INTERVAL ERROR OBSERVER-BASED AIRCRAFT ENGINE ACTIVE FAULT TOLERANT CONTROL METHOD

Abstract

The present invention provides an interval error observer-based aircraft engine active fault tolerant control method, and belongs to the technical field of aircraft control. The method comprises: tracking the state and the output of a reference model of an aircraft engine through an error feedback controller; compensating a control system of the aircraft engine having a disturbance signal and actuator and sensor faults through a virtual sensor and a virtual actuator; observing an error between a system with fault of the aircraft engine and the reference model through an interval error observer, and feeding back the error to the error feedback controller; and finally, using a difference between the output of the reference model of the system with fault and the output of the virtual actuator as a control signal to realize active fault tolerant control of the aircraft engine.

Description

TECHNICAL FIELD

The present invention relates to an interval error observer-based aircraft engine active fault tolerant control method, belongs to the technical field of aircraft control, and particularly relates to an active fault tolerant control method which is applied when aircraft engines have actuator and sensor fault under disturbances.

BACKGROUND

As the only power plant of an aircraft, an aircraft engine directly affects the safety, reliability and economy of the aircraft. Although reliable control system design can reduce the incidence of system fault, the actual system has complex structure and operates at high intensity; the factors that may cause the fault in the system are increased greatly; the types of the fault become increasingly diverse; and the fault of components is inevitable. As a driving component of the aircraft engine, the actuator is closely related to the state adjustment of the system. The actuator has large workload and complicated structure, and easiest to fail. Once the actuator fails, the entire system is collapsed, which causes a serious impact. The sensor is responsible for receiving and transmitting information of the aircraft engine system. The presence of sensor fault directly affects the safety and the reliability of the system. Therefore, it is of great significance to improve the fault-tolerant capacity of the system and ensure the stability and performance criteria of the system after the fault. Traditional fault tolerant control methods also face new challenges.

In general, fault tolerant control research methods are classified into two categories: passive fault tolerant control and active fault tolerant control. The idea of passive fault tolerant control is to pre-design a controller based on pre-judged possible faults, and passive fault tolerant control is a controller design method based on a robust control technology. When a fault occurs, the designed controller is called to keep an entire closed-loop system insensitive to the fault, thereby achieving the stability of the system. However, as the system becomes more and more complex, the types and number of the faults that may occur are increased. Therefore, the traditional passive fault tolerant control has great limitations, that is, all possible fault conditions need to be considered in advance, resulting in certain conservation of the controller. To reduce the conservation of the control system, active fault tolerant control that reconfigures the system becomes a research hotspot. The idea of active fault tolerant control is to realize online fault compensation by readjusting the parameters of the controller or reconfiguring the system after the fault occurs. That is, when there is no fault, the system is operated normally; and once the fault occurs, the system automatically adjusts or reconfigures a control law. The aircraft engine can be generally described as a linear-parameter-varying (LPV) system. Existing research results use a gain self-scheduling H-infinite optimization method when processing active fault tolerant control of the LPV system having actuator and sensor faults. The method readjusts the controller parameters when the system has the fault, thereby increasing the complexity of the system design. In addition, the control system of the aircraft engine is often interfered by noise signals. The existing methods have no ideal solution for the active fault tolerant control of the sensors and actuator faults of the aircraft engine when processing the interference signals.

SUMMARY

The technical problem of the present invention is: when the aircraft engine have the actuator and sensor faults and the control system is affected by noise signal interference, to solve the defects of the existing control method, the present invention provides an interval error observer-based aircraft engine active fault tolerant control method which can ensure that the aircraft engine can track a reference model without changing the structure and parameters of the controller. Namely, the reconfigured system has the same state and output as an original fault-free system, realizes a desired control objective, enables the system to have the capability to eliminate the faults autonomously, enhances the operating reliability of the aircraft engine and reduces maintenance cost of the aircraft engine.

The technical solution of the present invention is:

An interval error observer-based aircraft engine active fault tolerant control method comprises the following steps:

step 1.1: establishing an affine parameter-dependent aircraft engine linear-parameter-varying (LPV) model

{dot over (x)}_p(t)=[A₀+ΔA(θ)]x_p(t)+[B₀+ΔB(θ)]u_p(t)+d_f(t)

y _p(t)=C_px_p(t)+v(t)   (1)

where R^m and R^{m n} respectively represent a m-dimensional real number column vector and a m-row n-column real matrix; state vectors x_p=[Y_nl Y_nh]^T ∈ R^nx, Y_nl and Y_nh respectively represent variation of relative conversion speed of low pressure and high pressure rotors; n_x represents the dimension of a state variable x; n_y represents the dimension of an output vector y; n_u represents the dimension of control input u_p; control input u_p=U_pf ∈ R^nu is a fuel pressure step signal; output vectors y_p=Y_nh ∈ R^ny, A₀ ∈ R^nx×nx, B₀ ∈ R^nx×nx and C_p ∈ R^nx×nx are known system constant matrices; d_f(t) is a disturbance variable; the relative conversion speed n_h of the high pressure rotor of the aircraft engine is a scheduling parameter θ∈ R^p; system variable matrices ΔA(θ) and ΔB(θ) satisfy −ΔA≤ΔA(θ)≤ΔA and −ΔB≤ΔB(θ)≤ΔB; ΔA ∈ R^nx×nx is an upper bound of ΔA(θ); ΔB ∈ R^nx×nx is an upper bound of ΔB(θ); ΔA≥0; ΔB≥0; a state variable initial value x_p(0) satisfies x₀≤x_p(0)≤x₀; x₀, x₀ ∈ R^nx are respectively known upper bound and lower bound of the state variable initial value x_p(0); d, d ∈ R^nx are known upper bound and lower bound of an unknown disturbance d_f(t); sensor noise v(t) satisfies |v(t)|<V; V is a known bound; V>0;

step 1.2: representing the reference model of the fault-free system of the aircraft engine as

{dot over (x)}_pref(t)=A₀x_pref(t)+B₀u_pref(t)

y _pref(t)=C_px_pref(t)   (2)

where x_pref ∈ R^nx is a reference state vector of the fault-free system; u_pref ∈ R^nu is control input of the fault-free system; y_pref ∈ R^ny is a reference output vector; an error feedback controller of the fault-free system of the aircraft engine is designed according to the aircraft engine LPV model established in the step 1.1;

step 1.2.1: defining an error e_p(t)=x_pref(t)−x_p(t) between the affine parameter-dependent aircraft engine LPV model and the reference model of the fault-free system of the aircraft engine to obtain error state equations of the fault-free system:

ė_p(t)=[A₀+ΔA(θ)]e_p(t)+[B₀+ΔB(θ)]Δu_cp(t)−ΔA(θ)x_pref(t)−ΔB(θ)u_pref(t)−d_f(t)

ε_cp(t)=C_pe_p(t)=v(t)   (3)

where Δu_cp(t)=u_pref(t)−u_p(t) and ε_cp(t)=y_pref(t)−y_p(t);

step 1.2.2: representing state equations of the upper bound ē_p and the lower bound e_p of the error vector e_p as:

{dot over (e)}_p(t)=[A₀−LC_p]ē_p(t)+[B₀+ΔB]Δu_cp(t)+Lε_cp(t)+|L|V−d(t)+ΔA|x_pref(t)|+ϕ_p(t)

{dot over (e)}_p(t)=[A₀−LC_p]e_p(t)+[B₀−ΔB]Δu_cp(t)+Lε_cp(t)−|L|V−d(t)−ΔA|x_pref(t)|−ϕ_p(t)   (4)

where ē_p, e_p ∈ R^nx are respectively the upper bound and the lower bound of the error vector e_P, i.e., e_p(t)≤e_p(t)≤ē_p(t); ϕ_p(t)=ΔA(ē_p⁺(t)+e_p⁻(t)), ē_p⁺=max {0, ē_p}, ē_p⁻=ē_p⁺−ē_p, e_p⁺=max{0, e_p}, e_p⁻=e_p⁺−e_p; L ∈ R^nx×ny is an error gain matrix of the fault-free system and satisfies A₀−LC_p ∈ M^nx×nx; M^nx represents a set of n_x-dimensional Metzler matrix; |L| represents taking absolute values of all elements of the matrix L;

step 1.2.3: respectively setting e_pa=0.5(ē_p+e_p) and e_pd=ē_p−e_p; rewriting the formula (4) as:

{dot over (e)}_pd(t)=[A₀−LC_p]e_pd(t)+2ΔBΔu_cp(t)+ϕ_pd(t)+δ_pd(t)

{dot over (e)}_pa(t)=[A₀−LC_p]e_pa(t)+B₀Δu_cp(t)+LC_pe_p(t)+δ_pa(t)   (5)

where

ϕ_pd(t)=2ΔA(e_p⁺⁽t)+e_p⁻(t))

δ_pd(t)=2|L|V−d(t)+d(t)+2ΔA|x_pref(t)|  (6)

δ_pa(t)=−Lv(t)−0.5(d(t)+d(t))

step 1.2.4: representing the output of the error feedback controller as:

Δu_cp(t)=K_ae_pa(t)+K_de_pd(t)   (7)

representing the gain matrix of the error feedback controller as K_d, K_a ∈ R^nx×nx; setting e_x(t)=e_p(t)−e_pa(t), −0.5e_pd(t)≤e_x(t)≤0.5e_pd(t), and then

{dot over (e)}_pa(t)=[A₀+B₀K_a]e_pa(t)+B₀K_de_pd(t)+LC_pe_x(t)+δ_pa(t)   (8)

step 1.2.5: rewriting formulas (5) and (8) as:

$\begin{array}{cc}{\stackrel{.}{\xi }}_p\left(t\right)=G_p\left(t\right){\xi }_p\left(t\right)+{\delta }_p\left(t\right)& \left(9\right)\\ G_p\left(t\right)=\left[\begin{array}{cc}{A}_0-{\mathrm{LC}}_p& 0\\ B_0K_d& A_0+B_0K_a\end{array}\right]+A_{pd}\left(t\right)\mathrm{where}{\xi }_p\left(t\right)={\left[{e_{pd}\left(t\right)}^T,{e_{\mathrm{pa}}\left(t\right)}^T\right]}^T,{\delta }_p\left(t\right)={\left[{\left({\delta }_{pd}\left(t\right)+2\stackrel{\_}{\Delta B}\Delta u_{cp}\left(t\right)\right)}^T,{{\delta }_{\mathrm{pa}}\left(t\right)}^T\right]}^T\mathrm{and}\mathrm{then}& \left(10\right)\\ \left[\begin{array}{c}{\phi }_{pd}\\ LC_pe_x\end{array}\right]=A_{pd}\left[\begin{array}{c}{e}_{pd}\\ e_{\mathrm{pa}}\end{array}\right]& \left(11\right)\end{array}$

step 1.2.6: S^m×m representing an m-dimensional real symmetric square matrix; setting a matrix E,F ∈ S^2nx×2nx; E,F>0 representing that each element in E,F is greater than 0; constant λ>0; and obtaining a matrix inequality:

G _p ^T E+EG _p +λE+F<0   (12)

namely, setting each element in G_p^TE+EG_p+λE+F to be less than 0; solving the matrix inequality (12) to obtain the gain matrices K_d, K_a of the error feedback controller so as to obtain the error feedback controller from (7);

step 1.3: describing the aircraft engine LPV model having disturbance and actuator and sensor faults as:

{dot over (x)}_f(t)=[A₀+ΔA(θ)]x_f(t)+B_f(γ(t))u_f(t)+d _f(t)

y _f(t)=C_f(ϕ(t))x_f(t)+v(t)   (13)

where x_f ∈ R^nx is a state vector of a system with fault; u_f ∈ R^nu is the control input of the system with fault; y_f ∈ R^ny is an output vector of the system with fault; B_f(γ(t)) and C_f(ϕ(t)) are respectively actuator and sensor faults, expressed as

B _f(γ(t))=[B₀+ΔB(θ)]diag(γ₁(t), . . . , γ_n(t))

C _f(ϕ(t))=C_pdiag(ϕ₁(t), . . . , ϕ_n(t))   (14)

where 0≤γ_i(t)≤1 and 0≤ϕ_j(t)≤1 respectively represent the failure degree of the i th actuator and the jth sensor; γ_i=1 and γ₁=0 respectively represent health and complete failure of the i th actuator; ϕ_j is similar; diag(γ₁, γ₂, . . . , γ_n) represents a diagonal matrix with diagonal elements γ₁, γ₂, . . . , γ_n; diag(ϕ₁, ϕ₂, . . . , ϕ_n) is similar; setting γ(t) and ϕ(t) estimated values respectively as {circumflex over (γ)}(t) and {circumflex over (ϕ)}(t), and then

B _f(γ(t))=B_f({circumflex over (γ)}(t))+B_f(Δγ(Δγ(t))

C _f(ϕ(t))=C_f({circumflex over (ϕ)}(t))+C_f(Δϕ(t))   (15)

where Δγ(t)=γ(t)−{circumflex over (γ)}(t) and Δϕ(t)=ϕ(t)−{circumflex over (ϕ)}(t) are respectively errors of estimation of γ(t) and ϕ(t); a virtual actuator and a virtual sensor are respectively designed according to the actuator and sensor faults;

step 1.3.1: designing the virtual sensor as:

{dot over (x)}_vs(t)=A_vs(θ)x_vs(t)+B_f({circumflex over (γ)}(t))Δu(t)+Qy_f(t)

{circumflex over (y)}_f(t)=C_vsx_vs(t)+Py_f(t)   (16)

where

A _vs(θ)=A₀+ΔA(θ)−QC_f({circumflex over (ϕ)}(t))

C _vs =C _p −PC _f({circumflex over (ϕ)}(t))   (17)

where x_vs ∈ R^nx is a state variable of a virtual sensor system; Δu ∈ R^nu is a difference in control inputs of a fault model and a fault reference model; ŷ_f ∈ R^ny is an output vector of the virtual sensor system; Q and P are respectively parameter matrices of the virtual sensor;

step 1.3.2: an LMI region S₁(ρ₁, q₁, r₁, θ₁) representing an intersection of a left half complex plane region with a bound of −ρ₁, a circular region with a radius of r₁ and a circle center of q₁ and a fan region having an intersection angle θ₁ with a negative real axis; representing a state matrix A_vs of the virtual sensor as a polytope structure; A_vsj=A₀+ΔA(θ_j)−Q_jC_f({circumflex over (ϕ)}(t)), where θ_j represents the value of the jth vertex θ; A_vsj represents the value of the state matrix A_vs of the virtual sensor of the jth vertex; a necessary and sufficient condition for eigenvalues of A_vsj to be in S₁(ρ₁, q₁, r₁, θ₁) is that there exists a symmetrical matrix X₁>0 so that the linear matrix inequalities (18)-(20) are established, thereby obtaining a parameter matrix Q_j of the virtual sensor of the corresponding vertex;

$\begin{array}{cc}\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]X_1+{X_1\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]}^T+2{\rho }_1X_1<0& \left(18\right)\\ \left[\begin{array}{cc}-r_1X_1& q_1X_1+\left[\begin{array}{c}{A}_0+\Delta A\left({\theta }_j\right)-\\ Q_jC_f\left(\hat{\phi }\left(t\right)\right)\end{array}\right]X_1\\ q_1X_1+{X_1\left[\begin{array}{c}{A}_0+\Delta A\left({\theta }_j\right)-\\ Q_jC_f\left(\hat{\phi }\left(t\right)\right)\end{array}\right]}^T& -r_1X_1\end{array}\right]<0& \left(19\right)\\ \left(\begin{array}{c}\sin {\theta }_1\left\{\begin{array}{c}\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]X_1+\\ {X_1\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]}^T\end{array}\right\}\\ \cos {\theta }_1\left\{\begin{array}{c}{X_1\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]}^T-\\ \left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]X_1\end{array}\right\}\end{array}\begin{array}{c}\cos {\theta }_1\left\{\begin{array}{c}\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]X_1-\\ {X_1\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]}^T\end{array}\right\}\\ \sin {\theta }_1\left\{\begin{array}{c}\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]X_1+\\ {X_1\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]}^T\end{array}\right\}\end{array}\right)<0& \left(20\right)\end{array}$

selecting Q_j of a vertex corresponding to θ_j as a parameter matrix of the virtual sensor;

step 1.3.3: representing the parameter matrix P of the virtual sensor as:

P=C_pC_f^†  (21)

where † represents pseudo-inversion of the matrix;

step 1.3.4: designing the virtual actuator as

{dot over (x)}_va(t)=A_vax_va(t)+B_vaΔu_c(t)

Δu(t)=Mx_va(t)+NΔu_c(t)

y _c(t)={circumflex over (γ)}_f(t)+C_px_va(t)   (22)

where

A _va =A ₀ +ΔA(θ)−B_f({circumflex over (γ)}(t))M

B _va =B ₀ +ΔB(θ)−B_f({circumflex over (γ)}(t))N   (23)

where x_va ∈ R^nx is a state variable of the virtual actuator system; Δu_c ∈ R^nu is the output of the error feedback controller; y_c ∈ R^ny is an output vector of the virtual actuator system; M and N are respectively parameter matrices of the virtual actuator;

step 1.3.5: an LMI region S₂(ρ₂, q₂, r₂, θ₂) representing an intersection of a left half complex plane region with a bound of −ρ₂, a circular region with a radius of r₂ and a circle center of q₂ and a fan region having an intersection angle θ₂ with a negative real axis; representing a state matrix A_va of the virtual actuator as a polytope structure; A_vaj=A₀+ΔA(θ_j)−B_f({circumflex over (γ)}(t))M_j, where θ_j represents the value of the jth vertex θ; A_vaj represents the value of the state matrix A_va of the virtual actuator of the jth vertex; a necessary and sufficient condition for eigenvalues of A_vaj to be in S₂(ρ₂, q₂, r₂, θ₂) is that there exists a symmetrical matrix X₂>0 so that the linear matrix inequalities (24)-(26) are established, thereby obtaining a parameter matrix Mⁱ of the virtual actuator;

$\begin{array}{cc}\left[A_0+\Delta A\left({\theta }_j\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]X_2+{X_2\left[A_0+\Delta A\left({\theta }_j\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]}^T+2{\rho }_2X_2<0& \left(24\right)\\ \left[\begin{array}{cc}-r_2X_2& q_2X_2+\left[\begin{array}{c}{A}_0+\Delta A\left({\theta }_j\right)-\\ B_f\left(\hat{\gamma }\left(t\right)\right)M_i\end{array}\right]X_2\\ q_2X_2+{X_2\left[\begin{array}{c}{A}_0+\Delta A\left({\theta }_j\right)-\\ B_f\left(\hat{\gamma }\left(t\right)\right)M_i\end{array}\right]}^T& -r_2X_2\end{array}\right]<0& \left(25\right)\\ \left(\begin{array}{c}\sin {\theta }_2\left\{\begin{array}{c}\left[A_0+\Delta A\left({\theta }_j\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]X_2+\\ {X_2\left[A_0+\Delta A\left({\theta }_j\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]}^T\end{array}\right\}\\ \cos {\theta }_2\left\{\begin{array}{c}{X_2\left[A_0+\Delta A\left({\theta }_j\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]}^T-\\ \left[A_0+\Delta A\left({\theta }_j\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]X_2\end{array}\right\}\end{array}\begin{array}{c}\cos {\theta }_2\left\{\begin{array}{c}\left[A_0+\Delta A\left({\theta }_j\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]X_2-\\ {X_2\left[A_0+\Delta A\left({\theta }_j\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]}^T\end{array}\right\}\\ \sin {\theta }_2\left\{\begin{array}{c}\left[A_0+\Delta A\left({\theta }_j\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]X_2+\\ {X_2\left[A_0+\Delta A\left({\theta }_j\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]}^T\end{array}\right\}\end{array}\right)<0& \left(26\right)\end{array}$

selecting M_j of a vertex corresponding to θ_j as a parameter matrix of the virtual actuator;

step 1.3.6: representing the parameter matrix N of the virtual actuator as:

N=B_f^†B_p   (27)

where † represents pseudo-inversion of the matrix;

step 1.4: designing an interval error observer according to the aircraft engine LPV model having disturbance and actuator and sensor faults and the reference model of the system with fault;

step 1.4.1: representing the reference model of the aircraft engine system having disturbance and actuator and sensor faults as:

{dot over (x)}_ref(t)=A₀x_ref(t)+B_f({circumflex over (γ)}(t))u_ref(t)

y _ref(t)=C_f({circumflex over (ϕ)}(t))x_ref(t)   (28)

where X_ref ∈ R^nx is a reference state vector of the system having disturbance and actuator and sensor faults; u_ref ∈ R^nu is control input of the system having disturbance and actuator and sensor faults; y_ref ∈ R^ny is a reference output vector of the system having disturbance and actuator and sensor faults;

step 1.4.2: defining an error e(t)=x_ref(t)−x_f(t) between the aircraft engine LPV model having disturbance and actuator and sensor faults and the reference model of the aircraft engine to obtain error state equations of the system with fault of the aircraft engine based on the LPV model:

{dot over (e)}(t)=[A₀+ΔA(θ)]e(t)+B_f({circumflex over (γ)})Δu(t)−B_f(Δγ)u_f(t)−ΔA(θ)x_ref(t)−d_f(t)

ε_c(t)=c_f(ϕ(t))e(t)−C_f(Δϕ)x_ref(t)−v(t)   (29)

where Δu(t)=u_ref(t)−u_f(t) and ε_c(t)=y_ref(t)−y_f(t);

step 1.4.3: representing state equations of an upper bound e and a lower bound ē of the error e between the aircraft engine LPV model having disturbance and actuator and sensor faults and the reference model of the aircraft engine as:

{dot over (e)}(t)=[A₀−LC_f(ϕ(t))]e(t)+[B₀+ΔB]Δu_c(t)+L[ε_e(t)+C_px_vu+(C_p−PC_f(ϕ(t)))x_vs]+|L|V−d(t)+ΔA|x_ref(t)|+ΔB|u_ref|+ϕ(t)

{dot over (e)}(t)=[A₀−LC_f(ϕ(t))]e(t)+[B₀−ΔB]Δu_c(t)+L[ε_c(t)+C_px_va+(C_p−PC_f(ϕ(t)))x_vs]−|L|V−d(t)−ΔA|x_ref(t)|−ΔB|u_ref|−ϕ(t)   (30)

where ϕ(t)=ΔA(ē_v⁺(t)+e_v⁻(t)), e_v is a difference among the error state variable of the system with fault of the aircraft engine based on the LPV model, the state variable of the virtual actuator and the state variable of the virtual sensor; the upper bound of e_v is ē_v(t)=ē(t)−x_va(t)−x_vs(t); the lower bound of e_v is e_v(t)=e(t)−x_va(t)−x_vs(t); A₀−LC_f ∈ M^nxnx;

step 1.4.4: setting e_a=0.5(ē+e), e_d=ē−e, and obtaining the interval error observer from (30);

ė_d(t)=[A₀−LC_f(ϕ(t))]e_d(t)+2ΔBΔu_c(t)+ϕ_d(t)+ϕ_d(t)+δ_d(t)

ė_a(t)=[A₀−LC_f]e_a(t)+B₀K_ae_a(t)+B₀K_de_d(t)+δ_a(t)+LC_px_va+L(C_p−PC_f)+LC_fe(t)   (31)

where

ϕ_d(t)=2ΔA(e_v⁺(t)+e_v⁻(t))

δ_d(t)=2|L|V−d(t)+d(t)+2ΔA|x_ref(t)|+2ΔB|u_ref(t)|  (32)

δ_a(t)=−Lv(t)−0.5(d(t)+d(t))

step 1.5: using the aircraft engine state variable x_f(t) of the aircraft engine LPV model having disturbance and actuator and sensor faults, the output variable y_f(t), the reference model state variable x_ref(t) of the system with fault, the virtual actuator state variable x_va(t) and the virtual sensor state variable x_vs(t) as inputs of the interval error observer; using the interval error observer output e_a(t), e_d(t) as the input of the error feedback controller; using the error feedback controller output Δu_c(t) as the input of the virtual actuator; inputting the difference between the reference model output u_ref(t) of the system with fault and the virtual actuator output Δu(t) as a control signal into the system with fault of the aircraft engine, thereby realizing active fault tolerant control of the aircraft engine.

Compared with the existing technology, the interval error observer-based aircraft engine active fault tolerant control method designed by the present invention has the advantages:

(1) In the active fault tolerant control of the LPV system having actuator and sensor faults, the traditional gain self-scheduling H-infinite optimization method is often used. The method readjusts the controller parameters when the system has the fault, thereby increasing the complexity of the system design. The active fault tolerant control method proposed by the present invention can reconfigure the system which simultaneously has actuator and sensor faults without redesigning the controller.

(2) When the system has actuator faults and sensor faults, the method proposed by the present invention can enable the reconfigured system to have the same state and output as the original fault-free system.

(3) The method proposed by the present invention considers the problem that often appears in the noise signal interference of the control system in actual engineering, and improves the robustness of the control system.

DESCRIPTION OF DRAWINGS

FIG. 1 is an overall structural diagram of a system.

FIG. 2(a) and FIG. 2(b) are respectively contrasts of trajectories of H=0,Ma=0,n₂=94% aircraft engine LPV model states x_p1(t) and x_p2(t) and trajectories of fault-free reference model states x_pref,1(t) and x_pref,2(t).

FIG. 3 is a flow chart of an error feedback controller algorithm.

FIG. 4(a) and FIG. 4(b) are respectively the estimated curves of error states e_p1(t) and e_p2(t), upper bound states ē_p1(t) and ē_p2(t) and lower bound states e_p1(t) and e_p2(t) of H=0,Ma=0,n₂=94% aircraft engine fault-free system.

FIG. 5 is a varying curve of an actuator fault factor γ₁ and a sensor fault factor ϕ₁.

FIG. 6(a) and FIG. 6(b) are respectively the contrasts of trajectories of aircraft engine states x_f1(t) and x_f2(t) at H=0,Ma=0,n₂=94% under both disturbances and actuator and sensor faults, and trajectories of fault-free reference model states x_pref,1(t) and x_pref,2(t).

FIG. 7(a) and FIG. 7(b) are respectively the estimated curves of aircraft engine error states e_pf1(t) and e_pf2(t), upper bound states ē_p1(t) and ē_p2(t) and lower bound states e_p1(t) and e_p2(t) at H=0,Ma=0, n₂=94% under both disturbances and actuator and sensor faults.

FIG. 8 is a flow chart of a virtual sensor algorithm.

FIG. 9 is a flow chart of a virtual actuator algorithm.

FIG. 10 is a flow chart of an interval error observer algorithm.

FIG. 11(a) and FIG. 11(b) are respectively the contrasts of trajectories of aircraft engine states x₁(t) and x₂(t) at H=0,Ma=0,n₂=94% after active fault tolerant control and trajectories of fault reference model states x_ref,1(t) and x_ref,2(t).

FIG. 12(a) and FIG. 12(b) are respectively the estimated curves of aircraft engine error states e₁(t) and e₂(t), and upper bound states ē₁(t) and ē₂(t) and lower bound states e₁(t) and e₂(t) of an error observer at H=0,Ma=0,n₂=94% after active fault tolerant control.

DETAILED DESCRIPTION

The embodiments of the present invention will be further described in detail below in combination with the drawings and the technical solution.

The overall structure of the present invention is shown in FIG. 1, and comprises the following specific steps:

step 1.1: establishing an affine parameter-dependent aircraft engine LPV model; and taking relative conversion speed n₂ of a high pressure rotor of the aircraft engine as a variable parameter θ to normalizing the speed n₂=88%, 89%, . . . , 100% , i.e., θ ∈ [−1,1], to obtain a model:

$\begin{array}{cc}{\stackrel{.}{x}}_p=\left[A_0+\Delta A\left(\theta \right)\right]x_p\left(t\right)+\left[B_0+\Delta B\left(\theta \right)\right]u_p\left(t\right)+d_f\left(t\right)y_p=C_px_p\left(t\right)+v\left(t\right)\mathrm{where}& \left(33\right)\\ A_0=\left[\begin{array}{cc}-2.6748& -0.6877\\ 1.0704& -4.4672\end{array}\right],\Delta A\left(\theta \right)=\left[\begin{array}{cc}0.5199\theta & -2.4061\theta \\ 0.1049\theta & -0.8365\theta \end{array}\right]B_0=\left[\begin{array}{c}0.0033\\ 0.0012\end{array}\right],\Delta B\left(\theta \right)=\left[\begin{array}{c}-0.0004\theta \\ -0.0001\theta \end{array}\right]C_p=\left[\begin{array}{cc}0& 1\end{array}\right]& \left(34\right)\end{array}$

the state variable initial value is x_p(0)=[0, 0]^T; the upper bound and the lower bound of a disturbance variable d_f(t) are d, d ∈ R^nx;

$\overline{d}=-\underset{¯}{d}=\left[\begin{array}{c}0.001\\ 0.001\end{array}\right];$

and the sensor noise bound is V=0.01. ΔA(θ) and ΔB(θ) have established −ΔA≥ΔA(θ)≤ΔA and −ΔB≤ΔB(θ)≤ΔB.

$\begin{array}{cc}\stackrel{\_}{\Delta A}=\left[\begin{array}{cc}0.5199& 2.4061\\ 0.1049& 0.8365\end{array}\right],\stackrel{\_}{\Delta B}=\left[\begin{array}{c}0.0004\\ 0.0001\end{array}\right]& \left(35\right)\end{array}$

Step 1.2: representing the reference model of the fault-free system of the aircraft engine as

{dot over (x)}_pref(t)=A₀x_pref(t)+B₀u_pref(t)

y _pref(t)=C_px_pref(t)   (36)

where the state vector of the reference model is a constant value x_pref(t)=[4,2]^T. At H=0, Ma=0 and n₂=94%, contrasts of trajectories of aircraft engine LPV model states x_p1(t) and x_p2(t) and trajectories of fault-free reference model states x_pref,1(t) and x_pref,2(t) are shown in FIG. 2. An error feedback controller of a fault-free system of the aircraft engine is designed, and an algorithm flow of the error feedback controller is shown in FIG. 3.

Step 1.2.1: defining an error e_p(t)=x_pref−x_p between the affine parameter-dependent aircraft engine LPV model and the reference model of the fault-free system of the aircraft engine, with an initial value is e_p(0)=x_pref(0)−x_p(0)=[4, 2]^T.

Step 1.2.2: representing state equations of the upper bound ē_p and the lower bound e_p of the error vector e_p as:

{dot over (e)}_p(t)=[A₀−LC_p]e(t)+[B₀+ΔB]Δu_cp(t)+Lε_cp(t)+|L|V−d(t)+ΔA|x_pref(t)|+ϕ_p(t)

{dot over (e)}_p(t)=[A₀−LC_p]e(t)+[B₀−ΔB ]Δu_cp(t)+Lε_cp(t)−|L|V−d(t)−ΔA|x_pref(t)|−ϕ_p(t)   (37)

where e_p(0)=[−50, −50]^T, ē_p(0)=[50, 50]^T and ϕ_p(t)=ΔA(ē_p⁺(t)+e_p(t)). At H=0, Ma=0 and n₂=94%, the estimated curves of error states e_p1(t) and e_p2(t), upper bound ē_p1(t) and ē_p2(t) and lower bound e_p1(t) and e_p2(t) of aircraft engine fault-free system are shown in FIG. 4. From A₀−LC_p ∈ M^nx×nx, the error gain matrix of the fault-free system can be obtained

$\begin{array}{cc}L=\left[\begin{array}{c}-5\\ 20\end{array}\right]& \left(38\right)\end{array}$

Step 1.2.3: respectively setting e_pa=0.5(ē_p+e_p), e_pd=ē_p−e_p to obtain

{dot over (e)}_pd(t)=[A₀−LC_p]e_pd(t)+2ΔBΔu_cp(t)+ϕ_pd(t)+δ_pd(t)

{dot over (e)}_pa(t)=[A₀−LC_p]e_pa(t)+B₀K_ae_pa(t)+B₀K_de_pd(t)+LC_pe_p(t)+δ_pa(t)

ϕ_pd(t)=2ΔA(e_p⁺(t)+e_p⁻(t))   (39)

δ_pd(t)=2|L|V−d(t)+d(t)+2ΔA|x_pref(t)|

δ_pa(t)=−Lv(t)−0.5(d(t)+d(t))

where ē_p1 and e_p1 (ē_p2 and e_p2) respectively represent the first (second) element of ē_p and e_p, and e_x,1 and e_x,2 respectively represent the first element and the second element of e_x.

$\begin{array}{cc}2\left({\overline{e}}_{p1}^{+}+{\underset{\_}{e}}_{p1}^{-}\right)=\{\begin{array}{cc}2{\overline{e}}_{p1}=e_{\mathrm{pd},1}+2e_{\mathrm{pa},1}& {\overline{e}}_{p1}\ge 0,{\underset{\_}{e}}_{p1}\ge 0\\ 2\left({\overline{e}}_{p1}-{\underset{\_}{e}}_{p1}\right)=2e_{pd,1}& {\overline{e}}_{p1}\ge 0,{\underset{\_}{e}}_{p1}<0\\ -2{\underset{\_}{e}}_{p1}=e_{\mathrm{pd},1}-2e_{\mathrm{pa},1}& {\overline{e}}_{p1}<0,{\underset{\_}{e}}_{p1}<0\end{array}2\left({\overline{e}}_{p2}^{+}+{\underset{\_}{e}}_{p2}^{-}\right)=\{\begin{array}{cc}2{\overline{e}}_{\mathrm{p2}}=e_{\mathrm{pd},2}+2e_{\mathrm{pa},2}& {\overline{e}}_{p2}\ge 0,{\underset{\_}{e}}_{p2}\ge 0\\ 2\left({\overline{e}}_{p2}-{\underset{\_}{e}}_{p2}\right)=2e_{\mathrm{pd},2}& {\overline{e}}_{p2}\ge 0,{\underset{\_}{e}}_{p2}<0\\ -2{\underset{\_}{e}}_{p1}=e_{\mathrm{pd},2}-2e_{\mathrm{pa},2}& {\overline{e}}_{p2}<0,{\underset{\_}{e}}_{p2}<0\end{array}& \left(40\right)\end{array}$

Step 1.2.4: representing the output of the error feedback controller as

Δu_cp(t)=K_ae_pa(t)+K_de_pd(t)   (41)

representing the gain matrix of the error feedback controller as K_d, K_a ∈ R^nx×nx; setting e_x(t)=e_p(t)−e_pa(t), −0.5e_pd(t)≤e_x(t)≤0.5e_pd(t), and then

{dot over (e)}_pa(t)=[A₀+B₀K_a]e_pa(t)+B₀K_de_pd(t)+LC_pe_x(t)+δ_pa(t)   (42)

Step 1.2.5: rewriting (39) and (42) as

$\begin{array}{cc}{\stackrel{.}{\xi }}_p\left(t\right)=G_p\left(t\right){\xi }_p\left(t\right)+{\delta }_p\left(t\right)& \left(42\right)\\ G_p\left(t\right)=\left[\begin{array}{cc}{A}_0-{\mathrm{LC}}_p& 0\\ B_0K_d& A_0+B_0K_a\end{array}\right]+A_{pd}\left(t\right)\mathrm{where}{\xi }_p\left(t\right)={\left[{e_{pd}\left(t\right)}^T,{e_{\mathrm{pa}}\left(t\right)}^T\right]}^T{\delta }_p\left(t\right)={\left[{\left({\delta }_{pd}\left(t\right)+2\stackrel{\_}{\Delta B}\Delta u_{cp}\left(t\right)\right)}^T,{{\delta }_{\mathrm{pa}}\left(t\right)}^T\right]}^T,\mathrm{and}\mathrm{then}\left[\begin{array}{c}{\phi }_{pd}\\ LC_pe_x\end{array}\right]=A_{pd}\left[\begin{array}{c}{e}_{pd}\\ e_{\mathrm{pa}}\end{array}\right]2\stackrel{\_}{\Delta A}\left({\overline{e}}_p^{+}\left(t\right)+{\underset{\_}{e}}_p^{-}\left(t\right)\right)=A_{pd1}\left[\begin{array}{c}{e}_{pd}\\ e_{po}\end{array}\right],{\mathrm{LC}}_pe_x=A_{pd2}\left[\begin{array}{c}{e}_{pd}\\ e_{\mathrm{pa}}\end{array}\right]& \left(44\right)\\ A_{pd1}=\stackrel{\_}{\Delta A}\left[\begin{array}{cccc}{a}_{11}& 0& a_{13}& 0\\ 0& a_{22}& 0& a_{24}\end{array}\right],A_{pd2}=\left[\begin{array}{cccc}0& a_{31}& 0& 0\\ 0& a_{41}& 0& 0\end{array}\right]A_{pd}=\left[\begin{array}{c}{A}_{pd1}\\ A_{pd2}\end{array}\right]& \left(45\right)\end{array}$

All possible combining forms are considered: (a_ll, a₁₃) ∈ {(1, 2), (2, 0), (1, −2)}, (a₂₂, a₂₄) ∈ {(1,2),(2,0),(1,−2)} and (a₃₁, a₄₁) ∈ {(−2.5,10),(2.5,−10)}.

Step 1.2.6: S^m×m representing an m-dimensional real symmetric square matrix; setting a matrix E,F ∈ S^2nx×2nx; E,F>0 representing that each element in E,F is greater than 0; constant λ>0; and obtaining a matrix inequality:

G _p ^T E+EG _p +λE+F<0   (46)

namely, setting each element in G_p^TE+EG_p+λE+F to be less than 0; converting the matrix inequality (46) to a linear matrix inequality (LMI), and multiplying the left and right sides of the inequality (46) by E⁻¹ to obtain

$\begin{array}{cc}{E}^1G_p^T+G_pE^1+\lambda E^1+F_p\prec 0& \left(47\right)\\ G_p\left(t\right)=\left[\begin{array}{cc}{A}_0-LC_p& 0\\ 0& A_0\end{array}\right]+A_{pd}\left(t\right)+\left[\begin{array}{c}0\\ B_0\end{array}\right]KK=\left[\begin{array}{cc}{K}_d& K_a\end{array}\right]& \left(48\right)\end{array}$

introducing W=KE⁻¹, and then converting inequality (46) into the LMI; using an LMI tool kit to obtain

K _d=[−0.0014 −0.0002]

K _a=[−14.5130 −21.8837]

Step 1.3: describing the aircraft engine LPV model having disturbance and actuator and sensor faults as:

{dot over (x)}_f(t)=[A₀+ΔA(θ)]x_f(t)+B_f(γ(t)u_f(t)+d_f(t)

y _f(t)=C_f(ϕ(t))x_f(t)+v(t)

B _f(γ(t)=[B₀+ΔB(θ)]diag(γ₁(t), . . . , γ_n(t))

C _f(ϕ(t))=C_pdiag(ϕ₁(t), . . . , ϕ_n(t))   (50)

where state variable initial values x_f(0)=[0,0]^T, B_f(γ(t)) and C_f(ϕ(t)) are respectively actuator and sensor faults; and the actuator fault factor γ₁ and the sensor fault factor ϕ₁ decay from 1 to 0.2 in the 5th to 6th seconds, as shown in FIG. 5. At H=0, Ma=0 and n₂=94%, contrasts of trajectories of states x_f1(t) and x_f2(t) of the aircraft engine having disturbance and actuator and sensor faults and trajectories of fault-free reference model states x_pref,1(t) and x_pref,2(t) are shown in FIG. 6. Estimated curves of error states e_pf1(t) and e_pf2(t) and upper bounds ē_p1(t) and ē_p2(t) and lower bounds e_p1(t) and e_p2(t) of the aircraft engine having disturbance and actuator and sensor faults are shown in FIG. 7. A virtual sensor and a virtual actuator are respectively designed according to the actuator and sensor faults, and algorithm flows are respectively shown in FIG. 8 and FIG. 9.

Step 1.3.1: designing the virtual sensor as

{dot over (x)}_vs(t)=A_vs(θ)x_vs(t)+B_f({circumflex over (γ)}(t))Δu(t)+Qy_f(t)

{circumflex over (y)}_f(t)=C_vsx_vs(t)+Py_f(t)   (51)

where

A _vs(θ)=A₀+ΔA(θ)−QC_f({circumflex over (ϕ)}(t))

C _vs =C _p −PC _f({circumflex over (ϕ)}(t))   (52)

where x_vs ∈ R^nx is a state variable of a virtual sensor system; Δu ∈ R^nu is a difference in inputs of a fault model and a fault reference model; ŷ_f ∈ R^ny is an output vector of the virtual sensor system; Q and P are respectively parameter matrices of the virtual sensor.

Step 1.3.2: selecting an LMI region S₁(10, −4.5,15, π/6) and solving LMIs (53)-(55)

$\begin{array}{cc}\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]X_1+{X_1\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]}^T+2{\rho }_1X_1<0& \left(53\right)\\ \left[\begin{array}{cc}-r_1X_1& q_1X_1+\left[\begin{array}{c}{A}_0+\Delta A\left({\theta }_j\right)-\\ Q_jC_f\left(\hat{\phi }\left(t\right)\right)\end{array}\right]X_1\\ q_1X_1+{X_1\left[\begin{array}{c}{A}_0+\Delta A\left({\theta }_j\right)-\\ Q_jC_f\left(\hat{\phi }\left(t\right)\right)\end{array}\right]}^T& -r_1X_1\end{array}\right]<0& \left(54\right)\\ \left(\begin{array}{c}\sin {\theta }_1\left\{\begin{array}{c}\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]X_1+\\ {X_1\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]}^T\end{array}\right\}\\ \cos {\theta }_1\left\{\begin{array}{c}{X_1\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]}^T-\\ \left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]X_1\end{array}\right\}\end{array}\begin{array}{c}\cos {\theta }_1\left\{\begin{array}{c}\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]X_1-\\ {X_1\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]}^T\end{array}\right\}\\ \sin {\theta }_1\left\{\begin{array}{c}\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]X_1+\\ {X_1\left[A_0+\Delta A\left({\theta }_j\right)-Q_jC_f\left(\hat{\phi }\left(t\right)\right)\right]}^T\end{array}\right\}\end{array}\right)<0& \left(55\right)\end{array}$

obtaining a parameter matrix of a virtual sensor of a corresponding vertex

Q ₁=[−15.4224; 24.4935]

Q ₂=[8.5894; 33.1359]  (56)

Step 1.3.3: representing the parameter matrix P of the virtual sensor as

P=C_pC_f^†=5   (57)

where † represents pseudo-inversion of the matrix.

step 1.3.4: designing the virtual actuator as

{dot over (x)}_va(t)=A_vax_va(t)+B_vaΔu_c(t)

Δu(t)=Mx_va(t)+N Δu_c(t)   (58)

y _c(t)={circumflex over (y)}_f(t)+C_px_va(t)

where

A _va =A ₀ +ΔA(θ)−B_f({circumflex over (γ)}(t))M

B _va =B ₀ +ΔB(θ)−B_f({circumflex over (γ)}(t))N   (59)

where x_va ∈ R^nx is a state variable of the virtual actuator system; Δu_c ∈ R^nu is the output of the error feedback controller; y_c ∈ R^ny is an output vector of the virtual actuator system; M and N are respectively parameter matrices of the virtual actuator.

Step 1.3.5: selecting an LMI region S₂(1.5,−2,8, π/6) and solving LMIs (60)-(62)

$\begin{array}{cc}\left[A_0+\Delta A\left({\theta }_j\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]X_2+{X_2\left[A_0+\Delta A\left({\theta }_j\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]}^T+2{\rho }_2X_2<0& \left(60\right)\\ \left[\begin{array}{cc}-r_2X_2& q_2X_2+\left[\begin{array}{c}{A}_0+\Delta A\left({\theta }_j\right)-\\ B_f\left(\hat{\gamma }\left(t\right)\right)M_i\end{array}\right]X_2\\ q_2X_2+{X_2\left[\begin{array}{c}{A}_0+\Delta A\left({\theta }_j\right)-\\ B_f\left(\hat{\gamma }\left(t\right)\right)M_i\end{array}\right]}^T& -r_2X_2\end{array}\right]<0& \left(61\right)\\ \left(\begin{array}{c}\sin {\theta }_2\left\{\begin{array}{c}\left[A_0+\Delta A\left(\rho \left(t\right)\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]X_2+\\ {X_2\left[A_0+\Delta A\left(\rho \left(t\right)\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]}^T\end{array}\right\}\\ \cos {\theta }_2\left\{\begin{array}{c}{X_2\left[A_0+\Delta A\left(\rho \left(t\right)\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]}^T-\\ \left[A_0+\Delta A\left(\rho \left(t\right)\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]X_2\end{array}\right\}\end{array}\begin{array}{c}\cos {\theta }_2\left\{\begin{array}{c}\left[A_0+\Delta A\left(\rho \left(t\right)\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]X_2-\\ {X_2\left[A_0+\Delta A\left(\rho \left(t\right)\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]}^T\end{array}\right\}\\ \sin {\theta }_2\left\{\begin{array}{c}\left[A_0+\Delta A\left(\rho \left(t\right)\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]X_2+\\ {X_2\left[A_0+\Delta A\left(\rho \left(t\right)\right)-B_f\left(\hat{\gamma }\left(t\right)\right)M_i\right]}^T\end{array}\right\}\end{array}\right)<0& \left(62\right)\end{array}$

obtaining a parameter matrix of a virtual actuator of a corresponding vertex

M ₁=[3690.6 −4333.2]

M₂=[2170.5 2186.6]  (63)

Step 1.3.6: representing the matrix N of the virtual actuator as

N=B _f ^† B _p=5   (64)

Step 1.4: designing an interval error observer, wherein an algorithm flow of the interval error observer is shown in FIG. 10.

Step 1.4.1: representing the reference model of the aircraft engine system having disturbance and actuator and sensor faults as

{dot over (x)}_ref(t)=A₀x_f(t)+B_f({circumflex over (γ)}(t))u_ref(t)

y _ref(t)=C_f({circumflex over (ϕ)}(t))x_ref(t)   (65)

where the state vector of the reference model is a constant value x_ref(t)=[4,2]^T.

Step 1.4.2: defining an error e(t)=x_ref−x_f between the aircraft engine LPV model having disturbance and actuator and sensor faults and the reference model of the aircraft engine with an initial value of the error e(0)=x_ref(0)−x_f(0)=[4, 2]^T. At H=0, Ma=0 and n₂=94%, the contrasts of trajectories of aircraft engine states x₁(t) and x₂(t) and trajectories of fault reference model states x_ref1(t) and x_ref2 (t) after active fault tolerant control are shown in FIG. 11.

Step 1.4.3: representing state equations of an upper bound ē and a lower bound e of the error e between the aircraft engine LPV model having disturbance and actuator and sensor faults and the reference model of the aircraft engine as

{dot over (e)}(t)=[A₀−LC_f(ϕ(t))]e(t)+[B₀+ΔB]Δu_c(t)+L[ε_c(t)+C_px_va+(C_p−PC_f(ϕ(t)))x_vs]+|L|V−d(t)+ΔA|x_ref(t)|+ΔB|u_ref+ϕ(t)   (66)

{dot over (e)}(t)=[A₀−LC_f(ϕ(t))]e(t)+[B₀−ΔB]Δu_c(t)+L[ε_c(t)+C_px_va+(C_p−PC_f(ϕ(t)))x_vs]−|L|V−d(t)−ΔA|x_ref(t)|−ΔB|u_ref|−ϕ(t)

where ϕ(t)=ΔA(ē_v⁻(t)+e_v⁻(t)), ē_v(t)=ē(t)−x_va(t)−x_vs(t), e_v(0)=[−50, −50]^T and ē_v(0)=[50,50]^T. e_v(t)=e(t)−x_va(t)−x_vs(t). The gain matrix L of the observer satisfies A₀−LC_p ∈ M^nx×nx.

Step 1.4.4: setting e_a=0.5(ē+e), e_d=ē−e, and obtaining the interval error observer from (66)

{dot over (e)}_d(t)=[A₀−LC_f(ϕ(t))]e_d(t)+2ΔBΔu_c(t)+ϕ_d(t)+δ_d(t)

{dot over (e)}_a(t)=[A₀−LC_f]e_a(t)+B₀K_ae_a(t)+B₀K_de_d(t)+δ_a(t)+LC_px_va+L(C_p−PC_f)+LC_fe(t)   (67)

where

ϕ_d(t)=2ΔA(e_v⁺(t)+e_v⁻(t))

δ_d(t)=2|L|V−d(t)+d(t)+2ΔA|x_pref(t)|+2ΔB|u_pref(t)|  (68)

δ_a(t)=−Lv(t)−0.5(d(t)+d(t))

At H=0, Ma=0 and n₂=94%, estimated curves of aircraft engine error states e₁(t) and e₂(t), and upper bound states ē₁(t) and ē₂(t) and lower bound states e₁(t) and e₂(t) of an error observer after active fault tolerant control are shown in FIG. 12.

Step 1.5: showing the overall structure that realizes the active fault tolerant control of the aircraft engine in FIG. 1.

Simulation results show that when the actuator and the sensor of the aircraft engine fail, an overshooting process occurs in states and outputs after active fault tolerant control, but the actuator and the sensor quickly return to a normal state. This indicates that the interval error observer-based aircraft engine active fault tolerant control method can ensure that the reconfigured system has the same performance criteria as the original fault-free system.

Claims

1. An interval error observer-based aircraft engine active fault tolerant control method, comprising the following steps: step 1.1: establishing an affine parameter-dependent aircraft engine linear-parameter-varying LPV model {dot over (x)}(t)=[A0+ΔA(θ)]xp(t)+[B0+ΔB(θ)]up(t)+df(t)yp(t)=Cpxp(t)+v(t)   (1)where Rm and Rm×n respectively represent a m-dimensional real number column vector and a m-row n-column real matrix; state vectors xp=[Ynl Ynh]T ∈ Rnx, Ynl and Ynh respectively represent variation of relative conversion speed of low pressure and high pressure rotors; nx represents the dimension of a state variable x; ny represents the dimension of an output vector y; nu represents the dimension of control input up; control input up=Upf ∈ Rnu is a fuel pressure step signal; output vectors yp=Ynh ∈ Rny, A0 ∈ Rnx×nx, B0 ∈ Rnx×nx and Cp ∈ Rny×nx are known system constant matrices; df(t) is a disturbance variable; the relative conversion speed nh of the high pressure rotor of the aircraft engine is a scheduling parameter θ∈ Rp; system variable matrices ΔA(θ) and ΔB(θ) satisfy −ΔA≤ΔA(θ)≤ΔA and −ΔB≤ΔB(θ)≤ΔB; ΔA ∈ Rnx×nx is an upper bound of ΔA(θ); ΔB ∈ Rnx×nu is an upper bound of ΔB(θ); ΔA≥0, ΔB≥0; a state variable initial value xp(0) satisfies x0≤xp(0)≤x0; x0, x0 ∈ Rnx are respectively known upper bound and lower bound of the state variable initial value xp(0); d,d ∈ Rnx are known upper bound and lower bound of an unknown disturbance df(t); sensor noise v(t) satisfies |v(t)|<V; V is a known bound; V>0;step 1.2: representing the reference model of the fault-free system of the aircraft engine as {dot over (x)}pref(t)=A0xpref(t)+B0upref(t)ypref(t)=Cpxpref(t)   (2)where xpref ∈ Rnx is a reference state vector of the fault-free system; upref ∈ Rnu is control input of the fault-free system; ypref ∈ Rny is a reference output vector; an error feedback controller of the fault-free system of the aircraft engine is designed according to the aircraft engine LPV model established in the step 1.1;step 1.2.1: defining an error ep(t)=xpref(t)−xp(t) between the affine parameter-dependent aircraft engine LPV model and the reference model of the fault-free system of the aircraft engine to obtain error state equations of the fault-free system: {dot over (e)}p(t)=[A0+ΔA(θ)]ep(t)+[B0+ΔB(θ)]Δucp(t)−ΔA(θ)xpref(t)−ΔB(θ)upref(t)−df(t)εcp(t)=Cpep(t)−v(t)where Δucp(t)=upref(t)−up(t) and εcp(t)=ypref(t)−yp(t);step 1.2.2: representing state equations of the upper bound ēp and the lower bound ep of the error vector ep as: {dot over (e)}p(t)=[A0−LCp]ēp(t)+[B0+ΔB]Δucp(t)+Lεcp(t)+|L|V−d(t)+ΔA|xpref(t)|+ϕp(t){dot over (e)}p(t)=[A0−LCp]ep(t)+[B0−ΔB]Δucp(t)+Lεcp(t)−|L|V−d(t)−ΔA|xpref(t)|−ϕp(t)   (4)where ēp, ep ∈ Rnx are respectively the upper bound and the lower bound of the error vector eP, i.e., ep(t)≤ep(t)≤ēp(t); ϕp(t)=ΔA(ēp+(t)+ep−(t)), ēp+=max {0,ēp}, ēp−=ēp+−ēp, ep+=max{0,ep}, ep−=ep+−ep; L ∈ Rnx×ny is an error gain matrix of the fault-free system and satisfies A0−LCp ∈ Mnx×nx; Mnx represents a set of nx-dimensional Metzler matrix; |L| represents taking absolute values of all elements of the matrix L;step 1.2.3: respectively setting epa=0.5(ēp+ep) and epd=ēp−ep; rewriting the formula (4) as: {dot over (e)}pd(t)=[A0−LCp]epd(t)+2ΔBΔucp(t)+ϕpd(t)+δpd(t){dot over (e)}pa(t)=[A0−LCp]epa(t)+B0Δucp(t)+LCpep(t)+δpa(t)   (5)whereϕpd(t)=2ΔA(ep+(t)+ep−(t))δpd(t)=2|L|V−d(t)+d(t)+2ΔA|xpref(t)|  (6)δpa(t)=−Lv(t)−0.5(d(t)+d(t))step 1.2.4: representing the output of the error feedback controller as: Δucp(t)=Kaepa(t)+Kdepd (t)   (7)representing the gain matrix of the error feedback controller as Kd, Ka ∈ Rnx×nx; setting ex(t)=ep(t)−epa(t), −0.5epd(t)≤ex(t)≤0.5epd(t), and then {dot over (e)}pa(t)=[A0+B0Ka]epa(t)+B0Kdepd(t)+LCpex(t)+δpa(t)   (8)step 1.2.5: rewriting formulas (5) and (8) as: