Carbon Resources Conversion 4 (2021) 28–35

Contents lists available at ScienceDirect

Carbon Resources Conversion

journal homepage: www.sciencedirect.com/journal/carbon-resources-conversion

Design and simulation of gas burner ejectors

Mario Pichler a, *, Florian Wesenauer a, Christian Jordan a, Stefan Puskas b, Bernhard Streibl c,
Franz Winter a, Michael Harasek a
a
Institute of Chemical, Environmental & Bioscience Engineering, TU Wien, Getreidemarkt 9/166, 1060 Vienna, Austria
b
Wienerberger AG, Wienerbergerplatz 1, 1100 Vienna, Austria
c
DrS3 – Strömungsberechnung und Simulation e.U., Simmeringer Hauptstraße 397/141, 1110 Vienna, Austria

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

Keywords: Ignition within gas burner ejectors can lead to off design conditions and has significant influence on the burner
Self-ignition behavior. Thus ignition in the ejector should be prevented. In the present study the influence of combustion
Flammability limits reactions on the performance of gas burner injectors is investigated. To investigate if ignition is possible,
Reacting jet
simulated ignition delay times, using a detailed reaction mechanism, are compared to predicted mean residence
Ejector
times of the gas in the ejector. Gas burner ejectors are designed using one dimensional analytic equations, based
Jet pump
Computational fluid dynamics on energy and momentum conservation equations and conventional isentropic equations. 1D results are
compared to 2D computational fluid dynamics (CFD) simulations, to take into account non-ideal mixing effects
along the ejector. Results are validated with experiments with air at room temperature. 1D results show very
good agreement not only with CFD simulations for the case of non-reactive flows, but also with performed
experiments.
It is shown that the assumption of ideal mixing along the ejector and thus the comparison of the ignition delay
time to the gas mean residence time, to predict ignition in the ejector, is not valid. Ignition in the ejector is
possible, even if the ignition delay time is more than thirty times higher than the mean residence time. In
addition to that, it is shown, that ignition and the choice of reaction mechanism have significant influence on the
predicted gas burner ejector performance. Thus, the accurate prediction of ignition delay time and the use of a
detailed reaction kinetic are mandatory to correctly predict the burner ejector behavior.

1. Introduction
1.1. Ejector
Ejectors are devices where a high-pressure fluid (primary stream)
transfers energy to a low-pressure fluid (secondary stream), entraining
the low-pressure fluid and mixing the two streams in the mixing section.
The term jet pump is used when the ejector is operated using incom­
pressible fluids, whereas the term ejector is generally used when using
compressible fluids (Sun & Eames [1]). The performance and design of
such devices was subject of several investigations over the last years
[1–3].
A comprehensive overview of different approaches for the design of
ejectors, namely analytical, experimental and computational, is given by
Yadav et al. [4].
Kumar and Ooi used the Fanno flow concept to describe the frictional
compressible flow in the mixing zone of the ejector [5]. They extended
the model proposed by Chow and Addy [6] and improved the prediction
of the entrained gas ratio by up to 4 %.
Namkhat and Jugjai [7] theoretically and experimentally investi­
gated the entrainment of primary air in a self-aspirating gas burner. A
good agreement from experiment to theory was found.
Kumar et al [8] developed an ejector model based on the constant
rate of kinetic energy change (CRKEC) methodology. They found an
excellent agreement with experimental results, but further studies for off
design conditions are needed.
However, none of these have studied the influence of ignition within
the ejector. Depending on the exact design and conditions, reaction
might take place in the mixing section of the ejector when using gaseous
fuels (e.g. pure methane or natural gas) as high-pressure fluid and air at
elevated temperatures as low-pressure fluid. This might lead to off-
design conditions for ejectors, which are designed to be operated
solely under non-reactive conditions. Goal of this study is to predict if
ignition is possible in a given ejector by comparing ignition delay times
to the mean residence time of the gas in the ejector.

* Corresponding author.
E-mail address: mario.pichler@tuwien.ac.at (M. Pichler).

https://doi.org/10.1016/j.crcon.2021.01.006
Received 30 October 2020; Received in revised form 22 December 2020; Accepted 7 January 2021
Available online 16 January 2021
2588-9133/© 2021 The Authors. Publishing services by Elsevier B.V. on behalf of KeAi Communications Co. Ltd. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
M. Pichler et al. Carbon Resources Conversion 4 (2021) 28–35

Nomenclature v velocity, m/s

V̇ volume flow, m3/s
a area, m2
at nozzle cross section area at critical conditions, m2 Greek Symbols
A pre-exponential factor, 1/s β exponent of temperature in Arrhenius equation, –
c speed of sound, m/s γ ratio of heat capacities, –
cp specific heat capacity, J/(kg K) τ residence time, s
d diameter, m τigni ignition delay time, ms
dt nozzle diameter at critical conditions, m τT ratio of temperatures, –
E activation energy, J/mol Φ Fuel-air equivalence ratio, –
f0 – f4 heat capacity model constants, – Ω mass flow ratio, –
L length of ejector subsection, m
Subscripts
M Mach number, –
exp experimental value
Mi molar weight of specie I, kg/mol
i initial conditions of high pressure fluid
ṁ mass flow, kg/h
stag Stagnation property
p pressure bar, Pa
x position: ejector outlet
ℛ universal gas constant, J/(mol K)
0 position: initial conditions of low pressure fluid
T temperature, K
1 position: start of constant area mixing zone
Ta activation temperature, K

1.2. Ignition
The definition of ignition is very important to be able to verify if
combustion takes place in the ejector. It is discussed in the following
section.
To ignite a mixture a certain minimum amount of energy has to be
introduced to a mixture of combustible material (either gas, liquid or
solid) and air (providing oxygen). At increasing temperatures highly
more reactive radicals are formed, which then further react and can then
lead to complete combustion of the combustible material.
Two different ignition initialization processes can be distinguished
when talking about ignition:
• induced or external ignition (e. g. spark ignition in a petrol engine)
• self-ignition (e. g. diesel engine, gas turbine)
If an flammable, homogeneous mixture of fuel and air is locally
exposed to a high enough temperature (T) and pressure (p), ignition will
take place within a certain time. After ignition, the fuel (here only
gaseous fuel, mainly natural gas, is considered) will react within a very
short time (in order of milliseconds) with oxygen contained in the air,
leading to a sudden temperature increase and, depending on the case,
pressure increase. If the released heat of reaction is high enough, then
the reaction will continue further without the external addition of en­
ergy, until either oxygen or fuel is completely consumed. This process is
called self-ignition (Joos [9]). Ignition of the fuel/air mixture in gas
burner injectors investigated in the present study is achieved by self-
ignition.
1.3. Ignition delay time
As ignition is always a transient process, combustion will only take
place after a certain ignition delay time τigni , even if the ignition con­
ditions are met. The ignition delay time is in order of microseconds to up
to seconds at the minimum autoignition temperature (MAIT), depending
on the conditions. For example the ignition delay time τigni of methane at
the MAIT (898 K) and at an equivalence ratio Φ of 0.5, is in the order of
20 seconds (Reid, Robinson, & Smith [10]).
fuel/air mixture. Pressure is increased in the inert gas chamber, until the
membrane bursts, leading to a shock wave running through the system.
The shock increases pressure and temperature, which leads to ignition.
The ignition delay time is then defined as time from reflection of the
shock wave to start of radical formation, more precisely the intersection
of the tangent of radical curve and the initial radical concentration.
Often CH- (431 nm) or OH- radicals (308 nm) concentrations, as in­
dicators for ignition, are measured via chemiluminescence (Joos [9]).
A different approach was used by Reid, Robinson, & Smith [10] who
measured ignition delay times and minimum autoignition temperatures
of methane/air mixtures using stirred cylindrical (0.5 l volume) and
unstirred spherical vessels (0.8 l volume). They used two thermocouples
to monitor the gas temperature inside the reactor. The ignition delay
time was determined by the rise of temperature in the vessel. The pre­
mixed gas has entered the vessel at room temperature and reached vessel
temperature in about three seconds. A sharp distinction between igni­
tion and non-ignition was found. A change of 1–2 K of the initial vessel
temperature was sufficient to move from ignition to non-ignition.
The ignition delay time τigni of methane in air can be calculated ac­
cording using empirical correlations (Eq. (1)), for temperatures of 1400
K – 2050 K, an equivalence ratio Φ of 0.5–2.0 and concentrations of
[CH4] < 3.6 × 10− 5 mol/cm3 (Petersen, Röhrig, et al. [11]).
( )
E
τCH4 = 4.05 × 10− 15 [CH 4 ]0.33 [O2 ]− 1.05 exp (1)
RT
where τCH4 p− 0.72 . Here E is the activation energy in J/mol (E = 216.8
kJ/mol), ℛ is the universal gas constant, [CH4] and [O2] are concen­
trations in mol/cm3 and p is the pressure.
For more accurate predictions Petersen, et al. [12] suggest to adjust
the activation energies E according to Table 1. Ignition delay decreases
for increasing equivalence ratio Φ (Joos [9]).
Experimental ignition delay times and derived equations for the
ignition delay times of C2H2/O2/Ar, C2H4/O2/Ar and C2H6/O2/Ar
Table 1
Suggested activation energies E for methane (Petersen et al. [12]).
T, K pav, bar E, kJ/mol

1451–2001 0.71 173.7

1.3.1. Ignition delay time: experimental 1407–1625 10.9 185.9
For gaseous fuels often a shock wave tube is used for measuring the 1290–1407 10.9 151.1
ignition delay times. In a shock wave tube, a membrane separates two 1367–1659 19.6 167.8
1243–1367 19.6 109.7
chambers. One chamber is filled with inert gas, the other one with the

29
M. Pichler et al. Carbon Resources Conversion 4 (2021) 28–35

mixtures are presented in Petersen, Hall, Kalitan, & Rickard [13]. Cor­ and secondary gas (index 0) composition, temperature and pressure, the
relations for the ignition delay time of acetylene and ethylene are shown desired ratio of mass flow rates of secondary to primary stream Ω and
in Eqs. (2) and (3). the nozzle diameter d’x . From this the diameters d’x , d1 , d2 and d3 are
( ) calculated.
79.5
τC2 H2 = 5.2 × 10− 14 [O2 ]− 1 exp (2) The overall process is described by three subsequent sub-processes,
RT
namely constant pressure mixing, constant area mixing and diffusion.
(
26.6
) Equations presented by Keenan, Neumann, & Lustwerk [2] where
τC2 H4 = 3.3 × 10− 13 [C2 H4 ]0.19 [O2 ]− 0.95
[Ar]0.04 exp (3) modified such, that different molecular weights and heat capacities can
RT
be considered. Specific heat capacities of air and methane are modelled
While these simple equations can be used for a wide range of using well-known NASA polynomials (Gordon & McBride [16], equation
equivalence ratios Φ additional shock tube experiments have to be (4)). The gas heat capacities are assumed constant along the nozzle.
performed for mixtures of different hydrocarbons. Higher hydrocarbons
R( )
are usually more reactive than methane. Therefore, the ignition delay cp = f0 + f1 T + f2 T 2 + f3 T 3 + f4 T 4 (4)
time is much lower. Even low concentrations of higher hydrocarbons (e. M
g. propane) in methane, lead to a drastic decrease of the ignition delay Important dimensionless variables are the mass flow ratio Ω (equa­
time. Ignition delay time was simulated for a mixture of CH4 (15 wt.%), tion (5)) and the ratio of secondary to primary gas temperature τT
O2 (30 wt.%), N2 (55 wt.%) at 1 bar. The addition of 1 wt% of propane to (equation (6)). Here ṁi and ṁ0 are the mass flow through the nozzle and
methane, results in a decrease of the ignition delay time from 135.3 ms the entrained mass flow from the surroundings, respectively.
to 96.7 ms respectively.
ṁ0
Ω= (5)
ṁi
1.3.2. Ignition delay time: simulation
Using detailed chemical reaction mechanisms in perfectly stirred T0
reactor simulations, a more general description of the ignition delay τT = (6)
Ti
time can be achieved (Joos [9]). Detailed mechanisms can predict the
ignition delay time for a wide range of conditions with good accuracy. The pressure at position x, 1 and 2 is assumed to be constant and
One of the most commonly used detailed reaction mechanism for the equal to the pressure at the inlet of the ejector pipe px . The temperature
combustion of methane or natural gas is the GRI3.0 mechanism (Smith, of the gas mixture T1 = T2 = T3 is calculated iteratively using equation
et al. [14]). It is a mechanism designed and optimized for modelling the (4) and equations (7)-(8). Here the indices (e.g. i, 0, x) refer to the
combustion of natural gas, including NO formation and reburn chem­ respective position along the ejector, the superscripts h and l refer to
istry and consists of 53 species and 325 reactions. An even more detailed high and low pressure fluid respectively. The ratio of heat capacities γ of
reaction mechanism is the POLIMI C1-C3 HLT with 107 species and the mixture is calculated from equation (9).
2642 reactions. It is applicable for the pyrolysis, partial oxidation and clp0 ΩT0 + chpi Ti
combustion of hydrocarbon fuels up to three carbon atoms (POLIMI C1- T1 = (7)
Ωclp1 + chp1
C3 HLT [15]). In the presented study the ignition delay time is defined as
the time, where the simulated temperature increase ΔT/Δt takes a
clp1 Ω + chp1
maximum value. cp1 = (8)
1+Ω
2. Materials & methods γ l1 Ω + γ h1
γ1 = (9)
1+Ω
In the following chapter important assumptions, model equations,
concepts and simulative and experimental setups are discussed. First the The critical Mach number M*1 (equation (10)) at the end of the
model for the design of ejectors according to Keenan, Neumann, & constant pressure mixing section is calculated from the momentum
Lustwerk [2] is extended to consider gases with different molecular equation. The critical speed of sound c* is calculated using equation
weight and heat capacities. Then the experimental setup, conditions and (11). The critical Mach numbers M’* x and Mx are calculated from the
’’*

procedure is discussed briefly. The single cell simulation of ignition pressure ratios using equation (12) and equation (13). Here M’* x and Mx
’’*

delay times under various conditions (initial temperature and equiva­
lence ratios) elaborated in chapter 2.3. Finally the models, assumptions
and simulation domain, including boundary conditions, are discussed.
2.1. Ejector design
are the critical Mach numbers of high and low pressure fluid at position x
(Fig. 1), γhi and γ l0 are the ratios of heat capacities of the high and low
pressure fluid at stagnation conditions i and 0, pi and p0 are the initial
stagnation pressures of high and low pressure fluid and px is the pressure
at position x (Fig. 1).

The ejector geometry (schematic shown in Fig. 1) is based on the Mx’* c’* ’’* ’’*
x + Mx cx Ω
M1* = (10)
designed equations according to Keenan, Neumann, & Lustwerk [2]. The (1 + Ω)c*1
geometric proportions depend on a function of initial primary (index i)

Fig. 1. Ejector geometry and definition of important parameters.

30
M. Pichler et al. Carbon Resources Conversion 4 (2021) 28–35

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( ) l1
2(γ − 1) a’ γl − 1 ( 2 ) γ0 − 1
*
c = cp T (11) = M1* 1 + 0 1 − M1* (22)
γ+1 a1 2
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ( )
√ γ l0 − 1 ( ) −
1
√ h ( ( )γhi −h 1 ) a3 γl − 1
√γ i + 1 px γi = M *
1 + 1 − M *2 0
(23)
’*
Mx = 1− (12) a’ 3
2 3
γ hi − 1 pi
From here the inlet diameter d’’x , the constant area mixing diameter
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√
√ l ( ( )γ0 −l 1 )
l d1 , the outlet diameter d3 and the critical nozzle diameter dt are calcu­
Mx’’* =
√γ 0 + 1
1−
px γ 0
(13) lated from Eqs. (24)–(27).
γl0 − 1 p0 √̅̅̅̅̅̅̅̅̅̅̅̅̅̅
a’’
The Mach number M1 is then calculated from the conventional ’’ ’
dx = dx 1 + x’ (24)
ax
relation between M and M* (equation (14)).
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅
a1
M1 = M1*
1
(14) d1 = dt (25)
γ1 − 1
( * )2 at
1 − 2 ( M 1 − 1)
√̅̅̅̅̅̅̅
If M1 < 1, no shock takes place in the constant area mixing section, 4a3
d3 = (26)
thus there is no change in properties (T, M, p) from position 1 to position π
2. The critical Mach number M*3 (equation (16)), critical speed of sound √̅̅̅̅̅
at
c*3 (equation (17)) and velocity v3 (equation (18)) at the end of the dt = dx’ (27)
a’x
ejector pipe are calculated using the stagnation pressure p3,stag (equation
(15)). When the ejector geometry is fixed, the mean residence time of the
( )γ γ−1 1 gas in the ejector can be calculated from the total ejector volume Vtotal
p1 γ − 1 2
(15) and the total volumetric gas flow V̇total using equation (28).
1
p3,stag =( ) = p1 1 M1 + 1
p
2
1
Vtotal
p3,stag
τ= (28)
V̇ total
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√ ( ( )γ1γ− 1 ) Assuming perfect mixing in the ejector, the mean residence time τ
√
√γ 1 − 1 p3 1
combined with the predicted ignition delay time τigni for the exact gas
*
M3 = √ 1− (16)
γ1 + 1 p3,stag conditions (temperature, pressure, composition) can be used to predict if
ignition takes place in the ejector.
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
2(γ1 − 1)
c*3 = cp1 T1 (17)
γ1 + 1 2.2. Experimental setup

v3 = M3* c*3 (18) Cold flow experiments have been carried out to validate the simu­
lation results. A schematic of the experimental setup is shown in Fig. 2.
To define the ejector geometry, the ratios of areas a’’x /a’x , at /a’x , at /a2 ,
Pressurized air, supplied by a compressor, is mixed with CO2 (used as
a’ /a2 and a3 /a’ are calculated from conventional isentropic relations tracer gas) and supplied to the ejector nozzle. The CO2 concentration in
according to Eqs. (19)–(23) respectively. Here a’x , a’’x are the areas of the supplied gas mixture, in the gas leaving the ejector pipe and in the
high and low pressure fluid at position x, a2 and a3 are the cross-section surroundings, is measured using two non-dispersive infrared (NDIR)
areas of the ejector pipe at position 2 and 3 and at is the critical cross- sensors (measurement range: 0 to 5000 ppm, accuracy: ±30 ppm). A
section are of the nozzle. typical test cycle was composed of three different phases. In each of the
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√ ( )γhi −h 1
( )1/γhi √
√̅̅̅̅̅̅̅̅̅̅( px
√ px
γ
i
) p 1 − pi
a’’x pi chpi γhi (γ l0 − 1) i
=Ω τT l √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ (19)
a’x p0 cp0 γl0 (γ hi − 1) ( )1/γl0 √ √ ( )γ0 −l 1
l
px
√ γ
p0
1 − pp0x 0

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )1/γhi √
√ ( )γhi −h 1
px √ γ

pi
1 − ppxi i
at
= ( ) h1 √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ (20)
a’x γ − 1
2
γhi +1
i
1 − γh2+1
i

( )
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( ) γ1 √ ( )γ1γ− 1
1 √ 1
p1 √ p1
(( ) )√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ 1−
γhi − 1 γ1 √
p0,3 p0,3
at p0,3 √ 1
= √ ( ) ( ) h1 √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
a1 pi (γ1 − 1)γhi √ l
√(1 + Ω) 1 + Ωτ cp0 2
γ − 1
i
1 − γh2+1
T ch γhi +1
pi i

(21) Fig. 2. Schematics of the experimental setup. 1 … mass flow controller; 2, 3 …

rotameter; 4 … ejector nozzle; 5, ejector pipe; 6 … membrane pump; 7, 8 …
CO2 sensors; 9, 10 … manometer.

31
M. Pichler et al. Carbon Resources Conversion 4 (2021) 28–35

phases the CO2 concentration was measured for at least 60 seconds to Table 3
ensure stable operating conditions. In the first phase the CO2-concen­ Reaction equations, pre-exponential factors, temperature exponent and activa­
tration in the surrounding was measured at the ejector outlet. No CO2 tion temperature of the Hyer mechanism (Ahmed et al. [18]).
was added to the high pressure fluid. In the second phase CO2 was added Reaction A β Ta
to the high pressure fluid using the mass flow controller. The CO2 con­ CH4 + 0.5O2 →CO + 2H2 0 15154.28
4.4 × 109
centration was measured using sensor 8. In the last phase the carbon
CH4 + H2 O→CO + 3H2 3.0 × 108 0 15154.28
dioxide concentration at the ejector outlet was measured using sensor 7.
CO + H2 O→CO2 + H2 2.75 × 1010 0 10064.77
From the difference in CO2 concentration between nozzle inlet and
CO2 + H2 →CO + H2 O 9.62 × 1010 − 0.85 15154.28
ejector pipe outlet, the mass flow ratio Ω can be calculated. In addition
H2 + 0.5O2 →H2 O 7.45 × 1013 − 0.91 20085.44
to that, the outlet velocity v3 was measured using a Pitot pipe. The
pressure at the ejector nozzle and after the rotameter 2 was monitored H2 O→H2 + 0.5O2 3.83 × 1014 − 1.05 49552.1

using manometers. C2 H6 + O2 →2CO + 3H2 4.2 × 1011 0 15,034

C2 H6 + 2H2 O→2CO + 5H2 3 × 108 0 15,034

2.3. Ignition delay

The ignition delay time is simulated for different initial conditions
(species concentrations, temperatures, pressure) using OpenFOAM®.
The used solver (chemFoam) is a solver for chemistry problems, designed
for use on single cell cases to provide comparison against other chem­
istry solvers.
Initial conditions, namely pressure p in Pa, temperature T in K,
species concentrations Yj in wt-% or mol-% and the constant property
(volume or pressure), are provided in an ‘initialConditions’ file. Thermo-
physical properties, such as heat capacities cp or the used equation of
state and reaction mechanism are defined in a ‘thermopysicalProperties’
file.
Important output of the simulation are Yj , T and p over time. The
possibility of using and comparing different reaction mechanisms and
the possibility of automated simulation and evaluation, makes Open­
FOAM® and the solver chemFoam a great tool to investigate ignition
delay times and flammability limits under various conditions.
Detailed reaction mechanisms, as described earlier, are usually not
practicable to apply in CFD because of the number of equations to solve.
Thus, usually specialized skeletal mechanisms, containing only few
species and reaction equations are used. In this study the frequently used
Jones-Lindstedt skeletal mechanism (6 species, 4 reactions, Jones &
Lindstedt [17], Table 2) and Hyer skeletal mechanism (7 species, 8 re­
actions, formulation according to Ahmed et al. [18], Table 3) are used in
addition to the detailed GRI3.0 and POLIMI C1-C3 HLT mechanism.
Reaction rate constants are modelled with an Arrhenius type
formulation (equation (29)).
(29)
Ta
β −
k = AT e T
here T is the temperature, A is the pre-exponential factor, β is the
exponent of the temperature and Ta is the activation temperature.
Values in Table 2 and Table 3 are given in the OpenFOAM® native unit
system kmol, m3, s, K.
reaction mechanisms and unreactive flows. Results are compared to the
1D ejector results. The influence of reactions and the different reaction
mechanisms on the ejector behavior is investigated.
The simulated domain consists of nozzle and ejector pipe placed in a
cylinder and a cylindrical box for the free jet (Fig. 3). A constant pressure
and temperature boundary condition was applied at the inlet. At the
outlet a constant pressure condition in conjunction with a pressur­
eInletOutletVelocity condition was used. This allows gas to enter and exit
the simulation domain. The rotational symmetry was considered
applying a wedge boundary condition at the front and back face. All
other patches were considered to be walls with a no slip boundary
condition for velocity and zero gradient condition for temperature. A
hex-dominant mesh was created as a 5◦ -wedge using the OpenFOAM®-
native mesh generation utility blockMesh. The meshes consist of 44 ×
103 cells.
Simulations have been carried out using the local time stepping
(LTS)-approach to reach steady state faster. Turbulence was modelled
using the k-omega Shear Stress Transport (SST) model. The partially
stirred reactor (PaSR) model was used to consider turbulence-chemistry
interactions. This model calculates a finite reaction rate based on the
chemical and turbulence time scales [19]. Temperature dependency of
the gas heat capacity and viscosity where considered using NASA
polynomials and Sutherlands Model, respectively.
3. Results & discussion
Model equations described in chapter 2.1 were used to design two
ejectors for different stagnation temperatures of the low pressure fluid.
To validate the 1D model and CFD simulations, experiments were

2.4. CFD simulation

Gas burner ejectors geometries with and without reactions are

simulated using Computational Fluid Dynamics (CFD). Simulations are
carried out using the OpenFOAM®-solver reactingFoam and a 2D-
axisymmetric approach for both, reactive flows, using different

Table 2
Reaction equations, pre-exponential factors, temperature exponent and activa­
tion temperature of the Jones-Lindstedt mechanism (Jones & Lindstedt [17]).
Reaction A β Ta

CH4 0.5
+ 0.5O2 1.25
→CO + 2H2 4.4 × 10 11 0 15,096
CH4 + H2 O→CO + 3H2 3.0 × 108 0 15,096
H2 0.25 + 0.5O2 1.5 ↔ H2 O 6.8 × 1015 − 1 20,128 Fig. 3. Simulation domain and boundary description. 1 … high pressure fluid
CO + H2 O ↔ CO2 + H2 2.75 × 109 0 10,064 inlet; 2 … low pressure fluid in- and outlet; 3 … rotational axis; 4. Front and
back face.

32
M. Pichler et al. Carbon Resources Conversion 4 (2021) 28–35

carried out to determine the mass flow ratio Ω and outlet velocity v3.
Experiments were performed at surrounding temperature, using air as
high and low pressure fluid. CFD simulated gas residence times with and
without reactions were then compared to predicted ignition delay times
to check for ignition within the ejector. The influence of ignition and
different reaction mechanisms is then shown in chapter 3.4.

3.1. Ejector design

Ejectors were designed for a stagnation pressure and temperature of

the primary fluid and secondary fluid of pi = 1.5bar, Ti = 20◦ C and po =
1.01325bar T0 = 900, 1200 ◦ C respectively, a nozzle diameter of d’’x =
2.5mm and a mass flow ratio Ω = 8. The pressure at position x and
position 3 are set to px = 1.01275bar and p3 = 1.01325bar respectively.
While the diameters d’’x , d1 and d3 (see Fig. 1) can be calculated using
equations (4)-(27), there are no correlations for the length of the Fig. 4. Example of measured CO2 concentrations and velocities at the ejector
different injector sections. In this study a value of Lj /d1 = 1.2 was used outlet. Phase1: 0–66 s; Phase 2: 66–276 s; Phase 3: 276 – 420 s.
for the constant pressure mixing length. A value of Lj /d1 = 4 was used
for of the constant area mixing and the diffusion zone. Here Lj is the
length of the respective zone and d1 is the diameter of the constant area-
mixing zone. Resulting ejector dimensions, outlet temperatures T3 ,
outlet velocities v3 and mean residence times τ for two exemplary
ejectors are presented in Table 4.

3.2. Experiments

Experiments have been carried out with air at surrounding condi­

tions (p0 = 1 bar, T0 = 22 ◦ C) for the low pressure fluid, and compressed
air as a high pressure fluid using 3D-printed ejector parts. The experi­
ments were conducted as described in chapter 2.2. An example for the
measured concentrations and outlet velocities is shown in Fig. 4. Under
this conditions the calculated mass flow ratio using equations (4)-(27) is
Ω = 12.8 and the outlet velocity is v3 = 24.95m/s. The experiment was
carried out seven times under the same conditions. The average mass
flow ratio is Ωexp = 12.41 ± 1.0 (standard deviation: 0.27). The calcu­
lation of Ω is very sensitive to the measured CO2 concentration at the
ejector outlet, thus the tolerance of Ω is relatively high. The measured
average outlet velocity v3,exp = 28.46 ± 0.7m/s (standard deviation:
2.3m/s). Considering the accuracy, the measured average values do
match the predicted results.
3.3. Ignition delay
A comparison of the simulated ignition delay times, using the four
different reaction mechanisms and the experimental ignition delay times
(Petersen, et al. [12]) for a methane/air mixture at 19.4 bar is shown in
Fig. 5. Simulated results get more accurate when using more complex
reaction mechanisms. The complexity and accuracy increases from the
Jones-Lindstedt mechanism, to the Hyer mechanism, the GRI3.0
Table 4
Important results of the 1D ejector design.
Parameter, dimension Symbol T0 = T0 =
900 C 1200 C
◦ ◦
Fig. 5. Comparison of the simulated ignition delay times of a methane/air
mixture at 19.4 bar using the Jones-Lindstedt, Hyer skeletal mechanism, the
GRI3.0 mechanism and the POLIMI C1-C3 HLT mechanism to experimental
results from Petersen et al. [12].
mechanism and the POLIMI C1-C3 HLT mechanism. Here the ignition
delay time is defined as the time where the temperature increase ΔT/Δt
takes a maximum value.
As stated before, the ignition delay time and flammability limits are a
function of the exact conditions (species concentration, temperature and
pressure). In this study a number possible mixtures of fuel (99 wt% CH4,
1 wt% C3H8) and air/exhaust gas (with different O2 concentrations) at
different initial temperatures were simulated. The air/exhaust gas is
specified as a mixture of N2, O2, CO2 and H2O. The initial concentrations
of the species in the air/exhaust gas is changed based on the difference
between the oxygen concentration in air (23.135 wt%) and the cases
initial oxygen concentration. This is done according to the stoichiometry
of complete methane combustion with oxygen:
CH 4 + 2O2 →CO2 + 2H2 O

In doing so, the ignition delay time and flammability limits of all
diameter ejector pipe inlet, mm d’’x 47.5 50.3
possible Ω and different compositions of the low-pressure fluid were
diameter constant area mixing, mm d1 28.15 30.8
covered. Each simulated case was evaluated for ignition and ignition
diameter ejector pipe outlet, mm d3 28.9 31.8
delay time. Ignition was defined as a case where the following condi­
length of constant pressure mixing zone, Lx,1 33.8 36.9
tions are met:
mm
length of constant area mixing zone, mm L1,2 112.6 123.1
length of diffuser zone, mm L2,3 112.6 123.1 • temperature increase ΔTmax = Tmax − T0 ≥ 250K
outlet temperature, ◦ C T3 666.4 871.9 • ignition delay time τigni ≤ 0.2s
outlet velocity, m/s v3 53 54.7
mean residence time, ms τ 3.26 3.38 The ignition delay time was defined as the time where the

33
M. Pichler et al. Carbon Resources Conversion 4 (2021) 28–35

temperature increase ΔT/Δt takes a maximum value.

The simulation of flammability limits and ignition delay times was
carried out for all possible combinations of settings according to Table 5
using the POLIMI C1-C3 HLT reaction mechanism.
Out of the 33,936 simulations, 3381 cases have ignited. Ignited cases
at a given initial gas temperature where plotted in a ternary plot. Only
eight simulated cases have ignited for an initial temperature of 1000 K. It
is safe to assume that no ignition takes place below this temperature.
Ignited cases for an initial temperature of 1140 K, corresponding to
the ejector with T0 = 1200 ◦ C, are shown in Fig. 6. The color indicates
the ignition delay time for the respective mixture. The dashed grey line
represents all possible mixtures of natural gas (NG) and air, thus also
represents the ratio of mass flow rates of secondary to primary stream Ω.
This ratio can be calculated out of the desired natural gas concentration
[NG] according to equation (30).
1 − [NG]
Ω= (30)
[NG]
Each ejector design can be checked for possible ignition in the
ejector, when assuming perfect mixing and using equation (30), the
mean residence time in the ejector τ, the outlet temperature T3 and the
ignition delay time τigni (e.g. from a ternary plot). If τigni > τ it is assumed
that no ignition takes place within the ejector tube. If τigni < τ ignition
might happen within the ejector. Fig. 6. Ternary plot of natural gas (NG) oxygen (O2) and air/exhaust gas (Rest)
For the exemplary ejector with T0 = 900◦ C the predicted outlet including region of flammability and ignition delay times. An exemplary
temperature T3 is 939.5 K (666.4 ◦ C). For this temperature and a fuel designed ejector is marked with a red circle.
concentration of 11.1 wt% (corresponding to Ω = 8 according to
equation (30)) at the ejector outlet, no ignition takes place in the igni­
tion delay time study. Table 6
For a higher temperature of the secondary gas stream of T0 = 1200 CFD results for two different ejector designs. Reactions are not included.
◦
C the predicted outlet temperature T3 = 1145 K (871.9 ◦ C). At this no reactions T0 = 900 C
◦
T0 = 1200 C
◦

temperature and for a natural gas concentration of 11 wt% ignition can

mass flow ratio Ω, - 7.62 7.8
be achieved, as indicated in Fig. 6. However, when comparing the outlet velocity v3 , m/s 48.8 50.9
ignition delay time τigni = 123.6ms to the predicted mean residence time outlet temperature T3 , ◦ C 663.7 874.5
τ = 3.38ms it can be seen that, despite the point lying within the flam­ mean residence time τ, ms 3.4 3.3
mability limits, no ignition is likely to be achieved in the ejector.

3.4. CFd Results deviate significantly from 1D to CFD simulation when

including gas phase reactions, as shown in in Table 7. The outlet tem­
CFD simulations have been carried out for conditions and ejector perature T3 indicates that the fuel/gas mixture is igniting within the
designs according to Table 4. Local time stepping (LTS) was used to ejector tube. A detailed evaluation of the CFD simulation (Fig. 7) shows
reach steady state faster. Gas phase reactions were considered using the that ignition takes place in the entrainment zone of the ejector tube
detailed GRI 3.0 and the skeletal Hyer mechanism. The high pressure (indicated by temperature). When the secondary gas is sucked in, the
fluid was natural gas with (99 wt.% methane, 1 wt.% propane). The local conditions (temperatures, gas composition) differ significantly
secondary gas stream was air. Simulations were evaluated regarding the from the average values. Thus, in such mixing zones the ignition con­
ratio of mass flow rates of secondary to primary stream Ω, the temper­ ditions can be met and ignition is possible with very low ignition delay
ature T3 and velocity u3 at the ejector outlet and the mean residence times. This shows that the assumption of perfectly mixed gases is not
time τ of both, reacting and non-reacting flows. For each of the simu­ valid. Using a skeletal reaction mechanism, the ignition delay time is
lations, the presence or lack of ignition within the ejector was evaluated clearly under predicted (Fig. 5, Hyer), compared to the more accurate
for the different reaction mechanisms. detailed mechanism. This leads to even higher fuel conversion and outlet
Important results for the two different ejectors, excluding possible temperatures T3 .
reactions, are presented in Table 6. While the simulated mass flow ratios Results shown in Table 7 also reveal that ignition in the ejector has
Ω are slightly lower than the design specification, good agreement with significant influence on the mass flow ratio Ω, outlet velocity v3 and
the 1D results shown in Table 4 is achieved for the outlet velocity v3 , mean residence time τ. Using the detailed reaction mechanism the
outlet temperature T3 and the mean residence time τ. This also confirms
that a 2D-axysymetrical approach is sufficient to describe the flow and
important phenomena in the ejector. Table 7
CFD results for different reaction mechanisms and different secondary gas
temperatures T0 .
Table 5
Simulation settings for the simulation of flammability limits and ignition delay Including reactions T0 = 900 ◦ C T0 = 1200◦ C
times. Reaction mechanism Hyer GRI 3.0 GRI 3.0
Property min max Δ mass flow ratio Ω, - 4.12 5.67 5.8
outlet velocity v3 , m/s 60.82 53.2 57.7
natural gas, wt-% 0 50 0.5
outlet temperature T3 , ◦ C 1328.25 988.7 1410.3
O2, wt-% 0.1350 23.135 1.15
T, K 1000 1300 20 mean residence time τ, ms 3.15 3.45 2.75

34
M. Pichler et al.
Fig. 7. CFD simulated temperature field for the designed ejector with T0 =
900 ◦ C and different reaction mechanisms. (a) GRI3.0 mechanism, (b)
Hyer mechanism.
simulated Ω differs by 41 % compared to the design case. When using
the skeletal Hyer reaction mechanism, Ω drops by nearly 50 % (differ­
ence of 94 %) compared to the design case. As ejectors and gas burner
ejectors are often design to reach certain levels of Ω, the drop of leads to
performance far off the design specifications. The direct consequence of
a low mass flow ratio is that predicted concentrations in the ejector are
not reached. This leads to undesired fuel–air ratios. Lower mass flor
ratios also lead to lower gas velocities at the ejector pipe outlet. This
significantly influences free jet properties such as the penetration depth
in cross flow, the entrained gas mass in the jet or mixing of gases by the
free jet [20]. The GRI 3.0 mechanism is considered more accurate, thus
the CFD results using the detailed mechanism are considered the most
realistic amongst the tested simulation methods.
4. Summary and conclusion
The model equations based on the work of Keenan, Neumann, &
Lustwerk [2] have been successfully modified and applied to design gas
burner ejectors at low Mach numbers M1 ≪1 and pressure ratios pi /p0 .
1D results showed good agreement with 2D CFD simulations (when
reactions where not considered) and experiments with air at room
temperature. A detailed study of flammability limits and ignition delay
times for mixtures of natural gas (99 wt.% CH4, 1 wt.% C3H8) and air
was performed. Results showed that the ignition delay time decreases
with decreasing fuel content and increasing oxygen content. Comparing
the simulated ignition delay times (τigni = 123.6ms) to the mean resi­
dence time of the gas in the ejector (τ = 3.38ms) and assuming perfect
mixing along the ejector suggests, that no reaction takes place inside the
ejector. 2D axisymmetric CFD simulations, using the detailed GRI3.0
reaction mechanism, revealed that this assumption is not correct. Igni­
tion and combustion is possible, even if the ignition delay time is more
than thirty times higher than the mean residence time. Thus it is not
possible to predict ignition in the ejector by comparing ignition delay to
mean residence times. This shows that over simplification should be
avoided.
It was also shown that the simulated ejector performance depends on
the used reaction mechanism. To predict an accurate ejector behavior, it
is mandatory to correctly consider ignition delay. Thus, the use of
detailed reaction mechanisms is essential.
To verify the simulated gas burner and ignition behavior using
35
Carbon Resources Conversion 4 (2021) 28–35
natural gas and high temperatures of the secondary fluid, performing
experiments under this conditions is recommended in future studies.
CRediT authorship contribution statement
Mario Pichler: Investigation, Software, Validation, Methodology,
Writing - original draft, Visualization. Florian Wesenauer: Investiga­
tion, Validation. Christian Jordan: Data curation, Supervision. Stefan
Puskas: Project administration, Funding acquisition. Bernhard Streibl:
Formal analysis, Methodology, Project administration. Franz Winter:
Project administration, Funding acquisition. Michael Harasek: Project
administration, Supervision, Funding acquisition.
Acknowledgements
Financial support was provided by the Austrian research funding
association (FFG) within the research project “Entwicklung eines inno­
vativen Tunnelofen – Energiekonzeptes mit Reingasbrennern und
Energieeffizienter Prozesstechnik (TOREtech)” (FFG project # 865020).
References
[1] D.-W. Sun, I. Eames, Recent developments in the design theories and applications
of ejectors, J. Instit. Energy, 68 (1995) 65–79, 6.
[2] J.H. Keenan, E.P. Neumann, F. Lustwerk, An investigation of ejector design by
analysis and experiment, J. Appl. Mech. (1950) 299–309.
[3] C.J. Lawn, A simple method for the design of gas burner injectors, Proc. Instit.
Mech. Eng., Part C: J. Mech. Eng. Sci., 217 (2003). 237–246, 2.
[4] S.K. Yadav, K.M. Pandey, R. Gupta, Recent advances on principles of working of
ejectors: a review, Mater. Today:. Proc. (2020) 12.
[5] N.S. Kumar, K.T. Ooi, One dimensional model of an ejector with special attention
to Fanno flow within the mixing chamber, Appl. Thermal Eng., 65 (2014) 226–235,
4.
[6] W.L. Chow. A.L. Addy, Interaction between primary and secondary streams of
supersonic ejector systems and their performance characteristics, AIAA J., 2 (1964)
686–695, 4.
[7] A. Namkhat, S. Jugjai, Primary air entrainment characteristics for a self-aspirating
burner: model and experiments, Energy, 35 (2010) 1701–1708, 4.
[8] V. Kumar, G. Singhal, P.M.V. Subbarao, Realization of novel constant rate of
kinetic energy change (CRKEC) supersonic ejector, Energy, 164 (2018) 694–706,
12.
[9] F. Joos, Technische Verbrennung, Springer, Berlin Heidelberg, 2006.
[10] I.A.B. Reid, C. Robinson, D.B. Smith, Spontaneous ignition of methane:
Measurement and chemical model, Symposium (International) on Combustion, 20
(1985) 1833–1843, 1.
[11] E.L. Petersen, M. Röhrig, D.F. Davidson, R.K. Hanson, C.T. Bowman, High-pressure
methane oxidation behind reflected shock waves, Symposium (International) on
Combustion, 26 (1996) 799–806, 1.
[12] E.L. Petersen, J.M. Hall, S.D. Smith, J. de Vries, A.R. Amadio, M.W. Crofton,
Ignition of lean methane-based fuel blends at gas turbine pressures, J. Eng. Gas
Turbines Power 129 (2007) 937.
[13] E.L. Petersen, J.M. Hall, D.M. Kalitan, M.J.A. Rickard, Ignition Delay Time
Measurements of C, in Volume 1: Turbo Expo 2004, (2004).
[14] G.P. Smith, D.M. Golden, M. Frenklach, N.W. Moriarty, B. Eiteneer, M. Goldenberg,
C.T. Bowman, R.K. Hanson, S. Song, J. William C. Gardiner, V.V. Lissianski und Z.
Qin, GRI 3.0, visited on: 13.08.2019. Available: http://www.me.berkeley.edu/gri_
mech/.
[15] POLIMI C1-C3 HLT, visited on: 13 08 2019. Available: http://creckmodeling.chem.
polimi.it/menu-kinetics/menu-kinetics-detailed-mechanisms/menu-kinetics-c1-c3-
mechanism.
[16] S. Gordon, B. McBride, Computer program for calculation of complex chemical
equilibrium composition, rocket performance, incident and reflected shocks and
Chapman-Jouguet detonations, NASA Report SP-273 (1971).
[17] W.P. Jones, R.P. Lindstedt, Global reaction schemes for hydrocarbon combustion,
Combust. Flame, 73 (1988) 233–249, 9.
[18] G. Ahmed, A. Abdelkader, A. Bounif, I. Gökalp, Reduced chemical kinetic
mechanisms: Simulation of turbulent non-premixed CH4-Air flame, Jordan J.
Mech. Ind. Eng., 8 (2014) 66–74, 1.
[19] The OpenFOAM Foundation, OpenFoam v8 User Guide, 2020.
[20] U. Ketels, Possibilities for obtaining uniform temperature distribution in the tunnel
kiln setting, Ziegelindustrie Int. 7–8 (1990) 410–419.