Sunteți pe pagina 1din 56

MEKELLE UNIVERSITY

CFD Modeling of a Flow


Nozzle
Converging Diverging Subsonic-Supersonic

Mechanical Engineering
2000 E.C.

Alamin Oumer FST/


Adem Jaffer MU/R058/94
Advisors
Tesfalidet
Haileyesus
Acknowledgment
Abstract
Objective

1. To give physical understanding of flow field.

2. To evaluate the flow pattern of a given fluid inside a nozzle.

3. To get the changes that are taking place in different parameters,


parameters are Mach number, pressure, temperature, velocity and
density.

4. To get all important flow properties in real time, as the locator moves
from point to point in the flow field described by the characteristic
mesh.

5. Generation of Mach number (Mn), pressure (p/pc), temperature (T/Tc)


and density (ρ/ρc).

6. To find out reliability and efficiency of the CFD technique.


Contents

Introduction

1. Chapter One: CFD


1.1introduction to CFD
1.2fundamental concepts
1.2.1 continuum
1.2.2 flow field
1.2.3 Navier-Stokes equations
1.3CFD Methodology
1.3.1 Descritization method
1.3.2 Finite difference method
1.3.3 McCormack’s technique
1.3.4 Convergence and stability criteria

2. Chapter Two: Compressible Flow and


Qualitative Description of CV Flow
2.1 Compressible flow
2.1.1 Thermodynamics
2.1.2 Speed of Sound & Mach number
2.1.3 Stagnation properties
2.2 Introduction to CV nozzle
2.2.1 Qualitative description of flow and chocking
2.2.2 Flow pattern
2.2.3 Pressure variation
3. Chapter Three: CFD Analysis of CV Flow
3.1 Problem identification
3.1.1 Steady state
3.1.2 Time marching
3.2 Algorithm development
3.3 Validation and verification

4. Chapter Four: Conclusion and recommendations


4.1 CFD4.2 CV Nozzle

5. Literature Review

INTRODUCTION

In Computational fluid mechanics (CFD) a computer is used to solve a


large range of practical process in engineering and environment. Physical,
chemical and biological applications abound. In traditional manufacture and
design, a large proportion of time and money is spent building test rigs and
conducting experiments to validate engineering analyses. Computer
modeling has proven to be a cost-effective method of reducing time-to-
market, increasing value and reliability, and reducing dependence upon
empirical testing. Whole industries, such as automotive and aircraft
manufacturers, have embraced CFD and other modeling techniques as part
of the design cycle, and dramatic improvement in time-to-market, reliability
and safety are documented.
At the same time CFD is still a subject for research: our ability to
characterize the mechanics of turbulence, multi-phase and chemically
reacting flows are sometimes less than satisfactory. Moreover, the
spectacular growth in computer speed and power density still fails to match
the driving need to solve practical problems on ever-finer meshes. High
performance computing was and still is a major issue for the practitioner of
CFD. In the future we will see increasing demand for solutions for whole
assemblies and processes rather than single components.
Computational fluid dynamics (CFD) has grown from a mathematical
curiosity to become an essential tool in almost every branch of fluid
dynamics; from aerospace propulsion to weather prediction. CFD is
commonly accepted as referring to the broad topic encompassing the
numerical solution, by computational methods, of the governing equations
which describe fluid flow, the set of Navier-Stokes equations, continuity
and any additional conservation equations, for example energy.
As a developing science, Computational Fluid Dynamic has received
extensive attention throughout the international community science the
advent of the digital computer. The attraction of the subject is twofold.
Firstly the,
Chapter One

1. Computational Fluid Dynamics

CFD is predicting what will happen, quantitatively, when fluids flow, often
with the complications of:

 simultaneous flow of heat,


 mass transfer
 phase change
 chemical reaction
 mechanical movement (e.g. of pistons, fans, rudders),
 Stresses in and displacement of immersed or surrounding
solids.

1.1 Introduction to CFD

It is a computational technology that enables you to study the dynamics of


things that flow, using CFD; we build a computational model that represents
a system or device which in our case is converging-Diverging nozzle. Then
we apply the fluid flow physics to this virtual prototype. In a CFD a
computer is used to solve a large range of practical process in engineering
and the environment. In traditional manufacturing and design, a large
portion of time and money is spent building test rigs and conducting
experiments to validate engineering analysis. Computer modeling has
proven to be a cost effective method of reducing time to market, increasing
value and reliability, and reducing dependence up on empirical testing.
Whole industries, such as automotive and aircraft manufacturers, have
embraced CFD and other modeling techniques as part of the design cycle,
and dramatic improvements in time-to-market, reliability and safety are
documented.

CFD is still a subject for research: our ability to characterize the mechanics
of turbulence, multi phase and chemically reacting flows are sometimes less
than satisfactory. Moreover the spectacular growth in computer speed and
power density still fails to match the driving need to solve practical
problems on ever-finer meshes. High performance computing was and still is
a major issue for the practitioner of CFD. In the future we will see increasing
demand for solution for whole assemblies and process rather than single
components.

Computational fluid dynamics has grown from a mathematical curiosity to


become an essential tool in almost every branch of fluid dynamics, from
aerospace propulsion to whether prediction. CFD is commonly accepted as
referring to the broad topic encompassing the numerical solution, by
computational methods, of the governing equations which describe fluid
flow, the set of Navier-stokes equations, continuity equations and any
additional conservation equations, for example energy.

As a developing science, computational fluid dynamics has received


extensive attention throughout the international community since the advent
of the digital computer. The attraction of the subject is a twofold, firstly the
desire to be able to model physical experiment, for example whether systems
or hypersonic aerospace vehicles. Secondly the desire to be able to
investigate physical fluid systems more cost effective and more rapidly than
with experimental procedures.

There has been considerable growth in the development and application of


computational fluid dynamics to all aspects of fluid dynamics. In design and
development, CFD programs are now considered to be standard numerical
tools, widely utilized with in industry. As a consequence there is a
considerable demand for specialists in the subject to apply and develop CFD
methods through out engineering companies and research organizations.

In the past two decades a number of suites of general purpose CFD codes
have been developed and maintained worldwide. The best known of these
are based on the finite volume and finite element methods. These methods
are useful in solving the real life problem. So CFD is a tool which can apply
for the modeling of chemical engineering process equipment like reactor,
agitator, heat exchanger…

Using CFD we can solve the Navier-Stokes equation to get detailed


information about point to point variation of temperature, pressure, density,
velocity etc... Which is very much needed in accurate design of process
equipments?

1.2. Fundamental concepts

All fluids are composed of molecules in constant motion, however, in most


engineering applications we are interested I the average or macroscopic
effects of many molecules. It is these macroscopic affects that we ordinarily
perceive and measure.

1.2.1. Fluid continuum

Thus we treat a fluid as an indefinitely divisible substance, a continuum, and


do not concern ourselves with the behavior of individual molecules.

As consequence of the continuum assumption, each fluid property is


assumed to have a definite value at every point in space. Thus fluid
properties such as density, temperature, velocity and so on are considered to
be continues functions of position and time.

1.2.2. Flow field

Density could be defined at any point in a fluid. If density determinations


were made simultaneously at infinite number of points in the fluid, we
would obtain an expression for the density distribution as a function of space
coordinates
ρ=ρ(x, y, z,) at a given instant. Clearly density may vary with time, thus the
complete representation of density (the field representation) is given by

ρ=ρ(x, y, z, t). Similarly other properties could be dealt.

1.2.3 The Navier-Stokes equations

The Navier-Stokes equations are developed from Stokes-Fourier postulates


made by Navier, Stokes and Fourier for the derivation of the viscous-
conducting fluid model have a strong bearing on the modeling of a wide
range of fluids flow, including turbulent flows. The Stokes-Fourier
postulates are summarized below,

1. a fluid is a continuum in local equilibrium, molecular motions are


averaged and thus detailed information of the dynamics of molecular
collisions is lost. A model is required to recover the lost information.
(modeling requirement)
2. The diffusion of momentum and thermal energy by viscous fluid motion
is proportional to the rate of deformation and gradient of temperature,
respectively. (diffusion gradient)

3. A fluid is assumed to be isentropic ( Isentropic molecular collision


model)

4. A fluid is assumed to be homogenous. That is the viscous stress and heat


fluxes are not explicit functions of space and time.

5. when a fluid is at rest, the viscous stress is hydrostatic pressure


(consistency and reliability requirement)

6. When the flow is pure dilation, the averaged viscous stress is equal to the
pressure.

7. The model moduli (density, viscosity, specific heat, thermal conductivity


etc...) require experimental calibration and determination. (Uniqueness of
moduli).
The Navier-Stokes equations that are the governing equations of fluid flow
in CFD are developed using the Stokes-Fourier postulate, they are based on
the principles of conservation of mass, momentum and energy.
 u
Continuity equation  0
t x

Du p   u u j  2 u 
Momentum equation        i     ij j   g i
Dt x   x j xi  3
 xl 

Energy equation
DT   kT  D  u u j  2 ul  u i
c p      i     ij 
t xi t  xi xi  3 xl  x j

Equation of state   f ( P, T )

 U—component of velocity vector


 T—temperature

 P—pressure

 R—density

The independent variables are x, t spatial variable and time.

Fluid properties

µ--molecular dynamic viscosity

Cp—specific heat at constant pressure

K—thermal conductivity

gi—gravitational force in the i-th direction

1.3. CFD Methodology

CFD method of analysis employs the concept of control volume; control


volume is an arbitrary volume in space through which fluid flows. The
geometric boundary of the control volume is called control surface. The
control surface may be imaginary or real, it may be at rest or in motion.

CFD deals with detailed flow properties therefore it uses differential


approach than integral approach. The basic laws that we apply in out study
of fluid mechanics can be formulated in terms of infinitesimal or finite
systems and control volumes. As we might expect the differential and
integral equations look different. Both approaches are important in the study
of fluid mechanics.

In the differential case which we are interested in equations are differential


equations. Solutions of the differential equations of motion provide a means
of determining the detailed (point by point) behavior of the flow.

In general most applications of CFD take the same basic approach. Some of
the differences include problem complexity, available computer resource,
available expertise in CFD, and whether a commercially available CFD
package is used or a problem specific CFD algorithm developed. The
features to most CFD application can be summarized in the flow chart (fig.
1.1).

CFD methodology
1.3.1 Descritization

the process of discretization involves developing algebraic equations (based


on discrete points in the flow domain) to be used in place of partial
differential equations, of the various discretization techniques available for
the numerical solution of the governing differential equations we’ll be
employing later on the report the finite difference method.

1.3.2. The finite difference method

In this method the flow field is dissected in to a set of grid points and the
continuous functions (velocity, pressure…) are approximated by discrete
values of these functions calculated at grid points. Derivatives of the
function are approximated by using the differences between the function
values at local grid points divided by the grid spacing. The standard method
for converting the partial differential equations to algebraic equation is
through the use of Taylor series expansions.

Thus we can transform a partial derivative in to algebraic expansions


involving values of the variable at neighboring grid points. Procedures can
be used to develop approximations termed forward difference, backward
difference, and central difference representation methods. Applying the
method of finite difference method to the governing equations transfers the
differential equations in to a set of algebraic equations involving the physical
variables at the grid points. This set of equations is then solved by
appropriate numerical technique. The larger the number of grid points used,
the larger the number of equations that must be solved. For our modeling of
converging diverging nozzle we will use a method called McCormack’s
corrector and predictor technique, which is suitable for compressible flow
equation.

1.3.3. McCormack’s numerical technique


1.3.4. Convergence and stability criteria
Chapter Two

2. Compressible Flow and Qualitative Description of CV Flow

The flow in a converging diverging nozzle from subsonic to supersonic


speed is clearly compressible flow, therefore some revision about
compressible flow is discussed in the first section of this chapter, and in the
later section compressible flow and type of flows that could occur in a
converging diverging nozzle with a brief introduction are discussed
qualitatively.

2.1. Compressible flow

In this section we will see compressible flow; compressible flow implies


appreciable variations in density throughout a flow field. Compressibility
becomes important at high flow speeds or for large temperature changes.
Large changes in velocity involve large pressure changes, for an ideal gas
flow these pressure are accompanied by significant variations in both density
and temperature. Since two additional variables are encountered in treating
compressible flow, two additional equation of state must be applied to solve
compressible flow problems.

2.1.1. Review of thermodynamics

The pressure, density and temperature of a substance may be related to by an


equation of state. Although many substances are complex in behavior,
experiences shows that most gases of engineering interest, at moderate
pressure and temperature are well represented by ideal gas equation of state.

P=ρRT 2.1

Where R is constant for each gas,

Although no real substance behaves exactly as an ideal gas equation (2.1) is


in error by less than 1%, for air at room temperature and pressure as high as
30atm, the equation is less than 1%in error for temperature as low as 140K.
The ideal gas has other features that are simple and useful. In general, the
internal energy of a substance may be expressed as u=(v, T) where V=1/r,
the specific volume then simplifies to

u  u 
du  ( )V dT    dV
T  v  T
The specific heat at constant volume is defined cv

 u 
cv    so that
 T V
 u 
du  cv dT    dv
 v  T
For any substance that follows the ideal gas equation of state (eqn. 2.1), then
(ju/jT)T=0 and hence u=u(T), consequently

du=cvdT 2.2

Equation 2.2 implies that internal energy and temperature changes may be
related if cv, is known. Furthermore, since u=u(T).

The enthalpy of a substance is defined as


H=u+p/r. for ideal gas we have already stated in eqn. 2.1, and so h=u+RT.
Since u=u(t) for an ideal gas h also must be a function of temperature alone.

Now to obtain a relation between h and T, we express ‘h’ in its most general
form as

h=h(P,T) then

 h   h 
dh    dT    dp
 p  p  p  T
We have shown that for an ideal gas ‘h’ is a function of ‘T’ only.
 h 
Consequently    0 and,
 p  T

dh=CpdT (2.3)

Again since h is a function of T alone the above equation requires that c p be


a function of T only for an ideal gas.
The specific heat for an ideal gas has been shown to be functions of
temperature only their difference is constant for each gas. From
H=u+RT
We can write
dh=du+RdT
Now combing with equation 2.3, and using equation 2.2 we can write,
dh=cpdT=du+RdT and du=cvdT from 2.2
dh=cpdT=cvdT+RdT then

R=cp-cv (2.4)

The ration of specific heat is defined as

cp
k (2.5)
cv

Using the definition for k, we can solve equation(2.4) for either c p or cv in


terms of k and R
Thus R=cp-cv
Cp=R+cv
Substituting this value in to the above equation,
R=cp-cp/k
Then
c p  kR / k  1 (2.6a)
Using the same procedure we find the value of cv as

cv  R / k  1 (2.6b)

2.1.2. Speed of sound and Mach number

the speed of sound in a fluid is the speed of propagation of an infinitesimal


pressure wave, and is a finite number because of the compressibility of the
fluid. The time separation between when the flash of lightening is seen and
the associated thunder is heard is (about 3 seconds per kilometer) is evidence
of the finite speed of sound.
For and ideal gas one may show that the speed of sound, which is normally
given either the symbol c or ‘a’ is
1
 kP  2
c   kRT 
1
2    (2.7)
  

Where ‘k’ is the ratio of specific heats of the gas (k=1.4 for air), clearly we
can notice here that the speed of sound in a compressible flow varies from
point to point, because a ‘T’ varies from place to place.

Mach number

We define Mach number as the ratio of the local fluid velocity to the local
speed of sound,

V
M  (2.8)
C

The Mach number is an important quantity in flows for which the local fluid
velocity is greater than about 30% of the speed of sound (i.e. compressible
flow) of the medium.

When
M>0.3, high velocity gas flow for which the fluid compressibility is
important

Flows for which


M<1, are called subsonic
M=1 are called sonic
M>1, are called supersonic

Therefore in converging diverging nozzles, the nozzle is used to accelerate a


subsonic flow to supersonic flow by only varying the cross-sectional area.

2.1.3 Stagnation properties

Reference state, local isentropic stagnation properties; if we wish to describe


the state of a fluid at any point in a flow field, we must specify two
independent intensive thermodynamic properties (usually pressure and
temperature), plus the velocity at the point.

In our discussion of compressible flow we shall find it convenient to use the


stagnation state. The stagnation state is characterized by zero velocity,
stagnation properties at any point in a flow field are those properties that
would exit at that point if the velocity were reduced to zero. Consider the
point in a flow field where temperature is T, the pressure P, and the velocity
is V. the stagnation state at that point in a flow field would be characterized
by stagnation pressure P0, and stagnation temperature T0, and zero velocity.

To get local isentropic stagnation properties for the flow of an ideal gas, we
need to develop an expression describing the relationship among fluid
properties are specified. We develop the relationships among properties in
differential form. Then we integrate to obtain expressions for stagnation
conditions in terms of initial conditions corresponding to the actual flow at
the point.

The hypothetical decelerating process is shown schematically I fig 2.1. We


are interested in finding the stagnation properties for flow at point (1). To
find a relationship among the fluid properties during the deceleration
process, we apply the continuity and momentum equation to the stationary
differential stream tube control volume as shown below.

Fig 2.1 compressible flow in an envelop

a. Continuity equation

Uniform flow at each section

v x A         v x  v x  A  A (2.9)

b. Momentum equation
For no body force and frictionless flow becomes

The surface force acting on the infinitesimal control volume is and


substituting to continuity equation

We obtain
dp 2
 d (v x )  0 (2.10)

Here equation (2.10) is a relation among property during the deceleration
process. In developing this relation, we have specified a frictionless
deceleration process. Before we can integrate between the initial and final
(stagnation) states, we must specify the relation that exist between pressure
P, and density r, along the process since the deceleration process is
isentropic, then P and r for and ideal gas are related by the expression

p
 cons
k (2.11)

Our task now is to integrate equation 2.10, subject to this relation. Along the
stagnation stream line there is only a single component of velocity; v x is the
magnitude of the velocity. Hence we can drop the subscript I equation 2.10,
From equation 2.11 and 2.10 we get

 v2  dp 1 1
 d       k c k dp
 2  d

We can integrate this equation between initial state and the corresponding
stagnation state

0
v2 1
p0
1
v d  2   c k p
p
k
dp

To obtain
 k 1
 
v2  p k 1 k   p0 
1 k
c k k  1
2 k  1    p  
 
Since c1/k=r1/k/r

Since in this case we seek expression for stagnation pressure, we can rewrite
this equation as and by using equation 2.1,
For an ideal gas the stagnation pressure is
k k 1
p0  k 1 v 2 
 1  
p  2 kRT 

Also for an ideal gas the sonic speed was given earlier thus,
k
p0  k 1 2  k 1
 1  M  (2.12)
p  2 
We can readily obtain expression for other isentropic stagnation properties
by applying the isentropic relation given in 2.11,

For an ideal gas

T0 k 1 2
 1 M
T 2 (2.13)
1
0  k  1 2  k 1
 1  M 
  2  (2.14)

Critical condition

Stagnation conditions are extremely useful as reference conditions for


thermodynamic properties; this is not true for velocity since by definition
V=0 at stagnation. A useful reference value for velocity is the critical speed
at a Mach number unity.

Using the asterisk to denote conditions at M=1, then by definition

V*=C* (2.15)

At critical condition we use the equation 2.12-2.14, and substitute M=1,


k
p0  k  1  k 1

p *  2  (2.16)

T0 K  1 (2.17)

T* 2
1
0  K  1 K 1

  2  (2.18)
*

We can write the critical speed using the critical temperature or isentropic
stagnation temperature;

T*, from the above equation and from 2.15

2k
V*=c*= ( RT0 ) (2.19)
k 1

2.2 Introductions to converging diverging nozzle

A converging diverging nozzle is (CD nozzle) is a tube that is pinched in the


middle, making an hourglass shape. It is used as a means of accelerating the
flow of a gas passing through it to a supersonic speed.

Fig2.2 converging diverging nozzle,

The converging diverging nozzle was developed by a Swedish inventor


Gustaf de-Laval in the 19th century. The CD nozzle is assumed to have an
inlet, throat, and exit with rigid walls. It is a major component of rockets.

2.2.1. Qualitative flow behavior in CD nozzle.

Consider two cylinders connected with CD nozzle,


Fig 2.3

Cylinder a contains air at high pressure, The CD nozzle exhausts this air into
cylinder B,

The usual configuration of CD nozzle is shown in fig 2.4, gas flows through
the nozzle from a region of high pressure (P0) to one of low pressure (Pe).
The reservoir is usually big enough so that any flow velocities here are
negligible.

Fig 2.4 configurations of a CD nozzle,


Gas flows from the reservoir into the converging portion of the nozzle, past
the throat, through the diverging portion and then exhausts into the ambient
as a jet. The pressure of the ambient (pe) is referred as back pressure.

Now imagine controlling the back pressure, and measuring the resultant flow
rate through the CD nozzle. We expect that the lower we make the P e the
more mass flow you’ll get through the nozzle.

Chocking

The mass flow rate will increase only up to a point. If you lower the back
pressure enough you come to a place where the flow rate suddenly stops
increasing and doesn’t matter how much you lower the back pressure (P b).
Even if you make it vacuum, you can’t get any more mass flow out of the
nozzle. We say the nozzle has become ‘chocked’. Fig. 2.5.

Fig 2.5 choking

We can delay this behavior by making the nozzle throat bigger (hidden line
in fig 2.5) but eventually the same thing will happen. The nozzle will
become chocked even if we eliminated the throat altogether and just had a
converging nozzle.

The reason for such behavior has to do with the way the flows behave at
Mach number =1. When the flow speed reaches the speed of sound and
Mach number in an internal flow can only reach 1 at a minimum cross-
sectional area.
When the nozzle isn’t chocked, the flow through it is entirely subsonic and ,
if you lower back the back pressure further the flow speed eventually
reaches the speed of sound (M=1) away from the throat area where area is
minimum, and so the flow gets stuck. The flow pattern down stream of the
(in diverging section) can still change if you lower back pressure further, but
the mass flow rate is now fixed because the flow in the throat (and for that
matter in the entire converging section Is now fixed too.
The changes in flow pattern after nozzle is chocked become important if you
are concerned with exit velocity.

2.2.2. Flow pattern

Flow inside a converging diverging nozzle depends on the nozzle cross-


sectional area (which we see in the coming chapter), and the back pressure.
Here we will see the flow pattern due to back pressure variation; it could
behave like one of the following

1. Subsonic flow: flow through nozzle is completely subsonic for such


cases, and nozzle is not chocked. Fig 2.6

Fig 2.6 subsonic flow

In this case flow accelerates out of the high pressure reservoir through the
converging section reaching maximum subsonic speed at throat. The flow
then decelerates through the diverging section and exhausts to the ambient as
a subsonic jet. Lowering the back pressure in this state increases the flow
speed everywhere in the nozzle.

2. flow just chocked

If the back pressure is lowered and we eventually get to the situation


shown below in fig. 2.7
Fig 2.7 flow just chocked

Flow pattern exactly same as subsonic flow except that at the throat speed
has reached Mach 1, flow through the nozzle is now chocked since further
reduction can’t move the point of M=1 away from the throat. However the
flow pattern in the diverging section does change as you lower the back
pressure further.

3. shock in nozzle

If pb is lowered below that needed to just chock, the flow a region of


supersonic flow forms just downstream of the nozzle. Fig 2.8

Fig 2.8 shock in nozzle

Unlike subsonic flow, the supersonic flow accelerates as area gets bigger.
This region of supersonic acceleration is terminated by normal shock wave.
The shock wave produces a near instantaneous deceleration of the flow to
subsonic speed. This subsonic speed then decelerates through the remainder
of the diverging section and exhausts as a subsonic jet. In this regime if you
lower or raise the back pressure you increase or decrease the length of
supersonic flow in the diverging section before the shock wave.

4. Shock at exit

If you lower the back pressure you extend the supersonic region all the way
down the nozzle until the shock is sitting at the nozzle exit. Fig 2.9
Fig 2.9 shock at exit

Because you have a very long region of acceleration (the entire nozzle
length) in this case the flow speed just before the shock will be very large in
this case, however after the shock the flow in the jet will still be subsonic.

5. Over expanded

Lowering the back pressure further causes the shock to bend out into the jet,
fig 2.10. And a complex pattern of shocks and reflections is set up in the jet
which will now involve a mixture of subsonic and supersonic flow, or (if P b
is low enough) just a supersonic flow.

Fig 2.10 over expanded

Because the shock is no longer perpendicular to the flow near the nozzle
walls it deflects inward as it leaves the exit producing an initially contracting
jet. We refer to this as over expanded flow because in this case the pressure
at the nozzle exit is lower than that in the ambient (the back pressure). I.e.
the flow has been expanded by the nozzle too much.

6. Design condition
A further lowering of the back pressure changes and weakens the wave
pattern in the jet. Eventually we will have lowered the back pressure enough
so that it is now equal to the pressure at the nozzle exit. Fig 2.11

Fig 2.11 design condition

In this case the waves in the jet disappear altogether and the jet will be
uniformly supersonic. This situation, since it is often desirable is referred to
as the design condition.

7. Under expanded

Finally if we lower the back pressure even further we will create a new
imbalance between the exit and back pressure (exit pressure greater than
back pressure). Fig 2.12

Fig 2.12 under expanded

In this situation (called under expanded) what we call expansion waves that
produce gradual turning and acceleration to jet, from the nozzle exit, initially
turning the flow jet edges outward in a plume and setting up a different type
of complex wave pattern.

2.2.3. Pressure distribution


The plot for the above flow pattern behavior due to pressure distribution is
given in fig 2.13

Fig 2.13 pressure distribution

To understand how the pressure behaves you have to remember only a few
basic rules

 When the flow accelerates (sub or supersonically) the pressure drops


 The pressure rises instantaneously across a shock
 The pressure throughout the jet is always the same as the ambient (i.e.
the back pressure) unless the jet is supersonic and there are shocks or
expansion waves in the jet to produce pressure differences.
 The pressure falls across an expansion wave.

The labels on figure 2.13 indicate the back pressure and pressure distribution
for each of the flow regimes illustrated in figure 1.13. Notice how, once the
flow is choked, the pressure distribution in the converging section doesn't
change with the back pressure at all.
Chapter Three

3. Analysis of Converging-Diverging Nozzle Isentropic Flow, by


Using CFD

At the beginning of this chapter we will solve steady one dimensional flow
through converging-diverging nozzle of specified cross-sectional area
analytically, following this we’ll proceed to unsteady flow analysis of the
flow by using CFD technique. While using the CFD technique we will use
finite difference approximation method discussed earlier in chapter one and
employ Mac Cormack’s technique for predicting and correcting flow field
properties after some time and a number of iterations and then validate the
numerical solution with experimental data to see the accuracy of CFD
technique.

3.1. Steady one dimensional compressible flow in CD nozzle


It is already stated that gases at very high velocities are compressible, and
since the flow we are dealing with is compressible flow, and this
compressible flow in converging diverging nozzle affected by the cross
sectional area. And area is independent variable of time and only depends
only on the x-axis which is assumed to coincide with the nozzles axis of
symmetry. The fluid flow is assumed to be isentropic flow (i.e. negligible
friction & heat transfer).
Consider steady one dimensional, isentropic flow of any compressible fluid
through a converging diverging nozzle of arbitrary cross section, and area
depends only on x-axis, (Fig 3.1) to develop governing equations for this
flow; consider a control volume inside the converging diverging nozzle.

Fig 3.1.

3.1.1. Basic equations


a. Continuity equation

(3.1)

Assumptions
1. Steady flow
2. One dimensional flow

Using scalar magnitudes and differentiating the above equation, by dropping


the absolute values, gives the familiar form of continuity equation

(3.2)
b. Momentum equation
Integral form of momentum equation given in chapter one, for one
dimensional flow is given by

(3.3)
Assumptions
3. FBX=0

The surface force will be due to pressure forces at the surface ‘1 &2’ and to
distribute pressure force Rx, along the channel walls, substituting gives (by
using scalar magnitudes and dropping the absolute value sign)

(3.4)
c. First law of thermodynamics
integral form of first law was given in chapter two,

(3.5)
Where e or sometimes u is internal energy and is given by

Assumptions
4. Q=0 (isentropic)
5. Ws=0
6. Wshear=0 & w (other)
7. Effects of gravity are negligible

Under the assumptions the first law

But we know from continuity equation that the mass flow rate terms in
braces are equal, so they may be cancelled. We may also substitute h=u+Pv,

(3.5)

The above is called stagnation enthalpy defined in chapter two.


d. 2nd law of thermodynamics
Basic equation
(3.6)

For reversible adiabatic process, the equality holds and since the mass flow
rate terms are constant
S1=s2=s(3.7)
e. equation of state
P=ρRT
Assumption
8. Ideal gas

f. equation of process

(3.8)

These are governing equations for steady one dimensional isentropic flow of
an ideal gas.

3.1.2. Effect of variation of area on properties in isentropic flow

In chapter two we discussed the effect of the cross-sectional variation will


cause to the flow inside a converging diverging nozzle qualitatively, in this
section will see how the flow field variables are affected analytically with
the help of the basic equation given in the earlier subsection.
In considering the effect of area variation on fluid properties in isentropic
flow, we shall concern primarily with velocity and pressure. The changes in
isentropic flow of properties are undoubtedly is area variation. We wish to
determine the effect of a change in area A, on velocity and pressure. For a
change in dA in area of the converging diverging are dV & dP.
The differential momentum equation for isentropic flow in eqn. 3.4 reduces
to

(3.9)

(3.10)
And dividing by we obtain;

(3.11)

A convenient form of differential form of the continuity equation can be


obtained from equation. 3.3

(3.12)

Using equation 3.12 for dA/A gives

Substituting from equation 3.13 gives

Now recall that for isentropic process , so

(3.13)

From equation 3.13 we see that for M<1 (converging section of the nozzle)
an area change causes a pressure change of the same sign (positive dA
means positive dp form M<1):
For M>1, diverging section of nozzle) an area change causes a pressure
change of opposite sign.
Substituting from equation 3.11 in to eqn 3.13
(3.14)

From the eqn 3.14 we see that M<1 an area change causes a velocity change
of opposite sign (positive dA means a velocity change of the same sign.

Fig 3.2 effect of area


Subsonic –supersonic flow in converging diverging nozzle, the flow expands
isentropic ally to supersonic speed at nozzle exit, where the exit pressure,
temperature, velocity and Mach number are denoted by Pe,Te, ve,Me
respectively.
What of the remaining case M=1, further inspection of eqn 3.14 shows that
at M=1 dA/dv=0. This means the converging diverging nozzle area must
through a minimum or maximum at M=1. Inspection of fig 3.2. shows that
M=1 can be reached only in a throat, or section of minimum cross-section, a
phenomenon called chocking.
The flow in the nozzle for supersonic flow is locally subsonic in the
convergent section of the nozzle, sonic at the throat (minimum cross section)
and supersonic at the diverging section. The sonic flow (M=1) at the throat
means that the local velocity at this location is equal to the local speed of
sound.

(3.15)

Hence although the area of the nozzle changes as a function of distance


along the x, and therefore in reality the flow field is tow dimensional. The
assumption that the flow properties vary only with x; this is tantamount to
assuming uniform flow properties across a given cross-section. Such flow is
defined as quasi-one-dimensional flow.
3.1.3 Reference conditions for isentropic flow of an ideal gas in
converging diverging nozzle.

Expression for locally isentropic stagnation process for an ideal gas were
developed in chapter II,
These expressions were

(3.16)

(3.17)

(3.20)
As shown I section 3.1.2. Stagnation properties are constant throughout a
steady isentropic flow field.

Critical conditions

Flow properties at which Mach number is unity (throat), then the above flow
equations we substitute M=1,

(3.19)

(3.20)

(3.21)
For an ideal gas
The critical speed may be written either critical temperature T * or isentropic
stagnation temperature T0,
Thus

T* from equation 3.20 the above equation becomes


(3.22)
If the area at which the Mach number is unity is designated by A *, then it is
possible to express the contour of a passage in terms of area ratio.

Area ratio

Fig 3.3area ratio vs Mach number


3.2. Time Marching subsonic-supersonic

Dealing with time marching problems, initial conditions are required besides
boundary conditions. Initial conditions are usually arbitrary but taking it
close to steady state solution leads to good approximation of unsteady state
solution, we will look to these initial conditions in the later sections of this
chapter. Another important consideration to time marching problems is
convergence of the solution along with stability. We already discussed about
this in chapter one, CFL considerations and we will make use of this later.
Problem: subsonic-supersonic Converging diverging nozzle flow analysis
using CFD technique.
The set up

1. We will use the governing equations of fluid flow, and changing them
to discritized model by using finite difference method.

2. After using finite difference method the differential equation will


subsequently become in algebraic forms, but this algebraic forms are
tiresome to evaluate therefore we employ a numerical technique “Mac
Cormack’s corrector and predictor” to find numerical solution of
unsteady flow.

3. And later, we will validate the solution to experimental data, and


check its accuracy.

3.2.1. Governing equations

Assumption
1. Flow properties are uniform across a given cross section.
2. Isentropic flow
3. One dimensional flow
4. Non viscous flow

In reality nozzle flow is two dimensional flow but assuming one


dimensional flow because as area varies along x, flow field properties vary
in x, and y. but considering one dimensional flow gives very good
approximation.

Fig 3.4 control volume approach


a. Continuity equation
From integral form of continuity equation we have:

(3.25)

Let’s apply this to the control volume (i.e. slice of the nozzle flow where the
infinitesimal thickness of the slice is dx).
On the left side of the control volume, consistent with quasi-one dimensional
assumption (ρ,v,P,e) are uniform over area A, similarly on the right side of
the control volume(ρ+dρ,v+dv,p+dp,e+de).
The surface integral in eqn. 3.25

Hence the continuity equation becomes;

(3.26)

Eqn 3.26 is the partial differential form of the continuity equation suitable
for unsteady, quasi-one dimensional flows.
b. Momentum equation

(3.27)
Eqn. 3.27 is the integral form of the x-component of the momentum
equation written for in viscid (neglecting the viscous stress term) with no
body forces.
The surface integral in the above equation becomes

(3.28)
Eqn 3.28 is momentum equation appropriate for quasi-one dimensional flow
in non-conservative form.
c. Energy equation

In the integral form for one dimensional flow

(3.29)
Eqn 3.29 is the integral form of the energy eqn, for an ideal adiabatic flow
(q=0) with no body forces and no viscous effects.
From eqn 3.29

(3.30)
Eqn 3.30, is the none conservative form of the energy equation expressed in
terms of internal energy, appropriate for unsteady, quasi-one dimensional
flow.
Or substituting for Cv (expression for C was already given in chapter 2,
thermodynamics)

(3.31)

3.2.2 The finite difference methods and Mac Cormack’s

In CFD it is appropriate to use none dimensional variables like and


they vary from 1 to 0. Substituting the none dimensional variables in the
governing equations of the earlier section, and assuming the nozzle flow
comes from a large reservoir we proceed to letting L’ denote length of the
nozzle, denoting the speed of sound in the reservoir as a 0, sonic throat area
A*, (A is a function of x only it is not a function of time).
For our purpose lets designate the reservoirs pressure P0, temperature T0,
respectively. The reservoirs A- and hence the velocity approaches 0.
The governing equations appropriate for finite difference method are,
Continuity equation
(3.32)
Momentum equation

(3.33)
Energy equation

(3.34)

The above equations are expressed in finite difference expressions,


(discritization step of CFD), to implement the finite difference solution we
divide the x-axis (i.e. along the nozzle symmetry) to a number of grid points,
fig 3.5

Fig 3.5 Grid point distribution along nozzle

In the figure 3.5 the first grid point, labeled point 1, is assumed to be in the
reservoir, with denoting the spacing between grid points. The last point,
namely, that at the nozzle exit is denoted by N.
We are already familiar with finite difference method form chapter one,
The continuity equation becomes

(3.38)
Momentum equation

(3.39)
Energy equation
(3.40)

3.2.2 Mac Cormack’s technique

Referring to chapter one; we use the simplification from the finite difference
method, to obtain rate of change of flow field variables.
The time marching problems have issues of stability and convergence, it was
discussed in chapter one about CFL, hence we make use of this criteria to
find the time step that would result in stable solution of the numerical
method.

3.2.2.1. Calculation of the time step

The time step is calculated from courant number,

(3.41)
Where a is

(3.42)
And “v” is velocity;
Therefore, with c being the courant number, along with C 1, simply states
that must be less than, or at best equal to the next time step.

(3.43)
The other choice is to calculate at all grid points (i=1 to i=N) and then
choose the minimum value, (CFL requirement for stability).
Hence

(3.44)
Choose to be .
3.2.2.2. Predictor step

The predictor step uses the rate of change in each property form the finite
difference result and makes use of them in such a way (chapter one)
Density

(3.45)
Velocity

(3.46)
Temperature

(3.47)

3.2.2.3. Corrector step


The corrector steps make use of the rate change of flow field variables at
time, and is a back ward finite difference method.
For density

(3.48)
For velocity

(3.49)
For temperature
(3.50)
We go to the average of rate of change of fluid flow properties at time

Average

(3.51)

(3.52)

(3.56)
Now the corrected density, velocity and temperature respectively are given
by
Density is given by

(3.57)
Velocity is given by

(3.58)
Temperature is given by

(3.59)
Pressure is calculated from equation of state

Where R is universal gas constant (R=287 for air).

3.2.2.4. Boundary conditions


Not only is point 1an inflow boundary, is a subsonic inflow boundary. Weak
Mach waves which are propagating upstream and downstream. Mach waves
are travelling the speed of “a”.

i. Subsonic inflow boundary

We must allow on variable to float. The value of v1, changes with time and
is calculated from information provided by the flow field solution over the
internal points. We use linear extrapolation from point “2” and “3”; to
calculate v1
Slope=

Therefore

(3.60)
All other flow-field variables are specified. Since point “1” is viewed as
essentially the reservoir. (It is going to be 1 since we are using ratios)

ii. Supersonic out flow boundary

We again choose to use linear extrapolation based on the flow field values at
the internal points.

(3.61)

(3.62)

(3.63)

3.3. Algorithm Development


3.4. CONCLUSION

There are three methods that we can use to solve problems in fluid
mechanics these include experimental, theoretical and computational (CFD)
method. The idea of experimental testing is to evaluate the performance of
relatively inexpensive small-scale version of the prototype device. In
performing such test it is not always possible to simulate the true operating
condition of the prototype device. Theoretical method begins with the
development of a mathematical model involving differential equation for the
dependant variable. The most difficult part in this case is solving the
mathematical equations subjected to initial and boundary conditions.
In our project we have developed a Computational model that
represents a nozzle, using CFD technique. Then we apply the fluid flow
physics to this virtual prototype and the software which in our case is
MATLAB software outputs a prediction of the fluid dynamics.
Computational Fluid Dynamic (CFD) has advantage over the other ones that
computer simulation is free of some of the constraints imposed on the
experimental method for obtaining information upon which to base a design.
It is also possible the true operating condition of the prototype. It is also cost
effective than the other methods mentioned above.
On the other hand Computational method also have certain
limitations among these are computer shortage and speed, other limitations
arise owing to our inability to understand and mathematically model certain
complex phenomena.
In general we conclude that the result obtained while working with CFD
technique are satisfactory suggesting that purely geometric features may be
responsible, to a large extent, for the basic flow development. Indeed,
essential flow characteristics such as the velocity profile, temperature
profile, pressure profile, density profile and Mach number distribution in
both case i.e. within the nozzle and at the exit of the nozzle appears to be
fairly well predicted. Based on the methodology described in this thesis,
process design and analysis can be rapidly performed to scale geometry ,
evaluate design modifications, such as, to determine the shape of minimum
length nozzle that gives the same exit Mach number for a given throat
diameter and the nozzle performance in rocket propulsion or similar
applications can be evaluated rapidly.

RECOMMENDATION

In this project we would like to recommend few points. While


modeling a nozzle using CFD method we have made some assumptions.
One of the assumptions is that the flow is considered to be quasi-one
dimensional flows. This implies that we only consider the flow to be in one
direction i.e. the x-axis but in real case due to the variation of area there is
also a flow along the y-axis, so ignoring this factor will introduce some
errors in our calculation. Therefore while modeling a nozzle it would be
preferable to consider the flow along the y-axis.
We have also assumed the flow to be isentropic which a frictionless and
adiabatic i.e. with no heat loss to the surrounding which only depends on the
variation of area. This factor will also bring errors in our calculation.
Therefore it would be advisable to consider the above factors so that we can
model a much better and efficient nozzle.
The other factor that must be remembered is that the nozzle should be
properly chocked. The engineer should properly check the inlet pressure and
the back pressure so that the design could be safe. The back pressure should
not be too small or too large such that the pressure difference will not affect
the flow pattern. The flow pattern needs to be in design condition so that we
can achieve the maximum possible speed at the exit of the nozzle as
mentioned in chapter two. In the design condition the flow will be sub-sonic
at the convergent section sonic at the throat and super-sonic till the exit of
the nozzle.
The last but not the least point that we would like to recommend is
that for the proper growth of this CFD technique computer supply should be
enough with high speed computers. Also the people working on this area
should have enough knowledge of CFD and understanding of how to
mathematically model complex phenomena so that they can model
something with more value.

Chapter Four

4. Literature review

As one might expect the history of CFD is closely tied to the development of
the digital computer s. most problems were solved using methods that were
either analytic or empirical in nature until the end of World War II. Prior to
this time, there were a few pioneers using numerical methods to solve
problems. Of course, the calculations were performed by hand, and a single
solution represented a momental amount of work. Since the time, the digital
computers have been developed, and the routine calculations required in
obtaining a numerical solution are carried out with ease.

The actual beginning of CFD or the development of methods crucial to CFD


is a matter of conjecture. Most people attribute the first definitive work of
importance to Richardson (1910), who introduced point iterative schemes
for numerically solving Laplace’s equation and the bi-harmonic equation in
an address to the royal society of London. He actually carried out calculation
for the stress distribution in a masonry dam. In addition he clearly defined
the difference between problem that must be solved by a relaxation scheme
and those that we refer to as marching problems.
Richardson developed a relaxation technique for solving Laplace’s equation.
This scheme used data available from the previous iteration to update each
value of the unknown. In 1918, Liebmann presented an improved version of
Ricahrdson’s method. Liebmann’s method used values of the dependent
variable both at the new and old iterations level in each sweep through
computational grid. This simple procedure of updating the dependent
variable immediately reduces the convergence times for solving Laplace’s
equation. Bothe the Rechardson method and Liebmann’s scheme are usually
used in elementary heat transfer courses to demonstrate how apparently
simple changes in technique greatly improve efficiency.
Sometimes the beginning of modern numerical analysis is attributed to a
famous paper by courant, Fredish, and Lewy (1928). The acronomy CFL,
frequently seen in the uniqueness and existence questions were addressed for
the numerical solutions of partial equations. Testimony to the importance of
this paper is evidenced in its re-publication in 1967 in the original source for
the CFL stability requirement for the numerical solution of hyperbolic
partial differential equations.
In 1940, south well introduced a relaxation scheme that was extensively
used in solving both structural and fluid dynamics problems where an
improved relaxation scheme was required this method was tailored for hand
calculations, in that point residuals were computed and these were scanned
for the largest value. The point where the residuals were largest was always
relaxed as the next step in the technique. During the decades of the 1940s
and 1950s, southwell’s scheme to solve the incompressible, viscous flow
over a cylinder. This solution was obtained by hand calculation and
represented a substantial amount of work. Their calculation added to the
existing viscous flow solutions that began to appear in the 1930s.
During World War II and immediately following, a large amount methods
for solving problems in fluid dynamics. It was during this time that
Professor John Von Neumann developed his method for evaluating the
stability of numerical methods for solving time marching problems it is
interesting that Professor Von Newmann did not publish a comprehensive
description of his method. However O’ Brien, Hyman, and Kaplah (1950)
later presented a detailed description of the Von Neumann method. This
paper is significant because it presents a practical way of evaluating stability
that can be understood and used reliably by the scientists and engineers. The
Von Newman method is the most widely used technique in CFD for
determining stability. Another of the important contributions appearing at
about the same time was due to peter Lax (1954). Lax developed a technique
for computing fluid flows include shock waves that represent discontinuities
in the flow variables. No special treatment was required for computing the
shocks. This special feature developed by Lax was due to the use of
conservation law form of the governing equation and is referred to us shock
capturing.
At the same time, progress was being made on the development of methods
for both elliptic and parabolic problems. Frankel (1950) presented the 1 st
version of the successive over relaxation (SOR) scheme for solving
Laplace’s equation. This provided a significant improvement I the
convergence rate. Peacemann & Rachford (1955) and Douglas & Rachford
(1956) developed a new family of implicit methods for parabolic and elliptic
equations in which sweep directions were alternated and the allow step size
was unrestricted. These methods are referred to as alternating direction
implicit (ADI) schemes and were extended to the equations of the fluid
mechanics by Briley and Macdonland (1973) and Warming (1976, 1978).
This implementation provided fast efficient solves for the solution of the
Euller and Navier-stokes equations.
Research in CFD continued at a rapid pace during the decade of the sixties.
Early efforts at solving flows with shock waves used either the Lax approach
or an artificial viscosity scheme introduced by Von Newmann and
Richtmyer (1950). Early work at Los Alamos included the development of
schemes like the particle I cell (PIC) method, which used the dissipative
nature of the finite-difference scheme to smear the shock over several mesh
intervals (Evans and Harlow, 1957). In 1960, Lax and wnedroff introduced a
method for computing flows with shocks that was second-order accurate and
avoided the excessive smearing of the earlier approaches. The MacCormack
(1969) version of this technique became one of the most widely used
numerical schemes. Gary (1962) presented early work demonstrating
techniques for fitting moving shocks, thus avoiding the smearin associated
with the previous shock –capturing sches. Moretti and Abbett (1966) and
Moretti and Bleich (1968) applied shock-fitting procedures to
multidimensional supersonic flow over various configurations. Even today,
we see either shock-capturing or shock fitting methods used to solve
problems with shockwaves.
Godunov (1959) proposed solving multidimensional compressible fluid
dynamics problems by using a solution to a Riemann problem for flux
calculations at cell faces. (1979) showed how high-order schemes could be
constructed using the same idea. The intensive computational effort
necessary with the approach led Roe (1980) to suggest using an approximate
solution to the Riemann problem efficiency. (flux-difference splitting ) in
order to improve the efficiency. This substantially reduced the work required
to solve multidimensional problems and represents the current trend of
practical schemes employed on convection-dominated flows. The concept of
flux splitting was also introduced as a technique for treating convection-
Dominated flows. The concept of flux splitting was also introduced as a
technique for treating convection-Dominated flows. Steger & Warming
(1979) introduced splitting where fluxes were determined using an upwind
approach. Van Leer (1982) also proposed a new flux splitting technique to
improve on the existing methods. These original ideas are used in many o
the modern production codes, and improvements continue to be made on the
basic concept.
As part of the development of modern numerical methods for computing
flows with rapid variations such as those occurring through shock waves, the
concept of limiters was introduced. Boris & Book (1973) first suggested this
approach, and it has formed the basis for the nonlinear limiting subsequently
used in most codes. Harten (1983) introduced the idea of total variation
dimensioning (TVD) schemes. Others that also made substantial
contributions to the development of robust methods for computing
convecton-dominated flows with shocks include Enquist and Osher
(1980,1981), Osher(1984), Osher and chakravarithy (1983), is not an all-
inclusive list, the contributions of these dissipation with limiting as a major
factor in state of the art-schemes in use today.
Other contributions were made in algorithm development dealing with the
efficieny of the numerical techniques. Both improve the convergence rate of
iterative calculations. The muliti-grid approach was first applied to elliptic
equations by Federendo (1962), 1964) and was later extended to the
equations of fluid mechanics by Brandt (1972, 1977). At the same time,
Strides in applying reduced forms o the Euler and Navier-Stokes equation s
were being made. Murman and Cole (1971) made a major contribution in
solving the transonic small-disturbance equation by applying type dependent
differencing to the subsonic and wupesonci portins of flow field. The thin-
layer Navier-Stokes equations have been extensively applied to many
problems of interest, and the paper by Pulliam and Stegger (1978) is
representative of these applications. Also, the parabolized Navier-Stokes
(PNS) equations were introduced by Rudman and Rubin (1968), and this
approximate form of the Navier-Stokes equations has been used to solve
many supersonic viscous flow fields. The correct treatment of the stream
wise pressure gradient when solving the PNS equations was examined in
detail by vignerton et al (1978), and a new method of limiting the stream
wise pressure gradient in subsonic reginons was developed and in dominant
use today.
In addition to the changes in treating convection terms, the control-volume
or in treating convection terms, the control volume or finite-volume point of
view as opposed to the finite difference approach was applied to the
construction of difference methods for the fluid dynamic equations. The
finite volume approach provides an easy way to apply numerical techniques
to unstructured grids. With the development of methods that are robust for
general problems, large-scale simulations of complete vehicles are now a
common occurrence. Among contributions to this effort and Jameson and
Baker (1983), Shang et al. (1987), Obayashi et al. (1987) Yu et al (1987) and
Buning et al. (1988). At this time, the simulation of flow about a complete
air craft using the Euler equation is viewed as a reasonable tool for the
analysis and design of these vehicles. Most simulations of this nature are
still performed on serial vector computers. In the future, the full Navier-
Stokes equations will be used, but the application of these equations t entire
vehicles will only become an everyday occurrence when large parallel
computers are available to the industry.
The progress in CFD over the past 25 years has been enormous. For this
reason, to give credit to all who have contributed a number of review and
history papers that provide a more precise state of the art may be cited and
include those by Hall (1981), Jameson (1987) kutler (1993), Rubin and
Tnanehill (1992), and MacCormack(1993). In addition the focus’ 92 issure
of Aerospace America are dedicated to a review of the study of CFD should
also be mentioned inany brief history.
The development of any field is closely paralleled by the appearance of
books dealing with the subject. Early texts dealing with the CFD include
books by Roache (1972), Holt (1977), Chung (1978), Chow (1979), Patankr
(1980), Baker (1983), Peyret and Tyler (1983) Anderson et al. (1985), Orand
an Boris (1989) and Anderson (1995). The interested reader will also note
that occasional writings appear I the popular literature that discuss the
application of digital simulation to engineering problems. These applications
include CFD but do not usually restrict the range of interest to this single
discipline.

S-ar putea să vă placă și