Autoignition in stratified mixtures for pressure gain combustion [original]

Fatma Cansu Yücel, Fabian Habicht, Myles D. Bohon, Christian

Oliver Paschereit

Autoignition in stratified mixtures for pressure gain

combustion

Open Access via institutional repository of Technische Universität Berlin

Document type

Journal article | Accepted version

(i. e. final author-created version that incorporates referee comments and is the version accepted for

publication; also known as: Author’s Accepted Manuscript (AAM), Final Draft, Postprint)

This version is available at

https://doi.org/10.14279/depositonce-15713

Citation details

Yücel, F. C., Habicht, F., Bohon, M. D., & Paschereit, C. O. (2021). Autoignition in stratified mixtures for

pressure gain combustion. In Proceedings of the Combustion Institute (Vol. 38, Issue 3, pp. 3815–3823).

Elsevier BV. https://doi.org/10.1016/j.proci.2020.07.108.

This work is protected by copyright and/or related rights. You are free to use this work in any way permitted by

the copyright and related rights legislation that applies to your usage. For other uses, you must obtain

permission from the rights-holder(s).

Autoignition in Stratified Mixtures for Pressure Gain Combustion

Fatma Cansu Y¨

ucela,∗, Fabian Habichta, Myles Bohona, Christian Oliver Paschereita

aTechnical University of Berlin, Institute of Fluid Dynamics and Technical Acoustics, M¨uller-Breslau-Strasse 8, Berlin and 10623, Germany

Abstract

The reliable generation of quasi-homogeneous autoignition inside a combustor fed by a continuous air flow would

represent a milestone in realizing pressure gain combustion in gas turbines. In this work, the ignition distribution in-

side a stratified fuel–air mixture is analyzed. The ability of precise and reproducible injection of a desired fuel profile

inside a convecting air flow is verified by applying tunable diode laser absorption spectroscopy in non-reacting mea-

surements. High-speed, static pressure sensors and ionization probes allow for simultaneous detection of the flame and

pressure rise at several axial positions in reactive measurements with dimethyl ether as fuel. A second, exchangeable

combustion tube enables optical observation of OH∗intensity in combination with pressure measurements. Experi-

ments with three arbitrary fuel profiles show a set of ignition distributions that vary in shape, homogeneity, and the

number of simultaneous autoignition events. Although the measurements show notable variation, a significant and

reproducible influence of the fuel injection on the ignition distribution is observed. Results show that uniform au-

toignition leads to a coupling of the reaction front with the pressure rise and, therefore, induces a greater aerodynamic

constraint than non-uniform ignition distributions, which are dominated by propagating deflagration fronts.

Keywords:

homogeneous autoignition, fuel stratification, pressure gain combustion, shockless explosion combustor

∗Corresponding author:

Email address: [email protected]

(Fatma Cansu Y¨

ucel)

Preprint submitted to Proceedings of the Combustion Institute May 16, 2022

1. Introduction

Replacing the conventional isobaric combustion of the

gas turbine cycle by pressure gain combustion is a

promising concept to achieve improvements in cycle ef-

ficiency. Different approaches have been investigated,

such as pulse detonation combustion (PDC) [1] and ro-

tation detonation combustion (RDC) [2]. Both concepts

utilize propagating detonation wave(s), the high tem-

peratures and pressures of which present significant en-

gineering challenges. An alternative approach, called

shockless explosion combustion (SEC) [3], overcomes

these challenges by using quasi-homogeneous autoigni-

tion to achieve approximately constant volume combus-

tion without the presence of a detonation wave or me-

chanical constraints.

Homogeneous autoignition occurring in well mixed

combustible mixtures was originally referred to as ther-

mal explosion by Zel’dovich [4] as a concept of sponta-

neous flames, leading to an increase in pressure sim-

ilar to constant volume combustion. However, non-

uniformities in the mixture, such as variations in tem-

perature, pressure or equivalence ratio, can cause devi-

ations in ignition delay time and result in a propagating

autoignition front instead. The propagation velocity of

this autoignition front uai is inversely proportional to the

spatial gradient of the ignition delay time τai. Assum-

ing constant temperature Tand pressure pacross the

mixture, the ignition delay time τai is a function of the

equivalence ratio ϕsuch that uai can be expressed as:

uai = ∂τai

∂x!−1

= ∂τai

∂ϕ

∂x!−1

.(1)

Comparing uai to the speed of sound aleads to the di-

mensionless parameter ξ, where

ξ=a

uai

=a∂τai

∂ϕ

∂x.(2)

This allows for the description of different combustion

modes [4, 5]. A subsonic propagation of the reaction

front occurs when ξ > 1. For ξ=1 the propagation

velocity of the autoignition front is equal to the speed

of sound, allowing amplification by coupling and en-

abling deflagration-to-detonation transition (DDT). The

ideal process of thermal explosion, which occurs for

ξ=0, is not practical in experiments due to unavoid-

able perturbations of the initial conditions and mixture

inhomogeneity, leading to gradients in reactivity. How-

ever, ξ < 1 results in a quasi-homogeneous autoignition

leading to an approximate constant volume combustion

while avoiding DDT. Similar behavior in pressure rise

quasi-homogeneous

refilling of the tube

burned gas

stratified fuel-air mixture

burned gas

air-buffer quasi-homogeneous autoignition

p>p0

τ=const.

expansion wave

pressure wave

autoignition

Figure 1: Sketch of the single tube SEC test rig

can be observed in the case of multiple separate igni-

tion sources that ignite quasi-simultaneously [6]. This

regime of quasi-homogeneous autoignition and/or many

distributed ignition points is the objective for the imple-

mentation of SEC.

Along these principles, the realization of homoge-

neous charge compression ignition (HCCI) has been in-

vestigated intensively in the past [7]. Preventing engine

knock is a major challenge in HCCI and has been ad-

dressed by applying different methods, such as modifi-

cation of oxidizer and fuel properties, equivalence ra-

tio, exhaust gas recirculation, or engine parameters [8].

However, the implementation of this concept in a gas

turbine cycle for pressure gain combustion is a new ap-

proach.

The SEC is based on a periodic combustion process

as sketched in Fig. 1. The cycle begins with a strati-

fied, autoignitable fuel–air mixture throughout the com-

bustor (Fig. 1a). This stratification has been tailored

to compensate for the gradient in residence time, re-

sulting in a quasi-homogeneous autoignition (Fig. 1b).

The pressure rise during combustion induces a pressure

wave propagating downstream which is reflected as an

expansion wave when reaching the acoustically open

combustor outlet (Fig. 1c). When this wave reaches the

combustor inlet, the refilling process begins (Fig. 1d)

and the cycle restarts.

The objective of this work is to investigate the igni-

tion processes within a stratified fuel–air mixture. First,

the ability of injecting a defined mixture profile in a con-

vecting air flow is analyzed. Subsequently, the homo-

geneity of measured autoignition times and pressure rise

as a function of the fuel stratification are examined. Fi-

nally, the processes of autoignition homogeneity, flame

propagation, and process variability are studied in high-

speed images of OH∗chemiluminescence.

injection station

vaporizer

pressure regulator

FAT1P1P2P3P4P5

I1I2I3I4I5I6I7

preheater restriction convection tube combustor exhaust tube

solenoid valve

optical access combustor

1 2 34

Figure 2: Sketch of the test rig. Sensors: low-speed, static pressure sensors (FA, FF), thermocouples (T1, T2), high-speed, static pressure sensors

(P1–P5), ionization probes (I1–I8). The inset subfigure b) shows the exchangeable version of the combustor tube with optical access.

2. Experimental Setup and Measurement Procedure

A sketch of the test rig for the experimental investiga-

tion of autoignition of an axially stratified fuel–air mix-

ture in a convecting flow is shown in Fig. 2. The test

rig is composed of several sections, including reactant

injection, convection (0.5 m), combustor (0.5 m), and

exhaust (1 m) sections. All sections have an inner di-

ameter of 40 mm. The rig was originally designed by

Bobusch et al. [3, 9] and later used by Reichel et al. [10].

However, in these works, reproducibility was limited

and consistent homogeneous autoignition was difficult

to achieve. The control of the injection process has since

been improved by Y¨

ucel et al. [11].

In the reacting cases, a preheater is used to raise the

temperature of the constant air flow to 1023 K measured

at T1. Downstream of the preheater, the air flow is

forced through a restriction in order to prevent back-

flow of hot gases into the preheater due to ignition.

Fuel is injected via ten radial ports with 1 mm diam-

eter each, which are individually controlled by high-

speed solenoid valves (Staiger VA 204-716). A dome-

loaded pressure regulator (Swagelok RD6) is installed

upstream of the injection station to control the fuel

supply pressure. Two static pressure sensors FAand

FF(Festo SPTW) are installed to monitor the air and

fuel supply pressures. The fuel injection duration is

∆tinj =50 ms, and is divided into ten time windows,

each with a length of 5 ms. The number of open valves

is individually set for each time window defining the in-

jected fuel profile.

The modular setup allows for exchanging the stain-

less steel combustor for a quartz tube (Fig. 2b) in order

to achieve optical access. This configuration is used for

fuel concentration measurements and OH* chemilumi-

nescence imaging of the ignition distribution.

2.1. Fuel Concentration Measurements

Fuel concentration measurements using near-infrared

tunable diode laser absorption spectroscopy (TDLAS)

are conducted as proposed by Li et al. [12]. These

measurements are used to validate the control of the

injection geometry to achieve a desired mixture pro-

file within a defined time frame. This technique has

been used previously for time-resolved fuel concentra-

tion measurements in a similar configuration [10, 13].

While the combustion experiments in this work are con-

ducted with dimethyl ether (DME) as fuel, the concen-

tration measurements are done with methane to match

the absorption features around a wavelength of 1654 nm

utilizing the available laser. In the scope of this work,

we expect the variation in the injected mixture fraction

profile to be primarily controlled by turbulent diffusion

and mixing. Since turbulent fluctuations scale with the

Reynolds number, it is considered as the dominant mix-

ing parameter rather than molecular diffusion (which

is of the same order for both fuels). However, when

comparing the Reynolds numbers for the non-reacting

(approx. 48000) to reacting (approx. 6000) cases, it is

expected that the non-reacting cases will exhibit much

greater turbulent diffusion and a blurring of the mixture

profile. Lastly, the residence time is kept constant for all

measurements by matching the flow velocities, allowing

an equal amount of time to diffuse. Considering this, it

is reasonable to conclude that for the reacting cases the

resulting gradients are steeper than TDLAS measure-

ments reveal. While this prevents quantifying the exact

local equivalence ratio for the reacting DME cases, it

does allow for a qualitative measure of the reproducibil-

ity and accuracy of the injection scheme.

2.2. Reactive Measurements

For reactive measurements, the initial temperature is

monitored via two Type-K thermocouples. Five water-

cooled, high-speed pressure sensors are installed in the

combustor with a distance of 100 mm to record the static

pressure variation. The flame is detected via 8 ioniza-

tion probes that are mounted in the combustor and the

exhaust tube. Figure 2 shows the naming convention for

each sensor.

The temperature at the injection station remains con-

stant during the measurements. At the beginning of each

measurement, a gradient in wall temperature of about

50 K between sensors T1and T2is observed. Heat-

ing during the run increases T2by approximately 50 K.

However, the measurement data show no correlation be-

tween the ignition time throughout the measurement re-

gion and the measured wall temperature. Therefore, it

is reasonable to assume the impact of the transient wall

temperature on the ignition process to be negligible.

DME is used as fuel resulting in ignition times in

the range of 60 ms to 80 ms for the applied conditions

(p=1 atm, T=1023 K and 1 ≤ϕ≤2). This assures

autoignition of the convecting mixture inside the com-

bustor. The fuel supply pressure is FF=5.7 bar and the

equivalence ratio is controllable from ϕ=0 (all valves

closed) to ϕ=2 (ten valves open). The average fuel

mass flow rate was measured under steady state con-

ditions using a Coriolis mass flow meter. To assure a

gaseous state, the fuel is vaporized and guided through

a heated pipe (330 K) before injection.

The ignition behavior of DME is characterized by a

negative temperature coefficient (NTC) region, which

is studied in more detail by Burke et al. under high

pressure conditions [14]. However, calculating the rel-

evant ignition delay times with Cantera [15] for a zero-

dimensional constant volume reactor using the mech-

anism AramchoMech2.0 that has been validated for

DME-kinetics in previous works, reveal that all tests

were conducted outside the NTC region of DME. Op-

erating in the NTC region of DME would require ac-

counting for additional non-linear behavior of DME au-

toignition, and is therefore avoided.

The optically accessible section is composed of a se-

ries of four quartz tubes, each 120 mm long, supported

by stainless steel flanges fitted with one pressure sen-

sor each. An optical band-pass filter (CWL =310 nm,

FWHM =10 nm), an intensifier (Lambert Instruments

HiCATT) and a high-speed camera (Photron Fastcam

SA-Z) are used to detect the reaction zones by light

emission of OH∗intensity. The recorded high-speed im-

ages allow for observation of the ignition distribution at

87500 fps and a spatial resolution of 2.7 px/mm.

3. Results and Discussion

The results will be broken into two sections. First, the

control of the fuel injection profile will be investigated,

and three example contours will be discussed. The sec-

ond section will then examine the autoignition charac-

teristics of these profiles, focusing on the pressure rise

(as representing aerodynamic confinement) and corre-

0 20 40 60

tin ms

number of open valves

a) valve commands

0 20 40 60 80

tin ms

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

XCH4

b) measured injection curves

V-curve

Λ-curve

u-curve

Figure 3: Injection curve commands for V-, Λ- and u-curve (a) and

measured methane concentration 50 mm downstream of the combus-

tor inlet averaged over 150 cycles (b).

late the variation in pressure rise with direct observa-

tions of the homogeneity of autoignition.

3.1. Fuel Injection

The mass flow rates of fuel and air are set to match

the mixture bulk flow velocity for reacting experiments

(ubulk =18 m/s). Each control sequence is injected for

150 cycles with an operating frequency of 5 Hz. Three

different injection profiles are investigated at ambient

pressure and temperature: (i) Λ-curve, (ii) V-curve and

(iii) u-curve. The control sequences and the respec-

tive TDLAS measured, cycle averaged fuel concentra-

tion for the three trajectories are shown in Fig. 3.

The averaged results clearly show the capability of

replicating a desired fuel profile within the given time

span ∆tinj. The measured fuel profiles of individual cy-

cles show a standard deviation (std) of less than 5 %

throughout the injection. There is a clear smoothing ef-

fect, especially in the regions of high gradient. This is

expected and can be attributed to two effects: (i) turbu-

lent diffusion and (ii) shear layer effects. Diffusion will

smooth the sharp features of the injection profile (begin-

ning and end of injection period). The shear layer effect

near the wall causes a variation in the velocity profile

through the tube and induces a spatial distortion of the

injection profile. This phenomenon can only be mea-

sured as an integrated value across the tube through the

line-of-sight measurement. It is also important to men-

tion that due to inertia of the valves, there is a hysteresis

to the valve response when opening and closing, the ef-

fects of which are difficult to account. Currently, there

is no way to avoid these effects, and must instead be

accounted for when interpreting the reacting results.

For reacting tests, it is important to maintain a con-

stant cycle-averaged fuel flow rate, otherwise differ-

ences in pressure rise might occur due to variations in

total heat release. For this, the total valve-open time

82 84

∆τip

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

∆pin bar

72 76 78 80

tin ms

τpτi

Figure 4: An example pressure history during ignition event for pres-

sure sensors P1to P5resulting from the injection of the Λ-curve.

is maintained constant through the injection period and

averaged to eight open valves. This corresponds to

an average equivalence ratio of the fuel–air mixture of

ϕ=1.6. The variation of the integrated area of the

measured fuel concentrations shown in Fig. 3b is within

the measurement uncertainty. Therefore, the reactant

injection profile can be maintained and controlled with

reasonable certainty.

3.2. Autoignition Control

For the following investigations, the air mass flow is

held constant at 30 kg/h, resulting in a mean mixture

bulk velocity of ubulk =18 m/s. The three injection

profiles shown in Fig. 3a are applied to the reactive

measurements for 150 cycles with a firing frequency of

5 Hz. Figure 4 shows sample pressure histories of the

five distributed pressure sensors for a single cycle. The

ignition begins approximately 72 ms after the start of

the fuel injection. At peak 1in Fig. 4 the maximum

pressure rise is reached and then starts decreasing due

to expansion of the burned gas in upstream and down-

stream directions. Pressure waves that travel upstream

are reflected at the acoustically closed inlet and result

in a sharp rise in pressure (2). Pressure waves prop-

agating downstream are reflected at the tube outlet as

an expansion wave (3) that travels upstream. This ex-

pansion wave can be used to support the refilling of the

combustor.

For each cycle, the maximum relative pressure in-

crease ∆pmax is calculated as the difference between the

mean maximum pressure in peak 1and the pressure be-

fore ignition. The expression ’ignition time’ is intro-

duced and is calculated from the pressure data τpas the

time delay between the starting point of the injection

until the first increase in pressure exceeding a threshold

of 0.03 bar. A similar τiis derived from the ionization

2 4 6 8 10

0.2

0.4

0.6

0.8

∆τip in ms

∆pmax in bar

regime 1

regime 2

Fig. 4

u-curve

Λ-curve

V-curve

Figure 5: Pressure amplitude over delay ∆τip between autoignition

and pressure rise for each cycle.

probe data as labeled in Fig. 4. The difference between

these two ignition times is labeled ∆τip. ’Ignition delay

time’ is ambiguous due to the long injection period.

The amplitude of the pressure rise is plotted against

the coherency of the ignition (∆τip) in Fig. 5 for the three

injection profiles. There is quite a lot of cycle-to-cycle

variance for each specific fuel profile. However, there

is a notable shift towards higher pressure amplitudes

for lower ∆τip. Also, the three profiles tend to cluster

throughout this graph, indicating that the Λ-curve tends

towards smaller ∆τip while the V-curve is less homoge-

neous. The u-curve falls generally in between. Low

∆τip indicates a more quasi-simultaneous detection of

ignition and pressure rise, whereas higher ∆τip implies

a later or less homogeneous ignition compared. With

decreasing ∆τip, the resulting pressure amplitude is in-

creasing significantly. According to Oppenheim [16],

two characteristic modes of ignition can be generally

observed: (i) mild ignition and (ii) strong ignition. The

latter case can be described as a reaction front with a

detonation-like structure characterized by a sharp in-

crease in pressure. A mild ignition appears due to multi-

ple autoignition sources that propagate chaotically and

individually. All observed ignitions that are shown in

Fig. 5 can be categorized as mild ignitions differing

in the number of autoignition sources. This conclu-

sion is based on the comparison of the gradual pres-

sure increase (t≈73 ms in Fig. 4) subsequent to the

autoignition when comparing to the steep increase of

the reflected wave (t≈76 ms in Fig. 4) as suggested

by Bartenev and Gelfand [6]. However, comparing

regime 1 to regime 2 shows that under certain condi-

tions, multiple autoignition sources can initiate a reac-

tion front that is more likely to couple with the pressure

rise resulting in a greater pressure.

To characterize the pressure response within the com-

40 50 60 70

tin ms

-0.4

-0.2

0.2

0.4

0.6

∆pin bar

P1P5

40 50 60 70

tin ms

a) regime 2 b) regime 1

Figure 6: Dominant POD mode calculated from pressure data for

regime 2 (a) and regime 1 (b).

bustor, proper orthogonal decomposition (POD) was ap-

plied to the pressure data of 50 cycles for both regions

identified in Fig. 5. The dominant POD modes of the

pressure histories of sensors P1and P5are shown in

Fig. 6.

Figure 6a shows a much weaker pressure wave trav-

eling upstream the combustor with the speed of sound

detected by sensor P5and later by sensor P1respec-

tively. The propagation velocity of the ignition front is

slower such that limited coupling between the pressure

wave and heat release occurs. Figure 6b shows a nearly

simultaneous detection of the pressure increase by sen-

sors P1and P5. A gradual, non-sharp (compared to a

shock wave) rise in pressure is noticeable. The period of

elevated pressure was calculated to a mean value of 7.4

and 6.7 ms for V-curve and Λ-curve respectively with

a standard deviation of this period of 1.1 ms for both

injection profiles. The prolonged duration of pressure

increase results from a wider spatial distribution of the

ignition event for the V-curve whereas a more homoge-

neous ignition for the Λ-curve results in a more distinct

pressure rise. Due to the higher pressure amplitude in

regime 1, the expansion wave that arises from the re-

flection at the tube outlet is more pronounced resulting

in a lower pressure for 57 ms <t<62 ms. The period

of low pressure was calculated to be 5.5 and 5.8 ms with

a standard deviation of 0.3 ms. These calculations indi-

cate that there is little correlation between the duration

of the high pressure and the time span of low pressure.

Hence, this low pressure region is caused by reflection

of pressure waves at the acoustically open end of the

combustor and is therefore a function of acoustic time

scales only, which do not change for the conducted mea-

surements.

Based on the observed variations in pressure and ion-

ization probe histories for the ignition events as shown

in Fig. 5, the homogeneity in autoignition was found to

be primarily responsible for larger pressure rises. Be-

cause the specific fuel injection profiles used in this

study are not optimized to achieve consistent, homoge-

neous autoignition, they therefore show significant vari-

ation in the likelihood of a uniform autoignition which

results in the broadened distributions shown above.

To further explore the ignition process, the setup with

an optical access as shown in Fig. 2b is used to measure

OH∗intensity and the pressure simultaneously. Figure 7

shows x-t-diagrams of the OH∗intensity for the down-

stream part of section 1 and sections 2 and 3 of the opti-

cal setup for three sample shots. Every time step in the

x-t-diagrams represents an average OH∗intensity for all

pixels of a single image at the respective axial position.

649 snapshots are aligned for each figure. The verti-

cal shadows in the images correspond to the combustor

supports and the pressure traces (P2and P3) at these lo-

cations are overlaid. The white line represents the ig-

nition time τofor each axial position that is defined by

the OH∗intensity exceeding a threshold of 0.079. The

temporal position is aligned to the earliest ignition point

within the frame. Figures 7a and b show two different

cycles for the Λ-curve while the Fig. 7c on the right re-

sults from measurements with the V-curve. These three

cycles have been selected to be representative of the var-

ious visible ignition events observed for many cycles.

Lastly, it is important to recall, that there is a constant

bulk flow in the positive x-direction before ignition.

A uniform, highly distributed ignition front with mul-

tiple points of ignition is shown in Fig. 7a. Compared

with the other examples, the region after ignition shows

the highest OH∗intensity. Furthermore, the largest am-

plitude in the pressure signals are observed. The major-

ity of the high intensity OH∗occurs within a period of

about ∆t=2 ms. There also appears to be a propagation

of the gas in the upstream direction for x<200 mm and

in downstream direction for x>200 mm respectively.

The distributed ignition results in an aerodynamic con-

finement, which serves to increase the pressure more

than in the other examples. As the products expand, the

pressure falls and eventually reaches a local minimum

for ∆t≈3 ms. As previously mentioned, the pressure

wave propagating upstream is reflected at the tube inlet

and propagates downstream which is visible as a second

pressure peak in Fig. 7a at ∆t≈3.5 ms. At the acousti-

cally open tube outlet, this pressure wave is reflected as

an expansion wave that causes a low pressure region for

∆t>5 ms.

The ignition front in Fig. 7b is less uniform, and in-

stead propagates primarily from a single ignition point

at x≈100 mm. From this source, the ignition front

propagates in both axial directions. The propagation

0 100 200 300

xin mm

c) V-curve, regime 2

0 100 200 300

xin mm

-1

∆tin ms

a) Λ-curve, regime 1

0 100 200 300

xin mm

b) Λ-curve

0.0

0.5

1.0

OH∗intensity

-0.33

0.00

0.33

0.66

∆pin bar

Figure 7: x-t-diagrams for OH∗intensity and pressure histories for P2and P3for three example shots. The spatial range of each figure covers the

downstream part section 1 and the sections 2 and 3 of the optical accessible combustor (Fig. 2b). The white line represents the spatial distribution

of the ignition time τothat is defined by the OH∗intensity. The temporal positions of the three figures are adjusted by ∆t=t−τo,min.

velocity in the lab frame reference can be estimated

using the slopes of the fronts in the x-t-diagram as

ua,US ≈ −68 m/s and ua,DS ≈112 m/s for upstream (US)

and downstream (DS) direction, respectively. Account-

ing for the bulk flow velocity of ubulk ≈18 m/s, a nearly

constant propagation velocity of ua≈ −86 and 94 m/s

can be found, which suggests that the propagation of the

ignition front in Fig. 7b is a propagating turbulent flame,

rather than an autoignition front.

Figure 7c shows the distribution of OH∗intensity for

a measurement with the V-curve. The first ignition point

occurs close to or just outside of the image frame. Com-

pared with the first two examples, the autoignition oc-

curs much further downstream in the combustor due

to the richer mixture at the beginning of the injection.

Subsequently, a deflagration propagates upstream in the

combustor at ua,US ≈ −70 m/s in the lab frame refer-

ence. Simultaneously, a weaker pressure wave can be

seen in the pressure histories travelling at the speed of

sound. At ∆t=4 ms, a second deflagration front is

propagating downstream in the combustor which indi-

cates a second autoignition event has occurred further

upstream of the combustor (presumably also due to the

second high equivalence ratio region at the end of the

V-curve injection profile). In this case, the reduced

aerodynamic confinement due to multiple, separated ig-

nition points results in the smallest increase in pressure

of the three examples. The acceleration of the flow due

to the upstream ignition event results in the two prop-

agating fronts, and the curved trajectories in the fixed

laboratory frame.

From Fig. 7a to 7c, the inhomogeneity of the igni-

0 50 100 150 200 250 300 350

std(τo)

0.1

0.2

0.3

0.4

0.5

0.6

0.7

∆pmax in bar

Λ-curve

V-curve

u-curve

regime 1

regime 2

Figure 8: Pressure amplitude over standard deviation of the ignition

time std(τo) plotted for V-curve, Λ-curve and u-curve. The high-

lighted markers represent the three shots shown in Fig. 7.

tion front τoincreases, which can be expressed by an

increase of the standard deviation of the ignition time

std (τo). This increase in std (τo)corresponds to a de-

crease in the maximum pressure amplitude ∆pmax. Si-

multaneously, a decrease in the OH∗intensity and the

amplitude of the subsequent low pressure region ∆pmin

is visible.

Figure 8 shows ∆pmax as a function of std (τo)

for all measured cycles with the three injection pro-

files. The three examples presented in Fig. 7 are high-

lighted. Shots with high variance in the ignition front

(large std (τo)) correspond to a lower pressure rise than

those with more uniform ignition. The time delay be-

tween detecting a pressure rise and a combustion event

(from either an ionization probe or OH∗), as shown in

Figs. 5 and 8, shows a consistent trend. When the

time delay between the pressure rise and the combus-

tion event is coupled (as in regime 1), the distributed

heat release results in a greater pressure rise due to the

aerodynamic confinement compared with the uncoupled

regime 2. Achieving this type of autoignition event con-

sistently and routinely is the objective of utilizing the

SEC for pressure gain combustion. Towards this aim,

on-going work is focused on tailoring the specific injec-

tion profile to maximize this pressure rise as well as the

consistency in autoignition homogeneity.

4. Conclusion and Outlook

This work presented a novel combustion rig equipped

with a controllable reactant injection system designed

to deliver a prescribed, stratified charge of autoignitable

mixture. The reproducibility of the system for three in-

jection profiles was demonstrated. Then, the correlation

between pressure rise and distribution of ignition events

was shown.

It was observed that a temporal decoupling between

pressure rise and flame detection by ion probe was asso-

ciated with a lower overall pressure rise. Observing the

ignition process for different ignition regimes, broadly

classified as regimes 1 and 2, supported the conclusion

that a distributed, uniform autoignition event results in a

greater overall pressure rise. This work has shown that

these phenomena are repeatable and measurable, which

is an important requirement for the use of a shockless

explosion combustor as a pressure gain combustion de-

vice, where the concept of aerodynamic confinement

during autoignition is comparable with constant volume

combustion.

Using three arbitrary injection profiles in this work

demonstrated a variety of ignition distributions. This

work therefore serves as a starting point for the refine-

ment of injection profiles in order to maximize the pres-

sure rise and minimize the variability in ignition occur-

rence as well as to later implement closed-loop control

of the injection scheme. Despite the variability in the

profiles studied here, it is already seen that a significant

portion of ignition events follow the process in regime 1

and that the injection profile is very controllable. Both

are important results for progressing the controlled au-

toignition and studying these combustion events in strat-

ified mixtures.

Acknowledgments

The authors gratefully acknowledge the support of the

Deutsche Forschungsgemeinschaft (DFG) as part of

Collaborative Research Center CRC 1029 ”Substantial

efficiency increase in gas turbines through direct use of

coupled unsteady combustion and flow dynamics”. The

authors also wish to thank Andy G¨

ohrs and Thorsten

Dessin for their technical support.

References

[1] J. Gray, M. Lemke, J. Reiss, C. Paschereit, J. Sesterhenn,

J. Moeck, A compact shock-focusing geometry for detonation

initiation: experiments and adjoint-based variational data assim-

ilation, Combust. Flame 183 (2017) 144–156.

[2] R. Bluemner, M. D. Bohon, C. O. Paschereit, E. J. Gutmark,

Single and counter-rotating wave modes in an RDC, AIAA Pa-

per 2018-1608.

[3] B. C. Bobusch, P. Berndt, C. O. Paschereit, R. Klein, Shock-

less explosion combustion: An innovative way of efficient con-

stant volume combustion in gas turbines, Combust. Sci. Technol.

186 (10-11) (2014) 1680–1689.

[4] Y. B. Zeldovich, Regime classification of an exothermic reac-

tion with nonuniform initial conditions, Combust. Flame 39 (2)

(1980) 211–214.

[5] X. Gu, D. Emerson, D. Bradley, Modes of reaction front propa-

gation from hot spots, Combust. Flame 133 (1-2) (2003) 63–74.

[6] A. Bartenev, B. Gelfand, Spontaneous initiation of detonations,

Prog. Energ. Combust. 26 (1) (2000) 29–55.

[7] F. Zhao, T. N. Asmus, D. N. Assanis, J. E. Dec, J. A. Eng, P. M.

Najt, Homogeneous charge compression ignition (hcci) engines,

Tech. rep., SAE Technical Paper (2003).

[8] J. Eng, Characterization of pressure waves in hcci combustion,

Tech. rep., SAE Technical Paper (2002).

[9] B. C. Bobusch, Fluidic devices for realizing the shockless ex-

plosion combustion process, Ph.D. thesis, TU Berlin (2015).

[10] T. G. Reichel, J.-S. Sch¨

apel, B. C. Bobusch, R. Klein, R. King,

C. O. Paschereit, Shockless explosion combustion: Experimen-

tal investigation of a new approximate constant volume combus-

tion process, J. Eng. Gas Turb. Power 139 (2) (2017) 021504.

[11] F. C. Y¨

ucel, F. V¨

olzke, C. O. Paschereit, Effect of the switching

times on the operating behavior of a shockless explosion com-

bustor, in: Active Flow and Combustion Control 2018, Springer,

2019, pp. 121–134.

[12] H. Li, S. D. Wehe, K. R. McManus, Real-time equivalence ratio

measurements in gas turbine combustors with a near-infrared

diode laser sensor, P. Combust. Inst. 33 (1) (2011) 717–724.

[13] J.-S. Sch¨

apel, R. King, F. Y¨

ucel, F. V¨

olzke, C. O. Paschereit,

R. Klein, Fuel injection control for a valve array in a shockless

explosion combustor, in: ASME Turbo Expo 2018: Turboma-

chinery Technical Conference and Exposition, American Soci-

ety of Mechanical Engineers Digital Collection, 2018.

[14] U. Burke, K. P. Somers, P. O’Toole, C. M. Zinner, N. Mar-

quet, G. Bourque, E. L. Petersen, W. K. Metcalfe, Z. Serinyel,

H. J. Curran, An ignition delay and kinetic modeling study of

methane, dimethyl ether, and their mixtures at high pressures,

Combust. Flame 162 (2) (2015) 315–330.

[15] D. G. Goodwin, R. L. Speth, H. K. Moffat, B. W. We-

ber, Cantera: An object-oriented software toolkit for

chemical kinetics, thermodynamics, and transport pro-

cesses, https://www.cantera.org, version 2.4.0 (2018).

doi:10.5281/zenodo.1174508.

[16] A. K. Oppenheim, Dynamic features of combustion, Philos. T.

R. Soc. A 315 (1534) (1985) 471–508.