Fuzzy fault diagnosis based on fuzzy robust
v
-support vector classifier
and modified genetic algorithm
Qi Wu
⇑
Jiangsu Key Laboratory for Design and Manufacture of Micro-Nano Biomedical Instruments, Southeast University, Nanjing 211189, China
School of Hotel and Tourism Management, Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong
article info
Keywords:
Fuzzy
v
-support vector classifier machine
Triangular fuzzy number
Genetic algorithm
Fault diagnosis
Hybrid noises
abstract
This paper presents a new version of fuzzy support vector classifier machine (SVM) which can penalize
those hybrid noises to forecast fuzzy nonlinear system. Since there exist some problems of uncertain data
in many actual forecasting problem, the input variables are described as fuzzy numbers by fuzzy compre-
hensive evaluation. To solve the shortage of
e
-insensitive loss function for hybrid noises such as singular-
ity points, biggish magnitude noises and Gaussian noises, a novel robust loss function is proposed in this
paper. Then by the integration of the triangular fuzzy theory,
v
-SVC and robust loss function theory, fuzzy
robust
v
-SVC (FRv-SVM) which can penalize those hybrid noises is proposed. To seek the optimal param-
eters of FRv-SVC, genetic algorithm is also proposed to optimize the unknown parameters of FRv-SVC. The
results of the application in fuzzy car assembly line system diagnosis confirm the feasibility and the
validity of the FRv-SVC model. Compared with other SVC methods, FRv-SVC method has better classifier
precison for small sample with hybrid noises.
Ó 2010 Elsevier Ltd. All rights reserved.
1. Introduction
Recently, a novel machine learning technique, called support
vector machine (SVM), has drawn much attention in the fields of
pattern classification and regression estimation. SVM was first
introduced by Vapnik and his colleagues in 1995 (Vapnik, 1995).
It is an approximate implementation to the structure risk minimi-
zation (SRM) principle in statistical learning theory, rather than the
empirical risk minimization (ERM) method. This SRM principle is
based on the fact that the generalization error is bounded by the
sum of the empirical error and a confidence interval term depend-
ing on the Vapnik–Chervonenkis (VC) dimension. By minimizing
this bound, good generalization performance can be achieved.
Compared with traditional neural networks, SVM can obtain a un-
ique global optimal solution and avoid the curse of dimensionality.
These attractive properties make SVM become a promising tech-
nique. SVM was initially designed to solve pattern recognition
problems (Abbasion, Rafsanjani, Farshidianfar, & Irani, 2007;
Aydin, Karakose, & Akin, 2007; Bao, Liu, Guo, & Wang, 2005; Camci
& Chinnam, 2008; Chen, Li, Harrison, & Zhang, 2008; Hu, Cao, Xu, &
Li, 2007; Li & Shu, 2008; Lu & Li, 2007; Shieh & Yang, 2008; Wang &
Chiang, 2009; Widodo & Yang, 2007a; Widodo & Yang, 2007b; Wu,
Liu, Xiong, & Liu, 2009; Xiong, Liu, & Niu, 2005; Yang, Zhang, & Zhu,
2007). Recently, with the introduction of Vapnik’s
e
-insensitive
loss function, SVM has been extended to function approximation
and regression estimation problems (Cao, Zhang, & Hu, 2007; Chal-
imourda, Schölkopf, & Smola, 2004; Dong, Yang, & Wu, 2007; Jay-
adeva, Khemchandani, & Chandra, 2004; Min & Cheng, 2009;
Xiang, Zhou, An, Peng, & Yang, 2008; Yang, Jin, & Chuang, 2006).
In many real applications, the observed input data cannot be
measured precisely and usually described in linguistic levels or
ambiguous metrics. However, traditional support vector classifier
(SVC) method cannot cope with qualitative information. It is well
known that fuzzy logic is a powerful tool to deal with fuzzy and
uncertain data. Some scholars have explored the fuzzy support
vector machine (FSVM) (Chen et al., 2008; Shieh & Yang, 2008;
Wang & Chiang, 2009; Wu et al., 2009). There exists some uncer-
tain information in the fuzzy fault system, as influences the classi-
fier capability of the SVC models. Some SVC literatures did not
consider the optimal problem based on this uncertain information
(Abbasion et al., 2007; Aydin et al., 2007; Bao et al., 2005; Camci &
Chinnam, 2008; Hu et al., 2007; Li & Shu, 2008; Lu & Li, 2007;
Widodo & Yang, 2007a; Widodo & Yang, 2007b; Xiong et al.,
2005; Yang et al., 2007), and this will lead to the final diagnosing
results maybe not approaching the actual pattern values closely.
Moreover, some hybrid noises such as singularity points, biggish
magnitude noises and Gaussian noises exist in fuzzy fault pattern
series. The
e
-insensitive loss function can not handle better them.
And then, this paper proposes a novel fuzzy SVC, as can penalize
0957-4174/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.eswa.2010.09.101
⇑
Address: Key Laboratory for Design and Manufacture of Micro-Nano Biomedical
Instruments, Southeast University, Nanjing 211189, China. Tel./fax: +86 25
83792418.
E-mail address: wuqi7812@163.com
Expert Systems with Applications 38 (2011) 4882–4888
Contents lists available at ScienceDirect
Expert Systems with Applications
journal homepage: www.elsevier.com/locate/eswa