University of Pretoria etd – Makgata, K W (2005)

COMPUTATIONAL ANALYSIS AND

OPTIMISATION OF THE INLET SYSTEM OF A

HIGH-PERFORMANCE RALLY ENGINE

BY
KATLEGO WEBSTER MAKGATA

Submitted in partial fulfilment of the requirements for the degree of

Master of Engineering (Mechanical Engineering)
In the
Faculty of Engineering, Built Environment and Information Technology
University of Pretoria
May 2005
 University of Pretoria etd – Makgata, K W (2005)
ii

SUMMARY
Computational analysis and optimisation of the inlet system of a high-
performance rally engine

Katlego Webster Makgata

Study leaders: Prof. K.J. Craig; Mr. D. J. de Kock

Department: Department of Aeronautical and Mechanical Engineering,
University of Pretoria
Degree: Master of Engineering

In the car-racing world, whether it is motor cross, formula 1 or rally racing, there is a
need for maximum power through the race. While this is not possible through the
entire engine speed range, it is possible to manipulate the engine speed at which
maximum power is obtained by changing the engine configuration. Where restrictions
prevent engine reconfiguration to move the maximum power engine speed, there is
the option of increasing the engine power at the set engine speed. One of the most
effective ways to do this is to modify the intake system to allow for more air into the
engine, thus allowing for more fuel to be burnt and more power to be obtained. While
fuel economy suffers, this is sacrificed in the pursuit for maximum power. This
dissertation focused on improving the inlet system of a high-performance rally car
race engine by using computational fluid dynamics (CFD) and mathematical
optimisation techniques, the combination of which is called a computational flow
optimisation (CFO) system.

The history and background of intake systems is presented in the literature survey
section followed by a discussion on variable intake manifolds. Variable intake
manifolds are extensively used in modern production sports cars because of their
advantage of improving smooth power delivery through the entire engine speed range.
The theory of waves and their interactions is also given along with the theory of
engine tuning with regards to intake systems. Tuning takes advantage of the energy of
flow pulses to increase the amount of charge that is inducted into the engine. The
 University of Pretoria etd – Makgata, K W (2005)
iii

numerical methods used to solve for flow in intake systems are also presented as they
form the basis of the engine simulation code used in the study, namely EngMod4T.

The governing equations as well as other important aspects, such as grid generation,
boundary conditions and convergence, which relate to CFD are discussed. The
literature survey is concluded by a discussion on the mathematical optimisation
algorithm employed in this study.

The literature survey is followed by the presentation of the CFD work done in
determining the amount of air inducted by the intakes. The simulations are transient in
nature in order to capture the pulse interactions and their influence on the mass of air
inducted by the intakes. The first case considered is that of a single intake exposed to
atmosphere. To relate the results of the single intake simulation to a full-intake
simulation, the mass of inducted air is assumed to be equal for all four intakes. This
assumption was found to be flawed as shown by the simulation that followed that took
into consideration all four intakes also open to atmosphere. The simulation showed
that the intakes actually induct differing amounts of air and the total amount is less
than for four single trumpets. A more comprehensive simulation was conducted where
the airbox was included and the resulting total mass inducted showed that even less
air is inducted by this setup. The results of the latter were used as the basis of the
optimisation study that followed. Various airbox designs, obtained from the
optimisation software LS-OPT, were simulated and resulted in an improved airbox
design that inducts 6.2% more air than the original airbox. And since there is direct
relationship between mass of air inducted and engine power produced, it is expected
that the engine power would also increase by 6.2%.

The study demonstrates the successful implementation of a CFO system to solve a

complex industrial flow problem. With the increase of computing power and
increasing affordability of such systems coupled with the ease-of-use of commercial
CFD software, CFO should become more common in industrial application.

Keywords: airbox, computational fluid dynamics, computational flow optimisation,

inlet system, simulation, transient, mathematical optimisation.
 University of Pretoria etd – Makgata, K W (2005)
iv

ACKNOWLEDGEMENTS

My sincere thanks go to the following that contributed in their own way:

Professor Ken Craig and Danie de Kock. My family and friends who showed me
support throughout my study period. Thanks to Toyota Motorsport for donating their
facilities, time and for partially sponsoring this study. I would like to extend my
sincere thanks to the CSIR for also sponsoring the study and for their unbelievable
patience. And a special thank you to my mother for all her prayers.

In memory of my father, Philemon Bolayang Makgata, and my aunt, Dr. Princess

Tlou Mmolawa.
 University of Pretoria etd – Makgata, K W (2005)
v

TABLE OF CONTENTS

Abbreviations and Acronyms .............................................................. viii

List of Figures.......................................................................................... ix

List of Tables .......................................................................................... xii

CHAPTER 1 : Introduction.................................................................1

1.1 MOTIVATION ..............................................................................................1

1.2 OVERVIEW ..................................................................................................2

CHAPTER 2 : Literature Study..........................................................3

2.1 ENGINE INTAKE SYSTEMS......................................................................3

2-1.1 Introduction............................................................................................3
2-1.2 Overview and history of engine intakes .................................................3
2-1.3 Variable intake manifolds ......................................................................7
2-1.4 System Requirements .............................................................................9
2-1.5 Numerical methods ..............................................................................10
2-1.6 Tuning ..................................................................................................17
2-1.7 Excitation .............................................................................................22
2-1.8 Exhaust tuning effects ..........................................................................24
2-1.9 Variable back-pressure exhaust...........................................................26
2-1.10 Miscellaneous effects ...........................................................................26
2-1.11 Prediction of engine intake pressure using numerical methods ..........27
2-1.12 Combustion ..........................................................................................28
2-1.13 Conclusion ...........................................................................................29
2.2 COMPUTATIONAL FLUID DYNAMICS ................................................30
2-2.1 Introduction..........................................................................................30
2-2.2 Governing equations............................................................................31
2-2.3 Conservation of Mass ..........................................................................31
2-2.4 Conservation of Momentum.................................................................32
2-2.5 The Energy Equation ...........................................................................33
2-2.6 Two-Equation Model: k-ε Model .........................................................33
 University of Pretoria etd – Makgata, K W (2005)
vi

2-2.7 Near-Wall Treatment ...........................................................................35

2-2.8 Grid generation and GAMBIT .............................................................36
2-2.9 Boundary Conditions ...........................................................................41
2-2.10 Solution Algorithms .............................................................................42
2-2.11 Convergence ........................................................................................43
2-2.12 Background on the CFD solver used (FLUENT) ................................43
2-2.13 Conclusion ...........................................................................................44
2.3 MATHEMATICAL OPTIMISATION........................................................45
2-3.1 Introduction..........................................................................................45
2-3.2 Theory of optimisation .........................................................................46
2-3.3 Response surface methodology ............................................................46
2-3.4 Computational flow optimisation.........................................................49
2-3.5 Conclusion ...........................................................................................50

CHAPTER 3 : Intake Modelling .......................................................51

3.1 PROBLEM DEFINITION AND APPROACH ...........................................51

3-1.1 Problem definition ...............................................................................51
3-1.2 Modelling Approach ............................................................................52
3-1.3 EngMod4T............................................................................................54
3.2 SINGLE INTAKE WITHOUT AIRBOX ...................................................57
3-2.1 Grid Generation...................................................................................57
3-2.2 Boundary Conditions ...........................................................................59
3-2.3 Results and Discussion ........................................................................59
3-2.4 Conclusion ...........................................................................................61
3.3 FOUR INTAKES WITHOUT AIRBOX.....................................................62
3-3.1 Description of Problem........................................................................62
3-3.2 Grid Generation...................................................................................63
3-3.3 Boundary Conditions ...........................................................................65
3-3.4 Transient Modelling.............................................................................65
3-3.5 Results and Discussion ........................................................................66
3-3.6 Conclusion ...........................................................................................67
3.4 FOUR INTAKES WITH AIRBOX (BASE CASE)....................................68
3-4.1 Grid Generation and Boundary conditions .........................................69
3-4.2 Steady-state modelling and experimental validation...........................71
 University of Pretoria etd – Makgata, K W (2005)
vii

3-4.3 Transient Modelling.............................................................................73

3-4.4 Conclusion ...........................................................................................79

CHAPTER 4 : Intake Optimisation ..................................................80

4.1 INTRODUCTION .......................................................................................80

4.2 AIRBOX OPTIMISATION.........................................................................81
4-2.1 Candidate design variables and objective functions ...........................81
4-2.2 Mathematical formulation ...................................................................83
4-2.3 Results ..................................................................................................84
4-2.4 Off-design performance .......................................................................87
4-2.5 Conclusion ...........................................................................................89

CHAPTER 5 : Conclusions and Recommendations........................90

5.1 CONCLUSIONS..........................................................................................90
5.2 RECOMMENDATIONS.............................................................................91

REFERENCES........................................................................................93

Appendix A: Intake system....................................................................97

Appendix B: Input for EngMod4T, Gambit and Fluent.....................99

Appendix C: Experimental Setup .......................................................103

Appendix D: Journal Files ...................................................................105

Appendix E: Summary of Results .......................................................110

 University of Pretoria etd – Makgata, K W (2005)
Abbreviations and Acronyms viii

Abbreviations and Acronyms

This list contains abbreviations and acronyms as used in this report.

Abbreviation or Acronym Definition

ABDC After bottom dead centre

ANOVA Analysis of variance

ATDC After top dead centre

BBDC Before bottom dead centre

BTDC Before top dead centre

CAD Computer-aided design

CFD Computational fluid dynamics

CFO Computational flow optimisation

DNS Direct numerical simulation

DOHC Double overhead cam

FV Finite volume

ICE Internal combustion engine

LES Large eddy simulation

LFOPC Leap Frog Optimisation for Constrained Optimisation

PDE Partial differential equation

PISO Pressure-implicit with splitting of operators

RANS Reynolds-averaged Navier-Stokes

RPM Revolutions per minute

RSM Reynolds stress methods

SIMPLE Semi-implicit Method of Pressure-Linked Equations

SRSM Successive response surface method

VVT Variable valve timing

WOT Wide open throttle

 University of Pretoria etd – Makgata, K W (2005)
List of Figures ix

List of Figures

Figure 2-1: 1937 Riley hi-charge induction system [5] .................................................4

Figure 2-2: Induction tract of a Chrysler Valiant 5.2 litre V-8 [5] ................................5
Figure 2-3: Two twin-barrel carburettors with adaptor manifolds and trumpets on a
Morgan-TR four-cylinder engine [5] .....................................................................6
Figure 2-4: Below 5000rpm...........................................................................................7
Figure 2-5: 5000-5800 rpm............................................................................................7
Figure 2-6: Above 5800 rpm..........................................................................................7
Figure 2-7: Below 5000rpm...........................................................................................7
Figure 2-8: 5000-5800 rpm............................................................................................7
Figure 2-9: Above 5800 rpm..........................................................................................7
Figure 2-10: Propagation of a linear wave [12] ...........................................................12
Figure 2-11: Propagation of non-linear (a) pressure and (b) rarefaction waves along a
pipe [12] ...............................................................................................................14
Figure 2-12: The 1st Order Helmholtz Resonator (I-C Resonator) [11] ......................19
Figure 2-13: Effect of primary pipe length on volumetric efficiency for a four-cylinder
engine [12] ...........................................................................................................20
Figure 2-14: The 2nd Order Helmholtz Resonator (R-P Resonator) [11].....................21
Figure 2-15: Modelling of a resonance charging manifold two-degree-of-freedom
system [9].............................................................................................................22
Figure 2-16: Pressure at inlet valve vs. crank angle at different speeds. [Dotted line
depicts resonance speed (4500 rpm)] [9] .............................................................23
Figure 2-17: Modes of oscillation of 2nd Order system [9]..........................................24
Figure 2-18: Experimental exhausts on Ford V-8 racing engine [5] ...........................25
Figure 2-19: Phased burn combustion model [6].........................................................29
Figure 2-20: Ideal Otto cycle compared to the phased burn model [6] .......................29
Figure 2-21: Velocity profile as function of distance normal to the wall in the
turbulent boundary layer [24] ..............................................................................36
Figure 2-22: Cooper meshing scheme volume treatment [26].....................................39
Figure 2-23: Union of cube, a cylinder, and a triangular prism [26] ...........................39
Figure 2-24: Cooper meshing scheme [26]..................................................................40
Figure 2-25: Illustration of meshing types used in the study.......................................40
 University of Pretoria etd – Makgata, K W (2005)
List of Figures x

Figure 2-26: Progression of subregion in SRSM: (a) Only panning, (b) Only zooming
and (c) a combination of panning and zooming [30]...........................................48
Figure 3-1: Rendered 3-D Single intake showing components ...................................51
Figure 3-2: History of average pressure at inlet valve (Base design) ..........................53
Figure 3-3: Pressure at the inlet valve as simulated by EngMod4T (valve open) .......55
Figure 3-4: Temperature at the inlet valve as simulated by EngMod4T (valve open) 55
Figure 3-5: Sensitivity of inlet valve pressure to intake geometry changes ................56
Figure 3-6: Overall grid of single intake model...........................................................57
Figure 3-7: Zoomed-in view of intake .........................................................................58
Figure 3-8: Close-up of surface grid at the inlet port...................................................58
Figure 3-9: Rendered view showing boundary types by colour ..................................59
Figure 3-10: Mass flow rate at the inlet valve from the CFD simulation ....................60
Figure 3-11: Solid rendering of four-intake arrangement............................................63
Figure 3-12: Overall grid of four intakes without airbox ............................................64
Figure 3-13: Close-up view of the mesh of the four intakes........................................64
Figure 3-14: Boundary location ...................................................................................65
Figure 3-15: Mass of air inducted by individual intakes .............................................67
Figure 3-16: Original airbox geometry ........................................................................68
Figure 3-17: Views of overall grid...............................................................................69
Figure 3-18: Rendered view showing boundary types by colour (intakes and airbox)70
Figure 3-19: Flowbench experimental set-up ..............................................................71
Figure 3-20: Full geometry with simplified valves added and valve-lift pressure
boundary specification.........................................................................................72
Figure 3-21: Mesh extension for valve lift simulations ...............................................72
Figure 3-22: Comparison of CFD and flowbench results............................................73
Figure 3-23: Comparison of CFD and EngMod4T average static pressure prediction at
the inlet valve.......................................................................................................74
Figure 3-24: Pressure contours on walls at different flow times .................................75
Figure 3-25: Instantaneous velocity path lines in airbox and inlet port (Intake 1) ......75
Figure 3-26: Mass flow rate at the inlet valve from the CFD simulation ....................76
Figure 3-27: Mass of air inducted at individual intakes over one engine cycle ..........77
Figure 3-28: Shear stress and Y-plus plot for base case geometry ..............................78
Figure 3-29:Y-plus values on intake inner walls .........................................................79
Figure 4-1: Intake-airbox arrangement with variables.................................................81
 University of Pretoria etd – Makgata, K W (2005)
List of Figures xi

Figure 4-2: Results of airbox optimisation study.........................................................84

Figure 4-3: Illustration of the base and optimum airbox designs [mm] ......................85
Figure 4-4: Design variables for first iteration ............................................................86
Figure 4-5: Design variables for second iteration with last point simulated included
(point 8)................................................................................................................86
Figure 4-6: Simulated pressure for different engine speeds ........................................88
Figure 4-7: Mass flow rate response of base and improved airboxes at different engine
speeds...................................................................................................................88
 University of Pretoria etd – Makgata, K W (2005)
List of Tables xii

List of Tables

Table 2-1: Illustration of Porsche’s VarioRam System [8] ...........................................7

Table 2-2: Comparison of PowerFLOW results with Experiment results [18] ...........27
Table 2-3: Recommended values for empirical constants [21][24].............................35
Table 3-1: Summary of input data into EngMod4T.....................................................54
Table 3-2: Valve timing data .......................................................................................65
Table 4-1: Bounds for optimisation problem...............................................................83
Table 4-2: Parameter values and mass flow rate results of airbox optimisation .........85
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 1: Introduction 1

CHAPTER 1 : Introduction

1.1 MOTIVATION

The recent increase in computing power coupled with the declining cost of computers
has led to a rapid development in the science of computational fluid dynamics (CFD)
simulation. Simulation can be defined as the prediction and reconstruction of
behaviour of a product or physical situation under assumed or known conditions. It is
widely used in various engineering disciplines as a means of quickly determining or
improving the behaviour of a design, thus saving on experimental work that can be
expensive.

However, the emergence of simulation has not made experimental tests obsolete, as
these are important for verification during the design process. CFD as a simulation
tool has proved itself to be extremely useful over the years, especially since some
phenomena cannot be determined experimentally, for example molten steel flow
patterns. It has also been successfully used in the aerospace industry to predict
aerodynamic behaviour of aircraft components such as wings, canards, etc. These
simulations also yield derived flow information such as vorticity and residence time,
which are difficult to measure. CFD has also been extended for use in industrial
applications such as continuous steel casting, heat sink design (for cooling), building
ventilation and automotive engine cooling.

The use of CFD in prediction and improvement of system behaviour has, over the
years, gained acceptance in industrial application. This approach has worked well in
the improvement of a Formula 1 [1] and race car intake system, [2], [3], where the
results were also validated by road tests. Achieving race car intake performance
improvement without going the trial-and-error route is essential for economy and time
reasons, which is the reason that CFD and mathematical optimisation were employed
in this endeavour. Also, previous experimental attempts to improve the intake system
that is the subject of this study have yielded little success, hence the decision to use
numerical techniques.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 1: Introduction 2

The sole objective of this study is to determine and improve the flow in the engine
intake system of the concerned rally race engine through the use of the combination of
CFD and mathematical optimisation techniques. Understandably, the accuracy of
numerical methods is of great concern especially in CFD with the wide choice of
modelling options, all which give converged but different solutions. In this study, this
issue is addressed by comparing the results to flowbench experimental results.

1.2 OVERVIEW

A literature survey on the subject of intake flow, its history and existing intakes, is
presented in CHAPTER 2. Also included in this chapter is the theory of tuning, wave
interaction and numerical methods, concepts that are essential in order to understand
intake flow. The governing equations of flow are given in the section that discusses
CFD, together with a discussion of turbulence modelling as implemented in the
software used. Grid generation and the grid types employed in the study are also
presented. Mathematical optimisation and its formulation are also discussed in this
chapter. Only the optimisation technique that is relevant to the study is discussed in
detail.

An application of the CFD methods discussed above is given in CHAPTER 3 where

the intake is progressively modelled from the simplest geometry to the most complex.
The results are given and discussed in this section. Also included is the comparison
between numerical and experimental results as part of the validation effort.

CHAPTER 4 extends on the findings of CHAPTER 3. In this chapter, the

optimisation technique presented in the literature survey is applied to the problem,
and the optimisation loop, in conjunction with CFD, is performed to come to an
improved airbox design.

A summary of the work conducted in this study is given in CHAPTER 5. Conclusions

and recommendations for future work are also included based on the results obtained
in the study. Recommendations are based largely on the computational limitations and
numerical difficulties experienced in the study.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 3

CHAPTER 2 : Literature Study

2.1 ENGINE INTAKE SYSTEMS

2-1.1 Introduction

The basic function of engine intake systems is to provide the engine with a fresh air-
fuel mixture every cycle for combustion to take place. Different engine operation
requirements demand individualised intake systems. In other words, a vehicle that is
used for everyday purposes cannot have the same intake system as a racing vehicle,
though they may share the same engine platform. Racing engines require maximum
volumetric efficiency for increased power and torque, but this is not desirable as far as
economy is concerned. Everyday-use vehicles seldom require top-end power, thus
economy and driveability at lower speeds are more important in this instance. It is
therefore important that intake systems are designed to suit the purpose for which the
engine is intended.

This chapter begins by giving a basic overview of the intake system and its history,
followed by the system requirements for improved flow. The flow behaviour in the
intake is discussed next and thereafter, the exhaust gas effects are discussed. The last
section will describe engine intake performance prediction as applied to numerical
methods.

2-1.2 Overview and history of engine intakes

Engine intakes serve as flow passageways for the combustion of the air-fuel mixture
in the engine. Much attention has been focused in this area of development in an
attempt to maximise airflow into the engine, which would translate into better
performance. There is direct correlation between mass flow and power as shown by
Heywood [4]. The design of such systems has until recently been on a trial-and-error
basis, which is costly and time consuming. Early intake systems were based on any
commercially available pipes without regard for scientific reasoning pertaining to
interior flow. Many theories, largely by experience, were put forward regarding the
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 4

behaviour of the flow in the intake. These theories proved useful through the years
even when verification was necessary. Smith and Morrison [5] conducted extensive
research in this area and their work is still regarded as the basis for many scientific
designs. They proved most theories were actually justified and shed light onto other
factors to be considered when designing intake-exhaust systems. An example of an
early intake ram-induction attempt can be seen in Figure 2-1 as used on a 1937 Riley
1100cc and a 1500cc four-cylinder engine.

Figure 2-1: 1937 Riley hi-charge induction system [5]

The long induction ducts were used to try and take advantage of the ram effect.
Although the system was not a success due to erratic mixture formation arising from
wide temperature effects, more attempts on different engines did yield some success.
Chrysler Corporation had notable success with increased inlet length ducting as seen
in Figure 2-2. The long ducts make efficient use of the pressure pulsations to achieve
a torque increase of around ten percent (10%). The ducts, equal in length, were
selected to give an increase in torque over a 1500 rpm range with the maximum
occurring at 2800 rpm.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 5

Figure 2-2: Induction tract of a Chrysler Valiant 5.2 litre V-8 [5]

The layout of induction systems has not changed much since the early days. Basically
the system comprises an air inlet, load control device, carburettor/injector and inlet
tract (duct or manifold) leading into an inlet port. In some arrangements, like in a
carburettor engine, the inlet port serves the same purpose as the manifold. On the
other hand, the arrangement mentioned is normally different for racing engines where
the air inlet is usually replaced by a combination of a bellmouth-end type pipe and an
airbox-filter combination. The bellmouth-end pipe is normally referred to as a
“trumpet”. This arrangement is shown in Figure 2-3.

The airbox acts as a plenum chamber where the incoming air is diffused, thereby
increasing the pressure and reducing the velocity. Thus increased amounts of air can
be inducted with the use of the airbox. On Boxer (flat cylinder arrangement) and V-
type (V-arranged cylinders) engines, resonance intake manifolds may be used to boost
mid to high revving efficiency. A common plenum chamber feeds each bank of
cylinders through separate pipes [6].
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 6

Two pipes of different diameters connect the plenum chambers on the cylinder banks
and a valve controlled by an engine management system can close one of the pipes to
vary the amount of air going into each bank. The firing order is arranged such that the
cylinders breathe alternately from each chamber hence creating pressure waves
between them. If the frequency of pressure wave matches the engine speed, cylinder
filling is greatly improved and so is volumetric efficiency. By closing one of the
interconnecting pipes at low engine revs, the area and the frequency are reduced thus
enhancing mid-rev output.

Trumpets

Figure 2-3: Two twin-barrel carburettors with adaptor manifolds and trumpets
on a Morgan-TR four-cylinder engine [5]

Modern-day intake systems use innovative ways to achieve an even spread of torque
throughout the engine speed range. One ingenious design is the variable inlet tract
length manifold. This design varies the inlet tract length according to the engine
speed, thus running the engine at its optimum at all speeds. This system is widely
used by large car manufacturers such as Audi, BMW, Nissan, Volkswagen, Ferrari
and Porsche [8]. Motor manufacturer Mazda, has also used innovative induction
systems on their engines further showing the importance attached to intake systems
and their effect on power and torque [9], [10].
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 7

2-1.3 Variable intake manifolds

There are two kinds of variable intake manifolds namely; variable length intake and
resonance intake manifolds. Both manifolds make use of the geometry of intake
manifolds to reach the same goal. These manifolds are normally used on heavy
modern sedans that would benefit from increased torque. In addition, sports cars such
as the Ferrari 360 M and the 550 M make use of this technology for better driveability
at low speeds. By combining these manifolds with a system such as variable valve
timing (VVT), the engine can be made more flexible throughout its speed range with
improved fuel consumption.

The figures in Table 2-1 illustrate the 3-stage intake Porsche VarioRam system as
used on the 993 Carrera.

Table 2-1: Illustration of Porsche’s VarioRam System [8]

Figure 2-4: Below 5000rpm Figure 2-5: 5000-5800 rpm Figure 2-6: Above 5800 rpm

Figure 2-7: Below 5000rpm Figure 2-8: 5000-5800 rpm Figure 2-9: Above 5800 rpm
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 8

Below 5000 rpm, the long pipes and the resonance intake (item number 7) are
disabled and in the mid-range speed (5000-5800 rpm), the long pipes plus short-pipe
resonance intake are opened (valve 2) with one of the interconnected pipes of the
resonance intake pipes closed. At the top-end speed range, all the pipes are opened to
achieve maximum charging from the resonance intake.

Even with modern-day intake-exhaust system scientific design principles, many

arrangements are still a compromise between cost and engine efficiency. Unavoidable
engine design requisites are mostly responsible for a loss of efficiency, especially in
the intake stages. Valve-timing requirements, such as overlapping between inlet and
exhaust valve openings, are to blame for some loss of intake charge. The effect of
blow down is mostly evident in forced-induction engines where the momentum of the
incoming charge carries it through the open exhaust valve during overlapping.

As the cost of development is of concern to many designers, most intake system

designs are not optimised for reduced flow resistance and/or increased mass flow rate.
Modifications to intake systems can yield increased mass flow with only a slight
increase in power and torque. However, the cost and time required to arrive at the
optimised intake system can be enormous. It is because of this fact that many
designers cannot justify spending much time, effort and funds to optimise such
systems for mass production vehicles.

However, any slight increase in power and torque in the motor racing world could
mean the difference between winning and losing races. It is for this reason that
research and development has been most active in this segment of motoring. Intake
system designers harness exhaust gas energy to enhance mass flow into the cylinder
head by making use of the pressure wave motion of the exhaust process. However,
this effect can only be obtained if the exhaust pipe is tuned for that particular speed to
avoid exhaust gas motion from having a negative effect on performance. It is thus
imperative that one has a good understanding of the unsteady gas process before
commencing with the tuning of intake-exhaust systems. Smith and Morrison [5], Blair
[6] and Smith et al [7] provide an in-depth analysis and explanation of the
mechanisms that govern these gas processes.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 9

Found in Blair [6], are also some useful sections dealing with tuning as well as engine
computer (numerical) modelling. Because of modern-day computing capability, large
engine simulations can be performed and it is thus helpful to have some guidance in
this area to ensure credible and cost-effective results in the shortest time possible.

2-1.4 System Requirements

Intake manifolds usually comprise a load-control device (throttle body), a plenum and
runners that feed the charge to individual cylinders. Requirements in the design of
such systems include low resistance to the flow, good distribution of air-fuel charge to
the individual cylinders, tract lengths that can take advantage of the ram and tuning
effects as well as manifold heating for adequate fuel vaporisation.

One popular method to increase flow in the inlet port is through a gas-flowing
process, where the surface roughness of the inlet port is reduced substantially hence
reducing resistance to the flow. However, the effect this has on power output of an
engine is minimal when compared to other methods such as manifold tuning.

Manifold tuning takes advantage of the pressure pulses in the manifold to increase
mass flow. By making the manifold a certain length, it is possible for the highest
negative pressure wave to occur at the manifold exit (cylinder inlet) at chosen engine
speeds. When the valve is opened at this point, this pressure energy forces the air
mixture into the cylinder. Extensive research has gone into engine manifold and
airbox tuning as indicated by the amount of available literature in this regard.

The airflow quality in an airbox can be improved by reducing sharp corners as well as
ensuring that choking does not occur anywhere in the throat (i.e., negative entry
effects are minimised). Choking occurs when the local mach number reaches unity,
thus limiting the amount of intake charge inducted. De Vita et al [2],[3], have done
some research in this field and have successfully reduced the losses in the airbox by
using CFD as an aid. Although optimisation was performed, it was based on trial-and-
error thus not necessarily ensuring an optimum design. This study however, gives an
understanding of sources of flow losses and thus provides a starting point for the
mathematical optimisation process. By eliminating obvious design parameters that
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 10

give rise to losses, an optimum design can be reached in fewer iterations. This not
only saves time and effort, but also reduces expenses.

2-1.5 Numerical methods

Numerical methods for determining the pressure in the system have been developed
and used extensively over the years. These methods provide solutions to the complex
governing equations of fluid behaviour such as mass and heat transfer. Rapid
advances in computing power have made it possible for these equations to be solved
in an economic manner for complex flow problems where the equations may be non-
linear and the computational expense enormous. Numerous authors such as
Winterbone and Yoshitomi [11], Winterbone and Pearson [12], and Blair [6] have
done extensive research on this subject, with much success in accurately predicting
the transient pressure in intake systems. Their contributions are covered in this
section. Basically, the authors describe the intake flow in terms of wave theory.

Wave behaviour theory is largely based on sound theory and its propagation through a
medium after Newton, Laplace and other scientists discovered the correlation between
sound and wave actions. These wave action methods are applied to determine the
behaviour of waves within the engine intake system to predict the performance for
different configurations.

Intake charge within an intake pipe propagates in waveform because of the unsteady
pressure at the inlet valve due to piston movement as well as combustion in the
cylinder. Multiple waves are set up in the process and these waves move back and
forth in the intake, setting up pressure pulses. However, these waves do not normally
move in the same direction resulting in particle collisions. This phenomenon is known
as super-positioning. It is fundamental to understand this phenomenon as it occurs
continuously in the intake pipes and can be the cause of some inaccurate pressure
transducer readings in experimental work.

Basically two types of waves are continuously set up in engine manifolds namely,
linear and non-linear waves. These types of waves are discussed in detail.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 11

2.1.5.1 Linear waves

Linear waves are waves that propagate in a pipe system with disturbances whose
amplitude is small when compared to the undisturbed flow. This assumption leads to
the classical wave equation
∂ 2 ( ∆p ) 2 ∂ ( ∆p )
2

= a0 (2.1)
∂t 2 ∂x 2

This equation describes propagation through a homogenous medium at a single wave

speed, a0. This wave speed, independent of either waveform or direction of
propagation, can refer to the speed of light in electromagnetic waves in free space or
the speed of sound in acoustic waves. Analytical solutions to equation (2.1) can be
obtained in both the time and frequency domain as will be shown.

Time domain solution

To obtain the solution in the time domain Equation (2.1) is re-written as follows
∂ 2 pe 2 ∂ pe
2
= a0 , (2.2)
∂t 2 ∂x 2

where pe is the excess pressure above the undisturbed level, i.e., pe= p-p0, with
proposed general solutions of the form

pe = f ( t − x a0 ) + g ( t + x a0 ) (2.3)

or
pe = f ( x − a0t ) + g ( x + a0t ) , (2.4)

where the functions f and g represent longitudinal pressure waves, travelling in

opposite directions, in which the wave induces a velocity field parallel to the direction
of propagation. The sum of the individual waves gives the resultant pressure wave. By
considering movement only in the positive x-direction (g = 0), equation (2.4) reduces
to
pe = f ( x − a0t ) (2.5)
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 12

which remains constant for no wave amplitude dissipation hence the quantity
( x − a0t ) must be constant thus

x = a0t + const . (2.6)

At time t = 0 and still considering the positive x-direction, equation (2.4) gives an
arbitrary wave of the form f(x). At a later time t, the wave has an identical shape but
shifted by distance a0t as shown by Figure 2-10 and hence, waves can be tracked from
one end of a pipe to the other by simply taking note of the time delay. This method is
sometimes referred to as transmission line analysis because of the direct analogy to
transmission of power in an electrical transmission line without any losses.

Figure 2-10: Propagation of a linear wave [12]

The relationship between the excess pressure, pe, and the fluid velocity, u, in the
direction of propagation of linear waves is given by
pe m ρ 0 a0u = 0 (2.7)

The net velocity of induced by waves moving in the positive and negative x-directions
respectively, is of the following form
1
u=  f ( x − a0t ) − g ( x + a0t )  (2.8)
ρ 0 a0 

The mass flow rate is given by

 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 13

m& = ρ 0uF (2.9)

where F is the pipe cross-sectional area. By substituting equation (2.8) into (2.9) the
mass flow rate is redefined as
1
m& =  f ( x − a0t ) − g ( x + a0t )  (2.10)
Zc 

where Zc=a0/F is known as the pipe impedance, the inverse of which (Yc) is called the
pipe admittance.

Pearson, Winterbone and others have applied linear wave theory in the time domain to
solve pipe network systems and also wave propagation in engine intake manifolds
because of the fast solution times of such models. The assumptions of linear wave
theory are more appropriate for engine intake manifold flow because of the smaller
pressure wave amplitudes and velocities. Also, the flow in intake manifolds has
entropy levels that are nearly constant when compared to exhaust gas flow. These
methods are also very good in predicting volumetric efficiencies.

Frequency domain solution

Further understanding of the characteristics of wave propagation can be obtained by
describing the linear wave equations in the frequency domain. Expressions can then
be derived for the natural frequency of complex manifold systems. Since the system
considered in this study is one of a simple nature not requiring wave equation
solutions in the frequency domain, the reader is referred to Winterbone and Pearson
[12] for the full mathematical treatment and derivations.

2.1.5.2 Non-linear waves

The system considered in this study includes an airbox, whose flow behaviour cannot
be easily determined by the use of linear wave theory, hence requiring the use of non-
linear wave theory. A summary of non-linear wave theory is presented below.

Contrary to linear waves, as non-linear waves propagate, their amplitude is affected

and shocks can occur as shown by Figure 2-11.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 14

Figure 2-11: Propagation of non-linear (a) pressure and (b) rarefaction waves
along a pipe [12]

The equations that describe the waves will be presented in their summarised form for
both simple and compound waves. The full derivation of the equations is dealt with in
detail by Winterbone and Pearson [12]. Simple waves are waves that propagate as a
result of a single disturbance, while compound waves are a result of wave super-
positioning.

Simple waves
The particle or gas velocity as a function of pressure is of the form
(κ −1) 2κ
2a0  p  
u=   − 1 + u0 (2.11)
κ − 1  p0  

 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 15

while the speed of propagation, c, of a point on the wave is the sum of the local gas
velocity and the local speed of sound to give the following
(κ −1) 2κ
2a0  κ + 1  p  
c=    − 1 + u0 , (2.12)
κ − 1  2  p0  


where κ is the ratio of specific heats of the medium through which the wave is
travelling. The above equations are known as the Earnshaw’s equations. From
equation (2.12), the propagation speed of points in a wave increases as the pressure
increases, and eventually the pressure wave steepens and a shock is formed as seen in
Figure 2-11. After the shock has formed, it is maintained and equations (2.11) and
(2.12) are no longer valid. The shock wave characteristics are described by the
following equations
 p 
 p − 1
u = a0  0  + u0 (2.13)
1
 κ (κ − 1)  κ + 1  p  2

   + 1 
 2  κ − 1  p0  
and
1
  κ − 1   κ + 1  p  2

c = a0     + 1  + u0 (2.14)
 2κ   κ − 1  p0  

For air, with a ratio of specific heats, κ = 7/5, equations (2.11) to (2.14) become
for no shock:
 p 1 7 
u = 5a0   − 1 + u0 (2.15)
 p0  

and
  p 1 7 
c = a0  6   − 5 + u0 . (2.16)
  p0  

while for shock waves:

 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 16

 p 
 p − 1
u = 5a0  0  + u0 (2.17)
1
  p  2

 7 6 + 1 
  p0  
and
1
 1   p  2

c = a0   6 + 1  + u0 . (2.18)
 7   p0  

Compound waves
Expressions for the resultant pressure and velocity of compound wave systems can
also be derived. Consider two non-linear waves travelling towards one another in
stationary air (At pressure p0 and speed of sound a0) with absolute crest pressures of
p1 and p2. As they overlap, the absolute pressure and particle velocity of the resultant
wave at the overlap region become pt and ut respectively. The resultant particle
velocity in that region is given by
 p 1 7   p 1 7 
ut = 5a0   − 1 − 5a0  2  − 1
1
(2.19)
 p0    p0  

which is essentially the summation of the two wave particle velocities thus
ut = u1 + u2 . (2.20)

Equation (2.20) is also valid for linear waves with the negative sign in equation (2.19)
just an indication of the direction in which the wave is travelling, the convention
being left to right is positive while right to left is negative.

Method of Characteristics
Non-linear waves can be analysed through the use of the Method of Characteristics
since the partial differential equations do not yield closed analytical solutions. This
method aided in the simulation of engine performance while highlighting major
features of unsteady flow in engine manifolds. Besides having difficulty in dealing
with different fluids, two other major drawbacks of the method are [12]:
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 17

1. Conservation of properties, i.e., continuity, energy and momentum, is not

always achievable between two pipe ends, especially where there are large
pipe cross-section changes or entropy variations.
2. The method has difficulty in handling large pressure waves, which can be a
problem especially when dealing with exhaust systems that have diffusers.

To overcome these defects, finite difference (volume) methods, as found in most

computational fluid dynamics codes (CFD), are used while also making it possible to
find solutions in more than one dimension. The solution algorithms for these methods
can be written in conservative form and they can handle large pressure waves and
shocks with little numerical difficulties.

2-1.6 Tuning

Engine tuning can be loosely defined as a process of getting a system to respond in a

desired manner at specified engine speeds while taking into account the interactions
between the intake and exhaust systems. The many factors involved in engine tuning
have led to tuning developing into a science with methods for calculating wave
actions explored in more detail for increased accuracy. There are two distinct types of
inlet tuning namely, inertia and resonance tuning [11].

First-Order Helmholtz resonator

Inertia charging occurs due to the individual runners and throttle bodies per cylinder
and their geometry, while resonance tuning is largely influenced by plenums.

Cylinder charging at the tuning speed is dependent on the system resonance

frequency, which in turn depends on whether the inlet valve is closed or open. This
resonance is the vibration of air column in the intake manifold and different vibration
behaviours have been identified. When the valve is closed, the pipes resonate as a
quarter wave organ pipe with the frequency given by [11]:
a
f = (2.21)
4L

where a = speed of sound of air in the manifold [m/s]

 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 18

L = length of the primary pipe [m]

f = frequency [Hz]

However, when the valve is open, the system resonates as a Helmholtz resonator. This
mode of vibration occurs when the column of air in the pipe moves in-phase against
the air in the cylinder. An analogy can be made to the spring-mass system in which
the spring is the compressibility of the air in the cylinder and the mass is the mass of
the column of air. The natural frequency of a first-order Helmholtz resonator is given
by equation (2.22):
1
a  S 2
f =  LV  (2.22)
2π  

where S = cross-sectional area of primary pipe [m2] and

V = volume of resonator chamber [m3].

Alternatively, the volume V in the equation above can be taken as the mean cylinder
volume during the induction stroke, thus equation (2.22) becomes:
1 1
a  2S  2  r − 1 2
f =  LV   r + 1 (2.23)
2π

where r = nominal compression ratio.

Figure 2-12 is an illustration of the first-order Helmholtz resonator (I-C resonator).

Here the gas dynamics are modelled as a spring-mass system with the air in the
cylinder modelled as an air spring (Spring) and the air in the manifold as an air piston
(Mass).
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 19

Figure 2-12: The 1st Order Helmholtz Resonator (I-C Resonator) [11]

Such a resonance system results in tuning at certain engine speeds with the peak
pressure occurring at the inlet valve before it closes. Although not explicitly stated, it
can be assumed that most of the publications that appear on frequency tuning are
based on equations (2.21) and (2.23).

Through immense experimental work on a four-cylinder engine, Winterbone and

Pearson [12] have shown that the primary pipe length has a major influence on the
volumetric efficiency of an engine as shown in Figure 2-13. The standard manifold
pipe length of the engine was 0.4 m, which was increased and decreased at intervals
of 0.05 m.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 20

Figure 2-13: Effect of primary pipe length on volumetric efficiency for a four-
cylinder engine [12]

The volumetric efficiency plots exhibit two peaks at different engine speeds and it can
be seen that the engine speed at which the first peak occurs, for all plots, changes very
little when compared to the second peak. As the pipe is lengthened, the second peak
increases due to the increased amount of air in the manifold. At the same time, the
resonant frequency is reduced (see equation (2.21)) making the engine lose power at
higher speeds. On the other hand, shortening the manifold pipe results in the second
peak moving to the higher engine speed range with reduced volumetric efficiency.
Inertia tuning, brought about by manifold resonance effects, accounts for the shifting
of the second peak through the engine range. It is evident from the figure that by
combining the two tuning modes; one can achieve a steady increase of volumetric
efficiency and torque over a wide engine speed range, hence the emergence of
variable intake manifolds in modern vehicles.

Second-Order Helmholtz Resonator

The Second-Order Helmholtz resonator (R-P resonator), applicable only to boxer and
V-type engines, comprises a resonance tube and a plenum. Referring to Figure 2-14,
the air in the plenum is modelled as a spring (Spring A) and the air in the resonance
tubes as an air piston (Mass A).
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 21

Figure 2-14: The 2nd Order Helmholtz Resonator (R-P Resonator) [11]

In the boxer and V-type configurations, charging is achieved by a combination of

inertia and resonance charging when using a resonance-charging manifold. Thus
combining both the I-C and R-P resonators, modelling of these engine types is
possible. The configuration in both types of engines is symmetrical in nature and thus
the system can be simplified by only considering one bank of cylinders and plenums.
The modelling process from the engine configuration to the final simplified
symmetrical two-degree-of-freedom model can be seen in Figure 2-15.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 22

Figure 2-15: Modelling of a resonance charging manifold two-degree-of-freedom

system [9]

2-1.7 Excitation

The piston movement during the induction stroke as well as the inlet valve changing
area, provide the necessary excitation required for inertia and resonance charging. In
first-order systems, there is an increase in phase shift between the excitation and the
spring, with an increase in excitation frequency. At the resonance speed, this results in
a quarter period phase shift when the pressure is observed in the manifold (see Figure
2-16).
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 23

Figure 2-16: Pressure at inlet valve vs. crank angle at different speeds. [Dotted
line depicts resonance speed (4500 rpm)] [9]

In a resonance charging manifold, two resonance engine speeds exist as shown in

Figure 2-17. At low engine revolutions per minute (rpm) and primary resonant rpm,
both mass A and B are in phase and with the result that when both springs are
compressed, a higher mass flow rate is expected at the inlet valve.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 24

Figure 2-17: Modes of oscillation of 2nd Order system [9]

At the secondary resonant rpm, masses A and B are almost 90 degree out of phase. At
this resonance rpm, less variance in the mass flow rate than at the primary resonant
rpm is expected. This is because the direction of flow in the intake manifold and
plenum chamber, as indicated by springs A and B, is always opposite though not
necessarily equal. Reverse flow can be expected at the inlet valve due to these wave
interactions.

2-1.8 Exhaust tuning effects

Although outside the scope of this study, an understanding of exhaust gas dynamics is
important. It is inevitable that exhaust gases and the fresh inducted mixture will have
some sort of interaction since some valve overlap occurs. It is thus important that this
interaction does not have an undesirable effect on engine performance. In fact,
exhaust gases can actually aid in the induction process. Smith and Morrison [5] and
Blair [6] also cover this aspect of tuning in detail.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 25

During the exhaust process in a four-stroke engine, the inlet valve usually opens
before top dead centre (BTDC). During this valve overlap, the exhaust gas motion
causes pressure waves that, depending on the exhaust pipe configuration and engine
speed, can aid or restrict intake flow. Normal exhaust pipes for sports cars collect
exhaust pulses from individual cylinders and combine them into a larger pulse, with a
corresponding lower pressure behind the pulse. This low pressure actually helps in
drawing more air/fuel mixture into the cylinder from intake manifolds. This is called
scavenging. This wave phenomenon enables late inlet valve closing times, i.e., after
bottom dead centre (ABDC), during the compression stroke because of the
momentum of the incoming air. This momentum ensures better cylinder filling.

Exhaust tuning takes into account these effects to increase mass flow and requires an
understanding of the mentioned processes. It is because of these interactions that
tuned exhaust manifolds have a different configuration to original exhaust systems.
Certain restrictions such as engine bay space influence the geometry of the exhaust
system such that the final tuned system may not necessarily be the best system.

Figure 2-18: Experimental exhausts on Ford V-8 racing engine [5]

Figure 2-18 is an excellent example of good exhaust designs that cannot be practically
implemented into a race car due to space limitations. However, such experimental
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 26

designs are still necessary in order to gain a better understanding of an engine’s

performance. These designs eventually lead to a final design that attempts to
incorporate the most positive aspects of the experimental designs with some
compromises due to limitations.

2-1.9 Variable back-pressure exhaust

Scavenging works best at certain engine speeds and this speed is fixed for different
pipe lengths. That is; at low engine speeds, shorter pipes give the best scavenging
performance and vice-versa at high engine speeds. It is because of this reason that an
increasing number of modern super cars (e.g., Aston Martin Vanquish, Ferrari 550
Maranello, 360 Modena, Mitsubishi Lancer Evolution VII and Lamborghini Diablo
6.0) now employ the use of a variable back-pressure exhaust. This works in a similar
manner to the variable intake manifold, the only difference being that it works to
expel spent gases while the intake serves to provide fresh air mixture.

The system usually comprises two different lengths of exhaust pipes with a valve that
can switch between them. The system can therefore satisfy both low and high-speed
output requirements with reduced noise levels and harmful gas emissions.

2-1.10 Miscellaneous effects

Exhaust silencers and end-pieces have also been subject to design with the aim of
reducing flow losses and hence ensuring lower pressures in the flow path. Reduction
of these losses also ensures smoother passage of exhaust gases therefore requiring less
piston work during the exhaust stroke.

Increased engine power output leads to increased stress on engine components and
this aspect should not be overlooked when power increase is required as it could lead
to severe engine damage. On the other hand, factory stock engines have a safety factor
designed into them for power to be increased by a certain amount without much
engine wear. However, because this engine threshold is difficult to determine except
by test-to-failure methods, engine components expected to experience increased stress
such as pistons, connecting rods, big-end bearings and cylinder head gaskets must be
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 27

replaced with stronger components. In the case of turbocharged engines with nitrous
oxide, ceramic coating for various components such as pistons, valves, cylinder head
and cylinder walls reduces wear and friction and also reduces heat build-up in the
combustion chamber.

2-1.11 Prediction of engine intake pressure using numerical methods

Accurate numerical methods are required in order to correctly simulate the flow
behaviour in a system. These methods aid in predicting some of the complex flow
phenomena or processes that occur in an engine. Although 0-D and 1-D engine
simulation techniques offer quick turnaround times when simulating pressure in the
combustion chamber, these techniques do not give qualitative information about the
flow such as squish, swirl or tumble flow. It is only recently that complex 3-
dimensional processes such as combustion have been modelled, and simulated
successfully in CFD [13], [14], [15].

However, where cost is a concern, these techniques do offer a viable option with an
acceptable compromise between accuracy and cost. This can be achieved by coupling
the 0-D or 1-D engine (such as EngMod4T® [16]) package to a CFD code with the
pressure solution being updated at every time-step [17]. A cheaper alternative is to use
pressure data extracted from these engine simulation codes and enforce it as a
boundary condition in the CFD code. This has proved to be successful as
demonstrated by results from PowerFLOW® 3.0 for an intake port with inlet valve
lifts of 7 mm, 9 mm and 11 mm as seen in Table 2-2 [18]. The table shows that there
is an excellent agreement between the PowerFLOW and experimental results.

Table 2-2: Comparison of PowerFLOW results with Experiment results [18]

Volume Flow Rate (l/s) Swirl Moment Ratio
Experiment PowerFLOW Experiment PowerFLOW
7 mm 43.5 42.1 0.7 0.66
9 mm 46.5 47.6 1.00 1.00
11 mm 47.8 50.9 1.18 1.25
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 28

PowerFLOW® does not solve the Navier-stokes equations directly, but instead models
the microscopic physics of the particles that compose the macroscopic fluid. This
method contrasts with traditional CFD methods in that it works on the principle of
simplifying microscopic behaviour while still achieving desired macroscopic
behaviour. The particle distributions are represented with integers that allow the fluid
dynamics to be processed exactly on digital computers. Although this fluid dynamics
code does not take the traditional CFD approach, it is nevertheless worth mentioning.

It is also possible to realise engine performance improvement by the sole use of 1-D
engine simulation codes as demonstrated by Wirtz and Mazzoni [19] and
Papaconstantis [20]. These authors managed to achieve an improvement in power and
torque of approximately 5 % to 7 %. The time taken for these simulations is only a
fraction of the time that a CFD simulation would take, and thus 1-D simulations are
desirable for quick and inexpensive results. The major drawback of 1-D simulation
results is that they give the user no insight into the flow behaviour such as swirl and
separation, which is the advantage that CFD has over these methods.

2-1.12 Combustion

Spark ignition internal combustion engine operation is illustrated by the Otto cycle
[4]. This cycle describes the thermodynamic events that occur in a closed cycle but is
however based on constant volume combustion, which is not the case in real engine
operation. Combustion is thus modelled as a time-dependent process while taking into
account the ignition-timing advance and the finite time required for combustion [6].

When a combustion mixture is ignited at 25° BTDC, there is an ignition delay of

about 10° (i.e., from 335° to 345° crank angle). The actual period of heat release takes
place over a burn duration, b, of approximately 50° (from 345° to 395° crank angle),
over which the mass fraction of fuel burned has the profile as seen in Figure 2-19.
This can be referred to as the phased burn cycle. In Figure 2-20, the ideal Otto cycle is
compared to the phased burn cycle and the deviation is evident, with the phased burn
cycle more suited for real engine operation. This process is implemented into 1-D
engine simulation codes by taking into consideration the ignition advance, ignition
delay period and duration of burn.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 29

Figure 2-19: Phased burn combustion model [6]

Figure 2-20: Ideal Otto cycle compared to the phased burn model [6]

2-1.13 Conclusion

In this section, the history and development of intakes as well as their function were
presented. The equations describing wave behaviour in intakes were discussed as well
as the theory behind tuning. These governing equations are important as they serve as
the basis of the 1-D simulations that provide data for the CFD (discussed in the
section that follows) simulations.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 30

2.2 COMPUTATIONAL FLUID DYNAMICS

2-2.1 Introduction

Fluid flow is governed by the conservation of three fundamental quantities namely;

mass, momentum and energy. The description of these conservation laws in
mathematical form yields non-linear partial differential equations, which are famously
known as the Navier-Stokes equations. Although these Navier-Stokes equations have
been known since the 1800s, they have been too difficult to solve for arbitrary flows
until the development of the modern computer. The science involved in solving these
equations numerically on the computer is known as computational fluid dynamics
(CFD).

In order to solve the Navier-Stokes equations, the flow domain for the specific
problem is first discretised. Domain discretisation, using mesh or grid generating
techniques, involves breaking up the selected control volume into smaller volumes
and solving the Navier-Stokes equations over each volume. For simple geometries
such as rectangular shapes, discretisation is simple but gets more difficult as the
geometry becomes more complex. Also, to accurately capture flow phenomena such
as turbulence, fine grids are required at the region of interest leading to an increased
number of mesh elements. This results in a large system of non-linear equations
requiring large computing memory. However, with the ever-increased computing
capability available today, a lot of complex flow problems can now be solved.

Commercial CFD codes such as FLUENT, PHOENIX, CFX, CFD++ and Star-CD
have also developed to a point that the user need not be an expert on CFD in order to
successfully put it to use. The end-user must however have a good knowledge of fluid
dynamics for a successful simulation with credible results to be obtained. Therefore,
used correctly, CFD codes can reduce time and cost of experiments in product
development or process improvement.

The rest of the chapter discusses the various areas of CFD in the following order:
• Governing equations
• Grid generation techniques
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 31

• Boundary conditions that define a CFD problem

• Solution algorithms and convergence criteria
• Basic background on the CFD solver used (FLUENT)

2-2.2 Governing equations

The governing equations of fluid behaviour are given in equations (2.24) and (2.25)
[21], [22]. These equations are given for compressible flow, but can be easily
simplified for incompressible flow. In the Eulerian system, the particle derivative is
described as follows:
∂ v
= + (V ⋅ ∇ )
D
(2.24)
Dt ∂t

where:

(V ⋅ ∇ ) = divV = ∂u + ∂v + ∂w .
v v
(2.25)
∂x ∂y ∂z

This particle derivative will be used in the sections to follow to present the Navier-
Stokes equations in conservative form.

2-2.3 Conservation of Mass

The equation for conservation of mass in conservative form is given as:

∂ρ
( )
r
+ ∇ ⋅ ρV = 0 . (2.26)
∂t
r
where ρ is the density and V is the vector velocity of the fluid.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 32

2-2.4 Conservation of Momentum

The equations for conservation of momentum in the three Cartesian directions are
presented.

∂ ( ρ u ) ∂ ( ρ u ) ∂ ( ρ uv ) ∂ ( ρ uw )
2

+ + +
∂t ∂x ∂y ∂z
∂p ∂  r ∂u  ∂   ∂v ∂u  
=− +  λ∇ ⋅ V + 2 µ  +  µ  +   + (2.27)
∂x ∂x  ∂x  ∂y   ∂x ∂y  
∂   ∂u ∂w  
µ  +  + ρ gx
∂z   ∂z ∂x  

∂ ( ρ v ) ∂ ( ρ uv ) ∂ ( ρ v ) + ∂ ( ρ vw)
2

+ +
∂t ∂x ∂y ∂z
∂p ∂   ∂v ∂u   ∂  r ∂v 
=− +  µ  +   +  λ∇ ⋅ V + 2 µ  + (2.28)
∂y ∂x   ∂x ∂y   ∂y  ∂y 
∂   ∂w ∂v  
µ  +  + ρ g y
∂z   ∂y ∂z  

∂ ( ρ w ) ∂ ( ρ uw ) ∂ ( ρ vw ) ∂ ( ρ w )
2

+ + +
∂t ∂x ∂y ∂z
∂p ∂   ∂u ∂w   ∂   ∂w ∂v  
=− + µ +  + µ +  + (2.29)
∂z ∂x   ∂z ∂x   ∂y   ∂x ∂z  
∂  r ∂w 
 λ∇ ⋅ V + 2 µ  + ρ gz
∂z  ∂z 

These equations can be rewritten as a single vector equation using indicial notation:
v
D ρV v ∂   ∂vi ∂v j  v
= ρ g − ∆p +  µ  +  + δ ij λ divV  (2.30)
Dt ∂x j   ∂x j ∂xi  
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 33

2-2.5 The Energy Equation

The energy equation, which in essence is the first law of thermodynamics, is given in
its most economic form as follows:
D p  Dp ∂u
ρ e +  = + div ( k ∇T ) + τ ij' i (2.31)
Dt  ρ  Dt ∂x j

where the viscous stresses are given by the stress tensor:

 ∂ui ∂u j 
τ ij' = µ  + . (2.32)
 ∂x j ∂xi 

2-2.6 Two-Equation Model: k-ε Model

Engineering flow problems encountered in CFD are often turbulent, thus requiring
more effort to solve. Turbulent flows are described by disorderly (or ‘chaotic’) motion
of fluid particles in space and time. The degree of unsteadiness of the flow is often
subject to the complexity of the geometry. Flow fluctuations bring into contact
particles of differing momentum resulting in velocity reductions due to viscous
effects; subsequently the kinetic energy is reduced [23]. This process is said to be
dissipative.

Turbulence also increases mixing of particles; thereby differing concentrations of

conserved quantities are mixed. This process is called turbulent diffusion. Turbulence,
depending on the application, can have desirable effects, for example; chemical
mixing and increased wall heat transfer coefficients for better heat transfer rates than
in laminar flow. In other applications, turbulence is required though it may lead to
losses. Such an application is flow into the combustion chamber where mixing of fuel
and air is desired, but because of turbulence, flow velocity is reduced.

Most compressible viscous flow is turbulent, thereby increasing the complexity of the
governing equations. Direct numerical simulation (DNS) of turbulence is also
computationally expensive; therefore the alternative is to model the phenomenon.
Turbulence modelling is based on the principle that unsteady and irregular high-
Reynolds-number flow averaged over time is steady and predictable. This principle
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 34

leads to a set of non-linear differential equations known as the Reynolds-Averaged

Navier-Stokes (RANS) equations. For high-Reynolds number compressible flow;
viscous and turbulent effects are only significant near the walls, as a result the
turbulent models need only be applied to a small section of the computational domain.

Turbulent models comprise, in order of complexity, among others the zero-equation

model, one-equation model (turbulent kinetic energy), two-equation model (e.g., k-ε
model), Reynolds stress model (RSM) and large eddy simulation (LES). Direct
numerical simulation is the highest level of turbulence simulation, as it requires the
exact solution of the full Navier-Stokes equations without averaging or
approximation. Since the cost of simulation is directly proportional to Re3 [24] and
also due to computer capability limitations, DNS is currently only possible for flows
of low Reynolds (Re) numbers and simple geometrical domains.

As there are many suggested turbulence models today, only the best-suited model for
each application need be used and for this study the k-ε model proved sufficient for
the simulations. The k-ε model is a two-equation model coupling the turbulent kinetic
energy equation with the equation for modelling the rate of change of dissipation (ε).
The resulting two models are given below for fully elliptic high-Reynolds-number
flow [21].

Dε ∂  vt ∂ ε  ε ∂u i  ∂u i ∂u j  C 2ε 2
≈   + C1vt  +  − (2.33)
Dt ∂ x j  σ ε ∂x j  k ∂x j  ∂ x j ∂ xi  k

Dk ∂  vt ∂k  ∂u i  ∂u i ∂u j 
=   + vt  +  −ε (2.34)
Dt ∂x j  σ k ∂x j  ∂x j  ∂x j ∂xi 

where σk = vt/vK and σε = vt/vε are effective “Prandtl numbers” relating eddy diffusion
of k and ε to the momentum eddy viscosity. The eddy viscosity is modelled as
follows:
Cµ k 2
vt ≈ (2.35)
ε
µt
where vt =
ρ
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 35

Recommended values for attached boundary layer calculations of the five empirical
constants are given in Table 2-3.

Table 2-3: Recommended values for empirical constants [21][24]

Cµ C1 C2 σk σε
0.09 1.44 1.92 1.0 1.3

The values in Table 2-3 are however, not universal and need to be modified for
different problems such as jet, wakes, and recirculating flows. The combination of the
equations (2.33) to (2.35), empirical constants in Table 2-3, continuity (2.26) and the
momentum relations (equations (2.27) to (2.29)) yields a system of equations that are
simultaneously solved with turbulent closure to as the k-ε model. Near the wall, the k-
ε model is not valid and thus a curve-fit incorporating the laminar sublayer (law of the
wall) and the logarithmic layer (log law), known as a wall function, is used. The k-ε
computations begin at a certain distance yp away from the wall. The log-law belongs
to a class of models known as wall functions.

2-2.7 Near-Wall Treatment

Due to the viscous and sub-layer damping effects in the turbulent boundary layer, the
Reynolds number is significantly lower in this region. At high Reynolds number
flows, the k-ε model requires a very fine grid at the turbulent boundary layer to solve
for the flow properties in this region. This is due to the fact that the boundary layer is
very thin at high Reynolds numbers. Due to the computational effort a fine grid
requires, wall functions are used near the wall.

The relations given in (2.36) are for the velocity profile in the turbulent boundary
layer [24].
n + n+ ≤ 5
+ vt 
u = = 1 (2.36)
κ
( )
uτ  ln n + + B n + > 30
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 36

where vt is the mean velocity parallel to the wall and uτ is the shear velocity given
by:

τw
uτ = (2.37)
ρ

where τw is the shear stress at the wall, κ is the von Karman constant (κ = 0.41) and B
is an empirical constant related to the viscous sublayer (B = 5).

In equation (2.36), n+ is the dimensionless distance from the wall given by

ρu τ n
n+ = (2.38)
µ

In most literature n+ is also referred to as y+. The log-law is valid for: 30 < y+ < 60.

Figure 2-21: Velocity profile as function of distance normal to the wall in the
turbulent boundary layer [24]

Figure 2-21 illustrates the velocity profile as a function of distance normal to the wall.
Dashed lines are from corresponding equations whereas the solid line constitutes
averaged experimental data.

2-2.8 Grid generation and GAMBIT

The grid generation process or meshing involves dividing the flow domain into
smaller control volumes over which the discretised Navier-Stokes equations are
solved. Grid (mesh) types can be classified into two categories namely; structured and
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 37

unstructured grids. The mentioned types of grids find use in different applications and
are used in the meshing process in this study.

Structured grids consist of grid lines with a characteristic of not crossing or

overlapping. The position of any grid point is uniquely identified by a set of two (2-D)
or three (3-D) dimensional indices, e.g., (i, j, k). Unstructured grids make no
assumption about any structure in the grid definition and usually consist of triangular
(tetrahedral (tet) in 3D) elements.

A numerically generated structured grid or mesh, is understood here to be the

organised set of points formed by the intersections of the lines of a boundary
conforming to a curvilinear coordinate system. The prime feature of such a system is
that some coordinate line (surface in 3D) is coincident with each segment of the
boundary of the physical region.

The use of coordinate line intersections to define the grid points provides an
organisational structure that allows all computations to be done on a fixed square grid
when partial differential equations of interest have been transformed so that the
curvilinear coordinates replace the Cartesian coordinates as the independent variables.
This grid frees the computational simulation from restriction to certain boundary
shapes and allows general flow solvers to be written in which the boundary shape is
specified simply by input.

Grid generation for the purposes of this research takes place in FLUENT’s [25] pre-
processor, GAMBIT [26]. GAMBIT is a versatile pre-processor that can support a
large variety of commercially available computer-aided design (CAD) platforms. Raw
geometry can be imported from these CAD packages into GAMBIT where it is
operated on (i.e., the necessary simplification for CFD purposes are made), in
preparation for meshing.

Various meshing schemes are available in GAMBIT and are used where suitable in
the grid generation process. Most of these schemes are unique to GAMBIT and are
not necessarily documented, as they may be modifications of existing meshing
schemes. GAMBIT allows the user to specify any volume to be meshed, although the
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 38

shape and topological characteristics of the volume determine which mesh schemes
can be used.

Different meshing schemes such as Hex, Hex/wedge and Tet Hybrid, can be specified
on volumes to be meshed. The Hex mesh is composed of hexahedral elements only
constituting a fully structured mesh. The Hex/wedge mesh comprises mainly of
hexahedral elements with wedge elements where necessary. However, Hex and
Hex/wedge elements do not apply to any shape volume, as opposed to Tet Hybrid
meshes. The versatility of the Tet Hybrid element makes it appreciable for use where
the volume is complex and none of the other meshes can be applied. Solution
inaccuracies associated with the Tet Hybrid element type are, however, more
significant than for any of the other element types. Each element type is associated
with a volume-meshing scheme and only those schemes used in this study will be
discussed.

Cooper meshing scheme

When this scheme is enforced on a volume, GAMBIT treats the volume like a
cylinder with end-caps. In the actual volume, the faces at these locations are
representative of the end-caps and are called source faces as in Figure 2-22. They are
so-called because their face meshes are projected through the volume after certain
mesh operations have been performed by GAMBIT as is illustrated by Figure 2-24 (a)
to (b).
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 39

Figure 2-22: Cooper meshing scheme volume treatment [26]

The Cooper scheme involves, in order of occurrence, the following steps:

1. Meshing the non-source faces

2. Imprinting the source faces on one another
3. Meshing the source faces
4. Projecting the source faces’ node patterns through the volume to produce a
volume mesh

The steps described can be illustrated by the example in Figure 2-23, which shows the
union of a prism, cylinder and a rectangle.

Figure 2-23: Union of cube, a cylinder, and a triangular prism [26]

The various Cooper mesh scheme steps are clearly illustrated in Figure 2-24.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 40

Figure 2-24: Cooper meshing scheme [26]

Figure 2-25 is a typical grid used in this research illustrating the different meshing
schemes used.

T etrahed ral
m eshin g

C o op er m esh ing
schem e

T etrahed ral
m eshin g

Figure 2-25: Illustration of meshing types used in the study

 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 41

2-2.9 Boundary Conditions

Equations (2.26) to (2.31) presented in the preceding section remain the same
regardless of the flow field, i.e., external flow, internal flow, two-phase flow (with
some modifications), etc. In mathematics, any solution to a set of partial differential
equations (PDE’s) requires a set of boundary conditions for closure and the solution
of the governing equations is no exception.

CFD simulations largely depend on the boundary conditions specified, hence correct
boundary specification improves convergence to a correct solution. Incorrect
boundary and initial conditions, however, can give convergence although not to a
correct solution. There is a wide variety of boundary types available in FLUENT [25],
but only those used in this study will be given.

1) Flow inlet and exit boundaries

Pressure inlet: Used to define the total pressure and other scalar quantities at
flow inlets.
Pressure outlet: Used to define the static pressure at flow outlets (and also
other scalar variables, in case of backflow). The use of a pressure outlet
boundary condition instead of an outflow condition often results in a better
rate of convergence when backflow occurs during iteration. Note that when
backflow occurs, this boundary acts like a pressure inlet boundary.

2) Wall and symmetry:

Wall: Used to define a solid-fluid interface where viscous flow is considered,
thus applying a no-slip condition. The boundary condition on a surface
assumes no relative velocity between the surface and the gas immediately at
the surface.
Symmetry: Used to define surfaces at which normal velocity and normal
gradients of all other variables are zero. This boundary type is essential where
the geometry is symmetrical in nature, and only half the domain is modelled.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 42

2-2.10 Solution Algorithms

The governing equations of fluid flow are particularly difficult to solve because of
their non-linear nature. Much work has been done in numerical methods to solve for
these types of equations. Some proven and popular methods worthy of note are
SIMPLE, SIMPLE-C, SIMPLER, QUICK and PISO [27]. These methods are
appreciated because of their robustness when applied to a variety of problems. In this
study, steady-state and transient flows are solved using the SIMPLE and PISO
algorithms respectively.

The acronym, SIMPLE, stands for Semi-Implicit Method of Pressure-Linked

Equations, and describes the iterative procedure by which a solution to discretised
equations is obtained. This method is well suited to steady-state solution computation.
The iterative procedure is the pseudo-transient treatment of the unsteady governing
equations in a discrete form to obtain a steady-state solution. SIMPLE was introduced
by Patankar and Spalding (1972) [27] and was described in 1980 by Patankar [27].
The SIMPLE algorithm has a limitation in that new velocities and corresponding
fluxes do not satisfy the momentum balance after the pressure-correction equation is
solved. As a result, the calculation must be repeated until the balance is satisfied. The
PISO algorithm improves the efficiency of this calculation by performing two
additional corrections namely, neighbour and skewness correction.

PISO, which stands for Pressure-Implicit with Splitting of Operators, is a pressure-

velocity coupling scheme that is part of the SIMPLE family of algorithms. This
scheme is based on the higher degree of the approximate relation between the
corrections for pressure and velocity. In highly distorted meshes, the approximate
relationship between the correction of mass flux at the cell face and the difference of
the pressure corrections at the adjacent cells is very rough. An iterative process is
required to solve for the pressure-correction gradient components along cell faces
since they are not known beforehand leading to the introduction of a process named
skewness correction [24]. Here, the pressure-correction gradient is recalculated and
used to update the mass flux corrections after the initial solution of the pressure-
correction equation is obtained. This process significantly reduces convergence
difficulties associated with highly distorted meshes and allows FLUENT to obtain a
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 43

solution on a highly skewed mesh in approximately the same number of iterations as

required for a more orthogonal mesh. The PISO algorithm is used in this study for
calculation of the unsteady gas behaviour due to its efficient nature and suitability for
transient computations as recommended in FLUENT [25].

2-2.11 Convergence

Convergence of a flow or heat problem can be judged by observing the normalised

residuals. Residuals are numerical imbalances from the solved governing equations
resulting from an incomplete solution during the iterative process. The solution
process can be terminated when the normalised residuals fall below a specified value,
which is generally 10-3. However, in some cases even with the convergence criteria
(as far as normalised residuals are concerned) satisfied, the solution may not
necessarily be a correct one. To avoid such instances, quantities such as mass flow
rate, static pressure and heat flux can be monitored at a location in the flow domain
that is deemed to be important. The monitored quantity is observed until the change
from iteration to iteration is negligible thus ensuring good convergence.

2-2.12 Background on the CFD solver used (FLUENT)

FLUENT is a finite volume (FV) solver that can handle a wide variety of flow
problems such as external flow, internal flow, and two-phase flow. All modes of heat
transfer can also be solved by the CFD code. This code is used for all CFD analyses
throughout this study.

Modern CFD packages are user-friendly with improved user and code interfacing.
Thus understanding the underlying principles of flow is important in order to put the
code to good use. FLUENT has a facility for coding to make repetitive simulations, as
in an optimisation loop, more efficient and hence save time. This facility makes use of
a journal file, which is a file containing a list of text commands that set-up a CFD
model and runs the simulation. Post processing can also be performed with the use of
these journal files.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 44

The code developers validate and verify chosen simulated flow cases (mostly from
industrial users) against experimental data, consequently ensuring a validated code for
users. Thus, the code is very suitable for a wide-variety of industrial problems
including the one that is the subject of this study, as will be seen in the chapters that
follow.

2-2.13 Conclusion

As discussed in this section, the solution to the Navier-Stokes equations cannot be

determined analytically, hence it is solved at discrete points in the flow domain by an
iterative process. The various important aspects of grid generation process were also
discussed as well as boundary condition specification. Solution algorithms used in
solving the governing equations were briefly discussed as well as the solution
convergence. The code, FLUENT, used in the study was also discussed where its
journal file function’s efficiency was highlighted in being the main point making
FLUENT suitable for optimisation problems. Mathematical optimisation and its
application to CFD is discussed in the next section.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 45

2.3 MATHEMATICAL OPTIMISATION

2-3.1 Introduction

Optimisation is a process that involves changing parameters of a design in order to

arrive at a desired objective. One approach in design that is commonplace is changing
the design until an improvement is achieved. However, this “optimisation” process
based on intuition and experience does not always yield desired results. This is mainly
due to the fact that some design objectives may cross-influence one another and
intuitively manipulating these objectives may prove to be an impossible task. This has
led to the development of mathematical optimisation techniques to complement the
design process. In this study, these techniques are applied to the airbox design in order
to achieve a design that would give the maximum airflow through the intakes.

Complex non-linear multidisciplinary optimisation problems are often encountered

and with ever-improving computer performance, it has become feasible to solve these
problems. These problems are, however, often sensitive to design changes and
inherent computer round-off errors, due to memory limitations. These may make it
difficult to obtain the function gradients necessary for finding the optimum.

Response surface methodology (RSM) [28] is one optimisation methodology that is

statistical in nature. RSM constructs smooth approximations to functions in a multi-
dimensional space. By selecting optimally placed designs throughout the design
space, RSM constructs approximate surfaces thereby reducing local noise effects.
This method allows the designer to fully exploit the global information, giving
freedom to try alternative design formulations. For example, the design response can
be fixed while minimising or maximising other design criteria. The latter process can
be reversed, whereby constraints are varied and the response to the changes is
monitored.

Successive response surface methodology (SRSM) [29] is used in this study in an

attempt to maximise the airflow through the four engine intakes, by shrinking the
design space and formulating successive sub-problems thereby “zooming-in” on the
optimal design. Use is made of a software package LS-OPT [30] that has the response
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 46

surface methodology (RSM) as its base optimisation method. A formal introduction to

optimisation and problem formulation is given in the sections that follow.

2-3.2 Theory of optimisation

Mathematical optimisation is used to find the optimal values of selected design

variables to obtain the minimum or maximum of an objective function. In more
complex problems, more than one objective function is necessary, thus making the
process computationally expensive. Design variables, x=(x1, x2,…, xn)T, are
parameters that can vary within a specified design range and are mostly subject to
constraints or bounds.

The general standard mathematical form of the above paragraph is presented as

follows:

minimise f ( x ) , x ∈ R n

subject to
g j ( x ) ≤ 0, j = 1, 2,..., m

and
hk ( x ) = 0, k = 1, 2,..., p

where f, g and h are functions of design variables x=(x1, x2, …, xn)T. f(x) is the
objective function which is the quantity to be maximised or minimised. The functions
g(x) and h(x) are the inequality and equality constraints respectively.

2-3.3 Response surface methodology

The single-most important requirement of the response surface method is the analysis
of design points to determine their response. Since simulation-based problems can be
expensive and time-consuming to evaluate, it is desirable that the design set be chosen
appropriately in order to be as efficient as possible. Use of made of the theory of
experimental design (Design of experiments or DOE), of which LS-OPT uses the D-
optimality criterion [30]. The following formulation of a least-squares problem will
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 47

provide understanding of the working of the successive response surface methodology

(SRSM) that is used throughout the study.

Assume a single response variable y having independent variables x. The exact

relationship between these quantities is given by
y = η ( x) (2.39)

If the functional relationship is now approximated (e.g., polynomial approximation), it

follows that
η ( x) ≈ f ( x) (2.40)

The approximating function f is assumed to be a summation of basis functions:

L
f ( x ) = ∑ b jφ j ( x ) (2.41)
j =1

where L is the number of basis functions φ j used to approximate the model of

interest.

The sum of the square error to be minimised is given by the following

P  
∑ {[ y ( x }
P L

p ) − f ( x p )] 2
= ∑ 
p =1 
[ y ( x p ) − ∑ b jφ j ( x p )]2  (2.42)
p =1 j =1 

where P is the number of sampling points, y is the exact functional response at the
sampling points xj as given in equation (2.42) and b = [b1 , b2 , K , bL ] are unknown
T

constants to be determined whose solution is

b = ( X T X ) −1 X T y = CX T y (2.43)

where X = [ X uj ] = [φ j ( xu )] . (2.44)

In a linear approximation, used in the study, the relative value of each coefficient bi
represents the relative importance of the corresponding variable xi. Therefore, a
variable with a small value of b j and/or a large confidence interval becomes a
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 48

candidate for screening; hence only significant variables are retained for the
optimisation cycle.

The number of simulation points, P, is selected using the formula

P = int[1.5(n + 1)] + 1 where n is the number of selected design variables. To obtain a
best estimate of the coefficients bi in the vector b, the D-Optimality criterion is
employed to determine the location of the P simulation points.

The successive response surface method (SRSM) constructs linear response surfaces
on a subdomain from the subset of sampling points selected using the D-optimal
criterion. The resulting successive sub-problems are solved with a multi-start variant
of the leapfrog dynamic trajectory method, LFOPC [31].

The subdomain (or subregion) size is adjusted every iteration as the optimum is
approached by shrinking and panning as seen in Figure 2-26. Design points falling
outside of the subdomain space are handled by constructing an auxiliary problem to
minimise infeasibility and bring the design within the feasible design region.

pan zoom pan & zoom

x1 rU,0
subregion(1) x1 range r1 (0)
x (0) x (0)

x1 rL,0 x (1)
x (1)
x (0) = x (1)
range r1 (1)
(2)
subregion

(a) x2 (b) x2 (c) x2

Figure 2-26: Progression of subregion in SRSM: (a) Only panning, (b) Only
zooming and (c) a combination of panning and zooming [30]

In Figure 2-26, x(0) is the starting design point that is at the centre of the design
region. The user specifies the initial design range value ri(0) that is used to determine
the lower and upper bounds of the design subregion, (xirL,0), xirR,0)). The size of each
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 49

successive subdomain is established through the use of a contraction and a zoom

parameter.

2-3.4 Computational flow optimisation

The emergence of affordable and powerful computers has encouraged the growth of
what is called computational flow optimisation (CFO). This is a field of
computational science where computational fluid dynamics (CFD) and mathematical
optimisation are used in tandem to find solutions to complex flow problems. These
are largely simulation-based problems that tend to be cumbersome and computational
intensive to find solutions to.

Numerous CFO problems have been successfully tackled through the use of different
optimisation techniques such as gradient-based methods, neural networks and
successive response surface methods. Craig et al, [29], [32]-[36], and Kingsley [37]
have also successfully used CFO to find solutions to problems ranging from reduction
of head injury during a car crash [29], to the reduction of automotive pollution in an
urban environment [35].

When describing an engineering flow problem as an optimisation problem, care must

be exercised in the selection of the objective function as well as the design variables.
Insight is required in this regard so as to avoid over-defining a problem by, for
instance, selecting too many design variables or constraints. The more design
variables involved, the more computationally expensive it becomes to complete the
optimisation, thus variables with little effect on the objective function should be
ignored in the optimisation. The following steps are general steps of a simulation-
based CFO problem using SRSM as the optimisation technique.

• Specify the design variables and constraints.

• Specify optimiser settings, e.g., order of approximation, affordable number of
points.
• Simulate the base (initial) case objective function value in CFD. The optimiser
gives suggested designs by perturbing the initial the design parameters within
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 2: Literature Study 50

the design space and the function values of the suggested designs are
evaluated.
• Build a surface function through the obtained objective function values and
then find the minimum point on the surface by using the LFOPC optimisation
algorithm. The optimiser also gives predicted responses at the optimal design.

As the design space is shrunk, the above process is repeated until the objective
function satisfies specified convergence criteria. The objective function can be subject
to numerical noise that can be caused by grid resolution, data extraction from the
completed solution or computer round-off errors. Therefore the SRSM is ideal for use
in simulation-based problems (e.g., CFD) because of its robustness.

2-3.5 Conclusion

This section provided background on mathematical optimisation and one of the

methods, the successive response surface method (SRSM) implemented in LS-OPT,
was presented. The SRSM combined with CFD, together making up a CFO system,
forms the basis of the optimisation study in the chapters that follow.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 51

CHAPTER 3 : Intake Modelling

3.1 PROBLEM DEFINITION AND APPROACH

3-1.1 Problem definition

The intake system considered in this study is that of a Toyota Celica 2000cc Double
Overhead Cam (DOHC) engine. The individual intake system components are shown
in Appendix A. The standard production intake (not shown) has been modified for
improved induction by making use of a series of “ram” pipes starting from the lower
part of the front bumper extending all the way to the airbox. Modifications have also
been made to other components of the engine, but are outside of the scope of this
study.

The objective of the exercise is to solve the fluid flow at wide-open throttle (WOT)
through the trumpet, front cover and body manifold (comprising two intake valves;
and a single injector); collectively called the intake (shown in Figure 3-1). Similar
work has been performed before by Craig and de Kock [38], [39], but in this study the
boundary conditions are more detailed (as discussed further in this section) and
therefore are expected to provide more accurate results.

Trumpet

Injector location
Manifold
Inlet valves

Port

Figure 3-1: Rendered 3-D Single intake showing components

 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 52

It is required that from the CFD solution the following be determined:

• The mass flow rate into the head.
• The flow behaviour in the intake due to transient effects.

The engine speed considered for the simulation is 6500rpm and is the speed at which
the input pressure and temperature data are obtained. These input data used at the
manifold outlet are obtained from a simulation for the considered geometry in the
four-stroke engine simulation package EngMod4T [16]. EngMod4T is based on the 1-
D gas dynamics equations with simplifications, as described in section 2-1.5, and
correlations for complex 3-D flow phenomena.

3-1.2 Modelling Approach

The modelling approach taken in this study was of a bottom-up nature, whereby the
system is modelled from the simplest set-up to the most complex arrangement as
limited by computational resources. This was done in three steps as follows; a single
intake without airbox, four intakes also without the airbox and finally the airbox was
included for a more comprehensive analysis. Due to limited computational resources,
only the section from the cylinder head up to the airbox inlet was considered for the
CFD modelling. By excluding all the geometry (i.e., pipes, air-filter, etc.) downstream
of the airbox, and then assuming the flow conditions at the airbox inlet, a lot of time
was saved through modelling. The modelling of air mixed with the fuel at the
injection point is difficult, and thus an assumption was made that the flow medium is
homogenous air. In sections that follow, the same approach to the simulations was
employed with the only difference being the model as discussed next.

Firstly, a converged second-order steady-state CFD solution was obtained by

assuming a pressure drop across the intake. The purpose of the steady-state solution
was to initialise the flow in the intake before the unsteady solution was computed, and
also to obtain the outflow temperature boundary condition.

For the unsteady flow computation, the pressure outlet boundary was changed to a
varying pressure boundary, with the pressure data obtained from EngMod4T
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 53

(described in the next section) used as an input. To ensure periodic flow, five engine
cycles (5x720° crank angle) were solved to ensure cyclic behaviour. Figure 3-2 below
shows the history of the pressure at the inlet valve against flow time for five engine
cycles.

40000

30000

20000

10000
Pressure [Pa]

0
0 240 480 720 960 1200 1440 1680 1920 2160 2400 2640 2880 3120 3360 3600

-10000

-20000

-30000

-40000

Crank angle [°]

Figure 3-2: History of average pressure at inlet valve (Base design)

The broken lines on the plot depict the end of an engine cycle, and it can be seen that
the pressure profiles become similar from cycle to cycle. At this point the flow was
considered periodic and thus the pressure repeatable from cycle to cycle. This can be
referred to as cyclic convergence. For all the CFD simulations that follow, it was
ensured that cyclic convergence was achieved.

The computed unsteady mass flow rate and pressure at the exit of the inlet port are
monitored and written to a file, where after the total mass inducted was computed and
used as the single-most important factor of the CFD output. Where the geometry has
more than one intake, each one was monitored and the combined effect was used as
the final quantifying factor.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 54

3-1.3 EngMod4T

During an engine cycle, the pressure at the inlet valve is unsteady due to pulsing and
other effects. The time-variant pressure data are obtained from the four-stroke engine
simulation package, EngMod4T [16], which simulates a full engine cycle for a
specified number of cylinders. Input to the program includes parameters such as the
exhaust system configuration, compression ratio, inlet/exhaust valve geometry, valve
timing, firing order, etc., that fully define a combustion engine’s configuration and
operation.

Table 3-1 lists some of the input data used for EngMod4T. A more complete table
(Table B-1) can be seen in Appendix B. The exhaust configuration with the
dimensions required for input into EngMod4T is also shown in Figure B-1 (Appendix
B).

Table 3-1: Summary of input data into EngMod4T

Bore 86.00 mm
Stroke 86.00 mm
Conrod length 148.00 mm
Compression ratio 13.5:1
Exhaust temperature 800 °C
Firing order 1-3-4-2
Valve open 0 – 0.007213 s
Valve close 0.007213 – 0.018593 s
Engine cycle time 0.018593 s

Figure 3-3 and Figure 3-4 below show the simulated pressure and temperature
respectively, as given by EngMod4T, at the inlet valve during the valve open period.
Although the outputs from the program are given with respect to the flow time, they
are plotted against crank angle for easier interpretation.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 55

50000

40000

30000
Pressure [Pa]
20000

10000

0
0 30 60 90 120 150 180 210 240 270 300
-10000

-20000

-30000

Crank angle [°]

Figure 3-3: Pressure at the inlet valve as simulated by EngMod4T (valve open)

The temperature boundary condition is only used when reverse or backflow occurs at
the inlet valve. Otherwise, when flow enters the inlet valve (i.e., leaves the
computational domain), the temperature at this boundary is calculated from the
interior of the domain. Once the valve closes, an adiabatic wall boundary condition is
applied at the inlet valve boundary.

400

350

300
Temperature [°C]

250

200

150

100

50

0
0 30 60 90 120 150 180 210 240 270 300
Crank angle [°]

Figure 3-4: Temperature at the inlet valve as simulated by EngMod4T (valve

open)
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 56

Note that the same pressure and temperature data, i.e., Figure 3-3 and Figure 3-4, are
used throughout the study for all the model configurations. The motivation for using
the same data was obtained though a sensitivity study conducted in EngMod4T, as
shown in Figure 3-5. As shown in the figure, the pressure is only slightly sensitive to
changes in operating intake geometry. The pressure was observed for large and small
volume airboxes (volume determined from constraints of the optimisation study in
section 4-2.2), as well as for the case where the intakes are open to the atmosphere,
i.e., no airbox.

40000 Small Airbox - 19 L

No Airbox
Large Airbox - 25 L
30000

20000
Pressure [Pa]

10000

0
0 25 50 75 100 125 150 175 200 225 250 275 300
-10000

-20000

-30000
Crank Angle [°]
Figure 3-5: Sensitivity of inlet valve pressure to intake geometry changes

From the above results, it was decided to use the airbox pressures, as the simulations
are conducted mainly on geometry containing airboxes. It is believed that the error
being made by assuming the same pressure history is negligible.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 57

3.2 SINGLE INTAKE WITHOUT AIRBOX

3-2.1 Grid Generation

The trumpet and the manifold were created in GAMBIT [26] from supplied technical
drawings. The port, however, had to be scanned because of its complexity. Supplied
scanned data (in DXF format, later converted to IGES) of the inlet port was used to
generate the 3-D CAD drawing. Aztech CNC Manufacturing Solutions obtained the
data by scanning an inverse of the silicon mould of the inlet port using a Cyclone
Series II scanner from Renishaw [40]. The software used to get the scanned data to
surfaces is Paraform 2.1 [41]. The surfaces were then exported to GAMBIT, where a
volume was generated by adding the trumpet and manifold. Refer to Figure 3-1 for
the rendered 3-D illustration of the intake geometry.

The grid of the volume was generated in GAMBIT, where the grid was expanded at
the trumpet inlet to allow for the proper boundary specification. Due to symmetry,
only half of the volume was meshed. The expanded half-spherical domain was added
to minimise the influence of the inlet boundary condition on the flow in the trumpet.
In section 3.4, a more realistic airbox is used together with multiple intakes
(cylinders). The number of grid cells was in the order of 216000. Different views of
the grid can be seen in Figure 3-6, Figure 3-7 and Figure 3-8. The grid was then
imported into the Computational Fluid Dynamics (CFD) package, FLUENT for case
set up and solving for the flow.

Figure 3-6: Overall grid of single intake model

 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 58

Figure 3-7: Zoomed-in view of intake

Where possible, a structured mesh was used like in the manifold section. The mesh
consisted mainly of tetrahedral elements mostly in the domain representing the
atmosphere. The mesh gets denser from the outside of the domain towards the trumpet
vicinity so as to capture important flow features especially during unsteady
simulations. Tetrahedral elements were also used in the inlet port section because of
its irregular geometry as shown in Figure 3-8.

Figure 3-8: Close-up of surface grid at the inlet port

 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 59

The mesh at the inlet port is dense so that the flow is this area is resolved to finer
detail than in other areas of the computational domain. Since tetrahedral cells are
known to give less accurate results than hexahedral cells, the use of fine cells provides
increased accuracy at the expense of a high cell count. In the boundary layer area,
more tetrahedral cells are required to get same accuracy as lesser number of hex cells.

3-2.2 Boundary Conditions

Two pressure boundaries were specified at the inlet and outlet of the intake with a
specified pressure drop for the initial steady-state CFD simulation as seen in Figure
3-9. The pressure inlet boundary is used to represent atmospheric pressure. It is for
this reason that it is placed far from the trumpet mouth to avoid boundary
interference. At the pressure outlet boundary, a pressure drop is specified to initiate
intake flow.

Symmetry

Pressure inlet

Pressure outlet (inlet valve)

Figure 3-9: Rendered view showing boundary types by colour

3-2.3 Results and Discussion

The simulation performed in this section will be used mainly for comparison purposes
as the complexity of the simulations increases. Therefore only the mass flow rate
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 60

result will be discussed, as the pressure and velocity results are more relevant in the
full intake simulations in the sections that follow.

Mass flow rate

The mass flow rate at the inlet valve (outlet of CFD domain) is shown below in Figure
3-10. Negative mass flow corresponds to flow into the head. Therefore, ideally, the
entire flow rate plot should be below zero, implying no backflow. From Figure 3-10,
it can be noted that there is evidence of backflow as seen by the positive flow rate
section of the graph towards the end of the intake cycle. The losses due to this
backflow are undesirable and thus the valve opening could be timed such that it closes
when the pressure is at its peak (about 250° crank angle), to retain the high-pressure
pulse inside the cylinder.

0.05

0
0 30 60 90 120 150 180 210 240 270 300
Mass Flow rate [kg/s]

-0.05

-0.1

-0.15

-0.2
Crank Angle [°]

Figure 3-10: Mass flow rate at the inlet valve from the CFD simulation

The mass flow into the single intake can be extrapolated to give the total mass flow
into a full engine intake system. This results in a total inducted air mass of 2.751 g.
However, an assumption that the flow is equal for all intakes was necessary in order
to make the extrapolation. This assumption is however flawed since there it does not
account for the different flow losses that might occur from intake to intake.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 61

3-2.4 Conclusion

This section showed results for a base case of the intake geometry of a high
performance rally engine. The main result of the study is the mass flow rate, which
showed that further improvements could be made by reducing the amount of
backflow. The valve opening timing could serve as a good starting point in increasing
the mass flow rate into the cylinder. Furthermore, ensuring a smoother transition
between the injector point and the port body could reduce flow losses. To obtain more
accurate results, all four intakes will be considered in the next section (four-cylinder
engine) to gain insight into the effect that the intakes have on one another.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 62

3.3 FOUR INTAKES WITHOUT AIRBOX

In order to understand the impact of an airbox on engine performance, the next

configuration considered is that of an engine with the trumpets open to atmosphere.
This is necessary since some engines perform better without an airbox at some speeds
and also because an airbox has a characteristic of restricting wave expansion and
travel. This characteristic can be positive at higher engine speeds where wave
reflection against the airbox walls assists in charging because of the shorter
wavelengths.

The geometry considered in section 3.2 is extended to four intakes due to the expected
influence of the intakes on one another. Also, a volume is added above the intakes to
allow for pulses to travel into the atmosphere. Due to the firing order being taken into
account, intake charge “robbing” is expected. Charge “robbing” is the phenomenon
that occurs when the inlet of an intake opens while one of the other intakes is open.
This leads to a reduction in available charge mass to the latter. Three dimensional
flow effects as well as intake robbing are expected to change the mass flow rate as
predicted in the previous chapter.

3-3.1 Description of Problem

The gas flow at wide-open throttle (WOT) in the intake is solved. It is required to
determine the following:
• The mass flow rate into the head.
• The flow profile in the intake due to pulsing.

The geometry considered is the four trumpets as in the real engine. The firing order
(1-3-4-2) is also taken into account, and is used to determine the order of the valve
openings.

As in the previous section, the engine speed considered for the simulation is 6500rpm
and the same pressure data are also used as an input, although for each intake the
pressure data is unique as given by EngMod4T.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 63

3-3.2 Grid Generation

Due to the asymmetrical arrangement of the four intakes, the full geometry is
modelled. Figure 3-11 shows a solid rendering of the four-intake arrangement. The
grid was also expanded to allow for the implementation of the atmospheric boundary
condition as a mock representation of the surrounding volume of air as seen in Figure
3-12.

88 102 88

Figure 3-11: Solid rendering of four-intake arrangement [mm]

The surface grid is shown in Figure 3-12. Grid cells in the dome geometry (containing
the inlet boundary) get coarser further away from the trumpet because the flow in
these regions is not expected to have much of an effect on the flow in the intake. The
grid contains around 460 000 cells. A close-up of the mesh is shown in Figure 3-13.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 64

Figure 3-12: Overall grid of four intakes without airbox

Figure 3-13: Close-up view of the mesh of the four intakes

 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 65

3-3.3 Boundary Conditions

As in the single intake simulation, two pressure boundaries were specified at the inlet
and outlet of the intake respectively. The steady-state solution is obtained before the
unsteady flow is computed. For the full simulation, the outlet boundary at each of the
ports is changed to a varying pressure boundary during the respective valve-open
period. Thereafter, the boundary is switched to a wall in order to represent the valve-
closed cycle. This boundary change is performed in a sequence corresponding to the
firing order.

Pressure inlet

Pressure outlet (inlet valve)

Figure 3-14: Boundary location

3-3.4 Transient Modelling

The modelling approach described in section 3-1.2 was followed for this simulation.
The valve timing data of the engine are summarised in Table 3-2.

Table 3-2: Valve timing data

Total time
Valve open 0 – 0.0071632s 0.007213 s
Valve close 0.0071632 – 0.01841327s 0.01125007 s
Engine cycle time 0.01841327 s
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 66

It was necessary to obtain the valve timing data of the engine because of the need to
correctly simulate the full engine cycle as close to reality as possible. The
implementation of the valve timing into the simulation is described in detail in the
section that follows.

3.3.4.1 Engine firing sequence in FLUENT

For the full-engine simulation, the correct engine firing order must be carefully
implemented in the CFD solution process. This is achieved in FLUENT by changing
the pressure outlet boundary at each of the inlet ports to a varying pressure boundary
during which the EngMod4T pressure data are enforced for the valve-open period.
Thereafter, the boundary is switched to a wall in order to represent the valve-closed
cycle. The pressure at the inlet valve during the valve-closed period is dependent on
the FLUENT computations. The whole process of boundary changing and
specification is automated by the use of FLUENT’s scripting capability (journal file).
This journal file contains text commands that are read by FLUENT for execution at
user-specified intervals.

3-3.5 Results and Discussion

The mass flow rate into each intake is shown in Figure 3-15. It can clearly be seen
that the ingested mass of air into each intake is not equal. The total mass of air
inducted in each cycle obtained from the simulation is 2.619 g. This resultant inducted
mass is 4.8% less than that obtained from four times of that of the single intake
simulation in the previous section. This, along with Figure 3-15, confirms that indeed
the assumption (from the previous section) made that the four intakes ingest the same
amount of air is flawed.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 67

Mass intake at Individual Intakes

-6.35E-04
1 2 3 4
-6.40E-04
Mass Inducted [kg]

-6.45E-04

-6.50E-04

-6.55E-04

-6.60E-04

-6.65E-04

-6.70E-04
Intake number

Figure 3-15: Mass of air inducted by individual intakes

The atmospheric boundary condition (pressure inlet boundary) reflects pressure waves
travelling towards it (at the trumpet mouth). It is for this reason that the airbox is to be
explored to determine its influence on the resultant mass inducted.

3-3.6 Conclusion

The section considered four intakes, open to atmosphere, for analysis. The results
showed that more detail needs to be included in order to more accurately assess the
performance of the intake system. The airbox with its one inlet and asymmetrical flow
qualities was thus added as discussed in the next section.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 68

3.4 FOUR INTAKES WITH AIRBOX (BASE CASE)

The final implementation of the study, as far as the geometry is concerned, is the
addition of an airbox as part of the simulations to follow. The airbox geometry was
also obtained from Toyota Motorsport and forms a major component of the
parametric study. The original airbox (hereinafter referred to as the base
case/geometry) can be seen in Figure 3-16. To reduce meshing difficulties, some
detail in the geometry as seen in the figure was either removed or simplified in those
areas considered to have little impact on the flow. More pictures of the complete inlet
system, though not modelled, can be seen in Appendix A. The problem setup is
similar to the case without the airbox (previous section).

Figure 3-16: Original airbox geometry

The results obtained from this base case form the basis for comparison in the
optimisation study. It therefore is important that the predicted flow is accurate as far
as the CFD simulation is concerned.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 69

3-4.1 Grid Generation and Boundary conditions

In order to closely represent the intake port, simplified valve stems were added since
they reduce the flow area at the outlet and also offer increased flow resistance, which
could impact on the mass flow rate.

The mesh, containing around 400 000 elements, was generated in FLUENT’s pre-
processor GAMBIT [26] (see an example of the Gambit journal file in Appendix D).
The overall grid of the geometry can be seen in Figure 3-17. Note how the grid cells
are concentrated around the intake and the port geometry; this is mainly due to
interesting flow features that are expected to occur and a refined mesh in these areas
increases the chance of capturing these features.

Figure 3-17: Views of overall grid

 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 70

Two boundaries were specified at the inlet (pressure-inlet boundary in blue) and outlet
(pressure-outlet boundary in red) of the intake as seen in Figure 3-18. The location of
the inlet pressure boundary has some influence on the pressure inside the airbox and
intakes. However, the pressure differences due to boundary location have a negligible
effect on the power output of an engine as shown by Maynes et al [1]. This led to a
decision to place the boundary at the airbox inlet.

Pressure inlet

4
Pressure outlet (inlet 3
valves) 2
1

Figure 3-18: Rendered view showing boundary types by colour (intakes and
airbox)

A pressure drop is specified for the initial steady-state CFD simulation. The purpose
of the steady-state solution is to initialise the flow in the intake before the unsteady
solution is computed, and to obtain the outflow temperature boundary condition for
air density calculation. For the unsteady flow computation, the pressure outlet
boundary is changed to a varying pressure boundary, with the pressure data from
EngMod4T used as an input.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 71

3-4.2 Steady-state modelling and experimental validation

3.4.2.1 Experimental validation

As per accepted engineering practice, where possible, experimental tests are

performed to compare to simulation results so as to better judge the confidence level
in the simulation results. This validation route was also taken in this study, although
the experimental tests were limited to steady state since facilities to perform unsteady
flow tests were not available. Nevertheless, the steady-state results give a very good
idea of the accuracy of CFD simulations as will be shown by the results.

Experimental flow rate measurements were taken on a flowbench (SuperFlow-110

[42]), at a constant pressure drop of 1993.7 Pa (8 inch manometer reading), for
various valve lifts, as well as the case where the valves were removed completely to
allow for maximum flow. The experimental set-up can be seen in Figure 3-19, and
more detailed illustrations are presented in Appendix C, Figure C-1 to Figure C-4.
The valve lift measurements were taken at increments of 0.5 mm from 0 to 12mm
valve opening. Note, however, that for the transient simulations that follow, only the
no-valve case is used.

Airbox mounted on
flowbench.

Computer for data

capturing

SF 110 Flowbench

Figure 3-19: Flowbench experimental set-up

 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 72

CFD simulations of the experimental set-up were performed for comparison purposes
at four valve lift points (3, 6, 9 and 12 mm) were selected. The CFD model in Figure
3-18 was modified by the addition of simplified valves, at specified lifts, as shown in
Figure 3-20. Also shown in the figure is the location of the pressure outlet boundary
as an annular ring.

Pressure outlet

Figure 3-20: Full geometry with simplified valves added and valve-lift pressure
boundary specification

Figure 3-21 shows the added mesh volume (in red) to accommodate the valve lift
geometry. The added volume is one made up of structured elements thus not
increasing the total cell count by much.

Figure 3-21: Mesh extension for valve lift simulations

 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 73

3.4.2.2 Results of validation

The CFD and flow bench test results show satisfactory agreement as depicted in
Figure 3-22. For comparison purposes only, the no-valve geometry results are plotted
at a “valve-lift” point of 13 mm.

Flow Rate vs Valve Lift: Experiment vs. CFD

0.08

0.07

0.06
Flow rate [kg/s]

0.05
Test Valve-lift Results
CFD - no-valve
0.04
Test - no-valve
CFD Results
0.03

0.02

0.01

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Valve Lift [mm]

Figure 3-22: Comparison of CFD and flowbench results

The results for the valve lift cases differed by a maximum of 14.1% at a valve lift of
6mm and a minimum of 6.4% at 12mm valve lift. However, the results of the no-
valve case differed by only 1.6%, which is in excellent agreement. The deviation in
the valve lift cases can be attributed to the simplified valve, boundary placement as
well as the “missing” cylinder volume.

3-4.3 Transient Modelling

The FLUENT journal file for setting and running the transient case is listed in
Appendix D. It can be seen in Figure 3-23 that the pressure during the valve-closed
period as calculated by EngMod4T differs only from the CFD by the amplitude of the
waves. Even though the two simulation packages yield very similar results, there is
information such as flow patterns that only FLUENT can provide because it solves
the 3-D Navier-Stokes equations and also determines flow losses due to geometry that
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 74

are not predicted in EngMod4T. The principal losses in the intake are due to the
friction at the walls and also the fluid interactions such as swirling, vortices, etc.

Static Pressure at inlet valve

CFD vs EngMod4T on Intake 1
50000
EngMod4T
40000
CFD
30000

20000
Pressure [Pa]

10000

0
0 100 200 300 400 500 600 700 800
-10000

-20000

-30000 Valve open

Valve closed
-40000
Crank Angle [°]

Figure 3-23: Comparison of CFD and EngMod4T average static pressure

prediction at the inlet valve

The result of the deviation between the two results in Figure 3-23 is that the CFD
result is not fully cyclic within the 720° crank angle range because of the enforced
varying pressure boundary. The only way to remedy this would be to couple the CFD
and EngMod4T simulations continuously, or to replace the EngMod4T-generated
pressure boundary condition in FLUENT with a more accurate prediction of valve
motion and combustion in the cylinder (which is outside the scope of this study).

Figure 3-24 depicts contours of static pressure at different flow times. The different
shades on the walls indicate the strength of the pressure pulse, which determine the
amount of flow inducted in the system. In Figure 3-24 (a), the inlet valves for cylinder
number two are open as witnessed by the light shade at the inlet port mouth. By
observing time progression of the pressure contours, much information can be
extracted and used for flow quality improvement. Also, areas of pressure pulse
interaction become visible thus giving insight into the flow.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 75

(a) (b)

Figure 3-24: Pressure contours on walls at different flow times

As the flow enters the airbox, it is directed to the top, as seen in the path lines Figure
3-25, where some loss of kinetic energy occurs because of the collision with the roof
of the airbox. Although some energy has been lost, most particles stay attached to the
airbox walls and eventually get drawn into the intake, maintained by the suction
pressure at the inlet valves. Particles can be seen moving from the lower end of the
airbox and around the trumpet indicating some loss in energy translating to lower
mass flow at the intake. Also shown Figure 3-25, is a close-up view of the path lines
in the inlet port.

Figure 3-25: Instantaneous velocity path lines in airbox and inlet port (Intake 1)
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 76

Figure 3-26 depicts the predicted mass flow rate in intake number one. The mass flow
rate profile is similar for the other three intakes. The effect of the reverse flow (around
250° crank angle) is minimal because the integrated area under the graph above the
zero mass flow rate axis is relatively small compared to area below the axis. The
integrated area below the axis indicates the amount of mass flow inducted over the
crank angle range. Although this mass flow rate is mostly determined by the enforced
transient pressure boundary at the inlet valve, factors such as frictional losses due to
geometry, pressure pulses travelling through the airbox and cylinder-to-cylinder
charge robbing also contribute to the total mass flow rate.

Mass flow rate at Intake 1

0.05

0
0 60 120 180 240 300 360 420 480 540 600 660 720
Mass flow rate [kg/s]

-0.05

-0.1

-0.15

-0.2
Crank angle [°]

Figure 3-26: Mass flow rate at the inlet valve from the CFD simulation

Shown in Figure 3-27 is the amount of air that each intake inducts over one engine
cycle. The distribution of inducted air is not equal, with more air inducted by the two
middle intakes (Intakes 2 and 3). Note the negative axis used for mass inducted. The
variation is mostly due to the airbox shape and design.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 77

Mass intake at Individual Intakes

-6.10E-04
1 2 3 4
-6.20E-04
Mass Inducted [kg]

-6.30E-04

-6.40E-04

-6.50E-04

-6.60E-04

-6.70E-04

-6.80E-04
Intake number

Figure 3-27: Mass of air inducted at individual intakes over one engine cycle

Figure 3-28 shows the normalised shear stresses on the trumpet and airbox surfaces
over a full engine cycle (normalised to the average value over the engine cycle). Also
plotted in the figure are the y+ values (refer to section 2-2.7) on all intake inner
surfaces. It can immediately be seen that the contribution of wall shear stress in the
airbox to losses is less than that of the trumpet walls. This is expected as the highest
velocities occur in the intake. Also shown on the figure are the points where different
intakes open during operation (boxed numbers). The trumpet wall shear stresses show
little change when valve opening occurs. On the contrary, the y+ values are strongly
influenced by the valve opening and closing sequences.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 78

1.8
Trumpet shear stress
Y-plus
1.6 Airbox wall shear stress

1.4

1.2
Normalised values

0.8

0.6

0.4 2
3 4 1
0.2

0
0 120 240 360 480 600 720 840
Crank Angle [°C]

Figure 3-28: Shear stress and Y-plus plot for base case geometry

The oscillations in Figure 3-28 of y+ are similar to that of the trumpet shear stresses,
except at intake opening points where the y+ values increase before again following
the shear stresses’ oscillatory trend. These oscillations can be attributed to the
pulsating nature of the intake in a four-stroke internal combustion engine.

Adding to the solution confidence, are the actual area-weighted Y-plus values (not
normalised as in Figure 3-28) obtained through the engine cycle as shown in Figure
3-29.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 3: Intake Modelling 79

200

180

160
Y-plus values

140

120

100

80

60
0 60 120 180 240 300 360 420 480 540 600 660 720
Crank Angle [°C]

Figure 3-29:Y-plus values on intake inner walls

The values in Figure 3-29 range from just above 60 to below 200, which falls within
the acceptable y+ range for flow where heat transfer is not of critical importance as is
the case in this simulation.

3-4.4 Conclusion

The modelling and experimental validation of all four intakes with the airbox included
were presented in this section. The CFD and experimental results showed good
agreement and gave enough confidence to proceed with the optimisation study.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 4: Intake Optimisation 80

CHAPTER 4 : Intake Optimisation

4.1 INTRODUCTION

Since the base case results showed that an improvement could be achieved by
reducing the amount of reverse flow, various airboxes were used in an attempt to
maximize the mass flow rate. These airbox designs were obtained, not by trial and
error, but through the use of the optimisation software LS-OPT (refer to section 2.3).
The process involves selecting geometrical design variables and then varying them,
mathematically, to obtain geometrical different airboxes and solving for a full engine
cycle and obtaining the total mass flow. The total mass flow is then used as the
objective function that needs to be maximised. The results in this section were
presented at SACAM 2004 [43].
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 4: Intake Optimisation 81

4.2 AIRBOX OPTIMISATION

4-2.1 Candidate design variables and objective functions

4.2.1.1 Candidate design variables

Candidate design variables are possible variables that could be used in the
optimisation process and are initially considered before the final variables are
selected. The selection criterion is primarily based on the impact a design variable
might have on the objective function, but other factors such as ease of manufacturing
or even aesthetics are considered. Considered design variables are shown in Figure
4-1.

Inlet
Offset

Height Throat
diameter

Trumpet
clearance
4
3
2
1
Width
Figure 4-1: Intake-airbox arrangement with variables

Below are candidate design variables that were considered:

• Height on side of intake 1 (height at inlet pipe side remains constant to
accommodate pipe)
• Width
• Trumpet clearance
• Throat diameter
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 4: Intake Optimisation 82

• Offset
• Engine speed (RPM)
• Trumpet length/height
• Engine parameters such as valve seat angles, valve sizes, etc.

Due to the CFD computational expense, not all of the above variables could be
included in the optimisation study, especially investigation of airbox performance at
different engine speeds. Some variables, e.g., engine parameters and trumpet
geometry, were eliminated because of the race rules prohibiting any changes on them.
Nevertheless a sensitivity study on the influence of trumpet height was conducted
using EngMod4T and the results are shown in Figure B-2 (Appendix B), showing that
this variable is important in the performance of the engine. The throat diameter was
excluded because of the manufacturing complexity. The inlet offset was excluded
because of the geometrical modifications that would be required on the rest of the
system to accommodate the offset inlet.

4.2.1.2 Candidate objective functions

The following objective functions were also considered in the optimisation:

• Total mass of air inducted every cycle
• Mass of air inducted by each intake every cycle

The latter was considered because the amount of inducted air by each intake during an
engine cycle is not equal owing mainly to the airbox inlet not being positioned to
evenly distribute the flow. Due to the immense computational effort required to
simultaneously evaluate these objective functions in the optimisation study only the
total mass inducted was considered as the primary objective, while the individual
mass inductions were monitored.

4.2.1.3 Computational considerations

As discussed in section 2.3, the number of experimental points to be simulated is

determined by the following relation, int1.5[( N + 1)] + 1 , where N is the number of
variables. In this case, assuming only one engine parameter is considered, if all design
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 4: Intake Optimisation 83

variables (i.e., N = 8) are considered then 14 experimental points would be simulated

for each iteration. The available computing capacity would not give satisfactory
turnaround times considering a single CFD simulation would take, conservatively,
around 120 hrs (5 days) to evaluate on a Pentium 4, 2.4 GHz with 512 Mb RAM.
Therefore for one optimisation iteration, a total of 1680 hrs (70 days) would be
required.

It must also be noted that the convergence to an optimum solution often takes more
than two iterations, hence only three design variables (Airbox width, trumpet
clearance and airbox height) were chosen. This translates to seven (7) experimental
points requiring 840 hrs (35 days) to evaluate, which was deemed acceptable for the
study.

4-2.2 Mathematical formulation

Since use is made of mathematical optimisation techniques, it is required that the

problem be formulated in formal mathematical terms. The formulation is given below
where the objective function F ( x ) = − f ( x ) is to be minimised. The function f ( x ) is
the amount of airflow inducted into the engine, which is negative in nature hence, the
use of F ( x ) .

Minimise F ( x ) (4.1)
Subject to the bounds in Table 4-1,
Table 4-1: Bounds for optimisation problem
Variable name Design variable Min [mm] Max [mm]
Height x1 100 210
Trumpet clearance x2 60 90
Width x3 0 100

and the constraint to avoid physical interference of the airbox roof with the trumpet
mouths (given in mm):

g1 = 50 + x 2 − x1 ≤ 0 . (4.2)
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 4: Intake Optimisation 84

4-2.3 Results

The mass flow rate is integrated to give the total inducted mass during an engine cycle
using the trapezoidal method of integration. The results are given in Figure 4-2 for the
7 experimental points of each iteration, and they clearly show that the geometry has
an influence on the mass inducted into the engine. An improvement of 6.2% can be
seen at experimental point 3 of the second iteration (Iter-2). Noting that engine
performance is directly proportional to mass ingested, an increase of 6.2% in engine
power is expected. The optimisation process was judged to have reached a
satisfactory solution after Iteration 2 since the results yielded little variance for the
simulated experimental points.
Point with 6.2% improvement over base
case
0.0028

0.00275
M a s s flo w ra te [k g /s ]

0.0027

0.00265

0.0026 Iter-1

Iter-2
0.00255
Last-pt
0.0025

0.00245
Base case
0.0024

0.00235

0.0023
0 1 2 3 4 5 6 7 8 9
Experimental point
Figure 4-2: Results of airbox optimisation study

The final (improved) airbox parameter values, along with those of the base case, are
listed in Table 4-2. Also listed are the results for total mass of air inducted for the two
designs. (See breakdown of individual intakes in Figure E-2, Appendix E).
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 4: Intake Optimisation 85

Table 4-2: Parameter values and mass flow rate results of airbox optimisation

Base Case Improved design

Height [mm] 210 155

Trumpet clearance [mm] 90 84

Width [mm] 240 340

Total mass inducted [g] 2.578 2.748

The original airbox along with the improved airbox is shown in Figure 4-3. The
improved airbox is wider and also has lower end-height (at intake 1). The trumpet
clearance did not change much from the original one and seems to not be as
influential as the other variables.

Height
kept
constant
210

155

84
90

240 340

(a) (b)
Base Case Improved

Figure 4-3: Illustration of the base and optimum airbox designs [mm]
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 4: Intake Optimisation 86

The design variable values are plotted in Figure 4-4 and Figure 4-5 for the two
optimisation iterations. These figures show all the designs that were explored before
the improved design was achieved. The width values are plotted relative to the
starting width (i.e. 240 mm).

Design Variable History Height

Iter 1 Trumpet clearance
250 Width - 240 mm

Base
200
Variable value [mm]

design

150

100

50

0
1 2 3 4 5 6 7
Experimental point

Figure 4-4: Design variables for first iteration

Design Variable History Height

Iter 2 Trumpet clearance
250 Width - 240 mm
Improved
design
Variable value [mm]

200

150

100

50

0
1 2 3 4 5 6 7 8
Experimental point

Figure 4-5: Design variables for second iteration with last point simulated
included (point 8)
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 4: Intake Optimisation 87

Note that in Figure 4-5 (and Figure 4-2), there are eight experimental points.
Experimental point 8 does not form part of the second iteration but was added for
comparison purposes. This point is the last point given by LS-OPT as a conclusion of
the optimisation. As LS-OPT uses response surfaces, this point is the best design
predicted from the surface obtained in iteration 2. The value in Figure 4-3 (green –
last point) is the simulated result corresponding to the design variables of the
prediction.

Although not clearly visible, the figures show that the design space has been reduced
by small margin. Consider the design space in the figures. In iteration 1 (Figure 4-4),
the range is from 0 to 210 mm but this range is reduced to 20 to 200 mm in iteration 2
(Figure 4-5). This indicates that the design space for this particular variable has been
reduced as the optimisation converges by the SRSM algorithm in LS-OPT. The same
region reduction process applies to the remaining two variables. This analysis of the
design history shows the practical application of the optimisation methodology
discussed in section 2-3.3.

4-2.4 Off-design performance

4.2.4.1 Problem layout

Since the airbox was optimised for a single engine speed of 6500 rpm, the
performance at other speeds may be affected, hence simulations using the improved
design were done at two other engine speeds to determine the mass flow rate. The
base case was also simulated at these engine speeds, which were chosen to be 5500
rpm and 7500 rpm.

4.2.4.2 Results and Discussion

EngMod4T was used to simulate the inlet boundary pressure at these engine speeds
for use in the CFD simulations, as shown in Figure 4-6. The 5500 rpm and 7500 rpm
pressure plots have a significant area above the zero pressure axis when compared to
the 6500 rpm plot. This indicates that reverse flow might have a significant influence
on the mass flow results.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 4: Intake Optimisation 88

60000

45000

30000

15000
Pressure [Pa]

0
0 30 60 90 120 150 180 210 240 270 300
-15000
5500 rpm
-30000 6500 rpm
7500 rpm
-45000

-60000
Crank angle [°C]

Figure 4-6: Simulated pressure for different engine speeds

The mass flow rate results for the base and improved airbox simulations are presented
in Figure 4-7 (See also a breakdown of the individual intakes in Appendix E). The
designs have very similar performance at the 5500 rpm point, while at the improved
airbox also performs better than the base airbox at both 6500 rpm and 7500 rpm.

0.0028

0.0027
Base
Optimum
0.0026
Mass flow rate [kg/s]

0.0025

0.0024

0.0023

0.0022

0.0021

0.002
5000 5500 6000 6500 7000 7500 8000
Engine Speed [rpm]

Figure 4-7: Mass flow rate response of base and improved airboxes at different
engine speeds
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 4: Intake Optimisation 89

Although the airbox was optimised only for the 6500 rpm engine speed, it is by sheer
co-incidence that the improved design also outperforms the original airbox at the off-
design points. Although each engine speed might have a different optimum airbox, the
improvement gained shown by the improved airbox design (i.e., improved design at
6500 rpm) at the off-design points is sufficient.

4-2.5 Conclusion

In this section, results of the optimisation of the airbox were presented. By combining
CFD with mathematical optimisation, an improvement on the original airbox
performance was achieved by only changing the geometry of the airbox. Off-design
performance of the improved design airbox was also assessed and an overall
improvement over the original airbox was found.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 5: Conclusions and Recommendations 90

CHAPTER 5 : Conclusions and Recommendations

5.1 CONCLUSIONS

The literature survey conducted in CHAPTER 2 highlighted the importance of proper

engine intake system design in order to achieve different engine characteristics (e.g.,
economy and driveability for everyday driving vs. maximum volumetric efficiency
with little regard for economy for racing purposes). With the amount of research
already gone into various intake configurations and modelling techniques required to
evaluate intake systems, the effort needed to set up the problem approach for this
particular study was reduced immensely.

This study took the bottom-up approach to modelling the intake system of a high-
performance rally engine as presented in CHAPTER 3. This meant firstly modelling a
single intake, then all four intakes excluding the airbox and then including the airbox
as a final step.

Firstly, a single intake was modelled with the assumption that the mass flow rate into
all the four intakes was equal and the flow was symmetrical. The results from this
simulation were mostly for insight purposes and since the model was simple to set up,
it gave the opportunity to have a base starting point as the simulations and model
progressively became more complex. Although important information was gathered
from the simulation, the problem had been over-simplified and the model was thus
extended to include all four intakes but without the airbox.

The results obtained from the four intakes simulation pointed to the importance of
intake interaction and its influence on intake air mass flow. Features such as inter-
intake robbing were observed, and this prompted the inclusion of the airbox, as this
would affect pressure wave interactions with the inner airbox wall. The airbox as
expected, did have an influence on the mass flow rate inducted into the cylinders by
the intakes.
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 5: Conclusions and Recommendations 91

The final implementation of the study was optimisation of the existing airbox design
for the chosen engine speed, as outlined in CHAPTER 4. The investigation required
the selection of airbox geometrical parameters that would be varied, with the aid of
optimisation methods, to change the airbox design and determining the total mass
flow inducted during a full engine cycle. After two iterations, involving exploring
fifteen designs, a final design giving an improvement of 6.2% was obtained.

5.2 RECOMMENDATIONS

As far as recommendations are concerned, an optimisation of the intake geometry

could be performed where several objectives, such as mass flow into each intake or
the intake performance at different engine speeds, are considered. With the ever-
increasing computing capability, the number of design variables can be increased to
allow more complex airbox and intake geometries. These variables may include the
airbox inlet pipe throat diameter, radii of the edges, etc.

Also, experimental unsteady testing of the improved airbox is essential for proper
completion of the study while at the same time validating the optimisation findings.
This should be done at all engine speeds and finally, road-testing should be performed
to evaluate the effect of the optimised design on the actually rally car engine response.
This should be done under the same conditions for both airboxes in order to
objectively judge the influence of the change.

A more appropriate approach to the problem is to dynamically link the gas dynamics
boundary condition and the CFD solution. This requires extensive coding that would
enable the two packages, namely EngMod4T and FLUENT, to exchange boundary
information at each time-step during the unsteady computation. The computed CFD
boundary information would be exported to the 1-D gas dynamics solver and new
boundary data, i.e., pressure and temperature, would be computed and read back into
the CFD solver for use in the next time-step.

This linking would provide increased accuracy as far as boundary data are concerned,
and is done successfully with commercially available software such as Virtual 4-
Stroke as discussed in [1]. By using CFD moving mesh capability and directly
 University of Pretoria etd – Makgata, K W (2005)
CHAPTER 5: Conclusions and Recommendations 92

modelling combustion, the study would do away with the need for a 1-D engine
simulator. This however, is still not feasible with the current available computing
capability.

The study, in spite of its shortcoming of uncoupled computation as discussed, showed

that CFD and mathematical optimisation methods can be successfully married to
arrive at an improved design much more efficiently than with the traditional trial-and-
error method.
 University of Pretoria etd – Makgata, K W (2005)
REFERENCES 93

REFERENCES

[1] Maynes, BDJ, Kee, RJ, Kenny, RG, Fleck, R, Mackey, DO & Foley, L,
(2002), ‘Prediction of Formula 1 Engine and Airbox Performance using
Coupled Virtual 4-Stroke and CFD Simulations’, SAE Paper 2002-01-3318.
Proceedings of Motorsports Engineering Conference and Exhibition,
Indianapolis, USA.
[2] De Vita, A, Andreassi, L & Di Angelo, L (2001), ‘Experimental and
Computational Study for the Optimisation of Race Car Intake Air Flow’, SAE
Paper 2001-01-1852.
[3] De Vita, A, Di Angelo, L, Andreassi, L & Romagnuolo, S (2002), ‘CFD-
Aided Design of an Airbox for Race Cars’, SAE Paper 2002-01-2167,
International Body Engineering Conference & Exhibition and Automotive &
Transportation Technology Congress, Paris, France.
[4] Heywood, JB (1992), Internal Combustion Engine Fundamentals. McGraw-
Hill International, New York.
[5] Smith, PH & Morrison, JC (1968), The Scientific Design of Exhaust and
Intake Systems, Second Edition, G. T. Foulis & co, LTD. London.
[6] Blair, PG (1999), Design and Simulation of Four-stroke Engines, Society of
Automotive Engineers, Inc. Warrendale, Pa.
[7] Smith, LA, Fickenscher, T, & Osborne, RP (1999), ‘Engine Breathing –
Steady Speed Volumetric Efficiency and Its Validity Under Transient Engine
Operation’, SAE Paper 1999-01-0212, International Congress & Exposition,
Detroit, MI, USA.
[8] Variable Intake Manifold (2002), Retrieved: February 19, 2002, from
http://autozine.kyul.net/technical_school/engine/tech_engine_2.htm#Variable
Intake Manifold
[9] Hatamura, K, Morishita, K, Asanomi, K & Shiraishi, H (1987), ‘Mazda’s New
V-6 Gasoline Engine and its Innovative Induction System’, SAE Paper
871977.
[10] Sakono, T, Takizawa, S, Harada, S, Ikeda, T & Abe, H (1992), ‘Mazda’s New
Lightweight and Compact V-6 Engine’, SAE Paper 920677.
 University of Pretoria etd – Makgata, K W (2005)
REFERENCES 94

[11] Winterbone, DE & Yoshitomi, M (1990), ‘The Accuracy of Calculating Wave

Action in Engine Intake Manifold’, SAE Paper 900677.
[12] Winterbone, DE & Pearson, RJ (2000), Theory of Engine Manifold Design –
Wave action methods for IC engines, Professional Engineering Publishing
Limited, UK.
[13] Engström, J, Kaminski, CF, Aldén, M, Josefsson, G & Magnusson, I (1998),
‘Experimental Investigations of Flow and Temperature Fields in an SI Engine
and Comparison with Numerical Analysis’, SAE Paper 1999-01-3541,
Proceedings of International Fuels & Lubricants Meeting & Exposition,
Toronto, Canada.
[14] Anderson, MK, Assanis, ND & Filipi, SZ (1998), ‘First and Second Law
Analyses of Naturally Aspirated, Miller Cycle, SI Engine with Late Intake
Valve Closure’, SAE Paper 980889, Proceedings of the International
Congress & Exposition, Detroit, MI, USA.
[15] Jovanovic, Z & Petrovic, S (2000), ‘The Effect of Intake Flow Modelling on
Flame Front Shape and its Displacement in Cylindrical Combustion
Chamber’, Proceedings of the Eleventh International Multidimensional
Engine Modeling User’s Group Meeting, Detroit, MI, USA.
[16] Van Niekerk, CGJ (2002), EngMod4T – A four-stroke engine simulation
package.
[17] Polášek, M & Vitek, O (2001), ‘Combination of 0-D Simulation of an ICE
with Advanced CFD Methods’, pp. 33 – 42, Proceedings of the 12th
MotorSympo, International Symposium, Brno, Czech Republic.
[18] Exa Corporation, PowerFLOW 3.0 (2001), Retrieved: January 25, 2001,
http://www.exa.com/newsite/frames/powerflowmaster.html.
[19] Wirtz, R & Mazzoni, D (1998), ‘Application of WAVE in Motorcycle
Prototyping’, Proceedings of the Ricardo Software International User
Conference, Detroit, USA.
[20] Papaconstantis, C (2001) Improvement of a V6 2.5l SI Engine’s Performance
through the use of an Inlet Manifold that utilises Air Resonance Charging,
BEng Undergraduate thesis, Department of Mechanical Engineering,
University of Pretoria. South Africa.
[21] White, FM (1991), Viscous Fluid Flow, Second Edition, McGraw-Hill
International, New York.
 University of Pretoria etd – Makgata, K W (2005)
REFERENCES 95

[22] White, FM (1994), Fluid Mechanics, Third Edition, McGraw-Hill

International, New York.
[23] Tennekes, H & Lumley JL (1994), A First Course in Turbulence, The MIT
Press. Massachusetts.
[24] Ferziger, JH (2002), Computational Methods for Fluid Dynamics, 3rd Edition,
Springer, Berlin.
[25] Fluent Inc (2002), Fluent version. 6.0 Manuals, Lebanon, NH.
[26] Fluent Inc (2002), Gambit version.2 Manual, Lebanon, NH.
[27] Fletcher, CAJ (1991), Computational Techniques for Fluid Dynamics: Volume
II, Second Edition, Spring-Verlag, New York, USA.
[28] Myers, RH & Montgomery, DC (2002), Response Surface Methodology -
Process and Product Optimization Using Designed Experiments, 2nd Edition,
Wiley, New York, USA.
[29] Stander, N & Craig, KJ, (2002) ‘On the robustness of a simple domain
reduction scheme for simulation-base optimisation’, Engineering
Computations, vol. 19, no. 4, pp. 431-450.
[30] Stander, N, Eggleston, TA, Craig, KJ & Roux, WJ (2003), LS-OPT User’s
Manual Version 2, Livermore Software Technology Corporation, Livermore,
CA, USA.
[31] Snyman, JA (2001), ‘The LFOPC leap-frog algorithm for constrained
optimization’, Computers and Mathematics with Application, vol. 40, pp.
1085-1096.
[32] Craig, KJ, de Kock, DJ & Snyman, JA (1999), ‘Using CFD and mathematical
optimisation to minimize stack pollution’, International Journal for Numerical
Methods in Engineering, vol. 44, pp. 551-566.
[33] Craig, KJ, Venter, P, de Kock, DJ & Snyman, JA (1999), ‘Optimization of
structured grid spacing parameters for separated flow simulation using
mathematical optimization’, Journal of Wind Engineering and Industrial
Aerodynamics, vol. 80, pp. 221-231.
[34] De Kock, DJ, Craig, KJ & Snyman, JA (2000), ‘Using mathematical
optimization in the CFD analysis of a continuous quenching process’,
International Journal for Numerical Methods in Engineering, vol. 47, pp. 985-
999.
 University of Pretoria etd – Makgata, K W (2005)
REFERENCES 96

[35] Craig, KJ, De Kock, DJ & Snyman, JA (2001), ‘Minimizing the effect of
automotive pollution in urban geometry using mathematical optimization’,
Atmospheric Environment, vol. 35, pp. 579-587.
[36] Craig, KJ, de Kock, DJ, Makgata, KW & de Wet, GJ (2001), ‘Design
Optimization of a Single-Strand Continuous Caster Tundish Using RTD Data’,
ISIJ International, vol. 41, no. 10, pp. 1194-1200.
[37] Kingsley, TC (2001), Effectiveness of Computational Flow Optimisation in a
Dynamic Flow Process, BEng Undergraduate thesis. Department of
Mechanical Engineering. University of Pretoria. South Africa.
[38] Craig, KJ & de Kock, D (1998), Computational Investigation into Flow
Phenomena in Touring Car Engine Trumpet and Intake: Modelling of Base
Geometry, Report prepared for Toyota SA Marketing, Laboratory for
Advanced Engineering (Pty) Ltd, University of Pretoria, South Africa.
[39] Craig, KJ & de Kock, D (1998), Computational Investigation into Flow
Phenomena in Touring Car Engine Trumpet and Intake: Modifications-Phase
1, Report prepared for Toyota SA Marketing. Laboratory for Advanced
Engineering (Pty) Ltd, University of Pretoria, South Africa.
[40] Renishaw (2001), Retrieved: January 23 2001, from
http://www.renishaw.com/client/product/UKEnglish/PRD-1229.shtml.
[41] Paraform Inc. (2002), Paraform 2.1.
[42] SuperFlow (2003), SuperFlow SF-100 flowbench Manual.
[43] Makgata, KW, Craig, KJ & De Kock, DJ, (2004), ‘Computational analysis and
parametric study of a high-performance rally engine’, SACAM Paper 42, 4th
South African Conference on Applied Mechanics, Johannesburg region, South
Africa.
 University of Pretoria etd – Makgata, K W (2005)
Appendix A: Intake System 97

Appendix A: Intake system

The individual intake system components of the engine concerned are shown in
Figure A-1 to Figure A-5.

Figure A-1: Side view of airbox (firewall side)

Figure A-2: Side view of airbox (engine side)

Figure A-3: Air-filter housing

 University of Pretoria etd – Makgata, K W (2005)
Appendix A: Intake System 98

Figure A-4: Bottom view of airbox

Figure A-5: Complete intake system

 University of Pretoria etd – Makgata, K W (2005)
Appendix B: Input for EngMod4T, Gambit and Fluent 99

Appendix B: Input for EngMod4T, Gambit and

Fluent
Table B-1: Detail engine input data for EngMod4T
Crank shaft bearing type Plain
Bore 86.00 mm
Stroke 86.00 mm
Conrod length 148.00 mm
Compression ratio 13.5:1
Exhaust temperature 800 °C
Firing order 1-3-4-2
Exhaust valve
Valve open BBDC 75°
Valve close ATDC 45°
Valve 1mm open BBDC 56°
Valve 1mm before closed ATDC 26°
Dwell angle 5.5°
Seat outer diameter (dos) 29.45 mm
Seat inner diameter (dis) 27.40 mm
Port diameter at valve (dip) 25.70 mm
Port diameter and manifold 36.80 mm
Valve stem diameter (dst) 6.00 mm
Valve seat angle (phi) 45°
Length of exhaust in head 82.00 mm
Valve tappet clearance 0.25 mm
Inlet valve
Valve open BTDC 53°
Valve close ABDC 72°
Valve 1mm open BTDC 31°
Valve 1mm before closed ABDC 50°
Dwell angle 4.5°
Seat outer diameter (dos) 33.89 mm
 University of Pretoria etd – Makgata, K W (2005)
Appendix B: Input for EngMod4T, Gambit and Fluent 100

Seat inner diameter (dis) 31.62 mm

Port diameter at valve (dip) 30.07 mm
Port diameter and manifold 43.10 mm
Valve stem diameter (dst) 6.00 mm
Valve seat angle (phi) 45°
Length of exhaust in head 100 mm
Valve tappet clearance 0.23 mm
Exhaust system 4 into 2 into 1
Exhaust geometry See
Air fuel ratio (A/F) 12.7
RPM for maximum power 8000
Type of fuel Premium
 University of Pretoria etd – Makgata, K W (2005)
Appendix B: Input for EngMod4T, Gambit and Fluent 101

Infinite

Figure B-1: Exhaust system configuration

 University of Pretoria etd – Makgata, K W (2005)
Appendix B: Input for EngMod4T, Gambit and Fluent 102

Performance Curves
250

225

200
Torque [Nm] & Power [kW]

175

150

125
Power_25mm
100 Torque_25mm
Power_Basecase
75
Torque_Basecase
50 Power_45mm
Torque_45mm
25

0
1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 9000
Engine Speed [rpm]

Figure B-2: EngMod4T Engine performance results for different trumpet

heights
 University of Pretoria etd – Makgata, K W (2005)

Appendix 103

Appendix C: Experimental Setup

Manometer

Figure C-1: Cylinder head on the flowbench

Intake 1

Figure C-2: Base of airbox with trumpets bolted on the cylinder head
Figure C-2 shows the trumpets before the top of the airbox is mounted. Note that all
but one of the trumpets is blanked off to enable measurement of the flow through only
one trumpet.
 University of Pretoria etd – Makgata, K W (2005)

Appendix 104

Figure C-3: Another view of trumpet and airbox base assembly

Electronic meter
measuring valve lift

Cylinder head

Blanking off to
prevent leakage
Figure C-4: Cylinder head and airbox test set-up

Note the asymmetrical arrangement of the trumpets and also tap points where airflow
measurements are taken.
 University of Pretoria etd – Makgata, K W (2005)
Appendix D 105

Appendix D: Journal Files

D-1. Journal file used in the GAMBIT for building models

identifier name "t" new nosaveprevious
identifier name "location\inlets_v.dbs" old nosaveprevious *import file with inlet port surfaces
window modify noshade
/max angle = 13 deg

/defining variables
$clear = 66.4
$tclear =(90-$clear)+1
$rad = 50
$offset = 0
$height = 155
$ang = atan((210-$height)/470)
$wid = 78.6

coordinate activate "c_sys.1"

volume create width 470 depth 210 height (240+$wid) offset 235 105 (120+($wid/2)) brick
face create width (240+$wid) height 300 offset 150 0 (120+($wid/2)) zxplane rectangle
face move "face.630" offset 0 40 0
face move "face.630" dangle 60 vector 0 0 260 origin 0 40 0
volume split "volume.96" faces "face.630" connected
volume delete "volume.97" lowertopology
face create width 480 height 210 offset 240 105 0 xyplane rectangle
face move "face.634" offset 0 40 0
face move "face.634" dangle 25 vector 480 0 0 origin 0 40 0
volume split "volume.96" faces "face.634" connected
volume delete "volume.96" lowertopology
vertex cmove "vertex.1414" multiple 1 offset 0 0 100
vertex cmove "vertex.1429" multiple 1 offset 0 0 -100
vertex cmove "vertex.1430" "vertex.1431" multiple 1 offset -500 0 0
edge create straight "vertex.1431" "vertex.1430" "vertex.1432" "vertex.1433"
edge create straight "vertex.1433" "vertex.1431"
face create wireframe "edge.2045" "edge.2046" "edge.2047" "edge.2044" real

/rotating face at angle corresponding to height ratio "$ang"

face move "face.640" dangle $ang vector 0 0 360.727698 origin 470 210 -20.727698
volume split "volume.97" faces "face.640" connected
volume delete "volume.97" lowertopology
edge round "edge.2029" "edge.2048" "edge.2038" "edge.2011" "edge.2049" "edge.2050" \
"edge.2052" "edge.2032" "edge.2031" "edge.2017" "edge.2040" "edge.2037" "edge.2020" \
radius1 25
volume blend "volume.98"

vertex cmove "vertex.1454" multiple 1 offset 0 0 100

vertex cmove "vertex.1444" multiple 1 offset 0 0 -100
vertex cmove "vertex.1466" "vertex.1467" multiple 1 offset -500 0 0
edge create straight "vertex.1467" "vertex.1466" "vertex.1468" "vertex.1469"
edge create straight "vertex.1469" "vertex.1467"

/creating face for trumpet clearance "$tclear"

face create wireframe "edge.2109" "edge.2110" "edge.2111" "edge.2112" real
face move "face.671" offset 0 $tclear 0
volume split "volume.98" faces "face.671" connected
volume delete "volume.99" lowertopology
 University of Pretoria etd – Makgata, K W (2005)
Appendix D 106

/creating frustrum inlet "$rad"

volume create height 193 radius1 $rad radius3 55 offset 250 0 0 xaxis frustum
vertex create onedge "edge.2129" uparameter 0.75
vertex create onedge "edge.2129" uparameter 0.25
edge create straight "vertex.1480" "vertex.1481"
volume move "volume.99" offset 229.30706 27.63082 185
edge move "edge.2131" offset 229.30706 27.63802 185

/rotating the inlet "fixed angle"

volume move "volume.99" dangle -40 vector 0 0 100 origin 575.80706 27.63802 135

/offsetting the inlet from original "0" position "$offset"

volume move "volume.99" offset 0 0 -$offset connected

volume unite volumes "volume.99" "volume.98"

import acis "D:\Katlego\Master\CFD\Iter2\trump.sat" ascii
/volume delete "volume.105" "volume.102" "volume.101" "volume.106" \
/ "volume.110" "volume.109" "volume.113" "volume.114" lowertopology
volume unite volumes "volume.110" "volume.111"
volume unite volumes "volume.108" "volume.109"

/splitting the trumpets

volume split "volume.99" volumes "volume.108" connected bientity
volume split "volume.99" volumes "volume.110" connected bientity

/deleting the trumpets "solid"

volume delete "volume.108" "volume.117" lowertopology
volume delete "volume.110" "volume.121" "volume.118" "volume.120" \
lowertopology

volume split "volume.99" volumes "volume.112" connected bientity

volume split "volume.99" volumes "volume.113" connected bientity
volume split "volume.99" volumes "volume.114" connected bientity
volume split "volume.99" volumes "volume.115" connected bientity

volume split "volume.112" volumes "volume.104" connected bientity

volume split "volume.113" volumes "volume.105" connected bientity
volume split "volume.114" volumes "volume.106" connected bientity
volume split "volume.115" volumes "volume.107" connected bientity

window modify noshade

volume delete "volume.100" "v_volume.94" lowertopology

volume delete "volume.101" "v_volume.92" "volume.102" "v_volume.90" \
"volume.103" "v_volume.88" lowertopology

volume delete "volume.121" "volume.123" "volume.125" "volume.127" lowertopology

window modify face "face.137" "face.186" "face.232" "face.944" invisible

 University of Pretoria etd – Makgata, K W (2005)
Appendix D 107

D–2. Journal file used in the FLUENT simulations

fi/r-bc/*
fi/rea-prof/*

so/init/init-flow
so/iter 1000
file/write-case-data/D:\Katlego\Master\CFD\last-pt\lp_ss_1st_order.gz

so/set/dis pr 12
so/set/dis mo 1
so/set/dis ep 1
so/set/dis k 1
so/set/dis te 1
so/set/dis temp 1
def/mate/ch-cr/air air y ideal-gas

{DEFAULT VALUES USED}

so/iter 1000
file/write-case-data/*

def/mo/uns-1-o y

so/set/dis fl 22
so/set/dis pr 10
so/set/dis de 0
so/set/dis mo 0
so/set/dis ep 0
so/set/dis k 0
so/set/dis te 0

so/set/und-rel pr 1
so/set/und-rel mo 1
so/set/und-rel ep 1
so/set/und-rel de 1
so/set/und-rel k 1
so/set/und-rel te 1

so/mon/sur/c-m
so/mon/sur/set/p-1 pr p-1 () y 1 n y m-1-exp1_iter_2.out y "Mass Flow Rate"
so/mon/sur/set/p-2 pr p-2 () y 2 n y m-2-exp1_iter_2.out y "Mass Flow Rate"
so/mon/sur/set/p-3 pr p-3 () y 3 n y m-3-exp1_iter_2.out y "Mass Flow Rate"
so/mon/sur/set/p-4 pr p-4 () y 4 n y m-4-exp1_iter_2.out y "Mass Flow Rate"

so/mon/sur/set/m-1 pr p-1 () y 5 n y pr-1-exp1_iter_2.out y "Area-Weighted Average"

so/mon/sur/set/m-2 pr p-2 () y 6 n y pr-2-exp1_iter_2.out y "Area-Weighted Average"
so/mon/sur/set/m-3 pr p-3 () y 7 n y pr-3-exp1_iter_2.out y "Area-Weighted Average"
so/mon/sur/set/m-4 pr p-4 () y 8 n y pr-4-exp1_iter_2.out y "Area-Weighted Average"

(cx-gui-do cx-activate-item "MenuBar*MonitorsSubMenu*Surface...")

(cx-gui-do cx-activate-item "Surface Monitors*Frame2*Table2*PushButton11(Define)")
 University of Pretoria etd – Makgata, K W (2005)
Appendix D 108

(cx-gui-do cx-set-list-selections "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)"

'( 2))
(cx-gui-do cx-activate-item "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)")
(cx-gui-do cx-activate-item "Define Surface Monitor*PanelButtons*PushButton1(OK)")
(cx-gui-do cx-activate-item "Surface Monitors*Frame2*Table2*PushButton17(Define)")
(cx-gui-do cx-set-list-selections "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)"
'( 2))
(cx-gui-do cx-activate-item "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)")
(cx-gui-do cx-activate-item "Define Surface Monitor*PanelButtons*PushButton1(OK)")
(cx-gui-do cx-activate-item "Surface Monitors*Frame2*Table2*PushButton23(Define)")
(cx-gui-do cx-set-list-selections "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)"
'( 2))
(cx-gui-do cx-activate-item "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)")
(cx-gui-do cx-activate-item "Define Surface Monitor*PanelButtons*PushButton1(OK)")
(cx-gui-do cx-activate-item "Surface Monitors*Frame2*Table2*PushButton29(Define)")
(cx-gui-do cx-set-list-selections "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)"
'( 2))
(cx-gui-do cx-activate-item "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)")
(cx-gui-do cx-activate-item "Define Surface Monitor*PanelButtons*PushButton1(OK)")
(cx-gui-do cx-activate-item "Surface Monitors*Frame2*Table2*PushButton35(Define)")
(cx-gui-do cx-set-list-selections "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)"
'( 2))
(cx-gui-do cx-activate-item "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)")
(cx-gui-do cx-activate-item "Define Surface Monitor*PanelButtons*PushButton1(OK)")
(cx-gui-do cx-activate-item "Surface Monitors*Frame2*Table2*PushButton41(Define)")
(cx-gui-do cx-set-list-selections "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)"
'( 2))
(cx-gui-do cx-activate-item "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)")
(cx-gui-do cx-activate-item "Define Surface Monitor*PanelButtons*PushButton1(OK)")
(cx-gui-do cx-activate-item "Surface Monitors*Frame2*Table2*PushButton47(Define)")
(cx-gui-do cx-set-list-selections "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)"
'( 2))
(cx-gui-do cx-activate-item "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)")
(cx-gui-do cx-activate-item "Define Surface Monitor*PanelButtons*PushButton1(OK)")
(cx-gui-do cx-activate-item "Surface Monitors*Frame2*Table2*PushButton53(Define)")
(cx-gui-do cx-set-list-selections "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)"
'( 2))
(cx-gui-do cx-activate-item "Define Surface Monitor*Frame1*Frame1*DropDownList3(X Axis)")
(cx-gui-do cx-activate-item "Define Surface Monitor*PanelButtons*PushButton1(OK)")
(cx-gui-do cx-activate-item "Surface Monitors*PanelButtons*PushButton1(OK)")

(rpsetvar 'flow-time 0.000013293)

def/bc/zt p-1 po
def/bc/po p-1 y "cyl_prof_1" "p" n y "cyl_prof_1" "temp" n y n n n y 5 50
def/bc/zt p-3 wall
def/bc/zt p-4 wall
def/bc/zt p-2 po
def/bc/po p-2 y "cyl_prof_2" "p" n y "cyl_prof_2" "temp" n y n n n y 5 50
sol/set/ts 0.00004
sol/dti 64 20
def/bc/zt p-2 wall
sol/set/ts 0.00004
sol/dti 51 20
def/bc/zt p-3 po
def/bc/po p-3 y "cyl_prof_3" "p" n y "cyl_prof_3" "temp" n y n n n y 5 50
sol/set/ts 0.00004
sol/dti 64 20
def/bc/zt p-1 wall
sol/set/ts 0.00004
sol/dti 51 20
 University of Pretoria etd – Makgata, K W (2005)
Appendix D 109

def/bc/zt p-4 po
def/bc/po p-4 y "cyl_prof_4" "p" n y "cyl_prof_4" "temp" n y n n n y 5 50
sol/set/ts 0.00004
sol/dti 64 20
def/bc/zt p-3 wall
sol/set/ts 0.00004
sol/dti 51 20
def/bc/zt p-2 po
def/bc/po p-2 y "cyl_prof_2" "p" n y "cyl_prof_2" "temp" n y n n n y 5 50
sol/set/ts 0.00004
sol/dti 64 20
def/bc/zt p-4 wall
sol/set/ts 0.00004
sol/dti 51 20
file/write-case-data/*

*file location
 University of Pretoria etd – Makgata, K W (2005)
Appendix E 110

Appendix E: Summary of Results

2.8

2.75
Total Inducted Mass [g]

2.7

2.65

2.6

2.55

2.5

2.45
4 X Single Intake 4-Intakes-No- Airbox-Base Improved
Airbox Airbox design

Figure E-1: Total mass inducted for considered cases over one engine cycle

Single Intake
4-Intakes-No-Airbox
0.72 Airbox-Base
Improved Airbox design
0.7

0.68
Mass Inducted [g]

0.66

0.64

0.62

0.6

0.58

0.56
Intake 1 Intake 2 Intake 3 Intake 4

Figure E-2: Plot of individual intake flow induction for the four considered cases