SIMULATION OF FLUID FLOW IN INTERNAL COMBUSTION

ENGINES USING WAVE ACTION SIMULATION

by

VENKATA SUMAN MOVVA, B.Tech.

A THESIS

IN

MECHANICAL ENGINEERING

Submitted to the Graduate Faculty

of Texas Tech University in
Partial Fulfillment of
the Requirements for
the Degree of

MASTER OF SCIENCE

IN

MECHANICAL ENGINEERING

Approved

May, 2004
 ACKNOWLEDGMENTS

I like to express my sincere gratitude to Dr. Timothy Maxwell for introducing me

to the world of Engine Simulation. This work would not have been possible without his
constant guidance and encouragement. I would also like to thank my committee
members, Dr. Atila Ertas and Dr. Siva Parameswaran for their advice, suggestions and
time. I would also like to thank Dr. Thomas Burton for providing a scholarship for my
entire duration of study.
I would also like to thank my friends Deepak Sahini, Sandeep Singh, and Maaroof
Shahriar for helping me in preparation of this document.
Finally, I would like to express my gratitude to my parents and my brother for
their encouragement, support and well wishes.

11
 TABLE OF CONTENTS

ACKNOWLEDGMENTS ii
ABSTRACT v
LIST OF TABLES vi
LIST OF HGURES vii
NOMENCLATURE viii
CHAPTER
1. INTRODUCTION 1
2. LITERATURE SURVEY 5
3. GOVERNING EQUATIONS 9
3.1 One-Dimensional Inviscid Flow 10
3.2 The Continuity Equation 10
3.3 Momentum Equation 11
3.4 Energy Equation 13
3.5 Conservative Form of the Governing Equations 15
3.6 Numerical Methods 17
3.6.1 Method of Characteristics 17
3.6.2 Courant, Isaacson, and Rees Method (CIR) 17
3.6.3 Lax-Wendroff Schemes 18
4. METHOD OF CHARACTERISTICS 19
4.1 Conservation Equations 19
4.2 The Characteristic Equations 21
4.3 Non-Dimensional Characteristics Equations
and Riemann Variables 27
4.4 Numerical Solution of Non-Steady Flow Problems 29
4.5 Formulation of Characteristic
Equations for Numerical Solution 30
4.6 Generalized Characteristics 30

111
 4.7 Representation of Generalized
Riemann Variables at Boundaries 33
4.8 Simple Boundary Conditions 34
4.8.1 Closed End 34
4.8.2 Open End 35
4.8.3 Inflow into a System 35
4.9 Numerical Solution of the
Characteristics Equations 36
4.10 Zone of Dependence and Grid Structure 37
4.11 Stability Criterion 37
4.12 Evaluation of Riemann Variables 38
4.13 Computation Procedure 40
4.14 Partially Open-End Boundary Equations 40
4.15 Flows through a Valve 45
4.16 Cylinder Boundary Conditions 55
5. RESULTS 61
6. CONCLUSIONS AND FUTURE WORK 83
6.1 Conclusions 83
6.2 Future Work Required 83
REFERENCES 85
APPENDIX 88

IV
 ABSTRACT

Engine design is a long and costly process due to the large number of variables
that have an effect on the design. Engine Simulation plays a very important role in the
development of engines. It helps in analyzing the various engine configurations without
actually building the engine. Thus, Engine Simulation helps in reducing the cost and time
involved in developing a new engine. Many commercial codes are available in the market
to simulate engines. In recent years. Wave Action Models to simulate engines are
becoming popular due to their accuracy and robustness and are replacing the old Quasi-
Steady and Filling and Emptying Models. There are two parts in engine simulation, fluid
flow Simulation and Combustion and Heat Transfer Simulation.
In this thesis an attempt has been made to simulate the fluid flow in Internal
Combustion Engines and Engine manifolds using Wave Action Simulation. The Non-
Linear, Hyperbolic, Partial Differential Equations are solved using the Method of
Characteristics (MOC) which is the oldest and the most powerful method to formulate the
boundary conditions. A FORTRAN code is developed for obtaining the pressure and
velocity of the gas for various boundary conditions. Using the code the pressure and
velocity at any point in the manifolds can be obtained. The compression pressure of the
cylinder can also be estimated using the code. Pressure and Velocity distributions are
obtained from the results obtained from the code clearly show the Wave Action
Phenomenon.
The results obtained using the Code developed are in excellent agreement with
results obtained by other researchers.
 LIST OF TABLES

5.1 Input Conditions for a Simple Pipe Program 61

5.2 Sample Results Obtained from Simple Pipe Code (10 Nodes) 62
5.3 Sample Results Obtained from Simple Pipe Code (30 nodes) 64
5.4 Sample Results Obtained from Simple Pipe Code (50 nodes) 66
5.5 Sample Results Obtained from Velocity Code 70
5.6 Input Conditions to a Partially Open End Program 72
5.7 Sample Results Obtained from Partially Open End Code 72
5.8 Sample Input Conditions to Cylinder Program 77
5.9 Crank Angle Variation with Valve Area 78
5.10 Sample Results Obtained from the Cylinder Code 79

VI
 LIST OF HGURES

3.1 Control Volume for Homentropic Flow 10

4.1 Three-Dimensional Surface defining Characteristic Solution 24
4.2 Projection of Line on Characteristics Surface at c=0 24
4.3 Two Alternate Systems 31
4.4 Representation of Riemann Variables 33
4.5 Representation of Riemann Variables at Boundaries 34
4.6 Representation of Inflow Boundary 35
4.7 Zone of Dependence 37
4.8 Grid Structure 37
4.9 Evaluation of A at a Mesh Point 38
4.10 Partially Open-End Boundary Condition 41
4.11 Flow through a Valve 46
4.12 Control Volume for Engine Boundary 56
5.1 Pressure versus Time (10 Nodes) 63
5.2 Pressure versus Time (30 Nodes) 65
5.3 Pressure versus Time (50 Nodes) 67
5.4 Pressure at Pipe Entry versus Time 68
5.5 Pressure at Mid-Point versus Time 69
5.6 Velocity versus Time 71
5.7 Pressure versus Time for a Partially Open End Boundary (0.25) 73
5.8 Pressure versus Time for a Partially Open End Boundary (0.5) 74
5.9 Pressure versus Time for a Partially Open End Boundary (0.10) 75
5.10 Pressure versus Time for Different Nozzle Area Ratios 76
5.11 Variation of Cylinder Pressure with Crank Angle 80
5.12 Exhaust Pipe Pressure (Partially Open End) versus Crank Angle 81
5.13 Exhaust Pressure (Cylinder End) versus Crank Angle 82

vii
 NOMENCLATURE

a Speed of sound
a^ Speed of sound at reference pressure
a^ Stagnation speed of sound
A Non-Dimensional Speed of sound (aIa^^j.)
AQ Non-Dimensional speed of sound
\ Partially Open End Parameter A^ IA^^
C Velocity of Gas
C Specific heat at constant pressure
C„ Specific heat at constant volume
D Diameter
/ Friction factor
F Cross Section Area of pipe
m Mass flow rate
M Mach Number
p Pressure
Pg Stagnation Pressure
n , Reference Pressure
r ref
q Heat Transfer rate per unit mass
Q Heat transfer rate per unit time
r Cylinder pressure/Reference Pressure
r Pressure Ratio
m
T Temperature
TQ Stagnation Temperature
u Gas Velocity
U Non-Dimensional Velocity i//a^^^
[/„ Non-Dimensional velocity M/flip
X Non-Dimensional distance .v/L^^^
Z Non- Dimensional Time
K Ratio of Specific Heats
;^/iij Generalized Characteristics
;i^ fi Riemann variable of characteristic
^ Non-Dimensional Mass flow parameter

Vlll
 /f-i

^ Pressure function for valve —

'0
J
P Density
"^u Shear stress on pipe wall
, f
*P Nozzle area ratio, —^
F
W Valve-to-Pipe area ratio.

IX
 CHAPTER 1
INTRODUCTION

The first intemal combustion engine was developed by Otto way back in 1876.
After so many years of development it might be thought that the IC engine would have
been perfected. The contrary is the case as every decade seems to produce new
challenges. Even today a lot of money is spent on engine research to make the design
better, more reliable and more fuel efficient.
The design of engines remains a major challenge even today because the gas flow
in engine cylinders is unsteady. The scavenging process and the unsteady combustion
process add to the complexity of engine design. The shear number of variables which
have an influence on the intemal combustion engine make the development process a
long and costly procedure. Considerable capital is expended on the design, prototype
models and testing before the manufacturer can release an engine to the general public.
It probably requires little effort to convince the engine builder of the benefits of
engine simulation. A technique which would help engineers and researchers analyze
various engine designs without the expense and time of producing one off prototypes
would be an invaluable asset. This "asset" is the computer simulation and its use in
industry has become a mainstay over the past 30 years.
The essential elements involved in the prediction engine performance can be
classified as:
1. Air (Fluid) Flow Modeling,
2. Combustion and Heat Transfer in Engine Cylinders.
Air flow Modeling: It has long been realized that the performance of any
reciprocating engine is a function of the air and fuel flow through it. A naturally aspirated
diesel engine cannot produce a brake mean effective pressure of much in excess of 7 bar
simply because there is insufficient air to bum any more fuel. However, this power output
can be increased by supercharging, which effectively increases the air flow through the
engine and the fuel input can be increased roughly in proportion. Thus air flow plays an
important role in the power output of the engine.

I
engine and the fuel input can be increased roughly in proportion. Thus air flow plays an
important role in the power output of the engine.
The objective of this thesis is to discuss the various methods to simulate the air
(fluid) flow, advantages and disadvantages of each of the methods and to develop a
simulation model which will accurately predict the pressure and velocity of the gas in the
engine cylinder and in the inlet and exhaust manifolds. A FORTRAN code is developed
for the simulation of homentropic flow through constant area ducts using the equations
developed by Benson [1]. The various boundary conditions that have been considered for
the flow are given below:
1. Pipe with Open End,
2. Pipe with Closed End,
3. Pipe with Inflow,
4. Pipe with partially Open End,
5. Cylinder Boundary Condition,
6. Flow through Valves.
There has been and will always be a continuous effort by engine researchers to
develop computer models which are more efficient and accurate in representing the actual
physical processes occurring inside the intemal combustion engine. The knowledge of the
thermodynamic and fluid mechanics processes has increased and this leads to the
development of better simulation models which are sufficiently accurate to help with
design optimization. The models vary in their complexity and their ability to incorporate
critical phenomenon.
In general there are three types of engine simulation models that incorporate
varying levels of accuracy and completeness. The three types are:
1. Quasi-Steady Flow models,
2. Filling and Emptying models,
3. Wave Action models.
The level of complexity increases from the Quasi-Steady models to the Wave
Action models. Each model has its own advantages and disadvantages. The simulation
method that is adopted in this thesis is the Wave Action simulation which is the most
recent and the best method to simulate the flow as it takes into consideration the wave
motion inside the cylinder and the manifolds. The filling and emptying models and
Quasi-Steady models are primitive models and they do not consider the wave action in
the IC engine
Quasi-Steady models are relatively simple models that consider the engine and
turbocharger as a series of interconnected components. The performance of each of these
components is represented by its steady state characteristics, often obtained
experimentally. These components are connected by the air flow passing through them
and by the pressure ratios across them. This method is easily recognized by the fact that it
cannot predict power levels at various engine speeds. It can only predict the maximum
power potential. It also completely ignores the important wave tuning effect created by
intake and exhaust ducting dimension and cannot calculate the inertia ram effects of the
gas particles which occur before the intake valve closes. Quasi-Steady models provide
quick, cheap simulations but care has to be taken when extrapolating the data beyond the
valid range.
Filling and Emptying models are next in sophistication. These models use the
finite volume of the manifold sections and then calculate the mass flow rates across these
sections. The result is a set of first order non-linear ordinary differential equations. The
conditions in the cylinder and the manifolds are obtained by solving the differential
equations. These models are more realistic then the Quasi-Steady ones and require less
empirical data. However the computer time needed for their solution is greater than
Quasi-Steady. Filling and Emptying models are capable of predicting power levels across
differing engine speeds. However, like the Quasi-Steady models they also ignore the
wave tuning phenomenon and cannot calculate the inertia ram effect of the gas particles
which occur before the intake valve closes.
The significant influence of "tuning effects" created by intake and exhaust
ducting dimensions has been know to engine builders for many years. The Wave Action
Models were developed to take these effects into account. Wave action simulators are the
most sophisticated of the simulation models available for analysis of the gas exchange
processi They are extensively used in the industry for engine research and development
and should likewise be of extreme value to the high performance engine builder. Many of
the simulation programs available in the market use the "Wave Action Simulation
Method." This thesis will also be based on Wave Action Simulation to model the
pressure and velocity variations inside the engine cylinder and manifolds.
Waves occur in the manifolds of engines because of the periodic nature of the
induction and exhaust process. These processes act upon the compressible fluid in the
manifold, viz the air or exhaust gas, and pressure waves are transmitted through it. The
pressure waves travel at the local velocity of sound in all directions relative to the fluid.
The gas flow which operates inside the intemal combustion engine is very unsteady
because of the gases constant change of pressure, temperature and velocity throughout
the engine cycle. For example, when an intake valve is closed the gas velocity is zero.
When the valve opens the pressure differential begins to accelerate the gas particles into
the cylinder. This motion never occurs in a steady flow manner hence the discipline of
Unsteady Gas Dynamics is used to model the motion of gas particles inside the IC
engine.
 CHAPTER 2
LITERATURE SURVEY

The importance of simulation to the engine manufacturers has been emphasized in

the previous chapter. So to achieve better and faster simulations newer methods are being
proposed. This chapter looks into the history of development of simulation to the latest
trends available in the simulation field.
The flow behavior in the intake and exhaust systems of an IC engine plays an
essential role in determining the volumetric efficiency and residual gas fraction which, in
tum, have an important influence on fuel consumption and emissions. The study of wave
motion laid the foundation for the development of the wave action simulation. The study
of wave motion can be traced back to the sixth century BC when Pythagoras noted the
ratio of frequencies produced by a vibrating string as its length was changed. Much of the
early effect of wave motion focused on linear waves. Earashaw [2] was the first to
develop non-linear equations of superposition for sound waves. He showed that the
amplitude of the combined wave could be related to the individual wave by a seventh
power law. In 1858, Riemann [3] had proposed the Method of Characteristics for solving
of wave equations. This method was the dominated the field until 1950's when finite
difference methods started to be developed. De Haller [4] used a graphical method to
solve the set of simultaneous differential equations for the gas flows in the engine exhaust
systems. Jenny [5] extended the graphical calculations to include pipe wall heat transfer,
friction, entropy gradient and gradual pipe cross sectional area changes.
The mesh method of characteristics was the first finite difference technique to be
applied to the simulation of gas dynamics in engine manifolds [6]. This technique formed
the basis for development of computer programs by research groups and industrial
organizations through out the world. Many research papers have been written on the
application of mesh method of characteristics to the simulation of fluid flow in engine
manifolds. Benson [7] in 1971 described two complete cycle synthesis programs for
compression ignition engines. The programs include the intake pipes, the exhaust pipes
and turbocharger if it is fitted. The results obtained from the computer program were
checked with the experimental results. The results obtained were in good agreement with
the experimental and graphical calculations.
In 1981, S.C.Low and P.C. Baruah [8] presented a paper on a numerical scheme
for a computer aided design (CAD) package for the design and development of internal
combustion engines intake and exhaust manifold. The program was interactive and used
graphical and visual display units (VDU) facilities extensively. The basic concept was
presented in an earlier paper by S. C. Low. The program was applied to a commercial 4-
cylinder 4-stroke engine.
Benson [1] was the first to develop the code by solving the one- dimensional
formulations of the governing equations. This code was based on the mesh method of
characteristics. Different versions of this code are still in use even today. Blair [9] and
his co-workers at the Queen's University of Belfast developed a similar code.
After the development of the code for predicting the performance of the engine
was developed, the efforts were concentrated on using the code for inlet and exhaust
tuning and in turbocharger matching. Winterbone [10] applied the non- homentropic
method of characteristics to turbocharged diesel engines. He proposed a method of
analyzing the exhaust systems based on the transmission of available energy. He used it
to assess the various losses of energy as exhaust gas is transmitted down the pipe.
The code developed using the method of characteristics had several advantages and
disadvantages. The primary disadvantage was the time taken to run the code. When the
computation time became an issue the efforts of the researchers were concentrated on
improving the computation time and reducing the number of calculations involved.
Lakshminarayanan [11] used fourth-order Runge-Kutta method to solve the partial
differential equations. Ledger [12] investigated the various finite different schemes for
the solution of non-steady compressible flow equations. Pearson and Winterbone [13]
presented a computationally efficient simulation technique which is based on the
linearised one dimensional conservation equations. The technique was found to be 220
times faster then a comprehensive simulation program based on the method of
characteristics.
Some other disadvantages of the method of characteristics are that it is not
conservative, and hence there is no guarantee that the continuity, energy, or momentum
equations are solved correctly. The method also cannot cope with large pressure waves.
To overcome these disadvantage researchers began looking at the newer finite difference
and finite volume schemes which can overcome this defect. The problem with the finite
difference and volume techniques are that they require a researcher to have a high
mathematical background to be able to use them.
Some of the widely used finite difference schemes are the Lax-Wendroff Method
[14] and Richtmyer [15]. It was the start of a whole family of codes using the Lax-
Wendroff method. In Germany, industry and universities worked together for years in the
development and validation of the computer code taking this approach. Richtmyer [16]
proposed a two step approach to solve the equations of the unsteady gas flow. This
method is called the Two-Step Lax-Wendroff method. The MacCormack method [17, 18]
is another second-order accurate scheme that becomes identical to the Lax-Wendroff
scheme for the linear case. The scheme uses altemate forward and backward spatial
differences for the two steps in space (sometimes referred to as the 'predictor' and
'corrector' steps). The two-step Lax-Wendroff and MacCormack's methods are used
extensively in engine simulation codes in order to achieve second order accuracy.
The problem again with the second order accurate schemes discussed above is
that they produce spurious oscillations when shock waves and contact surfaces are
encountered. T.Bulaty and H. Niessner [19] presented a paper in 1984 in which they
suggest the addition of non-linear terms (flux limiters) to cope with such oscillations. The
Flux corrected transport (FCT) technique has been adopted by several research groups to
cope with oscillations. This technique appears to work satisfactorily for many flows in
engine manifolds [19, 20, 21, 22, 23]. It was shown by R.J Pearson and D.E.Winterbone
[24] that under some circumstances the FCT method can distort the solution and produce
large errors in mass conservation. The symmetric flux limiter due to Davis, based on the
total variation diminishing criterion (TVD) was shown to avoid such problems.
The search for newer methods to simulate the flow better and in a more accurately
is going on even today. The use of multidimensional method was also considered in
some papers [25] especially at the boundaries where the flow need not be one-
dimensional. The modeling of engine manifolds using one-dimensional techniques has
now reached a high level of maturity, and most of the basic phenomenon in the intake
and exhaust manifolds can be analyzed with a high degree of accuracy.
Even after the development of a number of schemes the MOC (method of
characteristics) is still popular and is widely used in many commercial codes. The MOC
provides insight into the physical wave propagation process and effective time-varying
boundary conditions. Yuhua Zhu and R.D. Reitz [26] used the MOC to develop a code
for the unsteady flow in internal combustion(IC) engines. They also applied it to predict
the residual gas fractions. The code was applied satisfactorily to predict the gas exchange
process of a spark ignition engine following blow-down and exhaust gas recirculation
(EGR) levels in a heavy duty diesel engine. One-Dimensional modeling is useful not only
in predicting the performance of the engines but is also very useful in conjunction with
multidimensional CFD codes such as KIVA since it provides more realistic boundary
conditions and shortens the computation time by utilizing 1-D modeling of intake and
exhaust systems rather than requiring 3-D computing.
 CHAPTER 3
GOVERNING EQUATIONS

In this chapter the equations used to evaluate the unsteady flow of gas in engine
manifold are derived. The equations that are obtained are partial differential equations
which can be solved by using a variety of methods which will be outHned in this chapter.
The advantages and disadvantages of these methods are also discussed.
The flow paths in internal-combustion engines can be extremely complex. These
paths are often simplified by considering the flow to be one-dimensional. In general, the
gas is highly turbulent and fluid frictional forces are present within the fluids and at the
walls. In this analysis the fluid is assumed to be inviscid (that the gas is sufficiently
dilute for intemal stresses to be ignored) and wall friction is also neglected. The gas is
assumed to be perfect. The gas flows in intemal combustion engines are non-steady and
changes in the fluid properties take place in space, x and time, t. However, under some
conditions the spatial rates of change of fluid properties are far greater than the temporal
one. These conditions are called quasi-steady flow conditions and steady flow gas
dynamic relations are used.
The entropy levels in the various parts of the flow may have significant effects. If
the entropy level is uniform everywhere and does not vary with space or time the flow is
called homentropic flow. If the entropy levels are not uniform, the flow is non-
homentropic. It is sometimes possible to subdivide parts of the flow in non-homentropic
flow, enabling homentropic relations to be used for each region. Suitable arrangements
must be made to patch together the regions at the interfaces. In this simulation, the flow is
homentropic and code is developed based on the homentropic flow equations. It is not
realistic to assume homentropic flow in the exhaust systems of intemal combustion
engines and non-homentropic models must be employed.
3.1 One-Dimensional Inviscid Flow
The fundamental equations of fluid mechanics are mathematical statements that
define the conservation of mass, momentum, and energy for a control volume. A
conservation law states that the rate of change of conserved flow property in a fixed
volume is the result of the net effect of the flux of the property across the boundary of the
volume and the change in that property due to intemal sources.
Consider the flow of a compressible fluid through an infinitesimal section of the
pipe in which the area of the cross-section perpendicular to the axis of the pipe varies, as
shown in the Figure 3.1. The area variation is gradual and the fluid properties are
approximately uniform across any cross-section and are taken as functions of 'x' and '/'
only. All these assumptions make the flow quasi-one-dimensional.

3.2 The Continuity Equation

The mass conservation law dictates that the rate of change of mass within the
control volume shown in Figure 3.1 is equal to the net mass flow through the element.

P dp.
dx
P
u U+ 5 AC
F dx

Figure 3.1 Control Volume for Homentropic Flow [27].

10
The rate of mass flow entering the control surface is

"/(.r) = puF. (3.1)

The rate of mass flow leaving the control surface is

m(x + dx)= puF + -~{puF)dx. (3.2)

ox
The rate of decrease of mass within the control surface ABCD is

—r-ipFdx). (3.3)

In any system the mass must be conserved,

=> puF + — (puF)dx-puF= - — (pFdx). (3.4)

ax dt
This simplifies to
d(puF), a
dx dx = -—(pFdx).
dt (3.5)
On expanding and rearranging
dp du dp pu dF
dt dx dx F dx

3.3 Momentum Equation

The momentum equation states that the sum of pressure forces and shear forces
on the control surface shown in Figure 3.1 equals the sum of rate of change of
momentum within the control volume and the net efflux of momentum out of the control
surface. The pressure forces on the control surface are equal to the sum of forces on the
end faces and the force on the side walls. The shear forces are due to friction at the wall.
The forces are given by the following expressions:
Pressure Forces

pF-(p + ^dx)(F-t^dx) + p — dx = -^(pF)dx+p — dx. (3.7)

dx dx dx dx dx

11
Shear Forces
2

T^nDdx---f^^nDdx. (3.8)

Where D is the equivalent diameter and the wall shear stress r„, is related to the friction
factor by the expression

/=-p^. (3.9)
-^P^r

The rate of change of momentum in the control volume

-J

— (pFudx). (3.10)
dt
The net efflux of momentum from the control surface is

(p + ^dx)(u+—dx)-(F + —dx)-pu^F = — (pu^F)dx. (3.11)

dx dx dx dx
Hence the momentum equation form the above equations is

(pF)dx + p dx- fpu^ -7tDdx = — (pFudx) + —-(pFu''-)dx. (3.12)

dx dx 2 dt dx
Expanding, dividing by F and changing the sign gives
dp du pu dF dp. . pu' 4 dp du du /-, nx
u(-^ + p — + - + u^) + f- -h-^-l-p—--HpM-- = 0. (3.13)
dt dx F dx dx 2 D dx dt dx
The term in the bracket is identically zero, from the continuity equation. Hence the
momentum equation becomes
du du 1 dp ^ ^ ,» , ..
— +M— + - + G = Q. (3.14)
dt dx p dx
Where G = f——^-= 2fu\u\ (3.15)

The term u/|u| or u\u\ is introduces to ensure that the wall friction always acts in opposite

direction to the direction of motion.

3.4 Energy Equation
The specific stagnation enthalpy is given by the expression

CT + ^ + ^ . (3.16)
2 p
The specific internal energy is given by the expression
2

C7 +y . (3.17)

The energy equation for a fixed volume shown in Figure 3.1 is

Q- ^s = -, + (Net efflux of stagnation enthalpy) (3.18)

dt
The heat transfer rate is
Q = qpFdx. (3.19)
where q is defined as the heat transfer energy per unit mass per unit time. The extemal

rate of shear work W^ is zero.

The time rate of change of intemal energy in the control volume is

^^^l{(pFdx)(CJ + ^)}. (3.20)

dt dt 2
The net efflux of stagnation enthalpy from the control surface is

^{(puF)(CJ + ^ + -^)}dx. (3.21)

dx 2 yO
Hence the energy equation becomes, after substitution of equations (3.19) to (3.21) into
equation (3.18)

qpFdx = -^{(pFdx)(CJ + '-^)} + — {(puF)(CT + '-^ + ^)}dx. (3.22)

dt 2 dx 2 p
The energy equation can be simplified using the continuity equation and momentum
equations. On expanding equation (3.22) and subtracting the continuity equation (3.6)
and writing

^ ^ l + u^, (3.23)
Dt dt dx

13
The energy equation becomes

^ ( C J +^ ) =, - i ^ ^ - ^ ^ . (3.24)
Dt ' 2 pdx pF dx

D jr ^ 1 du' u du'
T;:M=T^r- +T^^, (3.25)
Dt 2 2 dt 2 a;c '
du -, du
= u^-
— ++ ir
u-—-.
— (3.26)
dt dx
From the momentum equation (3.14)
// dp du T du ^
- ^ = M^- + »-^—+ (<G. (3.27)
p dx dt dx
Using equation (3.26) the above equation becomes
u dp D M' ^ ^
— - ^ = — ( — ) + MG. (3.28
pdx Dt 2

Substituting into equation (3.24) for ( ) yields

p dx

-—(CJ + —) = q + -—(-—) + uG -—-—. (3.29)

Dt 2 Dt 2 pF dx
or

-—(C^T) = q + uG - ^ - — . (3.30)
Dt pF dx

Consider the term — ( C T )

Dt "

|^(C.n = | ^ ( - i - ^ ) . (3.31)
Dt Dt K-1 p

- 1 .(1J^_4J^). (3.32)
K-1 p Dt p' Dt

_Li,(|l+„|l)-P(|l+„|l)). (3.33)
K-1 p dt dx p dt dx
From the continuity equation (3.6)

14
 d(Fu) F ^dp dp,
ax p dt dx
Further ,

p a~
- =— , (3.35)
P K

Substituting — (C^T),—-—and — into equation (3.30) produces, after simplification,

Dt dx p
the final form of the energy equation

^ + u ^ - a \ ^ + u^)-(K-l)p(q + uG)^0. (3.36)

dt dx dt dx
Thus, equations (3.6), (3.14) and (3.36) are the governing equations for the one-
dimensional flow of a compressible fluid in a pipe with area variation, wall friction, and
heat transfer. These relations are a set of non-linear hyperbolic partial differential
equations.

3.5 Conservative Form of the Governing Equations

Equations (3.2.5) (3.3.5) and (3.4.7) can be rearranged as follows
Momentum equation
d(pn^d(puF)^^ (3.37)
dt dx
Continuity equation
d{puF)^d(pu' + p)F _pdF^^y , ^^^^ (3.38)
dt dx dX 2
Energy equation
d(pe,F) ^ d(puKF) ^pp_^ (3.39)
dt dx

where e„ = C„r + - M ' and /JQ = e^ + ^ . (3.40)

° 2 p
Expanding and rearranging equations (3.5.1) (3.5.2) and (3.5.3) gives the
following equations

15
 dp ^ d(pu) ^ pudF _^
(3.41)
dt dx F dx
d(pu) , d(pu- + p) , pu- dF ,
r I 1 h pu = U . (3.42)
dt dx F dx
d(pe^ d(puh^ pu\ dF
•+- --I-- -pq = 0. (3.43)
dt dx F dx

Where G = / — — —
2 IMI D
Equations (3.5.5) (3.5.6) and (3.5.7) can be written in symbolic form as
dW dF(W)
(3.44)
dt' ox
where

p ' pu pu ' 0
W = pu ,F(W) = pu^ + p ,C(W) = pu^ d(\nF) ^ pG (3.45)
pu\ puh^ dx
P^o pq
When there is no pipe area variation, wall friction, or heat transfer the equation
reduce to
dW dF(W)
•+- = 0. (3.46)
dt dx
The presentation of equations in the above form is called the conservation law form [27]
since it is obtained from the integral conservation equations of mass, momentum, and
energy applied to the fixed control volume.
In general, it is not possible to obtain the analytical solutions for the hyperbolic
partial differential equations derived above and thus numerical techniques have to be
developed. Basically it is necessary to discretize the governing equations to form a set of
algebraic relationships which can then be solved on a computer. The goveming equations
define the propagation of information through the fluid in the manifolds as pressure
waves, which has a major effect on the engine performance and noise levels. It is
essential, therefore, to employ numerical methods that have the ability to resolve the
wave-action phenomenon if successful simulations of engines are to be performed.

16
3.6 Numerical Methods
3.6.1 Method of Characteristics
The theory of characteristics is a tool for handling general systems of partial
differential equations. The method was initially developed by Riemann [3]. This method
became the foundation for future studies in propagation of non-linear waves. This method
was used by Jenny [4] as a graphical technique to calculate flows in engine manifolds.
Benson [1] was the first person to use the technique as a numerical method in a computer
program for the simulation of flows in engine manifolds. He called it the Mesh Method of
Characteristics. This method was the dominant technique till the 1990's and even today is
used in many commercial engine simulation codes. The advantage of this method is that
it gives excellent insight into the propagation of waves in fluids. Even today it is the most
powerful approach to formulating boundary conditions in many modem engine
simulation codes. In recent days, this method has been replaced by the modern finite
difference schemes due to their accuracy and robustness. In this thesis, the mesh method
of characteristics will be developed for homentropic flow as this simplified approach
allows the clearest exposition of the significant features of this theory. The homentropic
approach itself can form an acceptable approximation of the flow in intake manifold,
where the energy and entropy levels of the gas do not vary greatly. However, the full
non-homentropic equations must be used for reliable predictions in exhaust systems.

3.6.2 Courant, Isaacson, and Rees Method (CIR)

The CIR method is equivalent to the mesh method of characteristics and uses the
same computational stencil as the mesh method of characteristics. Both the CIR method
and Mesh Method of Characteristics (MOC) are first-order accurate in space (due to
linear interpolation of characteristics) and time (because the time steeping procedure is
based on the forward Euler technique). Neither scheme is based on the conservation law
form of the equations. Roe [28] showed that discontinuities do not travel at the correct

17
speed. LeVeque [29] showed that any numerical scheme which is not based on the
conservation law form of the governing equations will not converge to a physically
consistent solution when discontinuities are present. CIR method and MOC also present
considerable problems when the variations in the gas properties are to be modeled
consistently.

3.6.3 Lax-Wendroff Schemes

These schemes were introduced by Lax and Wendroff. These schemes are based
on the conservative form of the Euler equations. Since these are based on the
conservative form the results obtained from these methods are more accurate and can
deal with discontinuities. The problem with these methods is that they cannot handle all
the shock waves that are encountered periodically in the exhaust manifold of the intemal
combustion engines. These schemes are used in many engine simulation codes.
There are many other high resolution schemes and Riemann Solvers which are
used to solve the hyperbolic partial differential equations but it is beyond the scope of
this thesis to discuss all of the methods.

18
 CHAPTER 4
METHOD OF CHARACTERISTICS

In this chapter, the Method of Characteristics is described and the equations used
in homentropic constant area ducts are derived. The boundary conditions that are
considered are given below:
1. Pipe with Open End,
2. Pipe with Closed End,
3. Pipe with Inflow,
4. Pipe with Partially open End,
5. Cylinder Boundary Condition,
6. Flow through Valves.

4.1 Conservation Equations

The three conservation equations are Continuity, Momentum, and Constant
Energy. The equations are given by equations (3.37), (3.38), (3.39). In order to simplify
the analysis, the following assumptions are made:
1. Area of cross section remains constant;
2. Friction is neglected;
3. Entropy is constant.
After making the above assumptions the continuity equation becomes
d{pu) dp
(4.1)
dx dt
The momentum equation becomes

±=n£l (4.2)
dx Dt

where — is the substantive derivative of velocity with respect to time

Dt

£l = ^ + u^, (4.3)
Dt dt dx

19
Expansion of Equation (4.1) yields
1 dp u dp du ^
—^ + - + — = 0, (4.4)
p dt p dx dx
Expansion of equation (4.2) gives
1 dp du du ^
—T7- + — + u — = 0. (4.5)
p dx dt dx
The speed of sound is related to the pressure, p and density, p and entropy, s by the
relation given below
2 .3p
a - (-^) (4.6)
dp'
For an ideal gas
2 Kp
(4.7)
a =

A reference speed of sound a^ is introduced, which corresponds to the speed of sound of

a gas at some reference pressure P^^^^to which the gas is insentropically expanded or

compressed from the local conditions (Speed of sound a and pressure p ).

The constant entropy condition leads to the following equations

iK
( „ XTx
(4.8)
ref \ " y

( \K-\
(4.9)
or
'ref v^^;
where K is the ratio of specific heats for a perfect gas
For isentropic flow p,,f,p,,f and a^ are constant. Equations (4.1.8) and (4.1.9) are written

in differential form as
dp _ 2K da (4.10)
p K-1 a
dp _ 2 da (4.11)
p K-1 a

20
Equations (4.4), (4.5), (4.10) and (4.11) are the conservative equations for homentropic
flow. Expressing equations (4.10) and (4.11) in terms of partial derivates and time'r' and
space ' .V ' gives the following equations
1 dp 2K 1 da
p dx K-ladx (4.12)

p dt K-la dt ' (4.13)

pdx K-ladx' (4.14)

Substitute for — — and —^ from (4.13) and (4.14) into (4.4) gives the continuity

equation in terms of ' a ' and ' M ':

2 da 2 da du ^
7 ^ + — r « ^ + « ^ = 0- (4.15)
K-ldt K-1 dx dx
Using equation (4.1.12) the momentum equation is written in terms of ' a ' and ' u '
2a da du du ^ ,, ^ ^^
K-ldx • + ^
dt^ + "7—=
dx 0. (4.16)

Two conservative equation (4.15) and (4.16) in terms of the local speed of sound 'a ' and
the particle velocity ' ; / ' are obtained. The objective now is to evaluate ' a' and ' ;<' at
each point ' x ' is the system at time 'r'.

4.2 The Characteristic Equations

Equations (4.15) and (4.16) form a set of quasi-linear hyperbolic partial
differential equations. A solution set of the form a = a(.v,r)and u = u(x,t) needs to be
obtained from these quasi-linear equations.
As discussed in the previous chapter these equations cannot be solved by
analytical methods and numerical solutions must be employed to obtain a solution. There
are many methods for solving the equations. The method that is used in this thesis is the
Method of Characteristics (MOC). A simple approach proposed by Benson [1] that

21
reduces the volume of algebraic manipulations generally involved with characteristic
theory is used to solve the equations.
The two separate equations of the solution, namely a = a(x,t) and u = u(x,t) are
transformed to another grouping c = c(x,t) andc = c(u,a). After transformation, a unique
relationship between ' c ' , '/<', ' a ' is sought such that the solution c = c(x,t) will give
numerical values for '//' and ' a ' at ' x ' and 'r'. If such a relationship can be
estabUshed, then it follows that the solution c = c(x,t) will represent a three-dimensional
surface. Special lines on this surface have the property that when projected onto the x -t
surface at c=0 the slope of the projected line has a numerical value equal to the value of c
at that point. Changes in the fluid properties (in this case 'a' and 'M') along the projected
line can then be examined. If a solution to a particular problem is obtained, a combination
of the substantive derivatives of the fluid properties along the projected line, together
with the slope of the line at a point, will be equivalent to the original equations. The set of
partial differential equations have been transformed to a set of ordinary differential
equations along the projected lines. These projected lines are called characteristics.
To obtain the characteristics solutions the original equations (4.15) and (4.16) are
K-1
transformed using linear combinations. Equation (4.15) is first multiplied by to

give the following expression

da da K-1 du ^ ,. ,~.
—-fH — + ——a — = 0. (4.17)
dt dx 2 dx
K-1
Equation (4.17) is then added to times equation (4.16) to give the following

,da , da. K-1 du _^^^^"^_n (A\9.\

dt dx 2 dt dx
The second Combination is

Equation (4.16) - - ^ x(Equation (4.15)

22
Direct solution of Equation (4.18) and (4.19) would lead to the form
a = a(x, t), (4.20)
" = u(x t). (4.21)
In order to obtain a characteristic solution it is assumed that 'a' and 'u' are uniquely
related by the expression
c = c(a, u). (4.22)
Then the solution to equations (4.18) and (4.19) will be of the form
c = c(x, t). (4.23)
This solution is represented by the curved surface bounded by the edge PQRS as shown
in Figure.4.1. In this solution the numerical value of c is represented by the height of the
surface above the x-t plane at c= 0. The surface projected on the x-t plane at c= 0 is
represented by the broken line bounded by the edges P'Q'R'S'
Consider a line AB on the surface bounded by PQRS and its projection on the x-t
plane at c=0, A'B'. The line A'B' is defined as a characteristic if at any point x, t the
slope of a line in the x-t plane at c = 0 is given by the numerical value c at that point. This
definition can be illustrated by reference to Figure 4.1 and Figure 4.2. Along the line AB
in the surface bounded by PQRS there is a point m. This point, when projected onto the x-
t plane at c=0 will he on the line A'B' at m'. The slope of A'B' at the point m' is given

hy(^y\) . If A'B' is a characteristic then this slope will be equal to the numerical

height mm'

23
 .J>-*--' B' IY
I

Distance

Figure 4.1 Three-Dimensional Surface defining Characteristics Solution [1]

langent lo /).'_,•

Time t

Figure 4.2 Projection of Line on Characteristics Surface at c-0 [1]

( ^ P ^ , = mm'. (4.24)

or, in general, along a characteristic in the x-t plane

,dx_ (4.25)

The fluid properties 'a' and 'M' along a characteristic are now examined. Since 'a' and
'M' are functions of 'x and 'r' throughout the flow field, there will exist some
functions a = |a(x,OL;^,, M = |M(X,OL;,„, along a characteristic.

24
Along a characteristic

/ , X ,da da , ,
(^«)....=(3^^^ + ;^^^),,,„ (4.26)

(^")..«.=(-^^? + - ^ ^ ) , , „ , . (4.27)

The total derivates along a characteristic are then

,da da da ,dx.

and
,du
,UM _duuu duuu dx
,ux,
(—dt W
'''""'= —
dt + —
dx (—)c.«.;
dt (4.29)
or
,^a, da da
( dt'"^'
- r X w =dt
^ + ^dx^ - (4.30)
and
,du, du du
{^\har=-^ + -:rC- (4.31)
rfr ""^ dt dx
The following three conditions then hold along a characteristic

(f)c.„.=c; (4.32)
,(ia, 3a da
{—)c,ar=—
• )char = +
+ — C- (4.33)
dt """^ dt dx
,du, du du
{—\^r=— + — c- (4.34)
dt "^"' dt dx
These equations represent the change in 'a' and 'M' with x and t along specially defined
curves called characteristics. The function c = c (a, u) now must be determined to satisfy,
along characteristics, the equations (4.32),(4.33) and (4.34) as well as the basic equations
(4.18) and (4.19) valid throughout the field.
If the function c is taken as c = M -i- a, then it follows from equations (4.25),(4.32)
,(4.33) and (4.34)

25
 (—) = c = ;, + «. (4.35)
dt
and

( ^ ) + ^ ^ ^ = 0. (4.36)
dt 2 dt
or

(-) =- ^ ^ . (4.37)
du 2
Equation (4.25) is replaced by a set of two characteristic equations. The first equation
(4.35) is called the direction-condition equation and gives the slope of the characteristic
in the x-t plane. The second equation relates 'a' and 'M' along a characteristic and is
called the compatibility condition.
If the function is taken as c = a-u, then from equations (4.25),(4.32),(4.33) and
(4.34) the direction condition is
{^) = c^u-a. (4.38)
dt
and the compatibility equation is
^ /£-l^ (4.39)
dt 2 dt

or
,da_ 1^ (4.40)
du 2
Equation (4.25) has been replaced by the two characteristic equations (4.38) and (4.40).
The complete solution is represented by the characteristic equations

(—) = u±a (4.41)

dt
'da\^^i<^ (4.42)
^du j 2
These are the homentropic non-steady equations of fluid flow.

26
4.3 Non-Dimensional Characteristic Equations and Riemann Variables
A direct representation of pressure can be obtained for homentropic flow if the
velocities are divided by a reference speed of sound a ,,which is defined as the speed of

sound corresponding to the isentropic expansion or compression from the local pressure p
and the speed of sound a to a reference pressure p^^^. This non-dimensionalizing

procedure is also extended to the distance x and time t by the introduction of a reference
length L^^^

The non-dimensional groups are

a ^, " ,, X ^ _^ref^
A= •,f/=—,x=-—,z = - (4.43)
"ref '^ref ^ref '^ref

The original characteristic equations were

dx (4.44)
(—) -u±a.
dt
(da\ K-1
= +- (4.45)
^du^
Now the above equations in non-dimensional form become

(—) = [/±A (4.46)

dZ
(da\ K-1 (4.47)
^du ^ = +-

A = ^ - ( P \(«--l)/2>c- (4.48)
^ref Pref

dA, K-1
Consider the expression (—7) = r"'
du 2
On integrating the above expression the following relation is obtained
K-1 (4.49)
A = — -U + C.

27
 At [/=01et A = A^C = A^A = A-i-^^U (4.50)
2
Thus, for the A characteristic

'dX]
= U + A. (4.51)
ydZ,
K-1
{^) = - (4.52)
dU 2
A= A+^^U (4.53)

The equation (—-) = -}-—— on following a similar approach and assuming that the

value of A at C/ = 0 is yff, becomes

(4.54)
.dz)
dA K-1
= +- (4.55)
dU

J3 = A-^^U (4.56)

Where A and j3 are called the Riemann Variables.

If A and U are replaced by the Riemann Variables A andyff, then

K-1
For the /i characteristic

(4.57)
dZ 2(K-iy 2(K-1)
For the y^ characteristic

(4.58)
dZ 2(K-1) 2(K-1) '^
These equations give the slopes of the A and /? characteristic.

28
Along a A characteristic, A is constant and hence the slope is a simple expression in the
form

( ^ ) A = "'-'Jy^. (4.59)

Where m = { - f ^ } A,n = - ^ - ^ (4.60)

2(K-1) 2(K-1)
Hence there is only one variable in this calculation, fi

Similarly, along a /? characteristic yS is a constant and hence the slope is again a simple
expression of the form
,dX,
(-—)p = nA-m. (4.61)
dZ
, 3-K , K-\-l , ^
where 7? = ,m={ ]B (4.62)
2(K-1) 2(K-iy^
Here the only variable is A. These new equations give a reduction in the number of
variables from two to one. The equations (4.59) and (4.61) lead to the development of a
method for direct numerical computation of the characteristic equation.

4.4 Numerical Solution of Non-Steady Flow Problems

The advent of high speed digital computer gave the promise of greater efficiency
in solving non-steady flow problems. Initially attempts were made to 'plot' the
characteristics on the computer, but because of the necessity to store large amount of
characteristics intersections both in the pipes and at the boundaries, the storage
requirements soon made such solutions impractical. To reduce the core store required in
the computer, special consideration was given to the reformulation of the characteristics
into more efficient form. This Method was derived by Benson and was presented in his
famous paper [6].

29
^^^^^^^^^^^^^^^^Q^^^lChmctc^^ Numerical Solution

The Riemann variables from equations (4.53) and (4.56) are A = A-^^^U
2
K—1
md/3 = A—^U. The slopes of the characteristics are given by (—),=U +A
dZ ^
,,dX
'dZ ^ - f/ - A. The value of A is obtained by adding equations (4.53) and (4.56)

A^(A + j3)/2. (4.63)

The value of U is obtained by subtracting equation (4.56) from (4.53)
U^(A-J3)/(K-1). (4.64)
Substituting the values of A and f/into the slope equations above gives
.dX /f + 1 3-K ,„

dX 3-K .. ^ K-\-l ,_

Assume,
2>-K _, , K+1
a= and b = (4 61)
2(K-1) 2(K-1) ^ ^
Equations (4.65) and (4.66) become

(-^)i=^^-«^- (4.68)

(—)p=aA-bl3. (4.69)

Equations (4.68) and (4.69) have a symmetry which leads to the possibility of replacing
the two equations by a single equation, with suitable adjustment of either the constants
( a'
_ > and 'b\ or the Riemann variables A andy^.

4.6 Generalized Characteristics

In Figure 4.3, two Z-X systems are shown; system A is the conventional system
and system B is the altemate system.

30
 Positive X Positive X
System A System B

Figure 4.3: Two Altemate Systems [1]

In the conventional system A the A and P characteristics are A^andfi^ and their slopes
are given by
HY 1
({^h).-H-«y^.= ^ (4.70)
dZ tan 6*,

(Ap),=aA,-bP,=-^ (4.71)
dZ tan^o

In system B the X-direction is reversed. However, the inclination of the characteristics

are made the same with respect to the original X-axis
tan^, =-tan4), (4.72)

tan^« = - t a n O , (4.73)

In system B the slopes of the A and P characteristics are

({^h)«=H-«)^«=-^ (4.74)
dZ tan O^
({^},),=ai,-Z,^,=-^ (4.75)

Compare the y^ characteristics in the system A with the A characteristics in system 5

tan^«=-tan<I),. (4-76)

1 (4.77)
tan dp tan <I)^

Hence
(4.78)
aA, -bP, = - ( H -aPB) = ^PB -H •

31
The equation is satisfied if A^ = J3^ and P^ = A^

Consider the expression {—-}^ =bA-a/3 for the system A

^ = K^P = PA (4.79)

( | - h . = H - " ' ^ . ~ . (4.80,

For the system B

A = Ag,p = pg (4.81)

{—}A, = H - aPs = —-— - (4.82)

dZ tanO^
If the identity Ag = ^^ andfig= A^ is used then for system B

{—h„ =b/3, -aA, =-——- (4.83)

dZ " tanO,
Now,

bP^-aA^ = --—-. (4.84)

tan^^

Since in system A, X is in opposite direction to X in system B.

Substitute Equation (4.84) and (4.85) into (4.83) gives
dx 1
dZ " ian Up

This equation gives the slope of the y5 characteristic in system A, the conventional
system. This leads to the replacement of two equations (4.68) and (4.69) by a single
equation and a new definition for the /land P characteristics. If a /icharacteristic is
defined as one with nominal slope in the same direction as the positive X-direction and a
P characteristic is one whose nominal slope is in the opposite direction, then

{^],=bA-aP. (4.87)
dZ

32
The generalized characteristics A^andA,, as shown in Figure 4.4 are now introduced.

Left Hand Right Han(J

End End

Figure 4.4: Representation of Riemann Variables

A /I, characteristic is a characteristic whose nominal slope is from left to right in a duct. A

/^i characteristic is one whose nominal slope is from right to left in a duct.

Finally U and A are obtained in terms of A^andA^^

u =A- - ^ , (4.88)
K -1

A -
A,, - ^
->
(4.89)

4.7 Representation of Generalized Riemann Variables at Boundaries

In any calculation, the value of the Riemann variables "entering' the boundary are
known and the problem is to evaluate the Riemann variable of the characteristic 'leaving'
the boundary. The known variable, entering the boundary is called A,„ and the unknown

Riemann variable is called/l,^„,. A convention of positive velocity, U towards a boundary

is adopted as shown in Figure.4.5, which gives

33
 Left hand end: .V Riglit liand end:
odd end (a) even end

^4 (

-^oul
Positive U Positive U
towards boundary towards boundary

Figure 4.5: Representation of Riemann Variables at Boundaries [1]

U_ K~Kui
(4.90)
K-1

\ A.„+/l„
m out
9
(4.91)

4.8 Simple Boundary Conditions

The simple boundary conditions are
1. Closed End
2. Open End
3. Inflow into a System

4.8.1 Closed End

At the closed end it is know that the velocity U is zero. The goal is to determine
the Riemann variable A^^^^ provided the value of A,„ is known.
Substitute the value of f/ = 0 in equation (4.90) we get that the

Kut - K • (4.92)

34
The above relations are used in the subroutine call for Closed End.

4.8.2 Open End

/ NCC-I)/:

Let A = a where p^. is the Pressure at the open end. Since at the
ref
V P-f J
open end the pressure is atmospheric pressure A = 1

^A=A^li^=i (4.93)
2

^„.=2-4. (4.94)

4.8.3 Inflow into a System

The equations for inflow into a system are little more involved then the equations
for the open end and closed end. Let the stagnation conditions in the reservoir from which
the gas is flowing into the pipe be (aoPo).The gas then expands insentropically from

(aoPQ)to a, p as shown in Figure 4.6. Application of the steady flow energy equation

across the inlet to the pipe from the reservoir gives

/f;
I Pipe end

Entropy s

Figure 4.6: Representation of Inflow Boundary [1]

a,0 _ a •-t--
w (4.95)
K-1 K-1 2

T 2 ^ 1 0
=> ar," -a + u~.
(4.96)

In non-dimensional form.

35
 A 2 A'' tC — 1 ^ ,n
\ =A--^ U\ (4.97)

Substitute the value of A and U gives

A,+A. Y K-1 Am -Aout

A>- = -I-- (4.98)
K-1
J

-A„..r-2\ U,A,„ + - ^',„-4Ao4 = 0. (4.99)

K-1 K-1 ^K-lj

Ku. = aA^n + ^ { M / - ( l - a = ) / l , / } . (4.100)

^3-K^ K-1
Where a = and b= 4 (4.101)
K+1 K+ 1

4.9 Numerical Solution of the Characteristic Equations

The numerical solution of the characteristic equations will now be formulated.
The method was proposed by Benson [l].The solution is performed wholly in the
position diagram, i.e., the Z-X field, which is subdivided into a rectangular grid system.
The grid pattern in fixed in the X-direction but the Z-coordinates of the grid lines are
adjusted according to the stability criterion for the numerical solution of the quasi-linear
hyperbolic partial differential equations. This results in flow being evaluated at irregular
time intervals on spatially fixed lines. In order to maintain a regular pattem of rectangles
the grid proportions AZ : AX are fixed for each time interval AZ. At a given time
Z = Z^ the Riemann variables are calculated at each mesh point. From the Riemann
variables at time Z, the time interval AZfor the next step is evaluated using the stability
criterion. The next time, Z = Zj +AZis then determined and the procedure repeated. The
stability criterion is based on the zones and the domains of dependence of characteristics.

36
4.10 Zone of Dependence and Grid Structure
In Figure 4.7, the region 2, 3, 4 is called the Zone of dependence of the point 4.
The line 3, 2 is called the domain of dependence of point 4. These properties of
characteristics will be used in selecting the grid stmcture as shown in Figure 4.8 in the Z-
X field. The grid or mesh points are numbered in the sequence shown. The first number
represents the time step Z, the second number the X-step. The grid proportions are fixed
by the stability criterion.

N. D. distance A"

Figure 4.7: Zone of Dependence [1]

3.1 3.2 3.3 3.4 3.5 3.6

N.D
TIME 2.1 2.2 2.3 2.4 2.5 2.6
Z

l-l 1.2 1.3 14 1.5 1-6

N.D Distance X

Figure 4.8: Grid Structure

4.11 Stability Criterion

The stability criterion states that to evaluate the conditions (A, U, ovA.P) at

points r+1, s from the conditions at r, s-1, and r, s+1 the mesh proportions must be

selected such that

37
 AZ 1
(4.102)
AX A + \U\
The mesh proportion will therefore vary throughout the flow field (Z-X diagram). The
standard procedure is to fix the length AX and let AZ vary in each time step so that all the
points at that time step lie in the zone of dependence of the previous time step. This
implies that the values of A and U should be evaluated at each mesh point and
AZ selected from the minimum value of
^
1. 'A
AX. (4.103)
A + \U\
V
In practice the pipe length, L, is subdivided into equal grid portions. If m is the number of
meshes in a pipe, then the actual length AL of a mesh will be

(4.104)
m
4.12 Evaluation of Riemann Variables
A relation is to be developed between the values of X and P at time Z, and the
new value of X at time Z+AZ, at any mesh point (R). This will then be extended to cover
the whole grid system. Consider the grid PQRS shown in Figure 4.9

Figure 4.9: Evaluation of X at a Mesh Point [1]

The values of X and P at P and S are known, and it is required to evaluate l at R. If
the length PT is small, then it is assumed that the X characteristic PT through point P has
a slope corresponding to the values of A- and P at P

38
 The slope of PT is then given by
,rfX,
^~J^h=bA-aP = bA^-aP^, (4.105)

where a = ~ ^ and b = - ^ ^ ± L .4 ^Q^^

2(K~1) 2(K-l) ^^-^^^^
The length ST is given by

5Z,= ^ (4.107)
^ r ^dX^
vA
dZ J^
AX
^^T=-r] ^- (4.108)

The value of A^ is obtained by the hnear interpolation between S and T

\=^s + -Z-- (\ - /I,). (4.109)

Substituting the value of SZ^ from equation (4.108) the following expression is obtained

^« = ^5 + — ( H - ""Pp )(^p - A) • (4.110)

By using all the relations developed above, a code in FORTRAN is developed tofindthe
Pressure and Velocity at any point in a Simple Pipe with Constant area of cross section
and entropy.
Given below, are the salient steps involved in the development of the code. The
boundary conditions discussed in this program are the (1) closed end, (2) open end, and
(3) Inflow into a pipe from a reservoir. The first step is to collect all the input data
required for the program. The input data required are the length of the pipe, the initial
pressure, initial temperature, Reference Pressure, Boundary conditions at Z =0 along the
pipe length, the boundary condition at X = 0 and X = L for all time Z, Number of Mesh
Points required (the more the mesh points the more accurate the calculation and more
time the program takes to run!), and the value of gamma (ratio of specific heats). Based
on this data the grid system is chosen by ensuring that AZ satisfies the stability criteria. It
is important to ensure that AZ satisfies the criterion for all mesh points.

39
 Next the nodes where the output is required and the simulation time in seconds
are determined. After entering the required input, the calculations are carried out
according to the relations developed above and the output at the required points is given.
The data is collected and exported to an Excel file from where the pressure and velocity
distribution graphs are obtained. The various boundary conditions are included in the
program as separate sub-routines.

4.13 Computation Procedure

The computation procedure that is adopted in the FORTRAN code is given below
1. Enter boundary values of A^aftdA^^at all the mesh points at time Z=0.
2. Evaluate A and U at all the mesh points.
3. Compute minimum value of AZ for all mesh points.
4. Start from r =2, s=2 evaluate values of A^ at all mesh points s=2 to s = m+L

5. Starting from r =2, s=l, evaluate values of /I,, at all mesh points s =1 lo s = m.

6. Enter the boundary equations for (A^),Jxom the known value oi(\^)^^.

7. Enter the boundary equation for (A^^),^^Jrom known value of (/l,)^^^,.

8. Repeat Steps (2) to (7) for next value of r until time Z, is equal to required time
for calculations.
9. Export the output file to Excel.
10. Plot the Pressure, Velocity profiles as a function of simulation time.

4.14 Partially Open-End Boundary Equations

The flow through a partially open end is a most important boundary condition and
can be used to simulate many devices encountered on engines, including the turbocharger
turbine.
The flow through a partially open end will be assumed to be isentropic. It is
required to relate the speed of sound A to the particle velocity U in the pipe immediately
adjacent to the partially open end. The reference pressure will be equated to the outside

40
constant static pressure. For subsonic flow, this will be equal to the throat pressure/^,
The processes are shown in Figure 4.10
Speed of
sound a
y'"
y / y P
"o
a y^---,

a,
"rer

Entropy j

Figure 4.10: Partially Open-End Boundary Condition [31]

Energy Equation

ar,~=a H u -a, + u, . (4.111)

Continuity Equation
puF^p,u,F,. (4.112)

From the basic relationship for speed of sound

-, Kp (4.113)
a" = ,

Dividing equation (4.111) by a^/ gives

A,'=A'+^U'-=A'-*^Ur (4.114)

For isentropic flow

a -1 JL^
— \K-\ (4.115)
'ref

Dividing equation (4.112) by yO,,^ and a^^and substituting for — f r o m equation

r^ref

(4.115), the continuity equation becomes

U^/K-\ = ^A/X-^U^ (4.116)

41
In the above equations, the following convention is used
FT"

0 = ^,A=—andf/=-^
'•ref 4, r
ref

Combining equations (4.114) and (4.116) gives

(A:--A\)
2_ K-1
U'^ (4.117)
-1
o-
For subsonic flow

P, 1.0.
P.^Pref ' A=^- = (4.118)
aref

(A^-1)
(/• _ K-1
(4.119)
/ / i \ 'K--1
-1
O^ v i y

For chocked flow in the throat (t)

(4.120)
The critical pressure ratio is given by the equations

( ^ \ ( . \ '/.-I ( . \ '%^-i
a (4.121)
KP'J vA.

{- (4.122)

The continuity equation (4.116) now becomes.

/K--1
—= 0 A (4.123)
A
From the energy equation (4.114), for sonic flow through the throat

A-+- -f/^ = A'+- -t/. (4.124)

42
 K+1
-I- (4.125)
K-1 K-1

Using equations (4.122) and (4.123), equation (4.125) becomes

K+1 r .\ - ^ A \ —,
<D^- = (4.126)
K-1 K-1
vA; \ A^ J

Equations (4.119) and (4.126) relate the critical ratio to the area ratio O for sub-

sonic and sonic flow. To make the equations compatible for the computer calculation the
following algebraic manipulations are made
Cross multiplication of Equation (4.119) gives the following

{/'(A^'^-" -<D') = - ^ (A' -1)^-. (4.127)

K-1

but,

A=A+^^U. (4.128)

Now replacing f/by 'A' and A.^ gives

2 (4.129)
(A„-A)^
K-1

' 2 ^ f 2 ^
(A -A)HA^''-''-^')= - ^ (A'-l)^\ (4.130)
K-1 ^K-1^

K-1
(4.131)
./(A) = (4-A)^(A/^'^-'^-0^)-V-(^'-l)^'-
Equation (4.131) cannot be solved analytically and numerical methods must be
employed. Let suffix 'n' be the first estimate and n+1 the next estimate. Then the first
2 1 /I — 1
estimate is taken as (A)„ = -^ and the step is (AA)„ = - ^ ^ •

Substitute the value of 'A' in equation (4.131)

If /(A)<0,(A)„„=(A)„-(AA)„ (4.132)

43
 If/(A)>0,(A)„,,=(A)„+(AA)„ (4.133)
_(AA)„
Set(AA)„„ = - ^ ^ , Equation (4.132) is now recalculated with (A)„ = (A)„,, and with

(AA)„ = (AA)„,, and tested as before. The iterative procedure is repeated until the required

accuracy is reached. When A is known then A„„, will be found out by using the relation

^««,=2A-/l,„. (4.134)
The same procedure is used for sonic flow to find out the value of A. The test for sonic
flow is the Mach number in the throat.

(4.135)

If the velocity (f/,) in the throat is greater than 1.0, then the flow is sonic.
For sonic flow
'•V
—=o • U = ^CA where C = (4.136)
A \ ^ Jcr

The goveming equation for the sonic flow is given by the expression

fA} = 02_
K+1 f A \
'A' K-l

.f Q. (4.137)
K-1 K-1
K^J A
V, ^ Jcr

For the solution of this equation again numerical methods are used and the first estimate

1+. 1
'K + 1 ^K + 1
is taken as (\ )„ = and(AA„)„=- The iterative procedure is

repeated until / is equal to predetermined error.

K^Jcr

The relation between A and U in Riemann Variables is given by the following expression

U_ (4.138)
A K+Kut J K-1
This on cross multipHcation and rearranging gives the following expression

44
 ^^ K-1U\
A„„, = 2 A A„ (4.139)
1 +K-IU
2 ~A
K+U
Substituting U/A= <t>C = ^(\) ^"-^ gives

' l - ^ * C ^
Kut = A„. (4.140)
l+ ^ ^ O C
v
K-1 K+V
1-~^^(A,/A\^ /.-I
(4.141)
i+^4)(A/A),.;'>'-i

K-1 f+1/
l-^-0(A/A)„ /-'
A,. ={ K —1 «•+!/
)A„ (4.142)
1+-^0(A/A)„ /-•

Using all the relation derived above a subroutine is developed in FORTRAN for the flow
through a partially open end.
The main program is same as the one for simple boundary conditions. The only change is
that one more input condition which is the area ratio (O) is also added.
The program is included in appendix. The pressure variations as a function of the
simulation time is obtained and are plotted in Microsoft excel.

4.15 Flows through a Valve

The constant pressure model is applied to the outflow through an exhaust valve
into the exhaust pipe (the normal flow direction) and backflow through an inlet valve into
the inlet manifold (the reverse flow direction).
In this model it is assumed that the gas leaves the cylinder with stagnation
pressure p and speed of sound a^ and expands isentropically as it passes through the
passage between the valve head and the valve seat. The second assumption is that the
isentropic expansion ceases when the gas stream reaches the valve throat. The gas is then

45
considered to expand adiabatically, but irreversibily at constant pressure, to fill the pipe
cross-sectional area, when the flow is subsonic in the throat, but with a drop in pressure
when it is choked in the throat. The processes are represented in Figure.4.11

^ . . . . ^

N\\s\\s\sN

Cylinder conditions - stagnation

^H

0
^"^

1 ^^-^y^P,ti
y^
y^A,
^ ^1

Entropy s Entropy 5
Subsonic flow Sonic (clioked) flow

Figure 4.11: Flow through a Valve [1]

The conditions upstream of the valve (0) are related to the conditions at the throat
(1) and the pipe (2), where the flow runs full. The basic equations are given by
Energy Equation:
K I T T K V 1
«o = « i " + ^ - " i ' = « 2 " + - 2 (4.143)

Continuity Equation:
yOjMlFi = P2M2F2 . (4.144)

From (0) to (1), the state changes are given by

K-\

/ _ ^-^ /" -o V2
a,0 _ £0 (4.145)
a, KP^J

The speed of sound is given by the relation

46
 2 P
a =K—
(4.146)

The following sign convention is used

F,
W = ^ ,7t =^ P 2 ^ - ,f/ = u. (4.147)
F, PoJ
A relationship between ;rand U for various values of \i/\s required.
For subsonic flow the pressure Pi= p,- Applying this to the continuity equation derived
above following relationship is obtained.
1 p2
M, = -U., . (4.148)
W P
From the relationship for the speed of sound,
/ A
(4.149)
A P^ v«^-y
But Pi = P2' applying all this to the continuity equation the following relation is obtained

r ^\
M, = — (4.150)
¥ V 'J
The energy equation is rearranged to give
K-1 T
« l ' = «0 (4.151)

Substituting the value of u^ derived above in the energy equation leads to the following

expression

K-1 1 ( \
«o^- u^ (4.152)
2 w
V ^y
Rearranging the above equation the following expression is obtained

-1I/2
( \
-1 (4.153)
¥ f \
2 "
K-1
\ ' J
V °y

47
From the energy equation on rearrangement

a2 K-1
= 1- (4.154)

£-1 = — and p, = p , . (4.155)

71

/ „ ^
a.
Substitute for and t / = ^ - in equation (4.153) the following expression is
V °y v ^ y
obtained

1/2
2 1
V-1 v (4.156)
n K-1 \n'
V 2
When the flow is just choked at the valve throat, u^=a^. The energy equation between
(0) and (1) is now
2 K+ 1 .
(4.157)

/ 1 \"2
a„ (4.158)
V 2 ,

At the sonic boundary, p^ = p^

K-\ K-\
z' „ ^ (K + 1\
a,0 _ [PA (4.159)
UJ yP-j [ 2 J
K-\
1/2
^P2^^ ^ 2 ^ (4.160)
7r =
AC + l
v^oy
Substitute for ;rin equation (4.156) the following relationship between f/and ^^is

obtained

48
 U
¥•• (4.161)
K+1
l-'^U'

When the pressure downstream of the valve is below the critical pressure, p, will

not be equal to p^^. The continuity equation now becomes

( \
": ^¥ ^2
«i (4.162)
a, \P-]
V ' y
Rearranging the above equation leads to the following expression
f \ f ^ \ f .. \ f \
U-, a. Po_ (4.163)
<3n Ch) Poj yP2j

From the energy equation

a. K-1 Uj
1- (4.164)
(3„

K+1
*0 _ (4.165)
V 2 ,

V.-
(4.166)
y_P0j
K+1

f \ / A f \
a^
Substitute the values of andf/ = ^ - , the above equation becomes
V °y Ch] yP.j On

K-1 ^ ^ ^
f r, \
1-
z' „ ^ % 2(/r-l)l
(4.167)
vPoy
^w ^K + lj u^

( \ /l":
Also 7t - El (4.168)
v^°y

49
 / O \ /[2(/f-l)]
l_^^(/3
•n = { V (4.169)
K+1 f/

This is the equation for choked flow in the throat.

When there is sonic flow in the Pipe then //, = a^^nd from the ellipse of energy

, K+1
«o =• u^. (4.170)
2 '
( 9 V
^ t / = ^ = (4.171)
K+1
V
In the derivation of the equations for flow through a valve one fundamental
assumption that the entropy is constant through out the flow is made, but in reality there
is an increase in entropy as gas flows from the cylinder to the pipe. The magnitude of the
entropy change depends on the area ratio of the valve to the pipe^//, and the overall
pressure ratio expressed as n. Hence even if it assumed that the entropy variation is small
enough to be neglected, it is still necessary to select a level at which to maintain the
entropy throughout the calculation.
Benson [1] derived the expression for taking care of the entropy variation and is
given by the following expression.

K-1 \
7r„ = •
"•AC
{^8).+^^- (4.172)

Where m is any arbitrary point where the entropy level is desired, a^^, a^p are two

possible entropy levels at point m. Tip is defined as follows

Fref
Tto^P (4.173)
Pc

The next step is to modify the above equations so that they can be adopted for solving in
a digital computer. If outflow from a cylinder to a pipe is considered, then A,„,>;i,„,
which means that U is negative.

50
Let A = P
,Ug-ulaf. , 71 =
' _PY ,LJ =u/ a
ref
"ref
V P-^ J

Equation (4.172) then becomes

K-1 f \
7r = • (f/fi) + ^„ (4.174)
v"-y
also, A«=A + ^ f /

Let C = —^represent the entropy change across the valve, then equation (4.174) on

rearranging gives the following expression

2C
- ^ B = -{T^-7^a)- (4.175)
K. 1

On substituting f/^in place of U in equation (4.156) gives

1/2
(4.176)
-u.
71 K-1 W i-'^u.
V

Substituting equation (4.175) into equation (4.176) gives

2£_ (7t-7l„)
K-1
71 K-1
' \7t^
4-il ^ K-1\2C ,
0. (4.177)

Let5 = C ( ; r - ; ^ „ ) , then

( 2
1/2 (B)
' -L-. K-1 = 0. (4.178)
/i(^) = -
7t /C-lU 1 ^(Bf
K-1
Equation (4.169) becomes

51
 2K
,/f+l/
/[2(A--1|| l-^(/'
7t = { ¥ (4.179)
,/c- + l ,
-u.
Substitute the value of (-Ug) the following expression is obtained

1 ' ^ - l / 2C ,
7T = •(
f 9 ^ /|2(/r-l)| l--^^(^-^J
V^ (4.180)
v'^ + ly 2C
(7z--^„)
K--1
(^ O ^'^/|2(ft:-
Let F = ^ and 5 is assumed above
,/f + l ,
Then the above equation simpHfies to

1 ?-5^
• t,n-\ _.2/f//f-l
/ K-1 = 0. (4.181)
2(71) =7t
B
K-1
The expression for the entropy change is derived by Benson [1] and is given by the
following expression

{^-^Uif-
C= (4.182)
71

(1-^B'r-
K-1 (4.183)
71
Neither equation (4.178) nor equation (4.181) can be solved analytically and numerical
methods must be employed as was employed for the flow through a partially open end.
For each value of (^there is a critical value ol7t = 7i^^. If 7i:^>7r^^then the flow is

subsonic, if TT^ < TT^^ there is sonic flow.

52
 The equation giving the relationship between Tt^^andy/is given by Benson [1]

[30] .The relation is given below and is used to determine weather the flow is subsonic or
sonic.

7r^,^=-^(2i^ + ^[l-{l + (K' -Dyr^f']) (4.184)

The above expression is used to determine whether the flow is subsonic or sonic and then
solve equation (4.178) or equation (4.181).
If the flow is subsonic, the initial value will be

1+5 . „. . . . . r 2 T
7r = - where S is given by the expression and the
v'^ + ly
1-S
increment will be taken as A;r =

S +7t
If the flow is sonic, then the initial value is given by ;r = ^and the increment is
2
5 -7t 1
ts.7Z = -. The first estimate for C is taken a s — .
4 71^
Now,
X ^(K-\)/2K: \(K-\)/2K , x(K--l)/2/f
P Pref ' Pref ^
TT = A (4.182)
P^) Pref Pc yP^ J

From this equation

^ ^K-\)I2K

Pc (4.183)
A = 7l = 4+^../2.
yP-fj
^ ^(.K-\)I2K

Pc (4.184)
A =27t -A.-
^Pref,

Hence the value of /l„„, is determined from the known values of p^, 4 and calculated
value ol7i. For inflow, the boundary equations corresponding to the partially open end
are used. The equations (4.131) and (4.142) are used but with A.„ multiplied by the

53
 ( Y/r-l)/2/f
t^ref
expression •To obtain A^^Jox inflow through the valve the following
Pc

transformation is used

/ \XK-\)I2K
A =(Aout ')PartiallyopenEnci P''f (4.185)
\ Pc
re J

The flow direction is checked by applying the following conditions

7i„ >1.0 then there is inflow

7t^ <1.0 then there is outflow

7t^ =1.0 there is no flow.

The mass flow rate through the valve is given by the expression
dm
dt puF. (4.186)

p is the pipe density , F is the pipe cross sectional area

dm,e . . , Pref ^ . U
ct, -'•'t''^''- '*'"'
A FORTRAN subroutine is developed using all the above equations. The input
din
parameters to the code are A^^andi//. The output is /l„,„and—'-. The results obtained
dt
from the code are given in the next chapter. The calculation procedure is given below
1. First the valve area ^^is tested.
2. If it is Zero, the valve is closed, if it greater than 1.0 then it is set
equal to unity, since the flow is controlled by pipe and not the valve.
3. The constants are then evaluated including;r^ , basing on the value of

7T^ the flow is determined as inflow, outflow and no flow.

4. If the flow direction is inflow then the partially open End Subroutine
is used. The input is changed accordingly as discussed above.
5. If the flow direction is outflow then the valve equations are used.
6. Next the flow is checked for Subsonic or Sonic flow.

54
 The relevant equations are used basing on the flow.
dm,
8. Finally A and are obtained from the code.
dt
4.16 Cylinder Boundary Conditions
In the cylinder boundary condition the logic control for the valve or port timings
must be included. The calculation of flow rates through the valve or ports will be
discussed below and then the evaluation of cylinder pressure will be considered. There
are a number of ways of evaluating the mass flow rate through valves and ports. Here
only the simplest form of calculation proposed by Benson [1]. In flow between a cylinder
and receiver the steady flow equations and isentropic pressure drop equations are
assumed to be valid. Let p^j be the pressure upstream and p.^ be the pressure downstream.
The mass flow rate is given by the expression

m = P2U2F-
2^2 2 '
(4.188)

The energy equation is

K-1
a^; =a{+- (4.189)

\lK \\IK
A ^Po,
P2 (4.190)
P2 = .'oi
\ Poi Poi 01

Combining equations (4.188), (4.189) and (4.190) gives the expression

nl/2
.2lk
2 \
1- II.
ma, 01 2K
_ Pi. (4.191)
^f =
PoA K-1 Poi Pox

The expression for ^^ is computed directly from the pressure ratio p^Jp^acro^s the

valve or port depending on the direction of flow.

When the flow is choked the above equation becomes
/ - N(/f + l ) / | 2 ( » f - l ) l

. = J ^ l . (4.192)

Equations (4.191) and (4.192) are used mamly for engine intake where the manifold

pressure is often constant.

55
To calculate the cylinder pressure the following assumptions are made in the calculation
of cylinder pressure
1. The pressure in the cylinder at any instant is almost uniform
2. As the flow is homentropic it is also assumed that the temperature in the
cylinder is uniform and equal to the stagnation temperature upstream of
valve.
3. The heat losses are assumed to be negligible
By making the above assumptions the lengthwise wave action in the cylinder is
effectively neglected. Since the cylinder volume will vary with time a control surface
must be defined round the cylinder. Figure 4.12 shows the engine cylinder with air
entering and gas leaving the cylinder.

Control surface
/
Cylinder
Air in Exhaust out
'nc'^'c'''c./'c.ec
'",,-''oa'''oa '"e.^o

Figure 4.12: Control Volume for Engine Boundary [3]

Applying the generalized first law of thermodynamics to the variable volume open
system gives the following equations.
3(^)rv • +, ... rum„ut(\)„^,-m.^(\).
^ _ ^ /u \ (4.193)
Q-W^-W^ =
dt

There is no shaft power so W, =0; there is no heat transfer (2=0 and the work done to

change the control volume is W^ =pX- Hence the first law is reduced to

_py ms^^m„u,(K)^-m^Me- (4-194)

dt

56
The terms in the above equation can be expressed with reference to the system shown in
Figure 4.12 as

(E)^=,ne,. = P^ (4.195)
K-1
d{El Pcdy,_^y dp ^
(4.196)
dt K-1 dt dt

Ka- (4.197)
A--1
r ^
dm
m„ = (4.198)
dt
V y
a^
K. = K = (4.199)
K-1
( dm\
m„ = (4.200)
ydt J

The first law for the open system is therefore

dm
( dm\ Ofl
dt
PcdV, PcdVc , ,. dp^ dt
+K + - (4.201)
dt K-1 dt dt K-1 K-1
On rearranging the expression in the required form is obtained

dp c_ dm ^ dm^ PcdK
Oa
•K (4.202)
dt ~ K dt
v^y dt
V y
This expression relates the rate of change of pressure to the flow rates, the stagnation
speed of sound, the cylinder pressure, the volume and the rate of change of volume.
'^ dm^
The mass flow rate is obtained directly from the valve calculation where the
dt
following expression was derived
U_ (4.203)
dt «„/ A'

57
 f \
dm
The mass flow rate of may be obtained from the valve calculation if there is an
dt
V y
intake pipe. The equation is given below
f \
dm ( dm\
(4.204)
dt ^dt J valve
\ y
The cylinder mass balance equation is
f \
dm^ dm dm\
(4.205)
dt dt y~^)e
\ y

dm
If there is a receiver then the value of is calculated from the receiver equations
dt
\ J
derived above.

Let r^ -^^ then equation (4.191) becomes

Pox
1/2
^2K'^
(4.206)
f{0 = ^K-1^

for r. >
AT-1-1

and
K+X
/ n \ 12(K--1)]
(4.206)
.f{0-^^K + lj

KiK-X
( 2 \
for r <

The mass flow rate is

dm, _ p.F^ (4.207)
f{rj-
dt a„

If the flow is outflow from the cylinder then the mass flow rate is given by the expression

58
 dm„ _-p,F2
f{rj- (4.208)
dt a^

Where p^ is the air pressure, a„ is the stagnation speed of sound, p,. is the cylinder

pressure, F^ is the instantaneous area of flow.

Benson [1] proposed the use of Euler's integration algorithm to calculate the
change in the cylinder pressure. The cylinder volume and change in cylinder volume is
given by the following expressions by Benson [1]
2
V, =F^r 1 + 71-7(rt'-sin^ 6')-cos 6*+ - (4.209)
CR-l]
where F^ is the cylinder cross sectional area, r the crank radius= stroke/2, n = connecting
rod length/crank radius, 6 = angle from TDC (top dead centre), CR is the compression
ratio. The rate of change of cylinder volume is given by the expression

dV. sin 2^
^ = Fr sin^-i-- 271N, (4.210)
dt 9 yj(n~ - s i n " ^

where A^ = Crank speed in rev/second.

The valve area is obtained by linear interpolation from a set of experimental
values given in the form of a tabular input. For the exhaust valve, the valve area ratio is
determined while for air ports the minimum area is obtained. After the valve areas have
been determined the cylinder pressure is calculated from the previous cylinder pressure
and the rate of change of pressure using Euler integration. Now i„„, is determined for the
dm
valve using the Valve subroutine by knowing A.^ and ij/. —^ is also determined from the

Valve subroutine. The cyUnder volume and the rate of change of cylinder volume is
determined from the equations derived above. Finally, the cylinder gas temperature and
the rate of change of cylinder pressure are determined.

59
 To run the program the user must input the following required data
1. Initial Pipe pressure
2. Cylinder Pressure at E.V.O
3. Cylinder Temperature at E.V.O
4. Reference Pressure
5. Pipe Length
6. Pipe Diameter
7. Number of Mesh Points
8. Ratio of specific Heats
9. E.V.O from T.D.C
10. Two-Stroke or Four-Stroke
11. Engine Speed
12. Number of points where output is required
13. Location of the output points
14. Duration of time in crank angle the code is to be run
15. Cylinder Diameter
16. Stroke
17. Connecting Rod Length
18. Compression Ratio
19. Air Pressure and Temperature if Air Valve is Fitted
20. From the Valve Opening Diagram [1], the number of input points are to be
specified
21. Angle from the opening of the valve
22. The corresponding valve areas
23. The above data is repeated for the Exhaust Valve also
24. Nozzle Area Ratio in case we use the partially open end routine.

60
 CHAPTERS
RESULTS

In this chapter, the results obtained by mnning the code developed using all the
equations derived in the previous chapter are discussed. The first code that is tested is the
code for a simple pipe with three boundary conditions which are the open end, closed end
and inflow. The input conditions are given in Table 5.1

Table 5.1.: Input Conditions for a Simple Pipe Program

Condition Value
Initial Pressure in the reservoir 1.5 bar
Initial temperature in the reservoir 300 K
Reference Pressure 1.0 bar
Length of the pipe 1.0 meter
Mesh size 10 nodes
Ratio of specific heats 1.4
Output Locations 1,5,11
Simulation time 0.05 seconds

Node 1 corresponds to the first point in the pipe, i.e., the pipe entry. Node 5
corresponds to the mid-point in the pipe. Node 11 corresponds to the open end of the pipe
.In this case it is assumed that one end of the pipe is open to a receiver and the other end
is the open end. The initial pressure in the reservoir is 1.5 bar. The result obtained on the
running the code is given below. Table 5.2 given below shows the simulation time in
seconds, and pressure at the three node points. As an example only the first 20 points are
shown.

Table 5.2: Sample Results Obtained from Simple Pipe Code (10 Nodes)

61
 Table 5.2: Sample Results Obtained from Simple Pipe Code (10 Nodes)

Time in Seconds Pressure at Node 1 Pressure at Node 5 Pressure at Node 11

2,88E-04 1.5 1.5
5.23E-04 1.5 1.5
7.58E-04 1.5 1.5
9.93E-04 1.5 1.5
1.23E-03 1.5 1.5
1.46E-03 1.5 1.5
1.70E-03 1.5 1.433893
1.93E-03 1.5 1.34982
2.17E-03 1.5 1.271702
2.40B-03 1.5 1.205961
2.64E-03 1.499337 1.15298
2.87E-03 1.49481 1.111406
3.11E-03 1.484391 1.079505
3.34E-03 1.469003 1.055567
3.58E-03 1.450709 1.050478
3.81E-03 1.431364 1.068997
4.04E-03 1.412952 1.101013
4.26E-03 1.396015 1.133636
4.48E-03 1.380748 1.16159

The pressure distribution obtained when the results are plotted is shown in Figure
5.1. The pressure at the open end of the pipe is a constant and is equal to 1.0 bar. Since
the pipe end is open to atmosphere, the pressure does not change through out the
simulation time. At the pipe entry, the pressure decreases at first and then remains
constant for a certain period of time and then decreases as the simulation time increases.
At the mid point, the wave action phenomenon can be felt with increased vigor because
the waves are reflected at the open end and come back because of this the pressure first
decreases and then increases and after certain point of time the effect becomes smaller
and smaller and the pressure gradually falls and reaches a constant value.

62
 Pressurf at
1 4 • ,__-^ ' pipe entry

Pressure at Mid-Point
1 ?

1-

\
0 8- \ Pressure at
Xopen end.

1
1

0-
0.01 0.02 0.03 0.04 0.05 0.06
Time(seconds)

Figure 5.1: Pressure versus Time (10 Nodes)

Next the same code with the same input conditions is mn, but the number of nodes
is increased from 10 to 30. Here the output locations are Node 1, Node 15, and Node 31.
Node 1 corresponds to pipe entry. Node 15 corresponds to mid-point of the pipe, and
Node 31 corresponds to the open end of the pipe. The output obtained is given in Table
5.3. Only the first 20 points are given. On increasing the number of nodes from 10 to 30,
the accuracy increases and the simulation time also increases. The number of points
where the pressure is obtained is increased. The data is exported to excel sheet and the
pressure distribution obtained is shown in Figure 5.3. The wave action can be more
clearly seen in this case, then the previous case when only 10 nodes were considered.
When 10 nodes were considered, the second pulse was not completely visible where as in
this case it is cleariy seen. There is no change in the pressure at the open end. The results

63
at the mid-point and pipe entry are much more accurate then those obtained when the
code was run with 10 nodes.

Table 5.3: Sample Results Obtained from Simple Pipe Code (30 Nodes)

Simulation Time in Seconds Pressure at Node 1 Pressure at Node 15 Pressure at Node31

9.60E-05 1.5 1.5
1.74E-04 1.5 1.5
2.53E-04 1.5 1.5
3.31E-04 1.5 1.5
4.09E-04 1.5 1.5
4.88E-04 1.5 1.5
5.66E-04 1.5 1.5
6.45E-04 1.5 1.5
7.23E-04 1.5 1.5
8.01E-04 1.5 1.5
8.80E-04 1.5 1.5
9.58E-04 1.5 1.5
1.04E-03 1.5 1.5
l.llE-03 1.5 1.5
1.19E-03 1.5 1.5
1.27E-03 1.5 1.5
1.35E-03 1.5 1.492557
1.43E-03 1.5 1.472444
1.51E-03 1.5 1.441361

64
 1.6

t \ ^ ^ Pressure at Mid-point
1.4

V i^y^ Pressure a t Pipe Entry

1.2

0.8
\ i Rressur
e at open

0.6

0.4

0.2

O01 0.02 0.03 0.04 0.05 0.06

Tiine(seconds)

Figure 5.2 Pressure versus Time (30 Nodes)

Next the number of Nodes is increased to 50. The output locations this time are
Node 1, Node 25 and Node 51. The output is now given for 30 locations because there is
no change in the pressure for the first 20 locations because the simulation time is short as
we increased the mesh size. From the pressure distribution, it is seen that there is not
much difference between the 30 nodes distribution and 50 nodes distribution. Increasing
the mesh size does increase the accuracy but the code takes more time and is not
desirable. Increasing the mesh size further will increase the simulation time further but
there will not be any appreciable changes in pressure values obtained.

65
 Table 5.4: Sample Results Obtained from Simple Pipe Code (50 Nodes)

Simulation time in Seconds Pressure at Node 1 Pressure at Node25 Pressure at Node51

5.76E-05 1.5 1.5
1.05E-04 1.5 1.5
1.52E-04 1.5 1.5
1.99E-04 1.5 1.5
2.46E-04 1.5 1.5
2.93E-04 1.5 1.5
3.40E-04 1.5 1.5
3,87E-04 1.5 1.5
4.34E-04 1.5 1.5
4.81E-04 1.5 1.5
5.28E-04 1.5 1.5
5.75E-04 1.5 1.5
6.22E-04 1.5 1.5
6.69E-04 1.5 1.5
7.16E-04 1.5 1.5
7.63E-04 1.5 1.5
8.10E-04 1.5 1.5
8.57E-04 1.5 1.5
9.04E-04 1.5 1.5
9.51E-04 1.5 1.5
9.98E-04 1.5 1.5
1.05E-03 1.5 1.5
1.09E-03 1.5 1.5
1.14E-03 1.5 1.5
1.19E-03 1.5 1.5
1.23E-03 1.5 1.5
1.28E-03 1.5 1.499039
1.33E-03 1.5 1.494862
1.37E-03 1.5 1.48533

The plot given below gives the pressure distribution for the entire simulation time.
From the graph once again the second pulses are seen clearly as was the case with 30
nodes. The pressure obtained is more accurate then obtained from the previous two cases.
The results are in excellent agreement with the values obtained by Benson [1].

66
 1.6

Pressure at
1.4 open end

1.2 N^
Pressure at Open ^u^ ^ ^

2 0.8
«
£
0.
0.6

0.4

0.2

0.01 0.02 0.03 0.04 0.05 0.06

Tiine( seconds)

Figure 5.3 Pressure versus Time (50 Nodes)

A comparative study of pressure at the pipe entry is done. The output obtained
from the above three programs at the pipe entry is plotted on a single sheet. A fair idea is
obtained as when to stop increasing the mesh size. When there is not much change in the
shape of the distribution it is safe to say that increasing the mesh size does not have an
appreciable effect on the accuracy of the output.

67
 1.6
1

- ^
1.4

^V- 10 Nodes

1.2 /
50 Nodes

30Nodes

5 0.8

0.6

0.4

<
0.2

0.01 0.02 0.03 0.04 0.05 0.06

Time(seconds)

Figure 5.4 Pressure at Pipe Entry versus Time

The pressure distribution at the midpoint of the pipe for 10, 30 and 50 nodes is
given in Figure 5.5. From the graph, it is observed that there is difference in the shape of
the graph for 10 nodes and 30 and 50 nodes. The distribution is more accurate for the 30
and 50 nodes and is less accurate for 10 nodes. It can be observed that even the amplitude
for the second and third pulses are different for 10 nodes. In this case the effect of
increasing the mesh size is very apparent. A completely different shape is obtained for 10
nodes and a different shape for 30 nodes and 50 nodes. There is not much difference
between the 30 nodes distribution and the 50 nodes distribution.

68
 1.6

1.4
ION odes

1.2

^ ^
30 Nodes
0.8 50 Nodes

0.6

0.4

0.2

0.01 0.02 0.03 0.04 0.05 0.06

Time(seconds)

Figure 5.5 Pressure at Mid-Point versus Time

The code which gives the velocity of the gas in the pipe in run for the same set of
input conditions. The code for this program is given in appendix. The velocity output is
obtained at Node 1, Node 25, and Node 51. The sample output is given in Table 5.5 and
the velocity distribution is given in Figure 5.6. From the plot, it can be seen that after
about 30 milliseconds the velocity reaches a constant value. The open end has an initial
velocity whereas the velocity at the pipe entry and mid point of the pipe is zero at the
start of the simulation. Once steady state is reached the velocity becomes almost equal at
all the three points.

69
 Table 5.5: Sample Results Obtained from Velocity Code

Velocity at Pipe Velocity at Mid

Simulation Time in Seconds Entry Point Velocity at Pipe End
2.88E-04 0 0 0.05963397
5.23E-04 0 0 0.05963397
7.58E-04 0 0 0.05963397
9.93E-04 0 0 0.05963397
1.23E-03 0 0 0.05963397
1.46E-03 0 0 0.05963397
1.70E-03 0 0.00680083 0.05963397
1.93E-03 0 0.015849531 0.05963397
2.17E-03 0 0.024701059 0.05963397
2.40E-03 0 0.032519102 0.05963397
2.64E-03 0.005323649 0.039090067 0.05963397
2.87E-03 0.014908791 0.044430077 0.05963397
3.11E-03 0.025886893 0.048645079 0.05963397
3.34E-03 0.036548078 0.051878631 0.05963397
3.58E-03 0.046190917 0.05600372 0.05963397
3.81E-03 0.05463621 0.062011838 0.05963397
4.04E-03 0.061670482 0.068628848 0.05963397
4.26E-03 0.067546636 0.074432284 0.05963397
4.48E-03 0.072475374 0.079023391 0.05963397
4.69E-03 0.076620519 0.082586586 0.05963397
4.90E-03 0.08011061 0.085376054 0.062021494
5.11E-03 0.083048075 0.087583005 0.068775535
5.32E-03 0.085516423 0.089346498 0.079101801

70
 300

Time (Seconds)

Figure 5.6 Velocity versus Time

The next code that is developed is for a pipe with a partially open end. This code
is given in appendix. The conditions that are assumed are given in Table 5.6. The
cylinder has a partially open end at one end and is closed at the other end. The output
obtained from the code is plotted and the pressure distribution in given in Figure 5.7.
From the graph it is seen that sudden opening of the partially open end causes a point
centered expansion wave to be generated. This wave is reflected at the closed end and
return back to the open end when a nonual expansion wave is reflected back into the
cylinder. The process continues until the cylinder is empty. The nature of discharge is
cleariy seen which is stepped. The constant pressure regions are the regions of steady
flow.

71
 Table 5.6: Input Condition to a Partially Open End Program

Condition Value
Initial Pressure 3.041
Initial Temperature 300K
Reference Pressure 1.0 bar
Length of the pipe 1.219 meter
Output Locations 1/12L and 11/12L (L is length of pipe)
Mesh Size 50
Ratio of Specific Heats 1.4
Node location of Output 5,45
Simulation time 50 Milliseconds
Opening Area 0.25 sq. meter

Table 5.7: Sample Results Obtained from the Partially Open End Code

Simulation Time in Milliseconds Pressure at location 1/12L Pressure at Location 11/12L

O.OOE-i-00 3.041 3.041
6.00E-05 3.040999 3.040999
1.14E-04 3.040999 3.040999
1.68E-04 3.040999 3.040999
2.22E-04 3.040999 3.040999
2.75E-04 3.040999 2.81284
3.29E-04 3.040999 2.652417
3.83E-04 3.040999 2.562515
4.37E-04 3.040999 2.517969
4.91E-04 3.040999 2.497936
5.45E-04 3.040999 2.489614
5.99E-04 3.040999 2.486372
6.53E-04 3.040999 2.48517
7.06E-04 3.040999 2.484744
7.60E-04 3.040999 2.484597

72
 3-
X

2.5 -
Fjessure at Node 45

•=• 2 -
>sure(ba

Pressure at Node 5
£
a 1 5 -

0 -
O01 0.02 0.03 0.04 0.05 0.06
Time (Seconds)

Figure 5.7 Pressure versus Time for a Partially Open End Boundary (0.25)

73
 The nozzle area ratio is changed to 0.5 and 0.10 and the pressure distribution is
obtained for the same set of input conditions. The Pressure distribution obtained is shown
in Figures 5.8 and Figure 5.9. The distribution obtained from the three nozzle area ratios
is plotted in a single sheet and is given in Figure 5.10. From the plot it is seen that the
discharge is a function of the nozzle area ratio and is independent of the initial conditions.
The larger the nozzle area ratio the larger is the fall in pressure. The results obtained
verify the results obtained by Benson [32].

3.5

^Pressure at Node
5
1 •

2.5

Pressure at N irifi
_^^'-'45

1.5

0.5

0.01 0.02 0.03 0.04 0.05 0.06

Simulation Time (Seconds)

Figure 5.8 Pressure versus Time for a Partially Open End (0.50)

74
3.5

'ressure at Node 5
„ ^ ^

Pressure a Node 45
2.5

1.5

0.5

0.01 0.02 0.03 0.04 0.05 0.06

Simulation Time (Seconds)

Figure 5.9 Pressure versus Time for a Partially Open End (0.10)

75
 0.01 0.02 0.03 0.04 0.05 0.06
Simulation Time (Seconds)

Figure 5.10 Pressure versus Time for Different Nozzle Area Ratios

76
 Next the valve and cylinder sub routines and used to find the pressure in the
cylinder. The code is given in appendix. For the cylinder, the input conditions are given
in Table 5.8.

Table 5.8: Sample Input Conditions to Cylinder Program

Condition Value
Initial Pressure in the Pipe 1.0 bar
Cylinder pressure at E.V.O 3.593 bar
Cylinder Temperature at E.V.O 762K
Reference Pressure 1.0 bar
Pipe Length 0.609 meter
Pipe Diameter 0.0508 meter
Mesh Size 12
Ratio of Specific Heats 1.4
EVO 136
Number of Cycles 4
Engine Speed 33.33 rev/sec
Output locations 1,13
Simulation Time 500 degrees
Cylinder bore 0.125 meter
Cylinder Stroke 0.130
Connecting Rod Length 0.273
Compression Ratio 14.0
Air Pressure 1.65 bar
Air Temperature 360 K
Opening Area 0.5

77
The variation of valve area with crank angle is given in Table 5.9 [3].

Table 5.9: Crank Angle Variation with Valve Area

Exhaust Valve Data Air Valve Data

Crank Angle(Degree) Valve area(sq.mt) Crank Angle(Degree) Valve Area(sq.mt)
0 0 0 0
1 0.00001408 4 0.00001544
9 0.00003658 12 0.00004036
17 0.00008644 20 0.00009555
25 0.00016884 28 0.000186
33 0.00028851 36 0.00031737
41 0.00042153 44 0.00046301
49 0.00054926 52 0.00060249
57 0.00066694 60 0.00073072
65 0.00071524 68 0.00084639
73 0.00086606 76 0.0009471
81 0.00094408 84 0.00103171
89 0.00103277 92 0.0011001
98 0.00106908 100 0.00115018
105 0.001089 108 0.00118184
113 0.00109185 112 0.00118862
117 0.00109185 124 0.00118862
125 0.001089 128 0.00118184
133 0.00106908 136 0.00115018
141 0.00103277 144 0.0011001
149 0.00094408 152 0.00103171
157 0.00086606 160 0.0009471
165 0.00071524 168 0.00084639
173 0.00066694 176 0.00073072
181 0.00054926 184 0.00060249
189 0.00042153 192 0.00046301
197 0.00028851 200 0.00031737
205 0.00016884 208 0.000186
213 0.00008644 216 0.00009555
221 0.00003658 224 0.00004036
229 0.00001408 232 0.00001554
231 0 236 0

78
 The output from the program is the cylinder pressure and the exhaust pipe
pressure at the cylinder end and the exhaust pipe pressure at the partially open end pipe.
The sample output and the pressure distribution are given in Table 5.10 and Figures 5.11,
5.12, and 5.13

Table 5.10: Sample Results Obtained from the Cylinder Code.

Crank Pressure in the Exhaust Pressure (Cylinder Exhaust Pressure(Partially

Angle(deg.) Cylinder End) Open end
136.3212 3.557593 7.81E-03
136.7617 3.54631
137.2021 3.535195
137.6425 3.524247
138.0829 3.513463
138.5233 3.502842
138.9637 3.492382
139.4041 3.48208
139.8445 3.471936
140.2849 3.461948
140.7254 3.452113
141.1658 3.44243
141.6062 3.432898 35.72247
142.0762 3.422883 4.440441 27.27097
142.6149 3.411599 3.785358 22.38544
143.1637 3.400324 3.373725 18.39338
143.7125 3.389271 3.012646 15.34337
144.2613 3.378438 2.716994 13.02845
144.8102 3.367823 2.477725 11.20926

From the graphs, it is seen that the cylinder pressure reaches a maximum value at
around 330 degrees. Cylinder residual waves can be seen in the exhaust pressure
diagram. These residual waves also appear in the partially open end of the exhaust pipe
but are very weak. The residual waves help in the scavenging of the cylinder.

79
Figure 5.11 Variation of Cylinder Pressure with Crank Angle.

80
 1.4 -|

1.32 •

•
1.24 •

\
1 16 •

• •

Khaut Pre isure

1 08 ^ p artially Op en END

1 - lBii^ft% ^^Sm

0.92 -

0.B4
130 170 210 250 290 330 370 410 450 490 530 570 610 650 690
Crank Angle (Degree)

Figure 5.12 Exhaust Pipe Pressure (Partially Open End) versus Crank Angle

81
 1.4
' •

•
1.32

' •

•
1.24

•£ 1.16

Exh aust Pres lure

.Cylinder End
£
a. 1.08 •'

0.92

0.84
130 170 210 250 290 330 370 410 450 490 530 570 610 650 690
CrankAngle{Degree)

Figure 5.13 Exhaust Pressure (Cylinder End) versus Crank Angle

Using the codes developed the pressure distribution of fluid for any type of engine
can be easily simulated. The only input that needs to be derived from experimental stage
is the valve data. Hence by using the codes given in the appendix the fluid flow in
engines can be simulated to a reasonable extent. A lot of assumptions have been made
and the results are only a rough estimate. In the next chapter work required to make the
code better and further improvements to simulate the flow in a much more efficient way
are described.

82
 CHAPTER 6
CONCLUSIONS AND FUTURE WORK

6.1 Conclusions
The following conclusions can be drawn from this thesis
1. The results obtained are in agreement with the results obtained by other
researchers. Similar results were obtained by Benson [1] and Winterbone [27].
The results obtained are only an approximation as the entropy change, friction and
heat transfer have been neglected.
2. A combustion model must be added to the code for predicting the exact cylinder
pressure developed. Without the combustion model the code only gives a rough
estimate of the pressure developed and is not close to actual pressure developed in
an operating cylinder.
3. The wave action phenomenon is clearly seen in the simple pipe, partially open
ended pipe and the exhaust pipes of IC engines.
4. Simulation time will become a question of concern as more complex boundary
conditions are added to the code
5. Higher order methods like the Lax-Wendroff method and TVD schemes should
be used to obtain higher accuracy and to shorten the simulation time. However at
the boundaries the MOC can only be used.

6.2 Future Work Required

In the previous chapter we have seen that the code developed was used for a
single cylinder, single exhaust pipe configuration. The first improvement that the author
suggests is to make the code adaptable to multi-cylinder and multiple-pipe systems. The
organization of the code will be similar to the simple single cylinder code that is already
developed. A loop has to be incorporated to check whether all the cylinders calculations
are complete, if not then the program must go back to the input conditions and accept the

83
mput conditions for the second cylinder. Another important thing to take care is the
junction boundary conditions and pipe branches.
The next big step would be to consider the entropy and cross section changes in
the engine and pipe systems. This would lead to additional complexity and some more
equations. In the code that is developed in this thesis we assumed constant entropy and
cross-section area.
After simulating the flow for varying entropy the code can then be modified to
include complex boundary conditions within pipe systems. Some of the complex
boundary conditions are throttle, compressor, supercharger, turbocharger, sudden
enlargement and contraction of pipes.
If all the above recommendations are implemented, then the code will work
satisfactorily for any type of engine with complex boundary conditions. The next step
would be to deal with issues hke speed and accuracy. To increase the speed and accuracy,
the author suggests the use of conservative methods which are more accurate and faster
than the traditional method of characteristics. The ultimate goal will be to simulate the
entire engine, so one should also look at the combustion modeling at some stage or the
other which is altogether a new topic.

84
 REFERENCES

1. Benson, R.S. (1982) The thermodynamics and gas dynamics of intemal combustion
engines (edited by J.H Horlock and D.E. Winterbone) Vol. 1 Clarendon Press,
Oxford.

2. Eamshaw, S. (1860) On the mathematical theory of sound, Phil. Trans. R. Soc, 150,
133-148.

3. Riemann, B. (1892) Gesammelte Mathematische Werke, Second Edition Teubner,

Leipzig

4. De Haller, P. The application of a graphical method to dynamics problems in gases.

Sulzer Technical Review, Vol. 27, No.l, 1945, pp.6-24.

5. Jenny, E. Unidimensional transient flow with consideration of friction, heat transfer

and change of section. The Brown Boveri Review, Nov.1950, pp.447-461.

6. Benson, R.S., Garg, R.D., and Woollatt, D.1964. A numerical solution to the
unsteady flow problem, Int.J.Mech.Sci., 6, 117-144

7. Benson, R.S. A comprehensive digital computer program to simulate a compressive

ignition engine including intake and exhaust systems. SAE paper 710773, 1971.

8. Low, S.C. and Baruah, P.C. A generalized computer aided design package for I.C
engine manifold system. SAE paper in SP-487 on Engine Control Modeling. Feb 23-
27, 1981, Detroit, Michigan.

9. Blair, G.P., and Gouldbum, J.R., 1967, The pressure time history in the exhaust
system of a high-speed reciprocating intemal combustion engine. SAE paper 670477.

10. Winterbone, D.E, Alexander, G.I., Nichols, J.R., Developments in methods of

considering wave action in pipes connected to I.C engines. Intemational symposium
on flows in intemal combustion engines-111.presented at The Winter Annual
Meeting of The American Society of Mechanical Engineers, Miami Beach, Florida,
November 17-22, 1985.

11. Lakshminarayanan,P.A., Prediction of gas exchange process in a single cylinder

internal combustion engine. SAE paper number 790359, 1979.

12. Ledger, D. Computational methods for unsteady compressible gas flows. Ph.D.
Thesis, April 1974, University of Manchester.

85
13. Pearson, R.J., and Winterbone, D.E., 1990. A rapid wave action simulation technique
for intake manifold design, SAE paper 900676.

14. Lax, P.D., and Wendroff, B., 1960. Systems of Conservation laws. Commun. Pure
Appl. Math., 13, pp. 217-237.

15. Richtymyer, R.D., and Morton, K.W., 1967. Difference Methods for initial value
problems, Wiley, New York.

16. Richtymyer, R.D. 1962. A survey of difference methods for non-steady fluid
dynamics, NCAR Technical Notes 63-2. National Center for atmospheric research,
Colorado

17. MacCormack, R.W. 1969 The effect of viscosity in hypervelocity and impact
cratering, AIAA paper 69-354.

18. MacCormack, R.W. 1971. Numerical solution of the interaction of a shock wave with
a laminar boundary layer. Proceedings of the second intemational conference on
Numerical Methods in fluid dynamics. Lecture notes in physics,8 Springer Verlag,
New York.

19. Bulaty, T. and Niessner, H. Calculation of one-dimensional unsteady flow in pipe

systems of I.C engines. In ASME annual Winter Meeting, New Orleans, Louisiana, 9-
14 December 1984.

20. Poloni, M., Winterbone, D. E. and Nichols, J.R. Calculation of pressure and
temperature discontinuity in a pipe by the method of characteristics and the two-step
differential Lax-Wendroff method. In international symposium on flows in internal
combustion engines, FED 62, ASME Annual Winter Meeting, Boston, 13-18
December 1987.

21. Winterbone, D.E., Pearson, R.J. and Zhao, Y. Numerical Simulation of intake and
exhaust flows in a high- speed multi-cylinder petrol engine using Lax-Wendroff
Method. IMechE Intemational Conference on computers in engine technology,
Cambridge, 10-12 September 1991 (Mechanical Engineering Publications, London).

22. Ferrari, G. and Onorati, A. Determination of silencer performance and radiated noise
spectrum by 1-D gas dynamic modeling. In Proceedings of FISITA Congress,
Beijing, China, 17-22 October 1994.

23. Lin, J., Schom, N. , Schemus, C. and Peng, L. Comparison studies on the method of
characteristics and finite difference methods one one-dimensional gas flows through
IC engine manifolds. SAE paper 960078, 1996.

86
24. Pearson, R.J., and Winterbone, D.E. 1998. The simulation of gas dynamics in engine
manifolds using non-linear symmetric difference schemes. Proc Instn. Mech. Engrs.
Vol 211 PartC.

25. Pearson, R.J., Bassett, M.D., Batten, P., Winterbone D.E. Two-dimensional
simulation of wave propagation in a three-pipe junction. Joumal of Engineering for
Gas Turbines and Power, October 2000, Vol 122/549-555.

26. Zhu, Y, Reitz, R.D. A 1-D gas dynamics code for subsonic and supersonic flows
applied to predict EGR levels in a heavy duty diesel engine. Intemational Joumal of
Vehicle Design, Volume 22 Issue3/4,1999,pp 227-252

27. Pearson, R.J., Winterbone, D.E. 2000. Theory of Engine manifold design-Wave
Action Methods for IC Engines. Professional engineering publication limited.

28. Roe, P.L. 1986. Characteristics -based schemes for Euler equations, Ann. Rev. Fluid
Mech., 18, 337-365.

29. LeVeque, R.J. 1992. Numerical Methods for conservation laws. Lectures in
mathematics ETH, Zurich.

30. Benson,R.S. 1972. Numerical solution of one dimensional non- steady flow with
supersonic and subsonic flows and heat transfer. Int.J.Mech.Sci, 14, 635-642.

31. John H. W. 1990.Intemal Combustion engineering science and technology, Elsevier

Applied Science, New York.

32. Benson, R.S. 1955. Discharge of gas from cylinder to atmosphere. The Engineer, 199,
546-548, 582-584, and 618-620.

87
APPENDIX

88
Given below is the code for a simple pipe with three possible boundary conditions
(1) Inflow into the pipe, (2) Open End, and (3) Closed End.
Program code
INTEGER S,SS,R
REAL GAMMA
DIMENSION CL1(2,51),CL2(2,51 ),N0UT(51 ),Z(2),PRESSUREOUT(2,51)
DIMENSION ZOUT (2)
c Input and Output to a File
OPEN (5, FILE='SAMPLE.TXT')
OPEN (6, FILE='S AMPLEOUT.TXT')
c READ the Initial Conditions
c PI: Initial Pressure
c Tl: Initial Temperature
c PR: Reference Pressure
c LPIPE: length of the Pipe
c M: Number of Mesh Points
c GAMMA: Ratio of Specific Heats,
c NODE: No. of Nodes
c TMAX: Simulaiton Time in Seconds.
READ (5,*) PI
READ (5,*) Tl
READ (5,*) PR
READ (5,*) LPIPE
READ (5,*) M
READ (5,*) GAMMA
READ (5,*) NODE
READ (5,*) (N0UT(N),N=1.N0DE)
READ (5,*) TMAX
ZOUT(1)=0.0
D 0 2 N = 1,N0DE
2 PRESSUREOUT(l,N) = PI
Write (6,*) ZOUT(l),(PRESSUREOUT(l,N),N = l.NODE)
PI =3.14159
Q =(GAMMA-1)/(2.0*GAMMA)
Ql = 1.0/Q
AA = (3.0-GAMMA)/(2.0*(GAMMA-1.0))
BB = (GAMMA+1.0)/(2.0*(GAMMA-1.0))
SS = 1+M
CL = (PI/PR)** Q
TR =T1/(P1/PR)** (2.0*0)
AREF = SQRT (GAMMA * 287.0 * TR)
XMESH = LPIPE/FLOAT(M)
XREF = LPIPE/FLOAT( M)
DELX = XMESH/XREF
DO 1 N= 1 ,SS
CL1(1,N) = CL
1 CL2(1,N) = CL
S= 1
READ(5,*) PO
AO = (PO/PR)**Q
S = SS

89
 R=l
Z(1) = 0.0
14 D 0 15S = 1,SS
DELZl=DELX/((CLl(R,S)+CL2(R,S))/2.0
4 +ABS(CL1 (R,S)-CL2(R,S))/(GAMMA-1))
IF (S.EQ. 1) DELZ = DELZl
IF(S.EQ.1)G0T0 15
IF (DELZl LT.DELZ) DELZ = DELZl
C Calculations for Next Time Step
15 CONTINUE
Z(R+1) = Z(R)+DELZ
DO 11 S=2,SS
11 CL1(R+1,S)=CL1(R,S)+(BB*CL1(R,S-1)-AA*CL2(R,S-1))
8*(CL1(R,S-1)-CL1(R,S))*DELZ/DELX
DO 12S = 1,M
12 CL2(R+1,S)=CL2(R.S)+(BB*CL2(R,S+1)-AA*CL1(R,S+1))
7 * (CL2(R,S+1)- CL2(R,S))*DELZ/DELX
S= 1
QQ = CL2(R+1,S)
c BOUNDARY CONDITIONS
CALL INFLOW(PP,QQ,A0,G)
CL1(R+1,S) = PP
S = SS
QQ = CL1(R+I,S)
CALL OPENEND(PP,QQ)
CL2(R+1,S) = PP
D0 29N = l,NODE
S = NOUT(N)
29 PRESSUREOUT(R+l,N)= (((CL1(R+1,S)+CL2(R+1,S))/2.0)**Q1)*PR
Z0UT(R+1)= Z(R+1)*XREF/AREF
WRITE (6,*) ZOUT(R+l),(PRESSUREOUT(R+I,N),N = l.NODE)
IF (ZOUT(R+l) GT.TMAX)GO TO 30
Z(R) = Z(R+1)
DO 31 S = 1,SS
CL1(R,S) = CL1(R+1,S)
31 CL2(R,S) = CL2(R+1,S)
GO TO 14
30 STOP
END

SUBROUTINE CLOSEDEND(PP,QQ)
PP =QQ
RETURN
END

SUBROUTINE OPENEND(PP,QQ)
PP = 2.0 - QQ
RETURN
END

SUBROUTINE INFLOW(PP,QQ,A0,G)
G = 1.4
A = (3.0-G)/(G+1.0)
B = 40*(G-1.0)/(G+1.0)
C = ((1.4+1.0)/(3.0-G))*QQ
IF (QQ.GE. AO) GO TO 1
PP = A * QQ+ SQRT(B*A0**2-(1.0-A**2)*QQ**2)
IF (PP.GT.C) PP= C
GO TO 2

90
 PP= 2.0*A0-QQ
RETURN
END

2) Given below is the code for calculating the velocity of the fluid at any point in a
simple pipe with three possible boundary conditions
1) Inflow into the pipe 2) Open End 3) Closed End.
Program code
INTEGER S,SS,R
REAL GAMMA
DIMENSION CLl(2,51 ),CL2(2,51 ),N0UT(51 ),Z(2)
DIMENSION ZOUT(2),VELOCITYOUT(2,51)
c Input and Output to a File
OPEN(5,FILE=S AMPLE.TXT')
0PEN(6, F1LE=:SAMPLEOUT.TXT)
c READ the Initial Conditions
c PI: Initial Pressure
c Tl: Initial Temperature
c PR: Reference Pressure
c LPIPE: length of the Pipe
c M: Number of Mesh Points
c GAMMA: GAMMA
c NODE: No. of Nodes
c TMAX: Simulaiton Time in Seconds.
READ (5,*) PI
READ (5,*) Tl
READ (5,*) PR
READ (5,*) LPIPE
READ (5,*) M
READ (5,*) GAMMA
READ (5,*) NODE
READ (5,*) (NOUT(N),N=l,NODE)
READ (5,*) TMAX
ZOUT(1)=0.0
D02N=1,N0DE
2
Write (6,*) ZOUT(l),(VELOCITYOUT(l,N),N = l.NODE)
PI =3.14159
Q =(GAMMA-1)/(2.0*GAMMA)
Ql = 1.0/Q
AA = (3.0-GAMMA)/(2.0*(GAMMA-1.0))
BB = (GAMMA+1.0)/(2.0*(GAMMA-1.0))
SS = 1+M
CL = (PI/PR)** Q
TR =T1/(P1/PR)**(2.0*Q)
AREF = SQRT (GAMMA * 287.0 * TR)
XMESH = LPIPE/FLOAT(M)
XREF = LPIPE/FLOAT( M)
DELX = XMESH/XREF
DO 1 N= 1 ,SS
CL1(1,N) = CL
1 CL2(1,N) = CL
S=l
READ(5,*) PO
AO = (PO/PR)**Q
S = SS

91
 R=l
Z(1) = 0.0
14 D 0 15S=1,SS
DELZl=DELX/((CLl(R,S)+CL2(R,S))/2.0
4+ABS(CLl(R,S)-CL2(R,S))/(GAMMA-l))
IF (S.EQ. 1) DELZ = DELZl
IF (S.EQ, 1) GO TO 15
IF (DELZl.LT.DELZ) DELZ = DELZl
C Calculations for Next Time Step
15 CONTINUE
Z(R+1) = Z(R)+DELZ
DO 11 S = 2,SS
11 CL1(R+1,S)=CL1(R,S)+(BB*CL1(R,S-1)-AA*CL2(R,S-1))
8*(CL1(R,S-1)-CL1(R,S))*DELZ/DELX
DO 12S = 1,M
12 CL2(R+1,S)=CL2(R,S)+(BB*CL2(R,S+1)-AA*CL1(R,S+1))
7 * (CL2(R,S+1)- CL2(R,S))*DELZ/DELX
S= 1
QQ = CL2(R+1,S)
c BOUNDARY CONDITIONS
CALL INFLOW(PP,QQ,A0,G)
CL1(R+1,S) = PP
S = SS
QQ = CL1(R+1,S)
CALL OPENEND(PP,QQ)
CL2(R+1,S) = PP
DO 28 N = 1,N0DE
S = NOUT(N)

28 VELOCITYOUT(R+l,N) = ((CLl(R+1,S)-CL2(R+1,S))/(K-1.0))*AREF

ZOUT(R+l)= Z(R+1)*XREF/AREF

WRITE (6,*) ZOUT(R+l),(VELOCITYOUT(R+l,N),N = l,NOD)

IF (ZOUT(R+l).GT.TMAX)GO TO 30
Z(R) = Z(R+1)
DO 31 S = LSS
CL1(R,S) = CL1(R+1,S)
31 CL2(R,S) = CL2(R+1,S)
GO TO 14
30 STOP
END

SUBROUTINE CLOSEDEND(PP,QQ)
PP =QQ
RETURN
END

SUBROUTINE OPENEND(PP,QQ)
PP = 2.0 - QQ
RETURN
END

SUBROUTINE INFLOW(PP,QQ,A0,G)
G = 1.4
A = (3.0-G)/(G+1.0)
B = 4.0*(G-1.0)/(G+1.0)
C = ((1.4+1.0)/(3.0-G))*QQ

92
 IF (QQ.GE. AO) GO TO 1
PP = A * QQ+ SQRT(B*A0**2-(1.0-A**2)*QQ**2)
IF (PP.GT.C) PP= C
GO TO 2
1 PP= 2.0*A0-QQ
2 RETURN
END

3) Program to calculate the pressure in a pipe with a partially open end boundary
condition

Program PARTIAL
INTEGER S,SS,R,T1,PR,LPIPE,MESH,NODE
REALGAMMA.Pl
DIMENSION CLl(2,51),CL2(2,51 ),N0UT(51 ),Z(2),PRESSUREOUT(2,51 ),ZOUT(2)
0PEN(5,F1LE=SAMPLE.TXT')
OPEN(6, FILE=SAMPLEOUT.TXT)
READ (5,*) PI
READ(5,*)T1
READ (5,*) PR
READ(5,*) LPIPE
READ(5,*) MESH
READ(5,*)GAMMA
READ (5,*) NODE
READ (5,*) (NOUT(N),N=l,NODE)
READ (5,*) ZMAX
ZOUT(1)=0.0
D 0 2 N = LNODE
2 PRESSUREOUT(l,N) = Pl
Write (6,*) ZOUT(l),(PRESSUREOUT(l,N),N = LNODE)
PI =3.14159
Q = (GAMMA-1)/(2.0*GAMMA)
Ql = 1.0/Q
AA = (3.0-GAMMA)/(2.0*(GAMMA-1.0))
BB = (GAMMA+1.0)/(2.0*(GAMMA-1.0))
M = MESH
SS = 1+M
CL= (PI/PR)** Q
TR = T1/(P1/PR)** (2.0*Q)
AR = SQRT (GAMMA * 287.0 * TR)
XMESH = LPIPE/FLOAT(M)
XREF = LPIPE/FLOAT( M)
DELX = XMESH/XREF
DO 1 N= 1 ,SS
CL1(1,N) = CL
1 CL2(1,N) = CL
S=l
c INPUT THE NOZZLE AREA RATIO
READ(5,*) PHI
AO = (PO/PR)**Q
S = SS
R=l
Z(1) = 0.0
14 D 0 15S=1,SS
DELZl=DELX/((CLl(R,S)+CL2(R.S))/2.0+
4 ABS(CL1 (R,S)-CL2(R,S))/(GAMMA-1))
IF (S.EQ. 1) DELZ = DELZl
1F(S.EQ.1)G0T0 15

93
 IF (DELZl.LT.DELZ) DELZ = DELZl
15 CONTINUE
Z(R+1) = Z(R)+DELZ
DO 11 S = 2,SS
11 CL1(R+1,S)=CL1(R,S)+(BB*CL1(R,S-1)-AA*CL2(R,S-1))
8*(CL1(R,S-1)-CL1(R,S))*DELZ/DELX
DO 12S = 1,M
12 CL2(R+1,S)=CL2(R,S)+(BB*CL2(R,S+1)-AA*CL1(R,S+1))
7 * (CL2(R,S+1)- CL2(R,S))*DELZ/DELX
S= 1
QQ = CL2(R+1,S)
CALL CLOSEDEND(PP,QQ)
CL1(R+1,S) = PP
S = SS
QQ = CL1(R+1,S)
CALL POE(PP,QQ,PHI,GAMMA)
CL2(R+1,S) = PP
D 0 2 9 N = l.NODE
S = NOUT(N)
29 PRESSURE0UT(R+1,N)= (((CL1(R+1,S)+CL2(R+1,S))/2.0)**Q1)*PR
Z0UT(R+1) = Z(R+1)*XREF/AR
WRITE (6,*) ZOUT(R+l),(PRESSUREOUT(R+l,N).N = l.NODE)
IF (ZOUT(R+l).GT.ZMAX)GO TO 30
Z(R) = Z(R+1)
D 0 31 S = l.SS
CL1(R,S) = CL1(R+1,S)
31 CL2(R,S) = CL2(R+1,S)
GOTO 14
30 STOP
END
SUBROUTINE CLOSEDEND(PP,QQ)
PP =QQ
RETURN
END
SUBROUTINE OPEEND(PP,QQ)
PP = 2.0 - QQ
RETURN
END
SUBROUTINE INFLOW(PP,QQ,A0)
A = (3.0-1.4)/(1.4+1.0)
B = 4 0 * (1.4-1.0)/(1.4+1.0)
C = ((1.4+1.0)/(3.0-1.4))*QQ
IF (QQ.GE. AO) GO TO 1
PP = A * QQ + SQRT(B*A0**2-(1.0-A**2)*QQ**2)
IF (PP.GT.C) PP = C
GO TO 2
1 PP = 2.0*A0-QQ
2 RETURN
END
SUBROUTINE POE(CLOUT,CLIN.PHIN.G)
C PHIN = NOZZLE AREA RATIO
DIMENSION A(6)
C CALCULATE CONSTANTS
A(1) = (G+1.0)/(G-1.0)
A(2) = 2.0/(G-1.0)
A(3) = 2.0*A(2)
A(4) = 2.0/(G+1.0)
A(5) = 1.0/A(2)
IF(CLIN.GT.1.0)GOTO10

94
 CLOUT = 2.0 - CLIN
GO TO 5
C INITIAL ESTIMATE TO SOLVE THE EQUATION USING NUMERICAL METHOD
10 AN = 0.5*(CLIN+1.0)
C NEXT ESTIMATE
DAN = 0.25*(CLIN-1.0)
C EQUATION FOR SONIC FLOW
1 FA = ((CLIN-AN)**2)*(AN**A(3)-PHIN**2)
5-(A(5)*PHIN**2)*(AN**2-1.0)
IF(ABS(FA).LT.0.0001) GO TO 2
IF(FA.LT.0.0)GOTO3
AN = AN +DAN
GO TO 4
3 AN = AN -DAN
4 DAN = 0.5 *DAN
IF(DAN .GT.0.0001) GO TO 1
2 CLOUT = 2.0 * AN - CLIN
U = (CLIN-CLOUT)/(G-1.0)
V = AN**A(2)*U/PHIN
IF( V.LE.1.0)GOTO5
AT = 0.5 +0.5 *SQRT(A(4))
DAT = 0.25 -0.25*SQRT(A(4))
6 X = 1.0/AT
FA = PHIN**2-(Y**A(3))*(A(1)-A(2)*Y**2)
IF(ABS(FA).LT.0.0001) GO TO 7
IF(FA.LT.0.0)GOTO8
AT = AT + DAT
GO TO 9
8 AT = AT -DAT
9 DAT = 0.5 * DAT
IF ( DAT.GT.0.0001) GO TO 6
7 A(6) = AT**A(1)
CON = A(5) *PHIN*A(6)
SKS = (1.0-CON)/(1.0+CON)
CLOUT = SKS*CLIN
5 RETURN
END

4) Program to calculate the pressure in the cylinder and in the exhaust pipe
PROGRAM CYLINDER
INTEGER R.S.SS
REAL GAMMA.MC
DIMENSION CLl(2.51),CL2(2,51 ),N0UT(51 ).Z(2),POUT(2,51 ).ZOUT(2)
COMMON AAIR.AREF,AVO,AZA(50).N1,CR.CYCLE1,DELZ
COMMON DPCDT,EVO,FAIR(50),ACYL,FEXH(50),FP,LAIR.1REV
COMMON IVOL,GAMMA,MC.NA.PA,PC,PI.PR.REV,RPM.STROKE.TA.THETA
COMMON TRAD,VC,XREF,ZCALC.ZREV,NE.AZE(50)
! DATA INPUT
OPEN(5.FILE =S AMPLE.TXT')
0PEN(6, FILE=SAMPLEOUT.TXT')
READ(5,*) P1.PC,TC,PR,LPIPE,DP
READ(5,*)MESH,GAMMA.EV0,AV0,CYCLE1,RPM
THETA = EVO
ZOUT(l) = THETA
ZCALC = 0.0
ZREV =0.0
REV = 1.0

95
 DATA OUTPUT
NO. OF LOCATIONS (NODE)
LOCATION(NOUT(N))
LENGTH OF RUN IN DEGREE CRANK ANGLE
READ(5.*) NODE
READ (5,*)(N0UT(N), N= l.NODE)
READ (5,*) ZMAX
! PRINT OUT THE INITIAL VALUES AT OUTPUT LOCATIONS
D 0 2N= l.NODE
2 P0UT(1.N)= PI
c PRINT out Initial Values
WRITE(6.*) ZOUT(l),PC.(POUT(l.N). N = 1 .NODE)
PI = 3.14159
! CALCULATE CONSTANTS
Q = (GAMMA-1)/(2.0*GAMMA)
Ql = 1.0/Q
AA = (3.0-GAMMA)/(2.0*(GAMMA-1.0))
BB = (GAMMA+1.0)/(2.0*(GAMMA-1.0))
M = MESH
SS = 1+M
CL = (PI/PR)** Q
! SET UP REFRENCE CONDITIONS
TREE = TC*(PR/PC)**(2.0*Q)
TRI = TREE
AREF = SQRT(GAMMA*287.0*TREF)
XMESH = LPIPE/FLOAT(M)
XREF = LPIPE/FLOAT(M)
! SETUPDMITIALCONDITIONS ATZ = 0
DELX = XMESH/XREF
DO 1 N= l.SS
CL1(1.N) = CL
1 CL2(1.N) = CL
! READ BOUNDARY CONDITONS
S=l
BOUNDARY CONDITION AT S = 1 , CYLINDER
READ DATA
Pv'OL = 1 CONSTANT VOLUME
LMR = 0 NO AIR INLET TO CYLINDER
IVOL=2 VARL\BLE VOLUME
lAIR = 1 AIR INLET TO CYLINDER
READ(5.*) rVOL. lAIR
GOTO(41.42),IVOL
41 READ(5.*) VC.PC.TC
AC = SQRT(GAMMA*287.0*TC)
GO TO 43
42 READ (5,*) DLA.STROKE.LROD.CR
Nl = 2.0*LROD/STROKE
TRAD =THETA*PI/180.0

FNN =SQRT(N1**2-SIN(TRAD)**2)

X = 0.5 *STROKE*( 1.0+N1 -FNN-COS(TRAD))

ACYL =0.25*PI*DIA*DIA
VC = ACYL * (X + STROKE/(CR-1.0))
AC = SQRT(GAMMA*287.0*TC)
MC =1.0E5*PC*VC/(287.0*TC)
DXDT = 0.5*STROKE*SIN(TRAD)*(1.0+COS(TRAD)/FNN)
DTHDT = 2.0*PI *RPM
DVCDT = ACYL*DXDT*DTHDT

96
 DPCDT = -GAMMA*DVCDT*PC/VC
IF(lAIR.EQ.0)GOTO43
READ(5.*) PA.TA
AAIR = SQRT(GAMMA*287.0*TA)
43 READ(5,*) NE
READ (5.*) (AZE(I),FEXH(I). I = l.NE)
FP = 0.25*PI*DP*DP
DO 44 1 = 1 . NE
44 FEXH(I) =FEXH(I)/FP
IF(IAIR.EQ.0)GOTO45
READ(5.*) NA
READ(5.*) (AZA(I).FAIR(I), 1 = l.NA)
45 S = SS
READ(5.*) PHI
R =1
Z(1) = 0.0
14 DO 15 S = l.SS
DELZl=DELX/((CLl(R.S)+CL2(R.S))/2.0
5+ABS(CLl(R,S)-CL2(R.S))/(GAMMA-l))
IF (S.EQ. 1) DELZ = DELZl
IF(S.EQ.1)G0T0 15
IF (DELZl.LT.DELZ) DELZ = DELZl
15 CONTINUE
Z(R+1) = Z(R)+DELZ
DO 11 S = 2,SS
11 CL1(R+1,S)=CL1(R.S)+(BB*CL1(R.S-1)-AA*CL2(R.S-1))
M*(CL1(R.S-1)-CL1(R.S))*DELZ/DELX
D 0 12S = 1.M
12 CL2(R+1,S)=CL2(R.S)+(BB*CL2(R.S+1)-AA*CL1(R.S+1))
Z* (CL2(R,S+1)- CL2(R.S))*DELZ/DELX
S=l
QQ = CL2(R+1.S)
c Boundary Conditions

CALL CYLNDR(PP.QQ)
CL1(R+1.S) = PP
S = SS
QQ = CLl (R+1.S)
CALL POE(PP,QQ.PHI.GAMMA)
CL2(R+1,S)= QQ
DO 29 N = l.NODE
S = NOUT(N)
29 P0UT(R+1.N) = (((CL1(R+1.S)+CL2(R+1.S))/2.0)**Q1)*PR
Z0UT(R+1) = THETA
WRITE(6.*) ZOUT(R+l).PC.(POUT(R+l.N).N= l.NODE)
IF (ZCALC.GT.ZMAX) GO TO 30
Z(R) =Z(R+1)
DO 31 S = l.SS
CL1(R,S) = CL1(R+1.S)
31 CL2(R.S) = CL2(R+1.S)
GO TO 14
30 STOP
END PROGRAM CYLINDER

SUBROUTINE POE(CLOUT.CLIN.PHIN.G)
C PHIN = NOZZLE AREA RATIO
DIMENSION A(6)
C CALCULATE CONSTANTS
A(1) = (G+1.0)/(G-1.0)

97
 A(2) = 2.0/(G-1.0)
A(3) = 2.0* A(2)
A(4) = 2.0/(G+1.0)
A(5)= 1.0/A(2)
IF (CLIN.GT. 1.0) GO TO 10
CLOUT = 2.0 - CLIN
GO TO 5
C INITIAL ESTIMATE TO SOLVE THE EQUATION USING NUMERICAL METHOD
10 AN = 0.5*(CL1N+1.0)
C NEXT ESTIMATE
DAN = 0.25*(CLIN-1.0)
C EQUATION FOR SONIC FLOW
1 FA = ((CLIN-AN)**2)*(AN**A(3)-PHIN**2)
5-(A(5)*PHIN**2)*(AN**2-1.0)
IF(ABS(FA).LT.0.0001) GO TO 2
IF (FA.LT.0.0) GO TO 3
AN = AN +DAN
GO TO 4
3 AN = AN -DAN
4 DAN = 0.5 *DAN
IF(DAN .GT.0.0001) GO TO 1
2 CLOUT = 2.0 * AN - CLIN
U = (CLIN-CLOUT)/(G-1.0)
V = AN**A(2)*U/PHIN
IF( V.LE.1.0)GOTO5
AT = 0.5 +0.5 *SQRT(A(4))
DAT = 0.25 -0.25*SQRT(A(4))
6 X=1.0/AT
FA = PHIN**2-(Y**A(3))*(A(1)-A(2)*Y**2)
IF(ABS(FA).LT.0.0001)GOTO7
IF (FA.LT.0.0) GO TO 8
AT = AT + DAT
GO TO 9
8 AT = AT -DAT
9 DAT = 0.5 * DAT
IF ( DAT.GT.0.0001) GO TO 6
7 A(6) = AT**A(1)
CON = A(5) *PHIN*A(6)
SKS = (1.0-CON)/(1.0+CON)
CLOUT = SKS*CLIN
5 RETURN
END

SUBROUTINE CYLNDR(CLOUT.CLIN)
REAL GAMMA.MC
DIMENSION TH(2).VCC(2)
COMMON AAIR.AREF.AVO.AZA(50),N 1 .CR.CYCLE 1 .DELZ
COMMON DPCDT,EVO.FAIR(50).ACYL,FEXH(50),FP,L^IR.IREV
COMMON IVOL.GAMMA.MC.NA.PA,PC,PI,PR,REV.RPM,STROKE.TA.THETA
COMMON TRAD,VC.XREF,ZCALC.ZREV.NE,AZE(50),A
DT = DELZ*XREF/AREF
DTHR = 2.0*PI*RPM*DT
TRAD= TRAD+DTHR
THETA = TRAD*180.0/PI
IF(THETA.GT. 180*CYCLE1) THETA = THETA-180.0*CYCLE1
ZCALC =ZCALC+DTHR*180.0/PI
ZREV = ZREV +DTHR * 180.0/PI
IREV =0.0
IF (ZREV.LE. 180.0 * CYCLE1*REV) GO TO 5

98
 IREV = 1
REV = REV+1.0
5 ALPHAE = THETA - EVO
IF (ALPHAE.LT.0.0) ALPHAE = 0.0
1=1
6 1 = 1+1
IF(ALPHAE.GT.AZE(I)) GO TO 6
PSI = FEXH(1-1)+(FEXH(I)-FEXH(M))*(ALPHAE-AZE(1-1))/
3(AZE(I)-AZE(I-1))
7 IF(IAIR.EQ.0)GOTO4
ALPHAA = THETA-AVO
IF(ALPHAA.LT.O.O) ALPHAA= 0.0
1= 1
8 1=1+1
EF (ALPHAA.GT.AZA(I)) GO TO 8
F2 = FAIR(1-1)+(FAIR(I)-FAIR(I-1))*(ALPHAA-AZA(1-1))/
C(AZA(I)-AZA(1-1))
IF (F2.LE.0.0) DMADT = 0.0
4 PC = PC+DPCDT*DT
IF(lREV.NE.l)GOTO10
TH(1) = EVO* PI/180
TH(2) = (AVO+AZA(NA-1))*PI/180
DO 11 N = 1.2
FNN = SQRT(N1**2-SIN(TH(N))**2)
X = 0.5* STROKE*(1.0+N1-FNN-COS(TH(N)))
11 VCC(N) = ACYL*(X+STROKE/(CR-1.0))
TC = PC * VCC(l) *TC /(PC*VCC(2))
PC = PC
IREV = 0
10 CALL VALVE(CLOUT.DMEDT.CLIN.PSI)
IF(IAIR.EQ.O) DMADT = 0.0
IF(IAIR.EQ.0)GOTO3
IF(F2.LE.0.0)GOTO3
CALLAIRFL(DMADT,PA.PC.AAIR,AC.GAMMA.F2)
3 MC = MC +(DMADT-DMEDT)*DT
GOTO(1.2).rVOL
1 DVCDT = 0.0
GO TO 9
2 FNN = SQRT(N1**2-SIN(TRAD)**2)
X = 0.5 *STROKE*(1.0+N1-FNN-COS(TRAD))
VC = ACYL*(X+STROKE/(CR-1.0))
DXDTH = 0.5 * STROKE*SIN(TRAD)*(1.0 +COS(TRAD)/FNN)
DTHDT = 2.0 * PI* RPM
DVCDT = ACYL*DXDTH*DTHDT
9 TC = 1.0E5 *PC * VC /(287.0*MC)
AC = SQRT(GAMMA*287*TC)
IF(DMADT.LT.O.O) A = AC
IF(DMADT .GT.0.0) A = AAIR
IF(DMADT.EQ.O.O) A = 0.0
DPCDT=-GAMMA*DVCDT*PC/VC-DMEDT*AC**2/(VC*1.0E5)+DMADT*A**2/(VC*1.0E5)
RETURN
END
SUBROUTINE VALVE(CLOUT.DMEDT.CLIN.PSI)
REAL GAMMA.Kl .K2.K3.MC,N,P1A
COMMON AAIR.AREF.AVO.AZA(50),AZE(50).N1,CR,CYCLE1.DELZ.DPCDT
COMMON EVO.FAIR(50).ACYL.FEXH(50).FP,IA1R,IREV
COMMONIVOL.GAMMA.MC,NA.NE.PA,PC,PI.REV,RPM,STROKE,TA.P
COMMON THETA.TRAD.VC.XREF.ZCALC.ZREV
PP = CLIN

99
 QQ = CLOUT
IF (PSLGT.0.0) GO TO 1
50 PP = QQ
DMEDT = 0.0
RETURN
1 IF(PSLGT.1.0)PSI= 1.0
GAMMA = 1.4
Kl = (GAMMA-1.0)/(2.0*GAMMA)
K2 = 2.0/(GAMMA-1.0)
K3 = 1.0/K1
S = SQRT(2.0/(GAMMA+1.0))
PIA = CLIN*(PR/PC)**K1
IF( PIA EQ. 1.0) GO TO 50
C = 1.0/PIA
D = PL^
4 IF(PlA.LT.1.0)GOTO2
CLIN = PIA
CALL POE(CLOUT.CLIN.PHI.GAMMA)
CLIN = PP *(PC7PR)**K1
CLIN = PIA * (PC/PR)**K1
GO TO 11
2 U = S*(SQRT(1.0+(GAMMA**2-1.0)*PSI*PSI)-1.0)/(PS1*(GAMMA-1.0))
C = SQRT(1.0-U**2/K2)/S
PICR = S - U/(K2*C)
C = 1.0/PIA
IF(PIA.LE.PICR) GO TO 3
XPI = 0.5*(1.0+S)
DXPI = 0.25*(1.0-S)
30 B = (XPI-PIA)*C
FNPl = (PSI/XPI)*SQRT(K2*(1.0/XPI**2-1.0))-(B*K2/(1.0-B*B*K2))
C = SQRT(1.0 - K2*B*B)/XPI
D = 1.0/C
IF(ABS(FNPD.LT.0.00001)GOTO 8
IF (DXPl.LT.0.000001) GO TO 8
IF(FNPI.GT.O.O) XPI = XPI +DXPI
IF(FNPI.LT.O.O) XPI = XPI -DXPI
DXPI = 0.5*DXPI
GO TO 30
3 XPI = 0.5 *(PLA+S)
DXPI = 0.25 *(S-PIA)
E = PSI*(2/(GAMMA+1.0))**((GAMMA+1.0)/(2.0*(GAMMA-1.0)))
40 B = (XPI - PIA)*C
FNPI = XPI**K3-E*((1.0-K2*B*B)/(K2*B))
C = SQRT(1.0-K2*B*B)/XPI
D = 1.0/C
IF(ABS(FNPD. LT.0.00001) GO TO 8
IF (DXPI.LT.0.000001) GO TO 8
IF (FNPI.GT.0.0) XPI = XPI-DXPI
IF(FNPLLT.O.O)XPI = XPI+DXPI
DXPI = 0.5 *DXPI
GO TO 40
8 CLOUT = 2.0* XPI*(PC/PR)**K1- CLIN
11 A = 0.5 *(CLOUT+CLIN)
U = (CLIN-CLOUT)/(GAMMA-1.0)
RP = A**(2.0*GAMMA/GAMMA-1.0)
DMEDT = -GAMMA*(PR/AREF)*FP*(U/A**2)*RP*1.0E5
RETURN
END
SUBROUTINE AIRFL(DMADT.PA,PC,AA1R,ACYL.G,F2)

100
 RM = PC/PA
IF(RM.GT.1.0)RM= 1.0/RM
IF(RM.EQ.l.O) DMADT =0.0
1F(RM.EQ.1.0)GOT0 1
D = G/(G-1.0)
RCR = (2.0/(G+1.0))**D
IF(RM.GT.RCR) GO TO 2
B = (G+1.0)/(2.0*(G-1.0))
FRM = G*(2.0/(G+1.0))**B
GO TO 3
2 R = 2.0/G
C = (G-1.0)/G
D = (2.0*G**2)/(G-1.0)
FRM = SQRT(D*(RM**R)*(1.0-RM**C))
3 IF(PC.GT.PA)GOT0 4
DMADT = PA*F2*FRM*1.0E5/AA1R
GOTO 1
4 DMADT = -PC*F2*FRM*1.0E5/ACYL
1 RETURN
END

101
 PERMISSION TO COPY

In presenting this thesis in partial fiilfilhnent of the requirements for a master's

degree at Texas Tech University or Texas Tech University Health Sciences Center, I

agree that the Library and my major department shall make it freely available for

research purposes. Permission to copy this thesis for scholarly purposes may be

granted by the Director of the Library or my major professor. It is understood that any

copying or publication of this thesis for financial gain shall not be allowed without my

further written permission and that any user may be liable for copyright infringement.

Agree (Permission is granted.)

-I r
Student Signature Date

Disagree (Permission is not granted.)

Student Signature Date