Energy xxx (2016) 1e9

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

A framework for identification of maintenance significant items in

reliability centered maintenance
Yang Tang a, *, Qingyou Liu a, b, Jiajia Jing c, Yan Yang d, Zhengwei Zou e
a
School of Mechatronic Engineering, Southwest Petroleum University, Chengdu 610500, China
b
Xi Hua University, Chengdu 610039, China
c
Safety, Environment, Quality Supervision & Testing Research Institute, CCDE, Guanghan 618000, China
d
School of Science, Southwest Petroleum University, Chengdu 610500, China
e
School of Computer Science, Southwest Petroleum University, Chengdu 610500, China

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

Article history: Identification of Maintenance Significant Items (MSI) is one of the key phases of the Reliability Centered
Received 6 July 2015 Maintenance (RCM), which is a screening phase where the number of items for analysis is reduced. But at
Received in revised form present, there is little systematic and convenient operation method to identify the MSI. In this paper, we
25 October 2016
presented a framework for identification of the MSI through combination of quantitative analysis with
Accepted 3 November 2016
Available online xxx
qualitative analysis. Firstly, we screened out part of non-MSI through the first screening which defines a
system boundary, set up a system hierarchy tree and do a Risk Analysis (RA) of a 2th order risk matrix.
And secondly, we omitted another part non-MSI in the second screening which carried out a Failure
Keywords:
Reliability centered maintenance
Mode and Effects Analysis (FMEA) and do a RA based on a 5th order risk matrix. Moreover, we carried on
Maintenance significant items a quantitative analysis by establishing evaluation indexes and scoring standard of the MSI based on the
Fuzzy Borda count method Analytic Hierarchy Process (AHP) and Fuzzy Borda Count method (FBC). Finally, we completed a case
Risk analysis study about the drilling pump to prove the feasibility and practicality of the method. This study is helpful
Failure mode and effects analysis for the applicable and effective Preventive Maintenance (PM) tasks.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction the system, there are some effects to safety of the people and the
environment; some effects to a task completion and some effects to
RCM is a systematic analysis method for planning the PM in the the economic cost [6,7]. But most of them have no a direct influence
field of the equipment management [1]. In order to ensure suc- on the overall system. When they are eliminated in time, the only
cessful implementation of the RCM, it needs not only a scientific consequence is the cost of corrective maintenance (CM) which is
theory and method as a guide, but also effective and convenient general lower than PM [8]. Thus, it is no necessary to analyze for all
technical tools [2]. Therefore, the present research and application components when making a maintenance decisions for a complex
of the RCM attaches great importance to supporting role of math- system based on the RCM theory. Moreover, the PM in the RCM is
ematical models and applicable analysis methods which have been used only for the MSI that is just a small part of all components in a
carrying out a classified research from different angles for the RCM system [9]. But it is the qualitative analysis method that is adopted
[3]. It is emphasized that effectiveness quantitative evaluation to identify the MSI in the traditional RCM just by virtue of the
applying the mathematical model enhances the accuracy of the experience of maintenance engineer and operating personnel. In
maintaining decision of the RCM as well as plays a very active role the analysis process above, they subjectively omit the non-MSI
for promoting the application and popularization of the RCM [4,5]. based on some qualitative terms, thereby leaving part are the MSI
Because a system or equipment generally composes of a large [10,11].
number of components, all of them have its own function, failure However, on the one hand, the boundary partition between the
mode and effects. For those function failures of the components in MSI and the non-MSI is vague by only using qualitative analysis
method above. Not only it can't prevent some MSI from missing, but
also not ensure that the failure of the MSI must have an influence
* Corresponding author.
on the system. On the other hand, there are multiple influence
E-mail address: tangyanggreat@126.com (Y. Tang). factors on whether one component is a MSI in system and their

http://dx.doi.org/10.1016/j.energy.2016.11.011
0360-5442/© 2016 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Tang Y, et al., A framework for identification of maintenance significant items in reliability centered
maintenance, Energy (2016), http://dx.doi.org/10.1016/j.energy.2016.11.011
2 Y. Tang et al. / Energy xxx (2016) 1e9

weight should be different during this evaluation, but they aren't filtered out the low risk system which is judged by the 2th order
considered in the traditional RCM. Moreover, it is the importance risk matrix in Fig. 3, and then omitted no risk and low risk sub-
level of the MSI that isn't identified from the results of the quali- system. Similarly, we can rule out no risk and low risk units and
tative analysis method. Therefore, in order to scientific and effective parts. Then we obtain the relatively significant items of the system
to screen the MSI in system, it is necessary to present a systematic to prepare for the next step screening.
framework for the RCM to improve practicality and operability
based on quantitative analysis and qualitative analysis, which also 2.2. Failure mode and effects analysis and risk analysis
is the purpose of this study.
The paper is organized as follows. In Section 2 we presented the FMEA is an inductive analysis tool to analyze failure mode and
qualitative analysis method to screen out the non-MSI in overall failure effect of the critical components whose failure will lead to
system by two screening process. Based on the APH and the FBC, undesirable consequences [16]. In the FMEA, all possible failure
the quantitative analysis method of the MSI is established in Sec- modes and all the possible consequences of every component also
tion 3. Then in Section 4, a case study about the drilling pump is be analyzed and classified with the severity of its failure mode and
done to obtain its MSI and important level. Finally, Section 5 pro- frequency of occurrence [17]. Based on the system hierarchy tree
vides some discussions and conclusions. and the results of the first screening, we can performed the FMEA
for the relatively significant items to obtain their function, failure
2. Method for screening out the non-MSI mode and failure effect on the system.
However, the different components have different failure modes
2.1. Definition system and the first screening which will lead to different PoF and CoF. In order to distinguish the
CoF, we divided it into four categories, namely, safety risk (SR),
Based on the method defining the MSI in the traditional RCM environmental risk (CR), economic cost risk (ECR) and maintenance
analysis, we had performed a blue print of the first screening of the cost risk (MCR) [18]. Therefore, we combine the experience of ex-
non-MSI in all system with the 2th order risk matrix, as shown in perts and field investigation to divide out the PoF in Table 1, and the
Fig. 1 [12]. CoF of four categories, as shown in Tables 2e5 [19].
Considering that the analyzed system may be associated with In Table 6, we created a record chart for the FMEA and the RA.
other systems, we need define the boundary and function of system We can fill in the FMEA results of every component and the cor-
analyzed in the first place in Fig. 1. As we all known, the compo- responding PoF categories and CoF categories by reference to the
sition of a system is general a relatively complex structure which above relevant tables.
consists of several subsystems, a subsystem consists of several
components, a component consists of several parts [13]. In order to 2.3. The second screening based on risk matrix
illustrate specifically the composition and hierarchy of a system and
be useful to the subsequent analysis, we set up a basic structure Because the PoF and the CoF in the last step analysis are divided
frame of the system hierarchy tree in Fig. 2. into 5 categories, we work out a 5
5 risk matrix to estimate level
In order to effectively and quickly to complete initial screening of risk, as shown in Fig. 4 [20]. In this 5th order risk matrix, the level
for components of every hierarchy in the system, the 2th order risk of the risk are divided into four types including high risk (H), me-
matrix was established, as shown in Fig. 3 [14]. This 2th order risk dium risk (M), low risk (L) and no risk (N). The risk matrix is suit-
matrix determines risk level of every component including High able for the safety risk, the environmental risk, the economic cost
Risk (H), Medium Risk (M) and Low Risk (L) based on its probability risk and the maintenance cost risk of the function failures of the
of failure (PoF) and consequence of failure (CoF) [15]. According to component.
the process diagram of the first screening in Fig. 1, we can first As an analysis item may existing multiple failure modes, their

Fig. 1. The first screening process.

Please cite this article in press as: Tang Y, et al., A framework for identification of maintenance significant items in reliability centered
maintenance, Energy (2016), http://dx.doi.org/10.1016/j.energy.2016.11.011
 Y. Tang et al. / Energy xxx (2016) 1e9 3

Fig. 2. The frame of system hierarchy tree.

Table 3
ER categories.

CoF Cat CoF

A No/Slight effect
B Minor effect
C Localized effect
D Major effect
E Massive effect

Table 4
ECR categories.

CoF Cat CoF Quantitative definition

A No/Slight effect DT 2h

B Minor effect DT 2e4 h
C Localized effect DT 4e8 h
Fig. 3. 2
2 Risk matrix. D Major effect DT 8e24 h
E Massive effect DT  24 h

Table 1
Probability of failure categories.
Table 5
PoF Cat PoF Quantitative definition MCR categories.
1 Totally unlikely Between 1 in 100000 days and 5000 per days CoF Cat CoF Quantitative definition
2 Unlikely Between 1 in 5000 days and 1000 per days
3 Possible Between 1 in 1000 days and 120 per days A No/Slight effect USD  2000
4 Probable Between 1 in 120 days and 30 per days B Minor effect USD 2 K-15 K
5 Expected Between 1 in 30 days and 1 per day C Localized effect USD 15 K-50 K
D Major effect USD 50 K-100 K
E Massive effect USD  100 K

Table 2
SR categories.

CoF Cat CoF Table 6

Record chart of the FMEA and the RA.
A No/Slight injury
B Minor injury S/N Part Function Failure Mode Failure Effect PoF CoF
C Major Injury
SR ER ECR MCR
D Single fatality
E Multiple fatalities

type and level of the risk will be not the same. In order to prevent Thus, in the process of the second screening, we will omit the non-
the MSI from omitting out in the second screening, we adopt a MSI which of all failure modes is no risk or low risk for the level of
conservative principle that as long as the level of any e risks of the evaluation of four risks. And the rest is regarded as the MSI which
analysis object is high or moderate risk, it is an MSI of the system. participate in the next step analysis.

Please cite this article in press as: Tang Y, et al., A framework for identification of maintenance significant items in reliability centered
maintenance, Energy (2016), http://dx.doi.org/10.1016/j.energy.2016.11.011
4 Y. Tang et al. / Energy xxx (2016) 1e9

The level of possible influence on personnel safety should be

considered in case of a unit or component failure, and its scoring
criteria are shown in Table 8.

2) Influence on environment safety (ES)

The level of possible influence on environment pollution should

be considered, in case of a unit or component failure and its scoring
criteria are shown in Table 9.

3) Influence on system functions (SF)

The influence on function of the whole system after any sub-

system or component failure should be a major concern, and its
scoring criteria are shown in Table 10.

4) Average failure rate (FR)

The Mean Time Between Failures (MTBF) value of a unit or

component can be calculated from the operational records and
Fig. 4. 5
5 Risk matrix.
reliability database of the relevant system or component. Its scoring
criteria are given in Table 11.
3. Method for determining level of the MSI
5) Maintenance costs (MC)
3.1. Establish evaluation indexes and scoring standard
For a unit or component, its structural complexity, maintenance
According to the investigation and research in many aspects, time and spare part costs should be accounted for maintenance
this report here presents that influence factor evaluating level of costs and its scoring criteria are shown in Table 12.
the MSI of a unit or component can be boiled down to four cate-
gories, namely, reliability factor, economy factor, monitoring 6) Downtime loss (DL)
availability factor and maintainability factor. And the four cate-
gories can be subdivided nine influence factors including influence This factor refers to an economic loss arising from downtime of
on personnel safety, influence on environment safety, influence on system or component and its scoring criteria shown in Table 13.
system functions, average failure rate, maintenance costs, down-
time loss, failure monitoring availability, downtime, complexity of 7) Failure monitoring availability (MA)
maintenance, as shown in Table 7 [21]. These influence factors are
regarded as general evaluation indexes defining the level of the This factor should be assessed based on the possibility of failure
MSI. When performing a practical analysis, we can choose the monitoring, monitoring complexity and monitoring cost on a unit
categories and the number of evaluation indexes based on the self- or component and its scoring criteria shown in Table 14.
characteristics of the analysis item.
In order to achieve a combination of qualitative analysis and 8) Downtime (DT)
quantitative analysis to evaluate the level of every MSI, it is
necessary to quantify the level of influence of all factors with a Downtime includes the time (in man-hour) for idleness, main-
scoring standard. From the reviews of maintenance engineer and tenance and restart, and its scoring criteria are shown in Table 15.
operating personnel, it is proposed that the influence level of the
evaluation indexes shall be classified into three to five grades 9) Complexity of maintenance (M)
depending on respective situation and scored with a 10 points
system [21]. The level of influence and scoring criteria of all eval- This factor is subject to degree of difficulty in assembly/disas-
uation indexes are stated as follows: sembly of a unit or component, complexity of system, degree of
difficulty for implementation (including height and surrounding
1) Influence on personnel safety (PS) environment) and supply of spare parts. Its scoring criteria are
shown in Table 16.

3.2. Scoring analysis based on the APH method

Table 7
Classification of influencing factors.
Weight coefficients of the nine influence factors evaluating the
Category Influencing factor

Reliability factor Influence on personnel safety

Influence on environment safety Table 8
Influence on system functions Scoring criteria of PS.
Average failure rate
S/N Level of influence Score
Economy factor Maintenance costs
Downtime loss 1 No hazard 0e2
Monitoring availability Failure monitoring availability 2 Slight hazard 2e6
Maintenance factor Downtime 3 Greater hazard 6e8
Complexity of maintenance 4 Significant hazard 8e10

Please cite this article in press as: Tang Y, et al., A framework for identification of maintenance significant items in reliability centered
maintenance, Energy (2016), http://dx.doi.org/10.1016/j.energy.2016.11.011
 Y. Tang et al. / Energy xxx (2016) 1e9 5

Table 9 Table 16
Scoring criteria of ES. Scoring criteria of M.

S/N Level of influence Score S/N Man-hours (h) Score

1 No/Slight effect 0e2 1 <12 0e2

2 Minor effect 2e6 2 12e24 2e6
3 Major effect 6e8 3 24e72 6e8
4 Massive effect 8e10 4 >72 8e10

Table 10 level of the MSI are different in an actual analysis. In order to

Scoring criteria of SF. calculate the total score of a component we first should figure out
S/N Level of influence Score the weight coefficient of every influence factors. The Analytic Hi-
1 No impact 0e1
erarchy Process (AHP) is a theory of measurement through pairwise
2 Slight impact 1e4 comparisons and relies on the judgments of experts to derive pri-
3 declined obviously 4e7 ority scales [22]. According to the analysis of this problem, we
4 complete loss 7e10 adopted the APH to calculate the weight coefficients of the nine
influence factors. The calculating flow of the APH is given as follows
[23]:
Table 11 Step 1: Construct a judgment matrixD through pairwise com-
Scoring criteria of FR.
parisons among the influence factors,
S/N MTBF(h) Score 2 3
1 >7000 0e2
u11 u12 …… u1n
6 u21 u22 …… u2n 7
2 3000e7000 2e4 D¼6
4 ……
7 (1)
3 1000e3000 4e6 …… …… …… 5
4 300e1000 6e8 un1 un2 …… unn
5 <300 8e10
In the judgment matrix D above, uij represents a relative
importance level of comparing influence factor No. i to influence
Table 12 factor No. j, and uji represents a relative importance level of
Scoring criteria of MC. comparing influence factor No. j to influence factor No. i. Thus, the
S/N Maintenance costs (US$) Score
value of uji is reciprocal to value of uij, namely, uji$uji ¼ 1. Table 17
exhibits the scale of the relative importance level.
1 <160 0e2
Step 2: Calculate a maximum eigenvalue lmax of the judgment
2 160e1600 2e6
3 1600e8000 6e8 matrix D by a system of homogeneous linear equations as follows:
4 >8000 8e10 8
>
> ðu  lÞu1 þ u12 u2 þ / þ u1n un ¼ 0
< 11
u21 u1 þ ðu22  lÞu2 þ / þ u2n un ¼ 0
(2)
>
> ////
Table 13 :
Scoring criteria of DL. un1 u1 þ un2 u2 þ / þ ðunn  lÞun ¼ 0
S/N Economic loss Score Step 3: Work out an eigenvector relative to the maximum
1 Few 0e2 eigenvalue lmax in Eq. (3):
2 Small 2e6
3 Large 6e8 W ¼ ð u1 ; u2 ; /; un Þ (3)
4 Huge 8e10
Step 4: Check consistency of the judgment matrix D
constructed:
Table 14
Scoring criteria of MA. CR ¼ CI=RI (4)
S/N Monitoring availability Score

1 High 0e2 CI ¼ ðlmax  nÞ=ðn  1Þ (5)

2 Middle 2e6
Where,
2 Low 6e8
3 No monitoring 8e10 CR - Random consistency ratio of the judgment matrix;
CI - General consistency index of the judgment matrix;
RI - Mean random consistency index of the judgment matrix.

Table 15
Scoring criteria of MA. Table 17
The fundamental scale of absolute numbers.
S/N Man-hours (h) Score
Definition Score
1 <12 0e2
2 12e24 2e6 Equally important 1
3 24e72 6e8 Moderately important 3
4 >72 8e10 Strongly more important 5
Very strong important 7
Extremely more important 9
Intermediate more important 2, 4, 6, 8

Please cite this article in press as: Tang Y, et al., A framework for identification of maintenance significant items in reliability centered
maintenance, Energy (2016), http://dx.doi.org/10.1016/j.energy.2016.11.011
6 Y. Tang et al. / Energy xxx (2016) 1e9

For judgment matrix of 1e10 step, the value of the RI is taken as Table 19
shown in Table 18 [23]. Expert evaluation form.

Step 5: Adjust the consistency and rank the weight coefficient. Part 121 Part 122 … Part ijk
When CR < 0.01, it is evident that the consistency of the judg- Expert 1 scoring
ment matrix D is satisfactory, and meaning the weight apportion- Expert 2 scoring
ment is rational; if not, the judgment matrix D should be adjusted …
until the required consistency is satisfied. Then, every value of the Expert N scoring

eigenvector W ¼ (u1, u2,/, un) of the judgment matrix D corre- Notes: i ¼ 1,2,3…, j ¼ 1,2,3…, k ¼ 1,2,3…, N ¼ 1,2,3….
sponds respectively to the weight coefficient of every influence
factor.
Step 6: Calculate the total score of every unit or component with Step2: Calculate fuzzy times fkn and fuzzy frequency Wkn. First of
a linear weighted mathematical model as follows: all, determine the fuzzy times fkn of the n th evaluated object Dn
ranked No. k position which is given by
X
n
Index ¼ mi ai (6) X
M
i¼1 fkn ¼ dkm ðDn Þmmn (7)
m¼1
Where:
Where,

n - Quantity of influencing factors; 1 If the nth evaluated object Dn ranked No: k position
dkm ðDn Þ ¼
mi - Score of influencing factor No. i; 0 Others
ai- Weight coefficient of influencing factor No. i. And secondly, we make a fuzzy times fkn statistical table as
shown in Table 20.
In order to eliminate the subjective influence of scoring expert, Finally, determine the fuzzy frequency Wkn which reflects the
it is necessary to invite at least 3 experts to score the same system difference of the score.
or component during the actual analysis process. We have made a
table frame to store the calculation results of every system or Wkn ¼ fkn =Rn (8)
component index as shown in Table 19. P
where, Rn ¼ fkn .
k
Step3: Calculate ranking score Qk. Its meaning is that the eval-
3.3. Importance ranking based on Fuzzy Borda Count method
uated object will be transformed a ranking position into a ranking
score.
In Section 3.2, we could obtain the total scoring results of every
component by several experts. But we would find in an actual 1
analysis that the total scoring of different expert for the same Qk ¼ ðN  kÞðN  k þ 1Þ (9)
2
component is different in Table 18. Thus, we cannot directly identify
the importance level of the MSI. In order to objectively integrate the Step4: Calculate Fuzzy Borda Count value FB(Dn). That is, to
scoring of all experts and accurately analyze the relative impor- determine the Fuzzy Borda Count score of the n th evaluation
tance level and priority sorting of every component, we select the object,
FBC according to a lot investigation and research on the various X
decision methods. The Borda Count method was put forward by FBðDn Þ ¼ Wkp Qk (10)
C.de Borda in 1784 to solve the voting problem in the early time k

[24]. In the classic Borda count judges only consider which alter- In the end, we can carry out sorting from big to small for N
natives are preferred to the others. But in this study on the level of evaluated objectDn through the Fuzzy Borda Count value FB(Dn).
the MSI we present a natural extension of the classic Borda count by Then we can carry on a screening for the MSI based on the Fuzzy
allowing the experts to show numerically how much some alter- Borda Count value FB(Dn) for the actual needs of subsequent anal-
natives are preferred to the others, evaluating their preference in- ysis of the RCM.
tensities [25]. This is possible by considering fuzzy binary relations
instead of ordinary binary relations. The following pre-requisites
4. Case study
are introduced to formalize the Fuzzy Borda Count [26,27].
Set Bm(Dn) to be evaluation value of evaluated object Dn under
The drilling pump is one of the three major components of oil
the evaluation method Pm (m ¼ 1,2,3 … M; n ¼ 1,2,3 … N), Here are
drilling rig as well as a key equipment of drilling fluid circulation
the analysis steps of the FBC:
system. It includes a lot of wearing parts and its working conditions
Step1: Calculate the degree of membershipmmn. When adopting
complicated. In the process of drilling, the drilling pump inject high
the mth evaluation method, calculate membership degreemmn of
pressure drilling fluid to downhole to cool the drill bit, clean the
each one of the evaluated objectDn which belongs to “superior”, the
drilling tools and carry out the cuttings. Moreover it may be as a
formula for the calculation of which is
.
mmn ¼ Bm ðDn Þ maxfB1 ðD1 Þ; B2 ðD2 Þ; :::; BM ðDN Þg Table 20
n
Fuzzy times statistical table frame.

D1 D2 … DN

Table 18 1 f11 f12 … f1N

Standard value of mean random consistency index RI. 2 f21 F22 … f2N
: : : … :
n 1 2 3 4 5 6 7 8 9 10 N FN1 FN2 … FNN

RI 0.00 0.00 0.58 0.9 1.12 1.24 1.32 1.41 1.45 1.49 Ʃ R1 R2 … RN

Please cite this article in press as: Tang Y, et al., A framework for identification of maintenance significant items in reliability centered
maintenance, Energy (2016), http://dx.doi.org/10.1016/j.energy.2016.11.011
 Y. Tang et al. / Energy xxx (2016) 1e9 7

Fig. 5. The hierarchy tree of drilling pump. 1 Power End; 2 Fluid End; 3 Auxiliary System; 11 Crank Module; 12 Bearing Assembly; 13 Cross Head Assembly; 21 Clack Box; 22 Piston
Assembly; 23 Intake Manifold Assembly; 31 Exhaust Manifold Assembly; 32 Spary Pump; 111 Bent Axle; 112 Axle Cover; 121 Bearing; 122 Bearing Suppor; 131 Connecting Rod;132
Crosshead; 133 Guide Plate; 211 Body; 212 Bonnet; 213 Bonnet Seal; 214 Valve Seal; 221 Piston; 222 Piston Rod; 231 Inlet Line; 232 Inlet Filter; 241 Outlet Line; 242 Outlet Filter; 311
Spary Pump; 312 Oil Filter.

power fluid of the bottom hole assembly or as a signal transmission Then we obtained the FMEA and the RA record chart about the
medium. So we must ensure that drilling fluid pumped is stability drilling pump system, as shown in Table 21.
and continuous. Otherwise it will cause neither wellbore collapse According to the analysis results of the PoF and the CoF of the SR,
nor great economic loss. Therefore, the reliability of drilling pump the ER, the ECR and the MCR in Table 21, we performed the second
system is directly related to the safety of drilling. We carry on the screening for 20 components of the drilling pump system by the
management for drilling pump based on the RCM theory. In order 5th order risk matrix in Fig. 4. After this analysis process, we
to ensure the effective implementation of the RCM method for the omitted the components that their entire failure mode is no risk
drilling pump, it is necessary to identify high risk failure modes and and low risk, and in the end we obtained 9 components of the high
screen out important components of drilling pump. So that work risk and the medium risk, namely, Bent axle (111), Bearing (121),
out effective maintenance decision to reduce the failure risk of the Crosshead (132), Guide plate (133), Body (211), Bonnet (212), Valve
drilling pump and improve the reliability of the oil and gas drilling seat (214), Piston (221), and Outlet line (241).
system. We made further analysis for the result of the previous step to
Based upon the analysis process of the important degree of obtain the relative importance level by the quantitative analysis
system components in this study, we definition System and Initial method in Section 3. First of all, we chose six evaluation indexes
Screening do a first screening for the drilling pump and establish including SF, FR, MC, MA, DT and M based on the analysis results
the hierarchy tree of the drilling pump system that is filtered as above and the actual situation of field application on the drilling
shown in Fig. 5. pump. And secondly, according to the scoring method of the rela-
In order to perform the second screening for the drilling pump tive important level in Section 3.2, we obtained the relative
system, we made the FMEA and the RA for these system compo- important level of the six evaluation indexes in Table 22.
nents based upon its hierarchy tree with the first screening in Fig. 5. We calculated the weight coefficient of the six evaluation

Table 21
The record chart of the FMEA and the RA of the drilling pump system.

S/N Part Function Failure mode Failure effect Pof Cof

Sr Er Ecr Mcr

111 Bent axle Converting rotary motion into reciprocating Breaking off Shut down and economic losses 2 A A C E
motion
112 Axle cover Plug shaft hole Loose and fall off Pollution of lubricating oil 2 A A A A
121 Bearing Reduce friction Locked dead and roller Shut down 3 A A B C
shedding
122 Bearing support Supporting role Bearing loose Damage to components 2 A A B B
131 Connecting rod Transmit reciprocating force Bending deformation Damage to components 2 A A B C
132 Crosshead Transform motion way Wear or broken Shut down 4 A A B B
133 Guide plate Guide for motion Wear or deformation Shut down 4 A A B B
211 Body Bearing high pressure liquid Crack or leak Hurt people and pollute the environment 3 B B C D
212 Bonnet Plug bonnet hole Thorn leakage Pump pressure reduced and slurry pulsation 4 A B A B
213 Bonnet seal Seal bonnet hole Leakage Pump pressure reduced and slurry pulsation 4 A A A A
214 Valve seat Open and close hydraulic channel Performance degradation Pump efficiency lower 4 A A B B
221 Piston Suction and discharge of slurry Corrosion or deformation Reduce the drilling speed 4 A A B C
222 Piston rod Connecting rod Connection instability Slurry hydraulic sharp fluctuation 2 A A B B
223 Cylinder sleeve Seal Seal failure Pump pressure reduced 3 A A B A
seal
231 Inlet line slurry suction passage Incoming air Drop in delivery 2 A A B C
232 Inlet filter Filtering slurry Blocked inlet line Drop in delivery and discharge pressure to 4 A A A A
reduce
241 Outlet line slurry discharge passage Pipe fracture Hurt people and pollute the environment 3 B B B C
242 Outlet filter Filtering slurry Block outlet line or fall off Suppress pump or downhole accidents 2 A A C A
311 Spary pump Pump into the lubricating oil Spray volume insufficient Bearing burnt 3 A A B B
312 Oil filter Filtering impurities Block pipeline Poor lubrication 2 A A B A

Please cite this article in press as: Tang Y, et al., A framework for identification of maintenance significant items in reliability centered
maintenance, Energy (2016), http://dx.doi.org/10.1016/j.energy.2016.11.011
8 Y. Tang et al. / Energy xxx (2016) 1e9

Table 22 Based upon scoring standard of the six evaluation indexes in

Scoring of pairwise comparisons of influence factors. Section 3.1, we first invited 5 experts to scoring for 9 components
SF FR MC MA DT M including Bent axle (111), Bearing (121), Crosshead (132), Guide
SF 1 7 5 6 7 7
plate (133), Body (211), Bonnet (212), Valve seat (214), Piston (221)
FR 1/7 1 1/4 1/3 1/2 2 and Outlet line (241). Then the total score of every component was
MC 1/5 4 1 3 4 4 calculated with Eq. (6) based on the weight coefficient of the six
MA 1/6 3 1/3 1 1/2 2 evaluation indexes in Table 22. We listed the total score of every
DT 1/7 2 1/4 2 1 4
component that are scored by 5 experts in Table 24.
M 1/7 1/2 1/4 1/2 1/4 1
In order to objectively and accurately analyze the relative
importance level of the MSI and priority sorting of every compo-
nent, we begin use the FBC to analyze the scoring results in
Table 23
Weight coefficient and priority ordering. Table 24. According to computation steps of the FBC in Section 3.3,
we calculated the statistic results of the fuzzy times fkn in Table 25.
S/N Index Weight coefficient Priority
We calculated the Fuzzy Borda Count value FB(Dn) of every
1 SF 0.5186 1 component in Table 26 based on the statistic results of the fuzzy
2 FR 0.0531 5 times fkn in Table 25.
3 MC 0.2043 2
4 MA 0.0825 4
According to the FB value in Table 26, we obtained importance
5 DT 0.1013 3 level ranking of the 9 components: Bent axle, Body, Piston, Valve
6 M 0.0402 6 seat, Outlet line, Crosshead, Bearing, Guide plate, Bonnet. At the
same time, we can perform again screening for the MSI in the
drilling pump system by setting a quantitative standard as well as
indexes with the AHP in Section 3.2. Construct the judgment matrix we can carry on to determine the PM tasks according to the
D by obtaining the relative importance level in Table 22. importance level of the MSI.
2 3
1 7 5 6 7 7
6 1=7 1 1=4 1=3 1=2 27 5. Conclusion
6 7
6 1=5 4 1 3 4 47
D¼6
6 1=6
7 (11)
6 3 1=3 1 1=2 277 RCM has been seen as a reliability and risk management
4 1=7 2 1=4 2 1 45 methodology used to determine PM requirements and optimize
1=7 1=2 1=4 1=2 1=4 1 maintenance system for a physical asset in its operating environ-
ment. The RCM methodology can be described by the following
The maximum eigenvalue lmax of the judgment matrix D is
three phases: Identification of MSI; Assignment of applicable and
6.4576 and the eigenvector is obtained as follow,
effective PM tasks for the MSI; Implementation and update of the
W ¼ ð u1 ; u2 ; /; un Þ PM tasks. Due to the MSI regarded as the object of the PM tasks, we
¼ ð 0:5186 0:0531 0:2043 0:0825 0:1013 0:0402 Þ have presented the framework for identification of the MSI in this
study based on combination of quantitative analysis with qualita-
(12)
tive analysis. It is a systematic analysis method that adopts the
AndCR ¼ 0.0876 < 0.1, the consistency of the judgment matrix D quantitative analysis thought to omit gradually the non-MSI in
is satisfactory, and the weight apportionment of six evaluation in- system to obtain the MSI. Moreover, the framework also analyzes
dexes is rational. The weight coefficient and priority ordering of the the importance level of the MSI and ranking them by the quanti-
six evaluation indexes are shown in Table 23. tative analysis method including two mathematical models: the

Table 24
The total scoring result of 9 components.

Bent axle (111) Bearing (121) Crosshead (132) Guide plate (133) Body (211) Bonnet (212) Valve seat (214) Piston (221) Outlet line (241)

Exp 1 8.5559 3.9346 4.4272 3.4869 5.5354 2.4975 5.6556 5.8668 4.3519
Exp 2 7.6304 4.065 4.7943 4.3326 6.3811 3.303 5.9088 5.4384 4.5462
Exp 3 6.8195 3.9346 4.3361 3.4258 5.2131 2.6519 5.3424 6.3323 4.2635
Exp 4 7.3461 4.8372 5.1585 4.6968 6.1543 3.2344 5.6036 5.7451 4.4367
Exp 5 6.9370 3.179 3.704 3.038 6.6916 3.1518 6.1426 6.4443 4.5462

Table 25
The statistic results of the fuzzy times fkn.

Bent axle (111) Bearing (121) Crosshead (132) Guide plate (133) Body (211) Bonnet (212) Valve seat (214) Piston (221) Outlet line (241)

1 5 0 0 0 0 0 0 0 0
2 0 0 0 0 2.64 0 0 1.62 0
3 0 0 0 0 0 0 2.21 1.71 0
4 0 0 0 0 1.41 0 1.65 0.71 0
5 0 0 1.97 0 0 0 0 0 0.66
6 0 0.66 1.05 0 0 0 0 0 1.74
7 0 1.5 0 1.21 0 0 0 0 0
8 0 0.53 0 0.91 0 0.45 0 0 0.6
9 0 0 0 0.44 0 1.55 0 0 0

Ʃ 5 2.69 3.02 2.56 4.05 2 3.86 4.04 3

Please cite this article in press as: Tang Y, et al., A framework for identification of maintenance significant items in reliability centered
maintenance, Energy (2016), http://dx.doi.org/10.1016/j.energy.2016.11.011
 Y. Tang et al. / Energy xxx (2016) 1e9 9

Table 26
The Fuzzy Borda Count value of 9 components.

Bent axle (111) Bearing (121) Crosshead (132) Guide plate (133) Body (211) Bonnet (212) Valve seat (214) Piston (221) Outlet line (241)

FB 36 3.34 8.61 1.77 23.47 0.23 18.44 22.75 5.88

APH and the FBC, which ensures the accuracy and rationality of the capability shortage in power systems. Energy 2016;113:1316e24.
[10] Ahmadi Alireza, So €derholm Peter, Kumar Uday. On aircraft scheduled main-
analysis results. Based upon the quantitative importance level of
tenance program development. J Qual Maint. Eng 2010;16(3):229e55.
the MSI, we can screen again for the MSI to simplify the PM tasks. [11] Cai Baoping, Liu Yonghong, Ma Yunpeng, Huang Lei, Liu Zengkai. A framework
A case from the oil and gas industry is presented to demonstrate for the reliability evaluation of grid-connected photovoltaic systems in the
the applicability of the framework for identification of the MSI. The presence of intermittent faults. Energy 2015;93:1308e20.
[12] Dehghanian Payman, Fotuhi-Firuzabad Mahmud, Aminifar Farrokh,
approach is, however, general and could also be used for other Billinton Roy. A comprehensive scheme for reliability centered maintenance
types of applications. We believe that by applying the systematic in power distribution SystemsdPart I: methodology. Power Deliv IEEE Trans
analysis method of the MSI, an improved basis can be established 2013;28(2):761e70.
[13] Aven Terje. A unified framework for risk and vulnerability analysis covering
for informing decision makers compared with the traditional both safety and security. Reliab Eng Syst Saf 2007;92(6):745e54.
method in RCM, as the importance of risk of failure mode and in- [14] Rausand Marvin. Risk assessment: theory, methods, and applications, vol. 115.
fluence factor is more adequately taken into account. John Wiley & Sons; 2013.
[15] Aven Terje, Renn Ortwin. On risk defined as an event where the outcome is
uncertain. J Risk Res 2009;12(1):1e11.
Acknowledgment [16] Sovacool Benjamin K, Kryman Matthew, Laine Emily. Profiling technological
failure and disaster in the energy sector: a comparative analysis of historical
energy accidents. Energy 2015;90:2016e27.
The work was supported by the Natural Science Foundation of [17] Matteson Schuyler. Methods for multi-criteria sustainability and reliability
China (Grant no. 51274171), the Graduate Student Innovation Fund assessments of power systems. Energy 2014;71:130e6.
of School of Mechatronic Engineering. Southwest Petroleum Uni- [18] Aven Terje, Zio Enrico. Some considerations on the treatment of uncertainties
in risk assessment for practical decision making. Reliab Eng Syst Saf
versity (CX2014BZ04).
2011;96(1):64e74.
[19] Aven Terje, Vinnem Jan Erik. On the use of risk acceptance criteria in the
References offshore oil and gas industry. Reliab Eng Syst Saf 2005;90(1):15e24.
[20] Selvik Jon T, Aven Terje. A framework for reliability and risk centered main-
[1] Rausand Marvin, Vatn Jørn. Reliability centred maintenance." Complex system tenance. Reliab Eng Syst Saf 2011;96(2):324e31.
maintenance handbook. Springer London; 2008. p. 79e108. [21] Subramanian Nachiappan, Ramanathan Ramakrishnan. A review of applica-
[2] Høyland Arnljot, Rausand Marvin. System reliability theory: models and sta- tions of Analytic Hierarchy Process in operations management. Int J Prod Econ
tistical methods, vol. 420. John Wiley & Sons; 2009. 2012;138(2):215e41.
[3] Jardine Andrew KS, Tsang Albert HC. Maintenance, replacement, and reli- [22] Saaty Thomas L. Decision making with the analytic hierarchy process. Int J
ability: theory and applications. CRC press; 2013. Serv Sci 2008;1(1):83e98.
[4] Carazas FG, de Souza Gilberto Francisco Martha. Risk-based decision making [23] Saaty Thomas L. "Analytic hierarchy process." encyclopedia of operations
method for maintenance policy selection of thermal power plant equipment. research and management science. Springer US; 2013. p. 52e64.
[24] García-Lapresta Jose  Luis, Martínez-Panero Miguel, Meneses Luis Carlos.
Energy 2010;35(2):964e75.
[5] Van der Weide Johannes AM, Pandey Mahesh D, van Noortwijk Jan M. Dis- Defining the Borda count in a linguistic decision making context. Inf Sci
counted cost model for condition-based maintenance optimization. Reliab Eng 2009;179(14):2309e16.
Syst Saf 2010;95(3):236e46. [25] Zeyuan Qiu, Dosskey Michael G, Kang Yang. Choosing between alternative
[6] Niu Gang, Yang Bo-Suk, Pecht Michael. Development of an optimized placement strategies for conservation buffers using borda count. Landsc Ur-
condition-based maintenance system by data fusion and reliability-centered ban Plan 2016;153:66e73.
maintenance. Reliab Eng Syst Saf 2010;95(7):786e96. [26] Buscaldi Davide, Rosso Paolo. Upv-wsd: combining different wsd methods by
[7] Zidan Aboelsood, El-Saadany EF. Incorporating customers' reliability re- means of fuzzy Borda count voting. In: Proceedings of the 4th international
quirements and interruption characteristics in service restoration plans for workshop on semantic evaluations. Association for Computational Linguistics;
distribution systems. Energy 2015;87:192e200. 2007.
[8] Zhou Xiaojun, Xi Lifeng, Lee Jay. Reliability-centered predictive maintenance [27] Levner Eugene, Alcaide David, Sicilia Joaquin. Multi-attribute text classifica-
scheduling for a continuously monitored system subject to degradation. tion using the fuzzy Borda count method and semantic grades."Applications of
Reliab Eng Syst Saf 2007;92(4):530e4. Fuzzy Sets Theory. Springer Berlin Heidelberg; 2007. p. 422e9.
[9] Min CG, Park JK, Hur D, Kim MK. A risk evaluation method for ramping

Please cite this article in press as: Tang Y, et al., A framework for identification of maintenance significant items in reliability centered
maintenance, Energy (2016), http://dx.doi.org/10.1016/j.energy.2016.11.011