Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: Petrochemical facilities and plants require essential ongoing maintenance to ensure high levels of reli-
Received 17 November 2011 ability and safety. A risk-based maintenance (RBM) strategy is a useful tool to design a cost-effective
Received in revised form maintenance schedule; its objective is to reduce overall risk in the operating facility. In risk assess-
24 April 2012
ment of a failure scenario, consequences often have three key features: personnel safety effect, envi-
Accepted 21 May 2012
ronmental threat and economic loss. In this paper, to quantify the severity of personnel injury and
environmental pollution, a failure modes and effects analysis (FMEA) method is developed using
Keywords:
subjective information derived from domain experts. On the basis of failure probability and consequence
Risk-based maintenance
Fault tree analysis
analysis, the risk is calculated and compared against the known acceptable risk criteria. To facilitate the
Failure modes and effects analysis comparison, a risk index is introduced, and weight factors are determined by an analytic hierarchy
Continuous catalytic reforming plant process. Finally, the appropriate maintenance tasks are scheduled under the risk constraints. A case study
of a continuous catalytic reforming plant is used to illustrate the proposed approach. The results indicate
that FMEA is helpful to identify critical facilities; the RBM strategy can increase the reliability of high-risk
facilities, and corrective maintenance is the preferred approach for low-risk facilities to reduce main-
tenance expenditure.
Ó 2012 Elsevier Ltd. All rights reserved.
0950-4230/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jlp.2012.05.009
Y. Wang et al. / Journal of Loss Prevention in the Process Industries 25 (2012) 958e965 959
and Giacchetta (2009). Khan’s team has made significant contri- and standby structures. This applies for a subsystem as well, which
butions to the development of the RBM strategy. Khan and Haddara consists of several facilities. The facility is regarded as the basic
(2003) proposed a complete framework for the RBM strategy, in element of risk assessment and evaluation.
which the probability of the unexpected event was determined
using fault tree analysis and the consequences involved the esti- 2.2. Risk assessment
mation of system performance loss, financial loss, human health
loss and environmental and/or ecological loss. On the basis of Risk can be seen as a natural consequence of refinery and
comparison results between the calculated risk and acceptable risk petrochemical production activities. It is impossible to eliminate all
criteria, intervals of preventive maintenance for key equipment risks, so risks are reduced to an acceptable level. Risk assessment
were obtained. To illustrate its applications in detail, the RBM requires the application of the appropriate techniques to analyze
strategy has also been applied to a number of situations including the risk of an unexpected failure scenario, which involves the
offshore oil and gas processing facilities, an ethylene oxide estimation of the likelihood (failure probability) and consequences
production plant, and a power plant (Khan & Haddara, 2004a; Khan (severity of the undesired failure scenario).
& Haddara, 2004b; Krishnasamy, Khan, & Haddara, 2005). Arunraj
and Maiti (2010) used risk as a criterion to select the appropriate 2.2.1. Estimation of failure probability
maintenance policy and the results showed that condition-based A failure scenario occurs when the plant fails to meet the
maintenance was suitable for high-risk equipment and corrective production requirements, including break down, reduction of
maintenance for low-risk equipment. To achieve the minimum risk output, inferior quality of output and even the occurrence of acci-
for the expected life of a liquefied natural gas plant, a risk-based dents such as fire and explosion. The failure scenario can be iden-
shutdown management strategy was considered (Keshavarz, tified according to the process features, operational conditions and
Thodi, & Khan, 2012). the safety management status of the plant. Considering a failure
In an RBM strategy, the risk of a particular failure scenario is scenario as a top event, a fault tree is constructed to delineate the
defined as the product of likelihood and consequences. These ways in which the top event can occur. Logic signs including “AND”
consequences have three key features: personnel safety effect, and “OR” gates are employed to graphically represent the relations
environmental threat and economic loss. Although economic loss among the top event, intermediate events and basic events.
can be directly measured in terms of money, it is a rather sensitive The failure probability of each basic event can be determined by
matter to assign financial cost to personnel injury and environ- statistics regarding adequate data associated with faults and acci-
mental pollution. Failure modes and effects analysis (FMEA) is dents, which are mainly collected from the operation, accident and
a widely used, effective tool to identify and assess how potential maintenance records of industrial plants (Hoyland & Rausand,
failures can affect the performance of a process. In the present 1995). When the failure probability of each basic event is known,
paper, to judge the severity of the personnel safety effect and the failure probability of the top event can be calculated using an
environmental threat, a semi-quantitative method of FMEA is AND/OR gate.
developed based on subjective information derived from experts.
The failure probability of a facility can be calculated using Y
n
a Weibull distribution model; then the risk can be calculated by PðANDÞ ¼ Pi (1)
multiplying failure probability with consequences. In risk evalua- i¼1
tion, the concept of a risk index is introduced to facilitate the
comparison between the calculated risk and the known acceptable Y
n
effects of a failure mode; the Occurrence is related to the proba- where MC is the maintenance cost, Cf is the fixed cost of the failure
bility of a failure mode occurring; the Detection indicates a failure’s scenario ($US), Cv is the variable cost per hour of downtime ($h1),
visibility that is the attitude of a failure mode to be identified by DT is the downtime, which includes the total time the plant would
controls and inspections. Then the Risk Priority Number (RPN) is be out of service as a result of the failure scenario (hours).
obtained from the product of these three parameters to measure The production loss (PL) can be estimated by multiplying
risk and severity of a failure mode. downtime (DT) and production loss per hour (PLPH, $h1).
components into high-octane components in the presence of a facility breaks down; and preventive maintenance activities are
a catalyst under appropriate conditions of temperature, pressure carried out at regular intervals in key facilities while they are
and time. A simplified flow diagram of an existing industrial plant is operating, its idea is to retain the key facilities in specific condition
shown as Fig. 1. The plant consists of two heat exchangers, four (Phan & Wang, 1996). Preventive maintenance activities often
reforming furnaces (Furnace A, C and D are contained in consist of a well-defined set of technical tasks, for example,
a combined furnace H201ACD, and Furnace B is a solo furnace inspection, repair, replacement, cleaning, lubrication, adjustment,
H201B), four series-connected moving-bed reformers, a catalyst and alignment. Normally, preventive maintenance is more effective
regenerator and other auxiliary facilities. Nearly 700,000 tones of than corrective maintenance because it is always to keep the key
refined naphtha are processed per year. Considering the whole CCR facilities in an available condition, so that large loss by unpredict-
plant as a system, it can be divided into four subsystems according able fails can be avoided. The purpose of this paper is to determine
to process characteristics; each subsystem consists of several the appropriate interval of periodic preventive maintenance using
facilities as shown in Fig. 2. the RBM strategy.
To ensure the safety and high reliability of the plant, mainte-
nance activities are necessary. The plant usually halts production 3.2. Risk assessment
for about 30 days to perform a major overhaul once every 3 years
even though no fault or failure has occurred; during the downtime Considering a situation where the “CCR system failed to perform
all the facilities are thoroughly examined and repaired. During the as required” as a top event, a fault tree is constructed (Fig. 3), in
operation cycle, corrective maintenance activities occur when which the failures of facilities are basic events.
The two-parameter Weibull distribution is a general failure First, 15 experts were selected from the operation, maintenance
distribution, Eq. (9), which can be fitted to a wide range of failure and management departments of the CCR plant. Then FMEA was
processes (except those that increase at an exceptionally fast rate). used to identify the failure modes of each facility as listed in Table 3;
these 15 domain experts were asked to make their own judgments
a1
a t on the severity of the personnel safety effect and environmental
lðtÞ ¼ , a>1; b>0 (9)
b b threat for each failure mode.
The opinions of the experts were obtained in an anonymous
where l(t) denotes the failure rate at time t, a is the shape manner. The various scenarios were conveyed to the experts via
parameter and b the scale parameter.
It is assumed that the facility failure process follows the two-
parameter Weibull distribution with a non-decreasing failure rate Table 2
l(t). The failure probability at a given time t can be determined from Parameters of the probability distribution function.
the cumulative distribution probability of the facility, Eq. (10). The Subsystem Facility Shape Scale MTBF Cumulative
mean time between failure (MTBF) can be found from Eq. (11). parameter a parameter /month distribution
b/month probability
a
t F(t) (during
FðtÞ ¼ 1 exp (10) 1 year)
b
Reforming E201A 2.1751 20.7732 18.3967 0.2615
reaction E201B 2.1751 20.7732 18.3967 0.2615
1 H201ACD 1.8267 17.5451 15.5953 0.3932
MTBF ¼ b,G 1 þ (11)
a H201B 2.0536 20.1409 17.8438 0.2920
R201 3.2659 17.5451 15.7204 0.2511
where G(x) is the gamma function. R202 3.2659 17.5451 15.7204 0.2511
The values of the two parameters a and b can be obtained from R203 3.2659 17.5451 15.7204 0.2511
R204 3.2659 17.5451 15.7204 0.2511
failure and maintenance records using maximum likelihood esti- Hydrogen K201 1.7765 12.8924 11.4696 0.5854
mations (Shin, Lim, & Lie, 1996). In our previous research, the recycling T201 2.8897 14.1356 12.5969 0.4636
probability distribution function of the reforming reaction unit of K202 A/Ba 2.2255 10.3058 9.1274a 0.2592a
the CCR plant was obtained by focusing on inadequate maintenance P202 A/Ba 1.7411 11.1343 9.9150a 0.2888a
Catalyst R301 1.9425 16.6722 14.7804 0.4102
(Hu, Cheng, Li, & Tang, 2009). In the present study, the two
regeneration
parameters a and b of Weibull distribution for facilities have been Separation of gas E202 2.0795 17.7460 15.7185 0.3581
revised as new operation data have been added. The cumulative and liquid D201 3.1255 27.2509 24.3797 0.0741
distribution probability during 1 year and MTBF of each facility can D204 2.1101 32.4832 28.7681 0.1151
now be calculated; the results are listed in Table 2. a
The facility is redundant; and the given time is 6 months.
Y. Wang et al. / Journal of Loss Prevention in the Process Industries 25 (2012) 958e965 963
Table 3
Risk of personnel safety effect and environmental threat.
Facility Failure mode Severity of consequences Personnel safety risk R1 Environmental risk R2
questionnaires. The experts gave score values for each failure (product of the CCR plant) and the production capacity per hour of
mode. The questionnaire responses were collected and reviewed, the plant. The results of the production losses and maintenance
then shared anonymously by the experts. The experts were also costs for each facility failure are given in Table 4.
able to revise their judgments. This process was repeated until
a consensus was reached among the experts. The personnel safety
3.5. Risk evaluation
risk and environmental risk was then calculated as the product of
the failure probability and consequence scores. The results are
Following discussion with domain experts based on the risk
listed in Table 3.
evaluation standards of Health Safety and Environment (HSE)
The fixed costs, variable costs and downtimes of the plants
management systsm of the CCR plant, the acceptable risk criteria of
(because of facility failures) were obtained from maintenance
the personnel safety effect and environmental threat can be set at
records. It is rare for two or more facilities within a subsystem to
1.2. Based on yearly maintenance expenditure and the production
break down at the same time. Therefore the production loss per
losses of the plant, the acceptable risk criterion of economic loss for
hour was obtained by identifying the price of the reformate
each of the four subsystems is calculated as $47,244, $23,622,
$47,244 and $23,622, respectively.
Table 4
Risk of economic loss.
Acknowledgments
In this paper, AHP was used to determine the three weight
factors as follows: personnel safety effect (w1 ¼ 0.3958), environ-
This work was supported by a grant from the National High
mental threat (w2 ¼ 0.3958) and economic loss (w3 ¼ 0.2084). The
Technology Research and Development Program of China (No.
detailed calculation procedures of AHP are not presented here. The
2009AA04Z402) and the Youth Science Foundation of China (No.
risk index calculations for each facility are shown in Table 5.
51105295).
Three facilities were found to violate the risk criterion:
combined furnace (H201ACD), solo furnace (H201B) and catalyst
regenerator (R301), which are underlined in Table 5. Therefore,
References
a new maintenance strategy must be developed to shorten the
interval of preventive maintenance for these three facilities. Alsyouf, I. (2007). The role of maintenance in improving companies’ produc-
tivity and profitability. International Journal of Production Economics, 105(1),
70e78.
3.6. Maintenance planning Apeland, S., & Aven, T. (2000). Risk based maintenance optimization: foundational
issues. Reliability Engineering & System Safety, 67(3), 285e292.
The approach that we adopted to lower the risk level of critical Arunraj, N. S., & Maiti, J. (2007). Risk-based maintenance e techniques and appli-
cations. Journal of Hazardous Materials, 142(3), 653e661.
facilities to meet the acceptable risk criterion was to reduce the Arunraj, N. S., & Maiti, J. (2010). Risk-based maintenance policy selection using AHP
failure probability via shortened intervals of preventive mainte- and goal programming. Safety Science, 48(2), 238e247.
nance. For a facility with a risk index of less than 1.0, the preventive Backlund, F., & Hannu, J. (2002). Can we make maintenance decisions on risk
analysis? Journal of Quality in Maintenance Engineering, 8(1), 77e91.
maintenance interval can be extended accordingly to reduce Bertolini, M., Bevilacqua, M., Ciarapica, F. E., & Giacchetta, G. (2009). Development of
maintenance expenditure. risk-based inspection and maintenance procedures for an oil refinery. Journal of
First, the maximum acceptable failure probability of each facility Loss Prevention in the Process Industries, 22(2), 244e253.
Dekker, R. (1996). Application of maintenance optimization models: a review and
was obtained through performing a reverse process of risk
analysis. Reliability Engineering and System Safety, 51(3), 229e240.
assessment and evaluation as mentioned above, to reduce risk level Farquharson, J. A., & Choquette, F. (5e9 August, 2002). Using QRA to make main-
below the acceptable risk criteria. Second, regarding reliability tenance trade-off decisions. In ASME pressure vessels and piping conference,
constraints, the corresponding maintenance interval was calcu- Vancouver, BC, Canada.
Harnly, J. A. (1998). Risk-based prioritization of maintenance repair work. Process
lated using the two-parameter Weibull distribution model. The Safety Progress, 17(1), 32e38.
results are listed in Table 6. Ho, W. (2008). Integrated analytic hierarchy process and its applications e a liter-
The results show that the calculated maximum acceptable ature review. European Journal of Operational Research, 186(1), 211e228.
Hoyland, A., & Rausand, M. (1995). System reliability theory: Models and statistical
failure probability of the five facilities, including 2nd hydrogen methods. New York: Wiley.
compressor (K202 A/B), pumps (P202 A/B), cooling unit (E202), Hu, H., Cheng, G., Li, Y., & Tang, Y. (2009). Risk-based maintenance strategy and its
separator (D201) and high pressure absorber (D204), are equal to applications in a petrochemical reforming reactor system. Journal of Loss
Prevention in the Process Industries, 22(4), 392e397.
1.0. This means that even though the five facilities do break down, Jovanovic, A. (2003). Risk-based inspection and maintenance in power and process
the consequence is slight and acceptable. Therefore, a reasonable plants in Europe. Nuclear Engineering and Design, 226(2), 165e182.
maintenance strategy for the five facilities is not preventive Keshavarz, G., Thodi, P., & Khan, F. (2012). Risk-based shutdown management
of LNG units. Journal of Loss Prevention in the Process Industries, 25(1),
maintenance but corrective maintenance on the occasions that 159e165.
they break down. Khan, F. I., & Haddara, M. M. (2003). Risk-based maintenance (RBM): a quantitative
approach for maintenance/inspection scheduling and planning. Journal of Loss
Prevention in the Process Industries, 16(6), 561e573.
4. Conclusions Khan, F. I., & Haddara, M. M. (2004a). Risk-based maintenance of ethylene oxide
production facilities. Journal of Hazardous Materials, 108(3), 147e159.
Khan, F. I., & Haddara, M. M. (2004b). Risk-based maintenance (RBM): a new
This paper presents a methodology for scheduling an RBM
approach for process plant inspection and maintenance. Process Safety Progress,
strategy for petrochemical plants. When quantifying the risk of 23(4), 252e265.
a failure scenario, a consequence analysis involves three features: Kjellen, U., Motet, G., & Hale, A. (2009). Resolving multiple criteria in decision-
personnel safety effect, environmental threat and economic loss. making involving risk of accidental loss. Safety Science, 47(6), 795e797.
Krishnasamy, L., Khan, F., & Haddara, M. (2005). Development of a risk-based
However, it should be noted that it is a sensitive matter to measure maintenance (RBM) strategy for a power-generating plant. Journal of Loss
personnel injury and environmental pollution in monetary terms; Prevention in the Process Industries, 18(1), 69e81.
Y. Wang et al. / Journal of Loss Prevention in the Process Industries 25 (2012) 958e965 965
Kumar, U. (1998). Maintenance strategies for mechanized and automated mining Rausand, M. (1998). Reliability centered maintenance. Reliability Engineering &
system: a reliability and risk analysis based approach. Journal of Mines Metals System Safety, 60(2), 121e132.
and Fuels, 46(11), 343e347. Shin, I., Lim, T. J., & Lie, C. H. (1996). Estimating parameters of intensity function and
Montgomery, R. L., & Berratella, C. (5e9 August, 2002). Risk-based maintenance: maintenance effect for repairable unit. Reliability Engineering & System Safety,
a new vision for asset integrity management. In n ASME pressure vessels and 54(1), 1e10.
piping conference, Vancouver, BC, Canada. Tan, J. S., & Kramer, M. A. (1997). A general framework for preventive maintenance
Patton, J. D. (1983). Preventive maintenance. New York: Instrument Society of America. optimization in chemical process operations. Computers & Chemical Engineering,
Phan, H., & Wang, H. (1996). Imperfect maintenance. European Journal of Operational 21(12), 1451e1469.
Research, 94(3), 425e438. Vaidya, O. S., & Kumar, S. (2006). Analytic hierarchy process: an overview of
Rao, B. (1996). Handbook of condition monitoring. Amsterdam: Elsevier. applications. European Journal of Operational Research, 169(1), 1e29.