Professional Documents
Culture Documents
One-Dimensional Engine Modeling and Vali PDF
One-Dimensional Engine Modeling and Vali PDF
AND VALIDATION
Final Report
KLK434-B
N07-09
16. Abstract
The use of one-dimensional CFD engine simulation is an essential tool to the engine development process. Engine design through
simulation can drastically reduce time needed to perform engine experiments and prototyping, as most engine experiments can be
simulated within the software. As long as a model can be validated to high degree of accuracy (i.e. +/- 5%), the model can be
used with a high level of confidence to optimize engine parameters. This work has used only a small fraction of WAVE’s
potential. More sophistication is possible in the areas of combustion kinetics, computational fluid dynamics, and emissions
abatement through the use of advanced features and co-simulation.
The intention of this research was to learn the process of one-dimensional CFD engine simulation and document what has been
learned about the process as well as the potential of the software for future work. The research opportunities available with this
software are endless and far-reaching. The software can also be used as an essential engine technology learning tool in technical
elective courses such as ME433 – Combustion Engine Systems. Any questions that may arise about standard engine operation or
combustion processes can be investigated with WAVE.
.
17. Key Words 18. Distribution Statement
Simulation; engines; Unrestricted; Document is available to the public through the National
Technical Information Service; Springfield, VT.
19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of Pages 22. Price
Unclassified Unclassified …
Form DOT F 1700.7 (8-72) Reproduction of completed page authorized
TABLE OF CONTENTS
1. INTRODUCTION..................................................................................................................... 4
2. APPROACH .............................................................................................................................. 5
3 METHODOLOGY .................................................................................................................... 6
Optimization .................................................................................................................... 17
Model Validation............................................................................................................. 27
Engine Modeling i
Engine Modeling ii
EXECUTIVE SUMMARY
The use of one-dimensional CFD engine simulation is an essential tool to the engine
development process. Engine design through simulation can drastically reduce time needed to
perform engine experiments and prototyping, as most engine experiments can be simulated
within the software. As long as a model can be validated to high degree of accuracy (i.e. +/- 5%),
the model can be used with a high level of confidence to optimize engine parameters. After that,
engine testing is only needed to validate the final prototype. This work has used only a small
fraction of WAVE’s potential. More sophistication is possible in the areas of combustion
kinetics, computational fluid dynamics, and emissions abatement through the use of advanced
features and co-simulation.
The intention of this research was to learn the process of one-dimensional CFD engine
simulation and document what has been learned about the process as well as the potential of the
software for future work. Students considering graduate school here at the University of Idaho
should be presented with the opportunity to use this software for engine research projects. The
research opportunities available with this software are endless and far-reaching. The software
can also be used as an essential engine technology learning tool in technical elective courses
such as ME 433. Any questions that may arise about standard engine operation or combustion
processes can be investigated with WAVE.
Engine Modeling 3
1. INTRODUCTION
The use of one-dimensional computation fluid dynamic (1D CFD) engine simulation software is
widespread throughout the engine development industry. This simulation method allows for
characterizing engine operation without the need for high-end processing and time-intensive
computations. Various commercial software packages are available; however, software costs
generally prohibit use in small organizations, primarily making them industry-specific software
packages. There is a great value in being able to use such engine development methods at the
collegiate level. The Ricardo Company has made this possible through Formula SAE
sponsorship. This thesis researched professional simulation methods and used Ricardo’s WAVE
software to research potential engine packages for the Formula SAE competition.
Simulation work has been applied in a case study of a naturally aspirated 600 cc motorcycle
engine. The objective of this case study was to maximize engine torque output at the most
useable engine speeds based on simulation predictions. Specifically, physical engine parameters
such as intake and exhaust geometries were varied to discover performance trends. While no
full-system optimization was performed, individual component parameters were investigated to
reveal their sensitivity to engine output. Upon completion of these component studies, a
complete engine package design was suggested based on the simulation predictions.
Engine Modeling 4
2. APPROACH
There are two primary engine simulation software packages used in the industry today: Ricardo
WAVE and GT-Power. Both software packages are similar in purpose and functionality. These
1D simulation packages require detailed input parameters to simulate engine operation as a
whole rather than using models that target certain engine sub-systems. Being that all engine
components work together as a system, it is advantageous to model the entire engine system
rather than individual subsystems. Ricardo WAVE was used for this research as it was provided
by Ricardo through Formulas SAE sponsorship.
Ricardo WAVE is more than just an engine simulation program. It is an engineering code
designed to analyze the dynamics of pressure waves, mass flows, energy losses in ducts,
plenums, and the manifolds of various systems and machines. When further-detailed analysis is
desired, WAVE can be coupled with various industry-standard software packages such as CFD
analysis programs and statistical analysis software. Ricardo also offers various external programs
that can be coupled with WAVE. [WAVE, 2007]
The basic operation of the WAVE code analyzes flow networks composed of ducts, junctions,
and orifices. Within this network of plumbing, engine cylinders, turbochargers / superchargers,
compressors, and pumps can be inserted. WAVE can simulate internal combustion engines as
well as other compressible-fluid flow systems. Once a simulation has completed, the post-
processing program, WavePost, allows for detailed analysis of the simulated engine operation.
The overall capabilities of WAVE extend beyond the simulation of engine operation. The
software bundle has a program specifically dedicated to acoustic and noise analysis. Other
secondary software uses include: Hybrid 1D/3D CFD engine simulation, thermal analysis,
controls systems, and combustion and emissions simulation. [WAVE, 2007]
Engine Modeling 5
3 METHODOLOGY
The use of Ricardo WAVE to develop engine simulations requires several computer programs
that vary in purpose from model setup and 3D modeling, to statistical analysis of model output.
The four programs used in this research are: WaveBuild, WaveMesher, WAVE, and WavePost.
WaveBuild is the pre-processor and GUI that is used to define all engine and simulation
parameters. All model and simulation parameters are defined from within this program with the
exception of external CAD geometries. A screen-capture of the GUI layout is shown in Figure 1.
The left portion of the screen lists the elements available to be inserted into the model such as
engine cylinders, ducts, and orifices. The right side of the screen is the visual portrayal of the
engine model layout in which all parameters can be modified. There are some geometries that
can be difficult and time consuming to construct with the elements available.
Engine Modeling 6
For this reason, WaveBuild allows complex geometries to be imported into the model from 3D
CAD files using a program called WaveMesher. WaveMesher allows the end-user to manipulate
3D cad models of complex geometries such as intake manifolds and exhaust headers, so that the
3D models can be broken-down into a usable one-dimensional form for the WAVE processor.
Figure 2 is a screen capture of the WaveMesher program in which an intake manifold has been
meshed for use in WaveBuild.
Figure 2: WaveMesher
Once the model is setup in WaveBuild and 3D models have been meshed, a simulation can be
started by running the WAVE code. WAVE is the solver that performs all the calculations
needed to simulate engine operation. It is a non-interactive program that runs in a DOS window
while streaming certain output data and simulation progress. The output data shown during
simulation runtime can be customized to show parameters of interest. These parameters can be
Engine Modeling 7
used as indicators to show whether or not the simulation is producing reasonable results,
allowing the simulation to be prematurely stopped if the model is not function properly.
Once a simulation is finished, a large output file is created that contains all data needed to
analyze the simulated engine operation. WavePost is the post-processor for WAVE simulations
that allows interpretation of simulation results. It allows for the creation of: time plots, sweep
plots, spatial plots (animated), and TCMAP plots (for turbines and compressors). Figure 4 is a
screen-capture of the WavePost post-processing software. A sample sweep plot is shown.
Engine Modeling 8
Figure 4: WavePost Post-Processor
Model Setup
The engine model is setup by defining some relatively basic inputs and then some more
advanced inputs that require some engine testing. The basic inputs are composed of engine
geometry and boundary conditions. Therefore, all dimensions from the intake and exhaust
ducting must be recorded and input. Likewise, manufacturer specifications for internal engine
geometry such as bore, stroke, connector rod length, wrist-pin offset and compression ratio must
be input. Initial conditions such as exhaust temperatures, intake temperatures, and wall
temperatures need to be input as reasonable values. These can be modified to higher accuracy
once actual engine measurements become available. A general flow diagram for the development
of a model is shown in Figure 5.
Engine Modeling 9
Figure 5: Simplified Model Setup Procedure
Within the WaveBuild GUI there are four basic junctions used to connect ducts [WAVE, 2007]:
Orifice: A mass-less and zero-length junction used to directly connect two ducts. The
diameter of the orifice defaults to match the connecting ducts but can be input
independently when necessary.
Ambient: Defines a condition in which one end of a duct is open to atmospheric pressure.
They are commonly used to define air filter openings and muffler outlets.
Engine Cylinder: Represents an internal combustion engine cylinder of either the 2-stroke
or 4-stroke type. Allows for modeling of advanced concepts such as 6 and 8 stroke
engines as well.
Y-Junctions: Defines a junction that contains more than two openings or other geometries
that cannot be adequately modeled with ducts and orifices. Used to model complex
geometries such as intake manifolds, mufflers, air filters, and exhaust collectors.
Engine Modeling 10
Some basic modeling conventions must be well understood to properly create a model using the
above-mentioned basic junctions. Y-Junctions should be used anytime that a desired geometry
cannot be achieved through the use of ducts and orifices. Any flow junction that has a branched
flow (i.e. manifolds and collectors) must be modeled with Y-Junctions. Two forms of Y-
Junctions exist: the simple Y-junction and the complex Y-Junctions. Simple Y-junctions are
represented by spherical volumes that distribute flow to multiple ducts. Simple Y-junctions are
rarely used because they are essentially an over-simplified complex y-junction. Complex Y-
junctions are represented by an arbitrarily shaped volume. A three-dimensional view allows for
the arrangement of ducts relative to the junction to be modeled appropriately. Additionally,
length properties from the duct opening to the back wall of the volume, discharge coefficients,
and area change parameters can be modified.
Once the basic inputs are defined, the advanced inputs that define the cylinder head need to be
defined. These inputs are: port flow coefficients, valve lift per crankshaft rotation, and
combustion modeling. Because the flow through the engine is highly dependent on the flow
through cylinder head ports, it is recommended that this information is measured directly from
the cylinder head as it will achieve higher accuracy than what would be expected from 1D CFD
modeling. These port flows can be measured on a flowbench. Typically, a pressure drop of 28”
of water is applied across one cylinder and volumetric flow (measured in CFM) through the
valves can be measured at incremental valve lifts. A spreadsheet provided by Ricardo converts
these measured flow values to useable port-flow coefficients. The second advanced input within
the cylinder head is the valve lift per crankshaft rotation. The valve lift per camshaft rotation can
be measured by incrementally rotating the camshaft and measuring valve lift with a dial
indicator. Then the opening points of the intake and exhaust valves relative to TDC must be
known to reference the camshaft rotation data to the crankshaft.
The third advanced input and one of the most important datasets needed to characterize an
engine model and validate simulation results comes from in-cylinder pressure measurements.
Once crank-resolved pressure traces are obtained for complete engine cycles at each operating
point, friction correlations and combustion models can then be determined. Unfortunately,
obtaining in-cylinder pressure data is not a simple task. Once a pressure sensor is physically
Engine Modeling 11
implemented into the cylinder head, special equipment is still needed to interpret the very rapid
sensor output. The University of Idaho Small Engine Research Facility currently does not have
this ability; thus in-cylinder pressure measurements were not available for this modeling
research.
Without in-cylinder pressure measurements, the combustion model had to be predicted based on
typical forced induction Wiebe function parameters. WAVE allows for three parameters in the
Wiebe correlation to be input: 10-90 percent burn duration, 50 percent burn point, and the Wiebe
exponent. [WAVE, 2007]
WEXP 1
AWI
BDUR (Equation 1)
W 1 e
where,
Engine Modeling 12
Table 1: Wiebe Function Parameters
Table 1 shows the Wiebe parameters determined by Blair for naturally aspirated engine
operation. Likewise, inputs based on Blair’s findings are included as well. The burn duration is
relatively constant but does increase slightly with engine speed. Ignition timing advance
increases with engine speed but the 50 percent burn location, CA50, remains relatively constant.
Lastly, the Wiebe exponent, WEXP, tends to decrease with engine speed.
Once these basic and advanced inputs have been satisfied, the simulation format must be setup.
WAVE allows for both transient and steady-state engine simulations. While transient simulations
can provided a more detailed representation of engine performance, steady-state simulations are
better suited for simulations in their early stages of development for various reasons. A
significant advantage of steady-state simulations is that parameters can be modified
independently for each engine speed simulated. Other important simulation parameters include:
cycles to run for convergence, convergence detection, and convergence tolerances. These are
some of many general simulation parameters. Another parameter allows for the flow-field to be
reinitialized in between simulation cases. This means that for each case the simulation runs, it
Engine Modeling 13
will start with the user-imposed initial conditions such as wall temperature and ambient
conditions of the gas. By default, this feature is turned off, as quite often the final gas conditions
from a prior case are closer to the next case than those initial conditions guessed by the user.
Another important general parameter is convergence detection. Convergence detection allows
the WAVE code to move on to the next case once a user-specified tolerance for convergence
(default is 1 percent) towards a solution is attained on the active case. Simulation duration is
another important parameter. This parameter sets the number of cycles for the engine to run
before settling on a solution. If auto-convergence is enabled, WAVE will move on to the next
case once convergence is reached regardless of the number of cycles specified. If convergence is
not reached by the number of specified cycles, the WAVE code will output a warning to let the
user know that convergence was not reached within desired tolerances and will continue on to
the next case.
There are certain limitations associated with 1D CFD flow analysis that require special attention.
Ricardo has provided a very comprehensive set of input parameters that allow users to highly
customize any flow model. One such limitation comes into play when trying to model ducts that
have significant geometric changes over very small length increments. An example of such
geometry is the bell-mouth of an intake runner. The 3-D flow phenomenon that occurs in such
geometry can not be accurately modeled by 1D code. In this case, discharge coefficients can be
used based on published findings for bell-mouth discharge coefficients. So, essentially flow
losses can be imposed on geometric instances that can’t be accurately modeled with 1D code.
Furthermore, when modeling tapered ducts, it is recommended to keep angles smaller than 7
degrees. This is due to the fact that flow separation, a phenomena not modeled in 1D code,
typically occurs at flow angles of 7 degrees or more. One such instance that was encountered in
this research was in the modeling of the flow restrictor that had a converging angle of over 20
degrees. If the restrictor is modeled exactly from its physical geometry, WAVE will over-predict
power output due to the inability to calculate flow losses due to flow separation.
Model Development
A model is only as accurate as the input data. This is a principle that must always be kept in
mind when attempting to validate a model. Likewise, it important to keep in mind that 1D
Engine Modeling 14
simulation is not a 100 percent accurate prediction of true gas dynamics. While good simulation
results can be within 5 percent of actual performance, the amount of time spent troubleshooting a
model while trying to perfectly match empirical data to simulation predictions should be limited.
One of the reasons why 1D simulation is the mainstream form of engine simulation is because it
allows for rapid development whereas a full-blown 3D CFD simulation could take weeks to
calculate engine operation at a single engine speed. Another thing that should be kept in mind is
that two identical engines will exhibit slightly different performance due to different conditions
such as: bearing wear, piston ring seal, valve seating, carbon build-up, and fuel composition.
These are all reasons why it is important to not expect perfect-matching predictions. What is
more important is that the simulation accurately predicts the shape of the torque curve as
opposed to the exact magnitudes. [Gurney, 2001]
It is recommended that initial validation and calibration of the model should be performed by
comparing predicted volumetric efficiency to that measured on the dynamometer. In order to
obtain volumetric efficiency calculations, an accurate air-mass flow sensor must be interfaced
with the engine and dynamometer data acquisition. The mass flow sensor that is equipped with
the dynamometer used for this research has calibration problems and could not be used for initial
model validation and calibration. Instead, torque output was the parameter used to validate the
model. While volumetric efficiency validation is recommended, torque validation could be
considered the next best method. Torque output validation is performed by comparing torque
output measured on a dynamometer to predicted torque output. Discrepancies between
predictions and measurements can arise from several sources. For example, if the model is not
accurately predicting the location of the pressure-wave-induced intake / exhaust torque peaks,
then the dimensions of the intake / exhaust geometry and intake / exhaust valve opening inputs
should be checked for accuracy. If the overall torque magnitude is offset a certain amount across
the entire engine speed range, engine parameters such as the combustion model, heat transfer and
friction correlations should be checked for accuracy. It should also be kept in mind when
troubleshooting a model that the erroneous input is most often due to an input that was guessed
due to a lack of data. If this is the case, the data needs to be acquired or modified to match
empirical results.
Engine Modeling 15
Once the model seems to be predicting reasonable results, it can be further refined. Boundary
conditions and initial conditions can be input to match temperatures and pressures measured at
different junctions during actual engine testing. This helps improve model accuracy and
decreases computational time needed for convergence.
Parametric Studies
Once a model has been validated, it can be used as a most valuable engine development and
design tool. While WAVE’s features extend far beyond performance development, this
discussion will primarily focus on the use of WAVE to develop a high-performance engine
package. A good starting point is to parameterize engine dimensions of interest such as intake
runner length and exhaust header primary length. By performing a parametric study on one
geometric dimension at a time, trends can be revealed that indicate near-optimal dimensions for a
desired torque output. Later studies can implement design of experiment (DOE) methods for full-
system optimization. The below figure shows an example of a geometric parameter study; in this
case showing torque output based on varied engine compression ratios.
Engine Modeling 16
Examining one parameter at time is valuable for investigating trends, but does not necessarily
tell the full story. This is especially true for parameters that have effects that are engine speed
dependent such as intake and exhaust duct lengths. For example, having an intake resonance
occur at the same engine speed as a poorly-timed exhaust pressure wave from the headers will
severely hinder the effect of the intake resonance. On the contrary, parameters such as tubing
diameter tend to have effects that are less dependent on other parameters.
Optimization
Before attempting to optimize an entire package, general trends, sensitivities, and near-optimal
values should be investigated. Additionally, it is good practice to implement physical constraints
on an optimization so that it will produce usable results. Such a constraint could be on header
primary lengths. If a short and lightweight package is demanded by the application, length
restrictions should be implemented in the experiment’s boundary conditions. WAVE allows
optimization through its experiment feature. The experiment feature allows you to choose as
many parameters as you want to be varied according to minimum and maximum acceptable
values (boundary conditions). Two basic DOE methods are available for WAVE experiments:
Full/Half Factorial (2-level) and central composite (3-level).
Once an experiment has been completed, any parameter can be minimized or maximized. For
example; if the user wants to optimize camshaft parameters, the lift multipliers, lobe anchors,
and duration multipliers could be set as varied parameters within the experiment. The experiment
post-processing screen is shown in Figure 7. The parameters included in the experiment are
located in the lower left of the screen and can be modified with the slide bars. Engine
performance graphs on the right will reflect changes in real-time as the slider bars are adjusted.
The list on the left side of the screen is a comprehensive collection of output graphs available. As
an example, the user can click on a torque output graph and have torque output maximized, with
WAVE indicating the experiment parameters that lead to that optimal performance. Conversely,
the user could attempt to minimize an output such as trapped residuals and determine what
camshaft parameters lead to that minimum. [WAVE, 2007]
Engine Modeling 17
Figure 7: WAVE Experiment Panel
Some parameters such as exhaust tubing diameter do not necessitate optimization due to the
limited sizes available. In other words an exact dimension produced from the optimization is not
Engine Modeling 18
useful if the tubing is only available in four different practical sizes. These are cases where basic
parametric studies may be best implemented.
The experiment feature does have some limitations. The parameter table that is used to vary
parameters against engine speed is not used in experiment mode. So if a control system is
modeled on the engine using the parameter table, it will not be implemented in experiment mode.
The best method of optimization with WAVE requires co-simulation. Co-simulation is the
modeling practice of simultaneously using two or more major software packages to come to one
result. Examples of this include: WAVE with Fluent CFD, WAVE with Reaction Design’s
KINetics (combustion analysis), and WAVE with LMS/OPTIMUS. The WAVE code can be
combined with a second program for higher-level modeling; in this case an optimization program
(i.e. LMS OPTIMUS) to run comprehensive parametric experiments to optimize a complete
engine system. Co-simulation typically requires state-of-the-art computing power to be used
practically. However, the results of using co-simulation optimization allow for a truly state-of-
the-art engine development program.
Experimental Verification
In order to properly validate an engine simulation model, the engine must be instrumented with a
wide array of sensors with data being logged through data acquisition. For example, if the model
assumes the same air-fuel ratio in each cylinder, air-fuel ratio should be logged for each cylinder
and fueling should be modified within the engine control unit (ECU) to match the model or vice
versa. The implementation of numerous sensors not only allows for simulation validation but
also acts as a form of engine diagnostics to provide confirmation that the engine is operating as
intended.
The dynamometer used for this research was a Land & Sea water-brake engine dynamometer
(dyno). This type of dyno is primarily used for steady-state testing which pairs up nicely with
engine simulation work. While eddy-current dynos have more precise control, a water-brake
dyno is adequate for many applications. The dyno is interfaced to Land & Sea dyno controls and
data acquisition. The controls are operated from a control room overlooking the engine test cell
equipped with Dynomax dyno control software. Cooling is performed through the use of a
Engine Modeling 19
cooling cart. The cooling cart provides enough cooling to do prolonged steady-state testing.
However, it is not able to exactly maintain constant coolant temperatures, which is often desired
for accurate benchmarking. Likewise, ambient conditions are not regulated and will typically
fluctuate slightly throughout testing sessions.
To allow for frequent engine testing on the dynamometer, a portable engine stand, shown in
Figure 8, was adapted to plug directly into the dyno controls and data acquisition systems. The
engine was coupled to the dyno at the transmission output shaft by a direct connection
constrained by two pillow block bearings.
The engine used for this research is controlled using the Motec M4 Pro standalone ECU. This
ECU can be adapted to practically any engine and is one of the most customizable ECU’s on the
market with an abundance of features. Advanced features used in this research include: cylinder-
to-cylinder fueling, fuel auto-calibration (“quick lambda”), and datalogging.
Due to the speed range of the dyno head, testing is typically performed at engine speeds from
4500 rpm to 12000 rpm. Engine coolant and exhaust temperatures are allowed to stabilize before
Engine Modeling 20
loading the engine and taking measurements. Torque measurements and sensor outputs are
recorded at 10,000 Hz while engine speed is slowly swept from low to high and then back to low
engine speed. A typical sweep would start at 5000 rpm and wide open throttle (WOT). From the
control room, the dyno controls are used to slowly increase engine speed all the way to 12,000
rpm. The dyno then slowly sweeps back to 5000 rpm. By sweeping through the engine speed
range in both directions, transient error effects can be revealed. For example, if the dyno operator
increments engine speed too rapidly, the dyno will report higher torque output then actual steady
state output. After running an engine speed sweep, the data can then be exported to a spreadsheet
program such as Excel for analysis. Additionally, the software allows for a replay animation that
graphically displays the sensor outputs and test conditions exactly as they were during the test
session.
The controls available from the control room include: throttle actuation, engine speed control,
and emergency shutdown. The dyno equipment is equipped with a data acquisition unit capable
of recording a large array of sensors during engine testing. Typical operating parameters
measured include: intake temperatures, exhaust gas temperatures, manifold pressure, exhaust
back-pressure, air mass flow, engine knock, torque, shaft RPM, throttle position, and air-fuel
ratio. The system is able to control speed to roughly +/- 50 RPM. Due mostly to RPM
fluctuations, a torque variation of about +/- 0.3 ft-lb is seen during testing at a give test point.
Engine Modeling 21
4. FINDINGS
The WAVE simulation software was provided through Formula SAE sponsorship and thus all
simulation work could only be directly applied to Formula SAE related engine development per
Ricardo’s software license. The competition engine used for the past three years at the University
of Idaho is the Suzuki GSX-R600. It is a high performance race-inspired sport-bike engine.
Formula SAE (FSAE) rules mandate that engine displacement does not exceed 600cc; making
600cc sports-bike engines a popular choice for Formula SAE competition vehicles. In order to
maintain competitive competition among teams, Formula SAE officials mandate that the intake
system must breathe through a single 20 mm diameter restrictor. This air restriction largely limits
the possible power produced and complicates the engine system design. The required layout of
the intake system according to official Formula SAE rules is shown in Figure 9. Only one
throttle body is allowed, and it must be located upstream of the flow restrictor. Formula SAE
rules also have a regulation for the exhaust system in that it must not produce more than 110 db
when measured at a distance of 0.5 meters from the exhaust outlet, with the sound meter oriented
at an angle of 45 degrees from the exhaust outlet axis on the horizontal plane. A schematic for
the sound test setup is shown in Figure 10. This exhaust rule also limits overall power output.
Engine Modeling 22
Figure 9: Formula SAE Intake Design Regulations
Engine Modeling 23
Engine Package Design
The primary design objective for all FSAE engines is maximum performance (maximum torque
at engine speeds where high torque output is desired). Optimal engine performance can be very
subjective. Often, the most desirable torque curve is determined by the type of course that the
vehicle will be racing on as well as driver preference. If a driver feels that more power is needed
at corner-exit, then it may be advantageous to increase torque at lower engine speeds at the
expense of some top-end power. Likewise, if the racecourse is very tight and gearing forces the
vehicle to drive at lower engines speeds more often than higher engine speeds, it may again be
advantageous for the engine to aim for high torque at lower engine speeds. Past experience with
the Formula SAE competition and Autocross racing experience suggests that the Formula SAE
courses are too small / tight to necessitate targeting torque increase at very high engine speeds.
Therefore, design objectives for this naturally aspirated engine package will target torque
increase at lower engine speeds more so than higher engine speeds.
Model Setup
The first basic inputs that need to be satisfied lie within the engine itself. Quantities such as bore,
stroke, compression ratio, and cylinder-head clearance must be defined. These inputs can
typically be found from manufacturer specifications published in service manuals. A method
used for the initial model setup involved starting with the engine itself and moving outwards;
from the port geometry, to the manifold / headers, to the exhaust outlet and intake inlet. The
nearest connection to the engine’s cylinder bores are the intake and exhaust ports. The centerline
distance from the valve to top of the ports had to be measured for both the intake and exhaust
sides. This is a difficult measurement to make due to the small port size and complex geometry.
This centerline distance was approximated using a piece of bent wire. Because the flow through
the ports will be defined by the valve flow coefficients determined in flowbench tests, only the
inlet / outlet diameter and centerline length must be defined. Once the port lengths are defined,
all external intake and exhaust components need to be precisely measured and input into the
model. Length measurements are critical for the intake and exhaust components as they affect
the pressure wave timing in the ducting. This is especially true for the intake runners and exhaust
header primaries. For this application, all dimensions of the intake and exhaust components were
retrieved from the original solid models (CAD models) used for manufacturing. Components
Engine Modeling 24
such as exhaust headers can be difficult to measure directly from the components, making solid
model measurements very valuable.
The intake manifold is one component that is often easier and more reliable to model as a 3D
mesh rather than using one-dimensional parameters. Because the modeling of complex y-
junctions can be cumbersome, a WAVE Mesher model was created to automatically calculate
junction parameters based on the imported CAD model. Without the CAD model, certain volume
calculations and area ratios at each manifold junction would have had to be calculated. The flow
restrictor is venturi design that has a large converging angle and a small diverging angle. The
restrictor was modeled as two tapered ducts joined by a very short 20 mm diameter duct in the
center. The throttle plate is represented by an orifice between two ducts with a parameterized
diameter. Each throttle position could then be represented by an effective diameter. For this
research, wide-open throttle (WOT) performance is only of interest; so this parameter remained
constant for most studies.
The air filter and muffler are modeled based on the supporting documentation that was packaged
with the WAVE software. The air filter is modeled by inserting a zero-length duct that connects
two y-junction volumes. The diameter of this duct is input as smaller than the actual physical
diameter to produce a certain pressure drop between the two volumes, often between 1-2 kPa for
most air filters. Precise modeling of the restrictor would require flowbench testing or pressure
drop measurement on the dyno to accurately model the pressure drop. Resonators and mufflers
are modeled similarly to their physical construction. The muffler designed for 2007 engine is a
tri-pass design. The tubes in the muffler are represented by ducts while the muffler canister (the
outer shell) is represented by a series of y-junction volumes. The volumes attach to the ducts by
orifices and baffles within the muffler. The WAVE documentation provides detailed
methodology for modeling mufflers.
To define the operation of the cylinder head, the valve lift per incremental camshaft rotation had
to be defined. The GSXR cylinder head is dual-overhead cam (DOHC) format, meaning that
there is a separate camshaft for the intake and exhaust valves. To make the measurement, valve
lift was measured on one camshaft at a time. The valve lift was measured by locating a dial
indicator directly above the face of the cam follower. Cam rotation was handled by attaching a
Engine Modeling 25
degree wheel to the end of the camshaft and rotating it one degree at a time. These results of
valve lift versus incremental camshaft rotation were related to crankshaft rotation based on
published valve opening points with reference to crankshaft position. The resulting valve lift per
crank revolution is shown in Figure 11.
0.3500
Intake Valve Lift
Exhaust Valve Lift
0.3000
0.2500
Valve Lift (in)
0.2000
0.1500
0.1000
0.0500
0.0000
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700
To finish defining the cylinder head operation, flow versus valve lift had to be defined. The input
accepted by WAVE is in the form of flow coefficients or discharge coefficients. These
coefficients require empirical flowbench data. The cylinder head was adapted to a flowbench and
volumetric air flow (in CFM) was measured at a pressure drop of 28” of H2O, the standard
pressure drop measurement used on flowbenches in the performance development industry. The
flow data was then converted to a usable flow coefficient format using an excel spreadsheet
provided within the Ricardo WAVE Help documentation. The results are shown in Figure 12.
Engine Modeling 26
Flowbench Testing
160.0
Flow (CFM) at 28" H2O Pressure Drop
140.0
120.0
100.0
80.0
60.0
Intake Port Flow
40.0
Intake Port Flow (REVERSE)
20.0 Exhaust Port Flow
Exhaust Port Flow (REVERSE)
0.0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Valve Lift (in.)
The third advanced input is the combustion model. Ideally, in-cylinder pressure data versus
crankshaft rotation should be recorded to calculate Wiebe function parameters. Unfortunately,
such data could not be obtained with currently available test equipment. The data published by
Blair along with the data that was input into the model can be found in Table 1 in Chapter 2.
Some intuition had to be used to determine the parameters due to the large difference in engine
operating speed and difference in engine type.
Model Validation
There are numerous parameters that can be used to validate an engine model. Parameters such as
torque output, volumetric efficiency, and air mass flow can all be used to evaluate overall engine
operation. As mentioned, torque output was chosen as the parameter to use for validation due to
test equipment limitations.
When using the torque output validation method, certain trends in the predicted torque curve can
be validated without too much trouble. The engine testing results will show torque peaks and
Engine Modeling 27
valleys that correspond to either intake or exhaust geometry. Those peaks and valleys should
occur at the same engine speeds that the simulation predicts. Because these pressure-wave-
induced torque peaks result strictly from intake / exhaust geometry and the intake / exhaust valve
lift profiles, it is important that these trends can be validated. If the predicted peaks and valleys
are not occurring at the correct engine speeds, geometry and valve lift profiles should be
investigated and modified for accuracy
When a torque validation was attempted for this case study, the predictions could not be
validated to a reasonable accuracy. To determine whether a given peak or valley was an effect of
the intake or exhaust, prior intake runner length and exhaust primary length experiments were
investigated to understand the associated trends. The modeled geometry of intake and exhaust
ducting did not seem to be at fault. To further troubleshoot the problem, a sensitivity study was
performed on intake valve opening location. The purpose of this sensitivity study was to
determine the extent of change in torque curve shape due to change in intake valve opening
location (in crank degrees). The valve opening point was originally input into the model based on
engine specifications of questionable accuracy that were found online. Figure 13 shows how well
the model output correlates to empirical results. The predictions are satisfactory considering the
limited amount of model input data available. A model cannot be expected to be highly accurate
without combustion analysis. The important thing is that this model can predict approximate
torque output as well the locations of peaks and valleys in the torque curve. The model does
predict the torque peaks seen at 5300 rpm, 8700 rpm, and 10300 rpm. If more empirical data was
available, the model could be correlated to errors as low 1-3 percent.
Engine Modeling 28
Figure 13: Torque Validation for Naturally Aspirated Model
Engine Modeling 29
simulation while plenum geometry is better designed through flowbench testing and 3D CFD
studies. Design considerations such as cylinder-to-cylinder flow distribution and flow losses
through the plenum are generally not predicted very well with one-dimensional simulation.
Model predictions have not shown the major cylinder-to-cylinder flow variations that occur due
to manifold plenum geometry. The orientation of the plenum inlet and the plenum shape highly
affect the cylinder flow distribution.
Engine Modeling 30
Figure 14: Intake Runner Length Study
It can easily be seen that long-length intake runners significantly improve performance at engine
speeds in the range of 6000-8000 rpm while causing severe losses above 10,000 rpm. The
current intake runner length is 7.4”. From this experiment, it was determined that an intake
runner length of 9-10 inches would be best for overall torque output. The mid-range torque is
high while high-rpm losses are minimal.
Engine Modeling 31
Figure 15: Intake Tubing Diameter Study
Engine Modeling 32
Figure 16 illustrates the effects of using tapered runners compared to the standard constant-
diameter intake runners. The results indicate that a tapered runner will show very similar
performance to that of constant-diameter runner with a diameter that is in between the diameters
of the tapered runner. It can be seen that the 1.250” to 1.625” taper behaves very similarly to the
1.500” constant-diameter runner. Simulation results indicate that tapered runners are not
necessary. However, flowbench results would be needed to validate this; as the 3D flow at
bellmouth inlet in the intake manifold is not adequately characterized by 1D models.
Engine Modeling 33
Figure 17: Design Options and Terminology for Inline 4 Exhaust Headers
Engine Modeling 34
Figure 18: Exhaust Header Primary Length Study
Engine Modeling 35
Figure 19: Exhaust Header Primary Tubing Diameter Study
The secondary tubing geometry does not have nearly the effect on torque output that the primary
tube geometry has. Figure 20 indicates that a larger secondary tubing diameter slightly improves
mid-range torque. However, the secondary tubing also has a large effect on sound output. The
tubing diameter can be guided by the muffler design knowing that for each incremental downsize
in tubing diameter, there is a small loss in mid-range torque. Figure 21 presents the effects of the
secondary length. The secondary length has only minor effects on torque output and should be
selected based on engine packaging constraints.
Engine Modeling 36
Figure 20: Secondary Tubing Diameter Study
Engine Modeling 37
Figure 22: Collector Length Study
An experiment was performed to determine if the y-junction length parameters for the exhaust
collector have a significant effect on the pressure wave tuning of the engine. The results shown
in Figure 22 indicated that those length parameters do in fact affect the pressure wave timing to a
similar extent that was seen in primary length study. However, this study was only performed for
model calibration reasons, not for engine design studies. The actual design of the collector
should be based on minimizing flow losses without concern of component length.
Engine Modeling 38
know that a 20” primary 4-1 design is not the optimal exhaust system. They want to know why a
20” length is chosen and why it’s better than alternative designs for the vehicle. With that said,
an exhaust system design that is guided by the Formula SAE competition and the team’s recent
goals for major weight reduction will be recommended. Gains available from a slightly heavier
design will also be presented for comparison purposes. Table 2 shows the geometry of three
designs for the exhaust headers: the current design on the 2007 engine, a recommended design
for 2008, and long-length alternative. The relative torque increases for these header designs is
presented at the end of this chapter.
* Can be designed exclusively for packaging due to engine output showing low-sensitivity to these parameters
Engine Modeling 39
Figure 23: Compression Ratio Study
Engine Modeling 40
throttle plate is ignored in the model so the actual bore diameter may need to be slightly larger
than predicted due to slight flow losses that may occur due to the throttle plate. A 35 mm
diameter throttle body is recommended based on these model predictions and the above
assumption.
Engine Modeling 41
5. CONCLUSIONS/RECOMMENDATIONS
The objective of the naturally aspirated case study was to investigate the potential for
performance increases for the 2007 Formula SAE competition vehicle, to be applied for the 2008
competition. Recommended engine parameter values have been compiled in Table 3. Some
values are recommended for reasons that are practical for team implementation. For example,
model results predict that a throttle body of 33-35 mm diameter would provide the best
combination of throttle response and power output. The current throttle body has a 38 mm
diameter. It is recommended that this design be reused as it is not practical for it to be redesigned
for only very marginal performance gains. The performance gains associated with the new intake
and exhaust system recommendations are shown in Figure 25. Line number 3 is the
recommended engine package for 2008 and line number 1 is the current 2007 design.
Engine Modeling 42
Table 3: 2008 NA Engine Recommendations
Manifold Type Side Inlet, Tapered-Log Type Plenum Center Inlet, Trapezoidal Distribution
Manifold
Header Type 4-1, for weight savings and 4-1, for weight savings and packaging
packaging
As previously discussed, the flow losses and flow distribution in the intake manifold are not
accurately modeled with the one-dimensional code unless flow-loss coefficients are input based
on test measurements. The 2007 manifold design utilized a side-entry manifold type. While the
one-dimensional code did not predict major flow-distribution problems, engine testing did. The
flow variations were so severe that if fueling was not controlled on a cylinder-to-cylinder basis;
one or more cylinders would not even operate due to the inaccurate fueling. Figure 26 shows the
two manifold types. The side-entry manifold, shown on the left, is similar to design that was
tested and currently installed on the 2007 vehicle. The center-inlet manifold, shown on the right,
provides a much more direct passage to all cylinder runners. It is recommended that the 2008
engine package uses a manifold of this type to provide better flow distribution and better overall
cylinder filling. The orientation of the inlet to the plenum can be modified and for sound test
reasons, should be directed away from the exhaust outlet.
Engine Modeling 43
Figure 26: Manifold Type Comparison (for illustrative purpose only)
Future work on the naturally aspirated model should be focused on the configuration of crank-
angle resolved measurements of in-cylinder pressure and pressure traces in the intake and
exhaust system. All data needed to properly define the engine model should be acquired. Once
the model is validated, the resulting accuracy of the model and the source of error should be
documented.
Engine Modeling 44
6. REFERENCES
Engine Modeling 45