CN105843080A - Intelligent nonlinear control system for hypersonic morphing aircraft - Google Patents

Abstract

The invention discloses an intelligent nonlinear control system for a hypersonic morphing aircraft and belongs to the aviation aerospace propulsion control technical field. According to the intelligent nonlinear control system of the invention, according to problems in outer-loop stable tracking control of the hypersonic morphing aircraft, with influence on modeling caused by a morphing structure and influence on tracking control performance caused by model parameter uncertainty and external unknown interference considered, the control system is divided into three sub systems according to the state variable characteristics of the aircraft; control signals are obtained sequentially through using a backstepping method; approximation is performed on the unknown interference through adopting an RBF neural network, so that the robustness of the controller can be ensured; and according to the problems of difficulty in the derivation of virtual control signals and differential expansion, dynamic surface control thinking is introduced to make improvement. With the above technical schemes of the invention adopted, the global stability of the closed-loop system can be guaranteed, and the system has high tracking performance and robust performance.

Description

Adjustable wing hypersonic aircraft Intelligent Nonlinear control system

Technical field

The present invention relates to flight control system, particularly relate to a kind of adjustable wing hypersonic aircraft Intelligent Nonlinear control System processed, belongs to aviation space flight Solid rocket engine technical field.

Background technology

The thrust power that hypersonic aircraft is used is to start without carrying the supersonic combustion punching type of oxidant Machine, aerodynamic arrangement is body/engine integration design.Between elastic and the flight propulsion system of hypersonic aircraft There is the strongest coupling, the nonlinear characteristic of dummy vehicle is extremely serious, and flight course has quick time variation, high ultrasonic Acutely, various uncertainties are serious, and traditional classical control method cannot meet well in the aerodynamic characteristic change of speed aircraft Flight control system stability and the performance requirement of strong robustness.The real-time of flight control system to be ensured, robustness And stability, response speed and control accuracy are had higher requirement, this greatly promoted advanced person control method and The development of control theory key technology.

Adjustable wing hypersonic aircraft is the hypersonic aircraft with scalable winglet, is to climb to solve to take off Rise that section lift is not enough, lift-drag ratio is too small, meet flying speed and the big feature of flying height envelope scope and the one that designs can Become rotor aircraft.Adjustable wing hypersonic aircraft has the feature of hypersonic aircraft and adjustable wing aircraft, Ke Yiyong concurrently In high speed flight at high altitude, the feature of adjustable wing aircraft can be made to select to stretch out according to different flight environment of vehicle and state of flight or It is to regain winglet, typically stretches out winglet when low-speed operations, improve lift, when high-speed flight, regain winglet to reduce resistance, Reduce fuel oil consumption.

Flight environment of vehicle, self aerodynamic characteristic complicated and changeable residing for adjustable wing hypersonic aircraft control system to flight The design of system brings a lot of technical difficult point.First, flight control system must is fulfilled for stability requirement.Large span is flown Envelope, serious external interference, the factor such as elastic deformation, high temperature and low density flow effect can have a strong impact on the stability of system. Second, flight control system must is fulfilled for robustness requirement.Under high dynamic pressure, high velocity environment, various external interference and internal ginseng Number change requires that flight control system must have stronger robustness.The residing atmospheric environment of aircraft is complicated so that flight Device quite sensitive, time variation is strong, it is easy to produce foundation structure distortion and parameter uncertainty；3rd, close coupling and non-linear Feature requires that system coordination controls.Adjustable wing hypersonic aircraft many employings Waverider or lifting body aerodynamic arrangement are to ensure The High Angle of Attack attitude of maneuvering flight, uses body/engine integration design not disintegrate when can ensure that high-speed flight.4th, Requirement of real-time.When high-speed flight, flight parameter has the feature of fierce fast time variant, and the control effect of pneumatic rudder face is on the contrary Falling sharply, SRT lengthens, it may appear that control latency issue.The reality of wing to be taken into full account deformation when controller designs Shi Xing, the complexity of control algolithm, to avoid controlling parameter too much, improve the speed of service of algorithm.5th, constraint bar to be met Part.Flight control system is ensureing while control accuracy, be also satisfied some constraintss, such as executing agency saturated about Bundle, the angle of attack and yaw angle constraint, climb and the hot-fluid constraint of reentry return section, arrange for ensureing housing construction intensity dynamic Pressure constraint and overload constraint etc..

Gain preset control method Application comparison in Control System Design is ripe and achieves certain achievement, when non- When linear system excursion is bigger, adopting and need to design multiple equalization point in this way, the stability of whole controller is difficult to It is guaranteed.Under hypersonic aircraft High Angle of Attack and high maneuvering condition, state of flight presents strong nonlinearity and high coupling, Gain preset method cannot meet the requirement of performance indications.

The dynamic inversion control method accurate modeling by controlled device Non-linear coupling characteristic, constitutes Non-linear coupling online Time-varying control device, to offset the Non-linear coupling time-varying characteristics of object, makes system become pseudo-linear system.But dynamic inversion Sensitive to modeling error, and under normal circumstances, nonlinear system Accurate Model is extremely difficult, once modeling and real system have difference Not, offseting of Non-linear coupling characteristic will have an impact, and causes the deterioration of control performance, it is impossible to ensures robustness.

Sliding-mode control, by designing discontinuous controller, forces system to produce sliding motion mode, once system Carrying out sliding mode, system will have Completely stableness to uncertain and interference.But, real system is due to switching device not Can there is inertia with avoiding, it is not accurate that variable structure system toggles causing actual sliding mode in different control logics Really occur on diverter surface, easily cause the violent shake of system, it is impossible to ensure the robustness of system.

In summary, for adjustable wing hypersonic aircraft, the strong nonlinearity existing due to himself and Uncertain external disturbance under flight environment of vehicle, above-mentioned existing control technology is all difficult to reach preferably to control effect.

Summary of the invention

The technical problem to be solved is to overcome prior art not enough, it is provided that a kind of adjustable wing is hypersonic flies Row device Intelligent Nonlinear control system, can effectively solve adjustable wing hypersonic aircraft strong nonlinearity characteristic and uncertain outside Flight control problem under portion's disturbance.

The present invention solves above-mentioned technical problem the most by the following technical solutions:

A kind of adjustable wing hypersonic aircraft Intelligent Nonlinear control system, is used for generating adjustable wing hypersonic flight Device

Control input quantity u, to ensure that adjustable wing hypersonic aircraft flying speed V, flying height H can quickly be followed the tracks of Flying speed reference value V to input_r, flying height reference value H_r；Described control system utilizes

Backstepping method designs, specifically include first be sequentially connected in series～the 3rd control subsystem and with First～the 3rd controls subsystem one to one first～third nerve network；Wherein,

First controls subsystem controls restrains specific as follows:

$v_1=-k_1{z}_1-{{\mathrm{\θ}}_1}^T{\mathrm{\ζ}}_1(x_1,{\mathrm{\μ}}_1,{\mathrm{\σ}}_1)+{\stackrel{\·}{y}}_r$

In formula, v₁It is the output of the first control subsystem；z₁It it is the tracking error of the first control subsystem；k₁For default system Number；For reference instructionFirst derivative；θ₁ ^Tζ₁(x₁,μ₁,σ₁) represent that first nerves network is in inputUnder output；θ₁、μ₁、σ₁It is respectively the weight of first nerves network, center and width；

Second controls subsystem controls restrains specific as follows:

$v_2=-k_2{z}_2-{{\mathrm{\θ}}_2}^T{\mathrm{\ζ}}_2(x_2,{\mathrm{\μ}}_2,{\mathrm{\σ}}_2)-z_1+{\stackrel{\·}{v}}_1$

In formula, v₂It is the output of the second control subsystem；z₂、z₁It is respectively the tracking of the second, first control subsystem by mistake Difference；k₂For predetermined coefficient；It is the output v of the first control subsystem₁First derivative；θ₂ ^Tζ₂(x₂,μ₂,σ₂) represent nervus opticus Network is in inputUnder output；θ₂、μ₂、σ₂It is respectively the weight of nervus opticus network, center and width；

3rd controls subsystem controls restrains specific as follows:

$u=G^{-1}(-k_3{z}_3-f-{{\mathrm{\θ}}_3}^T\mathrm{\ζ}(x_3,{\mathrm{\μ}}_3,{\mathrm{\σ}}_3)-z_2+{\stackrel{\·}{v}}_2)$

In formula, u is the 3rd control input quantity controlling the adjustable wing hypersonic aircraft that subsystem is exported；G is institute State the control gain matrix of adjustable wing hypersonic aircraft；z₃、z₂It is respectively the tracking error of the three, the second control subsystems； k₃For predetermined coefficient；It is the output v of the second control subsystem₂First derivative； For adjustable wing Hypersonic aircraft initial velocity V₀Three order derivatives,For adjustable wing hypersonic aircraft elemental height H₀Quadravalence Derivative；θ₃ ^Tζ(x₃,μ₃,σ₃) represent that third nerve network is in inputUnder output；θ₃、μ₃、σ₃It is respectively the The weight of three neutral nets, center and width.

Preferably, described first～third nerve network be RBF neural.

Further, control subsystem, the 3rd control subsystem are respectively arranged with a first-order dynamic filtering second Device: the first wave filter, the second wave filter；Second controls subsystem utilizes following formula approximate calculation

${\stackrel{\·}{v}}_1\≈{\stackrel{\·}{\stackrel{\‾}{v}}}_1=\frac{v_1-{\stackrel{\‾}{v}}_1}{{\mathrm{\τ}}_1},$

In formula,It is the output v of the first control subsystem₁Output after the first wave filter, τ₁It it is the first wave filter Time constant；

3rd controls subsystem utilizes following formula approximate calculation

${\stackrel{\·}{v}}_2\≈{\stackrel{\·}{\stackrel{\‾}{v}}}_2=\frac{v_2-{\stackrel{\‾}{v}}_2}{{\mathrm{\τ}}_2},$

In formula,It is the output v of the second control subsystem₂Output after the second wave filter, τ₂It it is the second wave filter Time constant.

The present invention is directed to the feature of adjustable wing hypersonic aircraft, utilize Backstepping method and combine nerve net Intelligent Nonlinear control system designed by network, and compared to existing technology, this control system has the advantages that

(1) Backstepping control design case method has good global stability, can effectively utilize nonlinear system Nonlinear characteristic inherently, the control system designed based on the method has when processing hypersonic aircraft problem Greater flexibility, can ensure its global stability effectively.

(2) in the case of control system exists internal interference and external disturbance simultaneously, neural network control method can have Effect offsets the interference impact on system so that adjustable wing hypersonic aircraft can response tracking signal fast and effectively, have Preferably tracking performance and robust performance.

(3) present invention utilize the most in the controls the thought of dynamic surface to solve differential expansion issues, significantly letter Change amount of calculation, improve real-time.

Accompanying drawing explanation

Fig. 1 is the structural principle schematic diagram of control system of the present invention；

Fig. 2 is the height tracing response curve of control system of the present invention；

Fig. 3 is the speed tracing response curve of control system of the present invention.

Detailed description of the invention

Below in conjunction with the accompanying drawings technical scheme is described in detail:

The thinking of the present invention is the feature for adjustable wing hypersonic aircraft, by Backstepping control system Method for designing also combines neutral net and designs a kind of adjustable wing hypersonic aircraft Intelligent Nonlinear control system, to solve Adjustable wing hypersonic aircraft flight control problem under strong nonlinearity characteristic and uncertain external disturbance.

Backstepping Control System Design method is as another using some state in hypersonic vehicle The virtual controlling input of a little states, compensates the effect of various uncertain factor while design controls input, uses and progressively pass The mode pushed away, and by Liapunov function design control law from front to back, to ensure the global stability of system.The method It is applicable to On-line Control, can reduce in the line computation time, there is unique advantage.But Backstepping control design case Method does not ensures that the robust performance of system.To this end, the present invention introduces neutral net in the controls, utilize neutral net Approach to uncertain disturbance term, to improve the robust performance of control system.

According to the design philosophy of Backstepping, in order to ensure adjustable wing hypersonic aircraft flying speed V and Flying height H can quickly trace into designated value V_rAnd H_r, adjustable wing hypersonic aircraft kinetic model is divided into external loop System, intermediate loop system and three subsystems of inner looping system.The design of controller is proceeded by, i.e. from external loop subsystem Design to inner looping subsystem direction from external loop subsystem.

As a example by certain type adjustable wing hypersonic aircraft, it uses blended wing-body layout, body profile to be triangle, greatly Swept back wing uses blended wing-body mode with fuselage, and elevator is arranged in trailing edge, and wing is deformable aerofoil, uses and is leading The stretch mode of winglet of wing both sides improves lift and flight efficiency.This aircraft longitudinal direction under hypersonic cruise flight condition Motion model is described as

$\stackrel{\·}{V}=\frac{Tcos\mathrm{\α}-D}{m}-g\×sin\mathrm{\γ}$

$\stackrel{\·}{\mathrm{\γ}}=\frac{L+Tsin\mathrm{\α}}{mV}-\frac{g\×cos\mathrm{\γ}}{V}$

$\stackrel{\·}{H}=Vsin\mathrm{\γ}$

$\stackrel{\·}{\mathrm{\α}}=q-\stackrel{\·}{\mathrm{\γ}}$

$\stackrel{\·}{q}=M_{yy}/I_{yy}$

In formula, quantity of state V, γ, α, q, H represent aircraft speed, flight track inclination angle, flying angle, pitching respectively Angular speed and flying height.L, D, T represent aircraft lift, resistance and motor power respectively, and M represents pitching moment, I_yyTable Show and longitudinally rotate inertia, ∫ (x2+z²) dm=I_yy；

Flight force and moment is expressed as:

L=0.5 ρ V²sC_L

D=0.5 ρ V²sC_D

M_yy=0.5 ρ V²sc[C_M(α)+C_M(δ_e)+C_M(q)]

In formula, ρ represents atmospheric density, and s is wing wetted area, C_L, C_D, C_MRepresent lift coefficient respectively, resistance coefficient and Pitching moment coefficient.

Under hypersonic average flight state, adjustable wing hypersonic aircraft intrinsic parameter quality m, pitching moment are used Property amasss I_yy, aircraft surfaces amasss s, chord-length c and there is perturbation, aerodynamic parameter C_L、C_DThere is also perturbation, use the rated value supposed attached Add a changes delta to represent the uncertainty of parameter, it may be assumed that

M=m₀(1+Δm)

I_yy=I₀(1+ΔI)

S=s₀(1+Δs)

C=c₀(1+Δc)

C_L=C_L0(1+ΔC_L)

C_D=C_D0(1+ΔC_D)

Output flying speed V and flying height H are used overall-finished housing linearization process, i.e. to flying speed V and flying Line height H respectively differential n and m time, until controlling to input η_cOr δ_eOccur in differential formula.Then have:

$\left\{\begin{array}{c}\stackrel{\·}{V}=f_V\left(x\right)\\ \stackrel{\·\·}{V}={\mathrm{\ω}}_1\stackrel{\·}{x}/m\\ \stackrel{\·\·\·}{V}=({\mathrm{\ω}}_1\stackrel{\·\·}{x}+{\stackrel{\·}{x}}^T{\mathrm{\Ω}}_2\stackrel{\·}{x})/m\end{array}\right.$

$\left\{\begin{array}{c}\stackrel{\·}{H}=f_h\left(x\right)\\ \stackrel{\·\·}{H}=\stackrel{\·}{V}sin\mathrm{\γ}+V\stackrel{\·}{\mathrm{\γ}}cos\mathrm{\γ}\\ \stackrel{\·\·\·}{H}=\stackrel{\·\·}{V}sin\mathrm{\γ}+2\stackrel{\·}{V}\stackrel{\·}{\mathrm{\γ}}\cos \mathrm{\γ}-V{\stackrel{\·}{\mathrm{\γ}}}^{2}sin\mathrm{\γ}+V\stackrel{\·\·}{\mathrm{\γ}}cos\mathrm{\γ}\\ H^{\left(4\right)}=\stackrel{\·\·\·}{V}\sin \mathrm{\γ}+3\stackrel{\·\·}{V}\stackrel{\·}{\mathrm{\γ}}cos\mathrm{\γ}-3\stackrel{\·}{V}{\stackrel{\·}{\mathrm{\γ}}}^{2}sin\mathrm{\γ}+\\ 3\stackrel{\·}{V}\stackrel{\·\·}{\mathrm{\γ}}\cos \mathrm{\γ}-3V\stackrel{\·}{\mathrm{\γ}}\stackrel{\·\·}{\mathrm{\γ}}sin\mathrm{\γ}-V{\stackrel{\·}{\mathrm{\γ}}}^3\cos \mathrm{\γ}+V\stackrel{\·\·}{\mathrm{\γ}}\cos \mathrm{\γ}\end{array}\right.$

In formula,ThereforeH⁽⁴⁾All containWithAgain because of ForC_m,δe=c_e(δ_e-α), SoH⁽⁴⁾Expression formula contain control input β_cAnd δ_e。

3 subdifferentials of output flying speed V and 4 subdifferentials of flying height h are expressed as:

${\left[\begin{array}{cc}\stackrel{\·\·\·}{V}& H^{\left(4\right)}\end{array}\right]}^T={\left[\begin{array}{cc}{V}_0^{\left(3\right)}& H_0^{\left(4\right)}\end{array}\right]}^T+\left[\begin{array}{cc}{b}_{11}& b_{12}\\ b_{21}& b_{22}\end{array}\right]{\left[\begin{array}{cc}{\mathrm{\β}}_c& {\mathrm{\δ}}_e\end{array}\right]}^T$

In formula:

$V_0^{\left(3\right)}=({\mathrm{\ω}}_1{\stackrel{\·\·}{x}}_0+{\stackrel{\·}{x}}^T{\mathrm{\Ω}}_2\stackrel{\·}{x})/m,$

$\begin{array}{c}{H}_0^{\left(4\right)}=3\stackrel{\·\·}{V}\stackrel{\·}{\mathrm{\γ}}\cos \mathrm{\γ}-3\stackrel{\·}{V}{\stackrel{\·}{\mathrm{\γ}}}^2\sin \mathrm{\γ}+3\stackrel{\·}{V}\stackrel{\·\·}{\mathrm{\γ}}\cos \mathrm{\γ}\\ -3V\stackrel{\·}{\mathrm{\γ}}\stackrel{\·\·}{\mathrm{\γ}}\sin \mathrm{\γ}-V{\stackrel{\·}{\mathrm{\γ}}}^{\left(3\right)}\cos \mathrm{\γ}+({\mathrm{\ω}}_1{\stackrel{\·\·}{x}}_0+{\stackrel{\·}{x}}^T{\mathrm{\Ω}}_2\stackrel{\·}{x})\sin \mathrm{\γ}/m\\ +V\cos \mathrm{\γ}({\mathrm{\π}}_1{\stackrel{\·\·}{x}}_0+{\stackrel{\·}{x}}^T{\mathrm{\Π}}_2\stackrel{\·}{x})\end{array},$

b₁₁=(ρ V²scω²/ 2m) cos α,

$b_{12}=-(c_e{\mathrm{\ρV}}^{2}s\stackrel{\‾}{c}/2{\mathrm{mI}}_{yy})(Tsin\mathrm{\α}+D_{\mathrm{\α}}-T_{\mathrm{\α}}cos\mathrm{\α}),$

b₂₁=(ρ V²scω²/ 2m) sin (α+γ),

$b_{22}=(c_e{\mathrm{\ρV}}^{2}s\stackrel{\‾}{c}/2{\mathrm{mI}}_{yy})\[Tcos(\mathrm{\α}+\mathrm{\γ})+L_{\mathrm{\α}}cos\mathrm{\γ}+T_{\mathrm{\α}}sin(\mathrm{\α}+\mathrm{\γ})-D_{\mathrm{\α}}sin\mathrm{\γ}\].$

According to above-mentioned speed, the differential equation of height, by disturbance adjustable wing hypersonic aircraft longitudinal dynamics mould Type can be to be expressed as form:

$\left\{\begin{array}{c}{\stackrel{\·}{x}}_1=x_2+{\mathrm{\φ}}_1\left(x_1\right)\\ {\stackrel{\·}{x}}_2=x_3+{\mathrm{\φ}}_2\left(x_2\right)\\ {\stackrel{\·}{x}}_3=f+Gu+{\mathrm{\φ}}_3\left(x_3\right)\end{array}\right.$

In formula, U=[η_c δ_e]^T, η_cFor engine throttle setting, δ_eElevator tilt value, φ_i(x_i) represent by pneumatic The composite interference that the uncertain and external interference that Parameter Perturbation causes is constituted.

The design object of controller is in the case of system exists unknown parameter perturbation and external interference so that adjustable wing Hypersonic aircraft energy tenacious tracking given speed reference signal and elevation references signalt≥0。

Before controller designs, make the following assumptions:

Assume 1: assume to control each element bounded of gain matrix G, i.e. there is constant g₁≥g₀＞ 0 so that g₀≤|| G||≤g₁。

Assume 2: assume speed and the elevation references signal V of expectation tracking_r(t)、H_r(t) continuous and derivable bounded, and there is height Order derivative all bounded.

According to the design philosophy of Backstepping, closed-loop system is divided into 3 subsystems, first subsystems Error is:

$\left\{\begin{array}{c}{z}_1=x_1-y_r\\ z_2=x_2-v_1\\ z_3=x_3-v_2\end{array}\right.$

In formula, v₁、v₂Virtual controlling for each subsystem inputs.

Carry out the specific design method in three loops separately below:

The first step (external loop): consider the external loop subsystem of closed-loop system

${\stackrel{\·}{x}}_1=x_2+{\mathrm{\φ}}_1\left(x_1\right)$

Tracking error vector z to first subsystem₁=x₁-y_rTemporally t derivation obtains

${\stackrel{\·}{z}}_1={\stackrel{\·}{x}}_1-{\stackrel{\·}{y}}_r=x_2+{\mathrm{\φ}}_1\left(x_1\right)-{\stackrel{\·}{y}}_r$

Wherein, φ₁(x₁) it is about state x₁Unknown nonlinear function, it is approached by present invention neutral net. The preferred RBF neural of described neutral net.RBF neural is the feed-forward type neutral net of a kind of function admirable, RBF network Arbitrary nonlinear function can be approached with arbitrary accuracy, and there is overall approximation capability, there is the strongest robustness, memory energy Power, non-linear mapping capability and powerful self-learning capability.

RBF neural is utilized to carry out approaching to obtain φ₁(x₁)=θ₁ ^Tζ₁(x₁,μ₁,σ₁)+ε₁, wherein, θ, μ, σ are respectively net The weight of network, center and width, ε₁=φ₁(x₁)-θ₁ ^Tζ₁(x₁,μ₁,σ₁) it is the neutral net error term of approaching generation, meet For normal number.Substituted into the equation after above-mentioned error derivation:

${\stackrel{\·}{z}}_1=x_2+{{\mathrm{\θ}}_1}^T{\mathrm{\ζ}}_1(x_1,{\mathrm{\μ}}_1,{\mathrm{\σ}}_1)+{\mathrm{\ϵ}}_1-{\stackrel{\·}{y}}_r$

Due to x₂=z₂+v₁, substitute into above formula and obtain:

${\stackrel{\·}{z}}_1=z_2+v_1+{{\mathrm{\θ}}_1}^T{\mathrm{\ζ}}_1(x_1,{\mathrm{\μ}}_1,{\mathrm{\σ}}_1)+{\mathrm{\ϵ}}_1-{\stackrel{\·}{y}}_r$

Design virtual controlling rule is as follows:

$v_1=-k_1{z}_1-{{\mathrm{\θ}}_1}^T{\mathrm{\ζ}}_1(x_1,{\mathrm{\μ}}_1,{\mathrm{\σ}}_1)+{\stackrel{\·}{y}}_r$

This virtual controlling rule is substituted into above-mentioned tracking error derivative obtain:

${\stackrel{\·}{z}}_1=-k_1{z}_1+z_2+{\mathrm{\ϵ}}_1$

Second step (intermediate loop): consider the intermediate loop subsystem of closed-loop system

${\stackrel{\·}{x}}_2=x_3+{\mathrm{\φ}}_2\left(x_2\right)$

Tracking error vector z to second subsystem₂=x₂-v₁Temporally t derivation obtains:

${\stackrel{\·}{z}}_2={\stackrel{\·}{x}}_2-{\stackrel{\·}{v}}_1=x_3+{\mathrm{\φ}}_2\left(x_1\right)-{\stackrel{\·}{v}}_1$

Similar to the first step, by a RBF neural to φ₂(x₂) approach, and by z₃=x₃-v₂Can follow the tracks of by mistake Difference derivative:

${\stackrel{\·}{z}}_2=z_3+v_2+{{\mathrm{\θ}}_2}^T\mathrm{\ζ}(x_2,{\mathrm{\μ}}_2,{\mathrm{\σ}}_2)+{\mathrm{\ϵ}}_2-{\stackrel{\·}{v}}_1$

In formula,

Design virtual controlling rule is as follows:

$v_2=-k_2{z}_2-{{\mathrm{\θ}}_2}^T\mathrm{\ζ}(x_2,{\mathrm{\μ}}_2,{\mathrm{\σ}}_2)-z_1+{\stackrel{\·}{v}}_1$

The derivative equation that virtual controlling rule substitutes into tracking error is obtained:

${\stackrel{\·}{z}}_2=-z_1-k_2{z}_2+z_3+{\mathrm{\ϵ}}_2$

3rd step (inner looping): consider the inner looping subsystem of closed-loop system

${\stackrel{\·}{x}}_3=f+Gu+{\mathrm{\φ}}_3\left(x_3\right)$

As can be seen from the above equation, state equation occurs in that the control input u of whole closed-loop system, therefore to controlling input U is designed, to error vector z₃=x₃-v₂Derivation, and utilize RBF neural to unknown nonlinear function phi₃(x₃) approach, Obtain tracking error derivative

${\stackrel{\·}{z}}_3={\stackrel{\·}{x}}_3-{\stackrel{\·}{v}}_2=f+Gu+{{\mathrm{\θ}}_3}^T\mathrm{\ζ}(x_3,{\mathrm{\μ}}_3,{\mathrm{\σ}}_3)+{\mathrm{\ϵ}}_3-{\stackrel{\·}{v}}_2$

In formula,For Making closed-loop system stable, design controls input as follows

$u=G^{-1}(-k_3{z}_3-f-{{\mathrm{\θ}}_3}^T\mathrm{\ζ}(x_3,{\mathrm{\μ}}_3,{\mathrm{\σ}}_3)-z_2+{\stackrel{\·}{v}}_2)$

Obtain after substituting into tracking error derivative

${\stackrel{\·}{z}}_3=-z_2-k_3{z}_3+{\mathrm{\ϵ}}_3$

The error equation of three above loop gained can be written as form:

$\left[\begin{array}{c}{\stackrel{\·}{z}}_1\\ {\stackrel{\·}{z}}_2\\ {\stackrel{\·}{z}}_3\end{array}\right]=\left[\begin{array}{ccc}-k_1& 1& 0\\ -1& -k_2& 1\\ 0& -1& -k_3\end{array}\right]\left[\begin{array}{c}{z}_1\\ z_2\\ z_3\end{array}\right]+\left[\begin{array}{c}{\mathrm{\ϵ}}_1\\ {\mathrm{\ϵ}}_2\\ {\mathrm{\ϵ}}_3\end{array}\right]$

Finally give control system of the present invention as shown in Figure 1, specifically include: the external loop that is sequentially connected in series, intermediate loop, Inner looping controls subsystem and controls subsystem three RBF neural one to one with three；Wherein,

It is specific as follows that external loop controls subsystem controls rule:

$v_1=-k_1{z}_1-{{\mathrm{\θ}}_1}^T{\mathrm{\ζ}}_1(x_1,{\mathrm{\μ}}_1,{\mathrm{\σ}}_1)+{\stackrel{\·}{y}}_r$

In formula, v₁The output of subsystem is controlled for external loop；z₁The tracking error of subsystem is controlled for external loop；k₁For in advance If coefficient；For reference instructionFirst derivative；Represent external loop neutral net In inputUnder output；θ₁、μ₁、σ₁It is respectively the weight of external loop neutral net, center and width；This is concrete In embodiment, k₁=3, θ₁It is the random number matrix of 6 row 2 row,σ₁=[0.2 0.3 0.2 0.3 0.1 0.1]^T, ζ₁=0.⁸。

It is specific as follows that intermediate loop controls subsystem controls rule:

$v_2=-k_2{z}_2-{{\mathrm{\θ}}_2}^T{\mathrm{\ζ}}_2(x_2,{\mathrm{\μ}}_2,{\mathrm{\σ}}_2)-z_1+{\stackrel{\·}{v}}_1$

In formula, v₂The output of subsystem is controlled for intermediate loop；z₂、z₁Be respectively intermediate loop, external loop controls subsystem Tracking error；k₂For predetermined coefficient；The output v of subsystem is controlled for external loop₁First derivative；Represent Intermediate loop neutral net is in inputUnder output；θ₂、μ₂、σ₂Be respectively intermediate loop neutral net weight, Center and width；In this specific embodiment, k₂=3, θ₂It is the random number matrix of 6 row 2 row, σ₂=[0.2 0.3 0.2 0.3 0.1 0.1]^T, ζ₂=0.8.

It is specific as follows that inner looping controls subsystem controls rule:

$u=G^{-1}(-k_3{z}_3-f-{{\mathrm{\θ}}_3}^T\mathrm{\ζ}(x_3,{\mathrm{\μ}}_3,{\mathrm{\σ}}_3)-z_2+{\stackrel{\·}{v}}_2)$

In formula, u controls the control input quantity of the adjustable wing hypersonic aircraft that subsystem is exported by inner looping；G is The control gain matrix of described adjustable wing hypersonic aircraft；z₃、z₂Be respectively inner looping, intermediate loop controls subsystem Tracking error；k₃For predetermined coefficient；The output v of subsystem is controlled for intermediate loop₂First derivative； For adjustable wing hypersonic aircraft initial velocity V₀Three order derivatives,The highest for adjustable wing hypersonic aircraft Degree H₀Fourth-Derivative；θ₃ ^Tζ(x₃,μ₃,σ₃) represent that inner looping neutral net is in inputUnder output；θ₃、 μ₃、σ₃It is respectively the weight of inner looping neutral net, center and width；In this specific embodiment, k₂=3, θ₃Be 6 row 2 row with Machine matrix number,σ₃=[0.2 0.3 0.2 0.3 0.1 0.1]^T, ζ₃=0.8..

In above-mentioned control system,

${\stackrel{\·}{v}}_1={\mathrm{\Σ}}_{j-1}^2\left(\frac{\∂v_1}{\∂x_j}{\stackrel{\·}{x}}_j\right)+\frac{\∂v_1}{\∂{\mathrm{\θ}}_1}{\stackrel{\·}{\mathrm{\θ}}}_1+\frac{\∂v_1}{\∂{\mathrm{\μ}}_1}{\stackrel{\·}{\mathrm{\μ}}}_1+\frac{\∂v_1}{\∂{\mathrm{\σ}}_1}{\stackrel{\·}{\mathrm{\σ}}}_1+{\mathrm{\Σ}}_{j-1}^2\left(\frac{\∂v_1}{\∂y_r^{(j-1)}}{y}_r^{\left(j\right)}\right)$

${\stackrel{\·}{v}}_2={\mathrm{\Σ}}_{i-1}^3\left(\frac{\∂v_2}{\∂x_i}{\stackrel{\·}{x}}_i\right)+{\mathrm{\Σ}}_{i-1}^2\frac{\∂v_2}{\∂{\mathrm{\θ}}_i}{\stackrel{\·}{\mathrm{\θ}}}_i+{\mathrm{\Σ}}_{i-1}^2\frac{\∂v_2}{\∂{\mathrm{\μ}}_i}{\stackrel{\·}{\mathrm{\μ}}}_i+{\mathrm{\Σ}}_{i-1}^2\frac{\∂v_2}{\∂{\mathrm{\σ}}_i}{\stackrel{\·}{\mathrm{\σ}}}_i+{\mathrm{\Σ}}_{i-1}^3\left(\frac{\∂v_1}{\∂y_r^{(i-1)}}\right)y_r^{\left(i\right)}$

It is therefore seen that to middle virtual signal v₁、v₂Direct differentiation is sufficiently complex, and therefore the present invention is further according to dynamic surface The thought controlled, by virtual signal v₁Obtained by first-order dynamic wave filterReplace v₁,For the output of wave filter, have:

${\mathrm{\τ}}_1{\stackrel{\·}{\stackrel{\‾}{v}}}_1+{\stackrel{\‾}{v}}_1=v_1,v_1\left(0\right)={\stackrel{\‾}{v}}_1\left(0\right)$

Just can be in the hope of by above formula:

${\stackrel{\·}{v}}_1\≈{\stackrel{\·}{\stackrel{\‾}{v}}}_1=\frac{v_1-{\stackrel{\‾}{v}}_1}{{\mathrm{\τ}}_1}$

In formula, τ₁For filter time constant, work as τ₁＞ 0 and taking fully enough hour,V can be infinitely close to₁。

Similarly, by virtual signal v₂Obtained by first-order dynamic wave filterReplace v₂,For the output of wave filter, Have

${\mathrm{\τ}}_2{\stackrel{\·}{\stackrel{\‾}{v}}}_2+{\stackrel{\‾}{v}}_2=v_2,v_2\left(0\right)={\stackrel{\‾}{v}}_2\left(0\right)$

Just can be in the hope of by above formula:

${\stackrel{\·}{v}}_2\≈{\stackrel{\·}{\stackrel{\‾}{v}}}_2=\frac{v_2-{\stackrel{\‾}{v}}_2}{{\mathrm{\τ}}_2}$

In formula, τ₂For filter time constant, τ₂＞ 0.

In order to verify the effect of the present invention, it is carried out numerical simulation checking.In order to describe the uncertainty of aircraft, imitative Middle lift coefficient reduces 20% than nominal value, and resistance coefficient increases by 20% than nominal value, meanwhile, introduces external disturbance, i.e. adds Enter the data noise of 10 sin (t).Utilize the hypersonic aircraft height obtained by control system of the present invention and speed Tracking response curve as shown in Figure 2,3.

Finding out from above-mentioned simulation result, the present invention has a characteristic that when there is larger interference, and the intelligence proposed is non- Linear control system can effectively offset the interference impact on system, adjustable wing hypersonic aircraft can fast and effeciently respond with Track signal, illustrates that this Intelligent Nonlinear control system has preferable tracing property and robustness.

Claims (4)

1. an adjustable wing hypersonic aircraft Intelligent Nonlinear control system, is used for generating adjustable wing hypersonic aircraft Control input quantity u, to ensure that adjustable wing hypersonic aircraft flying speed V, flying height H can quickly trace into input Flying speed reference value V_r, flying height reference value H_r；It is characterized in that, described control system utilizes Backstepping method Design, specifically include first be sequentially connected in series～the 3rd and control subsystem and control subsystem one by one with first～the 3rd First corresponding～third nerve network；Wherein,

First controls subsystem controls restrains specific as follows:

$v_1=-k_1{z}_1-{{\mathrm{\θ}}_1}^T{\mathrm{\ζ}}_1(x_1,{\mathrm{\μ}}_1,{\mathrm{\σ}}_1)+{\stackrel{\·}{y}}_r$

In formula, v₁It is the output of the first control subsystem；z₁It it is the tracking error of the first control subsystem；k₁For predetermined coefficient； For reference instructionFirst derivative；θ₁ ^Tζ₁(x₁,μ₁,σ₁) represent that first nerves network is in inputUnder output；θ₁、μ₁、σ₁It is respectively the weight of first nerves network, center and width；

Second controls subsystem controls restrains specific as follows:

$v_2=-k_2{z}_2-{{\mathrm{\θ}}_2}^T{\mathrm{\ζ}}_2(x_2,{\mathrm{\μ}}_2,{\mathrm{\σ}}_2)-z_1+{\stackrel{\·}{v}}_1$

In formula, v₂It is the output of the second control subsystem；z₂、z₁It is respectively the tracking error of the second, first control subsystem；k₂For Predetermined coefficient；It is the output v of the first control subsystem₁First derivative；θ₂ ^Tζ₂(x₂,μ₂,σ₂) represent that nervus opticus network exists InputUnder output；θ₂、μ₂、σ₂It is respectively the weight of nervus opticus network, center and width；

3rd controls subsystem controls restrains specific as follows:

$u=G^{-1}(-k_3{z}_3-f-{{\mathrm{\θ}}_3}^T\mathrm{\ζ}(x_3,{\mathrm{\μ}}_3,{\mathrm{\σ}}_3)-z_2+{\stackrel{\·}{v}}_2)$

In formula, u is the 3rd control input quantity controlling the adjustable wing hypersonic aircraft that subsystem is exported；G be described can Become the control gain matrix of wing hypersonic aircraft；z₃、z₂It is respectively the tracking error of the three, the second control subsystems；k₃For Predetermined coefficient；It is the output v of the second control subsystem₂First derivative；

For adjustable wing hypersonic aircraft initial velocity V₀Three order derivatives,For adjustable wing Hypersonic aircraft elemental height H₀Fourth-Derivative；θ₃ ^Tζ(x₃,μ₃,σ₃) represent that third nerve network is in inputUnder output；θ₃、μ₃、σ₃It is respectively the weight of third nerve network, center and width.

2. as claimed in claim 1 control system, it is characterised in that described first～third nerve network be RBF nerve net Network.

3. control system as claimed in claim 1, it is characterised in that control to divide in subsystem, the 3rd control subsystem second It is not provided with a first-order dynamic wave filter: the first wave filter, the second wave filter；Second controls subsystem utilizes following formula approximation meter Calculate

${\stackrel{\·}{v}}_1\≈{\stackrel{\·}{\stackrel{\‾}{v}}}_1=\frac{v_1-{\stackrel{\‾}{v}}_1}{{\mathrm{\τ}}_1},$

In formula,It is the output v of the first control subsystem₁Output after the first wave filter, τ₁It it is the time of the first wave filter Constant；

3rd controls subsystem utilizes following formula approximate calculation

${\stackrel{\·}{v}}_2\≈{\stackrel{\·}{\stackrel{\‾}{v}}}_2=\frac{v_2-{\stackrel{\‾}{v}}_2}{{\mathrm{\τ}}_2},$

In formula,It is the output v of the second control subsystem₂Output after the second wave filter, τ₂It it is the time of the second wave filter Constant.

4. control system as claimed in claim 1, it is characterised in that the control input of described adjustable wing hypersonic aircraft

Amount u=[η_c δ_e]^T, η_cFor the engine throttle setting of adjustable wing hypersonic aircraft, δ_eUltrasonic for adjustable wing height The elevator tilt value of speed aircraft.