You are on page 1of 111

Tom 52 (66) Fascicola 3 ISSN 1224-6077

SCIENTIFIC BULLETIN
of the POLITEHNICA University of Timioara, Romania Transactions of Mechanics

Proceedings of the 3nd Workshop on Vortex Dominated Flows Achievements and Open Problems
Timioara, Romania, June 1-2, 2007

Edited by S. Bernad, S. Muntean, R. Susan-Resiga

Editor in chief Prof.dr.ing. Victor BLOIU

Editors Dr.ing. Sandor BERNAD, Dr.ing. Sebastian MUNTEAN, Prof.dr.ing. Romeo SUSAN-RESIGA

Workshop Scientific Committee


Prof. R. RESIGA Charmain Dr. S. BERNAD Dr. S. MUNTEAN Prof. E.C. ISBOIU Prof. C. BERBENTE Prof. L. SANDU Prof. A. LUNGU Prof. L.I. VAIDA Prof. V. CMPIAN Dr. Th. POPESCU
Politehnica University of Timisoara

Romanian Academy Timioara Branch Romanian Academy Timioara Branch Politehnica University of Bucharest Politehnica University of Bucharest Technical University of Civil Engineering, Bucharest Dunrea de Jos University Galai Technical University, Cluj-Napoca Eftimie Murgu University, Reia Gheorghe Asachi Technical University, Iai

Editura ORIZONTURI UNIVERSITARE

Typesetting: Digital data supplied by editors Cover-Design: Romeo SUSAN-RESIGA, Sandor BERNADA, Dan NIU Editorial adviser: tefan KILYENI Desktop publishing: Valentina TEF

Tiparul executat la Imprimeria MIRTON 300125 Timioara, Str. Samuil Micu nr. 7 Telefon: 0256-272926, 0256-225684 Fax: 0256-208924

FOREWORD

The Academic Consortium for Research and Development on Fluid Dynamics (ACCORDFluiD), which brings together research teams from leading Romanian universities in a joint effort to address current challenges on vortex and swirling flows aero- and hydro-dynamics, has organized its third Workshop on Vortex Dominated Flows Achievements and Open Problems in Timioara June 1 2, 2007. This research consortium is supported by the Romanian National University Research Council grant No. 33 for the period 2005-2007, and it is aimed at: Fundamental research, focused on ellucidating the physics of vortex flows, and on developing new mathematical models and numerical techniques for both inviscid as well as unsteady turbulent swirling flows; Development of numerical techniques able to meet the special requirements for vortex flows, in particular swirling flows in adverse pressure gradient; Development of experimental techniques for vortex flows; Applications for turbomachines, naval propellers, wind engineering, combustion systems. The 2007 workshop brings together eight research teams from leading universities and research institutes in Romania: Politehnica University of Timioara; Romanian Academy Timioara Branch; Politehnica University from Bucharest; Aerospace Engineering School; Politehnica University from Bucharest; Power Engineering School; Technical Civil Engineering University; Bucharest; University Dunrea de Jos Galai; Technical University Cluj-Napoca; University Eftimie Murgu Reia; Technical University Gh. Asachi Iai. The two-day meeting, June 1-2, 2007, in Timioara is aimed at evaluating the current developments as well as at correlating the scientific efforts among the parteners. The 20 papers reviewed by the scientific committee, have been presented and discussed at the workshop then included in the present issue of the Scientific Bulletin of the Politehnica University of Timioara, Transactions on Mechanics. Although the ongoing grant ends in 2007, the Academic Consortium for Research and Development on Fluid Dynamics ACCORD-Fluid will continue its activity as a national research network. Prof.dr.ing. Romeo SUSAN-RESIGA

TABEL OF CONTENTS

Romeo SUSAN-RESIGA, Sebastian MUNTEAN, Vlad HASMAUCHI, Sandor BERNAD ............... Development of a Swirling Flow Control Technique for Francis Turbines Operated at Partial Discharge Andrei-Mugur GEORGESCU, Sanda-Carmen GEORGESCU, Sandor BERNAD, Costin Ioan COOIU ............................. COMSOL MULTIPHYSICS Versus FLUENT: 2D Numerical simulation of the stationary flow around a blade of the Achard Turbine Sandor BERNAD, Andrei-Mugur GEORGESCU, Carmen Sanda GEORGESCU, Danile BALINT, Romeo SUSAN-RESIGA .. 2D unsteady simulation of the flow in the Achard marine turbine Corneliu BERBENTE, Sterian DNIL, Marius STOIA-DJESKA ..................................................... A semianalitical method for the wing aerodynamics Marius STOIA-DJESKA, Sterian DANAILA, Corneliu BERBENTE ..................................................... A Physical And Theoretical Analysis of the Vortex breakdown on delta wings Constantin Viorel CMPIAN, Dorian NEDELCU ................................................................................... Cavitation tip clearance. Numerical simulation and experimental results Adrian STUPARU, Sebastian MUNTEAN, Daniel BALINT, Liviu ANTON, Alexandru BAYA ...... Numerical analysis of pump hydrodynamics at constant speed Eugen Constantin ISBASOIU, Petrisor STANESCU, Marius STOIA-DJESKA, Carmen Anca SAFTA, Georgiana DUNCA, Diana Maria BUCUR, Calin GHERGU... Swirling flows in the suction sumps. Experimental approach Mircea DEGERATU, Andrei GEORGESCU, Liviu HAEGAN, Costin Ioan COOIU, Rzvan- Silviu TEFAN, Lucian SANDU ......................................................................................................................... Some aspects about a vortex building model placed upwind an aeroelastic model in the boundary layer wind tunnel Adrian LUNGU . Free-surface turbulent flow around a lPg ship hull Adrian LUNGU ......................................................................................................................................... Uncertainties in the free-surface potential flow code solution Ana-Maria TOCU, Adrian LUNGU ......................................................................................................... Numerical flow investigation over a chemical tanker hull Mihaela AMORARITEI ............................................................................................................................ Performances data of propulsion Systems for high speed ships Dan OBREJA, Radoslav NABERGOJ, Liviu CRUDU, Sandita PACURARU ....................................... Simulation of the Ship Standard Manoeuvres Florin PACURARU, Dan OBREJA ........................................................................................................... Numerical and Experimental Investigation on a Tractor Tug Resistance Performance Corneliu BALAN, Diana BROBOANA, Catalin BALAN, Roland KADAR ........................................... On the stick slip boundary conditions at the wall of microchannels Corneliu BALAN, Diana BROBOANA .................................................................................................... Pressure constrain in vicinity of the separation point in planar, steady and isochoric motion: case inewtonian fluid

Florin BODE, Victor HODOR, Corina GIURGEA ................................................................................... Numerical investigation on a swirl burner with internal flue gas recirculation Victor HODOR, Florin BODE, Paula UNGURESAN, Claudiu RATIU .. CFD first prediction in designing a 50kw swirling burner within its combustion chamber Theodor POPESCU ................................................................................................................................... A foundation in distributions of Glauert theory

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Tom 52(66), Fascicola 3, 2007

3`nd Workshop on Vortex Dominated Flows Timisoara, Romania June 1 - 2, 2007

A SEMIANALITICAL METHOD FOR THE WING AERODYNAMICS


Corneliu BERBENTE, Prof.* Department of Aerospace Sciences E.Carafoli University Politehnica of Bucharest Sterian DNIL, Prof. Department of Aerospace Sciences E.Carafoli University Politehnica of Bucharest

Marius Stoia-Djeska, Assoc. Prof., Department of Aerospace Sciences E.Carafoli University Politehnica of Bucharest

*Corresponding author: Polizu 1, 011061, Bucharest, Romania Tel.: (+40) 21 023967, Fax: (+40) 21 3181007, Email: cberbente@yahoo.com ABSTRACT The Prandtl lifting line model is able to consider the main effects of the aerodynamic interaction between a parallel uniform flow and a wing of finite span with a sufficiently large aspect ratio. The differentialintegral equation for the vortex intensity can be reduced to an infinite system of linear equations for the coefficients of the Fourier series of circulation. However a collocation type method is not the most adequate one as it has to face chord or/and incidence jumps in several wing sections, starting with the wing ends. These jumps will cause instabilities altering the calculation results. In the following one presents a method able to avoid such jump instabilities by using a weighting integration over the wing span. KEYWORDS Lifting line, wing, aerodynamics, collocation, sail NOMENCLATURE An = coefficient in series expansion of circulation distribution = aspect ratio b = semi-wing span CD = induced drag coefficient CL = lift coefficient c = local wing chord y = spanwise coordinate on wing 1. INTRODUCTION Even today, when supercomputers and large computer codes are available, there is a need for simplified models that allow for an easy grasp of physical dominant effects. Lifting line theory is still useful because it provides a relatively simple method for the analysis of the aerodynamics of different wing shapes and allows a relatively quick comparison of the performances of different wing geometries. The primary results that may be obtained from the model are values of the induced drag coefficient and the lift-curve slope of a finite wing. The theory may be expanded to provide estimates of the change in the lift created by flaps and the rolling moment coefficient produced by an aileron deflection. 2. LIFTING LINE THEORY 2.1 The vortex system The lifting line model was essentially the replacement of the lifting wing by a model consisting of a system of vortices which imparted to the surrounding fluid a motion similar to the actual flow, and which sustained a force equivalent to the lift to be created. The vortex system can be divided into three main parts: the starting vortex; the trailing vortex system; and the bound vortex system. When a wing is accelerated from rest the circulation round it, and therefore the lift, is not produced instantaneously. Instead, at the instant of starting, at the sharp trailing edge the fluid is required to change direction suddenly while still moving at high speed. This high speed calls for extremely high accelerations and produces very large viscous forces, and the fluid is unable to turn round the trailing edge to the stagnation point (on the upper surface). Instead it leaves the surface and produces a vortex just above the trailing edge. The stagnation point moves towards the trailing edge, the circulation round the wing, and therefore its lift, increasing progressively as the stagnation point moves back. When the stagnation point reaches the trailing edge the fluid is no longer required to flow round the trailing edge. Instead it decelerates gradually along the aerofoil surface, comes to rest at the trailing edge, and then accelerates from rest in a

different direction. The vortex is left behind at the point reached by the wing when the stagnation point reached the trailing edge. Its reaction, the circulation round the wing, has become stabilized at the value necessary to place the stagnation point at the trailing edge The vortex which has been left behind is equal in strength and opposite in-sense to the circulation); round the wing and is called the starting vortex or initial eddy (Figure. 1)

the real physical wing in every way except that of thickness. This is the essence of finite wing theory. It is largely concerned with developing the equivalent bound vortex system which simulates accurately, at least a little distance away, all the properties, effects, disturbances, force systems, etc., due to the real aerofoil. The replacement bound vortex system must create the same disturbances, and this mathematical model must be sufficiently flexible to allow for the effects of the changed parameters. A real aerofoil produces a trailing vortex system. The hypothetical bound vortex must do the same.

Figure1. The initial eddy The pressure on the upper surface of a lifting wing is lower than that of the surrounding environment, while the pressure on the lower surface is greater than that on the upper surface, and may be greater than that of the surrounding environment. Thus, over the upper surface, the fluid will tend to flow inwards towards the root from the tips, being replaced by fluid which was originally outboard of the tips. Similarly, on the undersurface the fluid will either tend to flow inwards to a lesser extent, or may tend to flow outwards. Where these two streams combine at the trailing edge, the difference in spanwise velocity will cause the fluid to roll up into a number of small streamwise vortices, distributed along the whole span. These small vortices roll up into two large vortices just inboard of the wing-tips (Figure 2). The strength of each of these two vortices will equal the strength of the vortex replacing the wing itself.

Figure 3. Bound vortex system A consequence of the tendency to equalize the pressures acting on the top and bottom surfaces of an aerofoil is for the lift force per unit span to fall off towards the tips. The bound vortex system must produce the same grading of lift along the span. For complete equivalence, the bound vortex system should consist of a large number of spanwise vortex elements of differing spanwise lengths all turned backwards at each end to form a pair of the vortex elements in the trailing system. The varying span wise lengths accommodate the grading of the lift towards the wing-tips, the ends turned back produce the trailing system and the two physical attributes of a real wing are thus simulated (Figure 3). 2.2 Prandtl-Glauert formulation This model is an important example of using vortices for the representation of solid-fluid interaction by vortex singularities, a representation directly related to the viscous action within the boundary layer. Lets consider an incompressible flow around a wing of finite span as sketched in Figure 4.

Figure 2. Trailing vortices Both the starting vortex and the trailing system of vortices are physical entities which can be explored and discerned and seen if conditions are right. The bound vortex system, on the other hand, is a hypothetical arrangement of vortices which replace

S ( ) = ( 1) p p 2 Ap
1

has a slower convergence in terms of p2. In addition, at these points a chord discontinuity is present. In order to avoid the chord discontinuity at the wing ends, a first improvement was done by Carafoli, [1], by considering an approximate representation of the chord variation: Figure 4. The wing geometry The well known lifting line theory of Prandtl leads to the following differentialintegral equation:
1 ( y ) = kcV 4 V d dy d y y y b / 2
b/2

c0 sin = 0 + 22 cos 2 + 2 4 cos 4 c

(7 )

that smoothes the jump, but small modification in the wing planform appears. 3. PROPOSED FORMULATION Present improvement has several advantages: a) it conserves the wing real chord variation, leading to simpler formulas; b) gives suggestions to accelerate the convergence of the infinite system of equations for the coefficients A p, p = 1,2,; c) is easier to be adapted for any kind of calculations (including numerical calculation). A collocation method is equivalent to multiplying the equation (4) by a Dirac weight function centered on the collocation points, then integrating along the wing span. This approach has disadvantages at discontinuity points (in chord and/or angle of attack). There are many weight functions fitting the problem necessity. Will choose however functions of the same kind as appearing in the series (3). This corresponds to a Galerkin type selection what was proved to be very adequate in many other cases. By multiplying equation (4) with the weight functions sin n n = 1,2,3... that vanish at the wing ends, then integrating over the wing span, one gets a stable rapid convergent system of equations for An :
+ nH nn An = I n p (1 np ) H np Ap , 2 1
n = 1,2, K .

(1 )

Here denotes the angle of attack, c(y) the wing chord, V the upstream undisturbed velocity and ( y ) the circulation [2]. Defining by:
b y = cos , [0, ] , 2

(2)

Glauert assumed as a Fourier series:


= 2bV Ap sin n .
p =1

(3)

Now the problem consists in determination of the unknown coefficients Ap, p = 1,2,... . Introducing (3) in equation (1), one gets:

Ap sin p =
1

pAp sin p sin

(4 )

, = kc

2b

A certain number of points, called collocation points, are selected on the bounded vortex line (along the y axis). The equation is then imposed at each of these points and a set of linear algebraic equations, having Ap, p = 1,2,... as unknowns results. Under the form (4) the determination of the coefficients Ap, p = 1,2,3..., meets certain difficulties, both for the analytical and numerical calculations, due to instabilities at end stations = 0 and = . Thus, the term:
S ( ) =
1

(8)

I n = sin n ,
0

H np = H pn

sin n sin p . d = sin 0

(9)

pAp sin p sin

(5)

at wing extremities
S ( 0) = p 2 A p ,
1

(6)

np being the Kronecker symbol. The integrals In, Hnp, n, p = 1,2,3,...can be directly performed in many practical cases, for example for elliptical and polygonal wings. In case of symmetrical wing and incidence () one yields:

() = ( ) , () = ( ) ,

(10) (11) (12)

H np = 0 , for n + p = odd,
In = 0 ,

I 2 m 1 H 2 m 1,1 2 2m 1 + cos m , = = q 00 0 2m 1 2m( m 1)

(19)

for n = 2p = even.

m = 1,2,K
H 2 m 1 .2 p 1 = H 2 m 1, 2 p 3+ 4 0 ( 2 m 1) ( 2 m 1) 2 4 ( p 1) 2

A consequence of relations (10), (11), (12) is the vanishing of the coefficients An for even values of n:
A2 m = 0 , m = 1,2,3......

( m 1)(1 cos m cos p q 0 + ( m 1) 2 ( p 1) 2 m (1 + cos m cos p + m 2 ( p 1) 2

(20)

(13)

Then one needs only the system (8) for odd indices in the form [3]:
L2m1 A2m1 = I 2m1

(2 p 1)H
p=1, pm

2m1, 2 p1 A2 p1

(14)

H 2 m 1.2 p 1 = H 2 p 1, 2 m 1
where:
q = 1 ce , 0 = kc0 . 2b c0

(21)

where:
Ln =

(22)

nH nn

(15)

Taking into account that:


pm

Under assumption (subsequently confirmed) of a rapid convergence of series (3) of the circulation, we neglect in (14) terms A2 m + 3 with respect to A2 m 1 (four indices distance). On the other hand, the integrals H2m1,2p 1 are decreasing with the indices m, p (see relations (15)). For these reasons one may truncate the infinite sum in (10), starting with p = m + 2. Thus we can restrict ourselves to give only the coefficients A1 and A3:
A1 = I1L3 3I 3 H13 , A3 = 1 (I 3 H 31 A1 ) 2 L3 L1L3 3H13

lim

( m 1)(1 cos m cos p = 0 , p = 2,3,K ( m 1) 2 ( p 1) 2

(23)

For the first coefficient A1, one gets:


A1 = 00 =

(24) 30( 2 q ) + 16 0 (64 72q + 15q 2 ) 2 15 + 48 0 (7 5q) + 16 0 (64 72 q + 15q 2 )


2

(16)

In this way one can calculate wings of various shapes, where the presence of engines and fuselage can also be considered.
5. ASYMMETRIC WING NEAR WALL In the following one considers a wing not necessarily symmetric of span b/2 at a distance /2 with respect to an infinite vertical plan. This circumstance could appear in a wind tunnel with rectangular section and for a sail (Figure 5) fixed on a mast at a distance /2 from the deck of a vessel (practically such rigid metal sails are used to maneuver big vessels near the shore). In Figure 5 a hybrid combination between a soft sail made of linen and a solid sail, provided with a slat, rigid connected to the mast and. The angle of attack 0 is measured with respect to the zero lift axis, whereas the geometrical angle g is reported to the x axis. The positioning angles of slat and the soft sail are v and , respectively.

For a better accuracy, one can start with three coefficients: A1, A3, A5, using three equations corresponding to m =1, 2, 3, respectively. Knowing the coefficients A1, A3,, the aerodynamic forces can be evaluated. The lift coefficient results:
C z = A1 , =
b2 , S

(17)

where S and are the reference wing surface and the wing aspect ratio respectively. The coefficient of the induced drag , Cx , is:
C xi =
2 Cz

. n n z 1 + 3 3 A2 A2
n =1 1

A2

C2

A2
1

(18)

4. TRAPEZOIDAL WING For trapezoidal wings (Figure 4) the integrals above introduced can be evaluated in closed form:

I3 2 I 3r = (1 + sin 3 ) r 1 + 2 sin 2 cos 2 00 3

the index r standing for reported. The other integrals are expressed as functions of I1r and I2r, as follows:
L1 =

+ 0 I 1r , L3 = + 30H33r . 2 2

(28)

6 sin 2 5 sin H 33r = I 3r + 2 5 5 4 sin 2 3 r sin 2 3 3

(29)

Now one can calculate the first three coefficients of circulation:


2 A1 I L 3 0 I 3 r A1r = 1r 3 2 2 0 0 L1 L3 3 0 I 3r

,
(30)

Figure 5. The rectangular hybrid wing (VP) and the trapezoidal hybrid wing (VT) The system wingsolid infinite wall is equivalent from aerodynamic point of view with a wing in a mirror. As a consequence, the resulting motion takes place around a wing having a double span b. At y = / 2 one have a new jump in the chord distribution. Once more the proposed method is recommended. By denoting with the angle, defined by:
sin = , = arcsin 0, b b 2

A2 = 0 ,
A3 I A3r = 3r (1 0 A1r ) 0 0 L3

(25)

Integrals along the wing span ( [0, ]) , are calculated by extracting the portion ( /2, /2 ) out of a complete wing with chord c0 at y = 0. We have calculated the required integrals for a trapezoidal wing, at constnt incidence, 0. Then one gets as particular cases the rectangular and the triangular wings. The simplest case corresponds to elliptical wing. For the symmetrical wing the chord parameter variation, , defined by relation (4):
kc c kc = = 0 (1 r cos ) , 0 = 0 , q = 1 e . 2b 2b c0

We have taken in consideration four cases: a) the square hybrid wing, both with big slat (VP1a) and small slat (VP2a); b) the trapezoidal hybrid wing, with big slat (VT1a) and small slat (VT2a). Comparisons of the analytical calculations with experiments are given in Tables 1 and 2. As a conclusion, the lifting line theory is applicable for angle of attack 0 corresponding to interval (CZmin, CZmax). The angle 0 is measured with respect to the null lift axis of the hybrid wing system and depends on the curvature of the profile profiled mastsoft sailslat. Although this dependence is complicate, one can assume an almost constant slope of the curve Cz() for a fixed configuration.
6. CONCLUSIONS The good agreement between theory and experimental data for the slope dCZ/d0, the four cases presented in Table 1, especially for larger values of the aspect ratio , suggest a linear variation of CZ with 0 - the angle of the relative speed V. The rigid mast was profiled as a wing. Therefore the assumption of a almost unmodified direction of the null lift axis direction is confirmed. In exchange, one may see a significant displacement

(26)

Thus, one obtains for the integrals I2m-1, defined by relation (14):
I1 I 1r = 2(1 sin ) r cos 2 , 0 0

(27)

of null lift axis with wing planform, but not with the slat chord. As well, the angle 0 at Czmax keeps a value around 30 degree (Table 2). As regards the drag coefficient,Cx, at Czmax, more than a half (approx. 60%) is induced by the free vortex sheet, the rest belonging to friction effects. The two times increases of the angle 0max in comparison to usual wings can be explained through the slat effect as well by flexible wing adaptation to flow (Table 2).

The proposed method for solving the Prandtl equation, has the following advantages: it takes into account without problems the sudden chord variations or jumps (four jumps in case, of the wing in mirror); permits an exact calculations of the integrals leading to simple analytical expressions; a convenient arrangement of the system of equations giving the coefficients A1 , A3 , eventually A5 (A2 = 0 , A4 = 0) with a good accuracy.

Table 1. Comparisons analytical model with experimental data for the slope of lifting coefficient Model VP1.a. VP2a VT1.a VT2a r 0. 0. 0.7568 0.8000 0 0.4298 0.4030 0.4670 0.4418 3.689 3.920 5.834 6.491 A1r 0.7078 0.7208 0.5027 0.4941 A3r 0.1164 0.1235 0.0276 0.0185
-1 dCZ / d 0 (deg )

theory 0.0614 0.0627 0.0751 0.0769

exper.[3,4] (min error) 0,0682 0.682 0,0789 0.0769 (9.94%) (8.09%) (4.82%) (0.0%)

Table 2. Comparisons analytical model with experimental data for Cx for Czmax Cx for Czmax Model VP1.a. VP2.a VT1.a VT2.a (deg) Cz = 0 -23,5 -23.0 -12.7 -12.0 Czmax 7.0 8.0 18.0 17.0 (deg) 30.5 31.0 30.7 29.0 Czmax 2.05 2.25 2.25 2.50 imax (deg) 10.41 10.77 7.05 7.10 theory (indus) 0,372 0.423 0,277 0.310 experiment[3,4] 0,650 0.750 0,500 0.500 (57.2%) (56.4%) (55.4%) (62.0%)

REFERENCES 1. Carafoli, E., (1945), Theorie des ailes monoplanes denvergure finie, Analele Acad.Romne, Bucureti. 2. Carafoli E., Constantinescu V. N., (1981), Dinamica fluidelor incompresibile, Ed.Acad. R.S.R. Bucureti. 3. Berbente, C., (1973), Asupra unei metode pentru calcul aerodinamic al aripilor de anvergur finit n regim incompresibil, Bul. Inst. Politehnic Bucureti, XXXV, pp. 37-49. C., (2005) Experimental 4. Maraloi, Researches for Determining the Propulsion Parameters of the Hybrid-sail Ships, Revue

Roumaine des Sciences Techniques, Srie de Mcanique Applique, 50. 5. Berbente C., Maraloi C. (2006), Theoretical and Experimental Study Regarding the Aerodynamics of a Ship Sail System, U.P.B. Scientific Bulletin, Series D, 68, pp.17 - 32

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Tom 52(66), Fascicola 3, 2007

3`nd Workshop on Vortex Dominated Flows Timisoara, Romania June 1 - 2, 2007

A PHYSICAL AND THEORETICAL ANALYSIS OF THE VORTEX BREAKDOWN ON DELTA WINGS


Marius STOIA-DJESKA, Assoc. Prof.* Department of Aerospace Sciences Elie Carafoli POLITEHNICA University from Bucharest Sterian DANAILA, Prof. Department of Aerospace Sciences Elie Carafoli POLITEHNICA University from Bucharest

Corneliu BERBENTE, Prof., Department of Aerospace Sciences Elie Carafoli POLITEHNICA University from Bucharest

*Corresponding author: Polizu 1-6, 011061, Bucharest, Romania Tel.: (+40) 21 4023967, Fax: (+40) 21 3181007, E-mail: marius.stoia@rosa.ro wings. This non-linear effect was and is intensively exploited in the design of modern civil and combat aircrafts [1, 4, 5]. The leading edge vortex occurs when the flow encounters a sufficiently highly swept sharp leading edge and can be explained within the framework of inviscid flow theory by the Kutta condition. Increasing the angle of attack of the wing leads to an increase of the strength of the leading edge vortex. This is in fact a well-ordered vortical flow structure, steady and stable. Depending on the Mach number and geometrical parameters there is a critical angle of attack at which the internal structure of the vortex changes substantially and the vortex breakdown occurs. Vortex breakdown means a sudden increase of the cross section of the vortex core and the loss of the regularity of the flow. Further, vortex breakdown also marks the end of the favorable effects induced by the leading edge vortex. The phenomenon of vortex breakdown has been observed in wind tunnel experiments and real flights and can be investigated using computational simulations as well as theoretical approaches. This paper concerns with the analysis of the stability of a narrow vortex core. First, some experimental results are presented in order to explain the vortex break down phenomenon. Further the Euler equations for axi-symmetric flows are simplified using a number of assumptions based on the slenderness of the vortex core and on the conical flow structure inside the vortex. The resulting equations subjected to rational boundary conditions can be solved in closed form and the solutions obtained show two different features, resembling the jet-like stable region and the wake-

ABSTRACT The paper concerns with the vortex breakdown phenomenon. First, some experimental results are presented in order to explain the vortex break down phenomenon. Further the Euler equations for axially symmetric flows are simplified using a number of assumptions based on the slenderness of the vortex core and on the conical flow structure inside the vortex. The resulting equations can be solved in closed form and the solutions obtained show two different features, resembling the jet-like stable region and the wake-like after breakdown region of the flow. These features are discussed in the paper. KEYWORDS Vortex breakdown, vortex flow on delta wings NOMENCLATURE vr [m/s] radial velocity

v u M

[m/s] [m/s] [deg] [-]

tangential velocity axial velocity angle of attack the Mach number

Subscripts and Superscripts e external conditions 1. INTRODUCTION In aircraft aerodynamics and not only vortex flows play an important role. The low pressure induced on the upper side of highly swept wings by the leading edge vortices results in an important increase of the lift capabilities of small aspect ratio

like after breakdown region of the flow. These features are discussed in the paper.

2. FUNDAMENTALS OF DELTA WING AERODYNAMICS AND VORTEX BREAKDOWN The subsonic flow over a sharp edged delta wing at medium and large angles of attack is characterized by leading-edge separation. The dominant aspect of the flow is the vortical structure that forms on the upper side of wings with leading edge sweep angles greater than 45 deg. The flow separates when encountering the sharp leading edge

and forms a free shear layer which contains vorticity. Due to this embedded vorticity the free shear layer rolls up in a spiral fashion and the result is a steady primary vortex. The vortex has a viscous core but the flow outside the core is essentially inviscid. The vorticity from the entire length of the leading edge is convected through the free shear layer to the vortex core which increases in strength and in cross-section in the downstream direction. There is also a second vortex sheet lying above the primary vortex. This secondary vortex is due to the flow separation due to strong adverse pressure gradients in the spanwise direction.

a) = 10deg.

b) = 18 deg.

c) Vortex breakdown is imminent at = 18.3 deg. ,

d) Starboard vortex breakdown at = 18.5 deg.

Figure 1. The oil flow visualization of vortex breakdown at M = 0.85, LE = 650

Wind tunnel studies and numerical simulations have proven that [2, 3, 4, 5]: The presence of the stable primary vortex strongly affects the pressure and velocity distributions on the wing upper side, the predominant and beneficial effect being a low-pressure region under the vortex core.

This suction region gives the so called vortex lift effect that represents an essential feature of highly swept delta wings; As the angle of attack increases over a limit value a large scale vortex breakdown occurs

above the wing and the flow becomes fully unsteady. Figure 1 shows some results of the oil flow visualization study of vortex breakdown on delta wings at M = 0.85, LE = 650 obtained at High Speed laboratory, TU Delft. The white region represents approximately the low-pressure area and also delineates the position of the primary vortex. The vortex burst or breakdown mechanism is one of the unresolved problems in fluid mechanics. It means a sudden increase of the cross section of the vortex core and the loss of the regularity of the flow. Further, bursting involves a sudden decrease in the magnitude of the axial and circumferential velocities of the core. Detailed measurements show that the axial component of the velocity is increasing up to the breakdown point location (axial velocities more than three times the freestream velocity were measured) where it decreases abruptly. Upstream the bursting point the core has a jet-like structure while downstream the core has a wake-like structure that expands in the radial direction. It was discovered that the vorticity vector is first oriented in the axial direction and after the vortex breakdown reorients itself along a transverse direction [4, 2]. At lower angles of attack the primary vortex burst point is downstream the trailing edge and does not affect the vortex lift. The interest is now to control the flow so that the bursting point to remain downstream the trailing edge for medium and high alpha. It is nowadays known that the breakdown can be postponed by axial air blowing or, in other words by increasing the jet like axial velocity. Unfortunately, for aeronautical applications this approach is inefficient. We consider therefore that a better understanding of the vortex breakdown mechanism represents the first step towards new solutions for the flow control.
3. MATHEMATICAL MODEL AND CLOSED FORM SOLUTIONS 3.1. Simplified equations of the vortical flow It is now clear that the vortex breakdown mechanism is inviscid and thus the incompressible Euler equations can be used as the starting point for searching for rational but simplified flow models. Further, there are at least two completely different states of the vortex core that must be identified. In what follows we assume that the vortex core is isolated and its cross-sectional dimension is small relative to typical length scale along the axis of

vortex core. This assumption we made means that the vortex core is slender and slowly varying in axial direction. For incompressible inviscid flow the Euler equations in cylindrical coordinates are:
u vr 1 1 v + + vr + =0 x r r r u u 1 u 1 p + vr + + =0 u x r r x (1) vr vr 1 vr 1 2 1 p + vr + u + =0 u x r r r r u u 1 v 1 1 p u + vr + + u r u + =0 r r x r r

Under the assumptions we made, in the slender, narrow region with rotational axially-symmetric flow the equations reduce to:
vr u 1 + + vr = 0 x r r u u 1 p + vr + = 0 u x x r vr vr 1 1 p + vr u 2 + = u r x r r u u 1 u + vr + u r u = 0 r x r

( 2)

The unknowns of these equations are the velocity components and the pressure. These can be expressed as functions of x and the dimensionless radius r = r R ( x ) where R ( x ) denotes the radius of the cross-section of the core for x=const [3].. Then the pressure partial derivatives become:
p 1 p dp 1 dR p p = , = r + r x R r dx R dx r

(3)

and the same rule applies for the velocity components. Introducing these partial derivatives in (2) one obtain a new system of partial differential equations that contains product terms like

u u p or R . For conical flows inside , Ru x x x

the quasi-conical vortex core the velocity and the pressure do not depend on the axial direction. For non-conical flows but for slender and narrow vortex cores terms like those above can be neglected. If we do so then the system (2) reduces to a system of ordinary differential equations with the only independent variable r . The solution of the equations gives a qualitative insight in what happens inside the vortex core.

3.2. Boundary conditions At the axis of the core the radial velocity vanishes, i.e. vr = 0 for r = 0. At the exterior boundary of the vortex core the axial and circumferential components of the velocity and the pressure match continuously the outer flow values, i.e. u = U e ( x ) , v = Ve ( x ) , p = Pe ( x ) for r = 1. 3.3. Closed form solution The boundary conditions we assumed lead to the following closed solution

u = U e (1 V H ( r ) ) vr = V U e G ( r ) V2 v = SU e 1 H (r ) + H 2 (r ) 2 S p pe V V = V 1 + H ( r ) I ( r 2 2 2 U e

velocity and vorticity vectors are pointing in opposite directions. The streamline starts at infinity downstream close to the axis and runs upstream with an increasing spiral radius. When the axial component of the velocity becomes positive the spiraling streamline starts to run downstream but with a sudden increase of the spiral radius. This is the wake-like part of the solution and this can be observed on the rear part of the wing. We consider that the vortex breakdown mechanism is in fact responsible for the switch in the solution state.
4. CONCLUSIONS We have examined the physical behavior of the vortex flow on a delta wing with sharp and highly swept leading edge. Further we derived an analytical solution for an isolated slender vortex core. The solution has two branches. The jet-like solution develops in the upstream of the bursting point while the wake-like solution develops downstream this point. However, the mechanism responsible for the switch in the flow state from one to the other is still unclear. ACKNOWLEDGMENTS The present work has been supported from the National University Research Council Grant (CNCSIS) 33/2004. Oil flow visualization of vortex breakdown on delta wings at M = 0.85, LE = 650 have been obtained from High Speed laboratory, Technical University Delft. REFERENCES 1. Kuchemann, D., (1978) The Aerodynamic Design of Aircraft, Pergamon Press 2. Ekaterinas, J.A., Sciff, L.B. (1990) Numerical Simulation of the Effects of variation of Angle of Attack and Sweep Angle on Vortex breakdown over delta Wings, AIAA -90-3000-CP 3. Hall, M.G. (1961) A Theory for the Core of a Leading Edge Vortex, JFM, Vol. 11, pp. 209-227 4. Lamar, J.E. (2002) Understanding and Modeling Vortical Flows to Improve the Technology Readiness Level for Military Aircraft terms of reference, NATO RTO AVT-E-10/RTG 5. Rom, J. (1992) High Angle of Atack aerodynamics, Springer Verlag, New York

( 1)

(4) where we denoted the functions:

G (r ) =

(1 + r
(

1 2

1
,
1 + R 2 r

r R

H ( r ) = ln 1 + r 2 R2 I (r ) = H (r )
2

1 2

) )

1 2

1 1

1 + r 2 R 2

1 2

r 2 R 2

1 + R 2

1 2

R 2

and
V =W R'2 2 1 +
1 1 2 '2 2 1+ R 1 2 S 1 (5) 1 + 4 '2 R 1

1 R'2 2

The swirl ratio of the flow outside the vortex core is denoted with S = Ve U e and the parameter W = 1 . The components of the vorticity vector can now be derived analytically too and is left to the reader. The behavior of the solution (4) depends on the values of the parameter W. For W = 1 the velocity and vorticity vectors point in the same direction. A streamline coming from upstream enters the vortex core and continues downstream spiraling around the core axis with a slight reduction of the spiral radius. This is the jet-like part of the solution and it appears in the forward part of the wing. For W = 1 the

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Tom 52(66), Fascicola 3, 2007

3`nd Workshop on Vortex Dominated Flows Timisoara, Romania June 1 - 2, 2007

CAVITATION TIP CLEARANCE. NUMERICAL SIMULATION AND EXPERIMENTAL RESULTS


Constantin Viorel CAMPIAN, Prof.* Faculty of Engineering Eftimie Murgu University of Resita Dorian NEDELCU, Assoc. Prof. Faculty of Engineering Eftimie Murgu University of Resita

*Corresponding author: P-ta Traian Vuia, No. 1-4, 320085, Resita, Romania Tel.: (+40) 256 219134, Fax: (+40) 256 219134, Email: v.campian@uem.ro, d.nedelcu@uem.ro ABSTRACT The paper presents a flow numerical simulation and experimental results regarding cavitation tip clearance for Kaplan turbines. The analysed domain correspond to the gap between runner blades and runner chamber. Experimental results are obtained on the model and prototype. KEYWORDS Kaplan turbine, cavitation tip clearance, vortex flow 1. INTRODUCTION Improvements in hydrodinamics design of hydraulic turbine must generate a better control of cavitation phenomenon. In this paper cavitation tip clearance is analyzed. Tip clearance cavitation appears inside the gap between rotating blade and the runner chamber. The flow through the clearance is the results of the intrados/extrados pressure difference. When the boundary layer separation occurs, the tip clearance cavitation appears. Supplementary, a vortex attached to the blade tip is generated in the channels between the blades, which is the source for the vortex cavitation. Numerical simulation of the clearance water flow was made by Navier-Stokes method. The runner blades were manufactured with anti cavitation lips, in two geometrical solutions. Experimental investigations were performed on the model in the laboratory and on the prototype in power station.

Figure 1. The 3D model of the blade without anti cavitation lip

Figure 2. The 3D model of the blade with anti cavitation lip

2. NUMERICAL SIMULATION RESULTS 2.1. Geometrical model The 3D blade model without and with anti cavitation lip is done in the figure 1 and 2.

For anti cavitation lip were analyzed two geometrical solutions, where the second solution correspond to a shorter length of the anti cavitation lip comparative with first solution, beginning from the leading edge.

2.2. Numerical simulation The numerical simulation were performed with Navier-Stokes method [1] for blade without and with anti cavitation lip in two solutions. The numerical results are graphically presented in figures no. 3 to 6 for solution 1 and 7 for solution 2. Figure no. 8 presents comparative results between the two solutions. The pressure distributions are calculated at the same head for suction side of the blade where the anti cavitation lip is placed. Figure 6 Pressure distribution on the blade with anti cavitation lip solution 1 cavitation bubles near blade surface

Figure 3 Pressure distribution on the blade without anti cavitation lip top view Figure 7 Pressure distribution on the blade with anti cavitation lip solution 2 top view

Figure 4 Pressure distribution on the blade without anti cavitation lip front view

Figure 8 Comparative pressure distribution on the blades with anti cavitation lips The area close to the anti cavitation lip is the most important area for this study. Figure 8 shows that solution 2 for anti cavitation lip (with shorter length) is better than solution 1, because the cavitation pitting area will be smaller.

Figure 5 Pressure distribution on the blade with anti cavitation lip solution 1 top view

3. EXPERIMENTAL RESULTS 3.1. Experimental results on the model Experimental results were performed on the Iron Gates I turbine model [2]. The model has been installed on the test rig; the tests were performed, at the same operating conditions, for 3 blade configurations: blades without anti cavitation lips, figure 9, blades with long (solution 1) and short (solution 2) anti cavitation lips, figure 10 and 11.

The efficiency and the cavitation picture on the model blade are the same without and with anti cavitation lips (solution 1 and 2). 3.2. Experimental results on the prototype The experimental observations on the prototype runner blade were made after 8000 hours of operation. In the figures 12 and 13 are presented pictures with cavitation pitting areas and the long anticavition lip (solution 1). On the blade near the anti cavitation lip was observed a slightly cavitation erosion area, which is not visible on the figures.

Figure 9 Blade without anti cavitation lip

Figure 12 Cavitation pitting on the anti cavitation lip solutia 1

Figure 10 Blade with anti cavitation lip solution 1

Figure 13 Zoom on the cavitation pitting on the anti cavitation lip solutia 1 In the figure 14 is presented a picture with cavitation pitting areas and the short anticavition lip (solution 2). On the blades equiped with this solution of anticavitation lips, an area with depth cavitation pitting near the anticavitation lip appears.

Figure 11 Blade with anti cavitation lip solution 2

Figure 14 The cavitation pitting on blade and the anti cavitation lip solutia 2

4. CONCLUSIONS There are no significant cavitation differences between the blades with long, short and without anti cavitation lip, at the model tests. On the prototype, solution 2 of the anticavitation lip induce on the lip and blade a significance cavitation erosion, comparative with solution 1. The modified geometry of the solution 2 (short length) determine the peripheral vortex attachment, which generate the cavitation erosion on the blade. This cavitation behaviour was not possible to be observed on the model. The tip clearance cavitation analysed in the study can not be numerical reproduced with accuracy. The very small gap (5 mm) between the runner blade and runner chamber compared with the runner diameter (9500 mm) and the size blade compared with the small dimensions of the anticavitation lip limit the approximation of the phenomen by numerical simulation.

REFERENCES 1. VA TECH CFD Cavitation Research, Result Presentation in Zurich, April, 2002. 2. VA TECH PdF1 Report of the Task Force Cavitation, June, 2003

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Tom 52(66), Fascicola 3, 2007

3rd Workshop on Vortex Dominated Flows Bucharest, Romania June 1 - 2, 2007

NUMERICAL ANALYSIS OF PUMP HYDRODYNAMICS AT CONSTANT SPEED


Adrian STUPARU, Assist. Prof.* Department of Hydraulic Machinery Politehnica University of Timisoara Daniel BALINT, Assist. Prof. Department of Hydraulic Machinery Politehnica University of Timisoara Sebastian MUNTEAN, Senior Researcher Center of Advanced Research in Engineering Sciences Romanian Academy - Timisoara Branch Liviu ANTON, Prof. Department of Hydraulic Machinery Politehnica University of Timisoara

Alexandru BAYA, Prof. Department of Hydraulic Machinery Politehnica University of Timisoara

*Corresponding author: Bv Mihai Viteazu 24, 300223, Timisoara, Romania Tel.: (+40) 256 403692, Fax: (+40) 256 403700, Email: astuparu@mh.mec.upt.ro validated assumptions, one can develop a ABSTRACT methodology for computing the turbomachinery flow, so that very good and engineering useful This paper presents the numerical investigation of results are obtained [5]. the 3D flow in the impeller and the suction-elbow However, computing the real flow of a storage pump by using commercial code (unsteady and turbulent) through the whole storage FLUENT 6.0. First the investigated centrifugal pump requires large computer memory and CPU pump is described, then the equations that governs time even for our days computers. As a result, a the flow and the boundary conditions imposed and simplified simulation technique must be employed in the end the results of the flow simulation. to obtain useful results for pump analysis, using currently available computing resources. KEYWORDS storage pump, numerical investigation, turbulent flow NOMENCLATURE 2. COMPUTATIONAL DOMAIN, EQUATIONS AND BOUNDARY CONDITIONS The investigated storage pump has two stages with the impellers situated in opposition and a suctionelbow of complex geometry, Figure 1. Each impeller has five blades, Figure 2.

cp =
g

p p IN [-] gH
gravity

pressure coefficient [m/s2] [m2/s2]

k turbulent kinetic energy 1.

INTRODUCTION The progress in the field of Computational Fluid Dynamics (CFD) has made this technology an important tool in analysis and design of hydraulic turbomachinery. The turbomachinery flow is essentially unsteady due to the rotor-stator interaction. On the other hand, rigorously speaking, the geometrical periodicity of the rotor blade channels cannot be used since there are differences in the flow from one inter-blade channel to another. However, with carefully chosen and experimentally

Figure 1. Cross section through the storage pump

Figure 2. Isometric view of the storage pump impeller The computational domain includes only the impeller of the centrifugal pump. For the numerical investigation only one inter-blade channel is used because of the symmetry of the geometry. Figure 3 shows the 3D computational domain with boundary conditions corresponding to an inter-blade channel of the impeller. The computational domain is bounded upstream by an annular section (wrapped on the same annular surface as the suction outlet section, but different in angular extension) and extends downstream up to cylindrical patch, in order to impose the boundary conditions on the outlet section. Figure 4. Mesh generated on the 3D computational domain of the inter-blade channel For the flow analysis presented in this paper we consider a 3D turbulent flow model. A steady relative 3D flow is computed,
V = 0 d ( V ) = g p + V dt

(1a)

(1b)

The numerical solution of flow equations (1a) and (1b) is obtained with the expert code FLUENT 6.0, [3], using a Reynolds-averaged Navier-Stokes (RANS) solver. As a result, the viscosity coefficient is written as a sum of molecular viscosity and turbulent viscosity T, and the last term in the righthand-side of (1b) becomes [( + T )V ] . For the first case, when only the impeller domain is considered, we solve a relative flow, in a rotating frame of reference with angular speed = k ( k being the unit vector of the pump axis direction). By introducing the relative velocity

Figure 3. Isometric view of the inter-blade channel of the impeller The inter-blade channel domain is discretized with 500k cells using a structured mesh coupled with a special boundary layer discretization on the blade faces, see Figure 4.

W =V r

(2)

with r the position vector, the left hand side of (1b) becomes

(W ) + (WW ) + 2 W + t + ( r ) + r t

(3)

An important assumption used in the present computation is that the relative flow is steady. This

simplifies (3) by removing the first and last terms, and also allows the computation of impeller flow on a single inter-blade channel. The turbulent viscosity is computed using the RNG model. On the inlet surface of the impeller a constant velocity field was imposed normal on the surface. The velocity magnitude is computed using the flow rate of the operating point:

vI =

Q SI

(4)

head Hydraulic power Rotational speed Flow rate Pumping head Hydraulic power

Pu=gQH[k W] n [rot/min] Q [m /s] H [m] Pu=gQH[k W]


3

3294.9 8 1500 0.8 350 3432.2 7

For the outlet section the outflow condition is imposed. That means both flow and turbulent quantities remain unchanged downstream to the outlet section. On the periodic surfaces of the impeller the periodicity of the velocity, pressure and turbulence parameters were imposed:

3. NUMERICAL RESULTS The pumping head is determined using the following equation:

r r 2 v (r , , z ) = v r, + n , z b , 2 p(r , , z ) = p , K r, + n , z b

p v2 r r (v n )dS + z + outlet g 2 g H= r r ( ) v n dS
outlet

(5)

p v2 r r z + + inlet (v n )dS g g 2 r r (v n )dS


inlet

(6)

We investigated five operating points of the pump with the characteristics given in the table 1: Table 1: Operating points of the centrifugal pump Operating Parameter Symbol Value point Rotational n [rot/min] 1500 speed 3 Flow rate Q [m /s] 1 1 Pumping H [m] 319 head Hydraulic Pu=gQH[k 3128.2 power W] 73 Rotational n [rot/min] 1500 speed Flow rate Q [m3/s] 1.1 2 Pumping H [m] 300 head Hydraulic Pu=gQH[k 2941.9 power W] 5 Rotational n [rot/min] 1500 speed 3 Flow rate Q [m /s] 1.2 3 Pumping H [m] 265.45 head 2603.1 Hydraulic Pu=gQH[k W] 35 power Rotational n [rot/min] 4 1500 speed 3 Flow rate Q [m /s] 0.9 Pumping H [m] 336

And the error is calculated with the equation (7):

H calculat H cata log H cata log

100

(7)

The results of the numerical simulation compared with the ones given by the pump producer are presented in table 2 and figure 5: Table 2: Pumping head values resulted from the numerical simulation and given by the pump producer HFLUENT Hcatalog Q [m] [m] [%] [m3/s] 0.8 184.59 175 5.48 0.9 177.7 168 5.46 1 167.62 159.5 5.09 1.1 155.42 150 3.61 1.2 143.57 132.725 8.17

300 H(FLUENT) H(catalog) 250

100

90

200 80
eta [%] H [m]

150

eta(FLUENT) eta(catalog) 70

100 60

50

0 0.8

0.85

0.9

0.95

1.05 1.1 Q [m3/s]

1.15

1.2

1.25

1.3

50 0.8

0.85

0.9

0.95

1.05 1.1 Q [m3/s]

1.15

1.2

1.25

1.3

Figure 5. Pumping head vs. flow rate for the investigated operating points The efficiency of the impeller is determined using the equation:

Figure 6. Efficiency vs. flow rate for the investigated operating points

( = (

outlet

r vu v ndS

r r

inlet

M r r r vu v ndS M

The pressure coefficient is calculated with the following relation:


cp = p p IN gH
(9

(8)

The results of the numerical simulation compared with the ones given by the pump producer are presented in table 3 and figure 6: Table 3: Efficiency values resulted from the numerical simulation and given by the pump producer Q HFLUENT catalog FLUENT [m3/s] [m] [%] [%] 0.8 184.59 83 88.75 0.9 177.7 86 91.12 1 167.62 87 90.65 1.1 155.42 85 92.30 1.2 143.57 80 93.08

The distribution of the pressure coefficient along the pressure side and suction side of the blade for all five operating points are presented in figure 7 to 11

Figure 7. Pressure coefficient distribution along the blade for Q = 0.8 m3/s

Figure 8. Pressure coefficient distribution along the blade for Q = 0.9 m3/s

Figure 9. Pressure coefficient distribution along the blade for Q = 1.0 m3/s

Figure 10. Pressure coefficient distribution along the blade for Q = 1.1 m3/s

Figure 11. Pressure coefficient distribution along the blade for Q = 1.2 m3/s From figure 7 to 11 one can observe that the minimum value of the pressure coefficient is located on the leading ege and this value is decreasing while the flow rate goes up. The NPSHr (Net Positive Suction Head required) is defined as: In figure 12 the NPSHr distribution on the impeller blades is plotted for the five investigated operating points. It is well known, the maximum risk of cavitation occurs where the largest NPSHr value (corresponding to minimum pressure) is obtained. Due to the sharpness of the leading edge, the maximum NPSHr value appears at the junction between blade suction side and the crown near to the leading edge (the red spots in Figure 12). The value of the NPSHr rises with the rise of the flow rate and so the risk of cavitation is higher.

p V2 p NPSH r = i + i g 2 g g

(9)

where the variables with subscript (i) correspond to the averaged one on inlet section of the impeller.

Figure 12 NPSHr on the blades of the impeller for Q = 0.8 1.2 m3/s

4.

CONCLUSIONS The paper presents a numerical study of the 3D flow in the impeller of a storage pump. Comparison between numerical results and data given by the pump producer is presented. A particular attention is paid to the analysis of the flow on the impeller blade. The numerical simulation predicts a minimum pressure on the suction side near the crown, in agreement with experimental observations. The cavitation risk for the bigger flow rates is higher than for the smaller flow rates.

ACKNOWLEDGMENTS The authors acknowledge the support from the National University Research Council grants. All numerical computations have been performed at the Numerical Simulation and Parallel Computing Laboratory of the Politehnica University of Timisoara, National Centre for Engineering of Systems with Complex Fluids.. REFERENCES 1. Akhras, A., El Hajem, M., Morel, R., Champagne, J.Y., The Internal Flow Investigation of a Centrifugal Pump, Proceeding of Flucome, Sherbrooke, Canada, 2000 2. Fluent Inc., FLUENT 6. Users Guide, Fluent Incorporated, 2001. 3. Fluent Inc., Gambit 2. Users Guide, Fluent Incorporated, 2001.

4. Hofmann, M., Stoffel, B., Experimental And Numerical Studies On A Centrifugal Pump With 2d-Curved Blades In Cavitating Condition, Proceeding of Fourth International Symposium on Cavitation CAV2001, Pasadena, California, SUA 5. Muntean, S., Numerical methods for the analysis of the 3D flow in Francis turbine runners, PhD Thesis Politehnica University of Timisoara, Timisoara, 2002. 6. Sallaberger, M., Sebestyen, A., Mannschreck, E., Pinkas, W., Modern Pump Impeller Design With Consideration Of Non-Uniform Inlet Flow Fields, Proceedings of IAHR, Graz, 1999 7. Tamm, A., Ludwig, G., Stoffel, B., Numerical, Experimental And Theoretical Analysis Of The Individual Efficiencies of a Centrifugal Pump, Proceedings of ASME FEDSM01 2001 ASME Fluids Engineering Division Summer Meeting New Orleans, Louisiana, May 29 June 1, 2001 8. Tamm, A., Brten, A., Stoffel, B., Ludwig, G., Analysis Of A Standard Pump In Reverse Operation Using CFD, Hydraulic Machinery and Systems, 20th IAHR Symposyum, 2001. 9. Van Esch, B. P. M., Simulation of threedimensional unsteady flow in hydraulic pumps, Thesis University of Twente, Enschede, 1997.

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Tom 52(66), Fascicola 3, 2007

3`nd Workshop on Vortex Dominated Flows Timisoara, Romania June 1 - 2, 2007

SWIRLING FLOWS IN THE SUCTION SUMPS OF VERTICAL PUMPS. EXPERIMENTAL APPROACH

Eugen Constantin ISBASOIU, Prof* Department of Hydraulic Machinery Politehnica University of Bucharest Marius STOIA-DJESKA, Assist. Prof. Department of Aerospace Sciences Elie Carafoli Politehnica University of Bucharest Georgiana DUNCA, Assist. Department of Hydraulic Machinery Politehnica University of Bucharest

Petrisor STANESCU, MEng. S.C. AVERSA S.A. Bucuresti Carmen Anca SAFTA, Assist. Prof Department of Hydraulic Machinery Politehnica University of Bucharest Diana Maria BUCUR, Assist Department of Hydraulic Machinery Politehnica University of Bucharest Calin GHERGU, Prep. Department of Hydraulic Machinery Politehnica University of Bucharest

*Corresponding author: 313 Splaiul Independentei, 060024, Bucuresti, Romania Tel.: (+40) 21 4029 523, Fax: (+40) 21 4029 523, Email: ecisbasoiu@yahoo.com intake channel and the suction pump through the ABSTRACT sump. So the flow could not involve a large quantity of air, [1]. The experimental approach have the purpose to The swirling flow is influenced by the geometry determine the working conditions in axial and and the dimensions of the suction sump and, not at mixed flow pumps operation if the water level in the least, by the working conditions of the pump. the intake channel lowers. It will be obtained the The study on the physical model of the suction limit value of the suction head of installation pump conditions of a vertical pump deals with the inferior so as the pumps could have safety conditions of limit of the water level in the sump pump if the working. The physical model of the suction sumps level in the channel is fluctuating. Would be a of the pumps is described for two types of vertical swirling flow at the lowest water level in the suction pumps: NMV 1000RA and NMV 2000RA. sump or not? To respond at this question two types Experimental results combined with numerical of vertical pumps were proposed: NMV 1000RA simulations give us the behavior of the swirling and NMV 2000RA and their sumps suction. In [1] flows in a suction sump of a vertical pump. were mentioned the dimensions and the geometry of a suction sump for a vertical pump, the swirls types KEYWORDS and the protection elements to have a flow without Vertical pump, swirling flow, suction sump, vortexes. experimental results, numerical simulation, velocity distribution. 1. INTRODUCTION Suction sumps are hydro-mechanical equipments used in the pump stations to receive water flowing from the intake channel to the pumps suction aria. The pump functionality to the designed parameters is looking for a flow without swirls between the 2. PHYSICAL MODEL OF THE SUMP The flow study in the suction sump could be numerical [2], [3] or experimental using a hydraulic scale model [4], [5]. Tests for NMV 1000RA and NMV 2000RA pumps were made. The suction sump for NMV 1000RA pump was physical modeled on a scale of kl = 2.12, Figure 1 and a kl = 5 was used for NMV 2000RA, Figure 2. The scale

coefficients, kl, are the results of the hydraulic similitude, nq, for the particular case of equality between the speed triangle of the model pump and the real one. As a model pump was used a reference diameter of D = 400 mm.

the origin the water level in the suction sump, the pumping head is done as

p r V r2 H= + +a g 2 g

(1)

where pr is the pressure measured with the differential manometer placed on the out pipe of the pump, a is the difference of the level between the median manometer plane and the water level in the sump, Vr is the medium velocity of water in the discharge pipe of the pump. The flow rate, Q, was measured using the tighten section method by a diaphragm mounted on the discharge pipe of the pump. So:
Q= 2 p D 2 m 4

(2)

Figure 1. The suction sump of the NMV 1000RA pump (the dimensions of the prototype)

where D = 336 mm the discharge pipe of the pump; is the flow coefficient of the diaphragm; d m = with d = 269 mm the diaphragm D diameter; p is the pressure difference on the diaphragm; = 1000 kg/m3 water density at 200C temperature.
2

Figure 2. The suction sump of the NMV 2000RA pump (the dimensions of the prototype) 3. PARAMETERS MEASURED ON THE MODEL Classical methods were used to measure the working parameters, flow rate and pressure, of axial or diagonal pumps, Figure 3. The pumping head or the head of the pump, H, is defined as the difference between the discharge head of the pump, Hr, and the suction head of the pump, Ha. In the case of axial and mixed flow vertical pumps, if it is considered as

Figure 3 Measured parameters for axial and mixed flow pumps


4. TESTS RESULTS Five sets of measurements were done: two for the sump suction of NMV 1000RA pump and three for the sump of NMV 2000RA pump. The working measured parameters are in Table 1 with visual qualitative observations of the flow. During these tests the velocity components were measured in two perpendicular sections and directions of the suction sump. It could be seen in

Figure 1 and Figure 2 the measure sections numbered with 1, 2, 3 and 4. The velocities were measured with a Pitot-Prandtl Tube and a differential manometer. Displaying the velocity components any disturbance of the velocity distribution from the turbulent profile indicates the presence of a parasite flow induced by a swirl flow. Figure 4 and Figure 5 indicates no significant deviations from the turbulent velocity distribution in the sump of NMV 1000RA pump. No submerged swirl [1] was formed at the entrance of the impeller. h No. 1 2 3 4 5 The pump NMV 1000RA mm 372 333 333 333 850 740 850 850 pr MPa 0.44 0.25 0.25 0.25 1.9 1.9 1.4 1.4 Q m /s 0.42 0.62 0.62 0.04 0.628 0.586 0.629 0.625
3

Figure 6 and Figure 7 indicates a disturbance of the velocity components distribution and the swirling flows are imminent in these cases if the water level in the sump is changed (see Hsb in Table 1). The submerged head, Hsb, is the static pressure at the impeller entrance.

Table 1

H m 7.8 5.24 5.24 5.24 24.31 22.58 17.87 17.87

Hsb m 2.716 1.946 1.194 0.746 1.546 0.646 0.446 1.996

Observations without vortex without vortex with vortex with vortex without vortex with vortex with vortex without vortex

NMV 2000RA

Figure 4. Velocity distribution on horizontal axis (1 and 2 measure section) and vertical axis (3 and 4 measure section), NMV 1000RA, Hsb = 2.716 m, without vortexes.

Figure 5. Velocity distribution on horizontal axis and vertical axis, NMV 1000RA, Hsb = 2.716 m, without vortexes.

Figure 6. Velocity distribution on horizontal axis (1 and 2 measure section) and vertical axis (3 and 4 measure section), NMV 2000RA, Hsb = 0.646 m, with vortexes.

Figure 7. Velocity distribution on horizontal axis and vertical axis, NMV 2000RA, Hsb = 0.646 m, with vortexes.

A numerical simulation of the flow in the suction sump was continuing [1] in both cases of the sumps of NMV 1000RA and NMV 2000RA vertical pumps. The FLUENT computer fluid dynamic code was using and the k- model in turbulent flow was applied. The NMV 1000RA pump was numerical modeled for the flow rates of 1 m3/s and 3 m3/s. In these cases the numerical results did not indicate a swirling flow that could generate problems in the

pump operation. The NMV 2000RA pump was numerical modeled for the flow rates of 10 m3/s and 15m3/s. At a lower water level in the suction sump of the pump a swirling flow was observed, Figure 8, Figure 9 and Figure 10.

Figure 8 Velocity distribution in the vertical and median planes, x-z axis with 10 m3/s flow rate for NMV 2000RA pump

Figure 9 Velocity distribution in the horizontal plane, 10 m3/s flow rate, NMV 2000RA pump

Figure 10 Pathlines in the suction sump with 15 m3/s flow rate for NMV 2000RA pump
Politehnica of Timisoara, Trans on Mechanics, pp 17-22. Gustave A., s.a., (1996), Effects of approach flow condition on pump sump design, J. Hydr. Eng, ASCE, 122(9), pp 489-494. Schafer F., Hellmann D.-H., (2005) Optimization of approach flow conditions of vertical pumping systems by physical model investigation, Proc. of FEDSM2005, Huston, TX, USA, FEDSM2005-77347. Minisci E., Telib H., Cicatelli G., (2005) Hydraulic design validation of the suction intake of a vertical centrifugal pump station, by use of computational fluid dynamic (CFD) analysis, Proc. of FEDSM2005, Huston, TX, USA, FEDSM2005-77330. Iwano R., Shibata T., (2002), Numerical prediction of the submerged vortex and its application to the flow in pump sumps with and without a baffle plate, 9th International Symposium on Transport Phenomena and dynamics of Rotating Machinery, Honolulu, Hawaii. *** FLUENT 5 Users Guide, FLUENT Inc., 1998.

5. CONCLUSIONS The paper regarding the experimental study of the swirling flows in the intake channel suction sump of vertical pumps. Influence of the level upstream in the suction pump and the development of the swirling flow in the suction sump pump are observed. The physical model and the results of the tests give us: NMV 1000RA pump in the suction sump could work in good conditions without vortexes if the submerged head Hsb is more then 0.746 m. It is no recommended in pump operation at submerged heads less then 0.746 m to have a high flow rate. NMV 2000RA pump in the suction sump could work without vortexes if the submerged head Hsb is lower then 1.146 m and the flow rate is between 0.586 m3/s and 0.617 m3/s. REFERENCES 1. Isbasoiu, E.C., Munteanu T., s.a, (2005), Swirling flows in sumps of vertical pumps. approach; Sci. Bul. Of

2. 3.

4.

5.

Safta C.A., the suction Theoretical University

6.

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics

3rd Workshop on Vortex Dominated Flows Bucharest, Romania June 1 June 2, 2007

SOME ASPECTS ABOUT A VORTEX GENERATING BUILDING MODEL PLACED UPWIND AN AEROELASTIC MODEL IN THE BOUNDARY LAYER WIND TUNNEL
Mircea DEGERATU, Prof.* Hydraulic and Environmental Protection Department, Technical University of Civil Engineering Bucharest Liviu HAEGAN, Senior Lecturer Hydraulic and Environmental Protection Department, Technical University of Civil Engineering Bucharest Rzvan- Silviu tefan, Assistant Hydraulic and Environmental Protection Department, Technical University of Civil Engineering Bucharest Andrei GEORGESCU, Lecturer Hydraulic and Environmental Protection Department, Technical University of Civil Engineering Bucharest Costin Ioan COOIU, Assistant Hydraulic and Environmental Protection Department, Technical University of Civil Engineering Bucharest Lucian SANDU, Prof. Hydraulic and Environmental Protection Department, Technical University of Civil Engineering Bucharest

*Corresponding author: Blvd. Lacul Tei 124, sect. 2, 020396 Bucharest, Romania. Tel.: +40 21 243 3660, Fax: +40 21 243 3660, E-mail: mircead@hidraulica.utcb.ro Ud,x [m/s] x-axis dynamic displacement ABSTRACT velocity [m/s] y-axis dynamic displacement Ud,y Using as reference the assumptions synthesized velocity in the paper Dynamic wind tunnel tests for the gravity g [m/s2] Bucharest Tower Centre we have tried to evaluate the influence of a vortex generating building model placed upwind of the aero elastic model of the tower. Considering the actual construction development in Bucharest each aerodynamic study has to consider the influence of the built area that surrounds the studied building. Assuming different values of the free distance between the obstacle building and the studied one we tried to evaluate the structural influence of a building that is part of an urban agglomeration. KEYWORDS Wind engineering, Dynamic wind loads, Dynamic structural response NOMENCLATURE [m/s2] x- axis acceleration ax y- axis acceleration ay [m/s2] [m] x- axis dynamic amplitude Ax [m] y- axis dynamic amplitude Ay [Hz] x-axis frequency fx [Hz] y-axis frequency fy [s] x- axis period Tx [s] y- axis period Ty 1. INTRODUCTION The experimental studies performed, synthesized in this paper, followed the determination of the vibration behavior of a tall structure with dynamic response exposed to the atmospheric boundary layer and include the influence of vortices generated by another upwind positioned building. These experimental researches continue the topics of the paper Dynamic wind tunnel tests for the Bucharest Tower Center published in the Scientific Bulletin of the Politehnica University of Timisoara- Transactions on Mechanics. The simulated atmospheric boundary layer is the same as the one used in the above mentioned paper- i.e. a general power law profile with =0.23. The former paper presented the experimental tests performed in the boundary layer wind tunnel called TASL-1 from the Wind Engineering Laboratory- Technical University of Civil Engineering Bucharest for a 107 m tall building. We presented the following results: a). the building structure; b). the similitude criteria; c). the concept, design and experimental implementation of the

aero-elastic model of the building; d). the experimental conditions including atmospheric boundary layer simulation; e). the measurement methodology; f). the results obtained for the acceleration, frequency and dynamic amplitude both along wind and cross wind using as reference the building axis, for eight different wind directions. 2. EXPERIMENTAL RESULTS OBTAINED USING THE MODEL The measurements were performed for two different values of the wind/ building incidence angle using a 1:100 scaled model (first campaign: =95 - S-E wind- and the second campaign =5S-W wind) using Reynolds numbers of the order of 105. We analyzed the influence of the vortices generated by the structure situated upwind of the studied model. The model of the structure situated upwind is 80 cm long, 15 cm wide and 30 cm tall facing the incident air stream with the larger face crossing the wind direction in the experimental zone of the tunnel. This obstacle building was positioned in front of the model at three different distances: 90 cm, 180 cm and infinite value (the reference value, with no building upwards as shown in fig. 1a,b,c).

Fig. 1c We determined the vibration response characteristics of the aero-elastic 1:100 scale model using both x and y directions with accelerometers as accelerations (ax,ay), dynamic amplitudes of vibrations (Ax, Ay), frequencies (fx,fy), periods (Tx,Ty) and dynamic displacement velocities (Ud,x,Ud,y). These experiments were performed in order to observe the differences induced by the distance between the buildings as well as the wind simulated velocity profile. The results are divided into two groups with respect to the wind/ building incidence angle (SE95 and SW-5) each divided into three subgroups with respect to the position of the building situated upwind the model (90 cm, 180 cm, and ), so weve indexed our results with two arguments (SE 95-90, SE 95-180, SE 95- and SW 5-90, SW 5- 180, SW 5-). 3. EXPERIMENTAL RESULTS TRANSPOSED TO THE REAL BUILDING The vibration characteristics obtained by the measurements performed on the 1:100 scale model for the six different situations were transposed at the real scale using the proper similitude scales for each parameter. These scales were obtained using certain limitations for the wind action on aero elastic structures (just as the studied building is in our case). The transposed results at the natural scale represent the vibration characteristics along the x and y axis of the structure at a reference height of 100 m above ground. - The first ten diagrams present the vibration characteristics of the model due to loads induced by SE wind (SE 95-90, SE 95-180, SE 95-); - The last ten diagrams present the vibration characteristics of the model due to loads induced by SW wind (SW 5-90, SW 5- 180, SW 5-).

Fig. 1a

Fig. 1b

25.00 20.00 a x [m/s2] Model 15.00 10.00 5.00 0.00 0.00

0.25 0.2 a x [m/s2] Prototip

30.00 25.00

0.30 0.25 0.20 0.15 0.10 0.05 0.00 8.00

0.15 0.1 0.05 0 8.00

20.00 15.00 10.00 5.00 0.00 0.00

1.00

2.00

3.00

4.00
-7

5.00

6.00

7.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

Re x 10 pe Prototip Re x 10-5 pe Model SV 5 - ? SV 5 - 90 SV 5 - 180

Re x 10-7 pe Prototip Re x 10-5 pe Model SV 5 - SV 5 - 90 SV 5 - 180

6.00 5.00 Ax [mm] Model 4.00 3.00 2.00 1.00 0.00 0.00

600.00 500.00 Ax [mm] Prototip

7.00 6.00

700.00 600.00

Ay [mm] Model

400.00 300.00 200.00 100.00 0.00 8.00

4.00 3.00 2.00 1.00 0.00 0.00

400.00 300.00 200.00 100.00 0.00 8.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

Re x 10-7 pe Prototip Re x 10-5 pe Model SV 5 - ? SV 5 - 90 SV 5 - 180

Re x 10-7 pe Prototip Re x 10-5 pe Model SV 5 - ? SV 5 - 90 SV 5 - 180

13.00

0.13

13.00 12.00

0.13 0.12 0.11 0.10 0.09 0.08 8.00 fy [Hz] Prototip

12.00 fx [Hz] Model

0.12 fx [Hz] Prototip


fy [Hz] Model

11.00 10.00 9.00

11.00

0.11

10.00

0.10

9.00 0.00

1.00

2.00

3.00

4.00
-7

5.00

6.00

7.00

0.09 8.00

8.00 0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

Re x 10 pe Prototip Re x 10-5 pe Model SV 5 - ? SV 5 - 90 SE 5 - 180


SV 5 - ?

Re x 10-7 pe Prototip Re x 10-5 pe Model SV 5 - 90 SV 5 - 180

0.12

120.00

0.13 0.12

130.00 120.00 110.00 100.00 90.00 80.00 8.00

0.11 Tx [s] Model

110.00 Tx [s] Prototip


Ty [s] Model

0.10

100.00

0.11 0.10 0.09

0.09

90.00

0.08 0.00

1.00

2.00

3.00

4.00
-7

5.00

6.00

7.00

80.00 8.00

0.08 0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

Re x 10 pe Prototip Re x 10-5 pe Model


SV 5 - ?
SV 5 - 90
SV 5 - 180
SV 5 -

Re x 10-7 pe Prototip Re x 10-5 pe Model


SV 5 - 90
SV 5 - 180 Series4

Ty [s] Prototip

Ay [mm] Prototip

5.00

500.00

a y [m/s2] Prototip

a y [m/s2] Model

0.30 0.25 Udx [m/s] Model 0.20 0.15 0.10 0.05 0.00 0.00

0.30 0.25 Udx [m/s] Prototip 0.20 0.15 0.10 0.05 0.00 8.00
Udy [m/s] Model

0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 0.00

0.35 0.30 Udy [m/s] Prototip 0.25 0.20 0.15 0.10 0.05 0.00 8.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

Re x 10-7 pe Prototip Re x 10-5 pe Model SV 5 - ? SV 5 - 90 SV 5 - 180

Re x 10-7 pe Prototip Re x 10-5 pe Model SV 5 - ? SV 5 - 90 SV 5 - 180

30 25 a x [m/s2] Model 20 15 10 5 0 0.00 1.00 2.00 3.00 4.00


-7

0.3 0.25 a x [m/s2] Prototip 0.2 0.15 0.1 0.05 0 5.00 6.00 7.00 8.00 Re x 10 pe Prototip Re x 10-5 pe Model SE 95 - ? SE 95 - 90 SE 95 - 180

30 25
a y [m/s2] Model

0.30 0.25 0.20 0.15 0.10 0.05 0.00 8.00


a y [m/s2] Prototip
fy [Hz] Prototip
Ay [mm] Prototip

20 15 10 5 0 0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

Re x 10-7 pe Prototip Re x 10-5 pe Model


SE 95 -
SE 95 - 90
SE 95 - 180

9 8 7 Ax [mm] Model 6 5 4 3 2 1 0 0.00 1.00 2.00 3.00 4.00


-7

900.00 800.00 700.00


Ay [mm] Model

7 6

700.00 600.00 500.00 400.00 300.00 200.00 100.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 Re x 10-7 pe Prototip Re x 10-5 pe Model

600.00 500.00 400.00 300.00 200.00 100.00 0.00 5.00 6.00 7.00 8.00 Re x 10 pe Prototip Re x 10-5 pe Model SE 95 - ? SE 95 - 90 SE 95 - 180

Ax [mm] Prototip

5 4 3 2 1 0 0.00

SE 95 - ?

SE 95 - 90

SE 95 - 180

12 11 fx [Hz] Model 10 9 8 7 6 0.00

0.12 0.11 0.10 0.09 0.08 0.07 0.06 8.00 fx [Hz] Prototip

12 11.5 11
fy [Hz] Model

0.12 0.12 0.11 0.11 0.10 0.10 0.09 0.09 1.00 2.00 3.00 4.00 5.00 6.00 7.00 0.08 8.00

10.5 10 9.5 9 8.5

1.00

2.00

3.00

4.00
-7

5.00

6.00

7.00

8 0.00

Re x 10 pe Prototip Re x 10-5 pe Model SE 95 - ? SE 95 - 90 SE 95 - 180

Re x 10-7 pe Prototip Re x 10-5 pe Model SE 95 - ? SE 95 - 90 SE 95 - 180

0.15 0.14 0.13 Tx [s] Model 0.12 0.11 0.1 0.09 0.08 0.00

150.00 140.00 Tx [s] Prototip 130.00 120.00 110.00 100.00 90.00 80.00 8.00

0.12 0.115 0.11 Ty [s] Model 0.105 0.1 0.095 0.09 0.085 0.08 0.00

120.00 115.00 110.00 105.00 100.00 95.00 90.00 85.00 80.00 8.00 Ty [s] Prototip

1.00

2.00

3.00

4.00

5.00

6.00

7.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

Re x 10-7 pe Prototip Re x 10-5 pe Model SE 95 - SE 95 - 90 SE 95 - 180


SE 95 -

Re x 10-7 pe Prototip Re x 10-5 pe Model SE 95 - 90 SE 95 - 180 Series4

0.5 0.4 Udx [m/s] Model 0.3 0.2 0.1 0 0.00

0.5 0.4 Udx [m/s] Prototip

0.3 0.25 Udy [m/s] Model 0.2 0.15 0.1 0.05 0 0.00

0.3 0.25 0.2 0.15 0.1 0.05 0 8.00 Udy [m/s] Prototip

0.3 0.2 0.1 0 8.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

1.00

2.00

3.00

4.00
-7

5.00

6.00

7.00

Re x 10-7 pe Prototip Re x 10-5 pe Model SE 95 - ? SE 95 - 90 SE 95 - 180

Re x 10 pe Prototip Re x 10-5 pe Model SE 95 - ? SE 95 - 90 SE 95 - 180

4. CONCLUSIONS The experimental studies were performed for Reynolds numbers in the range [2.5 .... 7.5]* 107. These numbers are big enough to insure the self simillarity of the longitudinal velocity profile for the inertial domain which includes the fluctuating components of the velocity that induce the resonance excitement for this tall building category. The assumptions that will follow were made for Reynolds numbers in the range [6.5 .... 7.5]* 107 because this domain offers significant values for the aeroelastic behavior of the analyzed structure. Regarding the acceleration parameters, the vortices created upwind induce significant growth of the values referring to the x- axis for SE wind load and the y- axis for the SW wind load and decrease the acceleration values on the y- axis for SE wind load and the x- axis for SW wind load. Regarding the dynamic amplitudes, the vortices created upwards induce decreasing values reffering to both axis directions and both wind directions. As the distance between the obstacle building decreases, the amplitude values decrease is stronger. The present research confirms, once again, the necesity of performing this type of studies in order to get a good idea of the interaction between the

atmospheric boundary layer and the slender structures. The final caption presents the model placed in the experimental vein of the aerodynamic wind tunnel (Fig.2).

Fig. 2

REFERENCES 1. Degeratu, M., Georgescu, A., Hasegan, L., Cosoiu, C., Pascu, R., Sandu, L. 2006 Dynamic wind tunnel tests for the Bucharest tower center., Trans. on Mechanics, Sci. Bull. Politehnica University of Timisoara, Romania, Tom 51(65), Fasc. 3. 2. Davenport, A.G. 1968. The application of the boundary layer wind tunnel to the prediction of wind loading. Proc. of International Research Seminar of Wind Effect on Buildings and Structures, Toronto. 3. Degeratu, M. 2002. Atmospheric boundary layer (in Romanian), Timioara: Orizonturi Universitare. 4. Iamandi, C., Degeratu, M., Georgescu, A. 1996. A wind tunnel study for a TV tower in Bucharest Romania. International Conference on Urban Engineering in Asian Cities. Bangkok, Thailand. 5. Mondkar, D.P. & Powell, G.H. 1975. ANSR I General Purpose Program for Analysis of Nonlinear Structural Response. Report No. EERC 75-37. Berkeley, California.

6. Sandu, L., Degeratu, M., Haegan, L., Georgescu, A., Cooiu, C.I. 2005. Modelling of wind action on the Bucharest Tower Center International Building (in Romanian). Research report no. 427/2004. Technical University of Civil Engineering, Bucharest. 7. Romanian Standard Institute, 1990, Loads due to wind, STAS 10101/20. 8. Sandu, L., Degeratu, M., Hasegan, L., Georgescu, A., Stefan, R., Cosoiu, C., 2005, Actual and Perspective Research in Wind Engineering, Trans. on Mechanics, Sci. Bull. Politehnica University of Timisoara, Romania, Vol. 50 (64), pp. 2332. 9. Iamandi, C., Georgescu, A., Erbasu, C., 2003, Experimental Modelling of a Four Steel Tanks Battery, Proc. 11th International Conference on Wind Engineering, (T21), Lubbock, Texas, USA. 10. Iamandi, C., Georgescu, A., Erbau, C., 2002, Atmospheric Boundary Layer Change, Int. J. of Fluid Mechanics Research, Begell House Inc, New York, Vol. 29, ISSN 1064-2277.

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Tom 52(66), Fascicola 3, 2007

3`nd Workshop on Vortex Dominated Flows Timisoara, Romania June 1 - 2, 2007

FREE-SURFACE TURBULENT FLOW AROUND A LPG SHIP HULL


Adrian LUNGU*, Professor Department of Ship Hydrodynamics Dunarea de Jos University of Galati

*Corresponding author: 47 Domneasca Street, 800008, Galati, Romania Tel.: (+40) 236 495400, Fax: (+40) 236 495400, Email: adrian.lungu@ugal.ro the following paragraphs, the CFD process is composed of two distinct parts, (i) selection or development of a general-purpose CFD code and (ii) use of the CFD code for solution of a particular flow problem of interest. In general, the former occurs only at infrequent intervals when need arises to make large shifts in technology, whereas the latter must be followed for each simulation. Development of any general-purpose CFD code has several common elements. Specifically, formulation of the general initial and boundary value problem (IBVP) which is to be solved numerically using a CFD code, development of numerical methods for approximate solution of the IBVP, and documentation of the CFD code. Key issues in the formulation of the IBVP are in definition of the scope and level of flow description (e.g., RANS, LES, DNS), selection of governing partial differential equations and physical models for the fluid flow, and selection of a comprehensive set of initial and boundary conditions required to solve a wide range of applications. With regard to numerical methods, key issues include discretization of the continuous modeled PDEs, initial and boundary conditions, development of numerical algorithms for solution of the discretized modeled equations, and testing the algorithm in a CFD code. Generally speaking, there are two distinct approaches in representing the free-surface in a Reynolds-averaged Navier-Stokes (RANS) simulation, namely, interface tracking and interface capturing methods. Both approaches aim to compute the wave profile accurately because the wetted surface area appears in the calculation of the drag acting on ship motion. In interface tracking methods, a kinematic boundary condition is applied at the free-surface, and the governing equations are solved only for the water phase [1], [2]. Since the computational grid has to conform to the free-surface shape, this approach is not efficient for high Froude number flows in which breaking

ABSTRACT The paper presents a methodology for computing the 3D turbulent free-surface flow around a 10000 DwT LPG vessel. The flow is simulated by making use of a RANS-based finite volume approach. Closure to turbulence is attained through the K- SST turbulence model. Various flow cases were studied in an attempt of finding the most suitable hull forms from the wave-resistance point of view. KEYWORDS Free-surface flow, finite volume method, K- SST turbulence model 1. INTRODUCTION There is a significant effort in the marine industry to integrate the computational fluid dynamics (CFD) simulation capability in designing energy efficient ship hull forms while lowering the noise generated from them. Accurate simulation of turbulent free surface flows around surface ships has a central role in the optimal design of such naval vessels. The flow problem to be simulated is rich in complexity and poses many modeling challenges because of the existence of breaking waves around the ship hull, and because of the interaction of the two-phase flow with the turbulent boundary layer. Overall philosophy for the CFD process is given in this section as a set of procedures to guide engineers and scientists through the process of modeling fluid flow problems using a CFD code. Although some of the elements of the CFD process are relatively straightforward, development of a comprehensive process is useful for training nonexpert CFD users, establishing confidence in results from CFD codes, assessing risks in the use of CFD results in a design environment, and streamlining the task of obtaining CFD solutions leading to reduced manpower requirements. As described in

waves with high amplitudes are typical. On the other hand, the surface capturing methods are more versatile in handling a variety of free-surface flow conditions, in which the governing equations are solved for both the air and the water phases. Volume of fluid (VOF) [3], level-set [4] and front tracking [5] are the widely adopted techniques in this category. The turbulent flow structure in the vicinity of the free-surface can also be much more complex than the turbulent flow structure of single phase flows. For instance, it is known that for turbulent breaking waves, surface tension effects are important at smaller scales, and gravity is effective on larger scales [6]. Existing turbulence models, which have been mainly proposed for single phase flows, may not adequately represent the turbulence structure at the free-surface. In 1993, Menter [7] published a new shear stress transport k model which has remarkable advantages when compared with k and previous k models. The model is based on Bradshaws assumption that the principal shearstress is proportional to the turbulent kinetic energy, which is introduced into the definition of the eddyviscosity. The k models over-predict the turbulent shear stress in boundary layers under adverse pressure gradient. The previous k models give somewhat more realistic solutions for such flows, but the improvement is only partial. With the k SST model a significant improvement is reached. The k SST model also has the advantage of being insensitive to the free stream value. In the k SST model the k model is used near the wall and a k model, transformed to resemble a k model, is used outside of this region. The different sets of coefficients and the additional cross-diffusion term from the transformed k model are combined by blending or switching functions Fi in an intermediate region. The existence of a free-surface interacting with the turbulent boundary layer presents challenges for turbulence modeling. It is therefore necessary to assess carefully the performance of available turbulence models in literature for simulating turbulence flows with free-surface. The important dimensionless numbers defining the flow conditions for the free-surface flow around ships are the Froude and the Reynolds numbers, respectively. The Froude number is representative of the ratio of inertial forces to gravitational forces, whereas the Reynolds number represents the ratio of inertial forces to the viscous forces. A Froude number of 0.242 and a Reynolds number of 1.2107 defines the standard flow conditions

for the present case, which correspond to a ship speed of 16 knots. The ship hull is supposed to perform longitudinal (trim) and vertical (sinkage) free movements in waves. The characteristic length in Froude and Reynolds numbers is the distance measured between the perpendiculars of the ship. A series of nine speeds ranging from 14 knots to 18 knots around the required speed is considered. 2. GRID GENERATION In the past 30 years the technique of solving the Navier-Stokes (N-S hereafter) equations has made remarkable progress and the gradient of improvement has recently been decreased. The CFD technology is now putting more stress on customizing, for which the grid generation technique plays an important role. Since we have reliable solution method for the N-S equations, the simulation success relies on the grid generation to a larger extent than before. Actually a large part of the efforts is devoted to the generation when a 3D body of complex geometry is deal with. Although the gridless technique is continuously investigated, it still has some substantial difficulties, such as poor implementation of the conservation laws. Presently we must work within the framework of the method with grid system. Since the unstructured grids like those for the finite element method still have difficulties in developing into 3D cases, one can choose either the rectangular grids or the boundary-fitted, curvilinear structured grids. Although the structured grids find difficulties in the representation of the boundary of extreme complexity, it will be employed in the present study. A fine grid around the interface is desirable to minimize the smearing of the interface due to numerical diffusion. In addition, the VOF technique also benefits from adopting grid cells of hexahedral shape. Furthermore, the turbulent boundary layer, which is very thin due to the high Reynolds number, needs to be resolved with a mesh having a not too large aspect ratio. Although these requirements can be met easily for a relatively simple geometry, such as a free-surface piercing hydrofoil, the grid generation process for the current geometry proved to be a tedious one because of the existence of a bulbous fore, which can be seen in Fig. 1. A hybrid structured mesh consisting of about 2106 cells was generated for the present simulations. Fig.1 shows the overall grid structure adopted in the simulation. The grid generation process is as follows. First, the surface of the ship is described by an off-set file and then a boundary layer mesh is fitted all around it. The spacing of the nearest grid cells are such that it

is consistent with the wall function formulation. The computational domain is then meshed with quadrilateral shaped 2D cells, which are the basis for the 3D hexahedral cells in the domain. Finally, a local grid refinement (e.g. each 3D cell is divided into 4 cells, then into 16 cells, and so on) is applied around the free-surface and the hull extremities to provide reasonable resolution of the interface.

U i =0 xi

(2)

The N-S equations of motion can be written in the following form:

U jU i ij U i = Ri + + x j x j t

(3)

where ij is the total stress and for a Newtonian fluid can be written as:
ij = P ij + 2 Sij Skk ij
1 3

(4)

where Sij is the strain-rate defined as:


1 U U j Sij = i + 2 xi x j

(5)

Skk in (4) is zero for incompressible flow:


Z

1 U k U k + S kk = 2 xk xk

U k = x = 0 k

(6)

Figure 1. The structured grid topology 3. MATHEMATICAL MODEL The solver is a finite volume RANS (ReynoldsAveraged Navier-Stokes) one with a K- SST turbulence model. The convective terms are discretized with a Roe scheme and a second order explicit defect correction is used to approach second order accuracy. The rest of the terms are discretized with central differences. A local artificial time-step is added to the equations and the discrete coupled equations are solved using an ADI-solver. Solving the RANS equations will give the time average velocity and pressure, but since the time fluctuating velocity and pressure in general are much smaller in amplitude, knowing the average will usually be enough. 3.1.1. Equations The N-S equations can be solved numerically by resolving all scales for turbulent flows, which requires extremely dense grids to resolve the smallest turbulent length scales. In the present model, the continuity equation states that mass is conserved:
1 U i + =0 t xi

The RANS equations can be derived from (3) by splitting the instant velocity components Ui in time mean velocity ui, and time fluctuating velocity ui,

U i = U i + ui" ui + ui"

(7)

and the instant pressure, P, mean pressure in time, p, and time fluctuating pressure, p,

P = P + p '' p + p ''
The time mean of a variable is defined as:

(8)

= lim

1 2 T T

dt

(9)

The following rules of averaging apply for any two turbulent quantities 1 and 2:
" " ; 1 1 = 1 + 1 = 0; 1 = 1; 12 = 1

1''2 = 0;

1 1 ; 1 + 2 = 1 + 2 = s s

(10 )

" " 12 = 12 + 1 2

(1)

Taking the time average of the continuity equation gives:

Only incompressible flow is considered in the present study. That means that the changes in density are negligible. Then the continuity equation can be written:

U i U i ui = = =0 xi xi xi

(11)

Subtracting (11) from (2) gives that also the time fluctuating velocity fulfills the incompressible continuity equation:

ui'' =0 xi

(12)

procedure produces differential equations for the Reynolds stress tensor. But these equations produce some new unknowns and solving all these equations takes a lot of CPU time. Therefore, the Boussinesq approximation for incompressible flows
ui''u 'j' = T
ui u j + x j xi 2 + k ij 3

Then taking the time average of the N-S equations (first moving all the terms to the left hand side):
U jU i U P U i + U j Ri + i + t x j xi x j x xi j

(15)

is assumed valid and an equation for the turbulent kinetic energy, k, is created instead. The last term in (15) is needed to get the proper trace of the Reynolds stress tensor
ui''u 'j' = T
ui u j + x j xi 2 + k ui''u 'j' = 3

U jU i U i P + Ri + xi x j t t

(16 )

'' ' ' u j ui + u j ui U i + U j = U i + x j xi t x j

Thus, the Reynolds-averaged equations can be written as follows:


ui u j ui 1 p 2 k + + = Ri xi 3 xi t x j ui + u j + E x j x j xi

Ri +

p xi x j

u i + u j x j xi

(17)

Hence, the time averaged continuity equation and N-S equations for incompressible flow can be written as follows:
ui =0 xi
'' '' u j u i + u j ui ui = R 1 p + + i x j xi t

where E is the effective kinematic viscosity defined as:


E = + T ; = ; T = T

(18)

(13 )

x j

u u j i + x j xi

The k equation can be derived by subtracting the NS equations for the mean flow (13) from the N-S equations for the instantaneous flow (3), multiplying it with ui'' and taking the time average:
u k u j k + = ui''u 'j' i + x j x j t x j

( )

where:
=

(14)

k 1 '' '' '' 1 '' '' u u u p u j ; x j 2 i i j

(19)

3.1.2. Turbulence modeling

It is through the Reynolds stresses ui''u 'j' that the turbulence makes their imprint upon the mean velocity and pressure. The problem now is that the time averaging has created a tensor of unknowns. Since ui''u 'j' is symmetric there are six additional unknowns. This is called the closure problem. In order to solve the system more equations are needed. Additional equations can be formed by taking moments of the N-S equations. This can be done by multiplying the N-S equations by a fluctuating property and time average the product. This

ui''ui'' xk xk

The equation for k is not closed since it contains unknown quantities for ui''u 'j' , dissipation ( ) and turbulent diffusion (the last two terms on the right hand side). Closure of the k equation can be achieved by replacing the unknowns with approximations which are based on experimental data. The Reynolds stress tensor can be replaced by assuming that the Boussinesq approximation is valid. The last two terms in the first equation of the (19) set are usually combined and modeled as a gradientdiffusion term:

k 1 1 ui' 'ui' 'u 'j' p ''u 'j' = T k x j 2

(20)

still needs to be modeled and an equation for the turbulent dissipation can be created by taking
2 ui'' x j x j

(21)

contains the first-order Roe convective terms and the second order diffusive terms, while the secondorder flux corrections are used as an explicit defect correction. Each element in the tri-diagonal matrix is a 6x6 matrix. For each sweep a local artificial time-step is calculated based on the CFL and von Neumann numbers in all directions except the implicit one.
6. RESULTS AND DISCUSSIONS As said above, the aim of the present study was to investigate the flow features in order to determine the optimum shape of the hull form aimed at contributing the ship resistance. Two are the key features that are taken into account. The first is the front crest of the wave that develops around the ship stem, whereas the second is the flow structure around the stern. The first problem has relevance to the wave-making resistance problem. It is widely known that there is a significant effort in the marine industry to integrate the CFD simulation capability in designing energy efficient ship hull forms while lowering the noise generated from them. Accurate simulation of turbulent free surface flows around surface ships has a central role in the optimal design of such naval vessels. The flow problem to be simulated is rich in complexity and poses many modeling challenges because of the existence of breaking waves around the ship hull, and because of the interaction of the two-phase flow with the turbulent boundary layer. Moving a blunt piercing body through the free surface often results in a bow wave that breaks either in a plunging or in a spilling manner. Studying the breaking bow wave has relevance for naval architects involved in the design of hull forms since as much as 15% of the resistance of a full-form ship hull is associated with its breaking bow wave. The subject is difficult, however, since most solution techniques for waves modeling are valid only when the waves have sufficiently small slopes, their speeds are considerably smaller than the phase speed, and the particle accelerations are not large compared to the acceleration due to gravity. Besides, a challenge has always come from the uncertainty associated with the coexistent nonlinearities such as viscous interactions and freesurface tensions. Moreover, many difficulties arise when the air-water interface undergoes large deformations or becomes discontinuous, as shown in Figure 2, and a fully non-linear free-surface boundary condition is required.

and applying it on the instantaneous N-S equations (3) and finally averaging in time everything. This leads to quite a complex equation for containing considerably more double and triple correlations than the k equation. These correlations needs to be modeled and a far simpler equation is obtained
u 2 u j = C 1 ui' 'u 'j' i C 2 + + k k x j x j t + x j + T x j

( )

(22 )

where C 1 , C 2 and are modeled closure coefficients. Since the k model often does not produce satisfactory results, some other turbulence models have been developed. One of them is the k model that in some ways performs better then the k model and in some way worse. The relationship between and is =
0.09k

4. BOUNDARY CONDITIONS The boundary conditions used are described in Table 1. In the implementation two layers of ghost cells are used.

Table 1. Boundary conditions formulation No-slip Slip Inflow u ui = 0 ui ni = 0 ui = const . ui =0 B p p p p =0 =0 =0 B B B k =0 k = const. k k =0 B = const. = f (ut ...) =0 B

Outflow ui =0 B
p=0 k =0 B =0 B

5. NUMERICAL SCHEME The ADI scheme is used to solve the flow equations. The tri-diagonal systems that are solved

Figure 2. Free-surface profile around the ship stem An important issue of any free-surface flow under the breaking assumption is the separation that is exclusively provoked by the wave-induced effects. When an oncoming free-surface flow encounters a piercing body, the flow may separate as a result of a sort of blocking effect. The separated flow forms several vortices around the body, which are topologically rather similar to the horseshoe vortices generated around a juncture. The flow is completely three-dimensional and becomes more complicated as the vortices interact with the water wave and the boundary layer developed on the body surface. This convoluted three-dimensional flow can be found in many places where a piercing body intersects the water surface, such as the flow around appendages, antennas or periscopes mounted on submarine fuselages, etc. In spite of its engineering importance, there is no established method for estimating the wave resistance or wake characteristics because the detailed flow mechanism is not completely understood. Although some studies have been performed either theoretically or through conventional tank tests, the estimation of the wave resistance under the breaking assumption is hampered by the uncertainty associated with the complexity of the phenomenon. The heterogeneous character of the flow is due to the number of vortices originating upstream of the piercing body. These vortices result from the wave-induced separation due to the adverse pressure gradient in front of the body. Apart of the three dimensional separation that eventually takes place, complex free-surface deformations occur as a result of the breaking inception, contributing to the vorticity production in the immediate vicinity of the breaking wave-crest. Complex interactions that produce there unveil topological flow structures that resemble, up to some extent, to those generated by the

contingence between a separated flow around a juncture and the corresponding boundary layer. The phenomenon was firstly identified by Tahara and Stern [8]. Later, Muzaferija and Peric, [9] as well as Olivieri, A. et.al. [10], studied more extensively the phenomenon, being able not only to discuss the flow conditions and uncertainty analysis, but also to derive and verify a new topological rule for a surface-piercing body configuration. Improving the lines plan in the fore part of the ship usually requires trial computations, visualizations and comparisons of the friction coefficient field around the ship shoulders for all the considered speeds. Figure 3 shows the friction coefficient field computed around the hull at a speed of 16 Knot. Towing tank measurements have proven that hulls with a significant geometric gradient (i.e. low curvature) may be subjected of a high viscous resistance. Since areas with such a curvature should be avoided in the design process, the modification of the lines plan should be done interactively according to the numerical solution of the flow problem.

Figure 3. Friction coefficient around the hull The other area subjected to careful insight of the flow features regards the stern. This is because a poor definition of the forms may easily produce a strong wake, which eventually affects the propulsion performances. That is, blunt bodies with small bilge radius may determine strong vortical flow structures, accompanied by separations and reattachments. Usually the bilge vortex and the propeller jet are nearly coaxial. If, in addition, they are co-rotating an accumulation effect will occur. Such a completely three-dimensional viscous flow often leads to a non-uniform velocity field in the propeller, a fact that should be avoided for cavitational reasons, see [10] and [11]. In this respect, the discussion in conjunction with Figure 4 proposes an analysis of the particle paths plotted in the field around the aft part of the hull, as well as the topology of the corresponding streamlines. The figure also holds several transverse cuts in the fields of the turbulence quantity, T . As it may be seen in

the proposed figure, the flow is completely threedimensional, the horizontal projections of the streamlines being substantially different from an xplane to the other.

Figure 4. Streamlines topology computed around the ship stern The topological structure of the streamlines inside the boundary layer is characterized by the appearance of two pairs of divergence-coalescence lines just above the bulb stern. A nodal point of separation is placed in both sides of the symmetry plane. A coalescent line to which all the streamlines converge is extended over about a half of the draught. Although not ideal, it is worth mentioning that such a flow structure represents the best that we could get after a big deal of numerical computations. The wake of a ship with running propeller is quite different from the wake of a towed hull (nominal wake). The differences are not only related to the diminished momentum deficit but also to the interaction between the propeller-induced flow and the hull wake. A propeller in a uniform onset flow tends to accelerate the flow ahead as well as in the jet behind, and to weakly decelerate the flow outside these regions. Ahead of the propeller and outside the jet there is no propellerinduced rotation. For a qualitative analysis of the interaction between the propeller and the nominal wake, one must recall that vorticity, once generated, is carried with the flow and is spread by diffusion, on the understanding that at a high Reynolds number the convection dominates the diffusion. In this respect, Fig.5 depicts the vorticity iso-surfaces around the ship stern.

Figure 5. The vorticity iso-surfaces plotted around the ship stern Where in most towed-hull wakes the smallest axial velocity occurs in the symmetry plane, this is seldom the case in the hull wake; two minima away from the symmetry plane but at or near the free surface are commonly observed, as Figure 6 bears out. Apparently the viscous diffusion effects are so strong as to more than neutralize the acceleration effect. The enhancement of the vortex strength by stretching effects is also responsible for the acceleration of the flow in the central region near the free surface. As bilge vortices in a nominal wake are effective in reducing the wake peak in the top sector of the propeller disk, the stronger vortices in the propelled-hull wake have a proportional effect on the flow in the central region near the free surface.

Figure 6. The axial velocity iso-surfaces plotted around the ship stern
7. CONCLUSIONS The paper presents a methodology for computing the 3D incompressible turbulent free-surface flow around a 10000 DwT LNG carrier. A boundary fitted co-ordinate system is employed to allow an accurate formulation of the boundary conditions. Third-order upwind schemes are used to determine the location of the water particles laying on the free surface for the sake of speeding up the wave elevation. The numerical method proves the expected robustness in dealing with impact wave problems. ACKNOWLEDGMENTS The reported research could not be possible without the generosity of the Romanian National University Research Council (CNCSIS) which is greatly acknowledged for the provided financial support through the Grant A_C code 33/2005.

REFERENCES 1. Li, T., Matusiak, J. & Lehtimaki, R. (2000) Numerical simulation of viscous flows with free surface around realistic hull forms with transom. Int. J. Num. Meth. Fluids. 37, 601. 2. Rhee, S. H. & Stern, F. (2001) Unsteady RANS method for surface ship boundary layer and wake and wake field. Int. J. Num. Meth. Fluids. 37, 445. 3. Hirt, C. W. & Nichols, B. D. (1981) Volume of fluid method for the dynamics of free boundaries. J. Comp. Phys. 39, 201. 4. Osher, S. & Sethian, J. (1988) Fronts propagating with curvature dependent speed: Algorithms based on Hamilton Jacobi formulations. J. Comp. Phys. 79, 12. 5. Harlow, F. H. & Welch, E. J. (1965) Numerical calculation of time-dependent viscous incompressible flows of fluid with free surface. Phys. Fluids 8, 2182. 6. Brocchini, M. & Peregrine, D. H. (2001) The dynamics of strong turbulence at free surfaces Part 1: Description. J. Fluid Mech. 449, pp.225-238. 7. Menter, F.R. (1993) Zonal Two Equation k Turbulence Models for Aerodynamic Flows, 24th Fluid Dynamics Conference, AIAA paper-93-2906. 8. Tahara Y, Stern F. A large-domain approach for calculating ship boundary layers and wakes and wave fields for nonzero Froude number. Journal of Computational Physics 1996; 127: 348. 9. Muzaferija, S. & Peric, M. (1998) A two fluid Navier-Stokes solver to simulate water entry. In Proceedings of 22nd Symposium on Naval Hydrodynamics, Washington DC. 10. Olivieri, A., Pistani, F., Avanzini, A., Stern, F. & Penna, R. (2001) Towing tank experiments of resistance, sinkage and trim, boundary layer, and free surface flow around a naval combatant INSEAN 2340 model. IIHR Technical Report No. 421, The University of Iowa. 11. Larsson, L., Stern, F., Bertram, V. (2000) A workshop on numerical ship hydrodynamics. Proceedings of the Gothenburg 2000, Chalmers University of Technology, Gothenburg, Sweden.

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Tom 52(66), Fascicola 3, 2007

3`nd Workshop on Vortex Dominated Flows Timisoara, Romania June 1 - 2, 2007

UNCERTAINTIES IN THE FREE-SURFACE POTENTIAL FLOW CODE SOLUTION


Adrian LUNGU*, Professor Department of Ship Hydrodynamics Dunarea de Jos University of Galati

*Corresponding author: 47 Domneasca Street, 800008, Galati, Romania Tel.: (+40) 236 495400, Fax: (+40) 236 495400, Email: adrian.lungu@ugal.ro verification deals with mathematics and validation with physics) and activities (e.g., method of manufactured solutions and benchmark solutions for verification and use of experimental data for validation) defining V&V; discussion for validation of both codes and solutions; many case studies demonstrating use of the grid convergence index; single grid error estimation methods; and broader issues such as code quality assurance and certification. Such definitions, however, in the author viewpoint are inadequate. Quantitative metrics are needed for both verification and validation, and methodology is needed for combining errors and uncertainties. The AIAA Committee on Standards for CFD [3] and Guide for V&V of CFD Simulations [4] uses definitions from information theory for errors and uncertainties with emphasis on measurement of accuracy as opposed to estimation of errors and uncertainties; follows Roache phrases and expands considerably on his activities along with broad statements in defining validation, discusses mostly code, but also solutions and additionally discusses policy statements on experimental and numerical accuracy. Code verification activities measure accuracy in relation to benchmark analytical and ordinary and partial differential equation solutions for simplified problems along with software quality assurance: identify, quantify, and reduce errors in the computational model and its numerical solution. Model validation activities measure accuracy in relation to experimental data with emphasis on validation tiers based on unit problems, benchmark cases, subsystem cases, and complete systems: identify and quantify error and uncertainty in the conceptual and computational models, quantify the numerical error in the computational solution, estimate the experimental uncertainty, and compare the computational and experimental results. Solution verification largely follows Roache along with consistency and iterative convergence checks. Rig-

ABSTRACT The paper proposes a numerical investigation based on the boundary element method for solving the free-surface flow around the ship hull. Non-linear boundary condition is imposed not only on the hull, but also on the water free surface. The present paper summarizes the verification and validation (V&V) of a CFD simulation by analyzing the SHIPFLOW CFD code solutions, including some findings about the uncertainty estimates. Opportunities and challenges for achieving consensus and standard validation and certification methodology and procedures are discussed. KEYWORDS Free-surface potential flow, convergence tests, numerical issues

discretization

1. INTRODUCTION In spite of the ever-increasing need and importance for standards for computational fluid dynamics (CFD) uncertainty analysis/accuracy estimation, there are currently many viewpoints covering all aspects from basic concepts and definitions to detailed methodology and procedures. A similar situation existed for experimental fluid dynamics (EFD) uncertainty analysis ca. 1960 for which currently standards are widely accepted and available, although widespread use is still lacking. Pioneering work was done by Roache [1] who proposed the grid convergence index for estimating uncertainty due to grid and time step errors based on Richardson extrapolation using multiple solutions on systematically refined grids; thereby, providing a quantitative metric for verification. Ref. [2] expanded on this work through overall discussion of verification and validation, including: use of definitions for errors and uncertainties; phrases (e.g., verification deals with equations solved correctly and validation with correct equations and

orous implementation is impressive [5, 6]; nonetheless, these definitions are subject to same criticisms mentioned earlier. Another problem with such definitions is lack of an overall mathematical framework for V&V, which is considered essential in the authors viewpoint similarly as it is an essential and integral part of EFD uncertainty analysis. The literature also includes editorial policy statements [7], additional guidelines [8], and numerous case studies, which mostly focus on verification procedures for 2D problems. In general this literature follows approaches similar to that described above. The authors of [9] developed an alternative quantitative approach to solution validation specifically for already developed CFD codes for industrial applications (geometry and domain; models; initial, boundary and other conditions; fluid properties) and input parameters (such as iteration numbers and grid and time step sizes), which differs considerably from previous approaches. It is assumed that code verification and quality assurance issues have already been dealt with during code development. Similarly, if appropriate, it is assumed that model validation for simplified problems has also already been dealt with during model development. The philosophy is strongly influenced by EFD uncertainty analysis [10], including use of EFD definitions for errors and uncertainties. The methodology is based on concepts, definitions, and equations derived for simulation errors and uncertainties, which provide the overall mathematical framework. Verification procedures for estimating numerical errors and uncertainties include (1) the options of estimating the numerical uncertainty or the numerical error itself, which is used to obtain a corrected solution, and its uncertainty; and (2) the concept of correction factors based on analytical benchmarks. Previously developed validation methodology and procedures for estimating modeling errors and uncertainties [11] were extended to include the option of use of corrected solutions. The validation approach [9] has been shown successful in establishing intervals of V&V for RANS simulations for ship hydrodynamics through its use at recent Gothenburg 2000 Workshop on CFD in Ship Hydrodynamics [12-19]. A shortcoming of this and other V&V approaches is that the justification for uncertainty estimates at 95% confidence level is based on reasoning similar to that used for EFD bias uncertainties at the 0-order-replication level without additional statistical 1-order-replication level precision uncertainties, which in combination provides N-order-replication level and increased confidence for EFD uncertainty analysis. As with V&V there are many viewpoints on certification

[2]. Certification is defined as a process for assessing probabilistic confidence intervals for CFD codes for specific benchmark applications and certification variables. Presumably, range of applications requires interpolation and extrapolation methods. The approach combines previous V&V approach with extensions of concept of N-version testing for consideration bias uncertainties and use of reference values (experimental data and uncertainties) for estimating interval of certification. The present paper summarizes and combines the systematic check and verification of the SHIPFLOW CFD code approaches, including some findings about the uncertainty estimates. Opportunities and challenges for achieving consensus and standard V&V and certification methodology and procedures are discussed. 2. GENERAL FEATURES OF THE SOLVER A zonal approach is used in SHIPFLOW to compute the flow around the ship hull. The flow domain is divided into three zones, and a computational method is developed for each zone. The first zone covers the entire hull and a part of its surrounding free-surface. A free-surface potentialflow method of Rankine-source type is used. The second zone is a thin layer at the hull surface and a boundary layer method of the momentum integral type is used. The momentum integral equations are solved along streamlines traced from the potential flow solution. Finally, the third zone includes the aft part of the hull and extends about half a ship length downstream of the hull. It also covers about half a ship length in the radial direction. A NavierStokes method of the RANS type using the k model and a wall-law is used in zone three. The zones are computed in sequence and boundary conditions are generated for succeeding zones. The reason for the division of the flow field into zones is that the computational time may be reduced considerably compared to the global approach where the Navier-Stokes method is used in the entire computational domain. 3. RESULTS AND DISCUSSIONS Benefiting of a powerful computational tool, systematic computations were performed on a 10000 DwT LPG ship to determine the optimal hull forms from the resistance reduction point of view. The computational method is based on a hybrid BEMRANSE method that employs the zonal approach. The panels were distributed on the hull, as Fig, 1 shows, according to the lines plan provided by ICEPRONAV Galati. The numerical investigation

has been aimed at clarifying the influences that the numerical parameters may have on the accuracy of the computed solution. 41 computational cases were carried out for nine different speeds within a range of 4 knots between 14 and 18. Although five speeds are usually enough to determine the ship resistance, the research has been extended to a larger amount of pilot computations for the sake of clarifying completely the dependence between the flow characteristics and the physical, numerical and geometrical parameters, which were successively modified. To help the reader get a complete understanding of the aforementioned dependencies, the features of all the computational cases are tabulated in Table 1.
Z Y X

pu tat io

Computational parameter x/Lpp upstream Hull stations 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 30 35 40 45 49 50 52 55 48 x/Lpp downstream Speed [knot] 16 16 16 14-18 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 14-18 y/Lpp 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 0.5 0.6 0.7 0.8 0.9 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0

Vedere din prova

Vedere din pupa


Z X Y

Figure 1. Panel distribution on the hull Knowing the importance of the computational domain size on the overall accuracy of the numerical solution, we considered as appropriate to perform systematic computations aimed at clarifying the influence of each numerical parameter in the simulation process. First, the influence of the computational domain size at the downstream was studied. The theory of the numerical simulation of the free-surface flow suggests a minimum of 30 panels over a wave length as well as a minimum of two significant waves to be accommodated behind the hull. Under these circumstances, we considered six different domain sizes at the downstream ranging between 5 Lpp and 3.5 Lpp (computational cases No. 13, 8, 13 and 14). All the simulations were performed at 16 Knot, which represents the required nominal speed of the ship. Figs. 2 and 3 depict comparisons between the free-surface topologies computed for the domains with 1.5Lpp and 3.5 Lpp respectively 2.5 Lpp and 3.5 Lpp dimensions measured from the ship fore perpendicular. Since the domain length is expressed from the fore perpendicular, the real size of the downstream domain is smaller with a ship length, as the figures mentioned above show. Table 1. Computational parameters for the flow simulation

1 2 3 4-12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33-41

1.5 1.75 2.0 2.5 3.0 3.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5

0.7 0.7 0.7 0.7 0.7 0.7 0.25 0.3 0.4 0.5 0.6 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7

In the case of the smallest x/Lpp, the computational domain accommodates a wave only, the second being truncated around its half, whereas for the largest downstream domain, the maximum wave numbers is seven. Taking into account that to avoid the unwanted numerical interference between the hull and outflow boundary conditions induced by the Sommerfeld radiation condition (although Shipflow uses the Orlanski scheme) it is necessary to have a sufficiently extended domain at the downstream to allow the development of minimum two wave crests, as stated before. For this reasons all the following computations will be performed on a domain extended downstream till x/Lpp=2.5, measured in respect to the fore perpendicular.

Figure 2. Comparison between the free-surface topologies computed on domains of 1.5Lpp and 3.5Lpp at downstream

Figure 3. Comparison between the free-surface topologies computed on domains of 2.5Lpp and 3.5Lpp at downstream Fig.4 shows a comparison between the wave cuts drawn on the ship hull for a domain extending at the downstream between 1.5Lpp and 3.5Lpp in the six computational cases (1...4, 13 and 14) discussed. Although the variation of the domain size is significant, the wetted hull surface profiles are not revealing major differences. This fact may lead to the conclusion that the difference between the computed wave resistances is not significant either. This is valid only for the waves profiles on the hull, because if Fig.5 is analyzed, it may easily seen that thing change significantly behind the hull, where an important drop of the wave magnitude takes place as the computational domain increases in length. Moreover, the wave computed for the 3.5Lpp domain, (computational case 14), reveals a small phase shift. This fact is explained by a possible dissipation produced by the larger panel size at the

downstream, knowing that a rougher paneling induces a numerical damping of the computed solution. The computed values of the wave coefficient are tabulated in Table 2 for the six cases discussed above. Since the computational domain extends from 1.5Lpp (corresponding to the computational case 1) to 3.5Lpp (which corresponds to case 14), the numerical solutions are reported to that computed for the domain of 2.5Lpp in size at the downstream. The situation changes when the wave cuts on the ship hull are considered. As it may be seen in Fig.9, which presents a comparison between the wave profiles computed on the on the sip hull for all the upstream dimension between 0.25Lpp and 0.7Lpp, the magnitude of the first wave crest increases with the decrease of the upstream domain. Same conclusion applies to the second crest as well as to the following pseudo-crest. From the tabulated data it may be seen that the wave resistance coefficient decreases with the downstream domain size increase. It is important to observe that when the downstream extension decreases from 2.5Lpp to 1.5Lpp, Cw increases by 7.6%, whereas the downstream size increases form 2.5Lpp to 3.5Lpp, the decrease Cw is only 1.47%.
0.015

h/L
x/L downstream= 1.50 x/L downstream= 1.75 x/L downstream= 2.00 x/L downstream= 2.50 x/L downstream= 3.00 x/L downstream= 3.50

0.010

0.005

0.000

-0.005

-0.010 0.0

0.2

0.4

0.6

0.8

x/L

1.0

Figure 4. Comparison between the wave cuts on the hull computed on domains extending at the downstream from 1.5Lpp to 3.5Lpp

Table 2. Variation of the wave coefficient with the downstream size of the computational domain 1.5 1.75 2.0 2.5 3.0 3.5 x/Lpp 3 0.620592 0.595788 0.585007 0.581786 0.579638 0.573202 Cw10 Cw variation +7.6% +2.41% +0.55% 0 -0.37% -1.47% Table 3. Variation of the wave coefficient with the upstream size of the computational domain -0.25 -0.30 -0.40 -0.50 -0.60 x/Lpp 0.690257 0.660760 0.637831 0.603862 0.597301 Cw103 Cw variation +17.99% +12.95% +9.03% +3.22% +2.11% -0.70 0.585007 0

Based on this finding, we again came up to the decision that all the following computations in the present study be performed on a computational domain of a size at the downstream of 2.5Lpp, measured from the fore perpendicular.
0.015

h/L
x/L downstream= 1.50 x/L downstream= 1.75 x/L downstream= 2.00 x/L downstream= 2.50 x/L downstream= 3.00 x/L downstream= 3.50

0.010

0.005

0.000

-0.005

-0.010

domain, the wave coefficient increases significantly. It is important to point out that in the case of 0.25Lpp the difference between the wave coefficient and that corresponding computed for 0.7Lpp, is almost 18%. For larger sizes of the upstream the gap starts to decrease to 2% approximately, a value that corresponds to the 0.7Lpp case. Based on this finding all the following computations will be carried out for a domain having a size at the upstream equal to 0.7Lpp, measured form the fore perpendicular. Next, the influence of the lateral size of the computational domain on the wave coefficient is checked. Six domain breadths varying from 0.5Lpp and 1.0Lpp are considered (computational cases 20...24 and again 8 in Table 1). All the simulations are performed for the nominal speed of the ship.

0.0

1.0

2.0

3.0

x/L

Figure 5. Comparison between the wave profiles computed on a domain having the downstream size between 1.5Lpp and 3.5Lpp In the followings, the influence of the free surface extension at the upstream is studied. Six different computational domains extended in front of the ship from 0.25Lpp and 0.7Lpp (computational cases 15...19 and 8 in Table 1) are considered. All the simulations were performed for the nominal speed. Figures 6 and 7 show comparisons between the free-surface topologies computed for the domain sizes of 0.25Lpp and 0.7Lpp, respectively of 0.5Lpp and 0.7Lpp measured from the ship stem. As expected, the two figures prove that the variation of the domain size at the upstream does not determine significant modifications of the wave pattern at the downstream. Same conclusion may be withdrawn based on the analysis of Fig.8, which shows a comparison between the wave profiles computed for all the six upstream dimensions of the free-surface, ranging from 0.25Lpp and 0.7Lpp. A positive aspect is that regardless the domain size at the upstream the wave train does not modify the phase. Data in Table 3 prove that with the decrease of the upstream

Figure 6. Comparison between the free-surface topologies computed on domains of 0.25Lpp and 0.7Lpp at the upstream

Figure 7. Comparison between the free-surface topologies computed on domains of 0.5Lpp and 0.7Lpp at the upstream 1.00 0.5850070 0 55 0.589905 +0.83%

Table 4. Variation of the wave coefficient with the lateral size of the computational domain 0.50 0.60 0.70 0.80 0.90 y/Lpp 0.5329403 0.5230703 0.5154429 0.5282885 0.5568385 Cw103 Cw variation +9.77% +11.84% +13.96% +10.74% +5.06% Table 5. Variation of the wave coefficient with the panel resolution No. of stations 30 35 40 45 0.536393 0.550575 0.565203 0.573871 Cw103 Cw variation -8.31% -5.88% -3.38% -1.90% 50 0.589124 0.7%

Table 6. Variation of the wave coefficient with the ship speed Speed 14 14.5 15 15.5 16 0.329 0.3312 0.4513 0.5673 0.5845 Cw103
0.015

16.5 0.6255

17 0.7192

17.5 0.8382

18 1.2124

h/L
x/L upstream = -0.25 x/L upstream= -0.30 x/L upstream= -0.40 x/L upstream= -0.50 x/L upstream= -0.60 x/L upstream= -0.70

0.010

0.005

0.000

approximately, which corresponds to the 0.7Lpp case. For wider domains, the value of this gap decreases gradually to about 5%, a value that corresponds to the 0.9Lpp case. Based on this finding, the following simulations are performed on a domain of 1.0Lpp in breadth. It is worth mentioning that the size of 0.7Lpp which determines the smallest wave resistance coefficient represents the default value of the Shipflow.
0.015

-0.005

h/L
y/L = 0.50 y/L = 0.60 y/L = 0.70 y/L = 0.80 y/L = 0.90 y/L = 1.00

-0.010 -0.8 -0.6 -0.4 -0.2 0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

x/L

2.6

0.010

Figure 8. Comparison between the wave profiles computed on a domain having the upstream size between 0.25Lpp and 0.7Lpp
0.015

0.005

0.000

h/L
x/L upstream= -0.25 x/L upstream= -0.30 x/L upstream= -0.40 x/L upstream= -0.50 x/L upstream= -0.60 x/L upstream= -0.70

-0.005

0.010

-0.010
0.005

0.0

1.0

2.0

x/L

0.000

Figure 10. Comparison between the wave profiles computed on a domain having the lateral size between 0.5Lpp and 1.0Lpp In the following, a study on the influence of the panel number on the flow solution is proposed. Nine computational cases are proposed to which corresponds a hull discretization ranging from 33 to 55 stations (cases 25...32 and 8 respectively, in Table 1). 48 panels correspond to the case 8. All the simulations were performed for the nominal ship speed.
0.015

-0.005

-0.010 0.0

0.2

0.4

0.6

0.8

x/L

1.0

Figure 9. Comparison between the wave cuts on the hull computed on domains extending at the upstream from 0.25Lpp to 0.7Lpp Figures 10 and 11 present comparisons between the wave profiles computed in all the six lateral domain sizes. As it may be seen in the aforementioned figures, the variation of the lateral dimension of the domain brings forth significant changes of the wave patterns at the downstream, a fact that was somewhat expectable to happen. Again, it is worth mentioning that modifying laterally the domain size does not lead to a phase shift or to a wave-length modification. Figures in Table 4 prove that with the decrease of the lateral size of the domain a decrease of the wave coefficient is obtained. It is important to point out that the difference between the computed C w for the 0.5Lpp case and that corresponds to the 1.0Lpp case is of about 10%, then increasing to 14%

h/L
y/L = 0.50 y/L = 0.60 y/L = 0.70 y/L = 0.80 y/L = 0.90 y/L = 1.00

0.010

0.005

0.000

-0.005

-0.010 0.0

0.2

0.4

0.6

0.8

x/L

1.0

Figure 11. Comparison between the wave cuts on the hull computed on domains extending laterally from 0.5Lpp to 1.0Lpp

Fig. 12 bears comparisons between the freesurface topologies and between the pressure fields on the fore ship extremity computed for 30 and 48 panels over the ship hull. Although the plots do not emphasize major differences between the two numerical solutions, they exist, as the Fig.13 shows. Fig.13 presents a comparison between the wave cut profiles plotted on the hull based on computations over a hull paneled with 3055 stations. Even though the free-surface profiles display several differences on the hull, they vanish almost completely at the downstream, as Fig.14 bears out. Data in Table 5, which presents the variation of the wave coefficient with the panel resolution, were reported to the standard case computation with 48 panels on the hull (case 8 in Table 1). It is worthy to point out that when the hull panel number decreases, the wave coefficient decreases as well. In the case of a discretization with 30 panels, the difference between the wave coefficients is more than 8%, going down to 1.9%, for a number of 45 panels, then changing the sign till 0.83% for 55 panels on the hull. Because of that, the final conclusions of the present study will be drawn for the computed solution over a hull discretized with 48 panels. Based on the previous discussions, having the computational domain size previously investigated, establishing the discretization parameters (panel number, panel minimum size), a complete set of computations were performed for a speed range between 14 and 18 knots, with a modification step of 0.5 knots. The aim of these simulations is to establish the wave coefficient variation, see Table 6, for the optimum numerical treatment. Fig. 15...17 show the free-surface topologies computed for the nominal speed, in case 8 (Fig.15) compared to cases 4 (Fig.16) and 12 (Fig.17).

0.015

h/L
30 stations 35 stations 40 stations 45 stations 50 stations 55 stations

0.010

0.005

0.000

-0.005

-0.010 0.0

0.2

0.4

0.6

0.8

x/L

1.0

Figure 13. Comparison between the wave cuts on the hull computed for hull discretization with 30...55 de stations
0.015

h/L
30 stations 35 stations 40 stations 45 stations 50 stations 55 stations

0.010

0.005

0.000

-0.005

-0.010 -0.8 -0.6 -0.4 -0.2 0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

x/L

2.6

Figure 14. Comparison between the wave profiles computed for hull discretization with 30...55 de stations The free-surface flow theory prescribes the wave length as being: = 2LPP Fn 2 . Applying it to the ship speeds of 14, 16 and 18 knot respectively, for which the wave profiles were plotted, one may see the good agreement between the computed freesurface and that recommended by the wave theory. This can also be seen in Fig.18, which depicts a comparison between the wave profiles computed in computational cases 4, 8 and 12, therefore for the three speeds mentioned above. All the previous computations were performed on a uniform panel distribution. A last set of computations to which a hull and free-surface nonuniform discretization is further proposed. The aim of such purpose is to investigate the influence on the numerical solution of the panel clustering at the ship extremities, therefore in the areas where the hydrodynamic parameters are expected to display important gradients.

Figure 12. Comparison between the free surface topologies computed for hull discretization with 30 and 48 stations

Figure 15. Free-surface spectra computed in case 8

3341 in Table 1 and they correspond to the same speed range of 1418 knots, the speed step being again 0.5 knot. The numerical solutions are presented in the following figures. Thus, Fig.20 shows a comparison between the free-surface topologies computed in the cases 33, 37 and 41. Comparing them to those presented in Fig.15, 16 and 17, one may see that the waves are higher in magnitude.

Figure 16. Comparison between the free-surface spectra computed in cases 4 and 8

Figure 19. Comparison between the uniform and non-uniform panel distributions in the free-surface This fact will determine to a higher wave resistance, as the Table 7 and the Fig.21 prove. Fig.21 depicts a comparison of the wave-resistance graphs determined in the computational cases 412, and 3341, respectively. As it may be seen in the figure, the difference between the two sets of solutions is significant, a fact that may only be clarified after the experimental validation of the numerical simulation. 4. CONCLUSIONS The paper presented a methodology for computing the 3D incompressible free-surface flow around a 10000 DwT LPG ship. The computational method was based on a hybrid BEM-RANSE method that employs the zonal approach. A great deal of numerical simulations was performed to clarify the influences of the numerical and geometrical parameters on the overall accuracy of the Shipflow solver. ACKNOWLEDGMENTS The reported research could not be possible without the generosity of the Romanian National University Research Council (CNCSIS) which is greatly acknowledged for the provided financial support through the Grant A_C code 33/2005.

Figure 17. Comparison between the free-surface spectra computed in cases 8 and 12
h/L
0.015 v= 14 Nd v= 16 Nd v= 18 Nd

0.010

0.005

0.000

-0.005

-0.010 0.0 1.0 2.0

x/L

Figure 18. Comparison between the wave profiles computed in cases 8 and 12 Fig.19 comparatively presents the two panel distributions. The computations performed on the clustered panel distribution were denoted by

Figure 20. Comparison between the free-surface topologies computed on uniform and clustered panel distributions on the free-surface Table 7. The wave resistance variation with the ship speed Speed 14 14.5 15 15.5 16 0.5564 0.60721 0.65092 0.70339 0.72575 Cw103
0.0020

16.5 0.75364

17 0.79931

17.5 1.30101

18 1.9697

Cw
No Panel Clustering With Panel Clustering

0.0015

0.0010

0.0005

v [Knot]
14.0 14.5 15.0 15.5 16.0 16.5 17.0 17.5 18.0

Figure 21. The wave resistance variation with the ship speed. Computations on uniform and clustered panel distributions on the free-surface REFERENCES 1. Roache, P. (1994), Perspective: A Method for Uniform Reporting of Grid Refinement Studies, ASME J. Fluids Eng., Vol. 116, pp. 405-413. 2. Roache, P. J. (1998), Verification and Validation in Computational Science and Engineering, Hermosa publishers, Albuquerque, NM. 3. Rahaim, C.P., Oberkampf, W.L., Cosner, R.R., Dominik, D.F. (2003), AIAA Committee on Standards for CFD-Status and Plans, AIAA-2003-0844, 41st Aerospace Sciences Meeting, Reno, NV

4. AIAA (1998), Guide for the Verification and Validation of Computational Fluid Dynamics Simulations, AIAA-G-077-1998. 5. Roy, C. (2004), Verification of Codes and Solutions in Computational Simulation, Proceedings CHT-04, Norway. 6. Roy, C.J., Blottner, F.G. (2003), Assessment of One- and Two-Equation Turbulence Models for Hypersonic Transitional Flows, Journal of Spacecraft and Rockets, Vol. 40, No. 3, pp. 313-325. 7. Freitas, C.J., Ghia, U., Celik, I., Roache, P., Raad, P. (2003), ASMEs Quest to Quantify Numerical Uncertainty, 41st Aerospace Sciences Meeting, Reno, NV 8. Casey, M., Wintergeste, T. (eds.), (2000), ERCOFTAC Special Interest Group on Quality and Trust in Industrial CFD Best Practice Guidelines, Version 1, published by ERCOFTAC. 9. Stern, F., Wilson, R.V., Coleman, H., Paterson, E. (2001), Comprehensive Approach to Verification and Validation of CFD Simulations-Part 1: Methodology and Procedures, ASME J. Fluids Eng., Vol. 123, pp.793-802. 10. Coleman, H.W., Steele, W.G. (1999), Experimentation and Uncertainty Analysis for Engineers, 2nd Edition, Wiley, New York, NY. 11. Coleman, H.W., Stern, F., (1997), Uncertainties in CFD Code Validation, ASME J. Fluids Eng., Vol. 119, pp. 795-803. 12. Wilson, R.V., Stern, F., Coleman, H., Paterson, E., (2001), Comprehensive Approach to Verification and Validation of CFD Simulations-Part 2: Application for RANS Simulation of a Cargo/Container Ship, ASME J. Fluids Eng, Vol. 123, Issue 4, pp.803-810. 13. Wilson, R., Stern, F., (2002), Verification and Validation for RANS Simulation of a Naval Surface Combatant, Standards for CFD in the

14.

15.

16.

17.

18.

19.

20.

Aerospace Industry, AIAA 2002-0904 Aerospace Sciences Meeting, Reno, Nevada. Simonsen, C., Stern, F., (2003), Verification and Validation of RANS Maneuvering Simulation of Esso Osaka: Effects of Drift, Rudder Angle, and Propeller on Forces and Moments, Computers and Fluids, 32, pp. 13251356. Simonsen, C. and Stern, F., (2003), Flow Structure Around an Appended Tanker Hull Form in Simple Maneuvering Conditions, 8th International Conference on Numerical Ship Hydrodynamics, Busan, Korea. Kim, J., Paterson, E., Stern, F., (2003), Sub-Visual and Acoustic Modeling for Ducted Marine Propulsor, 8th International Conference on Numerical Ship Hydrodynamics, Busan, Korea. Weymouth G., Wilson, R., Stern, F., (2003), RANS CFD Prediction of Pitch and Heave Ship Motions in Head Seas, 8th International Conference on Numerical Ship Hydrodynamics, Busan, Korea. Xing, T., Kandasamy, M., Wilson, R., Stern, F., (2004), DES and RANS of Unsteady Free-surface Flows, 42nd AIAA Aerospace Sciences Meeting, Reno, NV, Division for Fluid Dynamics. Larsson, L., Stern, F., Bertram, V., (2003), Benchmarking of Computational Fluid Dynamics for Ship Flows: The Gothenburg 2000 Workshop, Journal Ship Research, Vol. 47, No. 1, pp. 63-81. Stern, F., Wilson, R., Shao, J., (2003), Statistical Approach to CFD Code Certification (Invited Paper), AIAA 2003-410 Applied Aerodynamics Special Session on CFD Uncertainty, 41st Aerospace Sciences Meeting, Reno, NV

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Tom 52(66), Fascicola 3, 2007

3`nd Workshop on Vortex Dominated Flows Timisoara, Romania June 1 - 2, 2007

NUMERICAL FLOW INVESTIGATIONS OVER A CHEMICAL TANKER HULL


Ana-Maria TOCU, Ph.D.Student* Department of Ship Hydrodynamics Dunarea de Jos University of Galati Adrian LUNGU, Professor Department of Ship Hydrodynamics Dunarea de Jos University of Galati

*Corresponding author: 47 Domneasca Street, 800008,Galati, Romania Tel.: (+40) 236 495400, Fax: (+40) 236 495400, Email: ana.tocu@ugal.ro

ABSTRACT A numerical method has been employed for the prediction of the flow past a chemical tanker hull. The computations were performed with the SHIPFLOW code that uses the zonal approach concept. In the zonal approach, the flow over the fore part of the hull is computed with a coupled potential flow-boundary layer method while the flow over the aft part with a RANS method. In the present work the RANS solver CHAPMAN is used. In CHAPMAN, the convective fluxes are evaluated using the approximate Riemann solver of Roe discretization. For the evaluation of the diffusive fluxes, central differences around the cell face centers are used. Flux-correction is used to increase the accuracy to second order in regions of smooth flow. The discrete equations are solved iteratively with ADI solver. The tri-diagonal system contains the first-order Roe convective terms and the second order diffusive terms, while the second-order flux corrections are used as an explicit defect correction. Turbulence is treated with the k- SST model without wall functions. The free surface geometry is obtained as the potential-flow solution and it is kept fixed for the solution of the RANS equations. KEYWORDS Potential flow, boundary layer, RANSE, viscous turbulent flow. NOMENCLATURE LPP Length between perpendiculars B Moulded breadth Mean Draught T

Instantaneous velocity components in Cartesian directions ui Time average velocity components in Cartesian directions, U i Fluctuating velocity components Cartesian directions P Instantaneous pressure Time average pressure, P p

xi Ui

Cartesian coordinates

u i"

in

p" Ri ij

Fluctuating pressure Volume force Total stress tensor Density Dynamic viscosity Turbulent dynamic viscosity Kinematic viscosity, Kroneckers delta

T ij

Reynolds stress tensor Strain-rate S ij

free-surface, energy

Turbulent kinetic energy Specific dissipation of turbulent kinetic Dissipation of turbulent kinetic energy Normal to surface Parameter direction crossing the boundary Frictional resistance coefficient Viscous pressure resistance coefficient Viscous resistance coefficient Wave resistance coefficient Total resistance coefficient Form factor

ni

B
CF CPV CV CW CT K

General variable Time average of Time

1. INTRODUCTION The design of the ship affects directly the quality, performance safety and environmental elements. It has a significant impact on production, operation, and maintenance costs. The initial design stage is important for a successful design. It is essential to start with as optimal parameters as possible. By using information from a hull database it is possible to perform parametric hull optimization at a very early stage of the design process. For hydrodynamic assessment of ship hulls in the optimization process, CFD codes play an important role. Their use in the optimization chain requires that they are accurate, fast, reliable, and that they communicate with the other components in the chain. In the first stage, the potential flow codes have been used in order to provide immediate results. Due to the substantially higher computational effort associated with RANSE computations, these required a greater effort to achieve solutions that lend them to practical application in an optimization environment. In the present work, some numerical solutions of flow around a realistic hull form are described. The computations were performed for the ship having the main particulars: LPP=109.8m, B=18m, T=6.3m, D=8.3m, Speed= 12knot. The lines plan is depicted in Fig.1. First, the tanker has been the subject of the calculations carried out based on the potential flow theory, testing the sensitivity brought by the choice of panel distribution and panel type. Then, the tanker has been used for testing the boundary layer and the Navier-Stokes viscous solvers.

The potential-flow theory has been used to compute waves patterns, wave resistance, lift and induced resistance and to provide the input to the boundary layer method, which predicts not only the transition, but also the boundary layer parameters on the fore half of the ship. For the aft part of the ship, the viscous flow was predicted using the RANS code with boundary conditions defined by the potential-flow treatment as well as by the boundary layer computation. SHIPFLOW includes two RANS codes, and the one used in the present work is CHAPMAN. In CHAPMAN, the equations for the momentum, pressure and turbulence (with a k- SST turbulence model) are fully coupled and solved. Roe discretization is employed for the convective fluxes and central differencing around the cell face center is used for the diffusive fluxes. A flux correction with a min-mod limiter is used for increasing the accuracy to second order in regions of smooth flow. The main attention was given to the stern where the viscous effects are significant. Solving the RANS equations will give the time average velocity and pressure. Knowing the average will usually be enough because of the time fluctuating velocity and pressure, which, in general, are much smaller in amplitude. 2. NUMERICAL METHOD 2.1. RANS equations Even if the Navier-Stokes equations can be solved numerically by resolving all scales, turbulent flows require extremely dense grids for the smallest turbulent length scales to be resolved. Therefore, we have to rely on the Reynolds decomposition in the foreseeable future. The continuity equation states that mass is conserved:

1 p U i + =0 t xi

(1)

Incompressible flow is considered, that is why the changes in density are negligible. The continuity equation can be written:

U i =0 xi

(2)

The Navier-Stokes equations of motion can be written in the following form: Fig.1. Lines plan

U jU i ij U i + = Ri + t x j x j

(3)

ij is the total stress and for a Newtonian fluid can


be written:

Then, taking the time average of the Navier-Stokes equations (first moving all terms to the left hand side):

U jU i U i P + + Ri + t x j xi x j

ij = P ij + 2 S ij S kk ij
Sij is the strain-rate defined as:

1 3

(4)

S ij =

U j 1 U i + 2 xi x j

U jU i U i P + Ri + t x j x j x j

U U j i + x j xi

(5)
=

U U j i + = x j xi u u j i + x j xi

(13)

Skk in (4) is zero for incompressible flow

u j ui + u "j ui" ui R + p + i x j xi x j t

S kk =

1 U ki U k + 2 x k x k

U k = x = 0 k

(6)

The Reynolds averaged Navier-Stokes equations can be derived from (3) by splitting the instant velocity components, Ui, in time mean velocity, ui, and time fluctuating velocity, ui" :

The time averaged continuity equation and NavierStokes equations for incompressible flow can be written:

u i =0 xi
u (u u + u u ) 1 p u + u + =R
i j i " " j i

(14)

U i = U i + u i" = u i + u i"

(7)

The instant pressure, P, components are the mean pressure, p, and time fluctuating pressure, p:

x j

x j x j

xi

where

P = P + p" = p + p"
The time mean of a variable is defined as:

(8)

(15)

= lim

1 T 2T

dt

(9)

The following rules of averaging apply for any two turbulent quantities 1 and 2 :

1 = 1 + 1" ;

1" = 0 ; 1" 2 = 0 ;

1 = 1 ;
1 1 ; = s s

1 2 = 1 2 ;

(10)

1 + 2 = 1 + 2 ;

" 1 2 = 1 2 + 1" 2 ;

Taking the time average of the continuity equation gives:


U i U i u i = = =0 xi xi xi

(11)

Subtracting (11) from (2) gives that also the time fluctuating velocity fulfills the incompressible continuity equation:
u i" =0 xi

(12)

2.2. Turbulence models The most popular versions of two equation models are the k- model, where is the rate at which turbulent energy is dissipated by the action of viscosity on the smallest eddies, [1], and the k- model, where is a frequency of the large eddies [2]. The k- model performs very well close to walls in boundary layer flows, particularly under strong adverse pressure gradients. However it is very sensitive to the free stream value of and unless great care is taken in setting this value, unrealistic results are obtained. The k- model is less sensitive to free stream values but generally inadequate in adverse pressure gradients and so Menter [3] has proposed a model which retains the properties of k- close to the wall and gradually blends into the k- model away from the wall. This model has been shown to eliminate the free stream sensitivity problem without sacrificing the k- near wall performance. The performance of two-equation turbulence models deteriorates when the turbulence structure is no longer close to local equilibrium. This occurs when the ratio of the production of turbulence energy to the rate at which it is dissipated at the small scales, departs significantly from its

equilibrium value, or equivalently when dimensionless strain rates (i.e. absolute value of the rate of strain times k/) become large. Various attempts have been made to modify two equation turbulence models to account for strong nonequilibrium effects. For example, the so-called SST (shear stress transport) variation of the Menter model, [3], leads to marked improvements in performance for non-equilibrium boundary layer regions such as may be found close to separation. The turbulent kinetic energy is over-predicted in regions of flow impingement and reattachment leading to poor prediction of the development of flow around leading edges and bluff bodies. Kato and Launder [4] have proposed a modification to the transport equation for which is designed to tackle this problem. The SST version of Menter k- based also offers a considerable improvement. Therefore, in the present work, the kSST model is used having k model near the wall and a k model, transformed to resemble a k model, outside of this region. 2.3. Boundary conditions The boundary conditions for computing the flow within the solution domain are requiring the no-slip condition on the hull surface for the velocity, a Neumann condition for the pressure, while for k and the following Dirichlet conditions are imposed: k = 0 , = f (u ,...) . In the symmetry plane, zerogradient Neumann conditions are imposed for all the variables. At the upstream the oncoming flow velocity is supposed constant, as k and are, whereas the pressure is extrapolated with zerogradient. At the downstream, the velocity k and are extrapolated with zero-gradient, while the dynamic pressure has the zero value. 3. COMPUTATIONAL STRATEGY An offset file based on the initial lines plan, as shown in Fig.1, defines the geometry of the ship. The offset file has been prepared in order to discretize the hull for the numerical simulation. The potential solving module has been used on different Reynolds numbers and in different calculation grids cases, as Table 1 proves. For instance, for different panels number, at the velocity v=11 knot, the results show that the level of grid dependency is large. Table 1. Computed wave coefficient Total number of Wave resistance panels coefficient Cw -2 0.159610 2920 0.100110-2 3325 v = 11 knot

0.940610-3 0.934610-3 0.937510-3

3805 4285 4765

The potential-flow method is used to solve the free surface problem and to provide the input to the boundary layer method. Steady state incompressible flow in a coordinate system that moves with the body is assumed. The free-surface problem is nonlinear since the free-surface boundary conditions are nonlinear and must be satisfied on the initially unknown wavy free surface. The solution method for the non-linear problem is used to linearize the free-surface boundary condition around a known base solution and to solve the problem in an iterative manner. In each iteration the wave height is computed from the linearized dynamic free surface boundary condition. The present analysis describes the flow around a three dimensional body, with free surface, with transom stern, and with sinkage and trim, at multiple ship speeds. For thin turbulent boundary layer computations, the momentum integral equations for the boundary layer are solved along streamlines traced from a potential-flow computation. For the viscous calculations, velocity components are extracted from the boundary layer computation and outside the boundary layer from the potential flow solution. The turbulent quantities are computed from analytical formulas based on the velocity profile in the inlet plane. In addition, the equations for the turbulent quantities are solved assuming a zero velocity gradient in the main flow direction. This step is used to obtain a smooth distribution of the turbulent quantities. 4. GRID GENERATION A 3D grid is generated by using suitable parameters for the RANSE solver. The potential flow and the boundary layer solution are utilized for computing the initial flow profiles. Because the Chapman solver does not use wall laws, the height of the cells closest to the hull surface should be very thin. The grid generator will automatically choose a suitable height. One single structured grid having the inner surface fitted to the hull will be created. In order to generate this single block grid, the program computes the outer boundaries, always controlling the functions to achieve orthogonality. It is noted that the grid covers only the aft half on one side of the hull (see Fig.2).

Fig.4 Free surface topology

Fig.2 Three dimensional grid topology generated around the aft part of the hull 5. NUMERICAL SOLUTIONS One of the most important issue of each numerical simulation output is the detailed analysis of the physical parameters that describe the flow. Therefore, an introspection within those parameters fields is proposed. In Fig.3, the dynamic pressure coefficient computed for 12 knots is shown. The negative pressure field observed at the aft part in the region of the bottom, under the bulb stern, causes flow suction, which cumulates to that created by the propeller. This behavior is confirmed by the physics behind the phenomenon.

Fig.5 Wave profile non-linearly computed for a speed of 12 knot It is worth mentioning that when the speed exceeds 13 Knots, the flow becomes unstable downstream the hull, where the transom determines the generation of wave crests of a significant magnitude. The increase of the speed determines an increase of the wave length, as expected. Verifications confirmed that the computed solution agree with the theory, i.e., =2LFn2. Various simulations were performed to check the performances of the lines plan. For being more precise, non-linear computations were employed for finding out the distribution of the pressure and velocity components on the body. The resulted pressure field shown in Fig.6, can be later on used in the hull form optimization process. The boundary layer module predicts transition and boundary layer parameters on the forward half of the ship. In Fig.7 the values for the boundary layer thickness are depicted.

Fig.3 Dynamic pressure coefficient on the hull. Some other results from the potential flow module, including the free surface elevation and the wave profile, computed for the 12 knots velocity case are depicted in Fig.4 and Fig. 5, respectively.

Fig.6 Pressure distribution on the hull

structure and wave propagation (i.e. without numerical diffusion or dispersion). The validation status on non-linear free-surface potential flow codes for wave resistance is considered to be in a relatively satisfactory state. Realistic estimates of total resistance can be achieved using a combination of a non-linear potential flow solution for the free surface, which once converged that can be used as a fixed free slip boundary for the RANSE calculation. Prediction of stern wave patterns, particularly with respect to transom sterns, is still a subject of research, but for certain vessels, a satisfactory position has been achieved.

Fig.7 Boundary layer thickness Through an iterative process, the RANS solver provides pressure and velocity distribution on the hull, as may be seen in Fig.9. The grid used for viscous computations has 120x30x60 grid nodes in the longitudinal, transverse and normal direction, respectively (see Fig.8). The iso-lines of the axial velocity component for some arbitrary slices are also depicted in Fig.10. Resistance results obtained from the viscous calculations are shown in Table 2. Table 2 CF 1.56810-3 CPV 0.674810-3 CV 2.24310-3 CW 0.736310-3 CT 2.97910-3 K 0.395 6. CONCLUSIONS For the development of new hull forms it has become increasingly important to predict numerically the hydrodynamic properties of a new vessel at an early stage of the design process. Reliability and accuracy of CFD calculations are essential prerequisites for the improvement and, finally, for the optimization of hull forms. On the basis of numerical flow field analysis, the most promising design alternatives must be chosen before model tests are conducted. The potential flow discussed earlier, has demonstrated accuracy in both wave pattern

Fig.8 Pressure coefficient in the aft hull obtained with the RANSE solver

Fig.9 Velocity vectors and pressure distribution on the aft body. Viscous calculations.

Fig. 10 Iso-wakes at different planes ACKNOWLEDGMENTS The reported research could not be possible without the generosity of the Romanian National University Research Council (CNCSIS) which is greatly acknowledged for the provided financial support through the Grant A_C code 33/2005. All the numerical computations have been performed at the Department of Hydrodynamics from the Dunarea de Jos University of Galati, Faculty of Naval Architecture. REFERENCES 1. Launder, B.E., Spalding,, D.B., (1974) The Numerical Computation of Turbulent Flows, Computer Methods in Applied Mechanics, vol. 3, pp. 269-289,1974

2. Wilcox, D.C., (1993) Turbulence Modeling for CFD, California 3. Menter, F.R., Zonal Two Equation k- Turbulence Models for Aerodynamic Flows, In 24th Fluid Dynamics Conference, Orlando, AIAA paper-93-2906 4. Kato, M., Launder, B. E. , (1993)The Modeling of Turbulent Flow Around Stationary and Vibrating Square Cylinders, Proc. 9th Symposium on Turbulent Shear Flows, Kyoto, pp. 10.4.1-10.4.6 5. Chakravarthy, S.R. , Osher, S., (1985) A New Class of High Accuracy TVD Schemes for Hyperbolic Conservation Laws, AIAA paper-85-0363 6. Deng, G.B., Queutey, P. Visonneau, M., Three-Dimensional Flow Computation with Reynolds Stress and Algebraic Stress Models 7. Dick, E., Linden, J., (1992) A Multigrid Method for Steady Incompressible NavierStokes Equations Based on Flux Difference Splitting, International Journal for Numerical Methods in Fluids, Vol.14, 1311-1323 8. Hellsten, A., (1998) Some Improvements in Menters k- SST Turbulence Model, In 29th Fluid Dynamics Conference, Albuquerque, NM, AIAA paper-98-2554 9. Merci, B., Vierendeels, J. , Reimslaugh, K., Dick, E., (1985) Computational Treatment of Source Terms in Two-Equation Turbulence models, AIAA Journal, Vol.38, No.11, 2085-2093 10. Schweighofer, J., Regnstrom, B., Starke, A.R., Tzabiras, G., (2005) Viscous-Flow Computations of Two Existing Vessels at Model and Full Scale Ship Reynolds Numbers, International Conference on Computational Methods in Marine Engineering, Barcelona, 2005 11. Flowtech International AB (2006) SHIPFLOW 3.3 - Users manual 12. Patel, V.C., Chen, H.C., Ju, S., (1988) Ship Stern and Wake Flows: Solutions of the Fully-Elliptic Reynolds-Averaged NavierStokes Equations and Comparisons with Experiments, Iowa Institute of Hydraulic Research Report No. 323 13. Patel, V. C., RODI, W., Sheuerer, G.., (1985), Turbulence Models for Near Wall and Low-Reynolds Number Flows: a

14. Review, AIAA Journal, vol.23, n.9, pp.1308-1319 15. Flowtech International AB (2006), XCHAP - theoretical manual, 2005 16. Xing, T., Kandasamy, M., Wilson, R. and Stern, F., (2004) DES and RANS of Unsteady Free-surface Flows, 42nd AIAA Aerospace Sciences Meeting, Reno, Nevada, 5-8 Jan 2004, Division for Fluid Dynamics. 17. Larsson, L., Stern, F., and Bertram, V., (2000), Benchmarking of Computational

Fluid Dynamics for Ship Flows: The Gothenburg 2000 Workshop, Journal Ship Research, Vol. 47, No. 1, March 2003, pp. 63-81. 18. Stern, F., Wilson, R., and Shao, J., (2003), Statistical Approach to CFD Code Certification (Invited Paper), AIAA 2003410 Applied Aerodynamics Special Session on CFD Uncertainty, 41st Aerospace Sciences Meeting, Reno, Nevada, 6-9 January 2003

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Special issue

3rd Workshop on Vortex Dominated Flows Bucharest, Romania July 1-2, 2006

PERFORMANCES DATA OF PROPULSION SYSTEMS FOR HIGH SPEED SHIPS


Mihaela AMORARITEI, Lecturer*, Naval Hydrodynamic Department Dunarea de Jos University of Galati

*Corresponding author: 47 Domnesca Street, 800008, Galati, Romania Tel.: (+40) 236 495400, Fax: (+40) 236 495400, Email: mamor@ugal.ro P propeller R waterjet rotor 1. INTRODUCTION During the last decades, the interest for high speed ships grew significantly worldwide. Fast ships can offer important advantages in commercial, military and superyacht market. The tendency toward increasing the ships speeds demands new types of propulsion devices which operate satisfactorily at high speeds, with higher efficiency, low noise and vibrations. Propellers design for high speed ships meets difficulties mainly due to presence of cavitation on propeller blades. For high speeds ranging up 20-25 knots, propeller cavitation cannot be avoided and the problem is to select and design a propeller able to accommodate the detrimental effects of cavitations: loss of thrust, erosions, noise and vibrations. In these cases, high speed propellers can be categorized in cavitating and supercavitating according to the development of cavitation. As an alternative for propeller cavitation and vibrations problems, waterjet propulsion systems are now applied to a wide range of ships types. It is found to be more expensive and more complicated than a propeller. However, with a speed range above 45 knots, waterjet propulsion systems are serious competitors to propellers. Hybrid propulsions systems of waterjet and conventional fixed, controllable pitch propellers or contrarotating propellers combine the bests of both. The propellers are used for normal cruising speeds, while the combination of waterjet with the propellers is used for maximum speed requirements. [1,2]. The propulsion device must be selected very early in the high ship design process and the type of propulsor can have a strong impact on the ship design itself [3]. The propulsion device must be selected by carrying out a study that balances the

ABSTRACT

The paper is focused on performances evaluation of propulsion systems for high speed ships. First, a discussion of applications of propellers and waterjet systems for high speed ships in light of advantages and disadvantages of these propulsion devices is given. A condensed survey on high speed propellers problems is presented. Secondary, an application related to propulsion performances for a ship operating in the speed range of 30-38 knots is discussed. Alternatively, the results for cavitating propellers and axial jet propulsion systems are presented. KEYWORDS Propeller, cavitation, supercavitating propeller, surface-piercing propeller, waterjet. NOMENCLATURE D [m] diameter revolution rate n [s-1] V [knots] ship velocity [m/s] advance velocity Va [-] propeller blade area ratio Ae/A0 P/D [-] propeller pitch ratio z [-] number of blades [N/m2] total static pressure p0 [N/m2] vapour pressure of water pv q [N/m2] dynamic pressure Q [m3/s] volume flow rate [kW] delivery power PD [-] propulsor efficiency
=
p0 pv [-] q V [-] J= A nD

cavitation number propeller advance coefficient specified speed

n s = 3.65

Q1 / 2 H 3/ 4

Subscripts and Superscripts

propulsive efficiency, costs, low noise and vibrations level. The paper presents aspects of performances evaluation of propulsion systems for high speed ships. First, a discussion of applications of propellers and waterjet systems for high speed ships in light of advantages and disadvantages of these propulsion devices is given. A condensed survey on high speed propellers problems is presented. Different types of cavitation are mentioned and the detrimental effects on marine propeller performances are discussed: loss of thrust, erosion, noise and vibrations. Secondary, an application related to propulsion performances for a ship operating in the speed range of 30-38 knots is discussed. Alternatively, the results for cavitating propellers and axial jet propulsion systems are presented. 2. HIGH SPEED PROPELLERS Cavitation, noise and vibrations are the major problems of propellers for high speed ships. These problems are caused by a combination of the operating conditions: the high speed of the ship, the rotational velocity of propeller, propeller submergence depth, the large delivered power to the propeller and the non-uniform wake-fields in which the propeller operates behind the ship. Cavitation - the phenomenon that water changes its phases into vapor in flow regions with very low pressures - has detrimental effects on propeller characteristics: loss of thrust and torque, erosions, noise and vibrations. For high speed above 20-25 knots, suppression of cavitation becomes impossible, and it is necessary to select and design several screw type propellers able to operate at low cavitation numbers, with satisfactory efficiency, cavitation and vibration characteristics. Criteria must be given on the type of cavitation to avoid and amount of thrust breakdown to be allowed [4]. According to the nature of cavities and to the particular location on the propeller, different types of cavitation occur on marine propellers: - vortex cavitation: trailing edge vortex cavitation, local tip vortex cavitation, leading edge vortex cavitation, hub vortex cavitation, propeller-hull vortex cavitation, - sheet cavitation (face / back cavitation), - bubble cavitation, - cloud cavitation - root cavitation. A sketch of different types of cavitation on a marine propeller is presented in Figure 1 (taken from [5]). These types of cavitations have different appearances and undesirable effects: some types of cavitation have major contributions to hull vibrations or high frequency noises, other types

cause pitting of the screw metal and accelerate blade erosion. Any great extent of cavitation affects the flow around propeller blades, causes a change in the lift and drag of the propeller blade sections and as a result, the thrust, the torque and the efficiency of the propeller are reduced.

Figure 1. Cavitation types. [5] Vortex cavitation requires the presence of vorticity. A detailed discussion regarding distinction between trailing, local and leading edge vortex cavitation is presented in [6]. The trailing vortex cavitation incepts downstream of the blade tip and it is determined by the radial loading distribution. The local tip vortex cavitation occurs in the vortices shed in the vicinity of the blade tip and it is influenced by the tip geometry. The leading edge vortex cavitation is determined by the vorticity produced at the leading edge, ussualy at high leading edge loading conditions. The leading edge vortex cavitation may have an appearance which is much like sheet cavitation. The end of tip vortex cavitation is far downstream and gives an excellence visual demonstration of the contraction of the race column behind the propeller. A fully developed tip vortex cavitation is a major contributor to hull excitations. In what extent the vortex cavitation is important for rudder erosion and for pressure fluctuation on the hull represents an important subject of concern [6]. Hub vortex cavitation occurs in the vortices shed in the vicinity of the blades roots. Generally, the hub vortex does not contribute to hull vibration unless it is very strong and excites the rudder [7]. Propeller hull vortex cavitation is a special form of vortex cavitation which extends from the propeller to the hull of the ship, at low speed and high loaded conditions. The propeller hull vortex cavitation occurs when a strong wake peak interacts with the propeller and it causes irregular vibrations and an extremely high noise level. Sheet cavitation is characterized by a continuous liquid/vapor interface attached to the blade surface. Sheet cavitation usually begins at the leading edge

(on either the face or the back side of the propeller blade), when the pressure distribution has an adverse pressure gradient. This happens when the angle of attack derivates from the ideal angle of attack: if the angle of attack is negative, sheet cavitations occurs on the propeller blade face, if the angle of attack is positive, sheet cavitations occurs on the back of the blade. When the propeller operates in non-uniform wake behind ship, the sheet cavitation extent varies as the propeller rotates. This periodic growth and collapse of sheet cavitation with the blade positions called intermittent or unsteady sheet cavitation was indicated as the cause of large forces exciting highly undesirable hull vibrations. In bubble cavitation, isolated spherical cavities are formed in the fluid at points where the pressure falls close the vapor pressure, near the midchord or maximum thickness of the blade. These cavities grow larger, move downstream with the fluid and collapse as they enter a region of higher pressure. The impact of collapsing bubble is very high and the bubble cavitation is considered the principal cause of erosion. The cloud or mist cavitation usually occurs behind the collapse of the sheet cavitation and causes blade erosion. When the tip vortex cavitation and the sheet cavitation pass a strong wake peak, they may break up into cloud and very small bubbles which collapse separately while moving with the fluid. The spread of sheet cavitation over the blade and the occurrence of bubble cavitation are accompanied by loss of thrust [8]. The effect of cavitation on hydrodynamic performances of a propeller is that increased power and rotational speed are required without a commensurable increase in the speed of the ship. Some types of cavitation are mainly responsible for large forces exciting hull vibrations: suction side sheet cavitation on the blade, tip vortex and collapsing sheet cavitation off the blade, propeller hull vortex cavitation. From the point of view of pitting of the blades and erosion the following types of cavitations have been identified as major contributors: bubble and cloud cavitation, sheet cavitation and unstable tip vortex cavitation. Generally, all designers of screw propellers for low and moderate speed ships endeavour to keep the propeller free of cavitation: the small amount of cavitation that does occur will not affect performances nor cause noise, vibrations and erosion damage. This does not apply for high speed propellers, many of which operate under appreciable cavitating conditions [9]. A propeller belongs into the high speed category, when both the advance and rotational speed are so high and it is extremely difficult to avoid cavitation, even under

ideal inflow conditions [10]. The definition excludes the highly loaded merchant ship propeller [11]. For high speed propellers, the most important problem is to control the influence of cavitation rather than attempt to suppress its occurrence. It is known that sheet cavitation is further divided in partial cavitation and supercavitation. A partial cavity is a cavity that is shorter than the chord length of the blade. A supercavity is a cavity that is longer that the chord length of the blade. According to the development of cavitation and to the flow regime in which trey operate, the high speed propellers can be clasified in: o cavitating propellers: Newton Rader propeller, Gawn Burrill propeller, o supercavitating propeller: fully submerged supercavitating propeller surface - piercing propeller. An important question in the initial design is whether a cavitationg or a supercavitating propeller is the most desirable for specified operating conditions of a high speed ship. The performance characteristics of high speed propellers depend to a considerable degree to the proper combination of the advance coefficient J and the cavitation number at which the propellers have to work. The recommended field of applications for supercavitating propeller was given in Figure 2 taken from [12].

Figure 2. Chart of practical use of supercavitating propeller [12] Experimental results have indicated that in order to have satisfactory supercavitating operations, the cavitation number of the blade section at 0.7 relative radius should been less than 0.045, and if the propeller operates at moderate speeds (35 to 50 knots), this requires a high rotational speed [12] . In the design range 0.045<<1, any propeller designed to fully cavitation have a considerable lower efficiency than a cavitating propeller. If partial cavitation cannot be avoided it is preferable to use special propellers designed for

such conditions. At present, relatively a few cavitating propellers series have been developed and tested for a wide range of cavitation number. The cavitating propellers belonging to the Gawn Burrill and Newton Rader series are used for high speed crafts with a good efficiency during cavitating conditions. Generally, in subcavitating regime the efficiency is lower than for corresponding conventional propeller. The Gawn Burrill series consists of three bladed propellers with elliptical developed blade outlines and segmental blade sections (Figure 3.a). Gawn Burrill series were designed to cover cavitation number from about 0.5 to a value corresponding to atmospheric pressure, pitch ratios P/D = (0.6 2) and blade area ratios Ae/A0 = (0.51 - 1.18) [13, 14].

supercavitating propellers are characterized by a very sharp leading edge and a very blunt trailing edge (Figure 4).

Figure 4. Flow around blade section for supercavitating propeller Major advantages of supercavitating propellers in comparison with cavitating propellers include relatively high efficiency during fully supercavitating conditions and reduction in noise and blade surface erosion as a result of vapor cavities which originate at the leading edge of the blade, extend and collapse downstream of the blade trailing edge. The supercavitating propellers require appendages like shafts, rudder and shaftstruts which lead to significant drag penalty. The supercavitating propellers cannot offer advantages in noncavitating conditions and they are not suitable for running at lower speeds, for extended periods at off design conditions. Some solutions have been proposed to avoid the lower efficiency under partially cavitating conditions and to extend the supercavitating propellers applications to lower speed. One is to ventilate the cavity on the back of the propeller blade by introducing air, the second is to modify the back of the supercavitating section through a tripping wedge or a tripping wire so that separation of flow will occur (Figure 5). Other solution is the combination between supercavitating propellers for high speed operations conditions and subcavitating propellers for low speeds.

Figure 3. Blade sections for cavitating propeller (a) Gawn Burrill, (b) Newton-Rader Newton Rader Series consist on three bladed propellers designed from a minimum cavitation number = 0.25 and cover pitch ratios P/D from about 1.04 to 2.08, blade area ratios Ae/A0 from about 0.48 to 0.95, [13,15]. The blade section shape used in Newton Rader series has the leading edge modified to achieve freedom from face cavitation (Figure 3b). The accepted effects of cavitation on Gawn Burrill and Newton Rader propellers are the changes in propeller hydrodynamic characteristics (thrust, torque, efficiency). The thrust and torque coefficients are given as functions of cavitation number, the advance coefficient and main propeller geometry (P/D, Ae/A0). The data from the systematic tests of model propeller series in cavitating environment are presented in graphical forms (Figure 10,11) and in form of polynomials for mathematical description [13],[16]. Using the method of the multiple linear regression analysis for each of series, polynomial equations for the series hydrodynamic characteristics have been obtained and the propeller of optimum efficiency can be automatically estimated using computed codes based on standard series. For very high speed ships (40 knots and higher), significant cavitation and blade erosion cannot be avoided and it is then advantageous to consider the use of supecavitating propellers. In a supecavitating propeller, the cavity on the back of blade has spread until it covers the whole of the back and it collapses far downstream to the trailing edge. The blade section shape in supercavitating propellers differs considerable from the conventional propeller. The

Figure 5. Modified supercavitating blade sections : tripping wedge and tripping wire [17] To predict the supercavitating propellers performances in the initial design, series chart for two, three and four blades propellers have been developed. Estimation of the optimum diameterrotation speed relationship and selection of number of blades on the basis of stress and vibrations are two important problems in initial design of the supercavitating propellers. Improved methods based on circulation theory, non-linear cavity flow theory and boundary surface panel technique have been extended to evaluate the hydrodynamics performances of supercavitating propellers.

A special type of supercavitating propeller is the surface-piercing propeller also called partially submerged supercavitating propeller or ventilated propeller. In this case, the propeller shaft is just above the free surface and the supercavitating propellers operate at partially submerged conditions. The advantages of surface piercing propellers in comparison with fully submerged propellers include the reduction of blade surface erosion since they are not susceptible to cavitation. The main obstacle for high speed propellers: cavitation is replaced by ventilation - the phenomenon which refers to introduction of air into cavitation area due to the presence of the free surface. A surface-piercing propeller can have a large diameter; its size is not limited by the blade tip clearance from the hull and by the vessel draught. The absence of underwater appendages (shafts, struts) leading to reduction or complete avoidance of appendage resistance. The disadvantages of surface-piercing propellers are the large blade impact forces when propeller blades enter water. The unsteady forces on propeller blades create relatively high vibrations and strength problems and a suitable distribution of skew may be defined to reduce the fluctuations of unsteady forces and reduce vibrations. The thrust and torque depend on propeller immersion, shaft angle and vessels trim. Partially submerged propellers have also problems with operations at low speeds and the artificially ventilation is a solution. Blade sections for surface piercing propellers include: wedge shaped sections, wedge shaped sections with cupped trailing edge similar to those used in fully cavitating conditions and a patented diamond back shape.

Figure 7. Trans-cavitating propeller blade [19] Most of high speed propellers are mounted on inclined shaft. A shaft angle of 10 degrees in high speed craft is commonly used for a sufficient aft draught, to place the propeller below the hull. In this case, the propeller operates in oblique flow which induces a cyclic variation of angle of attack and changes in propellers hydrodynamic characteristics. The cyclic variations of angle of attack increases with the oblique flow angle and the advance coefficient, the thrust and torque fluctuations become larger, the cavitation phenomena are intensified and contribute to propeller hull pressure augmentation. The fluctuating parts of the hydrodynamic forces and moments induced by propeller acting in oblique flow are very important with regard to noise and vibrations. ITTC [22] specified that the thrust fluctuation for a single propeller blade on highspeed vessel can attain values of 80 to 100 percent of mean thrust of one blade, but are reduced remarkably with decreasing cavitation number. The model test confirmed the expected result that fluctuating forces are reduced in fully cavitating propellers. The propeller forces in oblique flow: (propeller thrust along the shaft, torque, transverse force due to the oblique inflow) must be taken into account in prediction of the high speed ships performances. 3. WATERJET PROPULSION As an alternative for propeller cavitation and vibrations problems, waterjet propulsion systems are now applied to a wide range of ships types and are becoming serious competitors to propellers. The waterjet unit is an impeller or pump inside the hull, which takes water from outside in through a duct, adds energy and provides a jet reactive thrust of high velocity water expelled through a nozzle. The advantages are: higher efficiency, lower ship resistance due to absence of underwater appendages like shafts, rudder and shaftstruts. The absence of underwater appendages makes waterjet an ideal solution for shallow water operations, without any limitation on the size of pump. No reversing gearbox is required, the excellent manoeuvrability at all ship speeds, the ability to accelerate, reverse

Figure 6. Surface piercing propeller blade section shape - diamond back [18] Even in high speed ships, there are cases when the local relative velocities along the propellers blade varies with radius and the cavitation number near the blade tip is to small for sub-cavitating conditions, yet that near the hub is to big for supercavitating conditions. Taking into account the advantages and the disadvantages of supercavitating propellers, a trans-cavitating propeller represents a new solution. A trans-cavitating propeller (Figure 7 taken from [19]), adopts non-cavitating sections near the root (domain A) and supercavitating blade sections near the tip (domain B). The expected efficiency has been obtained in the design conditions.

and crash-stop is much better as compared to propeller propulsion. Absorption of engine power is independent of ship speed and no hull vibration or high speed cavitation ensures comfort on board. Advantages of waterjets versus propellers are safety in vicinity of other crafts, protection of marine life and environments and an expected reduction in acoustic signature. There are some important disadvantages: it is found to be generally more expensive and the propulsor itself is more complicated that a conventional propeller. The waterjet propulsion unit has higher weight than other propulsion devices: the weight of the water in the system above the free surface must be included in the system weight. The waterjet propulsion unit occupies considerable space inside the ship and the impeller access for inspections and repair is poor. The components of a waterjet propulsion system are schematically presented by Kruppa in [10] (Figure 8).

ns<300, mixed flow pump: 300< ns<500, axial flow pump: 500< ns<1200. The stator takes rotation out of the water and increase the pressure downstream the pump. The jet reactive of high velocity water is expelled through a nozzle which converts the pressure into kinetic energy. The excellent manoeuvrability is accomplished by directing the system's exit flow into any direction. The waterjet propulsion system requires optimization from the inlet to the exhaust nozzle and a lot of research has been carried out to optimize the inlet and the duct shapes, both experimentally as well as numerically. Waterjets are ideal for speeds in excess of 35 knots, where high pump efficiency provides higher speeds with lower fuel consumption, with much lower noise and vibration levels. Above about 35 knots, the advantages of waterjet versus propeller are a higher thrust and a rising propulsive coefficient PC (Figure 9)[24]. Below 32 knots, the propeller has the advantage.

Figure 9. Propulsion coefficient waterjet versus propeller taken from [24] 4. APLICATIONS AND RESULTS The application is related to propulsion performances for a ship operating in the speed range of 30-38 knots, ship which can be fitted with twin propellers or twin waterjets. The main characteristics of ship are presented in Table 1, and the wake and thrust deduction fractions have been set to zero. Table 1. Ship main characteristics Speed of ship [knots] 30 Resistance of ship [kN] 212 Number of propulsors 2

Figure 8. Schematic waterjet propulsion system taken from [10] The inlet location depends on the hull and waterjet propulsion system configuration, avoiding larger inlet losses and susceptibility of pump cavitation. The flow through the inlet duct can be characterized by the Inlet Velocity Ratio defined as ratio of the ship speed and the average axial velocity across the pumps inlet. Different types of pumps can be used: axial flow, centrifugal or mixed flow and these pumps have maximum efficiencies at certain range of specified speed ns as function of the rate of revolutions n, the volum flow rate Q and the total head of pump H. The following range of ns specify the three types of pumps generally used: centrifugal pumps: 40<

34 240 2

38 285 2

First, ship propulsion performances with cavitating propellers have been investigated. High speed ships frequently use either of Gawn Burrill and Newton Rader propellers for which the hydrodynamics performances al low cavitation numbers are known. Regression polynomials for

these series have been used to design optimum propellers at the speed design (34 knots) and to analyse their hydrodynamics performances in off design conditions. The results regarding evaluation of propulsion performances of cavitating propellers are presented in Table2. The hydrodynamics characteristics of designed Gawn Burrill and Newton Rader propellers are presented in Figures 10,11. The results for axial jet propulsion systems are presented in table 3 [26]. The diameters of rotors and the rotational speeds have been selected according with relation between power and ships speed for the most often used waterjets sizes. Table 2. Propellers performances Gawn-Burrill DP=1.22, Ae/Ao=1.18, P/D=1.279 V 30 34 38 nP [rpm] 852 962 791 0.656 0.658 0.613 PD 2490 3185 4590 0.99 0.77 0.61 Newton Rader DP=1.2, Ae/Ao=0.95, P/D=1.337 V 30 34 38 nP [rpm] 813 892 1000 0.642 0.661 0.663 D PD 2546 3174 4199
KT ,10KQ 0.6 0.5 0.4 0.3 0.2 0.1 0 0.6 0.8 1 1.2 1.4 J

KT ,10KQ 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0.6

Newton Rader P/D=1.337, Ae/Ao=0.95


KT (1) 10KQ (1) KT (0.77) 10KQ (0.77) KT (0.61) 10KQ (0.61)

J 0.8 1 1.2 1.4

Figure 11. Open water characteristics of Newton Rader propeller Table 3. Waterjet performances [26] Waterjet V 28 32 DR 0.95 1.02 nR [rpm] 780 740 Q 9.42 11.38 0.615 0.648 PD 2330 2851

36 1.09 700 13.29 0.678 3488

Gawn Burrill P/D=1.279, Ae/Ao=1.18


KT (1) 10KQ (1) KT (0.77) 10KQ (0.77) KT (0.61) 10KQ (0.61)

The propulsion devices efficiency has been plotted in Figures 12. In this particular case, below 34 knots, the advantages of propeller over waterjet are higher efficiency and low costs. It can bee seen that the propulsive efficiency of the selected waterjets exceeds that of the designed propellers about 34 knots and above. The Gawn Burrill propeller suitable for 30 knots has a lower efficiency at 38 knots. The propellers required to run above about 36 knots may be supercavitating.
0.7

0.65

Figure 10. Open water characteristics of Gawn Burrill propeller

0.6 Newton Rader Gawn Burill Waterjet

0.55

0.5 26 28 30 32 34 36 38 40 V [knots]

Figure 12. Propulsion coefficient waterjet versus propeller taken from [24] 5. CONCLUSIONS Generally, the propeller design for a high speed ship is a compromise of many considerations. High speed ships frequently use Gawn Burrill and

Newton Rader propellers for which the hydrodynamics performances al low cavitation numbers are known. For high speed ships in range from 40 to 80 knots, supercavitating propellers are considered to be the most fuel efficient propulsive. Surface piercing propellers are recognized as an outstanding propulsion device for small high speed crafts of limited draught. With a speed range above 45 knots waterjet propulsion systems offer distinct advantages over other propulsion devices and they are applied for a wide variety of ships: fast ferries, patrol boats, hydrofoils, motor yachts and small pleasure crafts. The propulsion devices in service for high speed ships will be a subject to continuous developments, the tendency toward increasing the ships speeds demanding new types of propulsion devices capable to operate satisfactorily at high speeds, with higher efficiency, low noise and vibrations. ACKNOWLEDGMENTS The author is grateful to Mr. Stefan Totolici (Bureau Veritas Romania) and to Mr. Dan Micu (ICEPRONAV Galati) for their scientific advices and for permission to publish the results regarding waterjets propulsion performances. REFERENCES 1. ***(2005), Proceeding of the 24th ITTC vol.I, The Propulsion Committee Report 2. ***,Waterjet Propulsion Systems, LIPS JET WARTSILA 3. ***,(1992), "Marine Engineering", Editor Roy Harrington, Newport News Shipbuilding 4. Cummin, R.,A.,Morgan,W.,B., Propeller Design Aspects of Large, High Ships, 5. ***(2002), ITTC Recommended Procedures, Testing and extrapolations methods. Cavitation Descriptions and Cavitation Appearance. 6. Kuiper, G., (2001), New Developments around Sheet and Tip Vortex Cavitation on Ships Propellers, CAV2001 7. Wijngaarden,E., (2000), The propeller as a source of noise and vibrations, 34th Wegemt School, Delft University of Technology, 8. Comstock, J.P., Editor, (1967), Principle of Naval Architecture 9. OBrien, T.,P., (1962), Marine Screw Propellers 10. Kruppa, C., (1967), High Speed Propellers. Hydrodynamics and Design, A three day intensive course for engineers, The University of Michigam

11. Kruppa, C., Practical Aspects in the Design of high Speed Small Propellers, 3rd Lips Symposium, Netherlands. 12. Tachmindji, A.,J.,Morgan,W.,B., (19), The Design and Estimated Performance of a series of Supercavitating Propellers 13. Kozhukharov, P.,G.., Zlatev , Z.,Z.,(1983), Cavitating Propellers Characteristics and their use in Propeller Design, The High Speed Surface Craft Conference, London 14. Blount,D.,L., Hubble,N., (1981), Sizing Segmental Section Commercially Available Propellers for Small Crafts, Propellers 81 Symposium, Virginia 15. Newton, R.,N., Rader, H.,P., (1960), Performance Data of Propellers for High Speed Craft, Meeting of the Royal Institution of Naval Architects 16. Kozhukharov, P.,G., (1986), Regression Analysis of Gawn Burrill Series for Applications in Computer Aided High Speed Propeller Design, RINA , 17. Venning, E., Haberman, W., (1962), Supercavitating Propeller Performances, SNAME, vol. 79, New York, 18. Ghose J.P., Gorkarn R.P., (2004), Basic ship propulsion, Indian Institute of Technology, Kharagpur 19. Yim,B., Kim,K.,S., Ahn,J.,W., Lee,J.,T., (1998), Design of Trans-Cavitating Propellers and Performance Analyses of the Test Result, Journal of Ship&Ocean Technology, Vol.2,No.1 20. ***(2002), Proceeding of the 23th ITTC vol.I, The Propulsion Committee Report 21. Young, Y.,L., (2002), Numerical Modeling of Supercavitating and Surface-Piercing Propellers, Ph.D Dissertation, University of Texas 22. ***(1987), Proceeding of the 19en ITTC, The Propulsion Committee Report 23. Brandau, J.,(1967), Aspects of Performance Evaluation of Waterjet Propulsion Systems and a Critical Review of State-of-the Art, SNAME Advance Marine Vehicles Meeting, Virginia 24. Alexander,K., (1995), Waterjet versus Propeller Engine Matching Characteristics, Naval Engineering Journal 25. Kuiper,G.,(2000), Cavitation, 34th Wegemt School, Delft University of Technology, 26. Amoraritei, M., Totolici, St., Micu, D.(1997), Consideratii cu privire la propulsia cu elice si propulsia cu jet a navelor rapide, Academia Navala Mircea cel Batran, Constanta

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Special issue

Workshop on Vortex Dominated Flows Achievements and Open Problems Timisoara, Romania, June 1 - 2, 2007

SIMULATION OF THE SHIP STANDARD MANOEUVRES


Dan OBREJA, Assoc. Prof. Naval Architecture Faculty Dunarea de Jos University of Galati Liviu CRUDU, Lect. Naval Architecture Faculty Dunarea de Jos University of Galati Radoslav NABERGOJ, Prof. Department of Naval Architecture, Ocean and Environmental Engineering University of Trieste Sandita PACURARU (POPOIU), Assist. Lect. Naval Architecture Faculty Dunarea de Jos University of Galati

*Corresponding author: Dunarea de Jos University of Galati, Romania Domneasca Street, no.111, Zip code: 800201,Tel.: (+40) 236 495 400, Email: dan.obreja@ugal.ro [m] Draught at aft perpendicular TA ABSTRACT TF [m] Draught at fore perpendicular [s] Natural roll period T To obtain a safe ship with optimum hydrodynamic performances represents a major U [Kn] Ship speed concern in the ship design and research activity. [m3] Volumetric displacement The occurrence of dangerous situations may have undesirable consequences for ship safety and the 1. INTRODUCTION critical situations should be carefully investigated. Three main factors must be taken into account in Special interest is related to the development of new order to consider the ship hydrodynamic safety hydrodynamic tools in order to investigate the problem: hull forms, environmental condition and manoeuvrability performances of the ship. In this ship operation. In this respect, it is very important paper, the numerical and experimental analysis of for designer to use powerful hydrodynamic tools in the standard manoeuvres characteristics have been order to investigate the ship safety, starting with the carried out on a fishing vessel. The mathematical initial design stage. It is obvious that the analysis of model is based on a set of differential equations of the dangerous situations as well as a deeper ship motions, which allows the determination of understanding of the complex hydrodynamic ships trajectory as well as kinematical and phenomenon are necessary. dynamical parameters of motions. Further The ships are considered similarly to six degree of developments can be performed with potential and freedom mechanical systems. In this paper, special viscous models in ship manoeuvrability problems. interest is related to the development of hydrodynamic tools in order to investigate the KEYWORDS manoeuvrability performances of the ship. Ship manoeuvrability, simulation, model tests The ship manoeuvrability studies the ship horizontal motions. The manoeuvrability concepts NOMENCLATURE gathers distinct nautical performances [1], such as: AR [m2] Rudder area - course keeping (describes the performances of AWL [m2] Area of load waterline the ship to maintain the course direction); B [m] Moulded breadth - course changing (describes how fast can change D [m] Stock propeller diameter the course direction ); GMT [m] Transverse metacentric height - speed changing and stopping. kxx [m] Roll radius of gyration If the ship moves at full speed the turning kyy [m] Pitch radius of gyration parameters are essential. In order to avoid an kzz [m] Yaw radius of gyration obstacle it is indicated to change the initial course KG [m] Vertical centre of gravity direction then to try to stop the ship. Still, the LBP [m] Length between perpendiculars stopping performances are important when the ship Lmax [m] Length overall LCG [m] Longitudinal centre of gravity

is slow acting. At normal speed it is important to know the course keeping performances. The theoretical manoeuvrability models determines the investigation of the motion from horizontal plane, in time domain. The main difficulty regarding simulation of the ship manoeuvres consists in determining the hydrodynamic forces applied on hull. The investigations have often nautical constraints (shallow water and channels), but the free surface influence and the fluid viscosity are neglected. The implementation of a computer code capable to simulate the ship manoeuvres in different operating conditions constitutes an important task in our research activity. In order to apply theoretical calculations in different design stages, it is necessary to confirm the software validity by means of experimental tests. The program reliability was checked on a fishing vessel, typical for the Mediterranean Sea, which has been tested in the Ship Hydrodynamics Laboratories of ICEPRONAV Galati [2]. The characteristics of the ship and of the experimental model are presented in Table 1, while a body plan of the vessel is shown in Figure 1. Table 1. Full scale ship and model main characteristics at full loading condition Main Full Model characteristics scale scale (1/12) Lmax 32.7 m 2.725 m LBP 25.0 m 2.083 m B 8.0 m 0.667 m 296.0 m3 0.171 m3 TF 2.42 m 0.202 m TA 2.74 m 0.228 m LCG 11.32 m 0.943 m 3.05 m 0.254 m KG GMT 0.65 m 0.054 m T 6.2 s 1.8 s kxx 2.46 m 0.205 m kyy 6.78 m 0.565 m kzz 6.9 m 0.575 m AWL 163.74 m2 1.137 m2 1.8 m 0.15 m D 2 2.88 m 0.02 m2 AR 12 Kn 1.8 m/s U

Figure 1. Body plan of the fishing vessel 2. MATHEMATICAL MODEL Taking into account ships motion equations related to a mobile orthogonal coordinate system Oxyz, connected to the ship and using the impulse and kinetic moment theorems [3] for the masses mi, having the speeds vi and vector radii ri then

Fi =
i =1 N i =1

d (mi vi ) . dt
N

(1)

(M
i =1

+ ri Fi ) = ri
i =1

d (mi vi ) . dt

(2)

where, Fi and M i are the external excitation forces and moments. If v0 (u,v,w) is the ships speed in the origin of the coordinate system and (p,q,r) the angular speed, then vi = v0 + ri . (3) Considering that the origin of the coordinate system is in the ships gravity centre G(xG,yG,zG) and X, Y, Z are the components of the external forces, K, M, N the components of the external moment, m is the ships mass and Ixx, Iyy, i Izz represent the inertia moments then, the system of the differential equations of motions (1) and (2) get the following simplified form

u X = m + qw rv t v Y = m + ru pw t w Z = m + pv qu t

(4)

p I xx + rq (I zz I yy ) t q M = I yy + pr (I zz I zz ) t r N= I zz + pq (I yy I xx ). t K=

(5)

When studying ships manoeuvrability, mainly the horizontal motions are taken into consideration. The vertical motions can be neglected. However, for high speed ships, when a significant tilt angle can appear during the gyration manoeuvre, roll motion can be considered, being coupled with the horizontal motions. In such conditions, if the origin of the coordinate system is on the ships symmetry plan, the system of the differential equations of motions becomes

simplified mathematical formulation can be obtained by using Taylors series expansion method. If only first order terms are considered then the linear form is obtained. Considering, as initial conditions, a straight trajectory and a constant speed of the ship (u = U) and neglecting the high order terms of the (8) and (9) equations of motions (r2 and rv), the linear formulation of the motions in the horizontal plane is obtained

& = mu & X e + X u u + X u& u & + Yr& r &= Ye + Yv v + Yr r + Yv& v & + rU + r &xG ) m(v

(11)

& + N r& r &= N e + N v v + N r r + N v& v & + mxG (v & + rU ). I zz r X u , X u& , Yv , Yr , Yv& , Yr& , N v , N r , N v& and
as hydrodynamic derivatives, characterizing the interaction of the wetted surface of the ship and the water. The above mentioned system can have the equivalent form

u X = m rv r 2 xG + przG t dr dp v Y = m + ru + xG zG dt dt t K= p v I xx mz G + ru t t

N r& are the hydrodynamic coefficients, known also

(6)

r v N= I zz + mxG + ru . t t

(7)

Furthermore, neglecting the roll motion (p =0), the system of the differential equations of motions (6) and (7), has a simplified form

& = X uu + X e (m X u& )u & + (mxG Yr& )r &= (m Yv& )v Yv v + (Yr mU )r + Ye & + (I zz N r& )r &= (mxG N v& )v N v v ( N r mxGU )r + N e .

(12)

u X = m rv r 2 xG t dr v Y = m + ru + xG dt t
N= r v I zz + mxG + ru . t t

Mention should be made that the first equation (surge motion) is not coupled with sway and yaw ones. Noting the vector s = {v, r} the coupled system of equation can be written in a matrix form

(8)

(9)

mxG Yr& m Yv& & mx N s I N & & v zz r G Yr mU Yv = s + Fe . N r mxGU Nv

(13)

The principal sources of the hydrodynamic excitation forces and moments are due to the propulsion and steering systems (Xe, Ye, Ne) as well as those on the submerged body due to an imposed manoeuvre of the ship (Xr, Yr, Nr):

X = Xe + Xr

Y = Ye + Yr N = Ne + Nr .

(10)

Noting the left side matrix M and the right one P, then equation (13) becomes r & = Ps + F . (14) M s e M is the matrix of masses and inertia moments of the ship and can be always reversed. If M 1 is the inversely of matrix M, multiplying equation (14) by M 1 one obtains the equivalent form r & = As + BF (15) s e where

Due to the fact that the external excitation forces and moments are multivariable functions, a

A = M 1 P B = M 1 .

(16)

Matrix A characterizes the internal dynamics of the system while P represents the potential damping matrix. The course keeping stability of the ship can be performed based on the investigation of the stability of the solutions of the motion equations system, when the straight constant speed and no external perturbations hypothesis are considered (the homogenous form of system 14) r & = As . (17) s One obtains the stability criterion which can be written as C = Yv ( N r mxGU ) + N v (mU Yr ) > 0 . (18) The hydrodynamic derivative N v is the decisive factor on the ships course stability. A bigger value of the projected longitudinal wetted surface of the ship will lead to a bigger value of N v derivative which means a better course stability of the ship. However, mention should be made that excessive course stability will affect the other manoeuvrability qualities of the ship (like gyration). On the other hand, if the ship is unstable, the course keeping will require continuous corrections to be applied to the steering system. Consequently, a compromise has to be found in order to finally reach general good manoeuvrability qualities. The hydrodynamic excitation forces and moments due to the rudder Fe (Xe, Ye, Ze) can be written in a linear form

The radius of the circular trajectory can be calculated with

R=

U UC . = r ( N v Y Yv N )

(22)

increase) or the rudder deflection angle, , can be increased (up to the critical value from where the lift force and moment diminished drastically). The manoeuvres at high deflections angles of the rudder require the consideration of the nonlinear hydrodynamic terms and the nonlinear inertial components. This leads to the utilization of nonlinear hydrodynamic models, which include the high order terms of Taylors series expansion of the hydrodynamic external excitation forces and moments. The nonlinear model of ships manoeuvrability, developed by Strom - Tejsen, consider the following equations [4]

One may observe that the radius R and ships stability parameter C are directly proportional, then, an excessive stability will lead to a bigger radius. In order to improve the manoeuvrability qualities, like turning ability, the surface of the rudder can be increased (which means the value of N will

&= (m X u& )u

Xe = X

Yv v + (Yr mU )r + Ye + f 2 (u , v, r , ) & + (I zz N r& )r &= (mxG N v& )v

X u u + X e + f1 (u , v, r , ) & + (mxG Yr& )r &= (m Yv& )v

(23)

Ye = Y N e = N where, Y
and

(19)

N v v ( N r mxGU )r + N e + f 3 (u , v, r , )

derivatives due to the rudder action and is the rudder angle. The first stage of the rudder deflection is followed by a transitory domain when all parameters a time dependent and then, by the stabilized regime when the accelerations of ship motions become zero. Consequently, the coupled equations of motion (14) can be written as follows P s + Fe = 0 (20) and the components of the vector s = {v, r} become

are the hydrodynamic

where the functions f1, f2, and f3 contain the nonlinear hydrodynamic derivatives. Using the notations

Yv v + (Yr mU )r + Ye + f 2 (u , v, r , )

X u u + X e + f1 (u , v, r , ) f 2' (u , v, r , ) =

f1' (u , v, r , ) =

(24)

N v v ( N r mxGU )r + N e + f 3 (u, v, r , )
the nonlinear system of equations becomes:

f 3' (u, v, r , ) =

v=

[(mxGU N r )Y + (Yr mU )N ]
(N vY
Yv N ) .
(21)

& = f 1' (u , v, r , ) (m X u& )u & + (mx G Yr& )r & = f 2' (u , v, r , ) (m Yv& )v & + (I zz N r& )r & = f 3' (u , v, r , ). (mxG N v& )v

(25)

r=

In principle, the nonlinear mathematical model can be reduced to a set of three first order differential equations, which can be resolved using numerical methods:

du = f1" [t , u (t )v(t ), r (t ), (t )] dt dv = f 2" [t , u (t )v(t ), r (t ), (t )] dt dr = f 3" [t , u (t )v(t ), r (t ), (t )]. dt

(26)

nonlinear regression procedure, in terms of hydrodynamic derivatives. In Table 2 we show the experimentally based static and dynamic hydrodynamic derivatives of the fishing vessel, at Froude number Fn = 0.40. Table 2. Experimental hydrodynamic derivatives Value G105 Derivative

The numerical solutions for the unknown speeds u, v and r for the time step (t + t ) can be obtained knowing the values of the speed u, v and r for the time step t

m X u& X vv / 2 X / 2 X v m Yv& mxG Yr& Yv Yvvv / 6 Yv / 2

3970.0 -459.5 -185.5 225.3 4688.2 1085.5 -4539.4 -9689.2 621.4 -462.1 10117.0 624.7 -111.4 8861.2 52.0 -1042.0 499.7 -1793.6 405.1 -263.4 -569.3 -6776.0 -278.0 33.7 -3111.4 -17.4

& (t ) u (t + t ) = u (t ) + t u &(t ) v(t + t ) = v(t ) + t v

In order to find the solutions (27) the values of the functions u(0), v(0) and r(0) for the time step t = 0 have to be known. Calculating the values of the speeds u(t), v(t) and r(t) at each time step, one may calculate: - the instantaneous values of the heading angle (t + t ) = (t ) + t r (t ) (28) - the trajectory of the ship when the fix coordinate system O0x0y0z0 is considered

&(t ). r (t + t ) = r (t ) + t r

(27)

mu Yr Yrvv / 2
Y Y / 6 Yvv / 2 Y0 mxG N v&

t [u (t ) cos (t ) v(t ) sin (t )] y 0 (t + t ) = y 0 (t ) +

x0 (t + t ) = x 0 (t ) +

(29)

t [u (t ) sin (t ) + v(t ) cos (t )]


- the instantaneous radius of the trajectory

u 2 (t ) + v 2 (t ) R(t ) = r (t )
- the instantaneous values of the drift angle

(30)

iz N r& Nv
N vvv / 6 N v / 2 N r mxG u N rvv / 2 N N / 6 Nvv / 2 N0

(t ) = arctg

v(t ) . U

(31)

The accuracy of the solutions is depending on the time step value t . The practical solutions of the mathematical models require the determination of the hydrodynamic derivatives which can be done using both theoretical and experimental methods. 3. EXPERIMENTAL MANOEUVRING DERIVATIVES In order to determine the hydrodynamic derivatives, experimental static and dynamic tests with Planar Motion Mechanism (PMM) were performed on a typical Mediterranean fishing vessel model, at the full loading condition [5]. As a result of the experimental investigations, the non-dimensional hydrodynamic forces and moments are expressed by Taylor series through a

4. MANOEUVRES SIMULATION OF THE FISHING VESSEL A simulation software for prediction of the manoeuvrability performances was implemented on Windows platform, using the main particulars of hull, propeller and rudder as basic input data. The

simulated manoeuvres can be performed for deep, calm and unrestricted water. Based on the experimental hydrodynamic derivatives, the manoeuvrability performances of the Mediterranean fishing vessel for the standard turning circle, zig-zag and spiral tests were determined. The turning test is a very important standard manoeuvre, required by navigation authorities. The predicted turning circles for 35 o rudder angle (starboard) obtained with the simulation code are depicted in Figure 2 and the calculated characteristics are shown in Table 3. The IMO recommendations are fulfilled with a sufficient margin [6].

program. The correlation between simulation and experimental results is satisfactory.


Port ,(deg)

First overshoot angle

t(s)
Second overshoot angle (simulation) (experiment)

Stdb

Tactical diameter Transfer

Figure 3. Zig-Zag manoeuvre 10/10 Table 4. Predicted and measured characteristics of the 10/10 zig-zag manoeuvre

S tead

y turn

ing ra

Characteristics
Advance
dius

Predicted 5.4 26.4 12.0

Measured 6.9 7.5 12.0

Drift angle

First Overshoot Angle [deg] Second Overshoot Angle [deg] Time to Second Execute [s]

Figure 2. Predicted turning circle Table 3. Predicted characteristics of the turning test Characteristics = 35 (St) Advance / Ship Length 3.59 Transfer / Ship Length Tactical Diameter / Ship Length Steady Diameter / Ship Length Speed Loss 1.52 3.26 3.25 54%

The spiral test gives information regarding the course keeping performances. Figure 4 depicts the results of spiral experimental model tests and theoretical predictions obtained with the computer program. The diagram of the rate of change of heading with respect to rudder angle suggests that the course of the ship is stable from dynamic point of view.
8 . (deg/s) 6

The zig-zag test gives information regarding the ship's response of the rudder and its yaw checking ability. Most common are the 10/10 and 20/20 tests, which are recommended by IMO. The experimental zig-zag (10/10) tests were performed in still water [2] with a free, selfpropelled and remote controlled model, in order to measure the course stability and the inertia performances. Figure 3 and Table 4 show the comparison of zig-zag results from experimental model tests and predictions with the computer

4 2 Port 0 -30 -25 -20 -15 -10 -5 -2 -4 -6 -8 0 5 10 15 20 25 30 Stbd

(deg)

Simulation Experiment

Figure 4. Spiral test

5. CONCLUSIONS The paper presents some results of the simulations of the standard manoeuvres for a fishing vessel. The computer cod is using Abkowitzs nonlinear mathematical model in its Strom - Tejsen formulation. The hydrodynamic derivatives have been obtained experimentally using a planar motions mechanism (PMM). Static and dynamic tests have been carried out. The correlation between the theoretical and experimental values is satisfactory but can be significantly improved. The actual investigations will be continued considering the complete numerical approach. As a first stage the potential theory assumptions will be taken into account. Calculating the velocity potential functions and using Bernoullis equation, the pressures distribution can be obtained. The integration over the wetted surface of the ship will lead to the determination of the hydrodynamic coefficients as well as ships motions in the horizontal plane. In the second stage the viscous RANS methods will be considered and the interdependences between hull - rudder - propeller will be progressively taken into account. During manoeuvres oblique turbulent flow occurs and phenomena like flow separations and vortexes fields have a strong influences on ships behaviour. The utilization of viscous models is, for the time being, limited due to the complexity of the grid generation requiring strong computer capacities in order to obtain accurate numerical solutions. However, further efforts have to be dedicated to this purpose to be able to reproduce the real phenomena and to provide reliable inputs in the design process. REFERENCES 1. Bertram, V., (2000), Practical Ship Hydrodynamics, ButterworthHeinemann 2. Obreja D. (2001), Theoretical Formulation and Experimental Verification of Design Indexes for the Safe Operation of Mediterranean Fishing Vessel at Sea, Research Project Ac 4323, ICEPRONAV Galati 3. Triantafyllou M., Hover F. (2002), Maneuvering and Control of Marine Vehicles, Department of Ocean Engineering, Massachusetts Institute of Technology Cambridge, Massachusetts 4. Strom-Tejsen, J., Chislett, M.S. (1966), A Model Testing Technique and Method

5.

6.

of Analysis for the Prediction of Steering and Manoeuvring Qualities of Surface Vessells, Report No.Hy-7, Lyngby Obreja, D., Messina, G., Crudu, L., Nabergoj, R. (2001), Investigations of Hydrodynamic Safety of a Fishing Vessel, International Symposium Technics and Technology on Fishing Vessels, Ancona IMO Resolution A751(18) adopted on 4 November 1993, Interim Standards for Ship Manoeuvrability

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Tom 52(66), Fascicola 3, 2007

3`nd Workshop on Vortex Dominated Flows Timisoara, Romania June 1 - 2, 2007

NUMERICAL AND EXPERIMENTAL INVESTIGATION ON A TRACTOR TUG RESISTANCE PERFORMANCE


Florin PACURARU, Assist. Lect. Naval Architecture Faculty Dunarea de Jos University of Galati Dan OBREJA, Assoc. Prof. Naval Architecture Faculty Dunarea de Jos University of Galati

*Corresponding author: Dunarea de Jos University of Galati, Romania Domneasca Street, no.111, Zip code: 800201,Tel.: (+40) 236 495 400, Email: dan.obreja@ugal.ro

ABSTRACT The study of the flow around the ship represents an important step in designing an optimised hull form concerning the low resistance criterion. The prediction of the ship resistance based on the CFD methods is validated with experiment results carried out in a towing tank. The comparison between numerical and experimental results obtained for a Tractor Tug of about 28 m in length, with an extended skeg is presented in this paper. The model tests have been performed at Galati University Towing Tank. The main attention is directed to the large appendages drag problem. The numerical and experimental results with and without appendages suggests a very important increase of tug resistance, generated by the large skeg. KEYWORDS ship resistance, numerical methods, model tests, appendages influence NOMENCLATURE

ABBREVIATIONS app appendage w with w/o without 1. INTRODUCTION Solving problems of ship hydrodynamics took a lot of attention during the last few decades. This is because of its importance to the design of different ships and also, the physical phenomenon of ship motion is not yet fully understood due to its complexity. The efforts of an naval architecture are determined by decreasing ship resistance and implicit the energy consumption. The approaching way of solving the problem is based on boundary element method, also known as panel method. Numerical simulation technique of free surface fluid flow around the ship, consist of potential flow aproach. The problem, modeled by field equation (Laplace) and boundary conditions is solved by an integral method [1] in which the sources are distribuited partly on the hull, and partly on the free surface. The integration is not made on the entire domain, only on the surface boundary. That is why it will be discretized only the body and the free surface, action that reduce the pre-processing job even if the method accuracy is inferior to methods based on differential equations for viscous flow. The presented method is widely used in hydrodynamic research field. Since Dawson [2] first proposed the modified Rankine Panel Method based on the Gadd [3] method linearizing the free surface condition in terms of the double-model solution, the applications published so far have shown the significant advantage of his method in simplicity and generality. Improvements, modifications and numerical studies on the method have been carried out by many researchers. Xia [4] applied Dawsons method to solve the wave problem of a ship with a

[-] [-]

velocity potential double model velocity potential

perturbation velocity potential due to presence of free surface [m] wavelength [m] wave heigh

w [-]

Fn [m] Froude number RT [kN] ship total resistance g [m/s2] acceleration of gravity

transom stern and to treat the flow around configurations with or without a lifting surface. He also tried to solve the fully nonlinear ship wave problems through iterations with the small perturbation approximations being employed at each iteration step. Improvements of the technique were introduced in the project by Ni and Kim [5]. Thanks to their work non-linear calculations could be carried out for all kind of ships. Further improvements in stability and robustness were introduced by Janson [6], who also made a theoretical investigation of the damping and dispersion errors of simulation technique. The objective of this work is to clarify the phenomena related to free surface flow around the SKD tractor tug hull which is moving in still water. When a ship moves through an undisturbed free surface it generates a wave system (transverse and divergent subsystems), which contributes to the total resistance. 2. MATHEMATICAL MODEL Assuming that the ship hull advances in the undisturbed water with a constant velocity U . The potential flow assumption of inviscid and irrotational flow is made for the steady flow around the hull. Equation field and boundary conditions are expressed in terms of velocity potential. The problem is numerically solved making use of distribution of Rankine sources on the boundary surface. The ship is specified in a coordinate system having its x-axis parallel to the longitudinal direction, the y-axis to the direction of starboard and the z-axis pointing upwards. The coordinate system has the same speed as the ship but does not follow the ship movements such trim and sink. The incompressibility of the fluid is generally assume in all physical modelling of phenomena in which fluid is water and where the relevance of wave pressure is negligible. The governing equation in potential flow method is Laplace equation (1): (1) 2 = 0 derived continuity equation (2) for incompressible flows: V =0 (2) For inviscid, irrotational flow the velocity V = (u , v, w) can be expressed as a gradient of a scalar function named the velocity potential: V = (3) Linearity of Laplace equation offers the possibility to combine elementary solutions (sources, sinks, doublets) to arbitrarily complicated solutions. Thus we can consider the total velocity potential as a sum

of double model velocity potential and the perturbation velocity potential due to presence of free surface: (4) = 0 + w The potential is subject to the several conditions on the hull and free-surface boundaries. Upstream disturbance by a moving ship vanishes at infinity: (5) lim = 0
r

Body boundary condition requires that no fluid particle penetrate the hull surface: (6) n = 0 Kinematic boundary condition is the mathematical formulation of the physical condition that a particle at the surface should remain at the surface all the time. (7) x x + y y z = 0 on z = The other condition to be satisfied on the free surface comes from the fact that the pressure on the free surface must be equal to the atmospheric pressure, which we assume that is constant. This condition is derived from Bernoulli equation. 2 2 g + 0.5 x2 + y + z2 U = 0 on z = (8)

The free-surface problem described above it is difficult to solve since the free surface boundary conditions are non-linear and must be satisfied on the initially unknown wavy surface. The solution for this problem is to linearize the free surface boundary conditions around a known solution and solve the problem in a iterative manner. In the first iteration the problem is linearized around the double model solution and then around the solution obtained from previous iteration.

z=0 Finally the radiation condition must be imposed (enforced) to avoid upstream waves. The radiation condition can not be described by an exact mathematical expression. It has to be enforced by an exact mathematical expressions. It has to be enforced by numerical trick. Forces an moments, including wave resistance are computed by integrating pressure over the ship hull.

02l wll + 2 0l 0 ll wl + g wz = 02l 0 ll on

(9)

3. NUMERICAL SIMULATIONS Several sets of numerical simulation were performed to analyse the hydrodynamic performances of SKD 753 tractor tug hull see Figure 1, and to determine the wave and total resistance of the ship. The ship hull is characterized by a very small length-beam ratio and a hard chine

shape with two knuckle. The main characteristics of the ship are presented in Table 1.
Table 1 Main dimensions of SKD 753 hull

Main characteristics Length of waterline Length between perpendiculars Breadth of waterline Draught mean Longitudinal center of buoyancy Volumetric displacement Wetted surface (bare hull) Wetted surface appendages Block coefficient of

Symbol

Full scale ship

LWL
L BP

29.04 m 28.17 m 11.063 m 3.2 m 14.503 m 672.78m 3


384.22 m 2 67.84 m 2

surface domain of 0.5 ship length upstream, 2 ship length downstream, the width of the free surface is 1 ship length. The hull is dicretize using 1800 panels, Figure 2. The free surface has a panel distribution of 30 panels per wave length in longitudinal direction. The wavelength ( ) of the wave generated by the ship depends on the Froude number and can be calculated by formula:
= 2 L Fn2

(10)

B TM LCB

S
S APP CB CM

The number of the panels in the transverse direction has been chosen such that the aspect ratio of free surface panels to be no more than three, Figure 3. The quality of the panel distribution on the linear surface potential flow can be judged by the using the double body resistance (without free surface). The double body resistance should be zero since the body is symmetric and no lift is generated, the calculated resistance will not be zero because of numerical error (7).

Midship section coefficient Waterline coefficient

0,667 0,887 0,831

CW

Figure 2 Panel arrangement on hull

Figure 3 Panel arrangement on free surface As it can be observed in diagrams 2-4, where comparisons between measurements and computed results are depicted, show a good agreement for moderated Froude numbers [8],[9]. For higher Froude numbers the difference between computed and measured results is rising. The authors believe that the discrepancy reasons comes from the hard chine hull shape which determine the imminence of strongly non-linear phenomena that can not be captured by the potential approach. This kind of phenomena has been observed in experimental investigation, Figure 6.

Figure 1 SKD 753 tractor tug lines plan Experience with the method shown that a minimum of 25-30 panels per unit length are necessary to obtain a solution which gives a good representation of wave pattern. The calculation are done for

RT[kN] 120
RT computed

100 80 60 40 20 0 0.15

RT measured

Fn 0.2 0.25 0.3 0.35 0.4

RT[kN] 400 350 RT measured 300 250 RT computed 200 150 100 50 0 Fn 0.2 0.2 0.2 0.3 0.3 0.3 0.3 0.3 4 6 8 2 4 6 8

Figure 2 Comparison between computed and measured resistance w app in ballast condition
RT[kN] 180 160 140 120 100 80 60 40 20 0 0.15 Fn 0.2 0.25 0.3 0.35 0.4 RT measured RT computed

Figure 5 Comparison between computed and measured resistance w app in full load condition

Figure 3 Comparison between computed and measured resistance w/o app in full load condition
RT[kN] 250
RT measured

Figure 6 SKD 753 flow for Froude 0.39

200 150 100 50

RT computed

0 0.24 0.26 0.28 0.3

Fn 0.32 0.34 0.36 0.38

Figure 4 Comparison between computed and measured resistance w app in ballast condition Figure 7 Comparison between pressure fields for the w and w/o appendage for Froude 0.31 Figure 7 depicts the pressure contours on the hull surface computed for Froude number 0.31. It can be noticed, in figure mentioned above, the influence of

the appendage presence on body pressure distribution by comparing the two cases: hull without appendage and hull with appendage. A comparison between appendage and no appendage case is proposed in figure 8 to show the computed free-surface spectrum drawn for Froude number 0.31.

0.09 0.08 0.07 0.06 0.05 0.04


full load no app full load app

H/L

0.03 0.02 0.01 0

-0.01 -0.02 -0.03 -0.04 -0.5 0 0.5 1 1.5 2

X/L

Comparison between wave cut profiles drawn for the w and w/o appendage in ballast condition Figure 8 Wave pattern comparison for or the w and w/o appendage Two comparisons of the wave cuts along the centre line are presented in figures 10,11 to prove the influence of the appendage on the wave profile. The height of the stern crest rises with 13.3 % in first figure and with 17.6 % in second.
4. CONCLUDING REMARKS A numerical procedure has been employed to introspect the flow around SKD 753 hull. The comparisons have shown a good agreement with experimental results for moderate Froude numbers. For higher Froude numbers the experiment investigation have shown the appearance of strongly non-linear phenomena such as wave breaking and flow detachment which can not be captured by the potential methods. The cause of this phenomena are large appendages, the fore body lines plan with a high angle of waterplane entrance, the wave pattern interference phenomena. The influence of the appendage on ship resistance has been studied by computing two cases: hull with appendage and hull with no appendage. The large appendages represent a very important factor of the ship resistance increasing. ACKNOWLEDGMENTS The present work has been carried out for SHIP DESIGN GROUP Galati, Romania. The authors wish to thank the staff of these companies for all their support. The authors wish to express their thanks to the staff of these companies for all their support and to prof. Adrian Lungu, the Dean of the Naval Architecture Faculty from University of Galati, for his valuable comments and observations during the experimental model tests.

0.09 0.08 0.07 0.06 0.05 0.04


ballast no app ballast app

H/L

0.03 0.02 0.01 0

-0.01 -0.02 -0.03 -0.04 -0.5 0 0.5 1 1.5 2

X/L

Comparison between the wave-cut profiles drawn for the w and w/o appendage in full load condition

REFERENCES 1. Lungu, A., Raad, P., Non-linear Potential Free-Surface Flow Around the Series-60 Ship Hull, 1998 ASME Fluids Engineering Division Summer Meeting, June 21 - 25, 1998 Washington DC 2. Dawson, C., A Practical Computer Method for Solving Ship Wave Problems, 2nd International Conference on Numerical Hydrodynamics, Berkley, 1977 3. Gadd, G.E., A Method of Computing the Flow and Surface Wave Pattern Around Hull Forms, Transactions of the Royal Institute of Naval Architects, Vol. 118, 1976, pp. 207-215 4. Xia, F., Larsson, L., A calculation method for the lifting potential flow around yawed surface piercing 3--D bodies, In Proceedings of the 16'th Symposium on Naval Hydrodynamics, pp. 583--597, 1986 5. Ni,S.Y., Hifgher Order Panel Methods for Potential Flows with Linear and Non-linear Free Surface Boundary Conditions, Phd Thesis, Chalmers University of Technology, Gothenburg

6. Janson, C-E., Potential Flow Panel Methods for the Calculation of Free Surface Flows with Lift, Ph.D. Thesis, Chalmers University of Technology, 1997 7. Mierlo, K.J., Trend validation of Shipflow based on the Bare hull upright resistance of the Delft Series, Master Theses, Delft Univeristy of Technology, 2006 8. Obreja, D., Popescu, G.., Pacuraru, F., (2004) Resistance Tests Report. Tractor Tug, Research Project No. 344/2004, Dunarea de Jos University of Galati 9. Obreja, D., Pacuraru, F. Popescu, G., Investigation of the Appendages Influence on Tractor Tug Resistance Performance, Scientific Bulletin of the Politehnica University of Timisoara, Transactions on Mechanics, pp.129-134, 2006

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Special issue

Workshop on Vortex Dominated Flows Achievements and Open Problems Timisoara, Romania, June 01 - 02, 2007

ON THE STICK SLIP BOUNDARY CONDITIONS AT THE WALL OF MICROCHANNELS


Corneliu BALAN*, Prof. Hydraulics Dept., REOROM Laboratory Politehnica University of Bucharest Catalin BALAN, Ph.D. Student Hydraulics Dept., REOROM Laboratory Politehnica University of Bucharest Diana BROBOANA, Assoc. Prof. Hydraulics Dept., REOROM Laboratory Politehnica University of Bucharest Roland KADAR, Ph.D. Student Hydraulics Dept., REOROM Laboratory Politehnica University of Bucharest

*Corresponding author: Splaiul Independentei 313, 060042, Bucharest, Romania Tel.: (+40) 21 4029705, Fax: (+40) 21 4029865, Email: balan@hydrop.pub.ro

ABSTRACT The present paper investigates the influences of slip boundary conditions on the flow in a 2D channel, at small Reynolds numbers. The numerical simulations are performed for two types of wall slip conditions: zero wall shear stress, respectively imposed slip velocity. The results evidence that changes in boundary conditions induce qualitative modifications within the flow field. Therefore, the method to alternate hydrophilic with hydrophobic walls is a promising technique for improve mixing in microchannels. KEYWORDS Channel hydrodynamics, wall slip, vorticity number, vortex, separation point NOMENCLATURE Re [-] Reynolds number Wo [-] vorticity number w [Pa] wall shear stress vw [m/s] wall (slip) velocity 1. INTRODUCTION Multiple applications of microfluidics are today related to biofluids and their flows in channels. One major target of the Lab-on-a-Chip

studies is the control of hydrodynamics and mixing of fluids in microchannels of complex geometries, literature being abundant in studies dedicated to these subjects, for example: (i) hydrodynamic mixing in micro-bifurcations, [1], [3]; (ii) length stabilization of velocity at the entrance in a microchannel and vortical structures in contractions flows, [9]; (iii) slip at the wall of channels, [7], [12]; (iv) techniques to mix immiscible fluids at low Reynolds flow regimes in channels of different configurations, [8], [10], [13]; (v) developing direct measuring procedure of velocity field, [6]. A general review of microchannel hydrodynamics is presented in [11]. The main goal of the present study is to establish proper flow conditions to maintain an uniform distribution of cells (maximum dimension of 10 m) in a straight channel of 800 m width, 50 m height and 10 mm length, for a flow rate in the range 10 l/min to 1000 l/min. The interest is to have no deposit and agglomeration of cells within the microchannel, therefore we have to achieve turbulence or to create vortical structures within the flow field. Since the maximum Reynolds number is expected to be around 100, no turbulence will be present in a straight channel, even thought the rheology of the fluid sample is non-Newtonian. Vortical structures can appear in laminar flows,

but only if the geometry is complex or/and Reynolds number is high enough or/and the fluid is non-Newtonian, see Fig. 1.

vortical structure

simulations are performed with FLUENT 6.0 code, steady Navier-Stokes solutions are obtained with coupled scheme for a Newtonian fluid with viscosity of 1 Pas and density of 1000 kg/m3. The channel is 1 mm width and 20 mm length, the upper wall being divided in segments of 1 mm long. On each segment belong to the upper wall we can consider different boundary conditions; in this paper alternate stick slip conditions are considered. In the rest of the walls no slip boundary condition is imposed.

vortical structure

Figure 1. Vortical structures in a T-channel bifurcation (3D stream lines distribution for a shear thinning fluid at Re = 600, simulation with FLUENT code 6.0). It is well known from classical fluid mechanics that tensions in a homogeneous fluid, therefore the tensions acting on particles in suspension or at solid walls, are giving by the pressure and vorticity distributions, see [2, 5]. Therefore, we have to generate non-uniform pressure and vorticity distribution in the channel, in order to improve the mixing of cells and to avoid their deposit. There are two possibilities in the case of straight channel: (i) to introduce small rigid bodies in the flow field or to increase the roughness of the channel walls, [11, 13]; this means that we create artificially condition to generate vortical structures; (ii) to alternate at the channel wall hydrophilic (no slip boundary condition) with hydrophobic (slip boundary condition) regions, see Fig. 2 (for details about slip and adherence conditions at the walls see [10, 11]). As far as we know, the second procedure is a novel and original technique, and we intend to investigate its capabilities and applications in this paper. 2. STICK SLIP BOUNDARY CONDITIONS The present paper is concerned with the channel flow simulations at small and medium Reynolds number. The 2D geometry is a contraction flow with aspect ratio 5:1, see Fig. 2;

Iso-velocity stream lines slip vw = 0 slip vw = 0 slip vw = 0

x - axis

v=0

slip v = 0

slip

v=0

slip

v=0

Vorticity number Wo > 1

Figure 2. Entrance flow in a microchannel (Re = 1, 2D simulations of Newtonian flow; noslip boundary conditions with exception of the upper wall): alternate hydrophobic hydrophilic regions on the upper wall, slip stick (vw = 0) boundary conditions. The aim of our investigations is to put in evidence differences in the flow field due to the imposed alternate stick (adherence, vw = 0) slip boundary conditions on the upper channel wall. The distributions of wall stress, wall pressure and vorticity number are obtained for different slip conditions. The slip conditions at the wall are of two categories: (i) imposed velocity slip (vw 0); (ii) imposed maximum admissible wall shear stress along the x - direction (in our simulations, w = 0). In Fig. 3 are shown

differences at the entrance in the channel due to the imposed stick slip constrains. The areas of vorticity number higher than one at the entrance in the channel are shown in Fig. 4. It is well known, [4], that kinematical restriction Wo > 1 represents a necessary condition for the existence of vortices in vicinity of wall at which the fluid adheres (i.e. hydrophilic walls). Simulations evidence that, even at small Reynolds number, the alternate stick slip wall conditions generate areas where possible vortical structures might be present (see also Fig. 2). This result is important and relevant for our application in microchannels hydrodynamics, respectively to create mixing between fluid layers in laminar flows, without grooving or increase dramatically the roughness of the walls.

At this moment of our study we are not interested in establishing the relation between the type of slipping condition, wall materials and fluid rheology. The first aim of investigation is only qualitative, with the target to explore the magnitude of flow changes within the channel due to alternate stick slip conditions at the wall.

no slip (stick vw = 0)

a) stick v = 0 slip (w = 0) stick (vw = 0)

v=0 b) slip stick

stick (vw = 0)

slip (vw = 0.5 vm)

v=0 c) stick slip

v=0

Figure 4. Areas at the entrance in the channel with Wo > 1 at Re = 1, for different boundary conditions at the upper wall (on the other walls the adherence boundary conditions was imposed, i.e. v = 0). In Fig. 5, Fig. 6 and Fig. 7 are represented the vorticity number, wall velocity, wall pressure and wall shear stress along the upper channel wall, for slip condition w = 0 and Re = 1. With the exception of the first segment from the entrance in the channels, the flow kinematics is perfect

Figure 3. Iso-pressure and iso-vorticity lines at the entrance in channel at Re = 1: a) no slip boundary conditions; b) slip (w = 0) stick (vw = 0) on the upper wall; c) stick (vw = 0) slip (vw = 0.5 vm - where vm in the mean velocity in the channel).

periodic along the x direction of the channel. No vortex at the entrance is obtained at this value of Reynolds number. As we expected, [4], on the segments with no slip Wo 1 (with the exception of the edges) and Wo < 1 where slip velocity is present. The points with Wo = 0 corresponds to the perfect plug flow in the neighbourhood of the wall and a local maximum of pressure, see Fig. 7.

between the two cases for pressure distribution and wall shear stress distribution on the segments of no slip conditions. Wall shear stress takes zero values only on surfaces with imposed slip velocity (not on the surface with adherence!), therefore we do not have any detachment of boundary layer.1
2.2x10
5

2.0x10 1.8x10 1.6x10

Wall pressure [Pa]

1.2

vorticity number wall shear stress

2.0x10

1.4x10 1.2x10 1.0x10 8.0x10

20000

1.0 15000

1.8x10

Wall shear stress [Pa]

Vorticity number [-]

0.8 10000 0.6

pressure wall shear stress slip - w = 0


0.008 0.009

6.0x10 4.0x10 2.0x10

1.6x10 0.007

slip
0.010

0.0 0.011

0.4

5000

x - upper wall [m]

0.2 0 0.0 0.013

0.014

0.015

0.016

0.017

0.018

Figure 7. Wall pressure and wall shear stress distribution: upper wall, alternate slip (w = 0) stick (vw = 0) boundary condition.
pressure wall shear stress

x - upper wall [m]

2.5x10 2.0x10

1.0x10 1.2x10
5

1.6 1.4 1.2

velocity wall shear stress

5.0x10 1.0x10
5

20000

0.0 -5.0x10 -1.0x10 0.015 0.016


3

8.0x10

Wall shear stress [Pa]

15000

Wall velocity [m/s]

1.0 0.8 0.6 0.4 0.2 0.0 0.013 5000 10000

6.0x10 0.012

slip - vw = 0.5 vmax


0.013 0.014

slip

x - upper wall [m]

Figure 8. Wall pressure and wall shear stress distribution: upper wall, alternate slip (vw = 0.5 vm) stick (vw = 0) boundary condition. One conclude from these simulations that alternate boundary conditions determine perturbation within the flow field, even at small Reynolds number. It is expected that modification of symmetry in pressure and vorticity distributions across and along the channel will introduce
See the paper included in this volume: Balan C., Broboana D., Pressure constrain in vicinity of the separation point in planar, steady and isochoric motion: Case I Newtonian fluid.
1

slip
0.014 0.015

slip
0.016 0.017

slip

0 0.018

x - upper wall [m]

Figure 6. Wall velocity and wall shear stress distribution: upper wall, alternate slip (w = 0) stick (vw = 0) boundary condition. Figure 8 discloses the wall variation of pressure and wall shear stress for the velocity slip condition vw = 0.5 vm. Comparison between Fig. 8 and Fig. 7 evidences no qualitative differences

Wall shear stress [Pa]

Wall pressure [Pa]

Figure 5. Vorticity number and wall shear stress distribution: upper wall, alternate slip (w = 0) stick (vw = 0) boundary condition.

1.6x10

5 4

1.4x10

1.5x10

Wall shear stress [Pa]

additional forces on the particles (cells) in suspension, the effect we are looking for our particular application. 3. ENTRANCE VORTEX
Wall shear stress [Pa]

4x10 3x10 2x10 1x10

no slip slip (w = 0) - stick (vw = 0) stick (vw = 0) - slip (w = 0) stick (vw = 0) - slip (v = vm/2)

Beyond the first critical value of the Reynolds number, a vortex is formed immediately at the entrance in the channel, see Fig. 9.
Separation Point

0
6

-1x10 -2x10

vw 0

vw = 0

-3x10

0.005

0.006

0.007

0.008

0.009

0.010

x - upper wall [m]

5.0x10 4.5x10 4.0x10 3.5x10 3.0x10 2.5x10 2.0x10

Figure 9. Location of the entrance vortex and separation point on the upper wall (stream lineas distribution at Re = 200; stick slip boundary conditions). The location of separation point1 is dependent not only on the value of Reynolds number (in this simulations the Reynolds number is kept constant at 200), but also on the imposed boundary conditions at the upper wall. In Fig. 10 are represented the distributions on the upper wall of the shear stress and pressure, for different alternate stick slip boundary conditions, in comparison to the solution with no slip. The separation point is always located within the first segment of the upper wall, independently on the type of boundary condition. If the stick condition is imposed on that segment (vw = 0) or the velocity wall slip is known, the separation point is defined by the classical condition w = 0. In the case of slip condition w = 0, the location of separation point is considered given by vx = 0, i.e. the point where the velocity component at the wall changes its sign, see Fig. 11.

Wall pressure [Pa]

no slip slip (w = 0) - stick (vw = 0) stick (vw = 0) - slip (w = 0) stick (vw = 0) - slip (v = vm/2)
0.005 0.006 0.007 0.008 0.009 0.010

x - upper wall [m]

Figure 10. Wall shear stress and wall pressure distributions at the entrance in micro-channel, for different boundary conditions.
600000 400000

WSS; no slip WSS; stick - slip vx; slip (w = 0) - stick

20 15 10

Wall shear stress [Pa]

0 -200000 -400000 -600000 0.0050 0.0052 0.0054

0 -5 -10 -15 -20 0.0060

separation point
0.0056 0.0058

x - upper wall [m]

Figure 11. Wall shear stress distribution and location of separation point, as function of different alternate stick slip boundary conditions.

Velocity [m/s]

200000

The comparison between numerical solutions put in evidence the decreasing of vortex length when alternate stick slip boundary conditions are imposed on the upper wall. The control and generation of vortical structures in channels is the main subject of study in micro-hydrodynamics. The present investigations propose a novel method to control the local hydrodynamics in vicinity of the channel walls. The existence of wall regions with different adhesion properties for the carrier fluid might be the solution to improve the mixing in a microchannel, without modifications of the geometry or/and roughness. 4. FINAL REMARKS The paper investigated the influences of slip boundary conditions on the flow structure in a 2D channel. The numerical simulations are performed for two types of wall slip conditions: (i) w = 0, respectively (ii) vw > 0. The results evidence that, even at small Reynolds number (characteristic flow regime for microchannels Re < 500), different imposed boundary conditions at the wall induce qualitative changes in the whole flow field. Therefore, the method to alternate hydrophilic with hydrophobic walls is a promising technique for improve mixing of fluids in microchannels. Of course, experimental investigations of channel flows under microscop are needed to confirm the results of numerical simulations. Such studies are in progress at REOROM Laboratory from UPB. The stick slip control of the fluids at the channel wall is a scientific topic of major impact in micro and nanofluid mechanics, especially if the working fluids are also sensitive to the electromagnetic fields, [10, 11]. The further investigations are focused to electro-rheological fluids and their flow in microchannels (in particular, the cells under study can be orientated in water by an electric field). ACKNOWLEDGMENTS The present study was supported by the CNCSIS Romanian grant no. 33/2006, Vortex Hydrodynamics Consortium REFERENCES 1. Baroud C.N., Willaime H. (2004) Multiphase flows in microfluidics, C.R. Physique 5, pp 547-555

2. Berker R. (1951) Sur certaines proprits de leffort qui sexerce sur une paroi en contact avec un fluide visqueux, C. R. Acad. Sci. Paris 232, pp 148-149 3. Bleris L.G., Garcia J.G., Arnold M.G., Kothare M.V. (2006) Model predictive hydrodynamics regulation of microflows, J. Micromech. Microeng. 16, pp 1792-1799 4. Broboana, D., Calin A., Muntean T., Balan C., Balan, C. (2005) Is the vorticity number a possible quantification of a vortex?, Sci. Bull. P.U. Timisoara Trans. Mech., Tom 50(64) Special Issue, pp 155-160 5. Caswell B. (1967) Kinematics and stress on a surface of rest, Arch. Rational Mech. Anal. 26, pp 385-399 6. Chao S-H., Holl M.R., Koschwanez, Carlson R.H., Jang L-S., Meldrum D.R. (2005) Velocity measurement in microchannels with a laser confocal microscope and particle linear image velocimetry, Microfluid Nanofluid 1, pp 155-160 7. Choi C-H., Ulmanella U., Kim J., Ho C-M., Kim C-J. (2006) Effective slip and friction reduction in nanograted superhydrophobic microchannel, 18, 087105 8. Husny J., Cooper-White J.J. (2006) The effect of elasticity on drop creation in T-shaped microchannels, J. Non-Newtonian Fluid Mech. 137, pp 121-136 9. Kang K., Koelling K.W. Lee L.J. (2005) Microdevice end pressure evaluations with Bagley correction, Microfluid Nanofluid, 10404-005-0067-2 10. Squires T.M., Quake S.R., Microfluidics: Fluid physics at the nanoliter scale, Rev. Modern Physics, 77, 2005, 977-1026 11. Stone H.A., Stroock A.D., Ajdari A. (2004) Engineering flows in small devices: microfluidics toward a Lab-on-a-Chip, Annu. Rev. Fluid Mech. 36, pp 381-411 12. Tretheway D.C., Meinhart C.D. (2004) A generating mechanism for apparent fluid slip in hydrophobic microchannels, Phys. Fluids 16, pp 1509-1515 13. Wang H., Wang Y., Zhang J. (2005) Influence of ribbon structure rough wall on the microscale Poiseuille flow, J. Fluids Eng. 127, pp 1140-1145

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Special issue

Workshop on Vortex Dominated Flows Achievements and Open Problems Timisoara, Romania, June 01 - 02, 2007

PRESSURE CONSTRAIN IN VICINITY OF THE SEPARATION POINT IN PLANAR, STEADY AND ISOCHORIC MOTION: CASE I NEWTONIAN FLUID
Corneliu BALAN*, Prof. Hydraulics Dept., REOROM Laboratory Politehnica University of Bucharest Diana BROBOANA, Assoc. Prof. Hydraulics Dept., REOROM Laboratory Politehnica University of Buchares

*Corresponding author: Splaiul Independentei 313, 060042, Bucharest, Romania Tel.: (+40) 21 4029705, Fax: (+40) 21 4029865, Email: balan@hydrop.pub.ro

ABSTRACT The paper is concerned with the investigation of 2D flow structure in the vicinity of separation point. The geometry under study is a T planar bifurcation with a remarkable vortex developed beyond the junction. Numerical computations of the Newtonian flow (Navier-Stokes solution at Re = 100) disclose in the neighbourhood of the separation point a particular wall pressure distribution, which suggests that Reynolds approximation for thin films is valid in this domain. The authors evidence a novel constrain for the pressure field at separation point of 2D vortices, i.e. the third derivative of pressure along the wall direction is zero in that point. KEYWORDS Separation point, vortex, pressure, Reynolds equation

vorticity,

wall

NOMENCLATURE Re [-] Reynolds number Wo [-] vorticity number [kg/m3] mass density 0 , [Pas] zero viscosity, viscosity function v [m/s] velocity field [1/s] vorticity field D [1/s] stretching [1/s] spin tensor n [-] normal direction to the wall p [Pa] pressure [Pa] first normal stresses difference N1 [Pa] second normal stresses difference N2

1. INTRODUCTION In 2D configurations, the flow separation areas consists of close vortices. If the vortex is located at the wall, its boundaries are established by separation (stagnation) points, defined as points at the wall where the shear stress is zero. Accordingly to the boundary layer theory, the separation point is located in the vicinity of the local maximum wall pressure and the stream line on which it lays separates the vortex from the main flow field, [11]. The aim of the present paper is to investigate the kinematics in vicinity of separation point and to establish possible supplementary constrains for the wall stresses. The fluid under study is Newtonian, but our target is to extend the results to viscoelastic flows. It is of interest for our applications in biofluid mechanics to obtain a simplified model of the flow in the neighbourhood of separation point, in order to detect more precisely its position at the wall of blood vessels. One main parameter of the flow kinematics is the vorticity number Wo, defined as the ratio between vorticity magnitude and stretching magnitude, curlv Wo := = . (1) 2 D 2 trD

In the very vicinity of the walls the motion is viscometric, therefore Wo 1. Actually, at any surface where the fluid adheres there is a perfect equilibrium between the vorticity magnitude and the shear strain rate, see [1] and [13]. At the same time, the tension at the surface of adherence is related to vorticity by the following relations: (2) t = p n + 2 0 n

for Newtonian fluids [2], and t = [ p + N 1 ( ) N 2 ( )]n + 2 ( ) n for viscoelastic fluids, [3], where =

(3) and

= 1 2 curlv . For details about the kinematics and dynamics of vertical structures in relation to vorticity field and flow separation see [7], [9] and [14]. Since it is well known that wall shear stress (WSS) and vorticity number are vanishing at separation point, a special attention in this study is given to the distribution of normal stresses at the wall where the vortex is located. At the end, a novel constrain for pressure distribution at separation point is proposed.

2. NUMERICAL CASE The present work is based on the Navier-Stokes solution in a 2D T geometry with relevant separation areas. The computations have been performed with FLUENT 6.0 code, at characteristic Reynolds number of 100 (Navier-Stokes and continuity equations are solved with the coupled scheme, at a convergence criteria of 10-9 for velocity components and continuity equation). In Fig. 1 are presented for comparison the numerical steady solutions obtained with the segregate and coupled procedures. One can observe major quantitative differences between solutions.
10
3

y* y**

4x10

3x10

WSS [Pa]

10

WSS dp*/dy segr. , coupl. ,

2x10

1x10

0 10
1

dp*/dy [Pa/m]

y
5

0.03

0.04

0.05

0.06

0.07

0.08

0.09

-1x10 0.10

y - direction [m]

Figure 1. Numerical solutions for the WSS and wall pressure gradient in the area of vortex I from the Fig. 2 (segregated and coupled solutions). In Fig. 2 are disclosed the vortical structures in the planar Tbifurcation with the constant width of 25 mm. The separation point (WSS = 0), D1, and the maximum wall pressure point (dp*/dy = 0), D2, are shown for the Vortex I.

Figure 2. Stream lines distribution in a T flow geometry at Re = 100 (Newtonian incompressible fluid in steady motion); detail with Vortex I and location of the points D1 (y = y*), respectively D2 (y = y**). The numerical case presented in Fig. 2 is considered in this study as representative case for the 2D vortices. The further analysis will be focused exclusively on the Vortex I characterization and the corresponding wall stresses distributions.

3. VORTEX INVESTIGATION Location of the separation point D1 is direct related to the velocity and vorticity distribution. The conditions for the existence of separation point at the wall of the channel are obtained numerically in Fig. 3 for the investigated flow. The simulations with the coupled scheme produce results in very good agreement with the theory.
800 600 1.0 200
Separation Point - D1
WSS Wo

1.2

0.8 0.6

0 -200 -400 -600 -800 -1000 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.0 0.10
vortex area y* = 60 mm

0.4 0.2

y - direction

Figure 3. Distributions along the wall for WSS and Wo; the area of Vortex I and location of separation point are well defined. In Fig. 4 are represented the computed velocity distributions in the channel and the corresponding regions with minimum vorticity magnitude. The representation of the computed iso-pressure spectrum discloses not only the point D2 (where wall pressure takes the maximum value, p = pmax), but also the existence of a particular iso-pressure lines, p = pcr, which actually separate the flow field, see Fig. 5 and Fig. 6. Moreover, that particular isopressure line corresponds to the separation point D1. So, D1 is also a point where pressure diverges at the wall. It seems that always the existence of a separation point at the wall is associated to a saddle pressure point (pressure critic point from Fig. 5) within the flow domain. We remark here that the computed ratio yp/ys (the relative location of D1 against D2) equals 0.4, which is very close to that obtained for an oblique stagnation-point flow by Dorrepaal [4], respectively yp/ys 0.396, see Fig. 7 (the origin of yp, respectively ys, coordinates is at the intersection of the tangent at the vortex I boundary with the wall).

Vorticity number - Wo [-]

400

WSS [Pa]

Figure 4. Velocity distributions across the channel width. The regions with Wo < 0.1 are marked in the flow field (at D1 the vorticity magnitude is zero).

Figure 5. The relative position of the pressure critic point against the separation point D1.

Figure 7. The relative position of the particular points D1 and D2.

Figure 6. Spectrum of pressure; the iso-pressure lines p = pcr separate the flow field (p = pcr corresponds to D1 and p = pmax corresponds to D2). Exploring the kinematics details of the flow in the very vicinity of the stagnation point, one observe that (i) the iso-pressure line p = pcr is almost normal to the wall at D1, (ii) the component vy of velocity is linear along the wall direction, see Fig. 8, and (iii) the component vx of velocity is constant up to a distance x = h from the wall, see Fig. 9. This remarkable flow structure suggests that in the neighbourhood of D1 the approximation of Reynolds for the motion in thin films is valid, i.e. hv y 3 p = 6 0 2 v x + h , y y y

(4)

Figure 8. The control domain in the very vicinity of D1; the spectrum of iso-velocity vy = constant are straight lines.

in the control domain of height h from the wall, see Langlois [8] for the mathematical expression of Reynolds equation. Because the right member of (4) is constant at h = constant, the relation is equivalent at D1 with the constrain 3 p = 0. (5) y 3
y = y*

22000 20000 18000 16000


pcr = 19410 Pa

800 600 400


y** = 77 mm

Pressure [Pa]

14000 12000 10000 8000 6000 4000 2000 0 0.03 0.04 0.05 0.06
pressure wss y* = 60 mm

200 0 -200 -400 -600 -800 -1000 0.10

0.07

0.08

0.09

y - direction [m]

Figure 10. Pressure and WSS distributions at the wall. Condition given by relation (5), with p = pcr, is valid at y = y* which indicates the coordinate of the point D1 at the wall (see also Fig. 3). This assertion is validated by numerical computations. In our case, the x-direction is the normal direction to the wall. The variation of vorticity magnitude is represented across the channel, on lines of y = constant, see Fig. 11 and Fig.12. In vicinity of the wall (x = - 0.0125 m) the vorticity distribution takes particular values, which confirm the existence of the critical points: (i) at D1 0 , and (ii) = 0 at D2. Figure 9. Details of the flow field in the neighbourhood of separation point D1. The distributions of normal and shear stresses along the wall from Fig. 10 confirms that relation (5) is valid at D1. Additional constrains at the stagnation (separation) point and criteria of the existence of planar flow separation areas are obtained also in [4] and [6]. Relation (5) at separation point D1 is completed with a supplementary restriction for vorticity at the point D2. By definition, the vorticity diffusion flux, , (6) = 0 n measures the rate at which vorticity is created at the boundary and diffused into the fluid [14]. In a 2D flow, vorticity diffusion has one single component ( ) and, because at the wall the Stokes approximation is valid, i.e. p = , (7) y has to be null at the point of maximum pressure.

Figure 11. Location of the critical points D1(L284) and D2(L266) (velocity distributions superimposed on the iso-pressure lines).

WSS [Pa]

700 600

L260, L280,

L266, L284,

L270 L285

Vorticity magnitude [1/s]

D2 500 400 300 200 100 0 D1 -0.0125 -0.0120 -0.0115 -0.0110 -0.0105 -0.0100

ACKNOWLEDGMENTS The present study was supported by the CNCSIS Romanian grant no. 33/2006, Vortex Hydrodynamics Consortium. REFERENCES 1. Broboana, D., Calin A., Muntean T., Balan C., Balan, C. (2005) Is the vorticity number a possible quantification of a vortex?, Sci. Bull. P.U. Timisoara Trans. Mech., Tom 50(64) Special Issue, pp 155160 2. Berker R. (1951) Sur certaines proprits de leffort qui sexerce sur une paroi en contact avec un fluide visqueux, C. R. Acad. Sci. Paris 232, pp 148-149 3. Caswell B. (1967) Kinematics and stress on a surface of rest, Arch. Rational Mech. Anal. 26, pp 385-399 4. Dorrepaal J.M. (2000) Is two-dimensional oblique stagnation-point flow unique?, Canad. Appl. Math. Quart., 8(1), pp 61-66 5. Dorrepaal J.M. (1992) The flow of a viscoelastic fluid near a point of re-attachment, ZAMP, 43, pp 707-714 6. Haller G. (2004) Exact theory of unsteady separation for two-dimensional flows, J. Fluid Mech., 512, pp 257-311 7. Jeong J., Hussain F. (1995) On the identification of a vortex, J. Fluid Mech. 285, pp 69-94 8. Langlois W.E. (1964) Slow viscous flow, Macmillan, New York 9. Lugt H.J. (1996) Introduction to vortex theory, Vortex Flow Press Inc., Potomac 10. Renardy M. (2006) Viscoelastic stagnation point flow in a wake, J. Non-Newtonian Fluid Mech., 138, pp 206-208 11. Schlichting H., Gersten K. (2000) Boundary layer theory, Springer, Berlin, Heidelberg 12. Thompson R. L., Souza Mendes P. R., Naccache M. F. (1999) A new constitutive equation and its performance in contractions flows, J. Non-Newtonian Fluid Mech. 86, pp 375-388 13. Truesdell C. (1988) On the vorticity numbers of monotonous motions, Arch. Rational Mech. Anal. 104, pp 105-109 14. Wu J.-Z., Ma H.-Y., Zhou M.-D. (2006) Vorticity and vortex dynamics, Springer, Berlin, Heidelberg

x - direction [m]

Figure 12. Distribution of vorticity magnitude on normal direction to the wall.


4. FINAL REMARKS Applications of the vortex theory are present in almost all the applied fields of fluid mechanics. The characterisation of vortical structures are of great importance especially in vicinity of the walls, the place of interaction between the fluid and deformable or rigid solid boundaries. The paper was focused to the study of stresses distribution at the wall, in presence of vortices. The authors obtained a supplementary condition for wall pressure at the point of separation, relation (5), which accomplishes the classical constrain for vorticity magnitude, i.e. = 0. The result was obtained assuming that the Reynolds thin film approximation is valid in vicinity of separation point, simultaneously with the validity at the wall of Stokes approximation. The hypothesis and analytical results were confirmed by the numerical simulations of a 2D Newtonian flow in a bifurcation, a configuration with remarkable vortical structure at the wall. In present, the authors investigate the kinematics and dynamics conditions at separation point in 3D bifurcations. In this case the numerical solutions are validated by the experiments performed in REOROM Laboratories. It is of interest for our further research (the computation of wall stresses in blood vessels) to test and to extend the validity of constrain (5) to viscoelastic flows, where the locations of separation (stagnation) and re-attachment points might be also dependent on the elastic normal stresses, see [5], [10] and [12].

Scientific Bulletin of the Polytechnic University of Timisoara Transactions on Mechanics Special issue

Workshop on Vortex Dominated Flows Achievements and Open Problems Timisoara, Romania, June , 2007

CFD FIRST PREDICTION IN DESIGNING A 50KW SWIRLING BURNER WITHIN ITS COMBUSTION CHAMBER

Victor HODOR, Prof.* Dep.of Thermal Engines and Equipments Technical University of Cluj-Napoca Paula UNGURESAN, Assist. Prof. Dep. of Thermal Engines and Equipments Technical University of Cluj-Napoca

Florin BODE, Assist.Prof. Dep. of Thermal Engines and Equipments Technical University of Cluj-Napoca Claudiu RATIU, Prof. Dep. of Hydraulics Technical University of Cluj-Napoca

*Corresponding author: Bv Muncii103-105, 300223, Cluj-Napoca, Romania Tel.: (+40) 264 410777, Email: victor_hodor@yahoo.com

ABSTRACT . The present paper regards to the first steps/results that were performed in order to design a best performing 50kW new swirling burner within its combustion chamber. CFD investigation/prediction method was used, in order to find out the way to mix air and methane gas in the proper combustion conditions. The device is energetically dimensioned up to a 3-5 m3/h CH4 flow rate (up to ~50kW). The gasodynamic geometry is axisimetricaly, and the turbulent air IN (for a requested premixed air-methane combustion rate) we have provided by imposing the swirled boundary condition (for the IN velocity). We have done the investigations/ computations searching the maximum efficiency for an as large as possible INPUT domain an as low as possible flow-rate potential). As soon as we want to identify an original burner design (an as simple as possible in manufacturing requirements) once again, we want to underline, that the CFD numerical investigations might neglect the species/combustion module at the very first stage of the overall burners geometry prediction. KEYWORDS Burner, CFD Gasodynamic-geometry prediction, Swirling, Turbulent combustion, Combustion chamber, RANS, RNG k- ABBREVIATIONS CFD Computational Fluid Dynamics INPUT Boundary conditions at the very entrance in the interest contour BQG Burner Quested/investigated geometry OBMR Output burner mixing region 1. INTRODUCTION Computational Fluid Dynamics (CFD) is efficiently used to find-out the design of original new gasoodynamical equipment, and today no real progress in design can be made without using CFD. With the same objectives, CFD have developed in turbulent combustion modeling, following a variety of approaches and distinct modeling strategies. We first have to eliminate evident improper gasodynamic flow iINPUT constrains, and focus our investigations on two guess overall burner geometries. Consequently to each select overall geometry, we try to select two or three shapes,

proper to be revised (as position, magnitude and/or as in-between proportion), in order to be able to compare results for the same boundary conditions (from a minimum to a maximum flow rate limit). 2. NUMERICAL SIMULATIONS RESULTS We have done so far, and then selected two geometries and some of the related computational results, in order to enface the

benefit (we have found out). In a first image denoted with a we have introduced the methane in longitudinal direction (to draw comparisons), and in the second image denoted with b, we show the advantage becoming from the idea to introduce the methane in a radial direction to the air stream. In Fig.1a and 1b below, we represent the related result for the velocity.

Fig.1a. First BQG -velocity field By examine the related results, we can see that the average velocity, at the front of the OUTPUT burner region is varying around the desired range of 0,4m/s (the methane flame speed of reaction). Even that, -the both gasodynamic geometryes seems to be proper

Fig.1b. Second BQG -velocity field by the mean of the velocity field, we have to revue extra parameters/conditions required by the conditions of a performant / complete and stabile combustion process.

Fig.2a. First BQG Stream Function In the Fig.2a and b (above), is to be seen the Stream Function field related to the same boundary conditions and to the same two BQG.

Fig.2a. Second BQG Stream Function It is to be seen that in the second gasodynamic geometry (Fig.2b) the Stream Function fluctuation (alternance of the increase and the decrease of

its values) in the OBMR, give us more confidence regarding the capability of ensuring the best mixing ratio requested to ensure the proper stoichiometric ratio around 1/10. Of great interest should be also the Kinetical energegy magnitudes, as it is important in

mixing process at the OBMR, but it is to be avoided by the means of enegy consumptions and the chance to introduce instabilities. That is why, in the next figure (Fig.3a and b) we enface the Kinetic Energy distribution.

Fig.3 a. First BQG Kinetic Energy By analysing these, it is very easy to be seen that for the first BQG, the high magnitudes of Kinetic Energy are grouped imidiatlety after the air swireled entrance, but for the second BQR the the KE have the highest values in the

Fig.3 b. Second BQG Kinetic Energy immerging zone of the methane in the air stream which gives us more trust in the second BQG.

Fig.4 a. First BQR preassure field By analyzing the pressure field in both cases, we may drought out an opinion concerning the local and the overall pressure within the combustion chamber. The magnitude of the pressure at OUTPUT boundary (at the exit from the combustion chamber should be greater than

Fig.4 a. Second BQR preassure field the pressure losses of the flue pass of the boiler.

3. CONCLUSIONS

By cumulating the forward results, we may conclude that the confidence -the chance / probability to ensure the best combustion conditions are in the second Burner Quested Geometry, so the investigations might continue abandoning the first BQG. ACKNOWLEDGMENTS We express our gratitude to the Numerical and Parallel Computing Laboratory from the Polytechnic University of Timisoara, for the opportunity they have gives us to use FLUENT, within the frame of a national consortium research program on Hydrodynamics of the Vortex and its Applications. REFERENCES 1.Batchelor, G.K., An Introduction to Fluid Dynamics, Cambridge Univ.Press, 1983. 2.H. Bchner, H. Bockhorn, Measurement and Simulation of Combustion Noise emitted from Swirl Burners with different Burner Exit Geometries, Division of Combustion Technology, Universitt Karlsruhe, 2006. 3.Fluent Inc. FLUENT 6 Users Guide, Fluent Incorporated, Lebanon, 2001. 4.Ferziger, J.H., Milovan, P., Computational Methods for Fluid Dynamics, Springer, ISBN 3-540-65373-2, 1999. 5.Hodor V., Dinamica gazelor, Ed. Casa Cartii de tiin, ISBN-973-9404-54-5, Cluj, 1999. 6.J. B. Bell, M. S. Day, A. S. Almgren, R. K. Cheng, I.Shepherd, Numerical Simulation of Premixed Turbulent Methane Combustion, Lawrence Berkeley National Laboratory Berkeley, California, USA, 2000. 7.Kenneth Kuan-yun Kuo, Principles of Combustion, John Wiley & Sons, 1986. 8.Launder, B.E., Spalding, D.B., Mathematical Models of Turbulances, Academic Press, New York, 1972. 9.Mihaescu, M., Computational Aeroacustic Based on Large Eddy Simulation and Acoustic Analogies, ISBN 91-628-6443-2, Lund, 2005. 10. Pantakar, S., Numerical Heat Transfer and Fluid Flow, Hemisphere Publ., MacGrow Hill, 1980. 11. Romeo S. Resiga, Numerical simulations in Fluid Mechanics,SeminarNotes, Timisoara,1993

Scientific Bulletin of the Politehnica University of Timisoara Transactions on Mechanics Tom 52(66), Fascicola 3, 2007

3`nd Workshop on Vortex Dominated Flows Timisoara, Romania June 1 - 2, 2007

A FOUNDATION IN DISTRIBUTIONS OF GLAUERT THEORY


Theodor POPESCU, lecturer Department of Fluid Mechanics, Fluid Machines and Drives Technical University Gheorghe Asachi of Iasi

*Corresponding author: Bvd Dimitrie Mangeron 63, 700050, Iasi, Romania Tel.: (+40) 232 278683/2138, Fax: (+40) 232 242109, Email: thpopescu2003@yahoo.com
version of Glauert model combined with blade element theory. The main aerodynamic codes are based on BEM [2, 5]. Some authors suggest that BEM is the only model suitable for engineering applications [1]. The core of Glauert theory is the model of active disk. According to the terminology introduced in [4], TAD is a simplified operational model. The source model is TPNB. The connection between TPNB and TAD is represented in Figure 1.

ABSTRACT Theory of active disk (TAD) is the core of the classic model of Glauert. TAD is a simplified operational model derived from the physical model of the propellers theory with finite number of blades modeled as lifting lines (TPNB). In this paper, the mathematical foundation of TAD is revised involving the theory of distributions. It is established an explicit mathematical connection between the generalized mathematical model of TPNB and the generalized mathematical model of TAD. KEYWORDS Glauert theory, distributions, propeller theory NOMENCLATURE [m/s] absolute velocity c relative velocity w [m/s] transport velocity cT [m/s]

Figure 1. The connection between TPNB and TAD The physical model of TAD derives directly from the physical model of TPNB by increasing the number of blades to infinite, while the local load is maintained constant for each radius. The propeller with finite number of blades is transformed in a propeller with infinite number of blades, namely an active disk. Due to the infinite number of blades, the accepted mathematical model of TAD consists in the equations of inviscid fluid dynamics with symmetry about the propeller axis. This approximation by convergence will be referred as the physical map T ph . For the time being, this convergence does not have a mathematical sense. The goal of this paper is to establish an explicit mathematical connection between the generalized mathematical model of TPNB and the generalized mathematical model of TAD.

{ f } [-]

regular distribution

Subscripts and Superscripts r radial direction tangential direction x1 axial direction ABBREVIATIONS TAD Theory of Active Disk TPNB Theory of the Propeller with finite Number of Blades modelled as lifting lines 1. INTRODUCTION The state-of-art in Horizontal Axis Wind Turbines design are the complex codes that include specialized modules for aerodynamics, structural dynamics, etc. The preferred aerodynamic model is BEM (Blade Element / Momentum) an improved

2. THE GENERALIZED MATHEMATICAL MODEL OF TPNB Let us consider the case of the free propeller with N blades. The non-inertial frame attached to the lifting line k and the field of local aerodynamic reactions f k are represented in Figure 2.
k k k ,x2 ,x3 Figure 2. The non-inertial frame O; x1

div{w( x c )} +{ ( x c )} +{cO x grad { p} + mk


k =0 N1 Ck

( 4)

The parametric equations of the lifting lines are

c2 N 1 div w + p c = P k k C 2 k =0

( 5)

The perturbation introduced by the lifting line k is modeled through the Dirac-type singulaties f k k , m k k and P k k . The relations between
C C C

the densities of the singularities are similar to the lifting surface case: m k = x k f k , P k = cT , f k

(6)

( )

( x1 ,r, ) x1 , Ck : 2 , k = 0 r R ,R , = = k [ b v] k N

3. THE AERODYNAMIC FORCES ON THE ACTIVE DISK Since the blades are identical, the local aerodynamic reactions f k are equal in norm at the same radius. Therefore:

( 1)
N1 k =0

f kCk

The natural working space is N 1 ' 3 k ; by R is denoted the nonDe C ; R U k =0


k k k inertial frame O; x1 ,x2 ,x3

f10 = B At ( k ) fr0 ( 0, r, 0 k =0 Rb f
N1 Rv

( 7)

B = [ e1 , e2 , e3 ] is the base of the non-inertial frame

) attached to the blade

0. In the generalized theory of lifting surface, the restriction map is injective [4]. The property holds in the case of the lifting line. Hence, we will work in

0 0 1 R and A ( ) = 0 cos sin . According to the 0 sin cos physical map T ph , the products N f10 , N f r0 and
N f 0 must be kept constant in (7), while N is increasing to infinite. To see if the limit make sense mathematically, we rewrite (7) as

D'

; R . The multiplication holds as a

irregular intrinsic operation [3]. In order to avoid the problems arising from the partial lack of associativity, we will assume an isochore motion. The generalized equations of TPNB are [4]: div { w} = 0 2) (
N

N1 k =0

f kCk

div{wc} +{c} = grad { p} +


k=0

o f1 N1 Rv 2 o = B At ( k ) ( 0, r, k ) fr N o k =0 Rb f

( 3)

where

N f10 o def N f r0 o def N f 0 , fr = , f = f1 = 2r 2r 2r


o def

(8)

In conclusion, the propeller thrust is equal with the active disk thrust.
4. THE AERODYNAMIC MOMENT ON THE ACTIVE DISK From (6) it derives the momentum density on the

2 We notice that At ( k ) ( 0, r, k ) is a Riemann k=0 N


2

N1

sum for

t A ( ) ( 0, r,) d. Therefore: 0
N 1

lifting line C k :
k k m1 ( r ) = r f0 ( r ) , mrk = 0, m ( r ) = r f10 ( r ) (13)

N k =0 N f k =const

lim

Ck

o f1 Rv 2 o t f r ( 0, r , ) d rdr B A ( ) Rb 0 o f

(9)

As expected, the moment of local aerodynamic reactions is the same for all lifting lines. Following the path from the previous section, we define:

m1 =

o def

0 o def N m0 o def N m0 N m1 r , mr = , m = 2r 2r 2r

(14)

If the products N f10 , N f r0 and N f 0 are


0 and maintained constant, then the products N m1 0 remain constant too. Therefore: N m N 1 N k =0 N f k =const

The limit from the left term will be denoted as the mathematical map T f . T f is the mathematical correspondent of T ph . From (9), we notice that T f transforms the system of N lifting lines in the annulus ( S AD ) : {( x1 ,r, ) x1 ,r [ Rb ,Rv ] , [ 0,2]} . In order to preserve the classic terminology, we agree to say that S AD is an active disk instead an active annulus. Then, (9) becomes:
N 1 k Tf f C k k =0
o o = = S AD f y y d f ( ) ( ) y S AD

lim

mk C k =

(9)

o m1 Rv 2 o t mr ( 0, r , ) d rdr B A ( ) o Rb 0 m

The field f ( r ) = f1 ( r ) e1 + f r ( r ) er + f ( r ) e defined in (8) represents the equivalent aerodynamic reaction on the active disk. Consequently, the resultant of equivalent aerodynamic forces on the active disk S AD is:
Rv o F ( S AD ) = f d = f1 2r dr e1 R S AD b
o

This relation can be rewritten as:


N 1 k Tf m C k k =0
o
o o (15) = m ( y ) ( y ) d y = m S AD S AD

(10)

The resultant of aerodynamic forces on the system of N lifting lines is:


N 1 k N 1 k F UC = f ds k =0 k =0 Ck

The field m ( r ) = m1 ( r ) e1 + mr ( r ) er + m ( r ) e defined in (14) represents the equivalent moment of aerodynamic reaction on the active disk. Consequently, the resultant moment of equivalent aerodynamic forces on the active disk S AD is:
Rv o M ( S AD ) = m d = m1 2r dr e1 R S AD b
o

(16)

(11)

The comparison between (10) and (11) leads to:


Rv N 1 k 0 = F1 ( S AD ) = F1 C N f1 dr U k =0 Rb

The resultant of aerodynamic forces on the system of N lifting lines is:

(12)

N 1 k N 1 k M UC = m ds k =0 k =0 Ck

(17)

The comparison between (16) and (17) leads to:


Rv N 1 k 0 M1 ( S AD ) = M1 UC = N r f dr k =0 Rb

T f is based on physical reasons. Therefore, its


(18)

field of application is restricted to the singular distributions concentrated on the lifting lines. It can not be seen how can be mapped T f to a regular distribution or to a singular vortex distribution. It is necessary to extend the T f to D' map T0 : D
T0 ( ) =
def

In conclusion, the axial moment of the propeller is equal with the axial moment of the active disk.
5. THE AERODYNAMIC POWER OF THE ACTIVE DISK The transport velocity formula is cT = cO + x , where cO = 0 for the turbine and

;R .

The path is suggested by (7). First, we define the

;R D

; R by:

cO = U 0 for the screw. From (6), it derives:


0 Pk ( r ) = cO , f k + , mk = cO f10 m1

1 2

2 0

( x1, r, ) d

(25)

) (

(19)

Obviously, T0 is linear and continuous. We now define TAD : D ' map of T0 :


def

It follows P denote: P (r ) =
o

(r ) = P (r ),
0

k = 0, N 1 . Let us

; R D'

; R as the adjoint

def

N P0 ( r ) 2r
f10 , N f r0

(20) and N
f 0

(TAD f , ) = ( f ,T0)
Let see how
2 0 0

(26)
TAD

transforms

regular

If the products N

are

distributions. If

{ f } is a regular distribution, then:

maintained constant, then N P 0 remain constant too. Consequently:


N 1 k Tf P C k k =0
o o = = S AD P y y dy P ( ) ( ) S AD

(TAD { f } , ) = f ( x1, r, )
1 2
2 0

(21)

( x1, r, ) d d rdr dx1


2 2

The power of the aerodynamic forces on the lifting lines is:


N 1 N 1 k = P C P ds U k =0 C k k =0

By the substitution = , it follows:


1 2 1 2

(22)

f ( x1, r , )

( x1, r , ) d d

The comparison with the power of the equivalent aerodynamic forces on the active disk
P ( S AD ) =

f ( x1, r , )d ( x1, r , ) d
0

S AD

P ( y ) d y

(23)

Renaming the variables, we derive:

(TAD { f } , ) = ({a0 ( f )}, )


a0 ( f ) = 1 2
2 0

(27)

leads to:
N 1 k P UC = P ( S AD ) k =0

(24)

f ( x1, r, ) d

is the first coefficient of

In conclusion, the power of the propeller is equal with the power of the active disk.
6. THE GENERALIZED MAP TAD

the Fourier series in generated by f . Let now see how TAD transforms the singular
N 1

distribution

k =0

f k Ck . Hence:

The map T f is the mathematical correspondent in the space D'

; R of the physical map T ph .

N 1 k TAD f C k k =0
Rv N 1 k =0 N 1 k =0

= ,
t

N 1 k T b C k k =0
f10 0 fr 0 f

def =

N 1 N k =0
N b =const

lim

k k b C

(32)

Rb

B TAD A ( ) k , ( 0, r, )

( r ) ( r ) dr = ( r )

N 2 R

Rv 2

b 0

f10 ( r ) B At ( ) ( 0, r, ) d f r0 ( r ) dr 0 f r ( )

N 1 k N 1 k We notice that TAD . T = k b b C k C = = 0 0 k k Consequently, (31) shows that the bounded vorticity is conserved by TAD : the product N b remains constant when N is increasing to infinite. We assume the following generic parametric

equations for the free vortex sheet S k :

(S ) :
k

y = y k ( t , s ) , t [ 0, + ) , s [ Rb , Rv ]

it follows
N 1 k o TAD f = f S AD Ck k =0

the curves of s = const. being the vortex lines. Then


(28)
kf = 0 f y,k t y,k t
0 y,k t = y,t

(33)

Similarly,
N 1 k TAD m Ck k =0 o N 1 k o m , T = S AD AD P C k = P S AD k =0

Since

k 0 0 y,k t y,s = y,t y,s , and

for

k = 0, N 1 , we have
N 1 k =0 0 y1, t 0 f t 0 0 y,0 = B A ( k ) y2,t t y,t dsdt 0 0 y,t 0 Rb y 3,t Rv

We conclude that TAD is the extension of T f to D'

;R .

kf Sk

7. THE VORTEX SYSTEM ASSOCIATED TO THE ACTIVE DISK

We denote by

k b

the bound vortex on the lifting

Following the path from the previous section, we derive:


N1 k T , k AD f = S k=0
Rv 2 0 y1, t N 0 f t 0 0 y2,t y,0 t y,t ( t, s, ) ddsdt B A ( ) 0 2 y,t 0 Rb 0 0 y3, t

line C k , by kf the free vortex field generated by the lifting line C k and by S k the free vortex sheet springing from C k . Then:
k rot {c } = b k =0 N 1 C
k

+ kf
k =0

N 1

Sk

(29)

k 0 b We notice that ( r ) = b ( r ) er ( k ) . Following the same steps as in section 3, we derive:

We now denote
z ( t , s, ) = At ( ) y ( t, s )

(34)

k TAD b Ck = b S AD k =0
o

N 1

(30)

Let us notice that z, = A,t ( ) A ( ) At ( ) y 0 . It follows

where b ( r ) =
o

def

0 b

(r )

(e y , y
0 1

z, = e1 y 0

and

0 ,t

y,0 s . It derives that (34) are the

( z, , z,t z,s ) =

2r

er

(31)

parametric equations of the axis-symmetric domain D derived through rotating S 0 around the propeller axis. Since dz = ( z, , z,t z,s ) d dtds , the

Let us denote

above integral is equal to


where
f ( , t , s ) =
o def

z,t z,t

( , t , s ) dz ,

becomes

TAD u, j = (TAD u ), j . It derives:


div{TAD ( w)} = 0

( )

div{w} = 0 .

It

is

obviously

that

N 0 f ( t, s ) 2

0 y,0 t y,s

( z, , z,t z,s )

(35)

(38)

It follows
TAD k f k = f D
k =0 S N 1

Therefore, the map TAD is compatible with the generalized mass balance equation. The generalized momentum balance equation (3) has the following equivalent form [4]:
N1 ( c,c ) p 1 ( c,cT ) + + rot{c} w= f k k (39) grad C 2 k=0

(36)

where D is the distribution generated by the characteristic function of the domain D . d Since it follows that f ( s) = b , ds N 1 k N 1 k . Consequently, = TAD T f Sk f S k k =0 k =0 (36) shows that the free vorticity is conserved by the map TAD . Finally, from (30) and (36) it follows:
o N1 N1 k o k + = TAD b S + f D k k b f AD C S k =0 k=0

We recall that TAD { f } = {a0 ( f )} . But


a0 ( f g) = a0 ( f ) a0 ( g) 2 + aj ( f ) aj ( g) + bj ( f ) bj ( g)
j=1

(37)

We conclude that the vortex system associated to the active disk consists in the bounded vortex annulus S AD and the free vortex domain D .
8. THE COMPATIBILITY BETWEEN TAD AND THE VORTEX SYSTEM ASSOCIATED TO THE ACTIVE DISK

It follows that TAD is not compatible with the generalized momentum equation. The reason of this lack of compatibility is the momentum equation nonlinearity. This lack of compatibility implies that the momentum equations with axial symmetry can not be accepted as the second TAD equation. This explains the cause of the contradictions arising in Glauert theory.
9. CONCLUSIONS TAD is a mathematical tool that allows a rigorous construction of TAD based on the source model of TPNB and a critical analysis of the hypothesis currently accepted. . The goal of the future paper is to establish a rigorous alternative to the Prandtl correction, based on the above operatorial approach. ACKNOWLEDGMENTS The author acknowledge the support from the National University Research Council grants (CNCSIS A-Consortium 33/2007). REFERENCES 1. Dumitrescu H., Georgescu A., et al. (1990) Aerodynamic Propeller Design (in Romanian). Ed. Academiei, Bucuresti 2. Moriarty P.J., Hansen A.C. (2005) AeroDyn Theory Manual. NREL-2005 3. Oberguggenberger M. (1992) Multiplication of Distributions and Applications to Partial Differential Equations. Longman Scientific&Technical and John Wiley&Sons

implies E (TAD x ) = 0 . Consequently, TAD will be compatible with the generalized model of TPNB if TAD ( E ( x ) ) = E (TAD x ) for each equation of the model. The compatibility defined above has the following physical meaning: TAD is compatible with the balance equation transformed equation TAD ( E ( x ) ) = 0 is the same balance equation written for the motion with symmetry about the propeller axis. The main equation of TPNB are (2) and (3). The equations (4) and (5) derives from (3). Since the motion is isochore, the mass balance equation (2) E ( x) = 0 if the

If E ( x ) = 0 is an equation that belongs to the generalized mathematical model of TPNB, we say that TAD is compatible with E ( x ) , if E ( x ) = 0

4. Popescu Th. (2003) Research in the Theory of Axial Rotor with Applications to Propellers and Fans (in Romanian). PhD Thesis, Technical University Gheorghe Asachi of Iasi 5. *** VEWTDC - Verification of European Wind Turbine Design Codes, Final report (2002). ECN-C-01-055