Energy Conversion and Management 198 (2019) 111155

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

Combustion in wavy micro-channels for thermo-photovoltaic applications – T

Part I: Effects of wavy wall geometry, wall temperature profile and reaction
mechanism
Zakaria Mansouri
Alternative Energies and Atomic Energy Commission, DRT/LITEN/DTBH, 17 Avenue des Martyrs, 38000 Grenoble, France

ARTICLE INFO ABSTRACT

Keywords: The output power of a Micro-Thermo-Photovoltaic (MTPV) system strongly depends on the micro-combustor
CFD geometry and the micro-flame stability. In this context, numerous numerical and experimental works have been
Wavy micro-channel conducted and focused on geometries limited to circular and planar micro-channels (MCs) with and without
Thermo-photovoltaic sudden expansion. Therefore, the present work proposes a novel MC configuration with wavy walls. The use of
Micro-flame stability
wavy walls increases the MC surface and consequently the heat transfer characteristics of this system could be
Reaction mechanisms
improved. In this study, two-dimensional numerical simulations on CH4–air premixed flames in the novel
configuration have been carried out. The wavy MCs are heated with a wall temperature gradient in the
streamwise direction. The investigations have been carried out with a MC height H = 600 μm, a mixture velocity
U = 1 m/s and the equivalence ratio Φ = 1. Moreover, the effects of various parameters on the flame behavior
are reported: the wavy wall geometries, the maximum values of the temperature profile and the kinetic me-
chanisms of CH4 combustion. The results are original, since the effect of the wavy wall geometries are not
available in the literature. In addition, the effects of the other parameters are not well documented in the
literature, especially the CH4 combustion mechanisms. It is found that the flame location is sensible to the wavy
wall geometry. As the number of waves increases, the flame moves towards the MC exit. The maximum value of
the temperature gradient has an opposite effect. As the maximum temperature increases, the flame moves to-
wards the MC inlet. The results show also that the flame location, shape and temperature are strongly affected by
the different reaction mechanisms.

1. Introduction micro-channels (MCs), the absorber, the emitter, the photovoltaic (PV)
cell and the heat sink. Within this system, the MC serves as a combustor
Micro-Thermo-Photovoltaic (MTPV) systems have received sub- which increases the temperature of the emitter forming an optical he-
stantial attention over the recent years due to the advances in the ated source (an average temperature of 1100 °C). Some of the thermal
micro-fabrication capability. They are efficient and economically viable emission (emission of photons) is then absorbed by the PV cell. This
power systems for commercial and space applications. For commercial later converts the radiated photons into electricity.
applications, they can be used to replace batteries of laptops and ta- The MC is the key component of the MTPV system, which sig-
blets. They can also provide a continuous supply of power in off-grid nificantly affects the output power and energy conversion efficiency,
homes. For space applications, they are proposed to supply power to Zuo et al. [1]. The MC faces generally the flammability and flame
orbital micro-satellites by replacing a fraction of the PV solar cells. quenching issues, Durisch et al. [2]. This makes the flame stability the
MTPV system transforms the chemical energy to thermal energy and most important issue and the key parameter to improve the thermal
then to electricity. The chemical and thermal energies are supplied by performance of the MC. Recently, several classic and advanced methods
combustion of a gaseous fuel, hydrocarbons or hydrogen (H2), in a are used to address the flame stability, such as: sudden expansion,
micro-combustor. The use of hydrocarbon fuels is beneficial due to their Akhtar et al. [3]; bluff-body, Yan et al. [4]; catalytic surface, Li et al.
energy densities (an average of 40 MJ/kg), which is much higher than [5]; porous media, Pan et al. [6] and external heating, Maruta et al. [7].
the actual best batteries (an average of 1 MJ/kg). MTPV system is The last method is the focus of the current study.
generally composed of the main parts shown in Fig. 1, which are: the MC with an external heating is found also in the literature as MC

E-mail addresses: [email protected], [email protected].

https://doi.org/10.1016/j.enconman.2018.12.105
Received 5 October 2018; Accepted 22 December 2018
Available online 17 January 2019
0196-8904/ © 2019 Elsevier Ltd. All rights reserved.
Z. Mansouri Energy Conversion and Management 198 (2019) 111155

Fig. 1. Schematic illustration of a Micro-Thermo-Photovoltaic system.

with a controlled temperature profile or gradient, Di Stazio et al. [8] suppression can be due to a few important elementary reactions, which
and Kamada et al. [9]. The majority of the reported MCs are focused on play a dominant role in contributing to the heat release rate. In the
cylindrical tubes (circular micro-reactors) and parallel plates (planar external temperature profile works, Wang et al. [16] conducted tran-
micro-reactors). These geometries make the MC a simple system, which sient numerical simulations of premixed CH4–air combustion in a
is a great advantage to investigate complex combustion chemistry. narrow channel. They imposed a fixed hyperbolic tangent temperature
Moreover, the inner diameter/height of the MC is smaller than the profile along the MC wall, starting from 300 K and ending with 1300 K
ordinary quenching diameter/height of the flame. The MC is heated by or 1400 K. Depending on the wall temperature profile, two instable
an external heat source and a stationary temperature profile along the flames were observed: the oscillating flame and the FREI. The oscil-
inner surface of the MC is developed in the streamwise direction. The lating flame corresponds to the high wall temperature profile, whereas
external heat source can be an electric heater, Nakamura et al. [10], or the FREI corresponds to the low wall temperature profile. Kang et al.
a hydrogen burner, Kikui et al. [11]. The MC operates under premixed [17] conducted transient numerical simulations of premixed CH4–air
condition and the mixture temperature strongly depends on the tem- combustion in a planar MC. They investigated the effect of different
perature profile of the wall. In addition, the flow mixture within the MC wall temperature profiles on the flame dynamics for different inflow
is laminar and at a constant pressure. In the previous studies using velocity conditions. For low velocity, such as 0.2 m/s, the flame shows
circular MC reactor, Maruta et al. [7]; Di Stazio et al. [8] and Kamada spatial oscillations and even flame extinction. For high velocity, such as
et al. [9], three types of flames are observed: stable flames, flames with 1 m/s, the flame is stabilized for all the wall temperature profiles ex-
repetitive extinction and ignition (FREI) and weak flames, depending amined.
on the mixture velocity at the inlet. The stable flame is an ordinary In the light of the above review, none of the aforementioned studies
steady flame and observed at a high flow velocity (around 0.5–1.2 m/s). characterized the combustion and micro-flame stability in a wavy MC.
The FREI occurs at an intermediate flow velocity (around 0.3–0.45 m/ The majority of these works adopted circular or planar MC. In addition,
s). The weak flame is stable with quite weak chemiluminescence and there is insufficient knowledge on the effect of kinetic mechanisms on
observed at a low flow velocity (around 0.05–0.25 m/s). the characteristics and stability of the micro-flames. A greater number
The micro-flame dynamics and behavior strongly depend on other of simulated premixed CH4–air MC, such as Kang et al. [15], Wang et al.
parameters than the mixture velocity, such as: the dimension of the MC, [16], Kang et al. [17] and Kang et al. [18], used the DRM-19 reaction
the fuel type and the external temperature profile. In the MC dimension mechanism of Kazakove and Frenklach [19]. Only few studies used or
studies, Norton and Vlachos [12] studied numerically the effects of the compared other reaction mechanisms to characterize the combustion of
MC diameters and thickness of walls on combustion characteristics and CH4–air in MC. Thus, the need of the present study is to further increase
flame stability. They found that the wall thickness is very important as the knowledge on CH4 combustion in MTPV systems, using various
it determines the upstream heat transfer, which is necessary for flame reaction mechanisms. Also, to propose innovative MC configurations
ignition and stability. For large diameters, such as 4 mm, the reaction with the aim to improve the efficiency of MTPV systems by enhancing
rate is low and the flame spreads out over approximately the half of the the heat transfer characteristics. This goal could be achieved by in-
burner length. For smaller diameters, such as the 600 μm, reaction is creasing the heat exchange surface within the MC by using wavy walls.
more localized, with greater intensity, causing steeper temperature and Moreover, different wavy walls are proposed to assess the complexity of
composition gradients. Li et al. [13] conducted a 2D numerical study on the combustion within the MC. Furthermore, this paper uses CFD
CH4–air premixed combustion in both cylindrical MC and planar MC. (Computational Fluid Dynamics) to fulfill the above points. Three main
Their results showed that the planar channel with H = 1 mm represents parameters are investigated, which are: the wavy wall geometry, the
higher flame temperature than a cylindrical tube with D = 2 mm, over wall temperature profile and the reaction mechanisms of CH4.
a velocity range varying from 0.3 to 1 m/s. In the fuel type studies, Tang
et al. [14] fabricated and experimentally investigated a planar MC
operating with C3H8/H2–air mixture. They built also a 3D calculation 2. Numerical procedure
model to characterize the combustion process of the studied blended
flue. It is reported that the H2 addition can overcome the shortcomings 2.1. Geometric model and grid
of propane flame instability. When a small amount of hydrogen is
added, the flame location could be fixed. The H2 enriched fuel can In this study, four planar MC configurations are modeled in two-
further reduce the minimum quenching diameter of the MC. Kang et al. dimensions (2D). The first configuration consists of two parallel flat
[15] investigated numerically the suppression of the flame instability in plates, as shown in Fig. 2. This configuration is chosen to perform va-
a premixed CH4–air planar MC using H2/CO addition. By adding small lidation of the CFD calculations with the results in the literature. The
amounts of H2 or CO, flame instabilities present for the methane air MC length (L) and height (H) are 6 mm and 0.6 mm, respectively.
mixture in the form of FREI could be effectively suppressed. The Premixed CH4–air is fed to the MC. A fixed temperature profile is im-
posed along the MC walls to initiate and maintain the combustion. The

2
Z. Mansouri Energy Conversion and Management 198 (2019) 111155

unit tensor.
Energy conservation equation

2
· u ( E + p) = · T hi Di + µ ( u + ( u ) ) ·u I ·
i 3

u + Sh
(3)
Fig. 2. Computational domain of the flat MC along with the temperature profile
where E is the total energy, is the mixture conductivity, T is the
used.
temperature, hi is the enthalpy of the species i , Di is the diffusion flux of
species i and Sh is the source term of the mixture enthalpy.
Species conservation equation

( u Yi ) = Di + Ri (4)
where Yi is the mass fraction of species i and Ri is the net production
rate of species i by chemical reaction.
The details of these equations term by term can be found in Poinsot
and Veynane [22] and Alipoor and Mazaheri [23].

2.3. Numerical method

Fig. 3. Schematics of computational domains of the wavy MCs.

The simulations are done using the CFD code ANSYS-Fluent R19.0.
It is reliable CFD software, which involves broad physical modeling
wall temperature ramps from 300 K to a higher value at 1400 K over the abilities including chemical reactions in complex geometries, ANSYS
initial 1 mm of the MC length according to a hyperbolic tangent func- [24] and Mansouri et al. [25–28]. It uses the finite-volume method to
tion. solve the flow governing equations. In the present study, two solvers are
The other configurations consist of an upper flat wall and a lower used and compared to solve the governing equations. The first is den-
wavy wall, as shown in Fig. 3. The wavy wall is used to increase the sity-based solver for compressible flows. It is chosen to perform code
contact surface between the reactor (MC) and the absorber. Then, more validation with a similar compressible code of Kang et al. [18], using
energy will be absorbed and emitted to the PV cell due to the en- the MC configuration in Fig. 1. Note that the standard density-based
hancement of the heat transfer. Consequently, the MTPV system de- solver may have stability problems due to the considerable change in
livers more electric output compared to the basic system with flat wall. the mixture density. This causes the formation of unwanted acoustic
The wavy MC length (L) is 6 mm and its minimum and maximum wave speeds and small-magnitude flow velocities. To overcome these
heights are Hmin = 0.55 mm and Hmax = 0.65 mm, respectively. The problems, the low Mach number approach should be used. This can be
difference between these configurations is the number of waves at the done by decomposing the pressure into two terms: thermodynamic
lower wall, which are: 2-waves, 4-waves and 8-waves. Note that the pressure and hydrodynamic pressure, Alipoor and Mazaheri [23] and
previous hyperbolic temperature profile (Fig. 2) is imposed also along Tomboulides et al. [29]. ANSYS-Fluent uses an alternative method, the
the wavy MC walls. co-called Advection Upstream Splitting Method (AUSM), Liou [30,31]
For all MC configurations, structured grids of 273 × 27 are em- to solve the compressible flow without modifying the governing equa-
ployed. The grids are uniform and not refined in order to keep the same tions. It is proved to be reliable and effective for low Mach number
spatial distribution over the MC, where the flame can be stabilized at reactive flows. This method is adopted in the present study similarly to
any location through the MC. The cell size within these grids is 22 μm. Kang et al. [18]. The second solver is pressure-based solver for in-
This value ensures sufficient grid points within the flame thickness of compressible flows. It is widely used in the literature, Zuo et al. [1]; Li
around δf = 0.25 mm for the present case. Note that available studies of et al. [12] and Wang et al. [16] and used for the present study due to its
similar flat micro-channels employed grid sizes of 50 μm, Kang et al. low computational time compared to the previous solver. Pressure
[15], and 25 μm, Nair et al. [20] and Kishore et al. [21]. Moreover, no Implicit Split Operator (PISO) algorithm is employed to solve the
need to perform grid-independent simulations since the employed cell pressure-velocity coupling. It is originally developed for non-iterative
size is lower than those used in the literature. computation of unsteady compressible flows, but it has been adapted
successfully to steady-state problems. A second-order upwind scheme is
2.2. Mathematical model used to discretize the governing equations. The convergence criteria for
all equations are set to 10−5, except for the energy equation; it is set to
Before applying the mathematical governing equations, the fol- 10−6. The stiff chemistry solver is used to remedy the stability and
lowing assumptions are made: neglected wall thickness, laminar flow, convergence problems of the reacting flow.
steady state, compressible flow, reactive flow, no surface reaction, no
radiation and no Dufour effects. The conservation equations for mass, 2.4. Boundary conditions
momentum, energy, and species are summarized as follows:
Mass conservation equation At the inlet of the domain, a mass flux (m = 1.122 kg/m2/s)
boundary condition is applied, which corresponds to an inflow velocity
·( u ) = 0 (1)
of 1 m/s. The mixture temperature T = 300 K is used with mass frac-
where is the mixture density and u is the velocity vector. tions of incoming species (N2, O2 and CH4), that corresponds to the
Momentum conservation equation stoichiometry (Φ = 1). At the outlet, atmospheric pressure boundary
condition is imposed. At the MC walls, no-slip boundary condition and
2
(u · u ) = p+ · µ( u + ( u ) ) ·u I zero-flux diffusion for all species are used. In addition, a hyperbolic
3 (2)
tangent temperature profile is imposed along the upper and lower
where p is the absolute pressure, µ is the dynamic viscosity and I is the walls.

3
Z. Mansouri Energy Conversion and Management 198 (2019) 111155

The wall temperature ramps from 300 K to a higher value at 1400 K

over the initial 1 mm of the MC length to respect the distance needed
for a fully developed flow, as done by Kang et al. [18].

2.5. Combustion mechanisms

In this study, five kinetic mechanisms of CH4–air combustion are

employed and compared. The mechanisms (schemes) include: one
global reaction mechanism, on detailed full chemistry mechanism and
three reduced mechanisms. The global reaction mechanism is that of
Westbrook and Dryer [32]; it consists of two reactions (2-step), where
the reaction of CO2 and CO is reversible. This reaction mechanism is
used in a major part of this work including the code validation, effect of
wavy channel models and effect of temperature profile. The detailed
reaction mechanism is the so-called GRI 3.0 of Gregory et al. [33]. The
mechanism consists of 325 reactions that involve 53 species and de-
scribes the full chemistry of CH4–air combustion. The three reduced
mechanisms are those of Kazakove and Frenklach [19] (DRM-19 me-
chanism), Smooke et al. [34] (Smooke mechanism) and Yand and Pope
[35] (Skeletal mechanism). The DRM-19 mechanism consists of 84 re-
actions that involve 20 species. The Smooke mechanism consists of 46
reactions that involve 17 species. The Skeletal mechanism consists of 41
reactions that involve 16 species. All these mechanisms are truncated
from the detailed GRI 3.0 mechanism. Fig. 4. Simulations versus numerical data: centerline profiles of (a)
Temperature and (b) CH4 mole fraction.
3. Code validation

algorithms of each solver. It is interesting to mention that the density-

As far as the authors’ knowledge, there are no existing experimental
based solver used is explicit and present a strong interdependence,
works dealing with the detailed characteristic of micro-flames, such as:
(coupling) between density, energy, momentum and species. Whereas
flame temperature, heat release, flow velocity and species concentra-
the pressure-based solver is implicit and it takes momentum and pres-
tions within a MC. This is due to lack of accessibility to these char-
sure as the primary variables, then passes to the energy and species. It is
acteristics using actual instrumentations of measurements and optical
well known that the pressure-based solver requires less memory (sto-
diagnostic within such micro systems (MC diameter < 1 mm).
rage) compared to the density-based solver. Consequently, the calcu-
Meanwhile, micro-flame behavior could be investigated experimentally
lations using the pressure-based solver for the present study are 20
to study the flame shape, ignition, extinction and instabilities, Maruta
times faster than those of density-based solver. Regarding the tem-
et al. [7]; Di stazio et al. [8] and Kamada et al. [9]. Moreover, few
perature difference it should note that the adiabatic flame temperature
numerical works have based their code validation on experimental re-
of CH4–air at atmospheric pressure, initial temperature of 298 K and
sults from meso-scale reactors (reactor diameter > 10 mm), Kang et al.
stoichiometry condition is around 2263 K, Haynes [36]. The differences
[15] and Wang et al. [16]. Their results were compared with only a
between the adiabatic flame temperature and the compressible and
temperature profile at the centerline of the reactor. In the authors’ point
incompressible temperatures are 525 K and 385 K, respectively. These
of view, it is inappropriate to perform the validation as the same as the
large differences are due to the 2-step reaction mechanism which de-
pervious works cited above. This is due to the scale difference between
scribes a global chemistry inducing high combustion temperatures.
the targeted MC and the meso-scale reactors (10 times larger) and the
Thus, the incompressible temperature is more close to the adiabatic
insufficient data for validation. Therefore, the numerical study of Kang
flame temperature. As a conclusion, the pressure-based solver will be
et al. [18] is adopted to perform the code validation using the MC
used for the rest of the study due to its fast computational time and
model presented in Fig. 1.
acceptable temperature prediction.
Fig. 4 shows comparisons of centerline temperature (Fig. 4a) and
Fig. 5 illustrates contours of temperature (Fig. 5a), heat of reaction
centerline CH4 mole fraction (Fig. 4b) along the MC. Note that the re-
(Fig. 5b) and axial velocity (Fig. 5c) for both compressible and in-
sults from both solvers (density-based and pressure-based) are com-
compressible calculations. The figure gives a clear idea about the,
pared with those of Kang et al. [18]. Generally, the mixture is pre-he-
temperature distribution along the MC, flame structure and location
ated in the region x < 0.002 m (zone 1) due to the hyperbolic
and velocity acceleration across the flame zone. The incompressible
temperature profile of the walls. Then, the combustion take place at
gradients are shifted slightly in the streamwise direction compared to
around x = 0.0026 (zone 2) accompanied with a sudden rise of tem-
the compressible gradients. It was referred to the difference between
perature and a perfect consumption of the fuel. Downstream this lo-
the calculation algorithms of each solver. The three zones mentioned
cation (zone 3) the temperature decreases considerably till the MC exit
previously (Fig. 4) can be clearly distinguished now. In the pre-heating
without traces of methane. Particularly, a good agreement between the
zone 1, the walls temperature is significantly higher than the mixture
present compressible results and the numerical data from Kang et al.
temperature, so the heat transfers from the walls to the mixture. This
[18]. This is referred to the use of the same numerical method (low
leads to decrease the mixture density and then accelerating the fluid
Mach number approach) to model the MC. However, the in-
from 1 m/s to around 6 m/s. Since the mixture is heated from the walls
compressible results show slight differences compared to the com-
towards the MC center, ignition occurs near the walls and the flame
pressible calculations. Within the pre-heating zone a good agreement is
stabilizes at the MC center (zone 2), consistent with the experimental
found, but within the flame zone the incompressible profiles are shifted.
studies of Maruta et al. [7] and Di Stazio et al [8]. The ignition tem-
For example, the temperature peaks take the values 2761 K and 2621 K
perature of the CH4–air is approximately 873–923 K according to Ro-
at the locations x = 0.0026 m and x = 0.0027 m for the compressible
binson and Smith [37]. After the ignition, mixture combusts quickly
and incompressible profiles, respectively. The difference of temperature
releasing the heat of reactions, reaching 15–19 W at the MC center. The
value and location could be due to the difference of the calculation

4
Z. Mansouri Energy Conversion and Management 198 (2019) 111155

4.2. Effect of asymmetric temperature profiles

In a real MTPV system, the heating source (temperature profile) of

the MC is applied only at a single wall (the upper wall for the present
study). The lower wall of the MC is in contact with the absorber and its
temperature should be less than that of the upper wall. For this reason,
this sub-section aims to explore the effect of asymmetric temperature
profiles on the flame behavior. Thus, the boundary condition at the
lower wavy wall for all models is changed and a constant temperature
of 1200 K is imposed instead of the hyperbolic temperature profile.
Fig. 7 shows the effect of asymmetric temperature profiles on the
flame location presented by centerline temperature (Fig. 7a) and cen-
terline CO mole fraction (Fig. 7b) for all wavy MC. Note that the CO
species could be used to identify the flame location, as used by Man-
souri et al. [38]. For the present asymmetric condition, the flame is
located at x = 0.00278 m, x = 0.00298 m and x = 0.00369 m for the
configurations of 2-waves, 4-waves and 8-waves, respectively. These
Fig. 5. Comparison between the compressible and incompressible simulations: new flame locations are offset from the precedent locations. For ex-
(a) temperature distribution (b) flame location and (c) velocity field. ample, the 2-waves flame is shifted by 0.28 mm compared to it old
location using a symmetric profiles. The 4-waves and 8-waves flame
released heat causes a sharp rise in the mixture temperature taking a locations are shifted by 0.24 mm and 0.81 mm, respectively. Thus, the
cone shape and exceeding the adiabatic flame temperature, consistent 8-waves flame is strongly affected by the asymmetric condition com-
with the numerical study of Norton and Vlachos [12]. It causes also a pared to the other models. The noticed shift is due to the enlargement of
fast acceleration of the mixture velocity reaching 12 m/s in the MC the pre-heating zone caused by the low temperature of the lower wavy
center. Note that the micro-flame generated is not flat and it is slightly wall. In other words, the temperature gradient within the pre-heating
curved due to the walls effect. The flame thickness at the MC center is zone is not symmetric and lower than the previous symmetric gradient.
approximately 0.2 mm and near the walls is thicker, approximately In this case, the mixture needs more distance to be pre-heated before
0.25 mm. In the downstream zone 3, after the methane has been con- the ignition. Consequently, the combustion takes place far of the old
sumed, the combustion stops and the mixture cools down to approxi- locations (at symmetric condition) and the micro-flame stabilizes in
mately the wall temperature (1425 K). There is no significant radial or new locations.
axial gradient within this zone. Temperature and heat release contours are of interest to examine
the flame shape, since the temperature and CO profiles give information
only about the flame location. The effect of asymmetric temperature
4. Results and discussions profiles on the flame behavior (shape and location) illustrated by
temperature contours (Fig. 8) and heat of reaction contours (Fig. 9) for
4.1. Effect of the wavy wall model all wavy MC. In Fig. 8, the temperature gradient (yellow to red colors)
within the flame zone takes a cone shape and its location varies from
This sub-section compares three different wavy MC, namely, 2- one model to another. This behavior is due to the difference in the pre-
waves, 4-waves and 8-waves and shows their effects on the flame lo- heating zone from one model to another, which is strongly affected by
cation. The comparison is made while keeping the entire boundary and the asymmetric condition and the wall geometry. Regarding Fig. 9, the
operating conditions used in the code validation, such as mass flux, flame shape is clearly shown. The flame in all models is curved similar
equivalence ratio and temperature profile. The effect of wavy wall to the previous flame of the flat MC. It can be noticed also that the flame
geometry on the temperature and species mole fractions at the cen- shape is affected by the asymmetric distribution of temperature pro-
terline of the MC is shown in Fig. 6. It can be seen that all the plotted files. The flames are inclined consistent with the experimental results of
profiles of each quantity exhibit the same tends and take the same Di Stazio et al [8]. They reported that the detection of the position of
values, except they are slightly shifted from each other. This indicates this kind of flames could be difficult experimentally due to the un-
that the number of waves within the walls affects only the flame lo- certainty cause by the distortion. Moreover, since the present study is
cation and not the flame properties (i.e. temperature and species mole numerical, the detection of the position of this kind of flames is easy
fractions). As the number of waves increases, the flame moves following due to the availability to all physical quantities.
the streamwise direction (towards the MC outlet). If we consider that
the flame location corresponds to the temperature peak, then the flame 4.3. Effect of temperature profile values
is positioned at x = 0.0025 m, x = 0.00274 m and x = 0.00288 m for
the configurations of 2-waves, 4-waves and 8-waves, respectively. The temperature profile values are a key parameter affecting the
These locations show that the flame tends always to stabilize between flame location and behavior as reported by Wang et al. [16]. Two high
two waves (Fig. 3) close to the MC center for each configuration. temperatures (1500 K and 1600 K) are tested along with the old one of
Moreover, the shift of the flame location is about 8% from 2-waves to 4- 1400 K. Figs. 10–12 show the overview of the flame location and shape
waves and about 4.8% from 4-waves to 8-waves. This shift is directly in the 8-waves MC with the different prescribed wall temperature
related to the slopes of the waves. Thus, the increase of the number of profiles. From Fig. 10 it can be noticed that the increase of the wall
the waves leads to the increase of the slope (in absolute value) of the temperature leads to drag the flame towards the MC inlet. The flame
waves, which pushes the flame towards the MC outlet. Based on these locates at x = 0.00249 m, x = 0.00294 m and x = 0.00369 m for the
results, the proposed models give satisfactory flame locations, since profiles of 1600 K, 1500 K and 1400 K, respectively. It can be said that
they are near the middle of the MC and far of the exit. Note that the the higher temperature of the wall the earlier flame ignition. In addi-
flame location of 4-waves model is very close to the flame location of tion, the maximum temperature and the CO production within the
the flat model (x = 0.0027 m) employed in the code validation. Hence, flame are decreased. To explain this behavior, temperature distribution
the geometry is an important factor for MC design. and heat release contours are visualized in Figs. 11 and 12, respectively.
From Fig. 11 it can be seen that the pre-heating zone is shortened and

5
Z. Mansouri Energy Conversion and Management 198 (2019) 111155

Fig. 6. Centerline profiles of temperature and species mole fractions for different wavy channels configurations.

Fig. 8. Temperature distribution for different wavy MC models using asym-

metric temperature profiles.

temperature profile to another. Also, the heat release reduces within the
flame core and takes the values of 14.61 w, 13 w and 10.67 w for the
profiles of 1400 K, 1500 K and 1600 K, respectively. This explains the
reduction of the flame temperature with high wall temperature profile.
Fig. 7. Centerline profiles of (a) temperature and (b) CO mole fractions for Furthermore, the high wall temperature provokes combustion near the
different wavy MCs using asymmetric temperature profiles. wall and not only near the center of the MC, which is clearly shown for
the 1600 K case. This leads to reduce the heat release within the flame
more warmed for the high temperature profiles. Along with the early core since there is a small heat release that forms near the wall. The
ignition and when the flame confronts the high pre-heating zone, it will profiles of heat of reaction at the upper wall in Fig. 13 justify well the
stabilize at an earlier position near the center of the MC. Moreover, the reduction of the flame temperature. The heat of reaction of the 1600 K
increase of the wall temperature affects the flame shape, as shown in profile is almost tow times higher than the heat of reaction of the other
Fig. 12. The flame becomes more and more inclined from one high profiles.

6
Z. Mansouri Energy Conversion and Management 198 (2019) 111155

Fig. 9. Heat of reaction (flame) contours for different wavy MC models using Fig. 12. Heat of reaction (flame) contours within the 8-waves MC using the
asymmetric temperature profiles. different prescribed wall temperature profiles.

Fig. 13. Heat release profiles at the upper wall of the 8-waves MC using the
different prescribed wall temperature profiles.

including three reduced mechanisms and a full detailed one (previously

mentioned in Section 2.5). Note that the 8-waves model and the 1500 K
temperature profile are used to perform this investigation. The effect of
combustion mechanisms on the temperature and species mole fractions
at the centerline of the MC is shown in Fig. 14. The tested kinetic
mechanisms alter aggressively the flame temperature and location and
the species consumption/production. All profiles exhibit quite different
tends and take different values. To perform an easy comparison, Table 1
Fig. 10. Centerline profiles of (a) temperature and (b) CO mole fractions within
regroups some amounts of the quantities of Fig. 14 within the flame
the 8-waves MC for the different prescribed wall temperature profiles. zone and at the wavy MC outlet. It is clearly seen that the global 2-step
mechanism shows large deviations of all quantities from the more de-
tailed schemes. Its maximum temperature differs from the reduced
mechanisms by around 500 K and its CO mole fraction at the MC exit is
almost neglected. However, the results of the other detailed schemes
are competitive. For example, the flame temperature difference reaches
around 57 K between the GRI 3.0 and Smooke mechanism and it takes
higher values for the other mechanisms. The CO mole fraction within
the flame zone turns between 0.06 and 0.065 and at the MC outlet it is
around 0.01. For the O2 mole fraction, the only mechanism capable of
consuming the entire amount is the GRI 3.0. Thus, the reduced schemes
show small amounts of O2 at the MC outlet. These differences in tem-
perature and species values could be attributed to the absence of some
intermediate reactions within the reduced schemes, unlike the full de-
tailed mechanism. Apart from the previous discussed differences, the
Smooke reaction mechanism has the closest tends and values to the full
Fig. 11. Temperature distribution within the 8-waves MC using the different GRI 3.0 mechanism among the Skeletal and DRM-19 mechanisms.
prescribed wall temperature profiles. These later two mechanisms show close distributions to each other but
a considerable deviation from the GRI 3.0 mechanism.
Temperature (Fig. 15) and heat of reaction (Fig. 16) distributions
4.4. Effect of combustion mechanisms are of interest to examine the flame behavior for each reaction scheme.
From Fig. 15, it can be seen that the pre-heating zone differs from one
Only few studies compared various reaction mechanisms of CH4–air mechanism to another. The global 2-step mechanism shows larger pre-
to characterize the micro-combustion in MC, such as Kang et al. [18]. heating zone and larger temperature gradient (yellow to red colors)
This sub-section deals with four additional reaction mechanisms within the flame zone compared to the other schemes. The more

7
Z. Mansouri Energy Conversion and Management 198 (2019) 111155

Fig. 14. Centerline profiles of temperature and species mole fractions within the 8-waves MC using different kinetic mechanisms.

Table 1
Temperature and species values for the different combustion mechanisms.
2-Step Skeletal Smooke DRM-19 GRI 3.0

Flame location (m) 0.00294 0.00239 0.00201 0.00229 0.00196

Flame T (K) 2571 1970.09 1994.23 2114.53 2051.54
Outlet T (K) 1387.8 1362.08 1410.97 1392.31 1348.81
Flame CO (–) 0.02685 0.06496 0.06506 0.06007 0.06357
Outlet CO (–) 0.00062 0.01164 0.01002 0.01091 0.01009
Outlet O2 (–) 0.00958 0.00604 0.00963 0.00172 0.00004
Outlet CO2 (–) 0.09414 0.07689 0.08632 0.09141 0.08357

detailed schemes exhibit smaller pre-heating zones and smaller tem-

perature gradients within the flame zone. This is governed by the po-
sition of stability of the flame, as shown in Fig. 16. The GRI 3.0 and
Smooke flames are stabilized at the top of the third wave at around
0.002 m. Whereas the Skeletal and DRM-19 flames are stabilized be-
tween the third and the fourth waves at around 0.0023 m. This dis-
crepancy could be attributed to the fact that the Skeletal and DRM-19
predicts a lower burning velocity compared to the GRI 3.0 and thereby
Fig. 15. Temperature distribution within the 8-waves MC using different ki-
leads to a more downstream flame stabilization location. Moreover, the
netic mechanisms.
flames of the detailed mechanisms are curved but in the opposite di-
rection compared to the global 2-step flame. Also, they occupy the
whole cross section within the MC, unlike the global 2-step flame which most adaptable to simulate steady micro-flames since its prediction is
is not in contact with the walls. This of course is due to the simple very close to that of the GRI 3.0 and it needs less computational time.
chemistry of the global mechanism compared to the other detailed
schemes. Furthermore, it should note that the computational time of the 5. Conclusions
detailed schemes is several times higher than that of the global me-
chanism, especially the GRI 3.0 scheme. The reduced mechanisms In this work, an innovative MC configuration with wavy walls for
(Skeletal, Smooke, and DRM-19) take less than the half of GRI 3.0 thermo-photovoltaic applications is proposed. The innovative wavy
computational time. It can be concluded that the Smooke mechanism is walls could contain 2-waves, 4-waves or 8-waves. The wavy walls

8
Z. Mansouri Energy Conversion and Management 198 (2019) 111155

[2] Durisch W, Bitnar B, Mayor J-C, von Roth F, Sigg H, Tschudi HR, et al. Small self-
powered grid-connected thermophotovoltaic prototype system. Appl Energy
2003;74(1–2):149–57. https://doi.org/10.1016/s0306-2619(02)00141-1.
[3] Akhtar S, Kurnia JC, Shamim T. A three-dimensional computational model of
H2–air premixed combustion in non-circular micro-channels for a thermo-photo-
voltaic (TPV) application. Appl Energy 2015;152:47–57. https://doi.org/10.1016/j.
apenergy.2015.04.068.
[4] Yan Y, Yan H, Zhang L, Li L, Zhu J, Zhang Z. Numerical investigation on combustion
characteristics of methane/air in a micro-combustor with a regular triangular pyr-
amid bluff body. Int J Hydrogen Energy 2018;43(15):7581–90. https://doi.org/10.
1016/j.ijhydene.2018.02.168.
[5] Li Y-H, Hong J-R. Performance assessment of catalytic combustion-driven thermo-
photovoltaic platinum tubular reactor. Appl Energy 2018;211:843–53. https://doi.
org/10.1016/j.apenergy.2017.11.095.
[6] Pan JF, Wu D, Liu YX, Zhang HF, Tang AK, Xue H. Hydrogen/oxygen premixed
combustion characteristics in micro porous media combustor. Appl Energy
2015;160:802–7. https://doi.org/10.1016/j.apenergy.2014.12.049.
[7] Maruta K, Kataoka T, Kim NI, Minaev S, Fursenko R. Characteristics of combustion
in a narrow channel with a temperature gradient. Proc Combust Inst
2005;30(2):2429–36. https://doi.org/10.1016/j.proci.2004.08.245.
[8] Di Stazio A, Chauveau C, Dayma G, Dagaut P. Combustion in micro-channels with a
controlled temperature gradient. Exp Therm Fluid Sci 2016;73:79–86. https://doi.
org/10.1016/j.expthermflusci.2015.09.020.
[9] Kamada T, Nakamura H, Tezuka T, Hasegawa S, Maruta K. Study on combustion
and ignition characteristics of natural gas components in a micro flow reactor with a
controlled temperature profile. Combust Flame 2014;161(1):37–48. https://doi.
org/10.1016/j.combustflame.2013.08.013.
Fig. 16. Heat of reaction (flame) contours within the 8-waves MC using dif- [10] Nakamura H, Tanimoto R, Tezuka T, Hasegawa S, Maruta K. Soot formation
ferent kinetic mechanisms. characteristics and PAH formation process in a micro flow reactor with a controlled
temperature profile. Combust Flame 2014;161(2):582–91. https://doi.org/10.
1016/j.combustflame.2013.09.004.
ensure an enhanced heat transfer between the reactor and the absorber [11] Kikui S, Nakamura H, Tezuka T, Hasegawa S, Maruta K. Study on combustion and
of the MTPV system. Within the proposed MC configuration the micro- ignition characteristics of ethylene, propylene, 1-butene and 1-pentene in a micro
flow reactor with a controlled temperature profile. Combust Flame
combustion characteristics of 2D laminar premixed CH4–air flames 2016;163:209–19. https://doi.org/10.1016/j.combustflame.2015.09.029.
(Φ = 1) have been studied numerically. The main contributions of this [12] Li J, Chou SK, Yang WM, Li ZW. A numerical study on premixed micro-combustion
investigation could be concluded as follows: of CH4–air mixture: effects of combustor size, geometry and boundary conditions on
flame temperature. Chem Eng J 2009;150(1):213–22. https://doi.org/10.1016/j.
cej.2009.02.015.
(1) The micro-flame was found slightly curved due to the walls effect. [13] Norton DG, Vlachos DG. Combustion characteristics and flame stability at the mi-
The flame location is sensible to the wavy wall geometry. As the croscale: a CFD study of premixed methane/air mixtures. Chem Eng Sci
2003;58(21):4871–82. https://doi.org/10.1016/j.ces.2002.12.005.
number of waves increases, the flame moves towards the MC outlet. [14] Tang A, Deng J, Cai T, Xu Y, Pan J. Combustion characteristics of premixed pro-
(2) The results of the non-symmetric temperature profiles revealed that pane/hydrogen/air in the micro-planar combustor with different channel-heights.
the flame takes new locations compared with symmetric tempera- Appl Energy 2017;203:635–42. https://doi.org/10.1016/j.apenergy.2017.05.187.
[15] Kang X, Gollan RJ, Jacobs PA, Veeraragavan A. Suppression of instabilities in a
ture profiles cases. This difference is due to the enlargement of the
premixed methane–air flame in a narrow channel via hydrogen/carbon monoxide
pre-heating zone of the non-symmetric cases. addition. Combust Flame 2016;173:266–75. https://doi.org/10.1016/j.
(3) The increase of the wall temperature moves the flames towards the combustflame.2016.07.003.
MC inlet and reduces the flame temperature. The high wall tem- [16] Wang W, Liu J, Zuo Z, Yang W. Entropy generation analysis of unsteady premixed
methane/air flames in a narrow channel. Appl Therm Eng 2017;126:929–38.
perature provokes combustion near the wall where small heat re- https://doi.org/10.1016/j.applthermaleng.2017.07.176.
lease was found at this location. This reduces the heat release [17] Kang X, Gollan RJ, Jacobs PA, Veeraragavan A. Numerical study of the effect of wall
within the flame core and explains the reduction of the flame temperature profiles on the premixed methane–air flame dynamics in a narrow
channel. RSC Adv 2017;7(63):39940–54. https://doi.org/10.1039/c7ra07265a.
temperature. [18] Kang X, Gollan RJ, Jacobs PA, Veeraragavan A. On the influence of modelling
(4) The 8-waves model and the 1500 K temperature profile were used choices on combustion in narrow channels. Comput Fluids 2017;144:117–36.
to investigate the flame temperature and location and the species https://doi.org/10.1016/j.compfluid.2016.11.017.
[19] Kazakove A, Frenklach M. Reduced reaction sets based on GRI–Mech 1.2, a
mole fractions for each the CH4 reaction mechanism. The 2-step 19–species reaction set, < http://www.me.berkeley.edu/drm/ > , accessed
mechanism showed large deviations of all quantities. The results of September 2018.
the other detailed schemes were competitive. The GRI 3.0 and [20] Nair A, Kishore VR, Kumar S. Dynamics of premixed hydrogen-air flames in mi-
crochannels with a wall temperature gradient. Combust Sci Technol
Smooke flames are stabilized at the top of the third wave. Whereas 2015;187(10):1620–37. https://doi.org/10.1080/00102202.2015.1059326.
the Skeletal and DRM-19 flames are stabilized between the third [21] Kishore VR, Minaev S, Akram M, Kumar S. Dynamics of premixed methane/air
and the fourth waves. mixtures in a heated microchannel with different wall temperature gradients. RSC
Adv 2017;7(4):2066–73. https://doi.org/10.1039/c6ra27582f.
(5) The Smooke mechanism was found the most adaptable scheme to
[22] Poinsot T, Veynane D. Theoretical and numerical combustion. 2nd ed. Edwards;
simulate steady micro-flames since its prediction is very close to 2005.
that of the full chemistry scheme (GRI 3.0). [23] Alipoor A, Mazaheri K. Studying the repetitive extinction-ignition dynamics for lean
premixed hydrogen-air combustion in a heated microchannel. Energy
2014;73:367–79. https://doi.org/10.1016/j.energy.2014.06.027.
Declaration of interests [24] ANSYS®. Academic Research. Release 19.0. Help system, FLUENT user guide.
ANSYS Inc.; 2018.
The authors declare that they have no known competing financial [25] Mansouri Z, Boushaki T, Aouissi M. Detached eddy simulation of non-reacting
swirling flow in a vortex burner. Int J Heat Technol 2017;35(3):594–602. https://
interests or personal relationships that could have appeared to influ- doi.org/10.18280/ijht.350316.
ence the work reported in this paper. [26] Mansouri Z, Boushaki T, Aouissi M, Gökalp I. Computational investigation of a
swirled premixed burner using hybrid RANS-LES method. Progress in hybrid RANS-
LES modelling Springer International Publishing; 2018. p. 243–54. https://doi.org/
References 10.1007/978-3-319-70031-1_20.
[27] Mansouri Z, Aouissi M, Boushaki T. Detached Eddy simulation of high turbulent
[1] Zuo WEJ, Han D, Jin Y. Numerical investigations on thermal performance of swirling reacting flow in a premixed model burner. Combust Sci Technol
double-layer four-channel micro combustors for micro-thermophotovoltaic system. 2016;188(11–12):1777–98. https://doi.org/10.1080/00102202.2016.1211888.
Energy Convers Manage 2017;150:343–55. https://doi.org/10.1016/j.enconman. [28] Mansouri Z, Boushaki T. Experimental and numerical investigation of turbulent
2017.08.029. isothermal and reacting flows in a non-premixed swirl burner. Int J Heat Fluid Flow

9
Z. Mansouri Energy Conversion and Management 198 (2019) 111155

2018;72:200–13. https://doi.org/10.1016/j.ijheatfluidflow.2018.06.007. [34] Smooke MD, Puri IK, Seshadri K. A comparison between numerical calculations and
[29] Tomboulides AG, Lee JCY, Orszag SA. J Sci Comput 1997;12(2):139–67. https:// experimental measurements of the structure of a counterflow diffusion flame
doi.org/10.1023/a:1025669715376. burning diluted methane in diluted air. Symp (Int) Combust 1988;21(1):1783–92.
[30] Liou M-S. A sequel to AUSM: AUSM+. J Comput Phys 1996;129(2):364–82. https://doi.org/10.1016/s0082-0784(88)80412-0.
https://doi.org/10.1006/jcph.1996.0256. [35] Yang B, Pope SB. Treating chemistry in combustion with detailed mechanisms—In
[31] Liou M-S. A sequel to AUSM, Part II: AUSM+-up for all speeds. J Comput Phys situ adaptive tabulation in principal directions—Premixed combustion. Combust
2006;214(1):137–70. https://doi.org/10.1016/j.jcp.2005.09.020. Flame 1998;112(1–2):85–112. https://doi.org/10.1016/s0010-2180(97)81759-2.
[32] Westbrook CK, Dryer FL. Simplified reaction mechanisms for the oxidation of hy- [36] Haynes WM. Handbook of chemistry and physics. 96th ed. CRC; 2015.
drocarbon fuels in flames. Combust Sci Technol 1981;27(1–2):31–43. https://doi. [37] Robinson C, Smith DB. The auto-ignition temperature of methane. J Hazard Mater
org/10.1080/00102208108946970. 1984;8(3):199–203. https://doi.org/10.1016/0304-3894(84)85001-3.
[33] Smith Gregory P, Golden David M, Frenklach Michael, Moriarty Nigel W, Eiteneer [38] Mansouri Z, Aouissi M, Boushaki T. Numerical computations of premixed propane
Boris, Goldenberg Mikhail, Thomas Bowman C, Hanson Ronald K, Song Soonho, flame in a swirl-stabilized burner: effects of hydrogen enrichment, swirl number
Gardiner William C Jr., Lissianski Vitali V, ZQ. GRI-Mech 3.0, < http://www.me. and equivalence ratio on flame characteristics. Int J Hydrogen Energy
berkeley.edu/gri_mech/ > , accessed September 2018. 2016;41(22):9664–78. https://doi.org/10.1016/j.ijhydene.2016.04.023.

10