Turbulent Rotating Disk Flow with Inward Throughow

S ebastien Poncet, Marie-Pierre Chauve, and Patrice Le Gal

I.R.P.H.E., UMR 6594 CNRS-Universit e dAix-Marseille I & II Technop ole Ch ateau-Gombert, 49 rue F. Joliot-Curie, BP 146, 13384 Marseille c edex 13 - FRANCE - Fax. 33 (4) 96 13 97 09 The evolution of the entrainment coecient K of the rotating uid in a rotorstator cavity with an inward throughow and prerotation is studied according to the ow parameters. Measurements are realized in water for a turbulent Batchelor type of ow with two separated boundary layers on rotating and stationary disks by the means of a laser Doppler anemometer and the results are compared to those performed by pressure transducers. We show that the entrainment coecient K depends on a local ow rate coecient Cqr according to a 5/7 power law whose coecients depend on the boundary conditions. A theoretical analysis conrms this behavior of K .

I.

INTRODUCTION

[1] was the rst to determine the ratio K between the mean azimuthal velocity V of the turbulent ow in a rotor-stator cavity, and that of the disk r at the same radius, where is the angular velocity of the rotating disk and r the local radius. He recommended a value of K equal to 0.5 and independent of the radial location. Then [2] proposed a ow pattern divided into three layers: two of them localized in the vicinities of the walls and a third one, the core, located between them. He calculated a theoretical value of K = V /r = 0.512, in the core and measured an experimental value equal to 0.357. He interpreted this discrepancy by the existence of a shear stress in the small radial gap between the rotating disk and the xed cylindrical endwall. [3] performed a similarity analysis by solving the system of dierential equations relative to the stationary axisymmetric ow between two innite disks. He

Electronic address: poncet@irphe.univ-mrs.fr,Tel.33(4)96.13.97.75 Electronic address: chauve@irphe.univ-mrs.fr,Tel.33(4)96.13.97.76 Electronic address: legal@irphe.univ-mrs.fr,Tel.33(4)96.13.97.79

2 specied also the formation of a non-viscous core in solid body rotation conned between the two boundary layers, which develop on the disks. This division of the ow in three distinct zones was the subject of an intense controversy: [4] found indeed that the tangential velocity of the uid can be zero everywhere apart from the rotor boundary layer. [5] proved nally the existence of a multiple class of solutions discovered numerically in fact by [6]. [7] studied the eects of the disk speed and of the inter-disk space h, on the structure of the turbulent ow by pressure and velocity measurements. They proposed constant values of K along a radius and found that K is sensitive to the variations of h. [8] found that K tends to 0.5 in the case of a turbulent ow between a stationary and a rotating disk of innite radius. In the eighties, [9] studied the ow in a rotor-stator system with a large radial gap and showed that the value 0.313 is a good estimation of K for 0.3 < r/R2 < 0.5 (r/R2 is the ratio between the local radius and the rotating disk radius) and that for greater radius, K is an increasing function of r. The measurements performed by [10] extend the self-similar ow area for all the radii lower than 0.75 and also show that K depends strongly on the radial position beyond this radius. By adding a ow prerotation equal to the disk velocity at the periphery, the radial dependence of K is much stronger. Thus, it appears that the evolution of the entrainment coecient K of the rotating uid along the radius is largely inuenced by the boundary conditions. [11] showed numerically, that the closed or partially closed inter-disk ows, can be classied in two families according to the inter-disk space h. [12] then [13] will conrm later this classication by their experimental contributions. When h is larger than the boundary layer thicknesses, the boundary layers are separated, the averaged ow is classically separated into three zones and belongs to the Batchelor type family. The rst zone is relative to the boundary layer developed on the xed disk: it is called the B odewadt layer. The tangential velocity of the ow varies between K r in the core to zero on the stationary disk. Note that this B odewadt layer is highly unstable and was the subject of several detailed studies by [14], [15], [16] and [17]. The second zone is dened by a tangential speed equal to K r and a quasi zero radial velocity: the core. The third zone is associated with the boundary layer which develops on the rotor: it is the Von K arm an or Ekman layer. In this zone, the tangential velocity varies from r on the rotating disk to that of the core K r. When a radial inow or outow is imposed, fewer results have been published. [18] measured the average velocity proles in the case of rotor-stator systems with an imposed outward

3 throughow and kept the same classication as specied by [19] in the case of closed ows. These proposed also a qualitative representation of the average pseudo-streamlines that they observed from their numerical simulations. They show that, in the case of a turbulent Batchelor ow with a weak throughow, the ow at the periphery preserves the properties of the ow without ux with the ux passing through the B odewadt layer and compressing the core. When approaching the center of the disk, the Ekman layer, which was centrifugal, becomes centripetal at a certain radius. A stagnation line is thus created on the rotor. It is similar to that observed by [10]. With a strong centripetal ow, both boundary layers are centripetal. [20] proposed a one-dimensional model to calculate K and introduced a global coecient of throughow rate. They validated their model by experimental measurements and showed the dominating inuence of a centripetal ow in the determination of K and on the distribution of pressure along the radius. But their model remains complex and dicult to use. [21] realized a comparative study between experimental results relative to a rotating centripetal ow and the computations obtained with a numerical model developed from an asymptotic approach. But the limitations inherent to the turbulence models and to the representations of the boundary conditions, do not allow to carry out reliable predictions. Later [22] proposed some numerical calculations of turbulent rotating ows based on the use of a new modelling of the Reynolds stress tensors. But there also, the authors emphasis the too high laminarization of the ow in comparison with the expected results. The interest in this kind of ow is multiple. Of course it has a major interest in turbomachinery, but examples where rotation and throughow are associated can also be found in geophysics or astrophysics. Moreover, from a fundamental point of view, the rotor-stator problem is one of the simplest conguration where exact solutions of the Navier-Stokes equations can be found and where rotation inuences turbulence modelling: this is the reason why they often become benchmarks for numerical simulations. Our present study relates to the determination of K when a centripetal ux is added on the turbulent rotating uid. Following the analysis performed by [23], we show that a local ow rate coecient Cqr is the similarity parameter of the ow that can be used directly to calculate the entrainment coecient K of the uid. As these values are connected to the pressure gradient across the cavity, a comparison between velocity and pressure measurements is performed. The paper is divided as follows: 2 is the description of the experimental set-up whereas 3 is devoted to the analytical model that we deduced from friction consid-

4
z
Pump

c
STATOR

interdisk space

h
ROTOR

Breakthrough crown

R2 R3

Tank

R1

Motor

Figure 1: Schematic representation of the experimental set-up and notations. d = 55 mm is the radius of the central opening, R1 = 38 mm and R2 = 250 mm the inner and outer radii of the rotating disk and R3 = 253 mm the outer radius of the cavity whose height h can be adjusted between 0 and 12 mm. The groove depth c is xed to 5 mm.

erations. This analysis leads to the determination of the behavior of K versus Cqr . In 4, the experimental results are presented and discussed before concluding in 5.

II.

EXPERIMENTAL SET-UP

The geometrical characteristics of our rotating disk cavity are displayed in gure 1. It consists of a cylindrical cavity composed by a xed disk (the stator) and a smooth rotating disk (the rotor). A xed shroud encloses the cavity. The rotor and the central hub attached to it rotate at the constant angular velocity . The ow depends mainly on three control parameters: the aspect ratio G, the global Reynolds number ReR2 and the dimensionless ow rate Q dened as follows: G = h/R2 ,
2 / , Q = Q/(R2 ), where is the kinematic viscosity of water, R2 the rotating ReR2 = R2

disk radius and Q the throughow rate. The height of the cavity h is variable between 0 and 12 mm. The radial gap e = R3 R2 is xed to 3 mm (R3 the outer radius of the cylindrical housing). As the disk rotates, water is sucked through the central opening of the cavity situated above the hub and whose radius d is equal to 55 mm. The incoming uid is entrained in rotation while passing through a breakthrough crown mounted underneath the rotor and linked to it, which enables to increase the tangential velocity of the uid, and

5 consequently limits the inuence of the non rotating cylindrical wall. This prerotation is achieved through 48 holes having a diameter of 10 mm calibrated in order to entrain enough uid when the disk rotates. It ranges between 0.43 and 0.5 for the considered values of the ow control parameters. A pump permits to impose a centripetal and variable ux Q. The measurement of the ow rate is performed by an electromagnetic ow-meter, located at the exit of the cavity. The rotation of the disk is given by a 5.5 KW electric servomotor. A variable speed numerical controller controls the angular velocity . The precision on the measure of the angular velocity and on the ux is better than 1%. In order to avoid cavitation eects, the cavity is maintained at a pressure of 2 bars. Pressurization is ensured by a tank-buer and is controlled by two pressure gauges. The temperature is also maintained constant (23 C ) by a special water cooling device in order to keep constant the density and the kinematic viscosity of water. The measurements are performed by the means of a laser Doppler anemometer (LDA) of the Dantec company and also by pressure transducers. The LDA technique is used to measure the tangential velocity in the vertical plane (r, z ) at a given azimuthal angle. The main defect of this non intrusive method is due to the size of the probe volume (0.8 mm in the axial direction) that alter the quality of the velocity eld measurements in the boundary layers. Pressure is measured by 6 piezoresistive transducers. These transducers are highly accurate (0.05% in the range 10 to 40 C ) and combine both pressure sensors and temperature electronic compensations. They are xed to the stator at the following radial positions 0.093, 0.11, 0.14, 0.17, 0.2 and 0.23 m located on two rays. Previous pressure measurements by embarked pressure gauges ([23]) showed that the pressure on the rotor and the one on the stator are the same within 2.5%. This is in fact a direct consequence of the Taylor-Proudman theorem, which forbids axial gradients in fast rotating ows.

III. A.

ANALYTICAL MODEL Governing equations

The turbulent ow is governed by the Navier-Stokes equations written in cylindrical coordinates (r, , z ). We note (Vr , V , Vz ) the components of the velocity eld and P the pressure. As we saw previously, for large h compared to the boundary layer thicknesses, the

6 mean ow is of Batchelor type with a central core and two boundary layers ([7]). The aim of our analysis is to describe the ow in the central core. In this region, it is well known that Vr Vz 0 (see [20]). Moreover, the tangential velocity remains constant along the z axis and so depends only on the radial position (see g.2). Thus, it appears that the totality of the ux passing through the cavity is conned in the boundary layers. So the Navier-Stokes equation for the tangential component reduces to the balance of the centrifugal force and of the radial pressure gradient: V2 /r = P/r. Then, if we dene the following dimensionless quantities: r = r/R2 , P = P/(1/2(R2 )2 ), we obtain: dP /dr = 2K 2 r . The reference pressure is measured at the outer radial position r = 0.92. Thus, we dene a pressure coecient, as follows : Cp (r ) = P (r ) P (r = 0.92). Finally the resulting equation that will permit later a comparison between the pressure eld and the velocity eld, writes: dCp (r )/dr = 2K 2 r (1)

B.

Denition of the local ow rate coecient Cqr

The challenge is to nd an equation linking the rate of rotation of the uid K versus h, and Q. It is recalled that we consider a ow with two turbulent boundary layers separated by a central rotating core. First we seek an expression for the thickness of the Ekman layer E . Following the analysis of [24], we suppose that the velocity proles evolve according to the classical one-seventh power law in the boundary layers. Thus, the friction coecient Cf , equal by denition to the normalised shear stress 0 , can be given by the
1/4 Deans formula ([25]): Cfmean = 0.073 Re mean , where Remean is the Reynolds number based

on Umean the mean velocity on the cross-section. In our case, we consider that the mean tangential velocity in the Ekman layer is Umean = (K + 1)r/2 and the characteristic length is the Ekman layer thickness E . Thus, the friction coecient is given by: K +1 r Cf 2
1/4

1/4

(2)

By noting the angle formed by the shear stress at the wall 0 and the tangential direction, the tangential component is expressed by:
0

K +1 = 0 cos r 2

7/4

1/4

(3)

7 The radial component of the shear stress is obtained by the balance between the centrifugal force and the shear stress in a volume element having a height equal to E :
r 0 = 0 sin = 2 rE

(4)

As the angle of streamlines inclination remains constant along the radius r (experimentally veried), we nd the expression for the Ekman layer thickness : E r K +1 2
7/5

r2

1/5

(5)

Secondly, we suppose that the radial friction is controlled in the B odewadt boundary layer by the radial ux QB , with QB < 0. Now, by considering that the radial velocity in the rotating core is zero, the continuity equation reads: VE E = Qt QB 2r (6)

Qt is the total superimposed ux. VE is a bulk velocity in the Ekman layer, proportional to the maximum value of the velocity reached in this layer, with the proportionality coecient : VE = r. Using (6) and the expression of VE , we nally get:
/5 Re1 r (Qt QB ) K=2 2r3 5/7

(7)

/5 3 with the expression of the new similarity parameter Cqr = Qt Re1 r / (2r ), where Rer =

r2 / is the local Reynolds number. The nal expression can then be expressed as follows: K = 2(a Cqr + b)5/7 1 (8)

with a and b two experimental constants. b is obtained from the case without ux Qt = 0. Note that, for r = R2 , the expression of Cqr is the same as the one dened in [20].

IV.

EXPERIMENTAL RESULTS A. Velocity proles

Figure 2 represents the axial proles of the dimensionless tangential velocity V = V /r for various experimental congurations. We analyze the inuence of the radial position, and of the Reynolds number, for a given aspect ratio G = 0.048 and a ow rate Q = 10317.

8
1

(a)

(b)

(c)

0.9

0.9

0.9

0.8

0.8

0.8

0.7

0.7

0.7

0.6

0.6

0.6

z*0.5
0.4

0.5

0.5

0.4

0.4

0.3

0.3

0.3

0.2

0.2

0.2

0.1

0.1

0.1

0.5

1.5

0.2 0.4 0.6 0.8

1.2

0.2 0.4 0.6 0.8

1.2

V*

Figure 2: Inuence of the Reynolds number on the axial proles of the dimensionless tangential velocity for G = 0.048 and Q = 10317: ReR2 = 1.038 106 , ReR2 = 2.076 106 , 2 ReR2 =

4.151 106 - (a) r = 0.44 - (b) r = 0.68 - (c) r = 0.98.

In each case, the existence of the three characteristic zones: the two boundary layers and the rotating core, is conrmed. Moreover, between r = 0.44 and r = 0.68, the ow is approximatively self-similar. However, as it can be observed on gure 2, K increases close to the center because of the conservation of the angular momentum. In particular, for r = 0.44, K can be greater than one. This means in fact that the uid rotates faster than the disk and consequently the Ekman layer becomes centripetal. Note that contrary to the contrarotating disks experiments of [26] or [27], no inner shear layer is attached to the stagnation point as in our experiment, the vorticity proles do not exhibit any local maximum. Closer to the periphery, at r = 0.98, the ow is largely inuenced by the boundary conditions. The inuence of the local Reynolds number velocity and of the dimensionless ow rate Q on the tangential velocity is presented in gure 3 for G = 0.036. For Q = 10317 and a given radius r = 0.68, the dimensionless tangential velocity in the core K decreases for increasing values of ReR2 . On the other hand, for ReR2 = 1.038 106 , at a given radius r = 0.56, K increases for increasing values of Q . Measurements have been realized for G equal to 0.024, 0.036 and 0.048 and compared in gure 4 for ReR2 = 4.15 106 and Q = 10317, where it is clearly visible that K is not sensitive to such variations of G, as the ows remain in the same regime (turbulent with separated boundary layers).

Figure 3: Evolution of K with (a) ReR2 , for G = 0.036, Q = 10317 and r = 0.68 - (b) Q , for G = 0.036, ReR2 = 1.038 106 and r = 0.56.

Figure 4: Radial evolution of K for ReR2 = 4.15 106 and Q = 10317 for dierent values of G: 2 G = 0.024, G = 0.036, G = 0.048.

B.

Pressure distributions

To complete the experimental analysis of the ow, we performed also pressure measurements by the means of 6 pressure transducers located on the stator on two rays because of geometrical constraints. In gure 5a, we plot the coecient Cp versus r for few relevant cases. As expected, the pressure decreases towards the center of the cavity: Cp is then always negative. Moreover, at a given radial position and for a given angular velocity, it can be observed that Cp decreases for increasing values of the ow rate Q . On the other hand, for a given value of Q , Cp increases for increasing values of ReR2 . According to relation 1, we can determine the entrainment coecient K from the value

10

Figure 5: (a) Radial distributions of the pressure coecient Cp for various ow conditions - (b) Radial proles of K : comparison between the values of K obtained from the pressure measurements and from laser anemometry, for various ow conditions (Q = 10317).

of Cp . To calculate the derivative of Cp , it is rst necessary to perform a polynomial t of the curves Cp versus r . Then, we calculate by nite dierence the derivative of Cp in order to obtain the value of K . Figure 5b compares the variations of K according to the dimensionless radius r for the data series obtained by the pressure sensors and from the LDA measurements. The results are in excellent agreement. The small dierences coming in fact from the calculation of the derivative of the coecient Cp .

C.

5/7 power law

Relation (8) has been tested for many values of Q, and r and for the three values of the inter-disk space h. Figure 6 shows that the whole set of the experimental results are in very good agreement with a polynomial t: K = 2(5.9 Cqr + 0.63)5/7 1. In particular, gure 6 shows that K is independent of the aspect ratio G as three dierent G = 0.024, 0.036, 0.048 congurations have been studied. The explanation of this behavior comes from the fact that the mean ow remains in each case a Batchelor type ow with separated turbulent boundary layers. The 5/7 law has also been validated in the range of Reynolds numbers [6.92 105 4.15 106 ] and of ow rates [1247 18215]. Because of experimental constraints, it was not possible to get a value of Cqr greater than 0.2 while keeping the boundary layers turbulent.

11

Figure 6: 5/7 power law giving K versus Cqr .

For low ux and rather large radius (small values of Cqr ), the ow tends to a self-similar Batchelor type ow, where the entrainment coecient K is constant. The asymptotical value of K for Q = 0 is 0.438. This value is higher that the one obtained in the case of a large radial gap (K = 0.313, [9]), mainly because of the inuence of our prerotation conditions. On the other hand, when one approaches the center of the disk for strong axial ux, the dependence in the power 5/7 becomes dominant. We report also on gure 6 the velocity measurements of [21] in the case of weak prerotation. As it can be seen, these data follow also a 5/7 power law with however smaller coecients as expected from the absence of prerotation in this experiment.

V.

CONCLUSION

We studied in this experimental work the behavior of the entrainment coecient K of the turbulent ow in a rotor-stator cavity, as a function of the Reynolds number, of the throughow rate and of the aspect ratio of the cavity. In particular, we determined an analytical law, which enables to calculate the entrainment coecient K versus a local ow rate coecient Cqr . This law has been determined analytically and has been validated by extensive pressure and velocity measurements for dierent values of the inter-disk space h and in a large range of Reynolds number and ow rates. Moreover, we have also determined the structure of the rotating disk ow when the inward ux is added. For a weak throughow, the ow in periphery preserves the properties of the case without ux: the Ekman boundary

12 is centrifugal and the B odewadt boundary layer is centripetal. These layers are separated by a central rotating core. But, for a strong throughow, the ow in the Ekman boundary layer becomes centripetal as the core rotates faster than the rotating disk.

VI.

ACKNOWLEDGEMENT

This research was supported by SNECMA Moteurs, Large Liquid Propulsion, Vernon (France) under grant 2003.021.G.

[1] A.J. St epano. Pompes centrifuges et pompes h elices. Trans. A.S.M.E., 54:334352, 1932. [2] F. Schultz-Grunow. Der Reibungswiderstand rotierender Scheiben in Gehausen. Z. Angew. Math. Mech., 5:191204, 1935. [3] G.K. Batchelor. Note on a class of solutions of the Navier-Stokes equations representing steady rotationally-symmetric ow. Quat. J. Mech. and Appl. Math., 4:2941, 1951. [4] K. Stewartson. On the ow between two rotating coaxial disks. Proc. Camb. Phil. Soc., 49:333341, 1953. [5] H.O. Kreiss and S.V. Parter. On the swirling ow between rotating coaxial disks: existence and uniqueness. Commun. Pure Appl.Math., 36:5584, 1983. [6] G.L. Mellor, P.J. Chapple, and V.K. Stokes. On the ow between a rotating and a stationary disk. J. Fluid. Mech., 31:95112, 1968. [7] J.W. Daily and R.E. Nece. Chamber dimension eects on induced ow and frictional resistance of enclosed rotating disks. A.S.M.E. J. Basic Eng., 82:217232, 1960. [8] P. Cooper and E. Reshotko. Turbulent ow between a rotating disk and a parallel wall. A.I.A.A. J., 13:573578, 1975. [9] A.Z. Szeri, S.J. Schneider, F. Labbe, and H.N. Kaufman. Flow between two rotating disks part 1. basic ow. J. Fluid. Mech., 134:103131, 1983. [10] D. Dijkstra and G.J.F. Van Heijst. The ow between two nite rotating disks enclosed by a cylinder. J. Fluid. Mech., 128:123154, 1983. [11] G.N. Lance and M.H. Rogers. The axially symetric ow of a viscous uid between two innite rotating disks. Proc. R. Soc. London A, 266:109121, 1962.

13
[12] A. Sirivat. Stability experiment of ow between a stationary and a rotating disk. Phys. Fluids A, 3:26642671, 1991. [13] L. Schouveiler, P. Le Gal, and M.-P. Chauve. Instabilities of the ow between a rotating and a stationnary disk. J. Fluid. Mech., 443:329350, 2001. [14] O. Savas. Stability of B odewadt ow. J. Fluid. Mech., 183:7794, 1987. [15] J.M. Lopez. Characteristics of endwall and sidewall boundary layers in a rotating cylinder with a dierentially rotating endwall. J. Fluid. Mech., 359:4979, 1998. [16] L. Schouveiler, P. Le Gal, M.-P. Chauve, and Y. Takeda. Spiral and circular waves in the ow between a rotating and a stationary disk. Exps. Fluids, 26:179187, 1999. [17] G. Gauthier, P. Gondret, and M. Rabaud. Axisymmetric propagating vortices in the ow between a stationary and a rotating disk enclosed by a cylinder. J. Fluid. Mech., 386:105126, 1999. [18] J.W. Daily, W.D. Ernst, and V.V. Asbedian. Enclosed rotating disks with superposed throughow. Technical Report 64, M.I.T, department of civil engineering, 1964. [19] J.M. Owen and R.H. Rogers. Flow and Heat Transfer in Rotating-Disc Systems - Vol.1: Rotor-Stator Systems. Ed. Morris, W.D. John Wiley and Sons Inc., New-York, 1989. [20] J. Kurokawa and T. Toyokura. Study on axial thrust of radial ow turbomachinery. pages 3140, Tokyo, 1972. 2nd International JSME Symposium Fluid Machinery and Fluidics. [21] R. Debuchy. Ecoulement turbulent avec aspiration radiale entre un disque xe et un disque tournant. PhD thesis, Universit e des Sciences et Technologies de Lille, 1993. [22] L. Elena and R. Schiestel. Turbulence modeling of conned ow in rotating disk systems. A.I.A.A. J., 33:812821, 1995. [23] R.M. Gassiat. Etude exp erimentale d ecoulements centrip` etes avec pr erotation dun uide conn e entre un disque tournant et un carter xe. PhD thesis, Universit e de la M editerran ee Aix-Marseille II, 2000. [24] H. Schlichting. Boundary-Layer Theory. McGraw-Hill, New-York, 7th edition, 1979. [25] K. Nakabayashi, O. Kitoh, and Y. Katoh. Similarity laws of velocity proles and turbulence characteristics of couette-poiseuille turbulent ows. J. Fluid. Mech., 507:4369, 2004. [26] J.M. Lopez, J.E. Hart, F. Marques, S. Kittelman, and J. Shen. Instability and mode interactions in a dierentially driven rotating cylinder. J. Fluid. Mech., 462:383409, 2002. [27] F. Moisy, O. Doar e, T. Pasutto, O. Daube, and M. Rabaud. Experimental and numerical study

14
of the shear layer instability between two counter-rotating disks. J. Fluid. Mech., 507:175202, 2004.