CEAS Space Journal (2020) 12:385–395

https://doi.org/10.1007/s12567-020-00307-7

ORIGINAL PAPER

Post‑test analysis of the LAPCAT‑II subscale scramjet

Sebastian Karl1 · Jan Martinez Schramm1 · Klaus Hannemann1

Received: 12 July 2019 / Revised: 5 February 2020 / Accepted: 9 March 2020 / Published online: 19 March 2020
© The Author(s) 2020

Abstract
A subscale flight experiment configuration propelled by a Mach 8 supersonic combustion ramjet (scramjet) was designed
within the framework of the European Commission co-funded Long Term Advanced Propulsion Concepts and Technologies
II project. The focus of this design exercise was to verify by ground testing the ability of the proposed scramjet engine to
produce adequate thrust for hypersonic level flight. Experiments performed in DLR’s High Enthalpy Shock Tunnel Göttingen
confirmed precedent CFD predictions of the total thrust and demonstrated the operability of the vehicle. Yet, significant
discrepancies between the CFD analyses, which were performed to design the vehicle, and subsequent detailed measure-
ments of the pressure distribution in the combustor were observed and could not be resolved so far. This paper focuses on
a further analysis of these residual discrepancies. It was found that the CFD predictions of the combustor pressure distribu-
tion are sensitive to the configuration of the intake boundary layer. Particularly, different assumptions for the location of
the laminar to turbulent boundary layer transition strongly influence cross-flow structures which develop on the intake and
which are able to trigger different combustion modes in the combustor. While the effect on total vehicle performance remains
limited, a significant impact on the structure and magnitude of the surface pressure distribution was observed. i.e., the large
combustor peak pressures, which occur in the experiment, can be explained by the occurrence of a strong shock train in the
vicinity of the combustor wall.

Keywords Scramjet · CFD · Turbulence modeling · Intake flow

Abbreviations SSFE Small-scale flight experiment

AUSMDV Advection upstream splitting method, upwind TAU​ DLR Navier–Stokes flow solver
flux splitting scheme TS Terminal shock
BS Bow shock (semi-strut H2 injection) TSP Temperature-sensitive paint
CFD Computational fluid dynamics p Pressure
EU-FP7 7th Framework Program of the EU P0 Total pressure
HEG High Enthalpy Shock Tunnel Göttingen q Surface heat flux
LAPCAT​ Long Term Advanced Propulsion Concepts u Flow velocity
and Technologies
MR2 Large-scale Mach 8 cruise configuration in
LAPCAT​ 1 Introduction
MUSCL Monotonic Upwind Scheme for Conservation
Laws The European Commission co-funded research project LAP-
RSM Reynolds stress model (menter) CAT-II [1] explored technological fundamentals of hyper-
RBS Reflected bow shock sonic air transport. It was found that scramjet propulsion
RTS Reflected terminal shock systems appear to be the most practical option to ensure
SA Spalart–Allmaras turbulence model long-range cruise at flight regimes beyond a Mach num-
ber of approximately six. Nevertheless, neither the present
technological state-of-the-art nor the empirical experience
* Sebastian Karl with these engines allows conclusions about their practi-
[email protected]
cal applicability for large-scale and long-range vehicles.
1
DLR Institute of Aerodynamics and Flow Technology,
Bunsenstr. 10, 37083 Goettingen, Germany

13
Vol.:(0123456789)
386 S. Karl et al.

Substantial further development, testing and qualification 2 SSFE vehicle layout

are still required to prepare potential future applications.
As a part of this effort, a small-scale flight experiment The basic layout of the 1.5-m-long hydrogen-fueled vehi-
configuration (SSFE) was derived from the LAPCAT MR2 cle is shown in Fig. 1. The design was driven by the opti-
full-scale vehicle concept for testing in high-enthalpy facili- mal integration of a high-performance propulsion unit
ties at DLR and ONERA at flight Mach numbers of about within an aerodynamically efficient wave rider design,
M = 8. The design of a small engine is particularly difficult. while guaranteeing sufficient volume for tankage, pay-
This is due to the unfavorable ratio of wetted surface to vol- load and other subsystems [5]. The optimization of the
ume and the relatively long combustor needed for complete aerodynamic shape, the propulsion system, the injection
fuel consumption. To compensate for these adverse effects, concept and initial performance predictions were entirely
the SSFE scramjet operates near an equivalence ratio of performed by means of CFD analyses [6].
one and fuel mixing was carefully optimized by a two-stage The inward turning intake was designed based on a
multi-strut injection concept. modified Busemann template flow field using a 3D stream
The ground-based testing in DLR’s HEG facility [2] con- tracing method. This template flow field consists of a
firmed the operability of the vehicle. Good agreement of 5° internal conical deflection followed by an isentropic
the general performance characteristics (thrust at different axisymmetric internal compression and a conical terminal
angles of attack and equivalence ratios) between the CFD- shock which redirects the compressed flow in free stream
based design and the experiments was achieved. However, direction [3]. The intake has an elliptical capture area in
the experiments also revealed consistent and significant the yz-plane with a ratio of major to minor axes of three
deficiencies of the CFD results regarding the pressure dis- and an overall contraction ratio of about nine. An insert
tribution inside the combustor (see Sect. 3). Further analyses with an elliptical shape in the xy-plane, indicated by the
of the intake flow [3] showed a strong dependency of the black line on the left part of Fig. 1, was added in front
combustor inflow conditions on the boundary layer transition of the streamline traced intake surface to enable smooth
characteristics on the intake. This leads to the hypothesis integration with the aerodynamic waverider design of the
that the boundary layer properties on the intake can have a vehicle on the windward side (Fig. 1 (bottom view)). The
substantial impact on the flow structure inside the combus- shape of the upstream part of the waverider windward side
tor. The results of detailed heat flux measurements using dictates the upstream boundary of this insert. The intake
temperature-sensitive paint [4] which recently became avail- feeds an elliptical combustor in which a two-stage multi-
able allow for the first time to prescribe a realistic boundary strut injector system was installed. The elliptical shape
layer transition behavior to the CFD analyses. This paper of the internal combustor flow path provides an optimum
investigates the effect of intake boundary layer transition compromise between wetted area and efficient fuel pen-
on the combustor operation and discusses the resulting dif- etration and mixing. The nozzle was laid out in two sec-
ferences with respect to the original vehicle design which tions. The first nozzle section (upstream, light green) has
was based on the assumption of turbulent boundary layers an area ratio of three and blends the elliptical combustor
on the entire intake surface. CFD predictions of the detailed cross section into a circular shape. During ramjet mode,
flow properties inside the combustion chamber are assessed this nozzle is used as an additional combustor and oper-
based on available high-resolution pressure measurements ates in thermally chocked mode. The second nozzle sec-
in Sect. 5. Main characteristics of the combustion process tion (downstream, dark green) was stream traced from an
and the flow structure inside the combustor are postulated axisymmetric isentropic expansion and truncated to a suit-
based on a combination of CFD results and available experi- able length, resulting in an expansion ratio of about ten.
mental data.

Fig. 1  Schematic of the small-

scale scramjet flight configura-
tion under investigation. Blue:
intake, red: combustion cham-
ber, green: thrust nozzle, gray
translucent: waverider

13
Post‑test analysis of the LAPCAT‑II subscale scramjet 387

One of the main challenges of the combustor design was Table 1  Free stream conditions
to control the boundary layer separation due to adverse pres-
Mach number 7.355
sure gradients in the vicinity of the combustor entrance. The
Static pressure 2060 Pa
vehicle does not include boundary layer control devices or
Static temperature 263 K
bleeds. Thus, the boundary layer entering the combustor is
Density 0.02717 kg/m3
already affected by the strong compression on the intake and
Flow velocity 2396 m/s
is therefore sensitive to separation. To remedy this issue, a
Unit Reynolds number 3.8E6 1/m
diverging combustor geometry and controlled and distrib-
Angle of attack −2 °
uted heat release with a staged injection scheme were fore-
Total pressure 17.73 MPa
seen in the design. A combination of an upstream injection
stage consisting of two semi-struts and a downstream stage
with a central full strut was found to provide the best config-
uration concerning combustion efficiency, flame anchoring and numerical predictions of the global performance char-
and separation control [6]. The resulting combustor layout acteristics was achieved.
is shown in Fig. 2. The distribution of the injected hydrogen Besides the successful demonstration of the operability
at design conditions is indicated by the isosurfaces of 5% of the vehicle, the wind tunnel tests revealed consistent
mass fraction. The two semi-struts inject about 65% of the and significant deficiencies of the CFD predictions of the
gaseous hydrogen fuel through a Mach 2 nozzle normal to pressure distribution inside the combustor. A comparison
the main flow direction. The full strut injects the remain- of surface pressure distributions between the CFD design
ing fuel through 4 Mach 2 nozzles at an angle of 15° with case and the available experimental results is shown in
respect to the main flow direction. The hydrogen injected Fig. 3. The results correspond to an angle of attack of
by the semi-struts covers the outboard regions, whereas the − 2° and an equivalence ratio of one. The CFD design was
full strut distributes fuel into the central part of the combus- based on the assumption of fully turbulent flow (no bound-
tor. This arrangement also minimizes critical boundary layer ary layer transition), and the Spalart–Allmaras turbulence
disturbances in the upstream central part of the combustor model was applied.
wall which is most sensitive to flow separation. The location of the surface pressure measurements is
indicated by the color of the symbols and the schematic in
the figure. Red symbols correspond to the centerline on the
3 Previous observations intake and thrust nozzle, and black symbols correspond to
the combustor side wall. Due to design constraints of the
Comprehensive ground-based testing in the HEG facility wind tunnel model, no experimental pressure data are avail-
of DLR confirmed the CFD predictions concerning the net able at other locations (e.g., combustor centerline). Because
thrust and the general operability of the small-scale vehi- the intake pressure distribution is not affected by the fuel-
cle [2]. The free stream conditions of the experiments are on operation of the combustor, only side wall pressures are
summarized in Table 1 and correspond to a flight velocity shown for the fuel-on case for clarity (right subfigure). The
of Mach 7.4 at an altitude of 27 km. The test time in the axial location of the intake leading edge, the combustor
HEG shock tunnel was about 3 ms, and the total pres- entrance, the semi-strut injection ports, the full-strut injec-
sure and enthalpy at the present flow conditions amount to tion ports and the thrust nozzle entrance are at x = 0, 440,
17.73 MPa and 3.24 MJ/kg, respectively. The ground tests 463, 610 and 763 mm, respectively. The pressure scales of
were complemented by detailed CFD analyses of the wind the fuel-off and fuel-on parts in Fig. 3 are different to clearly
tunnel flow, and good agreement between experimental highlight the differences in the fuel-off case.

Fig. 2  Combustor layout with

injection scheme. Injectors are
shown in red, and the computed
hydrogen distribution is indi-
cated by the brown isosurfaces
of 5% mass fraction

13
388 S. Karl et al.

Fig. 3  Fuel-off (left) and fuel-on (right) surface pressure distributions of the design case compared to experimental data; data are normalized by
the free stream total pressure, P0

Large differences between CFD prediction and experi- present investigation was to quantify the effect of the transi-
mental measurements of the surface pressure distribution tion location on the flow structure in the combustor and to
on the combustor side wall occur for the fuel-on case. The identify the impact on the CFD prediction of the combustor
strong experimental peak at an axial coordinate of 650 mm surface pressure distribution.
in Fig. 3 (right) is not reproduced by the numerical analyses
of [2]. The location of this peak correlates with the position
of the second injector stage. The level of the experimental 4 Numerical model used for the present
peak pressure in the combustor exceeds by far the maximum investigations
theoretical pressure rise due to hydrogen combustion. This
theoretical maximum is indicated by the Rayleigh line in the All numerical simulations in the present study were per-
right part of Fig. 3 (p/P0 of about 24 × 10−3). It was evaluated formed with the hybrid structured–unstructured DLR
using conservation of total mass, momentum and energy in Navier–Stokes solver TAU [7]. The TAU code is a sec-
a diverging duct with heat addition (Rayleigh flow). The ond-order finite-volume flow solver for the Euler and
assumed heat addition corresponds to complete fuel con- Navier–Stokes equations in their integral forms, using eddy
sumption at an equivalence ratio of one. The equivalent 1D viscosity, Reynolds stress or detached and large eddy simula-
flow properties before heat addition are computed from the tion for turbulence modeling.
CFD solution at the combustor entrance by stream-thrust For the present investigation, we employed the
averaging. The experimental exceedance of the Rayleigh Spalart–Allmaras one-equation eddy viscosity model [8] and
pressure indicates the presence of a strong shock system the Wilcox stress-omega Reynolds stress model [9] with a
which was absent in the initial CFD results. correction to avoid unphysical turbulent production across
Additional CFD investigations of the isolated intake flow strong compression shocks [10]. The choice of turbulence
[3] revealed a strong dependence of the flow structure at models was motivated by the reproduction of the numeri-
the combustor entrance on the assumption of the boundary cal setup used for the vehicle design and the previous test
layer transition location on the intake surface in the fuel-off analysis (Spalart–Allmaras) and by the application of a more
case (corresponding results of the present investigation in advanced model (Reynolds stress 7-equation).
Fig. 6). This unexpectedly large effect is due to the accumu- The AUSMDV flux-vector splitting scheme was applied
lation of low-momentum flow along the center plane of the together with MUSCL gradient reconstruction to achieve
intake driven by cross-flow effects. This cross-flow strongly second-order spatial accuracy. Hydrogen combustion
depends on the local boundary layer thickness which is was modeled with a finite-rate chemistry approach. The
affected by the transition location. The main objective of the fluid is considered to be a reacting mixture of thermally

13
Post‑test analysis of the LAPCAT‑II subscale scramjet 389

perfect gases, with a transport equation solved for each of 5 Results

the individual species. The chemical source terms in this
set of transport equations are computed from the law of 5.1 Intake flow
mass action by summation over all participating reactions.
The forward reaction rate is computed using the modified The results of recent measurements of the heat flux distribu-
Arrhenius law, and the backward rate is obtained from the tion on the complete intake surface by temperature-sensitive
equilibrium constant, which is derived directly from the paint (TSP) [4] were used to prescribe a realistic location
partition functions of the participating species. A modi- of the laminar to turbulent boundary layer transition line
fied Jachimowski reaction mechanism [11] for hydro- as an input for CFD investigations. The error of the TSP
gen–air mixtures was applied for this investigation. This data is below 0.035% for the measured temperature [15]
mechanism includes both hydrogen peroxide (­ H2O2) and which results in an estimated error for the surface heat flux
the perhydroxyl radical ­(HO2) and has been shown to be of less than 8%. A resulting numerical heat flux distribution
applicable over a wide range of pressures, densities and (RSM case) is shown in Fig. 4 together with an indication
equivalence ratios. Laminar reaction rates were used, and of the assumed transition line. The figure shows a top view
no specific model for the influence of turbulent fluctua- of the front section of the complete vehicle (see Fig. 1, view
tions on the chemical production rates was employed. in negative z-direction). For clarity, heat flux contours are
The accuracy of the present numerical approach was only shown on the intake surface (blue surface in Fig. 1).
validated and assessed based on canonical test cases and The assumed boundary layer transition line is shown on all
the HyShot scramjet configuration [12–14]. external and internal surfaces which are visible in the 3D
The free stream conditions are summarized in Table 1 top view.
and are representative for the SSFE ground tests in the A quantitative comparison of numerical and experimen-
HEG shock tunnel of the German Aerospace Center, DLR tal TSP heat flux profiles along three cuts at 50-, 200- and
[2]. They correspond to a flight altitude of approximately 350-mm downstream of the leading edge is shown in Fig. 5.
27 km. The transition line from Fig. 4 (bottom) was used in the
During ground testing in short-duration facilities the turbulent computations; hence, the state of the intake bound-
wall temperature remains approximately constant. The ary layer in the cuts upstream of 200 mm is always laminar.
test time in HEG was about 3 ms. Hence, the wall tem- The agreement in the laminar zone (cuts at 50 and 200 mm)
perature is fixed to 300 K for the viscous computations. is very good. Minor differences between the CFD curves
The computational domain comprises a half model of the are related to the application of different grid densities. The
vehicle (Fig. 1). The flow inside the hydrogen feeding cut through the turbulent zone at 350 mm (see also Fig. 4)
system was computed separately, and the outflow profiles shows that the Spalart–Allmaras (SA) model over-predicts
of the supersonic injection nozzles were prescribed as a the experimental heat flux level by about 40%. Improved
Dirichlet condition on the strut surfaces. The unstructured agreement is achieved when the shock-corrected Reynolds
computational grids consist of about 19 × 106 points/vol- stress model (RSM) is applied.
umes for the computational case using the experimental The flow properties at the combustor entrance show a
transition line and RSM turbulence model and of about significant dependency on the boundary layer properties of
7.4 × 106 points for all other cases. Boundary layers are
discretized with prismatic layers employing a dimension-
less wall spacing of y+ = O(1). The grid convergence of
the present computational setup for external and intake
flows was demonstrated in [3]. The external and combustor
flows were treated in segregated manner to reduce compu-
tational cost. The presence of combustion does not have an
upstream influence on the intake flow. Hence, the combus-
tion chamber and thrust nozzle (red and green regions in
Fig. 1) were treated in a separate computational zone. The
inflow conditions were prescribed as a Dirichlet condition
(specification of the complete flow state, i.e., velocities,
partial species densities and pressure) from a nose-to-tail
computation of a complete vehicle. The combustor outflow
was then prescribed as a Dirichlet condition in the back to Fig. 4  CFD surface heat flux distribution on the intake surface (top)
and locus of the transition line (bottom), light gray, downstream of
the nose-to-tail grid. x = 200 mm: turbulent regions, dark gray: laminar regions, black
lines: locus of the cuts in Fig. 5

13
390 S. Karl et al.

The figure shows the front part and intake of the vehi-
cle in a front-top view. The velocity contours are blanked
above 2000 m/s to highlight the extents of the boundary
layer and low-momentum structures. The left part of Fig. 6
shows skin friction patterns and low-momentum structures
resulting from two different turbulence models and bound-
ary layer transition assumptions (RSM transitional and
SA turbulent representing the vehicle design case). The
difference in boundary layer thickness is small; however,
the structure of the near-wall cross-flow and the pockets
of decelerated fluid are clearly affected. Yet, streamlines
that are off-set by 2 mm from the wall coincide with the
Busemann design as shown in the right part of this figure.
This confirms that the cross-flow phenomenon is limited
to small layer close to the walls.
The cross-flow is initiated by the sweep in the wall geom-
etry located at the intersection of the insert and the stream-
traced compression surface (indicated by the dashed blue
line in Fig. 6). The skin friction lines turn sharply inward
at this location. The driving mechanism is related to the
Fig. 5  Heat flux distribution along axial cuts as indicated in Fig. 4; conservation of the tangential velocity at oblique shocks.
laminar solution (lam) included for reference; experimental values are The shock originating at the intersection of the insert and
TSP measurements the inclined compression surface is misaligned in a near-
wall region due to the presence of entropy (blunt leading
the intake. This is related to the development of a pocket of edges) and boundary layers. Here, the local Mach number
decelerated fluid along the symmetry plane of the intake. and shock angle deviate from the inviscid ideal design based
Near the combustor entrance, this low-momentum pocket on the template flow field resulting in a misalignment of the
covers about 20% of the cross section of the internal flow flow direction downstream of the shock. A detailed analysis
path. The flow structure is caused by fluid from inside of the cross-flow properties and the related driving mecha-
the boundary layer, which is forced to entrain into the nisms was performed in previous studies [3].
bulk flow near the symmetry plane. This entrainment of The cases shown in Fig. 6 show a remarkable differ-
low-momentum boundary layer fluid is driven by a strong ence of the flow structures at the combustor entrance (last
converging cross-flow. This situation is illustrated by the downstream cut of axial velocities) between the two con-
skin friction lines and axial velocity contours in Fig. 6. sidered cases.

Fig. 6  Skin friction lines at the intake surface (left) and streamlines 2-mm off-set from the wall (right) with contours of axial velocity for differ-
ent turbulence models and transition locations

13
Post‑test analysis of the LAPCAT‑II subscale scramjet 391

5.2 Combustor pressure distributions which was used during vehicle design (results in Fig. 3, SA
model, fully turbulent) are visible for the fuel-on results. A
Pressure distributions in the combustor and on the intake series of strong pressure peaks develops between x = 550
(fuel-off) and in the combustor (fuel-on) are shown in Figs. 7 and 600 mm. The magnitude of the numerical peak pres-
and 8. The results in Fig. 7 represent the case of transitional sures now corresponds to the experimental observations and
flow and the application of the shock-corrected RSM model. exceeds the Rayleigh pressure. This indicates the presence
The agreement at fuel-off conditions is slightly improved of a strong shock structure in the vicinity of the combustor
(reduced over-prediction of surface pressure at x = 500 mm), side wall. The surface pressure distribution is characterized
but remarkable differences compared to the numerical setup by the presence of strong gradients. This is indicated by

Fig. 7  Surface pressure distribution for the RSM model/transitional flow. Left: centerline (red) and combustor side wall (black); right: combustor
side wall (black and red)

Fig. 8  Surface pressure distribution for the SA model/transitional flow

13
392 S. Karl et al.

the red symbols in Fig. 7 which correspond to the pressure 539 N for the transitional RSM case and 529 N for the tran-
range in a region of ± 2 mm around the centerline of the sitional Spalart–Allmaras case compared to 580 N for the
combustor side wall (black line in the combustor schematics experiment.
and mounting location of the pressure sensors of the wind
tunnel model). Nevertheless, the location of the shock train 5.3 Flow field structure
is predicted too far upstream by the numerical simulation.
The pressure distributions in Fig. 8 represent numeri- The strong differences between combustor surface pressure
cal results for the application of the Spalart–Allmaras tur- distributions which occur for slightly modified numerical
bulence model and a reduced reaction mechanism which boundary conditions are caused by the occurrence of dif-
excludes ­H2O2 and ­HO2 [16]. The assumption of the intake ferent flow field structures inside the combustor. Figures 9
boundary layer transition is identical to the results in Fig. 7. (3D view) and 10 (top view) illustrate the flow field in one
The impact on the fuel-off pressure distribution is limited, half of the symmetric combustor for the RSM/transitional
but, again, significant changes occur in the fuel-on case. computational case for which the best agreement with the
The general pressure level between x = 550 and 750 mm is experiments was observed (Fig. 7).
strongly reduced, and, contrary to the plateau shape of the The gray isosurfaces of large negative velocity diver-
design case in Fig. 3, pressure peaks are visible which indi- gence are used to visualize shock structures. Green isosur-
cated the presence of supersonic flow in the vicinity of the faces of negative axial velocity show flow separation zones.
surface. The brown surface (H2 mass fraction of 0.1) indicates the
The fuel-on surface pressures in the nozzle region (down- fuel distribution. The sonic line (blue surface) separates
stream section, x > 750 mm) are similar for all numerical supersonic and subsonic flow regions. The Mach number
setups in Figs. 3, 7 and 8. Differences are only visible in distribution is shown on a cut plane through the centerline
the combustor region between x = 550 and 750 mm. This of the combustor. The surface pressure results in Figs. 3, 7
indicates that, despite remarkable pressure deviations in the and 8 are located at the intersection line of this cut plane
combustor region, the vehicle performance (thrust) is not and the outboard combustor wall. The combustor entrance
strongly affected by the different modeling assumptions. is at x = 400 mm, and the full-strut injector is located at
This is confirmed by the evaluation of the total net thrust x = 600 mm.
from the nose-to-tail computations (difference of the vehicle Different flow structures are labeled in the top view of
drag force between fuel-on and fuel-off conditions). This Fig. 10. The terminal shock (TS) entering the combustor
is 536 N for the design case (Spalart–Allmaras turbulent), is a feature of the Busemann template flow (see Sect. 2) in
which the intake design was based on. It is a conical shock
wave with the tip anchored at the focal point of the waves
emanating from the intake compression surface. It aligns
the flow direction after the isentropic intake compression
to the axial direction of the combustor. This shock wave
is strongly affected by the presence of the low-momentum
fluid which enters the combustor due to the presence of the
intake cross-flow (see axial velocity contours in Fig. 6).
After passing the bow shock caused by the hydrogen injec-
tion at the semi-struts (BS), the terminal shock is reflected at
the combustor wall (RTS) and causes flow separation in the
vicinity of its impingement point (TS separation). A strong
shock system is located between a second interaction of the
Fig. 9  Flow field structure reflected terminal shock (RTS) and the reflected bow shock

Fig. 10  Top view of the flow

field structure (TS terminal
shock, RTS reflected terminal
shock, BS bow shock of the
semi-strut hydrogen injection,
RBS reflected bow shock)

13
Post‑test analysis of the LAPCAT‑II subscale scramjet 393

(RBS) and the semi-strut injector. The major part of the

combustor flow is subsonic downstream of this shock system
(red Mach number coloring).
The flow structure which produces the prominent pres-
sure peaks as shown in Fig. 7 is a strong shock train which
develops in the vicinity of the combustor wall (see Fig. 10).
This shock train is initiated at the global normal shock which
separates the super- and subsonic combustion zones. The
reason of the occurrence of the shock train is the strong
gradient of flow properties upstream of the normal shock
system. Close to the wall, the flow is not affected by com- Fig. 12  Surface pressure distribution of the shock train model prob-
bustion and, therefore, consists of cold air. The combustion lem
zone starts further away from the wall (around the brown
­H2 isosurface) and is characterized by hot gas at a reduced
Mach number. The Mach number decrease is primarily due air jet downstream of the initial shock exceeds the value
to the combustion-driven temperature rise. The pressure rise of the hot air/H2O mixture by more than a factor of 2 (see
due to the presence of the normal shock is larger for the cold comparison of black and red lines in Fig. 12). The result
flow close to the wall than for the hot flow in the combustion is the occurrence of a similar flow pattern as for an under-
zone. The resulting pressure mismatch downstream of the expanded jet in which the static pressure in the cold air jet
normal shock relaxes in form of a shock train. The physical adapts to the surrounding reference value. This reference
mechanism is illustrated by means of a model problem in pressure in Fig. 12 is resulting from a computation of the
next Sect. 5.4. pure hot air and ­H2O flow at the same boundary conditions.
This confirms that the shock train formation in the SSFE
5.4 Model problem for the occurrence of the shock combustor is most probably related to the presence of flow
train property gradients upstream of a strong shock system which
results in subsonic flow downstream of it.
A generic two-dimensional model problem as illustrated
in Fig. 11 is considered to clarify the mechanism of shock
train formation. A mixture of air and H ­ 2O at 2000 K (repre- 5.5 Comparison of flow field structures
sentative for the combustion products in the combustor flame for the different numerical model setups
zone) is injected into a slightly diverging channel and forms
a normal shock wave to match the specified exit pressure at The occurrence of the shock train resulting in the reproduc-
the outflow boundary. Close to the bottom wall, a thin layer tion of experimental peak pressures was only achieved for
of air at 1500 K is injected. This layer is representative for the application of the RSM turbulence model and a realistic
the flow conditions outside of the combustion zone close to assumption for the transition line. Other modeling setups
the wall in the upstream part of the SSFE combustor. Both result in different combustion modes (subsonic smooth or
layers share the same inflow velocity and static pressure. supersonic at low pressures) as discussed in Sect. 5.2. The
The results in Figs. 11 and 12 show the presence of a occurrence of these modes is triggered by small changes in
shock train and the associated strong pressure oscillations the flow topology as illustrated by the schematic flow field
downstream of the initial normal shock wave. Due to the structures on the cut plane through the combustor center
different upstream gas properties, the pressure in the cold in Fig. 13. The gray isosurfaces (shocks) and the blue

Fig. 11  Model problem for the

occurrence of the shock train
(isolines: Mach number)

13
394 S. Karl et al.

Fig. 13  Flow field structures for the applied numerical modeling setups (reference: RSM/transitional; supersonic: SA/transitional; subsonic: SA/
turbulent, red ellipse: footprint of decelerated fluid from intake cross-flow)

isosurface (sonic line) from Figs. 9 and 10 are reproduced ignites, and the combustion zone is extended up to the wall.
as gray and blue lines, respectively. The flow becomes subsonic at this separation-driven com-
The top part of the figure represents the reference case bustion zone and remains subsonic in downstream direction.
(RSM/transitional, details in Fig. 10) for comparison pur- This results in the smooth pressure variation shown in Fig. 3.
poses. The main flow features are discussed in Sect. 5.3. For all cases, the angle and configuration of the terminal
In the supersonic case (Spalart–Allmaras model/transi- shock is affected by the pocket of decelerated fluid which
tional), the impingement of the terminal shock at the com- enters the combustor. The footprint of this pocket is high-
bustor wall and the associated separation bubble are shifted lighted by the red ellipses in Fig. 13. This shock config-
in downstream direction (TS separation). The separation uration and the strength of the associated terminal shock
bubble is smaller, and the effect of this separation, to push separation (TS) influence the geometrical location of the
the hydrogen jet and the associated combustion zone away combustion zone and the associated flow structure and pres-
from the wall, is less pronounced. The flame zone is closer to sure distribution at the combustor wall.
the wall and the gradient of cold to hot flow at the beginning
of the shock train in the reference case is less pronounced.
Here, no shock train is formed and the flow remains super- 6 Conclusion
sonic until it reaches a single shock further downstream.
This results in a reduced surface pressure at the combustor The large discrepancy between the observed numerical
wall in the supersonic flow region up to x = 600 mm (see surface pressure distributions in Figs. 3, 7 and 8 is due
Fig. 8). to the occurrence of different flow structures close to the
In the subsonic case (Spalart–Allmaras model/turbulent, combustor side wall. Supersonic flow causes low pressures
used during vehicle design), the impingement location of the with visible peaks, and subsonic flow results in a smooth
terminal shock is shifted upstream. The associated separa- variation of pressure. Only an in-between mode, in which
tion bubble (TS separation) is larger and entrains hydrogen a local shock train occurs, reproduces the experimental
into the air flow close to the wall. This entrained hydrogen peak pressures. The combustion modes are triggered by the

13
Post‑test analysis of the LAPCAT‑II subscale scramjet 395

geometrical distribution of the flame zone. This distribu- 2. Hannemann, K., Martinez Schramm, J., Karl, S., Laurence, S. J.:
tion is affected by the configuration of the terminal shock Free flight testing of a scramjet engine in a large scale shock tun-
nel, AIAA 2015-3608. In: 20th AIAA International Space Planes
of the intake flow entering the combustor and the strength and Hypersonic Systems and Technologies Conference, Glasgow
of a local separation bubble at its impingement point on the (2015)
combustor side wall. The structure of the terminal shock is 3. Karl, S., Steelant, J.: Crossflow phenomena in streamline-traced
altered as it passes through a pocket of decelerated fluid at hypersonic intakes. J. Propul. Power 34(2), 449–459 (2018)
4. Martinez Schramm, J., Karl, S., Ozawa, H., Hannemann, K.: Heat
the combustor entrance. Cross-flow, which occurs close to flux measurements by means of ultra-fast temperature sensitive
the intake surface, generates this low-momentum fluid. The paints on the intake of the small scale flight configuration of the
structure of the cross-flow pattern is strongly affected by the LAPCAT II vehicle in the high enthalpy shock tunnel Göttingen.
turbulence model and transition assumption for the intake In: HiSST: International Conference on High-Speed Vehicle Sci-
ence Technology, Moscow, Nov. 26–29 (2018)
flow. By this mechanism, the numerical simulation results 5. Murray, N., Steelant, J., Mack, A.: conceptual design of a mach
for the combustor pressure distribution are very sensitive to 8 hypersonic cruiser with dorsal engine. In: Proceedings of the
the modeling of the intake boundary layer. Sixth European Symposium on Aerothermodynamics for Space
The combustor pressures which are experimentally Vehicles, ESA SP-659, European Space Agency, ISBN 978-92-
9221-223-0 (2009)
observed exceed the Rayleigh pressure and clearly indicate 6. Langener T, Steelant J., Karl S., Hannemann K.: Layout and
the presence of a strong shock train close to the surface. design verification of a small scale scramjet combustion cham-
Present numerical simulations using a realistic assumption ber. In: Proceedings of the XXI International Symposium on
for the laminar to turbulent transition on the intake repro- Air Breathing Engines (ISABE 2013), ISABE Paper 2013-1655
(2013)
duce the peak pressure level. Yet, the location of the peak 7. Langer, S., Schwöppe, A., Kroll, N.: The DLR flow solver TAU—
pressures is predicted too far upstream. Generally, the flow status and recent algorithmic developments, AIAA Paper 2014-
structure in the combustor is characterized by a complex 0080 (2014)
interplay of turbulent flow, shock and expansion waves and 8. Spalart, P.R., Allmaras, S.R.: A one equation turbulence model
for aerodynamic flows. La Recherche Aerospatiale 1, 5–21 (1994)
combustion. Hence, the application of scale-resolving turbu- 9. Wilcox, D.C.: Turbulence Modeling for CFD. DCW Industries
lence models together with improved boundary layer transi- Inc., La Canada (2006)
tion prediction in the regions which are not accessible by the 10. Karl, S., Hickey, J.P., Lacombe, F.: Reynolds stress models for
experiment (e.g., upper part of the combustor) are likely to shock: turbulence interaction. In: 31st International Sympo-
sium on Shock Waves 1, 511–517, Springer (2018) https​://doi.
lead to improved predictions in the future. org/10.1007/978-3-319-91020​-8
Despite the large discrepancies of surface pressure on the 11. Gerlinger, P.: An implicit multigrid method for turbulent combus-
combustor wall, the pressure distributions on the thrust noz- tion. J. Comput. Phys. 167, 247–276 (2001)
zle and the integral vehicle performance are insensitive to 12. Karl, S.: numerical investigation of a generic scramjet configura-
tion, Ph.D. Dissertation, Faculty of Mech. Eng., Technical Uni-
the different modeling assumptions used in the present study. versity Dresden (2011) [available online] URL: http://nbn-resol​
ving.de/urn:nbn:de:bsz:14-qucos​a-68695​. Accessed Jan 2017
Acknowledgements Open Access funding provided by Projekt DEAL. 13. Hannemann, K., Karl, S., Martinez Schramm, J., Steelant, J.:
Methodology of a combined ground based testing and numerical
Open Access This article is licensed under a Creative Commons Attri- modelling analysis of supersonic combustion flow paths. Shock
bution 4.0 International License, which permits use, sharing, adapta- Waves 20(5), 353–366 (2010). https​://doi.org/10.1007/s0019​
tion, distribution and reproduction in any medium or format, as long 3-010-0269-8
as you give appropriate credit to the original author(s) and the source, 14. Laurence, S.J., Karl, S., Schramm, J.M., Hannemann, K.: Tran-
provide a link to the Creative Commons licence, and indicate if changes sient fluid-combustion phenomena in a model scramjet. J. Fluid
were made. The images or other third party material in this article are Mech. 722, 85–120 (2013)
included in the article’s Creative Commons licence, unless indicated 15. Martinez Schramm, J., Edzards, F., Hannemann, K.: Calibration
otherwise in a credit line to the material. If material is not included in of fast-response temperature sensitive paints for their application
the article’s Creative Commons licence and your intended use is not in hypersonic high enthalpy flows, new results in numerical and
permitted by statutory regulation or exceeds the permitted use, you will experimental fluid mechanics XI, pp. 141–151, Springer (2016)
need to obtain permission directly from the copyright holder. To view a 16. Gaffney, R.L., White, J.A., Girimaji, S.S., Drummond, J.P.: Mod-
copy of this licence, visit http://creat​iveco​mmons​.org/licen​ses/by/4.0/. eling turbulent chemistry interactions using assumed pdf methods.
In: 28th AIAA/SAE/ASME/ASEE Joint Propulsion Conference
and Exhibit (1992)

References Publisher’s Note Springer Nature remains neutral with regard to

jurisdictional claims in published maps and institutional affiliations.
1. Steelant, J.: Sustained Hypersonic flight in europe: technology
drivers for LAPCAT II. In: 16th AIAA/DLR/DGLR International
Space Planes and Hypersonic System Technologies Conference.
Bremen, Germany, AIAA 2009-7240 (2009)

13