Sei sulla pagina 1di 9

Available online at www.sciencedirect.

com

Proceedings of the Combustion Institute 37 (2019) 3703–3711


www.elsevier.com/locate/proci

An experimental and computational study of


hydrogen–air combustion in the LAPCAT II supersonic
combustor
A. Vincent-Randonnier a, V. Sabelnikov a,b, A. Ristori a, N. Zettervall c,
C. Fureby c,∗
a ONERA – The French Aerospace Lab., F-91761 Palaiseau, France
b Central Aerohydrodynamic Institute (TsAGI), 140180 Zhukovsky, Moscow Region, Russia
c Defense Security Systems Technology, The Swedish Defense Research Agency – FOI, SE 147 25 Tumba, Stockholm,

Sweden
Received 30 November 2017; accepted 31 May 2018
Available online 5 September 2018

Abstract

Dual-mode ramjet/scramjet engines are considered a promising propulsion system for the next genera-
tion commercial high-speed transport flight vehicles. In this study we combine experimental measurements
of high-speed (subsonic and supersonic) combustion at different operating conditions in the LAPCAT-II
dual-mode ramjet/scramjet combustor with Large Eddy Simulations (LES) using finite rate chemistry mod-
els and skeletal H2 –air combustion chemistry. The combustor geometrically consists of four sections, and
experiments have been realized for wall injection of H2 in a Mach = 2 vitiated air-flow for total pressures
and temperatures of p0 = 0.40 MPa, 1414 K < T0 < 1707 K, and a fixed equivalence ratio of φ=0.15. For
this p0 the combustor is over-expanded, and the transition from supersonic to subsonic flow occurs at the
beginning of the fourth combustor section. The flow and combustion diagnostics include measurements of
p0 and T0 upstream of the combustor, wall-pressure profiles and Schlieren as well as OH∗ chemiluminescence
imaging. The computational set-up consists of the full combustor, from the nozzle into the dump-tank. The
computational model is composed of a compressible finite rate chemistry LES model, using the mixed sub-
grid flow model and the Partially Stirred Reactor (PaSR) combustion model, together with a novel skeletal 22
step H2 –air reaction mechanism. Qualitative as well as quantitative comparisons between experiments and
simulations show reasonably good agreement, but most importantly reveal a high sensitivity of both the LES
predictions and the experiments to T0 . The LES results are further used to describe the underlying mech-
anisms of flow, wall-injection, mixing, self-ignition and turbulent combustion, and how these interrelated
processes are modified by increasing the total temperature under otherwise identical conditions.
© 2018 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Supersonic combustion; Dual-mode ramjet/scramjet; Finite-rate chemistry LES; Skeletal chemistry

∗ Corresponding author.
E-mail address: fureby@foi.se (C. Fureby).

https://doi.org/10.1016/j.proci.2018.05.127
1540-7489 © 2018 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
3704 A. Vincent-Randonnier et al. / Proceedings of the Combustion Institute 37 (2019) 3703–3711

1. Introduction and background validation of the methods, and used together to


provide a more comprehensive understanding
The global incentive to consider the re- of the flow, mixing and combustion phenomena
introduction of commercial high-speed transport occurring under dual-mode conditions in this
to reduce the duration of antipodal flights to less combustor.
than four hours require novel civil hypersonic
aircraft. Among a plethora of technical, environ-
mental, human and economic factors to consider, 2. Experimental set-up
the propulsion plant and its integration in the flight
vehicle is a crucial component. Designing stable The ONERA-LAERTE facility equipped with
air-breathing high-speed combustion systems the LAPCAT-II dual-mode ramjet/scramjet com-
remains a major challenge, where fuel must be bustor, its operating conditions and the measure-
mixed, ignited, and burned within a few millisec- ment procedures have been described in detail in
onds. The large-scale and extreme thermodynamic [10]. A schematic of the facility and combustor
conditions in hypersonic aircraft make experiments sections are presented in Fig. 1(a)–(c). The refer-
costly and challenging, [1]. Data are often limited ence position x = 0 m corresponds to the throat
to combustion images and sparse measurements of of the Mach 2.0 nozzle. Experiments were per-
pressure, temperature and thrust with conventional formed for total temperatures 1414 K < T0 <
techniques. Validated computational models are 1707 K, at an equivalence ratio of φ = 0.15, and
the most promising option to bridge the gap be- a total pressure of p0 = 0.40 ± 0.01 MPa. The
tween ground-testing and full-scale flight-testing. inflow air-heating was performed by H2 –air com-
Moreover, these have the additional benefit of bustion and oxygen (O2 ) replenishment. The flow
providing a full-field view of the flow physics. in the combustor is overexpanded and the transi-
Injection of fuel, e.g., hydrogen, H2 and small tion from supersonic flow to subsonic flow occurs
hydrocarbons, in a supersonic hot air-stream and in the downstream part of the combustor (in the
its subsequent mixing, self-ignition and transition third combustor section) at x ≈ 0.75–0.80 m.
into turbulent combustion have been studied ex- H2 was injected at sonic velocity both from the
perimentally, e.g., [2–5], and computationally, e.g., upper and lower combustor walls at x = 0.20 m
[6–9], for different injector–combustor configura- through two flush-mounted 2.0 mm porthole in-
tions. This is a very challenging problem that must jectors. The operating conditions are presented in
be better understood before reliable supersonic Table 1 in terms of p0 and T0 , air- and fuel mass-
propulsion systems using this technique can be re- flows (mair and mH2 ), φ, inlet velocity, tempera-
alized. More specific issues include: (i) the short ture and density (vin , Tin and ρ in ), as well as fuel
combustor residence time combined with the slow temperature and density (ρ H2 and TH2 ). The flow
ignition kinetics at low and moderate tempera- and combustion diagnostics include measurements
tures, (ii) interactions between shock-waves and a of p0 and T0 upstream of the combustor, pres-
chemically reacting plume, (iii) reduced efficiency sure profiles along the combustor, Schlieren imag-
of fuel-air mixing due to compressibility effects, ing (12 kHz at 1 μs exposure time), and OH∗ visu-
(iv) combustion instabilities, particularly at high alization (4 kHz at 4 μs exposure time) in the three
flight-altitudes, and (v) the total temperature, T0 , windows in Fig. 1(c).
and equivalence ratio, φ, effects on self-ignition The combined OH∗ chemiluminescence and
and thermal choking. To minimize internal losses, Schlieren images Fig. 1(d), provide information
transverse fuel injection without cavity stabilization about the flow physics and flame stabilization
is preferred [9], requiring that flame stabilization is mechanisms, and how these attributes change
controlled either by an upstream recirculation re- with operating conditions. The shock-structure and
gion, the formation of coherent structures with fuel boundary layer separation are very sensitive to T0,
and air supporting a turbulent diffusion flame, or and appear to dominate the ignition and flame sta-
by shock-induced combustion. bilization mechanisms. For low T0 combustion oc-
Here we combine experimental investigations curs in the low-speed region in the third combustor
of supersonic combustion at different conditions in section, at intermediate T0 either unsteady combus-
the LAPCAT-II dual-mode ramjet/scramjet com- tion with the combustion fronts oscillating between
bustor [10], with finite rate chemistry Large Eddy x ≈ 0.30 m and x ≈ 0.45 m or thermal choking is
Simulation (LES) combined with skeletal H2 –air observed, and at high T0 shock-induced combus-
combustion chemistry. The operating conditions tion is observed.
are chosen to be representative of dual-mode
ramjet/scramjet combustion. The objectives are to
investigate transverse mixing, self-ignition, flame 3. LES models, numerical methods and
stabilization and compressible turbulent combus- computational set-up
tion, and the dependence of these phenomena on
T0 at fixed values of φ. Results from experiments The LES model is based on implicitly-filtered
and LES will be compared, providing reciprocal transport equations for mass, momentum, energy
A. Vincent-Randonnier et al. / Proceedings of the Combustion Institute 37 (2019) 3703–3711 3705

Fig. 1. (a) General plan of the LAERTE facility, (b) photo of the LAPCAT-II combustor, (c) geometry of the LAPCAT-II
combustor, and (d) combined OH∗ chemiluminescence and Schlieren images of representative combustion situations at
T0 = 1697 K, Table 1.

Table 1
Characteristic properties of the LAPCAT-II combustor simulations.
Case p0 T0 mair mH2 φ vin Tin ρ in ρ H2 TH2 Combustion
[Pa] [K] [kg/s] [g/s] [m/s] [K] [kg/m3 ] [kg/m3 ] [K]
1 397740 1414 0.2931 1.53 0.15 686 1199 0.614 0.176 294 Subsonic
3a 401550 1505 0.2931 1.47 0.15 707 1276 0.583 0.173 297 Unsteady supersonic
3b 400900 1506 0.2931 1.49 0.15 708 1277 0.582 0.173 299 Thermal choking
4a 410220 1697 0.2931 1.40 0.15 751 1439 0.525 0.165 306 Shock-induced
4b 402550 1707 0.2931 1.37 0.15 754 1448 0.515 0.163 304 Shock-induced

and species mass-fractions, [11], together with ther- a weighted average of the fine-structure and sur-
mal and caloric equations-of-state and constitu- rounding reaction rates using the reacting volume
tive equations. The thermal and caloric equations- fraction, γ ∗ . This model has been widely used in
of-state are obtained under the presumption of a combustion LES, and is validated for laboratory
mixture of thermally perfect gases using tabulated combustors [17–18], afterburners [19], gas turbine
formation enthalpies and specific heats. The consti- combustors [20], as well as scramjet combustors
tutive equations are those of a linear viscous mix- [21].
ture with Fickian diffusion and Fourier heat con- The LES-PaSR model equations are solved
duction. The viscosity is calculated using Suther- using a fully-explicit finite-volume code based
land’s law, whereas the thermal conductivity and on the OpenFOAM C++ library, [22]. High-
species diffusivities are computed from the viscos- order monotonicity-preserving reconstruction of
ity utilizing constant Prandtl and species Schmidt the convective fluxes and central differencing of
numbers, respectively [12]. The unfiltered reaction the diffusive fluxes [23], are combined with a total
rates in the filtered species mass-fraction equations variation diminishing based Runge–Kutta time in-
results from Guldberg–Waage law of mass-action, tegration scheme to result in a second-order accu-
involving summation over all reactions, with reac- rate algorithm. The chemical source terms in the
tion rates obtained from Arrhenius rate laws [13]. species transport equations are evaluated using an
The unresolved transport terms, or the subgrid operator-splitting approach together with a Rosen-
stress tensor and flux vectors, in the filtered trans- brock solver [24]. The code is density based, fully
port equations are closed by the mixed model [14]. compressible, and stability is imposed using com-
The filtered reaction rates are modeled using the pact stencils and by enforcing conservation of ki-
Partially Stirred Reactor (PaSR) model [15], which netic energy with a Courant number limitation of
is a multi-scale model based on the observation 0.5.
[16] that combustion often takes place in dispersed The computational set-up is shown in Fig. 1(c),
fine-structure regions surrounded by low reaction and starts at the de Laval nozzle and ends
rates. The filtered reaction rates are estimated as in a dump before the exhaust water-cooling.
3706 A. Vincent-Randonnier et al. / Proceedings of the Combustion Institute 37 (2019) 3703–3711

Fig. 2. Comparison of (a) laminar flame speeds, su , (left axis) and adiabatic flame temperatures, T, (right axis) at T = 300 K
and p = 1 atm, and ignition delay times, τ ign , at 1 atm (b) and (c) 4 atm, for H2 –air mixtures. Legend: ( ) M1, [34], ( )
Z22, [29], ( ) J20, [32], ( ) D7, [33], ( ) K30, [31], and experimental data (black symbols) from [30–31,35–36].

Hex-dominant grids with ∼26 and ∼77 million at 1 atm and 300 K, and τ ign at 1 atm and 4 atm,
cells, refined along the walls and in the vicinity of respectively. Concerning su , Z22 is in reasonable
the injectors and the downstream plume are used. agreement with the experimental data and the pre-
The Index of Quality [25], was used to assess the dictions from K30, J20 and M1. For Tflame , all
sensitivity of the solution to the grid, and reveals mechanisms produce similar predictions, with the
that 87% and 92% of the kinetic energy was re- largest deviations at around stoichio-metric condi-
solved, respectively, rendering both grids suitable tions. For τ i gn at 1 atm, Fig. 2b, all mechanisms per-
for LES. The lowest values of LESIQ occur in the form similarly above 1000 K but with large devia-
high-speed injector flow. Dirichlet boundary condi- tions below. The simplified chemistries of D7 and
tions are used for all variables at the inlet, and at the M1, neglecting the H2 O2 and HO2 chemistry, re-
sonic H2 -injectors. Moreover, at the outlet, a wave- sult in inabilities to reproduce the bending of the
transmissive boundary condition is used [26], given ignition curve within the crossover region, and with
a dump-pressure of 1.0 bar, and a no-slip subgrid the profiles for J20 and K30 predicting higher τ ign
wall-model [27], is employed to model the near-wall when the temperature decreases, than Z22. At 4 atm
flow. The grid is adapted to the wall model, and uses Fig. 2(c), the limitation in predictions of τ ign in the
y+ values of between 40 and 100. The measured crossover region and below for the different mecha-
wall temperature is used to provide an estimate of nisms compared to Z22 is even more obvious, with
the LES wall temperature of 800 K. either over-or underpredictions of τ ign as compared
with the experimental data. In addition, Z22 is also
the only investigated mechanism able to satisfac-
4. Hydrogen-air combustion chemistry torily reproduce the experimental data at the low-
est temperatures. Significant differences in τ ign be-
H2 –air combustion is here modeled using a low ∼1000 K between the mechanisms will impact
novel H2 –air reaction mechanism that uses the flame anchoring and stabilization within the com-
H2 –O2 structure from the pathway centric mech- bustor.
anism in [28], together with three additional fuel The reason for the discrepancies between the
breakdown reactions. This result in the Z22 mech- detailed mechanism predictions and the experi-
anism, using 9 species and 22 irreversible reactions mental data below ∼1000 K is under debate, with
[29], see Supplementary material A. The mecha- the issue of non-homogeneous conditions being
nism development was performed using the lam- one criticism of the current way of modeling τ ign at
inar flame speed (su ), flame temperature (Tflame ) low temperatures. Investigators [36], have observed
and ignition delay time (τ ign ) as targets. Addi- two distinct ignition phenomena depending on
tional effort during the mechanism development the temperature conditions. Z22 is designed to
was to create a mechanism that could match τ ign at take the possibly non-homogeneous conditions
and below the crossover temperature interval, 800 into account when using the standard model for
K < T < 1200 K, where the ignition chemical kinet- estimating τ ign but a secondary approach could be
ics is complex and controversial, as are the ignition to include a nonhomogeneous condition into τ ign
phenomena [30]. The detailed mechanisms by Kon- when simulating current detailed mechanisms.
nov et al. [31], K30, and Jachimowski [32], J20, the
skeletal mechanism by Davidenko et al. [33], D7, as
well as the global mechanism by Marinov [34], M1, 5. Results and discussion
are employed for comparisons together with exper-
imental data sets [30–31, 35–36]. A first appraisal of combustion in the
Figure 2(a)–(c) compare predictions and exper- LAPCAT-II combustor is provided in Fig. 3
imental data for su (left axis) and Tflame (right axis) which presents experimental combustion images
A. Vincent-Randonnier et al. / Proceedings of the Combustion Institute 37 (2019) 3703–3711 3707

Fig. 3. Experimental flame images and volumetric renderings of thermal radiation heat-flux from LES for (a) Case 1
(T0 = 1414 K), (b) Case 3a (T0 = 1505 K) and (c) Case 4a (T0 = 1697 K).

[10], and volumetric renderings of the thermal ra- from being attached to the upper and lower walls,
diation heat-flux from the LES of Cases 1, 3a and respectively, with a frequency of ∼50 Hz.
4a. The thermal radiation heat-flux is defined as Figure 4 presents perspective views of the flow-
q(x ) = τ εσ T 4 Ff →x in which τ is the atmospheric field in the LAPCAT-II combustor for (a) Case 1,
transmissivity, σ the Stefan–Boltzman constant, ε (b) Case 3a and (c) Case 4a in terms of iso-surfaces
the emissivity, T the flame temperature and Ff –x the of the second invariant of the velocity-gradient
view factor. As evident from Table 1, p0 , mair , φ are tensor, λ2 [37], colored by the H2 mass-fraction,
the same for all cases, resulting in gradually increas- YH2 , and the pressure, p, and volumetric render-
ing inflow axial velocities, vx , and temperatures, ings of the temperature, T. For all cases, the bound-
Tin , with increasing T0 . For Case 1 (at T0 = 1414 K) ary layer flow structures (longitudinal and hairpin
both the experimental combustion image and the vortices, and streaks) develop halfway down the
LES volumetric rendering show that combustion nozzle, further developing along the first combus-
starts in the subsonic region in the third combustor tor section until the fuel-injectors are reached. In
section, just after the 1° to 3° transition (in the front of each fuel-inject-or a bow-shock is formed
second window) at 0.75 < x < 0.80 m. For Case (at x ≈ 0.198 m), and in front of this a Lambda
3a (at T0 = 1505 K) the LES volumetric rendering shock is formed (at x ≈ 0.193 m). A horseshoe vor-
and the experimental combustion image reveal that tex develops in the recirculation region between the
combustion occurs much earlier, towards the end bow- and Lambda shocks as a result of the flow
of the first (straight) combustor section or in the stagnation and boundary-layer separation. This
beginning of the second combustor section (first vortex traverses downstream, parallel to the fuel-
window). Experiments suggests a combustion front rich plumes from the injectors.
oscillating between 0.30 m < x < 0.45 m, whereas For Case 1 most of the injected H2 remains
LES predicts that the onset of combustion is more within the plumes as they are convected down-
gradual, oscillating between 0.28 m < x < 0.40 m stream, and only a small fraction of H2 is engulfed
at ∼60 Hz, together with pockets of more intense in the horseshoe vortices. The H2 -rich plumes
combustion occurring intermittently. For Case consist of a counter-rotating vortex pair along
4a (at T0 = 1697 K) both LES and experiments the average trajectory of each H2 -plume, originat-
shows that combustion takes place even earlier: ing at the leading edges of the jet shear-layers,
The OH∗ and Schlieren images of Fig. 1(d) shows and jet shear-layer vortices enfolding the jets, and
that combustion starts abruptly at x ≈ 0.26 m developing due to Kelvin–Helmholtz instabilities
due to shock-induced ignition, whereas the LES in the jet-shear layers. Near the injectors, these
predictions suggest that the combustion inception flow structures typically appear as incoherent S-
appears more complicated, driven by shock- shaped side vortex arms and vortex rollers, but
induced ignition, with intermittent combustion downstream they organize into full -shaped vor-
occurring also in front of the injectors, beneath tex loops. Along the second and third combus-
the bow-shocks and partially in the horseshoe tor sections the vorticity field gradually weaken,
vortices surrounding the jets. All LES suggests whilst the H2 -plumes spread across the combustor
that the upper and lower flame branches alternate cross-section. The boundary layers separate about
3708 A. Vincent-Randonnier et al. / Proceedings of the Combustion Institute 37 (2019) 3703–3711

Fig. 4. Instantaneous distributions of the second invariant of the velocity gradient tensor, λ2 , colored by the H2 mass
fraction, YH2 , in green, iso-surfaces of the pressure, p, in magenta, and volumetric renderings of the temperature, T, in a
color scale from black through red, orange, yellow to white for (a) Case 1 (T0 = 1414 K), (b) Case 3a (T0 = 1505 K) and
(c) Case 4a (T0 = 1697 K). In (d) and (e), respectively, details of Case 3a and Case 4a are shown.

halfway down the third combustor section, at 0.75 In Case 4a the flow is again very similar to
m < x < 0.80 m, resulting in subsonic flow re- that of Cases 1 and 3a but with an even thinner
gions near the upper and lower walls, promoting boundary layer due to the higher inflow velocity.
ignition, flame stabilization and turbulent combus- The flow is dominated by the H2 -plumes, and their
tion. These regions occur at different locations on embedded counter-rotating vortex pairs, jet shear-
the two walls, resulting in an asymmetry manifested layer vortices, S-shaped side vortex arms, span-
as a thicker combustion zone along the upper or wise vortex rollers and vortex loops. Due to the
lower wall, intermittently switching to the other higher inflow temperature, ignition takes place eas-
wall. ier Fig. 2(b) and (c), resulting in that combustion
For Case 3a the approach flow to the fuel in- occurs earlier, and simultaneously across the full
jectors is very similar to that of Case 1 but with combustor width. Due to the volumetric expan-
a marginally thinner boundary layer due to the sion and temperature increase, the vortical struc-
higher inflow velocity. Similar flow structures as tures thicken in the combustion region, resulting in
observed for Case 1 also develop downstream of a different vorticity field in the first two combus-
each injector, i.e., a counter-rotating vortex pair, tor sections. Asymmetry in the combustion prop-
jet shear-layer vortices, S-shaped side vortex arms, agation is again observed similar to Case 3a. No
spanwise vortex rollers and -shaped vortex loops. axial combustion oscillations are observed for this
The higher T0 facilitate ignition Fig. 2(b) and (c), case. Combustion, however, occurs intermittently
resulting in combustion starting already towards already in the recirculation region upsteam of the
the end of the first combustor section. Combus- injectors.
tion occurs initially along the sides of the H2 - Figure 5 shows numerical Schlieren images
plumes, gradually widening towards the center to (based on the derivative of the refractive index)
cover the full combustor width downstream of the combined with OH mass-fraction, YOH , images to
first combustor section. Further downstream, com- match the experimental images in Figs. 1(d) for
bustion widens from the boundary layer to gradu- Cases 3a and 4a. From these figures, the upstream
ally cover almost the full combustor height in the shock-train and the bow- and Lambda shocks are
third and fourth combustor sections. Again, asym- visible, as well as the barrel-shocks, H2 -plumes and
metry in the combustion propagation is observed, reflected shocks downstream of injection. A mix-
resulting in a thicker combustion zone along one ing region occurs downstream of the fuel injectors,
of the walls, but periodically switching to the other the length of which appears strongly dependent on
wall. Combustion seems to oscillate also in the ax- the total temperature, with Case 1 (not shown) pre-
ial direction, with the flame front moving between senting by far the longest mixing region and Case
x ≈ 0.28 m and 0.40 m with a frequency of ∼60 Hz. 4a the shortest. The length of the mixing region
Pockets of more intense combustion are also ob- is also the particular feature that departs the most
served intermittently. Due to the volumetric expan- between the experiments and the LES. There can
sion and temperature increase, the vortical struc- be many reasons for this, including the accuracy of
tures thicken in the high-temperature combustion the reaction mechanism Fig. 2, wall-surface rough-
region modifying the mixing rates. ness, the accuracy of the LES turbulence models,
A. Vincent-Randonnier et al. / Proceedings of the Combustion Institute 37 (2019) 3703–3711 3709

Fig. 5. Combined numerical Schlieren and OH image for (a) Case 3a and (b) Case 3a.

Fig. 6. (a) Time-averaged wall-pressures, pw , along the lower combustor wall and (b) cross-sectional averaged Mach
number, Ma, and normalized heat-release, Q/Qmax , and streamthrust, s/snozzle . Legend: (×), (◦) and ( ) experimental data
for Cases 1, 3a and 4a, respectively, and LES predictions for ( ), ( ) and ( ) Cases 1, 3a and 4a, respectively, on the
26 million cell grid. In (b) solid lines denote Ma, dotted lines denote Q/Qmax and dashed lines denote s/snozzle .

wall-models and combustion models, and the captured by all LES. Comparison between mea-
accuracy of the characterization of the operating sured and predicted T0 and p0 at x ≈ 0.10 m shows
conditions. excellent agreement for all cases indicating that the
Based on Figs. 4 and 5 ignition occurs in the approach flow is well captured.
low-speed region in the third combustor section for For the cross-sectional averaged profiles in Fig.
Case 1, whereas for Case 4a combustion occurs as 6(b) we find that Q/Qmax occur in different locations
a consequence of shock-induced ignition of both for the different cases: For Case 1 Q/Qmax occurs in
the H2 –air mixture in the main fuels plumes and the third (subsonic) combustor section, whilst for
in the horseshoe-vortex systems. For Case 3a com- Cases 3a and 4a, Q/Qmax occur in the second (su-
bustion follows from a combination of turbulent personic) combustor section, but with different ax-
and coherent structure mixing, and gradual heating ial distributions. Ma remains almost constant un-
from multiple reflected shocks resulting in unsteady til Q/Qmax increases due to combustion after which
combustion. Ma gradually increase for Cases 3a and 4a pend-
Figure 6(a) and (b) shows profiles of the wall- ing the fourth combustor section is reached, after
pressure, pw , from the experiments and the LES, which Ma decreases towards Ma ≈ 0.8. Moreover,
and cross-sectional averaged profiles of the Mach s/snozzle is nearly constant through the combustor
number, Ma, normalized heat-release, Q/Qmax , and for Case 1 but increases along the fourth combus-
streamthrust, s/snozzle , in which s = p + 12 ρv2x . Re- tor section. For Cases 3a and 4a a small increase in
garding pw in Fig. 6(a) we notice that the LES’ cap- s/snozzle is observed along the second and third com-
tures the trends of the experiments well, but lack bustor sections, after which s/snozzle increase along
somewhat in detail. For Case 1 pw increases slightly the fourth combustor section.
too early, indicating too early combustion, possi-
bly due to premature flow separation in the third
combustor section. For Cases 3a and 4a both LES’ 5. Summary and concluding remarks
and experiments reveal a considerable increase in
pw due to combustion. For Case 4a both the exper- This paper describes a combined experimen-
iments and LES’ expose a peak in pw at x ≈ 0.26. tal and computational study of a supersonic flow
While the experimental pw profiles are nearly con- and hydrogen–air combustion in the LAPCAT-II
stant along the second combustor section, the LES’ dual mode ramjet/scramjet combustor. The com-
pw profiles decrease linearly along the second com- bustor consists of four sections with gradually in-
bustor section. This difference may be explained by creasing cross-sectional area, and experiments and
a small step in the combustor at x ≈ 0.60 m not LES have been performed for total temperatures
included in the computational configuration. The of 1414 K < T0 < 1707 K at a total pressure
pressure recovery for x > 0.80 m is reasonably well of p0 = 0.40 MPa and an equivalence ratio of
3710 A. Vincent-Randonnier et al. / Proceedings of the Combustion Institute 37 (2019) 3703–3711

φ = 0.15. The diagnostics include reference mea- [5] D.J. Micka, J.F. Driscoll, Proc. Combust. Inst. 32
surements of p0 and T0 upstream of the com- (2009) 2397–2404.
bustor, wall-pressures at the combustor walls, and [6] H. Koo, P. Donde, V. Raman, Proc. Combust. Inst.
Schlieren and OH∗ chemiluminescence imaging. 33 (2011) 2203–2210.
[7] J.A. Fulton, J.R. Edwards, H.A. Hassan, et al., J.
The LES set-up consists of the whole combustor
Prop. Power 30 (2014) 558–575.
and the LES are performed using a compressible [8] A. Saghafian, L. Shunn, D. Philips, F. Ham, Proc.
finite-rate chemistry LES model. Combust. Inst. 35 (2015) 2163–2172.
For all cases, excellent agreement between pre- [9] K. Nordin-Bates, C. Fureby, S. Karl, K. Hanne-
dicted and measured reference values of p0 and T0 mann, Proc. Combust. Inst. 36 (2017) 2893–2900.
is found, ascertaining mutual agreement between [10] A. Vincent-Radonnier, Y. Moule, M. Ferrier, AIAA
the experiments and LES. For this p0 the com- 2014 (2014) 2931.
bustor is overexpanded, and both the LES and [11] T. Echekki, E. Mastorakos, Turbulent Combustion
experiments show that transition from supersonic Modeling: Advances, New Trends and Perspectives,
Springer, Netherlands, 2011.
to subsonic occur at the beginning of the fourth
[12] E. Giacomazzi, F.P. Picchia, N. Arciacono, in: Pro-
(and last) combustor section. By sweeping T0 dif- ceedings of the Thirtieth Italian Meeting on Com-
ferent flow and combustion states are observed: For bustion, Ischia, Italy, 2007.
T0 = 1414 K combustion occurs only in the sep- [13] S. Menon, C. Fureby, Encyclopedia of Aerospace En-
arated region far downstream. For T0 = 1505 K gineering, in: R. Blockley, W. Shyy (Eds.), John Wiley
combustion starts between the first and second & Sons, 2010.
combustor sections, with experiments showing an [14] R. Bensow, C. Fureby, J. Turbul. 8 (54) (2007) 1–17.
abrupt combustion front and LES a less sharp but [15] V Sabelnikov, C. Fureby, Combut. Flame 160 (2013)
longitudinally oscillating combustion front. For 83–96.
[16] M. Tanahashi, M. Fujimura, T. Miyauchi, Proc.
T0 = 1697 K combustion starts even earlier, with
Combust. Inst. 28 (2000) 529–535.
the LES predictions showing a somewhat to early, [17] K-J. Nogenmyr, C. Fureby, X.-S. Bai, P. Petersson,
intermittent, combustion inception compared to M. Linné, Combut. Flame 156 (2009) 25–36.
the experimental data. Matching wall-pressure in- [18] E. Fedina, C. Fureby, J. Turbul. 12 (24) (2011) 1–20.
creases due to combustion are observed for the LES [19] N. Zettervall, K. Nordin-Bates, E. Heimdal-Nilsson,
and the experiments. The LES and experiments C. Fureby, Combut. Flame 179 (2015) 1–22.
both reveal a strong sensitivity of the combustor to [20] E. Fedina, C. Fureby, G. Bulat, W. Maier, Flow Tur-
variations in T0 reminiscent to turbulent supersonic bul. Combut. 99 (2017) 385–409.
hydrogen-air combustion. The LES results and the [21] M. Chapuis, E. Fedina, C. Fureby, K. Hannemann,
S. Karl, J. Martinez Schramm, Proc. Combust. Inst.
OH∗ and Schlieren images are further utilized to
34 (2012) 2101–2109.
describe the flow, mixing, self-ignition and turbu- [22] H.G. Weller, G. Tabor, H. Jasak, C. Fureby, Comput.
lent combustion, and the influence of increasing T0 . Phys. 12 (1997) 620–631.
[23] D. Drikakis, C. Fureby, F.F. Grinstein,
M. Liefendahl, Implicit Large Eddy Simulation:
Acknowledgment Computing Turbulent Fluid Dynamics, Cambridge
University Press, 2007, p. 94.
The authors acknowledge the support of ON- [24] E. Hairer, G. Wanner, Solving Ordinary Differential
ERA and FOI, respectively; VS was also sup- Equations, II: Stiff and Differential-Algebraic Prob-
ported by the Grant of the Ministry of Education lems, 2nd Ed, Springer Verlag Berlin, Heidelberg,
and Science of the Russian Federation (Contract 1996.
[25] I. Celik, Z.N. Cehreli, I. Yavuz, ASME J. Fluids Eng.
no. 14.G39.31.0001 of 13.02.2017). The assistance 127 (2005) 949–958.
from N. Wikström is acknowledged. [26] T.J. Poinsot, S.K. Lele, J. Comput. Phys. 101 (1992)
104–129.
[27] C. Fureby, Whither Turbulence and Big Data for the
Supplementary materials 21st Century, Springer Verlag, 2015, p. 375.
[28] A. Larsson, N. Zettervall, T. Hurtig, et al., Energy
Supplementary material associated with this ar- Fuels 31 (2017) 1904–1926.
ticle can be found, in the online version, at doi:10. [29] N. Zettervall, C. Fureby, AIAA (2018) 2018-1146.
1016/j.proci.2018.05.127. [30] D.J. Beerer, V.G. McDonell, J. Eng. Gas Turb. Power
130 (2008) 051507.
[31] V.A. Alekseev, M. Christensen, A.A. Konnov, Com-
References but. Flame 162 (2015) 1884–1898.
[32] C.A. Jachimowski, NASA Technical Paper 2791,
[1] D. Andreadis, Ind. Phys. 10 (2004) 24. 1988.
[2] A. Ben-Yakar, R.K. Hanson, Proc. Combust. Inst. 27 [33] D.M. Davidenko, I. Gökalp, E. Dufour, P. Magre,
(2001) 2173–2180. AIAA 2003 (2003) 7003.
[3] T. Sunami, P. Magré, A. Bresson, F. Grisch, [34] N.M. Marinov, C.K. Westbrook, W.J. Pitz, in: Pro-
M. Orain, M. Kodera, AIAA 2005 (2005) 3304. ceeding of the Eighth International Symposium
[4] S.J. Laurence, S. Karl, J. Martinez Schramm, on Transport Properties, San Fransisco, CA, USA,
K. Hannemann, J. Fluid Mech. 772 (2013) 85–120. 1995.
A. Vincent-Randonnier et al. / Proceedings of the Combustion Institute 37 (2019) 3703–3711 3711

[35] R. Blumenthal, Experimentelle Untersuchung und [36] S. Samuelsen, V. McDonell, M. Greene, D. Beerer,
Numerische Simulation der Selbstzuendung von Correlation of Ignition Delay with Natural Gas and
Kraftstoff/Luft-Gemischen im Stosswellenrohr unter IGCC type Fuels, DOE Award Number: DE-FC26-
Beruecksichtigung stroemungs-mechanischer Ein- 02NT41431, 2006.
fluesse Doctoral Thesis, Institut für Allgemeine [37] J. Jeong, F. Hussain, W. Shoppa, J. Kim, J. Fluid
Mechanik, RWTH Aachen, Germany, 1996. Mech. 332 (1997) 185–214.

Potrebbero piacerti anche