Sunteți pe pagina 1din 10

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

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

COMSOL MULTIPHYSICS VERSUS FLUENT: 2D NUMERICAL SIMULATION OF THE STATIONARY FLOW AROUND A BLADE OF THE ACHARD TURBINE
Andrei-Mugur GEORGESCU, Assoc. Prof.*
Hydraulics and Environmental Protection Department Technical Civil Engineering University Bucharest

Sanda-Carmen GEORGESCU, Assoc. Prof.


Hydraulics and Hydraulic Machinery Department University Politehnica of Bucharest

Sandor BERNAD, Senior Researcher


Center of Advanced Research in Engineering Sciences Romanian Academy - Timisoara Branch

Costin Ioan COOIU, Assist. Prof.


Hydraulics and Environmental Protection Department Technical Civil Engineering University Bucharest

*Corresponding author: Bd Lacul Tei 24, Sector 2, 020396, Bucharest, Romania Tel.: (+40) 212433660, Fax: (+40) 212433660, Email: andreig@mail.utcb.ro ABSTRACT Two-dimensional numerical modelling of the stationary flow around a blade of the Achard turbine, a new water-current turbine concept, is performed both with COMSOL Multiphysics 3.3 and with Fluent 6.0.1 software, in order to compare the results and the software capabilities. The k turbulence model has been selected and same geometry and boundary conditions were considered within computations. KEYWORDS Achard turbine, cross-flow marine turbine, NACA airfoil NOMENCLATURE

x, y, z

[m] Cartesian coordinates [grd] angle of attack [ ] tip speed ratio [m2/s] water cinematic viscosity [rad/s] rotational velocity [kg/m3] water density [grd] azimuth angle of the blade

Subscripts and Superscripts

D L *

drag (along flow direction) lift (cross flow direction) dimensionless variable

1. INTRODUCTION In 2001, the Geophysical and Industrial Fluid Flows Laboratory (LEGI) of Grenoble, France, launched the HARVEST Project (Hydroliennes Axe de Rotation VErtical STabilis), to develop a suitable technology for marine and river hydro-power farms using crossflow current energy converters piled up in towers [1, 2 & 8]. The hydro-dynamic of these systems is studied at LEGI with the support of the R&D Division of the EDF Group, while other laboratories of the RhneAlpes Region are charged with mechanical aspects (3S-INP of Grenoble and LDMS-INSA of Lyon), as well as with electrical aspects (LEG-INP Grenoble). In 2006, the Technical Civil Engineering University Bucharest, in collaboration with the University Politehnica of Bucharest and the Romanian Academy - Timisoara Branch, started the THARVEST Project, within the CEEX Program sustained by the Romanian Ministry of Education and Research [5]. The THARVEST Project aims to study experimentally and numerically the hydro-dynamics of a new concept

c cD cL c0 d F FD FL g H k p R Rec u = R V0 w

[m] [-] [-] [m] [m] [N] [N] [N] [m/s2] [m] [m2/s2] [Pa] [m] [ ] [m/s] [m/s] [m/s]

airfoil chord length drag coefficient lift coefficient airfoil mean camber line length airfoil maximum thickness hydrodynamic force drag force lift force gravity turbine height turbulent kinetic energy pressure turbine radius chord based Reynolds number transport velocity upstream velocity relative velocity

14

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

of water-current turbine, called Achard turbine, in collaboration with the LEGI partners involved in the HARVEST Project. The Achard turbine, a cross-flow marine or river turbine with vertical axis and delta blades is studied in France mainly with regard to marine applications, to extract energy from tidal currents in costal locations. But the Achard turbines are also suitable to be placed in big rivers, as the Danube, and to produce the desired power by summing elementary power provided by small turbine modules [6 & 8].

radial supports are shaped with NACA 4518 airfoils, while the circular rims are shaped with lens type airfoil. The turbine radius is R = 0.5 m, and the turbine height is H = 1 m. In Figure 2 we present the Achard turbine geometry generated in MATLAB (the upper and lower rims are not represented here). Along each delta blade, the airfoil mean camber line length c0 varies from 0.18 m at z = 0 , to 0.12 m at the extremities, where z = 0.5 m. Between the leading edge of the blades extremity and the leading edge of the blade at mid-height of the turbine, there is a 30 azimuth angle.

Figure 1. The Achard turbine [LEGI courtesy]


In Figure 1 we present the Achard turbine that is studied now in Romania. The turbine description and the generation of its geometry are explained within the section 2 of this paper. In this paper we focus on the 2D numerical modelling of the stationary flow around a blade of the Achard turbine. The blades are shaped with NACA airfoils. The 2D modelling corresponds to the flow around an airfoil in a horizontal cross-section of the runner, situated at one quarter of the turbine height, starting from the top. Different azimuth angles of the blade are analysed. Two types of airfoils are considered within the section 3: a curved one (NACA 4518) and a straight one (NACA 0018). The simulations are performed both with COMSOL Multiphysics 3.3 software and with Fluent 6.01 software. The results and the software capabilities are compared within the section 4 with experimental data. The paper conclusions are summarized in section 5. 2. ACHARD TURBINE DESCRIPTION The vertical axis Achard turbine from Figure 1 consists of a runner with three vertical delta blades, sustained by radial supports at the mid-height of the turbine, and stiffened with circular rims at the upper and lower part of the turbine. The blades and their

Figure 2. The Achard turbine geometry

At z = H 4 = 0.25 m, the horizontal crosssection of the runner gives airfoils with c0 = 0.15 m (the mean value of the mean camber line length along the delta wing), and with the chord length c = 0.1494 m (see section 3). Within this paper, the 2D computations correspond to the cross-plane placed at z = 0.25 m level (see Figure 3), where the
0.5 0.4 0.3 0.2 0.1

=0

y [m] 0
0.1 0.2 0.3 0.4 0.5 0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5

= 120

= 240

x [m]

Figure 3. Computational runner cross-section

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

15

three blade profiles have a mean camber line length of c0 = 0.15 m. The values of the azimuth angle of the blades in Figure 3 are = 0 o ; 120 o ; 240 o , in counter clockwise direction.
3. AIRFOIL SELECTION

Its dimensionless value is m = m c = 0.03758 . As percentage of the chord, m = 3.758% 4% , so the first digit of the NACA airfoil is 4.
(a) NACA 4518
0.02 0.01

y [m]

The runner blades are shaped with NACA airfoils of four-digit series [13], where the first digit is the maximum upper camber m (as percentage of the chord), the second digit is the distance p of the maximum upper camber from the airfoil leading edge (in tens of percents of the chord), and the last two digits describe the maximum thickness of the airfoil, d, as percent of the chord length. For the Achard turbine blades, we consider d = 18 %. In a xOy plane, the airfoil coordinates {x, y} are nondimensionalised with respect to the chord length c, with x = 0 at the leading edge and with x = 1 at the trailing edge (the dimensionless variables are denoted with an asterisk). The dimensionless coordinates x , y s of the airfoil mean camber line are defined as:

0.01 0.02

0.05

0.1

0.15

x [m]

(b) NACA 0018


0.02 0.01

y [m]

0.01 0.02

0.05

0.1

0.15

x [m]

Figure 4. Blade profiles for c0 = 0.15 m: (a) curved airfoil NACA 4518; (b) straight airfoil NACA 0018 Thus, the airfoil type corresponding to the Achard turbine blades is NACA 4518, an airfoil with the mean camber line along the runner circumference (Figure 4a). The computations from section 4, are performed for the profile NACA 4518 and also for the straight profile NACA 0018, which can be generated = 0 , since m * = 0 in (1). For the from (2) for y s NACA 0018 airfoil, the chord length is c = c0 = 0.15 m (Figure 4b). That last choice is due to the fact that experimental and numerical data are available for such a straight profile, corresponding to a Darrieus marine turbine, a vertical axis cross-flow turbine with two straight blades of 0.15 m chord length [3, 4, 7 & 9].
4. NUMERICAL RESULTS

ys =

m p
2

(2 p

1 x for 0 x < m , and

ys =

(1 p )

1 2 p + 2 p x - x 2 for
m x 1 ,

(1)

where p = 0.5 , since the mean camber line of the airfoil is along the circle of radius R = 0.5 m, as in Figure 3. So, the second digit of the NACA airfoil is 5, because the maximum upper camber is placed at a half-distance between the leading edge and the trailing edge. The coordinates x , y of the upper and lower surfaces of the NACA airfoil are defined by:

y = ys

d 0.29690 x 0.12600 x 20

2 3 4 0.35160x + 0.28430x 0.10150x .

(2)

The chord length c can be expressed upon the runner radius and the mean camber line length c0 :
c = 2 R sin (c0 2 R ) .

The numerical simulations are performed with COMSOL Multiphysics 3.3 [11] (a Finite Element Method based software), and with Fluent 6.01 [12] (a Finite Volume Method based software), using the k turbulence model. The stationary 2D flow around a blade of the Achard turbine is analysed both for the profile NACA 4518 (Figure 4a), for which the mean camber line is along the circle of radius R = 0.5 m, and for the straight profile NACA 0018 (Figure 4b), for which other results are available in literature.
4.1. Geometry of the models The 3D blades geometry is realised in MATLAB as in Figure 2. At a certain z-level, each blade profile is generated in 101 nodes on the upper airfoil surface, and 101 nodes on the lower airfoil surface, using (2). The dimensional (x, y) coordinates of those 202 nodes are then imported within the two software used for

(3)

For R = 0.5 m and c0 = 0.15 m, we obtain the mean chord length c = 0.1494 m at z = 0.25 m. The maximum upper camber is defined as:
m = R (1 cos(c0 2 R )) .

(4)

16

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

the flow simulation. The 2D flow domain extent in the xOy plane is scaled with respect to the NACA profiles mean camber line length that is c0 = 0.15 m. The origin of the system is placed on the profile chord, at one quarter from the leading edge. The rectangular flow domain extents ranges from 15c0 to + 25c0 on x-direction, and from 3.5c0 to + 3.5c0 on y-direction. Both COMSOL Multiphysics software and Fluent software use the same geometry of the flow domain. The discretization has the same number of nodes on the profiles, but the total number of cells is slightly different from one software to the other (see Figure 5). The meshing consists of triangular cells in COMSOL and quadrilateral cells in Fluent. For about 33000 cells, the resulting run-time in COMSOL exceeds 10 hours,

on a PC with 2 GB of memory and 3 GHz Intel processor, while in Fluent it is less than 25 minutes, on a PC with 4 GB RAM and 3.2 GHz dual-core Intel Xeon processor. Within both software, the following boundary conditions are considered: At the left side of the domain, on the water inflow boundary, a constant upstream velocity V0 = 4.71 m/s and a turbulent intensity of 2% are imposed. At the right side of the domain, on the water outflow boundary, a zero relative pressure is considered. On the upper wall, as well as on the lower wall of the domain, a slip symmetry condition is selected. On the profile surface, a logarithmic wall function is selected. In Fluent, a discretization of second order is used both for the velocity and pressure.

Figure 5. Zoom of the domain discretization at = 0 o , in COMSOL Multiphysics (upper image) and in Fluent (lower image)

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

17

The chord based Reynolds number is defined as Rec = w c , where w is the relative velocity. The tip speed ratio = R V0 has the imposed value = 2 . The relative velocity w on the blade at the leading edge is obtained by composing the upstream velocity V0 and the transport velocity u = R = V0 :
w = V0 1 + 2 cos + 2 = V0 1 + 4(1 + cos ) ,

(5)

where the azimuth angle defines the position of the blade around the circle, in counter clockwise direction. For our simulations, the Reynolds numbers exceed 7 10 5 for the whole range of the azimuth angle, 0 o ; 360 o , placing the phenomenon behaviour within the self modelling region with respect to the Reynolds number. In this paper, 2D numerical computations are performed on a fixed blade, the value of the upstream velocity being taken so that the Reynolds number on the fixed blade exceeds 10 5 , thus the flow may be assumed to have the same characteristics as in the real rotating case. The angle of attack , considered between the chord and the w-direction at the leading edge, is also related to the azimuth angle through the following relations:

within the range V0 w 3V0 . The velocity ratio w V0 upon the azimuth angle, that is w V0 = f ( ) , as well as the angle of attack variation = ( ) are presented in Figure 6. In this paper, most of the results are computed for discrete values of the pair of angles { ; } . Due to the symmetry, for the straight profile NACA 0018, the flow behaviour obtained at a certain positive value, is identical to the one obtained at the corresponding negative value.
4.2. COMSOL Multiphysics versus Fluent We present the results, namely the pressure field and/or the velocity field around the blades, obtained with both software, for the two types of profiles, NACA 4518 and NACA 0018, at different azimuth angles (see Figures 79). The results show that the flow behaviour description is similar in both COMSOL Multi-physics and Fluent software. Unfortunately, it is difficult to fit the same colour scheme on both software post-processors. Fluent seems to give more accurate results. For MATLAB users, the advantage of using COMSOL Multiphysics is obvious in pre-processing step, where the geometry can be generated directly with MATLAB sources. 4.3. Influence of the number of elements Within this paragraph we will discuss the influence of the number of computational cells. The simulations realised in COMSOL Multiphysics need less than 10 minutes when the grid consists of about 17000 cells, while the run-time is 60 times greater for twice as many cells (e.g. about 10 h for near 33000 cells). In Figure 10 we present the results obtained using COMSOL for the velocity field around the blade profiles NACA 4518 and NACA 0018, at = 30 o = 120 o , for about 16500 cells, and 32800 cells respectively. It seems that the flow behaviour (e.g. the vortices development) doesnt change substantially when multiplying by two the number of elements. So, to get a general view of the flow, computations with about 17000 cells can be acceptable due to the significant run-time economy. Of course, as long as pressure coefficients are concerned, the values obtained for 33000 cells are closer to the experimental existing results. Highly accurate computations are reported by Ervin et al [4] for the flow within a Darrieus marine turbine with two-blades of NACA 0018 profile. They used the TurbFlow software, which is less dissipative than COMSOL and Fluent. For more than 140000 cells, the computations are really time consuming (about 50 days) on a workstation with 8 GB of memory and 3.2 GHz dual-core Intel Xeon processor.

= arccos

1 V02 + w 2 c , for arcsin 2 V0 w 2R


1 V02 +w c , for + arcsin 2 V0 w 2R

)
)

( = arccos

0 o ; 180 o , and
2

(6)

180 o ; 360 o .
4 3.5 3 2.5 40

velocity ratio angle of attack

30 20 10 0

w/V0

2 1.5 1 0.5 0 0 90 180 270 360

-10 -20 -30 -40

Figure 6. Velocity ratio w V0 [] and angle of attack [grd] versus the azimuth angle [grd]
So, the resulting angle of attack varies within the range 30o ; 30o , more precisely, from 29.85 o

to + 29.85 o . For = 2 , the relative velocity (5) varies

18

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

Figure 7. NACA 4518 profile (left images) and NACA 0018 profile (right images), at = 0 o , in COMSOL Multiphysics (1st & 3rd row) and in Fluent (2nd & 4th row): pressure (first 2 rows) and velocity (last 2 rows)

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

19

Figure 8. Velocity field for NACA 4518 profile (left images) and NACA 0018 profile (right images), at = 60 o , in COMSOL Multiphysics (upper row) and in Fluent (lower row)

Figure 9. Pressure field (left images) and velocity field (right images) for NACA 4518, at = 120 o , in COMSOL Multiphysics (upper row) and in Fluent (lower row)

20

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

Figure 10. Velocity field around the NACA 4518 (left images) and NACA 0018 (right images), at = 30 o = 120 o , in COMSOL Multiphysics, for about 16500 cells (upper row) and 32800 cells (lower row)

Figure 11. Pressure field (left image) and velocity field (right image) around the NACA 4518, at = 30 o = 240 o , in COMSOL Multiphysics (about 33000 cells)

In Figure 11 we present the results obtained using COMSOL for the pressure field and the velocity field, around the blade profile NACA 4518, at the angle of attack = 30 o and azimuth angle = 240 o , for about 33000 cells. The flow behaviour can be compared with the one obtained for the blade placed in mirror position (with respect to the flow direction) in the Figure 10 (see left image, lower row).
4.4. Results at different angle of attacks

the flow around the NACA 4518 and NACA 0018 profiles, at different values of the angle of attack . The drag coefficient c D and the lift coefficient c L are computed for each profile, based on the elementary drag force (dFD ) distribution, and on the elementary lift force (dFL ) distribution around the profile boundary, using the following relations:
cD = 2 c w2

Within this paragraph we present some results obtained with COMSOL Multiphysics software for

dF

and c L =

2 c w2

dF

(7)

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

21

We mention that the above curvilinear integrals, which can be computed within the software preprocessor, give the resulting drag force and lift force on the profile, as:
FD = dFD and FL = dFL .

(8)

In Figure 13, our computed results for the NACA 4518 profile are plotted together with our results for NACA 0018, in order to show that there are no significant differences between the drag and lift coefficient values, due to the slight curvature of the 4518 airfoil with respect to the straight-one.
5. CONCLUSIONS

The numerical values of the drag coefficients and lift coefficients are compared in Figure 12 with the experimental data obtained at Sandia Laboratories [10] for the NACA 0018 airfoil, both in c D = c D ( ) and c L = c L ( ) representations, but also in polar diagram c D = c D (c L , ) . We notice that there are some differences between our computed values and the experimental ones, especially at low angles of attack (see the polar diagram).
NACA 0018
2 1.5 1.5 Sandia Labs [10] COMSOL results 1

0.5 0 0.5 0 50

[grd]

100

150

200

1.5 1

cL 0.5
0 0.5 1

0.5

50

[grd]

100

150

200

0.5

1.5

cD

Figure 12. Drag and lift coefficients versus the angle of attack (left images), and polar diagram (right image) for the NACA 0018 airfoil: numerical results in COMSOL and experimental data [10]
COMSOL Multiphysics results
0.8 NACA 4518 0.6 NACA 0018

0.4 0.2 0 30

20

10

[grd]

10

20

30

In this paper, 2D numerical computations are performed both with COMSOL Multiphysics 3.3 software and Fluent 6.0.1 software, in order to depict the stationary flow behaviour around a cross-section of a fixed blade of the Achard turbine. The value of the upstream velocity is taken so that the Reynolds number on the fixed profile exceeds 10 5 , thus the flow may be assumed to have the same characteristics as in the real rotating case. As mentioned above, both software produce similar results. Fluent results seem to be more accurate, but this is probably due to our greater experience in using Fluent. The run-time also seems to favour the use of Fluent for more complicated problems (i.e. greater amount of grid cells), but as the two computers used for the numerical simulations are far from having the same computational power, this statement must be observed with reserves. On the other hand, for profiles that can be generated from mathematical known equations (like the NACA profiles), the interoperability between COMSOL and MATLAB software gives a plus on the geometry generation aspect for COMSOL. Although global results with respect to the pressure coefficients on the airfoils agree well with experimental static results, it is obvious that such computational approaches cannot predict the dynamic stall phenomenon that is reported in the case of vertical axis cross-flow turbines. Moreover, 2D simulations can be acceptable for blades with constant crosssectional profiles along the z-axis, but do not permit an accurate description of the flow for vertical axis cross-flow turbines with varying blade cross-section along the z-axis like the delta blade of the Achard turbine. Those last two aspects are to be considered in further work performed by our team.
ACKNOWLEDGMENTS

1.5 1 0.5

0.5 1 1.5 30 20 10 0 10 20 30

Figure 13. Variations c D = c D ( ) and c L = c L ( ) for the NACA 4518 and 0018 airfoils

[grd]

Authors gratefully acknowledge the CEEX Programme from the Romanian Ministry of Education and Research, for its financial support under the THARVEST Project no. 192/2006. Special thanks are addressed to Dr Jean-Luc Achard, CNRS Research Director, and to PhD student Ervin Amet from LEGI Grenoble, France, for consultancy and documentation on the Achard turbine.

22

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

REFERENCES
1. Achard J.-L., Imbault D., Matre T. (2005) Dispositif de maintien dune turbomachine hydraulique. Brevet dpos, Code FR 05 50420, Titulaire: Institut National Polytechnique de Grenoble, France 2. Achard J.-L., Matre T. (2004) Turbomachine hydraulique. Brevet dpos, Code FR 04 50209, Titulaire: Institut National Polytechnique de Grenoble, France 3. Amet E., Pellone C., Matre T. (2006) A numerical approach for estimating the aerodynamic characteristics of a two bladed vertical Darrieus wind turbine, Scientific Bulletin of the Politehnica University of Timisoara, Transactions on Mechanics, vol 51 (65), pp 95-102 4. Amet E., Pellone C., Matre T., Achard J.-L. (2006) Dcrochage tourbillonnaire larrire des pales dune turbine Darrieus. In: 18me Congrs Franais de Mcanique, Grenoble, France, 9p. (to appear) 5. Georgescu A.-M., Georgescu S.-C., Bernad S. et al. (2006-2008) Interinfluence of the vertical axis, stabilised, Achard type hydraulic turbines (THARVEST), CEEX Programme, no. 192/20.07.2006, AMCSIT Politehnica Bucharest, Romania, Available at: http://hidraulica. utcb.ro/tharvest/ 6. Georgescu S.-C., Georgescu A.-M., Damian R. M., Achard J.-L. (2007) Past and future of water turbines in Romania. In: Proceedings of the 5th International Water History Association Conference on Pasts and Futures of Water, Tampere, Finland, 8p. (to appear)

7. Matre T., Achard J-L., Guittet L., Ploeteanu C. (2005) Marine turbine development: Numerical and experimental investigations. Sci. Bull. of the Politehnica University of Timisoara, Transactions on Mechanics, vol 50 (64), pp 59-66 8. Matre T., Antheaume S., Buvat C., Corre C., Achard J.-L. (2007) An innovative modeling approach to optimize the design configurations of marine (river) cross-flow current energy converters farm. In: Proc. of the 7th European Wave and Tidal Energy Conference, Porto, Portugal, 7p. (to appear) 9. Ploeteanu C. (2004) tude hydro-dynamique dune type dhydraulienne axe vertical pour les courants marins. Thse de Doctorat Institut National Polytechnique de Grenoble, France 10. Sheldahl R. E., Klimas P. C. (1981) Aero-dynamic characteristics of seven airfoil sections through 180degrees angle of attack for use in aerodynamic analysis of vertical axis wind turbines, SAND80-2114, Sandia National Laboratories, USA 11. *** (2006) COMSOL Multiphysics 3.3. Users Guide, COMSOL AB., Stockholm, Sweden 12. *** (2001) FLUENT 6. Users Guide, Fluent Inc., Lebanon, USA 13. *** (2006) UIUC Airfoil Coordinates Database. Version 2.0 (over 1550 airfoils), UIUC Applied Aerodynamics Group, Department of Aerospace Engineering, University of Illinois, Urbana-Champaign, USA, At: http://www.ae.uiuc.edu/m-selig/ ads/coord_database. html

S-ar putea să vă placă și