Anda di halaman 1dari 7

ADVANCED FINITE ELEMENT METHODS FOR FLUID DYNAMIC ANALYSIS OF SHIPS

E. Oate and J. Garca International Center for Numerical Methods in Engineering (CIMNE) Universidad Politcnica de Catalua Campus Norte UPC,08034 Barcelona,Spain e-mail:julio@cimne.upc.es, web page:http://www.cimne.upc.es Key words: Finite element, incompressible ows, free surface, nite calculus

Introduction
In recent years, the advent of advanced numerical schemes for the Navier Stokes (NS) equations and the rapid development of computer hardware have enabled a realistic prediction of fluid flow about ship forms. The troubles in accurately solving this problem are mainly due to the difficulty of solving numerically the incompressible fluid dynamic equations, which include significant non linearities and the obstacles in solving the constraint equation stating that at the free surface boundary the fluid particles remain on that surface which position is in turn unknown. Among the schemes developed over the last decade for the solution of the incompressible NS equations the fractional step schemes [1,2,3,4] yield highly accurate, pressure-stable [2] results by integrating in an explicit manner the advective terms of the NS equations. However, in most of the cases of interest for the naval architecture the time step imposed by the smallest elements may be orders of magnitude smaller than the time step required to obtain time-accurate results (physical time step). In some cases this implies tens of thousands of time steps per simulation, rendering the schemes impractical. Most of the artificial compressibility and preconditioned schemes suffer from the same shortcoming. On the other hand, the monolithic schemes treat, in general, the advective term in an implicit manner, which avoids the mentioned disadvantages. Nevertheless, these methods are very expensive from a computational point of view: the velocity and pressure discrete equations are coupled. This paper presents advances in recent work of the authors to derive a fractional step scheme based on the stabilized finite element method that allows overcoming the above mentioned problem. The starting point is the modified governing differential equations for the incompressible turbulent viscous flow and the free surface condition incorporating the necessary stabilization terms via a finite calculus (FIC) procedure developed by the authors [5,6,7,8]. To reach the mentioned objective, an implicit and uncoupled second order fractional step method based on the scheme originally proposed by Soto [9] is presented.

FIC formulation
We consider the motion around a body of a viscous incompressible fluid including a free surface. The stabilized finite calculus (FIC) form of the governing differential equations [10] for the three dimensional (3D) problem can be written as 1 rm =0 rm h j 2 x j
i i

on i, j = 1,2,3 on on j = 1,2,3 j = 1,2

(1a)

1 r rd h j d = 0 2 x j r r 1 h = 0 2 x j
j

(1b)

(1c)

where rm =
i

ui p + (uiu j ) + x xij t x j i j rd = r = ui xi i = 1,2,3 i = 1,2,3

+ ui t xi

In above ui is the velocity along the i-th global reference axis, p is the dynamic pressure, is the wave elevation and ij are the viscous stress tensor components. Let ni be the unit outward normal to the boundary and denoting by an overbar prescribed values, the boundary conditions for the stabilized problem are: u=u p=p and on u , 1 n j ij h j n j rm = t i 2
i

(2a) on p (2b)

1 1 u j n j = u n , n j ij gi h j nj r g i = t1 and n j ij si h j n j rm si = t2 on (2c) m 2 2 for t (t0,tf). The boundary has been considered split into three sets of disjoint components u, p and , the latter being the part where mixed conditions are prescribed. Vectors gi and si span the space tangent to . Finally, u and p are the two disjoint components of where Dirichlet and Neumann boundary conditions for the velocity are prescribed. Initial conditions have to be appended to problem (1)-(2).
i i

The underlined terms in eqs. (1)-(2) introduce the necessary stabilization for the numerical solution. Additional time stabilization terms can be accounted for in eqs. (1)(2) [5,6] although they have been found unnecessary for the type of problem solved here. The characteristic length distances hj represent the dimensions of the finite domain where balance of momentum and mass is enforced. The characteristic distances hj in eq. (1c) represent the dimensions of a finite domain surrounding a point where the velocity is constrained to be tangent to the free surface. Details of the derivation of eqs.

(1)-(2) and recommendations for computation of the stabilization parameters can be found in [5,11,12]. Eqs. (1)-(2) are the starting point for deriving a variety of stabilized numerical methods for solving the incompressible Navier-Stokes equations with a free surface. It can be shown that a number of stabilized finite element methods and new meshless methods allowing equal order interpolations for the velocity and pressure fields can be from the modified form of the momentum and mass balance equations given above [5-8].

Implicit Fractional Step Formulation


Let us discretize in time the stabilized momentum equation (1a) using the trapezoidal rule (or method) as
n + n + uin +1 uin p n + ij 1 r + uin + u n + ) + hj m = 0 ( j t x j xi x j x j 2
i

(3)

the superscripts n and refer to the time step and to the trapezoidal rule discretization parameter, respectively. For = 1 the standard backward Euler scheme is obtained, which has a temporal error of 0(t). The value = 0.5 gives the standard Crank Nicholson scheme, which is second order accurate in time 0(t2). A classical implicit fractional step method can be simply derived by splitting eq. (3). The resulting continuous problem, omitting the boundary and initial conditions for brevity, is as follows:
n ijn + r n+ uin+1 uin + ( uin + u nj + ) + p x 1 h j m = 0 t x j xi 2 xj j
i

(4a)

2 u* 1 rd n+1 n t ( p p ) = xi + 2 h j x xixi i j uin+1 = ui t ( p n+1 p n ) xi

(4b)

(4c)

where the terms denoted by an overbar are those calculated with the intermediate velocity i, which is introduced to allow the momentum splitting. The error due to taking implicit advective and viscous terms in (4a) can be shown [9] to be of the same order than the error of the stabilizing term and therefore, it has the same order of approximation than the original time discretization (3).

Monolithic Stabilized Scheme


At this point, it is important to introduce the associated matrix structure corresponding to the variational discrete form of (4) (see [5] for details of this derivation): M 1 ( Un+1 U n ) + K (U n+ )U n+ GPn = 0 t ( t + ) L ( P n +1 P n ) HU n + DU n +1 = 0 (5a) (5b)

1 (5c) (U n +1 U n +1 ) ( Pn +1 Pn ) = 0 t By taking n+1 from (5c) and inserting the result in (5a)-(5b), the following system of equations is obtained:

1 ( Un+1 U n ) + K (U n+ )U n+ GP n = E (U n+ ) t t ( L DM 1G )( P n+1 P n ) H (U n , P n ) = DU n +1

(6a) (6b)

The term E(Un+) in (6a) is the error coming from the implicit treatment of the advective and viscous terms, which is of order 0(t2). However, such term can be eliminated as in (6a)-(6b) by writing the following analog monolithic scheme: 1 (U n+1, i U n ) + K (U n+, i ) U n+,i GPn+1,i1 = 0 t t ( L DM 1G )( P n+1,i P n ) H (U n+1,i , P n+1,i ) = DU n+1,i M (7a) (7b)

Basically, in this final formulation the convergence of the block uncoupled solution is enforced by the first term of (7b), while the pressure stability is attained by the second term of the same equation. Finally, the wave elevation coupling effect is included in the monolithic scheme, obtaining the following system of equations: 1 (U n+1, i U n ) + K (U n+ , i ) U n+ ,i GPn+1,i1 = 0 t t ( L DM 1G )( P n+1, i P n ) H (U n+1, i , P n+1,i ) = DU n+1,i M M 1 ( Bn +1,i Bn ) + A(U n + ,i ) Bn + ,i R(U n + ,i , B n ) = MU3n + ,i t (8a) (8b) (8c)

where the Uin+ is the vector of the velocity along the i-th global reference axis.

Application Example
The application example here presented is the parametric analysis of a 79 catamaran designed by the company NAUTATEC. The main characteristics of this boat are listed next (see Figure 1): Length overall Length between perpendiculars Total breadth Extreme breadth of every hull Draught with appendages Hull depth Displacement Maximum sail area Passengers 23.95 m 23.35 m 11.75 m 1.65 m 1.31 m 2.40 m 26 730 Kg 222 m2 120

This study started from a fixed well-known geometry of the hulls, being the object of the study, the effect of the distance between hulls in the resistance and lateral forces. Three different alternatives of distance between hulls studied were: Case Distance between hulls centerlines Distance between hulls centerlines Name 79cata5_0 79cata4_5

10.1 m 9.1 m

Distance between hulls centerlines

8.1 m

79cata4_0

Figure 1. NURBS based geometrical definition of the 79catamaran. Different alternatives studied.

The analyses were carried out for a range of ship speed from 8 kn to 28 kn, being the real range from 8 kn to 16 kn. Regarding the drift angles, four cases were considered: 0.0 , 1.0, 2.5 and 5.0, being the project range from 0.0 to 2.5. The grid used for the analyses consisted of 450.000 linear tetrahedra. The fluid viscosity taken was 1.3 Kg/ms and the density 1024 Kg/m3. The k turbulence model with the extended law of the wall was used. The basic results of these analyses are presented next: Resistance i a Resistenc
45 40 35 30 79cata4_5der0 79cata4_5der1 79cata4_5der2,5 79cata4_5der5 0 -20 -40 -60 8 10 12 14 Nudos 16 18 20 22 24 26 28

KN

25

KN

20 15 10 5 0 8 10 12 14 16 18 Nudos 20 22 24 26 28

-80 -100 -120 -140 -160 -180

79cata4_5der0 79cata4_5der1 79cata4_5der2,5 79cata4_5der5

Sustentacin

Lift

Figure 2. Resistance and lift curves for the 79cata4_5 case.

Resistencia Resistance 30 25 20 KN 15 10 5 0 8 10 12 14 16 18 20 Nudos 22 24 26 28 79cata5_0der0 79cata4_5der0 79cata4_0der0 KN 35 30 25 20 15 10 5 0 8 10 12 14

Resistencia Resistance

79cata5_0der1 79cata4_5der1 79cata4_0der1

16

18 20 Nudos

22

24

26

28

Resistencia Resistance 35 30 25 KN 20 15 10 5 0 8 10 12 14 16 18 20 Nudos 22 24 26 28 79cata5_0der2,5 79cata4_5der2,5 79cata4_0der2,5 KN 45 40 35 30 25 20 15 10 5 0 8 10 12 14

Resistencia Resistance 79cata5_0der5 79cata4_5der5 79cata4_0der5

16

18 20 Nudos

22

24

26

28

Figure 3. Resistance curves for the different cases analyzed.

In order to evaluate de efficiency of the implicit algorithm, the CPU time needed to achieve the pseudo-steady state using both the explicit and implicit algorithms was evaluated. This test was carried out for three different grids. Using these, one single point of the 79_cata5_0 geometry analysis was run. The time step chosen for the implicit algorithm was 0.01L/V, being L the length of the ship and V the speed of the analysis. For this time step, no significant differences were found between the results obtained with the explicit or implicit algorithm. The results of this comparison are shown next. Grid characteristics 150 000 tetrahedra elements 300 000 tetrahedra elements 450 000 tetrahedra elements 600 000 tetrahedra elements Explicit / Implicit algorithm time ratio 1.1 : 1.0 1.3 : 1.0 2.2 : 1.0 4.1 : 1.0

Conclusions
An implicit second-order accurate monolithic scheme, based on the FIC formulation was presented to solve incompressible free surface flow problems. The final system of equations resulting from the time and space discretization is solved in each time step in an uncoupled manner. The numerical experience indicates that the formulation is very efficient for free surfaces flows, when the critical time step of the problem is some orders of magnitude

smaller than the time step required to obtain time-accurate results (physical time step), and that its time accuracy is excellent.

Acknowledgements
The authors want to thank the support of NAUTATEC (nautatec@nautatec.com) for this work. The authors also thank Dr. O. Soto for many useful discussions.

References
[1] Chorin. On the convergence of discrete approximation to the Navier-Stokes equations. Math. Comput., 23, 1969. [2] R. Codina. Pressure stability in fractional step finite element methods for incompressible flows. J. Comp. Phys., Submitted for publication, 2000. [3] J. Garca, E. Oate, H. Sierra, C. Sacco y S. Idelsohn. A stabilised numerical method for analysis of ship hydrodynamics. ECCOMAS 98 (Vol. II). Atenas 1998. [4] R. Lhner, C. Yang, E. Oate, and S. Idelsohn. An unstructured grid-based, parallel free surface solver. AIAA-97-1830, 1997. [5] E. Oate and J. Garca. Finite Element Analysis of Incompressible Flows with Free Surface Waves using a Finite Calculus Formulation. ECCOMAS 2001. Swansea UK 2001. [6] E. Oate. A stabilized finite element method for incompressible viscous flows using a finite increment calculus formulation. Comp. Meth. Appl. Mech. Engng. 182, 1-2, 355370, 2000. [7] E. Oate and J. Garca. A methodology for analysis of uid-structure interaction accounting for free surface waves. European Conference on Computational Mechanics (ECCM99). Munich, Germany, September 1999. [8] E. Oate and J. Garca. A stabilized finite element method for analysis of fluid structure interaction problems involving free surface waves. Proceedings of Fluid Structure Interaction Conference, Trondheim (1999). [9] O. Soto, R. Lhner, J. Cebral and R. Codina. A Time Accurate Implicit-Monolithic Finite Element Scheme for Incompressible Flow. ECCOMAS 2001. Swansea UK 2001. [10] E. Oate, J. Garca and S. Idelsohn, An alpha adaptive approach for stabilized finite element solution of advective-diffusive problems with sharp gradients, New Adv. In adaptive Comp. Met. In Mech., P. Ladeveze and J.T. Oden (Eds.), Elsevier (1998). [11] E. Oate, Derivation of stabilized equations for advective-diffusive transport and fluid flow problems, Comput. Meth. Appl. Mech. Engng., Vol. 151, 1-2, pp. 233-267 (1998). [12] E. Oate, J. Garca and S. Idelsohn, Computation of the stabilization parameter for the finite element solution of advective-diffusive problems with sharp gradients, Int. J. Num. Meth. Fluids, Vol. 25, pp. 1385-1407 (1997).

Anda mungkin juga menyukai