CN115659502A - Control surface fault diagnosis method of flying wing unmanned aerial vehicle based on neural network adaptive observer - Google Patents
Control surface fault diagnosis method of flying wing unmanned aerial vehicle based on neural network adaptive observer Download PDFInfo
- Publication number
- CN115659502A CN115659502A CN202211306320.0A CN202211306320A CN115659502A CN 115659502 A CN115659502 A CN 115659502A CN 202211306320 A CN202211306320 A CN 202211306320A CN 115659502 A CN115659502 A CN 115659502A
- Authority
- CN
- China
- Prior art keywords
- fault
- control surface
- control
- observer
- aerial vehicle
- 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.)
- Pending
Links
Images
Landscapes
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The invention discloses a control surface fault diagnosis method of a flying wing unmanned aerial vehicle based on a neural network adaptive observer, and belongs to the technical field of calculation, calculation or counting. Aiming at the comprehensive characteristics of strong coupling, strong nonlinearity and the like of the advanced layout flying wing unmanned aerial vehicle, decoupling processing and linear analysis are difficult to perform by adopting a conventional method after an actuator fails, the invention provides the fault diagnosis method based on the neural network adaptive observer, effectively processes the control surface characteristics of the flying wing unmanned aerial vehicle, and ensures the real-time performance and the fault estimation precision of the fault diagnosis system.
Description
Technical Field
The invention discloses an advanced layout unmanned aerial vehicle fault tolerance technology, particularly discloses a fault diagnosis method based on an RBF neural network adaptive observer, and belongs to the technical field of calculation, calculation or counting.
Background
In the last decade, the unmanned aerial vehicle with the advanced layout of multiple control surfaces has become a future development trend of the carrier-based unmanned aerial vehicle due to the advantages of attack and stealth. The landing environment of the unmanned aerial vehicle is very severe, the influence of sea waves, airflows and the like on the unmanned aerial vehicle during landing is more severe due to complex sea conditions, the control surfaces of the unmanned aerial vehicle with the flying wing layout are tightly distributed, the performance is enhanced, and meanwhile, the failure probability is increased to a certain extent, so that the failure occurrence probability of the unmanned aerial vehicle during landing is higher. In order to reduce the occurrence frequency of faults, a safer and more reliable fault-tolerant control system needs to be designed, and an effective fault diagnosis system is the key for optimizing the active fault-tolerant control.
On one hand, the multi-control-surface aerodynamic layout of the flying wing unmanned aerial vehicle provides sufficient hardware support for fault-tolerant control, and on the other hand, the compact control-surface layout of the flying wing unmanned aerial vehicle enables the channel coupling to be stronger and the nonlinear control characteristic to be more prominent. Therefore, for a multivariable system of the multi-control surface unmanned aerial vehicle with characteristics of nonlinearity, strong coupling, uncertainty and the like, the requirement of control tracking accuracy is difficult to meet by adopting a linear fault diagnosis and reconstruction method, and once a fault occurs and effective measures for detection, positioning, isolation and estimation are not timely taken, irreparable loss can be caused.
According to the control surface characteristics of the flying wing layout unmanned aerial vehicle, the redundant control surfaces of the unmanned aerial vehicle can be fully utilized by the active fault-tolerant method based on control redistribution, and the control redistribution method depends on the information acquisition of a fault diagnosis module, so that higher requirements on the precision and the real-time performance of fault detection and estimation are provided; meanwhile, the strong nonlinearity and strong coupling of the control surface of the flying wing aircraft layout bring great difficulty to fault diagnosis, and the current fault diagnosis research results aiming at the nonlinear characteristic of the control surface of the flying wing aircraft are few.
Fault detection and diagnosis methods have been developed to date and are generally classified into three categories, namely analytical model methods that rely on mathematical models, signal processing-based methods that do not rely on mathematical models, and knowledge-based methods. The analytical model method is most perfect in theoretical development and high in accuracy of fault diagnosis, but is high in processing difficulty and small in application range of nonlinear models and objects with strong interference, the fault diagnosis method based on signal processing is used more in detecting faults of aircraft sensors, the fault diagnosis method based on knowledge is independent of models, disturbance and unknown nonlinear dynamic items can be well processed, the method has great advantages in processing fault diagnosis problems of nonlinear systems compared with the analytical model method, and the method is a research hotspot in recent years, and is successfully applied to methods such as principal component analysis, genetic algorithm, neural network and fuzzy logic.
Compared with the traditional method, the neural network has stronger advantages on the fault diagnosis of the control surfaces with strong nonlinear and cross-coupling effects by virtue of good nonlinear fitting characteristics, and the RBF neural network can be better suitable for the fault estimation problem with higher requirements on real-time performance and local precision by virtue of better local approximation capability compared with other neural networks. On the other hand, the conventional fault diagnosis method based on the fault residual observer constructs a normal state observer approaching to a non-fault model, analyzes the fault residual signal of the observer and the output of the system after fault, puts high requirements on the construction accuracy of the normal state model, is difficult to process the disturbance and other model uncertainties in the flight process, and can greatly reduce the fault estimation precision.
Therefore, the invention aims to provide a fault diagnosis method based on a neural network adaptive observer, which solves the problem of control surface fault diagnosis considering the cross coupling effect of control surface deflection nonlinearity so as to overcome the problem that the conventional linear fault diagnosis scheme cannot be applied to the complex control characteristics of a flying wing unmanned aerial vehicle.
Disclosure of Invention
The invention aims to provide a neural network self-adaptive fault diagnosis observer aiming at the defects of the background technology, and the method adopts a self-adaptive online fault diagnosis framework combining an RBF neural network and the observer based on the fault diagnosis observer aiming at the faults of the actuator of the unmanned aerial vehicle with multiple control surfaces, so that the accuracy and the real-time performance of fault estimation are ensured.
The invention adopts the following technical scheme for realizing the aim of the invention:
an affine nonlinear mathematical model is established for the flying wing layout unmanned aerial vehicle, three-channel control moment coefficients of rolling, pitching and yawing are calculated by adopting a nonlinear dynamic inverse flight control law, a virtual control instruction is established, and the attitude angle and the attitude angular velocity of the flying wing aircraft are updated in real time by the control moment coefficient virtual control instruction output by an attitude inner loop;
the method comprises the steps of establishing an actuator model and a control surface fault model of the flying wing unmanned aerial vehicle, designing a fault diagnosis device based on a neural network self-adaptive network fault observer according to a body model, the actuator model and the fault model of the unmanned aerial vehicle, wherein the observer comprises unknown input, namely a fault to be estimated, estimating a fault item in the observer on line through a neural network so as to form a self-adaptive item of the observer, and when a system output error reaches an alarm threshold value, giving an alarm by the fault alarm device, and starting the self-adaptive fault diagnosis observer to perform on-line diagnosis of the control surface fault.
The affine nonlinear mathematical system of the flying wing unmanned aerial vehicle is
Wherein x is 1 ＝[φ，θ，ψ] T For state vectors in which the affine nonlinear dynamical system changes relatively slowly, x 2 ＝[p，q，r] T For a state vector with relatively fast change of an affine nonlinear dynamical system, phi, theta and psi are three attitude angles of a rolling angle, a pitching angle and a yaw angle of the unmanned aerial vehicle, p, q and r are three angular velocity components of corresponding attitude angles under a machine body coordinate axis system, and a virtual control command v = [ C ] is adopted lδ ，C mδ ，C mδ ] T ＝C(δ)，C lδ ，C mδ ，C nδ Controlling the torque coefficient for three channels of rolling, pitching and yawing, and C (-) is a control torque systemMapping of number to control surface deflection, f 1 (·)，f 2 (·)，g 1 (·)，g 2 (. Cndot.) are the state function and control function of the outer loop and inner loop, respectively.
Considering the influence of control surface deflection nonlinearity and cross coupling nonlinearity, fitting a nonlinear dynamic efficiency model of the control surface as follows:
wherein δ = [ δ ] l1 ，δ l2 ，δ l3 ，δ l4 ，δ r1 ，δ r2 ，δ r3 ，δ r4 ] T To control the amount of surface deflection, delta l1 ，δ l2 ，δ l3 ，δ r1 ，δ r2 ，δ r3 Respectively left and right elevon, delta l4 ，δ r4 Are respectively a left resistance rudder and a right resistance rudder,
the roll control moment coefficient and the pitch control moment coefficient of the ith control surface respectively,
a nonlinear fitting expression, p, of the yaw channel control moment coefficient of the ith control surface i3 ，p i2 ，p i1 ，p i0 Coefficient of cubic, coefficient of quadratic, coefficient of first order, constant term, Δ C, for a non-linear fit expression l (δ l3 ，δ l4 )，ΔC l (δ r3 ，δ r4 )，ΔC m (δ l3 ，δ l4 )，ΔC m (δ r3 ，δ r4 )，ΔC n (δ l3 ，δ l4 )，ΔC n (δ r3 ，δ r4 ) And the three channels are respectively used for controlling moment coefficients by the cross coupling of the left third lifting aileron and the right third lifting aileron and the resistance rudder at the same side.
Control surface faults of flying-wing drones are usually damage, seizing, saturation, floating, etc., parameterized as follows:
δ i inputting commands for controlling surface deflection, delta imax ，δ imin Upper and lower limits of control surface deflection constraint, respectively a (δ, t) is the control surface deflection output after the fault, and the relationship between the control surface deflection output and the control surface deflection output is summarized as follows:
wherein Λ = diag { α) 1 ，α 2 ，…，α 8 }， K＝diag{k 1 ，k 2 ，…，k 8 }，
α i Characterize the type of failure as i =1, the i-th control surface is completely deactivated (seized or floating), when α is i If =0, the i-th control surface is normal or partially failed. k is a radical of i For the remaining operating efficiency under damage failure, k i =0 denotes a floating fault in the i-th control surface, 0 < k i < 1 denotes the remaining operational capability of the i-th control surface in the event of a partial failure, k i And =1 indicates that the control surface is operating normally. Delta i (t F ) For control surfaces at t F A stuck position where the moment is completely failed.
The mapping relation between the deflection variable of the fault control surface and the three-channel control moment coefficient is
Therefore, under the fault state, the attitude angular velocity loop mathematical model of the flying wing unmanned aerial vehicle is described as follows:
wherein the content of the first and second substances,
epsilon is the system output errorAnd when the observer detects that the actuator has a fault, the alarm gives out a fault alarm and starts the neural network fault diagnosis observer.
The nonlinear adaptive observer is designed as follows:
wherein, L is an observer gain matrix designed according to the Hurwitz theorem,
is the estimated deflection of the fault control surface, and is used as an adaptive item of an observer to carry out online estimation by a neural network.
Considering the strong nonlinearity and strong coupling of a control efficiency model of the control surface and the requirements on the real-time property and the local precision of fault estimation, an RBF neural network is adopted as an adaptive adjustment mechanism of an observer to control surface fault information
Carrying out online estimation;
in order to reflect the fault information fully and accurately, the input of the estimator selects the actual output state y of the system and the output state of the observer
Reference input delta to control plane c . There is an ideal weight matrix W * ∈R n×q So that
Wherein
Is a radial basis function, i.e.
C j Is a center of Gaussian, b j As basis-width vector, RBF neural networkAnd reflecting the fault estimation error through the state approximation error of the adaptive observer, thereby realizing the adaptive adjustment of the weight matrix W.
Because the system state is measurable, the fault approximation error is only caused by the approximation error of the neural network, and the observed value dynamic error equation is obtained as follows:
when the control surface fault estimated by the neural network approaches to a real fault, designing a nonlinear system observer of the unmanned aerial vehicle, and only designing a gain matrix L to enable an error equation to be asymptotically stable, namely, the state error of the observer can reflect the estimated approximation degree of the fault to the real fault; according to the deflection relation of the formula fault control surface, the control surface fault estimated by the neural network is analyzed and compared with the expected deflection, so that fault parameters such as the fault type, the failure factor, the blocking position and the like of the fault control surface can be obtained, and the fault control method is used for fault-tolerant control based on redistribution.
By adopting the technical scheme, the invention has the following beneficial effects:
(1) The invention estimates the control surface fault by designing a fault diagnosis strategy based on a neural network adaptive fault observer aiming at the control surface deflection and cross coupling nonlinear characteristics of the flying wing unmanned aerial vehicle and the accuracy requirement and real-time requirement of fault information estimation, and provides an information basis for fault reconstruction.
(2) The method utilizes stronger nonlinear fitting characteristics of the neural network to carry out fault diagnosis on the control surface with strong nonlinear deflection effect, introduces the observer to provide error approximation basis for the neural network based on the deterministic equivalence principle, and is more suitable for effective fault estimation of the object compared with the traditional nonlinear method for linearizing the rudder deflection effect and being difficult to be embodied on the actuator, thereby improving the fault diagnosis capability.
Drawings
Fig. 1 is an architecture diagram of a fault tolerance method of a flying wing drone according to the present invention.
Fig. 2 is a diagram of a fault diagnosis based on a neural network observer according to the present invention.
Fig. 3 is a diagram of control surface input commands, output commands and dead-lock position estimation in the state of dead-lock failure of the first elevon at the right side of the beginning of 2 s.
Fig. 4 to 5 are graphs showing input commands, output commands and failure factor estimation of the control surface of the present invention in a state where the first elevon on the right side has a damage failure (failure factor 0.7).
Detailed Description
In order to facilitate understanding of those skilled in the art, the present invention will be further described with reference to the accompanying drawings.
The invention provides a control surface fault diagnosis method of a flying wing unmanned aerial vehicle, which solves the problem of fault diagnosis of an aircraft with a nonlinear control surface effect after an actuator has a fault.
Step one, establishing an affine nonlinear mathematical model for the flying wing layout unmanned aerial vehicle, calculating three-channel control moment coefficients of rolling, pitching and yawing by adopting a nonlinear dynamic reverse flight control law, constructing a virtual control command, and updating the attitude angle and the attitude angular velocity of the flying wing aircraft in real time by using the control moment coefficient virtual control command output by an attitude inner loop. The affine nonlinear mathematical system of the flying wing unmanned aerial vehicle is expressed as follows:
wherein x is 1 ＝[φ，θ，ψ] T For state vectors in which the affine nonlinear dynamical system changes relatively slowly, x 2 ＝[p，q，r] T The state vector with relatively fast change of an affine nonlinear dynamical system is represented by phi, theta and psi, the phi, theta and psi are three attitude angles of a rolling angle, a pitching angle and a yaw angle of the unmanned aerial vehicle, p, q and r are three angular velocity components of corresponding attitude angles under a body coordinate axis system, and a virtual control instruction v = [ C ] lδ ，C mδ ，C nδ ] T ＝C(δ)，C lδ ，C mδ ，C nδ Controlling the torque coefficient for three channels of rolling, pitching and yawing, C (-) is the mapping relation between the control torque coefficient and the control surface deflection, f 1 (·)，f 2 (·)，g 1 (·)，g 2 (. Cndot.) are the state function and control function of the outer loop and inner loop, respectively.
And step two, considering the influence of control surface deflection nonlinearity and cross coupling nonlinearity, and fitting the nonlinear dynamic efficiency model of the control surface as follows:
wherein δ = [ δ ] l1 ，δ l2 ，δ l3 ，δ l4 ，δ r1 ，δ r2 ，δ r3 ，δ r4 ] T To control the amount of surface deflection, delta l1 ，δ l2 ，δ l3 ，δ r1 ，δ r2 ，δ r3 Respectively left and right elevon, delta l4 ，δ r4 Are respectively a left resistance rudder and a right resistance rudder,
the roll control moment coefficient and the pitch control moment coefficient of the ith control surface respectively,
a non-linear fitting expression, p, for the yaw channel control moment coefficient of the ith control surface i3 ，p i2 ，p i1 ，p i0 Coefficient of cubic, coefficient of quadratic, coefficient of first order, constant term, Δ C, for a non-linear fit expression l (δ l3 ，δ l4 )，ΔC l (δ r3 ，δ r4 )，ΔC m (δ l3 ，δ l4 )，ΔC m (δ r3 ，δ r4 )，ΔC n (δ l3 ，δ l4 )，ΔC n (δ r3 ，δ r4 ) Three-channel control for cross coupling of left and right third lifting ailerons and same-side resistance rudderAnd (5) torque coefficient control.
Step three, the control surface faults of the flying wing unmanned aerial vehicle are usually damage, jamming, saturation, floating and the like, and the parameterized form is as follows:
δ i for controlling surface deflection input of command, delta imax ，δ imin Upper and lower limits of control surface deflection constraint, respectively a (δ, t) is the control surface deflection output after the fault, and the relationship between the control surface deflection output and the control surface deflection output is summarized as follows:
wherein Λ = diag { α) 1 ，α 2 ，…，α 8 }， K＝diag{k 1 ，k 2 ，…，k 8 }，
α i Characterizing the type of failure as i =1, the i-th control surface is completely deactivated (seized or floating), when α is i If =0, the i-th control surface is normal or partially failed. k is a radical of i For the remaining operating efficiency under damage failure, k i =0 denotes a floating fault in the i-th control surface, 0 < k i < 1 denotes the remaining operational capability of the i-th control surface in the event of a partial failure, k i And =1 indicates that the control surface is operating normally. Delta i (t F ) For control surfaces at t F A stuck position that is completely failed at that moment.
The mapping relation between the deflection variable of the fault control surface and the three-channel control moment coefficient is
Therefore, the mathematical model of the attitude angular velocity loop of the flying wing unmanned aerial vehicle under the fault state is described as
Wherein the content of the first and second substances,
and epsilon is a fault alarm threshold value of the system output error, when the observer detects that the actuator has a fault, the alarm gives out a fault alarm, and the neural network fault diagnosis observer is started at the same time.
Step four, designing the nonlinear adaptive observer as follows:
wherein, L is an observer gain matrix designed according to the Hurwitz theorem,
is the estimated deflection of the fault control surface, and is used as an adaptive item of an observer to carry out online estimation by a neural network.
Considering the strong nonlinearity and strong coupling of a control efficiency model of the control surface and the requirements on the real-time property and the local precision of fault estimation, an RBF neural network is adopted as an adaptive adjustment mechanism of an observer to control surface fault information
Carrying out online estimation; in order to reflect the fault information fully and accurately, the input of the estimator selects the actual output state y of the system and the output state of the observer
With control surface reference input delta c . There is an ideal weight matrix W * ∈R n×q So that
Wherein
Is a radial basis function, i.e.
C j Is a center of Gaussian, b j The RBF neural network reflects the fault estimation error through the state approximation error of the adaptive observer to form a base width vector, so that the adaptive adjustment of the weight matrix W is realized.
Step five, because the system state can be measured, the fault approximation error is only caused by the approximation error of the neural network, and the dynamic error equation of the observed value is obtained as follows:
when the control surface fault estimated by the neural network approaches to a real fault, designing a nonlinear system observer of the unmanned aerial vehicle, and only designing a gain matrix L to enable an error equation to be asymptotically stable, namely, the state error of the observer can reflect the estimated approximation degree of the fault to the real fault; according to the deflection relation of the formula fault control surface, the control surface fault estimated by the neural network is analyzed and compared with the expected deflection, so that fault parameters such as the fault type, the failure factor, the blocking position and the like of the fault control surface can be obtained, and the fault control method is used for fault-tolerant control based on redistribution.
The invention carries out fault diagnosis simulation on the flying-wing unmanned aerial vehicle, establishes a self-adaptive fault diagnosis observer based on a neural network in the simulation, and carries out the simulation process in MATLAB. FIG. 3 shows that when a stuck fault occurs on one control surface at a time other than 0, the neural network fault diagnosis observer can quickly and accurately estimate the stuck position; fig. 4 to 5 show that after the control surface has a damage fault, the neural network fault diagnosis observer can quickly and accurately estimate the remaining working capacity of the control surface, i.e., the failure factor, and the fault estimation result can still be quickly converged to a stable state under the condition that the control distribution command cannot be smoothly responded due to the fault. Therefore, the fault diagnosis method has good real-time performance and accuracy.
Claims (8)
1. A fault diagnosis method for a control surface of a flying wing unmanned aerial vehicle based on a neural network adaptive observer is characterized in that an affine nonlinear mathematical model is established for the flying wing unmanned aerial vehicle, three-channel control moment coefficients of rolling, pitching and yawing are calculated by adopting a nonlinear dynamic inverse flight control law, a virtual control instruction is established, and the attitude angle and the attitude angular velocity of the flying wing unmanned aerial vehicle are updated in real time by the control moment coefficient virtual control instruction output by an attitude inner loop;
the method comprises the steps of establishing an actuator model and a control surface fault model of the flying wing unmanned aerial vehicle, designing a fault diagnosis device based on a neural network self-adaptive network fault observer according to a body model, the actuator model and the fault model of the unmanned aerial vehicle, wherein the observer comprises unknown input, namely a fault to be estimated, estimating a fault item in the observer on line through a neural network so as to form a self-adaptive item of the observer, and when a system output error reaches an alarm threshold value, giving an alarm by the fault alarm device, and starting the self-adaptive fault diagnosis observer to perform on-line diagnosis of the control surface fault.
2. The method for diagnosing the control surface fault of the flying wing unmanned aerial vehicle based on the neural network adaptive observer is characterized in that the affine nonlinear mathematical system of the flying wing unmanned aerial vehicle is expressed as follows:
wherein x is 1 ＝[φ，θ，ψ] T For state vectors in which the affine nonlinear dynamical system changes relatively slowly, x 2 ＝[p，q，r] T The state vector with relatively fast change of an affine nonlinear dynamical system is represented by phi, theta and psi, the phi, theta and psi are three attitude angles of a rolling angle, a pitching angle and a yaw angle of the unmanned aerial vehicle, p, q and r are three angular velocity components of corresponding attitude angles under a body coordinate axis system, and a virtual control instruction v = [ C ] lδ ，C mδ ，C nδ ] T ＝C(δ)，C lδ ，C mδ ，C nδ Controlling moment coefficient for three channels of rolling, pitching and yawing, and C (-) controlling moment coefficient and controllingMapping of surface deflection, f 1 (·)，f 2 (·)，g 1 (·)，g 2 (. Cndot.) are the state function and control function of the outer loop and inner loop, respectively.
3. The method for diagnosing the faults of the control surfaces of the flying wing unmanned aerial vehicle based on the neural network adaptive observer is characterized in that the influences of control surface deflection nonlinearity and cross coupling nonlinearity are considered, and a nonlinear dynamic efficiency model of the control surfaces is fitted as follows:
wherein δ = [ δ ] l1 ，δ l2 ，δ l3 ，δ l4 ，δ r1 ，δ r2 ，δ r3 ，δ r4 ] T To control the amount of surface deflection, delta l1 ，δ l2 ，δ l3 ，δ r1 ，δ r2 ，δ r3 Respectively left and right elevon, delta l4 ，δ r4 Are respectively a left resistance rudder and a right resistance rudder,
the roll control moment coefficient and the pitch control moment coefficient of the ith control surface respectively,
a nonlinear fitting expression, p, of the yaw channel control moment coefficient of the ith control surface i3 ，p i2 ，p i1 ，p i0 Coefficient of cubic, coefficient of quadratic, coefficient of first order, constant term, Δ C, for a non-linear fit expression l (δ l3 ，δ l4 )，ΔC l (δ r3 ，δ r4 )，ΔC m (δ l3 ，δ l4 )，ΔC m (δ r3 ，δ r4 )，ΔC n (δ l3 ，δ l4 )，ΔC n (δ r3 ，δ r4 ) And the three channels are respectively used for controlling moment coefficients by the cross coupling of the left third lifting aileron and the right third lifting aileron and the resistance rudder at the same side.
4. The method for diagnosing the control surface faults of the flying-wing unmanned aerial vehicle based on the neural network adaptive observer is characterized in that the control surface faults of the flying-wing unmanned aerial vehicle are usually damage, seize, saturation, floating and the like, and the parameterized form is as follows:
δ i inputting commands for controlling surface deflection, delta imax ，δ imin Upper and lower limits of control surface deflection constraint, respectively a (delta, t) is the control surface deflection output after the fault, and the relationship is summarized as follows:
wherein Λ = diag { α) 1 ，α 2 ，…，α 8 }，K＝diag{k 1 ，k 2 ，…，k 8 }，
α i Characterizing the type of failure as i =1, the i-th control surface is completely deactivated (seized or floating), when α is i If =0, the i-th control surface is normal or partially failed. k is a radical of i For the remaining operating efficiency under damage failure, k i =0 denotes that the i-th control surface has a floating fault, 0 < k i < 1 denotes the remaining operational capability of the i-th control surface in the event of a partial failure, k i And =1 indicates that the control surface is operating normally. Delta i (t F ) For control surfaces at t F A stuck position where the moment is completely failed.
5. The method for diagnosing the control surface fault of the flying wing unmanned aerial vehicle based on the neural network adaptive observer according to claim 4,the method is characterized in that the mapping relation between the deflection variable of the fault control surface and the three-channel control moment coefficient is
Therefore, under the fault state, the attitude angular velocity loop mathematical model of the flying wing unmanned aerial vehicle is described as follows:
wherein the content of the first and second substances,
and epsilon is a fault alarm threshold value of the system output error, when the observer detects that the actuator has a fault, the alarm gives out a fault alarm, and the neural network fault diagnosis observer is started at the same time.
6. The method for diagnosing the control surface fault of the flying wing unmanned aerial vehicle based on the neural network adaptive observer is characterized in that the nonlinear adaptive observer is designed as follows:
wherein L is an observer gain matrix designed according to Hurwitz's theorem,
is the estimated deflection of the fault control surface, and is used as an adaptive item of an observer to carry out online estimation by a neural network.
7. The method of claim 5, wherein the control surface fault diagnosis of the flying wing unmanned aerial vehicle based on the neural network adaptive observer is performed by considering strong nonlinearity and strong coupling of the control efficiency model of the control surface and the faultThe estimated real-time performance and the local precision are required, an RBF neural network is adopted as an adaptive adjustment mechanism of an observer, and the control surface fault information is subjected to
Carrying out online estimation;
in order to reflect the fault information fully and accurately, the input of the estimator selects the actual output state y of the system and the output state of the observer
Reference input delta to control plane c . There is an ideal weight matrix W * ∈R n×q So that
Wherein
Each component of (a) is a radial basis function, i.e.
C i Is center of Gaussian, b j The RBF neural network reflects the fault estimation error through the state approximation error of the adaptive observer to form a base width vector, so that the adaptive adjustment of the weight matrix W is realized.
8. The method for diagnosing the control surface fault of the flying wing unmanned aerial vehicle based on the neural network adaptive observer is characterized in that the fault approximation error is only caused by the neural network approximation error due to the measurable system state, and an observed value dynamic error equation is obtained as follows:
when the control surface fault estimated by the neural network approaches to a real fault, designing a nonlinear system observer of the unmanned aerial vehicle, and only designing a gain matrix L to enable an error equation to be asymptotically stable, namely, the state error of the observer can reflect the estimated approximation degree of the fault to the real fault; according to the deflection relation of the formula fault control surface, the control surface fault estimated by the neural network is analyzed and compared with the expected deflection, so that fault parameters such as the fault type, the failure factor, the blocking position and the like of the fault control surface can be obtained, and the fault control method is used for fault-tolerant control based on redistribution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211306320.0A CN115659502A (en) | 2022-10-24 | 2022-10-24 | Control surface fault diagnosis method of flying wing unmanned aerial vehicle based on neural network adaptive observer |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211306320.0A CN115659502A (en) | 2022-10-24 | 2022-10-24 | Control surface fault diagnosis method of flying wing unmanned aerial vehicle based on neural network adaptive observer |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115659502A true CN115659502A (en) | 2023-01-31 |
Family
ID=84990431
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211306320.0A Pending CN115659502A (en) | 2022-10-24 | 2022-10-24 | Control surface fault diagnosis method of flying wing unmanned aerial vehicle based on neural network adaptive observer |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115659502A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116185057A (en) * | 2023-03-24 | 2023-05-30 | 西北工业大学 | Attitude fault-tolerant control method for wing body fusion flying unmanned aerial vehicle |
-
2022
- 2022-10-24 CN CN202211306320.0A patent/CN115659502A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116185057A (en) * | 2023-03-24 | 2023-05-30 | 西北工业大学 | Attitude fault-tolerant control method for wing body fusion flying unmanned aerial vehicle |
| CN116185057B (en) * | 2023-03-24 | 2023-09-01 | 西北工业大学 | Attitude fault-tolerant control method for wing body fusion flying unmanned aerial vehicle |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chen et al. | Human-in-the-loop consensus tracking control for UAV systems via an improved prescribed performance approach | |
| CN113568423B (en) | Intelligent fault-tolerant control method of four-rotor unmanned aerial vehicle considering motor faults | |
| CN111781942B (en) | Fault-tolerant flight control method based on self-constructed fuzzy neural network | |
| CN114879512B (en) | Spacecraft formation orbit fault-tolerant control method based on learning neural network sliding mode | |
| CN111880410A (en) | Four-rotor unmanned aerial vehicle fault-tolerant control method for motor faults | |
| Li et al. | Finite-time control for quadrotor based on composite barrier Lyapunov function with system state constraints and actuator faults | |
| CN115659502A (en) | Control surface fault diagnosis method of flying wing unmanned aerial vehicle based on neural network adaptive observer | |
| Wu et al. | Hierarchical fault-tolerant control for over-actuated hypersonic reentry vehicles | |
| Ma et al. | Deep auto-encoder observer multiple-model fast aircraft actuator fault diagnosis algorithm | |
| Li et al. | Learning-observer-based adaptive tracking control of multiagent systems using compensation mechanism | |
| CN115657479A (en) | Improved dynamic MOPSO-based control surface fault tolerance method for flying-wing unmanned aerial vehicle | |
| Ma et al. | Civil aircraft fault tolerant attitude tracking based on extended state observers and nonlinear dynamic inversion | |
| CN113341973B (en) | Course control method based on asymmetric phase difference of flapping wings | |
| Wu et al. | Elman Neural Network‐Based Direct Lift Automatic Carrier Landing Nonsingular Terminal Sliding Mode Fault‐Tolerant Control System Design | |
| CN115981265A (en) | Shipboard aircraft fault online detection method based on extended observer | |
| CN112631129B (en) | Fault-tolerant flight control method and system for elastic aircraft | |
| Zhao et al. | Active fault-tolerant strategy for flight vehicles: Transfer learning-based fault diagnosis and fixed-time fault-tolerant control | |
| Li et al. | Fault-tolerant aircraft control based on self-constructing fuzzy neural network for quadcopter | |
| Hu et al. | Improved adaptive compensation of variant fighter with multiple faults via extended observer | |
| CN114035597A (en) | Self-adaptive global sliding mode fault-tolerant control method based on Barrier function | |
| Emami et al. | Disturbance observer-based adaptive neural guidance and control of an aircraft using composite learning | |
| Singh et al. | Aerodynamic parameter estimation using two-stage radial basis function neural network | |
| Cui et al. | Reconfiguration control design of UAV against actuator faults based on control allocation method | |
| Dong et al. | Attitude compensation control for quadrotor under partial loss of actuator effectiveness | |
| Yang et al. | Projection operator-based fault-tolerant backstepping adaptive control of fixed-wing UAV against actuator faults |
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 |