Fuel 289 (2021) 119835

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Full Length Article

Ducted Fuel Injection: Experimental and numerical investigation on fuel

spray characteristics, air/fuel mixing and soot mitigation potential
F. Millo a, *, A. Piano a, B. Peiretti Paradisi a, L. Postrioti b, L. Pieracci b, A. Bianco c, F.C. Pesce d,
A. Vassallo d
a
Energy Department, Politecnico di Torino, Torino 10129, Italy
b
Università degli Studi di Perugia, Perugia 06125, Italy
c
POWERTECH Engineering, Torino 10127, Italy
d
PUNCH Torino S.p.A. (former GM), Torino 1029, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: Enhancing mixture preparation upstream of the premixed autoignition zone is a solution to reduce soot emission
Diesel formation in compression ignition engines. With this aim, in recent years, Ducted Fuel Injection (DFI) concept
Ducted Fuel Injection has been developed: DFI is based on the idea of injecting the fuel spray through a small cylindrical pipe within
Fuel spray
the combustion chamber at a certain distance from the nozzle injector hole. Recent research studies have
Mixing combustion
highlighted the high potential of this innovative concept for soot mitigation in both constant volume vessel and
engine-like operating conditions. However, the mechanisms driving the soot reduction have not yet been fully
understood. The aim of this research work is to further investigate the DFI concept, evaluating its impact on the
spray characteristics, on the air/fuel mixing and, therefore, on the soot formation phenomena. Firstly, an
experimental activity was carried out by means of a constant volume vessel test bench with optical accesses to
compare the spray evolution and sizing with and without duct adoption, over a wide range of vessel thermo­
dynamic conditions and injection pressures. After that, a simulation setup was defined in the commercially
available 3D-CFD software CONVERGE reproducing the experimental test bench, calibrating and validating the
spray model. Firstly, the spray model was calibrated and validated considering the same non-reacting conditions
exploited in the experimental analysis. Then, the calibrated spray model was used as a virtual tool to investigate
the air entrainment process. As a results, the DFI adoption increases the air entrainment in the near-nozzle region
caused by the high velocity spray that generates a pumping effect at the duct inlet. Moreover, mixing process is
also enhanced by the turbulence distribution at the duct exit resulting in a narrower distribution of equivalence
ratio at the ignition. As a consequence of the more effective air entrainment and improved turbulent mixing, DFI
remarkably mitigates soot emissions, with a reduction up to 80% with respect to free spray configuration in all
tested operating conditions.

lower their requirements thus reducing their cost and complexity. The
main objective of these advanced combustion strategies has been to
1. Introduction lower both NOx and PM emissions, and significant results have been
achieved for example by means of so called Low Temperature Com­
One of the crucial issues for a future generation of diesel engines is bustion (LTC) regimes [1,2]. In this context, an innovative combustion
the development of strategies for the abatement of pollutant emissions, strategy was developed by researchers from Sandia National Labora­
especially Nitrogen Oxides (NOx) and Particulate Matter (PM). Pursuing tories, the Leaner Lifted-Flame Combustion (LLFC) [3]. LLFC is based on
this aim, a combination of advanced combustion strategies and complex the fact that if the equivalence ratios distribution remains lower than or
aftertreatment systems are required. In particular, as far as in-cylinder equal to approximately two at the Lift-Off Length (LOL), the formation
approach to pollutant emissions control is concerned, innovative com­ of soot particles can be prevented. In addition, thanks to the high-
bustion strategies have been explored to meet the emission standards tolerance to Exhaust Gas Recirculation (EGR), LLFC can
avoiding the need of complex aftertreatment systems or, at least, to

* Corresponding author.
E-mail address: [email protected] (F. Millo).

https://doi.org/10.1016/j.fuel.2020.119835
Received 26 July 2020; Received in revised form 31 October 2020; Accepted 20 November 2020
Available online 18 December 2020
0016-2361/© 2020 Elsevier Ltd. All rights reserved.
F. Millo et al. Fuel 289 (2021) 119835

Abbreviations ID Ignition Delay

KH-RT Kelvin-Helmholtz and Rayleigh-Taylor
AMR Adaptive Mesh Refinement LLFC Leaner Lifted-Flame Combustion
CFD Computational Fluid Dynamics LOL Lift Off Length
CO Carbon Monoxide LTC Low Temperature Combustion
COV Covariance NOx Nitrogen Oxide
CRI Common Rail Injector NTC No-Time-Counter
CVV Constant Volume Vessel PAHs Poly-cyclic Aromatic Hydrocarbons
DFI Ducted Fuel Injection PDA Phase Doppler Anemometry
DT Dwell Time PM Particulate Matter
EGR Exhaust Gas Recirculation RANS Reynolds-Averaged Navier-Stokes
EOC End of Combustion RNG Re-Normalization Group
EOI End of Injection SMD Sauter Mean Diameter
ET Energizing Time SOC Start of Combustion
GDI Gasoline Direct Injection SOI Start of Injection
HC Hydro-Carbon TKE Turbulent Kinetic Energy

simultaneously control NOx emissions by reducing in-cylinder temper­ placed nearest to the injector tip (2.2 mm). In addition, a computational
atures [3]. Early investigations demonstrated the feasibility of LLFC at model was developed to further investigate the DFI concept suggesting
low loads, injecting at high pressure by means of a 2-holes injector; that the high-velocity core of the DFI spray entrains more air and de­
however, at high loads or when using high flow injectors, the LOL is velops elevated turbulence levels at the duct outlet, which contribute to
shortened, the re-entrainment of hot combustion products is promoted a more effective mixing lowering the equivalence ratios at the LOL.
and the interaction of adjacent sprays is detrimental, without achieving Svensson et al. in [13] analyzed the impact of duct geometric char­
LLFC regime [4,5]. Therefore, a further improvement in charge-gas acteristics on its performance, demonstrating the capability to reduce
mixing upstream the LOL was needed to sustain the LLFC under a soot formation only when the ignition was located downstream of the
wider range of engine operating conditions. With this aim, Ducted Fuel duct exit. The outcomes already presented were also confirmed by a
Injection (DFI) concept was proposed in recent years: DFI is based on the recent work from Tanno et al. [14], in which the authors tested the DFI
injection of the fuel through a small cylindrical pipe at predefined dis­ performances in a rapid compression machine, supported by extensive
tance from the nozzle, with the primary aim of extending the LOL, numerical simulation activities, showing the ability of the DFI to prevent
enhancing the mixing in the autoignition zone and thus reducing the soot formation thanks to a reduced equivalence ratio distribution, driven
equivalence ratio at ignition [6]. A preliminary study on the DFI ability by more air entrainment sent toward the spray core.
to abate soot emission has been carried out by Mueller et al. [6] testing a The previously mentioned research studies about the DFI technology
single duct geometry (3 mm inner diameter, 14 mm length and 2 mm have highlighted more than encouraging results regarding its ability to
stand-off distance from the injector tip) in comparison with a free spray abate soot formation, mainly thanks to a faster spray penetration, to a
jet, at different ambient temperatures. The results showed a reduction of consequent LOL extension and an ID increment. Moreover, experiments
an order of magnitude in soot luminosity by comparing DFI and free have suggested that duct should enhance the air/fuel mixing process.
spray injection conditions. Moreover, thanks to a faster penetration rate, However, additional effort should be spent to further investigate the
both the LOL and the Ignition Delay (ID) were extended by the duct mechanisms underlying the impressive soot reduction enabled by the
adoption, with a shorter combustion duration. By using the same Con­ DFI, toward the optimization of this new concept and its implementation
stant Volume Vessel (CVV), Gehmlich et al. in [7] investigated different in real engines. In this framework, the purpose of this study is to cover the
duct geometries confirming the ability of DFI to highly attenuate the still existing knowledge gap in terms of mixing and air entrainment by
soot mass, even at high temperatures (i.e. 1100 K), where the LOL was analyzing the difference in DFI spray evolution with respect to a free
anchored near the duct outlet section, suggesting that both the air spray configuration, by means of a reliable 3D-CFD spray model cali­
entrainment and the air/fuel mixing are enhanced within the duct. brated and validated against an extensive experimental dataset. The
Li et al. in [8,9] investigated the different DFI and free spray char­ calibrated spray model has been used to investigate and reveal several
acteristics in non-reacting conditions, confirming the DFI ability to in­ aspects of the development of the DFI jet, not directly measurable in an
crease the spray penetration rate, especially at high injection pressures. experimental campaign to further understand the reasons why the duct
In addition, the spray morphology was found to be highly affected by the strongly influences the spray penetration as well as the equivalence ratio
duct, showing a larger spray cone angle with respect to the corre­ at the ignition. To do so, the air entrainment in the near-nozzle spray
sponding free spray, and more fluctuating peripheries of the jet, thus region was quantified and its impact on the overall equivalence ratio was
creating a mushroom-shaped spray close to the duct outlet, especially at evaluated. In this regard, the TKE distribution was investigated at the
high ambient pressure (>30 bar) [9]. duct outlet to support the hypothesis of Mueller et al. [6], that “the duct
Nilsen et al. in [10] investigated the DFI concept in an optical enhances the turbulent mixing allowing an equivalence ratio reduction in
research engine demonstrating the ability of the DFI to break the usual the autoignition zone”, based on experimental evidences. In addition, as
trend of soot/NOx trade-off over different EGR rates, resulting in a far as the emission prediction is concerned, for the first time the detailed
strong soot attenuation, without any drawback for the other pollutants Particulate Mimic model has been used in DFI simulations to ensure a
(Fig. 11, [10]). An operating-parameter sensitivity study conducted in qualitative and quantitative soot emission prediction, going beyond the
[11] confirmed this trend of the DFI combustion, with comparable fuel intrinsic limit of the well-known Hiroyasu model adopted in [12].
conversion efficiency respect to the conventional diesel combustion. In The present work starts with the description of the experimental
another research study [12], Fitzgerald et al. combined numerical sim­ analysis carried at the Spray Lab of the Università di Perugia on non-
ulations and experimental tests to provide additional details regarding reacting sprays in a CVV, to characterize the DFI jet in terms of spray
the mechanism enabling soot reduction, highlighting as optimum duct penetration, drop sizing and velocity with respect to a corresponding
geometry the one featuring the smallest inner diameter (2 mm) and free spray. The calibration and validation processes of the 3D-CFD non-

2
F. Millo et al. Fuel 289 (2021) 119835

reacting spray model are then presented, maintaining constant all the
calibration parameters for both free spray and DFI configurations.
Therefore, a physical analysis of the simulated sprays is proposed to
enlighten the main differences in terms of air entrainment, air/fuel
mixing process occurring within and downstream the duct, and the
subsequent equivalence ratio distribution. Finally, the results of the 3D-
CFD combustion simulations, supported by a detailed chemistry solver,
are presented to preliminarily predict the DFI behavior in soot emission
abatement, as obtained in experimental campaigns on reacting sprays
[6,7,12].

2. Non-reacting spray

2.1. Experimental setup and results

A Bosch CRI1 common rail injector featuring a single, in-axis 180 µm

hole was used to investigate the interaction among the diesel fuel spray
and a prototype duct system in high temperature and pressure condi­
tions. The injector was fed with commercial diesel fuel pressurized by a
static pressure generator (Loccioni Mobility Thor) and actuated by a
programmable driver (Loccioni Mobility AEA006).

2.1.1. Hydraulic analysis

The hydraulic analysis of the tested prototype injection system was
carried out by a proprietary Injection Analyzer based on the Zeuch
Method, to derive simultaneously the shot-to-shot injection rate and
injected volume [15]. In order to measure the injection rate time-history
in a given operating condition (defined by the injection pressure level
and injector actuation strategy), the injection took place in a closed, Fig. 1. Top: Spray imaging layout; Bottom: PDA layout.
constant volume chamber filled with the same injected fuel. At the
injector actuation timing, the fuel pressure in the test chamber is con­
stant at a “base level” typically ranging between 20 and 60 bar for diesel
applications. During the injection process, the injection rate is propor­
tional to the chamber pressure rise rate and to the ratio of the chamber
volume to the fluid bulk modulus. Analyzing the chamber pressure time-
history it is possible to derive both the injection rate and the injected
volume relevant to each single injection process. At the end of the in­
jection process, after a proper stabilization period, the measuring
chamber pressure must be reset to the initial level so to repeat the
measurement for a consecutive injection event; to this end, a fast acting Fig. 2. Duct holder.
discharge valve evacuates the injected volume from the chamber by
opening a port towards a large downstream volume maintained at the dedicated CVV. The CVV features 3 rectangular optical accesses (50x100
pressure base level. This sequence of events defines the instrument mm) arranged in order to allow both a back-light optical layout for
operating cycle that in the current release can be operated up to 100 Hz global spray imaging and a classical Phase Doppler Anemometry (PDA)
acquiring simultaneously the injection rate and other dynamic signals layout for drop velocimetry and sizing measurements (110 deg scat­
such as rail pressure, injector current and needle lift. Further, in order to tering angle), as shown in Fig. 1.
obtain a reliable injected volume and injection rate measurements, the The common rail injector is installed on the CVV top, with the single
fluid evacuated from the test chamber is metered by a mass flow meter, jet evolving downward along the vessel main axis. The CVV was
thus overcoming the well-known difficulties related to the fluid bulk designed to operate up to 30 bar and 773 K thanks to a proper internal
modulus dependence on pressure and temperature and allowing a ceramic shield and to a continuous, hot air/nitrogen flow. The incoming
continuous self-calibration of the instrument. The Injection Analyzer is flow (typically 90 Nl/min) is heated by a 3.5 kW electric air gun installed
normally managed in order to perform parametric analyses of the in­ on the vessel side shell; the heater input power was controlled in closed-
jection apparatus by changing the operating conditions, among which loop with the vessel internal temperature measured 20 mm downstream
the parameters defining the actuation strategy (one of the n injector the nozzle exit. In order to investigate the interaction between the high
actuations or n – 1 dwell times, DT) or the injection pressure level. In a pressure spray and a duct, the prototype nozzle was equipped with a
given operating condition, a batch of identical injection events (e.g. 600) stainless steel duct-holder (Fig. 2), designed to accommodate different
is repeated, the first part of which (e.g. 300) is used to obtain a thermal duct designs at different distances from the nozzle hole.
stabilization of the system, while during the remaining events of the In the present work, a 2 mm internal diameter, 14 mm long duct was
batch all dynamic signals are recorded and analyzed to obtain the mean tested; the duct sharp inlet was positioned 2 mm downstream the nozzle
and individual events time-histories of injection rate and of the other exit (see Table 1).
acquired quantities. Further details about the used Injection Analyzer
are reported in [16]. Table 1
Selected duct geometry.
2.1.2. Spray global evolution and sizing analyses Length [mm] Inner diameter [mm] Stand-off distance [mm]
The overall spray evolution and the drop sizing analyses in high
14 2 2
temperature and pressure conditions were both carried out in a

3
F. Millo et al. Fuel 289 (2021) 119835

Table 2
Test plan for hydraulic, spray imaging and PDA analyses.
Hydraulic analysis

Injection Pressure [bar]

400 800 1200

10
Back Pressure Energizing Time 250 ÷ 2000
20
[bar] [µs]
30

High speed imaging analysis

Injection Pressure [bar]

400 800 1200

Vessel 10
Vessel Temperature 373, 573, 773 [K]
Pressure [bar] 20

Fig. 3. Experimental injection rate measurements at different rail pressures and

Phase Doppler Anemometry analysis
constant ET = 1000 µs.
Injection Pressure [bar]

400 800 1200

Vessel
10 Vessel Temperature 373, 773 [K]
Pressure [bar]

The optical layout used for spray imaging was a classical “Z-shaped”
Schlieren configuration (see Fig. 1, Top) used to allow the vapor phase
visualization in low temperature, non-evaporating conditions in the
CVV; removing the Schlieren knife the spray liquid phase could be
visualized. In high temperature conditions, both liquid and vapor phases
are clearly visible in simple backlight configuration. A Luminus CBT-140
LED was used as continuous light source, and 2 spherical mirrors were
used to parallelize light across the CVV and re-focus on the knife edge.
The spray evolution was recorded by a high speed camera (Vision
Research Miro M310) operated at 50 kframe/s, with 128x320 resolu­
tion. In each operating condition, 20 acquisitions were performed to
account for the shot-to-shot dispersion. The off-line analysis of frames
extracted from the video allowed the construction of time-dependent
Fig. 4. SOI zoom: injection rate (black) and injected volume (blue) at different
spray tip penetration and cone angle curves for both the liquid and rail pressures and constant ET = 1000 µs . Vertical bars indicate the computed
vapor phases, along with the corresponding spray boundary plots. spray start timings. (For interpretation of the references to colour in this figure
Further details about the image analysis procedure can be found in [17]. legend, the reader is referred to the web version of this article.)
For drop sizing and velocity analysis, a Dantec Dynamics 1-D PDA sys­
tem was used in the present research activity. The PDA system (see pressure increase causes a gradual reduction of the injection delay from
Fig. 1, bottom) is based on a BSAP80 processor, a 112 mm Fiber PDA the ET start, with less evident effects on the injector closing delay. The
transmitter and a 112 mm Hi-Dense receiver, which is specifically steady flow rate level is achieved for ET values between 400 and 500 µs
designed to operate with dense spray such as diesel or GDI jets. The light depending on the injection pressure level, confirming the end of ballistic
source is a Dantec FlowLite 1D (532 nm, 200 mW). Further details on the operation characterized by a relatively high shot-to-shot injected
main features of the system are reported in [18]. Both the PDA trans­ quantity dispersion. In order to correctly simulate the spray evolution,
mitter and receiver are positioned by a 3-axis Cartesian traverse system, the actual hydraulic Start Of Injection (SOI) was estimated from the
which allows the PDA measuring volume positioning in XY planes at injection rate measurement following the procedure reported in [16].
different distances from the nozzle exit along the vertical, downward According to this methodology the actual spray exit timing can be
oriented Z-axis. evaluated computing the injected volume as the injection rate time-
integral, and the actual SOI is coincident with the final zero-crossing
2.1.3. Experimental test plan and results event, as shown by the black vertical lines in Fig. 4. The correspond­
In order to explore the spray-duct interaction in a wide range of ing injection rate level is assumed as the initial value of the liquid in­
operating conditions, and to adequately validate the numerical model, jection flow rate adopted in the 3D-CFD Lagrangian simulation of the
the test plan reported in Table 2 was followed. spray evolution.
The injected volume and injection rate measurements were carried Sample sequences of the spray evolution in different operating con­
out with 10, 20 and 30 bar back pressure, with energizing time (ET) ditions are reported in Fig. 5. For the sake of brevity, only data per­
values up to 2000 µs. The obtained shot-to-shot dispersion of injected taining to the 20 bar vessel pressure operating conditions are shown. The
volume was appreciably low, with COV values below 0.5% in the linear first sequence (Fig. 5 – top) was obtained in non-evaporative conditions
operation zone for all the tested injection pressure levels. Consequently, (373 K) with the spray freely evolving in the CVV. In the second and
for the following spray evolution analysis the ET = 1000 µs was third sequences (Fig. 5 – middle and Fig. 5 – bottom, respectively), the
assumed, and the related experimental injection rate are depicted in free and ducted sprays developed in fully evaporative conditions (773
Fig. 3 for all the tested rail pressures. As can be seen, the injection K).

4
F. Millo et al. Fuel 289 (2021) 119835

Fig. 5. Spray evolution sequences, rail pressure = 1200 bar, ET = 1.0 ms. Top: free spray, vessel pressure = 20 bar, vessel temperature = 373 K. Middle: free spray,
vessel pressure = 20 bar, vessel temperature = 773 K. Bottom: ducted spray, vessel pressure = 20 bar, vessel temperature = 773 K.

The quantitative analysis of the ducted spray evolution evidences The drop sizing and velocity analysis was carried out along a linear
significant differences with the free spray, as outlined in Fig. 6 for some traverse along Y direction crossing the entire spray structure over a
of the examined operating conditions. The spray tip penetration velocity plane at 50 mm distance from the nozzle exit. The Y traverse was
downstream the duct is generally higher for the ducted configuration, positioned at X = 2 mm coordinate where the maximum data rate was
suggesting the missing momentum flux exchange with the surrounding obtained in a preliminary scan. In each measurement station in the
air in the duct to prevail on the spray-duct wall dissipative interactions. traverse, valid data pertaining to 150 consecutive shots were collected.
In terms of liquid spray structure shape, in non-evaporative conditions Moreover, the spray boundaries were determined according the data
the spray-duct interactions caused an evident increase of spray cone count criterion suggested in [19]. In Fig. 7, the raw results obtained in X
angle, from 15 to 20 deg, suggesting a more intense momentum flux = 2, Y = 0 in non-evaporative and evaporative conditions are reported
exchange and air entrainment flux. Conversely, in high vessel temper­ for the free spray configuration, in which time axis values are referred to
ature conditions the strong evaporation rate and the reduced air the ET start.
chamber density significantly thinned the spray liquid spray to such an Fig. 7 shows that in non-evaporative conditions a large number of
extent that global cone angle values in the range 9 to 12 deg were drops is detected after the End of Injection (EOI) (1500 µs aSOI). These
measured in all operating conditions and configurations. In addition, it drops, passing through the considered measuring station after the EOI
is worth to point out that the liquid spray cone angle is affected by a are progressively slowed down due to aerodynamic drag while their
higher fluctuation of the measured angle due to the duct insertion, diameter tends to remain constant due to both very slow evaporation
which is also perceivable observing the middle and bottom spray se­ and weak breakup mechanisms. Conversely, at vessel temperature equal
quences reported in Fig. 5. In fact, in the last sequences (from 300 µs to to 773 K, the spray evolution is completely different: drops are no longer
1500 µs aSOI), the internal liquid-portion of the spray appears thinner detected in the measuring station later than 1900 µs aSOI (400 µs after
and more distorted, suggesting also the presence of enhanced evapora­ the EOI) evidencing the drastic effect of evaporation mechanism in high
tion and more intense radial velocity components related to air temperature conditions. Further, in terms of drops axial velocity, the
entrainment in high temperature conditions. vessel temperature increase causes the spectrum of observed values to be

5
F. Millo et al. Fuel 289 (2021) 119835

Fig. 6. Liquid penetration (left) and global cone angle (right) for free spray and DFI configurations at different rail pressures. Test conditions: vessel pressure = 20
bar, vessel temperature = 373 K (top) – 773 K (bottom), ET 1000 µs.

significantly larger due to the reduced vessel air density and conversely containing the flow field data and the local thermodynamic conditions
the range of measured diameters is dramatically narrowed for both within the CVV was saved and then used to initialize the subsequent
improved breakup and promoted drop evaporation velocity. transient spray simulation for the injection conditions highlighted in
The comparison between free and ducted spray configurations in Table 2 – High speed imaging. As far as the temperature initialization of
terms of mean count data and drop sizing for the tested conditions vessel the duct walls, a steady-state condition was assumed considering the
pressure = 10 bar, vessel temperature = 773 K at different rail pressures duct walls temperature equal to the chamber temperature and the
is reported in Fig. 8. The most evident result is the remarkable decrease O’Rourke and Amsden model was employed to account for heat transfer
of drops count (− 58% at 1200 bar of rail pressure) obtained for the from duct wall to spray [21]. The inner chamber volume was auto­
ducted spray, in all the explored measuring stations: this evidence seems matically meshed by CONVERGE CFD using a patented cut-cell tech­
to confirm the enhanced drop evaporation rate due to the duct presence. nique [20]. Regarding the grid control, two fixed grid refinement
In terms of Sauter Mean Diameter (SMD) of the droplets, only marginal regions were considered: one covering the experimental imaging win­
reductions (lower than 0.5 µm) seem to be caused by the duct adoption, dow (minimum grid size 0.50 mm in the spray area); the other, only for
the most significant effect being the evident decrease in the residual DFI configuration, within the duct volume (minimum grid size 0.25 mm)
drops number downstream the duct rather than substantial modification to guarantee a reasonable minimum number of cells along the duct
in the drop diameter statistical distribution. diameter. In addition, an Adaptive Mesh Refinement (AMR) technique
(minimum grid size 0.25 mm) was applied in the chamber domain to
automatically refine the grid based on local temperature and velocity
2.2. Simulation setup gradients. Fig. 9 shows the chosen grid control settings in the DFI setup
during the injection event.
The 3D-CFD analysis was carried out by means of a commercially As far as the description of the liquid/gas coupling is concerned, a
available software, CONVERGE CFD [20]. Firstly, the experimental CVV Lagrangian and Eulerian approach was employed for the liquid and gas
was virtually reproduced considering two different configurations, the phase, respectively. The turbulence was simulated by means of a
free spray and the DFI, since this latter featured the duct and its holder Reynolds-Averaged Navier-Stokes (RANS) approach adopting a Re-
(please see Fig. 2). The numerical analysis was divided in two steps: a Normalization Group (RNG) k-ε model [22]. Several models of liquid
first steady-state simulation was carried out to initialize the local tur­ injection, spray breakup, evaporation, droplet drag and collision were
bulence inside the chamber; then a transient simulation was performed tested to properly reproduce the experimental data. In particular, Diesel
reproducing the injection event. In the steady-state analysis an ingoing #2 was selected as injected fuel and the blob injection model was
and outgoing air flows were considered, to replicate the experimental adopted [23], in which the initial injected droplets size is equal to the
test bench layout. When the steady-state was reached, the solution

6
F. Millo et al. Fuel 289 (2021) 119835

Fig. 9. Grid discretization and refinement for the ducted fuel injection
configuration.

Table 3
Summary of 3D-CFD model setup.
Turbulence model RNG k-ε
Fuel properties Diesel #2
Liquid Injection Blob model
Fig. 7. Raw drops data at X = 2, Y = 0, Z = 50 mm for free spray configuration. Droplet Turbulent Dispersion O’Rourke model
Test conditions: Rail pressure = 1200 bar, vessel pressure = 10 bar, vessel Spray Breakup model Calibrated KH-RT
temperatures = 373 K (non-evaporative conditions) − 773 K (evaporative Evaporation model Frossling with boiling model
conditions). Top: drop velocity. Bottom: sizing data. Droplet Drag Dynamic Drop drag
Collision model No collision model
Droplet/wall interaction Rebound/Slide
Heat Transfer model O’Rourke and Amsden

effective nozzle diameter. Due to the high pressure gradient across the
nozzle, the super-cavitation onset hypothesis was considered based on
Chaves et al. work [24]. To properly take into account this behavior and
to simulate the nozzle effective area shrinking, the discharge coefficient
value was chosen following the approach mentioned in [24]. Regarding
the atomization of the liquid and the breakup of the droplets, a cali­
brated Kelvin-Helmholtz and Rayleigh-Taylor (KH-RT) model was
implemented [25], calibrating the time and size constants of both the
models to match the experimental data in terms of liquid penetration
curve and droplets SMD. The Frossling evaporation model [26] was
included in the droplet boiling model; in addition, to account the pos­
sibility of drop shape variations, the dynamic drop drag model [27] was
adopted. To evaluate the droplet collision in the Lagrangian spray, both
O’Rourke algorithm [28] and no-time-counter (NTC) method [29] were
initially tested. However they were removed from the final calibrated
model since the significant increase of the CPU-time request (especially
for O’Rourke model) was not counterbalanced by a noticeable impact on
the spray evolution. Moving to the DFI configuration, an accurate
analysis of the wall spray impingement model was carried out to capture
the interaction between the spray parcels and the inner wall along the
duct length. Several wall film models were tested, and the model
developed by Naber et al. [30] was chosen as the more representative of
the experimental behavior. To summarize, all the implemented models
in the 3D-CFD setup are listed in Table 3.
The 3D-CFD spray model was calibrated on a selected free spray
reference condition, in order to both correctly characterize the spray
Fig. 8. Free vs. ducted spray configurations at different rail pressures. Test evolution for the whole tested range, and properly capture the experi­
conditions: vessel pressure = 10 bar, vessel temperature = 773 K. Top: count mental trends with and without the duct maintaining constant the spray
data. Bottom: sizing data. model constants. The reference case selected for the calibration features
a rail pressure equal to 1200 bar, a vessel temperature of 573 K and a

7
F. Millo et al. Fuel 289 (2021) 119835

Fig. 10. Liquid penetration curve for the free spray (red) and DFI (blue) configurations: numerical results (solid lines) compared with experimental data (dotted
lines). Test conditions: rail pressure = 800 bar (left) – 1200 bar (right); vessel pressure = 20 bar; vessel temperature = 773 K. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 11. Pressure distribution evolution within the duct, black line represents the jet profile. Test conditions: rail pressure = 1200 bar, vessel pressure = 20 bar,
vessel temperature = 573 K.

vessel pressure of 20 bar. The calibrated spray model was then validated experiments and simulation results are reported in the Appendix A, in
considering the conditions in Table 2 – High speed imaging, for both free which a more than satisfactory agreement could be highlighted.
spray and DFI configurations. The complete comparison between
2.3. Physical analysis

Fig. 10 shows the liquid penetration curves in the free spray and in
the ducted configurations, comparing the experimental data with the
3D-CFD simulation results. The chosen test conditions have same
chamber air density (vessel pressure = 20 bar, vessel temperature = 773
K) but different rail pressures (left: 800 bar, right: 1200 bar). Since the
duct used in the experimental activity was made of stainless steel
without any optical access in the inner region, the comparison between
experimental and simulated liquid penetration in DFI configuration can
be made only downstream the duct exit; therefore a grey box named
‘Duct’ was added in Fig. 10. It can be highlighted that the duct leads to a
faster spray penetration for the whole optical access window, and the
trend is correctly captured also by the developed numerical model.
However, this latter shows a slight overestimation of penetration rate
close to the experimental imaging limit. The highlighted trend suggests
that the duct narrows the jet section respect to the free spray, and lowers
the aerodynamic drag causing a higher momentum conservation and, as
Fig. 12. Mass flow rate of the entrained air into the spray: comparison between
free spray (red) ducted spray (blue) considering only the upper part of the spray a consequence, a faster spray penetration.
area. t* represents the time when the DFI jet impinges on the duct inner wall. Another interesting feature of duct introduction is a strong modifi­
Test conditions: rail pressure = 1200 bar, vessel pressure = 20 bar, vessel cation of the gas flow pattern close to the spray. Analyzing the spray
temperature = 573 K. (For interpretation of the references to colour in this propagation, it is known that a free spray is surrounded by a constant
figure legend, the reader is referred to the web version of this article.) pressure equal to the imposed chamber pressure [31]. On the contrary,

8
F. Millo et al. Fuel 289 (2021) 119835

Fig. 14. Pressure distribution and air flow streamlines at duct outlet. Black
Fig. 13. Velocity vectors distribution at 450 µs aSOI colored with the axial points represent the liquid fuel parcels. Test conditions: rail pressure = 1200
velocity component (Vz): comparison between free spray (left) and ducted jet bar, vessel pressure = 20 bar, vessel temperature = 573 K.
(right). Test conditions: rail pressure = 1200 bar, vessel pressure = 20 bar,
vessel temperature = 573 K. are depicted in Fig. 13, considering Reference test conditions. It is well-
known that in free spray configuration, the ambient air enters into the
in DFI configuration, the interaction between the spray and the air is spray boundaries with a predominantly radial velocity component [35],
prevented along the duct length, and the high velocity jet generates a as also shown in Fig. 13 – left. Moreover, due to the high momentum of
low pressure region within the duct, as in the ejector working principle the spray liquid core, the air entrainment is mainly located on the pe­
[32]. The evolution of the pressure distribution inside the duct is ripheries of the jet, thus resulting in a challenging air/fuel mixing near
depicted in Fig. 11 for the test conditions: vessel temperature = 573 K, the spray centerline. On the contrary, Fig. 13 – right shows that in the
vessel pressure = 20 bar and rail pressure = 1200 bar, hereafter defined ducted spray an annular entrained air flow has a strong velocity
as Reference test conditions. At SOI, the pressure inside the duct is component parallel to the spray direction, thus confirming the results
constant and equal to the imposed chamber pressure (20 bar). Then, at from [36]. In addition, considering the vectors at the duct inlet it is clear
50 µs aSOI, a region of high pressure is visible inside the duct on the jet that the surrounding air is directed not only on the peripheries of the
front: in fact, the spray evolution pushes and compresses the air inside spray, but also near the liquid core, suggesting a more efficient air/fuel
the duct, resulting in a higher local pressure with respect to the sur­ mixing, as also highlighted in Fig. 12.
rounding chamber. In addition, a progressive decrease of pressure arises Summing up, the suction effect due to duct adoption enhances the air
upstream the jet front (clearly visible at 150 µs and 250 µs aSOI) caused entrainment downstream the nozzle tip, and this air flow shows a strong
by the high velocity spray generating an air pumping effect at the duct directionality along the spray direction and is quite uniformly distrib­
inlet, and resulting in a faster liquid penetration as well. A quantitative uted on the radial dimension of the spray.
analysis on the amount of entrained air into the spray was carried out When the spray reaches the duct outlet, the velocity gradient be­
comparing the results between the free spray and DFI liquid jet, tween the incoming high-speed ducted spray and the quasi-stationary
considering the same axial extension defined as the length at which the air chamber creates local turbulent eddies (shearing vortexes) near the
liquid spray impinges on the duct inner wall. The results are depicted in duct sharp edges, as can be seen in Fig. 14 – left, where the pressure
Fig. 12, in which the evolution of the air mass flow rate through the distribution and the air flow streamlines are depicted at 80 µs aSOI. In
spray profile is presented for Reference test conditions. addition, the local static pressure is also shown, highlighting a toroidal
It is worth noting that, as suggested by the pressure evolution, the distribution of low pressure that could enhance the air entrainment at
ducted injection leads to an almost 2 times higher air entrainment in to the duct outlet as well. Going ahead in the injection event (130 µs aSOI),
the liquid spray. However, the DFI air entrainment curve shows a lower the toroidal vortex moves with the spray evolution and becomes more
rising slope with respect to free spray configuration. More in detail, the intense, dragging the surrounding chamber air into the developed vortex
entrained air mass flow rates starts to diverge at time equal to t*, that ring, and creating the typical DFI mushroom-shaped spray head inves­
corresponds to the time when the DFI jet impinges on the duct inner tigated in [9].
wall. Going ahead in the injection event, the free spray reaches almost To evaluate the local turbulence distribution, the time-evolution of
immediately a stationary value of entrained air in the region close to the the predicted Turbulent Kinetic Energy (TKE) is depicted in Fig. 15
injector tip. On the contrary, the DFI jet shows a slower rising transient comparing the free spray and the DFI configurations, for Reference test
before reaching the steady-state. In fact, right after the SOI, the DFI jet conditions. Comparing free spray and DFI configurations, it is worth
confinement effect results in a smaller available contact surface with the mentioning that a progressive strong enhancement of the TKE is detec­
surrounding air. Nevertheless, this effect becomes less important when ted downstream the duct outlet, where the vortex ring were previously
the progressive pressure decrease (highlighted in Fig. 11) caused by the highlighted, that could promote the air/fuel mixing in this region.
high velocity spray, starts to generate a strong air suction effect at the Regarding the region downstream the injector tip, Fig. 16 shows an
duct inlet. The steady-state value of the entrained air mass flow rate in enlargement of the TKE distribution at a fixed instant of time (450 µs
DFI configuration is reached approximately 250 µs aSOI, when the aSOI).
pressure layering within the duct becomes stable as well (Fig. 11, last First of all, the Reynolds number was evaluated to characterize the
frame). Moreover, since the near-nozzle region is the most influent in the flow regime inside the duct. The Reynolds number is defined as:
air/fuel mixing [33,34], the higher air entrained in this region, high­
ρuD
lighted in Fig. 12, could enhance the entire mixing rate, even if the duct Re = ,
μ
prevents the air entrainment along the wall length.
To further analyze the air flow rate, the velocity vectors of the where ρ is the ambient air density, D is the inner duct diameter, u the
gaseous phase at 450 µs aSOI for both the free spray and the ducted jet

9
F. Millo et al. Fuel 289 (2021) 119835

Fig. 15. Evolution of the TKE distribution comparing free spray and DFI configuration. Test conditions: rail pressure = 1200 bar, vessel pressure = 20 bar, vessel
temperature = 573 K.

upstream of the duct exit plane also play a critical role”. Starting from
this result, Mueller et al. in [6] have supposed that the duct enhances the
turbulent mixing allowing an equivalence ratio reduction in the auto­
ignition zone. In this regard, Fig. 17 shows the evolution of the equiv­
alence ratio distribution obtained through the 3D-CFD numerical
analysis for free spray and DFI configurations, considering Reference
test conditions. It is noteworthy that free spray jet leads to a high
equivalence ratio area corresponding to the liquid core of the spray due
to the already mentioned low air/fuel mixing near the spray centerline.
On the contrary, the ducted jet shows a narrower equivalence ratio
distribution grouped in the range [0.8 – 1.4], highlighting the potential
of the high efficient turbulent mixing and the enhanced air entrainment
enabled by the duct adoption. In addition, extremely lower equivalence
ratio could be highlighted in the near-nozzle region with respect to free
spray, thanks to the higher entrained air flow rate, shown in Fig. 12.

3. Reacting spray
Fig. 16. Enlargement of the TKE distribution of Fig. 15 at nozzle tip comparing
free spray and DFI configurations. Test conditions: rail pressure = 1200 bar, 3.1. Numerical analysis
vessel pressure = 20 bar, vessel temperature = 573 K.
The spray model, successfully calibrated and validated in non-
mean centerline velocity of the spray inside the duct and µ the dynamic reacting conditions, was then used to analyze the DFI behavior in
viscosity of the ambient air. In the Reference test conditions for the reacting spray conditions, by means of the SAGE detailed chemistry
selected duct geometry (Table 1), the obtained Re is equal to 2.4 × 105, solver implementation. The Skeletal Zeuch chemical mechanism
that is in line with the results obtained in similar non reacting conditions compiled of the n-heptane oxidation scheme, with additional soot re­
in [8]. Knowing that pipe flow becomes turbulent for Re > 2300, a actions from Mauss’s work [37], was adopted. It consists of 121 species
highly turbulent flow is expected within the duct, as confirmed by and 593 reactions; moreover, the inclusion of Poly-cyclic Aromatic
Fig. 16. A radial distribution of TKE is clearly visible within the duct that Hydrocarbons (PAHs) as soot precursors enables the detailed Particulate
reflects the velocity boundary layer from the duct walls to the spray axis, Mimic model [38] for more accurate soot estimation. The combustion
with a consequent improvement of the air/fuel mixing rate within the and emission models are summarized in Table 4.
duct. These findings are in line with the numerical analysis carried out in As far as the volume discretization is concerned, a fixed grid
[12] on reacting sprays. In the mentioned work, Fitzgerald et al. showed refinement was placed in the region in which the spray evolves and in
the effect of duct adoption on the TKE evaluated on the spray centerline the duct region, obtaining a base grid size in the spray area of 0.50 mm
(Figure 22, [12]), highlighting that the DFI resulted in a lower turbu­ and 0.25 mm inside the duct. In addition, AMR technique was applied to
lence inside the duct with respect to free spray configuration. However, automatically adjust the grid resolution based on the curvatures of the
Fig. 16 suggests that despite of lower values of TKE near the ducted velocity and temperature fields, obtaining a minimum grid size of 0.25
spray centerline, the radial distribution of turbulence could enhance the mm. The turbulence and the spray models were not modified with
air/fuel mixing respect to the free spray. Actually, this hypothesis agrees respect to the non-reacting simulation activity and are listed in Table 3.
with the findings obtained in [7] where the soot attenuation ability of Differently from non-reacting spray analysis, the chamber was consid­
the DFI has been observed, even when the flame LOL is anchored near ered quiescent.
the duct exit, highlighting that “enhanced mixing and/or entrainment The combustion analysis was performed sweeping the air

10
F. Millo et al. Fuel 289 (2021) 119835

Fig. 17. Evolution of the Equivalence Ratio distribution comparing free spray and DFI configurations. Test conditions: rail pressure = 1200 bar, vessel pressure = 20
bar, vessel temperature = 573 K.

The numerical results in terms of heat release rate and soot emission
Table 4
mass are shown in Fig. 18 for the tested operating conditions.
Summary of 3D-CFD combustion and emission models.
The heat release rate and the soot mass predictions are in good
Combustion SAGE detailed chemistry solver agreement with the experiments carried out in [6,7,12], with an
Chemical kinetics Skeletal Zeuch mechanism
impressive soot reduction for each tested chamber condition, thus con­
Soot formation/oxidation Particulate Mimic model
firming the reliability and the predictive capabilities of the developed
3D-CFD model. More specifically, the heat release rate comparison
shows an extension of the ID due to the duct adoption except for the
Table 5 1100 K case (Fig. 18 – right), in which the start of combustion occurs
Test matrix for reacting spray analysis.
inside the duct, requiring an optimization of duct geometry. However,
Rail Pressure [bar] Vessel Density [kg/m3] Vessel Temperature [K] this does not lead to soot mass increment since the highly efficient
1200 22.2 900 mixing enabled by duct adoption prevents the soot formation, con­
1200 22.2 1000 firming the experimental outcomes obtained in [7]. The delayed Start of
1200 22.2 1100 Combustion (SOC) at 900 and 1000 K, causes a more intense DFI pre­
mixed combustion phase and a quasi-steady combustion slightly higher
temperature and pressure within the CVV, keeping constant the chamber respect to the free spray causing a faster burnout at the EOI, as experi­
density, as highlighted in Table 5. In addition, the highest available rail mentally observed in [6,7,12]. Looking at the total soot mass predicted
pressure value was selected to be as close as possible to typical engine by the Particulate Mimic model, it is evident that the DFI configuration
operating conditions. is able to substantially abate soot emissions respect to the free spray,

Fig. 18. Heat release rate (top) and soot mass (bottom) comparison between free spray (red) and DFI (blue) configurations. Test conditions: vessel temperatures: 900
K (left), 1000 K (center), 1100 K (right); rail pressure = 1200 bar, air density: 22.2 kg/m3. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)

11
F. Millo et al. Fuel 289 (2021) 119835

Fig. 19. Evolution of the soot mass distribution comparing free spray and the DFI configurations. Test conditions: rail pressure = 1200 bar, vessel temperature =
1100 K, air density = 22.2 kg/m3.

Results concerning NOx, CO and HC emissions evaluated at the end

of combustion are presented in Fig. 20. A slight decrease in terms of NOx
emissions can be detected for DFI, thus suggesting the DFI ability to
break the traditional soot/NOx tradeoff as experimentally investigated
by Nilsen et al. (Fig. 11, [10]; Fig. 16, [11]), allowing high EGR rates to
attenuate NOx formation without any drawbacks on soot emissions.
Moreover, similar CO and HC emissions values are obtained for free
spray and DFI configurations. Concluding, the preliminary results ob­
tained for all gaseous pollutant emissions are promising, without
showing any detrimental effect of DFI.

4. Conclusions

In this paper, an assessment of the impact of Ducted Fuel Injection

(DFI) adoption on fuel spray characteristics, on air/fuel mixing and,
therefore, on soot formation phenomena was carried out. First, an
extensive experimental campaign was performed in a constant volume
vessel at several non-reacting thermodynamic conditions, to evaluate
the influence of ducted jet on liquid penetration length, spray cone angle
and droplet sizing. As a result, ducted configuration highlighted a faster
spray tip penetration with respect to free spray, suggesting that the
missing momentum flux exchange with the surrounding air in the duct
to prevail on the spray-wall dissipative interactions. In addition, the
comparison between free and ducted spray in terms of mean count data
and drop sizing under evaporative conditions highlighted a remarkable
decrease of drops count for the ducted spray, confirming an enhanced
drop evaporation rate due to the duct presence.
Fig. 20. NOx, CO and HC emissions as a function of the vessel temperature for Afterwards, the experimental data were used to calibrate and vali­
the test matrix in Table 5. date a 3D-CFD spray model for both free spray and ducted injection,
highlighting a high predictive capability in reproducing the spray
with a complete attenuation at 1.8 ms aSOI for all the tested conditions. development. Therefore, the spray model was used to investigate the
Contour plots of the evolution of the local soot mass distributions are different physical mechanisms enabled by DFI, as the increment of the
shown in Fig. 19 for vessel temperature equal to 1100 K case, comparing entrained air in the near-nozzle region due to the reduced pressure
the free spray and the DFI configurations. within the duct and to the improved air/fuel mixing enhanced by the
At the beginning (0.7 ms aSOI) the total soot mass quantities are high toroidal vortex at the duct exit. The combination of these phe­
comparable for the two configurations as well as the spatial extents of nomena was found to be beneficial in terms of equivalence ratio
soot mass distribution. Going ahead in the combustion process (1.3 ms reduction, reaching the goal to limit the spray region at duct exit below
aSOI), the soot formation in the free spray jet is still rising, showing a 2, thus paving the way for suppressing soot formation. Finally, to eval­
larger area of high soot mass density respect to the previous frame. On uate the DFI potential for soot emissions abatement, reacting conditions
the contrary, the DFI configuration highlights a reduction of soot mass were investigated by means of the already developed 3D-CFD model.
density at the jet front. Finally, towards the EOC (1.7 ms aSOI), the soot Results have shown that thanks to the DFI adoption, an impressive soot
mass is almost completely oxidized in the DFI configuration, while an reduction could be achieved without any drawbacks on the other pol­
extensive area of high soot mass density remains still visible in the free lutants, in good agreement with the combustion evolution trends
spray configuration. already reported in literature.

12
F. Millo et al. Fuel 289 (2021) 119835

CRediT authorship contribution statement Appendix A. Model validation

F. Millo: Conceptualization, Methodology, Investigation, Writing - Once the model was calibrated considering one test reference con­
review & editing, Supervision, Project administration. A. Piano: dition, it was validated on the whole experimental test matrix, (see
Conceptualization, Methodology, Investigation, Writing - original draft, Table 2), for both free spray and DFI configurations. Figs. 21A–24A
Writing - review & editing, Supervision. B. Peiretti Paradisi: Investi­ show the liquid penetration curves comparison between experiments
gation, Writing - original draft. L. Postrioti: Conceptualization, Meth­ and simulation results for free spray and ducted jet, considering two rail
odology, Investigation, Writing - original draft, Writing - review & pressure levels (800 and 1200 bar), two vessel pressure levels (10 and
editing, Project administration. L. Pieracci: Investigation. A. Bianco: 20 bar) and two vessel temperatures (573 and 773 K). The experimental
Conceptualization, Methodology, Investigation, Writing - review & liquid penetration curve is presented in terms of averaged values of
editing. F.C. Pesce: Conceptualization, Supervision, Project adminis­ twenty scans. Two different grey boxes were added on each plot: on one
tration. A. Vassallo: Conceptualization. side, since the duct was not optically accessible, no comparison can be
made within it; on the other side, the experimental imaging window
Declaration of Competing Interest length limits the comparison at 50.0 mm.
A more than satisfactory agreement between experiments and sim­
The authors declare that they have no known competing financial ulations could be highlighted for both free spray and DFI configurations.
interests or personal relationships that could have appeared to influence A slight discrepancy can be observed close to the imaging limit, espe­
the work reported in this paper. cially at higher vessel temperatures in Figs. 22A and 24A; however the
error was considered as acceptable for automotive engine applications
Acknowledgements taking into account the typical piston bore size. In addition, the DFI
ability to increase the liquid penetration rate is experimentally
Computational resources provided by hpc@polito, which is a project confirmed in all the tested conditions, and correctly reproduced by the
of Academic Computing within the Department of Control and Com­ developed 3D-CFD numerical model.
puter Engineering at the Politecnico di Torino (http://hpc.polito.it).

Fig. 21A. Liquid penetration curve for the free spray (red) and DFI (blue) configurations: numerical results (solid lines) compared with experimental data (dotted
lines). Test conditions: rail pressure = 800 bar (left) – 1200 bar (right); vessel pressure = 10 bar, vessel temperature = 573 K.

Fig. 22A. Liquid penetration curve for the free spray (red) and DFI (blue) configurations: numerical results (solid lines) compared with experimental data (dotted
lines). Test conditions: rail pressure = 800 bar (left) – 1200 bar (right); vessel pressure = 10 bar, vessel temperature = 773 K.

13
F. Millo et al. Fuel 289 (2021) 119835

Fig. 23A. Liquid penetration curve for the free spray (red) and DFI (blue) configurations: numerical results (solid lines) compared with experimental data (dotted
lines). Test conditions: rail pressure = 800 bar (left) – 1200 bar (right); vessel pressure = 20 bar, vessel temperature = 573 K.

Fig. 24A. Liquid penetration curve for the free spray (red) and DFI (blue) configurations: numerical results (solid lines) compared with experimental data (dotted
lines). Test conditions: rail pressure = 800 bar (left) – 1200 bar (right); vessel pressure = 20 bar, vessel temperature = 773 K.

References vaporizing conditions. Fuel 2020;268:117192. https://doi.org/10.1016/j.

fuel.2020.117192.
[10] Nilsen CW, Biles DE, Yraguen BF, Mueller CJ. Ducted fuel injection versus
[1] Dec JE. Advanced compression-ignition engines—understanding the in-cylinder
conventional diesel combustion: an operating-parameter sensitivity study
processes. Proc Combust Inst 2009;32(2):2727–42. https://doi.org/10.1016/j.
conducted in an optical engine with a four-orifice fuel injector. SAE Int J Engines
proci.2008.08.008.
2020;13(3). https://doi.org/10.4271/03-13-03-0023.
[2] Musculus MPB, Miles PC, Pickett LM. Conceptual models for partially premixed
[11] Nilsen CW, Biles DE, Mueller CJ. Using ducted fuel injection to attenuate soot
low-temperature diesel combustion. Prog Energy Combust Sci 2013;39(2-3):
formation in a mixing-controlled compression ignition engine. SAE Int J Engines
246–83. https://doi.org/10.1016/j.pecs.2012.09.001.
2019;12(3). https://doi.org/10.4271/03-12-03-0021.
[3] Pickett LM, Siebers, DL. Non-sooting, low flame temperature mixing-controlled di
[12] Fitzgerald RP, Svensson K, Martin G, Qi Y, Koci C. Early investigation of ducted fuel
diesel combustion, SAE Technical Paper 2004-01-1399, 2004, doi: 10.4271/2004-
injection for reducing soot in mixing-controlled diesel flames. SAE Int J Engines
01-1399.
2018;11(6):817–33. https://doi.org/10.4271/2018-01-0238.
[4] Polonowski CJ, Mueller CJ, Gehrke CR, Bazyn T, Martin GC, Lillo PM. An
[13] Svensson KI, Martin GC. Ducted fuel injection: Effects of stand-off distance and
experimental investigation of low-soot and soot-free combustion strategies in a
duct length on soot reduction, SAE Technical Paper 2019-01-0545, 2019, doi:
heavy-duty, single-cylinder, direct-injection, optical diesel engine. SAE Int J Fuels
10.4271/2019-01-0545.
Lubr 2012;5(1):51–77. https://doi.org/10.4271/2011-01-1812.
[14] Tanno S, Kawakami J, Kitano K, Hashizume T. Investigation of a novel leaner fuel
[5] Gehmlich RK, Dumitrescu CE, Wang Y, Mueller CJ. Leaner lifted-flame combustion
spray formation for reducing soot in diffusive diesel combustion-Homogenizing
enabled by the use of an oxygenated fuel in an optical CI engine. SAE Int J Engines
equivalence ratio distribution in the lift-off region-, SAE Technical Paper 2019-01-
2016;9(3):1526–43. https://doi.org/10.4271/2016-01-0730.
2273, 2019, doi:10.4271/2019-01-2273.
[6] Mueller CJ, Nilsen CW, Ruth DJ, Gehmlich RK, Pickett LM, Skeen SA. Ducted fuel
[15] Piano A, Millo F, Postrioti L, Biscontini G, Cavicchi A, Pesce FC. Numerical and
injection: A new approach for lowering soot emissions from direct-injection
experimental assessment of a solenoid common-rail injector operation with
engines. Appl Energy 2017;204:206–20. https://doi.org/10.1016/j.
advanced injection strategies. SAE Int J Engines 2016;9(1):565–75. https://doi.
apenergy.2017.07.001.
org/10.4271/2016-01-0563.
[7] Gehmlich RK, Mueller CJ, Ruth DJ, Nilsen CW, Skeen SA, Manin J. Using ducted
[16] Postrioti L, Buitoni G, Pesce FC, Ciaravino C. Zeuch method-based injection rate
fuel injection to attenuate or prevent soot formation in mixing-controlled
analysis of a common-rail system operated with advanced injection strategies. Fuel
combustion strategies for engine applications. Appl Energy 2018;226:1169–86.
2014;128:188–98. https://doi.org/10.1016/j.fuel.2014.03.006.
https://doi.org/10.1016/j.apenergy.2018.05.078.
[17] Brizi G, Postrioti L, van N. Experimental analysis of SCR spray evolution and sizing
[8] Li F, Lee C-F, Wu H, Wang Z, Liu F. An optical investigation on spray macroscopic
in high-temperature and flash boiling conditions. SAE Int J Fuels Lubr 2019;12(2).
characteristics of ducted fuel injection. Exp Therm Fluid Sci 2019;109:109918.
https://doi.org/10.4271/04-12-02-0006.
https://doi.org/10.1016/j.expthermflusci.2019.109918.
[18] Postrioti L, Cavicchi A, Brizi G, Berni F, Fontanesi, S. Experimental and Numerical
[9] Li F, Lee C-F, Wang Z, Liu F, Lu G. Optical investigation on impacts of ambient
Analysis of Spray Evolution, Hydraulics and Atomization for a 60 MPa Injection
pressure on macroscopic spray characteristics of ducted fuel injection under non-

14
F. Millo et al. Fuel 289 (2021) 119835

Pressure GDI System, SAE Technical Paper 2018-01- 0271, 2018, doi: 10.4271/ [30] Naber JD, Reitz RD. Modeling engine spray/wall impingement, SAE Technical
2018-01-0271. Paper 880107, 1988, doi: 10.4271/880107.
[19] Hung DLS, Harrington DL, Gandhi AH, Markle LE, Parrish SE, Shakal JS, Sayar H, [31] Musculus MPB, Kattke K. Entrainment waves in diesel jets. SAE Int J Engines 2009;
Cummings SD, Kramer JL. Gasoline fuel injector spray measurement and 2(1):1170–93. https://doi.org/10.4271/2009-01-1355.
characterization – a new SAE J2715 recommended practice. SAE Int J Fuels Lubr [32] Porter JL, Squyers RA. A Summary/Overview of Ejector Augmentor II – Technical
2009;1(1):534–48. https://doi.org/10.4271/2008-01-1068. Report, Off. Nav. Res. Rep. ATC #R-91100/9CR-47A, 1979.
[20] Richards KJ, Senecal PK, Pomraning E. CONVERGE 2.4. Madison, WI: Converg. Sci; [33] Rhim DR, Farrell PV. Characteristics of air flow surrounding non-evaporating
2018. transient diesel sprays, SAE Technical Paper 2000-01-2789, 2000, doi:10.4271/
[21] Amsden AA. KIVA-3V: A Block-Structured KIVA Program for Engines with Vertical 2000-01-2789.
or Canted Valves, LA Rep. LA-13313-MS; 1997. [34] Bruneaux G. Mixing process in high pressure diesel jets by normalized laser
[22] Orszag SA, Yakhot V, Flannery WS, Boysan F, Choudhury D, Maruzewski J, Patel B. induced exciplex fluorescence part I: Free jet, SAE Technical Paper 2005-01-2100,
Renormalization group modeling and turbulence simulations, Near-Wall Turbul. 2005, doi: 10.4271/2005-01-2100.
Flows 1031–1046, 1993. [35] Bruneaux G, Causse M, Omrane A. Air entrainment in diesel-like gas jet by
[23] Reitz RD, Diwakar R. Structure of high-pressure fuel sprays, SAE Technical Paper simultaneous flow velocity and fuel concentration measurements, comparison of
870598, 1987, doi: 10.4271/870598. free and wall impinging jet configurations. SAE Int J Engines 2012;5(2):76–93.
[24] Chaves H, Knapp M, Kubitzek A, Obermeier F, Schneider T. Experimental study of https://doi.org/10.4271/2011-01-1828.
cavitation in the nozzle hole of diesel injectors using transparent nozzles, SAE [36] Singh G, Sundararajan T, Shet USP. Entrainment and mixing studies for a variable
Technical Paper 950290, 1995, doi:10.4271/950290. density confined jet. Numer Heat Transf Part A Appl 1999;35(2):205–24. https://
[25] Reitz RD, Bracco FV. Mechanisms of breakup of round liquid jets. Encycl Fluid doi.org/10.1080/104077899275335.
Mech 1986;3:233–49. [37] Zeuch T, Moréac G, Ahmed SS, Mauss F. A comprehensive skeletal mechanism for
[26] Amsden AA, O’Rourke PJ, Butler TD, I.K., A Computer Program for Chemically the oxidation of n-heptane generated by chemistry-guided reduction. Combust
Reactive Flows with Sprays, Los Alamos Natl. Lab. Rep. LA-11560-MS; 1989. Flame 2008;155(4):651–74. https://doi.org/10.1016/j.
[27] Liu AB, Mather D, Reitz RD. Modeling the effects of drop drag and breakup on fuel combustflame.2008.05.007.
sprays, SAE Technical Paper 930072, 1993, doi :10.4271/930072. [38] Kazakov A, Frenklach M. Dynamic modeling of soot particle coagulation and
[28] O’Rourke PJ. Collective Drop Effects on Vaporizing Liquid Sprays, Ph.D. Thesis; aggregation: implementation with the method of moments and application to high-
1981. pressure laminar premixed flames. Combust Flame 1998;114(3-4):484–501.
[29] Schmidt DP, Rutland CJ. A new droplet collision algorithm. J Comput Phys 2000; https://doi.org/10.1016/S0010-2180(97)00322-2.
164(1):62–80. https://doi.org/10.1006/jcph.2000.6568.

15