SlideShare a Scribd company logo
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
1
Fault detection based on novel fuzzy modelling
Zhen Luo 1,2,
Huajing Fang 2
,Xiaoyong Liu 2
and Xuquan Chen 1
,2
1
Guangxi Special Equipment Supervision and Inspection Institute, Guangxi 530219, P.
R. China
2
Department of Control Science and Engineering, Huazhong University of Science and
Technology, Wuhan 430074, P. R. China
ABSTRACT
The Fault detection which is based on fuzzy modeling is investigated. Takagi-Sugeno (TS) fuzzy model can
be derived by structure and parameter identification, where only the input-output data of the identified
system are available. In the structure identification step, Gustafson-Kessel clustering algorithm (GKCA) is
used to detect clusters of different geometrical shapes in the data set and to obtain the point-wise
membership function of the premise. In the parameter identification step, Unscented Kalman filter (UKF) is
used to estimate the parameters of the premise’s membership function. In the consequence part, Kalman
filter (KF) algorithm is applied as a linear regression to estimate parameters of the TS model using the
input-output data set. Then, the obtained fuzzy model is used to detect the fault. Simulations are provided to
demonstrate the effectiveness of the theoretical results.
KEYWORDS
Unscented Kalman filter, Kalman filter, data driven, Gustafson-Kessel clustering algorithm, fuzzy
modeling, fault detection
1. INTRODUCTION
Since fault detection (FD) &diagnosis technique is essential to improve the safety and reliability
of dynamic systems, recently more and more attention has been paid to FD. Many FD methods
have been developed to detect and identify sensor and actuator faults, such as analytical
redundancy [1-3], a neural network [4], parameter identification method based on Fourier
Transform [5], testing the covariance matrix of the innovation sequence [5], testing the eigen
values of the sample covariance matrix [6]. All of these methods mentioned above are based upon
model. Models with good accuracy are necessary to improve the correct diagnostic of faults.
However, with the rapid development of industrial technology, the modern industrial processes
have become more and more complicated and large-scale. It is thus difficult to construct an
effective and explicit physical model to characterize these dynamical systems. Sometimes, it is
even impossible to model nonlinear systems by analytical equations [7]. Here, the fuzzy modeling
has been extensively used to model complex nonlinear system through a set of measured input-
output data [8]. The Takagi-Sugeno (TS) model, which uses the fuzzy modeling technique,
approximate the nonlinear system by smoothly interpolating affine local models. Each local
model contributes to the global model in a fuzzy subset of the space characterized by a
membership function [9]. Then, based on the established system model, FD can be carried out.
A TS fuzzy model is usually constructed in three steps:
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
2
Step 1: fuzzy clustering
Among various clustering algorithm, the Gustafson-Kessel clustering algorithm (GKCA) [10] has
been widely studied and applied by many researchers. The Gustafson-Kessel (GK) algorithm,
which is the fuzzy generalization of the Adaptive Distance Dynamic Clusters algorithm, searches
for ellipsoidal clusters. An advantage of the GK algorithm over Fuzzy C-means (FCM) [11] is
that GK can detect clusters of different shape and orientation in one data set.
Multidimensional fuzzy sets can be derived by clustering in the product space, but they are
generally difficult to be dealt with. In order to solve this problem, projected one-dimensional
fuzzy sets are usually preferred.
Step 2: estimate the parameters of the premise’s membership function
In order to obtain a fuzzy model, the premise’s membership functions must be expressed in a
form that allows computation of the membership degrees for input data not contained in the data
set. Each update estimation of the parameter vector corresponding to a nonlinear equation is
computed from the previous estimate and the new input data (here the input data are the point-
wise values of the membership function). To achieve this step, Unscented Kalman filter (UKF) is
used to approximate the point-wise defined membership functions by some suitable nonlinear
functions, for example normal function. The UKF as an improvement to the extended Kalman
filter (EKF) is one of the most widely used approach to analyze the stochastic nonlinear systems
[12]. This method is based on the unscented transform (UT) technique, a mechanism for
propagating mean and covariance through a nonlinear transformation [13, 14]. The state vector is
represented by a minimal set of carefully chosen sample points, called sigma points, which
approximate the posterior mean and covariance of the Gaussian random variable with a second
order accuracy[15, 16]. In contrast, the linearization technique used in the EKF can only achieve
first order accuracy. Generally, the prediction accuracy of UKF is better than EKF when the
model is highly nonlinear. The UKF has the same level of computational complexity as that of
EKF, both of which are within the order ( 3)
O L . Furthermore, Since the nonlinear models are
used without linearization, the UKF is not necessary to compute the Jacobian or Hessians
matrices [17, 18].
Step 3: estimation of the TS parameters
For linear dynamic systems with white process and measurement noise, the Kalman filter is an
optimal estimator [19]. The KF is used as a linear regression to efficiently choose the parameter
values of the consequent part (TS parameters) of the fuzzy model from the input-output data of
the identified system.
[20] proposes a fuzzy-modeling scheme combining GKCA and KF. GKCA is used to detect
clusters of different geometrical shapes in the data set and to obtain the point-wise membership
functions of the premise. After that, a KF is first used to estimate the parameters of the premise’s
membership function. Then, the KF is also used as a linear regression to efficiently choose the
parameter values of the consequent part (TS parameters) of the fuzzy model from the input-output
data of the identified system. KF process is a linear recursive minimum mean-square estimation
procedure. However, it can only deal with linear equation. In this sense, in order to obtain the
linear part, α -cut of the considered membership function is taken. Obviously, there are some
drawbacks with this algorithm
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
3
In order to obtain membership function for the premise’s fuzzy sets, the multidimensional fuzzy
set defined point-wise in the row of the partition matrix is projected onto the regressor. Division
of each membership function into two sets ( sets are obtained for each regressor and sets are
obtained for all regressors). It needs large amount of computation
The selection of α relies entirely on experience. Too much or too small will influence on
membership function approximation. Moreover, the cut data causes the loss of useful
information.
The straight line which is used to approximate each set is not very good fit of the data.
In this paper, we point-wise define the premise membership function for the regressor, each
membership function as a sets (c sets are obtained for each regressor and cn sets are obtained for
all regressors). This will halve the amount of computation. Obviously, it is more accurate to
approximate the point-wise defined membership function by some suitable nonlinear function.
UKF is utilized to approximate the nonlinear function. A common FD approach is to keep
tracking residuals of measurement and compare them against a set threshold value. The residual
can be generated by the comparison of actual sample points and estimated sample points .
The structure of this paper is as follow. In section 2, the TS fuzzy models principle is
summarized. Section 3 gives new fuzzy modeling algorithm. Fault detection is presented in
section 4. Section 5 presents some simulation results. Conclusions are given in section 6.
2. TS FUZZY MODELS
In this section, the TS fuzzy models principle is summarized. The TS fuzzy model can represent
or model any unknown nonlinear system ( )
y f x
= , on the basis of some available input-output
data 1 2
[ , ..., ]T
t t t nt
x x x x
= and t
y .
In the TS fuzzy model, the rule consequents are crisp function of the model inputs
i
R : if x is ( )
i
A x , then 1,2,...,
T
i i i
y a x b i c
= + = (1)
where is the dimensional input invariable,is the output invariable. and are the dimensional
TS parameters. denotes the th rule and is the number of rules in the rule base. is the premise
multivariable membership function of the th rule. is the parameter vector of the th rule
Multi dimensional fuzzy sets can be derived by clustering in the product space. But it is generally
difficult to be interpreted, so projected one-dimensional fuzzy sets are usually preferred.
The TS fuzzy system is described as a set ofc fuzzy rules where the i th rule is as follows:
i
R : if 1
x is 1 1
( )
i
A x and …and n
x is ( )
in n
A x , then 1,2,...,
T
i i i
y a x b i c
= + = (2)
The final output of the TS fuzzy model for an arbitrary input sample can be calculated using the
following expression:
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
4
1
1
( )( )
( )
c
T
i i i
i
c
i
i
x a x b
y
x
β
β
=
=
+
=
∑
∑
)
(3)
where ( )
i x
β represent the firing strength of the i th rule, and it has calculated by the following
equation:
1
( ) ij
n
i A
j
x
β µ
=
= ∏ (4)
ij
A
µ is the membership function of the fuzzy set ij
A .
3. NEW FUZZY MODELING ALGORITHM
The loop of establishing and training T-S fuzzy model by Gustafson-Kessel clustering
algorithm (GKCA) and UKF is as follows:
Fig.1 Flow chat of the new fuzzy modelling
Yes
Start
c=2
GKC
KF for
consequence
parameters
UKF for premise
membership
function
Performanc
es
End
No
c=c+
1
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
5
A detailed description of the algorithm is provided next.
1. From the input–output sequences 1
{( , )}N
k k t
x y = , partition the data into a set of local
linear submodels by using GKCA in the product space X Y
× .
2. Obtain the membership functions for the premise variables by using cluster projections
and UKF.
Estimate the consequent parameters by KF algorithm.
3.1 FUZZY CLUSTERING
Fuzzy clustering is an important tool to identify the structure in data. The Gustafson-Kessel(GK)
algorithm, which is the fuzzy generalization of the Adaptive Distance Dynamic Clusters
algorithm, searches for ellipsoidal clusters. An advantage of the GK algorithm over Fuzzy C-
means (FCM) is that GK can detect clusters of different shape and orientation in one data set.
The original objective functional for the GK algorithm is the following:
1 1
( ) ( )
c N
m T
it t i i t i
i t
J z v M z v
µ
= =
= − −
∑∑ (5)
The following restrictions hold
1
1 1
N
m
it
t
t N
µ
=
= ≤ ≤
∑
(6)
where is a positive-definite symmetric matrix related to the covariance matrix of the th
prototype, is a weighting exponent that determines the fuzziness of the resulting cluster(for a
crisp , fuzzy model , but typically ). is the number of data points, is the number clusters, is the
th data point, is the th cluster centre, is the degree of the membership of the th data point in the
th cluster centre.
3.2 PREMISE MEMBERSHIP FUNCTION
UKF as an improvement to extended Kalman filter (EKF) is the most widely used approach to
analyze the stochastic nonlinear system and show good performance in many cases[12]. In this
sense, we propose to use UKF as a nonlinear regression as follow: consider cn sets, each set
represents the nonlinear part of the point-wise set of a certain premise’s membership function (for
example Gauss type membership function). So we obtaincn parameter vector. In each set, we will
have j
N data (samples), where j denotes the j th set.
Then, each set can be modeled by the following measurement equation
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
6
2
2
( , ) 1,2...
( )
exp( ) 1,2...
2
j j j j
t t t
j j
j
t
t j
j
A h x v j cn
x d
v t N
θ
σ
= + =
−
= − + =
(7)
where [ , ]
j j j
d
θ σ
= is the parameters vector, j
t
v is the measurement noise, j
N is the number of
data(samples) in the j th set and the j denotes the j th nonlinear regression.
j
θ will be considered as a state variable, so the state equation will be
1 1
j j j j
t t t
F w
θ θ − −
= + (8)
where is the value of the state variable at the moment , is the state noise, is the state transition
matrix we assume that and are uncorrelated zero mean Gaussian random vectors and their
covariance matrices are and
The procedure for implementing the UKF can be summarized as follows:
Step 1: sigma points calculation
1 1 1 1
[ , ( ) ]
j j j
t t t t
n P
χ θ θ λ
− − − −
= ± +
) ) )
(9)
Step 2: prediction
| 1 1
j j j
t t t
F
χ χ
− −
=
| 1 1| 1
j j j
t t t t
F
θ θ
− − −
=
) )
| 1 1| 1
j j j jT
t t t t
P F P F Q
− − −
= +
)
(10)
Step 3: update.
| 1 | 1
( )
j j
t t t t
A h θ
− −
=
) )
| 1 | 1
( )
j j
l t t l t t
h
γ χ
− −
=
2
| 1 | 1 | 1 | 1
0
( )( )
j j
t t
n
j c j j j j T
l l t t t t l t t t t
A A
l
P A A R
ω γ γ
− − − −
=
= − − +
∑
) )
)
2
| 1 | 1 | 1
0
( )( )
j j
t t
n
j c j j j j T
l t t t l t t t t
A
l
P A R
θ
ω χ θ γ
− − −
=
= − − +
∑
) )
)
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
7
1
j j j j
t t t t
j j
t A A A
K P P
θ
−
=
) )
| 1 | 1
( )
j j j j
t t t t t t t
K A A
θ θ − −
= + −
) ) )
1
| 1 j j
t t
j T
t t t t t
A A
P P K P K
−
−
= −
) ) )
(11)
where, 2
0 / ( ) (1 )
c
n
ω λ λ α β
= + + − + , / 2( ) 1...2
m c
l l n l n
ω ω λ λ
= = + = , 2
( 1)
n
λ α
= − . c
l
ω is a set
of scalar weights, and n is the state dimension; the parameter α determines the spread of
the sigma points around x
)
and is usually set to 1 4 1
e α
− ≤ ≤ , The constant β is used to
incorporate part of the prior knowledge of the distribution of x , and for Gaussian
distributions, 2
β = is optimal.
Often it is expected that the system parameters do not vary or, if they do, the variation is
much slower than that of the system state. So, we will take j
F I
= .
3.3 ESTIMATION CONSEQUENT PARAMETERS
In this section, KF algorithm is used to compute the consequent parameters from the data
set and the estimated premise’s membership functions.
From (3), we have
1
( )( )
c
T
i i i
i
y x a x b
ψ
=
= +
∑
) (12)
where
1
( )
( )
( )
i
i c
i
i
x
x
x
β
ψ
β
=
=
∑ is the normalized activation value of the i th rule.
[ ] [ ] [ ] [ ]
1 2 1 1
( ) 1 ( ) 1 ... ( ) 1 ...
T
c c c
y x x x x x x a b a b
ψ ψ ψ
 
=  
)
[ ]
1 1...
T
c c
a b a b
Φ = , [ ] [ ] [ ]
1 2
( ) 1 ( ) 1 ... ( ) 1
c
x x x x x x
ψ ψ ψ
 
Ω =   (13)
The measurement equation can be taken as the following form
t t t
y v
= Ω Φ +
) (14)
Then, j
θ will be considered as a state variable, so the state equation will be
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
8
1 1
t t t
F w
− −
Φ = Φ + (15)
Now, we can use KF to estimate the TS parameter vector t
Φ as follows:
| 1 1| 1
t t t t
F
− − −
Φ = Φ
) )
| 1 1| 1
T
t t t t
P FP F Q
− − −
= +
1
| 1 | 1
( )
T T
t t t t t t t t
K P P R −
− −
= Ω Ω Ω +
| | 1 | 1
( )
t t t t t t t t t
K y
− −
Φ = Φ + −Ω Φ
) ) )
| | 1 | 1
t t t t t t t t
P P K P
− −
= − Ω (16)
where t
Φ
)
is the estimated value of t
Φ . Also, we will take F I
= .
4. PROCESS FAULT DETECTION
The fuzzy FD system is based on fuzzy models identified directly from data. A model is used to
estimate the nominal output signals.
y y
∆ = −
)
(17)
where y is the output of the system and y
)
is the output of the model in normal operation. When
any component of ∆ is bigger than a certain threshold, the system detects a fault.
5.SIMULATION
The sampling period of training data and testing data are all 0.05h. We choose 800 data points
without fault and 600 data points with fault 1 after collecting 200 data points without fault. The
number of clusters 12
c = .
0 100 200 300 400 500 600 700 800 900
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
x
membership
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
9
Fig.2 the red solid line indicates the actual membership function, the black dotted line indicates
the premise membership function based on UKF, the blue dashed line indicates the premise
membership function based on KF
Fig.3 the actual sample point without fault and estimated sample point without fault
Fig.4 the residual with fault
Fig.5 the actual sample point without fault and the estimated sample point with fault
0 100 200 300 400 500 600 700 800 900
0.185
0.19
0.195
0.2
0.205
0.21
0.215
0.22
0.225
0.23
0.235
no fault model output
sampling point
variable
NO
1
estimated value without fault
actual value without fault
0 100 200 300 400 500 600 700 800 900
0
0.005
0.01
0.015
sampling point
residual
0 100 200 300 400 500 600 700 800 900
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
sampling point
variable
NO
1
actual value without fault
estimated value with fault
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
10
Fig.6 the residual with fault
The membership functions is given in Fig.2. The red solid line indicates the actual membership
function, the black dotted line indicates the premise membership function based on UKF, the blue
dashed line indicates the premise membership function based on KF. It can be seen from the
figure that both approaches track the true state. However, the obtained premise membership based
on UKF is more close to the true value than based on KF. It means that estimating the premise
membership based on UKF is superior to based on KF.
In Fig.3, we can see that the actual sample point can be tracked well by the estimated sample
point. Fig.4 shows that the absolute value of the residuals which is derived by subtracting the
estimated value of the actual value is lower than the threshold. The simulation results consistent
with the actual case.
When the fault 1 occurs at the step 200, the simulation results are given in Fig.5-6. The estimated
sample points have deviated from actual sample points after 200 steps in Fig5 due to the fault, but
the actual sample point can be tracked well by the estimated sample point after the step 500
because of feedback effect. The fault makes absolute value of the residuals exceed the threshold
at step 206 which is given in Fig.6. I.e. the fault can be detected after occurrence of 6 steps.
6.CONCLUSION
In this paper, a fault detection algorithm based on data driven is proposed. The proposed
algorithm is composed of two steps: (1) fuzzy modeling contain fuzzy clustering, determination
of premise membership functions and TS parameters. (2) Fault detection based on the established
model. In the first step, a fuzzy modeling scheme on the basis of the GKCA and UKF to estimate
the premise membership, KF is used to estimate the TS parameter. The performances of the
proposed algorithm are demonstrated on the TE challenge problem.
0 100 200 300 400 500 600 700 800 900
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
detection of the occurance of the fault
sampling point
r
e
s
id
u
a
l
residual
threshold
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
11
REFERENCES
[1] Zhang YM, Li XR, Ieee(1997)”Detection and diagnosis of sensor and actuator failures using
interacting multiple-model estimator”, In: 36th IEEE Conference on Decision and Control: Dec 10-12
1997; San Diego, Ca; pp4475-4480.
[2] Maybeck PS(1999)” Multiple model adaptive algorithms for detecting and compensating sensor and
actuator/surface failures in aircraft flight control systems”, International Journal of Robust and
Nonlinear Control, 9(14):pp1051-1070.
[3] Rago C, Prasanth R, Mehra RK, Fortenbaugh R(1998), Ieee:”Failure Detection and dentification and
Fault Tolerant control using the IMM-KF with applications to the Eagle-Eye UAV”. In: 37th IEEE
Conference on Decision and Control: Dec 16-18 ; Tampa, Fl; pp4208-4213.
[4] Alessandri A(2003)”Fault diagnosis for nonlinear systems using a bank of neural estimators”.
Computers in Industry , 52(3)pp271-289.
[5] Perhinschi MG, Lando M, Massotti L, Campa G, Napolitano MR, Fravolini ML(2002), Acc”On-line
parameter estimation issues for the NASA IFCSF-15 fault tolerant systems”. In: 20th Annual
American Control Conference (ACC): May 08-10 2002; Anchorage, Ak; pp191-196.
[6] Hajiyev C, Caliskan F(2005)” Sensor and control surface/actuator failure detection and isolation
applied to F-16 flight dynamic”. Aircraft Engineering and Aerospace Technology , 77(2)pp152-160.
[7] Mendonca LF, Sousa JMC, da Costa J(2009)” An architecture for fault detection and isolation based
on fuzzy methods” Expert Systems with Applications 36(2)pp1092-1104.
[8] Amann P, Perronne JM, Gissinger GL( 2001)”Frank PM: Identification of fuzzy relational models for
fault detection”. Control Engineering Practice, 9(5)pp555-562.
[9] Skrjanc I, Blazic S(2005)” Agamennoni O: Identification of dynamical systems with a robust interval
fuzzy model”. Automatica , 41(2)pp327-332.
[10] Gustafson D, Kessel W(1979)”Fuzzy clustering with a fuzzy covariance matrix.” Proceedings of
IEEE CDC, San Diego, CA,pp 761-766.
[11] Nascimento S, Mirkin B, Moura-Pires F(2003)” Modeling proportional membership in fuzzy
clustering”. Ieee Transactions on Fuzzy Systems , 11(2)pp173-186.
[12] Xiong K, Zhang HY, Chan CW(2006)” Performance evaluation of UKF-based nonlinear filtering”.
Automatica , 42(2)pp261-270.
[13] Julier SJ, Uhlmann JK(1997)” A new extension of the Kalman filter to nonlinear systems”, vol. 3068.
[14] Julier SJ, LaViola JJ(2007)” On Kalman filtering with nonlinear equality constraints”. Ieee
Transactions on Signal Processing, 55(6)pp2774-2784.
[15] Kandepu R, Foss B, Imsland L(2008)”Applying the unscented Kalman filter for nonlinear state
estimation”. Journal of Process Control, 18(7-8)pp753-768.
[16] Julier SJ, Uhlmann JK(2004)”Unscented filtering and nonlinear estimation”. Proceedings of the Ieee ,
92(3)pp401-422.
[17] Li PH, Zhang TW, Ma B(2004)” Unscented Kalman filter for visual curve tracking”. Image and
Vision Computing, 22(2)pp157-164.
[18] Dini DH, Mandic DP, Julier SJ(2011)”A Widely Linear Complex Unscented Kalman Filter”. Ieee
Signal Processing Letters, 18(11)pp623-626.
[19] Simon D, Simon DL(2010)” Constrained Kalman filtering via density function truncation for turbofan
engine health estimation”. International Journal of Systems Science, 41(2)pp159-171.
[20] Chafaa K, Ghanai M, Benmahammed K(2007)” Fuzzy modelling using Kalman filter”. Iet Control
Theory and Applications, 1(1)pp58-64.
[21] Downs J J, Vogel E F(1993)”A plant-wide industrial process control problem”. Computers and
Chemical Engineering, 17(3)pp245-255.
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
Authors
Zhen Luo She was born in 1983. Sh
Science and Technology. Her research interests include
and fault diagnosis of networked control sys
Huajing Fang was born in 1955. He is
and Technology. His research interests
dynamics and control of autonomous
control.
Xuquan Chen was born in 1984.
Science and Technology. His research interest is signal processing.
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
She is a Ph.D. candidate in Huazhong University of
research interests include Kalman filtering state estimation
networked control sys tems.
was born in 1955. He is a Professor in Huazhong University of Science
His research interests include complex networked control systems,
mous vehicle’s swarm, fault diagnosis and fault-tolerant
was born in 1984. he is a Ph.D. candidate in Huazhong University of
Science and Technology. His research interest is signal processing.
Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014
12

More Related Content

What's hot

A Mathematical Programming Approach for Selection of Variables in Cluster Ana...
A Mathematical Programming Approach for Selection of Variables in Cluster Ana...A Mathematical Programming Approach for Selection of Variables in Cluster Ana...
A Mathematical Programming Approach for Selection of Variables in Cluster Ana...
IJRES Journal
 
FUZZY CONTROL OF A SERVOMECHANISM: PRACTICAL APPROACH USING MAMDANI AND TAKAG...
FUZZY CONTROL OF A SERVOMECHANISM: PRACTICAL APPROACH USING MAMDANI AND TAKAG...FUZZY CONTROL OF A SERVOMECHANISM: PRACTICAL APPROACH USING MAMDANI AND TAKAG...
FUZZY CONTROL OF A SERVOMECHANISM: PRACTICAL APPROACH USING MAMDANI AND TAKAG...
ijfls
 
Parallel Machine Learning
Parallel Machine LearningParallel Machine Learning
Parallel Machine Learning
Janani C
 
PROTECTOR CONTROL PC-AODV-BH IN THE AD HOC NETWORKS
PROTECTOR CONTROL PC-AODV-BH IN THE AD HOC NETWORKSPROTECTOR CONTROL PC-AODV-BH IN THE AD HOC NETWORKS
PROTECTOR CONTROL PC-AODV-BH IN THE AD HOC NETWORKS
Zac Darcy
 
IRJET- Domestic Water Conservation by IoT (Smart Home)
IRJET- Domestic Water Conservation by IoT (Smart Home)IRJET- Domestic Water Conservation by IoT (Smart Home)
IRJET- Domestic Water Conservation by IoT (Smart Home)
IRJET Journal
 
Heuristic approach to optimize the number of test cases for simple circuits
Heuristic approach to optimize the number of test cases for simple circuitsHeuristic approach to optimize the number of test cases for simple circuits
Heuristic approach to optimize the number of test cases for simple circuits
VLSICS Design
 
Symbol Based Modulation Classification using Combination of Fuzzy Clustering ...
Symbol Based Modulation Classification using Combination of Fuzzy Clustering ...Symbol Based Modulation Classification using Combination of Fuzzy Clustering ...
Symbol Based Modulation Classification using Combination of Fuzzy Clustering ...
CSCJournals
 
EFFECT OF TWO EXOSYSTEM STRUCTURES ON OUTPUT REGULATION OF THE RTAC SYSTEM
EFFECT OF TWO EXOSYSTEM STRUCTURES ON OUTPUT REGULATION OF THE RTAC SYSTEMEFFECT OF TWO EXOSYSTEM STRUCTURES ON OUTPUT REGULATION OF THE RTAC SYSTEM
EFFECT OF TWO EXOSYSTEM STRUCTURES ON OUTPUT REGULATION OF THE RTAC SYSTEM
ijctcm
 
Financial Time Series Analysis Based On Normalized Mutual Information Functions
Financial Time Series Analysis Based On Normalized Mutual Information FunctionsFinancial Time Series Analysis Based On Normalized Mutual Information Functions
Financial Time Series Analysis Based On Normalized Mutual Information Functions
IJCI JOURNAL
 
DOTNET 2013 IEEE CLOUDCOMPUTING PROJECT A fast clustering based feature subse...
DOTNET 2013 IEEE CLOUDCOMPUTING PROJECT A fast clustering based feature subse...DOTNET 2013 IEEE CLOUDCOMPUTING PROJECT A fast clustering based feature subse...
DOTNET 2013 IEEE CLOUDCOMPUTING PROJECT A fast clustering based feature subse...
IEEEGLOBALSOFTTECHNOLOGIES
 
Single Channel Speech De-noising Using Kernel Independent Component Analysis...
	Single Channel Speech De-noising Using Kernel Independent Component Analysis...	Single Channel Speech De-noising Using Kernel Independent Component Analysis...
Single Channel Speech De-noising Using Kernel Independent Component Analysis...
theijes
 
A Combined Approach for Feature Subset Selection and Size Reduction for High ...
A Combined Approach for Feature Subset Selection and Size Reduction for High ...A Combined Approach for Feature Subset Selection and Size Reduction for High ...
A Combined Approach for Feature Subset Selection and Size Reduction for High ...
IJERA Editor
 
Ag04606202206
Ag04606202206Ag04606202206
Ag04606202206
IJERA Editor
 
A Modified KS-test for Feature Selection
A Modified KS-test for Feature SelectionA Modified KS-test for Feature Selection
A Modified KS-test for Feature Selection
IOSR Journals
 
Sensitivity analysis in a lidar camera calibration
Sensitivity analysis in a lidar camera calibrationSensitivity analysis in a lidar camera calibration
Sensitivity analysis in a lidar camera calibration
csandit
 
Farsi character recognition using new hybrid feature extraction methods
Farsi character recognition using new hybrid feature extraction methodsFarsi character recognition using new hybrid feature extraction methods
Farsi character recognition using new hybrid feature extraction methods
ijcseit
 

What's hot (16)

A Mathematical Programming Approach for Selection of Variables in Cluster Ana...
A Mathematical Programming Approach for Selection of Variables in Cluster Ana...A Mathematical Programming Approach for Selection of Variables in Cluster Ana...
A Mathematical Programming Approach for Selection of Variables in Cluster Ana...
 
FUZZY CONTROL OF A SERVOMECHANISM: PRACTICAL APPROACH USING MAMDANI AND TAKAG...
FUZZY CONTROL OF A SERVOMECHANISM: PRACTICAL APPROACH USING MAMDANI AND TAKAG...FUZZY CONTROL OF A SERVOMECHANISM: PRACTICAL APPROACH USING MAMDANI AND TAKAG...
FUZZY CONTROL OF A SERVOMECHANISM: PRACTICAL APPROACH USING MAMDANI AND TAKAG...
 
Parallel Machine Learning
Parallel Machine LearningParallel Machine Learning
Parallel Machine Learning
 
PROTECTOR CONTROL PC-AODV-BH IN THE AD HOC NETWORKS
PROTECTOR CONTROL PC-AODV-BH IN THE AD HOC NETWORKSPROTECTOR CONTROL PC-AODV-BH IN THE AD HOC NETWORKS
PROTECTOR CONTROL PC-AODV-BH IN THE AD HOC NETWORKS
 
IRJET- Domestic Water Conservation by IoT (Smart Home)
IRJET- Domestic Water Conservation by IoT (Smart Home)IRJET- Domestic Water Conservation by IoT (Smart Home)
IRJET- Domestic Water Conservation by IoT (Smart Home)
 
Heuristic approach to optimize the number of test cases for simple circuits
Heuristic approach to optimize the number of test cases for simple circuitsHeuristic approach to optimize the number of test cases for simple circuits
Heuristic approach to optimize the number of test cases for simple circuits
 
Symbol Based Modulation Classification using Combination of Fuzzy Clustering ...
Symbol Based Modulation Classification using Combination of Fuzzy Clustering ...Symbol Based Modulation Classification using Combination of Fuzzy Clustering ...
Symbol Based Modulation Classification using Combination of Fuzzy Clustering ...
 
EFFECT OF TWO EXOSYSTEM STRUCTURES ON OUTPUT REGULATION OF THE RTAC SYSTEM
EFFECT OF TWO EXOSYSTEM STRUCTURES ON OUTPUT REGULATION OF THE RTAC SYSTEMEFFECT OF TWO EXOSYSTEM STRUCTURES ON OUTPUT REGULATION OF THE RTAC SYSTEM
EFFECT OF TWO EXOSYSTEM STRUCTURES ON OUTPUT REGULATION OF THE RTAC SYSTEM
 
Financial Time Series Analysis Based On Normalized Mutual Information Functions
Financial Time Series Analysis Based On Normalized Mutual Information FunctionsFinancial Time Series Analysis Based On Normalized Mutual Information Functions
Financial Time Series Analysis Based On Normalized Mutual Information Functions
 
DOTNET 2013 IEEE CLOUDCOMPUTING PROJECT A fast clustering based feature subse...
DOTNET 2013 IEEE CLOUDCOMPUTING PROJECT A fast clustering based feature subse...DOTNET 2013 IEEE CLOUDCOMPUTING PROJECT A fast clustering based feature subse...
DOTNET 2013 IEEE CLOUDCOMPUTING PROJECT A fast clustering based feature subse...
 
Single Channel Speech De-noising Using Kernel Independent Component Analysis...
	Single Channel Speech De-noising Using Kernel Independent Component Analysis...	Single Channel Speech De-noising Using Kernel Independent Component Analysis...
Single Channel Speech De-noising Using Kernel Independent Component Analysis...
 
A Combined Approach for Feature Subset Selection and Size Reduction for High ...
A Combined Approach for Feature Subset Selection and Size Reduction for High ...A Combined Approach for Feature Subset Selection and Size Reduction for High ...
A Combined Approach for Feature Subset Selection and Size Reduction for High ...
 
Ag04606202206
Ag04606202206Ag04606202206
Ag04606202206
 
A Modified KS-test for Feature Selection
A Modified KS-test for Feature SelectionA Modified KS-test for Feature Selection
A Modified KS-test for Feature Selection
 
Sensitivity analysis in a lidar camera calibration
Sensitivity analysis in a lidar camera calibrationSensitivity analysis in a lidar camera calibration
Sensitivity analysis in a lidar camera calibration
 
Farsi character recognition using new hybrid feature extraction methods
Farsi character recognition using new hybrid feature extraction methodsFarsi character recognition using new hybrid feature extraction methods
Farsi character recognition using new hybrid feature extraction methods
 

Similar to Fault detection based on novel fuzzy modelling

The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...
The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...
The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...
Waqas Tariq
 
The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...
The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...
The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...
CSCJournals
 
Cone Crusher Model Identification Using Block-Oriented Systems with Orthonorm...
Cone Crusher Model Identification Using Block-Oriented Systems with Orthonorm...Cone Crusher Model Identification Using Block-Oriented Systems with Orthonorm...
Cone Crusher Model Identification Using Block-Oriented Systems with Orthonorm...
ijctcm
 
Cone crusher model identification using
Cone crusher model identification usingCone crusher model identification using
Cone crusher model identification using
ijctcm
 
On Selection of Periodic Kernels Parameters in Time Series Prediction
On Selection of Periodic Kernels Parameters in Time Series Prediction On Selection of Periodic Kernels Parameters in Time Series Prediction
On Selection of Periodic Kernels Parameters in Time Series Prediction
cscpconf
 
ON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTION
ON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTIONON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTION
ON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTION
cscpconf
 
Direct digital control scheme for controlling hybrid dynamic systems using AN...
Direct digital control scheme for controlling hybrid dynamic systems using AN...Direct digital control scheme for controlling hybrid dynamic systems using AN...
Direct digital control scheme for controlling hybrid dynamic systems using AN...
XiaoLaui
 
Machine Learning.pdf
Machine Learning.pdfMachine Learning.pdf
Machine Learning.pdf
BeyaNasr1
 
2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile
2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile
2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile
Zac Darcy
 
2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile
2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile
2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile
Zac Darcy
 
SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...
SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...
SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...
ijistjournal
 
SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...
SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...
SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...
ijistjournal
 
Q UANTUM C LUSTERING -B ASED F EATURE SUBSET S ELECTION FOR MAMMOGRAPHIC I...
Q UANTUM  C LUSTERING -B ASED  F EATURE SUBSET  S ELECTION FOR MAMMOGRAPHIC I...Q UANTUM  C LUSTERING -B ASED  F EATURE SUBSET  S ELECTION FOR MAMMOGRAPHIC I...
Q UANTUM C LUSTERING -B ASED F EATURE SUBSET S ELECTION FOR MAMMOGRAPHIC I...
ijcsit
 
An Automatic Clustering Technique for Optimal Clusters
An Automatic Clustering Technique for Optimal ClustersAn Automatic Clustering Technique for Optimal Clusters
An Automatic Clustering Technique for Optimal Clusters
IJCSEA Journal
 
Dynamic Kohonen Network for Representing Changes in Inputs
Dynamic Kohonen Network for Representing Changes in InputsDynamic Kohonen Network for Representing Changes in Inputs
Dynamic Kohonen Network for Representing Changes in Inputs
Jean Fecteau
 
Dycops2019
Dycops2019 Dycops2019
Dycops2019
Jéssyca Bessa
 
Compositional testing for fsm based models
Compositional testing for fsm based modelsCompositional testing for fsm based models
Compositional testing for fsm based models
ijseajournal
 
Human Activity Recognition Using AccelerometerData
Human Activity Recognition Using AccelerometerDataHuman Activity Recognition Using AccelerometerData
Human Activity Recognition Using AccelerometerData
IRJET Journal
 
F017533540
F017533540F017533540
F017533540
IOSR Journals
 
State Estimation based Inverse Dynamic Controller for Hybrid system using Art...
State Estimation based Inverse Dynamic Controller for Hybrid system using Art...State Estimation based Inverse Dynamic Controller for Hybrid system using Art...
State Estimation based Inverse Dynamic Controller for Hybrid system using Art...
XiaoLaui
 

Similar to Fault detection based on novel fuzzy modelling (20)

The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...
The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...
The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...
 
The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...
The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...
The Positive Effects of Fuzzy C-Means Clustering on Supervised Learning Class...
 
Cone Crusher Model Identification Using Block-Oriented Systems with Orthonorm...
Cone Crusher Model Identification Using Block-Oriented Systems with Orthonorm...Cone Crusher Model Identification Using Block-Oriented Systems with Orthonorm...
Cone Crusher Model Identification Using Block-Oriented Systems with Orthonorm...
 
Cone crusher model identification using
Cone crusher model identification usingCone crusher model identification using
Cone crusher model identification using
 
On Selection of Periodic Kernels Parameters in Time Series Prediction
On Selection of Periodic Kernels Parameters in Time Series Prediction On Selection of Periodic Kernels Parameters in Time Series Prediction
On Selection of Periodic Kernels Parameters in Time Series Prediction
 
ON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTION
ON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTIONON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTION
ON SELECTION OF PERIODIC KERNELS PARAMETERS IN TIME SERIES PREDICTION
 
Direct digital control scheme for controlling hybrid dynamic systems using AN...
Direct digital control scheme for controlling hybrid dynamic systems using AN...Direct digital control scheme for controlling hybrid dynamic systems using AN...
Direct digital control scheme for controlling hybrid dynamic systems using AN...
 
Machine Learning.pdf
Machine Learning.pdfMachine Learning.pdf
Machine Learning.pdf
 
2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile
2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile
2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile
 
2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile
2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile
2-DOF Block Pole Placement Control Application To: Have-DASH-IIBITT Missile
 
SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...
SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...
SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...
 
SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...
SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...
SYSTEM IDENTIFICATION AND MODELING FOR INTERACTING AND NON-INTERACTING TANK S...
 
Q UANTUM C LUSTERING -B ASED F EATURE SUBSET S ELECTION FOR MAMMOGRAPHIC I...
Q UANTUM  C LUSTERING -B ASED  F EATURE SUBSET  S ELECTION FOR MAMMOGRAPHIC I...Q UANTUM  C LUSTERING -B ASED  F EATURE SUBSET  S ELECTION FOR MAMMOGRAPHIC I...
Q UANTUM C LUSTERING -B ASED F EATURE SUBSET S ELECTION FOR MAMMOGRAPHIC I...
 
An Automatic Clustering Technique for Optimal Clusters
An Automatic Clustering Technique for Optimal ClustersAn Automatic Clustering Technique for Optimal Clusters
An Automatic Clustering Technique for Optimal Clusters
 
Dynamic Kohonen Network for Representing Changes in Inputs
Dynamic Kohonen Network for Representing Changes in InputsDynamic Kohonen Network for Representing Changes in Inputs
Dynamic Kohonen Network for Representing Changes in Inputs
 
Dycops2019
Dycops2019 Dycops2019
Dycops2019
 
Compositional testing for fsm based models
Compositional testing for fsm based modelsCompositional testing for fsm based models
Compositional testing for fsm based models
 
Human Activity Recognition Using AccelerometerData
Human Activity Recognition Using AccelerometerDataHuman Activity Recognition Using AccelerometerData
Human Activity Recognition Using AccelerometerData
 
F017533540
F017533540F017533540
F017533540
 
State Estimation based Inverse Dynamic Controller for Hybrid system using Art...
State Estimation based Inverse Dynamic Controller for Hybrid system using Art...State Estimation based Inverse Dynamic Controller for Hybrid system using Art...
State Estimation based Inverse Dynamic Controller for Hybrid system using Art...
 

Recently uploaded

BRAIN TUMOR DETECTION for seminar ppt.pdf
BRAIN TUMOR DETECTION for seminar ppt.pdfBRAIN TUMOR DETECTION for seminar ppt.pdf
BRAIN TUMOR DETECTION for seminar ppt.pdf
LAXMAREDDY22
 
john krisinger-the science and history of the alcoholic beverage.pptx
john krisinger-the science and history of the alcoholic beverage.pptxjohn krisinger-the science and history of the alcoholic beverage.pptx
john krisinger-the science and history of the alcoholic beverage.pptx
Madan Karki
 
cnn.pptx Convolutional neural network used for image classication
cnn.pptx Convolutional neural network used for image classicationcnn.pptx Convolutional neural network used for image classication
cnn.pptx Convolutional neural network used for image classication
SakkaravarthiShanmug
 
Generative AI leverages algorithms to create various forms of content
Generative AI leverages algorithms to create various forms of contentGenerative AI leverages algorithms to create various forms of content
Generative AI leverages algorithms to create various forms of content
Hitesh Mohapatra
 
CEC 352 - SATELLITE COMMUNICATION UNIT 1
CEC 352 - SATELLITE COMMUNICATION UNIT 1CEC 352 - SATELLITE COMMUNICATION UNIT 1
CEC 352 - SATELLITE COMMUNICATION UNIT 1
PKavitha10
 
Engineering Drawings Lecture Detail Drawings 2014.pdf
Engineering Drawings Lecture Detail Drawings 2014.pdfEngineering Drawings Lecture Detail Drawings 2014.pdf
Engineering Drawings Lecture Detail Drawings 2014.pdf
abbyasa1014
 
spirit beverages ppt without graphics.pptx
spirit beverages ppt without graphics.pptxspirit beverages ppt without graphics.pptx
spirit beverages ppt without graphics.pptx
Madan Karki
 
The Python for beginners. This is an advance computer language.
The Python for beginners. This is an advance computer language.The Python for beginners. This is an advance computer language.
The Python for beginners. This is an advance computer language.
sachin chaurasia
 
Welding Metallurgy Ferrous Materials.pdf
Welding Metallurgy Ferrous Materials.pdfWelding Metallurgy Ferrous Materials.pdf
Welding Metallurgy Ferrous Materials.pdf
AjmalKhan50578
 
artificial intelligence and data science contents.pptx
artificial intelligence and data science contents.pptxartificial intelligence and data science contents.pptx
artificial intelligence and data science contents.pptx
GauravCar
 
CHINA’S GEO-ECONOMIC OUTREACH IN CENTRAL ASIAN COUNTRIES AND FUTURE PROSPECT
CHINA’S GEO-ECONOMIC OUTREACH IN CENTRAL ASIAN COUNTRIES AND FUTURE PROSPECTCHINA’S GEO-ECONOMIC OUTREACH IN CENTRAL ASIAN COUNTRIES AND FUTURE PROSPECT
CHINA’S GEO-ECONOMIC OUTREACH IN CENTRAL ASIAN COUNTRIES AND FUTURE PROSPECT
jpsjournal1
 
Properties Railway Sleepers and Test.pptx
Properties Railway Sleepers and Test.pptxProperties Railway Sleepers and Test.pptx
Properties Railway Sleepers and Test.pptx
MDSABBIROJJAMANPAYEL
 
CompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURS
CompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURSCompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURS
CompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURS
RamonNovais6
 
ISPM 15 Heat Treated Wood Stamps and why your shipping must have one
ISPM 15 Heat Treated Wood Stamps and why your shipping must have oneISPM 15 Heat Treated Wood Stamps and why your shipping must have one
ISPM 15 Heat Treated Wood Stamps and why your shipping must have one
Las Vegas Warehouse
 
LLM Fine Tuning with QLoRA Cassandra Lunch 4, presented by Anant
LLM Fine Tuning with QLoRA Cassandra Lunch 4, presented by AnantLLM Fine Tuning with QLoRA Cassandra Lunch 4, presented by Anant
LLM Fine Tuning with QLoRA Cassandra Lunch 4, presented by Anant
Anant Corporation
 
AI assisted telemedicine KIOSK for Rural India.pptx
AI assisted telemedicine KIOSK for Rural India.pptxAI assisted telemedicine KIOSK for Rural India.pptx
AI assisted telemedicine KIOSK for Rural India.pptx
architagupta876
 
一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理
一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理
一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理
ecqow
 
22CYT12-Unit-V-E Waste and its Management.ppt
22CYT12-Unit-V-E Waste and its Management.ppt22CYT12-Unit-V-E Waste and its Management.ppt
22CYT12-Unit-V-E Waste and its Management.ppt
KrishnaveniKrishnara1
 
官方认证美国密歇根州立大学毕业证学位证书原版一模一样
官方认证美国密歇根州立大学毕业证学位证书原版一模一样官方认证美国密歇根州立大学毕业证学位证书原版一模一样
官方认证美国密歇根州立大学毕业证学位证书原版一模一样
171ticu
 
哪里办理(csu毕业证书)查尔斯特大学毕业证硕士学历原版一模一样
哪里办理(csu毕业证书)查尔斯特大学毕业证硕士学历原版一模一样哪里办理(csu毕业证书)查尔斯特大学毕业证硕士学历原版一模一样
哪里办理(csu毕业证书)查尔斯特大学毕业证硕士学历原版一模一样
insn4465
 

Recently uploaded (20)

BRAIN TUMOR DETECTION for seminar ppt.pdf
BRAIN TUMOR DETECTION for seminar ppt.pdfBRAIN TUMOR DETECTION for seminar ppt.pdf
BRAIN TUMOR DETECTION for seminar ppt.pdf
 
john krisinger-the science and history of the alcoholic beverage.pptx
john krisinger-the science and history of the alcoholic beverage.pptxjohn krisinger-the science and history of the alcoholic beverage.pptx
john krisinger-the science and history of the alcoholic beverage.pptx
 
cnn.pptx Convolutional neural network used for image classication
cnn.pptx Convolutional neural network used for image classicationcnn.pptx Convolutional neural network used for image classication
cnn.pptx Convolutional neural network used for image classication
 
Generative AI leverages algorithms to create various forms of content
Generative AI leverages algorithms to create various forms of contentGenerative AI leverages algorithms to create various forms of content
Generative AI leverages algorithms to create various forms of content
 
CEC 352 - SATELLITE COMMUNICATION UNIT 1
CEC 352 - SATELLITE COMMUNICATION UNIT 1CEC 352 - SATELLITE COMMUNICATION UNIT 1
CEC 352 - SATELLITE COMMUNICATION UNIT 1
 
Engineering Drawings Lecture Detail Drawings 2014.pdf
Engineering Drawings Lecture Detail Drawings 2014.pdfEngineering Drawings Lecture Detail Drawings 2014.pdf
Engineering Drawings Lecture Detail Drawings 2014.pdf
 
spirit beverages ppt without graphics.pptx
spirit beverages ppt without graphics.pptxspirit beverages ppt without graphics.pptx
spirit beverages ppt without graphics.pptx
 
The Python for beginners. This is an advance computer language.
The Python for beginners. This is an advance computer language.The Python for beginners. This is an advance computer language.
The Python for beginners. This is an advance computer language.
 
Welding Metallurgy Ferrous Materials.pdf
Welding Metallurgy Ferrous Materials.pdfWelding Metallurgy Ferrous Materials.pdf
Welding Metallurgy Ferrous Materials.pdf
 
artificial intelligence and data science contents.pptx
artificial intelligence and data science contents.pptxartificial intelligence and data science contents.pptx
artificial intelligence and data science contents.pptx
 
CHINA’S GEO-ECONOMIC OUTREACH IN CENTRAL ASIAN COUNTRIES AND FUTURE PROSPECT
CHINA’S GEO-ECONOMIC OUTREACH IN CENTRAL ASIAN COUNTRIES AND FUTURE PROSPECTCHINA’S GEO-ECONOMIC OUTREACH IN CENTRAL ASIAN COUNTRIES AND FUTURE PROSPECT
CHINA’S GEO-ECONOMIC OUTREACH IN CENTRAL ASIAN COUNTRIES AND FUTURE PROSPECT
 
Properties Railway Sleepers and Test.pptx
Properties Railway Sleepers and Test.pptxProperties Railway Sleepers and Test.pptx
Properties Railway Sleepers and Test.pptx
 
CompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURS
CompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURSCompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURS
CompEx~Manual~1210 (2).pdf COMPEX GAS AND VAPOURS
 
ISPM 15 Heat Treated Wood Stamps and why your shipping must have one
ISPM 15 Heat Treated Wood Stamps and why your shipping must have oneISPM 15 Heat Treated Wood Stamps and why your shipping must have one
ISPM 15 Heat Treated Wood Stamps and why your shipping must have one
 
LLM Fine Tuning with QLoRA Cassandra Lunch 4, presented by Anant
LLM Fine Tuning with QLoRA Cassandra Lunch 4, presented by AnantLLM Fine Tuning with QLoRA Cassandra Lunch 4, presented by Anant
LLM Fine Tuning with QLoRA Cassandra Lunch 4, presented by Anant
 
AI assisted telemedicine KIOSK for Rural India.pptx
AI assisted telemedicine KIOSK for Rural India.pptxAI assisted telemedicine KIOSK for Rural India.pptx
AI assisted telemedicine KIOSK for Rural India.pptx
 
一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理
一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理
一比一原版(CalArts毕业证)加利福尼亚艺术学院毕业证如何办理
 
22CYT12-Unit-V-E Waste and its Management.ppt
22CYT12-Unit-V-E Waste and its Management.ppt22CYT12-Unit-V-E Waste and its Management.ppt
22CYT12-Unit-V-E Waste and its Management.ppt
 
官方认证美国密歇根州立大学毕业证学位证书原版一模一样
官方认证美国密歇根州立大学毕业证学位证书原版一模一样官方认证美国密歇根州立大学毕业证学位证书原版一模一样
官方认证美国密歇根州立大学毕业证学位证书原版一模一样
 
哪里办理(csu毕业证书)查尔斯特大学毕业证硕士学历原版一模一样
哪里办理(csu毕业证书)查尔斯特大学毕业证硕士学历原版一模一样哪里办理(csu毕业证书)查尔斯特大学毕业证硕士学历原版一模一样
哪里办理(csu毕业证书)查尔斯特大学毕业证硕士学历原版一模一样
 

Fault detection based on novel fuzzy modelling

  • 1. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 1 Fault detection based on novel fuzzy modelling Zhen Luo 1,2, Huajing Fang 2 ,Xiaoyong Liu 2 and Xuquan Chen 1 ,2 1 Guangxi Special Equipment Supervision and Inspection Institute, Guangxi 530219, P. R. China 2 Department of Control Science and Engineering, Huazhong University of Science and Technology, Wuhan 430074, P. R. China ABSTRACT The Fault detection which is based on fuzzy modeling is investigated. Takagi-Sugeno (TS) fuzzy model can be derived by structure and parameter identification, where only the input-output data of the identified system are available. In the structure identification step, Gustafson-Kessel clustering algorithm (GKCA) is used to detect clusters of different geometrical shapes in the data set and to obtain the point-wise membership function of the premise. In the parameter identification step, Unscented Kalman filter (UKF) is used to estimate the parameters of the premise’s membership function. In the consequence part, Kalman filter (KF) algorithm is applied as a linear regression to estimate parameters of the TS model using the input-output data set. Then, the obtained fuzzy model is used to detect the fault. Simulations are provided to demonstrate the effectiveness of the theoretical results. KEYWORDS Unscented Kalman filter, Kalman filter, data driven, Gustafson-Kessel clustering algorithm, fuzzy modeling, fault detection 1. INTRODUCTION Since fault detection (FD) &diagnosis technique is essential to improve the safety and reliability of dynamic systems, recently more and more attention has been paid to FD. Many FD methods have been developed to detect and identify sensor and actuator faults, such as analytical redundancy [1-3], a neural network [4], parameter identification method based on Fourier Transform [5], testing the covariance matrix of the innovation sequence [5], testing the eigen values of the sample covariance matrix [6]. All of these methods mentioned above are based upon model. Models with good accuracy are necessary to improve the correct diagnostic of faults. However, with the rapid development of industrial technology, the modern industrial processes have become more and more complicated and large-scale. It is thus difficult to construct an effective and explicit physical model to characterize these dynamical systems. Sometimes, it is even impossible to model nonlinear systems by analytical equations [7]. Here, the fuzzy modeling has been extensively used to model complex nonlinear system through a set of measured input- output data [8]. The Takagi-Sugeno (TS) model, which uses the fuzzy modeling technique, approximate the nonlinear system by smoothly interpolating affine local models. Each local model contributes to the global model in a fuzzy subset of the space characterized by a membership function [9]. Then, based on the established system model, FD can be carried out. A TS fuzzy model is usually constructed in three steps:
  • 2. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 2 Step 1: fuzzy clustering Among various clustering algorithm, the Gustafson-Kessel clustering algorithm (GKCA) [10] has been widely studied and applied by many researchers. The Gustafson-Kessel (GK) algorithm, which is the fuzzy generalization of the Adaptive Distance Dynamic Clusters algorithm, searches for ellipsoidal clusters. An advantage of the GK algorithm over Fuzzy C-means (FCM) [11] is that GK can detect clusters of different shape and orientation in one data set. Multidimensional fuzzy sets can be derived by clustering in the product space, but they are generally difficult to be dealt with. In order to solve this problem, projected one-dimensional fuzzy sets are usually preferred. Step 2: estimate the parameters of the premise’s membership function In order to obtain a fuzzy model, the premise’s membership functions must be expressed in a form that allows computation of the membership degrees for input data not contained in the data set. Each update estimation of the parameter vector corresponding to a nonlinear equation is computed from the previous estimate and the new input data (here the input data are the point- wise values of the membership function). To achieve this step, Unscented Kalman filter (UKF) is used to approximate the point-wise defined membership functions by some suitable nonlinear functions, for example normal function. The UKF as an improvement to the extended Kalman filter (EKF) is one of the most widely used approach to analyze the stochastic nonlinear systems [12]. This method is based on the unscented transform (UT) technique, a mechanism for propagating mean and covariance through a nonlinear transformation [13, 14]. The state vector is represented by a minimal set of carefully chosen sample points, called sigma points, which approximate the posterior mean and covariance of the Gaussian random variable with a second order accuracy[15, 16]. In contrast, the linearization technique used in the EKF can only achieve first order accuracy. Generally, the prediction accuracy of UKF is better than EKF when the model is highly nonlinear. The UKF has the same level of computational complexity as that of EKF, both of which are within the order ( 3) O L . Furthermore, Since the nonlinear models are used without linearization, the UKF is not necessary to compute the Jacobian or Hessians matrices [17, 18]. Step 3: estimation of the TS parameters For linear dynamic systems with white process and measurement noise, the Kalman filter is an optimal estimator [19]. The KF is used as a linear regression to efficiently choose the parameter values of the consequent part (TS parameters) of the fuzzy model from the input-output data of the identified system. [20] proposes a fuzzy-modeling scheme combining GKCA and KF. GKCA is used to detect clusters of different geometrical shapes in the data set and to obtain the point-wise membership functions of the premise. After that, a KF is first used to estimate the parameters of the premise’s membership function. Then, the KF is also used as a linear regression to efficiently choose the parameter values of the consequent part (TS parameters) of the fuzzy model from the input-output data of the identified system. KF process is a linear recursive minimum mean-square estimation procedure. However, it can only deal with linear equation. In this sense, in order to obtain the linear part, α -cut of the considered membership function is taken. Obviously, there are some drawbacks with this algorithm
  • 3. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 3 In order to obtain membership function for the premise’s fuzzy sets, the multidimensional fuzzy set defined point-wise in the row of the partition matrix is projected onto the regressor. Division of each membership function into two sets ( sets are obtained for each regressor and sets are obtained for all regressors). It needs large amount of computation The selection of α relies entirely on experience. Too much or too small will influence on membership function approximation. Moreover, the cut data causes the loss of useful information. The straight line which is used to approximate each set is not very good fit of the data. In this paper, we point-wise define the premise membership function for the regressor, each membership function as a sets (c sets are obtained for each regressor and cn sets are obtained for all regressors). This will halve the amount of computation. Obviously, it is more accurate to approximate the point-wise defined membership function by some suitable nonlinear function. UKF is utilized to approximate the nonlinear function. A common FD approach is to keep tracking residuals of measurement and compare them against a set threshold value. The residual can be generated by the comparison of actual sample points and estimated sample points . The structure of this paper is as follow. In section 2, the TS fuzzy models principle is summarized. Section 3 gives new fuzzy modeling algorithm. Fault detection is presented in section 4. Section 5 presents some simulation results. Conclusions are given in section 6. 2. TS FUZZY MODELS In this section, the TS fuzzy models principle is summarized. The TS fuzzy model can represent or model any unknown nonlinear system ( ) y f x = , on the basis of some available input-output data 1 2 [ , ..., ]T t t t nt x x x x = and t y . In the TS fuzzy model, the rule consequents are crisp function of the model inputs i R : if x is ( ) i A x , then 1,2,..., T i i i y a x b i c = + = (1) where is the dimensional input invariable,is the output invariable. and are the dimensional TS parameters. denotes the th rule and is the number of rules in the rule base. is the premise multivariable membership function of the th rule. is the parameter vector of the th rule Multi dimensional fuzzy sets can be derived by clustering in the product space. But it is generally difficult to be interpreted, so projected one-dimensional fuzzy sets are usually preferred. The TS fuzzy system is described as a set ofc fuzzy rules where the i th rule is as follows: i R : if 1 x is 1 1 ( ) i A x and …and n x is ( ) in n A x , then 1,2,..., T i i i y a x b i c = + = (2) The final output of the TS fuzzy model for an arbitrary input sample can be calculated using the following expression:
  • 4. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 4 1 1 ( )( ) ( ) c T i i i i c i i x a x b y x β β = = + = ∑ ∑ ) (3) where ( ) i x β represent the firing strength of the i th rule, and it has calculated by the following equation: 1 ( ) ij n i A j x β µ = = ∏ (4) ij A µ is the membership function of the fuzzy set ij A . 3. NEW FUZZY MODELING ALGORITHM The loop of establishing and training T-S fuzzy model by Gustafson-Kessel clustering algorithm (GKCA) and UKF is as follows: Fig.1 Flow chat of the new fuzzy modelling Yes Start c=2 GKC KF for consequence parameters UKF for premise membership function Performanc es End No c=c+ 1
  • 5. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 5 A detailed description of the algorithm is provided next. 1. From the input–output sequences 1 {( , )}N k k t x y = , partition the data into a set of local linear submodels by using GKCA in the product space X Y × . 2. Obtain the membership functions for the premise variables by using cluster projections and UKF. Estimate the consequent parameters by KF algorithm. 3.1 FUZZY CLUSTERING Fuzzy clustering is an important tool to identify the structure in data. The Gustafson-Kessel(GK) algorithm, which is the fuzzy generalization of the Adaptive Distance Dynamic Clusters algorithm, searches for ellipsoidal clusters. An advantage of the GK algorithm over Fuzzy C- means (FCM) is that GK can detect clusters of different shape and orientation in one data set. The original objective functional for the GK algorithm is the following: 1 1 ( ) ( ) c N m T it t i i t i i t J z v M z v µ = = = − − ∑∑ (5) The following restrictions hold 1 1 1 N m it t t N µ = = ≤ ≤ ∑ (6) where is a positive-definite symmetric matrix related to the covariance matrix of the th prototype, is a weighting exponent that determines the fuzziness of the resulting cluster(for a crisp , fuzzy model , but typically ). is the number of data points, is the number clusters, is the th data point, is the th cluster centre, is the degree of the membership of the th data point in the th cluster centre. 3.2 PREMISE MEMBERSHIP FUNCTION UKF as an improvement to extended Kalman filter (EKF) is the most widely used approach to analyze the stochastic nonlinear system and show good performance in many cases[12]. In this sense, we propose to use UKF as a nonlinear regression as follow: consider cn sets, each set represents the nonlinear part of the point-wise set of a certain premise’s membership function (for example Gauss type membership function). So we obtaincn parameter vector. In each set, we will have j N data (samples), where j denotes the j th set. Then, each set can be modeled by the following measurement equation
  • 6. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 6 2 2 ( , ) 1,2... ( ) exp( ) 1,2... 2 j j j j t t t j j j t t j j A h x v j cn x d v t N θ σ = + = − = − + = (7) where [ , ] j j j d θ σ = is the parameters vector, j t v is the measurement noise, j N is the number of data(samples) in the j th set and the j denotes the j th nonlinear regression. j θ will be considered as a state variable, so the state equation will be 1 1 j j j j t t t F w θ θ − − = + (8) where is the value of the state variable at the moment , is the state noise, is the state transition matrix we assume that and are uncorrelated zero mean Gaussian random vectors and their covariance matrices are and The procedure for implementing the UKF can be summarized as follows: Step 1: sigma points calculation 1 1 1 1 [ , ( ) ] j j j t t t t n P χ θ θ λ − − − − = ± + ) ) ) (9) Step 2: prediction | 1 1 j j j t t t F χ χ − − = | 1 1| 1 j j j t t t t F θ θ − − − = ) ) | 1 1| 1 j j j jT t t t t P F P F Q − − − = + ) (10) Step 3: update. | 1 | 1 ( ) j j t t t t A h θ − − = ) ) | 1 | 1 ( ) j j l t t l t t h γ χ − − = 2 | 1 | 1 | 1 | 1 0 ( )( ) j j t t n j c j j j j T l l t t t t l t t t t A A l P A A R ω γ γ − − − − = = − − + ∑ ) ) ) 2 | 1 | 1 | 1 0 ( )( ) j j t t n j c j j j j T l t t t l t t t t A l P A R θ ω χ θ γ − − − = = − − + ∑ ) ) )
  • 7. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 7 1 j j j j t t t t j j t A A A K P P θ − = ) ) | 1 | 1 ( ) j j j j t t t t t t t K A A θ θ − − = + − ) ) ) 1 | 1 j j t t j T t t t t t A A P P K P K − − = − ) ) ) (11) where, 2 0 / ( ) (1 ) c n ω λ λ α β = + + − + , / 2( ) 1...2 m c l l n l n ω ω λ λ = = + = , 2 ( 1) n λ α = − . c l ω is a set of scalar weights, and n is the state dimension; the parameter α determines the spread of the sigma points around x ) and is usually set to 1 4 1 e α − ≤ ≤ , The constant β is used to incorporate part of the prior knowledge of the distribution of x , and for Gaussian distributions, 2 β = is optimal. Often it is expected that the system parameters do not vary or, if they do, the variation is much slower than that of the system state. So, we will take j F I = . 3.3 ESTIMATION CONSEQUENT PARAMETERS In this section, KF algorithm is used to compute the consequent parameters from the data set and the estimated premise’s membership functions. From (3), we have 1 ( )( ) c T i i i i y x a x b ψ = = + ∑ ) (12) where 1 ( ) ( ) ( ) i i c i i x x x β ψ β = = ∑ is the normalized activation value of the i th rule. [ ] [ ] [ ] [ ] 1 2 1 1 ( ) 1 ( ) 1 ... ( ) 1 ... T c c c y x x x x x x a b a b ψ ψ ψ   =   ) [ ] 1 1... T c c a b a b Φ = , [ ] [ ] [ ] 1 2 ( ) 1 ( ) 1 ... ( ) 1 c x x x x x x ψ ψ ψ   Ω =   (13) The measurement equation can be taken as the following form t t t y v = Ω Φ + ) (14) Then, j θ will be considered as a state variable, so the state equation will be
  • 8. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 8 1 1 t t t F w − − Φ = Φ + (15) Now, we can use KF to estimate the TS parameter vector t Φ as follows: | 1 1| 1 t t t t F − − − Φ = Φ ) ) | 1 1| 1 T t t t t P FP F Q − − − = + 1 | 1 | 1 ( ) T T t t t t t t t t K P P R − − − = Ω Ω Ω + | | 1 | 1 ( ) t t t t t t t t t K y − − Φ = Φ + −Ω Φ ) ) ) | | 1 | 1 t t t t t t t t P P K P − − = − Ω (16) where t Φ ) is the estimated value of t Φ . Also, we will take F I = . 4. PROCESS FAULT DETECTION The fuzzy FD system is based on fuzzy models identified directly from data. A model is used to estimate the nominal output signals. y y ∆ = − ) (17) where y is the output of the system and y ) is the output of the model in normal operation. When any component of ∆ is bigger than a certain threshold, the system detects a fault. 5.SIMULATION The sampling period of training data and testing data are all 0.05h. We choose 800 data points without fault and 600 data points with fault 1 after collecting 200 data points without fault. The number of clusters 12 c = . 0 100 200 300 400 500 600 700 800 900 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x membership
  • 9. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 9 Fig.2 the red solid line indicates the actual membership function, the black dotted line indicates the premise membership function based on UKF, the blue dashed line indicates the premise membership function based on KF Fig.3 the actual sample point without fault and estimated sample point without fault Fig.4 the residual with fault Fig.5 the actual sample point without fault and the estimated sample point with fault 0 100 200 300 400 500 600 700 800 900 0.185 0.19 0.195 0.2 0.205 0.21 0.215 0.22 0.225 0.23 0.235 no fault model output sampling point variable NO 1 estimated value without fault actual value without fault 0 100 200 300 400 500 600 700 800 900 0 0.005 0.01 0.015 sampling point residual 0 100 200 300 400 500 600 700 800 900 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 sampling point variable NO 1 actual value without fault estimated value with fault
  • 10. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 10 Fig.6 the residual with fault The membership functions is given in Fig.2. The red solid line indicates the actual membership function, the black dotted line indicates the premise membership function based on UKF, the blue dashed line indicates the premise membership function based on KF. It can be seen from the figure that both approaches track the true state. However, the obtained premise membership based on UKF is more close to the true value than based on KF. It means that estimating the premise membership based on UKF is superior to based on KF. In Fig.3, we can see that the actual sample point can be tracked well by the estimated sample point. Fig.4 shows that the absolute value of the residuals which is derived by subtracting the estimated value of the actual value is lower than the threshold. The simulation results consistent with the actual case. When the fault 1 occurs at the step 200, the simulation results are given in Fig.5-6. The estimated sample points have deviated from actual sample points after 200 steps in Fig5 due to the fault, but the actual sample point can be tracked well by the estimated sample point after the step 500 because of feedback effect. The fault makes absolute value of the residuals exceed the threshold at step 206 which is given in Fig.6. I.e. the fault can be detected after occurrence of 6 steps. 6.CONCLUSION In this paper, a fault detection algorithm based on data driven is proposed. The proposed algorithm is composed of two steps: (1) fuzzy modeling contain fuzzy clustering, determination of premise membership functions and TS parameters. (2) Fault detection based on the established model. In the first step, a fuzzy modeling scheme on the basis of the GKCA and UKF to estimate the premise membership, KF is used to estimate the TS parameter. The performances of the proposed algorithm are demonstrated on the TE challenge problem. 0 100 200 300 400 500 600 700 800 900 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 detection of the occurance of the fault sampling point r e s id u a l residual threshold
  • 11. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 11 REFERENCES [1] Zhang YM, Li XR, Ieee(1997)”Detection and diagnosis of sensor and actuator failures using interacting multiple-model estimator”, In: 36th IEEE Conference on Decision and Control: Dec 10-12 1997; San Diego, Ca; pp4475-4480. [2] Maybeck PS(1999)” Multiple model adaptive algorithms for detecting and compensating sensor and actuator/surface failures in aircraft flight control systems”, International Journal of Robust and Nonlinear Control, 9(14):pp1051-1070. [3] Rago C, Prasanth R, Mehra RK, Fortenbaugh R(1998), Ieee:”Failure Detection and dentification and Fault Tolerant control using the IMM-KF with applications to the Eagle-Eye UAV”. In: 37th IEEE Conference on Decision and Control: Dec 16-18 ; Tampa, Fl; pp4208-4213. [4] Alessandri A(2003)”Fault diagnosis for nonlinear systems using a bank of neural estimators”. Computers in Industry , 52(3)pp271-289. [5] Perhinschi MG, Lando M, Massotti L, Campa G, Napolitano MR, Fravolini ML(2002), Acc”On-line parameter estimation issues for the NASA IFCSF-15 fault tolerant systems”. In: 20th Annual American Control Conference (ACC): May 08-10 2002; Anchorage, Ak; pp191-196. [6] Hajiyev C, Caliskan F(2005)” Sensor and control surface/actuator failure detection and isolation applied to F-16 flight dynamic”. Aircraft Engineering and Aerospace Technology , 77(2)pp152-160. [7] Mendonca LF, Sousa JMC, da Costa J(2009)” An architecture for fault detection and isolation based on fuzzy methods” Expert Systems with Applications 36(2)pp1092-1104. [8] Amann P, Perronne JM, Gissinger GL( 2001)”Frank PM: Identification of fuzzy relational models for fault detection”. Control Engineering Practice, 9(5)pp555-562. [9] Skrjanc I, Blazic S(2005)” Agamennoni O: Identification of dynamical systems with a robust interval fuzzy model”. Automatica , 41(2)pp327-332. [10] Gustafson D, Kessel W(1979)”Fuzzy clustering with a fuzzy covariance matrix.” Proceedings of IEEE CDC, San Diego, CA,pp 761-766. [11] Nascimento S, Mirkin B, Moura-Pires F(2003)” Modeling proportional membership in fuzzy clustering”. Ieee Transactions on Fuzzy Systems , 11(2)pp173-186. [12] Xiong K, Zhang HY, Chan CW(2006)” Performance evaluation of UKF-based nonlinear filtering”. Automatica , 42(2)pp261-270. [13] Julier SJ, Uhlmann JK(1997)” A new extension of the Kalman filter to nonlinear systems”, vol. 3068. [14] Julier SJ, LaViola JJ(2007)” On Kalman filtering with nonlinear equality constraints”. Ieee Transactions on Signal Processing, 55(6)pp2774-2784. [15] Kandepu R, Foss B, Imsland L(2008)”Applying the unscented Kalman filter for nonlinear state estimation”. Journal of Process Control, 18(7-8)pp753-768. [16] Julier SJ, Uhlmann JK(2004)”Unscented filtering and nonlinear estimation”. Proceedings of the Ieee , 92(3)pp401-422. [17] Li PH, Zhang TW, Ma B(2004)” Unscented Kalman filter for visual curve tracking”. Image and Vision Computing, 22(2)pp157-164. [18] Dini DH, Mandic DP, Julier SJ(2011)”A Widely Linear Complex Unscented Kalman Filter”. Ieee Signal Processing Letters, 18(11)pp623-626. [19] Simon D, Simon DL(2010)” Constrained Kalman filtering via density function truncation for turbofan engine health estimation”. International Journal of Systems Science, 41(2)pp159-171. [20] Chafaa K, Ghanai M, Benmahammed K(2007)” Fuzzy modelling using Kalman filter”. Iet Control Theory and Applications, 1(1)pp58-64. [21] Downs J J, Vogel E F(1993)”A plant-wide industrial process control problem”. Computers and Chemical Engineering, 17(3)pp245-255.
  • 12. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 Authors Zhen Luo She was born in 1983. Sh Science and Technology. Her research interests include and fault diagnosis of networked control sys Huajing Fang was born in 1955. He is and Technology. His research interests dynamics and control of autonomous control. Xuquan Chen was born in 1984. Science and Technology. His research interest is signal processing. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 She is a Ph.D. candidate in Huazhong University of research interests include Kalman filtering state estimation networked control sys tems. was born in 1955. He is a Professor in Huazhong University of Science His research interests include complex networked control systems, mous vehicle’s swarm, fault diagnosis and fault-tolerant was born in 1984. he is a Ph.D. candidate in Huazhong University of Science and Technology. His research interest is signal processing. Circuits and Systems: An International Journal (CSIJ), Vol. 1, No.3, July 2014 12