Vibration Sensor based Intelligent Fault Diagnosis
System for Large Machine Unit in Petrochemical
Industry
Qing-Hua Zhang
Guangdong Petrochemical Equipment
Fault Diagnosis Key Laboratory
Guangdong University of Petrochemical
Technology
Maoming, China
fengliangren@vip.tom.com
Aisong Qin
College of Information Engineering
Taiyuan University of Technology
Taiyuan, China
aisong.qin@lab.gdupt.edu.cn
Lei Shu*
Guangdong Petrochemical Equipment
Fault Diagnosis Key Laboratory
Guangdong University of Petrochemical
Technology
Maoming, China
lei.shu@lab.gdupt.edu.cn
Guoxi Sun
Guangdong Petrochemical Equipment Fault Diagnosis Key
Laboratory
Guangdong University of Petrochemical Technology
Maoming, China
guoxi.sun@lab.gdupt.edu.cn
Longqiu Shao
Guangdong Petrochemical Equipment Fault Diagnosis Key
Laboratory
Guangdong University of Petrochemical Technology
Maoming, China
longqiu.shao@lab.gdupt.edu.cn
Abstract: In this paper, to satisfy the need of fault monitoring,
dynamic real time vibration monitoring and vibration signal
analysis for large machine unit in petrochemical industry, which
cannot realize real-time, online and fast fault diagnosis, an intel-
ligent fault diagnosis system is developed using artificial immune
algorithm and dimensionless indicators, innovated with a focus on
reliability, remote monitoring and practicality, and be applied to
the Third Catalytic Flue Gas Turbine in a petrochemical enter-
prise and have got good effects.
Key Words-fault diagnosis
; time-domain vibration signals;
artificial immunity algorithm; dimensionless indicators;
immune detector
I. INTRODUCTION
Fault diagnosis is an area which is gaining increasing impor-
tance in rotating machinery. Along with the continuous advance
of science and technology, the structures of rotating machinery
become increasingly to be larger scale, higher speed and more
complicated, which result in higher probability of various fail-
ure in practice. In case one of the most critical components of
machinery or equipment breakdown, it can not only cause
enormous economic loss, but also can easily lose many people's
life. It is important to enable reliable, safe and efficient opera-
tion of large-scale and critical rotating machinery, which re-
quires us to achieve accurate and fast diagnosis of fault which
has occurred.
In recent years, lots of researchers have conducted consi-
derable effort on intelligent fault diagnosis system and devel-
oped a variety of diagnosis methods, which are applied to dif-
ferent objects. In [1], Meng et al. studied the method of infor-
mation fusion based on fault tree expert system, NN diagnosis
system, and mechanism model validation system used in the
fault diagnosis for ship nuclear power plants. In [2], Liu et al.
presented a fuzzy neural network and RBF neural network to do
the distributed local diagnosis and multi-source information
fusion technology for the global integrated diagnosis, which can
diagnose more typical accidents of PWR. In [3], Chen et al.
designed a fault diagnosis system for elevators which is com-
posed of slave MCU, master fault diagnosis system and GPRS
communication system. In [4], Zhang et al. introduced a mod-
eling of the remote intelligent fault diagnosis system (RIFDS)
based on multi-Agent theory with the application to avionic
devices. In [5], Tian et al. used the kernel independent com-
ponent analysis (KICA) in the performance test and fault di-
agnosis for the gearbox. In [6], Xu et al. proposed an informa-
tion fusion method for simultaneous fault diagnosis based on
random set theory. In [7], Jing and Zhang established a neural
network model for diagnosing diesel engine faults using both
the adaptive resonance theory (ART) and the back propagation
(BP) neural network to diagnose and identify the multiple faults
that occur during the operation of a diesel engine. In [8], Zhang
et al. analyzed vibration signals of induction motors by conti-
nuous wavelet transform (CWT) which can alleviate frequency
aliasing in CWT and determine whether or not rub fault hap-
pens. In [9], Zhou and Ye presented a composite fault detection
method based on signal singularities detected by wavelet anal-
ysis. In [10], Hu et al. proposed a simultaneous fault diagnosis
method based on multi-regression least square support vector
machine (LS-SVM) model. However, afore-mentioned diag-
nosis models can be very time-consuming and complex that
cannot realize real-time, online and fast fault diagnosis.
As far as the authors are aware, the method which is based on
artificial immune algorithm and dimensionless indicators has
not yet been utilized for fault diagnosis in the field of rotating
machinery. In this paper, an intelligent fault diagnosis system
based on negative selection algorithm (NSA) of artificial im-