Journal of Loss Prevention in the Process Industries 23 (2010) 622e629

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries

journal homepage: www.elsevier.com/locate/jlp

Study and application of Reliability-centered Maintenance

considering Radical Maintenance
Dacheng Li*, Jinji Gao
Engineering Research Center of Chemical Technology Safety of the Ministry of Education, Beijing University of Chemical Technology, P. O. Box 130,
15 Beisanhuan East Road, Beijing, 100029, China

a r t i c l e i n f o a b s t r a c t

Article history: Radical Maintenance (RM) takes the root causes of failures as executive objects to make maintenance
Received 22 January 2010 decisions. A Root Cause Analysis (RCA) is carried out on the function failure of equipment to perform RM and
Received in revised form to create a maintenance plan using the combined methods of Failure Mode, Effects and Criticality Analysis
13 June 2010
(FMECA), and Fault Tree Analysis (FTA). Evaluation criteria and matrixes of criticality are used to evaluate the
Accepted 18 June 2010
criticality level of failure modes. Moreover, Minimum Cut Sets (MCS) and importance calculation are applied
to analyze the fault tree quantitatively. Based on our research, the concept and the analysis process of
Keywords:
Reliability-centered Maintenance (RCM) considering RM are proposed and then applied to an actual
Reliability-centered Maintenance (RCM)
Radical Maintenance (RM)
engineering project in the petrochemical industry. The results of the application are presented through
Root cause analysis (RCA) a maintenance strategy based on the project analysis by establishing the evaluation criteria and matrixes of
Criticality analysis (CA) criticality for petrochemical rotating machines and a fault tree of compressor vibrations. The direct causes of
Fault tree the induction of vibration faults in rotating machinery are classified according to the fault mechanism and
frequency domain features in this paper. The research shows that using RM in the traditional RCM can help
assign maintenance resources rationally and improve the quality of maintenance strategies.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction 2. Concept of RM

Reliability-centered Maintenance (RCM) analyzes the functions
and failures of a system and identifies the consequences of these
failures to implement preventive measures using a standardized
logical resolution procedure (Moubray, 1997). Traditional RCM
focuses on the function failures of a system. It determines a mainte-
nance plan by considering failure probability statistics and conse-
quences; the analysis does not involve in-depth research to identify
the failure mechanism and the real causes of function failures. Radical
Maintenance (RM), on the other hand, considers the fundamental
factors that cause failures and investigates the elements affecting the
reliability and safety of a system and equipment. In this article, Failure
Mode, Effects and Criticality Analysis (FMECA), and Fault Tree
Analysis (FTA) were used to implement RCA on a system failure to
perform RM, which was then combined with RCM analysis to
improve the quality of maintenance strategies. The proposed theory
and analysis procedure of RCM considering RM were applied to an
ethylene plant in China. An ethylene refrigerant compressor was
taken as a demonstration model, and its failure modes were studied
to conduct CA and establish a fault tree.
* Corresponding author. Tel.: þ86 010 64442310, þ86 13693080880 (cell).
E-mail address: lidacheng1983@hotmail.com (D. Li).
As a type of Proactive Maintenance (PM), RM involves the
detection and prediction of the root causes of failures and then
takes measures to eliminate them or the conditions that induce
them (Gao, 2005). The executive object of RM is to identify the root
cause, which is the basic factor that may induce failures in the main
body of the fault cause, such as the components of a machine.
For example, a steam turbine with intense vibrations could be
diagnosed as having an unbalanced rotor. Conventional mainte-
nance involves conducting field dynamic balancing or dismem-
bering the unit and transporting the rotor to a factory to improve its
dynamic balance. On the other hand, RM further investigates
the reasons leading to rotor imbalance. If the fault is due to rotor
scaling, for example, flushing the rotor under low speeds can solve
the imbalance. However, a better method is to monitor and govern
the quality of water of the boiler to avoid the scaling of the rotor by
steam generated from the boiler. The RM involves any type, or
a combination, of preventive maintenance works, differing only in
specific objects (Cheng, Jia, Zhao, Jia, & Gan, 2003).
Generally, RM is not considered for all equipment because the
latter either has little effect on the system and production or does
not require high reliability. It is only for important equipment that
RM is implemented to analyze the root causes of fault, thereby

0950-4230/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jlp.2010.06.008
 D. Li, J. Gao / Journal of Loss Prevention in the Process Industries 23 (2010) 622e629
ensuring reliability and safety, as well as the maximization of
their service life, from root cause generation to function failure
occurrence (Gan, Cheng, & Jia, 2004). Moreover, RM execution
requires corresponding techniques for monitoring and detection;
relevant maintenance jobs are also indispensable. Therefore, labor
and materials are necessary for conducting RM. The relationship
between investments and benefits should be considered in deter-
mining whether or not RM must be executed. All function failures
do not have effective root causes, as some are induced by human
factors or periodic exhaustion. Although root causes occur in this
type of failure, they do not play a substantive role in the mainte-
nance method. In this situation, conventional maintenance should
be carried out by considering the function failures in the formula-
tion of the service plan.
3. Concepts and the analysis procedure of RCM
considering RM
3.1. Execution of RM
Consisting of a series of structural problem-solving methods
aimed at methodically identifying the root causes of failure, RCA
execution corrects and eliminates root causes that may minimize
failure recurrences instead of simply focusing on obvious failure
symptoms (Anthony, 2004; Cameron, Holmes, & Chen, 2008). Its
basic elements include material, equipment, environment, manage-
ment, and operating method, among others. Various RCA techniques
include Five Whys, Failure Mode and Effects Analysis (FMEA), FTA,
Pareto Analysis, Bayesian Inference, and the Ishikawa Diagram.
Of these, FMEA is a type of systematic reliability analysis that
determines all the possible failure modes of a product from bottom
to top and ensures the influence of the fault on the system (Cai &
Wu, 2004). This is followed by Criticality Analysis (CA) (Guo, Gao,
Yang, & Kang, 2009), which aims to classify every failure
mode determined by FMEA according to the associated influence of
criticality level and number/probability. Both techniques are
unified in this paper, thereby forming FMECA. FTA is a deductive
method that assumes system failures from top to bottom and
analyzes the possible reasons for a failure. As such, a combination of
the complementary FMECA and FTA techniques has been adopted
to conduct RCA. Here, the highly critical failure modes are captured
and used as the FTA objects (top event). An inverted tree diagram
illustrates the deduction of direct fault causes from top to bottom,
showing the logical relationship between hardware/software
factors (various bottom events) and system failures (top event).
Fig. 1. The flow chart of RCIMS.
623
The fault factors that may exist in the system include component
failure, environmental impact, human mistakes, and procedure
handling. After establishing the fault tree, the combination modes
and spread paths affecting the top event by each bottom event are
qualitatively analyzed to identify the possible failure modes of the
system. And the degree of the effect is calculated quantitatively to
obtain the probability of the system failure.
All Minimum Cut Sets (MCS) are determined to show the
weakest parts of the system. Many MCS and elementary events
in an FTA exist because modern systems have become larger
and increasingly complicated. In this paper, root causes were found
from FTA elementary events according to the historical data of the
equipment and the situation of the facilities in the same industry
and type. The elementary events, which associated with high
value in the importance calculation (Institute of integrated security
engineering research, 1989) by using frequency data, were
considered to be the most important to the system. Therefore, the
maintenance measures must first be taken for them.
3.2. Analysis procedure of RCM considering RM
Reliability-Centered Intelligent Maintenance System (RCIMS) is
used based on a system of monitoring and detection by applying
the concept of RCM and RM to set up a self-organizing mechanism.
The mechanism includes users, maintenance plan management,
spare parts management, and the maintenance department among
others.
Fig. 1 shows the flow chart of the RCIMS, which consists of six
parts. The first part involves basic information collection, which
includes information on the related equipment, process fluids, and
maintenance records. The second part involves system division.
The production system can be divided into several interconnected
but relatively independent sub-systems. The third part has to do
with an importance evaluation of the facilities. According to the
evaluation criteria of importance, the facilities are divided into
three risk levels, namely, high, medium, and low. The fourth part
involves FMECA, which aims to study facilities associated with
high and medium risks and then to analyze their failure modes and
causes. Here, CA is conducted for the failure modes. The fifth part
has to do with FTA, which analyzes failure modes associated with
high and medium-criticality. The sixth part involves maintenance
strategy formulation. Against the failure modes, corresponding
maintenance strategies are proposed considering the fault causes.
The three anterior parts are the same as the flow chart of traditional
RCM (Guo, Gao, & Kang, 2008). As such, this paper does not
E L M H H H E L M H H H
expound on them.

Pro bability of failure

Pro bability of failure

To improve the informationization and intellectualization of D L M M H H D L M M H H
the maintenance decision system, the researchers took the type of
petrochemical ethylene device as the unit with which to establish C L L M H H C L L M H H
a failure database of general petrochemical rotating machinery
B L L M M H B L L M M H
based on published literature. The database includes device types,
common and pivotal failure modes, and fault causes. Unified coding A L L L M H A L L L M H
with an update function was used to enhance the efficiency
I II III IV V I II III IV V
and accuracy of the FMECA and FTA. The database conforms to the
Consequence of safety Consequence of environment
actual operations of the petrochemical equipment, thereby
providing comprehensive failure mode analysis.
E M M H H H E L M M H H
When FMEA is used in RCIMS analysis, the functions of the

Pro bability of failure

Pr o b a b i l i t y o f f a i l ur e
equipment in the production process are first clearly defined. The D L M M H H D L L M M H
next step involves finding matching failure modes and causes from
the established database. If no matched record in the database exists, C L L M M H C L L L M M
the vacancy is supplemented according to the actual situation of the
B L L L M H B L L L L M
same type of equipment in the ethylene plant and the industry in
general. In this paper, a CA on each failure mode was added to the A L L L M M A L L L L L
traditional FMEA to sort the criticality of failure modes to assign
I II III IV V I II III IV V
rationally the service resources during maintenance strategy
formulation. The results of the CA for an ethylene plant are shown in Consequence of production loss Consequence of maintenance cost
the matrix. The abscissa of the matrix corresponds to the conse- Fig. 2. Matrix of criticality for petrochemical plant.
quences of failure, while the ordinate shows the possibility of failure.
The criticality evaluation was divided into four aspects: safety,
environment, production loss, and maintenance cost. Fig. 2 shows The traditional RCM procedure includes FMEA, provides the
the matrix of criticality, and Table 1 shows the corresponding criteria prerequisites for RCA, and enhances the mutual compatibility of
of evaluation. The value of the final criticality is the maximum value RCM with RM. The FTA model was applied to the FMECA results to
of four aspects as represented by. find the basic elements that caused the failure; thus, attention
Criticality ¼ MAX (safety criticality, environment criticality, was transferred from studying the main body of the fault cause to
production loss criticality, maintenance cost criticality). searching for the root cause. The top event of the fault tree and the

Table 1
Evaluation criteria of criticality analysis.

Item Level Criteria

Probability of Failure (PF)
E Occurs 5 times in a maintenance cycle
D Occurs 4 times in a maintenance cycle
C Occurs 3 times in a maintenance cycle
B Occurs 2 times in a maintenance cycle
A Occurs 0e1 time in a maintenance cycle
V More than 1 loss of life in a large accident
IV Dies immediately (including deaths from acute poisoning) or within 30 days after the accident

Consequence of Safety (CS)
III Physically disability or disfigurement; Loss of auditory, visual, or other organ function;
Serious hazard to human health
II Mild or moderate damage to the human body or face; Mild or short term disability in auditory,
visual, or other organ function; Mild or moderate hazard to human health
I No damage to the human body or face; No injuries to the auditory, visual, or other organ function;
No hazard to human health

Consequence of Environment (CE)
V Significant changes in the structure of ecosystem or losses in ecological function; Larger social and international
influence, must be corrected with the assistance of the local government and international authorities
IV Large output of pollutants; High toxicity of pollutants; Large social influence but can be resolved
with the support of local department and regional authorities
III Complex factors of pollution, many types of pollutants are produced, and the output is large;
Pollutants are either highly toxic or difficult to degrade; The company’s efforts are needed to resolve the issue
II Simple factors of pollution; Less types of pollutants are produced and the output is low;
Pollutant toxicity is lower; Can be resolved within the system
I No waste water, waste gas, waste residue, dust, odor, noise, vibration, radiation, electromagnetic waves,
or other adverse effects are produced; Within the permissible range of environment protection indices

Consequence of Production loss (CP) V No production and shutdown time over 24 h

IV No production and shutdown time from 8 to 24 h
III No production and shutdown time under 8 h
II Reduction of output or product quality
I Less or no effect

Consequence of Maintenance cost (CM) V >RMB50,000

IV RMB20,000-RMB50,000
III RMB10,000-RMB20,000
II RMB5,000-RMB10,000
I <RMB5,000
 D. Li, J. Gao / Journal of Loss Prevention in the Process Industries 23 (2010) 622e629 625

corresponding direct causes were obtained through FMECA. Then, Table 2

under the instruction of experts, the direct causes were dug deeper Criticality analysis for the ethylene refrigeration compressor.

to uncover the basic factors that may induce failures to form the Failure mode PF CS Level CE Level CP Level CM Level Criticality
elementary events of the fault tree. Abnormal vibration D II M I L V H V H H
The failure modes, direct fault causes, and root causes are High temperature D II M I L V H V H H
listed to create a maintenance strategy considering RM. Techniques of bearing
Seal leakage D III M I L V H V H H
for monitoring and detection, as well as radical measures, were
High temperature C I L I L II L III L L
then obtained by analyzing the root causes. An effective preventive of discharge
maintenance strategy and distinguishing methods for the direct Unit can not start C I L I L III M II L M
fault causes were also found. Routine maintenance, such as sched- normally
uled maintenance and breakdown maintenance, was formulated for Unit can not be jiggered B I L I L III L III L L
Abnormal shutdown C I L I L V H V H H
the general equipment and low-criticality failure modes. Finally, all Surge C III M I L V H V M M
these maintenance works were arranged in a maintenance scheme.
If definite root causes or effective maintenance scenarios against
root causes were not found, then the normal RCM analysis decision systems can not be operated normally or will have to be shutdown.
should be carried out for direct fault causes. When this happens, the entire production is disrupted.

4. An application case
In this case study, a 200,000-ton/year ethylene plant in China was
taken as a demonstration project. As an example, CA and FTA were
carried out for an ethylene refrigerant compressor that was driven by a
steam turbine. The structure form of the compressor is a three-section
centrifugal, and the flow rates from the first to the third sections are
7117, 15,920, and 27,826 m3/h, respectively. The process flow diagram
of the ethylene refrigerant system is shown in Fig. 3. The function of
the ethylene refrigerant compressor is to provide the cryogen in
three temperature grades (63  C, 75  C, and 101  C) to the plant.
If the supply of ethylene cryogen is reduced or interrupted by some
malfunctions in the compressor, the demethanization and cracking
4.1. FMECA
The failure modes of the ethylene refrigerant compressor that
could be learned from the failure database and maintenance records
were “Abnormal vibration,” “High temperature of bearing,” “Seal
leakage,” “High temperature of discharge,” “Unit can not be jiggered,”
“Unit can not start normally,” “Abnormal shutdown,” and “Surge.”
The corresponding direct causes of these failure modes were also
obtained with the guidance of experts. The CA for the ethylene
refrigerant compressor is shown in Table 2. There are four high-
criticality modes, two medium-criticality modes, and two low-criti-
cality modes for the compressor.
The ethylene plant contains 88 important facilities classified
according to their importance evaluation. A total of 549 kinds of

Fig. 3. The process flow diagram of ethylene refrigerant system.

626 D. Li, J. Gao / Journal of Loss Prevention in the Process Industries 23 (2010) 622e629

failure modes belonged to the facilities. Of these, 65 were high-
criticality modes, accounting for 12% of the total; 231 were medium-
criticality modes, accounting for 42%, and 253 were low-criticality
modes, accounting for the rest of the total.
4.2. FTA
FTA was carried out on the FMECA results. The failure modes with
high and medium-criticality were considered for the top events of FTA,
while the low-criticality failure modes were not considered for the root
cause. For example, based on the CA in Table 2, “Abnormal vibration” is
a high-criticality failure mode of the ethylene refrigeration compressor.
The direct causes that produce the vibration faults of the rotating
machinery can be divided into nine categories, namely, rotor self,
shafting, supporting system, electromagnetic force, frictional impact,
hydrodynamic excitation, resonance, component, and others. Based on
the frequency domain features, the vibration fault is divided into nine
types falling under three categories: subsynchronization, synchroni-
zation, and supersynchronization (Gao, 1993). The classifications
were based on the fault mechanism; therefore, the correspondent
relationships between fault features and their direct reasons were
quite clear. For example, precise sub-harmonic vibrations can only
occur in nonlinear systems, and only the four causes in the above
nine categories, namely, supporting system, frictional impact, hydro-
dynamic excitation and resonance, may induce it. Therefore, in engi-
neering practice, the other five causes can be excluded if the dominant
frequency of vibration is discovered as precisely sub-harmonic. This
greatly reduces the detection range of the failure cause. Fig. 4 shows the
fault tree of the failure mode “Abnormal vibration (synchronization),”
and Table 3 shows its corresponding elementary events.
The fault tree consists of 71 elementary events found to be the
root causes that induced the vibration of the ethylene refrigera-
tion compressor. Through a logical operation using Boolean
algebra, the fault tree is determined to be composed of 71 first-
order MCS-[X1],[X2],[X3],.,[X69],[X70],[X71]. Therefore, all the
71 elementary events were considered to be the most important
and the weakest parts of the fault tree system. In actual opera-
tions, controlling these aspects must be strengthened, and
necessary protective measures must be taken. Of these, the events
associated with high value of probability importance and criti-
cality importance, are considered the pivotal events, making the
implementation of RM first.
The failure frequency (times/year) of each elementary event qi
(i stands for i th elementary events) was determined by the experts
using statistical method and their experiences according to the
historical data of the equipment in this project and the facilities in
the same industry and type. The failure probability of the top event
PðTÞ can be calculated by the following equation:
PðTÞ ¼ PðE1 WE2 W.WEn Þ (1)
where E1 ; E2 ; .; En represent all MCS in the system. And the
probability importance Ipr ðiÞ and criticality importance Icr ðiÞ of the
elementary event can be calculated using the following equations:
vPðTÞ
Ipr ðiÞ ¼ (2)
vqi
qi
Icr ðiÞ ¼ Ipr ðiÞ (3)
PðTÞ

A1 A2

B1 B2 B3 B4 B5 B6
B2

X1 X2 X3 X4 X13 X14 X15 X16 X17 X24 X25 X26 X29 X30 X29 X30 X36 X41 X42 X43

C1 C2 X4 X18 X19 X20 X21 X22 X23 X27 X28 C3 C4 X37 C5 X44 C3
C3
X5

D1 D2 X31 X32 X33 X34 X35 X38 X39 X40

X6 X7 X8 X10 X11 X12

X9

A3 A4 A5 A6 A7

B7 B8 B9 B10 B11 B12 B13 B14 B15 X69 X70 B16

X45 X46 X47 X48 X49 X50 X51 X46 X55 X56 X57 X58 X59 X60 X61 X62 X64 X65 X66 X67 X59 X68 X71
C3
C6 X52 X53 X54 X63
B2 C3
C3
Symbols

Result event Basic event Logical AND

Logical OR Subtree transfer

Fig. 4. Fault tree for “Abnormal vibration (synchronization)”.

 D. Li, J. Gao / Journal of Loss Prevention in the Process Industries 23 (2010) 622e629 627

Table 3 Table 3 (continued)

Elementary events of the fault tree for “Abnormal vibration (synchronization)”.
Symbol Event description
Symbol Event description X44 Bolt loosening
T Abnormal vibration (synchronization) X45 Rotor dynamic balance without installing coupling
A1 Radial rotation vector exciting force induced by rotor self X46 Over standard centering of coupling installation
A2 Vibration induced by supporting system X47 Oversized coupling clearance by wear
A3 Exciting force induced by shafting X48 Improper selection of coupling material
A4 Frictional impact between rotor and stator X49 Improper heat treatment process for coupling
A5 Resonance X50 Low strength design of coupling
A6 Hydrodynamic excitation X51 Over standard manufacture quality of coupling
A7 Others X52 Coupling pitting/abrasion after long runtime
B1 Eccentric mass of rotor X53 Coupling easy damage by shaft deformation during installation
B2 Shaft bending X54 Aged elastic element of coupling
B3 Shaft cracking X55 Without considering change of thermal centering during installation
B4 Eccentric radial bearing X56 Centering deviation by uneven settlement of foundation
B5 Radial bearing damage X57 Oversized clearance of radial bearing
B6 Loose support X58 Over standard centering between shaft and stator
B7 Unbalanced shafting X59 Loose separator
B8 Coupling in low accuracy or damage X60 Irrational axial location dimension
B9 Bad angular centering X61 Improper adjustment of axial clearance between rotor and stator
B10 Radial dry friction between rotor and stator X62 Oversized shaft displacement by working fluid with liquid
B11 Axial partial dry friction between rotor and stator X63 Thrust bearing wear
B12 Structure resonance X64 Single unit structure design did not avoid critical speed
B13 Rotor/supporting system resonance X65 Rotor design did not avoid critical speed
B14 Multi-machine (compressor and turbine) resonance X66 Running speed of rotor close to critical speed
B15 Unsmoothed flow rate by separator inclining X67 Multi-machine (compressor and turbine) design did
B16 Influence of near vibration source not avoid critical speed
C1 Impeller scaling X68 Improper separator installation
C2 Blade damage X69 Sharp change of compressor load
C3 Casing deformation X70 Compressor evacuation
C4 Excessive string of bearing X71 Bad design of vibration isolation
C5 Bad lubrication for bearing
C6 Rotor preload by casing deformation
D1 Blade abscission
D2 Blade corrosion
X1 Irrational deign of rotor structure
X2 Uneven material of rotor The failure frequency, probability importance and criticality
X3 Rotor dynamic balance out-of-tolerance importance of each elementary event are shown in Table 4.
X4 Misplacement of connector installation
According to Eq. (1), the failure probability of the top event is
X5 Deteriorating water quality
X6 Impurity impacts blade
calculated as follows:
X7 Low strength design of blade
X8 Improper selection of blade material PðTÞ ¼ PðE1 WE2 W.WEn Þ
X9 Forging defect of blade ¼ Pðq1 þ q01 q2 þ q01 q02 q3 þ . þ q01 q02 .q070 q71 Þ
X10 Poor corrosion resistance of blade
X11 Working fluid with acidity ¼ 0:004 þ ð1  0:004Þ  0:005 þ .þ
X12 Working fluid with water þ ð1  0:004Þ  ð1  0:005Þ  .  ð1  0:00401Þ
X13 Low strength design of shaft
X14 Improper selection of shaft material  0:0041 ¼ 0:3005
X15 Improper heat treatment process for shaft
X16 Improper storage of shaft
X17 Poor thermal stability of shaft
X18 Oversized preload of shaft 4.3. Formulation of maintenance strategy
X19 Shaft natural bending after long runtime
X20 Over standard centering of casing installation Table 5 shows the RM aspect of the maintenance strategy for the
X21 Uneven expansion of shaft when heating vibration faults of the ethylene refrigeration compressor. Through
X22 Insufficient warm-up for stream turbine
FTA, B1, which is “Eccentric rotor mass,” is found to be the direct
X23 Uneven expansion of casing when heating
X24 Over standard manufacture quality of shaft cause of the failure mode “Abnormal vibration (synchronization).”
X25 Stress concentration at shaft shoulder The routine preventive maintenance is repair, rotor dynamic
X26 Easy crack in the coincidence of vibration node and shaft shoulder balance, and then strengthening the condition monitoring of the
X27 Start and stop unit frequently
rotor vibration. The fundamental fault mechanism is not consid-
X28 Shaft fatigue cracking after long runtime
X29 Over standard manufacture quality of bearing
ered, thus the same fault can be expected to happen again after
X30 Over standard centering of bearing installation some time. One of the root causes, which is screened out according
X31 Improper heat treatment process for casing to the result of importance calculation in Table 4, that induced B1 is
X32 Casing deformation after long time use X5 “Deteriorating water quality”. Meanwhile, RM supervises the
X33 Artificial impact on casing
quality of the circular cooling water and observes and adjusts
X34 Irrational design of bearing structure
X35 Mutation of working condition changes in water quality, such as pH value and conductivity, to
X36 Insufficient load capacity design of bearing name a few. Therefore, similar failures can be prevented from
X37 Bearing operation in overload occurring again. X22, “Insufficient warm-up for stream turbine,” is
X38 Insufficient injection of lubrication oil
the root cause of B2, “Shaft bending.” The FTA shows that sufficient
X39 Impurity in lubrication oil
X40 Over standard temperature, pressure of lubrication oil
warm-up can be used to eliminate the fault, thus disallowing
X41 Oversized support clearance dynamic balance. Some root causes such as X54, “Aged elastic
X42 Poor welding of anti-loosing cushion element of coupling” do not have effective RM measures to prevent
X43 Bad frame fastening their recurrence. As such, on the one hand, the quality of spare parts
 628

X1

X4
X3
X2

X6
X5

X9
X8
X7

X12
X11
X10

X15
X14
X13

X18
X17
X16

X21
X20
X19

X24
X23
X22

X27
X26
X25

X29
X28

X32
X31
X30

X35
X34
X33

X38
X37
X36

X41
X40
X39

X44
X43
X42

X47
X46
X45

X50
X49
X48

X52
X51

X55
X54
X53

X58
X57
X56

X61
X60
X59

X64
X63
X62

X67
X66
X65

X69
X68

X71
X70
Table 4

Symbol
elementary event.

0.01

0.01
0.01

0.01
0.005
0.004

0.004
0.005

0.004
0.004

0.004
0.004

0.004
0.004

0.005

0.008
0.0058

0.0043

0.0058

0.0043

0.0042
0.0074

0.0041
0.00426
0.00407
0.00435

0.00444

0.00412
0.00417

0.00404
0.00417

0.00667
0.00405
0.00588

0.00435
0.00404

0.00667
0.00444
0.00444

0.00455
0.00444

0.00571
0.00407

0.00407
0.00421

0.00488
0.00526

0.00413
0.00455

0.00416
0.00541

0.00541
0.00526

0.00571

0.00465
0.00431
0.00464

0.00571
0.00424
0.00421

0.00571
0.00412

0.00588

0.00415
0.00556

0.00419
0.00418

0.00406
0.00421

0.00401
0.00571
Frequency
Ipr ðjÞ

0.703

0.703

0.703
0.7023

0.7066
0.7066
0.7066

0.7026
0.7023

0.7025
0.7024

0.7023
0.7023
0.7023

0.7024
0.7026
0.7023

0.7023
0.7023
0.7024

0.7023
0.7024

0.7042
0.7023
0.7036

0.7026
0.7026
0.7023

0.7066
0.7042
0.7026

0.7036
0.7027
0.7026

0.7025
0.7035
0.7024

0.7029
0.7032
0.7024

0.7024
0.7027

0.7024
0.7033

0.7033
0.7032
0.7051

0.7028
0.7035
0.7025

0.7025
0.7028
0.7026

0.7035
0.7025

0.7036
0.7035
0.7024

0.7034
0.7025
0.7036

0.7025
0.7047
0.7024

0.7025
0.7025

0.7035
0.7023
0.7025

0.7024
0.7023
Icr ðjÞ

0.0117

0.0117

0.0117

0.0097

0.0098
0.00935

0.02351

0.02351
0.02351

0.00935

0.00996
0.00951
0.01017

0.00935
0.00935
0.00935

0.00975
0.01038
0.00935

0.00935
0.00963

0.00944
0.00975
0.00935

0.01563
0.00947
0.01377

0.01038
0.01017
0.00944

0.01563
0.01038

0.01064
0.01038
0.02351

0.01337
0.00951
0.01358

0.00951
0.00984

0.01141
0.01231

0.01266
0.00956
0.01064

0.01231
0.01877
0.00972

0.01337
0.01005
0.01266

0.01088
0.01337
0.01085

0.01337
0.00991
0.00984

0.01358
0.01337
0.00963

0.01301
0.01005
0.01377

0.01735

0.00977
0.00982

0.01337
0.00949
0.00984

0.00958
0.00937
Failure frequency, probability importance and criticality importance of each

Table 5
Maintenance strategy (RM part) of an ethylene refrigeration compressor.

Direct cause Feature Distinguishing method for Root cause Radical Maintenance
a similar vibration fault
Eccentric rotor mass Rotation vector Shaft vibration tends to be zero in Deteriorating water quality Strengthening the supervision of circular
exciting force low speed, but the vibration by rotor cooling water to guarantee the quality
bending is equal to the initial bending rate
Shaft bending High vibration disappears gradually with Insufficient warm-up for Sufficient warm-up for stream turbine;
extended warm-up; vibration in low stream turbine dynamic balance is not allowed
speed is equal to the initial bending rate
Unbalanced shafting None Rotor dynamic balance Dynamic balance must be done when
D. Li, J. Gao / Journal of Loss Prevention in the Process Industries 23 (2010) 622e629

without installing coupling changing a new coupling

Bad angular centering
Separator inclining
Radial bearing damage
Loose support
Axial alternating Vibration changes significantly with the load;
force the larger load the larger vibration
Circumference Vibration changes significantly with
alternating force the gas load
Alternating cycle Random vibration waves
bearing force
Vibration along the loose direction
is easily large; horizontal and vertical
critical speed are similar; measuring
amplitude of different point can
determine the loose source
Without considering the
change of thermal centering
during installation
Artificial impact on casing
Impurity in lubrication oil
Insufficient injection
of lubrication oil
Bad frame fastening
Re-centering considering the
thermal deformation
Taking protective measures to prevent
casing deformation by external force
Monitoring the lubrication system to
ensure the quality of lubrication oil
Monitoring the lubrication system to
ensure the injection rate of lubrication oil
Monitoring the vibration of the frame to
prevent loosening
 D. Li, J. Gao / Journal of Loss Prevention in the Process Industries 23 (2010) 622e629
should be ensured, and, on the other hand, preventive maintenance
work, such as changing the coupling periodically, is also
indispensable.
Maintenance strategies should be brought into the manage-
ment system of equipment informationization in an enterprise
after establishment to form a dynamic closed loop system of
maintenance management. The entire system should depend
on a correlative administration and a conformable network. The
administration is comprised of members from the company’s
leadership, equipment management department, workshop
managers, and maintenance personnel. A static and dynamic
database of facility maintenance information must be estab-
lished in the network to store, analyze, and share maintenance
data. Personnel in each department should be responsible for
their work and communicate in a timely manner on the
information platforms built in the network. Doing so will help
establish a set of information feedback mechanisms according to
the management system. On the other hand, the maintenance
strategy must be adjusted and determined periodically based
on the actual running status of equipment to achieve the
informationization and intellectualization of equipment main-
tenance management.
5. Conclusions
Compared with routine maintenance, RM takes failure root
causes as the implemental objects. The actual conditions of
equipment operation are adjusted through various monitoring and
detection techniques to correct and eliminate the basic factors that
cause failures, such as materials, equipment, the environment,
and management, among others. This is conducted to discover and
solve failures effectively and reduce the recurrence of similar fail-
ures. In addition, RCA can be carried out smoothly by combining
FMECA and FTA for failures, thus strengthening the combination of
RM with traditional RCM. The establishment of evaluation criteria
and matrixes of criticality for petrochemical rotating machines
allows for equitable distribution of service resources. The root
causes of failures can then be obtained effectively through the
foundation of a fault tree. The maintenance strategies considering
RM contain service plans for almost all basic factors that may
induce function failure. Of these, some maintenance scenarios may
be found in the conventional strategies based on the operator’s
experience. However, these maintenance works were often ignored
in the ordinal operation, because there was no specific RCA process
to make operators clearly understand the relationship between
629
failures and root causes. The factors, which should be improved
first to reduce the probability of the failure effectively, can be
determined through the comparison of probability importance and
criticality importance. In this paper, the direct causes that induce
vibration faults in rotating machinery were classified according
to the fault mechanism involved. Furthermore, the faults were
differentiated based on their characteristics in the frequency
domain. With knowledge of the dominant frequency of faults, the
possible main causes were approximately determined, and some
causes were excluded. The specific project analysis in this paper
shows that applying RM in the traditional RCM improves the
quality of the maintenance strategy and enhances the practicability
and purpose of RCM.
Acknowledgements
The authors would like to acknowledge the support of the
National Natural Science Foundation of China (Approved Grant
No.50635010 and 50975018).
References
Anthony, M. D. (2004). A statistical comparison of three root cause analysis tools.
Journal of Industrial Technology, 20(2), 1e9.
Cai, X. X., & Wu, C. Q. (2004). Application manual of modern machine design method
(1st ed.). BeiJing: Chemical Industry Press.
Cameron, M., Holmes, W., & Chen, K. (2008). The development of the SIRF
Roundtables root cause analysis method. In J. J. Gao, L. Jay, N. Jun, M. Lin, &
J. Mathew (Eds.), Proceedings of the 3rd world congress on engineering asset
management and intelligent maintenance systems conference; 2008 Oct 27e30
(pp. 207e220). Beijing: Springer.
Cheng, Z. H., Jia, X. S., Zhao, J. M., Jia, Y. X., & Gan, M. Z. (2003). Proactive mainte-
nance and RCM. Ordnance Maintenance Engineering Research, 4, 3e7.
Gan, M.Z., Cheng, Z.H., & Jia, X.S. (2004). The thought about study and application of
proactive maintenance. Annual Meeting of Repairing Technology Academic
Commitee in China Shipbuilding Engineering Institute.
Gao, J.J. (1993). Approach to mechanism and diagnosis method of vibration faults in
high speed turbomachinery.
Gao, J.J. (2005). Informatization and intellectualization of the engineering asset
management. National Conference for Device Management.
Guo, L. J., Gao, J. J., & Kang, J. X. (2008). Investigation on reliability centered maintenance
about rotating machine. In J. J. Gao, L. Jay, N. Jun, M. Lin, & J. Mathew (Eds.),
Proceedings of the 3rd world congress on engineering asset management and intelligent
maintenance systems conference; 2008 Oct 27e30 (pp. 538e592). Beijing: Springer.
Guo, L. J., Gao, J. J., Yang, J. F., & Kang, J. X. (2009). Criticality evaluation of Petro-
chemical equipment based on fuzzy comprehensive evaluation and a BP neural
network. Journal of Loss Prevention in the Process Industries, 22, 469e476.
Institute of integrated security engineering research. (1989). FTA safety engineering
(1st ed.). BeiJing: China Machine Press.
Moubray, J. (1997). Reliability-centred maintenance (2nd ed.). New York: Industrial
Press Inc.