Aircraft control response data fitting method, system and computer program product based on neural network

Abstract

The present application discloses a neural network-based aircraft control response data fitting method, system, and computer program product, belonging to the field of flight simulation. The method comprises: obtaining historical flight parameter data, constructing a neural network model, and training the neural network model based on the historical flight parameter data; setting the aircraft's initial environmental state and aircraft state, and determining the aircraft's airspeed, angle of attack, and sideslip angle based on the environmental state and aircraft state; obtaining the aircraft's real-time state data and control data; constructing real-time input data, and predicting it through the trained neural network model to obtain acceleration, angular acceleration, and engine speed change rate; updating the aircraft state and engine speed based on the predicted acceleration, angular acceleration, and engine speed change rate to achieve aircraft control response data fitting. The method of the present application can reduce complexity, improve the real-time performance, accuracy, and dynamic adaptability of aircraft control response data fitting, and effectively simulate the response characteristics of the aircraft under different control scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of flight simulation, and in particular to a neural network-based aircraft control response data fitting method, system and computer program product. Background Art

[0002] Traditional fixed-wing aircraft dynamics simulation treats the aircraft as a six-degree-of-freedom rigid body. During simulation, the aircraft's control system must first be restored to simulate the deflection angles of the aircraft's various wing control surfaces under a given operating signal. Aerodynamic coefficients such as lift coefficient and drag coefficient are then obtained through table lookup based on the current airspeed, altitude, Mach number, and control surface angle. Traditional methods require accurate aircraft mass and moment of inertia. If these parameters are missing or the errors are large, the dynamic equations cannot be accurately solved. Obtaining aerodynamic coefficients relies on extensive wind tunnel testing, which is costly and time-consuming. This is especially true for new or modified aircraft, where the experimental data accumulation cycle significantly restricts simulation progress. The aircraft control system (FCS) logic must be fully restored within the simulation system to derive the deflection angles of the aircraft's control surfaces, and then the corresponding aerodynamic coefficients are obtained through table lookup based on the aircraft's current aerodynamic shape. Summary of the Invention

[0003] The purpose of this application is to overcome the shortcomings of the prior art and provide a neural network-based aircraft control response data fitting method, system and computer program product to reduce data dependence and improve modeling efficiency and generalization.

[0004] In a first aspect, the present application provides a method for fitting aircraft control response data based on a neural network, comprising the following steps:

[0005] Acquiring historical flight parameter data, building a neural network model, and training the neural network model based on the historical flight parameter data to obtain a trained neural network model;

[0006] Setting an initial environmental state and aircraft state of the aircraft, and determining the aircraft's airspeed, angle of attack, and sideslip angle based on the environmental state and the aircraft state;

[0007] Obtain real-time status data and control data of the aircraft;

[0008] Constructing real-time input data, and predicting the real-time input data using the trained neural network model to obtain predicted acceleration, angular acceleration, and engine speed change rate;

[0009] The aircraft state and engine speed are updated according to the predicted acceleration, angular acceleration, and engine speed change rate, thereby achieving fitting of aircraft control response data.

[0010] Optionally, obtaining historical flight parameter data, constructing a neural network model, and training the neural network model based on the historical flight parameter data to obtain a trained neural network model includes:

[0011] Acquiring historical flight parameter data, constructing a ground coordinate system and an aircraft body coordinate system, and performing analytical calculations on the historical flight parameter data based on the ground coordinate system and the aircraft body coordinate system to obtain aircraft response data;

[0012] determining input data according to the historical flight parameter data and the aircraft response data, and performing data enhancement on the input data based on image conservation to obtain enhanced input data;

[0013] Constructing a training set and a validation set based on the enhanced input data;

[0014] Construct a neural network model, train the neural network model using the training set, test the neural network model using the validation set, determine the model weight when the prediction performance is optimal, and obtain a trained neural network model.

[0015] Optionally, the input data includes: remaining fuel amount, airspeed, angle of attack, sideslip angle, axial angular velocity around the aircraft body, lateral angular velocity around the aircraft body, normal angular velocity around the aircraft body, roll angle, pitch angle, engine speed, longitudinal position of the joystick, lateral position of the joystick, pedal position and throttle lever position.

[0016] Optionally, setting an initial environmental state and an aircraft state of the aircraft, and determining the airspeed, angle of attack, and sideslip angle of the aircraft according to the environmental state and the aircraft state includes:

[0017] Set the initial environment state and aircraft state of the aircraft;

[0018] determining a speed of the aircraft relative to the airflow in a ground coordinate system according to the environmental state and the aircraft state;

[0019] Converting the velocity of the aircraft relative to the airflow in the ground coordinate system to the velocity of the aircraft relative to the airflow in the body coordinate system;

[0020] The real-time airspeed, real-time angle of attack and real-time sideslip angle of the aircraft are determined according to the speed relative to the airflow in the aircraft body coordinate system.

[0021] Optionally, updating the aircraft state and engine speed according to the predicted acceleration, angular acceleration, and engine speed change rate to achieve fitting of aircraft control response data includes:

[0022] Determine the velocity in the body coordinate system based on the predicted acceleration;

[0023] Determine the speed in the ground coordinate system according to the speed in the body coordinate system;

[0024] Updating the aircraft's ground coordinate system position according to the velocity in the ground coordinate system;

[0025] Determine the angular velocity in the body coordinate system according to the predicted angular acceleration;

[0026] Determine the angular velocity in the ground coordinate system according to the angular velocity in the body coordinate system;

[0027] Updating the aircraft ground coordinate system attitude according to the angular velocity in the ground coordinate system;

[0028] The engine speed is updated according to the predicted engine speed change rate to achieve the fitting of the aircraft control response data.

[0029] Optionally, the neural network-based aircraft control response data fitting method further includes: when the aircraft is climbing or descending in a straight line and the lateral force on the aircraft is ignored, starting the longitudinal balance numerical calculation to optimize the real-time status data and control data of the aircraft.

[0030] Optionally, when the aircraft is climbing or descending in a straight line and ignoring the lateral forces on the aircraft, the longitudinal trim numerical calculation is started to optimize the real-time status data and control data of the aircraft, including:

[0031] Set the trim conditions and initialize the trial airspeed, trial joystick longitudinal position, trial engine speed and the corresponding trial range;

[0032] determining a theoretical pitch angle according to the trim condition and the trial airspeed, and determining a normal gravity acceleration and an axial gravity acceleration in an aircraft coordinate system based on the theoretical pitch angle;

[0033] Using the trained neural network model to predict the normal aerodynamic acceleration, calculating the difference between the normal aerodynamic acceleration and the normal gravitational acceleration, and adjusting the trial airspeed so that the absolute value of the difference between the normal aerodynamic acceleration and the normal gravitational acceleration is less than an error threshold;

[0034] using a trained neural network model to predict axial aerodynamic acceleration, calculating the difference between the axial aerodynamic acceleration and the axial gravitational acceleration, and adjusting the trial engine speed so that the absolute value of the difference between the axial aerodynamic acceleration and the axial gravitational acceleration is less than an error threshold;

[0035] Using the trained neural network model to predict the lateral angular acceleration around the aircraft body, adjusting the longitudinal position of the trial joystick so that the absolute value of the lateral angular acceleration around the aircraft body is less than an error threshold;

[0036] Check whether all accelerations are balanced. If so, end the longitudinal trim numerical calculation; if not, return to determine the normal gravity acceleration and axial gravity acceleration in the body coordinate system and continue iteration.

[0037] Optionally, determining a theoretical pitch angle according to the trim condition and the trial airspeed, and determining a normal gravitational acceleration and an axial gravitational acceleration in an aircraft coordinate system based on the theoretical pitch angle includes:

[0038] Using trim conditions and trial airspeed Determine the theoretical pitch angle , the balancing conditions include: target adverse wind speed Target glide angle , target angle of attack , theoretical pitch angle The expression is:

[0039]

[0040] in, is the glideslope angle of the aircraft relative to the air, which is expressed as:

[0041]

[0042] in, is the aircraft ground speed, To try out the airspeed, is the target glide angle;

[0043] The aircraft ground speed The expression is:

[0044]

[0045] Among them, A, B, and C are parameters, and the expressions are as follows:

[0046]

[0047]

[0048]

[0049] in, is the target glide angle, is the target reverse wind speed, To try to find the airspeed;

[0050] Based on the theoretical pitch angle Determine the axial gravitational acceleration in the body coordinate system , the expression is:

[0051]

[0052] in, is the acceleration due to gravity;

[0053] Based on the theoretical pitch angle Determine the normal gravitational acceleration in the body coordinate system , the expression is:

[0054]

[0055] in, is the acceleration due to gravity.

[0056] In a second aspect, the present application also provides a neural network-based aircraft control response data fitting system for executing the neural network-based aircraft control response data fitting method as described in any one of the first aspects, comprising: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the neural network-based aircraft control response data fitting method as described in any one of the first aspects.

[0057] In a third aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the neural network-based aircraft control response data fitting method as described in any one of the first aspects.

[0058] The present application provides a neural network-based aircraft control response data fitting method, system, and computer program product. By directly using historical flight parameter data to train the neural network model, the neural network training relies only on real flight parameter data in flight records, without the need for wind tunnel test data. This can significantly reduce data acquisition costs and time costs, avoid experimental errors, and enhance the model's generalization ability and dynamic adaptability. By directly using the trained neural network to predict acceleration, angular acceleration, and engine speed change rate, it bypasses force calculations and does not rely on mass and moment of inertia information. It can accurately capture the complex mapping relationship between aircraft control input and response data, and achieve rapid and accurate estimation of key parameters of the aircraft's motion state. By iteratively adjusting trial parameters to balance aerodynamic force with gravity, it can optimize the data fitting accuracy for specific flight phases and improve the stability of the flight state. The present application can significantly improve the real-time, accuracy, and dynamic adaptability of aircraft control response data fitting, effectively simulate the response characteristics of the aircraft under different control scenarios, provide high-precision data support for flight simulation, control performance evaluation, driving training, etc., and help improve the reliability and safety of aircraft design, testing, and operation.

[0059] In order to make the above features and advantages of the present invention more clearly understood, embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0061] Figure 1 This is a flowchart of a neural network-based aircraft control response data fitting method provided in one embodiment of the present application.

[0062] Figure 2 This is a flowchart of step S1 in the neural network-based aircraft control response data fitting method provided in one embodiment of the present application.

[0063] Figure 3 This is a flowchart of step S2 in the neural network-based aircraft control response data fitting method provided in one embodiment of the present application.

[0064] Figure 4 This is a flowchart of step S5 in the neural network-based aircraft control response data fitting method provided in one embodiment of the present application.

[0065] Figure 5 This is a flowchart of step S6 in the neural network-based aircraft control response data fitting method provided in one embodiment of the present application. DETAILED DESCRIPTION

[0066] To make the purpose and technical solutions of the embodiments of the present application clearer, the technical solutions of the embodiments of the present application will be clearly and completely described below in conjunction with the drawings of the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the described embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0067] In one embodiment, see Figure 1 The present application provides an aircraft control response data fitting method based on a neural network, which may include the following steps: step S1 to step S5.

[0068] Step S1: Acquire historical flight parameter data, construct a neural network model, train the neural network model based on the historical flight parameter data, and obtain a trained neural network model.

[0069] Step S2: setting the initial environmental state and aircraft state of the aircraft, and determining the aircraft's airspeed, angle of attack, and sideslip angle based on the environmental state and aircraft state.

[0070] Step S3: Acquire the real-time status data and control data of the aircraft.

[0071] Step S4: constructing real-time input data, and predicting the real-time input data through the trained neural network model to obtain predicted acceleration, angular acceleration, and engine speed change rate.

[0072] Step S5: updating the aircraft state and engine speed according to the predicted acceleration, angular acceleration, and engine speed change rate to achieve fitting of the aircraft control response data.

[0073] In the aircraft control response data fitting method based on neural network of the present application, by directly using historical flight parameter data to train the neural network model, the neural network training only depends on the real flight parameter data in the flight record, and does not require wind tunnel test data, which greatly reduces the data acquisition cost and time cost, enhances the generalization ability and dynamic adaptability of the model, and can avoid experimental errors; by directly using the trained neural network to predict acceleration, angular acceleration, and engine speed change rate, it bypasses the force calculation and does not rely on mass and moment of inertia information. It can use the powerful nonlinear fitting ability of the neural network to accurately capture the complex mapping relationship between aircraft control input and response data, and realize the rapid and accurate prediction of key parameters of the aircraft motion state. The method of the present application can significantly improve the real-time, accuracy and dynamic adaptability of aircraft control response data fitting, can effectively simulate the response characteristics of the aircraft under different control scenarios, provide high-precision data support for flight simulation, control performance evaluation, driving training, etc., and help improve the reliability and safety of aircraft design, testing and operation.

[0074] In step S1, see Figure 1 In step S1, historical flight parameter data is obtained, a neural network model is constructed, and the neural network model is trained based on the historical flight parameter data to obtain a trained neural network model.

[0075] As an example, see Figure 2 , step S1 may include the following steps: step S11 to step S14.

[0076] Step S11: Acquire historical flight parameter data, construct a ground coordinate system and an aircraft coordinate system, and perform analytical calculations on the historical flight parameter data based on the ground coordinate system and the aircraft coordinate system to obtain aircraft response data.

[0077] Step S12: Determine input data based on historical flight parameter data and aircraft response data, and perform data enhancement on the input data based on image conservation to obtain enhanced input data.

[0078] Step S13: Construct a training set and a validation set based on the enhanced input data.

[0079] Step S14: Construct a neural network model, train the neural network model using the training set, test it using the validation set, determine the model weights when the prediction performance is optimal, and obtain a trained neural network model.

[0080] As an example, in step S11, the flight record includes the flight parameter data of each sampling point during the flight process, and the historical flight parameter data can be obtained from the flight record. The historical flight parameter data may include: the remaining fuel amount; ,airspeed , angle of attack , sideslip angle , pitch angle , roll angle , yaw angle , engine speed n , aircraft speed v , angular velocity around the body axis p , lateral angular velocity around the body q , angular velocity around the body normal r , joystick longitudinal position a , joystick lateral position b , pedal position c , throttle lever position d .

[0081] Furthermore, a ground coordinate system is constructed with due east (e), due north (n), and the ground perpendicular (d) as the three axes to describe the aircraft's absolute motion relative to the ground. The positive X-axis, along the aircraft's longitudinal axis, points toward the nose; the positive Y-axis, along the aircraft's lateral axis, points toward the right wing; and the positive Z-axis, along the aircraft's vertical axis, points toward the belly. This body coordinate system is used to describe the aircraft's forces and motion.

[0082] As an example, the ground coordinate system and the body coordinate system can be transformed by the aircraft attitude and coordinate transformation matrix. The aircraft attitude includes: pitch angle , roll angle , yaw angle .

[0083] Furthermore, based on the ground coordinate system and the aircraft coordinate system, the historical flight parameter data is analyzed and calculated to obtain aircraft response data, which includes: angular acceleration around the aircraft axis , lateral angular acceleration around the body , angular acceleration around the body normal , body axial acceleration , lateral acceleration of the body , body normal acceleration , engine speed change rate and fuel consumption rate .

[0084] Specifically, the axial angular velocity of the aircraft body can be calculated for each frame in the historical flight parameter data. p , lateral angular velocity around the body q , angular velocity around the body normal r The angular acceleration around the body axis is obtained by taking the difference of , lateral angular acceleration around the body , angular acceleration around the body normal Angular acceleration around the body axis The expression is:

[0085]

[0086] in, is the angular velocity around the body axis at time t, is the angular velocity around the body axis at time t-1, is the angular velocity around the body axis at time t+1, is the sampling period.

[0087] As an example, the lateral angular acceleration around the body The expression is:

[0088]

[0089] in, is the lateral angular velocity around the body at time t, is the lateral angular velocity around the body at time t-1, is the lateral angular velocity around the body at time t+1, is the sampling period.

[0090] As an example, the angular acceleration around the body normal The expression is:

[0091]

[0092] in, is the normal angular velocity around the body at time t, is the normal angular velocity around the body at time t-1, is the normal angular velocity around the body at time t+1, is the sampling period.

[0093] Furthermore, the aircraft speed v Split it in the ground coordinate system to get the north speed of the aircraft in the ground coordinate system , eastward speed , vertical speed , the north speed of the aircraft in the ground coordinate system , eastward speed , vertical speed The quotient of the velocity difference and the sampling period is used to obtain the north acceleration in the ground coordinate system. , eastward acceleration , vertical acceleration . North acceleration in ground coordinate system The expression is:

[0094]

[0095] in, is the northward velocity at time t, is the northward velocity at time t-1, is the northward velocity at time t+1, is the sampling period.

[0096] As an example, the eastward acceleration in the ground coordinate system The expression is:

[0097]

[0098] in, is the eastward velocity at time t, is the eastward velocity at time t-1, is the eastward velocity at time t+1, is the sampling period.

[0099] As an example, the vertical acceleration The vertical acceleration in the ground coordinate system is only caused by aerodynamic force, and the influence of gravity acceleration needs to be subtracted. The expression is:

[0100]

[0101] in, is the vertical velocity at time t, is the vertical velocity at time t-1, is the vertical velocity at time t+1, is the sampling period, g is the acceleration due to gravity.

[0102] Furthermore, the north acceleration in the ground coordinate system is transformed into , eastward acceleration , vertical acceleration Converted into the body acceleration in the body coordinate system, the body acceleration includes: body axial acceleration , lateral acceleration of the body , body normal acceleration , the expression is:

[0103]

[0104] in, is the roll angle, is the pitch angle, is the yaw angle.

[0105] Similarly, the axial velocity in the body coordinate system can be obtained , lateral speed , angular acceleration around the body axis , lateral angular acceleration around the body , angular acceleration around the body normal .

[0106] As an example, the engine speed can be directly n The engine speed change rate is obtained by taking the quotient of the difference and the sampling period , the expression is:

[0107]

[0108] in, is the engine speed at time t, is the engine speed at time t-1, is the engine speed at time t+1, is the sampling period.

[0109] As an example, the remaining fuel level can be directly After the difference, the quotient is taken from the sampling period to obtain the fuel consumption rate , the expression is:

[0110]

[0111] in, is the remaining fuel quantity at time t, is the remaining fuel at time t-1, is the remaining fuel at time t+1, is the sampling period.

[0112] As an example, in step S12, input data is determined based on historical flight parameter data and aircraft response data, and the input data includes: remaining fuel ,airspeed , angle of attack , sideslip angle , angular velocity around the body axis p , lateral angular velocity around the body q , angular velocity around the body normal r , roll angle , pitch angle , engine speed n , joystick longitudinal position a , joystick lateral position b , pedal position c and throttle lever position d .

[0113] As an example, according to the law of conservation of image, when the sideslip angle , roll angle , angular velocity around the body axis p , angular velocity around the body normal r , lateral speed , pedal position c , joystick lateral position b When the mirror image is flipped and other variables remain unchanged, the corresponding lateral acceleration of the body , angular acceleration around the body axis , angular acceleration around the body normal The input data can be mirrored to increase the amount of data and improve the degree of training generalization. The expression is:

[0114]

[0115] in, is the sideslip angle after mirroring, is the roll angle after mirroring, is the angular velocity around the body axis after mirroring, is the normal angular velocity around the body after mirroring, is the lateral velocity after mirroring, is the lateral acceleration of the body after mirroring, is the angular acceleration around the body axis after mirroring, is the normal angular acceleration around the body after mirroring, is the horizontal position of the joystick after mirroring, This is the mirrored pedal position.

[0116] As an example, the enhanced input data includes: remaining fuel ,airspeed , angle of attack , sideslip angle , angular velocity around the body axis p , lateral angular velocity around the body q , angular velocity around the body normal r , lateral speed , roll angle , pitch angle , engine speed n , joystick longitudinal position a , joystick lateral position b , pedal position c and throttle lever position d , Side slip angle after mirroring , Roll angle after mirroring , angular velocity around the body axis after mirroring , the normal angular velocity around the body after mirroring , lateral speed after mirroring , lateral acceleration of the body after mirroring , angular acceleration around the body axis after mirroring , the normal angular acceleration around the body after mirroring , Horizontal position of the joystick after mirroring , mirrored pedal position .

[0117] As an example, in step S13, a training set and a validation set are constructed based on the enhanced input data. is the training set, is the validation set.

[0118] As an example, in step S14, a neural network model is constructed, and the neural network output layer data is set to include: the body axial acceleration , lateral acceleration of the body , body normal acceleration , angular velocity around the body axis , lateral angular velocity around the body , angular acceleration around the body normal and engine speed change rate . The neural network model is trained iteratively using the training set. The activation function can be the hyperbolic tangent function (tanh), which is used to introduce nonlinear transformations in the hidden layer to obtain nonlinear relationships. The loss function can use the mean square error loss (MSE) to optimize the training process and evaluate the predictive performance of the neural network model. The training can be iterated for N rounds (epochs), and the number of epochs can be adjusted based on the model prediction results and hardware conditions. In each iteration (i.e., an epoch), the training set data is forward propagated through the neural network to calculate the predicted value, and then the network weights are updated through the backpropagation algorithm to minimize the mean square error loss on the training set.

[0119] Furthermore, after each epoch, the validation set is used for forward propagation, and the mean square error loss on the validation set is calculated. After completing N epochs of training, the mean square error loss of each epoch on the validation set is compared, and the weight corresponding to the minimum mean square error loss is selected as the final training result to obtain a trained neural network model.

[0120] As an example, the neural network model may adopt a BP neural network.

[0121] In step S2, see Figure 1 In step S2, the initial environmental state and aircraft state of the aircraft are set, and the real-time airspeed, angle of attack, and sideslip angle of the aircraft are determined based on the environmental state and aircraft state.

[0122] As an example, see Figure 3 , step S2 may include the following steps: step S21 to step S24.

[0123] Step S21: Setting the initial environment state and aircraft state of the aircraft.

[0124] Step S22: determining the speed of the aircraft relative to the airflow in the ground coordinate system according to the environmental state and the aircraft state.

[0125] Step S23: converting the velocity relative to the airflow in the aircraft ground coordinate system into the velocity relative to the airflow in the aircraft body coordinate system.

[0126] Step S24: determining the real-time airspeed, real-time angle of attack, and real-time sideslip angle of the aircraft based on the speed relative to the airflow in the aircraft coordinate system.

[0127] As an example, in step S21, the initial environmental state and aircraft state of the aircraft are set. The environmental state is an environmental variable, which may include: north wind speed , east wind speed The aircraft state may include: aircraft motion state variables and aircraft engine state variables, the aircraft motion state variables include: ground coordinate system position , ground coordinate system speed , ground coordinate system attitude , body angular velocity The aircraft engine state variables include: engine speed n , Remaining fuel .

[0128] As an example, in step S22, the ground coordinate system velocity , North wind speed , east wind speed Calculate the speed of the aircraft relative to the airflow in the ground coordinate system, wherein the speed of the aircraft relative to the airflow in the ground coordinate system includes: the north speed of the aircraft relative to the airflow in the ground coordinate system , the eastward speed of the aircraft relative to the airflow in the ground coordinate system , the vertical velocity of the aircraft relative to the airflow in the ground coordinate system , the expression is:

[0129]

[0130] in, is the north velocity in the ground coordinate system, is the eastward velocity in the ground coordinate system, is the vertical velocity in the ground coordinate system.

[0131] As an example, in step S23, based on the coordinate transformation matrix, the velocity of the aircraft relative to the airflow in the ground coordinate system is converted into the velocity relative to the airflow in the body coordinate system, wherein the velocity relative to the airflow in the body coordinate system includes: the axial velocity of the aircraft relative to the airflow in the body coordinate system , the lateral speed of the aircraft relative to the airflow in the body coordinate system , the normal velocity of the aircraft relative to the airflow in the body coordinate system , the expression is:

[0132]

[0133] in, is the roll angle, is the pitch angle, is the yaw angle.

[0134] As an example, in step S24, the real-time airspeed of the aircraft can be determined based on the speed relative to the airflow in the aircraft coordinate system. , the expression is:

[0135]

[0136] in, is the axial velocity of the aircraft relative to the airflow in the body coordinate system, is the lateral velocity of the aircraft relative to the airflow in the body coordinate system, is the normal velocity of the aircraft relative to the airflow in the body coordinate system.

[0137] Furthermore, the real-time angle of attack of the aircraft can be determined based on the velocity relative to the airflow in the aircraft coordinate system. , the expression is:

[0138]

[0139] in, is the axial velocity of the aircraft relative to the airflow in the body coordinate system, is the normal velocity of the aircraft relative to the airflow in the body coordinate system.

[0140] Furthermore, the real-time sideslip angle of the aircraft can be determined based on the speed relative to the airflow in the aircraft coordinate system. , the expression is:

[0141]

[0142] in, is the axial velocity of the aircraft relative to the airflow in the body coordinate system, is the lateral velocity of the aircraft relative to the airflow in the body coordinate system.

[0143] In step S3, see Figure 1 In step S3, the real-time status data and control data of the aircraft are obtained.

[0144] As an example, obtain the real-time status data of the current frame of the aircraft, which may include: real-time remaining fuel quantity , real-time body angular velocity , real-time ground coordinate system attitude , real-time engine speed .

[0145] As an example, the real-time control data of the current frame of the aircraft can be obtained through the user device input, and the real-time control data may include: the real-time longitudinal position of the joystick , real-time joystick lateral position , real-time pedal position , real-time throttle lever position .

[0146] In step S4, see Figure 1In step S4, real-time input data is constructed and predicted by the trained neural network model to obtain predicted acceleration, angular acceleration, and engine speed change rate.

[0147] As an example, real-time input data is constructed based on the real-time airspeed, angle of attack, sideslip angle, real-time status data and control data of the aircraft, and the real-time input data includes: real-time remaining fuel amount , real-time angular velocity around the body axis , real-time lateral angular velocity around the body , real-time normal angular velocity around the body , real-time roll angle , real-time pitch angle , real-time engine speed , real-time joystick longitudinal position , real-time joystick lateral position , real-time pedal position , real-time throttle lever position , real-time airspeed , real-time angle of attack , real-time sideslip angle .

[0148] Furthermore, the real-time input data is normalized and input into the trained neural network model to obtain the predicted acceleration, angular acceleration, and engine speed change rate. The predicted acceleration includes: the axial acceleration of the body , lateral acceleration of the body , body normal acceleration The predicted angular acceleration includes: angular acceleration around the body axis , lateral angular acceleration around the body , angular acceleration around the body normal .

[0149] As an example, the normalization of real-time input data may include scaling data with a clear physical range to the interval [0, 1] using minimum and maximum scaling; converting continuous data into a standard normal distribution with a mean of 0 and a standard deviation of 1; and normalizing angular data in radians or mapping it to a specific interval through trigonometric functions.

[0150] In step S5, see Figure 1 In step S5, the aircraft state and engine speed are updated according to the predicted acceleration, angular acceleration, and engine speed change rate to achieve fitting of the aircraft control response data.

[0151] As an example, see Figure 4 , step S5 may include the following steps: step S51 to step S57.

[0152] Step S51: Determine the velocity in the body coordinate system according to the predicted acceleration.

[0153] Step S52: Determine the speed in the ground coordinate system according to the speed in the body coordinate system.

[0154] Step S53: Update the aircraft's ground coordinate system position according to the velocity in the ground coordinate system.

[0155] Step S54: Determine the angular velocity in the body coordinate system according to the predicted angular acceleration.

[0156] Step S55: Determine the angular velocity in the ground coordinate system according to the angular velocity in the body coordinate system.

[0157] Step S56: Update the aircraft's ground coordinate system attitude according to the angular velocity in the ground coordinate system.

[0158] Step S57: updating the engine speed according to the predicted engine speed change rate to achieve fitting of the aircraft control response data.

[0159] As an example, in step S51, according to the predicted axial acceleration of the aircraft body , lateral acceleration of the body , body normal acceleration Determine the velocity in the body coordinate system, the velocity in the body coordinate system includes: axial velocity , lateral speed , normal velocity , the expression is:

[0160]

[0161] in, Indicates the axial velocity of the previous frame, Indicates the lateral velocity of the previous frame, represents the normal velocity of the previous frame, is the sampling period.

[0162] As an example, in step S52, based on the coordinate transformation matrix, the axial velocity in the body coordinate system is converted to , lateral speed , normal velocity Converted to the speed in the ground coordinate system, the speed in the ground coordinate system includes: the north speed in the ground coordinate system , eastward velocity in the ground coordinate system , vertical velocity in the ground coordinate system , the expression is:

[0163]

[0164] in, is the real-time roll angle, is the real-time pitch angle, is the real-time yaw angle.

[0165] As an example, in step S53, according to the north speed in the ground coordinate system , eastward velocity in the ground coordinate system , vertical velocity in the ground coordinate system Update the aircraft's ground coordinate system position. The expression is:

[0166]

[0167] in, 、 、 is the real-time ground coordinate system position, 、 、 is the updated ground coordinate system position, is the sampling period.

[0168] As an example, in step S54, according to the predicted angular acceleration around the body axis , lateral angular acceleration around the body , angular acceleration around the body normal Determine the angular velocity in the body coordinate system, the velocity in the body coordinate system includes: the angular velocity around the body axis , lateral angular velocity around the body , angular velocity around the body normal , the expression is:

[0169]

[0170] in, is the angular velocity around the body axis in the previous frame, is the lateral angular velocity around the body in the previous frame, is the normal angular velocity around the body in the previous frame, is the sampling period.

[0171] As an example, in step S55, based on the coordinate transformation matrix, the angular velocity around the axis of the body in the body coordinate system is converted to , lateral angular velocity around the body , angular velocity around the body normal Converted to the angular velocity in the ground coordinate system, that is, the Euler angular acceleration, the expression is:

[0172]

[0173] in, 、 、 is the Euler angular acceleration, is the real-time roll angle, is the real-time pitch angle, is the real-time yaw angle.

[0174] As an example, in step S56, the aircraft ground coordinate system attitude is updated according to the Euler angular acceleration, and the expression is:

[0175]

[0176] in, 、 、 is the updated ground coordinate system posture, is the real-time roll angle, is the real-time pitch angle, is the real-time yaw angle, is the sampling period.

[0177] As an example, in step S57, according to the predicted engine speed change rate Update engine speed , the expression is:

[0178]

[0179] in, is the real-time engine speed, is the sampling period.

[0180] Furthermore, steps S3 to S5 are looped to continuously update the aircraft status and engine speed, thereby achieving dynamic fitting of the aircraft control response data.

[0181] In one example, the present application provides a neural network-based aircraft control response data fitting method, which may also include step S6: when the aircraft is flying in a straight climb or straight descent and the lateral force of the aircraft is ignored, the longitudinal balance numerical calculation is started to optimize the real-time status data and control data of the aircraft.

[0182] As an example, see Figure 5 , step S6 may include the following steps: step S61 to step S66.

[0183] Step S61: Setting the trim conditions, respectively initializing the trial airspeed, trial joystick longitudinal position, trial engine speed and the corresponding trial range.

[0184] Step S62: Determine a theoretical pitch angle according to the trim conditions and the trial airspeed, and determine the normal gravity acceleration and the axial gravity acceleration in the aircraft coordinate system based on the theoretical pitch angle.

[0185] Step S63: Use the trained neural network model to predict the normal aerodynamic acceleration, calculate the difference between the normal aerodynamic acceleration and the normal gravitational acceleration, and adjust the trial airspeed so that the absolute value of the difference between the normal aerodynamic acceleration and the normal gravitational acceleration is less than the error threshold.

[0186] Step S64: Use the trained neural network model to predict the axial aerodynamic acceleration, calculate the difference between the axial aerodynamic acceleration and the axial gravitational acceleration, and adjust the trial engine speed so that the absolute value of the difference between the axial aerodynamic acceleration and the axial gravitational acceleration is less than the error threshold.

[0187] Step S65: Use the trained neural network model to predict the lateral angular acceleration around the aircraft body, and adjust the longitudinal position of the trial joystick so that the absolute value of the lateral angular acceleration around the aircraft body is less than the error threshold.

[0188] Step S66: Check whether all accelerations are balanced. If so, end the longitudinal trim numerical calculation; if not, return to step S62 to continue iteration.

[0189] As an example, in step S61, when the aircraft is climbing or descending in a straight line and the lateral force on the aircraft is ignored, that is, the lateral force on the aircraft is ignored and the roll angle is maintained. , when the aircraft is climbing or descending in a straight line, the longitudinal trim numerical calculation can be started. Specifically, at a given target angle of attack Target glide angle , target reverse wind speed Under these conditions, the aircraft trim state is determined by trial and error by adjusting the airspeed, longitudinal position of the joystick, and engine speed so that the aerodynamic lift, engine thrust, and gravity acting on the aircraft are balanced, that is, the aircraft acceleration and angular acceleration are both close to 0.

[0190] As an example, the trimming conditions can be set according to user requirements, and the trimming conditions include: target angle of attack Target glide angle , target reverse wind speed Since the remaining fuel quantity can affect the total mass of the aircraft and thus affect the aerodynamic acceleration, the trim conditions can also include: target remaining fuel quantity .

[0191] Furthermore, the initial value of the trial airspeed is set to , and its corresponding trial range is ; Set the initial value of the longitudinal position of the trial joystick to , the corresponding trial range is ; You can set the initial value of the engine speed to , the corresponding trial range is .

[0192] In one example, the lower limit of the trial range can be set to the minimum value in the training set, the upper limit of the trial range can be set to the maximum value in the training set, and the initial value can be set to the middle value of the trial range.

[0193] As an example, in step S62, the theoretical pitch angle is determined based on the geometric relationship between the adverse wind speed, airspeed, glide angle, angle of attack and pitch angle. Specifically, the target adverse wind speed in the trim condition can be used. Target glide angle , target angle of attack , try to find the airspeed Determine the theoretical pitch angle , the expression is:

[0194]

[0195] in, is the aircraft's glide path angle relative to the air.

[0196] As an example, the aircraft's glide angle relative to the air The expression is:

[0197]

[0198] in, is the aircraft ground speed, To try out the airspeed, is the target glide angle.

[0199] As an example, aircraft ground speed The expression is:

[0200]

[0201] Among them, A, B, and C are parameters that can be determined according to the trim conditions and trim airspeed. The expressions are as follows:

[0202]

[0203]

[0204]

[0205] in, is the target glide angle, is the target reverse wind speed, To test the airspeed.

[0206] Furthermore, based on the theoretical pitch angle Determine the axial gravitational acceleration in the body coordinate system , the expression is:

[0207]

[0208] in, is the acceleration due to gravity.

[0209] Furthermore, based on the theoretical pitch angle Determine the normal gravitational acceleration in the body coordinate system , the expression is:

[0210]

[0211] in, is the acceleration due to gravity.

[0212] As an example, in step S63, the normal aerodynamic acceleration is predicted using the trained neural network model. When the longitudinal trim calculation is enabled, the normal aerodynamic acceleration is the normal acceleration of the aircraft body, which refers to the normal acceleration of the aircraft body caused by aerodynamic forces (gravity has been eliminated). The specific method for predicting the normal aerodynamic acceleration can be referred to the specific method for predicting the normal acceleration of the aircraft body in step S4 and will not be repeated here.

[0213] Furthermore, the normal aerodynamic acceleration is calculated The normal acceleration due to gravity is difference , the expression is: . Determine the normal aerodynamic acceleration The normal acceleration due to gravity is difference Is the absolute value of less than the error threshold? ,like , then the adjustment of the trial airspeed is ended, otherwise, the normal aerodynamic acceleration is determined The normal acceleration due to gravity is difference By adjusting the trial airspeed, the absolute value of the difference between the normal aerodynamic acceleration and the normal gravity acceleration is less than the error threshold; if , then the trial airspeed Too large, the current trial airspeed Set it as the maximum value of the corresponding trial range, and take the average of the maximum and minimum values ​​in the corresponding trial range as the new trial airspeed ;like , then the trial airspeed If the current trial airspeed is too small, set the current trial airspeed to the minimum value of the corresponding trial range, and take the average of the maximum and minimum values ​​in the corresponding trial range as the new trial airspeed. ; Based on the new trial airspeed Re-determine the normal gravity acceleration and axial gravity acceleration in the body coordinate system; adjust the trial airspeed cyclically until it meets the requirements. Finish.

[0214] As an example, in step S64, the trained neural network model is used to predict the axial aerodynamic acceleration. When the longitudinal trim numerical calculation is enabled, the axial aerodynamic acceleration is the aircraft axial acceleration, which refers to the axial acceleration of the aircraft caused by aerodynamic forces (gravity has been eliminated). The specific method for predicting the axial aerodynamic acceleration can be referred to the specific method for predicting the aircraft axial acceleration in step S4 and will not be repeated here.

[0215] Furthermore, the axial aerodynamic acceleration is calculated The axial acceleration due to gravity difference , the expression is: . Determine the axial aerodynamic acceleration The axial acceleration due to gravity difference Is the absolute value of less than the error threshold? ,like , then the adjustment of the engine speed is finished, otherwise, the axial aerodynamic acceleration is judged. The axial acceleration due to gravity difference The positive and negative of the engine speed is adjusted to make the axial aerodynamic acceleration The axial acceleration due to gravity difference The absolute value of is less than the error threshold; if , it indicates that the thrust is too large, set the current trial engine speed to the maximum value of the corresponding trial range, and take the average of the maximum and minimum values ​​in the corresponding trial range as the new trial engine speed ;like , it indicates that the thrust is too small, set the current trial engine speed to the minimum value of the corresponding trial range, and take the average of the maximum and minimum values ​​in the corresponding trial range as the new trial engine speed ; Cycle and adjust the engine speed until it meets the Finish.

[0216] As an example, in step S65, the trained neural network model is used to predict the lateral angular acceleration of the aircraft body. When the longitudinal trim numerical calculation is started, the lateral angular acceleration around the aircraft body is the aircraft pitch acceleration. The specific method for predicting the lateral angular acceleration around the aircraft body can refer to the specific method for predicting the lateral angular acceleration around the aircraft body in step S4, and will not be repeated here.

[0217] Further, the predicted lateral angular acceleration around the body is determined Is the absolute value of less than the error threshold? ,like , then the adjustment of the longitudinal position of the trial joystick is ended. Otherwise, the predicted lateral angular acceleration c around the body is judged to be positive or negative, and the predicted lateral angular acceleration c around the body is adjusted by adjusting the longitudinal position of the trial joystick to be Is the absolute value of less than the error threshold? , it indicates that the longitudinal position of the trial joystick is too large and the pitch angle of the aircraft tends to increase. The current longitudinal position of the trial joystick is set to the maximum value of the corresponding trial range, and the average of the maximum and minimum values ​​in the corresponding trial range is taken as the new longitudinal position of the trial joystick. ;like , it indicates that the longitudinal position of the trial joystick is too small and the pitch angle of the aircraft tends to decrease. The current longitudinal position of the trial joystick is set to the minimum value of the corresponding trial range, and the average of the maximum and minimum values ​​in the corresponding trial range is taken as the new longitudinal position of the trial joystick. ; Circularly adjust the longitudinal position of the joystick until it meets the Finish.

[0218] As an example, the error threshold The setting should ensure that when the absolute value of the acceleration and angular acceleration is less than the error threshold When , the corresponding cumulative changes in velocity and angular velocity can be ignored. For example, the error threshold Can be set to .

[0219] As an example, in step S66, since each adjustment of the trial value in steps S63 through S65 will slightly affect the equilibrium state required in the other steps, adjusting the trial longitudinal joystick position in step S65 can bring the aircraft's pitch acceleration close to zero. However, since the trial airspeed adjustment in step S63 is iteratively performed under the assumption that the longitudinal joystick position remains unchanged, the trial airspeed in step S63 needs to be fine-tuned again after step S65.

[0220] As an example, with each round of adjustment, the absolute value of the overall error tends to get smaller and smaller. Therefore, a comprehensive check is required to determine whether to perform another round of iterative trial and error. Specifically, check whether all accelerations are balanced. If so, terminate the longitudinal trim value calculation; if not, return to step S63 and calculate again using the new trial value obtained after a round of trial and error. 、 and If all three values ​​are less than the error threshold, the longitudinal trim numerical calculation is terminated; otherwise, the process returns to step S62 and continues iteration. By adjusting only the airspeed, longitudinal joystick position, and engine speed, the aerodynamic forces, engine thrust, and gravity acting on the aircraft are balanced, thereby achieving a stable flight state with acceleration and angular acceleration approaching zero.

[0221] The present invention relates to a neural network-based aircraft control response data fitting method. The method enhances the physical symmetry of the input data through mirror conservation, generates symmetrical data by utilizing the mirror mapping characteristics of flight parameters, expands the dataset size without relying on additional real data, avoids the model overfitting problem caused by data scarcity, and ensures the rationality and effectiveness of data enhancement by maintaining the mirror physical consistency of aerodynamic forces and control quantities. The method also achieves accurate mapping of velocities and accelerations in different reference frames based on the aircraft attitude angle and coordinate transformation matrix through a dynamic conversion mechanism between the ground coordinate system and the body coordinate system, thereby improving the model's dynamic description capability for complex flight attitudes. The method also achieves automatic dynamic balance of aerodynamic forces, thrust, and gravity through data-driven methods without the need to establish complex force balance equations, thereby keeping the acceleration prediction error of the model under stable flight conditions within a threshold range and significantly enhancing the physical rationality and engineering practicality of the model under specific working conditions. The present invention can achieve high-precision and high-robustness fitting of aircraft control response data while reducing modeling complexity, providing technical support with both efficiency and reliability for scenarios such as flight control system design and virtual flight test verification.

[0222] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The order of execution of these sub-steps or stages is not necessarily sequential, but can be executed in rotation or alternation with other steps or at least a portion of the sub-steps or stages of other steps.

[0223] In another embodiment, the present application also provides a neural network-based aircraft control response data fitting system, which includes: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement any of the above-mentioned neural network-based aircraft control response data fitting methods.

[0224] In another embodiment, the present application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the various steps of the neural network-based aircraft control response data fitting method provided in the above embodiment.

[0225] The computer-executable instructions for implementing the methods of the present application may be written in any combination of one or more programming languages. These computer-executable instructions may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer-executable instructions are executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer-executable instructions may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or electronic device.

[0226] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a cathode ray tube (CRT) or a liquid crystal display (LCD) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; feedback provided to the user can be any form of sensory feedback (e.g., visual feedback or tactile feedback); and input from the user can be received in any form, including acoustic input, voice input, or tactile input.

[0227] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as data electronics), or a computing system that includes middleware components (e.g., application electronics), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0228] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0229] Although the present application has been disclosed above with reference to the embodiments, they are not intended to limit the present application. Anyone with ordinary knowledge in the technical field may make slight changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be determined by the scope of the appended patent application.

Claims

1. A neural network-based aircraft control response data fitting method, characterized in that: The following steps are involved: Acquiring historical flight parameter data, building a neural network model, and training the neural network model based on the historical flight parameter data to obtain a trained neural network model; Setting an initial environmental state and aircraft state of the aircraft, and determining the aircraft's airspeed, angle of attack, and sideslip angle based on the environmental state and the aircraft state; Obtain real-time status data and control data of the aircraft; Constructing real-time input data, and predicting the real-time input data using the trained neural network model to obtain predicted acceleration, angular acceleration, and engine speed change rate; updating the aircraft state and engine speed based on the predicted acceleration, angular acceleration, and engine speed change rate; Under the given target angle of attack, target glide angle, and target adverse wind speed, the aircraft airspeed, longitudinal joystick position, and engine speed are adjusted to achieve zero aircraft acceleration and angular acceleration, completing the aircraft longitudinal trim calculation. The invention comprises the following steps: setting the trim conditions, respectively initializing the trial airspeed, the trial joystick longitudinal position, the trial engine speed and the corresponding trial range; determining the theoretical pitch angle according to the trim conditions and the trial airspeed, and determining the normal gravity acceleration and the axial gravity acceleration in the body coordinate system based on the theoretical pitch angle; using the trained neural network model to predict the normal aerodynamic acceleration, calculating the difference between the normal aerodynamic acceleration and the normal gravity acceleration, and adjusting the trial airspeed so that the absolute value of the difference between the normal aerodynamic acceleration and the normal gravity acceleration is less than the error threshold .... The axial aerodynamic acceleration is predicted by the model, the difference between the axial aerodynamic acceleration and the axial gravitational acceleration is calculated, and the trial engine speed is adjusted so that the absolute value of the difference between the axial aerodynamic acceleration and the axial gravitational acceleration is less than the error threshold; the lateral angular acceleration around the fuselage is predicted using the trained neural network model, and the longitudinal position of the trial joystick is adjusted so that the absolute value of the lateral angular acceleration around the fuselage is less than the error threshold; it is checked whether all accelerations are balanced. If so, the longitudinal balancing numerical calculation is terminated; if not, the normal gravitational acceleration and axial gravitational acceleration in the fuselage coordinate system are determined to continue the iteration.

2. The aircraft control response data fitting method based on neural network according to claim 1, characterized in that: Acquiring historical flight parameter data, building a neural network model, and training the neural network model based on the historical flight parameter data to obtain a trained neural network model, including: Acquiring historical flight parameter data, constructing a ground coordinate system and an aircraft body coordinate system, and performing analytical calculations on the historical flight parameter data based on the ground coordinate system and the aircraft body coordinate system to obtain aircraft response data; determining input data according to the historical flight parameter data and the aircraft response data, and performing data enhancement on the input data based on image conservation to obtain enhanced input data; Constructing a training set and a validation set based on the enhanced input data; Construct a neural network model, train the neural network model using the training set, test the neural network model using the validation set, determine the model weight when the prediction performance is optimal, and obtain a trained neural network model.

3. The aircraft control response data fitting method based on neural network according to claim 2, characterized in that: The input data includes: remaining fuel, airspeed, angle of attack, sideslip angle, axial angular velocity around the aircraft body, lateral angular velocity around the aircraft body, normal angular velocity around the aircraft body, roll angle, pitch angle, engine speed, longitudinal position of the joystick, lateral position of the joystick, pedal position and throttle lever position.

4. The neural network-based aircraft control response data fitting method according to claim 1, characterized in that: Setting an initial environmental state and an aircraft state of the aircraft, and determining the aircraft's airspeed, angle of attack, and sideslip angle according to the environmental state and the aircraft state, including: Set the initial environment state and aircraft state of the aircraft; determining a speed of the aircraft relative to the airflow in a ground coordinate system according to the environmental state and the aircraft state; Converting the velocity of the aircraft relative to the airflow in the ground coordinate system to the velocity of the aircraft relative to the airflow in the body coordinate system; The real-time airspeed, real-time angle of attack and real-time sideslip angle of the aircraft are determined according to the speed relative to the airflow in the aircraft body coordinate system.

5. The aircraft control response data fitting method based on neural network according to claim 1, characterized in that: Updating the aircraft state and engine speed according to the predicted acceleration, angular acceleration, and engine speed change rate to achieve fitting of aircraft control response data, including: Determine the velocity in the body coordinate system based on the predicted acceleration; Determine the speed in the ground coordinate system according to the speed in the body coordinate system; Updating the aircraft's ground coordinate system position according to the velocity in the ground coordinate system; Determine the angular velocity in the body coordinate system according to the predicted angular acceleration; Determine the angular velocity in the ground coordinate system according to the angular velocity in the body coordinate system; Updating the aircraft ground coordinate system attitude according to the angular velocity in the ground coordinate system; The engine speed is updated according to the predicted engine speed change rate to achieve the fitting of the aircraft control response data.

6. The neural network-based aircraft control response data fitting method according to claim 1, characterized in that: Determining a theoretical pitch angle according to the trim condition and the trial airspeed, and determining a normal gravity acceleration and an axial gravity acceleration in an aircraft coordinate system based on the theoretical pitch angle, including: Using trim conditions and trial airspeed Determine the theoretical pitch angle , the balancing conditions include: target adverse wind speed Target glide angle , target angle of attack , theoretical pitch angle The expression is: in, is the glideslope angle of the aircraft relative to the air, expressed as: in, is the aircraft ground speed, To try out the airspeed, is the target glide angle; The aircraft ground speed The expression is: Among them, A, B, and C are parameters, and the expressions are as follows: in, is the target glide angle, is the target reverse wind speed, To try to find the airspeed; Based on the theoretical pitch angle Determine the axial gravitational acceleration in the body coordinate system , the expression is: in, is the acceleration due to gravity; Based on the theoretical pitch angle Determine the normal gravitational acceleration in the body coordinate system , the expression is .

7. A neural network-based aircraft control response data fitting system, characterized in that: The neural network-based aircraft control response data fitting system includes: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the neural network-based aircraft control response data fitting method as described in any one of claims 1 to 6.

8. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the neural network-based aircraft control response data fitting method according to any one of claims 1 to 6 are implemented.

Images