Fault Detection for Linear Switched Systems Based on Multiple
Kalman Filters Method
WANG Jiawei, SHEN Yi
1
, ZHANG Miao
Department of Control Science and Engineering, Harbin Institute of Technology, Harbin 150001, China
1
E-mail: shen@hit.edu.cn
Abstract: The switched system is a very important type of hybrid systems, so research of the switched system has attracted
extensive attention over the past two decades. A switched system is composed of a certain number of subsystems with a switched
signal. With the increment in the number of the subsystems and complexity of each subsystem, the probability of the entire
system failure is also multiplied. So fault detection for switched system has become extremely necessary. Existing fault detection
approaches are mostly based on the robust fault detection filter. There is a small number of articles about using Kalman filter to
detect faults of the switched system. A multiple Kalman filters method is proposed for fault detection of the entire switched
system, and the characteristic of the fault of switched system is analyzed systematically. Finally, an example is given to prove
the feasibility of the proposed method.
Key Words: Switched system, multiple Kalman filters, fault detection, intermittent fault
1 Introduction
For enhancing the reliability of system or device, fault
detection technique was invented about forty years ago.
After decades of development, the fault detection technique
has become an international hot research field; many
methods are proposed [1-3] and have been applied in a lot of
automatic control systems and industrial processes
successfully. Among these methods, the model-based
approaches are thought to be a large class of the earliest and
most mature approaches. Kalman filter method is a kind of
commonly used model-based fault detection method, whose
principle is to use the innovations between real output signal
and predicted output as a fault detection indicator. For many
linear systems and some special nonlinear systems, Kalman
filter method has been verified to be very effective.
However, for complex systems, such as hybrid systems for
detection, the results obtained are relatively limited.
The switched system is a very important kind of hybrid
systems, which are constituted by some continuous-time or
discrete-time subsystems as well as a subsystem switched
rule. It is more effective to describe complex engineering
systems with inherently multimodal or variable structures as
switched systems, such as biochemical processes, the
computer control systems. Compared with general systems,
switched systems have the characteristics of complexity and
peculiarity. Therefore, the research of switched systems is
especially important, and has attracted extensive attention
over the past two decades. Many achievements obtained
mainly focus on the stability analysis of switched systems
[4-5]. The problems of fault detection for switched systems
also began to receive much attention. In [6-7], robust H
fault detection filters are designed for minimizing estimation
errors between the residuals and faults, and reducing the
effect of the disturbances and model uncertainties. In
addition, in literature [8-9], full dimension state observer is
used for fault detection of switched systems. The above
*
This work is supported by the National Natural Science Foundation of
China (No. 61273162 and No. 61201310).
methods can be classified as observer based approach,
which also is a kind of model-based methods However,
there is not a lot of literatures involved using Kalman filter
for fault detecting of switched systems. In recent paper [10],
Kalman filter is used for switched systems fault detection
from the view of the parameter identification. Therefore, for
this purpose, we try to use the multiple Kalman filters
method to detect the fault of switched systems from the view
of state estimation in the paper.
The main body of this paper is arranged as below: in the
second part, preliminary knowledge including the basic
concepts and structure types of the switched system are
introduced; in the third part, firstly, intermittent
characteristic analysis of the fault of the switched system is
discussed, then for a subsystem of the switched systems,
Kalman filter is designed for fault detection, and extended to
the entire switched system to form multiple Kalman filters
method; in the fourth part, a simulation case is given to
prove the feasibility of multiple Kalman filters method; the
conclusion follows in the fifth part of the paper.
2 Preliminary
Before discussing the problem of switching system fault
detection using multiple Kalman filters, we should introduce
related background knowledge of switching system briefly.
Switched systems are a significant class of hybrid systems,
which comprise several different subsystems and a switched
signal to determine which subsystem is activated at some
specified time. Switched systems which are formed by N
discrete subsystems are described as:
( 1) ((),(),())
() ((),())
k F xk uk vk
yk G xk nk
V
V
(1)
where
()
k denotes the switched system state vector, ()uk
is the control input vector,
()vk and ()nk are external
signal vectors which denote the disturbance or unknown
input signals,
()yk is the output vector. And
^`
():0, 1,2,kNn
V
fo " means the switched signal
which is a piecewise constant function.
1N ! is the number
3URFHHGLQJVRIWKHQG&KLQHVH&RQWURO&RQIHUHQFH
-XO\;LDQ&KLQD
6328