Flight test method for global dynamic aerodynamic modeling of fixed-wing aircraft

Abstract

The invention relates to the technical field of aircraft aerodynamic model modeling, in particular to a flight test method for fixed-wing aircraft global dynamic aerodynamic modeling, which comprises the following steps: arranging a sensor on an aircraft, manually controlling the aircraft to fly, and optimizing control parameters according to sensor response data. Enabling the aircraft to autonomously fly by adopting the optimized control parameters; determining whether the autonomous flight of the aircraft meets an identification test condition based on the preset trajectory and the actual trajectory of the autonomous flight; when identification test conditions are met, the aircraft is controlled to carry out multiple times of alternate level flight and excitation flight, and sensor response data are obtained to serve as identification test data; determining an independent variable and a to-be-identified parameter, and based on the identification test data, establishing a nonlinear aerodynamic model by constructing a regression quantity, iteratively screening the regression quantity and a least square method; errors in pneumatic modeling can be reduced, and the test efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of aircraft aerodynamic force model modeling, in particular to a flight test method for global dynamic aerodynamic force modeling of a fixed-wing aircraft. BACKGROUND

[0002] In order to meet the growing military and civilian demand, in the face of more complex flight environment, the development of aircraft has become the core of engineering to promote the innovation of aviation technology and enhance the national strategic competitiveness. The development of aircraft generally includes a sequential and iterative process: starting from wind tunnel test or numerical simulation method to obtain aerodynamic model, then simulation development, control law design and flight test are carried out in turn, in order to meet the demand, the aerodynamic model and control law are iteratively optimized. Therefore, obtaining accurate aerodynamic model is the key step to support the iteration of aerodynamic shape, control law design and performance verification of aircraft.

[0003] The data required for aircraft aerodynamic modeling can generally be obtained by wind tunnel test, numerical simulation and flight test. Among them, wind tunnel test is widely used in large projects, especially for some advanced aircraft with complex geometry, which can provide a reliable source of data and accurately evaluate the stability and control characteristics of the aircraft. However, due to the need to make scaled models, build complex test benches and calibrate high-precision sensors, the resource cost is high, and the cycle is often long. With the continuous development of computer technology and numerical algorithm, numerical simulation method has been widely used in aircraft design, and even replaced part of the wind tunnel test, but for some complex flow field scenes of unsteady separation or multi-physical field coupling, its accuracy may not meet the modeling requirements. Flight test refers to the flight of aircraft in real flight environment. Since wind tunnel test and numerical simulation are difficult to truly reproduce some complex movements and conditions in flight process, the data obtained by flight test has higher authenticity and reliability. These data not only can verify and correct the results of wind tunnel test and numerical simulation, but also are beneficial to build a more comprehensive aerodynamic model. However, real flight also faces high cost and high risk, which needs to invest a lot of funds in personnel, equipment and airspace coordination, and needs to consider the risks of crash, personnel casualties and equipment loss when testing new aircraft and verifying new technology. In view of the advantages and disadvantages of different aerodynamic modeling methods, a combination strategy of multiple methods should be adopted in the development process of aircraft, and the most suitable modeling method should be selected flexibly according to different types of aircraft.

[0004] In order to use real data for aerodynamic modeling, flight test and system identification algorithm can be combined to reduce the scale and improve the efficiency of flight test through reasonable test design, and to obtain accurate nonlinear aerodynamic model with as little data as possible.

[0005] Aircraft system identification algorithms design excitation inputs and record dynamic responses to construct high-precision aerodynamic models based on flight test data, realistically reflecting flight dynamic characteristics. Aircraft system identification can be carried out at both the frequency domain and time domain levels. The core of the frequency domain method is to utilize the frequency response function (FRF) and spectral data, while the time domain method is based on time-series responses. Commonly used basic methods include least squares, stepwise regression, and maximum likelihood estimation. Furthermore, multi-level software support frameworks such as ESIMA, SIDPAC, and RealSysid have been developed based on these methods. In recent years, aircraft system identification technology has been put into practical application. In parameter optimization, based on state-space models and prediction error method (PEM) technology, system identification experiments were conducted on the Sentera Vireo aircraft, focusing on solving the overparameterization problem of the lateral dynamics model. In nonlinear dynamic identification, from a time-domain analysis perspective, the Sequential Threshold Least Squares (STLSQ) algorithm and the Sparse Identification Nonlinear Dynamics (SINDy) method were used to conduct system identification experiments on the Skywalker X8 aircraft, achieving efficient sparse modeling of aerodynamic damping terms. In lightweight platform adaptation, addressing challenges such as weakened disturbance immunity and difficulties in low-frequency mode identification in flying-wing aircraft, existing technologies include frequency-domain system identification methods that combine open-loop and closed-loop approaches. Currently, most related system identification experiments rely on manual excitation input, and their automation level needs improvement. Summary of the Invention

[0006] In view of the above problems, the present invention provides a flight test method for global dynamic aerodynamic modeling of fixed-wing aircraft, which solves the technical problems of significant human error and insufficient efficiency in aerodynamic modeling in the prior art.

[0007] This invention provides a flight test method for global dynamic aerodynamic modeling of fixed-wing aircraft, comprising the following steps: Step S1: Install sensors on the aircraft, manually control the aircraft to fly, optimize the control parameters based on the sensor response data, and use the optimized control parameters to enable the aircraft to fly autonomously. Step S2: Determine whether the autonomous flight of the aircraft meets the identification test conditions based on the preset trajectory and the actual trajectory of autonomous flight. If the identification test conditions are not met, the experiment ends. Step S3: When the identification test conditions are met, control the aircraft to perform multiple alternating level flight and stimulated flight to obtain sensor response data as identification test data; Step S4: Determine the independent variables and the parameters to be identified. The independent variables include: airflow angle, attitude angular velocity, and rudder deflection; the parameters to be identified include: aerodynamic force and aerodynamic torque. Step S5: Based on the identified experimental data, a nonlinear aerodynamic model is established by constructing regression parameters, iteratively screening regression parameters, and using the least squares method. The nonlinear aerodynamic model is used to calculate the parameters to be identified based on the independent variables.

[0008] Preferably, in step S1, the step of manually controlling the flight of the aircraft specifically includes: the pilot inputting commands to make the aircraft perform 20 rolls and pitches in succession; The specific steps for optimizing control parameters based on sensor response data include: using the ArduPilot automatic parameter tuning algorithm to optimize PID controller parameters online, determining the control gain of the PID controller, and obtaining optimized control parameters.

[0009] Preferably, step S2 specifically includes: Step S2-1: The aircraft flies to the preset hovering point and performs a hovering action. If the actual hovering trajectory matches the preset hovering trajectory, the next verification task is performed; otherwise, the experiment ends. Step S2-2: The aircraft follows the preset waypoints and flies along a rectangular route. If the actual route trajectory matches the preset rectangular route trajectory, it is determined that the identification test conditions are met.

[0010] Preferably, step S3 specifically includes: In automatic flight mode, the aircraft enters the predetermined test route and maintains level flight. After a set time, the aircraft performs longitudinal excitation. After the excitation ends, the rudder deflection returns to the reference state, and the aircraft performs free response. After the response ends, the flight control system re-intervenes and resumes level flight. After a period of time, the aircraft performs lateral excitation. After completing the free response, it resumes level flight again and continues to fly along the route until it re-enters the test route. The above test process is repeated multiple times. The longitudinal and lateral excitations are implemented using an excitation control system, which includes a state monitor, an excitation trigger, and a rudder deflection superposition unit. The specific steps for the aircraft to execute the longitudinal or lateral excitations include: (1) Perform configuration initialization, including mapping the servo control channel, setting the flag variables indicating the flight phase and defining the parameters of the excitation signal; (2) The status monitor continuously monitors the flight status. When the status monitor confirms that the aircraft has been flying stably for a preset time, it sends a start signal to the excitation trigger. (3) After receiving the start signal, the excitation trigger starts timing and sends a start signal to the rudder deflection superimposed unit. After the timing reaches the preset time, the excitation trigger sends a stop signal. After receiving the start signal, the rudder deflection superimposition unit superimposes the excitation rudder deflection amplitude onto the reference rudder deflection calculated by the flight control according to the preset excitation signal, and performs rudder surface deflection; the rudder deflection superimposition unit stops working after receiving the stop signal.

[0011] Preferably, in step S4, the airflow angle includes the angle of attack. and sideslip angle The attitude angular velocity includes the dimensionless roll angular velocity. Dimensionless pitch angular velocity and dimensionless yaw rate The rudder deflection includes the elevator deflection angle. and aileron deflection ; The aerodynamic force includes the drag coefficient. Lateral force coefficient and lift coefficient The aerodynamic torque includes the rolling moment coefficient. Pitch moment coefficient and yaw moment coefficient .

[0012] Preferably, step S5 specifically includes: Step S5-1: Determine the constraints of the candidate regressors and determine the original set of regressors based on the constraints of the candidate regressors; Step S5-2: Process the original set of regressors using the orthogonalization method to obtain the orthogonalized set of candidate regressors; Step S5-3: Based on statistical criteria and stopping rules, regression quantities are gradually added or deleted from the orthogonalized candidate regression quantity set, and the unknown parameters of the nonlinear aerodynamic model are estimated using the least squares method, finally obtaining the nonlinear aerodynamic model represented by the screened regression quantities.

[0013] Preferably, in step S5-1, the constraint on the candidate regressor is: Longitudinal independent variables include 、 、 The horizontal independent variables include 、 、 、 Among the vertical independent variables, 、 、 The maximum order was determined to be 4, 3, and 3 respectively; the total order of the regressor formed by all vertical independent variables did not exceed 5; among the horizontal independent variables, 、 、 The maximum order is determined to be 3, 2, and 2 respectively; the total order of the regressor formed by all horizontal independent variables does not exceed 4. Step S5-2 specifically includes: The orthogonalized set of candidate regressors is constructed using the following expression:

[0014]

[0015] in, Indicates the first An orthogonalized candidate regressor Indicates the first One original regression quantity, Indicates the first The original regression quantity at the th... Orthogonalization coefficients in the direction of the orthogonalized candidate regressors Indicates the first An orthogonalized candidate regressor express and The inner product, express and The inner product, This represents the total number of original regression values.

[0016] Preferably, step S5-3 specifically includes: Step S5-3-1: Initialize the expression of the nonlinear aerodynamic model to contain only offset terms. ; Step S5-3-2: For each orthogonalized candidate regressor ,calculate The mean-reduced correlation coefficient between the measured value z of the dimensionless aerodynamic coefficient and the value of the aerodynamic coefficient. Choose the one with the largest mean-to-correlation coefficient. Incorporate the current nonlinear aerodynamic model and calculate the model's coefficient of determination. Changes, local statistics and predicted squared difference ; If local statistics If it exceeds the critical value of 20, then it is confirmed that... Incorporate the current nonlinear aerodynamic model; otherwise, Remove from the current nonlinear aerodynamic model; like If the change is greater than 0.5%, then it is confirmed that... Incorporate the current nonlinear aerodynamic model; otherwise, Remove from the current nonlinear aerodynamic model; If orthogonalized candidate regressors are added or removed, the least squares method is used to estimate the unknown parameters, and the latest estimated parameters are obtained. ; Step S5-3-3: Return to step S5-3-2 until... The change is less than 0.5% or PSE Stop candidate regressors when their value is at its minimum. By iterating through the data, we finally obtain the nonlinear aerodynamic model represented by the selected regression quantities.

[0017] Preferably, in step S5-3-2, the determination coefficient Local statistics and predicted squared difference The expressions are as follows:

[0018]

[0019]

[0020] in, Denotes the coefficient of determination. Represents the sum of squares of regression. This represents the total sum of squares of deviations. Represents the sum of squared residuals. Represents the estimated parameters transpose, Represents the regressor matrix transpose, Represents the measured value of the dimensionless aerodynamic coefficient. Indicates the number of samples. This represents the average value of z. express transpose, This represents a local statistic. Indicates the first , Let the variance of the j-th estimated parameter be denoted as . This represents the predicted squared difference. This means that PSE is defined as the expression on the right. This represents the response value estimated by the model. Indicates transpose. This represents the upper bound of the mean squared error of the prediction. This indicates the number of items in the current model.

[0021] Preferably, in step S5-3-2, the process of solving for the latest estimated parameters... The expression is:

[0022] It represents the dimensionless aerodynamic coefficient. , , , The measured value is calculated using the following expression:

[0023] in, Indicates the estimated parameters, This indicates that the inverse matrix is ​​obtained. Indicates the total mass of the aircraft. These represent the angular velocity components of the three axes, Indicates dynamic pressure. Indicates the wing reference area. Indicates thrust. These represent the angular velocity components of the three axes, Indicates triaxial angular acceleration, Indicates wingspan, Indicates the mean aerodynamic chord length. , , These represent the moments of inertia of the three axes, It represents the product of inertia.

[0024] Compared with the prior art, the present invention has at least the following beneficial effects: (1) This invention achieves dynamic flight testing without human intervention by deploying sensors on the aircraft and utilizing autonomously optimized control parameters, directly collecting and extracting dynamic response data of the aircraft from actual flight. The acquired data can fully reflect the complex nonlinear aerodynamic characteristics of the aircraft in real aerodynamic environments, thus providing an effective basis for accurate modeling and analysis.

[0025] (2) In response to the design requirements of aerodynamic identification flight tests for steady state and excitation input, this invention designs an autonomous identification experimental algorithm with steady state judgment and random trim rudder deflection register functions in the modeling step. During the flight test, it can autonomously identify and switch between level flight and excitation phases, realize batch automatic triggering and repeated execution of excitation maneuvers, effectively avoid the uncertainty caused by manual intervention, improve the standardization of the test process and the consistency of data collection, and ensure the high quality of identification test data.

[0026] (3) In view of the highly nonlinear aerodynamic characteristics exhibited by aircraft when performing maneuvers, this invention introduces a screening and pruning mechanism for aerodynamic model structures. During the identification and modeling process, regression terms that can accurately describe the nonlinear aerodynamic characteristics can be automatically screened, which greatly improves the accuracy and generalization ability of the model, thereby effectively ensuring the consistency and repeatability of the model description and meeting the dynamic modeling requirements under complex flight maneuvers. Attached Figure Description

[0027] The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of the invention.

[0028] Figure 1 The flowchart of the flight test method for global dynamic aerodynamic modeling of fixed-wing aircraft provided by the present invention is shown.

[0029] Figure 2 This is a layout diagram of the flight platform equipment provided by the present invention.

[0030] Figure 3 This is a schematic diagram of the amplitude and power spectrum of the excitation signal provided by the present invention.

[0031] Figure 4 The identification test flowchart provided by the present invention.

[0032] Figure 5 A schematic diagram of the autonomous incentive control framework provided by the present invention.

[0033] Figure 6 This is a schematic diagram of the screening and tailoring mechanism for the aerodynamic model structure provided by the present invention.

[0034] Attached reference numerals: 1- Pitot tube with wind vane, 2- Pitot tube bracket, 3- 6s battery, 4- Airspeed indicator, 5- Data transmission, 6- Flight control, 7- Atmospheric data conversion module, 8- GPS, 9- 3s battery, 10- Parachute compartment, 11- Electronic speed controller, 12- Receiver. Detailed Implementation

[0035] To better understand the above-described objectives, features, and advantages of the present invention, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other. Furthermore, the present invention can be implemented in other ways different from those described herein; therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0036] This invention, based on a time-domain analysis framework of system identification and combined with autonomous flight test data, provides a flight test method for global dynamic aerodynamic modeling of fixed-wing aircraft. Based on a flight test platform, the ArduPilot automatic parameter tuning algorithm is used to optimize PID controller parameters online, enabling the aircraft to achieve autonomous flight capability. Then, an autonomous excitation control algorithm is deployed to enable the aircraft to autonomously execute excitation maneuvers, avoiding human intervention. Finally, nonlinear characteristics are introduced through Taylor series expansion, and the model structure is determined and unknown parameters are estimated based on the system identification algorithm, thereby constructing high-precision longitudinal and lateral aerodynamic models. This invention innovatively simplifies the aircraft system identification test process, significantly improves test repeatability, and enhances the efficiency and accuracy of nonlinear aerodynamic modeling of aircraft.

[0037] To illustrate the effectiveness of the method proposed in this invention, the following detailed description of the above technical solution is provided through a specific embodiment, such as... Figure 1 As shown, a flight test method for global dynamic aerodynamic modeling of fixed-wing aircraft is disclosed. The specific implementation steps are as follows: Step S1: Install sensors on the aircraft, manually control the aircraft to fly, optimize the control parameters based on the sensor response data, and use the optimized control parameters to enable the aircraft to fly autonomously. This step first equips the aircraft with the ability to collect data and autonomously execute flight missions. The onboard commercial flight control system, Global Navigation Satellite System (GNSS), and atmospheric data acquisition module can collect data such as angular velocity, acceleration, airflow angle, and airspeed during flight, which are then used for subsequent aerodynamic modeling. Next, the roll and pitch controller parameters are adaptively optimized based on rapid roll and pitch attitude changes input by the pilot. The control parameters are then verified using hovering and trajectories to ensure stable aircraft flight, as detailed below.

[0038] The structure of the aircraft of the present invention is as follows: Figure 2 As shown, the aircraft has a pitot tube bracket 2 at its front end, which holds a pitot tube 1 with a wind vane. The aircraft contains a 6s battery 3, an airspeed indicator 4, a data transmission system 5, a flight control system 6, an air data conversion module 7, a GPS 8, a 3s battery 9, a parachute compartment 10, an electronic speed controller 11, and a receiver 12. The air data conversion module 7, powered by the 3s battery 9, is used to convert the airflow angle data collected by the pitot tube 1 with the wind vane. The 6s battery 3 powers the airspeed indicator 4, flight control system 6, GPS 8, electronic speed controller 11, and receiver 12.

[0039] Deploying the above equipment on the flight platform enables efficient acquisition of three-axis acceleration components. Three-axis angular velocity components airspeed V, airflow angle The ability to collect flight data such as PWM values ​​for each servo channel. These are the angle of attack and the sideslip angle, respectively.

[0040] The airborne commercial flight control system, Global Navigation Satellite System (GNSS), and atmospheric data acquisition module can collect data such as angular velocity, acceleration, airflow angle, and airspeed during flight, which can be used for subsequent aerodynamic modeling. In some embodiments, the flight controller 6 is equipped with an accelerometer and a gyroscope, capable of acquiring three-axis acceleration and three-axis angular velocity. The airspeed tube 1, airspeed meter 4, and atmospheric data conversion module 7 are capable of acquiring airspeed and airflow angle. To capture high-frequency dynamics and improve model identification accuracy, the sampling rate of each flight data is set to 100Hz.

[0041] In this step, the pilot manually controls the aircraft's flight. Specifically, the pilot inputs commands to make the aircraft perform 20 rapid rolls and pitches in succession. The sensors collect the corresponding response data, and the control parameters are optimized based on the response data to determine a set of control gains.

[0042] In some embodiments, the ArduPilot automatic parameter tuning algorithm can be used to optimize the PID controller parameters online, determine the control gain of the PID controller, and use the optimized control parameters to control the aircraft, enabling the aircraft to have autonomous flight capabilities.

[0043] Step S2: Determine whether the autonomous flight of the aircraft meets the identification test conditions based on the preset trajectory and the actual trajectory of autonomous flight. If the identification test conditions are not met, the experiment ends. After completing the parameter optimization in the previous steps, the aircraft performs autonomous flight, meaning the remote control device commands the aircraft to switch to automatic flight mode. This step uses hovering and zonal flight to verify whether the aircraft's autonomous flight meets the identification test conditions.

[0044] The specific steps include: (1) The aircraft flies to the preset hovering point and performs a hovering action. If the actual hovering trajectory matches the preset hovering trajectory, the next verification task will be performed; otherwise, the experiment will end. (2) The aircraft follows the preset waypoints and flies along the rectangular route. If the actual route trajectory matches the preset rectangular route trajectory, it is determined that the identification test conditions are met.

[0045] In some embodiments, the verification steps of the present invention further include: to ensure flight safety, the flight control program sets a fault protection mechanism: when the aircraft loses contact with the remote control device or the flight distance exceeds the safe range, the system automatically triggers a return to the preset point and performs a hovering; in other emergencies, the pilot can switch back to the manual flight model at any time to make manual intervention.

[0046] Step S3: When the identification test conditions are met, control the aircraft to perform multiple alternating level flight and stimulated flight to obtain sensor response data as identification test data; In this step, an autonomous flight test is conducted according to the system identification requirements. The aircraft's baseline state is set to level flight, and then the preset excitation control is automatically executed on the set flight path. The excitation control system consists of three parts: a state monitor, an excitation trigger, and a rudder deflection superimposed unit. When the state monitor detects that the aircraft has been autonomously flying at level speed on the set flight path for a predetermined time, it sends a signal to the excitation trigger to start the excitation. Then, the rudder deflection superimposed unit adds the excitation amplitude to the reference rudder surface according to the preset excitation design. After the set time is reached, the excitation trigger stops the excitation, as described in detail below.

[0047] like Figure 4 As shown, the identification test procedure is as follows: In automatic flight mode, the aircraft enters the predetermined test route and maintains level flight; after a set time, the aircraft performs longitudinal excitation, and after the excitation ends, the rudder deflection returns to the reference state, and the aircraft performs free response; after the response ends, the flight control system re-intervenes to control and resumes level flight; after a period of time, the aircraft performs lateral excitation, completes free response, and resumes level flight again, continuing to fly along the route until it re-enters the test route, and repeats the above test process.

[0048] The present invention employs an excitation control system to achieve the longitudinal and lateral excitation. Figure 5 The excitation control system of the present invention is shown, which includes a state monitor, an excitation trigger and a rudder deflection superimposed unit.

[0049] The specific steps to achieve the aforementioned vertical or lateral excitation include: (1) Perform configuration initialization, including mapping the servo control channel, setting the flag variables indicating the flight phase and defining the parameters of the excitation signal; The mapped servo control channel refers to the explicit assignment of control signals output by the flight controller to the corresponding physical servo, so that each servo can act according to the signals of the designated channel.

[0050] The flag variables for the flight phase are variables that distinguish the current state of the aircraft, and may include: preparation for excitation, start of longitudinal excitation, control surface maintain reference state, start of lateral excitation, and excitation end.

[0051] The parameters of the excitation signal include the excitation signal waveform, pulse width, and amplitude. To ensure sufficient dynamic response of the aircraft during testing, the longitudinal excitation signal waveform of this invention is a 3-2-1-1 multi-step signal, and the lateral excitation signal waveform is a double-step signal. The waveforms and power spectra of the 3-2-1-1 multi-step signal and the double-step signal are as follows: Figure 3 As shown.

[0052] (2) The status monitor continuously monitors the flight status. When the status monitor confirms that the aircraft has been flying stably for a preset time, it sends a start signal to the excitation trigger. (3) After receiving the start signal, the excitation trigger starts timing and sends a start signal to the rudder deflection superimposed unit. After the timing reaches the preset time, the excitation trigger sends a stop signal. After receiving the start signal, the rudder deflection superimposition unit superimposes the excitation rudder deflection amplitude onto the reference rudder deflection calculated by the flight control according to the preset excitation signal (a 3-2-1-1 multi-step signal in the longitudinal direction and a double step signal in the lateral direction), and performs rudder surface deflection; the rudder deflection superimposition unit stops working after receiving the stop signal.

[0053] After completing the above-mentioned alternating level flight and stimulated flight, the sensor response data is acquired as identification test data for subsequent aerodynamic model modeling and optimization.

[0054] Step S4: Determine the independent variables and the parameters to be identified. The independent variables include: airflow angle, attitude angular velocity, and rudder deflection; the parameters to be identified include: aerodynamic force and aerodynamic torque. This step determines the input independent variables and the output parameters to be identified for the nonlinear aerodynamic model, as detailed below.

[0055] The parameter to be identified in this invention is the aerodynamic force of the aircraft. and aerodynamic torque To allow for independent analysis and applicability to different measurement systems, dimensionless coefficients can be used, as shown in equations (2.1) and (2.2). Therefore, the problem is transformed into determining the drag coefficient. Lateral force coefficient Lift coefficient Rolling torque coefficient Pitch moment coefficient and yaw moment coefficient The expression for the dimensionless aerodynamic coefficient.

[0056] (2.1) (2.2) in, Indicates aerodynamic force, Indicates aerodynamic torque. Indicates dynamic pressure. Indicates the wing reference area. Indicates wingspan, Indicates the mean aerodynamic chord length. Indicates the drag coefficient. Indicates the lateral force coefficient. Indicates the lift coefficient. Indicates the rolling moment coefficient. Indicates the pitching moment coefficient. This represents the yaw moment coefficient.

[0057] Based on dimensional analysis, the aerodynamic coefficients are modeled and assumed. Since this invention is for low-speed testing of full-size aircraft, the Mach number and Reynolds number are not considered. The functional form of the aerodynamic coefficients can be obtained as shown in equation (2.3).

[0058] (2.3) in, Indicates that the index is aerodynamic coefficient, Indicates that the index is A function of the aerodynamic coefficient. These are the angle of attack and the sideslip angle, respectively. For the rudder deflection angle, To stabilize the shaft rotation speed, For characteristic length, This is airspeed. These represent the indices for drag, lateral force, lift, roll, pitch, and yaw coefficients, respectively.

[0059] In this step, the present invention first determines the independent variables of the input to the nonlinear aerodynamic model, which are used to construct regression terms and iteratively filter regression terms in subsequent steps.

[0060] This invention is based on the functional form of aerodynamic coefficients, using airflow angle, attitude angular velocity, and rudder deflection as independent variables, wherein the airflow angle includes the angle of attack. and sideslip angle Attitude angular velocity includes dimensionless roll angular velocity. Dimensionless pitch angular velocity and dimensionless yaw rate Rudder deflection includes elevator deflection angle. and aileron deflection .

[0061] Among them, the vertical independent variables include , , The horizontal independent variables include , , , .

[0062] Step S5: Based on the identified experimental data, a nonlinear aerodynamic model is established by constructing regression parameters, iteratively screening regression parameters, and using the least squares method. The nonlinear aerodynamic model is used to calculate the parameters to be identified based on the independent variables.

[0063] In this step, the present invention constructs longitudinal and lateral nonlinear aerodynamic models of the aircraft based on the obtained experimental data. To improve the standardization and unified comparison of aerodynamic analysis, the aerodynamic model is transformed into a model with six dimensionless aerodynamic coefficients, and nonlinear characteristics are introduced into the model structure through Taylor series expansion. To filter out excessive nonlinear terms that may be introduced by Taylor series expansion and potential data collinearity problems, orthogonalization techniques are used to generate a candidate set of regressors. Then, according to the selected statistical criteria and stopping rules, regressors are gradually added to and removed from the candidate set of regressors, ultimately resulting in a concise and explanatory aerodynamic model.

[0064] When an aircraft performs large or rapidly changing maneuvers, nonlinear characteristics can be introduced through Taylor series expansion, retaining some nonlinear terms to extend the linear model. (Using drag coefficients...) For example, it can be expanded into the polynomial form shown in equation (2.4).

[0065] (2.4) in, Indicates the drag coefficient. This represents the drag coefficient under reference conditions. express , Indicates the increment of angle of attack. express , Indicates pitch angular velocity, Indicates reference airspeed. express , express , express .

[0066] Because Taylor expansion introduces too many nonlinear terms, they need to be filtered. If too many terms are retained, the accuracy of the estimated parameters will decrease; if too few terms are retained, the description will be insufficient.

[0067] Since it is difficult to determine the model structure from numerous possibilities through theoretical analysis, a system identification algorithm is introduced to design an aerodynamic model structure screening and pruning mechanism, enabling it to accurately describe the nonlinear characteristics of aerodynamics. Figure 6 The workflow of the screening and tailoring mechanism for aerodynamic model structures is demonstrated, and is described in detail below.

[0068] (1) Generate the original set of regression quantities First, determine the constraints on the candidate regressors, as follows: Longitudinal independent variables include 、 、 The horizontal independent variables include 、 、 、 Among the vertical independent variables, 、 、 The maximum order was determined to be 4, 3, and 3 respectively; the total order of the regressor formed by all vertical independent variables did not exceed 5; among the horizontal independent variables, 、 、 The maximum order of regressions is determined to be 3, 2, 2, 2 respectively; the total order of regressions formed by all horizontal independent variables does not exceed 4. The constraints on the alternative regressions are shown in Table 1.

[0069] Table 1

[0070] Based on the above constraints, all possible product combinations are generated programmatically as ordinary polynomials about the independent variables. All these ordinary polynomials are then used as the set of original regressors, where the elements are... , Indicates the first One original regression quantity, This represents the total number of original regression values.

[0071] (2) Construct an orthogonalized set of candidate regressors In this step, an orthogonalization technique is used to construct a set of candidate regressors, which helps to solve the problem of data collinearity.

[0072] The original regressors are processed by Gram-Schmidt orthogonalization, that is, by constructing a set of orthogonalized candidate regressors using equations (2.5) and (2.6), where the elements are... , Indicates the first A number of orthogonalized candidate regressions.

[0073] (2.5) (2.6) in, Indicates the first An orthogonalized candidate regressor Indicates the first One original regression quantity, Indicates the first The original regression quantity at the th... Orthogonalization coefficients in the direction of the orthogonalized candidate regressors Indicates the first An orthogonalized candidate regressor express and The inner product, express and The inner product of.

[0074] Through the above steps, this invention processes the original regression values ​​one by one. This transforms it into a new term that is orthogonal to all previously transformed regressors. Ultimately, all of them This constitutes an orthogonalized set of candidate regressors.

[0075] (3) Iterative screening of regression parameters In this step, the present invention selects a set of regression quantum quantities from the orthogonalized candidate regressors of the established nonlinear aerodynamic model based on statistical criteria and stopping rules. Parameter identification is then performed based on the determined aerodynamic model structure to complete the nonlinear aerodynamic modeling. First, the mean-reduced correlation coefficient between the candidate regressors and the dependent variable is calculated. ， Select the subset with the highest correlation from the candidate regressors and remove the mean correlation coefficient. The expression is given by equation (2.7). For the statistical standard, a coefficient of determination is selected. Local statistics and predicted squared difference The results are obtained from equations (2.8), (2.9), and (2.10), respectively.

[0076] ,...,n

[0077] (2.7) (2.8) (2.9) (2.10) in, This represents the mean-free correlation coefficient. This represents the value of the j-th regressor in the i-th sample. This represents the mean of the j-th regressor. Represents the measured value of the dimensionless aerodynamic coefficient. This represents the mean of the dimensionless aerodynamic coefficients. This represents the sum of squares of the mean of the j-th regressor. This represents the sum of squared values ​​of the dimensionless aerodynamic coefficients after removing the mean. Indicates the total number of samples; Denotes the coefficient of determination. Represents the sum of squares of regression. This represents the total sum of squares of deviations. Represents the sum of squared residuals. Represents the estimated parameters transpose, Represents the regressor matrix transpose, Represents the measured value of the dimensionless aerodynamic coefficient. Indicates the number of samples. This represents the average value of z. express transpose, This represents a local statistic. Let j represent the j-th estimated parameter. Let the variance of the j-th estimated parameter be denoted as . This represents the predicted squared difference. This means that PSE is defined as the expression on the right. This represents the response value estimated by the model. Indicates transpose. This represents the upper bound of the mean squared error of the prediction. This indicates the number of items in the current model.

[0078] The coefficient of determination measures how closely the estimated value matches the measured value. Adding regressors to the model will always increase the coefficient of determination, but as more significant regressors are included in the model, The magnitude of the change will gradually decrease, therefore The variation range should not be less than 0.5%; local statistics Used to evaluate the significance level of newly introduced regressors; when it is below the critical value of 20, the regressor is considered insignificant and removed from the model; the predicted squared difference is used to achieve a balance between model fit performance and complexity. Stop iterating through the selection of regression values ​​when the minimum value is reached.

[0079] After each addition or deletion of a regressor, the least squares method is used to estimate the unknown parameters. This works by minimizing the sum of squared errors between the measured and model values ​​to solve for the parameters. Solve using equation (2.10).

[0080] This represents the dimensionless aerodynamic coefficient. , , , The measured value, which cannot be measured during flight, is calculated using other measured values ​​during flight, as shown in equation (2.11).

[0081] (2.10) (2.11) in, Indicates the estimated parameters, This indicates that the inverse matrix is ​​obtained. Indicates the total mass of the aircraft. These represent the angular velocity components of the three axes, Indicates dynamic pressure. Indicates the wing reference area. Indicates thrust. These represent the angular velocity components of the three axes, Indicates triaxial angular acceleration, Indicates wingspan, Indicates the mean aerodynamic chord length. , , These represent the moments of inertia of the three axes, It represents the product of inertia.

[0082] The specific steps for iteratively screening regressors include: (3.1) Initialize the expression of the nonlinear aerodynamic model to contain only offset terms. ; (3.2) For each orthogonalized candidate regressor By calculating the mean-reduced correlation coefficient r between it and the dependent variable z iz The candidate regressors with the highest correlation are selected and added to the current nonlinear aerodynamic model, and the coefficient of determination of the model is calculated. Changes, local statistics and predicted squared difference ; If local statistics If it exceeds the critical value of 20, then it is confirmed that... Incorporate the current nonlinear aerodynamic model; otherwise, Remove from the current nonlinear aerodynamic model; like If the change is greater than 0.5%, then it is confirmed that... Incorporate the current nonlinear aerodynamic model; otherwise, Remove from the current nonlinear aerodynamic model; If orthogonalized candidate regressors are added or removed, the least squares method is used to estimate the unknown parameters, and the latest estimated parameters are obtained. ; (3.3) Return to step (3.2) until Stop candidate regressions when the change is less than 0.5% or the PSE value is at its minimum. By iterating through the data, we finally obtain the nonlinear aerodynamic model represented by the selected regression quantities.

[0083] Finally, this invention yielded a nonlinear aerodynamic model represented by the selected regression variables, as shown in Table 1. The input variables of the nonlinear aerodynamic model include the longitudinal independent variables. 、 、 Horizontal independent variables 、 、 、 The output includes the parameters to be identified, namely the six dimensionless aerodynamic coefficients.

[0084] The technical solution of this invention acquires dynamic response data of an aircraft, directly reflecting the nonlinear aerodynamic characteristics during actual flight. Addressing the design requirements for steady-state and excitation inputs in aerodynamic identification flight tests, the invention designs an autonomous identification test control algorithm with steady-state judgment and random trim rudder deflection register functions. This enables batch, repeated, and automatic execution of excitation maneuvers under unmanned intervention, effectively avoiding human interference and improving the standardization level and data consistency of the test. Furthermore, considering the nonlinear aerodynamic characteristics exhibited by the aircraft during maneuvers, the invention also constructs a screening and tailoring mechanism for aerodynamic model structures, enabling the model to accurately describe the nonlinear characteristics of aerodynamics and possess good repeatability and consistency.

[0085] While the specific embodiments of the present invention depict actions or steps in a particular order, this should be understood as requiring such actions or steps to be performed in the shown specific order or sequential order, or requiring all illustrated actions or steps to be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations. The above descriptions are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention.

[0086] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A flight test method for global dynamic aerodynamic modeling of a fixed-wing aircraft, characterized in that, The method comprises the following steps: Step S1, arranging a sensor on an aircraft, manually controlling the flight of the aircraft, optimizing control parameters according to sensor response data, and making the aircraft fly autonomously by using the optimized control parameters; Step S2, determining whether the autonomous flight of the aircraft meets the identification test condition based on a preset trajectory and an actual trajectory of the autonomous flight, and ending the test if the identification test condition is not met; Step S3, when the identification test condition is met, controlling the aircraft to perform multiple alternating steady flights and excitation flights, and obtaining sensor response data as identification test data; Step S4, determining independent variables and parameters to be identified, wherein the independent variables include airflow angle, attitude angular velocity and rudder deflection, and the parameters to be identified include aerodynamic force and aerodynamic moment; Step S5, based on the identification test data, a nonlinear aerodynamic force model is established by constructing a regression quantity, iteratively screening the regression quantity and using a least square method; The nonlinear aerodynamic force model is used to calculate the parameters to be identified according to the independent variables.

2. The flight test method for fixed-wing aircraft global dynamic aerodynamic modeling according to claim 1, characterized in that, In step S1, the step of manually controlling the flight of the aircraft specifically comprises: a pilot inputting an instruction to make the aircraft perform 20 roll and pitch actions in sequence; The specific steps of optimizing the control parameters according to the sensor response data comprise: using an ArduPilot automatic parameter tuning algorithm to optimize PID controller parameters online to determine the control gain of the PID controller, and obtaining the control parameters.

3. The flight test method for fixed-wing aircraft global dynamic aerodynamic modeling according to claim 2, characterized in that, Step S2 specifically comprises: Step S2-1, the aircraft flies to a preset hovering point to perform a hovering action, if an actual hovering trajectory coincides with a preset hovering trajectory, the next verification task is performed, otherwise the test is ended; Step S2-2, the aircraft flies along a preset rectangular flight path following a preset waypoint, if an actual flight path trajectory coincides with a preset rectangular flight path trajectory, it is determined that the identification test condition is met.

4. The flight test method for fixed-wing aircraft global dynamic aerodynamic modeling according to claim 3, characterized in that, Step S3 specifically comprises: In the automatic flight mode, the aircraft enters a predetermined test flight path and maintains a steady flight state; after a set time, the aircraft performs a longitudinal excitation, the rudder deflection returns to a reference state after the excitation is completed, and the aircraft performs a free response; after the response is completed, the flight control system re-intervenes in the control to restore the steady flight; after a period of time, the aircraft performs a lateral excitation, and after completing the free response, the steady flight is restored again, and the aircraft continues to fly along the flight path until it re-enters the test flight path, and the above test process is repeated multiple times; The longitudinal excitation and the lateral excitation are realized by using an excitation control system, and the excitation control system comprises a state monitor, an excitation trigger and a rudder deflection adder; the steps of the aircraft performing the longitudinal excitation or the lateral excitation specifically comprise: (1) configuration initialization, including mapping a rudder control channel, setting a flag variable indicating a flight stage and defining parameters of an excitation signal; (2) the state monitor continuously monitors the flight state, and sends a start signal to the excitation trigger when the state monitor confirms that the aircraft has stably flown for a preset time length; (3) the excitation trigger starts timing after receiving the start signal, sends a start signal to the rudder deflection adder, and sends a stop signal after the timing reaches a preset time. The rudder deflection superimposer superimposes the excitation rudder deflection amplitude to the reference rudder deflection calculated by the flight control according to the preset excitation signal after receiving the start signal, and executes the rudder deflection; the rudder deflection superimposer stops working after receiving the stop signal.

5. The flight test method for fixed-wing aircraft global dynamic aerodynamic modeling according to claim 4, characterized in that, In step S4, the airflow angle comprises an angle of attack and a sideslip angle , the attitude angular velocity comprises a dimensionless roll angular velocity , a dimensionless pitch angular velocity and a dimensionless yaw angular velocity , the rudder deflection comprises an elevator deflection angle and a aileron deflection angle ; The aerodynamic forces include a drag coefficient , a side force coefficient , and a lift coefficient . The aerodynamic moments include a roll moment coefficient , a pitch moment coefficient , and a yaw moment coefficient .

6. The flight test method for fixed-wing aircraft global dynamic aerodynamic modeling of claim 5, wherein, The step S5 specifically includes: Step S5-1, determining a candidate regression quantity constraint, and determining an original regression quantity set according to the candidate regression quantity constraint; Step S5-2, processing the original regression quantity set by using an orthogonalization method to obtain an orthogonalized candidate regression quantity set; Step S5-3, adding or deleting regression quantities from the orthogonalized candidate regression quantity set based on statistical standards and stopping rules, and estimating unknown parameters of the nonlinear aerodynamic force model by using a least square method, to finally obtain a screened nonlinear aerodynamic force model represented by regression quantities.

7. The flight test method for fixed-wing aircraft global dynamic aerodynamic modeling of claim 6, wherein, In step S5-1, the candidate regression quantity constraint is: The longitudinal independent variables include 、 、 The transverse independent variables include 、 、 、 In the longitudinal independent variables, 、 、 The maximum order of the longitudinal independent variables is determined as 4, 3, 3 respectively; the total order of the regression quantity composed of all longitudinal independent variables does not exceed 5; in the transverse independent variables, 、 、 The maximum order of the transverse independent variables is determined as 3, 2, 2 respectively; the total order of the regression quantity composed of all transverse independent variables does not exceed 4; Step S5-2 specifically includes: The orthogonalized candidate regression quantity set is constructed by the following expression: wherein, represents the orthogonalized candidate regressor, represents the original regressor, represents the original regressor, orthogonalization coefficient of the original regressor in the direction of the orthogonalized candidate regressor, represents the inner product of and represents the inner product of and represents the total number of original regressors.

8. The flight test method for fixed-wing aircraft global dynamic aerodynamic modeling according to claim 7, characterized in that, Step S5-3 specifically includes: Step S5-3-1, initialize the expression of the nonlinear aerodynamic model to contain only the offset term ; Step S5-3-2, for each orthogonalized candidate regressor , calculate the de-mean correlation coefficient between the dimensionless aerodynamic force coefficient and the measured value z , select the candidate regressor corresponding to the largest de-mean correlation coefficient , add the current nonlinear aerodynamic force model, calculate the determination coefficient of the model the change of the local statistics and the prediction square difference ; If the local statistic is greater than a threshold 20, then it is confirmed that the current nonlinear aerodynamic model is added, otherwise it is removed from the current nonlinear aerodynamic model; If the change is greater than 0.5%, then confirm to add the current nonlinear aerodynamic force model, otherwise delete the current nonlinear aerodynamic force model from the current nonlinear aerodynamic force model; If an increase or deletion of the candidate regressors for orthogonalization occurs, the least square method is used to estimate the unknown parameters to solve the latest estimated parameters ; Step S5-3-3, go back to step S5-3-2 until the change of the value of the candidate regression amount is lower than 0.5% or PSE the value of the candidate regression amount is the smallest, stop the iteration of the candidate regression amount and finally obtain the screened nonlinear aerodynamic force model represented by the regression amount.

9. The flight test method for fixed-wing aircraft global dynamic aerodynamic modeling of claim 8, wherein, In step S5-3-2, the determination coefficient , the local statistic , and the prediction square error are expressed as follows, respectively. in, Denotes the coefficient of determination. Indicates the sum of squares of regression. This represents the total sum of squares of deviations. Represents the sum of squared residuals. Represents the estimated parameters transpose, Represents the regressor matrix transpose, Represents the measured value of the dimensionless aerodynamic coefficient. Indicates the number of samples. This represents the average value of z. express transpose, This represents a local statistic. Indicates the first , Let the variance of the j-th estimated parameter be denoted as . This represents the predicted squared difference. This means that PSE is defined as the expression on the right. This represents the response value estimated by the model. Indicates transpose. This represents the upper bound of the mean squared error of the prediction. This indicates the number of items in the current model.

10. The flight test method for fixed-wing aircraft global dynamic aerodynamic modeling of claim 9, wherein, In step S5-3-2, the expression of the latest estimated parameter is given by is the measured value representative of the dimensionless aerodynamic coefficient , , , the calculation expression is: wherein denotes an estimated parameter, denotes obtaining an inverse matrix, denotes the total mass of the aircraft, denote the three-axis angular velocity components, denotes the dynamic pressure, denotes the wing reference area, denotes the thrust, denote the three-axis angular velocity components, denotes the three-axis angular acceleration, denotes the wingspan, denotes the mean aerodynamic chord, , , denote the three-axis moments of inertia, denotes the inertia product.

Images