Evaluation of sloshing resistance performance for LNG carrier insulation system based on fluidstructure interaction analysis
 Author: Lee ChiSeung, Cho JinRae, Kim WhaSoo, Noh ByeongJae, Kim MyungHyun, Lee JaeMyung
 Organization: Lee ChiSeung; Cho JinRae; Kim WhaSoo; Noh ByeongJae; Kim MyungHyun; Lee JaeMyung
 Publish: International Journal of Naval Architecture and Ocean Engineering Volume 5, Issue1, p1~20, 31 March 2013

ABSTRACT
In the present paper, the sloshing resistance performance of a hugesize LNG carrier’s insulation system is evaluated by the fluidstructure interaction (FSI) analysis. To do this, the globallocal analysis which is based on the arbitrary LagrangianEulerian (ALE) method is adopted to accurately calculate the structural behavior induced by internal LNG sloshing of a KC1 type LNG carrier insulation system. During the global analysis, the sloshing flow and hydrodynamic pressure of internal LNG are analyzed by postulating the flexible insulation system as a rigid body. In addition, during the local analysis, the local hydroelastic response of the LNG carrier insulation system is computed by solving the local hydroelastic model where the entire and flexible insulation system is adopted and the numerical analysis results of the global analysis such as initial and boundary conditions are implemented into the local finite element model. The proposed novel analysis techniques can potentially be used to evaluate the structural integrity of LNG carrier insulation systems.

KEYWORD
Hydroelastic analysis , FSI analysis , KC1type LNG carrier insulation system , Sloshing , ALE method , Globallocal analysis.

INTRODUCTION
As the demand of natural resources such as natural gas and oil are tremendously increased, the huge size carrier such as liquefied natural gas (LNG) carrier is fabricated in nowadays. In accordance with this phenomenon, a lot of risk factors such as sloshing problem, crack propagation problem in midship section are emerged. Among these, the sloshing problem is considered as one of the most catastrophic problems since the structural failure of the LNG carrier leads to not only leakage of LNG but also tremendous loss of human and financial resources.
In order to overcome this problem, i.e., the insulation system which is consisted of lots of composites and austenitic stainless steels is adopted such as MARKIIItype, NO96type, KC1type insulation system. Among these, the KC1type insulation system which is fabricated by shipbuilding companies of Korea is designed to sustain the leakage of LNG as well as structural failure. Figs. 1 and 2 show the schematic of the membranetype LNG carrier and the KC1type insulation system. This insulation system encounters the severe sloshing loads during its oversea operation. Hence, it is essential to guarantee the structural safety of the insulation system during its design and fabrication.
For several decades, the hydroelastic analysis under sloshing has been analyzed by computational methods. In other words, the hydrodynamic pressure distributions as well as wave characteristics near the boundary condition are successfully evaluated based on finite element method (FEM) (Valtinsen, 1974; Nakayama and Washizu, 1980; Wu et al., 1998). However, in these researches, the structural response and behavior is not focused on, because the container is postulated as a rigid body and the interactive effect between interior fluid and exterior container is ignored. It is considered that the tremendous computational time and cost would be demanded when the structural deformations and interactive methods are adopted.
As an alternative method, the globallocal basis fluidstructure interaction (FSI) analysis has gained attention for addressing the sloshinginduced structural response problem (Mote, 1971; Mao and Sun, 1991; Cho and Lee, 2003; Cho et al., 2008). In these researches, the structural behavior as well as correlation between internal LNG flow and external container can be considered in a unified formulation. On the other words, the sloshinginduced flow and hydrodynamic pressure can be obtained during the global analysis. In addition, the structural stress/strain and deformation can be estimated during the local analysis. In the process of the local analysis, the container is considered as a deformable body (not a rigid body) and boundary/initial conditions at an arbitrary time are implemented into the local analysis procedures. For this reason, the computational time and cost can be saved during the globallocal analysis process.
Hence, in the present study, the globallocal method based hydroelastic analysis which is successfully derived by Cho et al. (2008) has been adopted to evaluate the structural safety as well as hydrodynamic pressure induced by interior LNG flow for the KC1type LNG insulation system. During the global analysis, the flow motion characteristics such as the flow velocity, hydrodynamic pressure and volume fraction which are used as an initial and boundary condition for the local analysis are obtained. In addition, during the local analysis, the structural behaviors of the actual sized KC1type insulation system such as effective stress/strain and deformation are estimated precisely.
NUMERICAL MODELING OF HYDROELASTIC PROBLEM FOR LNG CARRIER’S INSULATION SYSTEM UNDER SLOSHING
> Description of numerical modeling
The aforementioned hydroelastic problem has been analyzed by many scientists and engineers (Morand and Ohayon, 1995), especially Cho et al. (2008) who have successfully solved the hydroelastic problem of the MARKIII type LNG insulation system under sloshing. Prior to the FSI analysis for the KC1type LNG insulation system, the theoretical backgrounds such as problem descriptions, globallocal analysis method are discussed in Chapter 2 based on the authors’ previous study (Cho et al., 2008).
The hydroelastic response problem of the LNG carrier’s insulation system is a representative example of the FSI problem, namely, the deformation of the insulation system (structure) is induced by LNG sloshing (fluid). Fig. 3 shows the LNG carrier’s insulation system with length
L , breadthB , heightH , and thicknesst . The fill ratio of LNG (i.e., volume of LNG with respect to insulation system internal volume) is defined asn_{R} . The insulation system is applied to external excitation during the operation.By considering the insulation system as a threedimensional linear elasticity, the material and boundary domain can be written as Ω∈
R ^{3} andrespectively, where ∂Ω
_{D} is the essential boundary part and ∂Ω_{I} is the timevarying common interface contacting with interior LNG. The interface ∂Ω_{D} between the LNG and insulation system changes according to time. By neglecting the damping induced by the viscoelastic characteristics of the insulation ratio, the undamped dynamic displacement fieldu (x ;t ) is satisfied, and the initial and boundary condition can be represented aswhere
σ_{ij} is the Cauchy stress andρ andf are the density and body force of the insulation system, respectively,p is the hydrodynamic pressure caused by LNG,n is the normal unit vector in boundary ∂Ω, andT is the analysis time section.u ^{0} andare the initial value of the displacement and velocity fields, respectively.
u denotes the displacement constraint by the rigid dummy closure, which moves by the rigid body motion of the global rigid container.The LNG domain transformed into the movement of the free surface ∂Ω
^{S}_{F} can be expressed as Ω_{F} ∈R ^{3}. Moreover, for hydroelastic analysis of the LNG movement in the insulation system, the effects of viscosity and temperature change can be neglected due to the large amount of internal volume; hence, the sloshing of LNG can be considered as incompressible and inviscid. Therefore, the velocity of fluidV (x ;t ) can be written as follows.The momentum equation can be written as follows.
The initial and boundary conditions of Eqs. (5) and (6) can be represented as
where
ρ_{F} is the density of LNG,u (x ;t ) is the displacement of the insulation system,n ^{F} is the normal unit vector with regard to the flow boundary, and ∂Ω_{I} and ∂Ω^{S}_{F} are the boundaries of the insulation system and free surface of LNG, respectively. The tractionacting on the LNGfree boundary disappears when the flow is assumed to be inviscid (Cho and Lee, 2003). The total stress tensor of LNG
σ^{F} can be written asσ^{F} = ？p I , wherep is the hydrodynamic pressure of fluid andI is the unit tensor.The flow boundary can be used to obtain the transport equation and its initial condition; namely,
where
F_{ini} is the initial value of the volume fractionF . According to the Eulerian kinematic description, the volume fractions of the LNG and void are 1 (LNG flow region) and 0 (void region), respectively.> Globallocal analysis technique
In order to analyze the structural response of complex structures, a huge number of degrees of freedom are needed; hence, a large amount of computational time and cost is required. Thus, it is impossible to obtain the structural response for the entire structure as a whole. The globallocal analysis technique was introduced as an alternative method to investigate the structural response of regions of interest in structures (Mote, 1971; Mao and Sun, 1991). In globallocal analysis, the global analysis is carried out prior to the local analysis. The finite element (FE) model is simplified to save the calculation time. By using approximate mechanical information of the global structures such as the reaction force, the displacement can be obtained. Based on this information, local analysis of the region of interest is then performed. In local analysis, the FE model is fabricated as a real structural/material model, and specific mechanical information of specific regions can finally be acquired. The globallocal analysis technique is widely used in the structural analysis of complex structures due to its reliable analysis results and efficient computational time/cost.
Fig. 4 shows the globallocal approach for the hydroelastic analysis. The structural response of the righttop corner region of the LNG insulation system was investigated at critical time
t_{c} . Fig. 4(a) shows the simplified global rigidinsulation system sloshing model. The goal of the global analysis was to find the primary mechanical information of the rigidinsulation system sloshing problem: the total flow fieldV _{G} (x ;t ), pressure fieldp_{G} (x ;t ) and volume fractionF_{G} (x ;t ) of the internal LNG (Cho and Lee, 2004). These global analysis results can be used as an initial condition for the local analysis. Fig. 4(b) shows the local flexible hydroelastic model, which consists of the locally flexible insulation system domain Ω^{L} and local LNG domain Ω^{L}_{F} .In local hydroelastic analysis, the continuity and momentum equations can be written as
where
V _{L} is the flow velocity within the local flow region Ω^{L}_{F} with the local boundary ∂Ω^{L}_{F} and σ^{F}_{L} is the total stress in local region. The initial and boundary conditions can be represented as follows.The transport equation and its boundary condition can be written as
where
T^{τ} is the total analysis time in local analysis andτ is the time sectiont ？tc andF_{L} is the local volume fraction.？ _{L} is the displacement of the dummyrigid insulation system model.T^{τ} is the very short time during the global analysis timeT , and the surface tension effect in the free surfacecan be ignored due to the assumption of inviscid flow.
The undamped displacement
u _{L} of the flexible local model, which has a boundary condition ofwhere ∂Ω
_{D} is the essential boundary part in local domain and ∂Ω_{I} is the timevarying common interface contacting with interior LNG in local region, can be written as follows.The initial and boundary conditions of the above equation can be expressed as
where
R _{G} (x ;t_{c} ) and？ _{G} (x ;t_{c} ) are the displacement and velocity, respectively, of the local insulation system Ω^{L} by rigidinsulation system motion at critical timet_{c} .u _{L} (x ;τ ) is the displacement constraint by the rigid shell dummy. As shown in Fig. 4, the rigid shell dummy moves along insulation system motion.> Numerical approximation for FVMFEM coupling
Eqs. (12), (13), and (16) can be represented by a generalized form, that is,
where Φ is the dependent variable, Λ is the coefficient,
S _{Φ} is the source term dependent on the governing equation, andis the extended Euler region including the void region within the local insulation system. In other words, Φ=1, Λ =
ρ_{F} , andS _{Φ} = 0 in the continuity equation; Φ =V_{j}^{L} , Λ =ρ_{F} andS _{Φ} = ？∂p_{L} / ∂x_{j} in the momentum equation; and Φ =F_{L} , Λ=1, andS _{Φ} = 0 in the volume fraction equation, respectively. In this study, the threedimensional finite volume method (FVM) was adopted for space discretization, and the firstorder Eulerian scheme was applied for the spatial discretization and the time integration of the three transport equations, respectively. The extended Eulerian domainis discretized into a finite number of nonoverlapping control volume, and the time period
T^{τ} is divided intoN time intervals Δτ =T^{τ} /N withN +1 time stagesτ_{n} =n Δτ (n = 0,1,2,…,N ).Fig. 5(a) shows the threedimensional control volume of grid point
P and the six neighboring grid points. The volume and surface of the control volume areδV andA , respectively. By integrating Eq. (22) with regards to the control volumeδV and time increment Δτ and by applying the divergence theorem, next equation can be obtained to solve the working variableat time stage
τ_{n} _{+1}.Here,
is the volume integral of the source term
S _{Φ}.where
is the part of
independent of Φ,
is the coefficient of Φ
_{P} , andand
are the coefficients at arbitrary time step
τ_{n} .From Eq. (24), the discretized formula to obtain the variable
at time step
τ_{n} _{+1} can be written aswhere
a ^{0}_{P} = ΛδV /Δτ .a_{nb} is the function of the dependent variable. This variable is different from that of the discretization theory and remeshing. In this study, the power scheme was adopted to describe the spatial variation approximation (Patankar, 1980).The momentum equations in the discretized linear algebraic equations system, i.e., Eq. (25) are firstly solved to compute timestepwise flow velocities using the initial conditions or the previous timelevel values. It is evident from Fig. 5(b) that volume and surface integrals of the partially filled elements are carried out over the effective volume and the effective boundary, respectively, where the effective boundary consists of the element interfaces and the part of the fluid free surface.
On the other hand, the calculated velocities do not satisfy the continuity equation, hence, the pressures should be adjusted in each control volume element occupied by the fluid. From Fig. 6(a), the elementcentered pressure
p_{C} at the free surface and the pressurep_{N} of one of the neighboring elements inside the fluid can be written aswith
η =d_{C} /d . The neighbor element should be chosen such that the line connecting the centers of control elements is closest to the normal to the free surface (Hirt and Nichols, 1981).Once the pressure and velocity fields are calculated, the volume fraction equation in the discretized linear algebraic equation system, i.e., Eq. (25) is calculated to compute the elementwise constant volume fraction as shown in Fig. 6(b). In addition, in order to maintain the computational stability, the timestep size should not exceed the critical value given by
where
c is the speed of sound in fluid,u is the flow velocity, andh is the smallest distance between two adjacent element centers.By applying the isotropic finite element approximation technique to Eq. (18), the undamped displacement field
u _{L} can be represented as follows.where
？ _{L} is the undamped displacement field regarding to only one variable, and in the present study, the variable is time (Zienkiewicz et al., 2005). As a result, the following matrix equations can be acquired.where
F is the load vector by gravity and hydrodynamic pressure; this is written as follows.The time integration is carried out based on the central difference method (explicit method) and/or Newmark method (implicit method). However, the explicit method is widely adopted in dynamic structural problems of huge/complex structures due to the reduction of analysis time (Hu, 1997). The local analysis time
T^{τ} can be divided by the detail intervals of the finite numbers. The equation of motion (Eq. (29)) can then be replaced by the explicit time increment form such as through the following equation according to the lumped mass matrixAccording to the central difference method and dynamic relaxation method, which is based on the αdamping method, the velocity and displacement of each time step are derived as
where
α ？ 1.67ω _{max} Δτ . The size of the time increment step is properly decided from the following formula to consider the convergence and stability of the numerical analysis.where
ω _{max} is the largest element frequency in the finite element mesh (Cho et al., 2006; Cho et al., 2007).On the other hand, in the local hydroelastic problem, the LNG flow and structural motion influence each other at the common surface ∂Ω
^{L}_{I} . These physical interactions must satisfy the following kinematic and contact stress constraints (Schafer and Teschauer, 2001).The fluidstructure constraints can be numerically represented by direct standardization and the iterative discretized method. Applying the former method to complex structures is extremely difficult. Therefore, the latter method was adopted for this study. The two kinds of equations (i.e., fluid domain and structure domain governing equations) are separately calculated by using the iterative discretized method, and the interaction of the fluid and structure is calculated alternately. In order to overcome the mismatch of the mesh, the projection method was also adopted (Farhat et al., 1998).
In the local analysis for this study, the fixed Eulerian mesh was used for the LNG flow, and the complex and dense mesh was used for structures. In this case, the incompatible EulerianLagrangian interaction method can be used effectively. As shown in Fig. 7(a), the fluid mesh is larger than the physical domain of LNG, and the active and inactive domains are separated by the common interface ∂Ω
^{L}_{I} . On the other hand, the dummyrigid insulation system coupled model is implemented to the fixed Eulerian mesh, and its closed outer surface serves as a coupling surface Γ. In general in the EulerLagrange coupling, the coupling surface should be closed and set larger than the common fluidstructure interface, namely, Γ ⊃ ∂Ω^{L}_{I} .In order to transfer the hydrodynamic pressure
p_{L} of the flow field to the structure and the displacement？ _{L} and velocityof the structure to the flow field, the related variables must be interpolated. In this study, the structure mesh was denser than the fluid mesh; hence, the mesh pattern and basic function of the structure can be used as a standard mesh to interpolate the related variables. As shown in Fig. 7(b), the surface force
which is transferred to the surface element ∂Ω
^{L}_{K} of the structure interface, is implicitly interpolated as
n numbers of shape functionsψ_{k} , which is defined at the twodimensional master elementwhere
p_{L,k} is the pressure at nodek above the structure interfacewhich is transferred from
and
is the twodimensional master element.
The displacement and velocity of structure are transferred to the flow field using the incompatibility interpolation method (Harder and Desmarais, 1972). Here, note that the former adjusts the flow boundary of LNG, whereas the latter one specifies the flow boundary condition given in Eq. (15). In Fig. 7(b),
j andk represent thej th node ofand
k th node of. Eq. (35), that is, the kinematic constraint, can be written as follows using the
H coupling matrix (Beckert, 2003).The coupling in time can be done using either explicit or implicit approach, and the flowchart of the analysis algorithm in both approaches is presented in Fig. 8 (Cho et al., 2008). Based on the analysis algorithm, the userdefined subroutine is programmed. In this study, the explicit time coupling for the current problem with the critical timestep size determined by Eq. (27)
and (34). The numerical implementation of the local staggered hydroelastic analysis using the explicit incompatible Eulerian Lagrangian coupling method is carried out as follows.
(1) The initial structureLNG interface ∂ΩLI in the local model is specified and the initial free surface
is defined by assigning the volume fraction FL to each local Eulerian cell.
(2) Based on initial and boundary conditions, i.e., Eqs (19)(21), the structural dynamic equation system, namely, Eq. (18) is time integrated to solve the displacement uL and the velocity
of the local flexible insulation system.
(3) Next, the LNG interface
is moved by uL and specified with
as the interface flow boundary condition, i.e., Eq. (15) according to nonmatching interpolation scheme, namely, Eq. (38).
(4) Based on the adjusted active LNG material boundary, the flow boundary conditions and the initial condition, i.e., Eq. (14), the transport equations (Eq. (22)) are solved to find the local flow velocity VL, the volume fraction FL, and the hydrodynamic pressure pL.
(5) The free surface
and the structureLNG interface ∂ΩLI are updated, and then the LNG hydrodynamic pL is transferred to the structure interface
. And then, go to the next time integration.
DESCRIPTIONS OF KC1 TYPE INSULATION SYSTEM FE MODELS
> FE models
Figs. 9 and 10 show the local FE model and its component for the KC1 type LNG carrier’s insulation system. Fig. 11(a) shows the dummy element used as an interface, and Fig. 11(b) shows the coupled interface between the dummy and the membrane of the insulation system. The mechanical properties and element type of the KC1 type FE model are shown in Table 1.
> Boundary conditions
Fig. 12 shows the boundary conditions of the dummylocal KC1 insulation system model. If node sharing is applied to this model, the stress concentration may occur on the interface between the dummy and local model. In other words, the model types of the dummy and local models are rigid and deformable, respectively. Hence, the mismatch in deformation can cause an abnormal increase of stress. In order to overcome this problem, the contact element was adopted between the two models used in this study.
> Simplified FE model for anchor section
An anchor section exists between the insulation system and hull. The anchor combines the RPUF and membrane firmly. The surface of the anchor is made of SUS 304L, and inside is filled with RPUF. The implemented position and lamination structure of the anchor is shown in Fig. 13.
The anchor has extremely complex lamination structures, as well as others, and it might be necessary to create a mesh that is quite fine. However, this could induce a huge amount of calculation time to analyze the anchor region. In order to save the computational time and cost, the simplified FE model of the anchor section was introduced in this study. The two kinds of analysis scenarios were adopted as shown in Fig. 14.
Case A is the general FE model, which was fabricated by using element refinement. The connection area between the membrane and anchor was established as surfacetosurface contact. On the other hand, in Case B, the upper side of the anchor was eliminated for fast calculation. Instead, the additional thickness model with a coarser mesh size than the asis model was implemented to the upper side of the anchor. In addition, this model was unified with the membrane FE model. By using the additional thickness model, the calculation time can be improved. Moreover, the modeling time can be reduced due to the absence of the contact element. The related description such as number of elements and computational time will be discussed in Chapter 4.
GLOBALLOCAL HYDROELASTIC ANALYSIS RESULTS AND DISCUSSIONS
> Global analysis
Fig. 15 shows the FE model of the LNG carrier insulation system for global analysis. The dimensions of the FE model were 42.84
m × 37.4m × 27.2m (length × breadth × height). The filling level (η_{R} ), density (ρ_{F} ), and bulk modulus (K ) of the LNG were 95%, 500kg/m^{3} , and 1.44GPa , respectively. The FE model was fabricated based on a control volume element (Euler domain) and bilinear rigid shell element (Lagrangian domain). The numbers of each element are 135,608 and 48,366, respectively, as shown in Fig. 15(a). The analysis was carried out using MSC. Dytran and the interfaces between the LNG flow and rigid insulation system were fabricated using the arbitrary Lagrangian Eulerian (ALE) method.The insulation system was applied to a sinusoidal rolling excitation as shown in Fig. 15(b). The angular velocity was defined as
and the amplitude and the angular frequency were defined as a=3.87° and ω=2π/5.5
rad/sec , respectively, based on the actual measurement data in the shipping service. The rolling starts with the extremely small excitation until 5sec for the simulation purpose of the huge and heavy insulation system.The computational analysis time was set to 0
sec to 4014sec based on the actual measured data of the target LNG carrier. In the case of this target LNG cargo tank, the critical situation was occurred once per approximately 4000sec . The averaged time increment and the corresponding number of iterations were= 3.598×10^{3}
sec and 1,115,508, respectively.Fig. 16(a) and (b) show the volume fraction distribution and the sloshing flow of interior LNG at time 4010
sec . At this time, the insulation system inclined 12° in the clockwise direction. The choice of the critical time itself is not a main interest considering the goal of the present study, but the critical time was selected from the fact that the local region of concern is subjected to the remarkable hydrodynamic impact when the LNG sloshing flow is going up along the inner surface of the insulation system. It can be clearly seen that the rigid tank sloshing produces high LNG flow along the inner boundary of the insulation system, particularly at all the corners of the insulation system (Cho et al., 2008). According to the detailed numerical data, the peak flow velocity and the peak hydrodynamic pressure at the critical time are 1.26m/s and 0.142MPa , respectively.It is noted that the insulation system experiences the peak rolling amplitude equal to 22.1° in the rigid tank sloshing. Although the excitation from 5 to 30
sec is small and looks like not changeable as shown in Fig. 15(b), the cargo tank as well as LNG carrier are rolled to approximately 22° due to sloshing and ship motion coupling phenomenon. This has been already discussed by a few naval architects such as Faltinsen and Timokha (Faltinsen and Timokha, 2009). In their study, the increase of the ship motion amplitude is experimentally investigated and related governing equations are derived. Moreover, the concept of the antirolling ship is also proposed by controlling the coupling between sloshing and ship motion. In the present study, this coupling phenomenon is also observed during the simulation (global analysis).Fig. 17 shows the mapping of the LNG flow into the local problem region Ω
^{L}_{F} . The global flow velocity field V_{G} (x ;t_{c} ) and the global hydrodynamic pressure fieldp_{G} (x ;t_{c} ) of the finite volume mesh of the global model were mapped into the finite volume mesh of the local model. If the free surface is totally positioned in the internal/external region of the control volume, the volume fractionF_{L} is 1 or 0. The value of the volume fraction is decided as 0 <F_{L} < 1 according to the position of the free surface. The three kinds of calculated data (i.e., velocity field, pressure field, and volume fraction) were applied to the initial conditions of Eqs. (12)(17). The fluid field mesh of the fine local model consisted of 31589 uniform control volumes; the mechanical values ofρ_{F} andK used for LNG in the local model were the same as the values from the global analysis results.> Local analysis
Based on the data of the global analysis results, the local analysis was carried out. Figs. 18 and 19 show the local analysis results of the effective stress contour regarding Cases A and B. In both cases, the maximum effective stress occurred in the membrane region at 35
ms . The specific value of the maximum effective stress is listed in Table 2. In addition, the computational times for Case A and B are shown in Table 3. As shown in these tables, the differences between each case were found to be quite small and the computational time of the proposed FE model (Case B) is much less than the asis FE model (Case A); therefore, it is reasonable to choose the Case B type anchor model during local analysis.Through Table 2, it is confirmed that the average stress value of Case B is bigger (but not much) than the value of Case A. It is considered that this phenomenon might be caused by the differences of element types, i.e., shellcontact element type in Case A and shellhexahedral element type in Case B. However, these differences can be acceptable considering the computational time and cost as shown in Table 3. In this table, it can be found that the CPU time of Case A is much bigger than the Case B. Since it takes huge time to calculate the interface and contact regions in the Case A FE model. Hence, it is confirmed that the proposed FE model is more time and cost effective case when the FSI analysis of the LNG carrier cargo tank is carried out.
CONCLUDION REMARKS
In this study, the sloshing resistance performance for the KC1 type LNG carrier insulation system was evaluated by FSI analysis. The research results are listed below.
(1) FSI analysis was carried out based on the wellknown arbitrary LagrangianEulerian method, and other novel techniques for FSI analysis were introduced.
(2) The flow velocity, pressure, and volume fraction of the internal LNG of the KC1 insulation system can be calculated during the sufficient sloshing time based on the global analysis method. The obtained values are mapped into the local FE model of the KC1 insulation system.
(3) The maximum effective stress contours can be calculated based on the local analysis method. During local analysis, the arbitrary LagrangianEulerian method based FSI analysis is adopted. Moreover, a novel element model, i.e, additional thickness model without contact element, was implemented into the local KC1 insulation system in order to remarkably reduce the computational time and cost.
(4) Although validation of the analysis results through a comparison of the experiments was not carried out, the proposed analysis technique can potentially be used as a robust integrity assessment technique for LNG carrier’s insulation systems.

8. Farhat C., Lesoinne M., LeTallec P. 1998 Load and motion transfer algorithms for fluid/structure interaction problems with nonmatching discrete interfaces: momentum and energy conservation, optimal discretization and application to aeroelasticity. [Computer Methods in Applied Mechanics and Engineering] Vol.157 P.95114

[Fig. 1] Membranetype LNG carrier with insulation system.

[Fig. 2] Schematic of KC1type LNG carrier’s insulation system.

[Fig. 3] Partially filled insulation system subjected to external excitation (Cho et al., 2008).

[Fig. 4] Globallocal approach for hydroelastic analysis: (a) global rigidinsulation system sloshing model and (b) local flexible hydroelastic model (Cho et al., 2008).

[Fig. 5] Schematic representation of (a) threedimensional control volume element and (b) the effective volume and boundary of an element intersecting with the free surface (Cho et al., 2008).

[Fig. 6] Schematic representation of (a) pressure interpolation for an element intersecting with the free surface and (b) free surface interpolation within the extended Euler domain (Cho et al., 2008).

[Fig. 7] Schematic of (a) incompatible EulerianLagrangian coupling and (b) transfer through nonmatching fluidstructure interface (Cho et al., 2008).

[Fig. 8] Explicit and implicit fluidstructure coupling methods (Sigrist and Abouri, 2006; Cho et al., 2008).

[Fig. 9] Local model of KC1 type LNG carrier’s insulation system.

[Fig. 10] Components of KC1 LNG carrier’s insulation system.

[Fig. 11] Construction of coupling surface for KC1 type local model.

[Table 1] Material properties and element type of KC1 type model.

[Fig. 12] Construction of coupling surface for KC1 type local model.

[Fig. 13] Implemented position and lamination structure of anchor.

[Fig. 14] Anchor model types: (a) Case A (fine mesh model) and (b) Case B (simplified model).

[Fig. 15] LNG insulation system in rolling motion: (a) rigid insulation system model and (b) sinusoidal rolling excitation.

[Fig. 16] Global analysis results at t = 4010 sec: (a) LNG volume fraction and (b) LNG flow (unit: m/s).

[Fig. 17] Mapping of the LNG flow into the local problem region: (a) global view, (b) detailed view.

[Fig. 18] Effective stress contour of KC1 model Case A.

[Fig. 19] Effective stress contour of KC1 model Case B.

[Table 2] Maximum effective stress at 35 ms (local time) and 4010.035 sec (global time) (Unit: MPa).

[Table 3] Computational time for Case A and Case B.