Applied Energy 191 (2017) 689708

Contents lists available at ScienceDirect

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

Understanding the role of low reactivity fuel stratification in a dual fuel

RCCI engine A simulation study
Maciej Mikulski , Cemil Bekdemir
TNO Automotive, Automotive Campus 30, 5708 JZ Helmond, The Netherlands

h i g h l i g h t s

Multi-zone, detailed chemical kinetics model used to assess natural gas stratification in RCCI.

5 heavy-duty engine operating points served as basis for simulation.

Natural gas stratification is a promising strategy for partial load RCCI.

Efficiency gain, THC/CH4 emissions reduction, increased NOx.

a r t i c l e i n f o a b s t r a c t

Article history: Natural gas - diesel, Reactivity Controlled Compression Ignition (RCCI) is currently one of the most
Received 13 September 2016 promising combustion strategies for the next generation heavy-duty engines. A major issue to be
Received in revised form 30 December 2016 addressed for this dual fuel concept to become practically applicable is its low combustion efficiency
Accepted 27 January 2017
and associated high methane emissions in partial engine loads. Introducing gaseous fuel stratification
(by direct injection) is considered to be beneficial for combustion efficiency increase. Yet, the improve-
ment potential of such natural gas stratification in RCCI engines is unknown. Modeling studies are needed
Keywords:
to investigate this methane reduction measure because they can provide more insight to the combustion
Low temperature combustion
RCCI
process than measurements alone.
Dual fuel The objective of this study is to assess the potential of RCCI with direct injected low reactivity fuel in
Natural gas terms of thermal efficiency and methane emissions. A broad range of measured, single point injected gas,
Fuel stratification operating points served as a basis for the simulation study. TNOs in-house, multi-zone, chemical kinetics
Multi-zone model based model was validated for those points and used to generate simulation results for different natural
gas stratification profiles.
It is shown that gaseous fuel stratification in dual fuel RCCI helps to increase combustion efficiency and
decrease methane and carbon monoxide emissions. At the same time, nitrogen oxides increase and (at
some operating points) the emissions of non-methane hydrocarbons slightly increase too. The effect is
strongest for the low load cases showing maximum improvement of combustion efficiency by 11 per-
centage point for in-extensive natural gas stratifications. The improvement potential reduces with
increasing engine load. Thanks to the insights to the combustion process (given by the simulation
results), explanations for the observed trends are provided, and important phenomena are identified that
are associated with increased low reactivity fuel stratification.
2017 Elsevier Ltd. All rights reserved.

1. Introduction
Despite powertrain electrification and hybridization are rapidly
progressing into the transport sector, it is expected that internal
combustion engines will retain their position as the main propul-
sion technology for heavy-duty road transport and the maritime
Corresponding author.
E-mail address: maciej.mikulski@tno.nl (M. Mikulski).
sector. For decades, strict pollutant emission standards, and more
recently, the targets for greenhouse gas (e.g. CO2) reduction, forces
powertrain developers to seek for increasingly more sophisticated
measures. Especially the push for CO2 reduction strengthened the
attention for advanced combustion concepts. The CO2 benefit of
applying advanced combustion concepts is due to the potential
for higher thermal efficiency and use of alternative fuels, such as
bio-fuels and natural gas (NG).

http://dx.doi.org/10.1016/j.apenergy.2017.01.080
0306-2619/ 2017 Elsevier Ltd. All rights reserved.
690 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708

Nomenclature

Abbreviations PRF primary reference fuel

aTDC after top dead center RCCI reactivity controlled compression ignition
BMEP break mean effective pressure [bar] SOI start of injection [CA]
BR blend ratio [%] SPI single point injection (indirect to intake manifold)
CA crank angle TDC top dead center
CFD computational fluid dynamics THC total hydrocarbons
CH4 methane UHC unburned hydrocarbons
CHR cumulative heat released [J] U equivalence ratio []
CN cetane number m_ mass flow rate [kg/s]
CO carbon oxide CA10, CA50 crank angle at respectively 10% and 50% heat re-
CO2 carbon dioxide leased
DF dual fuel Ct mixing coefficient []
DI direct injection dp/dCA pressure rise rate [bar/CA]
EGR exhaust gas recirculation N engine rotational speed [rpm]
EVO exhaust valve opening P in cylinder pressure [bar]
FD fuel distribution [] T temperature [K]
HC hydrocarbons Vdisp displacement volume [m3]
HCCI homogeneous charge compression ignition Y mass fraction [kg/kg]
HD heavy-duty c specific heat ratio []
HPDI high pressure direct injection g efficiency [%]
HRR heat release rate [J/CA]
IMEP indicated mean effective pressure [bar] Subscripts
IVC intake valve closure comb combustion (efficiency)
LHV lower heating value [J/kg] d diesel
LTC low temperature combustion g gas
NG natural gas max maximum
NMHC non-methane hydrocarbons pump pumping (efficiency)
NOx nitrogen oxides res residuals at EVO
PCCI premixed charge compression ignition wall cylinder walls (losses)
PM particulate matter
PPR peak pressure rise [bar/CA]

Low temperature combustion (LTC) is a class of advanced com-
pression ignition concepts that was extensively studied in the past
decades. LTC strategies were proposed to reduce engine-out emis-
sions of nitrogen oxides (NOx) and soot, while achieving higher
thermal efficiencies than conventional diesel engines. Single fuel
LTC strategies include homogeneous charge compression ignition
(HCCI) and premixed charge compression ignition (PCCI), wherein
the fuel is fully premixed and partially premixed, respectively.
Those concepts were studied by many researchers, for example:
[15] and proved to be inapplicable in practice due to the lack of
control over the combustion process, leading to a limited load
range and unstable operation [6].
Reactivity controlled compression ignition (RCCI) is a more
recent LTC strategy in which two fuels are used to improve the
controllability and load range compared to HCCI and PCCI, while
retaining the vast advantage of high efficiency and low NOx and
soot emissions at engine-out [7]. Various combinations of low
and high reactivity fuels were used to demonstrate RCCI, amongst
others: gasoline - diesel [8], ethanol - diesel [9,10], gasoline - bio-
diesel [11] and gasoline - (cetane number improved) gasoline [12].
Nowadays, natural gas (NG) attracts increasingly more atten-
tion as the low reactive fuel alternative in RCCI. NG is a so-called
low carbon fuel which adds to the CO2 emission reduction. Addi-
tionally, good availability, lower price compared to diesel and
gasoline, and the potential to utilize renewable sources (e.g. bio-
methane) make the choice of NG in heavy-duty dual fuel RCCI engi-
nes attractive. NG-diesel RCCI is still relatively uncharted territory
[13]. Most of the works covering this topic are still 3D CFD model-
ing studies [14,15] or single-cylinder research engine experiments
[1618]. Regarding applications in multi-cylinder, heavy-duty pro-
duction engines, TNO presented a NG-diesel dual fuel engine
demonstrator working in RCCI mode, showing the possibility to
reach 51% indicated thermal efficiency with ultralow NOx emis-
sions [19]. Several challenges were identified in that study. A high
engine-out CH4 emission, especially at low loads, is the main draw-
back of NG-Diesel RCCI. Furthermore, variations of the input
parameters result in high operation instability.
NG-diesel RCCI has challenging improvement points for the
concept to become practically applicable. The most important ones
are:

further engine efficiency optimization,

CH4 emission reduction,

increasing load range,

combustion control to enable transient and robust operation
under real-world conditions.
Various measures have been tested to improve the RCCI concept
in general.
Proper NG-diesel blend ratio (BR) adjusting showed to be extre-
mely important [20]. Lower BRs (low NG, high diesel) are used to
assure better auto-ignition and lower CH4 emissions at low loads
[19,20]. This, on the other hand, leads to extensive wall impinge-
ment using early diesel injection strategies, therefore, resulting
in increased non-methane hydrocarbon (NMHC) emissions. Higher
loads enabled higher BR operation with increased thermal effi-
ciency; BR of 96% was achieved with around 49% indicated effi-
ciency [19]. Introducing uncooled Exhaust Gas Recirculation
(EGR) combined with external heating of the intake air was identi-
fied as a promising measure to assure low load operation [19].
 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708
Additionally, it was shown that RCCI is particularly sensitive not
only to the amount and temperature of the introduced external
EGR, but also largely to its chemical composition, which introduces
another factor increasing operational variations [21].
Hanson et al. [22] investigated the effect of apparent compres-
sion ratio (CR). This was achieved by varying Intake Valve Closure
(IVC) timing. BR was changed to control combustion phasing. For
earlier IVC timing, higher BRs and combustion efficiency were
achieved. Note however, that increased CR keeping BR at
unchanged level, resulted in significantly higher pressure rise rates
and maximum pressures. This causes problems for high load oper-
ation where durability limits were exceeded.
Diesel injection timing [2325] and injection pressure [2628],
were identified as main regulation parameters to assure optimum
combustion for all RCCI concepts. The extent to which this control
strategy can be utilized is limited by the injection timing range that
provides RCCI. Start of injection (SOI) at 30 aTDC was shown to be
a limit for RCCI combustion [19]. Further retarding the diesel injec-
tion decreased the time for mixing up to auto-ignition, which
caused the NOx emissions to rise rapidly. On the other hand Nie-
man et al. [23] showed that in a single injection strategy, advanc-
ing SOI timing from 80 aTDC resulted in a significant increase in
NOx emission, while retarding it caused a substantial increase in
hydrocarbon (HC) and carbon monoxide (CO) emissions.
Kokjohn et al. [29] discussed the concept of split diesel injection
in gasoline-diesel RCCI. Varying injection timings of both fuels
showed slight potential in reducing HC emissions compared to sin-
gle injection operation. At the same time significant increase in
NOx and Particle Matter (PM) were observed. The HC reduction
was mainly accounted to the reduced wall impingement when
advanced second injection was used. The research was later
extended by the study of Hanson et al. [22], which varied diesel
fuel mass between the two subsequent injections. Similar to the
results of the injection timing, NOx and PM increased with the
additional second injection. This is explained by increased fuel
stratification with later injections, creating richer and more reac-
tive zones which tend to ignite earlier, increasing in-cylinder tem-
perature. On the other hand double and triple pulse diesel injection
(without EGR) was also tested by TNO for NG-Diesel RCCI but over-
all it failed to show significant potential for efficiency improve-
ment in the multi-cylinder engine [19]. Note that standard
heavy-duty type injectors were used in this study which might
explain the different results obtained with respect to the discussed
gasoline-diesel research by Kokjohn et al. [29] and Hanson et al.
[22] performed on a single cylinder research engine.
Fuel stratification created by diesel injection is strongly affected
by in-cylinder mixing. This was first reported by Benajes et al. [30]
who performed a combined experimental and numerical study to
understand mixing on the auto-ignition process in RCCI. The
authors used experimental data as input to an in-house 1D spray
model to clarify the mixing process with respect to in-cylinder fuel
BR and injection timing. Recently, the authors of the present work
successfully validated an in-house multi-zone combustion model
on HD NG-diesel RCCI engine, where slow mixing was identified
as one of the reasons for slow combustion in low load cases [21].
Furthermore, a dependency between model mixing coefficient
and intake pressure (increasing with increasing load) was identi-
fied. Also 3D Computational Flow Dynamics (CFD) models [15]
were used to further investigate the influence of in-cylinder mixing
on fuel stratification. Due to high computational expenses, the
scope of those studies are typically restricted to a few simulation
cases.
An alternative way to control the reactivity of the in-cylinder
charge is direct injection (DI) of the low reactivity fuel. Already
in 2004, Reynolds and Evans [31] reported on the possibility of
improving emissions and performance of lean burn NG engines
691
through partial-stratification. In 2011, Munchi et al. [32] intro-
duced a combustion concept based on Westports two-way gas-
diesel injection. The test engine was operated on different dual fuel
strategies depending on load. Despite the wide scope of that work,
it did not extensively cover the RCCI regime.
Recently, Kokjohn et al. [33] used optical engine experiments
supplemented by simplified chemical kinetics computations to
determine the combined effect of diesel and direct injected gaso-
line stratification on the ignition delay in RCCI mode. The authors
introduced a so-called PRF (primary reference fuel) number to
describe the combined stratification of both fuels. The PRF number
was found to be a dominating factor on ignition delay and ignition
location within the cylinder. The equivalence ratio proved to have
smaller but still significant influence, whereas the temperature
gradient was identified as least influencing factor.
In 2015, Zoldak et al. [34] reported a CFD simulation study with
comparative evaluation of 3 strategies: (1) NG-diesel RCCI with
premixed NG, (2) high pressure direct injection (HPDI) or diesel
pilot-ignited natural gas concept, which injects a single diesel pilot
before TDC followed closely by DI of NG to generate a NG diffusion
flame, (3) NG-diesel RCCI with DI NG, introduced as further pro-
gression of RCCI with the aim of reducing the peak pressure rise
rate and further extending operating range.
A rather complicated, triple injection (diesel pre-injection fol-
lowed by DI gas event and diesel pilot injection) strategy was
reported as best, to achieve a 19.8% improvement in fuel consump-
tion relative to the base diesel case. This strategy also tempered the
peak heat release rate, resulting in reduced peak pressure rise rate
and Pmax (Maximum in-cylinder Pressure) relative to the pre-
mixed NG strategy. This study, however, focused only at one oper-
ating point which was high load 17 bar BMEP not exploring the
potential for lower loads.
The objective of the present study is to assess the potential of
direct injected low reactivity fuel in RCCI for the improvement of
thermal efficiency and reduction of CH4 emissions. More specifi-
cally, direct injection of NG and diesel in a heavy-duty RCCI engine
is evaluated using numerical simulations. Gaseous fuel DI strategy
is considered a promising path to explore efficiency increase and
load range extension for the RCCI concept. There are only few
(already mentioned above [32,34]) reports available on this topic.
A vision on the expected benefits is still lacking. Modeling studies,
which can provide more insight to the combustion process than
measurements alone, are required to build a vision. To accomplish
that, a much broader research scope is needed than usually is
affordable with detailed CFD studies. Especially the expected DI
NG potential for different loads and in-cylinder mixing conditions
needs to be investigated.
For the development of RCCI control strategies and increasing
the insight into the combustion process, TNO has developed its
own chemical kinetics based, multi-zone combustion model [35].
It captures the most important phenomena in RCCI, such as in-
cylinder fuel and temperature stratification, mixing, auto-ignition
and combustion chemistry. The model has undergone a detailed
validation process, proving prediction accuracy of in-cylinder pres-
sure and engine performance parameters within cycle-to-cycle
variations [21]. Its completeness and relatively fast computational
times allow to cover wide range of engine operating points and full
operating cycle within the conducted research, which is the main
advantage over detailed CFD approaches.
The model already proved to be suitable for simulating diesel
fuel distribution effect combined with in-cylinder mixing for single
point, port injected (SPI) NG and DI diesel RCCI [35,21]. Operating
points with homogeneous distribution of NG served as the refer-
ence to improve upon. The focus of this study is on operating con-
ditions with and without EGR since both regimes have proven their
advantages in different RCCI operation ranges. In the following,
692 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708
first the modeling approach is introduced briefly in Section 2, with
special attention to model validation and the concept of fuel strat-
ification in direct injection simulations. Then, in Section 3, the sim-
ulation results are presented and extensively discussed. A
summary and the conclusions are given in Section 4.
2. Modeling approach and definitions
2.1. The multi-zone model
The TNO in-house multi-zone model is used in this study. The
governing equations of this model have been given in [36]. In the
present implementation, the chemical reaction kinetics are solved
using Cantera libraries [37], which is an open source object-
oriented code that offers flexibility regarding implementation in
other software tools and the straightforward application of differ-
ent reaction mechanisms for different fuels. A combined mecha-
nism for n-heptane (as diesel surrogate) and methane
combustion is used in the present work. The reaction scheme pro-
posed by Peters et al. [38] is selected to model n-heptane chem-
istry. The C1-C3 part of the mechanism is substituted by GRI3.0
[39], which is particularly suited to model CH4 ignition with C2H6
dilution. The mechanism also contains the NOx scheme of GRI3.0.
Altogether this chemistry model adds up to 354 reactions among
65 species.
In the present work, a cylindrical configuration consisting of 13
zones with varying volumes is used. The zones near the cylinder
liner have the smallest volumes to better reflect auto-ignition
dynamics. All zones exchange mass and heat with their direct
neighbors. Additionally, they exchange heat with the cylinder
and piston head, whereas heat transfer to the cylinder liner only
occurs with the outermost zone. The modified Woschni heat trans-
fer correlation proposed by Chang et al. [40] for low temperature
combustion concepts is used. The wall surface temperatures of
the head, piston, and liner are separately calculated based on
engine load and speed, and coolant conditions. The models zonal
configuration and the exchange processes are shown schematically
in Fig. 1.
Due to the early fuel injection timing as typically applied in
RCCI combustion, the spray induced mixing period for both fuels
is assumed to be much shorter than the ignition delay period.
Therefore, fuel mass is introduced to each zone according to a pre-
Fig. 1. Graphical representation of the multi-zone model with cylindrical zonal configuration.
defined zonal Fuel Distribution (FD). The fuel injection processes of
the liquid and gaseous fuels are included as Gaussian mass flow
profiles as a function of the crank angle. Evaporative cooling of
the injected liquid fuel is taken into account. The molecular gradi-
ent based inter-zonal mixing model also acts on the introduced
fuel during the ignition delay period.
The multi-zone combustion model is sensitive to EGR composi-
tion. Differences in unburnt hydrocarbons (UHC), oxygen content,
and minor fractions of highly reactive species in the exhaust affect
the combustion efficiency and emissions of the next cycle. A valve
flow model takes that variability into account and accurately esti-
mates in-cylinder conditions at IVC. Both the intake and the
exhaust manifold compositions are refreshed every simulation
cycle. The simulations are run until convergence is reached, which
typically takes 46 cycles. The scheme of a multi-cycle simulation
is shown in Fig. 2.
With the above configuration, the average time to compute one
full engine cycle is in a range of half an hour on a standard i7 VPro
desktop computer. This translates to about 23 h until converged
results are available. Note that this time can be further reduced
by applying zone and mechanism reduction strategies and code
optimization techniques. Furthermore, for performing model cali-
bration and extensive simulation research, XCCI is run on a com-
puter cluster.
The valve flow and the wall temperature models were cali-
brated and validated by comparing the results with the engines
GT-Power model. Then, the entire multi-zone combustion model
was validated against measured engine data. The approach and
the results of that study are presented in [21].
2.2. Post-processing method
To validate the model and to investigate the influence of gas-
eous fuel stratification on the combustion process a number of
combustion control, emission and efficiency parameters are com-
pared. Those parameters are obtained by post-processing raw sim-
ulation results which consists of pressure, temperature and
composition in all zonal volumes for each CA of the cycle. Post-
processing is done carefully to make comparisons with engine test-
ing data as fair and easy as possible. The definitions of the most
important post-processed quantities are briefly highlighted in this
section.
 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708
Fig. 2. Schematic view of the multi-zone model simulations.
At the first post-processing stage, the raw simulation results are
zone-averaged to give average pressure, temperature and composi-
tion as a function of crank angle. The pressure rise rate (dp/dCA)
from the multi-zone model is highly oscillating due to a limited
number of sequentially igniting zones. The high gradients are
smoothened by reconstructing the pressure trace by solving the
first law of thermodynamics for a single zone closed system with
the same heat-loss model [35]. This is only done for the purpose
of estimating the maximum pressure rise rate, so all other quanti-
ties are derived from the original pressure trace. The heat release
rate (HRR) is calculated using a standard procedure based on the
first law of thermodynamics, well described by Heywood [41]. A
constant ratio of specific heats (c = 1.368) is used. The net HRR is
corrected with heat transfer to the walls (calculated from the
model), to give the gross HRR, which is presented in the following
sections. Furthermore, the gross HRR serves as the basis for calcu-
lation of the cumulative heat release (CHR) and the determination
of CA10 and CA50 (crank angle at 10% and 50% heat released,
respectively). The gross and net values of indicated mean effective
pressure (IMEP) are calculated by integrating the pressure signal
through the closed part of the cycle and the full engine cycle,
respectively. Indicated efficiencies are defined as:
1=2  IMEP V N
gInet=gross _ net=gross disp
_ NG LHV NG
; 1
mdiesel LHV diesel m
where Vdisp is the displacement volume, N the engine rotational
speed. The symbol m _ denotes specific fuel mass flow and LHV
stands for lower heating value (for diesel and NG, respectively). Fol-
lowing the convention in Eq. (1), pumping efficiency is calculated
as:
1=2  IMEPgross  IMEPnet  V disp  N
gpump 1  _ NG LHV NG
_ diesel LHV diesel m
: 2
m
Emissions in terms of ppm (parts per million) are calculated
from the zone averaged cylinder composition at EVO (exhaust
valve opening). The specific hydrocarbon content at this point
together with CO concentration is taken to calculate combustion
efficiency. This is done with the simplifying assumption that all
of the UHC at EVO, except CH4, account to diesel fuel particles.
Then, the combustion efficiency becomes:
_ CH4;res LHV CH4 m
m _ CO;res LHV CO m
_ NMHC;res LHV diesel
gcomb 1  _ _
: 3
mdiesel LHV diesel mNG LHV NG
693
The m _ terms with subscripts: CH4,res, CO,res, and NMHC,
res represent the concentrations of the corresponding species at
EVO recalculated with the total exhaust flow to the adopted con-
vention of mass flow (kg/s). Finally, the wall heat losses (denoted
as gwall) are evaluated by comparing the cumulative heat loss from
the model with the available chemical energy introduced with the
fuel (i.e. denominator in Eq. (3)).
2.3. Test case selection and model validation
For the present research, 5 engine operating points were cho-
sen. Cases 1, 3 and 4 represent a combined IMEP and blend ratio
(BR) variation from low to high load operation, without external
EGR. BR definition used in this work is energy based:
_ NG LHV NG
m
BR : 4
m _ NG LHV NG
_ diesel LHV diesel m
As discussed in Section 1, higher load RCCI operation allows for
larger substitution of diesel with natural gas, which is reflected in
the selected cases.
Additionally, case 2 was chosen to investigate gaseous fuel dis-
tribution influence in high EGR regimes at low load operation. Case
5 represents load extension to 18.5 bar IMEP using high external
EGR strategy. The main characteristics of the operating points are
presented in Table 1.
For each of the operating points the model was first carefully
validated using measured engine data. The measurements were
performed on a multi-cylinder, heavy-duty engine with common
rail diesel Direct Injection (DI) and Single Point Injection (SPI) of
NG (i.e. gaseous fuel was injected to the mixer in the air path to
achieve homogeneous gas-air mixture). Table 2 provides the most
relevant (in terms of simulation reproduction) parameters of the
test engine. For detailed specification of the validation measure-
ment test stand refer to the earlier work by Mikulski and Bekdemir
[21].
Many measured parameters served as input to the model.
Amongst others, the manifold pressures and temperatures, diesel
fuel temperature, consumption rate, and injection timing, engine
oil and cooling water temperatures. Additionally, air and gaseous
fuel mass flows along with intake manifold CO2 concentration
were measured and used to determine intake flow composition.
Two mixing related model parameters were calibrated to fit the
model to measurements; namely liquid fuel distribution (FD)
694 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708

Table 1
Characteristics of operating points chosen for the research.

Case N [rpm] IMEP [bar] BR [%] SOI [CA] EGR [%] ktotal [] kg [] Remarks
1 1002 3.2 60 50 2.6 4.3 Low load, without EGR
2 1002 3.2 60 30 48 1.8 3.0 Low load, with EGR
3 1002 6.9 75 50 2.6 3.4 Mid load, without EGR
4 1002 10.8 85 52 2.0 2.4 High load, without EGR
5 1501 18.5 93 55 44 1.05 1.14 High load, with EGR

traces. However, the current 13 zone configuration proves to be

Table 2
Specification of the engine used for validation of the model.
sufficient to capture the most important quantities, such as com-
bustion phasing.
Engine type Heavy-duty, compression ignition, 6 cylinders in line
Regarding emissions, the model predicts the trends of decreas-
Compression 14
ratio
ing THC, CH4 and CO for increasing engine loads. Furthermore, the
Fuel path 1 Diesel: Standard EN590, Cetane number = 52, LHV = 42.7 MJ/
kg, common rail system, multiple injection capability
Fuel path 2 Natural gas (NG): Dutch NG, LHV = 38 MJ/kg, single point,
intake manifold injection
Air path Turbocharger, intake throttle
EGR path High-pressure, cooling, valve

among the zones and inter-zonal mixing coefficient (Ct). Recent

optical engine research has shown that for early injection RCCI,
most of the direct injected fuel accumulates close to the cylinder
wall, with decreasing concentration towards the center of the
cylinder [8]. In the present study, the diesel fuel is imposed to
the zones taking into account this experimental fuel distribution
knowledge. The calibrated mixing coefficients and diesel fuel dis-
tributions that resulted in the best fits are given for all the cases
(15) in Table A1 (Appendix A). Note that all these reference cases
have homogeneous distribution of natural gas because the corre-
sponding reference experiments were performed with SPI system.
As an example, the fitted result for case 3 is shown in Figs. 35. In
the following sections, case 3 is consequently used to illustrate the
simulation setup for the low reactivity fuel stratifications. Addi-
tionally, a summary of the validation results for all the cases is pre-
sented in Fig. 6.
Fig. 4. Experimental and simulated in-cylinder pressure and heat release rate as a
The model proves to be able to capture the pressure trace function of CA for case 3. Cycle-to-cycle variations of the measured pressure are
within cycle-to-cycle variations for the cases used in this research indicated with the bars.
(Fig. 4). Heat release rates are represented well by the model,
which is seen at the matching measured and simulated CHR traces
(Fig. 5). The instantaneous HRR on the other hand, is locally over-
predicted as it can be seen in Fig. 4. This spiky behavior is caused
by the zonal nature of the model in combination with the limited
number of zones. Increasing the zonal resolution gives smoother

Fig. 3. Initial zonal volumes, diesel fuel distribution, and diesel equivalence ratio
for the validated case 3. Vertical grid lines indicate zone walls spacing between Fig. 5. Experimental and simulated cumulative heat release as a function of CA for
the walls represents actual configuration. case 3.
 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708 695

Fig. 6. Engine out emissions and control related parameters - model (brick) vs experiment (blue) for test cases 15. Error bars indicate mean cylinder-to-cylinder variations
when such data was available. Emission values are given relative to the experimental value of case 1. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)

ratio of CH4 compared with other HC is preserved. Note that the interchangeably. Different levels of low reactivity fuel stratification
NOx emissions are ultra-low in all cases (typically below 50 ppm) are studied and presented in Section 3. For a fair comparison of the
and close to measurement accuracy limits. The exact quantitative results, a reference setting is introduced, which is the homoge-
matching is not possible due to de fact that the model represents neous natural gas distribution (same as SPI) but using DI. This
the behavior of only one cylinder, whereas the measured emissions means that for each of the 5 cases, identical intake conditions
are averaged of all the cylinders. Taking into account the significant (pressure and temperature) are used to create their own DI refer-
cycle-to-cycle variations in measured pressures (Fig. 4) and ence simulation. The results of both SPI and DI homogeneous sim-
cylinder-to-cylinder variations in control parameters (Fig. 6), the ulations for case 3 are shown in Fig. 7. Note that the slight
obtained model results are considered sufficiently accurate. differences originate from different masses aspired to the engine.
A clear dependency is observed between engine load and the
model Ct value, with low load cases fitting well using a Ct of 10
and a value of 1900 for the highest load point. This trend corre-
sponds to higher inlet pressures and resulting higher valve flow
velocities at higher loads, which are important factors affecting
turbulent mixing. The mixing process is identified as one of the
main drivers for the combustion propagation towards the inner
part of the cylinder.
Diesel injection quantity and timing have a significant impact
on FD, whereas Ct remains unaffected. For example, the diesel fuel
stratification, and therefore the associated local mixture reactivity,
decreases with earlier SOI. These effects follow from complex spray
mixing physics. In the current approach, changing the diesel injec-
tion parameters requires adjusting the FD input to reach proper
reproduction of combustion parameters.
Intake temperature and EGR ratio changes are captured well by
the model without the need to adapt the calibration parameters.
For a more detailed discussion on the validation approach and
results, the reader is referred to a former publication [21]. Here,
only general remarks from that study are highlighted.

2.4. Setup of low reactivity fuel stratification

After successful model validation for SPI natural gas, the model
was used to simulate Direct Injection (DI) of natural gas. Here, the Fig. 7. Cylinder pressure as a function of CA for case 3. Model results for
terms low reactivity fuel, (natural) gas, and gaseous fuel are used homogeneous gas distribution realized by SPI and DI simulations.
696 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708
The DI natural gas case aspires only air in contrast to the SPI case
where a mixture of air and natural gas enters the cylinder. With
IMEP differences in the range of the engines cylinder-to-cylinder
variations, this effect was chosen not to be compensated in the pre-
sent research.
Then, a number of sweeps representing different gaseous fuel
stratifications are performed for each case in Table 1. On a real
engine, different levels of stratification can be introduced by shift-
ing gaseous fuel injection timing and changing injection pressure.
Earlier injection allows more time for the fuel to premix resulting
in a more homogeneous mixture. Similarly increasing injection
pressure will enhance fuel premixing in the pre-combustion phase.
In this simulation study, the investigated gaseous fuel stratifica-
tions are based on the results of the optical study on fuel distribu-
tions in RCCI [33]. This was already successfully applied to choose
diesel fuel distributions before [21]. According to the aforemen-
tioned optical study, low reactivity fuel with much earlier injection
and larger injected mass creates a similar distribution, however,
with substantially lower gradients than those observed for high
reactivity fuel injections.
Using this information, a set of different stratification sweeps is
build which is aimed at representing DI natural gas conditions.
First, the gas equivalence ratio in zone 13 (Ug(13)) is varied sub-
stantially while the gradient of the gas equivalence ratio (DUg) is
kept constant. This was done unless the chosen gradient could
not accommodate the total amount of fuel to be distributed within
the cylinder; in such event the gradient was increased accordingly.
Then, for one of the points from the first sweep, a second sweep is
performed, this time varying the gradient DUg. Such a sweep is
translated to a so-called Fuel Distribution (FDg, subscript g denotes
the gas) which is a 13 element vector containing the percentage of
total fuel introduced to the zones. This vector is the actual input of
the model. Figs. 8 and 9 illustrate the performed sweeps for case 3.
The full research matrix is provided in Appendix A along with
detailed data on equivalence ratios and corresponding FDs.
For all cases the gaseous fuel is injected at the same point,
namely 5 degrees CA after IVC. The injection duration is set at con-
stant 3.5 degrees CA. The temperature of the injected gas is kept at
300 K.
As stated in Section 2.3, the mixing coefficient (Ct) is not influ-
enced by the parameters of diesel injection. With the discussed DI
gas concept, the gaseous fuel is injected at much earlier crank
angle and presumably using significantly lower pressures com-
pared to diesel injection. Therefore, it is assumed that the gaseous
fuel injection does not change the model mixing coefficient (Ct)
Fig. 8. Gaseous fuel equivalence ratio as a function of zone number for case 3,
varying Ug(13). Vertical grid lines indicate zone walls spacing between the walls
represents actual configuration.
Fig. 9. Gaseous fuel Equivalence ratio as a function of zone number for case 3,
varying DUg. Vertical grid lines indicate zone walls spacing between the walls
represents actual configuration.
and the diesel fuel distribution. Hence, those parameters are the
same as for the validated reference cases. The corresponding values
can be found in Appendix A.
3. Results and discussion
In this chapter, the multi-zone combustion model results are
presented and discussed in detail. First, in Section 3.1, the simula-
tion results of different gaseous fuel stratifications are discussed
for the low load operating points, followed by the high load and
mid load operating points in Sections 3.2 and 3.3, respectively.
The analysis includes the control relevant quantities: CA10, CA50,
IMEP, PMAX, and PPR (max dp/dCA), supplemented by emissions
and efficiency indicators (Eqs. (1)(3)). Directly after the mid load
results, the effects that occur by utilizing low reactivity fuel strat-
ification at mid load are analyzed in further detail in Section 3.4.
Heat release rates, temperatures, and species concentrations of
individual zones are discussed to understand the cause of the strat-
ification effects. Specific remarks regarding the details of the low
and high load operating points are included in Appendix B.
In the following subsections, the results of all the cases from
Table 1 are presented in the same figure format. The horizontal
axis depicts the simulations having different fuel stratifications
by the percentage of total gaseous fuel injected to the 12th zone
FDg(12). This means that the higher FDg(12), the higher the fuel
stratification (less homogeneous). The red crosses represent the
Ug (13) sweeps, keeping the gradient constant, and the black cir-
cles represent the DUg sweeps. For detailed information on the
gaseous fuel distribution described by the FDg(12) parameter the
reader is referred to Appendix A.
3.1. Low load (case 1 and 2)
Varying gaseous fuel stratification shows good potential for
combustion efficiency improvement at low load as is shown in
Fig. 10. Significant increase in both combustion efficiency and
overall indicated efficiency is observed even for small fuel gradi-
ents. This can be explained by the fact that NG rich zones are at
the same time also rich in diesel. So, the high reactive diesel pro-
motes more complete combustion of the nearby low reactive NG,
which leads to a shorter burn duration (decreased CA50-CA10).
The amount of fuel associated with the most inner part of the
cylinder volume (below the diesel injector), which is most difficult
for combustion to propagate towards (except for the crevices
which were omitted in this study), is decreasing with increased
 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708 697

Fig. 10. Gaseous fuel stratification parameter study for case 1. Higher FDg(12) index values (horizontal axis) reflect increased stratification. Red crosses represent Ug (13)
sweep (constant gradient). Black circles represent DUg sweep. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

distribution. This effect also contributes to the observed efficiency
gain.
In the combustion efficiency plot it can be seen that the most
stratified test case does not further improve. This is caused by
the over-rich zone at the cylinder liner (zone 13). With increasing
fuel stratification the amount of oxygen, which was already insuf-
ficient for complete combustion in the homogeneous reference
case, reduces even further. This explains the increased CO concen-
tration. For the same reason the ignition delay increases with
increasing fuel stratification as is shown in the CA10 plot. Addi-
tionally, the lower specific heat of gaseous fuel compared to air
causes the first igniting zone to have a lower compression temper-
ature. This effect was also reported in earlier studies of conven-
tional dual fuel [42]. Note that the slightly stratified case shows
the opposite effect, earlier ignition (compare homogeneous (black)
point at FDg(12) = 0.33 and neighboring red point at FDg(12) = 0.37
on the CA10 plot (Fig. 10)).
Emission of NOx increases significantly with increased fuel
stratification. Fuel rich zones release heat at high rates during com-
bustion which leads to high local temperatures where NOx is gen-
erated. The same local in-cylinder temperature increase is
responsible for a slight increase of the heat losses to the walls.
Zones having a high temperature in combination with a large heat
exchange area, due to zonal volume increase during ignition, are
the main contributors to the heat exchange process. However,
the average in-cylinder temperature is similar to the homogeneous
case because the fuel lean zones remain relatively cold. This model
behavior is in accordance with the generally known NOx depen-
dency on local in-cylinder temperatures. In Section 3.4, this phe-
nomenon is further discussed by inspecting simulation results in
detail.
For case 1 (low load without EGR), significant gain in indicated
efficiency is reached, up to 7 percentage point increase compared
to the homogeneous reference case. This is mainly due to a better
combustion efficiency while combustion timing (CA50) is main-
tained at desired values similar to the reference case. Perhaps more
important, the emissions of THC and CH4 are massively reduced.
The reference CH4 emission of 4500 ppm reduced to an acceptable
level of 200 ppm.
For case 2 (low load with EGR), similar trends are observed as
for case 1, see Fig. 11. Using relatively low gaseous fuel stratifica-
tion, almost 11 percentage point gain in combustion efficiency is
achieved which corresponds to 7 percentage point increase of indi-
cated efficiency. Also here THC emissions, which mainly consists of
CH4, are greatly reduced. Again, NOx results show that higher fuel
stratification leads to significant increase in NOx production rates.
Over 12 times increase was recorded for one of the simulation
points. The best efficiency point generated 4 times larger NOx
emissions compared to the reference case, but in absolute numbers
is still below 20 ppm.
698 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708

Fig. 11. Gaseous fuel stratification parameter study for case 2. Higher FDg(12) index values (horizontal axis) reflect increased stratification. Red crosses represent Ug (13)
sweep (constant gradient). Black circles represent DUg sweep. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

These low load results show a great potential for the application
of for instance early direct injection of gaseous fuel for RCCI.
3.2. High load (case 4 and 5)
For case 4 (without EGR) only a minor increase in combustion
efficiency is achieved with slight fuel stratification, as is shown
in Fig. 12. The main reason is that the reference combustion effi-
ciency is already as high as 99.5%, leaving almost no room for
improvement. The high combustion efficiency and low THC and
CH4 emissions were already observed during the experimental
studies [19] and were attributed to high turbulent mixing rates
[21]. The minor gain in combustion efficiency comes at the cost
of lower indicated efficiency due to retarded ignition. NOx emission
rises significantly following the same rationale as for case 1 and 2,
the locally elevated combustion temperature associated with mix-
ture inhomogeneity increases NOx production. Note that very
slight stratification can still be beneficial regarding emissions,
meaning THC and CH4 reduction with acceptable NOx values, and
almost the same indicated efficiency (compare homogeneous ref-
erence (first point from the left) and neighboring stratified point
FDg(12) = 4.1 in Fig. 12).
For case 5 (high load with EGR), stratification of gaseous fuel
reduces the combustion efficiency and the indicated efficiency,
see Fig. 13. This goes together with an increase of THC, and espe-
cially CH4 and CO emissions. These results are in contrast with
the results at low and mid loads. Explanation of the differences
require more detailed analysis of the combustion process, which
will be discussed in Appendix B. Note that NOx emissions remain
on the same level or even reduce for stratified mixtures, mainly
due to lower combustion efficiency. On the other hand, despite
the lower combustion efficiency, both the peak pressure rise rate
and the maximum pressure are increasing (excluding the cases
where combustion efficiency was lower than 98%), which can be
limiting particularly at high loads.
3.3. Mid load (case 3)
For the mid load case, similar trends as shown for the low load
cases are observed in terms of combustion phasing, emissions and
efficiency. The potential for reduction of THC and CH4 emissions is
high and the potential for efficiency gain is significant, however,
much smaller than for the low load cases. The combustion effi-
ciency of the reference case is 91% against the best case value of
96%, which is similar to the best case efficiency at low load (see
Fig. 14). This supports the thesis that the remaining few percent
of fuel is associated with the zone closest to the cylinder liner, sub-
jected to wall impingement with the injected diesel. Another indi-
cation is the inability to reduce the THC emission below 1000 ppm.
Further investigation on this issue is given in Section 3.4.
 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708 699

Fig. 12. Gaseous fuel stratification parameter study for case 4. Higher FDg(12) index values (horizontal axis) reflect increased stratification. Red crosses represent Ug (13)
sweep (constant gradient). Black circles represent DUg sweep. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

3.4. Detailed analysis of selected cases
In this section, selected cases from the above presented sweeps
are analyzed in detail. This is done to gain better insight into the
physical and chemical processes as a function of space and time
that stem from low reactivity fuel stratification. Special emphasis
is put on explaining the following phenomena which are a function
of gaseous fuel stratification:
(a) Increasing ignition delay and reducing combustion duration;
(b) Increasing NOx emission;
(c) Significantly increasing combustion efficiency for low load
cases;
(d) Decreasing performance and decreasing combustion effi-
ciency for high load cases.
Since all the generally manifested effects are found in the mid
load case, this operating point was chosen to form the core of
the discussion in this section, with particular emphasis on explain-
ing points a and b. The specific remarks on the low load cases
(point c) and the high load cases (point d) are presented in Appen-
dix B.
For the mid load, the homogeneous gas distribution case
(FDg(12) = 3.3) is compared to one of the stratified gas cases
(FDg(12) = 5.8) from Fig. 14. In Fig. 15, the ignition retarding effect
of fuel stratification is clearly seen from the pressure traces of 3
different simulations. The corresponding cumulative heat release
plots are shown in Fig. 16. These confirm the increasing combus-
tion efficiencies with increasing fuel stratification as is presented
in the parameter sweeps in Fig. 14.
Average in-cylinder temperatures are shown in Fig. 17. About
100 K higher peak temperature is obtained for the most stratified
case compared to the homogeneous case. These average temper-
atures suggest that the engine is still operated in the low temper-
ature RCCI mode. To explain the increase of NOx emissions when
using fuel stratification, the local temperature should be studied,
which in the multi-zone model is the temperature of the individ-
ual zones. Fig. 18 shows the zonal temperature evolution as a
function of the crankangle for both cases. The fuel richest zones
reach the highest temperatures during combustion. Zones 11
and 12 of the stratified case have maximum temperatures of
approximately 2300 K, whereas these zones reach about 400 K
lower temperature in the homogeneous case. This large differ-
ence results in more than 10 times higher NOx concentrations
in those high temperature zones of the stratified case. On the
other hand, the fuel lean zones of the stratified case remain rel-
atively cold compared to the same zones in the homogeneous
case, and therefore produce less NOx, though it only partially
700 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708

Fig. 13. Gaseous fuel stratification parameter study for case 5. Higher FDg(12) index values (horizontal axis) reflect increesed stratification. Red crosses represent Ug (13)
sweep (constant gradient). Black plusses represent DUg sweep. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

compensates for the overall NOx emission increase. This is clearly
visible in Fig. 20.
Note that despite zone 13 (at cylinder wall) is fuel rich and
ignites early in both cases, it hardly generates NOx. The tempera-
ture becomes substantially high (1860 K and 2072 K for the homo-
geneous and stratified cases, respectively), however, a lack of
oxygen suppresses NOx formation. The O2 mass fraction, shown
in Fig. 21, drops to zero due to hydrocarbon ignition chemistry,
before NOx reaction triggering temperatures are reached. Later in
the cycle, around 400 crankangle degrees, the NOx concentration
increases due to transport of NOx from neighboring zones where
it is present at much higher concentrations. Zone 13 is also the
main CO contributor (not shown on the plots) because of the lack
of oxygen, which explains the increasing CO emissions for the
stratified cases in Fig. 14.
The combustion duration of the stratified cases is shorter. A
zonal heat release analysis (Fig. 19) shows higher heat release rates
during ignition of the fuel richer zones. This results in faster trans-
fer of heat and radical species to the neighboring zones, which
therefore ignite earlier. Eventually, the combustion process propa-
gates faster through the in-cylinder mixture, leading to more com-
plete combustion.
However, the main factor responsible for more complete com-
bustion and therefore reduction of THC emissions (especially
CH4) is a proper alignment of the high and low reactivity fuel con-
centrations. In the stratified cases, the natural gas and diesel have a
similar stratification profile, low concentration at the center of the
cylinder and high concentration at the cylinder wall. Since in both
the homogeneous as the stratified cases the innermost zones 1 and
2 have difficulty to combust, the stratified case still has a favorable
combustion efficiency due to the lower amount of natural gas in
those zones (see Fig. 22). Furthermore, the fuel richest zone (zone
13) has a lower CH4 emission, despite the fact that the equivalence
ratio is much higher than the homogeneous case. This is attributed
to the faster breakdown of CH4 at the higher temperatures, but as a
consequence the emission contains the intermediate species CO in
significantly higher numbers.
A significant increase in ignition delay is presented for all strat-
ified cases. The plots in Figs. 23 and 24 show that stratified gaseous
fuel creates temperature and oxygen concentration variations
among the zones, on top of the wall heat-loss induced temperature
variation which also exists for the homogeneous case. At CA 310
the temperature difference between the coldest (zone 13) and hot-
test (zone 1) zones is 12.9 K and 3.4 K for the stratified and the
homogeneous cases, respectively. The increased temperature dif-
ference using stratification is caused by two effects. First, the
injected gaseous fuel is at 300 K, so there is a direct cooling effect.
Then, more important, the mixture properties (cp/cv) change when
 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708 701

Fig. 14. Gaseous fuel stratification parameter study for case 3. Higher FDg(12) index values (horizontal axis) reflect increased stratification. Red crosses represent Ug (13)
sweep (constant gradient). Black circles represent DUg sweep. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

Fig. 15. Pressure as a function of CA for different gaseous fuel stratification Fig. 16. Cumulative Heat Released as a function of CA for different gaseous fuel
variations of Case 3. stratification variations of Case 3.
702 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708
Fig. 17. Temperature as a function of CA for different gaseous fuel stratification
variations of Case 3.
natural gas is added, which results in lower compression tempera-
tures. Zone 13 in the stratified case, where ignition starts, has also
6.2 K lower temperature than the homogeneous case. Besides the
Fig. 18. Temperature per specific zone as a function of CA for homogeneous FDg(12) = 3.3 (left) and stratified FDg(12) = 5.8 (right) gaseous fuel distributions (Case 3).
Fig. 19. Heat Release Rate per specific zone as a function of CA for homogeneous FDg(12) = 3.3 (left) and stratified FDg(12) = 5.8 (right) gaseous fuel distributions (Case 3).
Note that different scales are used on both plots.
difference in oxygen content, the temperature difference is already
enough to significantly change the ignition delay.
Comparing HRR (Fig. 19), CH4 (Fig. 22) and n-heptane mass frac-
tion evolution reveal that there are two ignition points present. The
diesel ignition comes first (small peak on the HRR plots Fig. 19),
before TDC, during which only minor part of gaseous fuel com-
busts. The main combustion event comes after approximately 12
CA. This is associated with gaseous fuel, which is subjected to
reforming during the pre-ignition phase, reaching its ignition con-
ditions. This angle was more or less constant for both stratified and
non-stratified cases, which proves that it was mainly affected by
the diesel fuel ignition point. The diesel ignition moment itself,
on the other hand, changes significantly with increasing fuel strat-
ification. This point is mainly influenced by temperature and oxy-
gen concentration during the diesel ignition delay period. Detailed
plots of temperature and oxygen mass fraction zonal distributions
during this period are in Figs. 23 and 24 respectively.
The detailed energy breakdown given in Fig. 25 provides value-
able information on how the fuel energy is transferred to useful
energy and losses. From the homogeneous reference case to the
stratified case FDg(12) = 5.8, the combustion efficiency improved
by 4.5 percentage point. This gain comes for 1.8 percentage point
in favor of the indicated efficiency. The wall heat-loss increased
slightly (0.5 percentage point), which corresponds with the
observed higher combustion temperature. The remainder is mainly
 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708 703

Fig. 20. NOx mass fractions per specific zone as a function of CA for homogeneous FDg(12) = 3.3 (left) and stratified FDg(12) = 5.8 (right) gaseous fuel distributions (Case 3).
Note that different scales are used on both plots.

Fig. 21. O2 mass fractions per specific zone as a function of CA for homogeneous FDg(12) = 3.3 (left) and stratified (right) FDg(12) = 5.8 gaseous fuel distributions (Case 3).

Fig. 22. CH4 mass fractions per specific zone as a function of CA for homogeneous FDg(12) = 3.3 (left) and stratified FDg(12) = 5.8 (right) gaseous fuel distributions.

lost by higher exhaust losses due to the shifted combustion phas-
ing. However, the diesel injection timing sensitivity of the RCCI
operation gives a possibility to mitigate this undesired shift. So,
for this particular operating point, there is potential to reach 48%
indicated efficiency.
4. Conclusions
In this work, a total of 5 operating points, representing load,
blend ratio and EGR variations, served as a basis for simulating
the influence of low reactivity fuel stratification on the perfor-
704 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708

Fig. 23. Temperature near diesel injection event per specific zone as a function of CA for homogeneous FDg(12) = 3.3 (left) and stratified FDg(12) = 5.8 (right) gaseous fuel
distributions (Case 3).

Fig. 24. O2 mass fractions near diesel injection event per specific zone as a function of CA for homogeneous FDg(12) = 3.3 (left) and stratified FDg(12) = 5.8 (right) gaseous fuel
distributions (Case 3).

Fig. 25. Energy distribution for different gaseous fuel stratification variations of Case 3.
 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708 705

mance and emissions of a dual fuel natural gas-diesel RCCI engine.
These baseline, single point gas injection, operating points were
carefully validated with engine test-bench results. An in-house
multi-zone combustion model using detailed chemical kinetics
was used to provide simulation results for the analysis of different
levels of low reactivity fuel stratification. The following main con-
clusions are drawn from the discussion of the results:
load RCCI. The engine could be operated on highly stratified low
reactivity fuel at low loads, gradually reduced towards more
homogeneous fuel distribution at higher loads.
Funding
This work was funded by TNOs internal research program.

Introducing low reactivity fuel stratification improves combus-

tion efficiency in dual fuel RCCI. This is caused by combustion
reaching further into the in-cylinder mixture and by the lower
fuel content at the center of the combustion chamber, which
typically is the worst combusting zone. The strongest effect is
seen at low loads, where the homogeneous reference had lar-
gest potential for improvement. This potential reduces with
increasing load.
Stratification of low reactivity fuel delays ignition by the diesel
fuel and at the same time shortens the combustion duration.
Both a locally lower oxygen content, and a lower local temper-
ature at the most reactive zones are responsible for the delay.
This effect is most prominent and negatively affecting the
results at the highest load using EGR; the combustion efficiency
decreases and, therefore, CO emission increases.
Low reactivity fuel stratification causes higher local peak tem-
peratures during combustion. As a result, the NOx emissions
increase and CH4 emissions decrease with increasing
stratification.
Overall, low reactivity fuel stratification, which could be
achieved by direct injection in the cylinder, is a promising strategy
to improve efficiency and reduce hydrocarbon emissions at part
Appendix A
See Table A1.
Appendix B. Detailed analysis (low and high load cases)
B.1. Specific remarks for low load cases
The observations regarding combustion efficiency gain, NOx
emission increase etc., as discussed in Section 3 for the mid load
case, are confirmed for the low load cases as well. To avoid repeti-
tions and to keep the paper concise, the detailed analysis presenta-
tion of case 1 and 2 is only focused on the main differences from
the mid load case.
At low load, the efficiency gain using stratified gaseous fuel is
much higher. This is attributed to the low combustion efficiency
of the reference case. The reason for this, is the difficulty of chem-
ical reactions to reach the inner part of the cylinder. Model-wise
this process is mainly controlled by the inter-zonal mixing
coefficient.
In case 1 (low load without EGR) for the reference simulation,
the three innermost zones did not combust at all, which is visible

Table A.1
The scope of the gaseous fuel distribution simulations. Validated diesel fuel distributions (Ud) and mixing coefficient (Ct) values for the reference case are given at the beginning
of each subsection.

No. Ug(13) [] DUg [] FD(13) [%] FD(12) [%] No. Ug(13) [] DUg [] FD(13) [%] FD(12) [%]
Ug(13) sweep DUg sweep
Case 1; Ud(13) = 2; DUd = 0.6; Ct = 10
1.0 0.22 0 2.0 3.3 1.5 0.33 20 3.0 5.0
1.1 0.25 20 2.3 3.7 1.6 0.33 10 3.0 4.9
1.2 0.29 10 2.6 4.2 1.7 0.33 7 3.0 4.8
1.3 0.33 10 3.0 4.9 1.8 0.33 5 3.0 4.7
1.4 0.50 10 4.5 7.2 1.9 0.33 4 3.0 4.7
Case 2; Ud(13) = 1.67; DUd = 0.4; Ct = 10
2.0 0.23 0 2.0 3.3 2.6 1.00 10 8.8 13.4
2.1 0.29 10 2.5 4.1 2.7 1.00 3 8.8 10.5
2.2 0.33 10 3.0 4.8 2.8 1.00 2 8.8 9.2
2.3 0.50 10 4.4 7.0
2.4 1.00 10 8.8 13.4
2.5 2.00 10 17.7 24.5
Case 3; Ud(13) = 1.25; DUd = 0.3; Ct = 700
3.0 0.28 0 2.0 3.3 3.5 0.50 10 3.6 5.8
3.1 0.33 10 2.4 3.9 3.6 0.50 5 3.6 5.5
3.2 0.40 10 2.9 4.7 3.7 0.50 4 3.6 53.8
3.3 0.50 10 3.6 5.8
3.4 0.67 10 4.9 7.6
Case 4; Ud(13) = 1.67; DUd = 0.4; Ct = 1600
4.0 0.40 0 2.0 3.3 4.5 1.00 10 5.0 7.6
4.1 0.50 20 2.5 4.1 4.6 1.00 7 5.0 7.3
4.2 0.67 10 3.4 5.2 4.7 1.00 5 5.0 7.0
4.3 1.00 10 5.0 7.6 4.8 1.00 4 5.0 6.7
4.4 2.00 10 10.0 13.9
Case 5; Ud(13) = 2; DUd = 0.4; Ct = 1900
5.0 0.88 0 2.0 3.3 5.5 1.25 20 2.9 4.5
5.1 1.00 50 2.3 3.7 5.6 1.25 50 2.9 4.6
5.2 1.25 20 2.9 4.5 5.7 1.25 100 2.9 4.7
5.3 2.00 10 4.6 6.3
5.4 5.00 5 11.4 9.6
706 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708

in the CH4 mass fraction plots in Fig. B.1. For the stratified simula- Case 2 (low load with EGR) shows similar behavior as case 1.
tion the combustion efficiency increased due to shorter combus- However, the lower performance gain using stratification com-
tion duration and, more important, due to lower gaseous fuel in pared to case 1 is caused by a large fraction of CH4 that remains
the badly combusting inner zones, also shown in Fig. B.1. in the outermost zone (zone 13 in Fig. B.2). The diesel fuel conver-

Fig. B.1. CH4 mass fractions per specific zone as a function of CA for homogeneous FDg(12) = 3.3 (left) and stratified FDg(12) = 4.9 (right) gaseous fuel distributions. Low load
without EGR (case 1).

Fig. B.2. CH4 mass fractions per specific zone as a function of CA for homogeneous FDg(12) = 3.3 (left) and stratified FDg(12) = 4.7 (right) gaseous fuel distributions. Low load
with EGR (case 2).

Fig. B.3. O2 mass fractions per specific zone as a function of CA for homogeneous FDg(12) = 3.3 (left) and stratified FDg(12) = 4.7 (right) gaseous fuel distributions. Low load
with EGR (case 2).
 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708
Fig. B.4. CH4 mass fractions per specific zone as a function of CA for homogeneous FDg(12) = 3.3 (left) and stratified FDg(12) = 6.4 (right) gaseous fuel distributions. High load
with EGR (case 5).
Fig. B.5. O2 mass fractions per specific zone as a function of CA for homogeneous FDg(12) = 3.3 (left) and stratified FDg(12) = 6.4 (right) gaseous fuel distributions. High load
with EGR (case 5).
sion consumes all the available oxygen before CH4 combustion
ends (Fig. B.3). Note that the CH4 concentration in zone 13, shown
in Fig. B.2, increases at the early stage of combustion. The interme-
diate n-heptane reactions, mainly alkyl radical decomposition to
C3H8 which further reacts to CH4, not able to continue due to lack
of oxygen is the reason for this.
B.2. Specific remarks for high load cases
In contrary to the low load cases, the high load cases show no
(or very small) performance increase using gaseous fuel stratifica-
tion. The homogeneous references have already a high combustion
efficiency. Adding fuel stratification does not significantly affect
the combustion evolution through the zones. Without the potential
for combustion efficiency increase, the indicated efficiency is
penalized by the increased ignition delay.
The introduction of high fuel stratification levels lead to a deteri-
oration of combustion efficiency, because zone 13 becomes over-
rich in fuel, especially in case 5. Note that this operating point has
a low air-to-fuel ratio (Table 1). Figs. B.4 and B.5 show CH4 and O2
mass fraction as a function of the crank angle for all the zones for
case 5. The fuel rich conditions and lack of oxygen are clearly visible.
The result is an overall increase of THC emissions, especially CH4,
accompanied by a drastic increase of CO emissions (see Fig. 14).
707
References
[1] Tomita E. Dual fuel HCCI combustion high octane and high cetane number
fuels. Int J Eng Res 2005;6:45363.
[2] Yao M, Zheng Z, Liu H. Progress and recent trends in homogeneous charge
compression ignition (HCCI) engines. Prog Energy Combust Sci
2009;25:398437.
[3] Manente V, Johansson B, Tunestal P, Cannella W. Effects of ethanol and
different type of gasoline fuels on partially premixed combustion from low to
high load. SAE paper 2010-01-0871; 2010.
[4] Azimov U, Tomita E, Kawahara N, Harada Y. Premixed mixture ignition in the
end-gas region (PREMIER) combustion in a natural gas dual-fuel engine:
operating range and exhaust emissions. Int J Eng Res 2011;12:48497.
[5] Musculus MPB, Miles PC, Pickett LM. Conceptual models for partially premixed
low-temperature diesel combustion. Prog Energy Combust Sci
2013;39:24683.
[6] Dempsey A, Walker N, Gingrich E, Reitz RD. Comparison of low temperature
combustion strategies for advanced compression ignition engines with a focus
on controllability. Combust Sci Technol 2014:21041.
[7] Reitz RD, Duraisamy G. Review of high efficiency and clean reactivity
controlled compression ignition (RCCI) combustion in internal combustion
engines. Prog Energy Combust Sci 2015;46:1271.
[8] Hanson RM, Kokjohn SL, Splitter DA, Reitz RD. Fuel effects on reactivity
controlled compression ignition (RCCI) combustion at low load. SAE paper
2011-01-0361; 2011.
[9] Splitter DA, Hanson RM, Kokjohn SL, Reitz RD. Reactivity controlled
compression ignition (RCCI) heavy-duty engine operation at mid-and high
highloads with conventional and alternative fuels. SAE paper 2011-01-0363;
2011.
[10] Srinivas P, Woo Y, Kook S, Hawkes ER. Ethanol utilisation in a diesel engine
using dual-fuelling technology. Fuel 2013;109. p. 597e607.
708 M. Mikulski, C. Bekdemir / Applied Energy 191 (2017) 689708
[11] Hanson RM, Curran S, Wagner RM, Reitz RD. Effects of biofuel blends on RCCI
combustion in a light-duty, multi-cylinder diesel engine. SAE paper, no. 2013-
01-1653; 2013.
[12] Splitter DA, Reitz RD. Fuel reactivity effects on the efficiency and operational
window of dual-fuel compression ignition engines. Fuel 2014;118:16375.
[13] Wei L, Geng P. A review of natural gas/diesel dual fuel combustion, emissions
and performance. Fuel Process Technol 2016;142:26478.
[14] Puduppakkam K, Liang L, Naik C, Meeks E, et al. Use of detailed kinetics and
advanced chemistry-solution techniques in cfd to investigate dual-fuel engine
concepts. SAE Int J Eng 2011;1(4):112749.
[15] Nazemi M, Shahbakhiti M. Modeling and analysis of fuel injection parameters
for combustion and performance of an RCCI engine. Appl Energy
2016;165:13550.
[16] Raihan MS, Guerry ES, Dwivedi U, Srinivasan KK, Krishnan SR. Experimental
analysis of diesel-ignited methane dual fuel low temperature combustion in a
single cylinder diesel engine. J Energy Engi 2014;141:19437897.
[17] Walker NR, Wissink ML, DelVescovo DA, Reitz RD. Natural gas for high load
dual-fuel reactivity controlled compression ignition in heavy-duty engines. J
Energy Res Technol 2015;4(137).
[18] Guerry ES, Raihan MS, Srinivasan KK, Krishnan SR, Sohail A. Injection timing
effects on partially premixed diesel-methane dual fuel low temperature
combustion. Appl Energy 2016;162:99113.
[19] Doosje E, Willems F, Beart R. Experimental demonstration of RCCI in heavy-
duty engines using diesel and natural gas. SAE technical paper, no. 2014-01-
1318; 2014.
[20] Benajes J, Molina S, Garca A, Monsalve-Serrano J. Effects of direct injection
timing and blending ratio on RCCI combustion with different low reactivity
fuels. Energy Convers Manage 2015;99:193.
[21] Mikulski M, Bekdemir C, Willems F. Experimental validation of a combustion
kinetics based multi-zone model for natural gas-diesel RCCI engines. In
Symposium for combustion control Aachen; 2016.
[22] Hanson RM, Kokjohn SL, Splitter DA, Reitz RD. An experimental investigation
of fuel reactivity controlled PCCI combustion in a heavy-duty engine. SAE
paper, no. 2010-01-0864; 2010.
[23] Nieman DE, Dempsey AB, Reitz RD. Heavy-duty RCCI operation using natural
gas and diesel. SAE technical paper, no. 2012-01-0379; 2012.
[24] Li J, Yang WM, An H, Zhao D. Effects of fuel ratio and injection timing on
gasoline/biodiesel fueled RCCI engine: a modeling study. Appl Energy
2015;155:5967.
[25] Curran S, Hanson R, Wagner R. Reactivity controlled compression ignition
combustion on a multi-cylinder light-duty diesel engine. Int J Eng Res 2012;3
(13):21625.
[26] Hariprasad T. Effect of injection pressure on performance of dual fuel diesel
engine. SAE technical paper, no. 2013-01-2887; 2013.
[27] Walker NR, Dempsey AB, Andrie MJ, Reitz RD. Use of low-pressure direct-
injection for reactivity controlled compression ignition (RCCI) light-duty
engine operation. SAE Int J Eng 2013;2(6):122237.
[28] Liu J, Yao A, Yao C. Effects of diesel injection pressure on the performance and
emissions of a HD common-rail diesel engine fueled with diesel/methanol
dual fuel. Fuel 2015;140:192200.
[29] Kokjohn SL, Splitter DA, Hanson RM, Reitz RD. Experiments and modeling of
dual fuel HCCI and PCCI combustion using in-cylinder fuel blending. SAE Int J
Eng 2010;2(2):2439.
[30] Benajes J, Molina S, Garca A, Monsalve-Serrano J. Effects of low reactivity fuel
characteristics and blending ratio on low load RCCI (reactivity controlled
compression ignition) performance and emissions in a heavy-duty diesel
engine. Energy 2015;2(90):126171.
[31] Reynolds C, Evans RL. Improving emissions and performance characteristics of
lean burn natural gas engines through partial-stratification. Int J Eng Res
2004;1(5).
[32] Munshi SR, McTaggart-Cowan GP, Huang J, Hill PG. Development of a partially-
premixed combustion strategy for a low-emission, direct injection high
efficiency natural gas engine. Proceedings of the ASME 2011 internal
combustion engine division fall technical conference, no. ICEF2011-60181,
October 2011.
[33] Kokjohn SL, Musculus MPB, Reitz RD. Evaluating temperature and fuel
stratification for heat-release rate control in a reactivity-controlled
compression-ignition engine using optical diagnostics and chemical kinetics
modeling. Combust Flame 2015;162(6):272942.
[34] Zoldak P, Sobiesiak A, Wickman D, Bergin M. Combustion simulation of dual
fuel cng engine using direct injection of natural gas and diesel. SAE Int J Eng
2015;2(8).
[35] Bekdemir C, Baert R, Willems F, Somers B. Towards control-oriented modeling
of natural gas-diesel RCCI combustion. SAE technical paper, no. 2015-01-1745;
2015.
[36] Egz U, Leermakers CAJ, Somers LMT, Goey LPH. PCCI combustion modeling
with a multi-zone approach including inter-zonal mixing. Proc Inst Mech Eng,
Part D: J Automob Eng 2013;227(9):131324
[37] Goodwin DG, Moffat HK, Speth RL. Cantera: an object-oriented software toolkit
for chemical kinetics, thermodynamics, and transport processes. Version
2.2.1.; 2016 <http://www.cantera.org>.
[38] Peters N, Paczko G, Seiser R, Seshadri K. Temperature cross-over and non-
thermal runaway at two-stage ignition of n-heptane. Combust Flame
2002;128(12):3859.
[39] Smith GP, Golden DM, Frenklach M, et al. GRImech 3.0 reaction mechanism.
Berkeley; 2012 <http://www.me.berkeley.edu/gri_mech/>.
[40] Chang J, Gralp O, Filipi Z, Assanis D, et al. New heat transfer correlation for an
HCCI engine derived from measurements of instantaneous surface heat flux.
SAE technical paper. no. 2004-01-2996; 2004.
[41] Heywood JB. Internal combustion engines fundamentals. Singapore: McGraw-
Hill International; 1988.
[42] Mikulski M, Wierzbicki W. Numerical investigation of the impact of gas
composition on the combustion process in a dual-fuel compression-ignition
engine. J Nat Gas Sci Eng 2016;31:52537.