Fire Safety Journal 92 (2017) 107–121

Contents lists available at ScienceDirect

Fire Safety Journal

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / fi r e s a f

A method for determining fire accidental loads and its application to

thermal response analysis for optimal design of offshore
thin-walled structures
Jung Kwan Seo a, *, Sang Eui Lee a, Joo Shin Park b
a
The Korea Ship and Offshore Research Institute, Pusan National University, Busan, South Korea
b
Central Research Institute, Samsung Heavy Industries Co., Ltd., Geoje, South Korea

A R T I C L E I N F O A B S T R A C T

Keywords: More than 70% of accidents that occur on offshore installations stem from hydrocarbon fire and explosion, and as
Accidental design load they involve heat and blast effects, they are extremely hazardous with serious consequences in terms of human
Fire risk analysis health, structural safety and the surrounding environment. To prevent further accidents, substantial effort has
Optimisation
been directed towards the management of fire and explosion in the safety design of offshore installations. The aim
Aluminium living quarter
of this paper is to present a risk-based methodology procedure to help determine the fire accidental design load of
an offshore installation (AL Living Quarter) in association with the thermal response characteristics for structural
optimisation. A probabilistic sampling approach with numerical fire simulations was taken to determine the fire
accidental load. To determine the optimisation of the thin-walled structures of the living quarter, an A60 based on
the results of thermal response analyses was conducted and the temperature distribution calculated. The analysis
results suggest incorporating both the design and safety planning aspects of offshore Living Quarter.

1. Introduction consequences of specific events within the framework of risk assessment

and management. Structural design and safety assessment both require
Oil and gas are important sources of energy and are mainly produced the identification of the characteristic actions and action effects of
in demanding oceanic and industrial environments with significant fire fire [16,17].
and explosion hazards. The topsides of offshore platforms are the struc- The thermal characteristics of steel are the main factors affecting
tures most likely to be exposed to hazards such as hydrocarbon fire and/ structural integrity in fire. Many researchers and reports have identified
or explosion [1,2]. More than 70% of accidents that occur on offshore that temperature varies with the thermal and mechanical properties
installations stem from hydrocarbon explosions and fires that involve associated with various steels and steel structures during fires. In the
blast effects and heat, and are thus extremely hazardous with serious conventional fire safety design of steel structures [11,14], it is usually
consequences in terms of human health, structural safety and the sur- assumed that a fire will heat every area of a structure with the same
rounding environment [3]. Fig. 1 shows examples of such accidents. The intensity. The heat intensity in terms of the heat flux is considered to
most significant fire and explosion events are those associated with hy- remain the same throughout the duration of a fire. However, this
drocarbon leaks from flanges, valves, seals, nozzles, etc. [4,5]. approach is too simplistic and does not describe the actual physics, which
Developing proactive measures to prevent the escalation of such also involve time- and space-variable radiation and convection. These
events thus requires detailed knowledge of the related phenomena and factors result in continuous changes over time in the amounts of
their consequences. The concern over the risk of fire is reflected in the combustible gas in fires.
current rules and designs for quantified risk assessment. Adequate The conventional fire safety design approaches are essentially
guidelines must be established for the assessment and management of composed of a series of regulations, standards and procedures. They must
risk [6–15]. therefore be supplemented by integrated fire safety design approaches
Generally, risk is defined as a product of frequency and consequence. that are in principle based on performance. Integrated fire safety design
Thus, the main task is to accurately calculate the frequency and requires taking advantage of fire computational fluid dynamics (CFD)

* Corresponding author.
E-mail address: [email protected] (J.K. Seo).

http://dx.doi.org/10.1016/j.firesaf.2017.05.022
Received 10 January 2017; Received in revised form 2 May 2017; Accepted 24 May 2017

0379-7112/© 2017 Elsevier Ltd. All rights reserved.

J.K. Seo et al. Fire Safety Journal 92 (2017) 107–121

Fig. 1. The Piper Alpha (left) and Deepwater Horizon (right) accidents.

Fig. 2. Flow chart of the design process.

simulations and nonlinear structural response analyses [18,19]. assessment (QRA) using KFX is conducted and the exceedance curves
Researchers have recently suggested that these procedures for quan- of temperature and heat flux are proposed.
titative fire risk assessment and management are associated with c. Heat transfer analysis of the AL thin-walled fire wall of Living
nonlinear structural response analyses, and have provided some exam- Quarter: to characterise the temperature distribution in the thin-
ples [20–26]. However, this approach has only been considered and walled structures under the design fire load, heat transfer analyses
examined in terms of steel frame structures in offshore platforms. The in steady and transient states are numerically conducted.
topsides of offshore platform typically include various modules (living d. Optimisation of AL thin-walled structures: to optimise the total
quarters, process modules, drilling modules, etc.) and their layout and height, a design of experiment (DOE) method, specifically a central
arrangements must be considered in terms of functional and safety composite design with a response surface method, is performed based
requirement. Most structural modules consist of basic structural members on the minimum temperature in a room of the thin-walled living
(beams, columns, plates, etc.), but living quarters are often thin-walled quarters, with a structural deflection.
structures (made from unstiffened or stiffened plate). The effect of acci-
dental design load and thermal response during fire should therefore 2. Optimal design of offshore thin-walled plates (living quarters)
be identified. due to fire
The objective of this study is to present a procedure for determining
fire accidental loads and thermal response analysis that is suited to the 2.1. Risk-based optimal design procedure
integrated fire safety design of offshore thin-walled structures (living
quarters). In this paper, the following are considered. This paper introduce fire design methodologies for offshore AL thin-
walled plate structures (living quarters). Fire loads are characterised by
a. An introduced risk-based optimal design procedure and fire risk- CFD simulations, followed by analyses of heat transfer and nonlinear
based design method. structural responses.
b. Fire accident simulation (Process and Utility area of the target Fig. 2 presents the procedure for a quantitative fire risk-based design
offshore installation): to define the design fire load, a quantitative risk method, as considered in the present study. Risk is defined as a product of
frequency and consequence. Thus, the main task is to accurately calculate

108
J.K. Seo et al. Fire Safety Journal 92 (2017) 107–121

Fig. 3. Schematic of offshore topside platform.

Table 1
Frequencies and probabilities of leak hole sizes [3].

Leak hole size Frequency (104) Probability (%)

Small (0.1–0.5 kg/s) 0.66 –

Intermediate (0.5–10 kg/s) 4.00 30.07
Medium (10–50 kg/s) 5.10 38.34
Large (>50 kg/s) 4.20 31.57

the frequency and consequences of specific events within the framework 3. Fire accident simulation
of risk assessment and management. Structural design and safety
assessment require the identification of both the characteristic actions 3.1. Scenario selection
(fire simulation) and action effects (heat transfer analysis and structural
analysis) of fire. Selecting credible fire scenarios is important in determining fire
The procedure for the fire risk-based design method includes the accidental loads. Numerous methods for selecting such scenarios are
following steps: 1) investigation of metocean data and functional re- available, including event tree analysis (ETA), standard random, strati-
quirements of target structure; 2) selection of credible fire scenarios with fied, Monte Carlo simulations (MCSs) and Latin hyper cube samplings
metocean and operation-related parameters using a probabilistic (LHSs) [27]. The LHSs use a probability density function for each vari-
approach; 3) obtaining fire loads (temperature, heat flux, etc.) of each able, which generates probabilistic scenarios with variable combinations.
fire scenario by computational fluid dynamics (CFD); 4) calculation of In this study, an LHS is used to select the fire scenarios, as it can effi-
fire accident frequencies of fire scenarios using historical data; 5) gen- ciently represent each whole scenario with a probabilistic approach. It
eration of fire (temperature, heat flux, flame length, etc.) exceedance can also cover all possible values even if they have very low probabilities.
curves with fire loads and frequencies and determination of design fire When selecting credible fire accident scenarios, all possible parame-
loads; 6) conducting structural consequence response analysis (or opti- ters that can have an effect on fire loads should be considered, such as
misation analyses of specific target structures) using the nonlinear finite temperature, heat flux and flame length. Fires can also be affected by
element method and/or an experiment, and investigating the conse- wind (direction and speed) and leak (rate, direction, duration and posi-
quences of fire accidents; 7) calculation of fire risk tion). In the case of a jet fire, scenarios should take into account the leak
[risk ¼ frequency  consequence]; and 8) decision making (redesign of rate, direction, duration and position associated with the release.
the structure or adoption of risk control options). In this study, credible jet fire scenarios are determined by sampling
techniques based on the following seven random variables, which satisfy
the minimum requirements: Wind speed (X1), Wind direction (X2), Leak
2.2. Target structure
rate (X3), Leak direction (X4), Leak position X (X5), Leak position Y (X6)
and Leak position Z (X7).
2.2.1. Offshore structure
When using these variables, the range of each must be defined and the
The target structure is a typical topside platform in the North Sea.
probability density functions (PDFs) identified so that a number of sce-
Fig. 3 shows the target topside platform including living quarters, utility
narios for fire simulation can be selected. Wind speed and direction are
area, process area and wellbay. It is assumed that the procedure is carried
obtained from wind rose data at the installation site.
out and focused on the living quarter external wall (which is functionally
Four leak hole sizes are defined by the manufacturing companies, as
designed for fire and blast resistance), and that the fire occurs in process
shown in Table 1. Small leak holes are assumed to be excluded from the

109
J.K. Seo et al. Fire Safety Journal 92 (2017) 107–121

Fig. 4. Probability density functions.

Table 2
Probabilities in leak direction.

Leak direction Probability (%)

þX 16.667
X 16.667
þY 16.667
Y 16.667
þZ 16.667
Z 16.667

scenarios due to their minor effect on temperature and heat flux. Fig. 4 test statistic was used to quantify how well the database follows a
shows the probability density of wind speed, wind direction and particular distribution. The better the distribution fits the data, the
leak rate. smaller the statistic. Table 3 summarises the PDFs of each parameter
It is assumed that six directions, þX, -X, þY, -Y, þZ and –Z, have equal applied. The Latin hypercube sampling technique shown in Fig. 7 is used
probabilities, as shown in Table 2. to select a number of credible fire scenarios. Table 4 represents the
The process area is assumed to be evenly divided into 10 sections in overview of twenty-five fire scenarios selected for the seven
the X direction, as shown in Fig. 5, and each section has the same random variables.
probability of a fire accident. The probability density of leak position on
the X axis is uniform (C1 ¼ 0.00002). Similar to the leak positions Y (X6) 3.2. Fire simulation
and Z (X7), it is assumed that the probability density is uniform
(C1 ¼ 0.000022 of Y and C1 ¼ 0.0000246 of Z), as shown in Fig. 6. The fire CFD simulation characterises the gas cloud dispersion, gas
The data analysis uses a database of selected offshore platforms and cloud temperatures and heat fluxes in a fire as time- and space-
the distribution functions are determined for all of the variable param- dependent. The fire load is physically correlated with the elevated tem-
eters. The probability density function that best represents the database peratures and heat fluxes in the gas cloud obtained by the fire CFD
was determined by a goodness of fit test. The Anderson–Darling http:// simulation [18]. The radiation and convection associated with fire are
www.sciencedirect.com/science/article/pii/S0010938X1200491X [28] important factors in characterising fire loads. A commonly adopted tool

110
J.K. Seo et al. Fire Safety Journal 92 (2017) 107–121

Fig. 5. Process area on X axis divided into 10 sections.

Fig. 6. Probability density of leak positions.

111
J.K. Seo et al. Fire Safety Journal 92 (2017) 107–121

Table 3
PDFs of each parameter.

Parameter PDF Values

Wind speed 2-Parameter Weibull α ¼ 9.904, β ¼ 2.307

Wind direction 3-Parameter Weibull α ¼ 169.3, β ¼ 1.758, γ ¼ 101.3
Leak rate 3-Parameter Weibull α ¼ 43.32, β ¼ 1.678, γ ¼ 0.5
Leak direction Uniform C1 ¼ 16.667
Leak position, X Uniform C1 ¼ 0.00002
Leak position, Y Uniform C1 ¼ 0.000022
Leak position, Z Uniform C1 ¼ 0.0000246

Fig. 8 illustrates a 3D CFD model of the offshore platform, and 0.1

million grids are chosen in this analysis. The detailed information of the
fuel in the fire simulation is given in Table 5.

3.2.1. Monitoring points

To obtain temperature and heat flux in the fire simulation, 473
monitoring points are selected in the target platform. To quantify the
effect on the living quarter, specific monitoring points are defined. Fig. 9
shows the definition of the monitoring points in the living quarter floor
level. To identify the temperature and heat flux topside, monitoring
points are spread out on each level of the deck.

3.2.2. Numerical results

For the selected 25 credible scenarios, fire simulations were con-
ducted using KFX. Fig. 10 illustrates the contours of jet fire in a numerical
fire simulation for Scenario 19. The fire was found to occur under the
living quarter, so it directly heated the living quarter floor and wall.
Fig. 11 describes temperature contour and the velocity vector field in
Scenario 19, which is one of the severest scenarios. It can be observed
that the temperature reaches 1235.9  C. Fig. 12 illustrates the tempera-
ture time histories in Scenario 19. Scenario 7 appears to have the highest
maximum temperature while Scenario 9 has the lowest.
Fig. 7. An example of Latin hypercube sampling technique [9].

3.2.3. Design fire load

for fire CFD simulations in current offshore industry practices is the KFX To generate a fire exceedance curve, both the fire load and fire fre-
code [29], which is a three-dimensional transient finite volume quency are required. Leak frequencies and ignition probabilities are
CFD programme. normally calculated as parts of a QRA, and details are described in

Table 4
Twenty-five scenarios selected from seven parameters.

Wind speed (m/s) Wind direction (deg) Leak rate (kg/s) Leak direction (deg) Leak position (m)

X Y Z

Scenario 1 10.738 317.45 8.4332 Y 13.0035 40.8152 18.7269

Scenario 2 15.5574 333.9594 45.9151 Y 19.005 22.6752 25.2405
Scenario 3 10.2382 182.7655 51.9601 Z 35.009 19.0472 33.3825
Scenario 4 8.4495 290.6708 77.1623 X 33.0085 20.8612 2.4428
Scenario 5 9.7658 174.3415 43.2156 -X 47.012 26.3032 10.5848
Scenario 6 6.3323 207.1678 14.0172 X 37.0095 35.3732 13.8417
Scenario 7 3.7256 268.6915 59.3864 -Y 3.001 15.4192 38.2678
Scenario 8 1.7183 190.9817 31.4897 X 41.0105 31.7452 36.6394
Scenario 9 6.7663 156.2801 27.2237 Z 9.0025 0.9072 39.8911
Scenario 10 7.1908 117.8756 25.1151 Y 23.006 4.5352 17.0985
Scenario 11 5.8834 20.9508 18.6757 -X 31.008 8.1632 31.7541
Scenario 12 11.8625 165.5789 4.3799 -Y 29.0075 11.7912 5.6996
Scenario 13 11.2749 64.5656 69.6392 -Z 1.0005 37.1872 23.6121
Scenario 14 2.9487 279.2427 33.6783 -Z 45.0115 9.9772 0.8144
Scenario 15 5.4123 146.1005 35.9254 -Z 15.004 39.0012 28.4973
Scenario 16 18.3903 215.284 40.6709 Z 25.0065 17.2332 4.0712
Scenario 17 12.5222 199.0896 38.2492 -Y 27.007 28.1172 7.328
Scenario 18 4.9087 354.1241 11.4165 -X 21.0055 42.6292 8.9564
Scenario 19 7.6104 240.5203 20.8622 Y 7.002 44.4432 15.4701
Scenario 20 14.2403 303.2607 48.811 Y 5.0015 24.4892 12.2133
Scenario 21 9.3138 258.8145 23.0009 -Y 11.003 33.5592 35.011
Scenario 22 8.0286 249.4622 29.3435 X 43.011 29.9312 21.9837
Scenario 23 4.3564 134.2974 63.9916 -Z 17.0045 13.6052 26.8689
Scenario 24 13.2906 231.894 55.4439 X 39.01 2.7212 30.1257
Scenario 25 8.8766 223.5046 16.4098 Z 49.0063 6.3492 20.3553

112
 Fig. 8. Geometry model (left) and CFD model (right) for fire simulation.

Table 5
Fuel information.

Name Molecular Formula Mole Percent (%)

Water H20 0.01

Nitrogen N2 0.38
Carbon dioxide CO2 2.24
Methane CH4 81.05
Ethane C2H6 5.71
Propane C3H8 5.32
N-Butane C4H10 3.11
N-Pentane C5H12 1.37
Hexane C6H14 0.55
Heptane C7H16 0.2
Octane C8H18 0.05
Nonane C9H20 0.01

Fig. 9. Definition of monitoring points in the living quarter floor level.

Fig. 10. The example of fire simulation in Scenario 19.

113
J.K. Seo et al. Fire Safety Journal 92 (2017) 107–121

Fig. 11. Example of temperature contour (left) and velocity field (right) in Scenario 19.

Fig. 12. Example of temperature to time history for Scenario 19.

Fig. 13. Fire exceedance curves of temperature and heat flux for living quarter wall.

114
J.K. Seo et al. Fire Safety Journal 92 (2017) 107–121

guidelines and standards such as references. It should be noted that leak

frequencies must be detailed for each sub-segment. The fire frequency
can be calculated using Eq. (1).

FFire ¼ FRelease x PIgnition (1)

where FFire ¼ fire frequency, FRelease ¼ leak frequency and

PIgnition ¼ ignition probability.

HSE [3,28] provides historical data of the leak hole size of the process
equipment in offshore installations. Leak frequencies are considered in
the present study, as the inventory of the process module is defined as the
mass of hydrocarbon (HC) in a segment (kg) and is calculated for
each segment.
Source or release models [30–33] are used to estimate the amount of
fuel released, or the rate of release of the fuel. These models are crucial in
Fig. 14. Example of how to determine design fire load at acceptance criteria, 104. the risk assessment process, as the release rate and quantity of fuel
released determine the size of the resulting cloud and hence the proba-
bility of ignition. The models are also used to find the initial sizes of fires
and explosions. The initial release rate through a leak depends mainly on
the pressure inside the equipment, the size of the hole and the phase of
release. Offshore hydrocarbon releases are usually gaseous, liquid and
two-phase. The gases can be hydrocarbons ranging from C1 to C4, while
liquids can be crude oil, diesel oil, aviation fuel and others. In this study,
the initial probability used is taken from a study by Cox et al. [34] and
from the Oil and Gas [35,36] model.
To determine the design fire load, the exceedance of temperature and
heat flux from the numerical results must be defined. The figures illus-
trate an example of fire exceedance curves of maximum temperature and
heat flux by using Cox and oil and gas models (see Fig. 13). Once the
exceedance curve is derived, the design fire load can be determined at an
acceptable level of exceedance. Fig. 14 gives an example of how to
determine design fire load. As shown in Fig. 14, the design fire load in
this project is 484.8  C at a 104 acceptance level.

Fig. 15. A typical section of a living quarter fire and blast wall.

Table 6
Dimensions of fire and blast wall.

Height (mm) Dim. Length (mm) Dim. Thickness (mm) Dim.

Fire wall, h1 45.0 Fire wall, l1 100.0 Fire wall, tFW 1.2
Air, h2 75.0 Air, l2 23.0 Fire plate, tFP 3.0
Insulation, h3 70.0 Insulation, l3 66.8 Blast wall, tBW 4.3
Blast wall, h4 100.0 Blast wall, l4 181.0 Blast plate, tBP 9.0
Angle bar, h5 75.0 Angle bar, l5 75.0 Angle bar, tAB 9.0

Fig. 16. Specific heat and thermal conductivity of Al6082 and mineral wool.

115
J.K. Seo et al. Fire Safety Journal 92 (2017) 107–121

Fig. 17. Reduction factor of mechanical properties for Al6082. Fig. 19. ISO standard and the design fire load temperature load curve.

Table 7 4. Optimisation of a thin-walled living quarter (fire and blast

Reduction factors for Al6082. wall) structure
Temperature ( C) Reduction factors (20  C)
4.1. Target living quarter fire and blast wall structure
Yield strength Modulus of elasticity

20 1 1 The benefits of aluminium in the design of offshore structures are

100 0.9 0.97
significant in terms of weight and corrosion resistance, and there have
150 0.79 0.93
200 0.65 0.86 been various attempts to design aluminium structures such as ladders and
250 0.38 0.78 helidecks. In this study, for the fire and blast wall, the living quarter is the
300 0.2 0.68 target structure. The target external wall structure, which has a fire rating
350 0.11 0.54
of A60, can be insulated, as shown in Fig. 15. Table 6 provides details of
550 0 0
the dimensions of the target structure.
It is important to identify the characteristics of the heat transfer ca-
Table 8 pacity of the target external wall, which has an impairment criterion in its
Mechanical properties of Al6082. fire rating of A60 [37–39]. The structure should also be optimised based
Properties Values on the numerical results of steady state and transient thermal analyses.
Tensile strength, ultimate (t < 5 mm) 290 (MPa)
ANSYS/WORKBENCH was used for thermal analysis and optimisation.
Tensile strength, ultimate (t > 5 mm) 310 (MPa)
Tensile strength, yield (t < 5 mm) 250 (MPa)
4.2. Finite element modelling
Tensile strength, yield (t > 5 mm) 260 (MPa)
Density 2700 (kg/m3)
Young's modulus 70 (GPa) 4.2.1. Material modelling
Coefficient of thermal expansion 23.4  106 (1/K) In steady and transient thermal analysis, material properties vary
Poisson's ratio 0.33 according to the corresponding temperature. The material is Al6082 and
Shear strength 210
the insulation is mineral wool. The material properties such as specific

Fig. 18. Boundary conditions of thermal analysis (left) and static structural analysis (right).

116
 Fig. 20. Generated mesh model for thermal (left) and static structural (right) analyses.

heat and thermal conductivity and the reduction factor of yield strength
and elastic modulus are defined as shown in Figs. 16 and 17. The specific
heat of mineral wool is assumed to be at a constant temperature of up to
1200  C. The mechanical properties of Al6082 are given in Tables 7 and
8 [40].

4.2.2. Boundary conditions

Applied boundary conditions for thermal and static structural analysis
are presented in Fig. 18. The length of the symmetry boundary is 4.6 m.
For the top and bottom sides, a fixed condition is applied and a gravity of
9.8 m/s2 acts on the structure in the Z direction.
The design fire load maximum temperature is 484.8  C. In steady
state thermal analysis, the maximum temperature is applied as a constant
temperature.
In transient thermal analysis, the fire accidental load curve, which is
developed from the previous CFD results and the ISO standard fire curve
Fig. 21. Applied parameters of the fire and blast wall, A60. (ISO-834), is applied [41] as shown in Fig. 19. This can enable an

Table 9
Range of applied parameters.

Parameters Scope

Fire wall height, h1 10 mm < h1 < 55 mm

Angle bar height, h2 30 mm < h2 < 100 mm
Insulation height, h3 20 mm < h3 < 100 mm
Length of fire corrugation, l1 80 mm < l1 < 200 mm

Table 10
Selected 25 scenarios for optimisation.

Design point h1 h2 h3 l1

D1 10 30 20 80
D2 10 30 100 80
D3 10 100 20 80
D4 10 100 100 80
D5 10 65 60 140
D6 10 30 20 200
D7 10 30 100 200
D8 10 100 20 200
D9 10 100 100 200
D 10 32.5 65 60 80
D 11 32.5 30 60 140
D 12 32.5 65 20 140
D 13 32.5 65 60 140
D 14 32.5 65 100 100
D 15 32.5 100 60 60
D 16 32.5 65 60 60
D 17 55 30 20 20
D 18 55 30 100 100
D 19 55 100 20 20
D 20 55 100 100 100
D 21 55 65 60 60
D 22 55 30 20 20
D 23 55 30 100 100
D 24 55 100 20 20
D 25 55 100 100 100

117
J.K. Seo et al. Fire Safety Journal 92 (2017) 107–121

Fig. 22. Sensitivities of parameters.

Table 11
Comparison of optimisation.

Item Original Optimum Ratio (%)

Fire Wall, h1 45.0 45.0 –

Air, h2 75.0 30.0 60.0%
Insulation, h3 70.0 74.0 5.7%
Blast Wall, h4 100.0 100.0 –
Total, h (mm) 294.3 253.3 13.9%
Fire Wall Length, l1 (mm) 100.0 105.0 5.0%
Weight, WTotal (kg) 116.2 114.5 1.5%
TR (oC) Steady 276.4 276.4 –
Transient 35.5 36.5 2.8%
δMax (mm) Steady 319.3 320.7 0.4%

Fig. 23. Temperature contours of original (left) and optimum (right) results.

118
J.K. Seo et al. Fire Safety Journal 92 (2017) 107–121

understanding and evaluation of the effect of an accurate fire curve in

both approaches within a risk-based design framework. The applied film
coefficient in the convection boundary condition is 5  104 W/mm2oC.
Transient thermal analysis is conducted during 1.0 h (equal to the
A60 rating).
Fig. 20 illustrates an example of generated mesh for thermal analysis.
The applied mesh size is 10 mm for thermal and static structural analyses.
Approximately 600,000 and 60,000 meshes are used in the thermal and
static structural analyses, respectively.

4.3. Optimisation

The aim of the optimisation study is to minimise the weight and

height of the wall and the temperature in the room by changing the di-
mensions of the wall. The design exploration uses the CCD method with
RSM in ANSYS [42]. The parameters of optimisation are given in Fig. 21,
and their range is given in Table 9. For this study, 25 cases with different
range of height and length dimensions are selected, and they are sum-
marised in Table 10.
Fig. 22 describes the sensitivities of each parameter to the weight and
height of the wall and the temperature in the room. The height of the
insulation is a primary parameter for reducing the temperature in a room,
and the length of the wall corrugation significantly increases the weight.
Table 11 shows the results of the optimisation. The total height h is
reduced by 13.9% and the total weight is reduced by 1.5%. Fig. 23
provides a comparison of the temperature contours of the original and
the optimum.
Fig. 24(a) represents the temperature distribution for the original and
the optimum models along the path P1 from the fire wall to the blast plate
where bolting is placed. Fig. 24(b) shows the path P2, which is 250 mm
from the angle bar in the Z direction where there is no bolting. In both
steady and transient thermal analyses, the temperatures around the blast
wall appear similar. Fig. 25 shows the total deformation and stress dis-
tribution for the original and optimum designs under thermal loads
calculated from the steady state thermal analysis.

5. Conclusions and remarks

The objective of this paper is to present a procedure for determining

fire accidental loads and thermal response analysis that is suited to the
integrated fire safety design of offshore thin-walled structures (living
quarters). The CFD code for fire simulation was used to determine fire
accidental loads, while the nonlinear FE code was used for the thermal
response and optimisation of the fire wall system for part of the living
quarter. Based on the results of this study, the following conclusions and
insights were achieved.

(a) A risk-based optimal design procedure was introduced to deter-

mine the design fire load, based on the numerical results. Twenty-
five credible scenarios were selected and then applied for the fire
simulations. The exceedance curves of maximum temperature and
heat flux were thus derived. Finally, at the acceptance probability
of 104, the design fire load was defined.
Fig. 24. Temperature distribution along paths P1 and P2 for the original and opti-
mum designs.
(b) To characterise the temperature distribution, thin-walled struc-
tures under the design fire load and heat transfer analysis in
steady and transient states were successfully completed for the
target fire wall structures. The new method was found to be useful
for analyses of heat transfer and the nonlinear structural responses
of offshore installations in fires, which are key elements for inte-
grated fire design in association with quantitative fire risk
assessment and management.
(c) Optimisation of AL thin-walled structures were suggested as A60,
taking into account the effect of thermal load. Steady state and
transient thermal analyses were performed to achieve this.

A follow-up study is on-going to apply the new procedure to the fire

119
J.K. Seo et al. Fire Safety Journal 92 (2017) 107–121

Fig. 25. Total deformation and von Mises stress distribution for original (left) and optimum design (right).

safety engineering and design of real offshore oil/gas installations, which [12] A. Ahmadn, S.A. Hassan, A. Ripin, M.W. Ali, S. Haron, A risk-based method for
determining passive fire protection adequacy, Fire Saf. J. 58 (2013) 160–169.
are composed of a combination of strong support members and a thin-
[13] S.J. Kim, J. Lee, J.K. Paik, J.K. Seo, W.H. Shin, J.S. Park, A study on fire design
walled plate panel system, in association with the optimum design of accidental loads for aluminum safety helidecks, Int. J. Nav. Archit. Ocean Eng. 8
passive fire protection. (2016) 519–529.
[14] J.A. Purkis, Fire Safety Engineering – Design of Structures, second ed., Elsevier,
New York, 2006.
Acknowledgement [15] J.K. Seo, D.C. Kim, Y.C. Ha, B.J. Kim, J.K. Paik, A methodology for determining
efficient gas detector locations on offshore installations, Ships Offshore Struct. 8 (5)
(2013) 524–535.
This research was supported by Basic Science Research Program [16] DNV (DNV-OS-C101), Design of Offshore Steel Structures, General (LRFD Method),
through the National Research Foundation of Korea(NRF) funded by the Det Norske Verita, Oslo, 2011.
Ministry of Science, ICT & Future Planning(NRF-2017R1A2B4004891) [17] S.E. Hirdaris, W. Bai, D. Dessi, A. Ergin, X. Gue, O.A. Hermundstad, R. Huijsmans,
K. Iijima, U.D. Nielsen, J. Parunov, N. Fonseca, A. Papanikolaou, K. Argyriadism,
and by the Ministry of Trade, Industry and Energy (MOTIE) and Korea A. Incecik, Loads for use in the design of ships and offshore structures, Ocean. Eng.
Institutes for Advancement of Technology (KIAT) through the Promoting 78 (2015) 131–174.
Regional Specialized (A010400243). [18] J.K. Paik, B.J. Kim, J.S. Jeong, S.H. Kim, Y.S. Jang, G.S. Kim, J.H. Woo, Y.S. Kim,
M.J. Chun, Y.S. Shin, J. Czujko, CFD simulations of gas explosion and fire actions,
Ships Offshore Struct. 5 (1) (2010) 3–12.
References [19] N.K. Shetty, C.G. Soares, P. Thoft-Christensen, F.M. Jensen, Fire safety assessment
and optimal design of passive fire protection for offshore structures, Reliab. Eng.
[1] J. Czujko, J.K. Paik, Paradigm change in safety design against hydrocarbon Syst. Saf. 61 (1998) 139–149.
explosions and fires, FABIG Newsl. 60 (2012) 20–38. [20] C.G. Soares, J.M. Gordo, A.P. Teixeira, Elasto-plastic behaviour of plates subjected
[2] J. Czujko, Design of Offshore Facilities to Resist Gas Explosion Hazard: Engineering to heat loads, J. Constr. Steel Res. 45 (2) (1998) 179–198.
Handbook, CorrOcean ASA, 2001. [21] C.G. Soares, A.P. Teixeira, Strength of plates subjected to localized heat loads,
[3] UK HSE, Offshore Hydrocarbon Releases Statistics and Analysis, UK HSE, 2002. J. Constr. Steel Res. 53 (2000) 335–358.
[4] D.P. Nolan, Handbook of Fire and Explosion Protection Engineering Principles for [22] B. Skallerund, J. Amdahl, Nonlinear Analysis of Offshore Structures, Research
Oil, Gas, Chemical, and Related Facilities, Noyes Publications, 1996. Studies Press, Hertfordshire, UK, 2002.
[5] A. Salem, Fire engineering tools used in consequence analysis, Ships Offshore [23] J.K. Paik, J. Czujko, J.H. Kim, S.I. Park, M.D.S. Islam, D.H. Lee, A new procedure for
Struct. 5 (2) (2010) 155–187. the nonlinear structural response analysis of offshore installations in fires, Trans.
[6] J.E. Vinnem, Offshore Risk Assessment: Principles, Modelling and Applications of SNAME 121 (2013) 224–250.
QRA Studies, second ed., Springer, London, UK, 2007. [24] J. Yanlin, B.S. Jang, Probabilistic fire risk analysis and structural safety assessment
[7] NORSOK Z-013, Risk and Emergency Preparedness Assessment, third ed., 2010. of FPSO topside module, Ocean. Eng. 104 (2015) 725–737.
Oslo, Norway. [25] J. Zhang, S. Lu, C. Li, R.K.K. Yuen, Performance of overall heat transfer coefficient
[8] D.P. Nolan, Handbook of Fire and Explosion Protection Engineering Principles for and exploring heat transfer through the ceiling vent of compartment fire in ship
Oil, Gas, Chemical, and Related Facilities, Noyes Publications, New Jersey, 1996. structures with A60 constructions, Ships Offshore Struct. 10 (3) (2015) 328–334.
[9] J.K. Paik, J. Czujko, Assessment of hydrocarbon explosion and fire risks in offshore [26] A.M. Salem, E.M. Dabess, A.A. Banawan, H.W. Leheta, Fire safety design of Nile-
installations: recent advances and future trends, IES J. Part A Civ. Struct. Eng. 4 (3) floating hotels, Ships Offshore Struct. 11 (5) (2016) 482–500.
(2011) 167–179. [27] K.Q. Ye, Orthogonal column Latin hypercubes and their application in computer
[10] J.K. Paik, J. Czujko, Engineering and design disciplines associated with experiments, J. Am. Stat. Assoc. 93 (444) (1998) 121–128.
hydrocarbon and fire risks in offshore oil and gas facilities, SNAME Trans. 120 [28] T.W. Anderson, D.A. Darling, A test of goodness-of-fit, J. Am. Stat. Assoc. 49 (1954)
(2012) 1–39. 765–769.
[11] J.M. Franssen, P.V. Real, Fire Design of Steel Structures, ECCS Eurocode Design [29] KFX, User's Manual for Kameleon Fire EX, Computational Industry Technologies AS,
Manuals, Ernst & Sohn, Berlin, Germany, 2010. Stavanger, Norway, 2013.

120
J.K. Seo et al. Fire Safety Journal 92 (2017) 107–121

[30] UK HSE, Marine risk Assessment, Offshore Technology Report No. OTO 2001/063, [36] UKOOA, Fire and Explosion Guidance Part 2: Avoidance and Mitigation of Fires,
Health and Safety Executive, 2001. Revision 2, 2006.
[31] CCPS, Guidelines for Chemical Process Quantitative Risk Analysis, American [37] DNV (DNV-OS-D301), Fire Protection, Offshore Standard, DNU, 2008.
Institute of Chemical Engineers, New York, 2000. [38] EUROCODE, Design of Aluminium Structures Part 1-2: General Rules – Structural
[32] D.A. Crowl, J.F. Louvar, Chemical Process Safety: Fundamentals with Applications, Fire Design, European Committee for Standardization CEN, Brussels, Belgium,
Prentice Hall International Series in the Physical and Chemical Engineering 1999.
Sciences, Prentice-Hall Inc., New Jersey, 2002. [39] Morgan, Fire Master Marine Plus Blanket, South America, 2014.
[33] M. Mike, Development of algorithms for predicting ignition probabilities and [40] Nedal Aluminium, Alloy Data Sheet EN AW-6082, Netherlands, 2015.
explosion frequencies, J. Loss Prev. Process Ind. 24 (2011) 259–265. [41] ISO (ISO 834–11:2014), Fire Resistance Tests – Elements of Building Construction –
[34] A.W. Cox, F.P. Lees, M.L. Ang, Classification of Hazardous Locations, Institution of Part 11: Specific Requirements for the Assessment of Fire Protection to Structural
Chemical Engineers, Rugby, UK, 1991. Steel Elements, International Organization for Standardization, 2014.
[35] OGP, Risk Assessment Data Directory: Ignition Probabilities, Report No. 434–6.1, [42] ANSYS Inc, ANSYS Mechanical Theory Reference: Release 15.0, 2014. Canonsburg
International Association of Oil & Gas Producers, London, UK, 2010. PA, USA.

121