You are on page 1of 8

Copyright 2013 Society of Automotive Engineers of Japan, Inc.

All rights reserved




Identification of Suitable Ignition Conditions for Stable Switch from SI to
HCCI of a Blowdown Supercharged Engine
- Towards One-step SI-HCCI Combustion Mode Transition -

Michael Jagsch
1)
Yasuo Moriyoshi
2)
Tatsuya Kuboyama
3)

1)-3) Chiba University, Graduate School of Engineering
1-33 Yayoi-Cho, Inage-Ku, Chiba, 263-8522, Japan (E-mail:mjagsch@graduate.chiba-u.jp)
Received on April 9, 2012

ABSTRACT: In this paper, the necessary in-cylinder conditions for the transition from spark ignion (SI) to
homogeneous charge compression ignition (HCCI) are analyzed. The hereby important factors for ignition time are
investigated analytically. A control oriented combustion model, that is validated by experimental data, is used for
sensivity analysis and a Monte-Carlo approach is taken to identify the important factors for transient HCCI and
combustion mode switch.
As a result, it was shown that the in-cylinder temperature plays a major role, exceeding other factors, such as the air-fuel
ratio. The analytical conclusions are illustrated by numerical simulations after which the feasibility for actuation will be
discussed. Final goal is to ensure a stable combustion in each cycle without crossing unstable combustion conditions.

KEYWORDS: Heat Engine, Compression Ignition Engine, Control, Numerical Calculation, combustion mode switch [A1]



1Introduction
Rising demands in fuel economy and stricter emission
regulations require continuous improvement of todays combustion
engines. A gasoline engine being world-wide in use, has compared
to diesel engine lower emissions, but in turn higher fuel
consumption. A Homogeneous Charge Compression Ignition
(HCCI) concept has been proven in the past decade to be able to
meet both requirements and ongoing research and efforts in
improvement imply the expectation of their use in the near future in
commercial passenger cars. Despite the promising advantages, still
several issues remain. An HCCI engine cannot cover all required
operation points, although efforts in increasing the operational range
have been intensively studied in the past. Furthermore, the inherent
difficulty in controlling an HCCI engine due to the fact that a direct
igniter does not exist, is still considered as a challenge and requires a
closed-loop control to ensure operational stability. This control
challenge becomes even more severe when switching between SI
and HCCI combustion modes is taken into consideration.
At Chiba University, an HCCI engine has been studied which
exploits the boosting potential of a blow-down pressure wave,
generated from one cylinder to supercharge another one with
exhaust gas through a connecting pipe in the exhaust system. To be
able to receive this pressure impulse, the connected cylinder must
re-open its exhaust valve after the intake stroke. This system is
called Blowdown supercharging system (BDSC) and Hatamura
(1)

has shown its potential to expand the operational area of HCCI to
higher loads.
A question remains in how to cover all needed operational
ranges of an ordinary passenger car, whose demand can be only met
by operating the engine in SI if needed and HCCI if possible. The
transition hereby should be as smooth as possible without any
remarkable fluctuations in IMEP and without any combustion
instabilities, which may lead to misfire. A reason for the difficulty of
a combustion mode switch is the big differences in engine
conditions, such as EGR or A/F-ratio between HCCI operation and
SI operation. HCCI requires a high EGR content to rise the
in-cylinder temperature during compression to ensure ignition but
also to keep the maximum pressure rise during combustion in an
acceptable range. The feature of not needing a spark makes it
possible to operate the engine in very lean modes and to omit the
intake throttle. Under such conditions, however, an SI engine would
not fire. The window for A/F-ratio is very narrow, as well as the
EGR rate is restricted for a stable operation. The challenge therefore
lies in enabling a switch from one combustion mode to the other, in
an as short as possible time and as smooth as possible. HCCI
ignition is very sensitive to in-cylinder temperature. Since during SI
combustion the temperatures of the exhaust gas and the in-cylinder
wall are much higher than in normal HCCI mode, a one-step switch
from SI to HCCI would lead to an undesired early and violent
combustion. This might lead in too high pressure rises, which
increases wall-heat losses and to misfire in the succeeding cycles
due to the resulting too low exhaust temperatures.
Regarding the higher temperature of the exhaust gas available
for the first HCCI cycle, it also becomes clear that a successful
M. Jagsch et al./International Journal of Automotive Engineering 4 (2013) 17-24
17
Research Paper
20134098

Copyright 2013 Society of Automotive Engineers of Japan, Inc. All rights reserved


switch could be obtained by strictly regulating the EGR content
during transition, in which in the first HCCI cycle the high exhaust
temperature is compensated for by keeping the EGR content lower
than in steady-state case. To enable this kind of switching strategy,
the effectiveness of a variable valve system has been demonstrated
through simulation in Roelle et al.
(2)
. An approach by variable valve
actuation was also presented by Kakuya et al
.(3).
In this method,
unstable combustion modes were identified in experiments, in which
neither SI nor HCCI can be operated stably. The initial combustion
mode was first driven to its operational limit, in which the unstable
area is the narrowest and passed through in a number of cycles, until
it reached the border for stable target combustion, before driving it to
the targeted operating point. In this multi-step method, however,
additional measures were necessary to stabilize the intermediate
SI-HCCI combustion, by using fuel injection strategies, and spark
assistance. Wu et al.
(4)
reported similar problems of intermediate
cycles, that lie neither in appropriate SI ranges, nor in stable HCCI
ranges. It was stressed that these cycles where prone to misfire and
spark assistance was needed. In the presented switch, negative valve
overlap (NVO) was realized within one cycle and the intake throttle
was opened rapidly. Similar to the concept of Kakuya et al.
(3)
, the
last SI cycle revealed an equivalence ratio of 1.0, but an EGR
content of 32% which implies a first drive to this operation point
before the actual switching process.
Contrary to NVO systems, the BDSC HCCI is at present
operated with a Honda VTEC variable valve system, which switches
between two distinct valve lifts and thereby between SI and HCCI
configuration. It thus allows the rapid actuation of the valve lifts and
with it, a direct jump into HCCI operation. Additionally, due to its
extended operation range, the frequency of switching remains low.
In order to find appropriate control strategies for switching, several
simplified HCCI engine models were presented in the past, in which
combustion was modeled as single Wiebe function
(5)
, and start of
combustion was determined by a simple knocking model, able to
cover a wide range of operation points. Chang et al.
(6)
implemented
such a simplified ignition model in a commercial 1-D code to
investigate the transient behavior of an HCCI engine. Additional
examples can be found in the literatures
(7)-(10)
. Further simplified
derivatives, based on the idea of Ravi et al.
(11)
, were also shown to
be able to capture transient HCCI in Widd et al.
(12)
.
In this paper, the feasibility of enabling a direct SI-HCCI
transition without falling into an operational gap will be analyzed.
This is motivated by the fact that SI delivers enough thermal energy
for the succeeding HCCI cycle to ignite, and presuming that a proper
cycle-to-cycle temperature control would after a smooth transition
eventually lead to stable HCCI mode. First, the concept of the
BDSC HCCI will be presented, after which a numerical model of
the engine on the 1-D commercial code BOOST from AVL will be
introduced. An ignition model, based on a correlation for ignition
delay times by He et al.
(13)
will be presented and approved by
experimental steady-state data. This ignition model will then be
investigated to extract the important variables for a stable ignition.
Goal is to find in-cylinder conditions, which minimize the
probability for misfire. The conclusions drawn in these analyses will
be validated on a 1-D simulation code and the feasibility of this
switching concept will be discussed by comparing the BDSC engine
with traditional NVO systems.
2Engine Concept
2.1. Blowdown supercharge

A sketch of the engine concept is presented in Figure 1. It
shows a 4-2-1 exhaust system, which connects cylinders pair wise
(#1 and #4, #2 and #3) through an exhaust pipe.


Fig. 1 Pressure wave propagating from cylinder #1 to #4

This configuration allows for one cylinder to capture a pressure
impulse from the corresponding one during its blowdown exhaust
phase, to increase the in-cylinder pressure and EGR rate before
compression. This pressure wave boosting scheme is indicated by an
arrow in Figure 1, in which cylinder #1 boosts cylinder #4.
Figure 2 displays the valve lifts of two coupled cylinders, in
which a phase shift of 360 CAD to each other is obvious. In this
figure the exhaust re-opening after intake can be seen and the
pressure impulse from one cylinder to the other is implied by grey
arrows. The exhaust valve of one cylinder opens at the time, when
the exhaust valve of the other is re-opened for EGR re-induction,
able to capture the pressure impulse.

#4 Cylinder
#1 Cylinder
Intake
Exhaust
EGR
Intake
Exhaust
EGR
Exhaust valve re-opening
for EGR re-breathing
Blow-down pressure wave
V
a
l
v
e
L
i
f
t
V
a
l
v
e
L
i
f
t
Crank Angle deg.ATDC

Fig. 2 Valve Lifts of cylinder #1 and #4

A great advantage of this system is that the exhaust re-breathing
scheme hardly has an impact on the intake stroke. Therefore its
influence on the volumetric efficiency is remarkably low, but in turn,
the EGR content can be controlled indirectly by the amount of
inducted fresh air. Specifications of the engine are given in Table 1.
M. Jagsch et al./International Journal of Automotive Engineering 4 (2013) 17-24
18

Copyright 2013 Society of Automotive Engineers of Japan, Inc. All rights reserved



Table 1: Specification of the Engine
3. Engine model
3.1. Base engine model

The 4-cylinder engine is modeled in the one-dimensional
commercial code BOOST from AVL. For the sake of low
complexity, two of the four cylinders were operated and the others
were switched off by deactivating the corresponding fuel-injectors
and keeping the valve lift fixed at their closing positions. A
minimum of two cylinders is required to enable the blowdown
boosting system.
In addition to the intake throttle, a second throttle was installed
on the exhaust side, to regulate the exhaust pressure and with it, to
control the in-cylinder EGR content. During experiments and
simulations, however, this valve was kept at a fixed angle.
In addition to that, a single-cylinder model was developed as a
tool for analysis of values that are not measurable in experiments.
Those are at most the in-cylinder temperature and mixture
composition. The investigation was accomplished by applying
measured pressure traces over one cycle and the mean temperatures
on the intake- and exhaust side to the cylinder model as boundary
conditions. The distance of the sensors from the ports and thus the
model boundaries from the ports were around 70mm on both, the
intake and exhaust sides. With this kind of modeling, fine-tuning
of model parameters is possible as well.

3.2. HCCI-specific sub-models

The conceptual difference of an HCCI requires the
consideration of HCCI-specific sub-models. Combustion can be
modeled as a standard single Wiebe function, but to predict the onset
of ignition, a knock-integral model (KIM), commonly also referred
to as the Livengood-Wu integral, was implemented. In this work, a
model based on the ignition delay for iso-octane by He et. al
(13)
was
used. Ignition delay has been reported to be depending
predominantly on pressure, temperature, oxygen mole ratio and
equivalence ratio. The resulting ignition correlation has been
compared to a wide range of experimental data that covers typical
HCCI operation and is therefore considered here appropriate for a
combustion mode switch analysis. After some modification, the
implemented ignition model in this work finally reads:


(1)


with P as the pressure, F/G the fuel by gas ratio, A/G the air by gas
ratio, R the universal gas constant and T the temperature. A
pre
is a
pre-exponential factor that has been tuned to account for the crank
angle as the integration variable, instead of time, and to fit the
knocking model to experiments. Additionally, gasoline as fuel
instead of iso-octance was thereby considered as well. The other
parameters were kept at their original values. Combustion sets in
when the value, integrated from a reference point (e.g.start of
compression), reaches unity. This ignition model was implemented
in BOOST via the Formula Interpreter block, with a trapezoidal
integration scheme and a fixed simulation time-step of 0.1 CAD.
The subsequent combustion was modeled by a single Wiebe
function, with a fixed Wiebe-exponent, and the combustion
durations are taken from experiments for each case.
Concerning the wall heat transfer, the HCCI combustion
process differs considerably from the SI case, due to the missing
flame front and exhibits therefore a different wall heat transfer
mechanism. The Hohenberg
(14)
correlation has been rated suitable
and preferred in this analysis. Further discussion on the topic can be
found in Chang et al.
(15)
and Soyan et al.
(16)
. With this configuration,
the knocking model, presented above was tested on the one-cylinder
engine model, with the pressure traces from steady-state experiments
applied as boundary conditions. The resulting CA50, the point at
which 50% of the total heat has been released, is compared to
experiments in Figure 3 for 26 steady-state cases. In these cases the
exhaust pressure has been varied between 1.25 bar, and 1.55 bar, and
the amount of fuel per cycle was between 12.8 mg and 17.2 mg,
with a resulting IMEP range between 420 kPa and 612 kPa.. The
engine speed was kept constant at 1500 RPM. Details are listed in
Table 2.

Table 2: Steady-State Engine Conditions

Cases Fuel [mg] Air/Fuel
[-]
Pressure
Exhaust
[bar]
EGR [%]
1 - 6 ~15.2 ~37.8 1.34-1.38 24-25
7-12 ~13.7 ~43.4 1.34-1.48 25.5-26.75
13-18 ~12.75 ~48.6 1.46-1.56 26.7-28.1
19-23 ~16.0 ~35.1 1.30-1.33 23.1-23.7
24-26 16.6-17.1 32.0-33.4 1.26-1.28 21.82-22.6



Fig. 3 Comparison of CA50 between simulation and experiment

Engine Type Inline 4 Cylinder
Bore x Stroke 86mm x 86mm
Compression Ratio 11.6
Fuel supply Port Injection
Fuel Gasoline (RON91)
}
|
.
|

\
|

|
.
|

\
|

|
.
|

\
|
=
u
u
u d
T R
E
G
A
G
F
P A KM
D C
B
pre
exp
M. Jagsch et al./International Journal of Automotive Engineering 4 (2013) 17-24
19

Copyright 2013 Society of Automotive Engineers of Japan, Inc. All rights reserved


Despite the simplicity of the employed combustion model, the
trend can be re-produced and the values for CA50 were
considered to be predicted with a reasonable accuracy.
As for the SI case, the Wiebe function has been applied as
well, whereas the Wiebe parameters have been set to correlate
with experimental data.
4Ignition Sensitivity Analysis
The idea is to analyze the target values to be controlled, since
CA50 can be only controlled indirectly. So, the question of what
exactly needs to be controlled, and which variables may not exceed
what bounds, remains. In this work, the simplified knocking model
has been chosen over a detailed chemistry model, due to its
simplicity and fast computation. It also allows a sensitivity analysis,
in which the most important factors for ignition can be extracted and
the model thereby simplified.
The sensitivity of a multivalued function can be defined and
normalized as (Warnatz et al.
(17)
, Chiang et al.
(18),(19)
)

u
y
y
u
u S
OP
OP
o
o
= ) ( (2)

with S as the sensitivity, y as the general output of the considered
function, and u as the general input. The subspcript OP denotes the
respective operation point which the functions is evaluated and
normalized at.
For calculation of the sensitivity knocking model, the inverse of
the ignition delay was analyzed with the Symbolic Toolbox from
MATLAB. The considered ranges are given in table 3, covering
typical HCCI conditions.

Table 3: Definition of the ranges for ignition analysis


The temperature and pressure of reference have been chosen to
be at the instant of 20BTDC and the considered operation points are
sampled randomly with a number of 1000 cases.
Figure 4 shows scatter plots of the ignition sensitivity of the
ignition model on the four different parameters. As expected, the
ignition is the most sensitive to in-cylinder temperature, with
normalized values between 21 and 17. The other three variables on
the other hand, show a constant normalized sensitivity over all
investigated operation points and are much lower than that of the
temperature. Not surprisingly do these values correspond to their
respective powers in the equation for ignition. It therefore can be
concluded that temperature plays a major role in HCCI combustion.
However the question of what values and what ranges for the four
variables are acceptable, has not been clarified as of yet.


Fig. 4 Sensitivity Analysis of HCCI ignition

5Ideal Ignition Condition Identification
In the previous section, the dependence of four parameters on
ignition was assessed, in which the temperature was shown to be the
most dominant factor. At this point suitable parameter ranges for all
four ignition parameters will be identified, in which a desired
ignition time can be expected and the probability of misfire and too
early combustion would be minimized. However, finding such
classifications is rather a complex task, since the ignition time
depends on those factors which can hardly be investigated separately
from each other. Therefore, a statistical, Monte-Carlo-like approach
in will be carried out, in which a range of initial values will be used
to predict start of combustion, and outliners and their cause will be
identified.
For this purpose, script in MATLAB was written, in which the
temperature-, and pressure traces are calculated from initial values at
40 CAD BTDC via polytropic compression, defined as

1
1
2
1
2
T
V
V
T
|
|
.
|

\
|
=


1
2
1
2
P
V
V
P
|
|
.
|

\
|
=

(3)

Here, the polytropic coefficient was set to 1.29, which incorporates
the charge composition as well as wall-heat losses. This value has
proven to be the best fit in this investigation, and its validity will be
shown in plots later. The confidence of using such a simplified
pressure and temperature model is gained from control-oriented
mean-value HCCI models which have been proven to capture the
transients of HCCI engine reasonably well.
Integration was carried out with a trapezoidal integration
scheme and a time-step of 0.1 CAD to keep the accuracy at a high
level. The values for initial conditions are again sampled randomly
within the same range, as given in Table 2. Since start of integration
for Livengood-Wu integral begins at 40BTDC back-calculation
from the values at 20BTDC to the values of temperature and
pressure at 40BTDC was necessary and this was done by applying
equations (3).
Variable Range
In-Cylinder Temperature (at 20BTDC) 790.0 870.0 [K]
In-Cylinder Pressure (at 20BTDC) 20.0 25.0 [bar]
Air/Gas 0.6 0.9 [-]
Fuel/Gas 0.02 0.04 [-]
M. Jagsch et al./International Journal of Automotive Engineering 4 (2013) 17-24
20

Copyright 2013 Society of Automotive Engineers of Japan, Inc. All rights reserved


Figure 5 shows the distribution of the outcome of the
investigation with a number of samples of 1000, in which misfire are
the cases around 380 CAD.

Fig. 5 Ignition Times under randomly set conditions as in Table 3

It can be seen that a dense accumulation is found in the CAD
range between 355 and 10 ATDC. Ignition cases with values beyond
10 ATDC are prone to misfire and should be avoided. This
assumption is physically reasonable, since the downward movement
of the piston cools the mixture and thus decreases the temperature
and pressure. In this example, almost 35% of the cases have misfired,
while the peak of the distribution curve is located around TDC,
which is a more than acceptable value.
The scatter plot for all the cases with ignitions within 10 CAD
BTDC and 10 CAD ATDC is displayed in Figure 6.


Fig. 6 Ignition Times under randomly set conditions as in Table 3

The dominance of the temperature for HCCI, as already pointed out,
becomes obvious again. While for the other variables a rather even
distribution can be seen, a strong influence on temperature is evident.
An accumulation of the ignition times becomes denser, the higher
the temperature is. Furthermore in this case study, an initial
temperature of around 850 K appears to ensure ignition within the
desired range, regardless of the values of the other parameters.
In a second analysis, the temperature range was chosen to be
between 850 and 870 K, while keeping the other values at the same
range as in the previous analysis, and again with a number of
samples of 1000. Figure 7 shows again a scatter plot of the ignition
time, in which no point of misfire can be detected.
Too early combustion cannot be detected in any of the cases,
which implies a temperature limit to be necessary for ignition. A
small slope of the temperature dependency exists, but the highest
values of 870 K did in this case not lead to too early ignition.

Fig 7 Scatter Plot of Ignition Times under randomly set conditions
within a tighter temperature range between 850K and 870K

The histogram in Figure 8 additionally shows that the frequency of
occurrence of almost 350, as previously found as misfire, has
completely vanished and the ignitions with the highest frequencies
have been shifted to a location of around 356 CAD.


Fig. 8 Ignition Times under randomly set conditions within a tighter
temperature range between 850K and 870K

This observation shows the necessity of precise temperature
control to keep combustion in a desired phasing. It furthermore
motivates to define a temperature window throughout the entire
combustion transition process, and thereby simplifying the
combustion transition strategy remarkably.
6Combustion mode switch
In the previous sections, suitable HCCI control strategies have
been devised by means of an ignition model. It was shown that the
in-cylinder temperature must be within a specified range to avoid
misfires or too early combustion. Controlling only the temperature
would furthermore make control straightforward, as other
combustion parameters can be neglected.
M. Jagsch et al./International Journal of Automotive Engineering 4 (2013) 17-24
21

Copyright 2013 Society of Automotive Engineers of Japan, Inc. All rights reserved


In the following, an SI-HCCI step will be demonstrated using a
2-cylinder full engine simulation model with the 1-D code BOOST
from AVL. The model includes the ignition model as previously
introduced. Combustion duration has been kept fixed as the focus of
this investigation is on ignition timing. Wall heat transfers in pipes
and cylinders, gas dynamics in intake and exhaust manifold
considered as well. The actuators however (throttles, VVA, etc.)
were treated statically and thus, by excluding any dynamics and
response time properties, the analysis remains un-biased and neutral.
The wall temperature was kept fixed, due to the slow response time
and therefore low influence of varying wall temperatures during the
transition. Fuel has been directly injected to ensure constant fueling
during the entire switch, but fuel evaporation has been disabled.
During SI, the intake has been throttled to a pressure of around
0.59 bar to keep the air-fuel ratio on a stoichiometric level. The
exhaust throttle has been kept fixed to ensure elevated exhaust
pressures during HCCI mode. The amount of fuel injected was 15.4
mg.
The first transition was accomplished through an instantaneous
full opening the re-breathing lift within one cycle and opening of the
intake throttle after the first HCCI cycle. With this measure, the first
HCCI cycles are stoichiometric and due to the low intake pressures,
more EGR gets into the cylinder which results in hotter in-cylinder
conditions. This should ensure that the cylinders not only undergo a
large transition in in-cylinder temperatures and pressures during the
switching process, but also in the in-cylinder contents, such as
air/fuel ratio, in order to demonstrate the influence of these
parameters during the switch.
This simple one-step switch actuation is illustrated in figure 9,
in which intake throttle opening timing, as well as the exhaust
re-opening events of both cylinders are displayed. Although intake
valve actuation has no function in this analysis, the lift curves are for
the sake of a better orientation depicted as dashed lines in this figure
as well.


Fig. 9 One-Step switch from SI to HCCI: Control Actuation

The result of the switch from SI to HCCI by simple switch
actuation is shown in Figure 10. For the evaluation of the ignition
model from the previous section, the calculation of CA50, with an
offset of 9.4 CAD is plotted in dashed lines in this figure too. The
initial conditions for the integral are taken from the BOOST results
for each cycle. It shows a good agreement with the BOOST results,
except for the early ignition cycles, which is something that can be
attributed to the enhanced wall heat transfer during compression,
which is not considered in the simplified ignition model. However,
once the temperature reaches normal levels again, such as in cycle
29 of cylinder #2, the simplified knocking model matches the
BOOST data well and follows its trend.


Fig. 10 One-Step switch from SI to HCCI: Results

The first HCCI cycle exhibits a too early CA50, implying a too early
ignition, which is assumed due to the hot exhaust gases. The
in-cylinder temperature at 20 BTDC, as given in the second plot
reveal extremely high values, which are due to the onset of
combustion happening earlier than this point. The intake throttle
opens after Cylinder #1 goes through two HCCI cycles before the
intake throttle opens, while cylinder #2 only has only one, which
explains the two cycles of early combustion phasing in cylinder #1.
In general, such early combustion phasing is undesired since a steep
pressure rise not only results in high noises, but also might lead to
uncontrolled knocking due to the resulting high pressures and
temperatures. For that reason, controlling the temperature is crucial
especially for the first HCCI cycle.
The IMEP of the switch is displayed the lowest plot in Figure
10, with an expected difference between SI and HCCI,
mainly due to the throttling losses in SI. This figure shows a
severe drop during the first HCCI cycles which are those with too
early combustion. This drop can be explained on one hand by the
lower effective compression ratio, which stems from the dislocated
combustion event. On the other hand because of high wall heat
losses, which arise due to the ongoing compression during
combustion, so that the temperatures in these cycles reach high
values which increase the wall heat transfer. This drastically reduces
M. Jagsch et al./International Journal of Automotive Engineering 4 (2013) 17-24
22

Copyright 2013 Society of Automotive Engineers of Japan, Inc. All rights reserved


the work output and the resulting exhaust temperatures. With this in
mind, it becomes clear that the amount of hot exhaust gases from the
last SI cycle must be reduced in the first HCCI cycles. To improve
the switch and to avoid the above mentioned problems, a precise
EGR and thereby temperature controller actuator is needed.
In this investigation, a variable re-breathing lift is suggested,
which in contrast to throttle bodies may allow a precise cylinder
individual control, making this kind of actuation attractive. Figure 11
shows the exhaust valve lift with an exemplary continuous variable
re-breathing lift.


Fig. 11 Variable exhaust valve actuation

In the following, the actuator dynamics are neglected again, and
full actuation within one cycle was assumed. The switch from SI to
HCCI was accomplished in such a way that the re-breathing lift was
set low in the first cycles, to reduce the exhaust gas contents and thus
the in-cylinder temperatures, then gradually opening it in the
successive cycles and for each cylinder individually. The intake
throttle has been opened after the first HCCI cycle for the same
reason explained above: The first HCCI cycles will be under
stoichiometric conditions and a large transition of the air/fuel ratio
can be examined.


Fig. 12 Complex actuation switch from SI to HCCI: Control
Actuation

Figure 12 shows the actuation of a temperature-oriented, rather
loosely calibrated, yet successful SI-HCCI transition. In the first
HCCI cycles, the re-breathing lift is set very low and the intake
throttle opens after the switch. With the presented actuation, cylinder
#1 even undergoes two stoichiometric cycles, until the higher intake
pressure increases the air/fuel ratio and that cylinder. Due to the
difference between the two cylinders, the step-wise opening of the
EGR lift has been calibrated for each cylinder individually so that
the in-cylinder temperatures could be kept in a-priori defined range,
which resulted in a desired combustion phasing for each cycle and
each cylinder.


Fig. 13 Complex actuation switch from SI to HCCI: Results

The results of this switch are shown in figure 13, and it reveals
some fluctuations in CA50. Especially cylinder #1 shows a lot of
variations, which might be explained due to the two cycles of
stoichiometric conditions, until the air/fuel ratio converges to its
steady-state value. Especially in cycle 31, the in-cylinder
temperature is lower than in the previous or successive cycle, but
there is also a large change in air/fuel ratio. Furthermore, once the
air/fuel ratio reaches convergence, the in-cylinder temperature seems
to correlate better with combustion phasing, which is especially
evident in the cycles after cycle 33. Variations observed here of
about 6 cad are in accordance with the findings from the previous
section. To reduce the early strong variations of cylinder #1 the
opening of the re-breathing lift may have been re-tuned such that the
temperature levels are higher and surmount the other influences.
Contrary to cylinder #1, cylinder #2 shows a rather smooth
transition which may be explained to the earlier reaching of the
steady-state air/fuel ration.
Additionally, the IMEP fluctuations were found to be very low
for both cylinders, as it can be seen in the lowest plot of Figure 13.

This switching example illustrates that controlling the temperature
M. Jagsch et al./International Journal of Automotive Engineering 4 (2013) 17-24
23

Copyright 2013 Society of Automotive Engineers of Japan, Inc. All rights reserved


precisely does not require to consider other values, such as A/F ratio
or pressure, making the switching process easier to understand and
easier to calibrate.
7Conclusions
In this paper, ideal ignition conditions were investigated by
means of a basic combustion model. The observations were then
further investigated on a 1-D full engine simulation model and the
following conclusions are drawn.

1) The identified ignition conditions suggest the necessity of
precise temperature control to control ignition time, while at the
same time neglecting other variables can be tolerated. By
considering this, finding control strategies and algorithm for the
combustion switch can be simplified. This is of high importance, as
the limiting factors for ideal ignition are the uncertainties over the
systems states, the complexity of the employed control algorithm
and the response properties of the actuators.
2) For the above reason, suitable actuators have to be
selected as a next step. Blow-Down Supercharge (BDSC) equipped
with a Honda VTEC, enables a direct switch into HCCI, but with the
likely consequence of too early ignition. As a possible solution to
overcome this problem, a more sophisticated variable valve lift
actuation was suggested and demonstrated in simulation.
3) If eventually the temperature can be regulated for each
cycle, then the number of unstable cycles can be reduced and the
robustness of the overall switching process increased.

REFERENCES
(1) K. Hatamura; A Study on HCCI (Homogeneous Charge
Compression Ignition) Gasoline Engine Supercharged by
Exhaust Blow Down Pressure, SAE Technical Paper, No.
2007-01-1873, 2007.
(2) M. Roelle, G. Shaver, J.C. Gerdes; Tackling The Transition:
A Multi-Mode Combustion Model of SI and HCCI for Mode
Transition Control
(3) H. Kakuya, S. Yamaoka, K. Kumano, S. Sato; Investigation
of a SI-HCCI Combustion Switching Control Method in a
Multi-Cylinder Gasoline Engine, SAE 2008-01-0792
(4) H. Wu, N. Collings, S. Regitz, J. Etheridge, M. Kraft:
Experimental Investigation of Control Method for
SI-HCCI-SI Transition in a Multi-Cylinder Gasoline Engine,
SAE 2009-10PFL-0546, 2009
(5) J.B. Heywood; Internal Combustion Engine Fundamentals,
McGraw-Hill Book, 1988
(6) K. Chang, A. Babajimopoulos, G. Lavoie, Z. Filipi, D.
Assanis; Analysis of Load and Speed Transitions in an
HCCI Engine Using 1-D Cycle Simulation and Thermal
Networks, SAE 2006-01-1087, 2006
(7) G. Shaver, M. Roelle, J.C. Gerdes; Modeling cycle-to-cycle
dynamics and mode transition in HCCI engines with variable
valve actuation; Control Engineering Practice 14, 2006
(8) D. Blom, M. Karlsson, K. Ekholm, P. Tunestal, R. Johansson;
HCCI Engine Modeling and Control using Conservation
Principles, SAE 2008-01-0789
(9) N. Killingsworth, S Aceves, D. Flowers, M. Kristic; A
Simple HCCI Engine Model for Control, Proceedings of the
2006 IEEE Internal Conference on Control Applications,
2006
(10) D.J. Rausen, A. G. Stefanopoulou, J-M. Kang, J. A. Eng,
T.-W. Kuo; A Mean-Value Model for Control of
Homogeneous Charge Compression Igntion (HCCI) Engines,
Journal of Dynamic Systems, Measurement and Control,
2005
(11) N. Ravi, M J. Roelle, A.F. Jungkunz, J. C. Gerdes; A
physically based Two-State Model for Controlling Exhaust
Recompression HCCI in Gasoline Engines,
IMECE2006-15331
(12) A. Widd, P. Tunestal, R. Johansson; Physical Modeling and
Control of Homogeneous Charge Compression Ignition
(HCCI) Engines, Proceedings of the 47
th
IEEE Conference
on Decision and Control, 2008
(13) X. He, M.T. Donovan, B.T. Zigler, T.R. Palmer, S.M. Walton,
M.S. Wooldridge, A. Atreya; An experimental and modeling
study of iso-octane ignition delay times under homogeneous
charge compression ignition conditions, Combustion and
Flame, No. 142, pp. 266-275, 2005
(14) G. Hohenberg, Advanced approaches for heat transfers
calculation, SAE technical paper, 790825,1979
(15) J. Chang, O. Gueralp, Z. Fillipi, D. Assanis, T.W.Kuo, P. Najt,
R. Rask; New Heat Transfer Correlation for an HCCI
Engine Derived from Measurements of Instantaneous Surface
Heat Flux, SAE 2004-01-2996, 2004
(16) H.S Soyhan, H. Yasar, H. Walmsley, B. Head, G.T. Kalghatgi,
C. Sorusbay; Evaluation of heat transfer correlations for
HCCI engine modeling, Applied Thermal Engineering, No
29, pp. 541-549, 2009
(17) J. Warnatz, U. Maas, R.W. Dibble: Combustion, Physical
and Chemical Fundamentals, Modelling and Simulation,
Experiments, Pollutant Formation, Springer 1996
(18) C.J. Chiang, A.G. Stefanopoulou; Sensitivity Analysis of
Combustion Timing and Duration of Homogeneous Charge
Compression Ignition (HCCI) Engines, Proceedings of the
2006 American Control Conference, 2006
(19) C.J. Chiang, A.G. Stefanopoulou; Sensitivity Analysis of
Combustion Timing of Homogeneous Charge Compression
Ignition Gasoline Engines, Journal of Dynamic Systems,
Measurement and Control, 2009
M. Jagsch et al./International Journal of Automotive Engineering 4 (2013) 17-24
24

You might also like