Proceedings of the 7th

World Congress on Intelligent Control and Automation

June 25 - 27, 2008, Chongqing, China

A Fault Detection and Isolation Scheme Based on Parity

Space Method for Discrete Time-Delay System*
Hongyu Wang, Zuohua Tian, Songjiao Shi and Zhenxin Weng
Department of Automation Shanghai Jiaotong University
Shanghai Jiaotong University
No.800 Dongchuan RD Shanghai, China
redrain@sjtu.edu.cn

Abstract - A Fault detection and isolation (FDI) scheme for
discrete time-delay system is proposed, which can not only detect
but also isolate the faults. A time delay operator ∇ is introduced
to deal with time-delay items. Residuals in the form of parity
space can be deduced from the recursion of the system equations.
Further more, a generalized residual set is created using the
freedom of the parity space redundancy. Thus, both fault
detection and fault isolation have been accomplished. The
proposed method has been verified by a numerical example.
Index Terms - fault detection and isolation; parity space; time-
delay system
I. INTRODUCTION
In recent years, fault detection and isolation (FDI)
results in Ref. [9] are extended. Both fault detection and fault
isolation method are proposed. The occurrence of the fault can
be detected timely and the position of the fault can be located
exactly. A numerical example is given to illustrate the design
method at the end.
II. MATHEMATICAL PRELIMINARIES
A time delay operator ∇ is defined according to Ref. [9].
∇f (k ) = f (k − 1) for any discrete-time function f . It is
easy to understand that
∇∇f (k ) = ∇ 2 f (k ) = f (k − 2) , ∇ w f (k ) = f (k − w) .
Consider a linear discrete time-delay system described by
­ v

problem in dynamitic system has been paid more and more ° x ( k + 1) = ¦ Ai x (k − i)

attention. A great number of methods for FDI have been ° i =0

proposed[1-5]. All of the FDI schemes are concerned with a core ° s

stage: the generation of the residual signals. The difference ® + ¦ B j [u(k − j ) + f (k − j )],
between the measurement of the system and its estimation is ° j = 0
called residual, whose values are zero or near to zero when no ° y (k ) = Cx (k ) (1)
fault occurs while differ distinctly from zero otherwise. °
Appropriate decisions such as the occurrence, magnification, ¯
type, location, etc. of the faults are called fault isolation,
where x ( k ) ∈ R is the state vectors, u( k ) ∈ R is
n p
which are achieved by residual evaluation.
the control vectors, y ( k ) ∈ R is the output
m
In the field of analytical model-based FDI techniques, the
analytical redundancy relations of the system are used to
vector, f ( k ) = [ f1 ( k ), " , f p ( k )] stands for the actuator
T
create residual signal. The approaches can be roughly
classified into observer-based approaches and parameter faults. fi ( k ), i = 1," , p is corresponding to the ith actuator
estimation approaches. Parity space approaches have been
proved to be structurally equivalent to the observer-based fault. Ai (i = 0," , v) , B j ( j = 0, " , s ) and C are constant
though the design procedures differ[6]. However, the parity
matrices with appropriate dimensions. Integers
space methodology using the temporal redundancy has its
advantages, especially in the discrete system. This method was v ≥ 0 and s ≥ 0 denote the number of time delays in the state
firstly generalized by the Ref. [5]. and control vectors respectively.
Time delays are inherent in many real physical processes Using the operator ∇ , the system (1) can be developed as
(i.e. mechanical and chemical processes, long transmission ­ x (k + 1) = A(∇) x (k ) + B (∇)[u(k ) + f (k )]
lines in pneumatic systems, power and water distribution ® , (2)
networks, air pollution systems etc.) Over the past two ¯ y (k ) = Cx (k )
decades, analysis and synthesis of dynamic time-delay where
systems have attracted a great deal of interests[7,8]. However,
there are relative fewer research results on FDI of time-delay ­° A(∇) = A0 + A1∇ + " + Av ∇v
systems[9,10]. ® . (3)
°̄ B (∇) = B0 + B1∇ + " + Bs ∇
s
This paper proposes a method to deal with the FDI
problem for the linear discrete-time systems with delays. The

*
This work is supported by National Natural Science Foundation of China (No. 60574081).

978-1-4244-2114-5/08/$25.00 © 2008 IEEE. 6411

 III. PARITY SPACE RESIDUAL GENERATION FOR FAULT Therefore, the residual r is independent of the system
DETECTION operating state x ( k − L) . Now, it is obviously that r is zero
The task of FDI is to design a residual signal which is
zero in a fault free case and non-zero when a fault occurs in when the system (1) works normally ( f ( k ) = 0 ) and may
the monitored system. Time delay implies that the state of the differ from zero when fault occurs ( f ( k ) ≠ 0 ). 
system for the next time step is not only determined by the The successful detection of a fault is followed by the
current state and the current input signals but also concerned fault isolation procedure which will distinguish (isolate) a
with the state and input signals of the former L step time particular fault from others. While a single residual signal is
intervals. The recursion of (2) from time instant k − L to sufficient to detect faults, a set of residuals (or a vector of
time instant k yields residual) is usually required for fault isolation. According to
Ref. [2], a commonly used scheme in designing the residual
§ y ( k − L) · § C · set is to make each residual sensitive to all but one fault, i.e.
¨ (k − L + 1) ¸ ¨ (∇) ¸
¨y ¸=¨ C ¸ x (k − L) ­ r1 = R( f 2 ," , f3 )
¨ # ¸ ¨ # ¸ °
¨ ¸ ¨ L ¸ ° #
y (k ) ¹ © CA (∇) ¹ °
© ® ri = R( f1 ," , fi −1 , fi +1 , " , f p ) (8)
(4)
ª § u( k − L ) · § f ( k − L ) · º °
«¨ » ° #
u(k − L + 1) ¸¸ ¨¨ f (k − L + 1) ¸¸ » °¯rp = R( f1 ," , f p −1 )
+ H Ǭ + ,
«¨ # ¸ ¨ # ¸» Where R (⋅) denotes some functional relation, which works as
«¨ ¸ ¨ ¸»
¬«© u(k ) ¹ © f (k ) ¹ ¼» the residual generator. This is defined as a generalized
structured residual set. The isolation can be performed by the
where
following logic:
§ 0 · ri ≤ thi ½°
¨ CB (∇) ¸
¨ 0 ¸ ¾ Ÿ f i ≠ 0 (9)
rj > th j for j = 1, " , i − 1, i + 1, " p °¿
H = ¨ CA(∇) B (∇) CB (∇) 0 ¸
¨ ¸ i
where th means the fault isolation threshold to the
¨ # # # ¸ corresponding fault.
¨ CA (∇) B (Δ)
L −1
# # " CB (∇) 0 ¸¹
© IV. NUMERICAL E XAMPLE
Consider a time delay system of the form (l), with
Equation (4) can be simplified as
Y (k ) = G (∇) x (k − L) + H (∇)[U (k ) + F (k )] (5) ª 0.15 0.324 0 º
Where
«
A0 = 0.102 0.258 0 »;
« »
§ y (k − L) · § C · «¬ 0 0 0.412»¼
¨ ¸ ¨ ¸
¨ y (k − L + 1) ¸ ¨ C (∇ ) ¸
Y (k ) = , G (∇) = , ª 0 0.265 0 º
¨ # ¸ ¨ # ¸ «
¨ ¸ ¨ L ¸ A1 = −0.406 0 0.308 » ; B = I3 ; C = I3 .
« »
© y (k ) ¹ © CA (∇) ¹ «¬ 0 −0.252 0.124 »¼
§ u( k − L ) · § f (k − L) · Using the approach presented in the previous sections, the
¨ ¸ ¨ ¸ above system can be transform into the form of (2). The
¨ u(k − L + 1) ¸ ¨ f (k − L + 1) ¸
U (k ) = , F (k ) = . system has three actuators. In order to achieve fault detection
¨ # ¸ ¨ # ¸ and isolation, a generalized residual set consists in three
¨ ¸ ¨ ¸ residual signals should be created as show in Fig.1.
© u( k ) ¹ © f (k ) ¹
According to Ref. [5], a residual signal (parity signal) can be
defined as:
r = W [Y (k ) − H (∇)(U (k ) + F (k )] (6)
W is chosen to satisfy the follow equation:
WG (∇) = 0 (7)

6412
 Fig. 1 Generalized residual set To investigate the applicability of the approach, the
In Fig.1, fi stands for the ith actuator fault, and it follow fault f is imposed on the system.
directly impacts the input signal ui . In order to locate which ­0, (t < 100)
f = [0, 0, f3 ]T ; f3 = ®
actuator has fault, ri should be independent with ui . This can ¯0.4, (t ≥ 100)
be realized by choose a proper W . It should be noted that It means that actuator 1 and actuator 2 have no faults, while
actuator 3 goes wrong after 100 second. It is assumed that the
there are more than one W which satisfies (7). The freedom
system is corrupted by white noise. The fault isolation
of W can be used to design the three residual generators. threshold is not zero, and it can be determined by the norm of
W can be got by solving (7). the noise. The result of the simulation is show in Fig.2.
W = [k1 ,
1.767∇k1 + 2.160k1 - 0.03768k 2 - 0.7173∇ 2 k2
-0.6968∇k 2 + 0.2520∇k3 - 0.01221k 4 - 0.2357∇k4
-0.1901∇3 k 4 - 0.4170∇ 2 k4 - 0.1074∇ 2 k5
-0.009721k5 - 0.1798∇k5 + 0.1133∇k6
+0.2068∇ 2 k6 + 0.1807512∇3 k6 ,
-0.308∇k 2 - 0.412k3 - 0.124∇k3 - 0.09979∇k4
-0.08162∇ 2 k4 - 0.2064∇k5 - 0.03819∇ 2 k5
+0.06224∇ 2 k6 - 0.1023∇k6 - 0.1697k6 ,
-6.667k1 - 0.6800k 2 + 2.7067∇k2 - 0.3703k4
+0.6968∇k4 + 0.7173∇ 2 k4 + 1.104∇k5
-0.2774k5 - 0.6821∇ 2 k6 + 0.1714∇k6 , (a) system input signals

k2 , k3 , k 4 , k5 , k6 ]
k1 , k2 ," k6 ∈ R
W is a 9 dimensions vector with 6 freedoms. Using (6), r can
be generated.
r = [c1 c 2 c3 ] × y +[ c 4 c5 c6 ] × (u + f )
= c1 y1 + c2 y2 + c3 y3 (10)
+ c4 (u1 + f1 ) + c5 (u2 + f 2 ) + c6 (u3 + f3 )
where c1 ," , c6 are polynomials concerned with
k1 ," , k6 and ∇ . Let
­c4 = 0
°
®c5 ≠ 0 (11)
°c ≠ 0 (b) system output signals
¯ 6
and solve it, the follows can be got:
[ k1 k2 k3 k4 k5 k6 ]
= [0 -0.2580 1 0 1 0]
Substituting it to (10) r1 can be achieved, which is
independent with f1 while concerned with f 2 and f3 . By the
same process, r2 and r3 can be designed.

6413
 time-delay system. The effectiveness of the proposed method
has been verified by a numerical example.
However, further studies are required which include the follow
aspects:
1) To determine an optimal recursion step L . Such that the
residuals can obtain a certain freedom to complete fault
isolation, while the computation is minimized.
2) To extend the fault isolation result in this paper. The sensor
faults and the actuator faults should be discerned.
3) To enhance the reliability and robust performance of the
FDI system.
REFERENCES
[1] P. M. Frank, X. Ding, “Survey of robust residual generation and
evaluation methods in observer-based fault detection systems,” Journal of
Process Control, vol. 7 PP. 403-424, Jun. 1997
[2] J. Chen, R. J. Patton, Robust Model-based Fault Diagnosis for Dynamic
Systems, Kluwer Academic Publishers, 1999.
[3] R. J. Patton, P. M. Frank, R. N. Clark, Issues of Fault Diagnosis for
Dynamic Systems, Springer, 2000.
[4] Venkat Venkatasubramanian, Raghunathan Rengaswamy, N Kavuri
Surya, Surya N. Kavuri, “A review of process fault detection and
diagnosis: Part I: quantitative model-based methods,” Computers and
(c) Fault signals Chemical Engineering, vol. 27, pp. 293-311, 2003.
[5] E. Y. Chow, A. S. Willsky, “Analytic pedundancy and the design of
robust fault detection systems,” IEEE Transaction on Automatic Control,
vol. 29, pp. 603-614, Jul. 1984.
[6] J. Gertler, “Analytical redundancy methods in fault detection and
isolation,” Proceedings of the IFACIIMACS Symposium SAFEPROCESS
91, Baden-Baden, pp.9-21, 1991
[7] L. Dugard, E. I. Verriest, Stability and Control of Time-delay Systems.
Lecture Notes in Control and Information Sciences, Springer, 1997.
[8] H. Yang, M. Saif, “Observer design and fault diagnosis for state retarded
dynamical systems,” Automatica, vol. 34, pp: 217-227, Feb. 1998.
[9] F. KRATZ, W. NUNINGER, S. PLOIX, “Fault detection for time-delay
systems: a parity space approach,” Proceedings of the American Control
Conference, Philadelphia, Pennsylvania, pp. 2009-2011, 1998.
[10]M. Zhong, H. Ye, G. Wang, “Multi-freedom design of fault detection
filter for linear time-delay systems,” 2004 8th Inernational Conference on
Control, Automation, Robotics and Vision, Kunming, China, pp. 1630-
1634, 2004.

(d) Residual signals

Fig.2 the simulation results
As is shown in Fig.2d, r1 and r2 differ from zero after
100 second obviously while r3 is near to zero still. It is
evident that the approach works well.
IV. CONCLUSION S
A fault detection and isolation scheme for discrete time-
delay system has been proposed. The scheme can not only
detect the faults but also isolate (locate) the faults. To fulfill
the FDI, a generalized residual set in form of parity space is
designed by the recursion of the system equations. Each
residual corresponds to an actuator fault. The actuator with
fault can be isolated from the normal ones exactly. A time
delay operator is used to deal with the problem brought by the

6414