Applied Energy 231 (2018) 549–561

Contents lists available at ScienceDirect

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

Optimization of performance and operational cost for a dual mode diesel- T

natural gas RCCI and diesel combustion engine
⁎
Ehsan Ansari , Mahdi Shahbakhti, Jeffrey Naber
Mechanical Engineering – Engineering Mechanics, Michigan Technological University, 1400 Townsend Dr., Houghton, MI 49931, USA

H I GH L IG H T S

• RCCI has better performance and lower operational cost in comparison to CDC mode.
• AtThemedium and high loads, RCCI showed better fuel economy, lower NOx and PM emissions.
• In optimum
optimum EGT range for dual mode RCCI-CDC operation is around 400 to 425 °C.
• RCCI enginescalibration, the CA50 retards as engine load increases.
• are useful in on-road applications and stationary applications.

A R T I C LE I N FO A B S T R A C T

Keywords: Diesel-NG fuel blends are increasingly being used in Reactivity Controlled Compression Ignition (RCCI) internal
RCCI combustion engines due to high Brake Thermal Efficiency (BTE), low NOx and PM emissions. It also has few
Engine optimization disadvantages such as high unburned Hydro Carbon (HC) and Carbon Monoxide (CO) emissions and relatively
Natural gas low Exhaust Gas Temperature (EGT). This study determines the optimum combustion mode between RCCI and
Dual fuel engines
Conventional Diesel Combustion (CDC) at different loads while meeting the Environmental Protection Agency
After-treatment system
(EPA) emission regulation. A Cost Function (CF) including Brake Specific Fuel Consumption (BSFC) and Brake
Specific Urea Consumption (BSUC) is considered and minimized in this study. The optimization of input vari-
ables is done between 3 and 12 bar Indicated Mean Effective Pressure (IMEP) engine load. The study aims to
calibrate the dual fuel diesel/NG RCCI engine to meet Tier 3 Bin 20 EPA standard, with or without after-
treatment system, while minimizing the cost of operation.
New parametric empirical models are developed and validated using experimental data from a light duty 1.9L
inline 4 cylinder Compression Ignition (CI) engine as a function of independent input variables. All the ex-
periments were conducted at Advanced Power System (APS) facility at Michigan Technological University. These
models predict HC, CO, PM and NOx emissions, EGT and BSFC. These models are then used to predict new
operating points to increase the population in the optimization process. The computed EGT is used to estimate
the Selective Catalyst Reduction (SCR) and Diesel Oxidation Catalyst (DOC) efficiencies to assess the emission
data with different input variables by considering the after-treatment system to see if they meet the tailpipe
emission regulation. The optimization results recommend using Diesel/NG RCCI at 7 to 12 bar IMEP operating
conditions and use CDC for below 7 bar IMEP operating condition.

1. Introduction
Natural Gas (NG) has been used as an alternative fuel for over
100 years [1]. Some of the recent applications include NG with diesel in
dual fuel Low-Temperature Combustion (LTC) regimes [2–19]. Due to
the high octane number of NG, it is possible to increase the compression
ratio and increase the thermal efficiency via extending the load limit
[2,20–22]. RCCI is an attractive dual fuel LTC combustion mode which
has shown significant improvement in reducing NOx and PM emissions
in comparison to single fuel diesel mode. Despite these advantages,
RCCI has relatively high HC and CO emissions and low EGT which
makes the reduction and oxidation process difficult for exhaust after-
treatment systems. An optimization is needed to precisely select the
RCCI input parameters (including injection strategy, injection pressure,

⁎
Corresponding author.
E-mail address: eansari@mtu.edu (E. Ansari).

https://doi.org/10.1016/j.apenergy.2018.09.040
Received 25 October 2017; Received in revised form 4 August 2018; Accepted 5 September 2018
0306-2619/ © 2018 Elsevier Ltd. All rights reserved.
E. Ansari et al. Applied Energy 231 (2018) 549–561

Nomenclature HCCI homogeneous charge compression ignition [–]

IMEP indicated mean effective pressure [bar]
ACAP advance combustion analysis program [–] ISFC indicated specific fuel consumption [g/kW h]
ANR ammonia nitrogen ratio [–] IVC intake valve close [°CAD]
BSFC brake specific fuel consumption [g/kW h] IVO intake valve open [°CAD]
BR blending ratio [%] λ lambda [–]
BSUC brake specific urea consumption [g/kW h] LHV lower heating value [MJ/kg]
BTDC before top dead center [°CAD] LTC low-temperature combustion [–]
BTE brake thermal efficiency [%] ṁ Diesel mass flow rate of diesel fuel [g/s]
CDC conventional diesel combustion [–] ṁ NG mass flow rate of NG fuel [g/s]
CF cost function [–] MAP manifold absolute pressure [kPa]
CFD computational fluid dynamic [–] MPRR maximum pressure rise rate [bar/deg]
CI compression ignition [–] NG natural gas [–]
CO carbon monoxide [g/kW h] O2 oxygen concentration [%]
COV coefficient of variation [%] NOx specific nitrogen oxides [g/kW h]
CR compression ratio [–] P pressure [kPa]
DI direct injection [–] PCCI premixed charge compression ignition [–]
DOC diesel oxidation catalyst [–] PFI port fuel injection [–]
DOE design of experiment [–] QLHV lower heating value [MJ/kg]
DPF diesel particulate filter [–] RCCI reactivity controlled compression ignition [–]
EGR exhaust gas recirculation [%] SCF standard cubic feet
EGT exhaust gas temperature [°C] SCR selective catalyst reduction [–]
EHC electrically heated catalyst [–] SI spark ignition [–]
EPA environmental protection agency [–] SOI start of injection [°CAD]
EVC exhaust valve close [°CAD] TDC top dead center [–]
EVO exhaust valve open [°CAD] TDI turbocharged direct injection [–]
γ isentropic coefficient [–] V volume [m3]

boost and Exhaust Gas Recirculation - EGR) to maximize the engine BTE
and minimize the tailpipe emissions to meet the U.S. Tier 3 Bin 20
emission regulation. Plenty of works are available in the literature for
RCCI engines [3–10,23–37], but only a few works are found in the
literature for optimization of RCCI combustion mode
[9,27,32,34,38–41]. The summary of the prior optimization studies in
RCCI combustion mode is provided in Table 1.
In recent studies [24,40], researchers from Universitat Politècnica
de València, conducted an experimental study to investigate the cap-
abilities of the diesel-gasoline RCCI for a single cylinder CI engine, that
was modified for dual fuel operation. The optimized dual fuel injection
strategy was changed as a function of engine load. It was observed that
at low engine load operating conditions, early double diesel injections
led to ultra-low PM and NOx emissions. At higher loads, early pilot
diesel and close to TDC main injection strategy was performed. The
diesel pilot injection increases the reactivity of the mixture in crevices
and squish region, where low reactive fuel gets trapped and the main
diesel injection acts as an ignition source. Due to the less mixing time,
higher soot level was reported. Moreover, the partial diffusion burn rate
provides precise control of combustion phasing, i.e., crank angle of 50%
fuel burnt (CA50). At full load operating conditions, “Dual-Fuel Diffu-
sion” (single delayed diesel injection) was used. Running the engine at

Table 1
RCCI optimization and parametric studies available in literature.
Target Diesel-gasoline RCCI Diesel-NG RCCI

Optimization study Experiential • RCCI/CDC – ethanol-gasoline/biodiesel – medium duty –

Emissions regulation is considered [Benajes-2016]
• Duty
DOE analysis and real-time optimization – Experimental – Light
– Emission regulation is NOT considered [Ansari-2016]
• Injection/Combustion strategy and piston bowl – Heavy Duty
– Emission regulation is considered [Benajes-2017]
• This work – Injection pressure and timing, boost, EGR, EGT, EPA emission
regulation and after-treatment is considered
Experimental and Numerical
Numerical • EGR, Equivalence Ratio – Heavy Duty – Emission regulation is
NOT considered [Kavuri-2016]
• BR, Injection Strategy, DI pressure, EGR – Numerical – Heavy Duty –
Emission regulation is NOT considered [Nieman-2012]
• considered
Piston Bowl shape – Light Duty – Emission regulation is NOT
[Hanson-2012]
• Regulation
Injection Strategy, BR, Spray Angle – Heavy Duty – Emission
is considered [Nazemi-2015]

Parametric study Experiential • considered [Splitter-2014]

Intake P, T and ∅ – Heavy Duty – Emission regulation is NOT • regulation
Input variable sweep – Experimental – Heavy Duty – Emission
is NOT considered [Dahodwala-2014,2015]
• regulation
Parametric study – Light Duty – Numerical – Emission
is NOT considered [Li-2014]
• regulation
DI injection timing sweep – Experimental – Heavy Duty – Emission
is NOT considered [Doosje-2014]
• NOT
DI Injection timing sweep – Light Duty – Emission regulation is
considered [Walker-2013]
• Transient study – Experimental – Light Duty - Emission
regulation is NOT considered [Hanson-2014]
Numerical • EGR and PFI fuel – Heavy Duty – Emission regulation is
considered [Yang-2013]
• Initial Temperature, Fuel Composition – Heavy Duty – Emission
regulation is NOT considered [Kakaee-2015]
• BR and SOI – Light Duty – Emission regulation is NOT
considered [Li-2015]

550
E. Ansari et al.
maximum rated torque at 14 bar IMEP was achievable through this
strategy without exceeding the mechanical constraints (i.e., Maximum
Pressure Rise Rate – MPRR) of the engine. However because of the
diffusion nature of the combustion, soot and NOx emissions did not
meet the EURO VI emission regulation [24,40].
Another study [42] from the University of Wisconsin-Madison
showed the mismatch between chemistry and the engine timescale
makes RCCI operation difficult at low speeds and high engine loads. In
this study, the authors explored the various levels of EGR and airfuel
ratio at low speeds and high engine loads on a diesel-gasoline RCCI
engine. A full factorial design of experiment of genetic algorithm was
used at 20 bar IMEP and 1300 rpm at three level of EGRs and equiva-
lence ratios. The result of optimization suggested that 50% of the ga-
soline blend ratio and low EGR is the optimum strategy to lower the
NOx formation and late SOI2 to meet the required load without ex-
ceeding the MPRR [42].
In another study [32] on a 0.47 L CI engine, the authors investigated
the effect of five input parameters – including Blend Ratio (BR), EGR,
temperature and pressure at Intake Valve Closing (TIVC and PIVC) and
Start of Injection (SOI) – on thermal efficiency, emission, ignition
timing and ringing intensity [32]. The results of genetic algorithm op-
timization showed that TIVC and EGR are the most dominant variables
affecting the engine performance and emission. It was observed that BR
and SOI are two dominant factors that affect NOx emission. Lowering
the EGR or increasing the TIVC increases the ringing intensity [32]. It
was also observed that RCCI combustion with advanced SOI and higher
BR shows higher thermal efficiency. In this study, they found a sweet
spot of 66.5% BR to achieve the best net indicated thermal efficiency
and low NOx emission [32].
In a study by Nieman, et. al. [9], a KIVA3V CFD code, was coupled
with CHEMKIN chemistry tool in a wide range of operating conditions
from 800 to 1800 RPM and 4 to 23 bar IMEP. The independent variables
that were investigated in genetic algorithm included: BR, DI injection
strategy, DI injection pressure and EGR percentage. The goal of this
Fig. 1. Engine schematic view.
551
Applied Energy 231 (2018) 549–561
optimization was to minimize the NOx, soot, CO, HC, ISFC and ringing
intensity in NG-diesel RCCI engine. The results of the optimization
showed that single late SOI decreased the MPRR and maximum in-cy-
linder pressure, and made it possible to reach 22 bar IMEP without
exceeding the mechanical constraint. It has been observed that proper
DI spray targeting is essential for complete RCCI combustion. Utilizing
two DI injections showed more optimal reactivity distribution in the
combustion chamber, resulting in better combustion and lower HC
emission [9].
High HC and CO emissions and relatively low EGT in RCCI engines,
bring a significant challenge for RCCI engines since the after-treatment
system’s efficiency depends on EGT. SCR and DOC are two of the main
exhaust after-treatment components for internal combustion engines.
Oxidation catalysts typically require high catalyst temperature of about
400 °C to oxidize the HC emission from methane fuel [43,44]. The DOC
has 95% conversion efficiency for HC and CO emissions at 450 and
200 °C, respectively [43,44]. The EGT of the light duty RCCI engines in
recent studies [25,44] were reported below 450 °C for the entire oper-
ating map.
In another study [16] from Sahand University of Technology, in-
depth simulation work was conducted to investigate the effect of diesel
injection strategies on NG/diesel light duty RCCI engine. It was ob-
served that increasing the BR increase the ignition delay due to the
lower reactivity of the entire mixture. This also helps to decrease the
NOx formation due to the lower pressure rise rate and the peak of heat
release rate. It was also observed that higher diesel quantity, in first DI
injection, advances the combustion phasing due to the faster heat re-
lease rate and higher pressure rise rate [16].
Most of the prior RCCI optimization studies [5,6,23,25–29,45] are
done for diesel-gasoline RCCI engines but little work is done for NG-
diesel RCCI engines. Table 1 shows the experimental/numerical para-
metric studies were completed on diesel-NG RCCI engines to investigate
the impact of the various control parameters (such as NG substitution,
EGR and injection strategy) on engine performance, combustion
E. Ansari et al. Applied Energy 231 (2018) 549–561

characteristics and emissions [2–11]. done in dry conditions.

Table 1 shows there is no optimization study in literature for NG-
diesel RCCI engines that considers “tailpipe” emissions, EGT and ex-
haust after-treatment efficiency. To the best of authors’ knowledge, this
is the first study undertaken to develop a combined empirical and
model-based optimization framework for the NG-diesel engine coupled
with SCR and DOC exhaust after-treatment systems. This study is ap-
plied on a light duty 1.9-liter NG-diesel RCCI engine with a target to
meet the US Tier 3 Bin 20 emission regulation. In addition, the cost of
operation (fuel + urea) of RCCI and CDC is determined at 1500 RPM
and 3 to 12 bar IMEP engine load.
This paper is organized as follows. The next section explains the
experimental setup along with design of experiments. Next, the ex-
perimental results are presented and utilized in DOE analysis. Then,
empirical models are developed to predict HC, CO, NOx, PM emissions,
AVL 415SE Smoke Meter was used to measure the soot content of
the exhaust gas. This AVL unit is capable of measuring as low as
0.01 mg/m3 of soot concentration in the exhaust gas, which makes it
suitable for dual fuel diesel-NG applications with low smoke emissions.
The measurement accuracy of the pressure sensors, flow meters, en-
coder, soot, and smoke meter are provided in Table 6.
The uncertainties in the presented results are calculated using the
Root Sum of Square (RSS) method and the uncertainty propagation
analysis is done according to NIST Technical Report 1297 (Taylor B.N
and Kuyatt). Where Y is the value of the calculated quantity, U re-
present the uncertainty of the variable and X is the measured variables.
2
∂Y
UY = ∑ ⎛ ∂X ⎞ Ux2
⎜ ⎟

⎝ i⎠ (2)
EGT and BSFC. These models are then used to create the required pool
of data for engine optimization. Finally, the optimization results are A summary of the uncertainty in calculated parameters for the en-
analyzed and conclusions are made. gine performance and emission metrics is reported in Table 7.

2. Experimental setup
3. Design of experiments
For this study, a 1.9L inline 4 cylinder Turbocharged Direct
Injection (TDI) Compression Ignition (CI) engine was used. The engine Around 120 experiments were conducted by sweeping the RCCI
is modified for NG-diesel RCCI operation, as shown in Fig. 1. The engine input variables. The range of the independent input variables is pro-
is connected to a 150 kW Dynamitic 8100 eddy-current absorption vided in Table 8.
dynamometer. This dynamometer has a maximum speed of 6000 rpm. The RCCI mode was performed with either one single DI or two DI
The brake power was measured by a strain gage load cell that de- injections. Single PFI injection in intake port was done when the intake
termined the brake torque on the dynamometer. valves are open. Since the PFI fuel (NG) in this study was in gas form,
The engine specifications are provided in Table 2. High-pressure there was no charge cooling effect due to the fuel evaporation, so there
and low-pressure EGR loops were installed between the intake and was no benefit in injecting of the NG while the intake valves were
exhaust ports of the engine. A small throttle valve in low-pressure EGR closed. On the other hand, based on the previous studies [47], the
was installed for precise control of EGR rate. The intake manifold was turbulence kinetic energy of the flow – during the intake stroke – helps
modified in order to install NG Port Fuel Injectors (PFI). This mod- for better mixture preparation; thus, open valve injection strategy was
ification allowed the injectors to spray the fuel into the intake ports. utilized.
Table 3 shows the specifications of DI and PFI injectors in this study. The injection pressure of DI diesel fuel was constant at 400 bar and
One AVL GH15D piezoelectric pressure transducer was used to sense NG was injected at 4.1 bar. Diesel only mode was performed with single
the in-cylinder high-speed pressure data. The high-speed data including diesel DI pulse and the timing was controlled to fix the combustion
in-cylinder pressure, Manifold Absolute Pressure (MAP) and fuel flow phasing of 50% mass fraction burned at 8°ATDC. The operation of RCCI
were recorded by Advance Combustion Analysis Program (ACAP) with engine was constrained by 10 bar/deg of MPRR as the limit to avoid
1-degree crank angle resolution. The combustion related parameters engine damage and Coefficient of Variation (COV) of IMEP by 5% was
such as the fraction of burned mass (B ) was computed by using the used for the limit to avoid unstable combustion.
following equation [46]: In this study, energy-based BR, according to Eq. (3), was used to
quantify the NG diesel fraction. ṁ is the fuel flow rate and QLHV is the
V (i−1) ⎞ V (i−1) lower heating value of the fuel.
B (i) = ⎡P (i)−⎛ ⎜ ⎟ γ × P (i−1) ⎤
⎢ ⎝ V (i) ⎠ ⎥ Vs (1)
⎣ ⎦ ṁ NG . Q LHVNG
BR% = × 100
P and V are the cylinder pressure and its corresponding volume, Vs is ṁ NG . Q LHVNG + ṁ Diesel . Q LHVDiesel (3)
the swept volume and γ is the average isentropic coefficient during
compression and expansion in each engine cycle. Low-pressure loop EGR was used in order to control NOx, com-
Other low-speed data including air flow, various temperatures were bustion phasing and MPRR as previously shown in Refs. [44,48] as an
logged using National Instrument (NI SCXI) hardware. effective means for RCCI combustion control. EGR was calculated based
The DI fuel in this work was Ultra Low Sulfur Diesel (ULSD), with on the oxygen concentration (O2%) difference between intake and ex-
fuel properties listed in Table 4. The PFI fuel was methane with a haust ports.
minimum purity of 99.5%.
A MotoTron MPC-555 Blackoak ECU was programmed to control Table 2
and monitor the engine during experiments. This controller was pro- Engine specifications.
grammed in order to enable the user to control the fuel injectors’ Cylinder arrangement I4
timing, duration, injection pressure and also the EGR throttle in real
Displacement 1.89 L
time.
Bore/stroke 79.5/95.5 (mm)
HORIBA MEXA-1600D/DEGR was used for exhaust emission mea- Compression ratio 17:1
surements. This system includes three different analyzers: CO/CO2 Max. power 66 kW (90 hp) @3750 RPM
analyzer (AIA-260), O2/CO2 analyzer (INA-260), and THC/NOx ana- Max. torque 210 Nm (155 ft-lb) @ 1900 RPM
lyzer (FCA-266). The FCA-266 analyzer used Flame Ionization IVO 354 (°BTDC)
IVC 169 (°BTDC)
Detection (FID) to measure the total HC in exhaust sample. Accuracy
EVO 162 (°ATDC)
and operating range of the emission analyzer for exhaust gas emission EVC 351 (°ATDC)
measurements are listed in Table 5. Sampling of the exhaust gasses was

552
E. Ansari et al. Applied Energy 231 (2018) 549–561

Table 3 Table 8
DI and PFI fuel injectors’ specifications. Range of variables in RCCI combustion mode.
Fuel injector specification BOSCH CRDI - DI BOSCH NGI2 - PFI Input variable Range Unit

# of injection holes 6 4 Pilot injection timing −55 to −50 °ATDC

Injector hole diameter (mm) 0.165 NA Main injection timing −15 to 0 °ATDC
Spray included angle (°) 144 15 NG energy based blend ratio 80–95 %
EGR 0–30 %
MAP 88–150 kPa
λ 1.1–3.6 –
Table 4
Load (IMEP) 3–12 bar
Fuel properties.
Fuel type ULSD CH4

Total sulfur (ppm) 7 0

Boiling point (°C) 184 −164
Cetane index number 48.7 0
Water content (ppm) 34 0
Higher heating value (MJ/kg) 45.68 55.5
Lower heating value (MJ/kg) 42.89 50.01

Table 5
Emission measurement system.
Gases Unit Measurement range Span Gases
Fig. 2. Injection strategy used at (A) medium (7, 8 and 9 bar IMEP) loads to
CO % 0–1 vol% to 10 vol% 1.00 reduce the HC emission and (B) at high (10, 11 and 12 bar IMEP) loads to
CO2 % 0–1 vol% to 16 vol% 13.99 decrease the MPRR.
NO/NOx ppm 0–100 to 1000/0–2000 to 5000 ppm 1553
O2 % 0–10 vol% to 25 vol% 4.005
THC ppm 0–100 to 1000/0–2000 to 20000 ppm 2100 maximum BTE) [44].

Table 6 4. Experimental results

Measurement accuracy of the sensors.
Sensors Resolution Unit
RCCI experimental data was collected for the test conditions in
Table 8. Fig. 2 shows the fuel quantity split between the pilot and main
AVL smoke meter 0.01 mg/m3 injections for different engine loads. This modification was done to
Pressure transducer 19 pC/bar avoid exceeding the MPRR mechanical constraint at higher engine
Encoder 1 deg
loads (i.e., 10, 11 and 12 bar IMEP) and decreasing the unburned HC at
Micromotion Coriolis flow meter lower engine loads (i.e., 7, 8 and 9 bar IMEP) to meet the emission
Maximum flow rate 108 kg/h
regulations.
Zero stability 0.002 kg/h
Mass flow accuracy ± 0.05 % The RCCI premixed burn rate is increased by increasing the pilot
Volume flow accuracy ± 0.05 % injection pulse width, while diffusion flame is decreased by decreasing
Mass & Volume repeatability ± 0.025 % the main injection pulse width. This modification increases the re-
Density accuracy ± 0.2 kg/m3
activity of the mixture in squish region and crevices and helps to burn
Air flow measurement all the premixed fuel [44]. Although this method is effective in reducing
LFE air flow rate accuracy ± 0.86 % the HC and CO emissions, it increases the MPRR and may exceed the
LFE air flow rate repeatability 0.1 %
Differential pressure sensor range 0–2.49 kPa
mechanical constraint at higher engine loads. To control the MPRR at
Differential pressure sensor accuracy ± 0.3 % higher engine loads, the injection strategy was modified to decrease the
Differential pressure sensor repeatability ± 0.05 % premixed burn rate by (i) decreasing the pilot injection pulse width and
(ii) increasing the diffusion flame by increasing the main injection pulse
width.
Table 7 Fig. 3 shows the BTE of RCCI combustion mode at different loads as
Uncertainty ranges, in calculated performance and emission parameters. a function of main injection timing (SOI2). 40% was the maximum BTE
Parameters Unit Range of Values Range of Uncertainty ( ± ) in this engine in RCCI mode, which was obtained at 12 bar IMEP at
−15°ATDC main injection timing.
BSFC g/kW h 182–740 1.66–11.16 Previous studies showed that the brake thermal efficiency of the
BTE % 10–40 0.3–1.0
Brake NOx g/kW h 0.2–7.0 0.1–0.4
RCCI mode at low load is diesel-like or worse [25,44]. The BTE of
Brake HC g/kW h 8–760 0.8–66 diesel-only mode at low loads showed better performance in compar-
Brake CO g/kW h 4–140 0.3–13 ison with the RCCI mode, as shown in Table 9. The three operating
points in Table 9 showed higher BTE in comparison to the RCCI mode,
which makes them strong candidates for the optimization process.
(O2 %)ambient −(O2 %)intake The EGT represents the temperature of the exhaust gas during the
EGR% = × 100
(O2 %)ambient −(O2 %)exhaust (4) engine operation. This temperature has a strong correlation with the
efficiency of after-treatment systems including SCR and DOC. The EGT
Single diesel-only mode operation was conducted with single diesel of the engine is plotted in Fig. 4 with respect to the SOI2 in RCCI mode.
fuel injection from 10 to 15 °BTDC to fix the combustion phasing The EGT values were logged by the thermocouple that was installed
(CA50) at 8°ATDC. This injection strategy was used to maintain the after the turbocharger (see Fig. 1). Fig. 4 shows that the maximum EGT
CA50 at optimum crank angle degree for minimum BSFC (i.e., is 480 °C, which was obtained at 11 bar IMEP and 0°ATDC main

553
E. Ansari et al. Applied Energy 231 (2018) 549–561

was observed [44].

5. DOE analysis, modeling procedure, and validation

The collected data was studied through DOE analysis to find out the
effectiveness and functional relationship between inputs and selected
outputs (HC, NOx, CO, PM, BSFC, and EGT). Two sets of DOE analysis
were performed with different inputs to find out the most effective
single and combined input variables. The first set of independent input
variables included SOI2, BR, and λ and the second set included SOI2,
EGR and λ . The results of DOE analysis are demonstrated in Table 10.
This table shows the effectiveness of the independent control variables
in RCCI combustion mode. The calculated “Confidence” in Table 10
shows how effective an input variable is for controlling the specific
outputs listed in the first column.
Fig. 3. Effect of main injection timing (SOI2) on BTE in RCCI mode.
The outputs were then formulated as a function of the most effective
input variables. The catalyst conversion efficiency (ηcatalyst ) of SCR and
Table 9 DOC was also formulated as a function of EGT [43,50,51].
Comparison of the BTE of the CDC and RCCI modes at 1500 RPM with
CA50 = +8°ATDC. HC , NOx , CO, PM , BSFC , EGT = fi (InputVariables ), i = 1, 2, …, 6
IMEP (bar) BTE (%) - CDC BTE (%) - RCCI
ηcatalyst = gj (EGT ), j = 1, 2, 3
3 23 15
4 30 23 A regression analysis was performed to find out the quantitative
5 31 27 relation between inputs to predict the outputs. In this analysis around
120 experimental data points were used; 75% of the experimental data
points were used to train the regression models and 25% were used to
validate the models. Eqs. (5) to (13) show the results of the regression
analysis to predict HC, NOx, CO, PM, BSFC and EGT for RCCI operating
conditions including pilot injection timing (−55°ATDC < SOI1 <
−50°ATDC), main injection timing (−15°ATDC < SOI2 < 0°ATDC),
λ (1.1–3.6), BR (80–95%), EGR (0–30%) and MAP (88–150 kPa).
Since HC and CO emissions are found to be a function of injection
strategy, two functions are proposed for single (λ > 3) and double
(λ < 3) injection strategies.

⎧ 4.93(SOI )2 + 201.4(SOI ) Single Injection (a)

⎪ 2
⎪ −907.84(λ ) −17573.4(λ )
⎪ + 612(MAP )−803.9(BR)
⎪
⎪ + 299.22(BR. λ )
⎪
HC (g/kW h) = 13.5(SOI 1) + 0.065(SOI 2)2 Double Injection (b)
Fig. 4. Effect of main injection timing (SOI2) on exhaust gas temperature (EGT) ⎨
⎪ −1.6(SOI 2) + 29.62(λ )2
in RCCI mode. ⎪
⎪ + 581.31(λ ) + 0.11(EGR)2
⎪
injection timing. It was also observed that EGT has a decreasing trend ⎪ −4.74(EGR)−0.21(MAP )
with retarding the main injection timing. ⎪
⎩ + 9.87(BR)−7.3(BR. λ )
NOx, HC, CO and PM emissions in RCCI mode are presented in
(5)
Figs. 5–8. These emissions are measured on volume basis and for dry
condition. For analyzing the HC emission, due to the much higher re-
activity of diesel fuel compared to the natural gas, only MWCH 4 is
considered in the unburned HC analysis and it is assumed all unburned
HC comes from premixed fuel (CH4) [49]. Due to the premixed nature
of RCCI combustion mode, engine operation at high load conditions is
difficult due to high MPRR. To control the MPRR, the injection strategy
needs to be modified at higher loads (Fig. 2). This change in injection
strategy increased the NOx and PM emission significantly at conditions
with IMEP ≥ 9 bar. Since RCCI is a low-temperature combustion mode,
HC and CO emissions were relatively high at 6 bar IMEP and lower
engine load operating conditions, which they are performed with single
injection strategy. These points showed 200 g/kW h and higher HC
emission. But HC and CO emissions showed a decreasing trend at higher
engine loads (7 bar IMEP and above). These points are performed with
double injection strategy and it showed significantly lower HC and CO
emissions. Double injection strategy could help to burn all premixed
fuel in squish area; thus, a significant reduction in HC and CO emissions Fig. 5. Effect of main injection timing (SOI2) on NOx emission in RCCI mode.

554
E. Ansari et al. Applied Energy 231 (2018) 549–561

CO (g/kW h)

⎧ 0.041(SOI )2 + 10.79(SOI ) Single Injection (a)

⎪ 2
⎪ + 0.001(SOI . λ )−280.449(λ )
⎪ + 388.278(λ ) + 35.72(MAP )
⎪
⎪ −47.36(BR) + 0.036(BR. λ )
⎪
= − 0.55(SOI 1) + 0.096(SOI 2)2 Double Injection (b)
⎨
⎪ + 0.647(SOI 2)−0.029(SOI 2. λ )2
⎪
⎪ + 44.75(λ )2−77.7(λ )−0.36(EGR)
⎪
⎪ −0.036(MAP ) + 0.44(BR)
⎪
⎩ −0.002(BR. λ ) (6)

The NOx model was developed and validated using experimental

Fig. 6. Effect of main injection timing (SOI2) on HC emission in RCCI mode. data. The NOx model as a function of input variables is demonstrated in
Eq. (7).

NOx (g/kW h) = −0.015(SOI 1) + 0.002(SOI 2)2−0.2(SOI 2)−0.09(λ )2

+ 4.71(λ )−0.006(EGR)2 + 0.14(EGR) + 0.002(MAP )
−2.24(BR)−4.25(BR. λ ) (7)

Diesel-NG RCCI combustion mode has extremely low PM emission.

But to control the MPRR at higher loads, it is essential to decrease the
premixed combustion and increase the diffusion flame burn rate. This
modification in double injection strategy increases the PM emission up
to 0.11 g/kW h at 12 bar IMEP. To this end, two models are proposed.
One model is developed to predict the PM emission up to 9 bar IMEP
and the second model is developed to predict the PM emission between
9 and 12 bar IMEP.

⎧ − 0.003(SOI 1) + 0.007(SOI 2)2 IMEP < 9bar (a)

⎪ 2
⎪ + 0.11(SOI 2) + 0.22(λ )
Fig. 7. Effect of main injection timing (SOI2) on CO emission in RCCI mode.
⎪ −0.81(λ ) + 0.002(EGR)2
⎪
⎪ −0.047(EGR) + 0.028(MAP )
⎪
⎪ −0.63(BR)−0.0003(SOI 2. λ )2
PM (g/kW h) =
⎨ 2.12(SOI 1) + 0.1(SOI 2)2 9bar ⩽ IMEP ⩽ (b)
⎪
⎪ −0.625(SOI 2)−474.3(λ ) 2 12bar
⎪
⎪ + 1232.44(λ ) + 0.87(EGR)2
⎪
⎪ −32.27(EGR) + 0.155(MAP )
⎪ −474.22(BR)−0.12(SOI 2. λ )2
⎩
(8)
Brake Specific Fuel Consumption (BSFC) is also predicted as a
function of input parameters:

BSFC (g/kW h) = −9.7(SOI 1) + 0.108(SOI 2)2 + 2.43(SOI 2)−4.47(BR)

+ 0.19(EGR)2−6.06(EGR)−0.54(MAP )−34.35(λ )2
Fig. 8. Effect of main injection timing (SOI2) on PM emission in RCCI mode.
+ 172.27(λ ) (9)

Table 10 The conversion efficiency of the oxidation catalyst and SCR depends
Characterizing effectiveness of control variables on RCCI outputs. The on the exhaust gas temperature. This efficiency increases by increasing
“Confidence” quantifies the level of the effectiveness and is calculated based on the catalyst temperature until the catalyst operating temperature (e.g.,
2 k factorial design method [66]. 300 °C) is reached. The experimental EGT data were collected through
Outputs Most effective variable/s Confidence thermocouples mounted before and after the turbocharger in the ex-
haust system. An EGT model was developed using RCCI exhaust gas
NOx SOI2, λ , EGR, BR. λ 91%, 88%, 69%, 69% temperature data after the turbine. This model predicts the EGT as a
HC λ , BR, SOI2, BR. λ 87%, 81%, 81%, 61%
function of RCCI input parameters. This model will be useful to de-
CO BR, λ , BR. λ , SOI2. λ 99%, 99%, 99%, 81%
BSFC λ , SOI2, EGR 89%, 81%, 78%
termine the efficiency of the after-treatment system. The EGT model is
PM SOI2, λ , SOI2. λ 88%, 83%, 84% presented in Eq. (10):

EGT (°C) = −1.61(SOI 1) + 0.076(SOI 2)2 + 3.64(SOI 2) + 14.7(λ )2

−78.4(λ )−0.062(EGR)2 + 2.21(EGR) + 1.48(MAP )
+ 3.28(BR)−2.37(BR. λ ) (10)

555
E. Ansari et al. Applied Energy 231 (2018) 549–561

Fig. 9 shows the experimental versus predicted results. The accuracy

of the predictions can be judged by how close the predicted values are
next to the reference line.
Fig. 10 shows the conversion efficiency of Cu/zeolite SCR catalyst
and DOC with respect to the EGT [43,52]. Performance of the SCR is
strongly dependent on the space velocity. The conversion efficiency of
the catalyst slightly decreases (< 10%) by increasing the space velocity
[52–55].
Eqs. (11) and (12) show the conversion efficiency of DOC for HC
and CO emissions respectively as a function of exhaust gas temperature.
These equations are parametrized by changing the coefficients of
Weibull distribution function to fit the experimental data from Fig. 10
[43,52].

(
Conversion EfficiencyHC = 100∗ 1−0.96∗e−( l ) −2
T k
) (11)

Where T is the EGT, k and l are constant and equal to 15 and 380, Fig. 10. Conversion efficiency HC (from methane), CO and NOx versus EGT
respectively. [43,55].

( T k
Conversion EfficiencyCO = 100∗ 1−e−( l ) ) (12) the exhaust gas temperature (Fig. 10) is shown in Eqs. (13a) and (13b).
The conversion efficiency of the SCR with respect to the EGT is split into
Where, k and l are equal to 12 and 170, respectively. two equations to improve the prediction accuracy.
The conversion efficiency of SCR (for NOx emission) with respect to

Fig. 9. Comparison between experimental and predicted data in RCCI mode. e in plots shows the average prediction errors.

556
E. Ansari et al.
Conversion EfficiencyNOx
⎧− 0.04∗T 2
+ 14.6∗T −1250 T ⩽ 170°C (a)
⎪ 2 2 Where BSUC is the brake specific urea consumption of the SCR at
= ⎝
(c1 )
⎠ ⎝
(
⎪ a1 ∗exp ⎛− T − b1 ⎞ + a2 ∗exp ⎛− T − b2
c2 ) ⎞⎠ T > 170°C (b)
⎨
⎪ + a ∗exp ⎛− T − b3 2⎞
⎪
⎩
3
⎝
( )
c3
⎠ (13)
Constants in Eq. (13b) are listed in Table 11.
In SCR, the reduction mechanism initiates by injection of urea-water
into the exhaust flow. The injected urea goes through a complex che-
mical process to produce NH3. This step is called hydrolysis process of
the molten urea which leads to producing ammonia and isocyanic acid
[54–56]. The global reactions for the hydrolysis process are:
(NH2)2 CO(aqueous) → (NH2)CO(molten) + H2O
(NH2)2CO(molten) → NH3 + HNCO
HNCO + H2O → NH3 + CO2
There are three different global reactions that take place in SCR
between NH3, NO, and NO2. The first global reaction is called standard
SCR reaction [54,56]. The Ammonia to NOx ratio (ANR) in this reaction
is equal to 1. This reaction can also take place at higher ammonia to
NOx ratio (ANR = 1.2) which leads to better performance of an SCR
system. This Standard SCR reaction is shown below and it requires equal
moles of NO and NH3 to react:
4NH3 + 4NO + O2 → 4N2 + 6H2O
The second global reaction, which is called Fast Reaction, is a
combination of series reaction over a Cu/zeolite SCR catalyst. This re-
action has the highest reaction rate among all SCR reactions [54,56].
4NH3 + 2NO + 2NO2 → 4N2 + 6H2O
Finally, the third SCR reaction is called Slow SCR which has the
lowest reaction rate, and is shown below [54,56]:
7 produce less NOx in comparison to fixed CA50 = +8°ATDC. In addi-
4NH3 + 3NO2 → N2 + 6H2 O
2 (19)
In this study, fast SCR chemical reaction was used to back-calculate
the required urea flow rate to react with the engine out NOx in each
operating condition.
DPF is used to trap the particulate to prevent particles to release in
atmosphere. DPFs usually have above 90–95% trapping efficiency
[57,58]. In this study, the raw values that are collected from AVL 415
are reduced by 95% to estimate the PM emission at tailpipe (after DPF).
6. Optimization of engine operation
An offline optimization framework was developed to minimize the
operational cost (i.e., fuel and urea) of the dual-mode RCCI-CDC en-
gine. Given there are several control variables used for the engine, a
large number of data points were required to find global optimum
operating points. The initial data points were generated through ex-
periments that covered only a limited operating space. To increase the
number of data points in this optimization, the predictive models from
the previous section were used to generate a rich data set that covered a
wide range of possible search space. To this end, new data points in
each load were predicted by the models. These data points were gen-
erated for various injection strategies, MAP, and EGR in each load. A
normalized cost function (J ) was considered and minimized in this
study to find out the optimum input parameters for 3 to 12 bar IMEP
engine operation at 1500 RPM. This optimization was conducted using
the constraint that the selected data points needed to meet the Tier 3
Bin 20 emission regulations. Tier 3 EPA emission standard is provided
in Table 12.
Applied Energy 231 (2018) 549–561
BSFC (g/kW h) BSUC (g/s)
J = W1 + W2
BSFCmax (g/kW h) BSUCmax (g/s) (20)
each operating condition. W1 and W2 are the weight factors in the cost
function. W1 and W2 are selected between 0.1 and 1 by considering the
cost of the fuel and urea. Since the urea consumption of the SCR system
is much smaller than the fuel consumption (below 2%), 0.2 is selected
for W2 and 1 is selected for W1 [59–61]. BSFCmax and BSUCmax are the
maximum fuel and urea consumption of the engine and SCR system,
determined for engine operation at each load condition.
7. Optimization results
An optimization program was developed that incorporated RCCI
(14)
relations (i.e., Eqs. (5)-(13)) in the cost function in Eq. (20). The opti-
(15) mization results showed that at low loads, single fuel diesel only mode
had better fuel economy in comparison to RCCI mode (from 3 to 6 bar
(16)
IMEP). However, at higher loads (between 7 and 12 bar IMEP) RCCI
showed better fuel economy, lower urea consumption, and overall
lower Normalized Cost Function (NCF). However, RCCI had higher HC
and CO emissions at higher loads in comparison to the CDC mode. The
most challenging part of the optimization was to minimize the tailpipe
emissions at 7, 8 and 9 bar IMEP due to the need for high enough EGT
between 375 and 435 °C. As it was described before, the conversion
efficiency of the after-treatment system is EGT dependent. The con-
version efficiency of HC is close to 100% at EGT ≥ 430 °C, but SCR
(17) efficiency showed decreasing trend at EGT ≥ 380 °C (Fig. 10). The so-
lution to decrease the HC emission was to optimize SOI1 and SOI2 for
different engine operating conditions. This modification is illustrated in
Fig. 2.
At 10, 11 and 12 bar IMEP, due to the high EGT (above 450 °C), the
(18) SCR has low conversion efficiency (see Fig. 11). To control the NOx
emission to meet the emission regulation, the combustion phasing of
the engine (CA50) in RCCI mode needed to retard at these loads. The
optimization results showed that properly adjusting CA50 helps to
tion, adjusting CA50 can be utilized to increase the EGT for better
conversion of unburned HC from methane to CO2 and H2O. The sum-
mary of the independent input variables at optimum conditions, pre-
dicted by models, is provided in Table 13.
The model-based optimization results for RCCI and experimental
results of CDC mode are shown in Table 14. These results are presented
by considering EGT after the turbine and corresponding after-treatment
efficiency. Table 15 shows the predicted EGT at optimum conditions
and the cost of operation ($/kW h) for RCCI and CDC modes at each
engine load. The cost is split based on fuel and urea consumption for
each case. The price of diesel fuel in Table 15 is taken from Alternative
Fuel Price Report published by U.S. Department of Energy in October
2016, and the price of NG is taken from the data by U.S. Energy In-
"$"
formation Administration [60,62]. The average of 8.2 3 from last
10 SCF
five years for NG fuels was selected in this optimization. The final op-
erating cost for RCCI mode was calculated based on diesel and NG fuel
combined with urea costs. EmissionLim is the specific emission based on
US Tier 3 Bin 20 emission regulation [63]. EmissionA.T is the predicted
emission by considering engine-out emissions and after-treatment
system.
To validate the model-based optimization in this work, new sets of
Table 11
Constants of SCR conversion efficiency model.
Index a b c
1 24.77 9.77 97.54
2 212.50 200.20 378.20
3 91.43 37.17 259.70
557
E. Ansari et al. Applied Energy 231 (2018) 549–561

Table 12 Table 15
Tier 3 certification Bins in EPA emission regulation [63] EGT and the cost of operation in each optimized condition.
Bin NOx PM* CO HCHO IMEP Mode CA50 EGT after turbine Cost of operation
mg/mi mg/mi g/mi mg/mi
RCCI CDC
Bin 160 160 3 4.2 4
Bin 125 125 3 2.1 4 NG + Diesel Urea Diesel Urea
Bin 70 70 3 1.7 4
Bin 50 50 3 1.7 4 (Bar) (–) °ATDC (°C) ($/kW h)
Bin 30 30 3 1.0 4 3 CDC 8 148 0.513 0.002 0.329 0.009
Bin 20 20 3 1.0 4 4 CDC 8 177 0.298 0.001 0.226 0.006
Bin 0 0 0 0 0 5 CDC 8 225 0.177 0.003 0.199 0.005
6 CDC 12 315 0.152 0.002 0.186 0.005
7 RCCI 8 389 0.122 0.002 0.177 0.005
8 RCCI 9 405 0.122 0.002 0.168 0.004
9 RCCI 11 435 0.108 0.002 0.161 0.004
10 RCCI 14 455 0.112 0.002 0.161 0.004
11 RCCI 15 470 0.110 0.002 0.167 0.004
12 RCCI 15 510 0.112 0.001 0.179 0.005

PM emissions: The results of experiments in this study (Table 14)

showed that PM emission at 7, 8, and 9 bar IMEP in RCCI mode is much
lower than the emission regulation even without Diesel Particulate
Filter (DPF). But in order to control the MPRR at higher loads, the
premixed burn rate was decreased by decreasing the pilot injection and
increasing the main injection amount. This modification in injection
strategy at higher loads causes an increase in the PM emission above the
PM limit in the emission regulation. Overall, the results show that it is
possible to eliminate the DPF in the RCCI engine for IMEP ≤ 9 bar, but
DPF is required for IMEP > 9 bar. The results in single fuel diesel mode
also showed the DPF is required to meet the emission regulations.
NOx emissions: In single fuel diesel-only mode, at 3 bar IMEP, due
Fig. 11. SCR and DOC conversion efficiency as a function of EGT. to low EGT (below 200 °C), SCR showed low conversion efficiency and
it did not mitigate the NOx emission. Even at 6 bar IMEP in CDC mode
Table 13 (CA50 = 8°ATDC), although EGT was high enough (around 300 °C) and
RCCI input variables predicted by the model for optimum conditions. the SCR had high conversion efficiency (around 98%), it did not meet
the regulation due to the high NOx concentration at this condition. The
IMEP SOI1 SOI2 MAP EGR
two potential solutions to reduce the NOx emission are i) running the
bar °TDC °TDC kPa % urea-SCR with higher ANR, which leads to higher urea consumption
7 −55 −8 90 18 and higher overall cost of operation; ii) retarding the combustion
8 −55 −6 115 18 phasing from 8 to 12°ATDC (The reported NOx at 6 bar IMEP in
9 −55 −5 120 21
10 −55 −4 130 23
Table 14 is at CA50 = 12°TDC).
11 −55 −3 150 23 In diesel-NG RCCI mode, RCCI combustion mode still needs SCR to
12 −55 −3 150 27 meet the Tier 3 Bin 20 emission regulation. At higher loads
(IMEP > 9 bar), NOx emission increases due to the modified injection
strategy (Fig. 2-B) to reduce the MPRR.
RCCI experiments were performed. Fig. 12 shows the results of new CO and HC emissions: one drawback of the RCCI engines, is high HC
experimental data and the values predicted by empirical models. After and CO emissions due to the low temperature of combustion. However,
the optimization, the tailpipe emissions need to be compared with Tier high conversion efficiency of oxidation catalyst reduces the CO emis-
3 Bin 20 light-duty emission regulation and specify which after-treat- sion to meet the emission regulation, while the required high light-off
ment system is required at optimum operating conditions in each en- temperature of the HC oxidation catalyst poses a significant challenge
gine load. Below is a summary of the findings. to meet the emission regulation [43,50,52,64]. Unburned HC emission

Table 14
Predicted and optimized RCCI and experimental CDC operating points at different engine loads. The subscripts “A.T” and “Lim” represent the tailpipe emission and
the EPA emission regulation limit, respectively.
IMEP Mode Normalized cost function HC HCA.T HCLim NOx NOxA.T NOxLim CO COA.T COLim PM PMA.T PMLim

(Bar) (–) (–) (g/kW h)

3 CDC 1.20 8.30 8.13 0.52 6.80 3.94 0.06 23.3 6.80 8.10 0.025 0.001 0.045
4 CDC 1.20 3.20 1.10 0.40 5.50 1.10 0.05 23.0 0.50 4.10 0.031 0.001 0.050
5 CDC 1.20 2.10 0.21 0.40 4.90 0.04 0.04 5.90 0.30 3.30 0.033 0.001 0.071
6 CDC 1.20 1.40 0.10 0.30 4.10 0.04 0.04 3.60 0.07 3.30 0.051 0.002 0.101
7 RCCI 1.08 2.10 0.30 0.30 1.50 0.02 0.03 5.10 0.09 3.30 0.003 0.0001 0.105
8 RCCI 1.05 3.30 0.30 0.30 0.50 0.01 0.03 5.20 0.09 3.10 0.003 0.0001 0.115
9 RCCI 0.92 2.20 0.29 0.30 0.30 0.01 0.03 5.40 0.05 2.60 0.003 0.0001 0.120
10 RCCI 0.86 5.00 0.19 0.30 0.30 0.02 0.03 3.50 0.03 2.60 0.016 0.0008 0.145
11 RCCI 0.57 9.30 0.10 0.30 0.22 0.01 0.03 4.60 0.04 2.60 0.038 0.001 0.145
12 RCCI 0.67 6.20 0.10 0.30 0.07 0.01 0.03 22.1 0.24 2.60 0.106 0.005 0.145

558
E. Ansari et al.
Fig. 12. Experimental vs. predicted PM, HC, NOx and CO tailpipe emissions, along with U.S. Tier 3 Bin 20 emission limits shown by black horizontal lines.
generally showed decreasing trend by increasing the load. However, to
control the HC emission at medium loads (7 to 9 bar IMEP), the BR
slightly decreased to 70% and the injection strategy was modified to
increase the premixed burn rate (Fig. 2-A). As shown in Table 14 in
single fuel diesel mode, HC emission at 3 and 4 bar IMEP did not meet
the emission regulation during cold start. A potential solution to meet
the emission regulations during cold start is to use Electrically Heated
Catalyst (EHC). This method will increase the EGT over 300 °C within
20 s [65], but it will add extra operational cost for the exhaust after-
treatment system.
8. Summary and conclusion
RCCI operation at a wide range of engine loads (3 to 12 bar IMEP)
was studied and optimized for a light-duty multi-cylinder diesel-NG
engine. This optimization study also included single fuel diesel only
mode due to better performance of the engine at low loads in CDC
mode. Energy based BR was changed from 80 to 95% in the RCCI mode.
Single injection and double injection strategies were used in RCCI
mode. A DOE analysis (23 factorial design) was performed to find out
the most effective input variables in controlling different outputs.
Around 120 experimental data points were used to perform the re-
gression analysis. This analysis helped to create the empirical correla-
tions to predict major RCCI outputs as a function of controllable inputs.
The empirical correlations are valid for the operating conditions in
Table 8. The resulting empirical models could predict BSFC, HC, CO,
and NOx, with average errors of 1.5, 0.6, 0.1 and 0.1 g/kW h, respec-
tively; EGT with an average error of 2.6 °C and PM with an average
error of 0.001 g/kW h.
The cost of operation was calculated and compared at each engine
load. A normalized cost function was defined in terms of BSFC and
BSUC. The minimum operational cost was found through model-based
optimization at each engine load. The optimization included constraints
to meet Tier 3 Bin 20 light-duty emission standard. This also helped to
determine the after-treatment system requirements. The major findings
from this study are:
559
Applied Energy 231 (2018) 549–561
• Single fuel diesel-only mode was the optimum combustion mode at
low loads (3–6 bar IMEP) conditions due to the better engine per-
formance and lower HC and CO emission, compared to the RCCI
mode. At 3 and 4 bar IMEP due to low EGT, reduction of HC and CO
emissions is the major challenge for the exhaust after-treatment
system.
• The cost (urea + fuel) of running the engine in RCCI mode at higher
engine loads (IMEP > 6 bar) is around 30–35% lower than that in
the CDC mode. Urea cost is around 0.5% to 3% of the total cost in
optimum operating conditions in this study.
• Given the engine needs to be calibrated for different engine speeds,
around 50% reduction in required engine experiments is anticipated
via the proposed model-based optimization in this study. The pro-
posed model based optimization reduced around 100 experiments in
this study. Overall, 120 engine experiments were conducted to op-
timize the dual mode diesel-NG engine at a fixed engine speed in
this study.
• The optimum CA50 retards as engine load increases. Retarding the
CA50 not only reduces the engine-out NOx emission but also in-
creases the EGT for better performance of the after-treatment
system.
• The optimum EGT range for dual mode RCCI-CDC operation is
around 400 to 425 °C. In this EGT range, both SCR and DOC operate
with high conversion efficiency to meet the EPA emission regula-
tion. The SCR conversion efficiency decreases below 90% at
EGT > 425 °C.
• At medium loads (7–9 bar IMEP), RCCI showed better fuel economy,
lower NOx and PM emissions but higher HC and CO emissions.
Within 7 and 9 bar IMEP, the PM emission was significantly low and
it could meet the emission regulation without using DPF. In this
IMEP range, due to the low EGT, the conversion efficiency of HC
oxidation catalyst was around 80–95% and it could meet the emis-
sion regulation right at the limit of the Tier 3 Bin 20 emission reg-
ulation.
• To control HC at medium loads (7 to 9 bar IMEP), injection strategy
was modified by increasing the pilot injection duration and
E. Ansari et al.
decreasing the main injection duration. This modification of injec-
tion strategy increases the reactivity of mixture in squish region and
crevices. This helps to consume all the premixed fuel and reducing
the unburned HC emission. This method is not effective at higher
loads since it increases the MPRR beyond the mechanical constraint.
• Modification of the injection strategy at higher loads (10 to 12 bar
IMEP) is essential to control the MPRR. Injection strategy was
modified by decreasing the pilot fuel and increasing the main in-
jection. This modification of injection strategy increased the PM
emission up to 0.11 g/kW h at 12 bar IMEP. Thus, DPF is required
for the operating condition with IMEP ≥ 10 bar to meet the emis-
sion regulation.
•
A significant reduction was observed in NOx emission in the RCCI
mode in comparison to the diesel mode. NOx emission can meet the
emission regulation with after-treatment for the entire engine loads
in this study. Although the EGT is relatively low, CO emission can
meet the emission regulation in all loads with the after-treatment
system.
Due to the lower cost of operation and higher thermal efficiency of
the dual fuel diesel/NG RCCI engine in compare to the single fuel CDC
mode, this type of engines not only are useful in light duty on-road
applications but also they can be useful in stationary applications such
as power generator. This study showed lower fuel and urea consump-
tion in a wide range of engine loads by utilizing a dual mode diesel-NG
RCCI engine. However, further research is needed on optimization of
the combustion chamber and also the development of oxidation catalyst
with operating temperature as low as 250 °C to improve RCCI at low
loads with low EGT and high HC emissions.
Acknowledgment
The authors thank Paul Dice from Michigan Technological
University for his technical support to calibrate the engine experimental
setup.
References
[1] Roberts TP, Phillips FC, Hunt AE, Jarboe WS. Report of a Committee of the
Engineers’ society of western pennsylvania, the utilization of natural gas. J Franklin
Inst 1885;119(1):48–66.
[2] Wannatong K, Akarapanyavit N, Siengsanorh S, Chanchaona S. Combustion and
knock characteristics of natural gas diesel dual fuel engine. SAE technical paper
2007-01-2047; 2007. http://doi.org/10.4271/2007-01-2047.
[3] Dahodwala M, Joshi S, Koehler E, Franke M, Tomazic D. Experimental and com-
putational analysis of diesel-natural gas rcci combustion in heavy-duty engines. SAE
Technical Paper 2015-01-0849; 2015. http://doi.org/10.4271/2015-01-0849.
[4] Dahodwala M, Joshi S, Koehler E, Franke M. Investigation of diesel and CNG
combustion in a dual fuel regime and as an enabler to achieve RCCI combustion.
SAE technical paper 2014-01-1308; 2014. http://doi.org/10.4271/2014-01-1308.
[5] Dempsey A, Walker N, Reitz R. Effect of piston bowl geometry on dual fuel re-
activity controlled compression ignition (RCCI) in a light-duty engine operated with
gasoline/diesel and methanol/diesel. SAE Int J Engines 2013;6(1):78–100. https://
doi.org/10.4271/2013-01-0264.
[6] Doosje E, Willems F, Baert R. Experimental demonstration of RCCI in heavy-duty
engines using diesel and natural gas. SAE technical paper 2014-01-1318; 2014.
http://doi.org/10.4271/2014-01-1318.
[7] Jia Z, Denbratt I. Experimental investigation of natural gas-diesel dual-fuel RCCI in
a heavy-duty engine. SAE Int J Engines 2015;8(2):797–807. https://doi.org/10.
4271/2015-01-0838.
[8] Kakaee A-H, Rahnama P, Paykani A. Influence of fuel composition on combustion
and emissions characteristics of natural gas/diesel RCCI engine. J Nat Gas Sci Eng
2015;25:58–65.
[9] Nieman D, Dempsey A, Reitz R. Heavy-duty RCCI operation using natural gas and
diesel. SAE Int J Engines 2012;5(2):270–85. https://doi.org/10.4271/2012-01-
0379.
[10] Salahi MM, Esfahaninan V, Gahrehghani A, Mirsalimi M. Investigating the re-
activity controlled compression ignition (RCCI) combustion strategy in a natural
gas/diesel fueled engine with a pre-chamber. Energy Convers Manage
2017;132:40–53.
[11] Karavalakis G, Gysel N, Hajbabaei M, Durbin T, Johnson K, Miller W. Influence of
different natural gas compositions on the regulated emissions, aldehydes, and
particle emissions from a transit bus. SAE technical paper 2013-01-1137; 2013.
http://doi.org/10.4271/2013-01-1137.
560
Applied Energy 231 (2018) 549–561
[12] Johnson DR, Heltzel R, Nix AC, Clark N, Darzi M. Regulated gaseous emissions from
in-use high horsepower drilling and hydraulic fracturing engines. J Pollut Eff Cont
2017;5:187. https://doi.org/10.4176/2375-4397.1000187.
[13] Javaheri A, Esfahanian V, Salavati-Zadeh A, Darzi M, Mirsoheil SM. Investigation of
natural gas composition effects on knock phenomenon in si gas engines using de-
tailed chemistry. Appl Mech Mater 2014;493:239–44. https://doi.org/10.4028/
www.scientific.net/AMM.493.239.
[14] Javheri A, Esfahanian V, Salavati-Zade A, Darzi M. Energetic and exergetic analyses
of a variable compression ratio spark ignition gas engine. Energy Convers Manage
2014;88:739–48. https://doi.org/10.1016/j.enconman.2014.09.009.
[15] Afkhami B, Kakaee AH, Pouyan K. Studying engine cold start characteristics at low
temperatures for CNG and HCNG by investigating low-temperature oxidation.
Energy Convers Manage 2012;64:122–8.
[16] Poorghasemi K, Khoshbakhti Saray R, Ansari E, Khoshbakht Irdmousa B,
Shahbakhti M, Naber J. Effect of diesel injection strategies on natural gas/diesel
RCCI combustion characteristics in a light duty diesel engine. Appl Energy. ISSN
0306-2619.
[17] Rahimi Boldaji M, Sofianopoulos A, Mamalis S, Lawler B. CFD Simulations of the
Effect of Water Injection Characteristics on TSCI: A New, Load-flexible, advanced
combustion concept. ASME Internal combustion engine division fall technical
conference. http://doi.org/10.1115/ICEF2017-3662.
[18] Afkhami B, Wang Y, Miers, S, Naber, J. Experimental study of flame stretch under
engine-like conditions. In: ASME 2017 Internal Combustion Engine Division Fall
Technical Conference, p. V001T03A015-V001T03A015.
[19] Shahsavan Martia, Hunter Mack J. Numerical study of a boosted HCCI engine fueled
with n-butanol and isobutanol. Energy Convers Manage 2018;157:28–40.
[20] Sevik J, Pamminger M, Wallner T, Scarcelli R, Reese R, Iqbal A, et al. Performance,
efficiency and emissions assessment of natural gas direct injection compared to
gasoline and natural gas port-fuel injection in an automotive engine. SAE Int J
Engines 2016;9(2):1130–42. https://doi.org/10.4271/2016-01-0806.
[21] Pamminger M, Sevik J, Scarcelli R, Wallner T, Wooldridge S, Boyer B, et al.
Evaluation of knock behavior for natural gas - gasoline blends in a light duty spark
ignited engine. SAE Int J Engines 2016;9(4):2153–65. https://doi.org/10.4271/
2016-01-2293.
[22] Genchi G, Pipitone E, Beccari S, Piacentino A. Knock resistance increase through the
addition of natural gas or LPG to gasoline: an experimental study. SAE technical
paper 2013-24-0100; 2013. http://doi.ogr/10.4271/2013-24-0100.
[23] Curran S, Hanson R, Wagner R. Effect of E85 on RCCI performance and emissions on
a multi-cylinder light-duty diesel engine. SAE technical paper 2012-01-0376, 2012.
http://doi.org/10.4271/2012-01-0376.
[24] Benajes J, Molina S, Garcia A. Effects of direct injection timing and blending ratio
on RCCI combustion with different low reactivity fuels. Energy Convers Manage
2015;99:193–209.
[25] Curran S, Hanson R, Wagner R, Reitz R. Efficiency and emissions mapping of RCCI
in a light-duty diesel engine. SAE technical paper 2013-01-0289; 2013. http://doi.
org/10.4271/2013-01-0289.
[26] Dempsey A, Curran S, Storey J, Eibl M, Pihl J, Prikhodko V, Wagner R, Parks J.
Particulate matter characterization of reactivity controlled compression ignition
(RCCI) on a light duty engine. SAE technical paper 2014-01-1596; 2014. http://doi.
org/10.4271/2014-01-1596.
[27] Hanson R, Curran S, Wagner R, Kokjohn S, Spiltter D, Reitz R. Piston bowl opti-
mization for RCCI combustion in a light-duty multi-cylinder engine. SAE Int J
Engines 2012;5(2):286–99. https://doi.org/10.4271/2012-01-0380.
[28] Hanson R, Curran S, Wagner R, Reitz R. Effects of biofuel blends on RCCI com-
bustion in a light-duty, multi-cylinder diesel engine. SAE Int J Engines
2013;6(1):488–503. https://doi.org/10.4271/2013-01-1653.
[29] Hanson R, Ickes A, Wallner T. Comparison of RCCI operation with and without EGR
over the full operating map of a heavy-duty diesel engine. SAE technical paper
2016-01-0794; 2016. http://doi.org/10.4271/2016-01-0794.
[30] Kokjohn S, Hanson R, Splitter D, Kaddatz J, Reitz R. Fuel reactivity controlled
compression ignition (RCCI) combustion in light- and heavy-duty engines. SAE Int J
Engines 2011;4(1):360–74. https://doi.org/10.4271/2011-01-0357.
[31] Li J, Yang W, Goh T, An H, Maghbouli A. Study on RCCI (reactivity controlled
compression ignition) engine by means of statistical experimental design. Energy
2014;78:777–87.
[32] Li Y, Jia M, Chang Y, Liu Y, Xie M, Wang T, et al. Parametric study and optimization
of a RCCI (reactivity controlled compression ignition) engine fueled with methanol
and diesel. Energy 2014;65:319–32.
[33] Lim J, Walker N, Kokjohn S, Reitz R. High speed dual-fuel RCCI combustion for high
power output. SAE technical paper 2014-01-1320; 2014. http://doi.org/10.4271/
2014-01-1320.
[34] Nazemi M. Modeling and analysis of reactivity controlled compression ignition
(rcci) combustion. MSc thesis 2015, Michigan Technological University.
[35] Pohlkamp K, Reitz R. Reactivity controlled compression ignition (RCCI) in a single-
cylinder air-cooled HSDI diesel engine. SAE technical paper 2012-32-0074; 2012.
http://doi.org/10.4271/2012-32-0074.
[36] Qian Y, Wang X, Zhu L, Lu X. Experimental studies on combustion and emissions of
RCCI (reactivity controlled compression ignition) with gasoline/n-heptane and
ethanol/n-heptane as fuels. Energy 2015;88:584–94.
[37] Splitter D, Hanson R, Kokjohn S, Wissink M, Reitz R. Injection effects in low load
RCCI dual-fuel combustion. SAE technical paper 2011-24-0047; 2011. http://doi.
org/10.4271/2011-24-0047.
[38] Lim O, Iida N, Cho G, Narankhuu J. The research about engine optimization and
emission characteristic of dual fuel engine fueled with natural gas and diesel. SAE
technical paper 2012-32-0008; 2012. http://doi.org/10.4271/2012-32-0008.
[39] Rahimi A, Fatehifar E, Saray RK. Development of an optimized chemical kinetic
E. Ansari et al.
mechanism for homogeneous charge compression ignition combustion of a fuel
blend of n-heptane and natural gas using a genetic algorithm. J Automob Eng
2010;224(9):1141–59.
[40] Benajes J, Garcia A, Monsalve J, Boronat V. Achieving clean and efficient engine
operation up to full load by combining optimized RCCI and dual-fuel diesel-gasoline
combustion strategies. Energy Convers Manage 2017;136:142–51.
[41] Nazemi M, Shahbakhti M. Modeling and analysis of fuel injection parameters for
combustion and performance of an RCCI engine. Appl Energy 2016;165:135–50.
[42] Kavuri C, Kokjohn S. Investigating air handling requirements of high load low speed
reactivity controlled compression ignition (RCCI) combustion. SAE technical paper
2016-01-0782; 2016. http://doi.org/10.4271/2016-01-0782.
[43] Williams S, Hu L, Nakazono T, Ohtsubo H, Uchida M. Oxidation catalysts for natural
gas engine operating under HCCI or SI conditions. SAE Int J Fuels Lubr
2009;1(1):326–37. https://doi.org/10.4271/2008-01-0807.
[44] Ansari E, Poorghasemi K, Irdmousa B, Shahbakhti M, Naber J. Efficiency and
emissions mapping of a light duty diesel - natural gas engine operating in con-
ventional diesel and RCCI modes. SAE technical paper 2016-01-2309; 2016. http://
doi.org/10.4271/20160.-2309.
[45] Hanson R, Ickes A, Wallner T. Use of adaptive injection strategies to increase the full
load limit of RCCI operation. J Eng Gas Turbines Power
2016;138(10):02802–102802.
[46] Bates S. Flame imaging studies of cycle-by-cycle combustion variation in a SI four-
stroke engine. SAE technical paper 892086; 1989. http://doi.org/10.4271/892086.
[47] Kuo T. Multidimensional port-and-cylinder gas flow, fuel spray, and combustion
calculations for a port-fuel-injection engine. SAE technical paper 920515; 1992.
http://doi.org/10.4271/920515.
[48] Kokjohn S, Hanson R, Splitter D, Reitz R. Experiments and modeling of dual-fuel
HCCI and PCCI combustion using in-cylinder fuel blending. SAE Int J Engines
2010;2(2):24–39. https://doi.org/10.4271/2009-01-2647.
[49] Pilusa TJ, Mollagee MM, Muzenda E. Reduction of vehicle exhaust emissions from
diesel engines using the whale concept filter. Aerosol Air Qual Res 12: 994–1006.
http://doi.org/10.4209/aaqr.2012.04.0100.
[50] Tang T, Zhang J, Shuai S, Cao D. Urea decomposition at low temperature in SCR
systems for diesel engines. SAE technical paper 2014-01-2808; 2014. http://doi.
org/10.4271/2014-01-2808.
[51] Tadrous T, Brown K, Towgood P, McConnell C. Development of passive/active DPF
system utilizing syngas regeneration strategy - retrofit, real life optimization and
performance experience. SAE technical paper 2010-01-0560; 2010. http://doi.org/
10.4271/2010-01-0560.
[52] Guo G, Warner J, Cavataio G, Dobson D. et al. The development of advanced urea-
561
Applied Energy 231 (2018) 549–561
SCR systems for tier 2 bin 5 and beyond diesel vehicles. SAE technical paper 2010-
01-1183; 2010. http://doi.org/10.4271/2010-01-1183.
[53] Shamim T. The effect of space velocity on the dynamic characteristics of an auto-
motive catalytic converter. SAE technical paper 2005-01-2160; 2005. http://doi.
org/10.4271/2005-01-2160.
[54] Song X. A SCR model based on reactor and engine experimental studies for a Cu-
zeolite catalyst” PhD thesis; 2013. Michigan Technological University.
[55] Prikhodko VY, Pihl J, Lewis S, Parks J. Effect of hydrocarbon emissions from PCCI-
type combustion on the performance of selective catalytic reduction catalysts. J Eng
Gas Turbines Power 2012;8. p. 134082804-082804-5.
[56] Song X, Parker G, Johnson J, Naber J, Pihl J. A modeling study of SCR reaction
kinetics from reactor experiments. SAE technical paper 2013-01-1576; 2013.
http://doi.org/10.4271/2013-01-1576.
[57] Wade WR, et al. Diesel particulate trap regeneration techniques. SAE Technical
Paper 810118; 1981. http://doi.org/10.4271/810118.
[58] Single wall diesel particulate filter (DPF) filtration efficiency studies using labora-
tory generated particles, Juan Yang, Mark Stewart, Gary Maupin, Darrell Herling,
Alla Zelenyuk. Pacific Northwest National laboratory, Richland, WA 99354, USA.
[59] Nishioka A, Sukegawa Y, Katogi K, Mamada H, Kowatari T, Mukai T, Babcock,
Yokota H. A study of a new aftertreatment system (2): Control of urea solution spray
for Urea-SCR. SAE technical paper 2006-01-0644; 2006. http://doi.org/10.4271/
2006-01-0644.
[60] Alternative Fuel Report Price. 2016; Available from: http://www.afdc.energy.gov/
publications/search/keyword/?q=alternative%20fuel%20price%20report.
[Accessed on Apr.10, 2017].
[61] Acharya R, Alam M, Boehman A. Fuel and system interaction effects on urea-SCR
control of NOx in diesel exhaust after treatment. SAE technical paper 2006-01-
0638; 2006. http://doi.org/10.4271/2006-01-0638.
[62] U.S. Energy Information Administration. 2016; Available from: https://www.eia.
gov/naturalgas/. [Accessed on Apr.10, 2017].
[63] Cars and Light-Duty Trucks - Tier 3. 2016; Available from: https://www.dieselnet.
com/standards/us/ld_t3.php. [Accessed on Apr.10,2017].
[64] Joo K, Jo J, Kim C, Lee J, Kim H. The study of NOx reduction using urea-SCR system
with CPF and DOC for light duty vehicle; the diesel NOx reduction system. SAE
technical paper 2008-01-1183; 2008. http://doi.org/10.4271/2008-01-1183.
[65] Kim C, Paratore M, Gonze E, Solbrig C, Smith, S. Electrically heated catalysts for
cold-start emissions in diesel after treatment. SAE technical paper 2012-01-1092;
2012. http://doi.org/10.4271/2012-01-1092.
[66] Montgomery DC. Design and analysis of experiments. 1979-2014. ISBN 978-1118-
14692-7, Arizona State University.