Aircraft landing performance analysis method and system based on runway perception-airborne QAR data fusion

Abstract

The invention relates to an aircraft landing performance analysis method and system based on runway perception-airborne QAR data fusion. The method comprises the following steps: collecting data during aircraft landing; key factors influencing the aircraft landing distance are analyzed, and variables remarkably influencing the aircraft landing distance are determined; establishing an aircraft landing behavior whole-process simulation model, obtaining simulation data, combining the simulation data with field data, and jointly establishing a mapping model from runway perception-airborne QAR data to an aircraft landing distance; according to runway perception-airborne QAR data before aircraft landing, an aircraft landing distance prediction value is obtained through calculation by using a mapping model, probability distribution of the aircraft landing distance is obtained through a probability model according to the aircraft landing distance prediction value, and then aircraft landing performance is analyzed. According to the method, the accuracy, adaptability and real-time performance of the aircraft landing distance prediction model are remarkably improved, scientific support is provided for airport aircraft landing operation safety, and safe operation of an airport runway is ensured.

Description

Technical Field

[0001] The present invention relates to the field of airport engineering, and in particular to an aircraft landing performance analysis method and system based on runway perception-airborne QAR data fusion, which are used to improve the safety of aircraft landing operation. Background Art

[0002] The landing phase is a critical phase in flight and also a phase where accidents are more likely to occur. On rainy days, the runway surface of an airport is prone to local or uneven accumulation of water. Under the action of the water film, the anti-skid performance of the runway is reduced, and the braking and maneuvering performance of the aircraft will also be reduced, leading to frequent accidents. Therefore, real-time and accurate analysis of aircraft landing performance based on the aircraft landing distance is crucial to ensure the safety of aircraft landing and runway operation and improve runway operation efficiency.

[0003] In the prior art, the following methods are usually used to predict the landing distance of an aircraft and analyze the landing performance of an aircraft: (1) Prediction method based on mathematical model: This method establishes a mathematical model of the required landing distance and its influencing factors under different runway pavement conditions, and makes corrections to obtain a prediction model for the landing distance of an aircraft on a wet runway or a contaminated runway. Although this method can make a preliminary estimate of the required landing distance of an aircraft, it ignores the physical process of the aircraft landing process, resulting in low prediction accuracy when facing extreme weather and complex meteorological conditions. (2) Prediction method based on empirical model: This method is based on the ESDU model and uses aircraft speed and empirical coefficients to predict the landing performance of an aircraft. However, the empirical coefficients in this empirical model are based on assumptions and experimental data, and have poor adaptability to different application scenarios. In addition, the model is difficult to update in a timely manner, resulting in low accuracy.

[0004] Therefore, the existing technology has many problems in aircraft landing distance prediction, such as over-simplification of the model, low adaptability, low accuracy and low real-time performance. In view of these problems, the present invention proposes an aircraft landing performance analysis method and system based on runway perception-airborne QAR data fusion, which significantly improves the accuracy, adaptability and real-time performance of the aircraft landing distance prediction model by combining simulation models and machine learning methods, and provides scientific support for the safety of airport aircraft landing operations. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide an aircraft landing performance analysis method and system based on runway perception-airborne QAR data fusion. By combining simulation models and machine learning methods, the accuracy, adaptability and real-time performance of the aircraft landing distance prediction model are significantly improved, providing scientific support for the safety of airport aircraft landing operations and ensuring the safe operation of airport runways.

[0006] The technical solution adopted to achieve the purpose of the present invention is an aircraft landing performance analysis method based on runway perception-airborne QAR data fusion, the method comprising:

[0007] S1, collect runway perception data, airborne QAR data and aviation meteorological data when the aircraft lands;

[0008] S2. Analyze the key factors that affect the aircraft landing distance and conduct sensitivity analysis to determine the variables that have a significant impact on the aircraft landing distance;

[0009] S3. Based on the runway perception data, airborne QAR data and aviation meteorological data, the variables that significantly affect the aircraft landing distance are used as key simulation elements to establish a simulation model of the entire aircraft landing behavior process;

[0010] S4. Setting the working conditions of the aircraft during landing in the whole process simulation model of the aircraft landing behavior, obtaining simulation data, combining the simulation data with the field data, and jointly establishing a mapping model of runway perception-airborne QAR data to aircraft landing distance, wherein the working conditions of the aircraft during landing include the variables that significantly affect the aircraft landing distance;

[0011] S5. According to the runway perception-airborne QAR data before the aircraft lands, a predicted value of the aircraft landing distance is calculated by using a mapping model, and the predicted value of the aircraft landing distance is used to obtain a probability distribution of the aircraft landing distance through a probability model, thereby analyzing the aircraft landing performance.

[0012] In addition, an aircraft landing performance analysis system based on runway perception-airborne QAR data fusion is also provided, which includes:

[0013] Data acquisition module, used to collect runway perception data, airborne QAR data and aviation meteorological data when the aircraft lands;

[0014] Aircraft landing variable analysis module, used to conduct sensitivity analysis on key factors of aircraft landing distance and determine the variables that have a significant impact on aircraft landing distance;

[0015] A modeling module, used to establish a simulation model of the entire landing behavior of the aircraft based on the data collected by the data acquisition module and taking the variables that significantly affect the landing distance of the aircraft as key simulation elements;

[0016] A mapping model establishment module is used to combine the simulation data obtained by the full-process simulation model of the aircraft landing behavior with the field data to jointly establish a mapping model from runway perception-airborne QAR data to aircraft landing distance;

[0017] The aircraft landing performance analysis module is used to use the mapping model to predict the aircraft landing distance according to the real-time monitoring runway perception-airborne QAR data, and obtain the probability distribution of the aircraft landing distance through the predicted value through the probability model, and finally determine the landing performance of the aircraft according to the aircraft landing distance probability function.

[0018] The present invention has the following advantages:

[0019] 1. The present invention effectively improves the accuracy of aircraft landing distance prediction under complex conditions and extreme weather conditions by combining field data with simulation data.

[0020] 2. The present invention deeply combines data with physical models, thereby improving the accuracy and adaptability of aircraft landing distance prediction.

[0021] 3. The present invention can realize real-time and accurate prediction of aircraft landing distance through runway perception data, airborne QAR data and aviation meteorological data. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 The present invention is a flow chart of the aircraft landing performance analysis method and system based on runway perception-airborne QAR data fusion.

[0023] Figure 2 Schematic diagram of the process for establishing a simulation model of aircraft landing behavior throughout the entire process. DETAILED DESCRIPTION

[0024] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] like Figure 1 As shown, the present invention provides an aircraft landing performance analysis method based on runway perception-airborne QAR data fusion, the method comprising:

[0026] S1. Collect runway perception data, airborne QAR data and aviation meteorological data when the aircraft lands, including:

[0027] S1.1. Connect the aircraft's internal QAR equipment to the data extraction tool, use professional software to read, decode and convert, and obtain various operation and status data of the aircraft during flight and landing. Extract key variables from the airborne QAR data, including speed, braking status, aircraft attitude, aircraft weight, etc. Airborne QAR data is usually stored in binary or proprietary formats, and is converted into a readable format (such as CSV or Excel) using specific airborne QAR data decoding software to facilitate subsequent analysis to select appropriate flight samples, select multiple aircraft landing data at different water film thicknesses, and ensure that the data includes a variety of different environmental conditions, such as sunny days, rainy days, and slippery runways.

[0028] S1.2. Obtain the meteorological data of the aircraft at landing through the meteorological station, airport meteorological department or third-party meteorological service provider, including wind direction and speed, temperature, humidity and rainfall, etc. Organize the meteorological data into a standard format, with each landing sampling time point having corresponding wind speed, temperature, humidity, etc., and ensure that the sampling time point is synchronized with the airborne QAR data.

[0029] S1.3. Obtain runway surface condition data through water film thickness sensors installed at the airport or on-site testing, including average water film thickness, runway friction coefficient, pavement temperature, sampling time, etc.

[0030] S1.4. Match the runway surface condition data, meteorological data and airborne QAR data according to the sampling time points to obtain time series data. Clean the collected data, use interpolation to fill in missing data and correct outliers. Then standardize all data. Use principal component analysis to reduce the dimension of time series data, find the principal components in the data, and map high-dimensional data to low-dimensional space.

[0031] S2. Analyze the key factors that affect the aircraft landing distance and conduct sensitivity analysis to determine the variables that have a significant impact on the aircraft landing distance, including:

[0032] S2.1. Use statistical tools to calculate the Pearson correlation coefficient between different variables and the aircraft landing distance. The value r is between -1 and 1. The closer the value is to 1 or -1, the stronger the linear correlation. Then calculate the significance level of the variable, represented by p. The smaller the p value, the more significant the correlation between the variable and the aircraft landing distance. It is considered that when the p value is less than 0.05, the correlation is significant, and the variables that have a significant impact on the aircraft landing distance are preliminarily screened out. The formula for calculating the Pearson correlation coefficient and significance level is:

[0033]

[0034] Among them, x i represents the value of the variable data point, Represents the value of the variable data point; y i represents the value of the aircraft landing distance data point, represents the average of the aircraft landing distance data points; r represents the correlation coefficient; n represents the number of samples. The corresponding p value is determined based on the t statistic.

[0035] S2.2. Establish a multiple regression model, calculate the regression coefficient of each variable, and then obtain the sensitivity coefficient through the standardization method to screen out the variables with the greatest impact on the aircraft landing distance. The formula is:

[0036]

[0037] in, represents the standardized regression coefficient, Represents the independent variable x j The standard deviation, σ y Represents the standard deviation of the dependent variable, aircraft landing distance.

[0038] Finally, several variables with great influence (significant influence) on the aircraft landing distance were selected, including wind speed, water film thickness, aircraft landing weight and target speed during the final approach phase.

[0039] S3, such as Figure 2 As shown in the figure, based on runway perception data, airborne QAR data and aviation meteorological data, a simulation model of the entire landing behavior of an aircraft is established, including:

[0040] S3.1. Establish the aircraft fuselage model, tire and landing gear geometry model, pavement model and aerodynamic system model of the simulation model. At the same time, establish the control subsystem to form the whole machine assembly, and perform the braking and rolling simulation of the aircraft landing process. The modeling is implemented as follows:

[0041] When establishing a multi-body dynamics model, it is necessary to first assume that all parts of the model are rigid bodies, establish models of different parts of the aircraft in CATIA, and establish communicators between different parts to form an aircraft assembly. The aircraft assembly mainly includes six main parts: fuselage subsystem, front landing gear subsystem, front landing gear tire subsystem, main landing gear subsystem, main landing gear tire subsystem and brake subsystem.

[0042] The landing gear structure in the landing gear geometric model is simplified into mass units, beam units and connector units for multi-body dynamics simulation, in which the elastic and damping parameters of the connector unit are set to simulate the landing gear hydraulic system.

[0043] The pavement in the pavement model is modeled with solid units. The pavement structure is composed of 32 cube structures of 4×8, each with a size of 5m×4.5m to cover the main wheel action area during aircraft take-off and landing. The width of the joint between blocks is 0.008m. The foundation layer of the pavement is modeled with the equivalent stiffness method, and grounding springs are established at the node positions to simulate the stiffness of the foundation soil.

[0044] The aerodynamic system model uses the water flow state equation to model the water film distribution. The Mie-GRUNEISEN state equation is:

[0045]

[0046] In the formula, P and P H are the water pressure and the Hugoniot shock pressure, e and e respectively. H are the specific internal energy and energy reference value, Γρ and Γ0 are the material parameters of water after impact and the initial values ​​of the material parameters, ρ and ρ0 are the density of water after impact and the initial density, respectively. s and U p is the shock wave velocity and particle wave velocity, c0 is the propagation speed of sound waves in the fluid, s is U s and U p The slope between .

[0047] The fluid-solid coupling method is used for modeling, and the water film is modeled based on the Abaqus CEL method. The water film uses the Euler grid, and the runway uses the Lagrangian grid. The water film thickness information cannot be used directly in the virtual prototype simulation. Therefore, it is also necessary to convert the water film thickness information into a pavement combination coefficient that can be used by the simulation software.

[0048] This embodiment integrates the simulation tools XFlow and Abaqus into one process through "Xflow-Abaqus joint simulation" to achieve coupled simulation of aerodynamics (fluid mechanics) and structural mechanics. During the joint simulation process, XFlow passes the fluid calculation results (such as surface pressure, fluid load) to Abaqus as boundary conditions or loading inputs of the structural model. After completing the structural calculation, Abaqus returns the structural deformation or displacement information to XFlow for updating the geometry and boundary conditions of the fluid domain.

[0049] The aircraft control system in the control subsystem controls the lateral and longitudinal behavior of the aircraft through Simulink in MATLAB, establishes a two-dimensional fuzzy controller based on error and error change rate, and establishes an aircraft lateral control system and ABS control system.

[0050] S3.2, in the simulation, set the speed, attitude and weight of the aircraft when it touches the runway, simulate the effect of water film thickness on tire friction and landing distance, adjust the air resistance according to the speed and angle of attack, consider the effect of wind speed and wind direction, and finally select several variables (wind speed, water film thickness, aircraft landing weight and target speed during the final approach phase) that have a greater impact on the aircraft landing distance as the key elements of the simulation;

[0051] S3.3. Repeatedly run the simulation to analyze the landing distance and braking deceleration under key factors, adjust the water film thickness, tire friction, braking system and aerodynamic parameters, ensure that the simulation results are consistent with the actual airborne QAR data and gradually optimize the model.

[0052] S4. According to the simulation model of the whole process of aircraft landing behavior, simulation data is obtained, and the simulation data is combined with the field data to jointly establish a mapping model of runway perception-airborne QAR data to aircraft landing distance, including the following steps:

[0053] S4.1. Obtain multi-condition simulation data through the full-process simulation model of aircraft landing behavior established in S3. The specific operations are as follows:

[0054] Different operating conditions are set through the simulation model of aircraft landing behavior to generate a wide range of simulation data samples. These operating conditions include the significant variables determined in S2, such as wind speed, water film thickness, aircraft landing weight, and target speed during the final approach phase, including changes in single factors and multiple factors of these variables. Through multiple simulation runs, landing distance data under different conditions are obtained. These simulation data not only cover common flight conditions, but also consider the aircraft landing performance under extreme conditions to ensure the diversity and breadth of the data. For example, by setting the water film thickness and wind speed, the simulation model simulates the impact of these factors on aircraft tire friction, braking deceleration, and landing distance, thereby ensuring that the simulation environment is consistent with the actual situation.

[0055] S4.2, the simulation data generated in step S4.1 is combined with the real data to jointly train the mapping model. The specific operations are as follows:

[0056] The simulation data generated in step S4.1 is combined with the actual QAR field data to form a complete data set, and the variables with significant impacts finally selected in step S2 are used as input feature data to train the XGBoost mapping model so that the model can accurately learn the relationship between key variables and landing distance. During the model training process, these significant variables are focused on, and the prediction accuracy of the mapping model is optimized by tuning the model's hyperparameters.

[0057] In order to eliminate the dimensional differences of different variables, this embodiment normalizes all data using the Z-score normalization formula:

[0058]

[0059] Where x is the original data, μ is the mean of the variable, and σ is the standard deviation of the variable.

[0060] The standardized data set is convenient for unified machine learning training. The standardized data set is then divided into training set, validation set and test set, usually 70% for model training, 15% for validation, and 15% for testing. After the data processing is completed, the XGBoost algorithm is used to build a mapping model to learn the relationship between airborne QAR data (such as speed, wind speed, braking force, etc.) and the landing distance of the aircraft. The XGBoost model is based on the gradient boosting tree and uses the mean square error (MSE) as the loss function. The formula is:

[0061]

[0062] in, is the landing distance predicted by the model, y i is the actual landing distance. By minimizing the loss function, the model prediction ability is gradually optimized.

[0063] S4.3. Optimize the mapping model parameters in S4.2. The specific operations are as follows:

[0064] By adjusting the hyperparameters of XGBoost, including learning rate, maximum tree depth, subsampling ratio, etc., the model performance was optimized through k-fold cross validation, and the mean square error (MSE) and determination coefficient (R 2 ) value, iteratively optimized according to feedback, and finally used the test set to verify the prediction accuracy of the model under different working conditions to ensure that the model can accurately predict the landing distance.

[0065] S5. Before the aircraft lands, the runway sensing data, airborne QAR data and aviation meteorological data are obtained, and the predicted value of the aircraft landing distance is calculated using the mapping model obtained in S4. Since there is a certain deviation in the data, the aircraft landing distance obtained by the mapping model is passed through the probability model to obtain the probability distribution of the aircraft landing distance. Assuming that the aircraft landing distance obeys the normal distribution, its predicted value obeys the probability density function of the normal distribution. The probability distribution and confidence interval can be used to judge the probability of aircraft landing risk in real time, and then evaluate whether the aircraft landing performance is good.

Claims

1. An aircraft landing performance analysis method based on runway perception-airborne QAR data fusion, characterized in that: include: S1, collect runway perception data, airborne QAR data and aviation meteorological data when the aircraft lands; S2. Analyze the key factors that affect the aircraft landing distance and conduct sensitivity analysis to determine the variables that have a significant impact on the aircraft landing distance; S3. Based on the runway perception data, airborne QAR data and aviation meteorological data, the variables that significantly affect the aircraft landing distance are used as key simulation elements to establish a simulation model of the entire aircraft landing behavior process; S4. Setting the working conditions of the aircraft during landing in the whole process simulation model of the aircraft landing behavior, obtaining simulation data, combining the simulation data with the field data, and jointly establishing a mapping model of runway perception-airborne QAR data to aircraft landing distance, wherein the working conditions of the aircraft during landing include the variables that significantly affect the aircraft landing distance; S5. According to the runway perception-airborne QAR data before the aircraft lands, a predicted value of the aircraft landing distance is calculated by using a mapping model, and the predicted value of the aircraft landing distance is used to obtain a probability distribution of the aircraft landing distance through a probability model, thereby analyzing the aircraft landing performance.

2. The aircraft landing performance analysis method based on runway perception-airborne QAR data fusion according to claim 1 is characterized in that: Step S1 includes: S1.

1. Connect the airborne QAR device to the data extraction tool, obtain various operation and status data of the aircraft during flight and landing through reading, decoding and conversion, and finally save it in a readable format; S1.

2. Obtain aviation meteorological data at the time of aircraft landing through weather stations, airport meteorological departments or third-party meteorological service providers, including wind direction and speed, temperature, humidity and rainfall; S1.

3. Obtain runway surface condition data, including water film thickness, runway friction coefficient, pavement temperature and sampling time, through water film thickness sensors installed at the airport or on-site testing; S1.

4. Match the airborne QAR data, aviation meteorological data and runway surface condition data according to the sampling time points, clean the collected data, use interpolation to fill in missing data and correct outliers, and then standardize all data.

3. The aircraft landing performance analysis method based on runway perception-airborne QAR data fusion according to claim 2 is characterized in that: Step S2 includes: S2.

1. Use statistical tools to calculate the correlation coefficients between different variables and aircraft landing distance, and preliminarily screen out the variables that have a significant impact on aircraft landing distance; S2.

2. Establish a multivariate regression model, calculate the sensitivity coefficients of all significant influencing variables, and evaluate the impact of each variable on the aircraft landing distance; S2.

3. Finally, the variables with the highest impact on the aircraft landing distance are screened out, including wind speed, water film thickness, aircraft landing weight and target speed during the final approach phase.

4. According to the aircraft landing performance analysis method based on runway perception-airborne QAR data fusion according to claim 3, step S3 comprises: S3.

1. Establish the aircraft fuselage model, tire and landing gear geometry model, pavement model and aerodynamic system model of the simulation model, and establish the control subsystem to form the whole aircraft assembly, and perform the braking and rolling simulation of the aircraft landing process; S3.2, in the simulation, set the speed, attitude and weight of the aircraft when it touches the runway, simulate the effect of water film thickness on tire friction and load braking force, adjust air resistance according to speed and angle of attack, consider the effect of wind speed and wind direction, and finally select several variables with a high degree of influence on the landing distance of the aircraft as the key elements of the simulation; S3.

3. Repeatedly run simulations to analyze key factors such as landing distance and braking deceleration, adjust water film thickness, tire friction, braking system and aerodynamic parameters, ensure that simulation results are consistent with actual airborne QAR data and gradually optimize the model.

5. The aircraft landing performance analysis method based on runway perception-airborne QAR data fusion according to claim 4 is characterized in that: The step S3.1 comprises: Establish models of different aircraft components in CATIA, and establish communicators between different components to form an aircraft assembly. The aircraft assembly includes a fuselage subsystem, a nose landing gear subsystem, a nose landing gear tire subsystem, a main landing gear subsystem, a main landing gear tire subsystem, and a brake device subsystem; The structure of the landing gear geometric model is subjected to multi-body dynamics simulation through mass units, beam units and connector units, wherein elasticity and damping parameters are set for the connector unit to simulate the landing gear hydraulic system; The pavement model is modeled by solid units. The pavement structure is composed of 32 cube structures of 4×8, each of which is 5m×4.5m in size to cover the main wheel action area during aircraft take-off and landing. The width of the joint between blocks is 0.008m. The foundation layer of the pavement is modeled by the equivalent stiffness method, and grounding springs are established at the node positions to simulate the stiffness of the foundation soil. The aerodynamic system model includes water film distribution modeling through the water flow state equation. The Mie-GRUNEISEN state equation is: In the formula, P and P H are the water pressure and the Hugoniot shock pressure, e and e respectively. H are the specific internal energy and energy reference value, Γ ρ and Γ0 are the material parameters of water after impact and the initial values ​​of the material parameters, ρ and ρ0 are the density of water after impact and the initial density, respectively. s and U p is the shock wave velocity and particle wave velocity, c0 is the propagation speed of sound waves in the fluid, s is U s and U p The slope between The modeling was carried out by fluid-solid coupling method, and the water film modeling was carried out based on the Abaqus CEL method. The water film used the Euler grid, and the runway used the Lagrangian grid. The control subsystem realizes the control of the lateral and longitudinal behaviors of the aircraft through Simulink in MATLAB, establishes a two-dimensional fuzzy controller based on error and error change rate, and establishes an aircraft lateral control system and an ABS control system.

6. The aircraft landing performance analysis method based on runway perception-airborne QAR data fusion according to claim 5 is characterized in that: Step S4 includes: S4.

1. According to the obtained full-process simulation model of aircraft landing behavior, different operating conditions for aircraft landing are set, including wind speed, water film thickness, aircraft landing weight and target speed during the final approach phase. The effects of water film thickness, wind speed and aircraft braking method on landing distance are tested by single factor and multi-factor combination, and multiple simulations are run to generate simulation data samples covering a wide range of conditions. S4.

2. Use the simulation data and field data as input, standardize and divide the data set, use the XGBoost algorithm to build a mapping model, train the model to learn the relationship between airborne QAR data and landing distance, and adjust the hyperparameters through the validation set to optimize the model performance; S4.

3. Optimize model performance by adjusting hyperparameters and using cross-validation to evaluate the MSE and R of the model. 2 The model is iterated and optimized based on the feedback, and finally the test set is used to verify the prediction accuracy of the model under different working conditions to ensure that the model can accurately predict the landing distance.

7. An aircraft landing performance analysis system based on runway perception-airborne QAR data fusion, characterized in that: include: Data acquisition module, used to collect runway perception data, airborne QAR data and aviation meteorological data when the aircraft lands; Aircraft landing variable analysis module, used to conduct sensitivity analysis on key factors of aircraft landing distance and determine the variables that have a significant impact on aircraft landing distance; A modeling module, used to establish a simulation model of the entire landing behavior of the aircraft based on the data collected by the data acquisition module and taking the variables that significantly affect the landing distance of the aircraft as key simulation elements; A mapping model establishment module is used to combine the simulation data obtained by the full-process simulation model of the aircraft landing behavior with the field data to jointly establish a mapping model from runway perception-airborne QAR data to aircraft landing distance; The aircraft landing performance analysis module is used to use the mapping model to predict the aircraft landing distance according to the real-time monitoring runway perception-airborne QAR data, and obtain the probability distribution of the aircraft landing distance through the predicted value through the probability model, and finally determine the landing performance of the aircraft according to the aircraft landing distance probability function.

8. The aircraft landing performance analysis system based on runway perception-airborne QAR data fusion according to claim 7, characterized in that: The whole process simulation model of aircraft landing behavior sets the working conditions of the aircraft during landing to obtain simulation data. The working conditions of the aircraft during landing include the variables that significantly affect the landing distance of the aircraft, including wind speed, water film thickness, aircraft landing weight and target speed during the final approach phase.

Images