You are on page 1of 16

a __

ELSEVIER

l!d
23

Computer methods in applied mechanics and engineering


Comput. Methods Appl. Mech. Engrg. 167 (1998) 17-32

Computational

study into the flow field developed of NACA 0012 airfoils


N. Ahmed, B.S. Yilbas*, M.O. Budair
of Petroleum & Minerals, 1998 Department, King Fahd University 16 January

around a cascade

Mechanical

Engineering

Dhahran 31261, Saudi Aruhia

Received

1998; revised 26 January

Abstract Numerical simulation of flow past airfoils is important in the aerodynamic design of aircraft wings and turbomachinery components. These lifting devices often attain optimum performance at the condition of onset of separation. Therefore, separation phenomena must be included if the analysis is aimed at practical applications. Consequently, in the present study, numerical simulation of steady flow in a linear cascade of NACA 0012 airfoils is accomplished with control volume approach. The flow field is determined by solving two-dimensional incompressible Navier-Stokes equations while the effects of turbulence are accounted for by the k--E model. Boundary layer developed at the suction and the pressure surfaces of the airfoil is investigated together with relevant pressure contours for different angles of attack and solidity. Separation point at the airfoil surface is predicted at high angles of attack. Pressure, lift and drag coefficients are computed and the results are compared with the predictions of isolated single NACA 0012 airfoil as well as the data available in the literature. However, the leading edge rotation is also introduced to determine the effect of leading edge rotation on stall inception of isolated airfoil. It is found that increase in solidity increases the angle of attack at which separation occurs and pressure, lift and drag coefficients are highly influenced by the angle of attack and the solidity. The results of leading edge rotation indicates that the drag coefficient reduces considerably while the lift coefficient increases. 0 1998 Elsevier Science S.A. All rights reserved.

1. Introduction During the first two decades of this century the need for an aerodynamic approach to the design of turbines and compressors was gradually realized. The design of wings and isolated airfoils was well understood and attempts had been made to apply the isolated airfoil approach to turbomachines. As a result of systematic improvement of the aerodynamic art and the compilation of cascade data, it became possible to design an efficient turbomachine. However, more work should be carried out to minimize some operating problems of turbo-machines, such as stalling and surging of compressors. With the development of high-speed computers, Computational Fluid Dynamics (CFD) is emerging as an equally important tool as compared to experiments. At present, CFD is rapidly growing and it can be anticipated that it will eventually replace the physical model testing in many of the cases. Consequently, studies into cascade using CFD becomes fruitful, since it minimizes the experimentation. Rhie and Chow [l] used numerical method to solve two-dimensional Navier-Stokes equations for incompressible flow. The k--E model was applied to turbulent flow with and without trailing edge separation. It was shown that CFD results agreed with the experimental measurements for NACA 0012 airfoils for attached flows at Re = 2.8 X lo6 and Re = 3.8 X 106. Jonnavithula et al. [2] conducted a numerical study for stall propagation in axial compressors. In the numerical study compressor blades were assumed as an isolated linear cascade of airfoils and stall propagation was simulated using vortex tracking method. Experimental results indicated that computational predictions were qualitatively correct.

* Corresponding

author.

004%7825/98/$19.00 0 1998 Elsevier Science S.A. All rights reserved. PII: SO0457825(98)00104-2

Davoudzadeh et al. [3] made a two-dimensional study for the unsteady flow in a linear cascade of J-79 stator blades. Prandtl mixing length model was used to account for the effects of turbulence and the study was focused on the stall inception portion of the overall rotatin g stall problems. Use of higher order turbulence models was suggested for future studies to reduce the discrepancies between computational predictions and measurements. Davidson and Rizzi [4] carried out a study to predict stall over a two-dimensional airfoil. They used a standard explicit Runga Kutta time marching code. They showed that for high Reynolds number (Re = 2 X IOh), Baldwin Lomax model failed to predict the stall whereas Algebraic Stress Model (ASM) did predict the stall at an angle of attack of 16 which was found to be in good agreement with experimental observations. On the experimental side, Day [5] conducted studies using hot wires on two laboratory test compressors to investigate the process leading to the formation of finite amplitude rotating stall ceils. Flow was analyzed both spatially and temporarily to show that model perturbations were not always present prior to stall. Hence, measurements confirmed that unsuspected importance of short length scale disturbances in the process of stall inception. Mathioudakis and Breugelmans [6] carried out measurement of the three-dimensional flow held within a single stage compressor operating in deep rotating stall. The overall features of big stall cells characterized by return flows with high tangential velocity upstream of the rotor had been confirmed. Radial velocities in the cells were found to be much smaller than the axial and circumferential components. Hoffman [7] made measurements on the lift and the drag characteristics and associated flow field over the suction surface of a NACA 0015 airfoil at Re = 2.5 X IO both with and without free stream turbulence (FST). The oil flow technique was used to visualize the flow patterns on the suction surface of the airfoil at different flow intensities. He found that increasing the FST from 0.25% to 9% resulted in an increase in peak lift coefficient of 30% with no measurable change in the slope of the lift coefficient against angle of attack curve. For the case of drag no appreciable change was noticed. Mehta et al. [S] obtained experimental results for separated flow over NASA GA(w)-1 airfoil having 2% trailing edge thickness. The fully separated flow was examined in terms of surface pressure distribution, skin friction, mean velocity profiles and the boundary layer integral properties. Yilbas et al. [9] examined the stall behavior of an isolated compressor rotor experimentally. They showed that averaged flow measurements during full span stall supported the view that the unstalled part of the blading must operate at flows beyond the partial stall zone region. On the other hand, the stall cell and reversed flow are limiting factors at peak performance of the airfoils. Past efforts in this field have met with limited success due to availability of the computing facilities when introducing moving boundaries. In early days. Tenant [IO] presented an interesting analysis for the two-dimensional moving wall diffuser with a step change in area. He concluded that the boundary layer separation delayed when the wall velocity exceeded the upstream flow velocity. An extensive test program was conducted by Modi et al. [I I] to control boundary layer for NACA 63-21X airfoil through leading edge rotation. They showed that, by leading edge rotation, lift increased considerably while some degree of reduction in drag was observed. On the other hand, separation control using moving surf was conducted by Hassan and Sankar [ 121 using a numerical scheme for laminar flow conditions. They showed that leading edge rotation of NACA 0012 airfoil increased the lift and reduced the drag considerably, but the effect of turbulence was left obscure. The present study consists of two parts. In the first part leading edge rotation of isolated NACA 0012 airfoil is considered, in this case, the effect of leading edge rotation on stall inception is investigated. Consequently, flow characteristics including flow field, lift and drag coefficients are computed for different rotational speeds and variable surface area of leading edge rotation. It should be noted that the present study is limited to numerical simulation of the problem, therefore, practicality of leading edge rotation is not primary concern. In the second part, simulation of steady flow in a linear cascade of NACA 0012 airfoils is introduced and the effects of solidity (c/s) and angle of attack (cy) on flow field are computed. In both cases, k--E model is employed to take account for the turbulence while parametric study is conducted to investigate the effects of solidity and the angle of attack on lift, drag and pressure coefficients. To develop foundations for the parametric study, twodimensional incompressible flow passing through the linear cascade is taken into account. In addition, predictions obtained from the present study are compared with the results obtained from previous studies [I, 131.

2. The governing The particular

equations form of the general transport equation which governs the system under study is presented

N. Ahmed et 01. I Comput. Methods Appl. Mech. Engrg.

167 (1998) 17-32

19

below. It is composed of a continuity and two momentum equations. It is assumed that no heat transfer takes place in the system so that energy equation is not required. Further, the flow Mach number considered is low enough to assume flow as incompressible. As a result of this, equation of state is not required. Therefore, the governing equations are: Mass conservation

(1)
Momentum conservation

where U, is the velocity density. 2.1. Turbulence


modeling

component

in the coordinate

directions

x,, p is the local pressure

and p is the fluid

A two-dimensional k--E model is used for the solution of the conservation equations, turbulence and its rate of dissipation. The turbulence kinetic energy is given by:

for the kinetic energy of

The isotropic

dissipation

rate of the turbulent

kinetic energy is given by

where p, = turbulent viscosity, which is CWfPpk2 rate and G is the rate of generation of turbulent kinetic

Second last term in Eq. (4), PE is the destruction energy and is given by G=p,[(z+zJ2].

(5)

At high Reynolds number where local isotropy prevails, the rate of dissipation, E is equal to the kinematic viscosity times fluctuating vorticity. The isotropic dissipation rate E is defined as:

au, du;

=Tigg
An exact transport equation can be derived from the Navier-Stokes equation for fluctuating vorticity, and thus for the dissipation, E. This contains complex correlation whose behavior is little known and for which fairly drastic model assumptions must be introduced in order to make the equation tractable. The outcome of this modeling is the c-equation (Eq. (4)). Together with the k-equation, e-equation forms the so-called k--E turbulence model. Generally, k--E model is valid in regions where the flow is entirely turbulent. Consequently, at high Reynolds numbers k--E model is used for the present computations and its constants are given in Table 1.
Table I Constants

for k-c

model

CP
0.09

C,l I .44

Cc2
1.92

% I .o

q
1.3

fP
I .o

j;
I .o

fi
1.o

E 0.0

20

N. Ahmed rt al.

I Cwnput.

Methods

Appl.

Mrch.

Engrg.

167 (1998)

17-32

2.2. The j&e

volume discretization

The calculation domain is divided into a number of non-overlapping control volumes such that there is one control volume surrounding each grid point. The differential equation is integrated over the control volume. Piecewise profiles expressing the variation of variable 4 are used to evaluate the required integrals. The result is the discretization equation containing the values of 4 for a group of grid points. The discretization equation obtained in this manner expresses the conservation principle for the finite control volume just as the differential equation expresses it for the infinitesimal control volume. 2.3. The discretization procedure

As described earlier, partial differential equations are to be discretized into algebraic equations by using appropriate approximation to obtain a numerical solution to the problem. The procedure followed is the Finite Volume Method. We will describe it in general curvilinear coordinates for the genera1 transport equation for an orthogonal coordinate system. In vector notation, the general transport equation for steady-state situation is given by v. (pU4) This equation Therefore = v. (r*v4, is integrated + s over the finite control volume around note P, as shown in Fig. 1. (7)

[V. W4)l

dv d5 =

sdvd5

Here, S, is the average value of S over the finite control volume and ,I+, u. are the components of velocity vector U in the orthogonal coordinate directions 5 and v, respectively. We represent the total flux across a face of the finite control volume by J for convenience. Focusing our attention on east face:

(10)
It is clear that total flux is composed De,, respectively. of a convective flux and a diffusive flux. We represent them by C, and

(11)
N NW ; i NE

n ..____ __________._____~___...........

Fig. I. A finite control

volume.

N. Ahmed et al. I Comput. Methods Appl. Mech. Engrg.

167 (1998) 17-32

21

(12)

To get the linear algebraic X$=&f@/? Hence, we have

equations,

the source term is approximately

linearized (13)

J, - J,v + J,, - J, = 6% + s, &) Arl A6

(14)

To make further progress, it is necessary to make profile assumptions about the variation of 4 within the finite control volume. For the diffusion flux a linear profile can be assumed. This results in the central discretization, i.e.

(15)
Central discretization is usually not appropriate for the convective flux and may result in non-physical oscillations in the solution. To make the discretization compatible with physical reality a hybrid scheme is used. Depending on the cell Peclet number it uses either an upwind or central discretization for the convective flux C,. Cell Peclet number is defined as

(16)
Using Hybrid C,=pu, C, = pu& C, = pu& Scheme: & + tip Ag , 2 A[, A.$, if -2<Pe<2 (17) (18) (19) at other faces of the finite control volume. Substituting these expressions in

if PC > 2 if P, < 2 are obtained

Similar expressions Eq. (14) it yields (Ap - $)A where

= A,&

+ A.,& + A,& + A,,&

+ S,

(20)

(21) (22)

(23)

(24) A,=A,+A,+A,+A,y (25)

Here us)< represents ub velocity at the east cell face. In this connection it should be clear that this velocity is not an interpolated one but rather it should be calculated at cell faces in contrast to other variables whose values are calculated at the center of the cells and at faces that values are obtained by interpolation. This arrangement is highly beneficial to avoid a non-physical oscillatory solution for the pressure field and to increase the accuracy.

3. Grid generation 3.1. Grid generation

and calculation

procedure

For the accuracy of the numerical scheme it is important that grid clustering should be located at large gradient regions in the flow. The grid was generated using algebraic equations for the boundary nodes and Laplace equation for the interior nodes with an effort to minimize the non-smoothness of the grid but at the same time having grid clustering at regions of larger gradients to obtain an economical and accurate solution. Fig. 2 shows the grid generated for the cascade of NACA 0012 airfoils. The iterative method was found to be very sensitive to the smoothness and orthogonality of the mesh generated. For non-smooth meshes heavy under-relaxation was required to prevent divergence of the solution. This large under-relaxation reduces the convergence rate with the consequence of increased computational effort.

Fig. 2. Grids used in the computation (a) Size 80 X 62, stagger angle = 30 and L./S = 0.55 for infinite cascade; angle = 30 and c/s = 0.83; (c) size = 80 X 62 for isolated airfoil.

(b) sire = 80 X 62, stagger

N. Ahmed

et al.

I Comput.

Methods

Appl.

Mech. Engrg.

167 (1998)

17-X?

23

3.2. Boundary

conditions

To solve the governing equations, comprising the model, boundary conditions are needed at each part of the domain boundary. The problem is solved in x-z plane and infinite linear cascade is considered. The magnitudes of u and w velocities are specified. The incoming flow has been considered to be turbulence free. As a result the values of k and E have been taken to be zero. For pressure boundary condition either the value of pressure at a boundary is needed or the value of flow rate perpendicular to the boundary is to be specified 1141. At the inlet the incoming flow rate to the domain is specified. u, v and m are specified and k = E = 0; and it is assumed that at the exit boundary, convection of flow variables is much larger than the diffusion so that no effect of the downstream values on the upstream flow field. As a result no boundary condition is needed at the exit boundary [ 141. To make our assumption strongly valid, we take the exit boundary at so large a distance (5 chord length) from the airfoil that there is no circulating flow at the exit boundary. $$=O and a4 -=0 dY

where C$ applies to all variables. No slip boundary condition is imposed at the solid boundary. For turbulence quantities, Law of Wall is used to determine their values in the first cell adjacent to the wall. These values serve as boundary conditions for the rest of the domain. 3.2.1. LUM?of the wall boundary conditions for standard k--E model If the law of the wall is applied, then, we suppose that the first computational the turbulent sublayer. At this point the velocity UP is parallel to the boundary

point close to the wall (P) is in and has a logarithmic variation:

u*, called the friction velocity, are defined as I/Z

and yl,

representing

a dimensionless

normal distance from point P to the wall,

and Y,:

= yy

PY,,Y

where r,, is the shear stress at the wall, K is the Von Karman constant, E is a roughness parameter and Y,, is the actual distance from the point P to the wall. It is the value of the dimensionless distance y,: that sets the limits between the different sublayers. For the turbulent layer y, is approximately between 10 and 400. The nose rotation has been considered to explore its potential advantages. The isolated airfoil at three different angles of attack ( 13, 15, 16) has been analyzed. Five percent of the chord on each of the suction and pressure surfaces of the airfoil has been specified with a wall velocity such that at the upper surface, this velocity is in the direction of the flow whereas at the lower surface it is in the direction opposite to that of the flow. The free stream velocity of the incoming flow is set at 50 m/s. Two values of tangential velocities as 50 m/s and 150 m/s, have been considered for the leading edge rotation velocity. 3.3. Calculation procedure

For the general variable 4 the solution to the discretized algebraic equations can be obtained using either direct or iterative methods. Direct method needs the algebraic equations to be linear. If, however, the equations are non-linear then an iterative method is necessary. A very effective algebraic equation solver is the Tri Diagonal Matrix Algorithm. In this method, algebraic

24

N. Ahmed et al. I Com~mt. Methods Appl. Med.

Etzgrg. 167 (1998) 17-32

equations for a row of nodes are solved simultaneously using Thomas Algorithm. Hence, the boundary point information is carried in a single iteration for that row. This process is carried out for each row successively. Most of the time we used this method for the solution of our problem. However, equation for pressure correction was solved using whole field method, since a simultaneous satisfaction of the continuity in the whole domain increased the convergence rate. Usually, several hundred sweeps were required to obtain a converged solution. If the pressure field is given, the solution to the momentum equations can be obtained by employing the method described above. Moreover, unless the correct pressure is employed, the resulting velocity field obtained from the solution of the momentum equations unable to satisfy the continuity equation. However, no explicit equation for pressure is given. This is particularly true for an incompressible flow. In this regard, several methods are available. Consequently, the SIMPLE procedure is used, which is basically an iterative process. Let a tentatively calculated velocity field based on a guessed pressure field p* is denoted by UT, nz. Let the correct pressure p is obtained from: p =p*
+p

(26)
correction in velocities n;, uh can be introduced in a similar manner: (27)

The corresponding UC= UT + Ll;

Making certain assumptions, given by:

the velocity

correction

formula for east face of the mesh element,

for example,

is

Now, discretizing the continuity equation for pressure correction:


(A, - S,,)p, =

equation

and using the velocity

correction

formulas,

one can obtain

an

A,>p,: + A,P:

+ 4~:

+ Amp:. + SC,

(30)

Thus, we have obtained an equation for pressure correction or in turn for pressure. The important steps to compute the flow properties are as follows: l Guess the pressure field p*. l Solve the momentum equations to obtain UT, u;. l Solve the pressure correction equation. l Calculate p by adding p to p*. l Solve equations for other variables 4 (e.g. turbulence kinetic energy) if they have a coupling with momentum equations. l Treat the corrected pressure as a new guessed pressure p *. Return to step 2 and repeat the whole procedure until a converged solution is obtained.

4. Results and discussions The flow in a cascade of NACA 0012 airfoils is studied at Re = 3.24 X 10. An H-type grid was used with 80 X 62 points per passage. Over the airfoil surface 49 points were distributed. In order to increase the accuracy of the calculations and for economy of computations, a high grid refinement was introduced near the solid boundary. The grid independent test was conducted and its results are shown in Fig. 3. Consequently, selection of 80 X 60 mesh points gives sufficient accuracy for the present case. The front and rear outer boundaries are located at 4 and 5 chord distances away from the body, respectively. Single passage periodicity assumption is used to simulate the infinite cascade. At the inlet, the incoming mass flow is specified together with velocity components and turbulence quantities. At the exit, the velocity components and turbulence parameters are extrapolated from the inner solution by assuming that the first derivatives of the flow properties are zero. The angle of attack is varied from 0 to 24 degrees gradually while solidity (c/s) ranges from 0.55 to 0.83.

N. Ahmrd et ul. I Comput. Methods A@.

Mech. Engrg.

167 (1998) 17-32

2.5

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

X/C Fig. 3. Grid independent test results.

Fig. 4 shows the velocity field around the isolated single NACA 0012 airfoil for three different angles of attack. It is obvious that boundary layer thickness is considerably small at the leading edge and increases along the chord. At the upper and lower surfaces of the airfoil, nearly symmetric boundary layer profiles are obtained at angle of attack (a) of 3 as expected. As the angle of attack increases the symmetry disappears and considerable difference occurs. The development of free shear layer at the trailing edge is quite apparent. This free shear layer diminishes in downstream because of the diffusion process. The spilling of the flow is evident at high angles of attack, because of the development of high pressure region on the lower surface of the airfoil at the leading edge. This results in a high velocity to be developed in this region. Therefore, the large difference between the pressures at the two surfaces of the airfoil generates a lift. As the angle of attack reaches to 16 and above, the flow cannot remain attached to the airfoil surface in the down stream because of a high turning angle and thickening of the boundary layer occurs. In this case, it detaches from the surface, i.e. separation is resulted. Fig. 5 shows the pressure coefficients for the two cases of leading edge rotation at different angles of attack. It is evident that up to the extent of 25% chord length, small increase in the pressure difference between the upper and the lower surfaces occur. Moreover, in comparison to the case of without rotation, the pressure on both the upper and lower surfaces has increased somewhat, but the difference in pressure is almost identical for the two cases (i.e. with and without rotation). It is observed that the effect of leading edge rotation is more pronounced for a higher angle of attack (16) as compared to a lower angle of attack. It is quite apparent that a high rotational velocity causes high pressure difference to occur between the two surfaces. Hence, a rotating velocity of 150 m/s is found to be more promising to increase the lift. However, another case has been considered in which rotational velocity has been set at 150 m/s, whereas the 11% chord length has been specified with a wall velocity. Fig. 6 shows the velocity vector for this case at an angle of attack of 16 which can be comparable to no leading edge rotation. It is observed that there is an increased flow spilling at the leading edge. The separation point has shifted downward towards the trailing edge and a reduction in the wake size is quite evident. However, there is no significant effect of rotation on the separation region and still large separation occurs close to the leading edge, therefore, the effect of leading edge surface rotation seems to be very localized. If the separation could be delayed to a significant extent then this method could be very advantageous. Fig. 7 shows the pressure coefficients for the two cases at different angles of attack. It can be seen that suction peak has increased in magnitude for 11% leading edge chord rotation. This has a positive effect on the lift generated. At the trailing edge, the pressure difference is essentially the same. On the pressure surface we observe a pressure kink. This is due to the fact that airfoil surface has a velocity upstream of this point whereas it has zero velocity downstream of this point.

26

(b)

Fig. 8 shows the lift and drag coefficients for the two cases. The case of leading edge rotation has been simulated for only three different angles of attack since at lower angle of attack there is no separation, so there is no effect of delaying of separation by this practice. It is observed that there is an increase in the lift in the case of the rotating nose as compared to the non-rotating nose airfoil. It is further observed that lift increases with the increase in the percentage of the length of the chord that rotates. The slope of the curve corresponding to the leading edge rotation nose airfoil retains the same slope as at lower incidence% i.e. similar to the case obtained from the experiment [ 11. On the other hand, decrease in drag occurs and the reduction in drag increases with increasing angle of attack. The reduction in drag is more pronounced for 11% chord surface rotation at leading edge. At 16 angle of attack the reduction in drag is quite signiticant. The slope of the drag curve reduces or the rotating surface length increases. This is due to the separation point moving further down stream as the rotating surface length increases. In the case of full surface rotation drag coefficient reduces considerably while life coefficient increases. Fig. 10 shows the flow field around the intinite cascade of NACA 0012 airfoils for C./S of 0.55 and 0.83, respectively. It is evident that increasing the solidity (c/s) effects the boundary layer developed around the airfoils. Separation starts earlier in the case of low solidity (c/s). The flow spilling at the leading edge is

N. Ahmed

et al.

I Comput.

Methods

Appl.

Mech.

Engrg.

167 (1998)

17-32

27

-3

-2

u"

-1

01

02

0.3

0.4

0.5

06

07

0.6

09

X/C Fig. 5. Pressure coefficient along with the chord for different speeds of leading rotation.

(4

Fig. 6. Flow field for the isolated airfoil (a) without and (b) with leading edge rotation.

28

N. Ahmcd

ef al.

Compuf.

Methods Appl. Mech. Engrg.

167 (19981 17-.Z.?

01

02

03

0.4

0.5

06

07

0.8

09

x/C
Ftg. 7. Pressure coefficient along the chord length for different rotatmg surface to chord ratio.

10

12

14

16

18

Fig. 8. Lift and drag coefficients

with angle of attack for different

rotating

surface to chord ratto.

N. Ahmed

et al.

I Comput.

Methods

Appl.

Mech.

Engrg.

167 (1998)

17-32

29

Fig. 9. Pressure

coefficient

along chord when full surface rotating

I~,, __--Zc-u-r-----/C/H A

---Lu

_-_/_A_-_

_a--A

Alpha=22

___Y--

de.3 -

--

,l_~_-------

CISd.55
__Iu--

NYU-

-M

Flow field for infinite cascade

at (Y= 22 and stagger = 30, for solidity ratios of (a) c/s = 0.55 and (b) c/s = 0.83

30

N. Ahrrwd

rt al.

I Com/xu.

Methods

Appl.

Mrch.

Eiz,qr,q. 167 (1998)

17-.?2

obvious. Due to symmetrical arrangement of airfoils the separation develops on all the airfoils at the same angle of attack. Fig. 11 shows the pressure coefficient along the chord of the cascade. The suction peak on the airfoil surface

Fig.

I I. Effect

of cascade

solidity

ratio

on the surface

mean

preaaure

coefficients

at different

angles

of attack

(a) CY= 6: (b) u = 12;

(c) a = 20.

N. Ahmed

et 01. I Cmput.

Methods

Appl.

Mrch.

Engrg.

167 (1998)

17-32

31

-*--

Cascade c/s = 0.83 Experiment Isolated ailroilll]

IO

12

14

16

18

20

Fig. 12. Effect of cascade

solidity ratio on (a) hft coefficient

and (b) drag coefficient

decreases as the solidity increases. This is due to the pressure suppression as a result of closely spaced airfoils in the cascade. Isolated single airfoil has the largest suction peak. Fig. 12 shows the lift coefficient corresponding to different c/s and angle of attack. Lift coefficient increases with increasing of angle of attack. The maximum obtainable lift becomes small as the solidity increases. This is due to the loss in suction peak at the upper surface as this is influenced by the pressure suppression of the neighboring airfoil. The angle of attack corresponding to separation point moves toward relatively higher values as the solidity increases. When comparing the present predictions with the previous results [ 1,131, they are in good agreement. However, the drag coefficient increases as the angle of attack increases and it reduces slightly with the increase in solidity ratio. Isolated airfoil has the largest drag for a given angle of attack. It is evident that as the separation is reached the drag increases sharply, whereas the lift drops for isolated airfoil. Moreover, cascade data show increasing trend for both drag and lift coefficients as the angle of attack increases towards 20. This is because no separation is resulted due to pressure suppression effect of neighboring airfoils. When comparing cascade data with previous study [13], it is evident that both results are in agreement.

5. Conclusions A simulation of the flow field around the isolated single with and without leading edge rotation and a linear cascade of NACA 0012 airfoils is carried out using a control volume scheme. Attached and separated flows have been computed using the k--E turbulence model. Lift, drag and pressure coefficients have been determined.

32

N. Ahrnrd

ct cd. I Comput.

Methods

Appl.

Mech.

Etqrg.

I67 (19988) 17-32

The conclusions derived for the flow field studied are as follows: l High degree of flow spilling occurs at the leading edge of the airfoil for high angle of attack. This, then, results in separation at the trailing edge. The point of separation moves towards the leading edge when c/s increases. The effect of pressure suppression is quite evident. l With the increase in incidence the adverse pressure gradient attains large values. l The boundary layer thickness increases towards the trailing edge as the angle of attack increases. The rate of this increase reduces at low solidity. l In the case of leading edge rotation, separation delays at high angle of attacks and diminishes as the total surface rotates. The conclusion obtained from the pressure, lift and drag coefficients may be listed as follows: l Pressure coefficient on the suction surface at the trailing edge attains higher values with incidence and with a decrease in the solidity. The present predictions give closer results to experimental findings as compared to previous studies. l Lift coefficient reduces as large separation occurs. However, this has not been predicted to be drastic as being observed in experiments. Further. the incidence angle at which a drop in the lift occurs has a slightly larger value as compared to available experimental data. This may be due to the steady state analysis employed in the computation. In this case once the separation occurs, the vortex developed stays as it forms rather than detaches from the surface and forming the vortex shed. Therefore, the lift coefficient predicted immediately after the separation may not agree well with experimental results. However, the present results obtained for isolated airfoil give closer values to experimental findings as compared to previous results. l As the solidity increases, the incidence at which maximum lift is obtained, increases. l It is found from the simulation results of leading edge rotation that slight increase in lift, but considerable decrease in drag occur in the case of nose rotation. However, the lift coefficient increases drastically while drag coefficient reduces as the full surface of the airfoil is rotated, in this case. separation vanishes.

Acknowledgment The authors acknowledge Arabia for this work. the support of King Fahd University of Petroleum and Minerals, Dhahran, Saudi

References
[ I] C.M. Rhie and W.L. Chow. Numerical study of the turbulent Ilow past an airfoil with trailing edge separation. AIAA J. 21( I I) (1983) 152%1532. [2] Thangam S. Jonnavithula and F. Ststo, Computational and experimental study of stall propagation m axial compressors, AIAA .I. (1990) 1945-1952. [3] F. Davoudzadeh, N.S. Liu, S.J. Shamroth and S.J. Thoren, Navier-Stokes solution of the turbulent flow lields about an isolated airfoil, AIAA J. 26 (I 990) 242-25 1. [4] L. Davidson and A. Rizri. Navier-Stokes stall predictions using an algebraic Reynolds stress model, J. Space Craft Rockets 29(6) (1992) 794-800. [S] I.J. Day, Stall inception in axial flow compressors, J. Turbomachinery I IS (1993) l-9. [6] K. Mathioudakis and F.A.E. Breugelmans, Three-dimensional flow in deep rotating cells of an axial flow compressor, J. Propulsion Power 4 (1988) 263-269. [7] J.A. Hoffman, Effects of free stream turbulence on the performance characteristics of an airfoil, AIAA J. 29(9) ( 1991 ) 1353-13.54. [8] J.M. Mehta and S. Goradia, Experimental studies of the separated flow over a NASA GA(w)-1 airfoil, AIAA J. 22(4) ( 1984) SS2-554. [9] Ali Koc, B.S. Yilbas and E. Baltacmglu, Some observations on stall behavior of an isolated compressor rotor, Mech. lncorp Engineer (1992) 7-14. [IO] J.S. Tennant, A subsonic diffuser with moving walls for boundary layer control. AIAA J. I I (1973) 240-242. [I I] V.J. Modi, J.L.C. Sun. T. Akutsu, P. Lake, K. McMilan, P.G. Swinton and D. Mullins, Moving-surface boundary-layer control for aircraft operation at high incidence, AIAA J. Aircraft 1X( I I ) (198 I) 963-968. [ 121A.A. Hassan and L.M. Sankar, Separation control using moving surf a numerical simulation, AIAA J. Aircraft 29(l) (1992) I31 - 139. [I31 D.W. Zingg and G.W. Johnston, Interactive airfoil calculations with higher-order viscous flow equations, AIAA J. 29(7) (1991) 1033-1040. [ 141 C. Hirch, Numerical Computation of Internal and External Flows, Wiley Series in Numerical Methods in Engineering, Vol. 2 (John Wiley and Sons Ltd., (Reprint), England, 1991).

You might also like