Denote
u
F
j
(t)=
u
F
1j
(t), u
F
2j
(t), , u
F
mj
(t)
T
= r
j
u(t)+s
j
u
s
(t)
where
r
j
= diagfr
j
1
, r
j
2
, , r
j
m
g, r
j
i
2½r
j
i
, r
j
i
s
j
= diagfs
j
1
, s
j
2
, , s
j
m
g, i 2 I(1, m), j 2 I(1, L)
u
s
(t)= u
s1
(t), u
s2
(t), , u
sm
(t)½
T
Then, the sets with above structure are defined by
D
r
j
= fr
j
jr
j
= diagfr
j
1
, r
j
2
, , r
j
m
g, r
j
i
2½r
j
i
, r
j
i
g
D
0
r
j
= fr
j
jr
j
= diagfr
j
1
, r
j
2
, , r
j
m
g, r
j
i
= r
j
i
,orr
j
i
= r
j
i
g
D
s
j
= fs
j
js
j
= diagfs
j
1
, s
j
2
, , s
j
m
g, s
j
i
=0 or s
j
i
=1g
ð4Þ
To have an uniform actuator fault model, the follow-
ing model is shown for all possible faulty modes L
u
F
(t)=ru(t)+su
s
(t)
ð5Þ
where r 2 D
r
j
and s 2 D
s
j
, j 2 I(1, L).
Remark 2. A more general model (5) is given in this arti-
cle than the ones in some studies
2,20,25–31
since most of
the aforementioned SMC strategies specifically suit just
one or two of the above fault models, which may be vio-
lated in practical situations.
Therefore, one can describe the dynamics of system
(1) with actuator faults (5) as
_
x(t)=(A + DA(x, t))x(t)+B
2
(ru (t)+su
s
(t)) + B
1
w(t),
z(t)=Cx(t)
ð6Þ
One standard assumption is that all the states of sys-
tem are available at every instant to ensure the achieve-
ment of the fault-tolerant objective. Besides, the
following assumptions are also made in the FTC design.
Assumption 1. The unparametrizable actuator stuck
fault is a piece-wise continuous bounded function, that
is, there exist an unknown positive constant
u
s
such
that k u
s
(t) k 4
u
s
.
Assumption 2. The rank(B
2
r) = rank(B
2
)=l for all
r 2 D
r
j
, j 2 I(1, L).
Assumption 3. Up to m l (14l4m 1) actuators
undergo stuck or outage fault, the rest of actuators can
still achieve a desired control objective. All actuators
are allowed to go through loss-of-effectiveness failures
simultaneously.
Remark 3. Assumption 1 is common and quite natural in
the robust FTC literature.
8
As discussed in Tang et al.,
33
Assumption 2 is necessary for compensating the stuck
faults or outage completely,
3,33
which introduces a condi-
tion of actuator redundancy in the system. Assumption 3
is a basic assumption which ensures the existence of a
feasible solution to the actuator failure accommodation
problem when the plant parameters and actuator fault
pattern are known.
3
Definition 1.
18
Let g
0
. 0 be a given positive number, for
any t . 0, then the following closed-loop system
_
j (t)=A
c
j(t)+B
c
w(t),
z(t)=C
c
j (t), j (0) = 0
ð7Þ
is said to be with an adaptive H
‘
performance index no
larger than g
0
, if the following inequality holds
Ð
‘
0
z
T
(t)z(t)dt4g
2
0
Ð
‘
0
w
T
(t)w(t)dt + t
ð8Þ
The aim of this article is to find a suitable sliding
manifold and design an adaptive sliding mode FTC law
such that the system (6) can be guaranteed to be quad-
ratically stable even in the presence of actuator failures,
unmatched uncertainties, and disturbance effects.
To facilitate control system design, the following
lemmas are presented and will be used in the later
developments.
Lemma 1.
34
Let T(s)=C
c
(sI A
c
)
1
B
c
, the closed-loop
system (7) is stable and T(s) satisfies k T(s) k \ g
0
if
there exist a positive definite matrix P = P
T
. 0 and a
positive scalar g
0
such that
PA
c
+ A
T
c
PPB
c
C
T
c
g
2
0
I 0
I
2
4
3
5
\ 0
Lemma 2 (Projection Lemma).
35
Given a symmetric matrix
F 2 R
n 3 n
and two matrices T
1
and T
2
of column dimen-
sion n, there exists a matrix X that satisfies
F + T
T
1
X
T
T
2
+ T
T
2
XT
1
\ 0
if and only if the following projection inequalities with
respect to X are satisfied
N
T
T
1
FN
T
1
\ 0, N
T
T
2
FN
T
2
\ 0
where N
T
1
and N
T
2
denote arbitrary bases of the null
spaces of T
1
and T
2
, respectively.
Table 1. Fault model.
Fault model
r
j
i
r
j
i
s
j
i
Normal 1 1 0
Outage 0 0 0
Loss-of-effectiveness .0 \10
Stuck 0 0 1
694 Proc IMechE Part I: J Systems and Control Engineering 227(9)