Applied Energy 94 (2012) 166173

Contents lists available at SciVerse ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Numerical study of a butanol/heptane fuelled Homogeneous Charge Compression

Ignition (HCCI) engine utilizing negative valve overlap
Sona Visakhamoorthy a, John Z. Wen a,, Siva Sivoththaman b, Charles Robert Koch c
a

Department of Mechanical and Mechatronics Engineering, University of Waterloo, 200 University Avenue West, Waterloo, ON, Canada N2L 3G1
Department of Electrical and Computer Engineering, University of Waterloo, 200 University Avenue West, Waterloo, ON, Canada N2L 3G1
c
Department of Mechanical Engineering, University of Alberta, Edmonton, AB, Canada T6G 2G8
b

a r t i c l e

i n f o

Article history:
Received 10 November 2011
Received in revised form 16 January 2012
Accepted 17 January 2012
Available online 16 February 2012
Keywords:
HCCI
Butanol
Heptane
Multi-zone
Model
Negative valve overlap

a b s t r a c t
The calibration and comparison to experimental data of a parallel computing multi-zone combustion
model for simulating operational characteristics of an n-butanol/n-heptane fuelled Homogeneous Charge
Compression Ignition (HCCI) engine utilizing the negative valve overlap (NVO) technology is described.
The model is calibrated using one experimentally characterized operating point and by taking into
account the major features of NVO. The model simulations at other four operating points closely match
the cylinder pressure trace and heat release rates of the experiments. The unburned hydrocarbon emission is predicted to a reasonable level while NOx (nitric oxide NO and nitrogen dioxide NO2) formation is
under-predicted. The difculty in mapping the NOx emission is attributed to the fact that the combustion
model operates within the closed cycle period of engines and therefore does not capture the complexity
of the charge stratication within the NVO equipped engine. Nevertheless, the trend of increasing NOx
levels with the increasing fraction of butanol in the fuel mixture is captured and the model is able to predict the pressure, heat release rates, and combustion phasing for the three fuel blends tested.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
Alternative energy technologies and renewable fuels are possible ways to reduce the transportation fuel usage. Homogeneous
Charge Compression Ignition (HCCI) engines are a promising extension of the current internal combustion engine technology in that
they offer a design where emission levels of Spark Ignition (SI) engines can be attained with the thermal efciency of Compression
Ignition (CI) engines [1]. In HCCI engines, which can be considered
as hybrid engines between SI and CI engines, a premixed fuelair
charge is compressed until it autoignites. This technology can help
reduce the environmental impacts of the increased transportation
fuel usage by offering reduced NOx (nitric oxide NO and nitrogen
dioxide NO2) and soot emissions while operating at near dieselengine efciencies. Another major benet offered by these HCCI
engines is the ability to combust a wide variety of fuels ranging
from low caloric value fuels such as biomass derived gases to natural gas, alcohols, gasoline, and diesel [15]. Furthermore, many of
these fuels can be combined as blends such as n-butanol/n-heptane,
ethanol/n-heptane, and n-butanol/gasoline [611].

Corresponding author. Tel.: +1 519 888 4567x38362; fax: +1 519 885 5862.
E-mail addresses: [email protected] (S. Visakhamoorthy), john.
[email protected] (J.Z. Wen), [email protected] (S. Sivoththaman),
[email protected] (C.R. Koch).
0306-2619/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.apenergy.2012.01.047

In HCCI engines a homogeneous airfuel mixture is compression ignited with a relatively large amount of charge dilution. This
results in low-temperature combustion, which subsequently produces low NOx and less soot due to a lack of fuel-rich ame regions
or localized high-temperature regions in the cylinder. Moreover,
higher thermal efciencies are possible due to the higher compression ratios required to autoignite these dilute airfuel mixtures.
However, one of the main challenges is that combustion phasing
of HCCI engines is controlled by chemical kinetics and as such
there is no specic ignition timing event as there is in SI or CI engines. Another challenge is the relatively high levels of Unburned
Hydrocarbon (UHC) and CO (carbon monoxide) emissions [12
14]. These emissions primarily arise from the charge in the crevice
regions and from a thin thermal boundary layer that forms along
the in-cylinder surfaces. Other challenges include higher noise levels due to rapid Heat Release Rates (HRRs) at high loads and a relatively narrow engine operating range [1].
To improve ignition and combustion processes and to reduce
the environmental impacts, various engine control strategies have
been implemented on HCCI engines. These techniques include
varying the level of the Exhaust Gas Recirculation (EGR), adjusting
the equivalence ratio, changing the intake temperature, and implementing Variable Valve Actuation (VVA). Negative valve overlap
(NVO) is a specic form of VVA where the exhaust valve is closed
before all the exhaust has been completely evacuated from the cylinder, which traps variable amounts of hot combustion products

S. Visakhamoorthy et al. / Applied Energy 94 (2012) 166173

[15]. The actual NVO period is dened as the time during which
both the inlet and exhaust valves are closed. This control strategy
for HCCI engines allows for charge preheating by the trapped exhaust gases, which promotes autoignition and improves performance [16]. The complicated chemical and physical processes
introduced by recycling the trapped exhaust gases do bring about
new challenges to theoretical and modeling studies.
To understand the details of the combustion process or to design model based control strategies for HCCI engines, appropriate
simulation models must be developed. These simulation models
can differ widely in terms of complexity and computational cost
and can be broadly categorized into one of three groups: single
zone models with chemical kinetics [17], multi-zone models with
chemical kinetics [18,19], and Computational Fluid Dynamics
(CFD) based models with chemical kinetics [20]. The simplest of
these three groups of models is the single zone model with chemical kinetics where the entire charge mass is treated as a single
lumped zone of homogeneous temperature, pressure, and species
concentration. Reaction rates and species evolution are solved
using chemical kinetics. These models tend to over-predict HRR
and pressure rise rates while under-predicting certain emissions
due to the assumptions inherent to such single zone models. For
example, they lack any means of representing crevice regions
and thermal boundary layers within the engine which are a significant source of CO and UHC emissions. To better differentiate
among the bulk charge, thermal boundary layer, and crevice regions, multi-zone models are used. Here, the in-cylinder charge
is broken down into concentric ring-like or individually lumped
zones which are treated as stirred reactors. Each zone is homogeneous in terms of species concentrations and temperature and
pressure distributions, but the zones are stratied from one another. This treatment allows for capturing the species and temperature gradients throughout the cylinder but requires more zones if
a larger temperature gradient exists across the cylinder.
Since single and multiple zone models operate within the closed
period of the engine cycle (between the Inlet Valve Closing (IVC)
and the Exhaust Valve Opening (EVO)) the initial conditions of
the combustion chamber (including the chemical compositions
and temperature) at IVC are set by the user. Due to the difculty
in ascertaining the exact values of these initial conditions and the
sensitive nature of HCCI combustion, this can lead to some inaccuracy in the model prediction. To reduce this error and improve the
inputs to the model, single zone models may be augmented with
full engine cycle simulations that calculate conditions up to IVC,
at which the single zone model takes over until EVO [21]. Similarly,
multi-zone models may be linked to a CFD solver such as KIVA
which has been used to determine the initial gas mixture and temperature distribution prior to combustion. Typically at a certain piston position, the multi-zone model takes over based on the
temperature prole that is calculated by the CFD code. Finally, there
are pure CFD models where the chemical kinetics is fully implemented throughout the closed cycle period [20]. In these cases mass
and heat transfer can be solved along with accurate turbulence
modeling of the compressed gases. The CFD modeling provides insight into the specic engine features which promote turbulence
such as in the squish area of the pistons. However, due to the level
of details involved and depending on the mesh size, these simulations can be very computationally intensive. In addition, signicant
efforts are needed to produce complicated meshes for different engine geometries. Nevertheless, some research has been done to parallelize these models across computing clusters to reduce the
computational time [20]. More fundamental work needs to be done,
however, for implementing the CFD code in designing model based
control strategies for HCCI engines.
As mentioned earlier, a major benet of HCCI engines is their
ability to combust a wide range of fuels. Fuels that have been

167

experimentally tested and numerically simulated include Primary

Reference Fuels (PRFs), natural gas, diesel, and methanol/ethanol/
dimethyl-ether blends [1,2224]. In this study, a HCCI engine
fuelled with three n-butanol/n-heptane blends is used to validate
a parallelized multi-zone model coupled with the detailed chemical kinetics. Since the fuel tested contains a fraction of large hydrocarbons (n-heptane), it exhibits low-temperature heat release [25].
This initial heat release occurs due to an increase in initial reaction
rates caused by increasing temperature. Under such conditions,
hydrogen is abstracted by oxygen from the fuel creating alkyl (R)
and hydroperoxyl (HO2) radicals. The alkyl radicals are consumed
via two pathways which become dominant in the low-temperature
heat release: alkylperoxy radical (RO2) production and olen production. At all temperatures, alkyl radical conversion to RO2 is faster than olen production. However, starting at approximately
700 K at 10 atm, the reverse RO2
production reaction becomes
dominant over the forward reaction thus increasing the production
of alkyl radicals [26]. These alkyl radicals are then consumed via
the olen pathway. This overall increase in olen production
reduces the fuel consumption leading to the end of the lowtemperature heat release region.
The objective of this work is to test and calibrate a recently
developed multi-zone HCCI model [27] on a NVO equipped HCCI
engine fuelled with varying ratios of bio-butanol/n-heptane and
to compare the predictions with experimental data. The utility of
the model for predicting the in-cylinder pressure, HRR, NOx and
unburned hydrocarbon emissions is examined.

2. Numerical simulation
The chemical mechanism chosen for this study is the one developed by Dagaut and Togbe [28]. This mechanism is based on the
amalgamation and reduction of two discrete models: an n-butanol
model [8] and two n-heptane models [29,30]. To predict the NOx
formation, the NOx mechanism from GRI-Mech 2.11 is added to
the above mechanism and this creates a reaction mechanism consisting of 200 species and 1805 reactions. The GRI-Mech 2.11 NOx
mechanism is selected as opposed to the that from GRI-Mech 3.0
since previous work [31,32] have indicated that GRI-Mech 3.0
over-predicts NOx levels in partially premixed ames.
For this study, a 10-zone model with the detailed chemical
kinetics is used to numerically solve the species, temperature
and pressure evolution within the cylinder. It has been reported
[27,33] that 10 zones are sufcient for solving the in-cylinder pressure trace, which is a major indicator of the engine performance.
The selection of the number of zones is exible but subject to the
availability of the number of CPUs for parallel computing. This
model operates within the closed cycle period of the engine and
steps through species, temperature and pressure evolution at each
incremental crank angle of one engine cycle. Heat transfer between
zones and the cylinder walls is considered but the mass transfer
between zones is not. There is, however, a mass loss due to
blow-by and this is implemented by an equivalent loss of mass
from all zones. Heat transfer coefcients are determined using
the Woschni correlation which accounts for bulk gas velocity
caused by the piston movement. Specically, in order to describe
the temperature gradient across the cylinder, a scaling factor is
chosen with the selected volume of the hottest core zone. The original multi-zone numerical model is parallelized using the domain
decomposition method and Fortran MPI (Message Passing Interface). This approach shares the computationally intensive step of
solving the species evolution for each zone across multiple processors. It is found that running the parallelized model on a quad-core
Intel i7 workstation results in runtimes reduced by approximately
half in comparison to single-core computation of the serial model

168

S. Visakhamoorthy et al. / Applied Energy 94 (2012) 166173

Table 1
Description of adjustable model parameters.

Table 2
Engine operating parameters.

Parameters

Parameter (unit)

Value

Intake temperature of the air fuel mixture

Cylinder wall temperature assumed to remain spatially and
temporally constant
Heat transfer scaling factor applied in the Woschni correlation
Blow-by losses
Fraction of the cylinder volume allocated to the core zone
Fraction of the trapped residual gases
Geometric ratio how rapidly zones get thinner as they approach the wall
Thermal width the temperature difference between the outermost and
core zones

Compression ratio ()
Bore (mm)
Stroke (mm)
Connecting rod to crank arm ratio ()
Displacement volume (l)
IVO, IVC (deg, after the bottom dead center)
EVO, EVC (deg, after the bottom dead center)
Engine speed (RPM)
EGR (%)
Coolant temperature (C)
Intake temperature (C)
Equivalence ratio ()
BVP (%)

12:1
97
88.9
3.5996
0.653
151, 21
100, 130
1021
0
6970
80
0.332, 0.345, 0.346, 0.357, 0.366
12, 17, 22

on the same workstation. Typical serial mode runtimes are 80 min

while parallelized runs take 40 min.
The model is calibrated against a given set of experimental data
with a number of parameters adjusted. Exact experimental values
for some of these parameters are generally unknown or difcult to
determine from the existing experiment which is why they need to
be calibrated. Once a suitable set of calibrated parameters has been
determined for a single engine operating point, these parameters
are left unchanged when the model is used as a predictive tool
for other operating points. Table 1 shows a list of the adjustable
parameters and their brief descriptions. Note: the geometric ratio,
which is dened as the ratio of the thicknesses of the neighboring
zones [34], is always set to 1 in this study.
3. Experimental data
The experimental apparatus is a single cylinder HCCI engine
combusting n-butanol and n-heptane blends [25]. The engine test
bed is a Ricardo Hydra Mark III using a Mercedes E550 cylinder
head with variable valve timing. The fuel is injected into the intake
runner. The intake air is preheated to a set temperature using a
600 W electrical band-type heater. The in-cylinder pressure is
measured using a Kistler piezoelectric transducer and a K-type
thermocouple with an accuracy of 2 C is installed in the intake
manifold and measures the intake temperature. Exhaust emissions
are characterized using a 5-gas emissions test bench (Horriba CLA510SS emission analyzer, which measures the gas concentrations
using the chemiluminescent method). The exhaust emissions are
sampled 5 cm downstream of the exhaust ports. NOx emissions
are measured with 1 ppm resolution, UHCs with 10 ppm resolution, and CO with 0.01% resolution. The crank angle position is recorded with 0.1 resolution using a BEI optical encoder.
Three Butanol Volume Percentages (BVPs) of the fuel at various
equivalence ratios are tested with the intake temperature and engine speed held constant. They are BVP 12 (representing a volume
fraction of 12% butanol in the fuel mixture), BVP17 and BVP 22. To
calibrate and validate the simulation, ve experimental operating
points are selected, i.e., BVP 12 (u = 0.332, 0.346), BVP 17
(u = 0.345, 0.357) and BVP 22 (u = 0.366). One operating point
(BVP 17 at u = 0.357) is used to calibrate the model while the other
four are used for model validation. The experimental operating
conditions of the engine are presented in Table 2. Some of these
parameters (listed in Table 1) are adjusted during the model calibration, as shown later.
4. Results and discussion
4.1. Model calibration
As mentioned earlier, the parameters listed in Table 1 need to be
calibrated using available experimental data since their values are

not directly measured and cannot be easily determined. The operating point with BVP 17 at u = 0.357 is used to calibrate the model. The
calibration process matches pressure traces during the compression
stroke by adjusting blow-by and matches ignition timing by varying
the intake temperature. Then other parameters such as the temperature stratication, fraction of residual gases, and heat transfer scaling factor are modied to achieve a small error between the model
prediction and experiment pressure trace over the entire CAD (Crank
Angle Degree) range. During calibration, the intake temperature is
not restricted to the measured intake manifold temperature since
the signicant amount of trapped gas, due to NVO, could cause preheating of the fresh airfuel mixture charge. Blow-by is found to be
negligible for this case and this is conrmed by literature [15]. It is
important to note that this initial calibration of model parameters
is conducted using the measured pressure trace only. The same set
of model parameters is then used to investigate the trend of HRR
and the formation of NOx and UHC, which have required a further
model calibration by adjusting the equivalence ratio, as discussed
in Section 4.2.
The uncalibrated and calibrated pressure traces from the initial
calibration are shown in Fig. 1 with the calibrated pressure trace
matching experimental data quite well. The uncalibrated pressure
trace does not take into account residual gases, blow-by losses, and
the temperature stratication. The calibrated model parameters
are specied in Table 3. It should be noted that the high residual
gas fraction of 0.29 (29%) is caused by early closing of the exhaust
valve to trap hot residuals. The cylinder wall temperature is not directly measured, but is estimated to be 20 above the coolant temperature. A relatively large core zone volume is selected with a
greater heat transfer scaling factor. This implies that for this initial
calibration case, there is a large volume of the hot fuel/air/gas mixture, mainly consisting of the residual gas, with a relatively uniform temperature in the cylinder. Between this core and the
cylinder wall, there is a signicant temperature gradient.
Using the calibrated pressure curve above, the HRR curve is calculated and shown in Fig. 2. Using the heat release analysis method
outlined in [35], the net HRR is calculated by accounting for any
heat transfer through the walls of the cylinder. The HRR curve generally shows adequate matching although the rst-stage ignition is
somewhat early and larger than experimental data. Experimentally, the rst-stage ignition peaks at around 15CAD bTDC (before
the Top Dead Center) whereas the simulation data indicates a peak
at 20CAD bTDC. Discrepancies in predicting the correct ignition
period may be due to the reaction mechanism or result from errors
in setting initial conditions. It can be attributed to the model limitation in fully describing the NVO effects as well, as discussed
later.
The measured NOx level for the calibration case is 5.4 ppm [25]
while the simulation result is 0.7 ppm. This under-prediction of the
NOx level can be explained by the thermal NOx formation. In fact,

169

S. Visakhamoorthy et al. / Applied Energy 94 (2012) 166173

60

80

BVP = 17
exp = 0.357
simul = 0.357

BVP = 17
exp = 0.357
simul = 0.357

70

Pressure (atm)

Pressure (atm)

60
50
40
30

50
40
30
20

20
10

10
-40

-20

20

40

-40

-20

20

40

CAD

CAD

Fig. 1. Uncalibrated (left) and calibrated (right) predictions of the pressure trace for BVP 17. The dashed lines are the simulation results and the solid lines are the measured
data.

Table 3
Calibrated model parameters.
Parameter (unit)

Value

Intake temperature (C)

Wall temperature (C)
Heat transfer scaling factor ()
Blow by loss
Core zone volume ()
Residual gas fraction ()
Thermal width (C)

122
90.0
40.0
Neglected
0.30
0.29
0.0

110

BVP = 17
exp = 0.357
simul = 0.357

90

HRR (J/CAD)

70
50
30
10

-40

-20

-10 0

20

40

experiments. Otherwise, the locations of model zones could be set

to better represent the temperature variation in cylinder. This
could result in a better prediction of NOx formation.
The predicted UHC emission of 226 ppm is much lower than the
measured 1780 ppm in experiment. This difference is likely due to
the thermal boundary layer in the numerical model being too thin.
In other words, combustion is more complete in the numerical
simulation and hence produces less UHC. This is observed during
the model calibration as the expansion stroke of the simulation
showed a slight deviation from the experiment. The deviation is
likely caused by the excess energy remaining in the charge and
can be reduced by increasing the size of the quenched thermal
boundary layer. However, increasing the size of this layer in the
simulation reduces the mass available for combustion and interferes with the proper combustion. Under experimental conditions,
the implementation of NVO facilitates combustion with fuel injections occurring during the NVO period [16]. With NVO, trapped
residuals can be oxidized during the NVO period and the fuel injected during this time facilitates combustion [16]. However, the
simulation model only operates between IVC and EVO and cannot
capture the NVO period and its associated phenomena. Thus, the
combustion facilitation effect of NVO is not fully captured in the
model and an increase in the size of the thermal boundary layer
is also restricted.

4.2. Model limitations in predicting NVO

CAD
Fig. 2. The HRR curves for the calibration case BVP 17. The solid line is the
measured data and the dashed line is the simulation result.

the predicted peak temperature in the cylinder ranges from 2100 K

to 2250 K for individual cases and all these values are above the
threshold temperature of the thermal NOx formation. The NVO
used in the experimental engine can introduce signicant amounts
of the inhomogeneity into the charge prior to ignition [36]. This
inhomogeneity occurs in both the temperature distribution and
species concentrations. Any fuel-rich regions caused by this inhomogeneity could cause localized hot spots and a higher production
of NOx. Those hot spots, especially when they occur across the cylinder and in individual zones, are not captured by the multi-zone
model due to its underlying assumption of complete mixing between trapped gasses and the inducted fuel/air charge. The model
also assumes the trapped residual gas with the same temperature
of the intake charge (although the intake temperature is adjusted
to account for the total energy balance). Since the in-cylinder temperature prole is difcult to measure, it is not available from the

Previous work [27] involving similar combustion modeling has

not been used to study experimental data from the NVO equipped
HCCI engines. Due to the implementation of NVO, there could be
variable amounts of trapped residual gases inside the cylinder
depending on the valve timing. Using the experimental valve timing and engine specications, the trapped gas percentage is determined using the ratio of the volume at the Exhaust Valve Closing
(EVC) to the volume at IVC and is found to be 29%. This level of
trapped gas is quite high but it is conrmed by other work using
this same engine [25]. Such high levels of the trapped gas are possible because the exhaust valve is closed before the piston reaches
TDC thus preventing all the exhaust gas from being evacuated. This
trapped gas could have a fairly high temperature [36] and this
should result in signicant charge preheating. In addition, the
trapped gas will modify the initial chemical compositions of the
fuel/air mixture in the cylinder. The inclusion of such a large percentage of exhaust gases in the subsequent intake charge has different effects with regards to the model development. Firstly, since the
multi-zone model does not account computationally for the temperature of trapped residual gases (it assumes the trapped gases

170

S. Visakhamoorthy et al. / Applied Energy 94 (2012) 166173

Table 4
Comparison of the model parameters before and after taking the NVO into account.
Original calibration
Parameter (unit)

New calibration accounts for NVO

Value Parameter (unit)

Intake temperature (C)

122
Heat transfer scaling factor () 40.0
Core zone volume ()
0.30
Thermal width (C)
0.0

Value

Intake temperature (C)

132
Heat transfer scaling factor () 1.0
Core zone volume ()
0.01
Thermal width (C)
30.0

are at the same temperature of the intake charge), the intake charge
temperature is specied as a manual input to the model. Secondly,
the initial calibration assumes at IVC, there is a homogeneous mixture throughout the cylinder and there is no any temperature stratication in the computational domain. However, in practice, NVO
could result in the signicant inhomogeneity and stratication of
the charge temperature and trapped residual gas at IVC [36]. This
can lead to fuel-rich regions which will ignite sooner than the assumed homogeneous charge. Alternatively, increasing the overall
equivalence ratio of the model could account for the earlier ignition
caused by fuel-rich regions and applying the temperature stratication could better mimic the experimental situation. Finally, this
model assumes that the residual gas is a complete combustion
products constituting of CO2, H2O and N2. However, the actual
residual gas contains also NOx and UHCs. These recycled NOx and
UHCs may contribute to combustion by effectively creating a
slightly richer mixture in the subsequent cycle. This effect may be
minimal for most operating points, but for points near the misre
limit, the recycling of unburned and partially burned reactants
can signicantly inuence subsequent cycles [37]. The energy
content of the intake charge should be increased when NVO is
present due to the heat energy of the trapped residuals and any
unburned or partially oxidized reactants.
With the increased energy content accounted, the model is
recalibrated for the case (BVP 17, u = 0.357) to improve matching
of the predicted cylinder pressure with experimental data while
relaxing constraints on the model parameters. The main difference
between the experimental and original simulation pressure curves
is a slight deviation during the expansion stroke likely due to more
energy remaining in the charge during expansion in the simulation. This can be improved by forcing more mass into the outer
zone to reduce the amount of mass which reacts. Doing so has
two effects: it increases total unburned reactant emissions and reduces the peak pressure since there is less mass available for combustion. To further improve the model predictions, the equivalence
ratio is increased by 14.8%. Not only does this aid in combustion,

but it also represents the increase in equivalence ratio caused by

trapped UHCs and any fuel-rich regions which may ignite early.
Normally, for engines not implementing NVO, the equivalence ratio is not adjusted and is set to the experimental value [27]. Admittedly, a 14.8% increase is fairly large. Nevertheless, this increase,
along with an increased mass in the outer zone, is the main adjustment between the new calibration and the original calibrated
results.
Some other minor changes in model parameters are listed in
Table 4. Note that the wall temperature and the residual gas fraction remain unchanged. The smaller heat transfer scaling factor is
used together with a smaller core zone volume and a greater thermal width. The new calibration results in excellent pressure trace
matching for the calibration case as shown in Fig. 3. Expansion
stroke pressure trace matching is also improved compared to the
original calibration. Using the pressure data, the HRR curve is calculated and matches the experimental results quite closely. In
addition to exhibiting a good peak behavior, the HRR curve shows
a signicantly improved rst-stage combustion timing and magnitude when compared to the experiment.
4.3. Model predictions
With this new set of calibrated parameters listed in Table 4, the
other four operating points of the engine are simulated. No other
parameters are changed during simulation except for BVP and
the equivalence ratio. All equivalence ratios are raised by 14.8%
above the experimental values.
For the four validation points the simulation results match closely to the experimental results except for a quite small deviation
during the expansion strokes of the predicted pressure traces, as
shown in Fig. 4. These simulation results capture the overall trends
of the system with a good prediction of the cool ame activity. For
example, changing from u = 0.332 to u = 0.346 with BVP12 shows
an increase in the peak pressure which is expected due to the richer mixture, although the low-temperature HRR remains constant, as shown in Fig. 5. All HRR curves in Fig. 5 show excellent
main-ignition timing and magnitudes and very good low temperature heat release timing and magnitudes. The effect of increasing
equivalence ratios on the cool ame activity for a given BVP will
be discussed later.
From the measured data, for a given BVP, HRR curves for various
equivalence ratios exhibit nearly identical low-temperature HRR,
but have varying high-temperature HRR. In other words, the equivalence ratio has little effect on the cool ame activity for a given
BVP. This same trend is observed in the simulation predictions.

BVP = 17
exp = 0.357
simul = 0.410

60

Pressure (atm)

40
30

70
50
30

20

10

10

-40

-20

90

50

HRR (J/CAD)

BVP = 17
exp = 0.357
simul = 0.410

110

CAD

20

40

-40

-20

-10 0

20

40

CAD

Fig. 3. Pressure and HRR curves for the improved calibration case. The solid lines are the measured data and the dashed lines are the simulation results.

171

S. Visakhamoorthy et al. / Applied Energy 94 (2012) 166173

60

60

BVP = 12
exp = 0.332
simul = 0.381

40
30
20

40
30
20

10

-40

-20

10

20

40

-40

-20

CAD

30
20

40

40
30
20

10
0

20

BVP = 22
exp = 0.366
simul = 0.420

50

Pressure (atm)

Pressure (atm)

40

-20

60

BVP = 17
exp = 0.345
simul = 0.396

50

CAD

60

-40

BVP = 12
exp = 0.346
simul = 0.397

50

Pressure (atm)

Pressure (atm)

50

10

CAD

20

40

-40

-20

20

40

CAD

Fig. 4. Predicted pressure traces for the four validation points. The solid lines are the measured data and the dashed lines are the simulation results.

Fig. 5. HRR curves for the four operating points. The solid lines are the measured data and the dashed lines are the simulation results.

172

S. Visakhamoorthy et al. / Applied Energy 94 (2012) 166173

Fig. 6. The simulated HRR curves with varying equivalence ratios for BVP 12.

Table 5
Experimental and simulation emissions.
NOx

UHC

BVP

Simulation
(ppm)

Experimental
(ppm)

Simulation
(ppm)

Experimental
(ppm)

12
12
17
17
22

0.322
0.346
0.345
0.357
0.366

0.60
0.54
0.55
0.53
0.52

5.5
5.4
5.3
5.4
5.1

492
503
521
531
558

1851
1746
1877
1780
1845

0.6

6
Simulation
Experimental

5.8
5.7

0.56

5.6
5.5

0.54

5.4
5.3

0.52

Experimental NOx (ppm)

0.58

Simulation NOx (ppm)

5.9

5.2
5.1

0.5
10

15

20

25

Butanol Volume Percentage

Fig. 7. NOx trends in experimental data and simulation for a number of engine
operating points listed in Table 5. The linear lines are plotted to show trends in both
the simulation (dotted) and the experiment (solid).

In addition to the two cases of u = 0.332 and 0.346 at BVP 12, three
more cases of u = 0.320, 0.340, and 0.360 are simulated. The HRR
curves for these ve cases are shown in Fig. 6, which shows that
varying the equivalence ratios for a xed BVP of 12 does not bring
about signicant changes in the low-temperature heat release.
The UHC predictions for the four predictive cases are on the order of 500 ppm. These levels are close to the experiment when compared to the rst set of calibration results and this is attributed to
the increased mass in the outer zone (i.e., the cooler thermal boundary layer). UHC levels are however still lower than the measured
data of around 1800 ppm. NOx emissions for the four validation

cases are approximately one order of magnitude lower than the

reported experimental values (about 5 ppm) as shown in Table 5.
As previously mentioned the model cannot capture the inhomogeneity in species concentrations, which leads to part of the model
discrepancy in predicting chemical species and temperature in
the engine. Considering the three mechanisms of NOx formation,
namely fuel, prompt, and thermal, NOx formation from fuel sources
is unlikely in this study as there is no fuel bound nitrogen. The
prompt NOx is not found in the simulation results and this agrees
with the previous nding that this mechanism was primarily
observed in fuel-rich ames where hydrocarbon groups react with
nitrogen to form Hydrogen Cyanide (HCN). The only effective mechanism, therefore, is thermal NOx. It is interesting to see that as BVP
is increased, NOx levels decrease (Fig. 7) and this is likely due to a
reduction in peak in-cylinder temperatures although no experimental temperature measurements are available to conrm this.
This same trend is observed in the simulation results. The NOx trend
can be also affected by the discrepancy in predicting certain chemical species such as N2 and O2 under engine conditions with NVO.
Again, the model has limitations in describing NVO equipped HCCI
engines. Nevertheless, even though the simulation results underpredict NOx levels, the model is useful for capturing overall trends
as shown in Fig. 7.
5. Conclusions
A numerical multi-zone combustion model of an n-butanol/
n-heptane fuelled HCCI engine utilizing NVO is studied. After calibrating the model to one operating point the simulations are run to
predict the pressure traces and heat release rates for a number of
operating points which are then compared to experimental data.
The model is calibrated to account for major effects of NVO on
the initial model conditions at IVC. The resulting calibrated model
produces pressure traces and the HRR curves that closely match
the experimental results at all ve operating points. However simulated NOx emissions are under-predicted compared to the experiment and this is attributed to the model limitation of not being
able to capture the signicant inhomogeneity of temperature and
chemical compositions in the practical engine equipped with
NVO. Similarly, the unburned hydrocarbon emissions are underpredicted but their predictions are improved when the major
effects of NVO are accounted for during model calibration. Since
the combustion model only operates for the closed valve period
of the engine cycle, it cannot completely capture the combustion
and emission effects of NVO. Nevertheless, calibration of the input
parameters of the model to account for some aspects of NVO does
result in excellent pressure trace and HRR predictions. This suggests that the model can be used as a predictive tool for investigating in-cylinder pressure, heat release, and combustion phasing in
NVO equipped HCCI engines fuelled with the different n-butanol/
n-heptane blends.
Acknowledgements
The authors of this paper would like to acknowledge the nancial support provided through the Auto21 project D301-DHC and
through NSERC.
References
[1] Yao M, Zheng Z, Liu H. Progress and recent trends in homogeneous charge
compression ignition (HCCI) engines. Progress Energy Combust Sci
2009;35:398437.
[2] Yamasaki Y, Kanno M, Taura Y, Kaneko S. Study on biomass gas HCCI engine.
SAE Int 2009;32:66.
[3] Haggith D, Sobiesiak A, Miller L, Przybyla G. Experimental indicated performance
of a HCCI engine fuelled by simulated biomass gas. SAE Int 2010;01:1081.

S. Visakhamoorthy et al. / Applied Energy 94 (2012) 166173

[4] Nathan SS, Mallikarjuna JM, Ramesh A. An experimental study of the biogasdiesel HCCI mode of engine operation. Energy Convers Manage
2010;51:134753.
[5] Maurya RK, Agarwal AK. Experimental study of combustion and emission
characteristics of ethanol fuelled port injected homogeneous charge
compression ignition (HCCI) combustion engine. Appl Energy 2011;88:
116980.
[6] Saisirirat P, Foucher F, Chanchaona S, Mounaim-Rousselle C. Spectroscopic
measurements of low-temperature heat release for Homogeneous Combustion
Compression Ignition (HCCI) n-Heptane/Alcohol mixture combustion. Energy
Fuels 2010;24:54049.
[7] Saisirirat P, Togbe C, Chanchaona S, Foucher F, Mounaim-Rousselle C, Dagaut P.
Auto-ignition and combustion characteristics in HCCI and JSR using 1-butanol/
n-heptane
and
ethanol/n-heptane
blends.
Proc
Combust
Inst
2011;33:300714.
[8] Dagaut P, Togbe C. Oxidation kinetics of butanol-gasoline surrogate mixtures in
a jet-stirred reactor: experimental and modeling study. Fuel 2008;87:331321.
[9] Maurya RK, Agarwal AK. Statistical analysis of the cyclic variations of heat
release parameters in HCCI combustion of methanol and gasoline. Appl Energy
2012;89:22836.
[10] Fathi M, Saray RK, Checkel MD. The inuence of Exhaust Gas Recirculation
(EGR) on combustion and emissions of n-heptane/natural gas fueled
homogeneous charge compression ignition (HCCI) engines. Appl Energy
2011;88:471924.
[11] Wu H-W, Wang R-H, Ou D-J, Chen Y-C, Chen T-y. Reduction of smoke and
nitrogen oxides of a partial HCCI engine using premixed gasoline and ethanol
with air. Appl Energy 2011;88:388290.
[12] Oakley A, Zhao H, Ladommatos N, Ma T. Experimental studies on controlled
auto-ignition (CAI) combustion of gasoline in a 4-stroke engine. SAE Int
2001;01:1030.
[13] Christensen M, Johansson B, Einewall P. Homogeneous charge compression
ignition (HCCI) using isooctane ethanol and natural gas a comparison with
spark ignition operation. SAE Int 1997:972874.
[14] Stanglmaier RH. TWR, Souder JS. HCCI operation of a dual-fuel natural gas
engine for improved fuel efciency and ultra-low NOx emissions at low to
moderate engine loads. SAE Int 2001;01:1897.
[15] Audet AD. Closed loop control of HCCI using camshaft phasing and dual
fuels. Edmonton: University of Alberta; 2008.
[16] Aroonsrisopon T. Analysis of stratied charge operation and negative valve
overlap operating using direct fuel injection in homogeneous charge
compression ignition engines. Madison: University of Wisconsin-Madison;
2006.
[17] Flowers D, Aceves A, Smith R, Torres J, Girard J, Dibble R. HCCI in a CFR engine:
experiments and detailed kinetic modeling. SAE Int 2000;01:0328.
[18] Aceves SM, Flowers DL, Westbrook CK, Smith JR, Pitz W. A multi-zone model
for prediction of HCCI combustion and emissions. SAE Int 2000;01:0327.
[19] Visakhamoorthy S, Tzanetakis T, Haggith D, Sobiesiak A, Wen JZ. Numerical
study of a homogeneous charge compression ignition (HCCI) engine fueled
with biogas. Appl Energy 2012;92:43746.

173

[20] Wang Z, Shuai S-J, Wang J-X. Multi-dimensional simulation of HCCI engine
using parallel computation and chemical kinetics. SAE Int 2008;01:0966.
[21] Ogink R, Golovitchev V. Gasoling HCCI modeling: computer program
combining detailed chemistry and gas exchange processes. SAE Int
2001;01:3614.
[22] Mack JH, Aceves SM, Dibble RW. Demonstrating direct use of wet ethanol in a
homogeneous charge compression ignition (HCCI) engine. Energy
2009;34:7827.
[23] Atkins MJ, Koch CR. The effect of fuel octane and dilutent on homogeneous
charge compression ignition combustion. Proc Inst Mech Eng Part D J
Automob Eng 2005;219:66575.
[24] Chen Z, Yao M, Zheng Z, Zhang Q. Experimental and numerical study of
Methanol/Dimethyl Ether dual-fuel compound combustion. Energy Fuels
2009;23:271930.
[25] Shahbakhti M, Ghazimirsaied A, Audet A, Koch CR. Combustion characteristics
of Butanol/n-Heptane blend fuels in an HCCI engine. in: Proceedings of
combustion institute Canadian section spring technical meeting. Ottawa;
2010.
[26] Zhao F, Asmus TW, Assanis DN, Dec JE, Eng JA, Najt PM. Homogeneous charge
compression ignition (HCC) engines: key research and development issues.
SAE Int 2003;03:646.
[27] Tzanetakis T, Singh P, Chen J-T, Thomson MJ, Koch CR. Knock limit prediction
via multi-zone modelling of a primary reference fuel HCCI engine. Int J Veh Des
2010;54:4772.
[28] Dagaut P, Togbe C. Experimental and modeling study of the kinetics of
oxidation of Butanol-n-Heptane mixtures in a jet-stirred reactor. Energy Fuels
2009;23:352735.
[29] Curran HJ, Gaffuri P, Pitz WJ, Westbrook CK. A comprehensive modeling study
of n-heptane oxidation. Combust Flame 1998;114:14977.
[30] Seiser R, Pitsch H, Seshadri K, Pitz WJ, Curran HJ. Extinction and autoignition of
n-heptane in counterow conguration. Proc Combust Inst 2000;28:202937.
[31] Naha S, Aggarwal SK. Fuel effects on NOx emissions in partially premixed
ames. Combust Flame 2004;139:90105.
[32] Barlow RS, Karpetis AN, Frank JH, Chen JY. Scalar proles and NO formation in
laminar opposed-ow partially premixed methane/air ames. Combust Flame
2001;127:210218.
[33] Flowers DL, Aceves SM, Martinez-Frias J, Dibble RW. Prediction of carbon
monoxide and hydrocarbon emissions in iso-octane HCCI engine combustion
using multizone simulations. Proc Combust Inst 2002;29:68794.
[34] Tzanetakis T. Multi-zone modeling of a primary reference fuel HCCI
Engine. Toronto: University of Toronto; 2006.
[35] Heywood JB. Internal combustion engine fundamentals. New York: McGrawHill; 1988.
[36] Rothamer DA, Snyder JA, Hanson RK, Steeper RR, Fitzgerald RP. Simultaneous
imaging of exhaust gas residuals and temperature during HCCI combustion.
Proc Combust Inst 2009;32:286976.
[37] Shahbakhti M, Koch CR. Characterizing the cyclic variability of ignition timing
in a homogeneous charge compression ignition engine fuelled with n-heptane/
iso-octane blend fuels. Int J Engine Res 2008;9:36197.