CN108375907B - Adaptive compensation control method of hypersonic aircraft based on neural network - Google Patents

Adaptive compensation control method of hypersonic aircraft based on neural network Download PDF

Info

Publication number
CN108375907B
CN108375907B CN201810262739.8A CN201810262739A CN108375907B CN 108375907 B CN108375907 B CN 108375907B CN 201810262739 A CN201810262739 A CN 201810262739A CN 108375907 B CN108375907 B CN 108375907B
Authority
CN
China
Prior art keywords
function
elevator
hypersonic aircraft
radial basis
follows
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810262739.8A
Other languages
Chinese (zh)
Other versions
CN108375907A (en
Inventor
胡庆雷
李梓明
郭雷
王陈亮
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beihang University
Original Assignee
Beihang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beihang University filed Critical Beihang University
Priority to CN201810262739.8A priority Critical patent/CN108375907B/en
Publication of CN108375907A publication Critical patent/CN108375907A/en
Application granted granted Critical
Publication of CN108375907B publication Critical patent/CN108375907B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Feedback Control In General (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)

Abstract

The invention discloses a hypersonic aircraft self-adaptive compensation control method based on a neural network, which comprises the following steps: establishing a longitudinal dynamics model of the hypersonic aircraft, and decomposing the longitudinal dynamics model into an attitude subsystem and a speed subsystem; establishing an elevator fault model of the hypersonic aircraft; constructing a smooth function to estimate nonlinear input saturation, and introducing a radial basis function neural network to estimate a nonlinear function in a longitudinal dynamics model of the hypersonic aircraft; and designing an adaptive compensation controller and a corresponding adaptive parameter updating law of the hypersonic aircraft by a backstepping method. The invention provides a radial basis function neural network adaptive compensation control method considering elevator faults and input saturation, solves the influence of various elevator faults and actuator saturation on an aircraft in the flight process of a hypersonic aircraft, and ensures the fault-tolerant capability and robustness of a system.

Description

Adaptive compensation control method of hypersonic aircraft based on neural network
Technical Field
The invention belongs to the technical field of aircraft control, and particularly relates to a hypersonic aircraft self-adaptive compensation control method based on a neural network.
Background
Hypersonic aircraft have attracted considerable commercial and military attention in recent years as a reliable and economical means of transport to adjacent spaces. However, due to its special configuration, unique flight conditions, extremely sensitive aerodynamic parameters of the hypersonic aircraft, and the high nonlinearity of its dynamics, all of which make the control design of the hypersonic aircraft very difficult.
The problem that an unknown nonlinear link exists in the system can be well solved by applying the adaptive compensation control of the radial basis function neural network, and the system can meet the stability requirement and simultaneously meet the corresponding control requirement. So far, control methods including robust control, sliding mode control, linear quadratic control and the like are applied to control design of a longitudinal model of the hypersonic aircraft, and compared with the mentioned control methods, the self-adaptive backstepping control provides an effective method for solving an unknown nonlinear model. On the one hand, in aircraft control, actuator saturation may cause control effect to deteriorate or even completely lose control, and in recent years, a control problem of a system with an input saturation characteristic has received great attention, and by constructing an auxiliary system, the system input saturation problem can be solved. However, when the system has an unknown delay link, the auxiliary system model is difficult to establish, great difficulty is caused to the stability analysis of the closed-loop system, and the problem that the unknown gain link exists in the system can be well solved by applying the adaptive compensation control. On the other hand, due to frequent operations and harsh working environments, the aircraft elevators may be affected by faults which are destructive for the aircraft, and in the control research of the present day, the establishment of a fault model is often assumed that each elevator has a fault only once, and the mode (control effect completely fails) and parameters of the fault are not changed. It is evident that this is an extreme case and that the type involved in a real aircraft elevator fault is complex. The elevator fault model provided by the patent can well cover various types of faults, has no limit requirement on the number of the faults, and is more practical.
Disclosure of Invention
The technical problem solved by the invention is as follows: the invention provides a radial basis function neural network adaptive compensation control method considering elevator faults and input saturation, which solves the problem that the aircraft is affected by various elevator faults and actuator saturation in the flight process of the hypersonic aircraft and ensures the fault tolerance capability and robustness of the system.
In order to achieve the technical purpose, the technical scheme of the invention is as follows:
a hypersonic aircraft self-adaptive compensation control method based on a neural network comprises the following steps:
s1: establishing a longitudinal dynamics model of the hypersonic aircraft, and decomposing the longitudinal dynamics model into an attitude subsystem and a speed subsystem;
s2: establishing an elevator fault model of the hypersonic aircraft;
s3: constructing a smooth function to estimate nonlinear input saturation, and introducing a radial basis function neural network to estimate a nonlinear function F in a longitudinal dynamics model of a hypersonic aircrafti(i＝1,2,3)；
S4: and designing an adaptive compensation controller and a corresponding adaptive parameter updating law of the hypersonic aircraft by a backstepping method.
Further, in S1, the longitudinal dynamics model is:
Figure GDA0001674925600000021
Figure GDA0001674925600000022
Figure GDA0001674925600000023
Figure GDA0001674925600000024
Figure GDA0001674925600000025
wherein V, h, gamma, alpha and q are respectively speed, height, track inclination angle, attack angle and pitching rate; m, ReMu and IyyRespectively the mass of the aircraft, the radius of the earth, the universal gravitation constant and the inertia moment; t, D, L and MyyRespectively representing thrust, drag, lift and pitching moment.
Further, in S1, the model of the pose subsystem is:
Figure GDA0001674925600000026
Figure GDA0001674925600000027
Figure GDA0001674925600000031
y＝x1
wherein the state variable x1＝γ，x2＝θp，x3＝q，θpThe pitch angle of the hypersonic aircraft; f. of1(x1,V)，f3(x1,x2,x3V) and g3(V) is a non-linear function processed by a radial basis function, f2And g2Is a known constant; u. ofj＝ejJ e N represents the jth elevator, N is a set of nonnegative integers,ejthe deflection angle of the jth elevator; djGain, sat (u) representing jth deflection anglej) Is a saturated non-linear function representing the yaw angle of the jth elevator.
The model of the speed subsystem is:
Figure GDA0001674925600000032
wherein f isV(x1,x2,x3h, V) and gV(x1,x2,x3H, V) is a nonlinear function processed by a radial basis function; u. ofVβ is the fuel equivalence ratio, sat (u)V) Is a saturated non-linear function representing the fuel equivalence ratio.
Further, in S2, the elevator fault model is:
Figure GDA0001674925600000033
where h ∈ N denotes the h-th failure, kj,h，
Figure GDA0001674925600000034
And
Figure GDA0001674925600000035
are constants determined according to the specific fault and occurrence time of the elevator, wherein0≤kj,hLess than or equal to 1, which represents the health index of the jth elevator when the jth elevator has the ith fault,
Figure GDA0001674925600000036
and
Figure GDA0001674925600000037
respectively represents the starting time and the ending time of the h fault of the j elevator, and
Figure GDA0001674925600000038
Figure GDA0001674925600000039
is a piecewise continuous bounded function for representing the additive fault part of the jth elevator in the h fault, vj(t) represents a control signal of an elevator deflection angle.
Preferably, in S3, the smoothing function is constructed based on the saturation of the elevator yaw angle input;
the smoothing function is of the form:
Figure GDA00016749256000000310
sat(uj)＝ψ(uj)+ψd(uj)
wherein,
Figure GDA00016749256000000311
ψd(uj) Is a bounded function;
Figure GDA00016749256000000312
and
Figure GDA00016749256000000313
each represents ujUpper and lower bounds of (a); psi (u)j)＝ψaujNamely sat (u)j)＝ψauj+ψd(uj),ψaIs a continuously bounded non-linear function.
Preferably, in S3, the smoothing function is constructed based on the fuel equivalence ratio input being saturated;
the smoothing function is of the form:
Figure GDA0001674925600000041
sat(uV)＝ψaVuV+ψdV(uV)
wherein,
Figure GDA0001674925600000042
ψdV(uV) Is a bounded function;
Figure GDA0001674925600000043
and
Figure GDA0001674925600000044
each represents uVUpper and lower bounds of (a); psiaV(uV)＝ψauVNamely sat (u)V)＝ψaVuV+ψdV(uV),ψaVIs a continuously bounded non-linear function.
Further, in S3, the method includes introducing a radial basis function neural network to estimate a nonlinear function F in the systemiSpecific forms of (i ═ 1,2,3) are as follows:
Figure GDA0001674925600000045
wherein, thetai∈RNIs the optimal weight vector, R, of N nodes in the radial basis functionNIs an N-dimensional real space, phii(ξi)＝[φi1(ξi),…,φiN(ξi)]T∈RNIs a vector of basis functions in the radial basis functions; delta ([ xi ])i) The error of the approximation is represented by,iis a constant number, wherein
Figure GDA0001674925600000046
οijAnd biThe center and width of the radial basis function, respectively; defining a constant
Figure GDA0001674925600000047
Figure GDA0001674925600000048
Is that
Figure GDA0001674925600000049
Is determined by the estimated value of (c),
Figure GDA00016749256000000410
to estimate the error, gmIs constant and 0 < gm≤min[inf{g1(V)},g2,inf{g3(V)},inf{gV(x1,x2,x3,h,V)}]。
Further, in S4, the specific form of the adaptive compensation controller and the corresponding adaptive parameter updating law for designing the hypersonic aircraft by the back-stepping method is as follows:
defining an error variable s1、s2、s3：
s1＝x1-γr
s2＝x2-x2d
s3＝x3-x3d
Wherein, γrDefining h as a control signal for the track pitch angle gammarFor the reference signal of height h, choose
Figure GDA00016749256000000411
To ensure that when gamma tracks its control signal gammarWhile h tracks its reference signal hr；x2dIs the first state equation of the attitude sub-system
Figure GDA00016749256000000412
Virtual control signal of x3dFor the attitude sub-system
Figure GDA00016749256000000413
The virtual control signal of (2);
the virtual controller of the attitude subsystem is designed as follows:
Figure GDA0001674925600000051
Figure GDA0001674925600000052
x2dthe corresponding adaptive parameter updating law is as follows:
Figure GDA0001674925600000053
x3dthe corresponding adaptive parameter updating law is as follows:
Figure GDA0001674925600000054
the actual controller design of the attitude subsystem is as follows:
Figure GDA0001674925600000055
vjthe corresponding adaptive parameter updating law is as follows:
Figure GDA0001674925600000056
Figure GDA0001674925600000057
Figure GDA0001674925600000058
definition VrIs a reference signal for the velocity V and,
Figure GDA0001674925600000059
for a tracking error of V, the controller of the velocity subsystem is:
Figure GDA00016749256000000510
uVthe corresponding adaptive parameter updating law is as follows:
Figure GDA00016749256000000511
Figure GDA00016749256000000512
Figure GDA00016749256000000513
wherein,
Figure GDA00016749256000000514
(t)＝[d1(ψa1u1+ψd1),d2(ψa2u2+ψd2),…,dn(ψanun+ψdn)]T，
Figure GDA00016749256000000515
Figure GDA00016749256000000516
Figure GDA00016749256000000517
and
Figure GDA00016749256000000518
are estimates of zeta and p respectively,
Figure GDA00016749256000000519
and
Figure GDA00016749256000000520
are each ζVAnd pVEstimate of (c), xi1＝(x1,γr,h,V)T，
Figure GDA0001674925600000061
ξV＝(x1,x2,x3,h,V)T，∈，ci，λiAnd muiThey are all normal numbers.
The invention has the beneficial effects that:
(1) compared with the fault model established in the existing aircraft design process, the fault model established in the invention is more suitable for the general situation of hypersonic aircraft elevator faults, can well cover various types of faults, and is more practical.
(2) Compared with the traditional adaptive control method of the hypersonic aircraft, the method solves the problem of input saturation of the actuator by establishing the smooth function, and is favorable for the direct use of the backstepping method in the adaptive design process.
(3) The invention solves the inconvenience caused by unknown nonlinear function in the design of the self-adaptive compensation control system by introducing the radial basis function neural network, and ensures that only one parameter needs to be updated in each step in the self-adaptive parameter updating law no matter how large the dimension of the optimal weight vector of the neural network by defining the function related to the optimal weight vector of the neural network, thereby greatly reducing the calculated amount.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a flow chart of the present method;
FIG. 2 is a system block diagram of the present method;
FIG. 3 shows tracking height reference signal h by height h in the methodrA graph of the Matlab simulation results of (a), wherein the abscissa represents time (in s) and the ordinate represents height (in ft);
FIG. 4 shows that the velocity V tracks the velocity reference signal V in the present methodrIn which the abscissa represents time (in s) and the ordinate represents speed (in ft/s).
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments that can be derived by one of ordinary skill in the art from the embodiments given herein are intended to be within the scope of the present invention.
As shown in fig. 1, the adaptive compensation control method for the hypersonic aircraft based on the neural network comprises the following steps:
s1: establishing a standard longitudinal dynamics model of the hypersonic aerocraft, and decomposing the model into an attitude subsystem and a speed subsystem;
a longitudinal dynamic model of a standard hypersonic aircraft is established as follows:
Figure GDA0001674925600000071
Figure GDA0001674925600000072
Figure GDA0001674925600000073
Figure GDA0001674925600000074
Figure GDA0001674925600000075
wherein V, h, gamma, alpha and q are respectively speed, height, track inclination angle, attack angle and pitching rate; m, ReMu and IyyThe mass, the earth radius, the universal gravitation constant and the inertia moment of the aircraft are respectively taken as m is 9375kg, Re20903500ft and Iyy＝7×106bf ft; t, D, L and MyyRespectively representing thrust, drag, lift and pitching moment.
The model of the attitude subsystem of the hypersonic aircraft is established as follows:
Figure GDA0001674925600000076
Figure GDA0001674925600000077
Figure GDA0001674925600000078
y＝x1
wherein the state variable x1＝γ，x2＝θp，x3＝q，θpThe pitch angle of the hypersonic aircraft; f. of1(x1,V)，f3(x1,x2,x3V) and g3(V) is a non-linear function processed by a radial basis function, f2And g2Is a known constant; u. ofj＝ejJ e N represents the jth elevator, N is a set of nonnegative integers,ejthe deflection angle of the jth elevator; djGain, sat (u) representing jth deflection anglej) Is a saturated non-linear function representing the yaw angle of the jth elevator.
The model of the velocity subsystem of the hypersonic aircraft is established as follows:
Figure GDA0001674925600000081
wherein f isV(x1,x2,x3h, V) and gV(x1,x2,x3H, V) is a nonlinear function processed by a radial basis function; u. ofVBeta is a fuel equivalence ratio, and the flight speed of the hypersonic aerocraft is mainly determined by the fuel equivalence ratio beta, so that the hypersonic aerocraft is selected as an input; sat (u)V) Is a saturated non-linear function representing the fuel equivalence ratio.
S2: establishing an elevator fault model of the hypersonic aircraft;
a general elevator fault model is established as follows:
Figure GDA0001674925600000082
where h ∈ N denotes the h-th failure, kj,h，
Figure GDA0001674925600000083
And
Figure GDA0001674925600000084
are constants determined according to specific faults and occurrence time of the elevator, wherein k is more than or equal to 0j,hLess than or equal to 1, which represents the health index of the jth elevator when the jth elevator has the ith fault,
Figure GDA0001674925600000085
and
Figure GDA0001674925600000086
respectively represents the starting time and the ending time of the h fault of the j elevator, and
Figure GDA0001674925600000087
Figure GDA0001674925600000088
is a piecewise continuous bounded function for representing the additive fault part of the jth elevator in the h fault, vj(t) represents a control signal of an elevator deflection angle. Compared with a common fault model, the fault model established by the invention has universality and can represent two different faults:
the first failure: when k is more than or equal to 0j,hAt ≦ 1, the jth elevator loses its partial effectiveness and suffers additional failure
Figure GDA0001674925600000089
The influence of (a) on the performance of the device,
Figure GDA00016749256000000810
the second failure: when k isj,hWhen the elevator is equal to 0, the elevator is completely out of control and is not controlled by the control signal any more,
Figure GDA00016749256000000811
s3: constructing a smooth function to estimate nonlinear input saturation, and introducing a radial basis function neural network to estimate a nonlinear function F in a longitudinal dynamics model of a hypersonic aircrafti(i＝1,2,3)；
For the case of elevator yaw angle input saturation, the form of the smoothing function is constructed as follows:
Figure GDA00016749256000000812
sat(uj)＝ψ(uj)+ψd(uj)
wherein,
Figure GDA00016749256000000813
ψd(uj) Is a bounded function;
Figure GDA00016749256000000814
and
Figure GDA00016749256000000815
each represents ujUpper and lower bounds of (a); psi (u)j)＝ψaujNamely sat (u)j)＝ψauj+ψd(uj),ψaIs a continuously bounded non-linear function.
For the case where the fuel equivalence ratio input is saturated, the form of the smoothing function is constructed as follows:
Figure GDA00016749256000000816
sat(uV)＝ψaVuV+ψdV(uV)
wherein,
Figure GDA0001674925600000091
ψdV(uV) Is a bounded function;
Figure GDA0001674925600000092
and
Figure GDA0001674925600000093
each represents uVUpper and lower bounds of (a); psiaV(uV)＝ψauVNamely sat (u)V)＝ψaVuV+ψdV(uV),ψaVIs a continuously bounded non-linear function.
Estimating hypersonic flight by introducing radial basis function neural networksNon-linear function F in longitudinal dynamics model of machinei(i ═ 1,2,3), which is expressed as follows:
Fi(ξi)＝θi Tφi(ξi)+Δ(ξi),|Δ(ξi)|≤i
wherein, thetai∈RNIs the optimal weight vector, R, of N nodes in the radial basis functionNIs an N-dimensional real space, phii(ξi)＝[φi1(ξi),…,φiN(ξi)]T∈RNIs a vector of basis functions in the radial basis functions; delta ([ xi ])i) The error of the approximation is represented by,iis a constant number, wherein
Figure GDA0001674925600000094
οijAnd biThe center and width of the radial basis function, respectively; defining a constant
Figure GDA0001674925600000095
Figure GDA0001674925600000096
Is that
Figure GDA0001674925600000097
Is determined by the estimated value of (c),
Figure GDA0001674925600000098
for estimation error, gm is constant, and 0 < gm≤min[inf{g1(V)},g2,inf{g3(V)},inf{gV(x1,x2,x3,h,V)}]. Several non-linear segments approximated by radial basis function neural networks are shown below:
Figure GDA0001674925600000099
Figure GDA00016749256000000910
in which ξ1＝(x1,γr,h,V)T，
Figure GDA00016749256000000911
ξV＝(x1,x2,x3,h,V)T。
S4: designing a self-adaptive compensation controller and a corresponding self-adaptive parameter updating law of the hypersonic aircraft by a backstepping method;
defining an error variable s1、s2、s3：
s1＝x1-γr
s2＝x2-x2d
s3＝x3-x3d
Wherein, γrDefining h as a control signal for the track pitch angle gammarFor the control signal of the height h, select
Figure GDA00016749256000000912
To ensure that when gamma tracks its control signal gammarWhile h tracks its control signal hr；x2dIs the first state equation of the attitude sub-system
Figure GDA00016749256000000913
Virtual control signal of x3dFor the attitude sub-system
Figure GDA00016749256000000914
The virtual control signal of (2);
the virtual controller of the attitude subsystem is designed as follows:
Figure GDA0001674925600000101
Figure GDA0001674925600000102
x2dcorresponding adaptationThe parameter updating law is as follows:
Figure GDA0001674925600000103
x3dthe corresponding adaptive parameter updating law is as follows:
Figure GDA0001674925600000104
the actual controller design of the attitude subsystem is as follows:
Figure GDA0001674925600000105
vjthe corresponding adaptive parameter updating law is as follows:
Figure GDA0001674925600000106
Figure GDA0001674925600000107
Figure GDA0001674925600000108
definition VrIs a reference signal for the velocity V and,
Figure GDA0001674925600000109
for a tracking error of V, the controller of the velocity subsystem is:
Figure GDA00016749256000001010
uVthe corresponding adaptive parameter updating law is as follows:
Figure GDA00016749256000001011
Figure GDA00016749256000001012
Figure GDA00016749256000001013
wherein,
Figure GDA00016749256000001014
(t)＝[d1(ψa1u1+ψd1),d2(ψa2u2+ψd2),…,dn(ψanun+ψdn)]T，
Figure GDA00016749256000001015
Figure GDA00016749256000001016
Figure GDA00016749256000001017
and
Figure GDA00016749256000001018
are estimates of zeta and p respectively,
Figure GDA00016749256000001019
and
Figure GDA00016749256000001020
are each ζVAnd pVEstimate of (c), xi1＝(x1,γr,h,V)T，
Figure GDA00016749256000001021
ξV＝(x1,x2,x3,h,V)T，∈，ci，λiAnd muiThey are all normal numbers.
The system block diagram of the control method based on the above steps is shown in FIG. 2, and the adaptive controller is controlled by the adaptive controller according to the given signal (height reference signal h)rAnd a velocity reference signal Vr) And comprehensively calculating the state information of the system to obtain control signals of the elevator and the fuel throttle valve so as to control the system, wherein the arrow direction represents the signal transmission direction.
In the embodiment, matlab simulation is performed on the method by selecting the following parameters: e is 0.1, c1＝40，c2＝7，c3＝5，cV＝5，λ1＝0.1，λ2＝0.1，λ3＝0.1，λ4＝0.15，λ5＝0.25，μ1＝0.3，μ2＝0.3，μ3＝0.3，μ4＝0.2，μ5＝0.2，μV1＝μV2＝μV3＝0.1，λV1＝λV2＝λV30.1. The width and center point of the radial basis function neural network are selected as follows: b1＝b2＝b3＝10，bV15. To o1＝(ο11,ο12,ο13,ο14)，ο1jAre respectively selected from the matrix
Figure GDA0001674925600000111
And arranged and combined to give a total of 3481 different omicron1Value, i.e. for phi1(ξ1)＝[φ11(ξ1),…,φ1N(ξ1)]T∈RNA total of 81 radial basis functions were used, with N being 81. To o2＝(ο21,ο22,ο23,ο24,ο25)，ο2jAre respectively selected from the matrix
Figure GDA0001674925600000112
And arranged and combined to give a total of 35243 differentO ° o2Value, i.e. for phi2(ξ2)＝[φ21(ξ2),…,φ2N(ξ2)]T∈RNA total of 243 radial basis functions are used, N243. To o3＝(ο31,ο32,ο33,ο34,ο35,ο36,ο37)，ο3jAre respectively selected from the matrix
Figure GDA0001674925600000113
And arranged and combined to give a total of 372187 different omicron-3Value, i.e. for phi3(ξ3)＝[φ31(ξ3),…,φ3N(ξ3)]T∈RNA total of 2187 radial basis functions were used, N2187. To oV＝(οV1,οV2,οV3,οV4,οV5)，οVjAre respectively selected from the matrix
Figure GDA0001674925600000121
And arranged and combined to give a total of 35243 different omicronVValue, i.e. for phiV(ξV)＝[φV1(ξV),…,φVN(ξV)]T∈RNA total of 243 radial basis functions are used, N243.
As shown in fig. 3 and 4, by Matlab simulation, a self-adaptive compensation control method based on a radial basis function neural network can be obtained, so that the control signals of the hypersonic aircraft can be tracked respectively by the altitude and the speed under the conditions of elevator faults and actuator saturation of the hypersonic aircraft, and the performance requirement of small enough tracking error can be met.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (9)

1. A hypersonic aircraft self-adaptive compensation control method based on a neural network is characterized by comprising the following steps:
s1: establishing a longitudinal dynamics model of the hypersonic aircraft, and decomposing the longitudinal dynamics model into an attitude subsystem and a speed subsystem;
s2: establishing an elevator fault model of the hypersonic aircraft;
s3: constructing a smooth function to estimate nonlinear input saturation, and introducing a radial basis function neural network to estimate a nonlinear function F in a longitudinal dynamics model of a hypersonic aircrafti，i＝1,2,3；
S4: designing a self-adaptive compensation controller of the hypersonic aircraft and corresponding self-adaptive parameter updating through a backstepping method;
in S4, the specific form of the adaptive compensation controller and the corresponding adaptive parameter updating law for designing the hypersonic aircraft by the backstepping method is as follows:
defining an error variable s1、s2、s3：
s1＝x1-γr
s2＝x2-x2d
s3＝x3-x3d
Wherein the state variable x1＝γ，x2＝θp，x3＝q，θpIs the pitch angle of the hypersonic aerocraft, q is the pitch rate, gammarDefining h as a control signal for the track pitch angle gammarFor the reference signal of height h, choose
Figure FDA0002630910150000011
To ensure that when gamma tracks its control signal gammarWhile h tracks its reference signal hr；x2dIs the first state equation of the attitude sub-system
Figure FDA0002630910150000012
Virtual control signal of x3dFor the attitude sub-system
Figure FDA0002630910150000013
The virtual control signal of (2);
the virtual controller of the attitude subsystem is designed as follows:
Figure FDA0002630910150000014
Figure FDA0002630910150000015
x2dthe corresponding adaptive parameter updating law is as follows:
Figure FDA0002630910150000016
x3dthe corresponding adaptive parameter updating law is as follows:
Figure FDA0002630910150000017
the actual controller design of the attitude subsystem is as follows:
Figure FDA0002630910150000021
vjthe corresponding adaptive parameter updating law is as follows:
Figure FDA0002630910150000022
Figure FDA0002630910150000023
Figure FDA0002630910150000024
defining a constant
Figure FDA0002630910150000025
Figure FDA0002630910150000026
Is that
Figure FDA0002630910150000027
Is determined by the estimated value of (c),
Figure FDA0002630910150000028
to estimate the error, gmIs constant and 0 < gm≤min[inf{g1(V)},g2,inf{g3(V)},inf{gV(x1,x2,x3,h,V)}]；
Definition VrIs a reference signal for the velocity V and,
Figure FDA0002630910150000029
for a tracking error of V, the controller of the velocity subsystem is:
Figure FDA00026309101500000210
uVthe corresponding adaptive parameter updating law is as follows:
Figure FDA00026309101500000211
Figure FDA00026309101500000212
Figure FDA00026309101500000213
wherein,
Figure FDA00026309101500000214
(t)＝[d1(ψa1u1+ψd1),d2(ψa2u2+ψd2),…,dn(ψanun+ψdn)]T，
Figure FDA00026309101500000215
Figure FDA00026309101500000216
and
Figure FDA00026309101500000217
are estimates of zeta and p respectively,
Figure FDA00026309101500000218
and
Figure FDA00026309101500000219
are each ζVAnd pVEstimate of (c), xi1＝(x1,γr,h,V)T，
Figure FDA00026309101500000220
ξV＝(x1,x2,x3,h,V)T，∈1，∈2，ci，λiAnd muiThey are all normal numbers.
2. The method of claim 1, wherein in S1, the longitudinal dynamics model is:
Figure FDA00026309101500000221
Figure FDA0002630910150000031
Figure FDA0002630910150000032
Figure FDA0002630910150000033
Figure FDA0002630910150000034
wherein V, h, gamma and alpha are respectively speed, height, track inclination angle and attack angle; m, Re, μ and IyyRespectively the mass of the aircraft, the radius of the earth, the universal gravitation constant and the inertia moment; t, D, L and MyyRespectively representing thrust, drag, lift and pitching moment.
3. The method of claim 2, wherein in S1, the model of the pose subsystem is:
Figure FDA0002630910150000035
Figure FDA0002630910150000036
Figure FDA0002630910150000037
y＝x1
wherein f is1(x1,V)，f3(x1,x2,x3V) and g3(V) is a non-linear function processed by a radial basis function, f2And g2Is a known constant; u. ofj＝ejJ e N represents the jth elevator, N is a set of nonnegative integers,ejthe deflection angle of the jth elevator; djGain, sat (u) representing jth deflection anglej) Is a saturated non-linear function representing the yaw angle of the jth elevator;
the model of the speed subsystem is:
Figure FDA0002630910150000038
wherein f isV(x1,x2,x3h, V) and gV(x1,x2,x3H, V) is a nonlinear function processed by a radial basis function; u. ofVβ is the fuel equivalence ratio, sat (u)V) Is a saturated non-linear function representing the fuel equivalence ratio.
4. The method of claim 3, wherein in S2, the elevator fault model is:
Figure FDA0002630910150000039
where h ∈ N denotes the h-th failure, kj,h，
Figure FDA00026309101500000310
And
Figure FDA00026309101500000311
are constants determined according to specific faults and occurrence time of the elevator, wherein k is more than or equal to 0j,hLess than or equal to 1, which represents the health index of the jth elevator when the jth elevator has the ith fault,
Figure FDA00026309101500000312
and
Figure FDA00026309101500000313
respectively indicating the starting time and the ending time of the h fault of the j elevator,
and is
Figure FDA00026309101500000314
Is a piecewise continuous bounded function for representing the additive fault part of the jth elevator in the h fault, vj(t) represents a control signal of an elevator deflection angle.
5. The method of claim 4, wherein the smoothing function is constructed based on the saturation of the elevator yaw angle input in S3.
6. The method of claim 4, wherein the smoothing function is constructed based on fuel equivalence ratio input saturation at S3.
7. The method of claim 5, wherein the smoothing function is of the form:
Figure FDA0002630910150000041
sat(uj)＝ψ(uj)+ψd(uj)
wherein,
Figure FDA0002630910150000042
ψd(uj) Is a bounded function;
Figure FDA0002630910150000043
and
Figure FDA0002630910150000044
each represents ujUpper and lower bounds of (a); psi (u)j)＝ψaujNamely sat (u)j)＝ψauj+ψd(uj),ψaIs a continuously bounded non-linear function.
8. The method of claim 6, wherein the smoothing function is of the form:
Figure FDA0002630910150000045
sat(uV)＝ψaVuV+ψdV(uV)
wherein,
Figure FDA0002630910150000046
ψdV(uV) Is a bounded function;
Figure FDA0002630910150000047
and
Figure FDA0002630910150000048
each represents uVUpper and lower bounds of (a); psiaV(uV)＝ψauVNamely sat (u)V)＝ψaVuV+ψdV(uV)，ψaVIs a continuously bounded non-linear function.
9. The method of claim 7 or 8, wherein in S3, the estimating of the hypersonic flight vehicle by introducing the radial basis function neural networkIn the longitudinal dynamics model of (1) a non-linear function FiThe specific form of i ═ 1,2,3 is as follows:
Fi(ξi)＝θi Tφi(ξi)+Δ(ξi)，|Δ(ξi)|≤i
wherein, thetai∈RNIs the optimal weight vector, R, of N nodes in the radial basis functionNIs an N-dimensional real space, phii(ξi)＝[φi1(ξi),…,φiN(ξi)]T∈RNIs a vector of basis functions in the radial basis functions; delta ([ xi ])i) The error of the approximation is represented by,iis a constant number, wherein
Figure FDA0002630910150000049
οijAnd biThe center and width of the radial basis function, respectively; defining a constant
Figure FDA00026309101500000410
Figure FDA00026309101500000411
Is that
Figure FDA00026309101500000412
Is determined by the estimated value of (c),
Figure FDA00026309101500000413
to estimate the error, gmIs constant and 0 < gm≤min[inf{g1(V)},g2,inf{g3(V)},inf{gV(x1,x2,x3,h,V)}]。
CN201810262739.8A 2018-03-28 2018-03-28 Adaptive compensation control method of hypersonic aircraft based on neural network Active CN108375907B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810262739.8A CN108375907B (en) 2018-03-28 2018-03-28 Adaptive compensation control method of hypersonic aircraft based on neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810262739.8A CN108375907B (en) 2018-03-28 2018-03-28 Adaptive compensation control method of hypersonic aircraft based on neural network

Publications (2)

Publication Number Publication Date
CN108375907A CN108375907A (en) 2018-08-07
CN108375907B true CN108375907B (en) 2020-11-10

Family

ID=63031779

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810262739.8A Active CN108375907B (en) 2018-03-28 2018-03-28 Adaptive compensation control method of hypersonic aircraft based on neural network

Country Status (1)

Country Link
CN (1) CN108375907B (en)

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109062055A (en) * 2018-09-10 2018-12-21 南京航空航天大学 A kind of Near Space Flying Vehicles control system based on Back-stepping robust adaptive dynamic surface
CN108919651B (en) * 2018-09-26 2020-09-01 北京航空航天大学 Hypersonic aircraft self-adaptive fault-tolerant control method considering input constraint of actuator
CN109508035B (en) * 2018-12-24 2021-11-26 南京邮电大学 Multi-region hierarchical unmanned aerial vehicle formation path planning method based on distributed control
CN110231774A (en) * 2019-04-08 2019-09-13 天津大学 Disturbance-observer becomes air intake duct hypersonic aircraft fuzzy coordinated control method
CN110187713A (en) * 2019-04-12 2019-08-30 浙江大学 A kind of longitudinally controlled method of hypersonic aircraft based on aerodynamic parameter on-line identification
CN110340898B (en) * 2019-08-22 2020-12-01 北京航空航天大学 Self-adaptive fault-tolerant control method for free floating space manipulator
CN110488852B (en) * 2019-08-28 2020-12-01 北京航空航天大学 Full-profile self-adaptive control method for hypersonic aircraft
CN110673472B (en) * 2019-09-09 2022-06-28 南京理工大学 Adaptive robust control method based on neural network compensation dead zone inversion error
CN111007724A (en) * 2019-12-19 2020-04-14 哈尔滨工业大学 Hypersonic aircraft designated performance quantitative tracking control method based on interval II type fuzzy neural network
CN111158398A (en) * 2020-01-15 2020-05-15 哈尔滨工业大学 Adaptive control method of hypersonic aircraft considering attack angle constraint
CN112558484A (en) * 2020-12-29 2021-03-26 彩虹无人机科技有限公司 Design method of controller for displaying and considering dynamic characteristics of aircraft actuating mechanism
CN113961010B (en) * 2021-08-26 2023-07-18 中国科学院合肥物质科学研究院 Tracking control method for four-rotor plant protection unmanned aerial vehicle

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5102072A (en) * 1990-11-19 1992-04-07 General Dynamics Corporation, Convair Division Adaptive gain and phase controller for autopilot for a hypersonic vehicle
US6882992B1 (en) * 1999-09-02 2005-04-19 Paul J. Werbos Neural networks for intelligent control
CN102880052B (en) * 2012-09-29 2014-12-31 西北工业大学 Time scale function decomposition based hypersonic aircraft actuator saturation control method
CN104898682B (en) * 2015-05-05 2017-10-31 南京航空航天大学 A kind of hypersonic aircraft reenters posture fault tolerant control method
CN105843077B (en) * 2016-04-07 2019-03-29 杭州电子科技大学 A kind of gain switching method of Spacecraft Rendezvous system and maximum domain of attraction estimation
CN106842912B (en) * 2016-11-30 2020-08-14 南京航空航天大学 Hypersonic speed maneuvering flight control surface saturation robust control method
CN106934120B (en) * 2017-02-23 2020-02-11 哈尔滨工业大学 Three-dimensional guidance law design method for intercepting hypersonic aircraft based on forward guidance
CN107479384B (en) * 2017-09-05 2019-10-22 西北工业大学 The non-backstepping control method of hypersonic aircraft neural network Hybrid Learning

Also Published As

Publication number Publication date
CN108375907A (en) 2018-08-07

Similar Documents

Publication Publication Date Title
CN108375907B (en) Adaptive compensation control method of hypersonic aircraft based on neural network
CN108919651B (en) Hypersonic aircraft self-adaptive fault-tolerant control method considering input constraint of actuator
CN110333728B (en) Heterogeneous fleet fault tolerance control method based on variable time interval strategy
CN106444799B (en) Four-rotor unmanned aerial vehicle control method based on fuzzy extended state observer and self-adaptive sliding mode
CN107065539B (en) Control surface fault self-adaptive fault-tolerant control method of flying wing layout aircraft
CN103558857B (en) A kind of distributed compound anti-interference attitude control method of BTT aircraft
Jiang et al. Adaptive fault-tolerant tracking control of near-space vehicle using Takagi–Sugeno fuzzy models
CN110879535B (en) Sliding mode fault-tolerant control method of T-S fuzzy UMV
CN105629734B (en) A kind of Trajectory Tracking Control method of Near Space Flying Vehicles
CN104898431A (en) Reentry aircraft finite time control method based on disturbance observer
CN112666833B (en) Vehicle speed following self-adaptive robust control method for electric automatic driving vehicle
CN113268084B (en) Intelligent fault-tolerant control method for unmanned aerial vehicle formation
CN105137999A (en) Aircraft tracking control direct method with input saturation
CN107203138B (en) Aircraft robust control method with saturated input and output
CN105182742A (en) Elastic body aircraft adaptive constrained tracking control indirect method
CN110488603B (en) Rigid aircraft adaptive neural network tracking control method considering actuator limitation problem
CN113777932B (en) Four-rotor self-adaptive sliding mode fault-tolerant control method based on Delta operator
CN105629988A (en) Anti-interference attitude control method of drag-free satellite
CN105629739B (en) Anti-interference control method is fed back in a kind of output without towing satellite relative displacement passage
CN113741188A (en) Fixed-wing unmanned aerial vehicle backstepping self-adaptive fault-tolerant control method under actuator fault
CN105182990A (en) Robust control method of three-DOF model helicopter with output limits
CN113126491A (en) Anti-interference tracking control design method based on T-S fuzzy interference modeling
Ye et al. Adaptive dynamic surface control of switched MIMO nonlinear systems with input saturation and its application to NSVs
CN111208733B (en) Adaptive compensation control method for aircraft control system under multi-azimuth turbulent wind disturbance
CN109240317A (en) The submarine earthquake detection flight node finite time configuration for considering propeller failure includes control method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant