Journal of Loss Prevention in the Process Industries 22 (2009) 392397

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries

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

Risk-based maintenance strategy and its applications in a petrochemical

reforming reaction system
Haijun Hu a, *, Guangxu Cheng a, Yun Li a, Yiping Tang b
a
Department of Process Equipment and Control Engineering, Xian Jiaotong University, Xian 710049, PR China
b
State Key Laboratory for Manufacturing Systems Engineering, Xi0 an Jiaotong University, Xian 710049, PR China

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

Article history: Risk-based maintenance (RBM) strategy can be used in developing a cost-effective maintenance plan to
Received 27 September 2008 make nancial and safety improvements in a petrochemical system. Most maintenance is imperfect in
Received in revised form process industries. Previous works on RBM did not consider imperfect maintenance and thus could
3 February 2009
hardly meet the engineering requirements of process industries. An improved RBM approach based on
Accepted 3 February 2009
the proportional age reduction model is proposed in this paper. The parameters of the failure distribution
and maintenance effect of equipment are estimated by maximum likelihood estimation. The nancial
Keywords:
risk for each subsystem is estimated and compared with the risk criteria. Under the risk constraints, the
Imperfect maintenance
Preventive maintenance proper interval of periodic preventive maintenance is determined for the equipment in high-risk
Risk-based maintenance subsystems. The case of a reforming reaction system in a petrochemical plant is studied using the
Proportional age reduction model improved RBM approach. The results show most equipment in this system is imperfectly repaired. The
Reforming reaction system imperfect nature of the periodic preventive maintenance means it needs to be carried out more
frequently.
2009 Elsevier Ltd. All rights reserved.

1. Introduction Haddara (2003) proposed a complete framework for RBM strategy.

Moreover, the probability of the unexpected event was determined
Process systems include much equipment that often runs under by fault tree analysis. The consequence was estimated as the mean
rigorous operation conditions. Because of aging, erosion, wear or of indexes of nancial loss, system performance loss, adverse
other reasons, equipment can unexpectedly fail, resulting in large effects on human health and environmental degradation. Reverse
nancial loss and even serious accidents and environmental fault tree analysis was used to determine the maximum failure
pollution. To avoid such situations, it is necessary to improve probability of the key equipment under risk constraints. According
equipment reliability and system availability by maintenance or to the reliability limits, the interval of periodic PM for key equip-
replacement. However, maintenance costs tend to be very high, ment was obtained under the condition that the default mainte-
accounting for approximately 2050% of the total operating budget nance effect is perfect. Khan and Haddara (2004a) applied the RBM
of process systems (Tan & Kramer, 1997). Thus cost-effective strategy to offshore oil and gas processing facilities to develop
maintenance strategies are required. a detailed maintenance plan for the safe and fault-free operation of
Preventive maintenance (PM) is effective since it can reduce the facilities. Khan and Haddara (2004b) also applied the meth-
operation costs and the probability of catastrophic consequences. odology to an ethylene oxide production plant to reduce safety
PM policies, as discussed in detail by Wang (2002), have been risks. Krishnasamy, Khan, and Haddara (2005) developed a risk-
widely applied in industries and scheduled by optimizations of based maintenance strategy for a power plant to reduce large
cost, availability and risk. Recently, risk-based maintenance (RBM) nancial risks. Apeland and Aven (2006) presented a Bayesian
strategies have received increasing attention from researchers. method for risk-based maintenance optimization as an alternative
Arunraj and Maiti (2007) reviewed the works on RBM and cata- to the probabilistic frameworks.
loged the risk analysis methods and RBM strategies. Khan and The RBM strategy has been effectively developed by Khans
group and other scholars. However, the effect of imperfect main-
tenance has not been considered in previous works even though it
* Corresponding author. would seriously inuence the expected equipment reliability and
E-mail address: [email protected] (H. Hu). system risks. Technical skills, maintenance resources and time are

0950-4230/$ see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jlp.2009.02.001
 H. Hu et al. / Journal of Loss Prevention in the Process Industries 22 (2009) 392397 393

always limited in process industries. Most maintenance only 3) Consequence estimation. There are two main parts to the
repairs some of the components in failing equipment. For example, economic consequence of equipment failure (halt, output
a maintenance action is scheduled to block the leaky tubes in a heat reduction, or lower quality of output): maintenance cost and
exchanger, but it may not address the erosion of the shell. There- production loss. The maintenance cost (MC) consists of xed
fore, the equipment is not restored to a good as new state and will costs (e.g., costs of purging, material discharge, inspection, and
deteriorate faster than equipment having had perfect maintenance component replacement) and variable costs (e.g., costs of labor,
or a complete replacement. As a result, it is necessary to consider maintenance materials, and vehicle rental) associated with
imperfect maintenance in the implementation of RBM strategy in downtime, while the production loss (PL) is the product of
process industries. downtime and production loss per day.
An improved RBM approach based on the proportional age
reduction (PAR) model is presented for the scheduling of tasks in an
MC Cf DT$CV ; (1)
equipment maintenance plan for which the criteria considered are
nancial risks. An extensive joint probability density function (PDF)
of successive failures and the nal survival is proposed to estimate PL DT$PLPD; (2)
the parameters of the failure distribution and maintenance effect.
where Cf is the xed cost of maintenance ($), DT is the downtime
Economic risk criteria are proposed on the basis of maintenance
(number of days), Cv is the variable cost of maintenance per day
expenditure. A periodic and imperfect PM plan for equipment in
($ day1), and PLPD is the production loss per day ($ day1).
high-risk subsystems is established to meet the risk criteria. The
Production loss is always far greater than maintenance cost in
case of a reforming reaction system in a petrochemical plant is
process industries. It rarely happens that two pieces of equipment
studied using the improved approach.
in a subsystem fail at the same time. Therefore, the failure conse-
quence of a subsystem can be estimated by the average of the
2. RBM strategy failure consequences of equipment in the subsystem.

Nearly 80% of the total risk in a process system is caused by 20% 4) Risk estimation. The risk for each subsystem is the product of
of equipment (API, 2000). The total risk can be reduced effectively its failure probability and consequence.
by RBM strategy, which prioritizes maintenance activities on the
basis of quantied risk due to equipment failures. The high-risk
equipment is maintained with greater frequency and thorough- 2.3. Risk evaluation
ness. There are slight differences in the implementation of RBM
strategies (Arunraj & Maiti, 2007; Khan & Haddara, 2003; Krish- The purpose of risk evaluation is to decide whether the risk is
nasamy et al., 2005). Generally, the RBM strategy includes acceptable. Risk criteria dene how much risk/probability in
subsystem identication, risk estimation, risk evaluation and a certain period (always one year) is acceptable. A detailed
maintenance planning. discussion of the risk measure and criteria was conducted by
Jonkman, Van Gelder, and Vrijling (2003). Economic risk criteria
vary with applications in different industries and countries. They
2.1. Subsystem identication are determined according to the yearly maintenance expenditure of
a system in this paper.
A complex process system may consist of many subsystems
identied by their functional and operational characteristics. Each 2.4. Maintenance planning
subsystem includes interrelated equipment/components. For
example, a stand-by pump system, a group of heat exchangers, or Subsystems with unacceptable risks are identied after risk
a compressor system in a reforming system can be identied as evaluation. Equipment in these subsystems is regarded as high-
a subsystem distinctly. The subsystem is the basic unit of risk risk equipment. Proper maintenance plans should be performed to
estimation and evaluation. reduce the equipment failure probability. On the basis of the risk
criteria, the maximum acceptable failure probability of each
2.2. Risk estimation subsystem is determined rst. Khan and Haddara (2003) and
Krishnasamy et al. (2005) applied reverse fault tree analysis to
Risk estimation comprises four steps. decide the maximum acceptable failure probability of equipment.
Khan and Haddara (2004b) proposed an optimization policy to
1) Failure scenario development. The failure scenario is developed determine the maximum failure probability of equipment by the
to determine the cause and effect relations between potential minimization of the quadratic maintenance cost. The new main-
events and nal failures according to the operational, physical tenance plan was scheduled by the reliability constraints of
and logical characteristics of the system. A failure scenario equipment. The default maintenance effect in the above works
occurs when the system fails to work as required; the system was considered to be perfect, though it is always imperfect in
may halt, output may be reduced, or output may be of lower practice.
quality.
2) Probability estimation. According to the failure scenario, a fault 2.4.1. Imperfect PM model
tree can be built, which may regard system failure as the top Imperfect maintenance restores equipment to a state some-
event, subsystem failures as intermediate events and equip- where between good as new and bad as old. Imperfect main-
ment failures as bottom events. The probability of the bottom tenance models were extensively discussed by Pham and Wang
event occurring is determined from the cumulative distribution (1996). The PAR model (Shin, Lim, & Lie, 1996) is an effective age
function of the equipment. Minimum cut sets of the fault tree model that assumes imperfect maintenance reduces the period of
are explored using Boolean algebra. The probabilities of the top operation of equipment from (tk  tk1) to r(tk  tk1), where
and intermediate events occurring are calculated with the help r(0  r 1) is the improvement factor and tk is the time of the kth
of the probability function of compatible events. failure. The PAR model was widely applied owing to its exibility.
394 H. Hu et al. / Journal of Loss Prevention in the Process Industries 22 (2009) 392397

On the basis of the PAR model, Jayabalan and Chaudhuri (1992) 2.4.3. Determining the maximum acceptable failure
proposed a cost optimization maintenance policy for a system probabilities of equipment
requiring assured reliability. Shin et al. (1996) presented an esti- In a fault tree of system failure, the probability function of the
mation procedure for the PAR model to obtain distribution top event can be described as
parameters and the improvement factor. Martorell, Sanchez, and
Serradell (1999) extended the PAR model by introducing an accel- Q f q1 ; q2 ; .; qM ; (8)
erated life model to describe the effect of working conditions. The where Q is the probability of the top event occurring, f($) is the
PAR model is also used to describe imperfect maintenance in this probability function, qi is the failure probability of the ith bottom
paper. event and M is the number of bottom events.
To decrease the probability of the top event occurring, the
2.4.2. Parameter estimations for the PAR model probability of the ith bottom event should be reduced from qi to qiqi
Parameters of the failure distribution and maintenance effect (0 <qi  1, 0 <i  M). The new probability of the top event is
are estimated from maintenance records. The two-parameter
Weibull distribution is a general distribution widely used to model Q 0 f q1 q1 ; .; qi qi ; .; qM qM ; (9)
processes such as failure due to aging and wear. It is assumed
where Q0is the new probability of the top event and qi is the factor
equipment failure follows the Weibull distribution with a non-
of the probability reduction of the ith bottom event.
decreasing failure rate of l1(t) in the rst maintenance stage. After
qi is the cumulative failure probability of the ith piece of
the kth effective maintenance (the (k 1)th stage), the equipment
equipment over a certain period dened by the risk criteria. qi can
failure rate is (Shin et al., 1996)
be determined by subjective judgments and optimizations of cost,
availability and risk. If the ith piece of equipment has a failure
 a1
a t  rtk process following an exponential distribution (a 1), it will not be
lk1 t l1 t  rtk ; a  1; b > 0; (3) preventively maintained. Therefore, its corresponding qi value
b b
cannot be reduced. Generally, it is assumed there are m pieces of
where b is a scale parameter and a is a shape parameter of the equipment in a subsystem, the failure processes of which follow
distribution. Weibull distributions with a > 1. It is further supposed that PM
The joint PDF for n successive failure times is (Shin et al., 1996) reduces the cumulative failure probabilities of the m pieces of
( ) equipment by the same proportion q (qi h q, 0 < i  m). If Q0 is rst
n
Y n
Y n
X determined by the risk criterion, q can be obtained by solving
f t1 ; t2 ; .tn fk tk lk tk exp  Ak tk ; (4) equation (9) with Newton iteration. The maximum failure proba-
k1 k1 k1
bility for the ith piece of equipment over a certain period is
where lk(tk) (a/b)((tk  rtk1)/b)a1, Ak(tk) ((tk  rtk1)/b)a 
((tk1  rtk1)/b)a, and tk is the time of the kth failure. PFi* q$qi 0 < i  m: (10)
However, an observation period may not necessarily end with
a failure, but with a survival. The joint PDF should be extended to 2.4.4. Altering the equipment maintenance plan
include a survival event. We let tn1 be the end time of the obser- Periodic PM policy is applied to enhance equipment reliability. If
vation period and the periodic interval is T, the cumulative failure probability of the
 equipment is 2 3
0 failure occurs at tn1 Z t
d :
1 survival occurs at tn1 Ft 1  exp4  lsds5
0
The conditional PDF of the event that a failure occurs at tn1 is 2
k1 Z
6 X
i1*T
f tn1 ln1 tn1 expf  An1 tn1 g: (5) 1  exp4  li1 sds
i0 i*T
The survival probability at tn1 is
3
8 9 Z
> > t
< Z tn1 = lk1 sds7
 5t > kT: (11)
Rtn1 jt1 ; t2 ; .tn exp  ls  rtn ds kT
>
: tn >
;

expf  An1 tn1 g: (6) We let DT be the period associated with risk criteria (always one
year) and T* be the period of perfect maintenance. The ith piece of
Therefore, the log likelihood function of all events in [0, tn1] is equipment should have

n
Fi j$DT  minf1; j$PFi g j 1.NS (12)
X
lnLa; b; r nlna  nalnb a  1 lntk  rtk1 ( )
k1
  T *  Ns DT *
n1
andFi T *  Ns DT  min 1; PFi ; (13)
X DT
 Ak tk 1  dlna  alnb
k1 where Fi($) is the cumulative distribution function of the ith piece
a  1lntn1  rtn : 7 of equipment and NS int[T*/DT].
Generally, the longer Ti is, the greater the savings in the PM
budget is. If all equipment in the system has the same perfect repair
Maintenance records from different observation periods are period of T* (Ni 1)Ti (which is common in petrochemical plants),
collected together to obtain more accurate estimation. The the objective of maintenance planning is to seek the minimum
parameters a, b, r are obtained by maximum likelihood estimation. value of Ni for each piece of equipment.
 H. Hu et al. / Journal of Loss Prevention in the Process Industries 22 (2009) 392397 395

Combined furnance Solo furnance High purity hydrogen gas

(H201ACD) (H201B)

Steam gas

Refining gas

High pressure
absorber (D204)

Reactor Separator Separator

Reactor Reactor Reactor
(R204) (D201) (D202)
(R201) (R202) (R203)

Heat exchanger Heat exchanger

(E201A,E201B) (E204)

Air cooler Heat exchanger

(A201) (E202)
Reformate
0.3 0.4MPa steam gas
3.5MPa steam gas M
Pump Hydrogen compressor
Hydrogen Steam turbine (P201A/B) (K202A/B)
compressor (K201) (T201)
Refined naphtha

Hydrogen gas

Fig. 1. Simplied ow diagram of the reforming reaction system.

3. Application of RBM strategy to a reforming reaction system naphtha per day is processed in the system. The plant usually
halts for several months to perform a major overhaul once every
3.1. System description 35 years. In the downtime, equipment is completely examined
and repaired. The major overhaul is considered to be perfect
The improved RBM approach is applied to the reforming maintenance. Routine maintenance including lubrication, the
reaction system in Sinopecs Luoyang Petrochemical Plant. Fig. 1 removal of dust, and the replacement of seals is considered to be
is a ow diagram of the process. Nearly 2500 tons of rened minimal maintenance. Corrective maintenance (CM) follows

Table 1
Parameters of the failure rate and improvement factor.

No. Subsystem Equipment Scale parameter Shape parameter Improvement Cumulative failure probability
b/month a factor r F(t) (during 1 year)
1 1th compressor subsystem Steam turbine (T201) 19.001 2.713 0.822 0.24979
Hydrogen compressor (K201) 17.711 1.895 1 0.38011
Steam feed pipeline 120.000a 1 N/A 0.09516

2 Heating-reaction subsystem Reactors (R201/R202/R203/R204) 40.181 3.154 0.897 0.02187

Combined furnace (H201ABC) 18.842 1.833 0.917 0.35426
Solo furnace (H201B) 20.774 2.147 0.815 0.26495
Rening gas feed pipeline 150.000a 1 N/A 0.07688

3 Heat exchanger subsystem Heat exchangers (E201A/E201B) 22.802 2.171 0.770 0.21977

4 Cooler subsystem Air cooler (A201) 40.574 2.835 1 0.03114

Heat exchanger (E202) 19.546 2.129 1 0.29807

5 Separation subsystem Separators (D201/D202) 45.285 4.038 0.823 0.00468

6 Pump subsystem Pumps (P201A/P201B) 12.329 1.772 0.604 0.24354b
7 2th compressor subsystem Hydrogen compressors (K202A/K202B) 15.068 2.307 0.727 0.11265b

8 Absorber subsystem High pressure absorber (D204) 33.083 2.100 1 0.11209

Heat exchanger (E204) 25.139 1.929 0.781 0.21348

9 Instrument subsystem Control valves & Detecting Instruments N/A N/A N/A 0.58590a
a
Cited from the literature (Shen, 2006).
b
Approximate running time is 6 months.
396 H. Hu et al. / Journal of Loss Prevention in the Process Industries 22 (2009) 392397

Failed to perform as
required

Failed to supply Failed to Failed to cool Failed to separate Failed to purify

hydrogen gas heat/cool material Failed to react outcome gas from oil hydrogen gas

Hydrogen Heat Combined

exchanger Heat
Compressor Reactor R202 Furnace Air cooler Separator exchanger
K201 failed E201A failed failed H201A CD A201 failed D201 failed E204 failed
failed

Heat Heat
steam turbine exchanger Reactor R203 Solo Furnace pump system High pressure
exchanger
T201 failed E201B failed failed H201B failed failed absorber
E202 failed
D204 Failed

2th
Steam feed Reactor R201 Reactor R204 Refining gas Separator compressor Instrument
pipeline failed failed failed feed pipeline D202 failed system failed system failed
failed

Fig. 2. Fault tree of system failure.

failures. Periodic PM is applied to some key equipment. CM and detailed maintenance records for all equipment from
PM are considered to be imperfect maintenance, since they November 1998 to August 2007. The parameters of failure
generally only repair failed or faulty components; for example, distribution and improvement factor of the equipment are
the replacement of tube banks, repair of cracks in shells, and estimated from the maintenance records or taken from refer-
laying of anticorrosive coatings for vessels. A high production loss ences (Shen, 2006). The results of the parameter estimations
and maintenance cost cannot be tolerated when the renery are given in Table 1. The results show four pieces of equipment
prot decreases with the soaring of oil prices. To reduce the designated K201, A201, E202, and D204 are always perfectly
economic risk due to unexpected failures, the RBM strategy is maintained while the remaining equipment is usually imper-
applied to determine the proper interval of periodic PM for each fectly repaired.
piece of equipment at a cost of probably low maintenance
expenditure. 3.3. Risk estimation

3.2. Subsystem identication and parameter estimations The scenario of system failure (system halt, output reduction, or
lower quality of output) is developed. A fault tree of the scenario is
The system can be divided into nine subsystems according constructed (Fig. 2). Generally, the intermediate events are failures
to their functions and operational characteristics (Table 1). The of subsystems. To simplify the modeling work, the failures of three
pump subsystem and the second compressor subsystem are the subsystems (instrument subsystem, pump subsystem and second
cold stand-by system with two pieces of equipment. There are compressor subsystem) are regarded as bottom events, so there are

Table 2
Risks of subsystems (exchange rate of US$1 6.9 RMB).

No. Subsystem Failure probability Failure consequence Risk (during 1 year)/$

(during 1 year) 1 1
Cf/$ DT/day Cv/$ day PLPD/$ day Total cost/$
1 1th compressor subsystem 0.5792 2173.9 1.5 144.93 43 478 67 609 39 159
2 Heating-reaction subsystem 0.5989 7246.4 5 289.86 43 478 226 090 135 400
3 Heat exchanger subsystem 0.3912 2898.6 3 289.86 43 478 134 200 52 500
4 Cooler subsystem 0.3199 1449.3 2 289.86 43 478 88 986 28 467
5 Separation subsystem 0.0093 2898.6 4 289.86 21739 91 014 846
6 Pump subsystem 0.4278 724.6 1 72.464 0 797 341
7 2th compressor subsystem 0.2126 1449.3 1 144.93 0 1594 339
8 Absorber subsystem 0.3016 1449.3 4 144.93 14 493 60 000 18 096
9 Instrument subsystem 0.5859 724.64 1 72.464 0 797 467
 H. Hu et al. / Journal of Loss Prevention in the Process Industries 22 (2009) 392397 397

Table 3 applied to enhance the reliability of key equipment in these

Intervals of periodic imperfect PM for key equipment. subsystems during the period of major overhaul (three years). The
No. Equipment Acceptable max Interval of Interval of Adjusted minimum times of periodic imperfect PM for equipment are
failure imperfect perfect Interval calculated using the reliability constraints. The theoretical intervals
probability PM/month PM/month Ti/day
of imperfect and perfect PM for equipment are compared in Table 3.
PF*i (during 1
year)
The results imply imperfect PM should be performed more
frequently than perfect PM is so that the risk criteria can be met.
1 Steam turbine 0.2232 9 9 270
(T201) The intervals are nally adjusted according to production.
2 Hydrogen 0.3397 9 9 270
compressor (K201)
4. Conclusions
3 Reactors (R201/ 0.0097 9 9 270
R202/R203/R204)
4 Combined furnace 0.1575 3.3 3.6 100 (1) The unexpected failure of equipment results in large nancial
(H201ACD) losses in process industries. The RBM strategy uses risk criteria
5 Solo furnace 0.1178 4 5.1 120 to schedule maintenance to save capital and guarantee safety.
(H201B)
However, it requires improvement since the effect of imperfect
6 Heat exchangers 0.1456 7.2 9 210
(E201A/E201B) maintenance is not considered in maintenance planning. This
paper proposes an improved RBM strategy based on the PAR
model. The case of a reforming reaction system is studied using
the improved approach. The results show that most equipment
a total of 21 bottom events in the tree. If no maintenance is per-
in the system is imperfectly repaired and three subsystems
formed, the cumulative failure probability of equipment is deter-
have intolerable risks. The imperfect PM should be performed
mined by
more frequently than the perfect PM is to satisfy the risk
  a  criteria when other conditions are the same.
t
Ft 1  exp  : (14) (2) This work is of benet to the implementation of RBM strategy
b
in various industries when equipment maintenance is not
In the cold stand-by system with two pieces of equipment, the always perfect. Future works may focus on the reduction of
pieces run alternatively with a period of one month. The period is failure consequence besides failure likelihood. Moreover, it
always far less than the equipment failure time, so half of the would be worth making comprehensive and quantitative
system operating time is used to approximately calculate the estimations of health, environmental and asset risks.
cumulative failure probability of the equipment. The failure
probabilities of subsystems are calculated using fault tree anal- Acknowledgments
ysis. The averages of the xed costs, variable costs and downtime
of subsystem failures are derived from the maintenance records. This work was supported by the National Natural Science
The production loss can be obtained from the operating expen- Foundation of China (No. 20676109).
diture of the petrochemical plant. Equipment failure may cause
system breakdown, decrease production quantity and quality, or
References
not disturb the production process. The production losses of
subsystems are quite different. The risk of each subsystem over Apeland, S., & Aven, T. (2006). Risk based maintenance optimization: foundational
one year is calculated. The results of risk estimation are given in issues. Reliability Engineering and System Safety, 67(3), 285292.
API. (2000). API 581: Based resource document-risk based inspection (1st ed.).
Table 2.
Washington, DC: American Petroleum Institute.
Arunraj, N. S., & Maiti, J. (2007). Risk-based maintenancedtechniques and appli-
cations. Journal of Hazardous Materials, 142(3), 653661.
3.4. Risk evaluation Jayabalan, V., & Chaudhuri, D. (1992). Cost optimization of maintenance scheduling
for a system with assured reliability. IEEE Transaction on Reliability, 41(1), 2125.
According to the maintenance cost and operation data from Jonkman, S. N., Van Gelder, P. H. A. J. M., & Vrijling, J. K. (2003). An overview of
quantitative risk measures for loss of life and economic damage. Journal of
2003 to 2006, the average expenditure of subsystems, nearly Hazardous Materials, 99(1), 130.
US$36 232 per year (exchange rate of US$1 6.9 RMB), is esti- Khan, F. I., & Haddara, M. M. (2003). Risk-based maintenance (RBM): a quantitative
mated and taken as the basic risk criterion. The very high risks of approach for maintenance/inspection scheduling and planning. Journal of Loss
Prevention in the Process Industries, 16(6), 561573.
some subsystems are difcult to reduce to the average level. A
Khan, F. I., & Haddara, M. (2004a). Risk-based maintenance (RBM): a new approach for
maximum risk criterion of US$72 464 per year is proposed process plant inspection and maintenance. Process Safety Progress, 23(4), 252265.
according to advice from three experts at the plant. The Khan, F. I., & Haddara, M. R. (2004b). Risk-based maintenance of ethylene oxide
subsystem risk over one year should be reduced to US$72 464 if it production facilities. Journal of Hazardous Materials, 108(3), 147159.
Krishnasamy, L., Khan, F., & Haddara, M. (2005). Development of a risk-based
is above the maximum risk criterion, or kept unchanged if below maintenance (RBM) strategy for a power-generating plant. Journal of Loss
the base risk criterion, or else reduced to US$36 232. Therefore, Prevention in the Process Industries, 18(1), 6981.
the yearly risks of the rst compressor subsystem and the heat Martorell, S., Sanchez, A., & Serradell, V. (1999). Age-dependent reliability model
considering effects of maintenance and working conditions. Reliability Engi-
exchanger subsystem should be reduced to US$36 232, while that neering and System Safety, 64(1), 1931.
of the heating-reaction subsystem should be reduced to Pham, H., & Wang, H. (1996). Imperfect maintenance. European Journal of Opera-
US$72 464. tional Research, 94(3), 425438.
Shen, X. (2006). Study on the reliability analysis of reforming & regeneration system in
CCR for safety operation. Master thesis, Xian Jiaotong University, China (in Chinese).
Shin, I., Lim, T. J., & Lie, C. H. (1996). Estimating parameters of intensity function and
3.5. Maintenance planning
maintenance effect for repairable unit. Reliability Engineering and System Safety,
54(1), 110.
The maximum acceptable failure probabilities (one-year dura- Tan, J. S., & Kramer, M. A. (1997). A general framework for preventive maintenance
tion) of three subsystems are determined rst. The maximum optimization in chemical process operations. Computers Chemical Engineering,
21(12), 14511469.
acceptable failure probability of equipment in these subsystems is Wang, H. (2002). A survey of maintenance policies of deteriorating systems. Euro-
obtained by numeric iteration (Table 3). A periodic PM plan is pean Journal of Operational Research, 139(3), 469489.