Document [original]

Investigation of Fluidic Devices for Mixing

Enhancement for the Shockless Explosion

Combustion Process

Bernhard C. Bobusch1, Phillip Berndt2Christian Oliver Paschereit1, and

Rupert Klein2

1Technische Universit¨at Berlin, Institut f¨ur Str¨omungsmechanik und Ttechnische

Akustik, – Hermann-F¨ottinger-Institut –

M¨uller-Breslau-Str. 8; D-10623 Berlin, Germany,

bernhard.bobusch,[email protected]

2Freie Universit¨at Berlin, Department of Mathematics, Geophysical Fluid Dynamics,

Arnimallee 6; D-14195 Berlin, Germany,

pberndt,[email protected]

Abstract. Fuel-air mixing is a crucial process in low emission com-

bustion systems. A higher mixing quality leads to lower emissions and

higher combustion efficiencies. Especially for the innovative constant vol-

ume combustion processes ”Shockless Explosion Combustion” (SEC) the

mixing of fuel and air is an important parameter, since the whole com-

bustion process is triggered and controlled via the equivalence ratio. To

enhance the passive scalar mixing, fluidic oscillators are investigated and

compared to the standard jet in crossflow fuel injection configurations.

The mixing quality of the different geometries is assessed in a water

test-rig by making use of planar laser induced fluorescence. After a short

introduction to the SEC-process, the test-rig and the different injection

configurations are introduced. To verify whether the mixing quality is

sufficient for the SEC-process, a numerical investigation using the ex-

perimentally determined unmixedness is conducted. It is not only shown

that the fluidic oscillators are able to enhance the mixing quality and

create an independence of the mixing quality from the jet in crossflow

momentum, but it is also verified in a first numerical calculation that

the achieved mixing quality might be good enough for the Shockless

Explosion Combustion process.

Keywords: mixing ·constant volume combustion ·Shockless Explosion

Combustion ·numerical combustion

1 Introduction

High quality mixing is one of the most crucial parameters in modern low emis-

sion combustion systems. This becomes even more crucial regarding advanced

homogeneous pulsed (constant volume) combustion approaches such as the ho-

mogeneous charge compression ignition (HCCI) for internal combustion engines

layered fuel-air mixture

filling of inert

buffer volume

quasi-homogeneous self ignition

suction wave

exhaust gas

pressure wave

Fig. 1: Process cycle of the shockless explosion combustor

or the shockless explosion combustion (SEC) process for gas turbine combustion.

Both of these systems rely on the auto-ignition of a homogeneous fuel-air-charge

inside a combustor, thus, the preparation of a perfectly mixed combustion charge

is the most important factor for these systems.

For the SEC-process not only is a perfectly homogeneous fuel-air mixture

needed, but in addition, a stratification of the equivalence ratio in the com-

bustor is necessary. An overview of this process is shown in Fig. 1. Like other

constant volume combustion processes, the shockless explosion combustion pro-

cess is based on a periodic combustion process. A standing pressure wave is

established inside the combustion tube. The moment this pressure wave reduces

the pressure at the tube inlet below the plenum pressure, the tube is filled with

compressor air (left). After filling a volume with pure air, fuel is added to the

combustion air at the inlet position until around 40% of the tube is filled with

a combustible mixture (top). The air volume is needed to separate the hot flue

gases of the previous cycle from the fresh fuel-air mixture. The suction wave is

reflected from the open end of the tube and travels upstream to the inlet (right).

Due to the hot air from the compressor, the mixture undergoes auto-ignition

(bottom). The equivalence ratio inside the combustion chamber is adjusted in

a way that the ignition delay matches the residence time of the mixture in the

tube. Thus, the entire fuel-air volume undergoes homogeneous auto-ignition and

the mixture is burnt instantaneously without any shock waves. In addition, the

ignition delay is adjusted to match the oscillation period. This means that the

combustion of the mixture occurs simultaneously with the pressure wave raising

the pressure at the tube inlet. The pressure wave is amplified and travels to the

end of the combustion tube, where it is reflected as a suction wave and restarts

the process.

This process has by implication very high demands on the mixing quality:

1. The mixing needs to be very fast to avoid regions where the ignition delay

is too short.

2. The mixing must be homogeneous in the radial plane.

3. The axial stratification of equivalence ratio demands for very low mixing in

the axial direction.

4. The mixing must be independent from the fuel volume flow (i.e., jet in cross-

flow momentum) to have a broad range of possible equivalence ratios for base

load, part load, and idle operation.

To create this type of mixing several different jet in crossflow configurations

are investigated in this work. Namely the tested configurations are round jet,

rectangular jet, slit, and spatially oscillating jet in crossflow. The jet in cross-

flow configuration is the easiest, most robust, and very common way to mix two

fluids. Over the past decades, the round turbulent jet in crossflow was widely

investigated as a steady jet (e.g., [1–4]) or as a modulated/pulsed jet (e.g., [5–7]).

The rectangular jet in crossflow is not as thoroughly described as the round jet,

but recently this configuration was analyzed more extensive as well (see [8, 9]).

The slit injection lies somewhere between these two configurations. It can be

seen as a two-dimensional jet in crossflow with infinite depth. For the spatially

oscillating jet, only a few publications can be found. Nathan et al. [10] exten-

sively investigated jet parameters and their impact on the mixing for an inline

configuration. Several oscillating devices, mechanical and fluidic, were tested. It

was shown that the oscillation enhances the mixing especially close to the fuel

inlet. Arnaud and Paschereit [11] tested the enhancement of scalar mixing due

to a spatial oscillation of a jet in crossflow in a water test-rig. They confirmed

the findings of Nathan et al. [10] and in addition found out that the spatial

oscillation reduces the dependency of the mixing quality on the jet in crossflow

momentum. These advantages make the spatially oscillating jet in crossflow in-

jection very promising for the demands of the shockless explosion combustion

process. To assess and compare the mixing quality of the different configurations

they were investigated in a water test-rig making use of a fluorescent dye.

Known from literature (e.g., [12, 13]), the mixing of water with dye is a good

indicator for the fuel-air mixing in combustion systems. In addition, the results

of the experiments are used in a numerical calculation to see if the mixing quality

is high enough to assure a reliable SEC-process.

The remainder of the paper is organized as follows: First the experimental

setup, measurement technique, and evaluation methods are presented. This is

followed by the experimental results and the numerical investigations.

2 Experimental Setup

To investigate the mixing performance of several geometries, a test-rig was build

employing water as the fluid. The entire geometry is based on the planned SEC

combustion tube. A schematic overview of the rig can be found in Fig. 2. A

more detailed cut through the dye inlet section can be found in Fig. 3. The

test-rig consists of a main combustion tube made of acrylic glass with an inner

diameter of 40 mm and a length of 800 mm. Upstream of this tube a 25 mm valve

is installed to open and close the main mass flow. It is followed by a diffuser which

Fig. 2: Sketch of the test-rig.

increases the inner diameter up to the mentioned 40 mm of the main tube (see

Fig. 3). It was assured that the flow is attached to the surface of the diffuser at

all times.

Downstream of the diffuser interchangeable disks represent the fuel, in this

case dye, inlet. The disk is colored yellow for a better visibility in Fig. 3. Pictures

and details of the investigated geometries can be found later in this section. The

flow rate of the dye can be adjusted using an electrically driven proportional

valve, which was installed upstream of the inlet disc plenum.

Two separated water circuits were realized for the injection of the main and

the dye flow. Both circuits are equipped with pressure driven one-way valves in

the circuit to generate the needed pressure upstream of the two main valves.

To investigate the mixing quality and determine the unmixedness parame-

ters of the different injection geometries, Planar Laser Induced Fluorenscence

(PLIF) was employed in three radial planes at different axial positions down-

stream of the dye injection. The images were taken with a high-speed camera

Fig. 3: Cut through the fuel inlet section of the water test-rig.

from the downstream side of the tube. A shutter frequency of 500Hz minimizes

the blurring of the image due to long exposure times.

21 different injection geometries including round, rectangular, and spatially

oscillating (fluidic) jets, as well as slits were investigated. Different cross-section

areas were employed to create a broad range of jet in crossflow momentums for

all injector disks as well as different frequencies for the fluidic oscillators. The

jet in crossflow momentum Jis defined as the squared ratio between the jet flow

velocity wjand the main flow velocity w0.

J=wj

w02

(1)

Note that for all four different injector geometries (i.e., round, rectangle, oscil-

lating, slit) different sizes and amounts of injection ports were tested. This was

done to investigate the behavior of these injectors over a wide range of jet in

crossflow momentums. Since the volume flow range for all configurations was

approximately the same, the different sizes result in different ranges of J. This

broad range is needed in the SEC process due to the fact that the fuel mass flow is

changed within each injection cycle, as mentioned in the introduction. Addition-

ally for distinct injection ports two different numbers of ports were investigated.

A summary of the different geometries, including the main flow parameters, can

be found in Tab. 1. The Reynolds number of the main flow was Re = 31280 and

thus, a fully turbulent flow was assured. To give an idea of the different injection

geometries a three-dimensional model of four of the used injector disks is shown

in Fig. 4. For better visibility of the fluidic oscillator, a cut was made through

the material to show the geometry.

The investigated spatially oscillating jets were created using fluidic oscilla-

tors. From literature [11], it is known that these devices can enhance the spa-

Table 1: Investigated injector geometries including the main parameters. All

units in mm or mm2respectively.

Disk Outlet geom. Outlets Type dhAOutlet Rejet;min Rejet;max J

1 round 13 hole 1.40 20.01 4479 11143 16.7-103.6

2 round 13 hole 1.90 36.86 3542 8371 5.7-34.2

3 round 13 hole 2.40 58.81 3059 7519 2.7-16.0

4 round 7 hole 2.00 21.99 6249 14485 13.2-82.4

5 round 7 hole 2.60 37.17 4916 11361 5.8-31.2

6 round 7 hole 3.30 59.87 3787 9984 2.2-15.0

7 rectangle 13 hole 1.26 16.33 3909 9493 15.8-93.4

8 rectangle 13 hole 1.69 21.99 3359 7138 6.4-32.6

9 rectangle 13 hole 2.00 26.06 2548 5732 2.6-13.4

10 rectangle 7 hole 1.73 12.09 5574 12668 14.1-87.9

11 rectangle 7 hole 2.29 16.06 4183 9983 4.5-31.0

12 rectangle 7 hole 2.71 18.97 3308 8127 2.4-14.7

13 rectangle 13 fluidic osc. 1.31 16.97 3430 8092 11.2-62.5

14 rectangle 13 fluidic osc. 1.78 23.18 2943 7223 4.4-26.7

15 rectangle 13 fluidic osc. 2.17 28.24 1889 5618 1.2-10.9

16 rectangle 7 fluidic osc. 1.79 12.54 4098 11088 8.6-62.6

17 rectangle 7 fluidic osc. 2.40 16.83 3303 9726 3.1-26.7

18 rectangle 7 fluidic osc. 2.87 20.09 2516 7828 1.3-12.2

19 slit 1 — 0.40 50.19 893 2728 8.2-76.3

20 slit 1 — 0.60 75.22 867 2701 3.4-33.3

21 slit 1 — 0.80 100.21 891 2674 2.0-18.4

tial mixing, especially close to the injection position. The oscillatores used were

described in detail in [14]. The oscillatory parameters of these devices were de-

termined by numerical simulations, which were carried out with a validated nu-

merical model [15]. The oscillation frequency was calculated from a time history

plot of the velocity inside the mixing chamber.

For each of these 21 configurations, 8 fuel flow rates were investigated at

the mentioned three axial positions. Recording 1632 snapshots for each of these

tests over 822,000 pictures were taken in total. Each picture was corrected for

background reflections and normalized using a homogeneous picture with the

maximum dye concentration. It was assured that the dye concentration was

low enough to be well within the linear regime of the fluorescence intensity.

From the normalized images containing pixel values between 0 and 1, the two

unmixedness paramters Uxand Utwere calculated, which represent the spatial

and the temporal mixing quality repectively. Based on the work of Danckwerts

[16], the parameters are defined as follows:

Ux=σ2

σ2

=σ2

C∗

∞(1 −C∗

∞),(2)

Fig. 4: Colored CAD-Models of the four different injector types (from left to

right): round holes (red), rectangular holes (green), slit (blue), and fluidic oscil-

lator (yellow).

where σ2

x=1

Ni−1

i=1 C∗(i)−C∗

∞2,(3)

and σ2

xdenotes the mixture variance of the temporally averaged concentration

field C∗(i), which is recorded by the Nicamera pixels and is defined in Eq. 3.

The variance immediately before the start of the mixing process (σ2

0) is calcu-

lated from the dye concentration C∗

∞. The spatial unmixedness parameter Ux

represents the average spatial mixing quality and gives a value between 0 (per-

fectly mixed) and 1 (not mixed at all). To investigate the temporal unmixedness,

the parameter Utis used, which employs the variance of all the concentration

records. It is defined as:

Ut=σ2

σ2

=σ2

C∗

∞(1 −C∗

∞),(4)

where σ2

t=1

NiNt−1

i=1

(C∗(i, t)−C∗

∞)2.(5)

Ntis the number of snapshots recorded during one measurement sequence (1632

in this work). According to Eq. 5, Utcaptures both spatial and temporal fluc-

tuations in the concentration. These unmixedness parameters can be used as a

first measure to evaluate the mixing quality of a given injector geometry.

For the transfer of the experimental results to the numerical investigations, an

additional analysis method was employed, which is presented in a next section.

0.6

0.8

1.2

1.4

Fig. 5: Averaged, normalized spatial concentration field for disk 13 (upper half)

and disk 1 (lower half).

3 Experimental Results

As a first impression of the results of the experiments, two pictures of the av-

eraged and normalized concentration field Cin the first measurement plane are

shown in Fig. 5. For a better visibility, only one half of the pictures is shown

for the two different configurations, whereby one of the two pictures was flipped

upside-down allowing them to be plotted in one figure. The top half of the pic-

ture corresponds to disk 13 (see Tab. 1), an example for a high mixing quality.

The lower half of the picture corresponds to disk 1 which shows a rather poor

mixing quality due to the different injection geometry. This is clearly visible in

the inhomogeneities of the concentration field. In addition to the pictures the

concentration histograms for the two measurements are shown in Fig. 6. It is

visible that the distribution of the pixel counts has a much sharper peak for disk

13 (see Fig. 6a) than for disk 1 (Fig. 6b). The vast amount of data (see Tab. 1)

demands the selection of the injection geometries for further investigations by

making use of the defined umixedness parameters. For the SEC-process, a fast,

reliable and high-quality mixing is necessary in the radial plane, while almost no

mixing is desired in the axial direction. From the steady injection point of view,

the latter is unquantifiable since no changes in the dye volume flow are present

in the current work. Accordingly, the spatial unmixedness parameter Uxis used

to investigate the overall mixing parameters of the different geometries.

The range of the spatial unmixedness parameter for the different injection

geometries is plotted in Fig. 7 for the first measurement plane 50 mm down-

stream of the injection point (see Fig. 2). For this plot, all the geometries of the

same type are grouped together to show the overall performance of such a con-

figuration, since the different injection parameters make a plot of the individual

0.6 0.8 1 1.2 1.4

0.2

0.4

0.6

0.8

1·104

C/C

Pixel Counts

(a) Disk 13 — fluidic osc. dh= 1.31 mm

0.6 0.8 1 1.2 1.4

1,000

2,000

3,000

C/C

Pixel Counts

(b) Disk 1 — holes; dh= 1.4 mm

Fig. 6: Histogram of normalized concentration for (a) disk 13 and (b) disk 1 at

x= 50 mm.

measurements rather unclear. Accordingly, disks 1 −6 are summed up in the

black range, 7 −12 in the red, 13−18 in the green, and finally the slits (19−21)

in the blue range. As one can see in this plot the slits and the round holes give a

lower mixing quality than the rectangular holes and the fluidic oscillators. The

rectangular holes show a better mixing quality (lower unmixedness parameter)

on the lower side of the range than the fluidic oscillators but a worse mixing

quality than the fluidic oscillators for the upper boundary of the range. Even

though the mixing quality is slightly better for the rectangular holes in some

cases, they fail for other injection situations. However, to assure the reliability

of the mixing system over a wide range of equivalence ratios, fuel mixtures, and

power levels, the fluidic oscillators show a more desirable behavior. Due to these

advantages, the fluidic oscillators are chosen to be analyzed in more detail in this

work. To gain a deeper insight into the mixing mechanisms of the oscillating jet,

two different sizes of fluidic oscillators (disks 13 −15 and disks 16 −18) have

been designed. Since the range of flow rates is defined by the dye pump, the two

different oscillators were manufactured with 3 different aspect ratios each. From

literature, it is known that the frequency of these oscillators scales linearly with

the volumetric flow rate [14, 17]. Accordingly, the dependency of the frequency

on the jet in crossflow momentum ratio Jis of quadratic nature. To show the

frequency ranges, Jis plotted against the frequency in Fig. 8. It is clearly visible

that even though the jet in crossflow momentum Jis in the same range for both

round rect fluidic osc. slit

8·10−3

Fig. 7: Spatial unmixedness parameter Uxat measurement plane x= 50 mm for

the four diffferent geometries: round jet, rectangular jet, fluidic oscillator, and

slit).

20 40 60 80

f in Hz

Disk 13

Disk 14

Disk 15

Disk 16

Disk 17

Disk18

Fig. 8: Frequency of the fluidic oscillators for the different geometries and differ-

ent values of J.

sizes, the oscillation frequency is different. Disks 16 −18, which consist of seven

larger fluidic devices give a lower frequency range than disks 13 −15.

Keeping this in mind, the mixing performances of disks 13 and 16 were com-

pared. For the same Jand, thus, the same volume flow, Reynolds number, and jet

penetration, they create different frequencies. To see the impact of this change

in injection geometry, the spatial unmixedness is plotted against Jin Fig. 9.

Two main conclusions can be made from these results. First, the mixing quality

of the smaller oscillators with a higher frequency is better close to the point

of injection. This might also be affected by the fact that for the smaller fluidic

oscillators, 13 oscillating jets are present, while only 7 were employed for disk 16.

However, it can be stated that a higher oscillation frequency may be favorable

for a faster mixing process. The second observation from Fig. 9 is that the mixing

quality of both configurations converge as the mixture is convected downstream.

10 20 30 40 50 60

0.5

·10−3

Disk 13; x=50mm

Disk 13; x=100mm

Disk 13; x=200mm

Disk 16; x=50mm

Disk 16; x=100mm

Disk 16; x=200mm

Fig. 9: Spatial unmixedness versus Jfor disks 13 and 16 for the three measure-

ment planes.

20 40 60 80 100

6·10−3

Disk 1

Disk 7

Disk 13

Fig. 10: Spatial unmixedness versus Jfor disks 1, 7, and 13 at x= 50 mm.

This confirms the findings of Lacarelle et al. [11] that the mixing enhancement

of the fluidic oscillators is most pronounced close to the injection plane.

Now, knowing the influence of the oscillating frequency, the influence of the

jet in crossflow momentum can be analyzed. To do so the spatial unmixedness

parameters of disks 1, 7, and 13 are plotted with respect to Jin Fig. 10. These

disks have been selected because they have a similar hydraulic diameter and rep-

resent three different injection geometries. As already mentioned, the injection

geometry for the SEC-process needs to be as independent from the jet in cross-

flow momentum as possible. From Fig. 10, it can be deduced that this desired

characteristic is best represented by the fluidic oscillators (disk 13). Even though

the rectangular holes (disk 7) have a higher mixing quality for some values of J,

they show a very pronounced dependency on the jet in crossflow momentum. An

even stronger dependency can be seen for the circular holes of disk 1. Keeping in

mind that the frequency increases with Jand that a higher frequency leads to

0 10 20 30 40

0.2

0.4

0.6

0.8

Delta x in mm

Rdiagonal

Rt·u

Fig. 11: Correlation coefficients for three spatial directions plus the axial direction

calculated using the Taylor hypothesis.

better mixing, the independence of the fluidic injection from Jis likely to stem

from the increase in the oscillating frequency. These findings again confirm the

results from Lacarelle et al. [11], who stated that fluidic oscillators may not only

enhance the passive scalar mixing between two fluids but also make that mixing

quality less dependent from the jet in crossflow momentum.

Following these observations, disk 13 is the best injection configuration of

the investigated geometries. In addition to the mixing quality of the different

geometries, the mixing parameters of the best configuration were included in a

numerical model of the SEC process to investigate if the mixing quality of the

given setup is good enough to assure a reliable Shockless Explosion Combustion.

The data of most promising configuration (injection disk 13) was selected

to be investigated by the numerical simulations. Since the most upstream auto-

ignition location (i.e., location of the combustible mixture) in the combustion

tube is at least located a distinct distance downstream of the injection point,

the data in the third measurement plane (x= 200 mm) was analyzed. From

the temporally averaged picture, the spatial variance was calculated according

toEq. 3. Since no data was available in the axial direction, except the individually

taken radial planes, the Taylor hypothesis was employed to gain an insight into

the size of the axial mixing structures of the flow.

To use the Taylor hypothesis, defined as

∂C

∂t =−u∂C

∂x x=x0

,(6)

which states that spatial fluctuations in an homogeneously turbulent field, where

the streamwise turbulent fluctuations are low (u0u), can be calculated from

temporal fluctuations at a given coordinate, using the mean streamwise velocity

u. To assure the homogeneous character of the fluctuations in the scalar mixing

field, the temporal and spatial correlations where compared. Since all of the

correlation curves are reasonably close to each other (see Fig. 11), the use of

the Taylor hypothesis is justified. Accordingly, from the individual snapshots of

concentration at different times in the measurement plane, an integral length

scale of the mixing fluctuations was calculated as

Λ=

∆x∗

R(∆x)d(∆x),(7)

where ∆x∗is defined as the first intersection of R(∆x) and the x-axis. This length

scale of 9.7 mm was then used together with the spatial variance to create an

equivalence ratio stratification where the worst possible mixing deviations were

imposed. This stratification was then used in the numerical simulation to assess

the reliability of the SEC-process.

4 Numerical Investigation

The effect of Gaussian noise with the observed standard deviation from the

experiment, σ= 2.424·10−3, on the homogeneous ignition of a stratified mixture

has been numerically investigated by the simulation of the dynamics of a mixture

with perturbed equivalence ratios.

The exact numerical setup is based on the internal state of a pipe in which

the SEC process is about to enter the combustion stage; prior to this state, gas

flowed from the pipe’s inlet at x= 0 to a position xin a time τflow that is

assumed to be given via a constant inflow velocity u0. This, in turn, is given

by the geometry of the pipe and fuel properties or can be arbitrarily chosen for

single shot ignitions, respectively. Here, it is chosen such that the gas can flow

40 cm within the range of the ignition delay times. Since the ignition shall occur

homogeneously, the ignition delay time of the gas that ends up at position x

must be chosen to be τ0+τflow, such that the remaining ignition delay time τat

the beginning of the simulation is τ0everywhere. This choice is done by altering

the equivalence ratio Φbetween stoichiometric (1.0) and lean (0.5) values. In the

numerical setup, perfect advection without diffusion is assumed, such that the

initial values can be calculated by selecting Φsuch that τ=τ0+τflow and then

solving the mixture in an isochoric 0D reactor for τflow +ητ0, where ηis some

fraction smaller than unity and chosen such that the assumption that the fluid

is only advected is not violated due to the pressure rise of the reaction. For this

simulation, η= 0.7 was chosen.

For the chemical kinetics, the latest fuel suggestion and associated reaction

mechanism for a 3 bar test rig, which will also be presented at AFCC 2014 [18],

namely a composition of

CH3OCH3+ 1.1 H2+ 0.8 CH4(8)

in air, has been used. The initial temperature was chosen to be 787.5 K.

Within the setup described above, the choices of Φwere perturbed with

different noise levels on equidistant points with 9.7 mm spacing and linear in-

terpolation in between. The resulting stratification was then placed into a 1D

domain as initial conditions for a fluid dynamic simulation that solves the Eu-

ler equations with a thermally perfect equation of state and the above reaction

mechanism. The in-house code is based on a second-order MUSCL extension of

the HLL approximate Riemann solver [19] and uses Strang splitting to couple

with an isochoric 0D kinetics code. Reflecting wall boundary conditions have

been employed for the left boundary, and a fixed pressure plenum at 3 bar for

the right one.

If the full range of equivalence ratios is used and an extremal (deterministic)

noise of alternating 3σand −3σis applied, the different sensitivities at the

extremal Φvalues are too high and a detonation develops near Φ= 1, traveling

towards the Φ= 0.5 boundary. The current quality of mixing is, therefore,

insufficient and requires improvement. To assess the possibilities with the current

mixing and estimate the required variation to allow usage of the full Φrange,

further simulations were conducted. They show two possible ways to achieve

homogeneous combustion:

By limiting the choice of Φto either 0.5–0.75 or 0.75 – 1.0, the sensitivity

is sufficiently reduced to ensure homogeneous combustion. Test case 1 in Fig. 12

shows the simulation results for the upper range. From the ignition delay plot,

one can observe that the ignition takes place throughout the domain in a 0.1 ms

window. In the heat map, the transition from blue to yellow/orange is the com-

bustion. The steeper lines are pressure waves in the burnt gas. The results for

the lower range of Φvalues are slightly better, with the ignition time varying by

only half the upper range’s value, but otherwise similar.

Reducing the variance by a factor of 2

3is mathematically equivalent to a

reduction of the confidence interval to a 2σenvironment. A simulation with

reduced variance can, therefore, be equivalently interpreted as being an improved

mixing quality or accepting 5% non-SEC ignitions. In both cases, the combustion

is almost completely homogeneous, as can be seen in test case 2 of Fig. 12.

5 Conclusion

The innovative constant volume combuston process SEC has very high and spe-

cific demands on the fuel-air mixing. In order to create the desired levels of

mixing quality, several geometries were investigated in a water test-rig. They

were analyzed employing planar laser induced fluorescence and evaluated by the

unmixedness parameter Ux.

It was shown that spatially oscillating jets in crossflow created by fluidic os-

cillators are able to get close to the needed mixing characteristics. In contrast to

a slit injection, the round, or the rectangular jet in crossflow, fluidic oscillators

allow the creation of a very high mixing quality already close to the fuel injection

plane. In addition, the mixing quality is independent from the jet in crossflow

momentum for these types of injectors. The results indicate that this is likely to

stem from the increasing oscillation frequency of the fluidic oscillators. A higher

jet in crossflow momentum leads to a higher volumetric flow in the oscillator

and, thus, a higher oscillating frequency. By comparing two different sizes of

1 bar 16 bar

0.0465 0.0470 0.0475 0.0480 0.0485 0.0490

500

1000

1500

2000

2500

3000

T[K]

t[s]

(a) Testcase 1: Φlimited to 0.75 – 1.0.

0 bar 14.5 bar

0.036 0.037 0.038 0.039 0.040 0.041 0.042 0.043 0.044

500

1000

1500

2000

2500

3000

T[K]

t[s]

(b) Testcase 2: Noise reduced to ±2σperturbations.

Fig. 12: Ignitions for the different simulations. The heatmaps show pressure evo-

lution, the other plot shows the temperature as a function of time overlain for

each of the computational cells.

fluidic oscillators, it was possible to show that higher frequencies are favorable

for the mixing process. However, since higher frequencies lead to smaller geome-

tries for the fluidic oscillators the pressure loss of the fuel injection system will

increase and might contradict the advantages of a higher mixing quality. Hence,

the optimal oscillating frequency must to be identified for every new injection

configuration.

In order to validate the achieved mixing quality in the highly demanding

SEC-process, a numerical simulation was conducted. Based on these calculations

it was possible to show that for an ideal filling charge of the SEC-combustor,

the identified perturbations in the equivalence ratio are likely low enough to

assure a reliable homogeneous auto-ignition. The experimental verification of

these findings and the further investigation of the injection and ignition process

will be the scope of future work. However, the presented work gives confidence

that the proposed SEC-process can be achieved reliably and efficiently.

Acknowledgement

The authors gratefully acknowledge support by the Deutsche Forschungsgemein-

schaft (DFG) as part of collaborative research center SFB 1029 ”Substantial

efficiency increase in gas turbines through direct use of coupled unsteady com-

bustion and flow dynamics”. In addition, the authors would like to thank the

CONFET for assistance in the lab and helpful discussions.

References

1. Keffer, J., Baines, W.: The round turbulent jet in a cross-wind. Journal of Fluid

Mechanics 15(4) (1963) 481–496

2. Kamotani, Y., Greber, I.: Experiments on a turbulent jet in a cross flow. AIAA

journal 10(11) (1972) 1425–1429

3. Andreopoulos, J.: On the structure of jets in a crossflow. Journal of Fluid Me-

chanics 157 (1985) 163–197

4. Smith, S.H., Mungal, M.G.: Mixing, structure and scaling of the jet in crossflow.

Journal of Fluid Mechanics 357 (February 1998) 83–122

5. Eroglu, A., Breidenthal, R.: Structure, penetration, and mixing of pulsed jets in

crossflow. AIAA journal 39(3) (2001)

6. Johari, H., Pacheco-Tougas, M., Hermanson, J.: Penetration and Mixing of Fully

Modulated Turbulent Jets in Crossflow. AIAA Journal 37(7) (July 1999) 842–850

7. Johari, H.: Scaling of Fully Pulsed Jets in Crossflow. AIAA Journal 44(11) (Novem-

ber 2006) 2719–2725

8. Weston, R., Thames, F.: Properties of aspect-ratio-4.0 rectangular jets in a sub-

sonic crossflow. Journal of Aircraft 16(10) (1979) 701–707

9. Plesniak, M.W., Cusano, D.M.: Scalar mixing in a confined rectangular jet in

crossflow. Journal of Fluid Mechanics 524 (February 2005) 1–45

10. Nathan, G., Mi, J., Alwahabi, Z., Newbold, G., Nobes, D.: Impacts of a jet’s

exit flow pattern on mixing and combustion performance. Progress in Energy and

Combustion Science 32(5-6) (September 2006) 496–538

11. Lacarelle, A., Paschereit, C.O.: Increasing the Passive Scalar Mixing Quality of

Jets in Crossflow With Fluidics Actuators. Journal of Engineering for Gas Turbines

and Power 134(2) (2012) 021503

12. Clarke, A., Gerrard, A., Holliday, L.: Some experiences in gas turbine combustion

chamber practice using water flow visualization techniques. Symposium (Interna-

tional) on Combustion 9(1) (1963) 878–891

13. Lacarelle, A., Sebastian, G., Paschereit, C.O.: A Quantitative Link Between Cold-

Flowscalar Unmixness and NOX Emissions in a Conical Premixed Burner. In:

ASME Turbo Expo 2010. Number X (2010) ASME GT2010–23132

14. Bobusch, B.C., Woszidlo, R., Bergada, J.M., Nayeri, C.N., Paschereit, C.O.: Exper-

imental study of the internal flow structures inside a fluidic oscillator. Experiments

in Fluids 54(6) (June 2013) 1559

15. Kr¨uger, O., Bobusch, B.C., Woszidlo, R., Paschereit, C.O.: Numerical Modeling

and Validation of the Flow in a Fluidic Oscillator. In: 21st AIAA Computational

Fluid Dynamics Conference, Reston, Virginia, American Institute of Aeronautics

and Astronautics (June 2013)

16. Danckwerts, P.: The definition and measurement of some characteristics of mix-

tures. Applied Scientific Research, Section A 3(1952) 279–296

17. Bobusch, B.C., Woszidlo, R., Kr¨uger, O., Paschereit, C.O.: Numerical Investiga-

tions on Geometric Parameters Affecting the Oscillation Properties of a Fluidic

Oscillator. Proceedings of the 21st AIAA Computational Fluid Dynamics Confer-

ence (2013)

18. Cai, L., Pitsch, H.: Tailoring fuels for shockless explosion engine (preliminary title).

Proceedings of AFCC (2014)

19. Harten, A., Lax, P.D., Van Leer, B.: On upstream differencing and godunov-type

schemes for hyperbolic conservation laws. SIAM Review (1983) 35–61