Escolar Documentos
Profissional Documentos
Cultura Documentos
Pr
ep
1. INTRODUCTION
rin
Abstract: This paper proposes a fault tolerant control method for input-affine nonlinear
systems using a nonlinear reconfiguration block (RB). The basic idea of the method is to insert
the RB between the plant and the nominal controller such that fault tolerance is achieved
without re-designing the nominal controller. The role of the RB is twofold: on one hand it
transforms the output of the faulty system such that its behaviour is similar to that of the
nominal one from the controllers viewpoint; on the other hand it modifies the control input to
the faulty system such that the stability of the reconfigured loop is preserved. The RB is realized
by a virtual actuator and a reference model. Using notions of incremental and input-to-state
stability (ISS), it is shown that ISS of the closed-loop reconfigured system can be achieved by
the separate design of the virtual actuator. The proposed method does not need any knowledge
of the nominal controller and only assumes that the nominal closed-loop system is ISS. The
method is demonstrated on a dynamic positioning system for an offshore supply vessel, where
the virtual actuator is designed using backstepping.
Nominal
Controller
Controller
5: General
Figure 4: General structure ofFigure
an AFDI
module structure of an AFDI module
Nominal
Plant
zf
Plant
Faulty
Faulty
zf
Plant
uf
yf
yf
Reconfiguration
Block
uf
uc
Nominal
Nominal
Controller
(a)
yc
uc
Nominal
Controller
(b)
Controller
(c)
5: General
an AFDI
module structure of an AFDI module
Figure 4: General structure of Figure
an AFDI
module structure ofFigure
6: General
Fig. 1. Fault-tolerant control using a reconfiguration block: (a) nominal loop, (b) faulty plant with nominal controller,
(c) reconfigured plant with nominal controller
Pr
ep
rin
2. PRELIMINARIES
sumed that the
the faulty system is measurable.
Faulty
d state ofFaulty
d
2
AFTC for a system with additive Lipschitz
nonlinearity
zf
z
Plant
subject to actuator faults
using a virtual actuator wasPlantThe field of freal numbers and the set of nonnegative reals
yfBarzegary (2013). Pedersen
uf
presented in Khosrowjerdi and
are respectively
denoted by R, and R0 . For any vector
yf
et al. (2014) proposed a new design method for the virtual x Rn , xT stands for its transpose and kxk = xT x
Reconfiguration
actuator based on absolute stability theory, which
was denotes its Euclidean norm. Also, the i-th entry of x is
tested for the reconfiguration of power systems subject to
Blockdenoted by xi . The infinity norm of x denoted by kxk is
faults in local controllers in emergency situations.
givenycby maxi |xi |. Given a measurable function u : R0
uf of polytopic linear parameterucvary- Rn , its
(essential) supremum is denoted by kuk which is
Fault tolerant control
defined
as: kuk := (ess)sup{ku(t)k, t 0}. The function
ing (LPV) systems subject
to
sensor
faults
using
virtual
Nominal
Nominal
u is essentially bounded if kuk < .
sensor was proposed
in
de
Oca
and
Puig
(2010),
where
r
r
Controller
Controller
the structure of the nominal
controller was assumed to The function : R R is called a class K function
0
be known. It was further assumed that the nominal con- denoted by K 0
if it is continuous, strictly increasing,
troller
of a structure
state feedback
combined
withstructure
an unbounded
6: General
of an AFDI
Figureconsists
5: General
of Figure
an AFDI
module
and module
satisfies (0) = 0. The function : R0
LPV observer. Tabatabaeipour et al. (2012) considered R
0 R0 is called a class KL function denoted by
the problem of control reconfiguration for continuous- KL if (, t) K and (r, t) 0 as t .
time LPV systems with both sensor and actuator faults
and without any assumptions about the structure of the Consider the following nonlinear system
Faulty
d
nominal controller.
In this
context input-to-state stability 2
x = f (x(t), u(t)), x(0) = x0 ,
z
properties of the reconfigured
investigated.
:
(1)
f
Plant system were
y(t) = h(x(t)),
In Tabatabaeipouru et al. (2014) the control reconfiguraf
yf with both sensor and where x(t) Rn is the state, u(t) Rm is the input,
tion for discrete-time
LPV systems
actuator faults were
considered
and both stability and y(t) Rq is the output. We use the following stability
Reconfiguration
performance of the reconfiguration block was investigated. definitions.
Block
In this paper we extend the idea of reconfigurable con- Definition 1. 0-global asymptotic stability (Sontag
yc block to input-affine (2008)) The system (1) with u(t) = 0, t R0 is called
trol design using u
a c reconfiguration
nonlinear dynamical systems. Only actuator faults are 0-globally asymptotically stable (0-GAS) if there exists a
Nominal
considered, and the reconfiguration
block is realized by function KL such that for all t0 and x(t0 ), the solution
a nonlinear virtual
actuator.
Using incremental stability of the system satisfies
r
Controller
properties, it is shown how to design the nonlinear virtual
kx(t)k (kx(t0 )k, t).
(2)
actuator independent of the nominal controller to achieve
Figure
General structure
of an
AFDIThe
module
ISS
of the6:reconfigured
closed-loop
system.
main con- Definition 2. Input-to-state stability (Sontag (2008))
tributions are given in Theorems 12, 13 and Corollary 14. The system (1) is called input-to-state stable (ISS) with
The proposed method does not require any information respect to (w.r.t.) the input u(t) if there exist some KL
about the nominal controller and only assumes that the and some K such that for all t0 and x(t0 ) and all inputs
nominal closed-loop system is ISS. The design of a fault- u(t), all solutions of the system satisfy
2
tolerant dynamic positioning system for an offshore supply
kx(t)k (kx(t0 )k, t) + (ku(t)k ).
(3)
vessel is utilised as case study. The design of the nonlinear Definition 3. Input-to-output stability (Sontag (2008))
virtual actuator is demonstrated using backstepping con- The system (1) is called input-to-ouput stable (IOS) w.r.t.
trol. The simulation results show the effectiveness of the the input u(t) and the output y(t) if there exist some
proposed method.
KL and some K such that for all t and x(t ) and
0
(4)
z = Cz x
where x Rn is the state, uc U Rm is the control
input, d D Rk is the input disturbance, y Rq is the
measured output, and z Rp is the controlled output.
rin
Pr
ep
(7)
(8)
x f = f (xf ) + Bf uf + Bd d
Pf : yf = xf
(15)
zf = Cz xf
where xf Rn is the state of the faulty system, uf U
Rm is the faulty control input, yf Rq and zf Rp are
uf
x r = f (xr , uc , yf ),
R : uf = hr (xr , uc ),
(16)
yc = hry (z, yf ).
+
()
P : Reference Model
uc
C : Controller
uc
rin
()
xf
x f = f (xf ) + Bf uf + Bd d
Pr
ep
x f
x = x
= f (
x) + Buc [f (xf ) + Bf uf + Bd d]
= f (
x) Bf (
x) [f (xf ) Bf (xf )]
+ (B Bf N)uc Bd d
= f (
x) Bf (
x) [f (
x x ) Bf (
x x )]
+ (B Bf N)uc Bd d
= (
x) (
x x ) + (B Bf N)uc Bd d (18)
= f (
(t0 ) = x0
x) + Buc , x
x
A :
uf = (
x) (xf ) + Nuc
yc = xc
(17)
where () , f () Bf ().
In the following, we show the conditions for ISS of the
difference system and we show that if the virtual actuator
is designed independently such that the difference system
is ISS, then the reconfigured closed-loop system is also ISS.
Theorem 12. (Reconfigured system stability) Consider the reconfigured closed-loop system (Pf , A , C ). If
the nominal closed-loop system L is ISS and the virtual
actuator A is designed such that the difference system
is ISS, then the reconfigured closed-loop system is ISS.
Proof. Introducing the new variable x the dynamics of
the closed-loop reconfigured system (Pf , A , C ) in new
variables is re-written by:
= f (
x
x) + Buc
P :
yc = xc
x c = fc (xc , yc )
C :
(19)
uc = (xc , yc )
x = (
x) (
x x )
:
+ (B Bf N)uc Bd d
which is graphically depicted in Figure 3.
By Assumption 10 the nominal closed-loop system L is
ISS, hence also the closed-loop system (P , C ) is ISS
xT Px + kx kkP(B Bf N)kkuc k
+ kx kkPBd kkdk
bkx k2 + kx kkP(B Bf N)kkuc k
+ kx kkPBd kkdk
(1 )bkx k2
(27)
1
for kx k b
(kP(B Bf N)kkuc k + kPBd kkdk), where
b = min (P), which proves ISS of the system w.r.t. d.
rin
Pr
ep
x
(
x x )
xT
x) (
x x )) xT
(26)
P ( (
Px
= R() , z1 + d
u , B1 RT ()D ()( z )
1
d
1,f
(z) :
d ))
+
M
()(
(R()
Kp z1 Kd z2
(38)
T T
where z = [zT
1 , z2 ] is a dummy vector representing either
the state of the faulty system or the state of the reference
model. The virtual actuator exploits the knowledge of the
magnitude of the fault through the input matrix B1,f .
rin
= R() d
(31)
M ()s = D ()(s + d
) + R()(B1 uc + w )
M ()(
d (R() d ))
(32)
The control law (34)-(35) has been selected because backstepping controllers are known to guarantee ISS with respect to input disturbances (Krstic et al., 1995).
where
A() =
Pr
ep
Let x , [
T , sT ]T be the state vector, uc the control input,
and d = w the wind generated disturbance. Then the
fault-free plant P is given by
z = x1
where
R() d
f (x) =
d
M1
) (
d
)
()D ()(s +
0
0
B() =
, Bd () =
M1
M1
()R()B1
()R()
Let = [
, s]T be the state of the closed-loop system, then
its dynamics reads
= A() + Bd ()d
(39)
The DP backstepping nominal control law is given by (Fossen and Strand, 1999)
= R() ,
+ d
(34)
1 T
Kd s
uc , B1 R () D ()( d
) Kp
+ M ()(
d (R() d ))
(35)
dynamwhere is the virtual control that stabilizes the
ics in the first step; Kp > 0, and Kd > 0 are diagonal
design matrices. The nominal closed-loop dynamics reads
=
+s
(36)
1
s = M1
()K
M
()(D
()
+
K
)s
(37)
p
I
1
M1
()Kp M ()(D () + Kd )
(40)
+ ( 1 2 )T P()Bd ()(d1 d2 )
= ( 1 2 )T Q()( 1 2 )
+ ( 1 2 )T P()Bd ()(d1 d2 )
V ( 1 , 2 ) + (kd1 d2 k)
(41)
where
Q() =
Kp
0
= QT () > 0 .
0 D () + Kd
LOA
B
[xG , yG , zG ]T
[l1 , l2 , l3 ]T
[1 , 2 ]T
M
76.2 [m]
18 [m]
[42, 0, 0]T [m]
[27.4, 17.2, 10.5]T [m]
[30, 30]T [deg]
diag
{5.3e6,
8.3e6, 3.7e9}
"
#
5.0e4
0
0
0
2.7e5 4.4e6
0 4.4e6 4.2e8
diag {1.4e5, 1.4e5, 1.4e5}
diag {2.5, 2.5, 2.5}
diag {2.5, 2.5, 2.5}
diag {1, 1, 1}
diag {25, 25, 25} [sec]
Damping matrix
K
Kp
Kd
Ta
Time constants
1
0
0
50
100
150
200
250
300
350
400
450
500
50
100
150
200
250
300
350
400
450
500
50
100
150
200
250
Time [sec]
300
350
400
450
500
x 10
2
4
6
0
5
x 10
5
10
15
0
1.08
1.1
North [m]
0.5
Time [sec]
0.5
400
300
200
1.12
0.494
0.493
0.492
rin
100
2
1
North [m]
46.5
N [m]
E [m]
[deg]
td = 55 sec
1.5
0
3
x 10
500
Xe [N]
Value
Ye [N]
Length overall
Beam
Centre of gravity
Moment arms
Angle of rotations
Mass-inertia matrix
Symbol
Ne [Nm]
Quantity
2.5
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1
East [m]
0
0.1
tf = 50 sec
45
0
200
400
Time [sec]
[-]
46
45.5
0.5
Pr
ep
East[m]
[deg]
600
1
0
100
200
300
Time [sec]
400
500
50
100
150
Time [sec]
200
250
300
7. CONCLUSIONS
In this paper a new method for FTC of nonlinear systems
subject to actuator faults using a nonlinear reconfiguration
block was proposed. The main idea of the method is to
achieve fault-tolerance without re-designing the nominal
controller by inserting the reconfiguration block between
the faulty system and the nominal controller. The proposed method does not need any knowledge of the nominal
controller and it is only assumed that the nominal closedloop system is input-to-state stable. It was shown that if
the virtual actuator is designed separately such that the
difference system is -ISS, then the reconfigured closedloop system is ISS. The effectiveness of the method is
shown on a case study of dynamic positioning system of
an offshore supply vessel, where the virtual actuator is
designed using the backstepping control technique.
REFERENCES
Pr
ep
rin
All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately.