Integrated deformation control method for aircraft facing dive strike
Through the integrated deformation control method of the aircraft facing dive strike, the problem that traditional control methods are difficult to ensure the deceleration performance and control stability in dive tasks is solved, and the precise control of the terminal speed of the aircraft is realized and the effective tracking of the guidance commands is achieved, ensuring the efficient completion of the task.
Patent Information
- Application Number
- CN202510336512.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-20
- Publication Date
- 2025-06-20
AI Technical Summary
Traditional open-loop deformation control is difficult to ensure deceleration performance and control stability during the dive task, especially when aerodynamic characteristics and external environment change drastically.
An integrated deformation control method for aircraft oriented to dive strikes is designed. By incorporating the deformation amount into the system control input, an integrated deformation control optimization framework is constructed, and an intelligent adaptive controller and parallel estimation model are adopted, and a sequence quadratic planning method is combined to optimize the deflection and deformation of the aerodynamic rudder surface.
It realizes precise control of the speed of the aircraft end and effective tracking of guidance instructions, improves the estimation performance of aerodynamic uncertainty and external interference, and ensures the efficient completion of dive strike missions.
Smart Images
Figure CN120178682A_ABST
Abstract
Description
Technical Field
[0003] The present invention relates to a method for controlling an aircraft, in particular to an integrated deformation control method for an aircraft facing dive strikes, and belongs to the field of aircraft control. Prior Art
[0004] During the terminal depression section of a reentry vehicle, it is necessary to dynamically adjust the drag to precisely control the flight speed to ensure the normal operation of the vehicle's detection system and complete the accurate strike on the target. However, the vehicle faces complex process constraints during the dive mission, and at the same time, the aerodynamic characteristics and the external environment change violently. These factors make it difficult for traditional open-loop deformation control to ensure deceleration performance and control stability, posing challenges to the design of control methods.
[0005] In the article "Adaptive super-twisting sliding mode control of variable sweep morphing aircraft" (Binbin Yan, Pei Dai, Ruifan Liu, Muzeng Xing and Shuangxi Liu, "Aerospace Science and Technology", 2019, 92: 198-210.), an adaptive control strategy was designed for the longitudinal model of a morphing aircraft, and the super-twisting sliding mode robust control strategy was used to resist the influence of uncertainties in the morphing model. The deformation amount of this method is set as an open-loop command that only depends on the altitude, which cannot adapt to the complex mission requirements during the dive section and cannot fully utilize the optimal aerodynamic configuration advantage of the morphing aircraft.
[0006] Object of the Invention
[0007] The present invention relates to an integrated deformation control method for an aircraft facing dive strikes. This method incorporates the deformation amount into the system control input, constructs an integrated optimization framework for deformation control, and ensures the effective execution of the dive strike mission. For the aircraft attitude dynamics model, an intelligent adaptive controller is designed to obtain the desired control moment, and a composite estimation strategy based on the prediction error and the tracking error is established through a parallel estimation model to improve the estimation performance of aerodynamic uncertainties and external disturbances and the command tracking accuracy. Considering the over-driven control problem caused by the increase in the input dimension, with the dive speed, deformation amount, and aerodynamic control surface deflection limit as the constraint conditions, and minimizing the control moment deviation, aerodynamic drag, and energy loss as the optimization objective function, an optimal allocation model is established, and the sequential quadratic programming method is used to solve the optimal aerodynamic control surface deflection and deformation amount. This method ensures the precise control of the terminal speed and the effective tracking of the guidance command through the integrated coordination of deformation and control, providing technical support for the efficient completion of the dive strike mission.
[0008] Content of the Invention (corresponding to the claims)
[0009] The technical solution adopted by the present invention to solve its technical problems is: an integrated deformation control method for aircraft facing dive strikes, which is realized through the following steps:
[0010] (a) Considering the six-degree-of-freedom nonlinear model of the variable aircraft as
[0011]
[0012] The state variables of the kinematic model are x = [V, α, β, γ V , p, q, r] T , and the control inputs are u = [δ e , δ a , δ r , δ Λ T , where V represents the flight speed, α, β, and γ V represent the angle of attack, sideslip angle, and bank angle respectively, p, q, and r represent the roll rate, pitch rate, and yaw rate respectively, δ e , δ a and δ r represent the elevator, aileron, and rudder deflections respectively, δ Λ represents the wing sweep angle, D, Y, and L represent the drag, side force, and lift respectively, M and N represent the roll moment, pitch moment, and yaw moment respectively, θ represents the flight path inclination angle, σ represents the flight path deviation angle, I xx , I yy and I zz represent the three-axis moments of inertia respectively, I xz represents the product of inertia between the x and z axes, m represents the mass of the aircraft, and g represents the acceleration due to gravity.
[0013] (b) The expressions for force and moment are respectively
[0014]
[0015] In the formula, Q represents the dynamic pressure, S w (δ Λ ) represents the wing reference area, c(δ Λ ) represents the mean aerodynamic chord length, b(δ Λ ) represents the wingspan, and the expressions for the aerodynamic coefficients are respectively
[0016]
[0017] In the formula, C L0 (V, α, δ Λ ), C Lq (δ Λ )、C D0 (V,α,δ Λ )、 C Yβ (α,δ Λ )、 C Yp (δ Λ )、C Yr (δ Λ )、 C M0 (V,α,δ Λ )、 C Mq (δ Λ )、C Nβ (α,δ Λ )、 C Np (δ Λ )、C Nr (δ Λ ) all represent aerodynamic parameters.
[0018] (c) Decouple the variant aircraft model to obtain the angle of attack subsystem, sideslip angle subsystem, and bank angle subsystem.
[0019] From equations (2) and (6), the angle of attack subsystem is written as
[0020]
[0021] where g q represents the pitch control coefficient, v q represents the pitch control moment, d α and d q represent pitch channel disturbances, and f α and f q are expressed as
[0022]
[0023] From equations (3) and (7), the sideslip angle subsystem is written as
[0024]
[0025] where g r represents the yaw control coefficient, v r represents the yaw control moment, d β and d r represent yaw channel disturbances, and f β and f r are expressed as
[0026]
[0027] From equations (4) and (5), the bank angle subsystem is written as
[0028]
[0029] where g p represents the roll control coefficient, v p represents the roll control moment, d γV and d p represent the roll channel disturbances, f γ and f p The expressions are respectively
[0030]
[0031] (d) Design an adaptive control method for the angle of attack subsystem.
[0032] Step 1: Considering the angle of attack dynamics, use a neural network to approximate the unknown aerodynamic function f α , we can get
[0033]
[0034] where represents the optimal weight estimated by the neural network, θ fα (x) represents the basis function vector,[[]] ε α represents the neural network approximation error,[[]] represents the design parameter.[[]]
[0035] Define the angle of attack tracking error e α = α - α d , α d represents the angle of attack reference command, and the virtual control quantity is designed as
[0036]
[0037] where k α > 0 represents the design parameter,[[]] represents the estimated value of f α ,[[]] represents the estimated value of the optimal weight vector ,[[]] represents the estimated value of D α .[[]]
[0038] Introduce a first-order filter as
[0039]
[0040] where σ α > 0 represents the filter parameter, qd Denote q c The value obtained after passing through a first-order filter.
[0041] Define the pitch rate tracking error e q = q - q d , and the derivative of the angle of attack tracking error is
[0042]
[0043] where
[0044] Design the filter compensation signal as
[0045]
[0046] where the compensation signal c q is given in the next step.
[0047] Define the compensated angle of attack tracking error as
[0048]
[0049] Define the prediction error as
[0050]
[0051] where λ α > 0 represents a design parameter.
[0052] Design the composite learning update law as
[0053]
[0054] where Γ α > 0, ρ zα > 0 and represent design parameters.
[0055] Design the disturbance observer as
[0056]
[0057] where P Lα > 0 represents a design parameter.
[0058] Step 2: Considering the pitch rate dynamics, use a neural network to approximate the unknown aerodynamic function f q , we can obtain
[0059]
[0060] where represents the optimal weight estimated by the neural network, represents the basis function vector, ε q represents the approximation error of the neural network, and represents the design parameter.
[0061] The designed pitch control moment is
[0062]
[0063] wherein, represents the estimated value of f q , represents the estimated value of the optimal weight vector , represents the estimated value of D q , and k q > 0 represents the design parameter.
[0064] The derivative of the pitch rate tracking error is
[0065]
[0066] wherein,
[0067] The designed filter compensation signal is
[0068]
[0069] Define the compensated pitch rate tracking error as
[0070]
[0071] Define the prediction error as
[0072]
[0073] wherein, λ q > 0 represents the design parameter.
[0074] The designed composite learning update law is
[0075]
[0076] wherein, Γ q > 0, ρ zq > 0 and represent the design parameters.
[0077] The designed disturbance observer is
[0078]
[0079] wherein, P Lq > 0 represents the design parameter.
[0080] (e) Design an adaptive control method for the sideslip angle subsystem.
[0081] Step 1: Considering the sideslip angle dynamics, use a neural network to approximate the unknown aerodynamic function f β , we can obtain
[0082]
[0083] where represents the estimated optimal weight of the neural network, represents the basis function vector, ε β represents the neural network approximation error, represents the design parameter.
[0084] Define the sideslip angle tracking error e β =β - β d , β d represents the sideslip angle reference command, and the virtual control quantity is designed as
[0085]
[0086] where k β >0 represents the design parameter, represents the estimated value of f β , represents the estimated value of the optimal weight vector , represents the estimated value of D β .
[0087] Introduce a first-order filter as
[0088]
[0089] where σ β >0 represents the filtering parameter, r d represents the value obtained after r c passes through the first-order filter.
[0090] Define the yaw rate tracking error e r =r - r d , and the derivative of the sideslip angle tracking error is
[0091]
[0092] where
[0093] Design the filtered compensation signal as
[0094]
[0095] where the compensation signal cr It is given in the next step.
[0096] Define the compensated sideslip angle tracking error as
[0097]
[0098] Define the prediction error as
[0099]
[0100] where λ β > 0 represents a design parameter.
[0101] Design the composite learning update law as
[0102]
[0103] where Γ β > 0, ρ zβ > 0 and represent design parameters.
[0104] Design the disturbance observer as
[0105]
[0106] where P Lβ > 0 represents a design parameter.
[0107] Step 2: Consider the yaw rate dynamics and use a neural network to approximate the unknown aerodynamic function f r , and we can get
[0108]
[0109] where represents the optimal weight estimated by the neural network, represents the basis function vector, ε r represents the neural network approximation error, represents a design parameter.
[0110] Design the yaw control moment as
[0111]
[0112] where represents the estimated value of f r , represents the estimated value of the optimal weight vector , represents the estimated value of D r , k r > 0 represents a design parameter.
[0113] The derivative of the yaw rate tracking error is
[0114]
[0115] where
[0116] The designed filter compensation signal is
[0117]
[0118] Define the compensated yaw rate tracking error as
[0119]
[0120] Define the prediction error as
[0121]
[0122] where λ r > 0 represents the design parameter.
[0123] Design the composite learning update law as
[0124]
[0125] where Γ r > 0, ρ zr > 0 and represent the design parameters.
[0126] Design the disturbance observer as
[0127]
[0128] where P Lr > 0 represents the design parameter.
[0129] (f) Design an adaptive control method for the bank angle subsystem.
[0130] Step 1: Considering the bank angle dynamics, use a neural network to approximate the unknown aerodynamic function f γ , we can get
[0131]
[0132] where represents the optimal weight estimated by the neural network, represents the basis function vector, ε γ represents the neural network approximation error, represents the design parameter.
[0133] Define the bank angle tracking error eγ = γ v -γ Vd , γ Vd represents the tilt angle reference command, and the virtual control quantity is designed as
[0134]
[0135] where k γ > 0 represents the design parameter, represents the estimated value of f γ , represents the estimated value of the optimal weight vector , represents the estimated value of D γ .
[0136] The first-order filter is introduced as
[0137]
[0138] where σ γ > 0 represents the filtering parameter, and p d represents p c obtained after passing through the first-order filter.
[0139] Define the roll rate tracking error e p = p - p d , and the derivative of the tilt angle tracking error is
[0140]
[0141] where
[0142] The filtered compensation signal is designed as
[0143]
[0144] where the compensation signal c p will be given in the next step.
[0145] Define the compensated tilt angle tracking error as
[0146]
[0147] Define the prediction error as
[0148]
[0149] where λ γ > 0 represents the design parameter.
[0150] Design the composite learning update law as
[0151]
[0152] where Γ γ > 0, ρ zγ > 0 and represent design parameters.
[0153] The designed disturbance observer is
[0154]
[0155] where P Lγ > 0 represents a design parameter.
[0156] Step 2: Considering the roll rate dynamics, using a neural network to approximate the unknown aerodynamic function f p , we can obtain
[0157]
[0158] where represents the optimal weight estimated by the neural network, represents the basis function vector, ε p represents the approximation error of the neural network, represents a design parameter.
[0159] The designed roll control moment is
[0160]
[0161] where represents the estimated value of f p , represents the estimated value of the optimal weight vector , represents the estimated value of D p , k p > 0 represents a design parameter.
[0162] The derivative of the roll rate tracking error is
[0163]
[0164] where
[0165] The designed filter compensation signal is
[0166]
[0167] Define the compensated roll rate tracking error as
[0168]
[0169] Define the prediction error as
[0170]
[0171] where λ p > 0 represents a design parameter.
[0172] Design the composite learning update law as
[0173]
[0174] where Γ p > 0, ρ zp > 0 and represent design parameters.
[0175] Design the disturbance observer as
[0176]
[0177] where P Lp > 0 represents a design parameter.
[0178] (g) Define v = [v q , v r , v p T , and the relationship between the control moment and the control input can be obtained as
[0179] G(x)u = v (64)
[0180] where G(x) represents the control efficiency matrix, and its expression is
[0181]
[0182] where and are expressed as
[0183]
[0184] where
[0185]
[0186] The optimal control allocation problem is formulated as
[0187]
[0188] where π1(x, u) = u - u min , π2(x, u) = u max - u,[[]] π5(x, u) = V - V min , π6(x, u) = V max -V, v des represents the desired control moment calculated by the controller, W1, W2, and μ represent the optimization weights, and u min and u max represent the amplitude limits of the pneumatic rudder deflection and deformation, and υ min and υ max represent the speed limits of the pneumatic rudder deflection and deformation, V min and V max represent the flight speed limit, and k represents the iteration sequence.
[0189] (h) The sequential quadratic programming algorithm is used to solve the optimization allocation problem online, and the iterative solution is expressed as
[0190] u k+1 = u k + a k d k (69)
[0191] where a k represents the iteration step size, represents the gradient of the objective function, and H k represents the Hessian matrix at the current iteration point. The convergence condition of the iteration is
[0192]
[0193] where ε c represents the convergence accuracy, represents the Lagrange multiplier, represents the gradient of the inequality constraint.
[0194] (i) According to the obtained control input u, the variant aircraft model (1)-(7) is returned to achieve speed control and guidance command tracking.
[0195] Advantages of the Invention
[0196] The beneficial effects of the present invention compared with the prior art are as follows:
[0197] (1) The present invention incorporates the deformation into the system control input, constructs an integrated deformation control framework, realizes the terminal speed control by assisting in adjusting the flight resistance through the deformation, and jointly ensures the effective execution of the guidance command by the pneumatic rudder operation.
[0198] (2) The present invention uses intelligent learning and a disturbance observer to online estimate the lumped uncertainty, introduces a parallel estimation model to construct the prediction error, and significantly improves the uncertainty estimation performance through the performance evaluation feedback mechanism.
[0199] (3) The present invention comprehensively considers factors such as speed constraints, execution constraints, and energy losses to establish an allocation model, and uses the sequential quadratic programming method to solve for the optimal aerodynamic control surface deflection and deformation amount, thus solving the over - actuated control problem caused by the increase in the input dimension. Description of the Drawings
[0200] Figure 1 Integrated Deformation Control Method for Aircraft Facing Dive Strike
[0202] The present invention discloses an integrated deformation control method for aircraft facing dive strike, belonging to the field of aircraft control. This method incorporates the deformation amount into the system control input, constructs an integrated optimization framework for deformation control to ensure the effective execution of the dive strike mission. For the aircraft attitude dynamics model, an intelligent adaptive controller is designed to obtain the desired control moment, and a composite estimation strategy based on prediction error and tracking error is established through a parallel estimation model to improve the estimation performance of aerodynamic uncertainties and external disturbances and the command tracking accuracy. Considering the over - actuated control problem caused by the increase in the input dimension, with the dive speed, deformation amount, and aerodynamic control surface deflection limit as constraint conditions, and minimizing the control moment deviation, aerodynamic drag, and energy loss as the optimization objective function, an optimal allocation model is established, and the sequential quadratic programming method is used to solve for the optimal aerodynamic control surface deflection and deformation amount. This method ensures the precise control of the terminal speed and the effective tracking of the guidance command through the integrated coordination of deformation and control, providing technical support for the efficient completion of the dive strike mission.
[0203] 8. Implementation Examples
[0204] Refer to Figure 1 , the integrated deformation control method for aircraft facing dive strike. The specific steps are as follows:
[0205] (a) Consider the six - degree - of - freedom nonlinear model of the variable - geometry aircraft as
[0206]
[0207] The state variables of the kinematic model are x = [V, α, β, γ V , p, q, r] T , and the control input is u = [δ e , δ a , δ r , δ Λ T , where V represents the flight speed, α, β, and γ V represent the angle of attack, sideslip angle, and bank angle respectively, p, q, and r represent the roll rate, pitch rate, and yaw rate respectively, and δ e , δ a , and δ rrespectively represent the elevator, aileron and rudder deflections, δ Λ represents the wing sweep angle, D, Y and L respectively represent drag, side force and lift, M and N respectively represent rolling moment, pitching moment and yaw moment, θ represents the flight path angle of climb, σ represents the flight path deviation angle, I xx 、I yy and I zz respectively represent the three-axis moments of inertia, I xz represents the product of inertia between the x and z axes, m = 1800 kg, g = 9.8 m / s 2 .
[0208] (b) The expressions for forces and moments are respectively
[0209]
[0210] where Q represents the dynamic pressure, S w (δ Λ ) represents the wing reference area, c(δ Λ ) represents the mean aerodynamic chord length, b(δ Λ ) represents the wing span, and the expressions for aerodynamic coefficients are respectively
[0211]
[0212] where C L0 (V,α,δ Λ ), C Lq (δ Λ ), C D0 (V,α,δ Λ ), C Yβ (α,δ Λ ), C Yp (δ Λ ), C Yr (δ Λ ), C Lβ (α,δ Λ ), C M0 (V,α,δ Λ ), C Mq (δ Λ ), C Nβ (α,δ Λ ), C Np (δ Λ ), C Nr (δ Λ ) all represent aerodynamic parameters.
[0213] (c) Decouple the variable aircraft model to obtain the angle of attack subsystem, sideslip angle subsystem, and bank angle subsystem.
[0214] From equations (2) and (6), the angle of attack subsystem is written as
[0215]
[0216] where g q = 4, v q represents the pitch control moment, d α and d q represent the pitch channel disturbances, and the expressions for f α and f q are respectively
[0217]
[0218] From equations (3) and (7), the sideslip angle subsystem is written as
[0219]
[0220] where g r = 20, v r represents the yaw control moment, d β and d r represent the yaw channel disturbances, and the expressions for f β and f r are respectively
[0221]
[0222] From equations (4) and (5), the bank angle subsystem is written as
[0223]
[0224] where g p = -12.5, v p represents the roll control moment, d γV and d p represent the roll channel disturbances, and the expressions for f γ and f p are respectively
[0225]
[0226] (d) Design an adaptive control method for the angle of attack subsystem.
[0227] Step 1: Considering the angle of attack dynamics, use a neural network to approximate the unknown aerodynamic function f α , we can obtain
[0228]
[0229] In the formula, represents the optimal weight estimated by the neural network, represents the basis function vector, ε α represents the approximation error of the neural network,
[0230] Define the angle of attack tracking error e α =α - α d , α d represents the angle of attack reference command, and the virtual control quantity is designed as
[0231]
[0232] In the formula, k α =20, represents the estimated value of f α , represents the estimated value of the optimal weight vector , represents the estimated value of D α .
[0233] Introduce a first-order filter as
[0234]
[0235] In the formula, σ α =0.05 represents the filtering parameter, q d represents q c obtained after passing through the first-order filter.
[0236] Define the pitch rate tracking error e q =q - q d , and the derivative of the angle of attack tracking error is
[0237]
[0238] In the formula,
[0239] Design the filtering compensation signal as
[0240]
[0241] In the formula, the compensation signal c q will be given in the next step.
[0242] Define the compensated angle of attack tracking error as
[0243]
[0244] Define the prediction error as
[0245]
[0246] where λ α = 10.
[0247] Design the composite learning update law as
[0248]
[0249] where Γ α = 25, ρ zα = 1.7 and
[0250] Design the disturbance observer as
[0251]
[0252] where P Lα = 1.8.
[0253] Step 2: Consider the pitch rate dynamics and approximate the unknown aerodynamic function f q , and we can get
[0254]
[0255] where represents the optimal weight estimated by the neural network, represents the basis function vector, ε q represents the approximation error of the neural network,
[0256] Design the pitch control moment as
[0257]
[0258] where represents the estimated value of f q , represents the estimated value of the optimal weight vector , represents the estimated value of D q , k q = 20.
[0259] The derivative of the pitch rate tracking error is
[0260]
[0261] where
[0262] Design the filter compensation signal as
[0263]
[0264] Define the compensated pitch rate tracking error as
[0265]
[0266] Define the prediction error as
[0267]
[0268] where λ q = 3.5.
[0269] Design the composite learning update law as
[0270]
[0271] where Γ q = 15, ρ zq = 2.5 and
[0272] Design the disturbance observer as
[0273]
[0274] where P Lq = 0.5.
[0275] (e) Design an adaptive control method for the sideslip angle subsystem.
[0276] Step 1: Considering the sideslip angle dynamics, use a neural network to approximate the unknown aerodynamic function f β , we can obtain
[0277]
[0278] where represents the optimal weight estimated by the neural network, represents the basis function vector, ε β represents the neural network approximation error,
[0279] Define the sideslip angle tracking error e β = β - β d , β d represents the sideslip angle reference command, and design the virtual control quantity as
[0280]
[0281] where k β = 0.35, represents f βThe estimated value, represents the optimal weight vector The estimated value, represents D β The estimated value.
[0282] Introduce a first-order filter as
[0283]
[0284] where σ β = 0.05, r d represents the value obtained by r c after passing through the first-order filter.
[0285] Define the yaw rate tracking error e r = r - r d , and the derivative of the sideslip angle tracking error is
[0286]
[0287] where
[0288] Design the filter compensation signal as
[0289]
[0290] where the compensation signal c r is given in the next step.
[0291] Define the compensated sideslip angle tracking error as
[0292]
[0293] Define the prediction error as
[0294]
[0295] where λ β = 3.8.
[0296] Design the composite learning update law as
[0297]
[0298] where Γ β = 31, ρ zβ = 0.8 and
[0299] Design the disturbance observer as
[0300]
[0301] where P Lβ= 3.2。
[0302] Step 2: Considering the yaw rate dynamics, a neural network is used to approximate the unknown aerodynamic function f r , we can obtain
[0303]
[0304] where represents the optimal weight estimated by the neural network, represents the basis function vector, ε r represents the approximation error of the neural network,
[0305] The yaw control moment is designed as
[0306]
[0307] where represents the estimated value of f r , represents the estimated value of the optimal weight vector , represents the estimated value of D r , k r = 15。
[0308] The derivative of the yaw rate tracking error is
[0309]
[0310] where
[0311] The filter compensation signal is designed as
[0312]
[0313] The compensated yaw rate tracking error is defined as
[0314]
[0315] The prediction error is defined as
[0316]
[0317] where λ r = 2.2。
[0318] The composite learning update law is designed as
[0319]
[0320] where Γ r = 12, ρ zr= 0.5 and
[0321] Design the disturbance observer as
[0322]
[0323] where P Lr = 2.1.
[0324] (f) Design an adaptive control method for the bank angle subsystem.
[0325] Step 1: Considering the bank angle dynamics, use a neural network to approximate the unknown aerodynamic function f γ , we can get
[0326]
[0327] where represents the optimal weight estimated by the neural network, represents the basis function vector, ε γ represents the neural network approximation error,
[0328] Define the bank angle tracking error e γ = γ v - γ Vd , γ Vd represents the bank angle reference command, and design the virtual control quantity as
[0329]
[0330] where k γ = 50, represents the estimated value of f γ , represents the estimated value of the optimal weight vector , represents the estimated value of D γ .
[0331] Introduce a first-order filter as
[0332]
[0333] where σ γ = 0.05, p d represents the value obtained after p c passes through the first-order filter.
[0334] Define the roll rate tracking error e p = p - p d , and the derivative of the bank angle tracking error is
[0335]
[0336] wherein,
[0337] the designed filtering compensation signal is
[0338]
[0339] wherein, the compensation signal c p is given in the next step.
[0340] Define the compensated roll angle tracking error as
[0341]
[0342] Define the prediction error as
[0343]
[0344] wherein, λ γ = 2.5.
[0345] Design the composite learning update law as
[0346]
[0347] wherein, Γ γ = 10, ρ zγ 3.5 and
[0348] Design the disturbance observer as
[0349]
[0350] wherein, P Lγ = 5.5.
[0351] Step 2: Considering the roll rate dynamics, use a neural network to approximate the unknown aerodynamic function f p , we can obtain
[0352]
[0353] wherein, represents the optimal weight estimated by the neural network, represents the basis function vector, ε p represents the neural network approximation error, L fp = 1.5.
[0354] Design the roll control moment as
[0355]
[0356] In the formula, denotes the estimated value of f p , denotes the estimated value of the optimal weight vector , denotes the estimated value of D p , and k p = 150.
[0357] The derivative of the roll angle rate tracking error is
[0358]
[0359] In the formula,
[0360] The designed filtering compensation signal is
[0361]
[0362] Define the compensated roll angle rate tracking error as
[0363]
[0364] Define the prediction error as
[0365]
[0366] In the formula, λ p = 2.2.
[0367] Design the composite learning update law as
[0368]
[0369] In the formula, Γ p = 25, ρ zp = 2.3, and
[0370] Design the disturbance observer as
[0371]
[0372] In the formula, P Lp = 3.7.
[0373] (g) Define v = [v q , v r , v p T , and the relationship between the control moment and the control input satisfies
[0374] G(x)u = v (64)
[0375] In the formula, G(x) denotes the control efficiency matrix, and the expression is
[0376]
[0377] In the formula, and The expressions of are respectively
[0378]
[0379] In the formula,
[0380]
[0381] The optimal control allocation problem is formulated as
[0382]
[0383] In the formula, π1(x,u) = u - u min , π2(x,u) = u max -u, π5(x,u) = V - V min , π6(x,u) = V max -V, v des represents the expected control moment calculated by the controller, W1, W2 and μ represent the optimization weights, u min and u max represent the amplitude limits of the aerodynamic rudder deflection and deformation, υ min and υ max represent the speed limits of the aerodynamic rudder deflection and deformation,, V min and V max represent the flight speed limits, and k represents the iteration sequence.
[0384] (h) The sequential quadratic programming algorithm is used to solve the optimization allocation problem online, and the iterative solution is expressed as
[0385] u k+1 = u k + a k d k (69)
[0386] In the formula, a k represents the iteration step size, represents the gradient of the objective function, and H k represents the Hessian matrix at the current iteration point. The convergence condition of the iteration is
[0387]
[0388] In the formula, ε c represents the convergence accuracy, represents the Lagrange multiplier, represents the gradient of the inequality constraint.
[0389] (i) Based on the obtained control input u, return the variant aircraft model (1)-(7) to achieve speed control and guidance command tracking.
Claims
1. A method for controlling the integrated deformation of an aircraft for dive strikes, characterized in that: The control method comprises: 1) Incorporate the deformation into the system control input and establish a six-degree-of-freedom nonlinear model of the variant aircraft; 2) Decouple the model into angle of attack subsystem, sideslip angle subsystem and roll angle subsystem; 3) Design an adaptive control method for the angle of attack subsystem to obtain the desired control torque; 4) Design an adaptive control method for the sideslip angle subsystem to obtain the desired control torque; 5) Design an adaptive control method for the roll angle subsystem to obtain the desired control torque; 6) Establish an optimal allocation model and use sequential quadratic programming method to solve the optimal aerodynamic control surface deflection and deformation.
2. The control method according to claim 1, characterized in that: Considering the nonlinear model of the variant aircraft with six degrees of freedom: The state variables of the kinematic model are x=[V,α,β,γ V ,p,q,r] T , the control input is u=[δ e ,δ a ,δ r ,δ Λ ] T , V represents the flight speed, α, β and γ V represents the angle of attack, sideslip angle and roll angle respectively, p, q and r represent the roll angle rate, pitch angle rate and yaw angle rate respectively, δ e , δ a and δ r represent the elevator, aileron and rudder deflection angles respectively, δ Λ represents the wing sweep angle, D, Y and L represent drag, side force and lift respectively. M and N represent the rolling moment, pitching moment and yaw moment respectively, θ represents the track inclination angle, σ represents the track deviation angle, and I xx ,I yy and I zz They represent the inertia moments of the three axes, I xz represents the inertia product between the x and z axes, m represents the mass of the aircraft, and g represents the acceleration due to gravity. The expressions for force and torque are In the formula, Q represents dynamic pressure, S w (δ Λ ) represents the wing reference area, c(δ Λ ) represents the average aerodynamic chord length, b(δ Λ ) represents the wing span, and the expressions of aerodynamic coefficients are where C L0 (V, α, δ Λ ), C Lq (δ Λ ), C D0 (V, α, δ Λ ), C Yβ (α, δ Λ ), C Yp (δ Λ ), C Yr (δ Λ ), C M0 (V,a,d Λ )、 C Mq (d Λ )、C Nβ (a,d Λ )、 C Np (δ Λ ), C Nr (δ Λ ) represent aerodynamic parameters.
3. The control method according to claim 1, characterized in that: The variant aircraft model is decoupled to obtain the angle of attack subsystem, sideslip angle subsystem and roll angle subsystem. According to equations (2) and (6), the angle of attack subsystem is written as In the formula, g q represents the pitch control coefficient, v q is the pitch control moment, d α and d q represents the pitch channel disturbance, f α and f q The expressions are According to equations (3) and (7), the sideslip angle subsystem is written as In the formula, g r represents the yaw control coefficient, v r is the yaw control moment, d β and d r represents the yaw channel disturbance, f β and f r The expressions are According to equations (4) and (5), the roll angle subsystem can be written as In the formula, g p represents the roll control coefficient, v p is the rolling control torque, and d p represents the roll channel disturbance, f γ and f p The expressions are 4. The control method according to claim 1, characterized in that: Design an adaptive control method for the angle of attack subsystem. Step 1: Considering the angle of attack dynamics, a neural network is used to approximate the unknown aerodynamic function f α , can be obtained In the formula, represents the optimal weight estimated by the neural network, represents the basis function vector, ε α represents the neural network approximation error, Represents design parameters. Define the angle of attack tracking error e α =α-α d , α d represents the angle of attack reference command, and the designed virtual control quantity is In the formula, k α >0 indicates design parameters, represents f α The estimated value of represents the optimal weight vector The estimated value of Indicates D α The estimated value of . The first-order filter is introduced as In the formula, σ α >0 indicates filtering parameters, q d Indicates q c The value obtained after passing through a first-order filter. Define the pitch rate tracking error e q =qq d , the derivative of the angle of attack tracking error is In the formula, Design the filter compensation signal as In the formula, the compensation signal c q Given in the next step. The angle of attack tracking error after compensation is defined as The prediction error is defined as In the formula, λ α >0 indicates design parameters. Design the compound learning update law as In the formula, Γ α >0,ρ zα >0 and Represents design parameters. The disturbance observer is designed as Where P Lα >0 indicates design parameters. Step 2: Consider the pitch rate dynamics and use a neural network to approximate the unknown aerodynamic function f q , can be obtained In the formula, represents the optimal weight estimated by the neural network, represents the basis function vector, ε q represents the neural network approximation error, Represents design parameters. The designed pitch control moment is In the formula, represents f q The estimated value of represents the optimal weight vector The estimated value of Indicates D q The estimated value of k q >0 indicates design parameters. The derivative of the pitch rate tracking error is In the formula, Design the filter compensation signal as The pitch rate tracking error after compensation is defined as The prediction error is defined as In the formula, λ q >0 indicates design parameters. Design the compound learning update law as In the formula, Γ q >0,ρ zq >0 and Represents design parameters. The disturbance observer is designed as Where P Lq >0 indicates design parameters.
5. The control method according to claim 1, characterized in that: Design an adaptive control method for the sideslip angle subsystem. Step 1: Considering the sideslip angle dynamics, a neural network is used to approximate the unknown aerodynamic function f β , can be obtained In the formula, represents the optimal weight estimated by the neural network, represents the basis function vector, ε β represents the neural network approximation error, Represents design parameters. Define the sideslip angle tracking error e β =β-β d , β d Represents the sideslip angle reference command, and the designed virtual control quantity is In the formula, k β >0 indicates design parameters, represents f β The estimated value of represents the optimal weight vector The estimated value of Indicates D β The estimated value of . The first-order filter is introduced as In the formula, σ β >0 indicates filtering parameters, r d Represents r c The value obtained after passing through a first-order filter. Define the yaw rate tracking error e r =rr d , the derivative of the sideslip angle tracking error is In the formula, Design the filter compensation signal as In the formula, the compensation signal c r Given in the next step. The sideslip angle tracking error after compensation is defined as The prediction error is defined as In the formula, λ β >0 indicates design parameters. Design the compound learning update law as In the formula, Γ β >0,ρ zβ >0 and Represents design parameters. The disturbance observer is designed as Where P Lβ >0 indicates design parameters. Step 2: Considering the yaw rate dynamics, a neural network is used to approximate the unknown aerodynamic function f r , can be obtained In the formula, represents the optimal weight estimated by the neural network, represents the basis function vector, ε r represents the neural network approximation error, Represents design parameters. The designed yaw control moment is In the formula, represents f r The estimated value of represents the optimal weight vector The estimated value of Indicates D r The estimated value of k r >0 indicates design parameters. The derivative of the yaw rate tracking error is In the formula, Design the filter compensation signal as The yaw rate tracking error after compensation is defined as The prediction error is defined as In the formula, λ r >0 indicates design parameters. Design the compound learning update law as In the formula, Γ r >0,ρ zr >0 and Represents design parameters. The disturbance observer is designed as Where P Lr >0 indicates design parameters.
6. The control method according to claim 1, characterized in that: Design an adaptive control method for the roll angle subsystem. Step 1: Considering the roll angle dynamics, a neural network is used to approximate the unknown aerodynamic function f γ , can be obtained In the formula, represents the optimal weight estimated by the neural network, represents the basis function vector, ε γ represents the neural network approximation error, Represents design parameters. Define the roll angle tracking error e γ =γ v -γ Vd , γ Vd Represents the reference command of the roll angle, and the designed virtual control quantity is In the formula, k γ >0 indicates design parameters, represents f γ The estimated value of represents the optimal weight vector The estimated value of Indicates D γ The estimated value of . The first-order filter is introduced as In the formula, σ γ >0 indicates filtering parameters, p d Indicates p c The value obtained after passing through a first-order filter. Define the roll angular rate tracking error e p =pp d , the derivative of the roll angle tracking error is In the formula, Design the filter compensation signal as In the formula, the compensation signal c p Given in the next step. The compensated roll angle tracking error is defined as The prediction error is defined as In the formula, λ γ >0 indicates design parameters. Design the compound learning update law as In the formula, Γ γ >0,ρ zγ >0 and Represents design parameters. The disturbance observer is designed as Where P Lγ >0 indicates design parameters. Step 2: Considering the roll rate dynamics, a neural network is used to approximate the unknown aerodynamic function f p , can be obtained In the formula, represents the optimal weight estimated by the neural network, represents the basis function vector, ε p represents the neural network approximation error, Represents design parameters. The design rolling control torque is In the formula, represents f p The estimated value of represents the optimal weight vector The estimated value of Indicates D p The estimated value of k p >0 indicates design parameters. The derivative of the roll rate tracking error is In the formula, Design the filter compensation signal as The roll angular rate tracking error after compensation is defined as The prediction error is defined as In the formula, λ p >0 indicates design parameters. Design the compound learning update law as In the formula, Γ p >0,ρ zp >0 and Represents design parameters. The disturbance observer is designed as Where P Lp >0 indicates design parameters.
7. The control method according to claim 1, characterized in that: The optimal allocation model is established, and the sequential quadratic programming method is used to solve the optimal aerodynamic control surface deflection and deformation. Define v = [v q ,v r ,v p ] T , the relationship between the operating torque and the control input satisfies G(x)u=v (64) In the formula, G(x) represents the control efficiency matrix, which is expressed as In the formula, and The expressions of i=e,a,r,Λ are respectively In the formula, The optimal control allocation problem can be formulated as In the formula, π1(x,u)=uu min ,π2(x,u)=u max -u, π5(x,u)=VV min ,π6(x,u)=V max -V,v des represents the desired steering torque calculated by the controller, W1, W2 and μ represent the optimization weights, and u min and u max Indicates the limit of the deflection and deformation of the aerodynamic control surface, υ min and max Indicates the speed limit of the deflection and deformation of the aerodynamic control surface, V min and V max represents the flight speed limit, and k represents the iteration sequence. The sequential quadratic programming algorithm is used to solve the optimization allocation problem online, and the iterative solution is expressed as u k+1 =u k +a k d k (69) In the formula, a k represents the iteration step length, represents the gradient of the objective function, H k Represents the Hessian matrix of the current iteration point. The convergence condition of the iteration is In the formula, ε c represents the convergence accuracy, represents the Lagrange multiplier, represents the gradient of the inequality constraint.