You are on page 1of 25

CASCADExpert

Reference Manual

Romeo RESIGA

March 2005
2
Chapter 1

Inviscid Incompressible
Irrotational Flows

1.1 Inviscid flows


Inviscid flow theory, if employed judiciously and with a clear understanding
of when it is valid and when it must be replaced by more general theories,
offers considerable simplification to fluid flow problems [8].
In particular, the differential equation governing momentum transport
in inviscid flow possesses only first-order spatial derivatives instead of the
second-order derivatives in the full viscous flow momentum equation. This
seemly minor difference allows the integration of momentum equation to
obtain a solution for pressure in irrotational flows.
For many inviscid flows, the initial conditions and the inlet/outlet con-
ditions are such that one can argue that vorticity must vanish completely.
Highly special though the property of irrotationality may seem to be, it is
given great practical importance by the consequence of Kelvin’s circulation
theorem that material elements of a uniform fluid set into motion from rest
remain without rotation unless they move into a region where viscous forces
are significant. In such irrotational flows the velocity field can be found by
solution of a simple linear equation (the Laplace equation) in regions with
no vorticity. This condition can be applied, for instance, to flow past airfoils
(if one ignores the thin downstream wake), for which the inviscid flow theory
can be used to obtain accurate solutions for the airfoil lift.

3
4 CHAPTER 1. III FLOWS

1.2 Irrotational flows


A flow is said to be irrotational if the vorticity vanishes everywhere, or

∇×v =0 (1.1)

A velocity field that can be expressed as

v = ∇φ (1.2)

satisfies the irrotationality condition (1.1). The scalar φ is called the velocity
potential, or sometimes the ”scalar potential”, to distinguish it from another
vector valued potential. One can prove also that all irrotational vectors must
have a representation of the form (1.2).

1.3 Incompressible flows


For any incompressible flow the divergence of the velocity everywhere van-
ishes. Vectors satisfying this condition are called solenoidal (or divergence
free). Integrating the restriction ∇ · v = 0 over a volume V with bounding
surface S and using the divergence theorem yields
Z
v · n dS = 0, (1.3)
S

where n is the outward unit normal of S. Equation 1.3 states that for any
solenoidal velocity field, the net volumetric flow rate over a closed surface
must vanish.
Any solenoidal vector can be written as the curl of another vector, or

v =∇×β (1.4)

where β is called vector potential. The vector v in (1.4) automatically satifies


the divergence free condition thanks to the vector identity ∇ · (∇ × β) = 0.
Note that the vector potential β is not defined uniquely, since we can add
any gradient of a scalar β + ∇α while satisfying (1.4).
For a two-dimensional flow in a plane x−y, the vector potential is oriented
normal to the plane. Since by definition, in a plane flow ∂/∂z = 0, it results
that the scalar gradient ∇α has to be in the x − y plane, but on the other
1.3. INCOMPRESSIBLE FLOWS 5

hand the vector potential is normal to the flow plane. As a result, for 2D
plane flows ∇α must vanish. The velocity vector potential becomes in this
case β = ψez , where ez is the unit vector normal to the x − y plane. The
function ψ(x, y) is called the stream function, for reasons to be explained be-
low. Substituting this expresion for β in (1.4) gives the velocity components
in the x− and y− directions, respectively, as:
∂ψ ∂ψ
vx = , vy = − . (1.5)
∂y ∂x
The gradient of ψ can thus be written as
∂ψ ∂ψ
∇ψ = ex + ey = −vy ex + vx ey . (1.6)
∂x ∂y
The scalar product of ∇ψ with the velocity v is

v · ∇ψ = 0

meaning that ψ = constant along the direction of the velocity. In other


words, ψ is constant along streamlines, thus the name stream function.
The strength Q of a stream tube can be written for an incompressible
flow as Z Z Z
Q≡ v · n dS = n · (∇ × β) dS = β · dx, (1.7)
A A C
where A is any open surface spanning the cross section of the stream tube
with bounding curve C and unit normal n. The last form of the above integral
follows from Stoke’s theorem.
For the case of two-dimensional flow in the x − y plane, let us consider
the rectangular surface A shown in Figure..., where one edge lies on the x − y
plane and has length L and the other edge is oriented in the z−direction and
has lenght b. The edge lying in the x − y plane ends at points P1 and P2 . The
value of the streamfunction on the two streamlines passing through these two
points is ψ1 and ψ2 , respectively. Since β = ψez , the edges of the surface
A lying tangent to the x − y plane make no contribution to the integral in
(1.7). Taking into account the direction of integration in (1.7), which is in
the positive z−direction on one side of A and in the negative z−direction on
the opposite side, we obtain

Q = b (ψ2 − ψ1 ) . (1.8)
6 CHAPTER 1. III FLOWS

The difference in value of streamfunction between any two streamlines is


equal to the volumetric flow rate passing between the two streamlines per
unit width normal to the plane of motion.

1.4 Flows that are both incompressible and


irrotational
A flow that is both irrotational and incompressible must satisfy both the
condition ∇ × v = 0 of zero curl and the condition ∇ · v = 0 of zero
divergence.
Using the vector identity ∇2 v = ∇(∇ · v) − ∇ × (∇ × v) we obtain for
an irrotational and divergence free velocity field

∇2 v = 0. (1.9)

Substituting (1.2) into the condition of zero divergence gives an equation


for the velocity potential as
∇2 φ = 0. (1.10)
Hence, for a flow that is both irrotational and incompressible, both the
velocity field and its scalar potential must satisfy Laplace’s equation. Any
function that is a solution of Laplace’s equation is said to be a harmonic
function.
We are particularly interested here in 2D plane flows. In this case, the
velocity components can be expressed using either the scalar velocity poten-
tial,
∂φ ∂φ
∇φ = ex + ey = v x ex + v y e y , (1.11)
∂x ∂y
or the streamfunction,

∂ψ ∂ψ
∇ψ = ex + ey = −vy ex + vx ey . (1.12)
∂x ∂y

From (1.11) and (1.12) we have ∇φ · ∇ψ = 0, and as a result the the stream-
lines and equipotential lines are orthogonal. The ensemble of streamlines and
equipotential lines is sometimes called the flownet, for reasons obvious from
Figure 1.1.
1.4. IRROTATIONAL AND INCOMPRESSIBLE FLOWS 7

We have already shown that the scalar velocity potential is a harmonic


function, and for 2D flows it satisfies the Laplace’s equation

∂2φ ∂2φ
+ = 0. (1.13)
∂x2 ∂y 2
On the other hand, applying the irrotationality condition to the 2D velocity
expressed with the streamfunction leads to

∂2ψ ∂2ψ
+ 2 = 0, (1.14)
∂x2 ∂y
meaning that the streamfunction is harmonic too, in 2D.
The complex function which has φ(x, y) as real part and ψ(x, y) as imag-
inary part is called complex potential, and it is the main starting point in
using the complex analysis for solving 2D inviscid, incompressible, and ir-
rotational flow problems. However, these powerful analytical tools are no
longer convenient for general geometries, when one has to employ numerical
techniques.
Although both φ and ψ satisfy the 2D Laplace equation, the accompany-
ing boundary conditions for the same geometry and velocity field are specific
for potential or streamfunction formulations. However, solving the problem
in φ or in ψ leads to exactly the same velocity field. For the present com-
puter code we have chosen to solve the boundary value problem
for the streamfunction, with the boundary conditions specific to
the flow in a linear cascade.

1.4.1 Flow in the neighborhood of a stagnation point.


Straining rate
For streaming motions around airfoils the velocity field has at least one stag-
nation point at or near the leading edge. The precise location of this stagna-
tion point depends on the body geometry, as well as on the angle of attack
of the incoming flow.
For an inviscid flow, the body surface is impenetrable, therefore the ve-
locity can have only tangent component to the body. As a result, the body
contour (in 2D) is itself a streamline. This streamline originates upstream,
and bifurcates at the leading edge stagnation point. We will examine below
the details of the flow in a small neighborhood of the stagnation point.
8 CHAPTER 1. III FLOWS

0.4

0.2

-0.2
y

-0.4

-0.6

-0.8

-1
-0.4 -0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
x

Figure 1.1: Equipotential lines and streamlines for cascade flow, [9].

Let us assume, without loss of generality, that the stagnation point is


located at origin. The velocity vanishes there and hence at the origin we
have
∂ψ ∂ψ
= 0, and = 0. (1.15)
∂x ∂y
Without loss of generality we may also suppose that ψ = 0 at the origin, and
therefore by Maclaurin’s theorem,
1 2 ∂2ψ
· µ ¶ µ 2 ¶ µ 2 ¶¸
∂ ψ 2 ∂ ψ
ψ= x 2
+ 2xy +y + ... (1.16)
2 ∂x 0 ∂x∂y 0 ∂y 2 0
Since the flow is incompressible, we have according to (1.5)
µ 2 ¶ µ ¶ µ ¶
∂ ψ ∂vx ∂vy
B= = =−
∂x∂y 0 ∂x 0 ∂y 0
For a mean rotational flow, with
µ 2 ¶
∂2ψ
µ ¶ µ ¶ µ ¶
∂ ψ ∂vy ∂vx
A1 = =− , and A2 = = ,
∂x2 0 ∂x 0 ∂y 2 0 ∂y 0

we have in general A1 + A2 6= 0. However, if the flow is irrotational then


A1 = −A2 = A and the streamfunction near the stagnation point is

A(x2 − y 2 ) + 2Bxy .
¤
ψ= (1.17)
2
1.4. IRROTATIONAL AND INCOMPRESSIBLE FLOWS 9
p
As a result, the velocity magnitude v = vx2 + vy2 around the stagnation
point is
sµ ¶
2 µ ¶2

∂ψ ∂ψ p
v= + = r A2 + B 2 , with r = x2 + y 2 . (1.18)
∂x ∂y

It follows that near the stagnation point,


sµ ¶2 µ ¶2 sµ ¶2 µ ¶2
∂v ∂vx ∂vx ∂vy ∂vy
= + = + , (1.19)
∂r ∂x 0 ∂y 0 ∂x 0 ∂y 0

where r is the distance measured from the stagnation point.


Note that the equation (1.17) can be rewritten by rotating the system of
coordinates with angle
√ α defined by A = C sin 2α and B = C cos 2α, with
straining rate C = A2 + B 2 [8, p. 125], as

ψ = Cx0 y 0 , (1.20)

with x0 = x cos α − y sin α, y 0 = x sin α − y cos α. This corresponds to the


flow normal to a straight line, which is the prototype for stagnation point
irrotational flow.
The incoming stagnation streamline intersects the straight line (which
coresponds to the local tangent to an arbitrary body) at right angle. The
velocity magnitude increases as we depart from the stagnation point as
∂v/∂r = C = straining rate.

1.4.2 Integration of the equation of motion


In the absence of viscosity the equation of motion becomes
Dv ∂v 1
≡ + v · ∇v = − ∇p + f (1.21)
Dt ∂t ρ
where the density ρ is constant for incompressible flows, and f is the body
force per unit mass. The vector identity 12 ∇(v · v) = v · ∇v + v × (∇ × v)
enables (1.21) to be written in the alternative form
µ 2 ¶
∂v v p
− v × (∇ × v) = −∇ + −f ·x , (1.22)
∂t 2 ρ
10 CHAPTER 1. III FLOWS

where v 2 = v · v. When v = ∇φ, this becomes


∂φ v 2 p
µ ¶
∇ + + − f · x = 0, (1.23)
∂t 2 ρ
showing that the quantity within brackets must be a function of time t alone,
F (t) say. The form of this unknown function is without significance,R because
we could define a new velocity potential φ0 such that φ0 = φ − F (t)dt,
∇φ0 = ∇φ, and thereby remove the function of t without affecting the
velocity distribution. It is customary to ignore the arbitrary function of t
and to write the integral of (1.23) as

∂φ v 2 p
+ + − f · x = constant (1.24)
∂t 2 ρ
throughout the fluid.
In particular, for steady flows with negligible body forces,
v2 p
+ = constant throughout the fluid, (1.25)
2 ρ
which is a particular case of the Bernoulli’s theorem. The relation (1.25)
provides an explicit expression for the pressure when the velocity distribution
is known.
In most of the practical applications the body force is due to gravity, i.e.
f = −gez , where z is the vertical axis. For incompressible flows, the quatity
ρv 2
ps = p + + ρgz (1.26)
2
is the so-called total or stagnation pressure of a fluid particle, [10, p. 52].
The total pressure ps of the fluid of incompressible flows is defined as the
sum of pressure, kinetic and geopotential energies. In frictionless flows ps /ρ
is constant for a fluid particle, i.e. the total energy per unit mass remains
constant.

Relative motions
Frequently it is advantageous to treat a flow with reference to a so-called
relative system of coordinates which undergoes a certain motion with respect
to a fixed absolute system. For example, in turbomachinery applications the
1.4. IRROTATIONAL AND INCOMPRESSIBLE FLOWS 11

relative system turns with the rotor about the centerline of the machine,
whereas the absolute system is rigidly connected with the stationary parts
or casing of the machine. At the normal operating or design point conditions
the rotor turns with constant angular velocity ω.
The relative velocity of a fluid particle, w, is related to the absolute
velocity v, by the relation

v =w+ω×r (1.27)

where r is the radius vector from the origin located on the rotation axis. If
R is the distance from the axis, then

|ω × r| = ωR. (1.28)

In turbomachinery literature this is called the transport velocity u = ωR.


For incompressible, steady and frictionless relative flow, the relative total
pressure
ρw2 ρω 2 R2
ps R = p + − + ρgz (1.29)
2 2
will remain constant along a particular relative streamline, [10] (Bernoulli’s
theorem for relative flows).

1.4.3 Maximum velocity / minimum pressure theorem


In an irrotational flow of an incompressible fluid in a bounded region V, the
maximum velocity magnitude and the minimum pressure must occur on the
boundary S of V, [2, p. 384], [8, p. 82].
In order to prove the velocity maximum theorem, it is convenient to use
the kinetic energy per unit mass (v · v)/2 = v 2 /2. Our task is equivalent to
proving that the maximum value of v 2 /2 must occur on the boundary S of
V.
If the velocity vector is harmonic (1.9) the Laplacian of v 2 /2 is positive
definite, for upon expanding we have
µ 2¶
2 v ∂vi ∂vi
∇ = ≥ 0. (1.30)
2 ∂xj ∂xj

The proof of the velocity maximum theorem follows by contradiction.


Thatis, we suppose that there exists some point P in the interior of V at
12 CHAPTER 1. III FLOWS

which v 2 /2 attains a local maximum. If R is an infinitesimally small region


surrounding the point P , with bounding surface ∂R, then from the assump-
tion that the specific kinetic energy attains a maximum value at P is follows
that v 2 /2 decreases as we progress away from P , or
µ 2¶
v
Z
n·∇ dS < 0, (1.31)
∂R 2
where n is the outward unit normal of ∂R. However, using the divergence
theorem, we can write
µ 2¶ µ 2¶
v v
Z Z
2
n·∇ dS = ∇ dV
∂R 2 R 2
We have already shown that the Laplacian of the specific kinetic energy is
positive definite, (1.30), so it is not possible for (1.31) to be satisfied. Thus,
no point in the interior of V exists at which the velocity magnitude attains
a local maximum. Of course, if the point P were on the boundary S of V,
then no region R exists that surrounds P but also lies entirely in V, and so
this contradiction does not apply.
Hence a maximum of the velocity magnitude v can occur only at a point
on the boundary. The occurence of a minimum of v at an interior point is
not excluded; indeed, stagnation points at which v has the smallest possible
value do occur in the interior of the fluid.
A related result can be obtained for the pressure. It follows from the
Bernoulli’s theorem that ∇2 p = −ρ∇2 (v 2 /2), and consequently using (1.30)
we obtain
∂vi ∂vi
Z Z
n · ∇p dS = −ρ dS < 0. (1.32)
∂R R ∂xj ∂xj
Now if p had a minimum value at some interior point in the fluid, n · ∇p
would be positive at allR points of a small closed surface surrounding that
point and the value of n · ∇p dS taken over the same surface would be
positive; which is impossible, according to (1.32). Hence any point at which
p has a minimum value must lie on the boundary, although a maximum may
occur at an interior point. The position of a minimum of p does not in
general coincide with the position of a maximum of v, although there is such
coincidence when the flow is steady and the variation of f · x is negligible,
[2].
These results have qualitative application to situations in which certain
physical phenomena occur at that place in the fluid where the pressure is
1.4. IRROTATIONAL AND INCOMPRESSIBLE FLOWS 13

a minimum or the speed is a maximum. For instance, cavitation occurs in


water when the absolute pressure falls below a critical value, and we infer
that, for given boundaries to the (irrotational) flow, cavitation will occur first
at some point on the boundary as the pressure everywhere is decreased.
14 CHAPTER 1. III FLOWS
Chapter 2

Hydrofoil geometry

The hydrofoil shape can be generated in several ways:

• Use analytically given foils, like NACA family, and compute the foil
contour on an arbitrary number of points, [5];

• Use airfoil/hydrofoil data bases, and import a generally small number


of points on the contour. Then, the hydrofoil contour must be recon-
structed, either by spline interpolation or by a least squares modal
representation, [3].

Although interpolation is in theory the most accurate reconstruction pro-


cedure for a given set of points, for streamlined bodies one may experience
difficulties due to the strong curvature variation at leading edge. As a result,
unwanted bumps may appear on the hydrofoil, with large impact on the flow
behavior. A better way to represent a hydrofoil shape is provided by a least

15
16 CHAPTER 2. HYDROFOIL GEOMETRY

squares approach, with a special set of ortho-normal functions, [3]


¡√ ¢√
m1 (x) = x − x 30
13 √ √
µ ¶
14 13
m2 (x) = x(1 − x) − x+ x 10
3 14 14
35 √ √
µ ¶
27 2 35 73
m3 (x) = x (1 − x) + x − x − x(1 − x) 1190
17 81 81 81
123 √ √
µ ¶
132 3 123 377 953 2
m4 (x) = x (1 − x) − x+ x+ x(1 − x) − x (1 − x) 442
13 748 748 748 748
671 √
µ
910 671 8 5723 2
m5 (x) = x4 (1 − x) + x− x − x(1 − x) + x (1 − x)
111 11830 11830 35 5915


10123 3
− x (1 − x) 10582
5915
221 √
µ
221 551 3467 2
m6 (x) = 108 x5 (1 − x) − x+ x+ x(1 − x) − x (1 − x)
11988 11988 5994 5994


10141 3 13025 4
+ x (1 − x) − x (1 − x) 962
5994 5994

Although the original set of points may not lie precisely on the reconstructed
curve, the result is nevertheless smooth and remarkably accurate most of the
time.
Chapter 3

Pressure computation

3.1 Pressure coefficient


The pressure coefficient is generally defined as
p − pref
cp ≡ 1 2
(3.1)
2
ρvref

3.1.1 Pressure coefficient with respect to inlet condi-


tions

p − pin
cp ≡ 1 2
(3.2)
2
ρvin
Using Bernoulli theorem,
cp = 1 − v̂ 2 sin2 αin (3.3)

3.1.2 Pressure coefficient with respect to outlet condi-


tions

p − pout
cp ≡ 1 2
(3.4)
2
ρvout
Using Bernoulli theorem,
cp = 1 − v̂ 2 sin2 αout (3.5)

17
18 CHAPTER 3. PRESSURE COMPUTATION

3.1.3 Pressure coefficient with respect to average con-


ditions

p − p∞
cp ≡ 1 2
(3.6)
2
ρv∞
where
sin2 α∞ sin2 α∞
· ¸ · ¸
1 2 1 2
p∞ = pin − ρv∞ 1 − = pout − ρv∞ 1 − (3.7)
2 sin2 αin 2 sin2 αout

Using Bernoulli theorem,

cp = 1 − v̂ 2 sin2 α∞ (3.8)

3.1.4 Pressure coefficient with respect to average con-


ditions

p − p∞
cp ≡ 1 2
(3.9)
2
ρv∞
where
sin2 α∞ sin2 α∞
· ¸ · ¸
1 2 1 2
p∞ = pin − ρv∞ 1 − = pout − ρv∞ 1 − (3.10)
2 sin2 αin 2 sin2 αout

Using Bernoulli theorem,

cp = 1 − v̂ 2 sin2 α∞ (3.11)
Chapter 4

Examples

4.1 Low Pressure Turbine (LPT) cascade


Figure 4.1 shows the aerodynamic tunnel at Notre Dame University, Indiana,
U.S., used to investigate a low pressure turbine (LPT) cascade from a Pratt
& Whitney turboengine. The LPT cascade is shown in Figura 4.2. The mid
blade has 30 presure taps on suction side, and 10 on pressure side. Measure-
ments were performed for Reynolds number Re = 104 . . . 105 and turbulence
intensity u0 /U∞ = 0.08% . . . 3%. Because the air speed was around 20 m/s,
the flow can be considered incompressible, and no significant thermal effects
are present.
The turbine blade profile was given by points, and is interpolated using
periodic splines. The cascade spacing is 0.88558 the axial chord projection.
The inlet angle is β IN = 55◦ . For an inviscid, incompressible and irrota-
tional flow computation the Kutta-Jukovski condition was imposed at the
maximum curvature point on trailing edge (best choice for rouded trailing
edges).
The streamlines for irrotational flow of a perfect (inviscid and incom-
pressible) fluid are shown in Figure 4.3. The computed outflow angle is
β OU T = 149.7◦ , and the corresponding flow deflection angle is
β OU T − β IN ≈ 95◦ .
The pressure distribution is given using the pressure coefficient,
¶2
p − pIN
µ
v
cp ≡ 1 2 = 1− IN
sin2 β IN ,
2
IN
ρ (v ) v x

19
20 CHAPTER 4. EXAMPLES

Figure 4.1: Aerodynamic tunnel at the University of Notre Dame, IN, U.S.,
Hessert Center for Aerospace Research.
4.1. LOW PRESSURE TURBINE (LPT) CASCADE 21

Figure 4.2: Low pressure turbine cascade model.


22 CHAPTER 4. EXAMPLES

OUT
β
frontul retelei

axa masinii
IN
β

Figure 4.3: Streamlines for the inviscid, incompressible and irrotational flow
in LPT cascade.
4.1. LOW PRESSURE TURBINE (LPT) CASCADE 23

shown as solid curve in Fig. 4.4. One can see that near trailing edge the
Kutta-Jukovski condition produces a sudden (artificial) acceleration, followed
by a deceleration until stagnation. Obviously, the real flow detaches from
the blade and generates a wake, but this phenomenon cannot be described
within the inviscid flow framework. Besides this drawback, our numerical
results agree quite well with the experiment.
For Reynolds number Re = 105 , Fig. 4.4, the computed pressure distribu-
tion follows the experimental data on approximately 75% of the chordlength.
Flow deceleration on the last 25% close to the trailing edge leads to flow
detachment on suction side. However, at large Reynolds (Re = 105 ) the
flow is re-attaching on the sucction side. On the pressure side the flow is
only accelerated, and as a result we have an excellent agreement with the
experiment, as expected.
For Reynolds number one order of magnitude lower, Fig. 4.4, the viscous
effects lead to early flow detachnment on the suction side at 50% chordlength.
The flow does not reattaches further downstream, resulting in a thick wake
with corresponding larger losses. On the pressure side the flow remains at-
tached.
24 CHAPTER 4. EXAMPLES

1
coeficientul de presiune

0
experimental (intrados), Re=1E5
experimental (extrados), Re=1E5
−1 numeric (solutia MEF)
re
at
de as
sp ar
rin e
−2 de
re

−3

−4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
coordonata axiala adimensionalizata
2

1
coeficientul de presiune

0
experimental (intrados), Re=1E4
experimental (extrados), Re=1E4
−1
numeric (solutia MEF)

−2

−3
desprinderea curentului
real pe extrados
−4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
coordonata axiala adimensionalizata

Figure 4.4: Pressure distribution on the LPT blade.


Bibliography

[1] R. Aris. Vectors, Tensors and the Basic Equations of Fluid Mechanics.
Prentice-Hall, 1962 (reprinted Dover Publications, 1989).

[2] G. K. Batchelor. An Introduction to Fluid Dynamics. Cambridge at the


University Press, 1967.

[3] I.-C. Chang, F. J. Torres and C. Tung. Geometric Analysis of Wing Sec-
tions. NASA Technical Memorandum 110346, 1995.

[4] J. P. Gostelow. Cascade Aerodynamics. Pergamon Press, 1984.

[5] C. L. Ladson, C. W. Brooks, A. S. Hill and D. W. Sproles. Computer


Program to Obtain Ordinates for NACA Airfoils. NASA Technical Mem-
orandum 4741, 1996.

[6] B. Lakshminarayana. Fluid Dynamics and Heat Transfer of Turboma-


chinery. John Wiley & Sons, 1996.

[7] R. I. Lewis. Turbomachinery Performance Analysis. Arnold, 1996.

[8] J. S. Marshall. Inviscid Incompressible Flow. John Wiley & Sons, 2001.

[9] R. Resiga. Numerical Fluid Mechanics. Orizonturi Universitare,


Timisoara, 2003 (in Romanian).

[10] M. H. Vavra. Aero-Thermodynamics and Flow in Turbomachines. John


Wiley & Sons, 1960.

25

You might also like