Professional Documents
Culture Documents
QRA On Leakage Failure of Bayesian Network
QRA On Leakage Failure of Bayesian Network
QRA On Leakage Failure of Bayesian Network
a r t i c l e i n f o a b s t r a c t
Article history: Submarine pipeline is the major transportation way of subsea oil and gas production. Due
Received 28 February 2016 to the internal and external factor, the failure probability of submarine pipeline is increas-
Received in revised form 26 May ing, which could lead to the spill accidents of oil and gas. Efficient risk analysis is vital
2016 for preventing and mitigating such potential accident. This paper presents a risk-based
Accepted 6 June 2016 accident model to conduct quantitative risk analysis (QRA) for leakage failure of subma-
Available online 16 June 2016 rine pipeline. Firstly, we employ bow-tie method to model the causal relationship between
pipeline leakage and potential accident scenarios. Subsequently, in order to overcome the
Keywords: difficulties of bow-tie in modeling uncertainties and conditional dependency, a Bayesian
Quantitative risk analysis network model for pipeline leakage is developed through mapping from the former bow-
Leakage failure tie. Meanwhile, an object-oriented Bayesian network that has a smaller and more clarified
Submarine oil and gas pipeline structure is also constructed by modularizing the primary Bayesian network. Eventually,
Bow-tie model the probability updating is implemented in risk analysis using Bayesian network when a
Bayesian network new evidence or observation occurs, and an experience learning from accident precursor
Accident precursor data data is also conducted through Bayesian approach. The proposed accident model based on
Bayesian network can provide a more case-specific and realistic analysis consequence com-
pared to bow-tie method, since it could consider the common cause failures and conditional
dependency in accident evolution process of pipeline leakage.
© 2016 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
1. Introduction leakage incidents of oil and gas pipeline between 2001 and
2011 in the UK continental shelf. According to International
The rising demand of energy accelerated the oil and gas explo- Association Oil and Gas Producer (OGP UK, 2010), the alloca-
ration from ocean, countries like American, Canada, China or tion of failure mechanisms for submarine pipeline is as shown
Japan have made exploration work in deep water area which in Table 1. Once a leakage occurs, it may cause severe fire
reached 3000 m water depth. Submarine pipeline is the major and explosion due to accident escalation, and pose a threat
transportation way of subsea oil and gas after exploration to human safety, environment, asset, and reputation (Fang
and exploitation. However, because of scouring caused by cur- et al., 2014). For instance, the leakage accidents of submarine
rent and wave, third-party damage and seaquake, or design pipeline of China Zhuhai and BP as shown in Fig. 1 are two
defect, submarine pipelines has a relative high probability representative examples.
of leakage failure (Zhang et al., 2011). As per the UK Health The leakage failure risk of submarine pipeline is unable to
and Safety Executive (HSE UK, 2011), there are about 1978 be eliminated, but preventive and mitigative measures can be
∗
Corresponding author.
E-mail address: offshore@126.com (G. Chen).
http://dx.doi.org/10.1016/j.psep.2016.06.006
0957-5820/© 2016 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
164 Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 163–173
Fig. 3 – SHIPP conceptual accident model (adapted from Rathnayaka et al., 2011).
166 Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 163–173
failure of submarine oil and gas pipeline may occur due to pipeline system, a safety barrier model for pipeline leakage
either internal or external factors. External factors include shown in Fig. 7 is developed. The occurrence of catastrophic
corrosion, damage caused by external loads, fatigue damage accident is usually due to failure of safety barriers succes-
caused by suspended span and natural disaster, while internal sively, not failure of single barrier (Skogdalen and Vinnem,
factors include material defect, weld-seam defect and auxil- 2012). In this case, there are six main safety barriers preventing
iaries failure (Jin et al., 2004). Corrosion is a main hazard factor escalation of a leakage event. The monitor and alarm barrier
that can lead to leakage, the corrosion may occur in internal is able to perceive the change of flux and pressure of pipeline
or outer surface of a pipeline, the reasons of corrosion include in normal work condition, and send alarm signals. The leak-
not removing the corrosion gas and impurities, not adding age detection system based on fiber optic is widely gaining
corrosion inhibitor, not pigging regularly, anticorrosive coat- acceptance in oil and gas industries (Aljaroudi et al., 2015b),
ing failure and Cathodic protection failure (Fang et al., 2014). and has two kinds of failure modes, one is that it does not per-
The external loads damage is also an important risk factor ceive when a leakage really occurs, the other is that it sends
when pipeline locates in offshore area, the dropped objects false alarm when a leakage does not occur. The emergency
from passing ships, Anchoring work of ships and fishing gear shutdown is implemented when the leakage is detected suc-
interaction etc. may cause serious damage to pipeline. In addi- cessfully, this process includes stopping pump and cutting
tion, if burial depth of pipeline is not enough, the suspended off globe valves, this process failure may involve no stopping
span will appear due to a harsh subsea environmental con- pump, no stopping pump timely, no cutting off globe valves or
dition, in this case, fatigue failure may happen because of no cutting off globe valves timely. Negative pressure protection
vortex-induced vibration of high frequency (Zhu and Chen, system of submarine pipeline locates between ashore valve
2009). Natural disaster including subsea earthquake, seabed chamber and outlet of pipeline, which is used in submarine
movement and typhoon is also not a negligible factor for caus- pipeline of hangzhouwan in china presently (Fang, 2014). It
ing pipeline leakage. The material and weld-seam defect are can pump the crude oil in pipeline into storage tanks or trans-
inherent defect caused by design or operation factor during portation station of oil, and keep internal pressure of pipeline
designing and constructing stages of pipeline, also a vulner- lower than pressure of ambient seawater in order to prevent
able spot that may lead to pipeline leakage when there are crude oil spill into seawater after emergency shutdown. The
external forces existing. Besides, failure of pipeline auxiliaries remaining crude oil in pipeline after emergency shutdown
such as flange, valve etc. due to design fault or aging may also will spill into seawater once the negative pressure protection
lead to a leakage occurrence or out of control. The primary system fail to work. Note that the spill volume of crude oil
events of FT shown in Fig. 6 and their probabilities (OGP UK, depends on the volume of pipeline in the case of failure of
2010; Participants, 2002; Hu et al., 2012) are as listed in Table 4. negative pressure protection system. When a spill accident
In the present study, the probabilities in Table 4 are referred of oil and gas happens, the emergency response barrier must
for educational purpose and only used to demonstrate the work, which includes plugging leakage hole of pipeline, main-
proposed methodology. tenance of pipeline and cleaning up spill oil. In the process
of cleaning up spill oil and gas, an ignition prevention bar-
4.2. Safety barrier of pipeline leakage rier is necessary for preventing ignition sources and detecting
concentration of flammable gas.
Based on SHIPP conceptual accident model (Rathnayaka et al., It is worth noting that the human factor, management and
2011), as well as considering the characteristics of submarine organization serve as a special barrier which affect the whole
168 Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 163–173
Fig. 8 – Bow-tie model of leakage failure of submarine oil and gas pipeline.
process of leakage accident evolution (Rathnayaka et al., 2013). barriers and corresponding failure probabilities are as shown
In this paper, for modeling conveniently, keeping it common to in Table 5 (PSA NO, 2014; Abimbola et al., 2014).
other barriers, which are in sequential order. During the above
process, the function of human, management and organiza- 4.3. Bow-tie model of pipeline leakage
tional factor is more prominent between monitor and alarm
and emergency shut down, including finding alarm signals, An ET is developed based on the sequential failure of safety
judging and verifying the leakage occurrence. Consequently, barriers shown in Fig. 7. It is assumed that the human response
the independent barrier of human response shown in Fig. 8 barrier is able to work successfully once monitor and alarm
is located between monitor and alarm and emergency shut barrier work correctly. If the human response barrier fails to
down. In the above safety barriers model, failure of different work, the emergency shutdown barrier will fail at the same
barriers leads to different consequences. The all above safety time, and once the emergency shutdown barrier fails, the
Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 163–173 169
5.2. Probability updating Table 5 – Safety barriers of event tree in Fig. 7 and their
probabilities (PSA NO, 2014; Abimbola et al., 2014).
Once the Bayesian network is mapped from bow-tie, the prob- Symbol Safety barriers Prior Posterior
ability updating can be conducted using Bayesian network probability probability
through introducing a evidence. The probability after updating S1 Monitor and alarm 3.0E-03 1.01E-01
is in terms of posterior probability, for example, when a cer- barrier
tain state event E is given, the posterior probability of event Xi S2 Human response 1.0E-02 9.84E-02
is calculated as P (Xi | E). The most common evidence used in S3 Emergency shutdown 2.0E-02 1
probability updating is the knowledge about top event or con- S4 Negative protection 4.0E-02 1
system
sequences (Khakzad et al., 2013a). In this paper, the top event
S5 Emergency response 5.5E-03 1
leakage failure is adopted as evidence for estimating posterior S6 Ignition protection 3.0E-02 1
probability of root nodes while the E end-state is adopted as
Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 163–173 171
Table 7 – Cumulative number of precursor events and accidents of a submarine oil and gas pipeline.
T/year 1-A 2-A 3-B 4-C 5-A 6-D 7-E 8-A 9-A 10-B 11-C 12-A 13-D 14-E 15-A 16-D 17-E
1 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0
2 2 1 1 0 2 0 0 0 0 1 0 1 0 0 0 0 0
3 4 2 2 1 2 0 0 0 1 1 1 2 0 0 0 0 0
4 4 2 2 1 3 1 0 1 1 2 1 3 1 0 0 0 0
5 5 3 3 1 3 1 0 1 1 2 2 3 1 0 1 0 0
6 7 3 4 1 4 1 0 2 1 2 4 5 1 0 1 0 0
7 8 4 5 2 4 2 1 3 2 3 4 5 1 0 1 0 0
8 8 4 6 2 5 2 1 3 2 4 5 5 2 0 1 1 0
9 10 5 8 3 6 2 1 3 2 5 5 6 2 1 2 1 0
10 12 6 10 3 7 3 2 4 3 6 7 8 3 1 2 1 0
172 Process Safety and Environmental Protection 1 0 3 ( 2 0 1 6 ) 163–173
S3
S4 Fig. 13 illustrates that the probability of consequence of safe
0.20 S5 state decrease gradually during ten years, this case indicate
S6 risk decision-making and preventive measures must be taken
0.15 for improving safe status of pipeline.
As a result of considering accident precursor data, the
0.10 probabilities of safety barriers and consequences are more
case-specific. Hence, the estimated outcome with posterior
0.05 probabilities is more accurate than prior probabilities. More-
over, dynamic change of these failure probabilities is also
0.00 represented using experience learning. Thus, the develop-
0 2 4 6 8 10 ment trend of safety status for pipeline can be predicted
Time-year
efficiently.
Fig. 11 – Dynamic failure probabilities of safety barriers.
0.00010 C
D Meanwhile, based on backward inference, the probability
0.00008 E updating is implemented when a new evidence is consid-
ered. In addition, an experience learning is also conducted
0.00006 with recorded accident precursor data, and dynamic failure
probabilities of safety barriers and consequences are obtained
0.00004 through this method.
In present study, when CPT is not amended, conditional
0.00002
dependency and common cause failures are not considered
simultaneously, Bayesian network had the same estimating
0.00000
0 2 4 6 8 10 value with bow-tie method. Otherwise, the Bayesian network
Time-year approach had a more big estimating value due to considering
more factors. In Bayesian network analysis, probability updat-
Fig. 12 – Dynamic probabilities of consequences in
ing indicated man-made drilling oil stolen, Seabed movement
dangerous state.
and auxiliaries aging are most probable factors for lead-
ing to a leakage occurrence. In addition, five most probable
0.01590 development patches of pipeline leakage are also identi-
fied through strength of influence of GeNIe software. The
A
probability updating also indicated that emergency response
0.01585
Probabilities of consequence
Furthermore, the study could provide an educational refer- Kujath, M.F., Amyotte, P.R., Khan, F.I., 2010. A conceptual offshore
ence for risk analysis of other ocean oil and gas equipment. oil and gas process accident model. Journal of Loss Prevention
in the Process Industries 23 (2), 323–330.
Meel, A., Seider, W.D., 2006. Plant-specific dynamic failure
Acknowledgment
assessment using Bayesian theory. Chemical Engineering
Science 61 (21), 7036–7056.
The authors gratefully acknowledge the financial support pro- OGP UK, 2010. Risk assessment data directory: Risers and pipeline
vided by science and technology development projects of release frequencies, http://www.ogp.org.uk/pubs/434-04.pdf.
Shandong province in China (Project No: 2014GSF120014) and Participants, O., 2002. OREDA Offshore Reliability Data Handbook,
postgraduate innovation engineering projects of china univer- DNV, PO Box.
PSA NO, 2014. Trends in risk level in the petroleum activity,
sity of petroleum (East china) (Project No: YCXJ2016057). The
http://www.psa.no/getfile.php/PDF/RNNP%202014/RNNP
first author also would like to express his gratitude to lecturer
2014 ENG.pdf.
Xiuquan Liu for providing help during preparing this paper. Rathnayaka, S., Khan, F., Amayotte, P., 2013. Accident modeling
and risk assessment framework for safety critical
References decision-making: application to deepwater drilling operation.
Proceedings of the Institution of Mechanical Engineers. Part O:
Journal of Risk and Reliability 227 (1), 86–105.
Abimbola, M., Khan, F., Khakzad, N., Butt, S., 2015. Safety and risk
Rathnayaka, S., Khan, F., Amyotte, P., 2011. SHIPP methodology:
analysis of managed pressure drilling operation using
Predictive accident modeling approach. Part I: Methodology
Bayesian network. Safety Science 76, 133–144.
and model description. Process safety and environmental
Abimbola, M., Khan, F., Khakzad, N., 2014. Dynamic safety risk
protection 89 (3), 151–164.
analysis of offshore drilling. Journal of Loss Prevention in the
Rausand, M., 2013. Risk assessment: theory, methods, and
Process Industries 30, 74–85.
applications. John Wiley and Sons.
Aljaroudi, A., Khan, F., Akinturk, A., Haddara, M., Thodi, P., 2015a.
Reason, J.T., Carthey, J., De Leval, M.R., 2001. Diagnosing
Risk assessment of offshore crude oil pipeline failure. Journal
“vulnerable system syndrome”: an essential prerequisite to
of Loss Prevention in the Process Industries 37,
effective risk management. Quality in health care 10 (suppl 2),
101–109.
ii21–ii25.
Aljaroudi, A., Khan, F., Akinturk, A., Haddara, M., Thodi, P., 2015b.
Reason, J., 1990. The contribution of latent human failures to the
Probability of Detection and False Detection for Subsea Leak
breakdown of complex systems. Philosophical Transactions of
Detection Systems: Model and Analysis. Journal of Failure
the Royal Society B: Biological Sciences 327 (1241), 475–484.
Analysis and Prevention.
Sklet, S., 2006. Safety barriers: Definition, classification, and
Bhandari, J., Abbassi, R., Garaniya, V., Khan, F., 2015. Risk analysis
performance. Journal of Loss Prevention in the Process
of deepwater drilling operations using Bayesian network.
Industries 19 (5), 494–506.
Journal of Loss Prevention in the Process Industries 38, 11–23.
Skogdalen, J.E., Vinnem, J.E., 2012. Combining precursor incidents
DNV-RP-F107, 2010. Risk Assessment of Pipeline Protection. Det
investigations and QRA in oil and gas industry. Reliability
NorskeVeritas.
Engineering and System Safety 101, 48–58.
HSE UK, 2011. Offshore Safety Statistics Bulletin,
Tan, Q., Chen, G., Zhang, L., Fu, J., Li, Z., 2013. Dynamic accident
http://www.hse.gov.uk/offshore/statistics/stat1011.htm.
modeling for high-sulfur natural gas gathering station.
Kalantarnia, M., Khan, F., Hawboldt, K., 2009. Dynamic risk
Process Safety and Environmental Protection 92 (6), 565–576.
assessment using failure assessment and Bayesian theory.
Pittsburgh, U.O., 2014. GeNIe 2.0. G. 2.0 (Ed.) 2.05219.0. University
Journal of Loss Prevention in the Process Industries 22 (5),
of Pittsburgh.
600–606.
Xue, L., Fan, J., Rausand, M., Zhang, L., 2013. A safety barrier-based
Kalantarnia, M., Khan, F., Hawboldt, K., 2010. Modelling of BP
accident model for offshore drilling blowouts. Journal of Loss
Texas City refinery accident using dynamic risk assessment
Prevention in the Process Industries 26 (1), 164–171.
approach. Process Safety and Environmental Protection 88 (3),
Yuan, Z., Khakzad, N., Khan, F., Amyotte, P., 2015. Risk analysis of
191–199.
dust explosion scenarios using Bayesian networks. Risk
Kawsar, M.R.U., Youssef, S.A., Faisal, M., Kumar, A., Seo, J.K., Paik,
analysis 35 (2), 278–291.
J.K., 2015. Assessment of dropped object risk on corroded
Fang, N., 2014. Study on risk analysis and emergency measures of
subsea pipeline. Ocean Engineering 106, 329–340.
submarine pipeline leakage accidents. China university of
Khakzad, N., Khan, F., Amyotte, P., Cozzani, V., 2013. Domino
petroleum, Qingdao.
Effect Analysis Using Bayesian Networks. Risk Analysis 33 (2),
Fang, N., Chen, G., Zhu, H., Meng, H., 2014. Statistical analysis of
292–306.
leakage accidents of submarine pipeline. Oil and Gas Storage
Khakzad, N., Khan, F., Amyotte, P., 2011. Safety analysis in
and Transportation 33 (01), 99–103.
process facilities: Comparison of fault tree and Bayesian
Hu, X., Duan, M., Guan, Y., 2012. Quantitative risk assessment of
network approaches. Reliability Engineering and System
deepwater submarine pipeline based on fuzzy bow-tie model.
Safety 96 (8), 925–932.
China Safety Science Journal 22 (03), 128–133.
Khakzad, N., Khan, F., Amyotte, P., 2012. Dynamic risk analysis
Jin, W., Zhang, E., Shao, J., Liu, D., 2004. Cause Analysis and
using bow-tie approach. Reliability Engineering and System
Countermeasure for Submarine Pipeline Failure. Bulletin of
Safety 104, 36–44.
Science and Technology 20 (06), 529–533.
Khakzad, N., Khan, F., Amyotte, P., 2013a. Quantitative risk
Zhang, X., Xie, L., Chen, G., 2011. Integrity management
analysis of offshore drilling operations: A Bayesian approach.
technique for submarine pipeline. Oil Field Equipment 40 (12),
Safety Science 57, 108–117.
10–15.
Khakzad, N., Khan, F., Amyotte, P., 2013b. Dynamic safety
Zhu, H., Chen, G., 2009. Integrated reliability assessment on
analysis of process systems by mapping bow-tie into Bayesian
VIV-induced fatigue failure for submarine pipeline system
network. Process Safety and Environmental Protection 91
due to span. China Offshore Oil and Gas 21 (2), 133–137.
(1–2), 46–53.