Sie sind auf Seite 1von 18

Applied Thermal Engineering 93 (2016) 279–296

Contents lists available at ScienceDirect

Applied Thermal Engineering


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 / a p t h e r m e n g

Research Paper

Calculation of Boil-Off Rate of Liquefied Natural Gas in Mark III tanks


of ship carriers by numerical analysis
Mario Miana a,*, Regina Legorburo a, David Díez a, Young Ho Hwang b
a ITAINNOVA, Instituto Tecnológico de Aragón, Materials & Components, María de Luna 7-8, 50018 Zaragoza, Spain
b
DONGSUNG FINETEC, 120, Hyeopdongdanji-gil, Miyang-myeon, Anseong-si, Gyeonggi-do 456-843, Republic of Korea

H I G H L I G H T S

• Four numerical approaches calculate Boil-Off Rate for Mark III tanks of an LNG ship.
• Approaches 1 and 2 define 2D and 3D models with selected temperature configurations.
• A reduced order model (ROM) is defined by an equivalent thermal conductivity.
• Approaches 3 and 4 apply this ROM to 2D and 3D sections of tanks.
• BOR varies with insulation thickness: 0.0919% for 270 mm and 0.0631% for 400 mm.

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

Article history: The heat flow from environment to LNG stored in Mark III of ship carriers is calculated in this paper by
Received 1 June 2015 numerical simulations. Four different approaches are defined and evaluated: Approach 1 starts from simple
Accepted 13 September 2015 2D numerical computations of heat fluxes over representative sections of the insulation barriers in 10
Available online 9 October 2015
specific temperature configurations defined by published data. Approach 2 evolves toward full 3D simu-
lations of insulations layers under the same temperature configurations. A Reduced Order Model is next
Keywords:
developed by calculating equivalent thermal conductivity for insulation barriers. This equivalent thermal
Liquefied Natural Gas
conductivity is applied in the fluid flow and heat transfer simulation from the environment to the LNG
Boil-Off Rate
Reduced Order Model in 2D and 3D models by Approaches 3 and 4, respectively. For a full ship with a capacity of 165,000 m3
Heat transfer with 270 mm thick insulation barriers, the obtained BOR and the overall heat transfer coefficient vary
Computational fluid dynamics from 0.895% and 0.0656 W/m2·°C for Approach 2 to 0.0945% and 0.0693 W/m2·°C for Approach 3. For Ap-
Finite element methods proach 4, the BOR and overall heat transfer coefficient are 0.0919% and 0.0674 W/m2·°C. When the thickness
of the insulation barrier is increased to 400 mm, these initial values are reduced to 0.0631% and
0.0453 W/m2·°C.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction pacity of pipeline systems; and third, the potential problems due
to unstable political situations when international pipelines entail
The growth of Natural Gas (NG) as an energy source after the the crossing of a number of countries and borders [2]. However, LNG
oil shock in the early 1970s led to an ever rising consumption of is not without its own problems. Liquefied Natural Gas is carried
Liquefied Natural Gas (LNG). By 2013, Natural Gas accounted for 25% at cryogenic temperatures near its saturation temperature, typi-
of the global energy consumption [1]. NG is transported from re- cally about −162 °C, and this implies a net heat transfer from the
serves to consumer zones by pipelines or shipped in liquid state or environment, in spite of the excellent insulation of ship tanks. This
as compressed gas. Liquefied Natural Gas covers 10% of the global heat flow yields the generation of a boil-off gas that must be vented
demand for NG, so the global LNG fleet has grown steadily: it stood out of the tank to avoid large rises in pressure that could damage
at 357 vessels in 2013 and 31 new vessels were scheduled for de- its mechanical structure. Moreover, evaporation is not homoge-
livery in 2014 [1]. neous because LNG is a mixture of hydrocarbons and nitrogen. The
The advantages of shipped LNG over pipeline systems are well most volatile components (nitrogen and methane) evaporate first,
known: first, a greater adaptability to cover the growing distances which is known as the LNG ageing or weathering. These methane
from reserves to consumer zones; second, the limiting export ca- losses imply worse liquid qualities when the ships arrive at the
regasification terminal [3]. The evaporation rate of LNG during ship-
ping is often characterized by the BOR parameter, which is the
* Corresponding author. Tel.: +34 976 01 11 57; fax: +34 976 01 18 81. percentage of evaporated LNG per day with respect to the initial
E-mail address: mmiana@itainnova.es (M. Miana). LNG loaded.

http://dx.doi.org/10.1016/j.applthermaleng.2015.09.112
1359-4311/© 2015 Elsevier Ltd. All rights reserved.
280 M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296

In summary, the BOR prediction is the main figure to measure ranging from 1.2 mm to 40 m can lead to creating grids of up to 7·1012
the heat transfer from the environment to LNG stored in tanks for cells to represent a quarter of a 40,000 m3 capacity ship tank.
shipping. The BOR parameter plays a major role in the economics As the number of computational cells is increased, the time to
of the LNG trading, since the evaporation of the methane con- run simulations is extended and the difficulties in analyzing such
tained in LNG involves significant energy and financial losses [4]. a detailed domain arise. By way of a comparison, one of the largest
To minimize this heat transfer, the walls of Cargo Containment numerical grids ever solved was composed of 4·1012 cells [21]. There-
Systems (CCS) are composed of superb insulation layers made of fore, detailed numerical simulations of all geometric features of LNG
perlite in NO-96 tanks or plywood, reinforced polyurethane foams tanks are not affordable. Moreover, a very large mesh also re-
(R-PUF) and membranes in Mark III tanks [5]. However, the calcu- quires an extended computational time to be converged, and the
lation of the overall heat flow through the insulation barriers is not detailed information provided by such an enormous amount of data
a trivial task, given the following difficulties: is not of interest for the analysis of design modifications, where
overall conclusions are more useful in the design stages than com-
• The large dimensions of the tanks (~40 m long) compared to the prehensive information about every point. Previous numerical studies
thickness of the insulation barriers (~0.27 m) or the thickness on heat transfer in LNG ship tanks [22–27] were based on finite dif-
of the stainless steel membrane in contact with the LNG ference approximations, only considering the conduction through
(~1.2 mm). the different insulation layers.
• The large number of involved heat transfer phenomena: LNG con- To overcome all these inconveniences, this paper proposes a set
vection inside of the tank; conduction through insulation layers; of 4 efficient approaches to predict the BOR in Mark III tanks com-
natural convection in ballast compartments of irregular shapes; bining detailed numerical simulations of fluid flow and heat transfer
forced convection from ship to water and air environment. and Reduced Order Models. Reduced Order Models (ROMs) are a
• The measurement of thermal properties for materials in a wide group of well-established technologies to reduce the size of the com-
temperature range, ranging from −162 °C to +45 °C. putational model with a minimum loss of accuracy [28]. ROMs have
• The lack of detailed validation results for on-board LNG tanks. been successfully applied in a wide number of applications includ-
ing transient heat transfer phenomena [29,30] and aerodynamic
To overcome these difficulties, numerical simulations are pro- loads [31].
posed in this paper as the basic tool to predict the BOR parameter The paper is structured as follows: Section 2 describes the
for a 165,000 m3 capacity LNG ship carrier with 4 Mark III tanks. problem to be investigated and Section 3 defines the developed
The scientific literature on numerical simulations for LNG tanks methodologies; Section 4 collates the results achieved by the dif-
covers different aspects among which are stress analysis [6,7], studies ferent approaches and judges their advantages and disadvantages.
on vibration caused by the sloshing phenomena [8], or the thermal From these results, a new design modification is proposed and evalu-
analysis of representative sections of a tank [9,10]. Chen et al. [11] ated consisting of an increase in the thickness of the insulation
analyze the pressure and temperature changes in LNG storage tanks barrier. Finally, Section 5 summarizes the main conclusions drawn
by means of dynamic models while References [12–15] estimate from this investigation and outlines the planned future work.
the BOR using computational fluid dynamics assuming an overall
heat transfer coefficient of 0.411 kJ/h·m2·°C. Experimental results
have been obtained only for very simplified geometries under lab- 2. Definition of the problem
oratory conditions [16–20].
It should be noted that the solution of the energy transport equa- 2.1. Description of Mark III tanks on LNG ship carriers
tion by numerical methods in any system needs the definition of
a computational grid able to capture the main geometric features, The problem to be investigated is the heat transfer from the ex-
such as the different insulation layers to be considered with Mark ternal environment to the LNG stored inside a 165,000 m3 capacity
III tanks. Although mesh sizes can be enlarged in zones with small ship carrier. The LNG ship contains 4 Mark III tanks following the
temperature changes, smooth transitions between the different gap arrangement and the prismatic shapes shown in Fig. 1 and having
sizes are recommended. The previously defined typical length scales the characteristic dimensions collected in Table 1.

(a)

(b) (c)

Fig. 1. (a) Tank layout; (b) Tank 1: dimensions and names of edges; (c) Tanks 2, 3 and 4: dimensions and names of Flat Panels (Flat Panel 2 is the base of the tank).
M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296 281

Table 1 Table 2
Characteristic dimensions of tanks. Thermal conductivity of solid materials.

Tank Insulation 1 2, 3 4 Material Temperature (°C) λ (W/m·°C)


thickness
Epoxy – 2.7
Net volume at 100% 270 mm 27,009 49,391 40,713 Plywood [10] 10 0.12
loaded (m3) 400 mm 26,335 48,353 39,826 Stainless steel [34] −163 10.146
Breadth (m) 34.4 38.5 38.5 Steel – 45
Height (m) 27.9 27.9 27.9 Triplex – 0.35
Length (m) 33.7 47.0 41.2 R-PUF [35] −160 0.013
−80 0.019
0 0.022
40 0.025
According to IMO regulations [32], ships for carrying liquefied Glass wool −160 0.014
−80 0.022
gases in bulk must be of the double hull type. The external hull is
0 0.031
50 mm thick steel and the inner hull is 18 mm thick steel. The space 40 0.035
between the inner and the outer hull is divided into ballast com-
partments and is also composed of 18 mm thick steel. These
compartments are assumed to be filled up with air. The insulation
barriers are directly attached to the inner hull by a mastic layer. The divided into two Flat Panels: FP 6, wetted by LNG, and FP 7, wetted
prismatic shape of the tanks is fitted with insulation layers by de- by NG. The name of each edge comprises two figures: the first one
fining four different bodies, called Flat Panels, 90° and 135° Corners indicates the angle of the edge and the second one represents the
and Trihedrons. These bodies are composed of a primary barrier, name of one of the attached Flat Panels. Edges starting from 1 cor-
directly in contact with LNG, and a secondary barrier which is thicker respond to 90° Corners while edges starting from 3 correspond to
than the primary barrier. Between these barriers, a 0.6 mm thick 135° Corners which are of the same length as the tank. Lastly, edges
impermeable triplex layer prevents the LNG from leaking through lining to the cofferdams are called CDAMS Edges, corresponding to
the outside. The primary barrier is composed of a 1.2 mm thick cor- 90° Corners, likewise.
rugated stainless steel membrane, a plywood layer and the R-PUF
layer. The secondary barrier is composed of R-PUF supported by two
2.2. Properties of the materials
plywood layers. The initial detailed description of these CCS can be
found in Roni and Chauvi [33]. Fig. 2 shows the assembly of the Flat
The insulation barriers and the external and inner hulls are com-
Panels and Corners.
posed of 7 different materials; their thermal conductivities are given
The nomenclature of the different Flat Panels and Corners is de-
in Table 2. The thermal conductivity of the plywood, the stainless
scribed in Fig. 1b and c. It depends on their relative position to the
steel and the R-PUF are obtained from References [10,34,35]. For
seawater or the air. The vertical walls are Flat Panel 1, and they line
the rest of materials, thermal conductivity was directly provided
to the cofferdam, which is the space between two consecutive tanks,
by suppliers. For the air filling the compartments between the inner
or directly to environment (seawater and air) at the prow and the
and the outer hull, density is modeled using Eq. (1), following the
stern of the ship. The bottom wall is named as Flat Panel 2, and the
Boussinesq approach:
top wall is Flat Panel 8. The sidewalls are Flat Panels 3–7, starting
from the bottom wall. The water line divides the vertical sidewall ρ = ρ0 [1 − β ⋅ ( T − T0 )] (1)
of Tanks 2, 3 and 4 into Flat Panel 4, under the water line, and Flat
Panel 5, over the water line. For Tank 1, the water line falls within where ρ0 is the reference density (1.209 kg/m3), β is the thermal ex-
FP 3, so there is no FP 4. The lateral inclined top wall is further pansion coefficient (0.00348 °C−1) and T0 is the reference temperature

(a) (b) (c)

Fig. 2. Insulation barriers: (a) Flat Panels; (b) 90° Corner; (c) 135° Corner and Trihedron.
282 M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296

Table 3 3. Numerical approaches for BOR calculation


Air properties as a function of temperature [36].
Four successive approaches are developed to obtain the heat flow
Temperature (°C) λ (W/m·°C) CP (J/kg·°C) μ (kg/m·s)
transferred from the environment to the LNG based on simplified
−20 0.02281 1005.74 1.620·10–5
numerical simulations and Reduced Order Models of heat transfer
0 0.02436 1005.90 1.722·10–5
20 0.02587 1006.36 1.821·10–5 through the insulation barriers of LNG tanks. The proposed ap-
27 0.02640 1006.60 1.854·10–5 proaches can be classified into two groups depending on their
35 0.02699 1006.92 1.893·10–5 working procedure and the applied geometrical description, as
45 0.02772 1007.39 1.940·10–5 shown in Fig. 3.
Approaches 1 and 2 use the same compounding method: both
approaches obtain the heat flux transferred through Flat Panels and
(15 °C). The specific heat, the thermal conductivity and viscosity are 90° and 135° Corners for some selected configurations through nu-
obtained by linear interpolation with temperature using the values merical simulations. These heat fluxes are then multiplied by the
tabulated in Table 3 [36]. total heat transfer area giving the net heat flow transferred through
global locations, for example, Flat Panel 1 or Edge 3–6. The sum of
2.3. Sailing conditions all heat flows transferred through the walls of the tanks will provide
the input heat flow that will be used to calculate BOR. The differ-
IMO regulations [32] set the sailing conditions considering that ence between Approaches 1 and 2 is that Approach 1 runs 2D
the temperature of the seawater is 32 °C and that of the air is 45 °C. numerical simulations of simplified geometries of Flat Panels and
The velocity of the ship is defined as 19 knots. The LNG tanks are Corners, while Approach 2 considers the full 3D geometries of Flat
considered to be filled up to 98% of capacity. The lower part is oc- Panels and Corners.
cupied by LNG at −162 °C, while the upper area is filled with NG. The second method applies the general framework of Reduced
Order Models (ROMs) to derive Approaches 3 and 4. Specifically, the
2.4. BOR definition ROM technology applied to this system searches for an equivalent
thermal resistance while only considering a single material along
BOR is defined as the percentage of the evaporated LNG mass the insulation layers. This material will be defined by equivalent
per day with respect to the initial loaded LNG mass [3], so the heat thermal conductivities to yield the same heat transfer simulations
flow (Q) received by LNG is translated into BOR by Eq. (2): as the above detailed numerical solutions. To achieve this, the
running of comprehensive numerical simulations of individual com-
Q ⋅ 24 ⋅ 3600 ⋅ 10−3 ponents like Flat Panels and 90° and 135° Corners is proposed. These
BOR = ⋅ 100 (2)
V ⋅ ρ ⋅ H vap comprehensive numerical simulations are the same as those cal-
culated in Approach 2.
where ρ is the density of LNG density and Hvap is the enthalpy of From these highly detailed results, an effective or integration pa-
vaporization of LNG (425 kg/m3 and 511 kJ/kg) at −162 °C and 1 bar. rameter is extracted. For the insulation walls of Mark III tanks, the
If BOR is calculated for each tank, Q and V are the total heat flow integration parameter is the equivalent to the thermal conductiv-
received by the LNG stored inside of each tank and the total volume ity of the materials that comprise the walls. An effective thermal
of this tank. However, BOR can be calculated for the full ship too, conductivity will be defined for Flat Panels and Corners to con-
so Q and V are the heat flow received by the total LNG transported dense out all heat transfer resistances at the different layers that
by the ship and the full ship capacity. compose the different bodies of the insulation barriers. Once more,

Fig. 3. Approaches for BOR calculation.


M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296 283

Table 4
Temperatures reported by Song et al. [25].

Configuration TLNG (°C) TIH (°C) Insulation barrier

A −162.00 20.02 FP 2, Edge 1–2


B −162.00 21.91 FP 3, Edge 1–3
C −162.00 29.19 FP 4, FP 5, Edge 1–4, Edge 1–5
D −162.00 31.51 FP 6, Edge 1–6
E −158.00 30.96 FP 7, Edge 1–7, Edge 3–8
F −158.00 32.44 FP 8, Edge 1–8
G −162.00 5.00 FP 1 CDAMS, Edge CDAMS
CR1 −162.00 27.44 Edge 3–3
CR2 −162.00 32.62 Edge 3–4
CR3 −162.00 38.09 Edge 3–6

the difference between Approaches 3 and 4 is that Approach 3 runs


2D numerical simulations of fluid flow and heat transfer in repre-
sentative 2D slices of ship arrangement, while Approach 4 considers
full 3D sections of the tanks and the ballast compartments.

Fig. 4. Temperature configurations defined by Song et al. [25].


3.1. Approach 1: 2D numerical simulations of FP and corners in
specific configurations
3.2. Approach 2: 3D numerical simulations of FP and corners in
The heat transfer from the environment to the LNG stored in specific configurations
MARK III tanks can be modeled as the heat flux through a planar
wall under different temperatures on either side. The different layers Approach 2 follows the same method as above, in other words,
that compose the tank walls act as like resistors in series for the the calculation of heat fluxes through Flat Panels and Corners in the
heat transfer, and therefore, applying Fourier’s law, the heat flux 10 specific configurations determined by Song et al. [25]. However,
transferred through the wall is simply the temperature difference Approach 2 increases the detailed level of the obtained results
divided by the total thermal resistance [37]: because the heat fluxes are achieved through 3D numerical simu-
lations of the complete geometries of Flat Panels and Corners,
TEnv − TLNG including the steel corrugations in contact with LNG, instead of the
q′′ = (3)
RT′′ characteristic 2D slices as in the previous section. The selected grids
are composed of 2.2 millions cells for Flat Panels and 700,000 cells
The total thermal resistance is the sum of the resistance of each for Corners. The obtained results are summarized in Section 4.1.
layer that composes the insulation barrier, and this resistance is a
function of the thermal conductivity of the material that com- 3.3. Definition of Reduced Order Models for heat transfer in Mark III
poses the layer and the thickness and the conductive shape factor tanks
of said layer. The dependence of the thermal conductivity of R-PUF
on temperature makes a direct calculation of the thermal resis- A realistic prediction of BOR can be obtained through detailed
tance by analytical formula like, for example, the planar wall numerical simulations of the fluid flow and heat transfer consid-
resistance [37], complicated. To solve this problem, Approach 1 is ering the external hulls, the ballast compartments, the inner hull,
divided into three steps. First, the heat fluxes through Flat Panels the metallic sheets that form the ballast compartments and the in-
and Corners are numerically calculated considering the properties sulation walls. The heat transfer from the environment to the inner
of materials defined in Section 2.2 and the temperature for LNG and hull will be simulated without any significant simplification, de-
for the inner hull obtained from the literature [25]. Trihedrons are fining a numerical grid to solve the continuity and the momentum
not considered due to their small heat transfer area by compari- and energy transport equations in these zones. The Reduced Order
son with Flat Panels or Corners. The second stage is the calculation Model is defined for the insulation walls, which are discretized
of the heat transfer areas of Flat Panels and Edges defined in Fig. 1b through a numerical grid of 10 cells in the normal direction, as shown
and c. The third step calculates the heat flows transferred through in Fig. 6.
the different walls by multiplying the previously obtained heat fluxes Such coarse mesh is not able to capture all geometrical fea-
and heat transfer areas, yielding the total heat flow which is finally tures of the Flat Panels and Corners. Thus, the Reduced Order Model
translated into BOR using Eq. (2). proposes the definition of an equivalent material for each insula-
The temperatures of the inner hull and the membrane re- tion body (Flat Panels and 90° and 135° Corners), characterized by
ported by [25] are split in Table 4 into 10 specific configurations the density and the heat capacity of the R-PUF as this accounts for
described in Fig. 4. LNG is considered to be at −162 °C while natural the highest percentage of the total volume of the barriers. The
gas presented at the top of the tank is at −158 °C. The tempera- thermal conductivity of this equivalent material is determined to
tures in Table 4 are then imposed as boundary conditions for the yield the same thermal resistance as is obtained for Flat Panels and
numerical model. The simplicity of this approach is applied to the Corners in Approach 2, following Eq. (4):
definition of the numerical domain too. Only the 2D representa-
tive sections of Flat Panels and 90° and 135° Corners shown in Fig. 5 q′′App.2
λEq = λR−PUF = λR−PUF ⋅ C Eq (4)
are investigated. The numerical calculations of heat fluxes are per- ′′
q App.2,Mod
formed using the Finite Element Method by the ABAQUS 6.12
software [38]. The numerical grids are composed of about 200,000 where λR-PUF is the thermal conductivity of R-PUF displayed in Table 2,
cells. The obtained values are presented and analyzed in Section 4.1. q″App.2 is the heat flux obtained in Approach 2 and q″App.2,Mod is the
284 M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296

(a) (b)

(c)

Fig. 5. 2D sections analyzed in Approach 1: (a) Flat Panel; (b) 90° Corner; (c) 135° Corner.

heat flux obtained in a “modified” Approach 2 where all layers but longitudinal slice of Tank 1. This conjugated fluid flow-heat trans-
the inner hull are only composed of R-PUF. The configuration with fer analysis is governed by the continuity, momentum and energy
the largest difference of temperature from the inner hull to LNG is transport equations [39]. These equations are solved by the ANSYS
selected, in other words, configuration D for Flat Panels and 90° FLUENT software, Release 15.0, using the 2D double precision solver
Corners and Configuration E for 135° Corners. As a result, the equiv- [40]. The fluid flow is assumed to be non-isothermal incompress-
alent parameter CEq integrates the exhaustive information from the ible turbulent flow, following the Navier–Stokes equations.
3D numerical models into a single parameter to characterize the Turbulence is modeled using the k–ε Realizable model [41] while
thermal resistance from inner hull to LNG. the turbulence generation by walls is calculated through En-
hanced Wall Functions [40]. Pressure and velocity fields are coupled
3.4. Approach 3: ROM and 2D numerical simulations of LNG tanks by the SIMPLE algorithm [42]. Second order spatial discretization
schemes are applied to the momentum, energy, turbulent kinetic
Approach 3 is based on 2D numerical simulations of fluid flow energy and turbulent dissipation rate, while pressure is discretized
and heat transfer from the environment to the LNG stored inside using the Body Force Weighted scheme [40].
the Mark III tanks. Four different domains are built as shown in Fig. 7: In all domains, the shell conduction from the inner hull to the
two symmetrical transversal slices of the Tanks 1 and 2, one lon- external hull through the different solid walls that form the ballast
gitudinal slice of the cofferdam between Tanks 1 and 2, and one compartments is considered by meshing of these slim walls with
M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296 285

The convective heat transfer coefficient is estimated from correla-


tion (6):

Nu = C NC ⋅ ( Ra ⋅ cos ( γ ) )
n
(6)

Correlation (6) assumes that convection is the result of buoy-


ancy forces alone and ignores the influence of sloshing LNG motion
inside the tanks. The CNC constant varies from 0.1 to 0.54; the ex-
ponent n varies from 1/4 to 1/3 depending on the orientation of the
wall and γ is the wall angle compared to the horizontal [39]. The
convective heat transfer varies in Tank 2 from 68.4 W/m2·°C for FP
8 to 1067.8 W/m2·°C for FP 3.

3.5. Approach 4: ROM and 3D numerical simulations of LNG tanks

Once the numerical model has been defined and applied in 2D


domains for Approach 3, it can be easily extended to 3D domains
for the development of the Approach 4. This section only de-
scribes significant differences with respect to Approach 3. First, the
ship is divided into three different domains, as shown in Fig. 8.
Tanks 2 and 3 are directly next to the cofferdams, so a single sim-
ulation is enough to get the heat flow transferred from the
environment to the LNG. Meanwhile, Tanks 1 and 4 sit against the
cofferdams or lie against the prow or the stern, respectively. Con-
sidering the symmetrical behavior of the heat transfer in Tank 1
(see the analysis of Fig. 14 in Section 4.2), just a single mesh of a
quarter of Tank 1 is built. Since Tank 4 is larger than Tank 1, the
symmetrical behavior of the heat transfer can be ensured too for
Fig. 6. Mesh of the Flat Panels and Corners in the Reduced Order Model (270 mm
Tank 4, so just a single mesh is built too for Tank 4. The thermal
thickness). boundary condition imposed at the vertical limiting wall will define
which the case is: if the convective heat transfer is applied to sea-
water or the air, the simulation obtains the heat transfer from
two cells along their thickness. The detailed geometries of the in- Tank 1 or 4 to the prow or the stern; if a constant temperature of
sulation barriers have been replaced by equivalent thermal 5 °C is set, the simulation represents the heat transfer from the
conductivities for Flat Panels and 90° and 135° Corners. The shell cofferdams.
conduction of the inner and outer hulls is directly modeled using The building of the computational grids is mainly based on the
18 and 50 mm thick steel meshed layers, respectively. The shell con- sweep of the 2D built for Approach 3, including the meshing of the
duction of the stainless steel membrane is not simulated horizontally; metal sheets that form the ballast compartments. Mapped con-
just a tiny thermal resistance is considered vertically, defined by formed meshes are selected for the most of the domain; zones as
1.2 mm thickness of stainless steel. The cell height attached to all corner compartments are meshed with pave schemes and swept
walls is 7 mm to allow an accurate capture of the thermal and viscous along the normal direction to reduce the high cell skewness. The
boundary layers. The structured mesh type is selected for most of selected meshes comprise 10.2 million cells for Tank 1 to 13.65
the regions except in zones with sharp angles. The selected meshes million cells for Tank 2. The external and the inner hulls are also
vary from 1.7 millions of quadrilateral cells for Tank 2 to 4.4 mil- included, together with the Flat Panels and Corners. Finally, the PISO
lions of cells for the cofferdam between Tanks 1 and 2. algorithm [43,44] with skewness correction has been selected for
The thermal boundary condition at the cofferdam walls imposes the pressure–velocity coupling.
a temperature of 5 °C, while convective boundary conditions are
applied at the external hull and at the membrane walls. For the ex- 4. Results and discussion
ternal hull walls, the air or the seawater temperature defined by
the IMO sailing conditions [32] is imposed and the convective heat 4.1. Results of Approaches 1 and 2
transfer coefficient (h) is obtained from correlation (5), which cor-
responds to the average Nusselt number for turbulent parallel flow A grid independence study is performed first to ensure the quality
over a flat plate [39]: of the results. For Approach 1, the number of cells from LNG to inner
hull is doubled without any significant difference in the heat fluxes
4 1
Nu = 0.037 ⋅ Re 5 ⋅ Pr 3 (5) for Configurations D and E. For Approach 2, several mesh resolu-
tions and domain extents are considered, resulting in the selected
where Nu is the Nusselt number, Re is the Reynolds number and domains displayed in the temperature contours shown in Fig. 9. The
Pr is the Prandtl number. Considering the length of the ship as the uniform thickness of all insulation layers for Flat Panels are high-
typical length scale (~200 m) and the air and water properties ob- lighted by constant temperature profiles for height. However, the
tained from Lemmon et al. [36] and Incropera and de Witt [39], the different hardwood pieces and the angle geometry yield a non-
convective heat transfer coefficient is 12.40 W/m2·°C for the walls uniform evolution of temperature from the top to the bottom of
in contact with air and 6529 W/m2·°C for the walls in contact with Corners.
water. These values are of the same order of magnitude as the typical Table 5 summarizes the heat transfer areas for the different Flat
values of the convective heat transfer coefficient collected in Panels and Corners in Tanks 1 to 4, and Table 6 displays the heat
Incropera and de Witt [39]. For the membrane walls in contact with fluxes obtained through the computational simulations per-
LNG, it is first assumed that LNG is at −162 °C for the entire tank. formed in Approaches 1 and 2. The heat flow transferred through
286 M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296

Fig. 7. 2D numerical domains simulated in Approach 3: (a) transversal section of Tank 1; (b) transversal section of Tank 2; (c) longitudinal section of cofferdam between
Tanks 1 and 2; (d) longitudinal section of Tank 1.

Fig. 8. 3D numerical domains simulated in Approach 4.


M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296 287

(a)

(b)

Fig. 9. Temperature contours (°C) of Flat Panels and 90° and 135° Corners in configuration D: (a) Approach 1; (b) Approach 2.
288 M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296

Table 5
Heat transfer areas for Approaches 1 and 2 (inner hull side).

Zone Body Configuration Tank 1 (m2) Tanks 2, 3 (m2) Tank 4 (m2)

Edge 1–8 C90 F 19.599 0.000 19.598


Edge 3–8 C135 E 34.939 48.680 42.690
Edge 1–7 C90 E 2.716 0.000 2.684
Edge 1–6 C90 D 13.566 0.000 18.818
Edge 3–6 C135 CR3 34.939 48.680 42.690
Edge 1–5 C90 C 4.809 0.000 9.910
Edge 1–4 C90 C 0.000 0.000 4.152
Edge 3–4 C135 CR2 34.939 48.680 42.690
Edge 1–3 C90 B 36.066 0.000 10.193
Edge 3–3 C135 CR1 34.939 48.680 42.690
Edge 1–2 C90 A 19.706 0.000 27.595
Edge CDAMS C90 G 96.462 306.978 107.014
Total Edges 332.679 501.699 370.724
FP 8 FP F 699.880 981.203 858.562
FP 7 FP E 94.015 528.690 462.591
FP 6 FP D 469.657 528.690 462.591
FP 5 FP C 160.252 1000.352 876.293
FP 4 FP C 0.000 406.864 356.395
FP 3 FP C 244.391 0.000 0.000
FP 3 FP B 1041.121 479.013 419.080
FP 2 FP A 703.807 1390.171 1216.430
FP 1 CDAMS FP G1 763.110 1892.325 946.163
FP 1 Prow/Stern FP C 763.110 0.000 946.163
Total FP 4939.342 7207.309 6544.266
Total Tank 5272.021 7709.009 6914.990

each insulation barrier can be obtained by multiplying the heat fluxes 4.2. Results of Approaches 3 and 4
by the corresponding heat transfer areas. Once the heat flows trans-
ferred from Flat Panels and Corners are obtained, the sum of all heat First, the equivalent thermal conductivities obtained from the
flows is applied to calculate the BOR following Eq. (2). It can be seen procedure reported in Section 3.4 are reported here. Table 7 com-
that 3D heat fluxes for Corners are 13%–16% lower than 2D results. pares the obtained heat fluxes for the selected configuration in the
The 2D model considers that the hardwood is continuous, while the original and modified Approaches 2 and the equivalent constant to
3D geometry correctly describes the distribution across the differ- be applied in Eq. (4). The last row shows the obtained heat flux when
ent blocks of hardwood and R-PUF. Thus, a larger R-PUF content the equivalent thermal conductivity is applied. It can be con-
increases the thermal resistance of the 3D corner and yields lower cluded that this order reduction is suitable to capture the global
heat fluxes. The difference for Flat Panels is about 2% and it is caused thermal behavior of the insulation barrier because the differences
by the simplification of the stainless steel corrugation. The 2D model when using this reduced model with respect to the detailed model
only considers a 1.2 mm thick stainless steel layer, but the 3D model are less than 0.17%.
takes into account the real shape of the corrugations, the wooden Once the equivalent thermal conductivities of Flat Panels and
pieces that support them and the air that fill the gaps between the 90° and 135° Corners are obtained, the numerical simulations can
membrane and the supporting pieces. be performed. The grid independence study has been performed
These differences are not directly translated in the BOR calcu- by doubling the number of cells inside of the insulation barriers,
lation. Fig. 10 compares the BOR obtained by the different approaches yielding evident differences in heat flows for all simulated domains
and for the different tanks. It can be seen that the contribution of in Approaches 3 and 4.
Corners to the total BOR varies from 4.9% to 6.5% because the heat Qualitative results are given in Figs. 11 and 12 through temper-
transfer area of Corners is about 6% of the total heat transfer area. ature and velocity contours in the different analyzed domains. The
In addition, it can be observed that larger tanks yield a lower BOR, maximum fluid velocity is located at the bottom compartments
as the surface area density (the surface to volume ratio) is reduced because they show the configuration that encourages the convec-
[45]. tive heat transfer, having the cold wall high up and the bottom wall
low down. The ballast compartments at the corners show signifi-
cant irregular velocity fields due to the instabilities caused by the
plumes generated from these cold walls [37]. The temperature con-
Table 6
tours reveal that the main heat transfer resistance is obviously
Heat fluxes obtained for Approaches 1 and 2.
generated by the insulation barriers because the drop in temper-
Configuration q″App.1 (W/m2) q″App.2 (W/m2) ature through the air compartments is noticeably lower than it is
FP C90 C135 FP C90 C135 through the Flat Panels and Corners, as it would be expected. The
A 13.467 12.828 – 13.191 11.322 – thermal bridges of the steel sheets comprising the ballast compart-
B 13.643 12.994 – 13.364 11.470 – ments can be identified in Fig. 13 by the significant rise in
C 14.333 13.646 – 14.037 12.048 – temperature in these figures.
D 14.557 13.858 – 14.255 12.235 –
The symmetry of the thermal field in Tank 1 along its longitu-
E 14.255 13.558 16.053 13.962 11.982 13.757
F 14.399 13.694 – 14.104 12.102 –
dinal section is demonstrated by Fig. 14. The profiles of the
G 12.122 11.555 – 11.872 10.195 – temperature difference for the inner and outer hulls are drawn start-
CR1 – – 15.971 – – 13.672 ing from the symmetry line. No significant differences are found for
CR2 – – 16.530 – – 14.152 distances lower than 5 m from the midpoint, so the different thermal
CR3 – – 17.134 – – 14.670
boundary conditions imposed at cofferdam walls and at stern or prow
M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296 289

(a) (b)

(c) (d)

Fig. 10. BOR obtained for Approaches 14: (a) Tank 1; (b) Tanks 2 and 3; (c) Tank 4; (d) Ship.

walls only influence in the region nearest to these zones. Thus, Tanks hull temperatures for configurations A–E and CR1–CR3, the results
1 and 4 can be divided by symmetry planes for Approach 4. shown for both Approaches 3 and 4 are the area-weighted average
Table 8 compares the numerical temperatures obtained from Ap- of the wall that separates the inner hull from the ballast compart-
proaches 3 and 4 with results reported by Song et al. [25]. Only a ments or the metal sheets. The temperatures in the configurations
rough comparison should be made because there are certain dif- show larger differences, especially for Approach 3, and for Config-
ferences between the simulated and the published cases. For uration F, the top wall.
example, the thermal conductivity of R-PUF is 0.04 W/m·°C in Song Table 9 compares the heat flow and BOR obtained from other
et al. [25]; the LNG temperature varies in Reference [25] from −162 °C available published data with results from Approach 4. The signif-
to −158 °C in the vapor phase, while current numerical simula- icant differences between the geometries of MOSS and Mark III tanks
tions are based on a constant temperature of −162 °C and the and the variations in thicknesses and material properties among
geometries of the compartments are different, as shown in Fig. 15. these references avoid a direct validation based on the reported
Reference [25] considers just five ballast compartments, while the values. However, it should noted that Approach 4 is close to the
current numerical model divides the fluid space into 12 compart- values reported from Song et al. [25], considering that only a quarter
ments. For this reason, Table 8 shows the area-weighted average of of the tank is simulated in Reference [25].
temperature for Approach 3 and the volume-weighted tempera- The calculation of BOR for Approach 3 is not straightforward
ture for Approach 4 contemplating the different compartments because Edges 1–3, 1–4, 1–5, 1–6 and 1–7 are not simulated while
identified in Fig. 15. Ballast compartments 2 and 4 show small dif- FP 2, FP 8 and Edges 1–2 and 1–8 in Tank 1 are simulated twice.
ferences between the published and the obtained results due to a The proposed solution is to consider the missing data as the average
good concordance between the different geometries. For the inner of the nearest simulated corners, while the repeated data are reduced
by taking the average value between the two simulated domains.
Moreover, the 2D domains do not represent the exact geometry, so
Table 7
the heat transfer area must be determined for the different edges
Heat fluxes obtained for the equivalent thermal conductivity.
based on the length of the tanks. Computational results are cast into
270 mm 400 mm
the different heat fluxes obtained at the membrane level, while the
FP C90 C135 FP C90 C135 heat fluxes obtained for Approaches 1 and 2 were calculated at the
q″App.2 (W/m )2
14.255 12.235 13.757 9.505 8.208 9.589 inner hull side. Thus, only overall results are shown in terms of the
q″App.2,Mod (W/m2) 13.578 9.191 10.725 9.155 5.904 7.875 BOR calculated for the heat transfer through Flat Panels and Corners
CEq 1.050 1.331 1.283 1.038 1.390 1.218 in Fig. 10 together with the overall BOR and the overall heat trans-
q″App.2, Eq R-PUF (W/m2) 14.280 12.265 13.771 9.504 8.175 9.586
fer coefficient in Table 10.
290 M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296

(a) (b)

Fig. 11. Contours obtained for Approach 3: (a) air temperature (°C); (b) air velocity (m/s).

The heat flow through Flat Panels is again the most important, The thermal resistances for the full ship obtained for Approach
but the relative importance of edges increases by 12% for Ap- 4 are 4.7197·10−5 °C/W for RT, 4.8392·10−4 °C/W for ROut→IH and
proach 3 and 16% for Approach 4 due to the rise in the equivalent 5.3112·10−4 °C/W for RIH→LNG. It can be concluded that the insula-
thermal conductivity. However, the overall heat transfer rate does tion barriers represents the 91.11% of the overall thermal resistance.
not change dramatically: Table 10 reveals that the largest differ- Finally, Approach 4 yields the estimation of the heat flow re-
ence is about 6.6% for Approaches 3 and 4 in Tank 1. In summary, quired to have cofferdams at 5 °C. The cofferdam between Tanks 1
the differences achieved in the overall BOR for the full ship lie below and 2 needs 206 kW; 342 kW are required for the cofferdam between
3% for all approaches. Tanks 2 and 3; while 346 kW must be provided for the cofferdam
The overall heat transfer coefficient is defined by Eq. (7): between Tanks 3 and 4, yielding an overall cooling power of 894 kW
for the full ship. This overall cooling power can be compared to the
Q heat flow reported in Reference [24], considering that reference data
U= (7)
ALNG ⋅ ( TEnv − TLNG ) are just for a quarter of a smaller tank, at colder sailing conditions
with different materials and thickness.
where Q is the overall heat flow received by each tank or by the
full ship, ALNG is the heat transfer area of the membrane in contact
with LNG, TEnv is the area-weighted average temperature of the en- 4.3. Evaluation of the 4 approaches
vironment, considering the respective surfaces exposed to seawater
and air and TLNG is the LNG temperature. The obtained value for the Approach 1 is based on a published procedure [25], but devoted
full ship in Approach 4 is 0.0674 W/m2·°C, which can be com- to a lower ship capacity of 137,000 m3 with different thermal in-
pared with the value of 0.07 W/m2·°C obtained from Eq. (8): sulation materials. Approach 1 can be calculated very quickly,
requiring less than 2 h in a standard computer server. Approach 2
1 provides more detailed information about heat transfer through Flat
U[14] =
1 t 1 (8) Panels and Corners than the previous approach, but it requires more
+ +
hLNG λ hEnv time to run (about 1 day) and it does not solve the previous dis-
advantage regarding the definition of the temperatures on the inner
Equation (8) was applied in Reference [14] for a material with hull side.
different thickness and constant thermal conductivity, so a differ- Approach 3 does not assume the simplifications regarding the
ent value of U is reported in Reference [14]. For the current case, inner hull temperatures and it gives an initial estimation of the con-
the thermal conductivity of R-PUF varies with temperature, so the vective and conductive heat transfer resistances through external
value at mid temperature (−80 °C) is applied in Eq. (8). Then, Eq. (8) and inner hulls and ballast compartments. However, the determi-
represents a good starting approach for BOR calculation, since con- nation of the heat transfer areas and the heat fluxes reduces the
duction through R-PUF layer is the most important thermal accuracy of this approach because it does not represent the 3D be-
resistance. However, if a detailed thermal analysis is required, the havior of the tanks. BOR is not directly calculated from numerical
proposed approaches of the current investigation yield a more com- simulations because it is still necessary to obtain average heat fluxes
prehensive description of the thermal fields inside the LNG ship and heat transfer areas from geometry files.
tanks. Approach 4 proposes the numerical simulation of heat transfer
These overall heat transfer coefficients can then be cast into global and fluid flow in 3D models of the different tanks, considering the
thermal resistances under the terms of Eq. (9): orientation of each tank to cofferdams or to the prow and the stern
of the ship. The meshing of the insulation layers cannot include all
TEnv − TLNG TEnv − TIH TIH − TLNG
Q= = = (9) geometric features such as the plywood, the triplex layers and the
RT ROut →IH RIH →LNG first and second barriers, so a Reduced Order Model is defined to
M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296 291

(a)

(b)

(c)

(d)

Fig. 12. Contours obtained for Approach 4: (a) air temperature (°C), orientation to cofferdam; (b) air velocity (m/s), orientation to cofferdam; (c) air temperature (°C), ori-
entation to prow/stern; (d) air velocity (m/s), orientation to prow/stern.
292 M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296

(a)

(b)

Fig. 13. Temperature contours (°C) at the inner wall obtained for Approach 4: (a) orientation to cofferdam; (b) orientation to prow/stern.

account for the different thermal conductivity of each insulation creasing the thickness of the secondary barrier. The temperatures
barrier. This approach does not assume any simplifications about defined for the selected configuration in Table 5 are applied at the
heat transfer areas. This increased model complexity also yields a membrane and inner hull sides, yielding the heat fluxes reported
rise in the additional data provided by the numerical simulations, in Table 7. The constants for equivalent thermal conductivities to
for example, the required power to keep cofferdams at 5 °C togeth- be applied on Flat Panels and Corners are obtained as the quotient
er with a full description of the heat transfer and fluid flow in the of the previously obtained heat fluxes. Next, the mesh of Flat Panels
ship. However, Approach 4 requires the longest times for pre- and Corners is increased in the various domains in Approach 4 and
processing and solving the different stages of numerical calculations. the new equivalent R-PUF materials are applied. The numerical
For example, 1 week is needed to run the calculations for the 5 dif- models described in Sections 3.4 and 3.5 are applied again, yield-
ferent numerical domains included in this task. ing the BOR and the overall heat transfer coefficient shown in Fig. 16
and Table 10.
4.4. Proposal and evaluation of a design modification: increasing the The conclusion is clear: an increase of the thickness of the in-
thickness of the secondary barrier sulation barriers means a reduction in the capacity of the ship by
2.2%, but the heat flow received by LNG is significant reduced (by
Approach 4 is applied to calculate the BOR when the thickness up to 67%) from the initial heat flow.
of the secondary R-PUF layer is increased from 160.4 mm to
290.4 mm. The objective of this modification is the reduction of the 5. Conclusions and future work
heat transfer to the LNG with a minimum decrease in the capacity
of the ship. These thicker insulation layers are directly supported Four approaches have been developed to calculate the Boil-Off
by the inner hull, so the external hull, the ballast compartments and Rate of Liquefied Natural Gas when shipped in Mark III tanks,
the inner hull remain unchanged from the previous section. from the simplest procedure using 2D simulations in 10 specific
The procedure is as follows. The numerical grids built for Flat configurations to 3D simulations of the fluid flow and heat trans-
Panels and 90° and 135° Corners in Approach 2 are modified by in- fer around the LNG tanks, applying Reducing Order Models to
M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296 293

Fig. 14. Temperature difference for the cofferdam and the stern or prow sides.

Table 8
Comparison of temperatures of ballast compartments and configurations of inner
hull with results for Tank 2 obtained by Song et al. [25].

Ballast compartment/ T[25] (°C) TApp.3 (°C) TApp.4 (°C)


Configuration

t1 38.67 41.20 36.35


t2 39.12 38.69 37.65
t3 36.67 32.93 30.55
t4 29.00 29.87 28.62
t5 25.84 29.84 27.55
A 20.02 27.62 23.09
B 21.91 26.92 23.15
C 29.19 30.48 25.67
D 31.51 34.92 30.24
E 30.96 37.74 36.40
F 32.44 41.03 25.90
G 5.00 2.76 0.12
CR1 27.44 27.86 26.70
CR2 32.62 28.20 26.32
CR3 38.09 34.90 35.22

Fig. 15. Ballast compartments for temperature comparison with Song et al. [25].

Table 9
Comparison of current study with previous investigations reported in scientific literature.

Reference Type of Tank volume (m3) Insulation λ (W/m·°C) Sailing conditions Results Comments
tank thickness
TAir Env TSea Env Q (W) BOR (%
(m)
(°C) (°C) Ev/day)

[14] Mark III 40,447 0.53 0.0605 5 0 157,458 0.1552


[22] MOSS 41,051 0.22 Unknown −18 0 99,820 0.1188 Tank volume estimated from overall dimensions
25 18 123,500 0.1470
45 32 136,000 0.1619
[23] MOSS 32,960 0.22 0.030 −18 0 101,500 0.1171
45 32 132,400 0.1527
[24] Mark III 138,000 0.24 0.0536 −18 0 89,700 – Heat flow received from Cofferdam; results for 1/4
(full ship) of tank
[25] Mark III 7,536 0.25 0.04 −18 0 16,317 0.0863 Volume estimated from Q and BOR; results for 1/4
45 32 18,972 0.1003 of tank
28 29 18,469 0.0977
Current Mark III 27,009 0.27 Varying 45 32 71,443 0.1074 Heat flow and BOR from App. 4
study 49,391 with T 104,197 0.0856
40,713 97,224 0.0969
294 M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296

Table 10
Comparison of BOR and the overall heat transfer coefficient obtained using the different Approaches and the different thicknesses.

270 mm 400 mm

App. 1 App. 2 App. 3 App. 4 App. 4

Tank 1 BOR (% mEv/mLd·day) 0.1091 0.1061 0.1145 0.1074 0.0765


U (W/m2·°C) 0.0670 0.0652 0.0703 0.0660 0.0458
Tanks 2, 3 BOR (% mEv/mLd·day) 0.0860 0.0836 0.0872 0.0856 0.0583
U (W/m2·°C) 0.0676 0.0657 0.0686 0.0673 0.0448
Tank 4 BOR (% mEv/mLd·day) 0.0955 0.0929 0.0990 0.0969 0.0660
U (W/m2·°C) 0.0676 0.0658 0.0701 0.0687 0.0457
Ship BOR (% mEv/mLd·day) 0.0921 0.0895 0.0945 0.0919 0.0631
U (W/m2·°C) 0.0675 0.0656 0.0693 0.0674 0.0453

Fig. 16. BOR obtained for 400 mm thick insulation.

define equivalent thermal properties that can describe the heat Acknowledgements
transfer resistance through the insulation layers. The obtained
results are in concordance with the available published data, so This dissemination work has been funded in part by the FEDER
these approaches can be used to estimate BOR or to provide a Operative Program for Aragon (2007–2013).
detailed description of the thermal fields inside the LNG ship
carriers. Nomenclature
There is a three-fold plan for future work. First, Approach 1
can be easily incorporated into a software application that can CNC Natural convection correlation constant
anticipate BOR in ship carriers using analytical calculations of CEq Equivalent thermal conductivity calculation con-
heat transfer through the different insulation layers if the shape stant, dimensionless
factor for each layer has already been obtained. Second, the mea- CP Heat capacity (J/kg·°C)
surement of experimental BOR data during real ship transportation g Gravitational acceleration (9.81 m/s2)
would provide very useful information to validate the current h Convective heat transfer coefficient (W/m2·°C)
numerical approaches. These measurements must include the ship HVap LNG enthalpy of vaporization (511 kJ/kg)
geometry, the evolution over time of LNG temperature, pressure L Characteristic length (m)
and composition, the air and seawater temperatures, the ship m Mass (kg)
speed and the thermal properties of the insulation barriers to Nu Nusselt number, dimensionless, Nu = h·L/λ
eliminate the effects of ageing in polyurethane foams. Lastly, de- Pr Prandtl number, dimensionless, Pr = ν/α
tailed numerical simulations of significant phenomena at the LNG Q Heat flow (W)
site including, for example, boiling, stratification, sloshing phenom- q″ Heat flux (W/m2)
ena or mixing processes if the ship carrier includes a reliquefaction R″T Thermal resistance (m2·°C/W)
plant would yield additional descriptions of the thermal behavior Ra Rayleigh number, dimensionless, Ra = g·β·ΔT·L3/ν·α
of LNG when shipped. Re Reynolds number, dimensionless, Re = W·L/ν
M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296 295

S Conductive shape factor, dimensionless [10] S.W. Choi, J.U. Roh, M.S. Kim, W.I. Lee, Analysis of two main LNG CCS (Cargo
Containment System) insulation boxes for leakage safety using experimentally
t Thickness (m)
defined thermal properties, Appl. Ocean Res. 37 (2012) 72–89.
T Temperature (°C) [11] Q.S. Chen, J. Wegrzyn, V. Prasad, Analysis of temperature and pressure changes
U Overall heat transfer coefficient (W/m2·°C) in liquefied natural gas (LNG) cryogenic tanks, Cryogenics 44 (2004) 701–709.
V Volume (m3) [12] M.M. Faruqe Hasan, A.M. Zheng, I.A. Karimi, Minimizing Boil-Off Losses in
liquefied natural gas transport, Ind. Eng. Chem. Res. 48 (2009) 9571–9580.
W Characteristic velocity (m/s) [13] M.S. Zakaria, K. Osman, A.A. Yusof, M.H.M. Hanafi, M.N.A. Saadun, M.Z.A. Manaf,
Parametric analysis of Boil-Off Gas rate inside liquefied natural gas storage tank,
Greek symbols J. Mechan. Eng. Sci. 6 (2014) 845–853.
[14] M.S. Zakaria, K. Osman, M.N. Musa, Boil-Off Gas formation inside large scale
α Thermal diffusivity (m2/s) liquefied natural gas (LNG) tank based on specific parameters, Appl. Mech.
β Thermal expansion coefficient (°C−1) Mater. 229–231 (2012) 690–694.
γ Wall angle with horizontal direction (rad) [15] M.S. Zakaria, K. Osman, M.N.A. Saadun, M.Z.A. Manf, M.H. MHanafi,
Computational simulation of Boil-Off Gas formation inside liquefied natural tank
λ Thermal conductivity (W/m °C) using evaporation model in ANSYS fluent, Appl. Mech. Mater. 393 (2013)
ν Kinematic viscosity (m2/s) 839–844.
ρ Density (kg/m3) [16] D. Boukeffa, M. Boumaza, M.X. Francois, S. Pellerin, Experimental and numerical
analysis of heat losses in a liquid nitrogen cryostat, Appl. Therm. Eng. 21 (2001)
Δ Difference
967–975.
[17] O. Khemis, M. Boumaza, M. Ait Ali, M.X. Francois, Experimental analysis of heat
Subscripts transfers in a cryogenic tank without lateral insulation, Appl. Therm. Eng. 23
(2003) 2017–2117.
0 Reference conditions
[18] O. Khemis, R. Bessih, M. Ait Ali, M.X. Francois, Measurement of heat transfers
App.1, App.2, Approach 1, 2, 3 or 4 in cryogenic tank with several configurations, Appl. Therm. Eng. 24 (2004)
App.3, App.4 2233–2241.
Cdam Cofferdam [19] T. Kanazawa, K. Kudo, A. Kuroda, M. Tsui, Experimental study on heat and fluid
flow in LNG tank heated from the bottom and the sidewalls, Heat Transf. Asian
Env Environment (seawater or air) Res. 33 (2004) 417–430.
Eq Equivalent [20] M. Belmedany, A. Belgacem, R. Rebiai, Analysis of natural convection in liquid
Ev Evaporated nitrogen under storage conditions, J. Appl. Sci. 8 (2008) 2544–2552.
[21] A. Hamon, One million cores, a breakthrough in CFD simulation, International
i Insulation layer Science Grid This Week (2013). <http://www.isgtw.org/feature/one-million
IH Inner hull -cores-breakthrough-cfd-simulation> (accessed 27.05.15).
Ld Loaded [22] Y.M. Kim, S.C. Ko, B.I. Chun, K.K. Kim, A study on the thermal design of the
cryogenic LNG carrier, Journal of the Korean Society of Marine Engineering 17
Prw/Stn Prow or stern (1993) 1–10.
[23] J.G. Kim, Y.M. Kim, C.S. Kim, A study on unsteady temperature distribution
Acronyms analysis of moss type LNG carrier by insulation system, J. Ocean Eng. Technol.
11 (1997) 159–168.
LNG Liquefied Natural Gas [24] J.H. Heo, Y.B. Lee, Heat flux calculation for thermal equilibrium of cofferdam
ROM Reduced Order Model in a LNG carrier, J. Soc. Nav. Arch. Korea 35 (1998) 98–106.
CCS Cargo Containment System [25] S.O. Song, J.H. Lee, H.P. Jun, B.Y. Sung, K.K. Kim, S.G. Kim, A study on the three
– dimensional steady state temperature distribution and bor calculation
R-PUF Reinforced Polyurethane Foam
program development for the membrane type LNG carrier, Journal of the Korean
FP Flat Panel Society of Marine Engineering 23 (2) (1999) 140–149.
C90 90° Corner [26] J.H. Lee, K.K. Kim, S.T. Ro, H.S. Chung, S.G. Kim, A study on the thermal analysis
C135 135° Corner of spray cooling for the membrane type LNGC during the cool-down period,
Trans. Korean Soc. Mech. Eng. B 27 (2003) 125–134.
BOR Boil-Off Rate, defined in Eq. (2) (% evaporated LNG [27] J.U. Heo, Y.J. Lee, J.R. Cho, M.K. Ha, J.N. Lee, Heat transfer analysis and BOG
mass/loaded LNG mass·day) estimation of membrane – type LNG cargo during laden voyage, Trans. Korean
NG Natural Gas Soc. Mech. Eng. A 27 (2003) 393–400.
[28] B. Shapiro, Creating reduced order modelling for electronic systems: an overview
FEM Finite Element Method and suggested use of existing model reduction and experimental system
identification tools, IEEE Trans. Compon. Packag. Technol. 26 (2003) 165–172.
[29] S. Somarathne, M. Seymour, M. Kolokotroni, C.F.D. Thermal, Simulation of a
typical office by efficient transient solution methods, Build. Environ. 40 (2005)
References 887–896.
[30] F. He, L. Ma, Thermal management of batteries employing active temperature
[1] International Gas Union, World LNG Report – 2014 Edition, International Gas control and reciprocating cooling flow, Int. J. Heat Mass Transf. 83 (2015)
Union, Office of the Secretary General, C/O Statoil ASA, Norway, 2015. 164–172.
<http://www.igu.org/sites/default/files/node-page-field_file/IGU%20-%20World [31] W.A. Silva, Discrete-time linear and nonlinear aerodynamic impulse responses
%20LNG%20Report%20-%202014%20Edition.pdf> (accessed 27.05.15). for efficient CFD analyses (Ph.D. dissertation), College William Mary,
[2] S. Cornot-Gandolphe, O. Appert, R. Dickel, M.F. Chabrelie, A. Rojey, The Williamsburg, VA, 1997.
Challenges of Further Cost Reductions for New Supply Options (Pipeline, LNG, [32] I.G.C. Code, International Code for the Construction and Equipment of Ships
GTL), in: 22nd World Gas Conference, Tokyo, Japan, 1–5 June, 2003. Carrying Liquefied Gases in Bulk, International Maritime Organization, 2003.
[3] D. Dobrota, B. Lalic, I. Komar, Problem of Boil-Off in LNG supply chain, Trans. [33] J. Roni, J. Chauvin, The General Electric – Technigaz Mark III Containment System,
Marit. Sci. 2 (2013) 91–100. GASTECH, Monaco, 1978.
[4] J. Romero Gómez, M. Romero Gómez, J. López Bernal, A. Baaliña Insua, Analysis [34] C.Y. Ho, T.K. Chu, Electrical resistivity and Thermal Conductivity of Nine Selected
and efficiency enhancement of a Boil-Off Gas reliquefaction system with AISI Stainless Steels, CINDAS Report 45, Thermophysical and Electronic
cascade cycle on board LNG carriers, Energy Convers. Manag. 94 (2015) 261– Properties Information Analysis Center, Lafayette, IN, 1977.
274. [35] C.J. Tseng, M. Yamaguchi, T. Ohmori, Thermal conductivity of polyurethane foam
[5] F. Deybach, T. Gavory, Very Large LNG Carriers: A Demonstration of Membrane from room temperature to 20 K, Cryogenics 37 (1997) 305–312.
Systems Adaptability, Gastech, Bangkok, 2008. <http://www.gtt.fr/wp-content/ [36] E.W. Lemmon, R.T. Jacobsen, S.G. Penoncello, D. Friend, Thermodynamic
uploads/2012/09/very-large-carrier-gastech-2008-pdf-0-6mo-mo-2012-04 properties of AIr and mixtures of nitrogen, argon, and oxygen from 60 to 2000 K
-11.pdf> (accessed 27.05.15). at pressures to 2000 MPa, J. Phys. Chem. Ref. Data 29 (2000) 331–385.
[6] D. Lee, S.H. Yoon, K.H. Kim, I. Choi, D.G. Lee, Composite anti–buckling structure [37] J.H. Lienhard IV, J.H. Lienhard V, A Heat Transfer Textbook, third ed., Philogiston
for the corrugations of liquefied hydrogen containers, Compos. Struct. 95 (2013) Press, Cambridge, MA, USA, 2008.
492–499. [38] SIMULIA, ABAQUS Analysis User’s Manual, Dassault Systèmes Simulia Corp.,
[7] Y.H. Yu, B.G. Kim, D.G. Lee, Cryogenic reliability of the sandwich insulation board Providence, RI, USA, 2012.
for LNG ship, Compos. Struct. 95 (2013) 547–556. [39] F.P. Incropera, D.P. de Witt, Fundamentals of Heat and Mass Transfer, fifth ed.,
[8] K.H. Kim, S.H. Yoon, D.G. Lee, Vibration isolation of LNG containment systems John Wiley & Sons, New York, USA, 2002.
due to sloshing with glass fiber composite, Compos. Struct. 94 (2012) 469–476. [40] ANSYS, ANSYS Fluent Theory Guide, Release 15.0, Canonsburg, PA, USA, 2014.
[9] H.B. Lee, B.J. Park, S.H. Rhee, J.H. Bae, K.W. Lee, W.J. Jeong, Liquefied natural [41] T.H. Shih, W.W. Liou, A. Shabbir, Z. Yang, J. Zhu, A new k–ε eddy–viscosity model
gas flow in the insulation wall of a cargo containment system and its for high Reynolds number turbulent flows – model development and validation,
evaporation, Appl. Therm. Eng. 31 (2011) 2605–2615. Computers & Fluids 24 (2005) 227–238.
296 M. Miana et al./Applied Thermal Engineering 93 (2016) 279–296

[42] S.V. Patankar, D.B. Spalding, A calculation procedure for heat, mass and [44] R.I. Issa, A.D. Gosman, A.P. Watkins, The computation of compressible and
momentum transfer in three-dimensional parabolic flows, Int. J. Heat Mass incompressible recirculating flows by a non-iterative implicit scheme, J. Comput.
Transfe. 15 (1972) 1787–1805. Phys. 62 (1986) 66–82.
[43] R.I. Issa, Solution of the implicity discretizated fluid flow equation by operator- [45] W.M. Rohsenow, J.P. Harnett, Y.I. Cho, Handbook of Heat Transfer, third ed.,
splitting, J. Comput. Phys. 62 (1986) 40–65. McGraw-Hill, New York, 1998.

Das könnte Ihnen auch gefallen