Aircraft parameter offline strategy library establishment method

By establishing a vehicle dynamics model and using wavelet neural network and deep forest to identify historical flight data, the problems of low accuracy of parameter information and uncertainty in the offline parameter strategy library of the aircraft are solved, and higher control is achieved Accuracy and robustness.

CN118915446AInactive Publication Date: 2024-11-08RES & DEV INST OF NORTHWESTERN POLYTECHNICAL UNIV IN SHENZHEN +2

Patent Information

Application Number
CN202410959229.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the construction of the offline parameter strategy library of aircraft, the problem of low accuracy of parameter information and strong uncertainty of model parameters leads to insufficient control accuracy and robustness.

Method used

By obtaining the speed, altitude, flight path angle, angle of attack and pitch angle velocity of the aircraft, establishing a dynamic model, using wavelet neural network and deep forest to identify historical flight data, correct the parameter interpolation table, and establish an offline parameter strategy library.

Benefits of technology

It improves the accuracy of the aircraft model parameters, reduces the uncertainty of the model parameters, enhances the control accuracy and robustness, has economic benefits and broad application prospects.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915446A_ABST
    Figure CN118915446A_ABST
Patent Text Reader

Abstract

The invention discloses an aircraft parameter offline strategy library establishment method, which relates to the technical field of aircraft parameter identification, and comprises the following steps of: establishing an aircraft kinetic model, and obtaining a curve approximate fitting expression and a measurement equation of force and torque according to the model; according to the curve approximate fitting expression of the force and the torque, obtaining a to-be-identified parameter; combining the measurement equation, the curve approximate fitting expression of the force and the torque and the to-be-identified parameters to obtain an identification model; based on the identification model, performing parameter identification on the historical flight data of the aircraft through a wavelet neural network to obtain a corrected parameter value; and correcting the original parameter interpolation table through the correction parameter value, performing learning fitting on the corrected interpolation table data, and establishing an offline parameter strategy library. Compared with a traditional parameter interpolation table, the method can provide more accurate high-speed aircraft model parameter information, and can solve the problem of strong uncertainty of aircraft model parameters.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of aircraft parameter identification, and in particular to a method for establishing an aircraft parameter offline strategy library. Background Art

[0002] With the rapid development of aerospace technology, the structure and mission requirements of high-speed aircraft are becoming more and more complex, and the accuracy requirements for control systems are becoming higher and higher. Due to the complex flight environment of high-speed aircraft across a large airspace and wide speed range, the aerodynamic and aerothermal environment changes dramatically, the aircraft aerodynamic and torque models are complex and highly nonlinear, and the model parameters are highly uncertain. Therefore, if the aircraft model parameters can be accurately estimated, it can not only enhance the robustness and adaptability of the system, but also improve the control accuracy of the system.

[0003] The construction of an aircraft offline parameter strategy library requires obtaining aircraft model parameter information. Existing acquisition methods include: (1) obtaining parameters by analyzing test data through wind tunnel tests, (2) using computational fluid dynamics for theoretical calculations, and (3) performing parameter identification based on measurement data from actual flight tests. However, wind tunnel tests are limited in size and can only use scaled models of aircraft. In addition, they are affected by factors such as support interference and tunnel wall interference, and are unable to simulate the actual complex flight environment. Therefore, the parameters obtained are greatly deviated from the actual real data. The inadequacy of aerodynamic theory will lead to large differences between the numerical results of computational fluid dynamics and the derivatives of aerodynamic parameters under actual flight conditions. The parameters are uncertain, which reduces the control performance of the aircraft. Actual flight tests are expensive and cannot be repeated multiple times, so the data sample size is small.

[0004] In summary, the construction of the aircraft offline parameter strategy library has the problems of low accuracy of the acquired parameter information and strong uncertainty in the constructed aircraft model parameters. Summary of the invention

[0005] The present invention provides a method for establishing an aircraft parameter offline strategy library, which solves the problems of low accuracy of parameter information in the existing parameter strategy library and strong uncertainty of model parameters.

[0006] The present invention provides a method for establishing an aircraft parameter offline strategy library, comprising the following steps:

[0007] Obtaining the speed, altitude, flight path angle, angle of attack and pitch velocity of the aircraft, and establishing an aircraft dynamics model based on the speed, altitude, flight path angle, angle of attack and pitch velocity of the aircraft;

[0008] Obtaining curve approximate fitting expressions and measurement equations of forces and moments according to the aircraft dynamics model;

[0009] The parameters to be identified are obtained by approximate fitting expressions of the force and moment curves;

[0010] The measurement equation, the curve approximate fitting expression of force and moment and the parameters to be identified are combined to obtain the identification model of force coefficient and moment coefficient;

[0011] Based on the identification model of force coefficient and moment coefficient, the parameters of the aircraft historical flight data are identified through wavelet neural network to obtain the corrected parameter values;

[0012] The original parameter interpolation table is corrected by correcting the parameter values;

[0013] Based on the deep forest, the corrected interpolation table data is learned and fitted to establish an offline parameter strategy library.

[0014] Preferably, the aircraft dynamics model is as follows:

[0015]

[0016] Where V, h, γ, α, and Q are the speed, altitude, flight path angle, angle of attack, and pitch velocity of the aircraft, respectively. are the derivatives of V, h, γ, α, Q, m, g, I y are the mass, gravitational acceleration and moment of inertia of the aircraft respectively; T, D, L, and M are the thrust, drag, lift, and pitch moment of the aircraft respectively.

[0017] Preferably, the curve approximate fitting expression of the force and moment is as follows:

[0018]

[0019] In the formula, δ e ,δ c and Φ are respectively the elevator deflection angle, canard deflection angle and fuel equivalence ratio, is the thrust fitting parameter, and is the thrust coefficient derivative, and is the derivative of the drag coefficient, and is the lift coefficient derivative, With c e is the torque coefficient derivative, S, With z T are the reference area, the average aerodynamic chord length and the thrust-torque coupling coefficient, and ρ are dynamic pressure and atmospheric density respectively, ρ0, h0 and h s is the fitting parameter of atmospheric density;

[0020] The measurement equation is as follows:

[0021]

[0022] in,

[0023]

[0024] Where n x 、n y 、n z is the triaxial overload signal of the projectile, ω x ,ω y ,ω z is the three-axis angular velocity of the projectile, and the component of the velocity in the projectile coordinate system is Moment of inertia I x =I y =I z is a constant, k is the current sampling time, and Δt is the sampling time interval.

[0025] Preferably, the parameters to be identified include thrust parameters and aerodynamic force and torque parameters, the thrust parameters include, β1, β2, β3, β4, β5, β6, β7, β8, the aerodynamic force and torque parameters include c e .

[0026] Preferably, the identification model of the force coefficient and the moment coefficient is as follows:

[0027]

[0028] Where, T is thrust, C is L , C D With C M They are lift coefficient, drag coefficient and pitching moment coefficient respectively.

[0029] Preferably, the wavelet neural network uses Mexican Hat wavelet function as wavelet element to replace traditional BP neural network neurons for forward propagation, and adds linear terms to compensate the output of the hidden layer;

[0030] The forward channel multi-input and multi-output expression of the wavelet neural network is:

[0031]

[0032] Where X = [x1, x2, …, x m ] T is the input, Y=[y1,y2,…,y n ] Tis the output, ψ(·) is the wavelet function, P is an n×m constant matrix, Q is an n×1 constant matrix, a and b are the one-dimensional wavelet function scaling and translation parameters, respectively, and w j is an n×1 weight matrix.

[0033] Preferably, the wavelet neural network performs parameter identification on the historical flight data of the aircraft by using a partial differentiation method;

[0034] Output layer Y = [y1, y2, ..., y q ] T For the input layer X = [x1, x2, ..., x p ] T The partial differential of can be expressed as:

[0035]

[0036] Among them, z is ψ(||aX-b||), y norm With x norm are the normalized parameters of output and input respectively.

[0037] Preferably, the aircraft historical flight data is obtained from a limited number of actual flight tests of the aircraft.

[0038] Preferably, the original parameter interpolation table is obtained through a wind tunnel test.

[0039] Preferably, learning and fitting the corrected interpolation table data based on the deep forest to establish an offline parameter strategy library includes the following steps:

[0040] The corrected interpolation table data is used as the data set, of which 70% is used as the training set and 30% is used as the test set;

[0041] The fixed sliding window step size of the multi-granularity scanning part is set to 1, the number of random forest and completely random forest base decision trees is set to 50, and the number of random forest and completely random forest base decision trees of the cascade forest part is set to 200;

[0042] The deep forest model is trained through the training set, and prediction is performed through the test set samples to obtain the offline parameter strategy library.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] The present invention derives an identification model for force coefficients and moment coefficients, and then performs offline identification on the historical flight trajectory data of the aircraft. Finally, the original aircraft parameter interpolation table is corrected according to the aircraft model parameters obtained by offline identification, which can provide more accurate high-speed aircraft model parameter information compared to the traditional parameter interpolation table. For the corrected parameter interpolation table, the interpolation table data is learned and fitted to establish an offline parameter strategy library. According to this offline parameter strategy library, the required parameters are searched to replace the nominal parameters, which suppresses the strong uncertainty of the model parameters, and has greater economic benefits and broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 A schematic diagram of an implementation flow of a method for establishing an aircraft parameter offline strategy library according to the present invention;

[0047] Figure 2 A historical flight trajectory time-Mach number curve diagram of an embodiment of the present invention;

[0048] Figure 3 A historical flight trajectory time-pitch rudder deflection angle curve diagram of an embodiment of the present invention;

[0049] Figure 4 A historical flight trajectory time-pitch angular velocity curve diagram of an embodiment of the present invention;

[0050] Figure 5 A historical flight trajectory time-attack angle curve diagram of an embodiment of the present invention;

[0051] Figure 6 It is a curve diagram of the fitting result of the wavelet neural network to the thrust according to an embodiment of the present invention;

[0052] Figure 7 It is a curve diagram of the fitting result of the wavelet neural network to the lift coefficient according to an embodiment of the present invention;

[0053] Figure 8 It is a curve diagram of the fitting result of the wavelet neural network to the resistance coefficient according to an embodiment of the present invention;

[0054] Fig. 9 It is a curve diagram of the fitting result of the wavelet neural network to the pitch moment coefficient according to an embodiment of the present invention;

[0055] Fig.10It is a curve diagram of prediction results of the thrust parameter β1 by the offline parameter strategy library of an embodiment of the present invention;

[0056] Fig.11 This is a curve diagram of prediction results of the thrust parameter β2 by the offline parameter strategy library of an embodiment of the present invention;

[0057] Fig.12 This is a curve diagram of prediction results of the thrust parameter β3 by the offline parameter strategy library of an embodiment of the present invention;

[0058] Fig.13 This is a curve diagram of prediction results of the thrust parameter β4 by the offline parameter strategy library of an embodiment of the present invention;

[0059] Fig.14 This is a curve diagram of prediction results of the thrust parameter β5 by the offline parameter strategy library of an embodiment of the present invention;

[0060] Fig.15 This is a curve diagram of prediction results of the thrust parameter β6 by the offline parameter strategy library of an embodiment of the present invention;

[0061] Fig.16 This is a curve diagram of prediction results of the thrust parameter β7 by the offline parameter strategy library of an embodiment of the present invention;

[0062] Fig.17 This is a curve diagram of prediction results of the thrust parameter β8 by the offline parameter strategy library of an embodiment of the present invention;

[0063] Fig.18 The offline parameter strategy library of the embodiment of the present invention is a resistance parameter The prediction result curve graph;

[0064] Fig.19 The offline parameter strategy library of the embodiment of the present invention is a resistance parameter The prediction result curve graph;

[0065] Fig. 20 The offline parameter strategy library of the embodiment of the present invention is a resistance parameter The prediction result curve graph;

[0066] Fig.21 The offline parameter strategy library of the embodiment of the present invention is used to calculate the lift parameter The prediction result curve graph;

[0067] Fig. 22 The offline parameter strategy library of the embodiment of the present invention is used to calculate the lift parameter The prediction result curve graph;

[0068] Fig.23 The offline parameter strategy library of the embodiment of the present invention is used to calculate the pitch moment parameter The prediction result curve graph;

[0069] Fig.24 The offline parameter strategy library of the embodiment of the present invention is used to calculate the pitch moment parameter The prediction result curve graph;

[0070] Fig.25 The offline parameter strategy library of the embodiment of the present invention is used to calculate the pitch moment parameter The prediction result curve graph;

[0071] Fig.26 The offline parameter strategy library of the embodiment of the present invention is used to calculate the pitch moment parameter c e The prediction result curve graph. DETAILED DESCRIPTION

[0072] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0073] See also Figure 1 The present invention provides a method for establishing an aircraft parameter offline strategy library, comprising the following steps:

[0074] Step 1: Get the speed, altitude, flight path angle, angle of attack and pitch velocity of the aircraft, and establish the aircraft dynamics model based on the speed, altitude, flight path angle, angle of attack and pitch velocity of the aircraft. The model is as follows:

[0075]

[0076] Where V, h, γ, α, and Q are velocity, altitude, flight path angle, angle of attack, and pitch velocity, respectively; m, g, and I are y are mass, gravitational acceleration and moment of inertia respectively; T, D, L, M are thrust, drag, lift and pitch moment respectively. Each force and moment can be approximated by curve fitting:

[0077]

[0078] In the formula, δ e ,δ c and Φ are respectively the elevator deflection angle, canard deflection angle and fuel equivalence ratio, is the thrust fitting parameter, and is the thrust coefficient derivative, and is the derivative of the drag coefficient, and is the lift coefficient derivative, With c e is the torque coefficient derivative, S, With z T are the reference area, the average aerodynamic chord length and the thrust-torque coupling coefficient, and ρ are dynamic pressure and atmospheric density respectively, ρ0, h0 and h s is the fitting parameter of the atmospheric density.

[0079] Step 2: Based on the aircraft dynamics model, specifically, the measurement equation is obtained based on the relationship between the three-axis overload and angular acceleration information and aerodynamic force, torque and thrust in flight mechanics. The overload and angular velocity information can be directly measured by the aircraft sensor. The measurement equation is as follows:

[0080]

[0081] Where n x 、n y 、n z is the triaxial overload signal of the projectile, ω x ,ω y ,ω z is the three-axis angular velocity of the projectile, which can be directly measured; the moment of inertia I x =I y =I z is a constant; the angle of attack α cannot be measured directly, but is obtained through the component of the velocity in the missile body coordinate system. Let its component in the missile body coordinate system be The angle of attack is then given by:

[0082]

[0083] It is given by the form of angular velocity difference:

[0084]

[0085] Where k is the current sampling time and Δt is the sampling time interval.

[0086] According to the curve approximation fitting expression (2) of force and torque, considering the aircraft model for the control system, the simplified parameters to be identified are:

[0087] Thrust parameters: β1, β2, β3, β4, β5, β6, β7, β8;

[0088] Aerodynamic force and torque parameters: c e .

[0089] Combining equations (2) and (3), the identification model of force coefficient and moment coefficient can be obtained as follows:

[0090]

[0091] Where, T is thrust, C is L , C D With C M They are lift coefficient, drag coefficient and pitching moment coefficient respectively.

[0092] Step 3: Design the wavelet neural network structure. Traditional parameter identification mainly includes maximum likelihood method, recursive least squares method and Kalman filtering. It usually requires accurate model initial value prior information and is not suitable for fast time-varying and highly uncertain air-breathing high-speed aircraft. The present invention uses wavelet neural network for parameter identification, which solves the problem that traditional parameter identification methods require accurate model initial value prior information.

[0093] The Mexican Hat wavelet function is used as the wavelet element to replace the traditional BP neural network neurons for forward propagation.

[0094] Add linear terms Compensate the output of the hidden layer. The forward channel multi-input and multi-output expression of the wavelet neural network is:

[0095]

[0096] Where X = [x1, x2, …, x m ] T is the input, Y=[y1,y2,…,y n ] T is the output, ψ(·) is the wavelet function, P is an n×m constant matrix, Q is an n×1 constant matrix, a j , b j are the one-dimensional wavelet function scaling and translation parameters, w j is an n×1 weight matrix.

[0097] The Levenberg-Marquardt (LM) algorithm is used for error back propagation network training:

[0098] Δx=-H -1 (x)g(x) (8)

[0099] Where H(x) is the Hessian matrix of the error function and g(x) is the gradient.

[0100]

[0101] In the formula, J(x) is the Jacobian matrix of the error function:

[0102]

[0103] Where x=(w1,w2,...w j ,a1,a2,...a j ,b1,b2,...b j ). Calculate Δx at each step to update x, and the final Δx expression is:

[0104] Δx=[J T (x)J(x)+μI] -1 J T (x)e(x) (11)

[0105] Where μ is the proportional coefficient and I is the unit matrix.

[0106] Step 4: Offline parameter identification based on wavelet neural network.

[0107] The partial differential method is used for parameter identification, and the output layer Y = [y1, y2, ..., y q ] T For the input layer X = [x1, x2, ..., x p ] T The partial differential of can be expressed as:

[0108]

[0109] where z is ψ(||aX-b||); y norm With x norm They are the normalized parameters of output and input respectively. Taking the input layer as an example, the normalization method is given:

[0110]

[0111] In the formula, and Respectively represent the maximum and minimum values ​​of the normalized range; x i,max With x i,min are the maximum and minimum values ​​of each dimension respectively.

[0112] The expressions in formula (12) are:

[0113]

[0114]

[0115] Among them, W is the j The weight matrix composed of , the other parameters are the same as mentioned above.

[0116] Step 5: Encrypt and expand the aircraft parameter interpolation table after offline identification and correction by wavelet neural network.

[0117] Based on wavelet neural network, a limited number of historical flight data are identified offline to obtain more accurate aircraft model parameters than the original parameter interpolation table. By expanding the corrected parameter interpolation table using the interpolation method, more uniform and sufficient aircraft model parameter information can be obtained.

[0118] Step 6: Use the encrypted and expanded parameter interpolation table as a sample data set and establish an offline parameter strategy library based on deep forest.

[0119] Deep forest is an integrated algorithm with random forest and completely random forest as basic units, which is mainly divided into two parts: multi-granularity scanning and cascade forest. The modified and expanded data samples are used as the original input data set, of which 70% of the data are used as training sets and 30% of the data are used as test sets. The fixed sliding window step size of the multi-granularity scanning part is set to 1, and the number of random forest and completely random forest base decision trees is set to 50; the number of random forest and completely random forest base decision trees in the cascade forest part is set to 200. The deep forest model is trained on the training set, and the test set samples are predicted to evaluate the accuracy and generalization performance of the offline parameter strategy library.

[0120] By means of online query, the required parameters are found to replace the nominal parameters according to the offline parameter strategy library, which suppresses the strong uncertainty of the model parameters and has great economic benefits and broad application prospects.

[0121] Example

[0122] The specific process of establishing the off-line parameter strategy library of a high-speed aircraft described in the present invention is explained below by using Matlab simulation.

[0123] First, based on the nominal parameters of the forces and moments in the aircraft dynamics model, model parameter perturbations and structural parameter perturbations (18) are added. r represents the actual value of the model parameter, C0 is its nominal value, P r is the actual structural parameter and P0 is its nominal value.

[0124]

[0125] Among them, C r1 =P r =0.1, C r2 =0.2,

[0126] The historical flight trajectory is obtained by simulating the aircraft model after adding perturbations, such as Figures 2 to 5As shown in the figure. The flight Mach numbers of each trajectory are almost the same, and the coverage range is the same, but at the same Mach number, the command angle of attack is different, and the coverage range of the rudder angle is also different. It can be found that the angle of attack intervals vary greatly, thus forming the total angle of attack interval as the envelope range of the parameter interpolation table. By performing offline parameter identification on a limited number of historical flight data, the original parameter interpolation table is corrected.

[0127] The initial simulation parameters of the aircraft body dynamics model are set as follows: h = 26212.8m, V = 2270.76m / s, γ = 0°, α = 3.6386°, Q = 0°, Φ = 0.2514, δ e =11.4635°.

[0128] Then two wavelet neural networks are used to identify the thrust parameters and aerodynamic force and torque parameters near a certain feature point of the historical flight trajectory data.

[0129] For the wavelet neural network of thrust parameter identification, the input is: Φα 3 , α 3 , Φα 2 , α 2 , Φα, α, Φ; the output is thrust T. The training fitting curve of wavelet neural network for thrust is as follows Figure 6 As shown, it can be seen that the training results have a high degree of overlap with the true values ​​and the fitting effect is good.

[0130] For the wavelet neural network for aerodynamic force and moment coefficient identification, the input is: α 2 , α and δ e ; Output: lift coefficient C L , drag coefficient C D and the pitch moment coefficient C M The training curve of wavelet neural network for aerodynamic force and moment coefficient is as follows: Figures 7 to 9 shown.

[0131] The comparison between the identification results and the true values ​​of the thrust parameters and the derivatives of the aerodynamic force and moment coefficients is shown in Tables 1 and 2. It can be seen that after being subjected to aerodynamic perturbations, the partial differential method can better identify the results from the flight data.

[0132] Table 1 Thrust parameter identification results

[0133]

[0134] Table 2 Aerodynamic coefficient derivative identification results

[0135]

[0136] Finally, the parameter interpolation table after correction and expansion of wavelet neural network offline identification is used as the data set, and learning and fitting are performed based on deep forest to establish an offline parameter strategy library. The test results of the offline parameter strategy library on the prediction set are as follows: Figures 10 to 26 The mean square error (MSE) and determination coefficient (R 2 ) results show that the maximum MSE does not exceed 0.1536, R 2 The minimum is not less than 0.98, and the prediction results are basically consistent with the actual values. Therefore, the trained parameter strategy library model still has good prediction accuracy in the case of unknown flight states, and the model generalization ability is very strong. Compared with the original parameter interpolation table, the deep forest strategy library contains more parameter information and has good extrapolation performance for adjacent flight states, providing a reference for high-speed aircraft flight parameter data mining.

[0137] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0138] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for establishing an aircraft parameter offline strategy library, characterized in that: The following steps are involved: Obtaining the speed, altitude, flight path angle, angle of attack and pitch velocity of the aircraft, and establishing an aircraft dynamics model based on the speed, altitude, flight path angle, angle of attack and pitch velocity of the aircraft; Obtaining curve approximate fitting expressions and measurement equations of forces and moments according to the aircraft dynamics model; The parameters to be identified are obtained by approximate fitting expressions of the force and torque curves; The measurement equation, the curve approximate fitting expression of force and moment and the parameters to be identified are combined to obtain the identification model of force coefficient and moment coefficient; Based on the identification model of force coefficient and moment coefficient, the parameters of the aircraft historical flight data are identified through wavelet neural network to obtain the corrected parameter values; The original parameter interpolation table is corrected by correcting the parameter values; Based on the deep forest, the corrected interpolation table data is learned and fitted to establish an offline parameter strategy library.

2. The method for establishing an aircraft parameter offline strategy library according to claim 1, characterized in that: The aircraft dynamics model is as follows: Where V, h, γ, α, and Q are the speed, altitude, flight path angle, angle of attack, and pitch velocity of the aircraft, respectively. are the derivatives of V, h, γ, α, Q, m, g, I y are the mass, gravitational acceleration and moment of inertia of the aircraft respectively; T, D, L, and M are the thrust, drag, lift, and pitch moment of the aircraft, respectively.

3. The method for establishing an aircraft parameter offline strategy library according to claim 2, characterized in that: The curve approximate fitting expression of the force and moment is as follows: In the formula, δ e ,δ c and Φ are respectively the elevator deflection angle, canard deflection angle and fuel equivalence ratio, is the thrust fitting parameter, and is the thrust coefficient derivative, and is the derivative of the drag coefficient, and is the lift coefficient derivative, With c e is the torque coefficient derivative, S, With z T are the reference area, the average aerodynamic chord length and the thrust-torque coupling coefficient, q and ρ are the dynamic pressure and the atmospheric density, ρ0, h0 and h s is the fitting parameter of atmospheric density; The measurement equation is as follows: in, Where n x 、n y 、n z is the triaxial overload signal of the projectile, ω x ,ω y ,ω z is the three-axis angular velocity of the projectile, and the velocity components in the projectile coordinate system are Moment of inertia I x =I y =I z is a constant, k is the current sampling time, and Δt is the sampling time interval.

4. The method for establishing an aircraft parameter offline strategy library according to claim 3, characterized in that: The parameters to be identified include thrust parameters and aerodynamic force and moment parameters, wherein the thrust parameters include β1, β2, β3, β4, β5, β6, β7, and β8, and the aerodynamic force and moment parameters include 5. The method for establishing an aircraft parameter offline strategy library according to claim 4, characterized in that: The identification model of the force coefficient and moment coefficient is as follows: In the formula, C L , C D With C M They are lift coefficient, drag coefficient and pitching moment coefficient respectively.

6. The method for establishing an aircraft parameter offline strategy library according to claim 1, characterized in that: The wavelet neural network uses Mexican Hat wavelet function as wavelet element to replace traditional BP neural network neurons for forward propagation, and adds linear terms to compensate the output of the hidden layer; The forward channel multi-input and multi-output expression of the wavelet neural network is: Where X = [x1, x2, …, x m ] T is the input, Y=[y1,y2,…,y n ] T is the output, ψ(·) is the wavelet function, P is an n×m constant matrix, Q is an n×1 constant matrix, a and b are the one-dimensional wavelet function scaling and translation parameters, respectively, and w j is an n×1 weight matrix, and N is the number of input samples.

7. The method for establishing an aircraft parameter offline strategy library according to claim 1, characterized in that: The wavelet neural network performs parameter identification on the historical flight data of the aircraft by using a partial differentiation method; Output layer Y = [y1, y2, ..., y q ] T For the input layer X = [x1, x2, ..., x p ] T The partial differential of can be expressed as: Among them, z is ψ(||aX-b||), y norm With x norm are the normalized parameters of output and input respectively.

8. The method for establishing an aircraft parameter offline strategy library according to claim 1, characterized in that: The aircraft historical flight data is obtained from a limited number of actual flight tests of the aircraft.

9. The method for establishing an aircraft parameter offline strategy library according to claim 1, characterized in that: The original parameter interpolation table is obtained through wind tunnel testing.

10. A method for establishing an aircraft parameter offline strategy library as claimed in claim 1, characterized in that: Based on the deep forest, the corrected interpolation table data is learned and fitted to establish an offline parameter strategy library, including the following steps: The corrected interpolation table data is used as the data set, of which 70% is used as the training set and 30% is used as the test set; The fixed sliding window step size of the multi-granularity scanning part is set to 1, the number of random forest and completely random forest base decision trees is set to 50, and the number of random forest and completely random forest base decision trees of the cascade forest part is set to 200; The deep forest model is trained through the training set, and prediction is performed through the test set samples to obtain the offline parameter strategy library.

Citation Information

Patent Citations

  • Aircraft sliding mode control method based on parameter identification

    CN117908385A

  • Method for rapidly estimating parameters of high-speed aircraft based on aerodynamic parameter experience

    CN118211511A

  • System and method for estimation of battery state and health

    US20220057451A1

Cited By

  • Parameter correction method and system of fixed-wing unmanned aerial vehicle semi-physical simulation system

    CN120065774A