SAE 2004-32-0029

JSAE 20044316

Residual Gas Measurements in a Utility Engine

B. P. Albert and J. B. Ghandhi
Engine Research Center, University of Wisconsin-Madison

2004 Small Engine Technology Conference

Graz, Austria
September 27-30, 2004
400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (724) 776-4841 Fax: (724) 776-5760 Web: www.sae.org

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise,
without the prior written permission of SAE.
For permission and licensing requests contact:
SAE Permissions
400 Commonwealth Drive
Warrendale, PA 15096-0001-USA
Email: permissions@sae.org
Fax:
724-772-4028
Tel:
724-772-4891

For multiple print copies contact:

SAE Customer Service
Tel:
877-606-7323 (inside USA and Canada)
Tel:
724-776-4970 (outside USA)
Fax:
724-776-1615
Email: CustomerService@sae.org
Copyright 2004 SAE International
Copyright 2004 Society of Automotive Engineers of Japan, Inc.
Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE.
The author is solely responsible for the content of the paper.
Printed in USA

2004-32-0029 / 20044316

Residual Gas Measurements in a Utility Engine

B. P. Albert and J. B. Ghandhi
Engine Research Center, University of Wisconsin-Madison

Copyright 2004 SAE International and Copyright 2004 Society of Automotive Engineers of Japan, Inc.

ABSTRACT
The residual gas fraction was measured in an air-cooled
single-cylinder utility engine by directly sampling the
trapped cylinder charge during a programmed misfire.
Tests were performed for a range of fuel mixture
preparation systems, cam timings, ignition timings,
engine speeds and engine loads. The residual fraction
was found to be relatively insensitive to the fuel mixture
preparation system, but was, to a moderate degree,
sensitive to the ignition timing. The residual fraction was
found to be strongly affected by the amount of valve
overlap and engine speed. The effects of engine speed
and ignition timing were, in part, due to the in-cylinder
conditions at EVO, with lower temperatures favoring
higher residual fractions. The data were compared to
existing literature models, all of which were found to be
lacking. The effect of the threshold value used to define
lift was investigated, and a modified form of the Yun and
Mirsky correlation, which was relatively insensitive to lift
threshold, was proposed.

INTRODUCTION
Small engine applications often demand engine designs
that emphasize power, which is governed by the
engines ability to breathe at wide-open throttle (WOT).
Thus, small engines are equipped with cams designed
to ensure maximum charging efficiency at WOT. The
resulting valve lift, phasing, and timing of high
performance cam profiles correspond to poor engine
performance at low speeds and light loads, which
manifests itself in poor combustion stability and high
pollutant emissions. The poor performance at low
speeds and light loads can, in part, be attributed to the
presence of high levels of exhaust gas residuals (EGR)
trapped in the cylinder. It is important to understand
how EGR affects the combustion process and pollutant
emissions during light load operation since residualtolerant cam designs could help improve the combustion
process and resulting emissions from small engines.
Previous research has documented how EGR affects
combustion stability and pollutant emissions, however,
the majority of past research has been conducted on

automotive engines [1].

Small engine intake
components and geometries are considerably different
than automotive engine intake systems. Automotive
applications typically have an intake plenum
downstream of the throttle from which individual intake
runners feed each cylinder.
Small engine intake
systems do not incorporate a large intake plenum. In
addition, the throttle on small engines is typically
mounted directly to the intake runner, as close as 10 cm
from the intake valve. This introduces variations in
intake fluid flow, flow control, and overall in-cylinder
charge mixing.
At part load operation the intake plenum on large multicylinder engines is at a known vacuum. The vacuum of
the intake plenum in relation to the barometric pressure
of the engine exhaust produces a known pressure
gradient. The pressure gradient acts as the primary
driver to backflow exhaust gases from the exhaust
runner back into the cylinder and intake runners. The
only intake component on a small engine that realizes a
vacuum is the intake runner. The relatively small
volume of the intake runner on a small engine negates
some of the potential for exhaust to backflow into the
intake system. This may introduce some variation in
regards to controlling trapped EGR, relative to the better
understood control of trapped EGR in automotive
applications. It is of interest to understand how well the
variables used to predict trapped EGR in automotive
engines work with small engines.
The objective of this paper is to investigate the effect of
valve timing on trapped residual fraction in a small utility
engine, and to assess how residual gas fraction models
formulated from automotive engine data fare for small
engines.
The three models used for comparative
analysis include: Yun and Mirsky [2], Fox, Cheng and
Heywood [3], and an ideal gas model. An in-cylinder
sampling valve was used to directly measure the
residual fraction in engine equipped with an adjustable
cam. Steady-state emission measurements of nitrogen
oxides (NOx), unburned hydrocarbons (uHC), carbon
monoxide (CO), carbon dioxide (CO2), and oxygen (O2)
were made for each test condition.
In addition,
measurement of the in-cylinder CO2 concentration was
performed to determine the residual fraction. Finally,

different fuel mixture preparation systems and spark

timings were tested to investigate how combustion
progress and charge mixedness influence EGR.

EXPERIMENTAL APPARATUS
ENGINE
The test engine used was a Briggs and Stratton Intek
engine (Model 110600). This is an air-cooled, singlecylinder, four-stroke, vertical shaft, overhead valve
engine that is rated at 4.8 kW. The engine specifications
are listed in Table 1.
Table 1 Engine Specifications
Bore

68.26 mm 2.6875 in

Stroke

46.00 mm 1.7969 in

Connecting Rod Length

88.57 mm 3.487 in

Displacement

167.4 cm3 11.58 in3

Compression Ratio

7.0:1

The engine used for this experiment was equipped with

an adjustable cam that allowed the opening and closing
timing of the cam lobes to be varied via an access panel
on the side of the engine. The production cam lobes
were split in half, giving two halves each having the
stock cam profile, that butterfly apart to vary the timing
and overall duration. Each lobe was limited to 60q of
advance or retard. Rotating the lobes clockwise or
counter-clockwise retards or advances the cam
timing/phasing relative to the stock design.
The engine head for this study was modified to
accommodate an in-cylinder pressure transducer and
the sampling valve. The sampling valve was installed in
the stock spark plug location. The engine head was
modified to allow a new spark plug location. Due to
space constraints a Champion RDJ7Y spark plug was
used in place of the standard Champion RC12YC spark
plug. A spark gap of 0.76 mm was used for all tests.
Tests were performed on the engine to compare the
performance of the new spark plug and its location
relative to the production plug and location. Ignition
timing and energy were held constant. The engine
suffered poorer performance at WOT with the new spark
plug and location, however, all part-load conditions
tested produced similar performance for both plugs.
Since the scope of this project was to investigate the
effect of EGR at part-load conditions, the new spark plug
and location was used for all tests in order to
accommodate the sampling valve.
Engine speed and load were controlled by a DC electric
dynamometer.
Combustion air to the engine was
delivered from compressed, dried shop air that was
regulated and filtered. The airflow rate was measured
with a Merriam laminar flow element (LFE) (model

50MH10). The test fuel was an EPA Tier II EEE fuel

supplied by Haltermann Products. Air-to-fuel ratios
(AFR) were measured with a wideband AFR analyzer
(Horiba model MEXA-110). In addition, a Horiba
emissions console measured the dry concentration of
NOx, uHC, CO, CO2, and O2. Both the AFR analyzer
and exhaust concentration measurements were used to
calculate fuel flow from the measured air flow.
Engine events and parameters were controlled using a
MotoTron engine control system. The MotoTron engine
control system was comprised of a Motorola PCM0312M
engine control unit (ECU), MotoTune version 8.4.2.5
software, and a RS232 communications cable. The
engine control system allowed for full control of the
engine events and parameters. For the scope of this
project, it was used for fuel injection timing and duration,
air injection timing and duration (using the air-assist fuel
injection, see below), ignition timing and energy, and to
activate the in-cylinder sampling valve.
The cylinder pressure was measured with a Kistler 6061
piezoelectric transducer coupled with a Kistler 5010
charge amplifier. The entire system was calibrated
using a dead-weight tester. The intake port pressure
was measured downstream of the throttle using an
Omega PX303-200A5V absolute pressure transducer.
The intake port pressure was used to peg the cylinder
pressure at BDC of the intake/compression stroke. The
ratio of the intake runner volume to displaced volume
was 0.21.
FUEL SYSTEMS
Three different fuel mixture preparation systems were
investigated. The objective of the different fueling
strategies was to investigate how in-cylinder fuel
distribution affects the amount of trapped EGR. The
different fuel systems were a carburetor, a carburetormounted liquid fuel injection system (CMLFI), and a
homogeneous mixture system (HMS). All three fuel
systems were supplied by the same fuel system.
The first mixture preparation system was a modified
stock carburetor. The main jet was replaced with a
modified jet that incorporated a needle valve for fuel flow
control. The carburetor had an idle bypass circuit to
allow operation at low loads; however, the idle bypass
had no flow control. The idle bypass circuit therefore,
controlled the AFR at highly throttled conditions. As a
result, AFRs were typically lean of stoichiometric when
the idle bypass was fueling the engine.
The second mixture preparation system investigated
was a carburetor-mounted liquid fuel injector (CMLFI).
The objective of the CMLFI system was to mimic
carburetor fuel delivery, but with the precise metering
capabilities of a fuel injector.
The CMLFI was
constructed from a carburetor body and a four-hole,
automotive-type fuel injector. The fuel injector was rated
at 8.62 kg/hr at 241 kPa gage fuel pressure. The float
bowl, float, float needle valve, and spring retaining

system of the stock carburetor were all removed to

install the injector, and all air bleed holes in the
carburetor were sealed with high temperature epoxy. A
retaining bracket was constructed to hold the fuel
injector in place and supply fuel to the fuel injector
orifice. The retaining bracket also integrated a siphon
tube that replaced the emulsion tube of the stock
carburetor. The siphon tube was approximately 20 mm
in length with a residence volume of 38 mm3. The
retaining bracket/siphon tube was mounted to the
carburetor throttle body. By mounting the fuel injector to
the throttle body of the carburetor via a siphon tube, fuel
delivery similar to the stock carburetor has been
observed [4].
Control of the fuel injector timing and pulse width was
accomplished via the engine control system. The fuel
injector timing and duration was calibrated to occur
centered on the intake stroke, at the time of maximum
piston speed. The injection duration was limited to 120
crank angle degrees by adjusting the fuel pressure (34.5
103.4 kPa gage) in order to maintain a fuel flow event
similar to a stock carburetor.
The third mixture preparation system investigated was a
homogeneous mixture system (HMS). The objective of
the HMS was to produce a completely vaporized, well
mixed mixture of air and fuel in order to ensure there
was no fuel/air stratification of the intake charge. A
homogeneous prevaporized mixture was achieved by
using an Orbital air-assist direct in-cylinder injector to
inject a well atomized spray approximately one meter
upstream of the throttle. The air-assist injector was
directed upstream into the air stream at a 45 angle.
The fuel-air mixture then entered a chamber that was
externally heated to promote vaporization. A similar
system has been found to produce excellent results in
optical studies [5]. The air injector duration and the fuelair delay (FAD) were maintained at 3 and 4 ms,
respectively. The fuel injector controlled the mass of
fuel delivered. An intake charge temperature (measured
just upstream of the throttle) of 52 C was found to be
sufficient to ensure complete vaporization. This intake
temperature was, therefore, used for all of the mixture
systems tested for consistency.
SAMPLING VALVE SYSTEM
An in-cylinder sampling valve (SV) was used to acquire
a sample of the trapped cylinder contents during fired
operation. The SV was designed to capture a sample
large enough to be representative of the in-cylinder
contents. The complete characterization of the SV can
be seen in [6, 7]. The SV was a poppet-style, solenoidactuated valve, constructed primarily from stainless steel
to reduce heat transfer to the solenoid. The SV installs
into a 14 mm spark plug hole and has a 10 mm diameter
poppet. The maximum lift of the poppet was limited to 1
mm due to the flow restriction within the valve body. A
Kaman Instruments model 0.5U displacement sensor
was mounted to the SV to monitor the poppet lift. The

sensor range was only 0.5 mm, therefore, it was only

used to determine valve opening and closing times.
The SV actuator was a Ledex-Dormeyer model 5SF-2X
low-profile linear solenoid. The engine control system
was programmed to actuate the SV during the
compression stroke of a programmed misfire. The
program would inhibit the spark while simultaneously
activating the SV. The number of fired cycles between
the sampled cycles was user adjustable. In this study
there were 15 and 10 fired cycles between sampled
cycles for wide open throttle (WOT) and part load
operation, respectively. The opening time and duration
were adjusted for each running condition to maximize
the sample flow through the CO2 analyzer.
A soap-bubble leak test was performed on the SV during
engine operation to assess the sealing of the valve seat.
The maximum speed (3060 RPM) and load (100%)
operating condition, which was considered to be the
worst case for leaks, was tested. The measured leak
flow rate was less than 2% of the sample flow rate, and
was deemed to be sufficient.
ENGINE OPERATING CONDITIONS
The engine was tested over a wide range of operating
conditions, see Table 2. Three speeds (3060, 1750,
1200 RPM), and three loads (WOT, 25, 10%) were
tested. For each speed/load condition, four valve
overlap periods were tested (0, 50, 90, 130 degrees),
and they are discussed more fully below. All three fuel
distribution strategies (carburetor, CMLFI, HMS) were
tested for each speed/load/valve timing condition. The
part load values were referenced to a maximum load
condition, which was defined as: 3060 RPM, WOT, AFR
of 12.5:1, carburetor fuel delivery, no intake air heating,
and an intake surge tank pressure at standard pressure
(101.3 kPa).
The following conditions were held
constant for all operating conditions: AFR of 12.5:1,
intake surge tank pressure at standard pressure (101.3
kPa), intake air temperature of 52 C, and exhaust surge
tank pressure at 1 bar (100.0 kPa).
Table 2 Operating Conditions
Ignition
Cam
Fuel System
Speeds (RPM)
Load (%)

Fixed / Optimized
Zero / Stock / Intermediate / Max
Carburetor / CMLFI / HMS
3060
1750 / 1200
100
25
10
25
10

The production version of this engine has a fixed ignition

timing of 19 BTDC. In this study tests were performed
for both the production ignition timing and an optimized
ignition timing for each operating condition. The ignition
timing for a given operating condition was optimized by
maximizing the load for a given engine speed and
throttle position, then re-adjusting the throttle with fixed

ignition timing to achieve the targeted load. Table 2

summarizes all of the operating conditions for this study.
The four cam timings tested will be termed the zero,
stock, intermediate, and maximum cases, respectively,
which refers to the amount of valve overlap. Table 3
summarizes the resulting cam events for each case,
where a valve lift of 0.25 mm (0.010 in) was chosen as
the reference value for determining the opening and
closing events. For this study, the valve overlap period
was symmetrically adjusted; for a given amount exhaust
valve close (EVC) advance there was an equal amount
of intake valve open (IVO) retard. The exhaust valve
open (EVO) and intake valve close (IVC) timings were
maintained constant for all cases with the exception of
the zero overlap case. Due to the design of the
adjustable cam, in order to achieve zero degrees of
overlap the exhaust cam lobe was slightly retarded and
intake cam lobe was slightly advanced. The full lift
profiles for the four cases can be seen in Fig. 1.
The WOT performance varied significantly with the cam
timing. The indicated mean effective pressures at 3060
RPM and WOT condition were 562, 740, 732, and 679
kPa for the zero, stock, intermediate and maximum
timings, respectively.

concentrations were corrected by a dry-to-wet factor.

The dry-to-wet factor for the exhaust measurement, Kexh,
was determined from the 5-gas emission measurements
in the standard method [6], but the dry-to-wet factor for
the compressed charge, Kcc, had to be evaluated based
on Kexh and equation (1) since CO2 was the only gas
constituent measured in the compressed charge. The
compressed charge dry-to-wet factor is then

RESIDUAL GAS FRACTION MODELS

The three residual gas fraction models investigated for
comparative analysis relative to the measured EGR
included: Yun and Mirsky [2], Fox, Cheng and Heywood
[3], and an ideal gas model.
The Yun and Mirsky correlation for estimating the
trapped residual fraction is

yr

IVC

EVO

EVC

OF
[q/m]

Zero

358

629

79

361

0.11

Stock

335

602

106

385

3.66

Int.

311

602

106

406

12.31

Max

297

602

106

425

27.26

RESIDUAL GAS ANALYSIS

RESIDUAL GAS FRACTION

xCO2 ,cc
xCO2 ,exh

where the compressed charge includes the inducted air

and fuel plus the trapped exhaust gas, and corresponds
to the sample that was acquired by the SV. The mole
fractions of CO2 in the compressed charge and exhaust
were determined from the measured concentrations of
CO2 in the compressed charge and exhaust. Since the
NDIR used to measure the CO2 concentration required
the measurements be performed on a dry basis, both

In

Ex

0
0

(1)

VEVC U EVC

VEVO U EVO

360
CA (deg)

Ex

720

0
0

360
CA (deg)

720

(3)

Ex

In

Lift (mm)

xr

1
ne

The residual gas mole fraction, xr, can be determined

from the ratio of the mole fraction of CO2 in the
compressed charge, xCO2 ,cc , to the mole fraction of CO2
in the raw engine-out exhaust, xCO2 ,exh via

VEVC PEVC

VEVO PEVO

Lift (mm)

IVO

(2)

1  xr

The derivation for equation (2) can be seen in Appendix

A. Under the standard assumption that the molecular
weight of the trapped and exhaust mixtures are the
same, the residual mole fraction is equivalent to the
residual mass fraction, yr.

Table 3 Cam timings tested. All crank angle values are

referenced to TDC of the compression stroke.

Case

>1  KEXH xr @

KCC

In

360
CA (deg)

Ex

Figure 1 Cam profiles investigated.

720

In

360
CA (deg)

720

The Fox, Cheng, and Heywood correlation for estimating

the trapped residual fraction is

yr

OF Pint
1.266

N Pexh

M Pint

0.87

Pexh  Pint

 0.632

Pexh

0.74

OF

0.3

0.2

0.1

(4)

rc

where N is the engine speed, OF is the overlap-factor

defined below, M is the equivalence ratio, and rc is the
compression ratio. The overlap factor is given by
IV EV
EVC

Dint Lint dT  Dexh Lexh dT

IVO
IV EV

Vdis

0.4

y r,carb

where VEVC and VEVO are the cylinder volume at EVC

and EVO, PEVC and PEVO are the cylinder pressure EVC
and EVO, and ne is the polytropic coefficient for the
expansion. For this study the expansion polytropic
coefficient was assumed equal to the compression
polytropic coefficient because of the retarded heat
release for the high residual fraction, part-load
conditions.

(5)

0.0
0.0

0.1

0.2
y r,HMS

0.3

0.4

Figure 2 Comparison of measured residual fraction for

the HMS and carbureted fuel delivery cases.

RESULTS
where Dint and Dexh are the intake and exhaust inner
valve seat diameters, Lint and Lexh are the lift of the
intake and exhaust valves, and Vdis is the displaced
volume of the engine.
The ideal gas model is a simple method to predict
residual gas fraction based on partial pressures. The
basic assumptions are that the fresh charge temperature
is equal to the measured intake temperature, and the
residual gas temperature is equal to the measured
exhaust temperature. It also relies on the measured
mass of fuel and air delivered to the engine. Under
these assumptions, the residual mass fraction is given
by
yr

mr
mr  mair  mfuel

(6)

with the residual mass, mr, determined from successive

application of the ideal gas law
mr

VIVC
RTexh

 mair  mfuel  RTINT

PIVC 
VIVC

(7)

The second term in the brackets represents the partial

pressure of the fresh charge at IVC, with the gas
constant, R, evaluated using the molecular weight of air.
It is important to note that the pressure values in all of
the above residual fraction models is the absolute
pressure. Thus, pegging the cylinder pressure correctly
is imperative in using these correlations.

EFFECT OF FUEL SYSTEM

The effect of the fuel system on the measured residual
fraction can be seen in Figs. 2 and 3. The data for the
HMS was chosen for the abscissa because the HMS
was found to produce the widest stable operating range
for the engine. The results from the CMLFI, Fig. 3, were
found to be in excellent agreement with the HMS results
for all of the conditions tested. At low to moderate
residual fractions the carburetor results agree very well
with the HMS results. However, for conditions with a
relatively high residual fraction the carburetor provides
slightly lower values of residual fraction. The difference
is, in part, thought to be the result of the poorer (relative
to the HMS) combustion quality observed with the
carburetor at high residual fractions.
EFFECT OF VALVE OVERLAP
Changing the valve overlap effect produced the
expected results higher levels of overlap produced
higher residual fractions. These results can be seen in
Fig. 4 for the HMS data. The effect of overlap was more
pronounced at the higher valve overlap conditions. The
effects of speed and ignition timing, which are included
in Fig. 4, are discussed separately below.

0.4

0.3

0.3

0.2

0.2

yr

yr,CMLFI

0.4

0.1

0.1

0.0

0.0
0.0

0.1

0.2
y r,HMS

0.3

0.4

Figure 3 Comparison of measured residual fraction for

the HMS and CMLFI fuel delivery cases.

EFFECT OF IGNITION TIMING

Figure 5 shows the effect of ignition timing on the
residual fraction for all of the data. There is a clear bias
to the data, with the advanced ignition timings providing
a higher residual fraction. The difference in combustion
phasing between the two cases is believed to be the
cause for this behavior. With the standard ignition timing
the part-load cases are highly retarded. This produces a
higher in-cylinder temperature and pressure at the time
of EVO, the latter being measured and the former

Valve Overlap
zero
stock
intermediate
max
0 1 2 3 4 5 6 7 8 9 10 11 12
Case

Figure 4 Comparison of valve overlap amount on

residual fraction for the HMS data only. The cases
correspond to speeds of 3060 (0-4), 1750 (5-8) and
1200 (9-12) RPM, with loads of 100% (0),
25%(1,5,9) and 10% (3,7,11). The odd numbered
cases are for the fixed ignition timing and the next
larger even number corresponds to the advance
ignition timing for the same case
observed by a similar relation in exhaust temperature.
The higher temperatures give rise to lower residual gas
density, and thus a lower residual fraction for an
equivalent volume. The higher pressure at EVO is
thought to aid in the blowdown and scavenging event.
EFFECT OF SPEED
The effect of engine speed was found to be significant.
Figure 6 shows data acquired with the HMS and the
stock ignition timing. The data for a given load and cam
timing are connected, and clearly show a first-order
effect of engine speed. The effect of speed is not
thought to be entirely due to transient effects. As the
engine speed increases from 1200 to 3060 RPM at a
fixed load, the exhaust temperature increases by almost
50%, reducing the residual gas density. Similarly, as the
speed increases the pressure at EVO increases due to
the reduction in heat transfer, setting up different blow
down phenomena. Additionally, there could be wave
dynamics in the runners that are affecting the charging
performance of the engine.

0.4

yr,Advanced Ignition Timing

HMS Data

0.3

0.2

0.1

0.0
0.0

0.1
0.2
0.3
y r, Stock Ignition Timing

0.4

Figure 5 Effect of ignition timing on the measured

residual fraction.

the exhaust temperature due to heat transfer to the walls

during gas exchange. This effect would result in an
under-estimation of the residual mass, and would be
more pronounced at high residual fraction, which is
consistent with Fig. 7.
Attempts to improve the
prediction capability of the Ideal Gas model using the
measured oil temperature as a surrogate for the
chamber wall temperature did not produce significantly
improved results.

0.40
0.35

yr

0.30
0.25

0.4

0.20
0.15

0.3

1000

1500

2000 2500
Speed [RPM]

3000

3500

Figure 6 Effect of engine speed on residual fraction.

The data are from the HMS, the filled symbols are
10% load and the open symbols are 25% load. The
valve overlap conditions are zero (square), stock
(circle), intermediate (hourglass) and maximum
(triangle).

y r,IG

0.10
0.2

0.1

0.0
0.0

COMPARISON OF MODEL PERFORMANCE

As described above, three models were evaluated for
their ability to predict the residual fraction. The Yun and
Mirsky and Ideal Gas models were implemented directly
as described. The Fox, Cheng and Heywood model,
however, could not be directly implemented because the
intake pressure, Pint, is not uniquely defined for small
engines with limited intake volume. For this model the
intake pressure was taken to be the pressure at bottom
dead center of the compression stroke.
The results of the Ideal Gas model, yr,IG, are plotted in
Fig. 7 against the measured residual fraction. The
correlation between the Ideal Gas prediction and
measured residual fraction is, in general, quite poor. At
low residual fraction the Ideal Gas model over-predicts
the residual fraction, while at high residual fraction the
model under-predicts the value.
The cause for this behavior is believed to be the
assumptions made regarding the temperatures of the
fresh and residual charges. The measured intake gas
temperature will be lower than the trapped fresh charge
temperature due to heat transfer from the hot walls
during the intake stroke. Therefore, the partial volume of
the intake charge is under-estimated resulting in a
higher than expected residual fraction. This problem is
expected to be worse at low residual fraction, and this is
consistent with the observation in Fig. 7. Also, the
residual gas temperature is expected to be lower than

0.1

0.2
yr

0.3

0.4

Figure 7 Comparison of ideal gas model to measured

results.
The comparison of the Yun & Mirsky model prediction to
the measured residual is shown in Fig. 8. The trend in
the results is captured reasonably by the Yun & Mirsky
model, but the model substantially under-predicts the
residual fraction. The basic premise of the Yun & Mirsky
model is that the residual gas undergoes a polytropic
expansion from EVO to EVC. The EVC to EVO density
ratio is then given by the second term in Eq. 3, and
when coupled with the cylinder volumes it results in the
necessary mass ratio.
The use of the polytropic
coefficient instead of the isentropic coefficient accounts
to a small degree for the effect of heat transfer.
In the real case, however, the effects of heat transfer will
be more significant. Therefore, the second term of Eq. 3
under-predicts the residual gas density. This result is
reflected in the trends of Fig. 8 where the predicted
residual fraction lies below the measured value.

the results.

0.4

MODEL SENSITIVITY
The prediction capabilities of all of the models was less
than desired, so an attempt was made to assess what
the best implementation of the model was within given
constraints. The two main factors that were investigated
were the threshold value used to define of valve opening
and closing, and the polytropic exponent used in the Yun
and Mirsky model.

yr,Y&M

0.3

0.2

The effect of these parameters was assessed by

investigating the root mean square (RMS) error for a
model. The RMS error was calculated for a data set as

0.1

0.0
0.0

0.1

0.2
yr

0.3

0.4

Figure 8 Comparison of Yun and Mirsky model to

measured results.
The results from the Fox, Cheng and Heywood model
are not very good, as shown in Fig. 9. The model
significantly over-predicts the residual fraction for all of
the valve overlap conditions except for the zero overlap
configuration.
This poor agreement is not overly
surprising given that the model has been applied well
outside its intended range. The overlap factor of the
data used in [3] was 0.52 < OF < 1.43, which is similar to
the zero overlap case, but well below the other cam
phasings considered here, see Table 3. Other factors,
such as the choice of the intake pressure, also affected
the results, but detailed investigation revealed that the
first term in Eq. 4, which contains OF, was dominating

1.2

Overlap
0 - zero
1 - stock
2 - alternate
3 - max

1.0

yr,FCH

0.8

33
33 2
22
33 3 2
222

1 1
1
1
1 1
1
1111
11
111
11
1
1
11 11
1
11
00
3
000 0 00
0
0
0
0
0
0
0
0
0
0
0
0
0
3 00
22 3
11 00
2
2222
2
22

0.6
0.4
0.2
0.0
0.0

0.1

0.2
yr

0.3

0.4

Figure 9 Comparison of Fox, Cheng and Heywood

model to measured results.

ErrorRMS

1
n

y
i 1

i
r

 y ri ,p

(8)

where n (=140) is the number of operating conditions,

and yr,p is the residual fraction predicted by the model.
The data, normalized by the value obtained with a 0.25
mm lift, are shown in Fig. 10. It is observed that the
RMS error decreases for lift thresholds less than 0.25
mm for both the Yun & Mirsky model and the Ideal Gas
model, therefore, better accuracy is obtained for low
threshold lift values.
As discussed previously, the polytropic exponent in the
Yun & Mirsky model does not accurately take into
account the heat transfer from the residual gas. A best
fit value for the exponent, nbf, in Eq. 3 was determined
for each value of lift threshold, and the results are shown
in solid lines on Fig. 10. The RMS error for the
optimized case was normalized by the value from the
direct implementation of the Yun and Mirsky model at a
lift threshold of 0.25 mm. There are two things to note.
First, the RMS error for the optimized Yun & Mirsky
model is everywhere lower than the standard
implementation.
Further, the error is relatively
insensitive to the lift threshold chosen. Secondly, the
best-fit exponent is always larger than the isentropic
value of 1.4. This indicates that the heat transfer is
under-predicted in this simple model as argued above.

Yun & Mirsky

Ideal Gas
Optimized Y&M

1.2

2.6

modified Yun and Mirsky method was relatively

insensitive to the value of lift threshold chosen.

2.4

ACKNOWLEDGMENTS

1.1

2.0

Support for this work was provided by the Wisconsin

Small Engine Consortium. The efforts of Matt Manthey
and Blake Suhre of MotoTron are gratefully
acknowledged.

0.8

1.8

REFERENCES

0.7

1.6

2.2
1.0

nbf

ErrorRMS / ErrorRMS(0.25 mm)

1.3

0.9

1.
2.

0.6
0.0

0.1

0.2

0.3
0.4
Lift [mm]

0.5

0.6

Figure 10 Comparison of the lift threshold level on the

RMS error of the residual gas predictions. The
Optimized Y&M model exponent is shown on the
right axis.

3.

4.

CONCLUSION
The residual gas fraction was measured in an air-cooled
single-cylinder utility engine by directly sampling the
trapped cylinder charge during a programmed misfire.
The residual fraction was determined by comparing the
trapped CO2 concentration to that in the exhaust. Tests
were performed for a range of fuel mixture preparation
systems, cam timings, ignition timings, engine speeds
and engine loads. The results were compared to
predictions from models available in the literature.
The residual fraction was found to be relatively
insensitive to the fuel mixture preparation system, but
was to a moderate degree sensitive to the ignition
timing. The residual fraction was found to be strongly
affected by the amount of valve overlap and engine
speed. The effects of engine speed and ignition timing
were, in part, due to the in-cylinder conditions at EVO,
with lower temperatures favoring higher residual
fractions.
None of the three models investigated was able to
accurately predict the residual fraction over the entire
range of conditions tested. The Fox, Cheng and
Heywood model performed well for very low overlap
conditions, but at the cam timing of the production
version of this engine it predicted values approximately
2u too high. The performance deteriorated at higher
overlap conditions due to the extrapolation of the model
past the calibration data set. The Ideal Gas and Yun
and Mirsky models were inherently limited due to heat
transfer to the cylinder walls during the gas exchange
period. An optimized exponent for the Yun and Mirsky
model was proposed as a function of the lift threshold
used to define the valve events. The accuracy of the

5.
6.

7.

Heywood, J.B., Internal Combustion Engine

Fundamentals. 1988: McGraw-Hill Inc.
Yun, H.J. and W. Mirsky, Schlieren-streak
measurements of instantaneous exhaust gas
velocities from a spark-ignition engine. SAE
Paper 741015, 1974.
Fox, J.W., W.K. Cheng, and J.B. Heywood, A
model for predicting residual gas fraction in
spark-ignition engines. SAE Paper 931025,
1993.
Strauss, S. and J.K. Martin, An investigation of
the intake manifold fuel transfer characteristics
of a four stroke engine. SAE Paper 2001-011213, 2001.
Probst, D.M. and J.B. Ghandhi, An experimental
study of spray mixing in a direct injection engine.
Int. J. Engine Res., 2003. 4(1): p. 27-45.
Foudray, H.Z., Scavenging measurements in a
direct-injection two-stroke engine, in Mechanical
Engineering. 2002, University of WisconsinMadison: Madison.
Foudray, H.Z. and J.B. Ghandhi, Scavenging
Measurements in a Direct-Injection Two-Stroke
Engine. SAE Paper 2003-32-0081, 2003.

APPENDIX
Starting from the overall stoichiometric balance
nair ,dry  nfuel

nexh

nexh,dry  nexh,H2O

the exhaust stream dry-to-wet factor is

nexh,dry
K EXH
.
nexh,dry  nexh,H2O

(A1)

(A2)

By definition, the residual fraction is

xr

nair ,dry

nr
 nfuel  nr

(A3)

from which one can write the compressed-charge dry-towet correction factor
KCC

nair ,dry  nfuel  xr [nexh,dry ]

nair ,dry  nfuel  xr [nexh,dry  nexh,H2O ]

Dividing the numerator and denominator by nexh,dry +

nexh,H2O and simplifying gives Eq. 2

(A4)