International Multidimensional Engine Modeling User’s Group Meeting at the SAE Congress 2003
Optimizing the Scavenging System for a Two-Stroke Cycle, Free Piston Engine
for High Efficiency and Low Emissions: A Computational Approach
S. Scott Goldsborough and Peter Van Blarigan
Sandia National Laboratories
ABSTRACT
A free piston internal combustion (IC) engine
operating on high compression ratio (CR) homogeneous
charge compression ignition (HCCI) combustion is being
developed by Sandia National Laboratories to
significantly improve the thermal efficiency and exhaust
emissions relative to conventional crankshaft-driven SI
and Diesel engines. A two-stroke scavenging process
recharges the engine and is key to realizing the
efficiency and emissions potential of the device. To
ensure that the engine’s performance goals can be
achieved the scavenging system was configured using
computational fluid dynamics (CFD), zero- and one-
dimensional modeling, and single step parametric
variations. A wide range of design options was
investigated including the use of loop, hybrid-loop and
uniflow scavenging methods, different charge delivery
options, and various operating schemes. Parameters
such as the intake/exhaust port arrangement, valve
lift/timing, charging pressure and piston frequency were
varied. Operating schemes including a standard uniflow
configuration, a low charging pressure option, a stratified
scavenging geometry, and an over-expanded (Atkinson)
cycle were studied.
The computational results indicated that a stratified
scavenging scheme employing a uniflow geometry, and
supplied by a stable, low temperature/pressure charge
will best optimize the efficiency and emissions
characteristics of the engine. The operating CR can be
maximized through substantial replacement of the
burned charge, while short-circuiting emissions can be
controlled by late fuel introduction. The in-cylinder flows
are important to both NOx and short-circuiting emissions
with inadequate mixing (and resulting temperature
stratification) the predominant driver of NO production,
and fuel penetration to the exhaust valve region the main
cause of unburned hydrocarbon emissions.
INTRODUCTION
In an effort to improve the fuel economy and exhaust
emissions of advanced electrical generators, Sandia
National Laboratories is developing a novel free piston IC
engine [1]. The approach utilizes a free piston, double-
ended cylinder arrangement with a linear alternator
integrated directly into the cylinder’s center section.
Lean (φ~0.35) HCCI combustion at alternating cylinder
ends is used to drive permanent magnets fixed to the
piston, back and forth through the alternator’s coils. The
alternator serves to generate useful electrical power, and
to control the piston’s motion by dynamically varying the
1
rate of electrical generation. Engine startup is also
achieved using the alternator. Charging of the engine’s
cylinders is accomplished using a two-stroke cycle
process. Figure 1 illustrates the engine concept.
Exhaust
Compressor
Intake
Alternator
Windings
Cooling
Fluid
Magnets
Piston
Delivery
Tank
Cooling
Fluid
Intake Manifolds
Figure 1 – Free Piston Engine-Generator
Critical to the engine’s operating efficiency and
emissions capabilities however, are the design and
performance of the scavenging system. Without
adequate charging the HCCI combustion process will be
degraded, becoming less efficient (e.g. non-constant
volume, incomplete, or autoigniting at a low CR) and
producing excessive emissions such as unburned
hydrocarbons (HC) and NOx. In addition, improper fuel
delivery can lead to short-circuiting emissions while
further reducing fuel economy. To address these
concerns a computational investigation was undertaken.
BACKGROUND
Conventional two-stroke engines are plagued by
problems of insufficient charging and high short-circuiting
emissions throughout parts of their operating regimes,
with this generally resulting from the wide range of
speeds and power outputs over which the engines
operate. For the free piston engine however, a much
more narrow range of operating speeds is expected to
be utilized. This is due to the electrical generating
scheme employed by the device; efficient generation will
be achieved by operating at a fixed oscillation rate.
Single speed operation significantly simplifies the
scavenging system design, in effect allowing the
charging process to be optimized about a specific
operating point.
However, several parameters are still critical to the
design. First, the scavenging process must ensure
rapid, TDC combustion, while high compression ratio
should be achieved before combustion initiates. In
addition, the local fuel mixtures must be lean and the gas
temperatures low enough so that NOx formation is
inhibited. Further, lean operation requires that losses in
the engine be minimized so that a high fraction of the
work output can be converted into useful electrical
power. As such, restrictions on the scavenging scheme
(e.g. low pumping power, low blowdown losses, etc.) are
imposed. Two more items are the desire of a long
stroke-to-bore ratio (for adequate TDC clearance at high
CR, and an advantageous surface area to volume ratio),
and mechanical simplicity (since no crankshaft is
available to operate a pump, or valves, etc.).
The thrust of the present study was to develop the
first steps towards configuring an optimal scavenging
system for the free piston engine, based on the efficiency
and emissions goals of the generator. Multi-dimensional
modeling was used to facilitate an understanding of the
gas flow process, and a single step parametric method
was employed to narrow the range of design possibilities.
COMPUTATIONAL TOOLS
KIVA-3V was used for this study where changes to
the code included calculating pertinent scavenging
parameters (e.g. scavenging and trapping efficiencies,
flow rates, etc.), defining the free piston motion,
modifying the intake boundary conditions to accept time
dependent values, and incorporating an intake charge
compression model as a subroutine to the code. HCCI
combustion and NOx generation were simulated using
chemical equations exclusively (4 reduced kinetic, and 6
partial equilibrium, with propane as the fuel); there were
no turbulent mixing parameters, as has been suggested
in Ref. [2].
A significant aspect of analyzing the KIVA results
was the capability to visualize gas motion through the
cylinder. To facilitate this the 3D post-processing
software Ensight was used.
To assess the pumping and friction losses in the
system, the charge compression and piston ring friction
processes were modeled. For the compressor
component this process was modeled assuming zero-
dimensional (0D) behavior (thermodynamic control
volumes allowing quasi-steady inlet and outlet flows)
where important results were the pressure and
temperature histories input into the KIVA-3V code. The
ring friction process (the dominant component of friction
in the free piston engine) was modeled using a one-
dimensional (1D) analysis of the ring-oil-wall system
(assuming Reynold’s equation is applicable) where
important results were the relative variations in friction
work due to changes in piston-cylinder configuration (e.g.
stroke, frequency(ƒ), etc.).
2
OPTIMIZATION PROCEDURE
A single step parametric method was adopted as the
main optimization tool since a wide range of geometric
and operating parameters was to be investigated. The
following procedure was used. First, the performance of
three different scavenging methods (loop, hybrid-loop
and uniflow) was investigated in order to determine the
capabilities and limitations of each. Parameters such as
the charging pressure, intake/exhaust area/timing, and
piston frequency were varied. High scavenging
efficiency (ηsc~0.9) (needed for high CR operation) and
high trapping efficiency (ηtr~1.0) were used as the
metrics. These initial simulations allowed a number of
design options to be eliminated from consideration, and
provided a knowledge base for configuring an optimal
arrangement.
Using the uniflow geometry several charge delivery
options (e.g. delivery tank size, tank temperature, etc.)
were then explored. The effects on both power
consumption and in-cylinder flows were determined.
Following these computations, four select operating
schemes (standard uniflow, low charging pressure / low
frequency, stratified scavenging, and over-expanded
cycle) were analyzed, this time with the overall thermal
efficiency and output emissions as the metrics.
COMPUTATIONAL RESULTS
Scavenging Methods
The arrangements for the three scavenging methods
are illustrated in Figure 2 with the desired flow patterns
for each arrangement included for clarity.
(a) Loop (b) Hybrid-Loop (c) Uniflow
Figure 2 – Scavenging Arrangements
Figure 3 presents the results of the parametric
simulations for these scavenging methods. As can be
seen, the loop and hybrid-loop options produce
unacceptable scavenging, however the uniflow geometry
provides a means of achieving the efficiency and
emissions goals of the engine.
 The expected benefits of these schemes, relative to
the standard arrangement, were:
1
Case II – decreased pumping power, with the
0.95
operating frequency reduced to ensure
adequate charging.
0.9 Case III – increased scavenging efficiency, with
Trapping Efficiency

short-circuiting controlled by late fuel

0.85
introduction; the higher operating
Uniflow
0.8 compression ratio will lead to improved cycle
thermal efficiency.
0.75
Case IV – increased work output through
0.7
recovered blowdown potential.
Loop
The metrics used here were thermal efficiency (ηTH)
0.65 Hybrid-Loop (including the work output of the cycle, the work input to
0.6
the compressor, and friction losses; the total delivered
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 fuel was used to account for fuel losses) and exhaust
emissions (including short-circuited fuel (C3H8) and NO.)
Scavenging Efficiency
For each of these cases a uniflow geometry (4
Figure 3 – Trapping Efficiency vs. Scavenging Efficiency
exhaust valves, 8 intake ports) was used with the
orientations and timings configured based on the findings
To maximize the scavenging performance for this
of the previous sections. For the stratified scavenging
arrangement (ηsc~0.85, ηtr~0.99), the following case four tall ports delivered the initial air charge, and
conditions should be met. The exhaust valve lift and four short ports introduced the premixed fuel-air charge.
timing should be adjusted so that the cylinder pressure A series of iterations was used to arrange the geometries
can blow down to the intake manifold pressure before
to achieve the desired performance (ηsc ~ 0.8, ηtr ~ 0.99
the intake ports open; this will allow maximum recharging
for Cases I, II, and IV; ηsc ~ 0.9, ηtr(fuel) ~ 0.99 for Case III;
with minimal trapping losses. The generation of plug-
power output ~ 20kW).
type flow is best achieved when the incline angles are set
to 0° and the swirl angles set uniformly to about 15°;
Case Case Case Case
there seems to be little change in performance when I II III IV
either an 8-port or 12-port configuration is used. The Bore [cm] 7.62 8.85 7.24 7.62
cylinder’s recharging can be maximized if the scavenging
Stroke [cm] 21.46 29.63 25.56 44.05
time is adequately adjusted; the most effective means of
achieving this is by modifying the piston’s frequency. Effective CR 19:1 19:1 26:1 19:1
EVO [CAD] 116 102 99 142
Charge Delivery System IPO (Air) [CAD] NA NA 132 NA
Simulations with the charge delivery system were
IPO (Fuel/Air) [CAD] 146 130 148 139
conducted to determine parametric effects on power
consumption and in-cylinder flows. The 0D/KIVA-3V IPC (Fuel/Air) [CAD] 208 222 207 214
calculations assumed an internal compressor IPC (Air) [CAD] NA NA 220 NA
arrangement, where the magnets attached to the EVC [CAD] 232 245 248 285
working piston function as a stepped compressor piston
Valve Diameter [cm] 2.54 2.86 2.38 2.54
within a concentric design. Parameters investigated
included the volumetric compression ratio of the Valve Lift [cm] 0.925 0.925 0.925 0.925
compressor, the delivery tank volume, the delivery tank Port Width (Fuel/Air) [cm] 2.00 2.27 2.20 1.93
temperature, and the compressor’s valve areas. The Port Width (Air) [cm] NA NA 1.63 NA
calculations indicated that a large tank supplying a
stable, low temperature charge and utilizing sufficient Port Height (Fuel/Air) [cm] 2.30 5.80 2.70 6.50
compressor flow area will maximize the performance of Port Width (Air) [cm] NA NA 4.70 NA
the scavenging system. Swirl Angle 15 15 15 15
Frequency [Hz] 45 33 45 45
Operating Schemes
Four different operating schemes were investigated Charging Pressure [bar] 1.2 1.1 1.2 1.2
with the objective to maximize the thermal efficiency of Intake φ 0.475 0.475 2.70 0.475
the engine cycle. The effects of the scavenging process
Table 1 – Geometric and Operating Parameters
on the thermodynamic cycle and output emissions were
studied. The schemes included: a standard uniflow
Table 1 lists the geometric and operating parameters
arrangement (Case I), a low frequency / low pumping
for these arrangements. (Here CAD is used to note the
pressure option (Case II), a stratified scavenging
port and valve timings. However, it is only a time
configuration (Case III), and an over-expanded cycle
notation since the free piston engine does not have a
(Case IV).
3
crankshaft to define the piston’s motion. (CAD = (t-tTDC) ·
ƒ ·360.)

Operating Results
The operating results are presented in Table 2. It 100
Case II
can be seen here that the overall thermal efficiency for
Case II is the same as for the standard configuration;

Pressure [bar]
Case III
however, the short-circuiting and NO emissions are
increased (+200% and +250%, respectively). On the
Case I
other hand, Case III has an increased thermal efficiency 10

(+10%), with higher short-circuiting losses (+300%) and Case IV

lower NO (-70%). (The higher short-circuiting emissions
for this case resulted from the imprecision of the iteration
process, and could probably be reduced.) Case IV also
operates with a higher thermal efficiency (+3%), but 1
greater short-circuiting and NO emissions (+200%) 100 1000
result. 3
Volume [cm ]

Case Case Case Case

Figure 4 – Cylinder Pressure vs. Cylinder Volume
I II III IV
ηsc 0.83 0.83 0.93 0.84 Figure 5 presents the flow visualization for the
standard uniflow case for reference. Here Ensight’s
ηtr 0.997 0.995 (A) 0.928 0.994
particle tracing routine was used to track the fresh and
(F/A) 0.988 burned charge flow through the engine. In this figure
mdel [g] 0.96 1.15 (A) 1.03 0.92 small blue spheres represent the fresh charge and small
(F/A) 0.16 orange spheres represent the burned charge. Trailing
lines indicate velocity.
W cycle [J] 480 790 644 697
Similar flow patterns were seen with the other three
W comp [J] 24.6 29.7 45.1 35.6 arrangements; however, two notable changes result as
W fric [J] 19.2 31.2 24.3 59.9 the port heights are increased for Cases II, III and IV.
ηTH 0.475 0.475 0.525 0.491 First, some of the burned charge becomes trapped just
above the piston, with some of this forced into the intake
C3H8 [ppm] 67 133 206 136
manifold as the piston begins its compression stroke.
NO [ppm] 281 707 94 560 For the short ports in Case III this reverse flow is
Power [kW] 18.7 23.1 24.6 25.7 substantial.
A more significant result is a change in bulk flow
Table 2 – Operating Results motion with reduced swirl and decreased in-cylinder
mixing. [For these runs the swirl ratio (SR) dropped from
The results of the simulations are explained by 2.9 for Case I to 1.9 for Case II, 2.7 for Case III and 1.7
studying the operating cycle and cylinder/port flows for for Case IV.] This affects both the short-circuiting and
the four arrangements. The effects of compressor and NO emissions.
friction work on the overall cycle are also important. In terms of short-circuiting, the change in in-cylinder
Figure 4 presents the thermodynamic cycles for the bulk motion and decreased mixing allows greater
four configurations. Comparing these runs, it can be penetration of fuel-rich mixtures to the exhaust valve
seen that for Case II the initiation of HCCI combustion region. In terms of NO emissions, greater temperature
occurs somewhat earlier in the stroke, with more over- stratification results during scavenging and compression,
compression of the burned charge after combustion. with this leading to pre-ignition near the cylinder core.
This leads to the higher NO emissions noted in Table 2, This can be seen in Figure 6 where the maximum
and will be discussed in greater detail below. The cylinder temperature is plotted versus instantaneous
thermodynamic cycle for Case III is very similar to the compression ratio over the engine cycle. For Case I the
standard case, however the increase in compression earliest ignition point is at CR=14:1, for Case II this
ratio before combustion is evident. This is due to the decreases to CR=11:1, and for Case IV this is closer to
lower bulk cylinder temperature at EVC, achieved CR=9:1. For these three arrangements the bulk of the
through more complete flushing of the burned charge. cylinder combusts, as determined by the maximum rate
For Case IV the change in the scavenging cycle is of pressure change, near CR=16:1. (Case III is included
considerable with no blowdown present (though there is in this plot for reference, where the bulk combustion
some pressure drop after EVO/IPO due to continued occurs at about 24:1.)
piston expansion). In this arrangement, there is also
early HCCI combustion with significant compression after
combustion, as seen in Case II.

4
 2500
SR CRop
Case I 2.9 16.8:1
Case II 1.9 15.7:1

Maximum Temperature [K]

Case III 2.7 24.2:1
2000
Case IV 1.7 16.1:1

1500

Case I

1000 Case III

Case II
Case IV

500
0 5 10 15 20 25

Compression Ratio
Figure 6 – Maximum Cylinder Temperature vs.
Compression Ratio

Although this pre-ignition problem has only a small

impact on the cycle efficiency (↓2%), the NO production
increases substantially since the burned charge is over-
compressed to TDC. This is most significant for Case II
since the piston’s velocity is slower and the over-
compression is sustained for a longer period of time.
With regard to fuel distribution for these
configurations, it seems that the in-cylinder flows
similarly dilute the incoming charges through the
scavenging and compression processes. This is
important for the stratified scavenging option. This point
is illustrated in Figure 7 where the maximum and
average dilution ratios for Cases I and III are plotted
versus CAD. It can be seen that even though there are
significant differences between these two arrangements,
the bulk flow dilution during compression yields similar
differences between the maximum and average dilution
ratios at the time of combustion. This seems to indicate
that fuel dispersion might not be problematic with the
stratified scavenging configuration, if a premixed charge
is used.

Case III
0.8
Dilution Ratio

0.6 Case I

Maximum EVC

0.4

0.2
Average

0
100 150 200 250 300 350

CAD
Figure 5 – Scavenging Flow Visualization vs. Crank Figure 7 – Maximum and Average Dilution Ratios vs.
Angle Degree [Case I] Crank Angle Degree
5
Compressor Work
Table 2 above shows that the compressor work
fraction decreases for Case II (-25%) while it increases
for Case III (+37%). This was expected due to the lower
charging pressure and higher delivery ratio, respectively.
However, with respect to ηTH these changes are offset by
changes in the thermodynamic cycle. For Case II the
conversion efficiency is decreased due to early ignition
problems, while for Case III it is higher due to the higher
compression ratio achieved.
Friction Work
Table 2 shows that the only significant change in
friction work fraction occurs with the over-expanded
configuration (+200%). This is due to the increased
piston velocity used for this arrangement. However, the
increased friction work is offset by an increase in the
conversion efficiency, in this case due to the additional
charge expansion.
Design Robustness
A final series of simulations was run with the four
operating schemes to assess the ‘robustness’ of these
designs. The effects of slight variations in the input
conditions of equivalence ratio and piston frequency
(±10%) were investigated in an attempt to simulate
fluctuations seen in actual engine operation.
From these calculations it was seen that the
changes in emissions are more significant for Case II,
relative to Case I. The thermal efficiency for Case II,
however, seems to be slightly more stable for these
variations. For Case III the thermal efficiency is also
stable, while the changes in emissions are comparable
to the standard configuration.
An important result of this simulation series was that
the over-expanded configuration (as designed) was
extremely sensitive to fluctuations in the operating
conditions. For the small changes investigated the
operating cycle became unstable, with large cycle-to-
cycle variations in delivery ratio and power output.
Similar inconsistent behavior has been observed in
actual Atkinson-cycle engine operation [3,4], however the
degree of variation was not as severe as with these
calculations. This may simply be a numerical problem
with the computational setup; however it may suggest
that rigorous control of the input conditions might be
required for effective, and stable operation with the over-
expansion scheme.
As an additional point, it was seen that the increases
in emissions can be substantial for some of the
variations. This may be important in actual engine
operation.
DISCUSSION AND CONCLUSIONS
Multi-dimensional, 0D and 1D modeling, and single
step parametric variations have been used to analyze
and optimize the scavenging system for a free piston
engine-generator, in order to ensure high efficiency
operation with low output emissions. KIVA-3V was
employed, along with models for the compressor and
friction processes. A range of design options was
6
investigated including the use of loop, hybrid-loop and
uniflow scavenging methods, different charge delivery
options, and various operating schemes. The post-
processing software, Ensight, allowed the in-cylinder and
port dynamics to be visualized and more thoroughly
understood. Using these tools, the overall array of
design possibilities was significantly narrowed, while
some interesting configurations were explored.
The results of the analyses indicated that the loop
and hybrid-loop methods as investigated here, cannot
achieve sufficient scavenging performance, while the
uniflow method, although it increases the mechanical
complexity of the engine, yields the most desirable
scavenging characteristics.
As calculated, an optimal arrangement employs a
stratified scavenging scheme supplied by a steady, low
temperature/pressure (~300K/1.2bar) charge. The
highest possible thermal efficiency should result;
however, control of fuel short-circuiting emissions,
especially over small variations in the engine’s operating
frequency, may prove challenging. In addition, the
means of supplying the fuel (carburetor or port injection)
has not been addressed, and this will require additional
study. On the other hand, this configuration seems to be
capable of providing adequate mixing during scavenging
and compression to enable rapid, TDC HCCI
combustion, while maintaining efficient performance as
the operating conditions vary slightly.
It was seen that the in-cylinder flow characteristics,
resulting from the scavenging process can significantly
affect the operating performance. The KIVA-3V
calculations suggested that in the premixed HCCI
operating mode, with low φ and moderate ηsc, the
production of NOx is more dependent on hot residual
initiated pre-ignition and subsequent over-compression,
than on the combustion of fuel-rich regions within the
cylinder.
In addition, changes in the flow patterns with
frequency variation can lead to large increases in the
short-circuiting emissions. Without adequate control,
these losses may become unacceptable. One option to
limit this may be to utilize low pressure, port injection,
late in the scavenging cycle, in combination with a
uniform intake manifold geometry. The injection timing
and duration could be dynamically adjusted depending
on the operating conditions, and as a result short-
circuiting emissions may be better managed.
REFERENCES
1. Van Blarigan, P., Paradiso, N. and Goldsborough, S.
“Homogeneous Charge Compression Ignition with a Free Piston: A
New Approach to Ideal Otto Cycle Performance,” SAE Paper 982484,
1998.
2. Kong, S. C., Marriott, C. D., Reitz. R. D. and Christensen, M.,
“Modeling and Experiments of HCCI Engine Combustion Using
Detailed Chemical Kinetics with Multidimensional CFD,” SAE Paper
2001-01-1026, 2001.
3. O’Flynn, G. T., Saunders, R. J. and Ma T. H., “Combustion
characteristics of an Otto-Atkinson engine using late inlet valve closing
and multi-point electric fuel injection,” SAE Paper 925107, 1992.
4. Raynes, S. H., “An Atkinson cycle engine for low pollution,” SAE
Paper 984064, 1998.