i represents the acceleration value of the i-th data point, Δθ i Represents the heading angle offset of the i-th data point, and n is the total number of data points.

3. The multidimensional data mining and analysis system based on flight simulation according to claim 2, characterized in that: The steps for obtaining the flight trajectory deviation distribution result are specifically as follows: Based on the trajectory deviation characteristic distribution result, the deviation angle data of the position points in the flight trajectory are extracted, the angle changes between adjacent position points are analyzed, the angle changes are serialized in combination with the time axis, and the trajectory points are assigned time tags and reordered to obtain the flight trajectory deviation angle sequence; Based on the flight trajectory offset angle sequence, the flight attitude angle is selected as a reference parameter, the difference between the offset angle and the attitude angle is calculated, the mapping relationship between the trajectory point and the difference is analyzed, the difference data is integrated, the time axis parameters are supplemented and classified and sorted, and a trajectory angle offset difference matrix is ​​generated; Based on the trajectory angle offset difference matrix, the difference outliers and corresponding position coordinates are extracted, the trajectory space model is drawn in combination with the trajectory offset angle sequence, the outliers are marked in the trajectory model, the offset difference and the trajectory point position relationship are matched, and the flight trajectory deviation distribution result is generated.

4. The multidimensional data mining and analysis system based on flight simulation according to claim 3, characterized in that: The steps for obtaining the action offset association matrix are specifically as follows: Based on the flight trajectory deviation distribution result, analyzing the absolute change rate of the control torque change value and the rudder surface offset value, screening the data points whose change rate exceeds the threshold value, and obtaining the control torque and rudder surface offset data group; By using the control torque and rudder surface offset data set, the change value is extracted and the difference amplitude is analyzed to identify the correlation factor of the corresponding relationship of the data points, and the formula is: Calculate the correlation score value of each set of data and generate a preliminary correlation score matrix; Among them, R represents the correlation score value, α and β are the weight coefficients of control torque and rudder surface deviation respectively, T k and S k represent the control torque and rudder surface offset in the kth data set, respectively, and v is the number of data sets; Based on the preliminary correlation score matrix, a data group with a score higher than a preset threshold is selected, and the data group is standardized to generate a motion offset correlation matrix.

5. The multidimensional data mining and analysis system based on flight simulation according to claim 4, characterized in that: The steps for obtaining the analysis results of the key characteristics of the action are specifically as follows: Based on the motion offset association matrix, the motion offset values ​​are extracted and matched with the corresponding association parameters, the cumulative distribution of each motion offset value is analyzed, the cumulative distribution values ​​and the association parameters are classified and integrated, and the distribution intervals of the high-frequency association parameters and the motion offset are recorded to generate a motion distribution characteristic data set; Based on the action distribution characteristic data set, the offset interval of the associated parameter frequency in the cumulative distribution is screened, the action position data corresponding to the offset interval is extracted, and the offset characteristics of each group of actions are sorted by interval size through the action position and offset characteristics, and the characteristic value of the offset weight is marked to generate a key action characteristic grouping table; Based on the key action characteristic grouping table, the distribution of weight characteristics in the action sequence is analyzed, the mapping relationship of the offset characteristic value is adjusted according to the position of the action sequence, and the mapped distribution characteristics are integrated to generate the action key characteristic analysis result.

6. The multidimensional data mining and analysis system based on flight simulation according to claim 5, characterized in that: The steps for obtaining the offset range of the key action are specifically as follows: Based on the analysis results of the key characteristics of the action, call the initial data set of the offset action distribution range, parse the key characteristic parameters of each action, analyze the trend range of parameter changes, determine the peak offset point in the action characteristics, and obtain preliminary analysis data of the action characteristic offset; Based on the preliminary analysis data of the motion characteristic offset, the distribution density of the motion offset points is calculated using the formula: Filter the offset point set whose distribution density exceeds a preset threshold to obtain the distribution range of high-frequency offset points; Where D is the distribution density value, P j is the weight of the offset point, W j is the feature weight of the action offset point, P is the mean of the offset point weight, and m is the total number of offset points; Based on the distribution range of the high-frequency offset points and superimposed with the motion characteristic trend range, the variation interval of the characteristic parameters is analyzed, the motion offset influencing factor is introduced, and the key motion offset range is obtained.

7. The multidimensional data mining and analysis system based on flight simulation according to claim 6, characterized in that: The steps for obtaining the key action offset prediction result are specifically as follows: Based on the key action offset range, extract the values ​​of the action offset direction and amplitude, classify and mark each action data according to the offset direction, assign the offset amplitude value to a continuous interval range, count the number of actions in each classification, and generate a key action offset classification data set; Based on the key action offset classification data set, the number of actions within the differentiated amplitude range is counted through each group of data in the offset direction classification, and integrated classification is performed according to the offset direction, and the number differences of differentiated offset directions within the amplitude range are compared to generate a key action offset distribution model; Based on the key action offset distribution model, the interval data of the offset direction and amplitude distribution ratio are extracted, the action data are sorted in time series according to the interval distribution trend, the change range and amplitude interval of the action offset direction in the future time period are predicted, and the key action offset prediction result is generated by combining the distribution trend and the predicted data.

8. The multidimensional data mining and analysis system based on flight simulation according to claim 7, characterized in that: The steps for obtaining the multi-dimensional training allocation scheme for flight simulation are specifically as follows: Based on the key action offset prediction result, extract the key action characteristic parameter set, including the action execution duration parameter, the action offset amplitude parameter, and the action execution accuracy parameter, and generate the initial set of key action characteristic parameters and task stage matching by analyzing the characteristic weights of the matching parameters and task stages; Based on the initial set of key action characteristic parameters and task stage matching, the characteristic matching degree distribution value is calculated using the formula: Optimize the influence of the matching value on the rudder adjustment distribution, and generate an optimized matching characteristic distribution matrix; Among them, M c represents the characteristic matching degree distribution value, T a Indicates the execution time of the action, D a Indicates the amplitude of the motion deviation, P a Indicates the action execution accuracy, Q1 is the action offset weight parameter, and Q2 is the action accuracy weight parameter; The optimized matching characteristic distribution matrix is ​​called, and the rationality of optimizing the control surface adjustment and the control moment distribution is analyzed and combined with the control moment distribution and the mission phase correction characteristics. By screening the optimal distribution combination, a multi-dimensional training allocation plan for flight simulation is generated.

Images