Quantitative risk assessment of fire accidents of large-scale oil tanks triggered
by lightning

Wu Dongyin, Chen Zhen

PII: S1350-6307(15)30154-0
DOI: doi: 10.1016/j.engfailanal.2015.11.029
Reference: EFA 2754

To appear in:

Received date: 18 August 2015

Revised date: 25 October 2015
Accepted date: 16 November 2015

Please cite this article as: Dongyin Wu, Zhen Chen, Quantitative risk assess-
ment of fire accidents of large-scale oil tanks triggered by lightning, (2015), doi:
10.1016/j.engfailanal.2015.11.029

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Cover letter

Institute of Mechanics, Chinese Academy of Sciences

North Fourth Ring West Road #15, Beijing, China, 100190
TEL: 0086-18810898256

PT
E-mail: chenzhen@imech.ac.cn

Oct 25, 2015

RI
Editors of Engineering Failure Analysis

SC
Dear Editors,

NU
I am a student of the Institute of Mechanics, Chinese Academy of Sciences, and now
am a doctoral candidate in safety engineering. After investigating fire accident cases
caused by lightning in oil tanks area and summarizing different categories of fire
MA
accident scenarios, a generic event tree was built to present the process of fire
development of oil tanks with the protection measures of a fire extinguishing system
and a fire brigade. A methodology was proposed to quantitatively assess the fire risk
caused by lightning. The application of the methodology to a case study provided the
D

probabilities of different forms of accidents for three categories.

The above-described work is very important and useful for fire safety and firefighting
TE

involved in the current development of crude oil reserves. Therefore, a manuscript

titled “Quantitative risk assessment for the fire accidents of large-scale oil tanks
P

triggered by lightning” was finished, and we wish that the manuscript can be
published in Engineering Failure Analysis.
CE

All of the authors of this research paper have directly participated in the planning,
execution, or analysis of the study. All of the authors of this paper have read and
approved the final version submitted. The contents of this manuscript have not been
AC

copyrighted or published previously. The contents of this manuscript are not under
consideration for publication elsewhere. The contents of this manuscript will not be
copyrighted, submitted, or published elsewhere while acceptance of the manuscript is
under consideration. There are no directly related manuscripts or abstracts, published
or unpublished, by any author(s) of this paper.
I appreciate your considering our manuscript for publication. I look forward to your
reply.

Sincerely yours,
Zhen Chen
 ACCEPTED MANUSCRIPT

Highlights

We construct the detailed fire accident scenarios with various factors.

We build probability models to quantitative analysis the effect of different factors.

PT

The failure probabilities were calculated based on ETA and FTA method.

RI
SC
NU
MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

Quantitative risk assessment of fire accidents of large-scale oil tanks triggered by lightning

Wu Dongyina, Chen Zhenb,*

a. School of Energy and Power Engineering, Xi’an Jiaotong University, Xi’an, China, 710049

PT
b. Department of Thermal Engineering, Tsinghua University, Beijing, China, 100084

RI
Abstract: Severe fire accidents may be triggered by lightning strikes in large-scale oil tank areas. A
methodology was proposed to quantitatively assess the fire risk. Based on previous relevant fire

SC
cases, three categories of fire accident scenarios are summarized: (A) fire accidents at the rim seal,
(B) fire accidents above the floating roof and (C) fire accidents by perforation of the float pan. A
generic event tree was built to present the process of fire development of oil tanks with the

NU
protection of a fire extinguishing system and a fire brigade. The probability models were proposed
to calculate the probabilities of lightning strike, tank wall and floating roof damage/perforation,
failure of the fire protection system and timely response of fire brigade. The application of the
MA
methodology to a case study provided the probabilities of different forms of accidents for three
categories. According to the results of the case study, the probabilities of rim seal fire, local pool
fire and full surface fire of large-scale oil tanks are 1.10×10-2, 10-8–10-6 and 2.17×10-9,
respectively, for Categories A and C, and the probabilities of rim seal fire, local pool fire and full
D

surface fire of large-scale oil tanks are 4.99×10-2、10-8–10-5 and 9.85×10-8, respectively, for
TE

Category B. The fire risk by lightning for Category B is relatively higher, which could be reduced
by improving the design of the float pan and enhancing the reliability of the components.
Compared with the hazard of local pool fire and full surface fire, the risk of rim seal fire is
P

unacceptable; in this case, an inert gas protection system can be introduced to reduce the risk.
CE

Keywords: large-scale oil tanks, lightning strike, quantitative risk assessment, accident scenarios,
fire accidents.
AC

1. Introduction
The construction of crude oil reserve facilities has great significance for national economic
security and the development of social stability. The use of large-scale steel tanks is one of the
important ways of storing crude oil reserves. The most common tanks being used and under
construction are floating roof tanks, with a capacity of 1×105 m3 or 1.5×105 m3 (Papavinasam S.,
2013; Wu T.Z., 2004). Crude oil poses the hazards of fire and explosion and is apt to have
accidents in the process of transport and storage, such as leakage and diffusion. Leaking oil or oil
vapor could result in an explosion or fire very easily when exposed to an ignition source. If there
was not timely and appropriate action to control such accidents effectively, more severe accidents
can result. For example, the Domino effect is well-known in major accidents in the chemical
process industries (Shaluf I.M. & Abdullah, S.A., 2011; Mishra K.B., Wehrstedt K.D., & Krebs H.,
2013; Antonioni G., Spadoni G. & Cozzani V., 2007).
According to the survey results by Chang (Chang J.I. & Lin C.C., 2006.), approximately 33%
of fire and explosion accidents (242 cases) in an area with tanks are triggered by lightning. To
reduce the risk of lightning, most countries have established standards of lightning protection for
crude oil storage, such as API RP545-2009 in the USA (America, 2009) and GB 15599-2009 in
 ACCEPTED MANUSCRIPT

China (China, 2009). Overall, these standards have a limited effect on reducing the probability of
accidents by lightning. Many oil tanks that meet the relevant standards still incur
lightning accidents every year, causing serious damage or injuries. Some typical cases are
presented in Table 1.
According to the survey results in Reference 2, the risk of fire and explosion of oil tanks

PT
triggered by lightning is relatively high. However, it is very difficult to reconstruct the accident
scenarios, which leads to scarce research results in this area.
Amos Necci (Necci A., Argenti F., Landucci G. & Cozzani V., 2014; Necci A., Antonioni G.,

RI
Cozzani V., Krausmann E., Borghetti A. & Nucci G.A., 2014.) studied hazardous material
accidents triggered by lightning strike, with a focus on the identification of the event sequences

SC
and the accident scenarios following lightning impact on atmospheric tanks. The past accident
cases were used to validate the calculation of the expected frequencies of the final scenarios for
different type of tanks. However, Amos Neccide concentrated primarily on rim seal fire research,

NU
without enough consideration of the other forms of accidents. Thus, Amos Neccide’s analysis of
accident scenarios was incomplete. Elisabetta Renni et al. (Renni E., Krausmann E. & Cozzani V.,
2010) studied the potential forms of accidents based on investigation and statistical analysis of the
MA
cases caused by lightning in the chemical process industries. Su Boni et al. (Su B.N., Huang H. &
Li Y.T., 2013) proposed a detailed model to provide a quantitative analysis of the risk of lightning
strikes and direct damage/perforation. The probabilities for different scales of tanks were
determined by the models, but without the analysis of the accident scenarios. Moreover, all of the
D

above studies did not consider the effects of fire brigade on the fire and explosion, which is
TE

obviously unreasonable. If the fire brigade arrives in time after the fire and explosion occurred, the
accidents could be effectively addressed in most cases, even if the tank fire protection system
failed to operate.
P

From the existing research results, there is still a lack of a comprehensive model, so the
CE

quantitative risk assessment of fire triggered by lightning is very difficult to perform. To address
this problem, this paper provides a summary of the main accident scenarios via the analysis of
accidents based on previous research results in the literature. The proposed method is applied to a
AC

case (crude oil reserve in Tianjin City, China) to provide a better assessment and management of
the risk of industrial facilities with respect to external hazards due to natural events.
2. Analysis for accident scenarios
Oil tanks struck by lightning can result in a variety of accidents. In accordance with the fire
cases listed in Table 1, some common features of these accidents can be summarized as follows:
 Leakage of oil or oil vapor.
 Local fire due to lightning.
 The automatic fire extinguishing system can effectively control and put out the early-rising
fire.
 If the fire protection system failed, a full surface pool fire may occur.
To assess the risk of major accidents triggered by lightning, a generic event tree is obtained,
as shown in Fig. 1, based on past accident cases involving atmospheric tanks storing hydrocarbons
or generic flammable materials.
After a detailed analysis of the cases in Table 1, there are three main accident scenarios for
oil tanks, as described below.
Category A: fire accidents at the rim seal
 ACCEPTED MANUSCRIPT

There is a circular space in the range of 200 to 500 mm between the floating roof and the
tank wall, which is required for good sealing (Hu H.Y., Liu B.Q., Liu Q.Z., Zhang T.T. & Gao X.,
2011). If the sealing is too tight, it would be adverse to the movement of the floating roof. If the
sealing is too loose, it would not control the oil and oil vapor leakage. As a result, a secondary
sealing device is used to protect the quality of the oil and the safety of the tanks.

PT
When an oil tank is struck by lightning, inductive charge will release according to some rules.
If the charge cannot be released in a timely manner, an electrical potential between the tank wall
and the floating roof forms. In this case, many components of the tank will become sources of

RI
spark, such as the conducting strip of the primary seal and the secondary seal, the connecting bolts
and the metal seal plate. Moreover, it is unavoidable that there is a flammable mixture between the

SC
primary seal and the secondary seal (Lang X.Q., Gao X., Gong H., Liu Q.Z. & Wang T., 2008;
Wang Z.G., 2007), which can easily be ignited by the spark, resulting in a local fire. In the case of
failure of the fire protection system, fire expansion and spreading will result.

NU
Category B: fire accidents above the floating roof
For the safety of oil storage and convenience of service, there are many components fixed at
the floating roof, for example, breathing valve, vent valve and entrance.
MA
During actual operation, the failure of the breathing valve or the vent valve will lead to
leakage of oil and oil vapor. When the floating roof drops, the oil sticking on the inner tank wall
will volatilize. If the seal between the rim seal and tank wall is poor, then oil and vapor can also
leak. This leakage will gather onto the floating roof and form a flammable and explosive mixture.
D

In the case of a lightning strike, such a mixture can easily cause a fire and explosion.
TE

The case of leakage of oil vapor forming a flammable mixture is considered to be a small
probability event, with a probability of 0.05 (Lang X.Q., Liu Q.Z. & Gong H., 2009; Luo A.M., Li
W.C. & Wu Z.Z., 2008).
P

Category C: fire accidents by perforation of the float pan

CE

Because of the intense heating effect of lightning current, it can cause metal damage or
perforation, thereby causing oil leakage. Meanwhile, the molten metals at high temperature can
easily ignite combustion. If the floating roof is damaged or perforated, the leakage oil is prone to
AC

lead to a fire and explosion incident. If the tank wall is damaged or perforated, a
massive oil spill may lead to a flowing fire. However, the steel of the tank wall is much thicker
than the floating roof and is positioned perpendicularly; as a result, the probability of tank wall
failure is minimal compared with that of the floating roof (Su B.N., Huang H. & Li Y.T., 2013).
Therefore, only the risk of floating roof damage by lightning is considered in the case study.
Normally, there is some space with high concentrations of oil vapor between the floating roof
and the oil surface. When the floating roof is perforated by lightning, this oil vapor will release
quickly and mix with air, forming an atmosphere conducive to fire and explosion. Once the
flammable mixture is ignited, jet fire may emerge from the surface of the floating roof. According
to Table 1, this type of accident often appears around the rim seal, and it has the risk of developing
into a full surface fire without the implementation of timely and effective control measures.
3. Quantitative calculation method
3.1 Probability of lightning strike
Most of the lightning discharge phenomena in nature occur between clouds. However,
cloud-ground discharges, even with relatively low probability, have a significant effect on
production and life styles. Such discharge forms are studied widely, due to the great hazard they
 ACCEPTED MANUSCRIPT

pose to life and property.

Cloud-ground discharges include positive stroke and negative stroke; most lightning strikes
are negative stroke. Positive strokes only appear in some special natural disasters. The current
distribution of lightning can be determined from the statistical data and has nothing to do with the
load resistance. The average annual number of thunderstorm days is 28.2 in Tianjin, which is in

PT
the middle of the thunderstorm region (the annual average number of thunderstorms is more than
15 days and less than 40 days).The probability that the current is greater than I can be calculated
using formula 1 (China, 1997), and the relationship between the current and the lightning current

RI
time is expressed as formula 2 (IEC, 2010).

lg P = − I / 88 (1)

SC
where I is the magnitude of current, kA, and P is the probability that the current is greater than I.

I (t / τ 1 )10

NU
i= ⋅ ⋅ exp(−t / τ 2 ) (2)
k 1 + (t / τ 2 )10
where k is the correction factor, and τ1 and τ2 are time constants of the initial and wake times,
MA
respectively. Considering only the first pulsing strike of lightning, the following values are set:
τ1=1.82 µs, τ2=285 µs, and k=0.986。
Integrating the time (t), the relation between the amount of discharged electricity and the
peak of the current (I) can be expressed as follows.
D

Q=I×2.827×10-4 s (3)
TE

To assess the risk of fire accidents due to lightning, the probability of lightning strike should
be computed first. According to IEC 62305-2 (IEC, 2010) proposed by the International
Electrotechnical Commission, the frequency (ND) of lightning strikes for a building can be
P

calculated using formula 4.

CE

N D = NG ⋅ AD ⋅ CD × 10 −6 (4)

The parameters in formula 4 are as follows.

AC

(1) NG: Density of lightning strikes

The density of lightning strikes (NG) is defined as the frequency of lightning strikes per
square kilometer every year, which can be determined from the data of lightning strikes. In the
case of no lightning monitoring system, the density of lightning strikes can be estimated using
formula 5.

N G = 0.1Td (5)

(2) AD: Equivalent receiving area

The oil tanks should be simplified to a regular three-dimensional model when establishing
the equivalent receiving area (AD). The equivalent receiving area can be calculated by formula 6,
where R and H are the radius and height of tank.

AD = π( R + 3H ) 2 (6)

(3) CD: Location factor of the building

The oil tank is not isolated in a real situation. Therefore, the effect of other tanks or buildings
must be considered when computing the frequency of lightning strikes. The location factor of
 ACCEPTED MANUSCRIPT

building (CD) is presented in Table 2.

Based on the above calculation method, the receiving area of a single tank is 3.38×104 m, and
the density of lightning strikes is 0.0352.
3.2 Direct damage or perforation
Lightning current has an intense heating effect, which can cause the fusion of metals

PT
contacted with the arc. When the tank wall or floating roof is damaged or perforated, it will lead to
leakage and escape of oil or oil vapor. The flammable materials will be ignited by the hot metal
heated by lightning current. In this case study, the probability of forming a fire source is given as 1

RI
if the metal was damaged or perforated. If not, the probability is given as 0.
After the tank wall or floating roof is contacted by the lightning arc, there will be a

SC
hemispherical groove of radius r on the face of metal because of the heat effect. When r is larger
than the thickness d of the metal, the tank wall or floating roof can be considered to be damaged or
perforated by the lightning strike. According to the results introduced by the lightning protection

NU
guide (DEHN, 2014), the heat released by lightning is in proportion to the electric charge, while
the size of the hemispherical groove is in proportion to the released heat. Thus, the volume of
hemispherical groove is proportional to the amount of electrical charge in the lightning strike. The
MA
probability of direct damage or perforation can be determined based on the current distribution, as
presented in Table 3.
3.3 Probability of fire protection system failure
According to the accident scenario analysis presented in section 2, it is clear that the fire
D

protection system is very important for controlling the early-rising fire in the tank area. In the case
TE

where the fire protection system failed, the propagation and upgrade of the degree of fire accidents
may occur. Therefore, the failure probability of a fire protection system is significant for the
quantitative risk assessment of lightning hazards for oil tanks. The fire protection system mainly
P

includes an automatic sprinkler system, an automatic foam extinguishing system and a manual
CE

foam extinguishing system in the crude oil reserve area. Based on the results of Amos Necci et al.
(Necci A., Argenti F., Landucci G. & Cozzani V., 2014; Necci A., Antonioni G., Cozzani
V., Krausmann E., Borghetti A. & Nucci C. A., 2013), the failure probability of a fire protection
AC

system can be analyzed and computed using the event tree method.
Taking a fixed automatic foam system as an example, the calculation method is briefly
introduced as follows.
The input data required for the quantitative analysis of the Top Event (Foam System Failed)
are as follows (Nucci C.A., Rachidi F., 2003):
Logical relationships among the system components, such as series systems, parallel systems, and
standby systems.
Type of component, such as repairable components, non-repairable components, and intermittent
operation components.
Components reliability data, such as the failure rate, mean time to repair, and test interval.
The FTA for a fixed automatic foam system is shown in Fig. 2; the fault trees related to the
subsystems were not included for the sake of brevity. More details on the method and the
procedure are reported in Reference (Nucci C.A. & Rachidi F., 2003). The final failure probability
obtained in the present analysis is listed in Table 4.
3.4 Response time of the fire brigade
The automatic foam extinguishing system is the first line of defense for a tank fire, followed
 ACCEPTED MANUSCRIPT

by the use of manual fire extinguishing equipment. These measures can cope with most fire
accidents and put out fire successfully. If the protection system fails, then the oil reserve only
relies on the nearby fire brigade to control and extinguish the fire.
According to the regulations of oil reserve design specifications in China (China, 2002), fire
crews and related equipment should arrive at the fire scene within 5 min. The telephone operator

PT
receives the fire alarm and will alert the firefighting team to arrive at the scene and put out the
fire. If the total response time is less than 5 min, then the fire brigade is considered to have arrived
at the fire scene in a timely manner and to have put out the fire successfully in the case study.

RI
Ed Claridgea and Michael Spearpoint (Claridgea E. & Spearpoint M., 2013) obtained values
for each of the following elements that make up the response times from 6,410 fire accidents.

SC
These times are classified as confirmed time (T1), preparation time (T2), and arrival time (T1). The
total response time of the fire service is equal to the sum of Ti.
(1) Times for the receipt of information (T1)

NU
Once a communication center receives a report that a fire incident has occurred, either by
direct communication from an automatic alarm system or from an emergency telephone call, etc.,
the time is logged until the incident location and the generic situation have been confirmed.
MA
As expected, the great majority of incidents notified, over 85%, were from emergency
telephone calls, according to the results of Literature 15. Based on the NZFS's proposal, the value
of T1 can be treated as a lognormal distribution, while the mean (µ) and standard deviation (σ)
values are 39 and 22, respectively.
D

(2) Times for preparation (T2)

TE

Once the communication center received fire alarm, the fire service will prepare for action
quickly. Next, the firefighters must assimilate the information provided and dress in structural
fire-fighting clothing before mounting the fire-fighting truck and leaving the station. Based on the
P

NZFS data of 3000 incidents, 20% of the firefighter response times occurred in less than 60 s.
CE

A log-logistic distribution best fitted each of the responses with the values given as career
firefighters; the fastest responses occurred in the order expected, with 95% of the responses for the
main cities occurring within 123 s (µ = 77), compared with 271 s for all of the career staff (µ =
AC

113) and 459 s for the volunteers (µ = 224).

(3) Times for arrival (T3)
The average travel speed that a fire truck can obtain to reach the fire incident depends on a
number of factors, including travel distance, road type, weather conditions, and traffic conditions.
Tommasson et al. (Berman G.L, 2009) and Yung (Yung, D., 2008) considered the factors
affecting the response time of fire truck and reported the most important factors to be the distance
travelled to the incident. Of the 5,309 structure fires, the distance for each incident calculated
using the Trans CAD software divided by the times taken for the truck to travel to the incident
provides the calculation of the average truck speed to each incident. These times and distances are
shown in Fig. 3.
3.5 Validation
3.5.1 Validation for probability model of lightning strike and direct damage/perforation
As the previous work by Necci and others (A. Necci, 2014), they proposed Monte Carlo
method to calculation the capture frequencies of lightning impact for process equipment. The
results based on MC method were compared with actual frequencies, which prove its accuracy and
effectiveness. Therefore, the probability model proposed by this work is applied to the case in
 ACCEPTED MANUSCRIPT

previous work for validation.

Table 5 lists the calculation of lightning frequencies of some sample tanks from a reference,
while P-MC denotes as the results by Monte Carlo and P-PM denotes as the results by probability
model proposed in this work. From the comparison, the results based the model of this work is

PT
closed to the results in the reference, and the relative error is always below 5%. Therefore, we
think the probability model of lightning strike and direct damage/perforation is accurate enough
for the study.

RI
3.5.2 Validation for probability model of fire protection system failure
In section 3.3, the probability model of fire protection system failure is put up based on FA

SC
method. In order to ensure the accuracy of the model, the failure probability values of the selected
events in Fig.2 obtained from the statistics (Necci C.A, 2003) are compared with those generated
using the probability model proposed in the work.

NU
From Fig.5, there is slight difference in the failure probability values of the selected events
obtained from the calculations and statistical data. The relative error for probability of all basic
MA
event is always below 10%. The biggest one is A4, Undetected major leak, but its failure
probability is very small, approximate 10-7, so its error has little effect on the accuracy of top
event. Therefore, we think the probability model of lightning strike and direct damage/perforation
is accurate enough for the study.
D

4. Case-study
TE

The paper mainly considers a crude reserve in Tianjin City, China, to analyze the risk of fire
accidents due to lightning. In total, the reserve has 1×106 m3 of crude oil. There are two tank
groups in the reserve. One group has six tanks, each with a volume of 1×105 m3, and the other one
P

has four tanks, each with a volume of 1×105 m3, along with two reserved tank zones. Every tank is
CE

21.8 m in height and has a diameter of 80 m, and the space between two tanks is 32 m, as shown
in Fig. 5. The main material of the floating roof and tank wall is #16 Mn steel, with thicknesses of
5 mm and 35 mm, respectively. According to the design standard of oil reserves in China, an
AC

automatic foam fire extinguishing system is installed in the tank area, which can continuously
supply a foam mixture for 30 min at the speed of 96 L/s. There are 12 fixed PC8 foam makers
around the top edge of the roof, which can supply environmentally friendly aqueous film-forming
foam at a concentration of 6%. In addition, there are two mobile manual foam lances for
emergency use. The supply speed of the fixed foam system is 24 L/s, and the supply lasts for 30
min. The distance between the reserve and the nearest fire brigade is 5.3 km.
The lay-outs of the floating roof, the fire protection system and the seal equipment are shown
in Fig. 6 and Fig. 7.
5. Results and discussion
An overall method is proposed in Section 3 and applied to perform a quantitative risk
assessment for the case in Section 4. First, the probability of lightning strike, direct damage or
perforation as well as fire protection system failure can be determined based on the given
information.
The results of the probability of direct damage or perforation are shown in Table 3, valued at
1.10×10-2 in the case considered. The probability of fire protection system failure was determined
based on the event tree analysis; the calculation is presented in Fig. 3 and Table 6. The response
 ACCEPTED MANUSCRIPT

time of the fire brigade includes the time for the receipt of information, the time for preparation,
and the time for arrival (the confidence level set at 90%); the detailed computation process will
not be presented here, but the results are listed in Table 7.
For Categories A and C, the probability of lightning strike can be calculated using formula 1,
which can also be treated as the probability of the ignition source caused by lightning. In addition,

PT
the fire risk of Categories A and C can be assessed based on the probability of fire protection
system failure and the fire brigade not arriving on time. For Category B, the flammable mixture
will be ignited when the tank is struck by lightning, without considering the probability of direct

RI
damage or perforation.
Finally, the results of different forms of accidents are listed in Table 7.

SC
The results in Table 7 indicate that the probability of rim seal fire is highest at 4.99×10-2
under the protection of a foam protection system and fire brigade.
The fire risk of Categories A and C is less than Category B, and the risk of Category B can be

NU
reduced by controlling the oil vapor on the floating roof via various measures, for example,
optimization of the design of the breathing valve and the vent valve, improving the air flow on
the floating roof, and checking the component of roof actively.
MA
From the results in Table 7, the probability of a local pool fire and full surface fire is lower
than 10-5 and 10-7, respectively, which represent extremely rare events, and the risk level could be
accepted. In contrast, the risk of a rim seal fire is higher, so some action should be taken to reduce
the risk. In accordance with the case, an inert gas protection system could be introduced to reduce
D

the risk of rim seal fire. The inert gas protection system can control 80-90% of fire occurrence
TE

according to relevant experience. With such a system, the probability of rim seal fire risk will be
reduced to approximately 10-3, and the safety of the oil tanks would be improved markedly.
6. Conclusions
P

A methodology was developed for the quantitative risk assessment for different accident
CE

forms triggered by lightning in an oil tank area. First, past similar fire cases were surveyed and
analyzed, with three typical accident scenarios proposed: fire accidents at the rim seal, fire
accidents above the floating roof and fire accidents by perforation of the float pan. Next, the
AC

generic event tree of fire accidents caused by lightning was built, considering the influence of a
nearby fire brigade. The probability of direct damage/peroration by lightning strike, fire protection
system failure and fire brigade arriving on time were calculated by the event tree analysis and
probabilistic method. Finally, the probabilities of a rim seal fire, local pool fire and full surface
fire produced by three different accident scenarios were determined.
Acknowledgments
The authors thank the National Natural Science Foundation of China (No. 51376148) for the
financial support of this work.
References
America, 2009. Recommended practice for lightning protection of aboveground storage tanks for
flammable or combustible liquids.
Antonioni G., Spadoni G., Cozzani V., 2007. A methodology for the quantitative risk assessment
of major accidents triggered by seismic events, Journal of Hazardous Materials, 147, 48-59.
Berman G.L, 2009. Review of Fire and Rescue Service Response Times. Fire Research Series
1/2009. Communities and Local Government Publications, London.
Chang J.I., Lin C.C., 2006. A study of storage tank accidents. Journal of loss prevention in the
 ACCEPTED MANUSCRIPT

process industries, 19(1), 51-59.

China, 1997. Overvoltage protection and insulation coordination for AC electrical
Installations.PRC National Standard.
China, 2002. Code for design of oil depot. PRC National Standard.
China, 2009. Safety rules of lightning with relation to petroleum and its facilities. PRC National

PT
Standard.
Claridgea E., Spearpoint M., 2013. New Zealand fire service response times to structure fires.
Procedia Engineering, 62, 1063-1072.

RI
DEHN, 2014. Lightning protection guide, revised 2rd edition.Germany.
Hu H.Y., Liu B.Q., Liu Q.Z., Zhang T.T., Gao X., 2011. Analysis on second sealing oil and gas

SC
space discharge in floating roof tank. China Safety Science Journal, 21(3), 106-109.
IEC, 2010. Protection against lightning-Part 2: Risk management.
Lang X.Q., Gao X., Gong H., Liu Q.Z, Wang T., 2008. Research on reducing oil gas concentration

NU
in large floating roof tank seals. Journal of Oil and Gas Technology. 30( 2), 618-619.
Lang X.Q., Liu Q.Z, Gong H., 2009. Research on crude oil evaporation and structure optimization
of floating roof rim seal. China Safety Science Journal, 19(5), 91-96.
MA
Luo A.M., Li W.C., Wu Z.Z., 2008. Explosion of solvent vapor in a ring partition of the floating
roof. Journal of loss prevention in the process industries, 21(6), 642-645.
Mishra K.B., Wehrstedt K.D., Krebs H., 2013. Lessons learned from recent fuel storage fires. Fuel
Processing Technology. 107:166-172.
D

Necci A., Antonioni G., Cozzani V., Krausmann E., Borghetti A., Nucci C. A., 2013. A model for
TE

process equipment damage probability assessment due to lightning. Reliability Engineering and
System Safety, 115, 115:91-99.
Necci A., Antonioni G., Cozzani V., Krausmann E., Borghetti A., Nucci G.A., 2014. Assessment
P

of lightning impact frequency for process equipment. Reliability Engineering and System Safety,
CE

130, 95-105.
Necci A., Argenti F., Landucci G., Cozzani V., 2014. Accident scenarios triggered by lightning
strike on atmospheric storage tanks. Reliability Engineering and System Safety, 127, 30-46.
AC

Necci C.A., Rachidi F., 2003. Interaction of electromagnetic fields with electrical networks
generated by lightning. IEEE Power and Energy Series, 34:425-478.
Papavinasam S., 2013. Corrosion Control in the Oil and Gas Industry, Chapter 2-Oil and Gas
Industry Network. Gulf Professional Publishing.
Renni E., Krausmann E., Cozzani V., 2010. Industrial accidents triggered by lightning. Journal of
Hazardous Materials, 184(1-3), 42-48.
Shaluf I.M., Abdullah, S.A., 2011. Floating roof storage tank boilover. Journal of loss prevention
in the process industries, 24(1): 1-7.
Su B.N., Huang H., Li Y.T., 2013. Study on Probability Calculation for Oil Tank Fire and
Explosion Caused by Lightning. China Safety Science Journal, 23(4), 79-83.
Wang Z.G., 2007. The reason and counter method of oil gas space explode and fire in the first and
second seal of oil floating roof tank. Fire Science and Technology, 26( 6), 654-655.
Wu T.Z., 2004. Summary of development on large scale vertical oil storage tank, Petro-Chemical
Equipment Technology, 25(3), 56-60.
Yung, D., 2008. Principles of Fire Risk Assessment in Buildings.Wiley, Chichester.
 ACCEPTED MANUSCRIPT

PT
RI
PFD: Probability of Direct damage/perforation
PF1: Probability of Automatic foam extinguishing system failure

SC
PF2: Probability of Manual foam extinguishing system failure
Pf3: Probability of Fire brigade not arriving on time
Fig. 1 Generic event tree of fire accidents for floating roof tanks triggered by lightning

NU
MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

Fixed automatic foam

system failure

PT
RI
SC
A5
A1 Foam A2 A6
supply Foam A7
unavailabl

NU
maker
e fails A4
B1
B2

A3
MA
Fig. 2 Fault tree for fixed automatic foam system
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

0.030

0.025

Probability density
0.020

PT
0.015

0.010

RI
0.005

SC
0.000
0 10 20 30 40 50 60 70 80 90 100
Response speed (km/h)

Fig. 3 The probability density of the response speeds

NU
MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

0.01
0.00398
Calculation 0.00386

8.83E-4
0.00116
0.00121
Reference
1E-3 8.95E-4 6.24E-4
6.09E-4

1.76E-4

PT
1.85E-4
9.57E-5 1E-4
9.75E-5 1.06E-4
Probability 1E-4

8.96E-6

RI
1E-5 9.62E-6

SC
1E-6

1.37E-7
1.5E-7
1E-7

NU
1E-8
A1 A2 A3 A4 A5 A6 A7 B1 B2
MA
Basic Events
Fig. 4 Comparison of the failure probability values generated by this study and statistical data
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

Office Building

Metering
Garage Room

Boiler Room

PT
Pump
Office Building Room

RI
fire-fighting
equipment

SC
sewage treatment
equipment reserve
zone
Emergency pool

NU
Fig. 5 Lay-outs of crude oil tank area
MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
MA
Fig. 6 Floating roof and foam protection system
D
TE
P
CE
AC
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
Fig. 7 Floating roof tank and sealing equipme

MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

Table 1 Typical fire accidents of floating roof oil tanks trigged by lightning
Date Location Description
The effect of the monolayer seal was so poor that oil leakage was severe. The oil
Maoming,
03/08/1995 vapor was ignited by lightning and resulted in fire. The fire was put out quickly by
China
the automatic fire protection system.

PT
Ontario, The mixture inside the tank reached explosion limits. After the floating roof was
19/07/1996
Canada struck by lightning, the explosion turned over the roof and caused a full surface fire.
The continuous hot weather promoted the volatilization of oil, and excess leakage

RI
Wuhan, oil vapor spilled from the failed breathing valve and was collected at the top of the
13/07/1998
China roof. The flammable vapor was ignited by spark due to lightning and spread into the

SC
inside of tank via the vent valve, causing a full surface pool fire.
Louisiana, Heavy rain caused the floating roof to be tripped, and the lightning ignited the
07/06/2001
USA exposed oil directly, resulting in a pool fire.

NU
When inductive charge was struck by lightning and produced discharge, the oil
Yizheng,
07/08/2006 vapor between the primary seal and the secondary seal was ignited and caused a rim
China
seal fire. The fire was put out quickly by the automatic fire protection system.
MA
The flammable mixture between floating roof and oil was ignited by lightning,
Baishawan,
07/07/2007 causing a rim seal fire. The fire was put out quickly by the automatic fire protection
China
system.
When inductive charge was struck by lightning and produced discharge, the oil
D

Ningbo,
05/03/2010 vapor between the primary seal and the secondary seal was ignited and caused a rim
China
TE

seal fire. The fire was put out quickly by the automatic fire protection system.
When inductive charge was struck by lightning and produced discharge, the oil
vapor between the primary seal and secondary seal was ignited and caused a rim seal
P

Dalian,
22/11/2011 fire. Meanwhile, the power supplies of the fire protection system were damaged.
China
CE

After the explosion, a rim seal fire developed into a local pool fire. The fire was put
out by the fire brigade in a timely manner.
AC
 ACCEPTED MANUSCRIPT

Table 2 The value of the location factor of the building (CD)

Exposure degree of buildings and surrounding objects CD
Obviously lower than surrounding objects 0.25
Approach to surrounding objects 0.5

PT
Obviously higher than surrounding objects 1
Isolated buildings on the hill 2

RI
SC
NU
MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

Table 3 Probability of lightning burning through the oil tank shell

Electric charge
Current peak need for
Thickness/mm required for peroration Probability
peroration (I/kA)
(Q/C)

PT
4 27 95 8.20E-02
5 45 158 1.10E-02
6 78 273 2.00E-04

RI
7 127 442 1.18E-06
8 191 666 6.96E-09

SC
9 271 943 4.11E-11
10 367 1276 2.42E-13

NU
MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

Table 4 Failure probability of events in Fig. 2

ID Event in Fig. 2 Failure probability
A1 Water supply 8.83×10-4
A2 Proportioning system 1.16×10-3

PT
A3 Values on foam and water lines 6.24×10-4
A4 Undetected major leak 1.37×10-7
A5 Detection system 8.96×10-6

RI
A6 Logic solver 1.76×10-4
A7 Actuation pumps 9.57×10-5

SC
B1 Foam supply 1.00×10-4
B2 Foam maker 3.98×10-3
Global system 7.03×10-3

NU
MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

Table 5 Comparison for results of some sample tanks from a reference

Volume Diameter Height Thickness P-MC P-PM
ID Error/%
/m3 /m /m /mm events/years events/years
1 250 7.7 7.5 5 4.36×10-2 4.19×10-2 3.90

PT
-2
2 500 7.8 11.0 6 4.94×10 4.75×10-2 3.85
-2
3 750 10.5 9.0 7 4.77×10 4.57×10-2 4.19
4 1000 15.0 6.0 9 4.57×10-2 4.73×10-2 -3.50

RI
5 2500 20.0 5.4 11 4.77×10-2 4.86×10-2 -1.89
-2
6 5200 25.0 11.0 11 6.15×10 6.35×10-2 -3.25

SC
-2
7 7634 30.0 10.8 12 6.49×10 6.64×10-2 -2.31
-2
8 9975 42.0 7.2 12 6.67×10 6.87×10-2 -3.00
9 12367 54.0 5.4 13 7.15×10-2 7.42×10-2 -3.78

NU
-2
10 16303 66.0 5.4 15 8.09×10 8.32×10-2 -2.84
MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

Table 6 Failure probabilities of the automatic fire extinguishing system

System ID System description Calculated value
PF1 Automatic foam extinguishing system 2.37×10-3
PF2 Manual foam extinguishing system 7.03×10-3

PT
PF2 Fire brigade not arriving on time 1.18×10-1

RI
SC
NU
MA
D
P TE
CE
AC
 ACCEPTED MANUSCRIPT

Table 7 Probabilities of different forms of accidents

Accident scenario Categories A and C Category B
Rim seal fire 1.10×10-2 4.99×10-2
Local pool fire 10-8–10-6 10-8–10-5

PT
Full surface fire 2.17×10-9 9.85×10-8

RI
SC
NU
MA
D
P TE
CE
AC