TASK QUARTERLY 9 No 1, 115–132

UNSTEADY SHOCK WAVE – TURBULENT

BOUNDARY LAYER INTERACTION
IN THE LAVAL NOZZLE
PIOTR DOERFFER 1, OSKAR SZULC 1
AND FRANCO MAGAGNATO 2
1 Institute of Fluid-Flow Machinery,
Polish Academy of Sciences,
Fiszera 14, 80-952 Gdansk, Poland,
{doerffer, osmark}@imp.gda.pl
2 Institute für Strömungslehre, Universität Karlsruhe,
Kaiserstr. 12, D-76128 Karlsruhe, Germany
[email protected]
(Received 2 September 2004)
Abstract: The flow in transonic diffusers and supersonic air intakes often becomes unsteady
due to shock wave-boundary layer interaction. Oscillations may be induced by natural separation
unsteadiness or forced by boundary conditions. Significant improvements of CFD tools, increased
computer resources and the development of experimental methods have again drawn the attention
of researchers to this topic.
Forced oscillations of a transonic turbulent flow in an asymmetric two-dimensional Laval
nozzle have been considered to investigate the problem. A viscous, perfect gas flow was numerically
simulated using SPARC, a Reynolds-averaged compressible Navier-Stokes solver, employing a two-
equation, eddy viscosity, turbulence closure in the URANS approach.
For time-dependent and stationary flow simulations, Mach numbers upstream of the shock
between 1.2 and 1.4 were considered. Comparison of computed and experimental data for steady
states generally gave acceptable agreement. In the case of forced oscillations, a harmonic pressure
variation was prescribed at the exit plane resulting in shock wave motion. Excitation frequencies
between 0Hz and 1024Hz were investigated at a constant pressure amplitude.
The main result of the work is the relation between the amplitude of shock wave motion
and the excitation frequency in the investigated range. Increasing excitation frequency resulted in
decreasing amplitude of the shock movement. At high frequencies, a natural mode of shock oscillation
(of small amplitude) was observed, which was insensitive to forced excitement.
Keywords: unsteady transonic flow, shock wave, nozzle flow

Nomenclature
X – the streamwise coordinate
Y – the crosswise coordinate
y + – non-dimensional distance from the wall

tq109m-e/115 11 V 2005 BOP s.c., http://www.bop.com.pl

116 P. Doerffer et al.

t– time
f– frequency
k– turbulent kinetic energy
τ– turbulent time scale
ε– the dissipation rate of turbulent kinetic energy
Tu – turbulence level at the inlet
µ– dynamic viscosity
µt– eddy viscosity at the inlet
γ– the ratio of specific heats, CP /CV
R – the gas constant
q– the wall heat flux
ρ– density
u, v– velocity components
E – total energy
Re – the Reynolds number
Ma – the Mach number
p(t)– pressure varying in time
Ap – amplitude of pressure varying in time
P – pressure level
P0 – total pressure at the inlet
T0 – total temperature at the inlet
CL – the lift coefficient
Cd – the drag coefficient
CFL – the Courant-Friedrichs-Levy number

1. Introduction
Recently unsteady transonic flows, mostly of two-dimensional geometries but
also in three-dimensional ones, have again received considerable attention and an
extensive description of inviscid and viscous flows can be found in literature. This
is due to the significantly improved experimental and numerical tools available
nowadays. In comparison to the one-dimensional approach, the present research can
reflect all details of the shock wave-boundary layer interaction structure.
Our investigations were carried out in a convergent-divergent nozzle. The case
was examined with Mach numbers upstream of the shock ranging from 1.2 to 1.4. An
ideal gas flow was considered with a Reynolds number of approximately Re = 1.4·106
based on the flow parameters in the throat [1].
Because unsteady and separated transonic flows were involved, numerical meth-
ods capable of solving time-dependent, mass-weighted, Reynolds-averaged, compress-
ible Navier-Stokes equations were considered. The SPARC finite volume computer
code [2] was used to calculate the results with the help of parallel computing (MPI)
on a Beowulf type cluster of PCs.
The experimental investigation concerned only stationary flow with steady
conditions at the nozzle outlet. Measurements included static pressures and a schlieren
visualisation of the flow structure. Only numerical results are available for the
unsteady cases. Steady state measurements and computations were carried out solely
to verify our numerical methods.
The main goal of the unsteady computations was to explore the influence of
frequency (ranging from 0Hz to 1024Hz) on the shock wave motion (for constant

tq109m-e/116 11 V 2005 BOP s.c., http://www.bop.com.pl

 Unsteady Shock Wave –Turbulent Boundary Layer Interaction. . . 117

amplitude pressure changes at the outlet). It was achieved with the help of a new
turbulence model [3, 4], utilizing a non-linear, cubic, Craft-Launder-Suga [5] formu-
lation. Spectral interaction between the driving unsteady phenomena and turbulence
was negligible due to low values of frequency.
Two new observations were made. A maximum of the amplitude was observed
at very low frequencies, where the moving shock wave passed beyond the stationary
shock boundaries. This happened only when the shock movement was very slow,
viz. below 50Hz. At approximately 500Hz, the shock wave response disappeared. For
higher excitations, a new natural oscillation appeared, with a much lower frequency,
which was insensitive to forced back pressure fluctuations. It was a shorter time step
that enabled us to capture the appearance of natural shock oscillations. The observed
shock movement was connected with the formation and shedding of vortices from the
separation zone downstream of the shock. Additionally, static pressure fluctuations
were analyzed at a few significant locations in the stream-wise direction.

2. Previous work
The majority of earlier work was done using simplified inviscid and laminar gas
dynamics theories. Just to mention a few, Adamson (1972, [6]) and later Adamson
and Richey (1973, [7]) used the general gas dynamics potential equation and the
Bernoulli equation with the small-disturbance theory and a similarity transformation
to analyze and compute numerically solutions of unsteady transonic flows with and
without planar shock waves in two-dimensional symmetric channels for stationary
and varying-in-time wall shapes. After abandoning the similarity solutions, the
proposed methodology was extended to arbitrary wall shapes and initial conditions
with oscillating back pressure (Richey and Adamson, 1976, [8]), curved shocks in
asymmetric channels (Chan and Adamson, 1978, [9]), and large-amplitude motion
(Adamson, Messiter and Liou, 1978, [10]). Later, a comparison was presented of the
shock wave motion due to the fluctuating back pressure and the boundaries’ movement
(Messiter and Adamson, 1984, [11]).
With the help of the small disturbance theory Rizetta and Chin (1979, [12])
investigated the effect of frequency on transonic flow over the NACA 64A010 airfoil
for three different types of motion. Bölcs, Fransson and Platzer (1989, [13]) utilized
full Euler equations to study numerically the large-amplitude motion of shock waves
in a quasi one-dimensional nozzle and in a two-dimensional channel with a sinusoidal
bump. Their investigations were later extended to an experimental and numerical
comparison of the shock motion for a fixed geometry of the nozzle, employing various
measuring techniques and a solution of the Euler equations (Ott, 1992, [14], and Ott,
Bölcs and Fransson, 1995, [15]).
Continuously increasing accessible computer power enabled the use of mass-
weighted Reynolds-averaged Navier-Stokes equations with proper turbulence models.
Extensive experimental and numerical investigations of natural and forced oscillations
of shock waves were carried out. A two-dimensional and fully turbulent, unsteady and
separated transonic flow over a thick circular arc foil was studied and computed with
help of an algebraic turbulence model by Seegmiller, Marvin and Levy (1978, [16]) and
compared with experimental data. Chen, Sajben and Kroutil (1979, [17]) measured

tq109m-e/117 11 V 2005 BOP s.c., http://www.bop.com.pl

118 P. Doerffer et al.

the natural oscillations of a shock wave in separated diffuser flow, while Liou and
Coakley (1984, [18]) compared measured self-excited and forced fluctuations of weak
and strong shock waves with numerical computations based on thin-layer Navier-
Stokes equations with the low-Reynolds k-ω 2 two-equation turbulence model. Shock
buffeting over bicircular airfoil was computed and compared with experiment by
Arnone, Liou and Povinelli (1995, [19]), on the basis of Navier-Stokes equations with
an algebraic Baldwin and Lomax turbulence model. Gerolymos, Vallet, Bölcs and Ott
(1996, [20]) investigated the influence of three-dimensional effects on the transonic
flow in a nozzle by means of a k-ε turbulence closure. More recently, Bron, Ferrand
and Fransson (2002, [21]) explored non-linear interactions in a two-dimensional nozzle
configuration using the Wilcox k-ω turbulence model. Very interesting results of the
Reynolds number’s influence on the unsteady flow through a critical nozzle have been
presented by Setoguchi, Matsuo and Kim [22].
All previous work concerning forced shock wave oscillations in nozzles underline
the fact that the amplitude of the shock movement depends on both the amplitude
and frequency of imposed back pressure changes. Moreover, the shock movement
amplitude decreases with increasing excitation frequency. An average unsteady shock
position for oscillating back pressure does not agree with the steady state position.
An unsteady flow field can be very different from a steady flow field.

3. The experimental arrangement and test conditions

The normal shock wave was located in the divergent part of the nozzle defined
by the adequate pressure level at the outlet. The upper wall of the divergent part
was convex, unlike the bottom wall, which was flat (see Figure 1). The Mach
numbers upstream of the shock were between 1.2 and 1.4, depending on the shock’s
position.

Figure 1. The test section

Glass windows at the divergent section of the nozzle allowed for the use of the
schlieren visualization method. Fifty seven pressure taps were uniformly spaced at
5mm intervals within the area of the shock wave motion. For ten positions of the shock
wave schlieren images were taken and static pressure distributions were measured.
Total pressure and temperature at the inlet were related to ambient conditions.
Pressure at the outlet was set by the control nozzles to achieve the desired shock
wave position.

tq109m-e/118 11 V 2005 BOP s.c., http://www.bop.com.pl

 Unsteady Shock Wave –Turbulent Boundary Layer Interaction. . . 119

4. The numerical method of solution

The present investigation was carried out with a cell-centered block-structured
SPARC code parallelized with a Message Passing Interface (MPI). This code solves
numerically compressible, mass-weighted, Reynolds-averaged Navier-Stokes equations
in generalized coordinates with several turbulence models. The algorithm uses a semi-
discrete approach, utilizing a finite-volume, density-based formulation for spatial
discretization and either an explicit Runge-Kutta type or an implicit LU-SSOR [23]
method for integration in time.
In order to accelerate convergence, the local time stepping and the implicit
residual averaging techniques are included in the explicit approach. Additionally, the
full multigrid (FMG) strategy with V -cycles is used. To damp numerical oscillations,
the artificial dissipation models is used, both the original blending of the 2nd and
4th-order Jameson-Smith-Turkel switch and a SLIP scheme as proposed by Tatsumi
et al. [24, 25]. This leads to good convergence rates and shock wave resolution. In
our study a very efficient implicit dual-time-stepping scheme was chosen for the time-
accurate simulations, which allowed us to utilize all acceleration techniques developed
for explicit algorithms.
From the various turbulence closures implemented in the code, the two-
equation, low-Reynolds turbulence model of Craft-Launder-Suga was chosen in both
the linear and the non-linear adaptive k-τ formulation. The τ = kε turbulent time scale
was chosen instead of the dissipation rate of turbulent kinetic energy, ε, to simplify
the implementation of the boundary condition in the code (k = τ = 0 at the walls).
The linear model was sufficient for time-independent solutions, while a new adaptive
turbulence approach was used for the unsteady calculations, based on a non-linear
stress-strain relation, which adapts itself automatically to the spatial and temporal
resolution of the computational domain.
Second-order, central-difference spatial discretization and the third-order,
three-stage Runge-Kutta method (CFL = 2.2) with a SLIP scheme and four-grid multi-
grid proved to be sufficient for steady state predictions. For a test case with fluctuat-
ing back-pressure, we utilized a switch scheme together with a second order, implicit,
dual-time-stepping formulation with the same Runge-Kutta method.
The four-block, body-fitted, structured computational grid was prepared with
the help of algebraic methods of orthogonal mesh generation and additional elliptical
smoothing (Figure 2). In order to find a grid-independent solution, three meshes with
different numbers of computational cells were considered.
The steady state computations were performed on coarse (4800), intermediate
(20 000) and fine (78 000) grids. The difference between solutions obtained for the
intermediate and the fine mesh was negligible. The intermediate mesh proved to be
adequate for unsteady cases and more economical as far as computational resources
were concerned.
The considered grids were very smooth and regular in the region of the
shock wave motion. Computational meshes were reconstructed after the initial tests
in accordance with the boundary layer development in the channel. For the grid-
independent solution, it was necessary for the non-dimensional distance of the first
point adjacent to the wall to be of the order of y + = 1 (so that it would lie deep in the

tq109m-e/119 11 V 2005 BOP s.c., http://www.bop.com.pl

120 P. Doerffer et al.

viscous sub-layer of the boundary layer). This value was of the order of unity for the
whole boundary layer, which was true for the fine as well as the intermediate mesh.
the inlet and outlet planes were situated more upstream and downstream compared
to the experimental test section in order to assure mesh orthogonality and improved
numerical accuracy in this region. Moreover, it prevented flow recirculation at the
outlet plane and led to better convergence of the numerical solution. The considered
grids were very smooth and regular in the region of shock wave motion.
Total pressure, total temperature and an imposed flow direction with isentropic
relations were used to obtain all information at the inlet. Measurements of the
characteristic length scale of turbulence and turbulence level at the inlet boundary
were not available. A turbulence level of Tu = 3% and a physically consistent value
of the ratio of eddy viscosity to dynamic viscosity of µt /µ = 10 were assumed on the
basis of previous investigations. We set the total temperature at T0 = 307K and the
total pressure at P0 = 100 432Pa for all the unsteady cases. Values measured at the
inlet were utilized for the steady state calculations.
Values of constant outlet static pressure for the steady cases were chosen to
obtain the same shock wave positions as in the experiment. Harmonic oscillations
were prescribed for the unsteady cases according to the following function: p(t) =
p̄ + Ap /2 sin(2π f t), where p̄ is the average unsteady pressure, Ap – a constant
amplitude, f – frequency, and t – time. The amplitude of pressure changes was taken
to be Ap = 5kPa, with an unsteady average exit pressure of p̄ = 70 273Pa. This method
of imposing a time-dependent boundary condition for the Runge-Kutta method time
advancement reduced the formal accuracy of the space-time method to the first order
locally and the second order globally, independently of the spatial operator [26].
The computed frequencies and the respective time steps are summarized in
Table 1. There were from 16 to a maximum of 125 time steps (in the dual-time-
stepping approach) per one period of oscillations.
The walls were defined with the no-slip of velocity and adiabatic (q = 0)
conditions; k = τ = 0 was set at the wall for the turbulent quantities.

Figure 2. The nozzle geometry and the intermediate mesh

tq109m-e/120 11 V 2005 BOP s.c., http://www.bop.com.pl

 Unsteady Shock Wave –Turbulent Boundary Layer Interaction. . . 121
Table 1. The unsteady flow parameters

Frequency Time step Amplitude Number of time

f [Hz] ∆t [s] Ap [kPa] steps per period
0.5 2.0 · 10−2 5 100
8 1.0 · 10−3 5 125
64 2.5 · 10−4 5 63
128 5.0 · 10−4 5 16
256 2.5 · 10−4 5 16
512 1.25 · 10−4 5 16
1024 6.25 · 10−5 5 16

The upstream-extended channel length assured the development of boundary

layer (and k and τ profiles minimizing the influence of the inlet boundary condition.
The outlet plane, far downstream from the region of interest, minimized the non-
physical constant pressure assumption.
Computations in SPARC are usually started from a simple, homogeneous
distribution with constant flow parameters throughout the flow-field. For the time
dependent cases, the flow was initialized together with fluctuating pressure at the
outlet on all the considered grids. The fluid was modeled as a viscous perfect
gas. The specific heat ratio was γ = 1.4, while the gas constant was equal to
R = 287m2 s−2 K. The dynamic viscosity was dependent on the temperature only,
according to Sutherland’s law. Turbulent and laminar heat exchange conditions
were based on the assumption of constant Prandtl numbers, laminar Pr = 0.72 and
turbulent Prt = 0.9.
The performed tests proved a reduction of the residuals by four orders of
magnitude to be sufficient convergence criteria for steady and unsteady cases (on
all meshes in the multigrid approach and for all transport equations considered, see
Figures 3 and 4).

Figure 3. Convergence history for steady-state Figure 4. Convergence history for unsteady
computation on the coarse, intermediate and computation on the intermediate mesh,
fine meshes, Ma = 1.3 f = 256Hz

The convergence judgment was also based on global parameters such as lift, CL ,
and drag, Cd , coefficients (treated only as integral parameters, insignificant in this
case). Obtaining a constant value of Cd for the steady flow and the periodic variation
of Cd for the unsteady cases is shown in Figures 5 and 6.

tq109m-e/121 11 V 2005 BOP s.c., http://www.bop.com.pl

122 P. Doerffer et al.

Figure 5. Time history of the drag coefficient, Figure 6. Time history of the drag coefficient,
Cd , for steady-state computation, Ma = 1.3 Cd , for unsteady computation, f = 256Hz

The imposition of fluctuating back pressure on the coarser meshes in the full
multigrid approach shortened the transient phase and thus minimized the compu-
tational effort. Data continued to be collected for analysis for at least one period
after the transient phase. For the higher frequencies the time taken by the flow to
stabilize was comparable to the period of fluctuations. It was found that more inner
steps (in fictitious time) were needed when fewer physical time steps per one period
of oscillations were used. For quasi-steady (very low) frequencies, the transient phase
practically disappeared.

5. A comparison of experimental and numerical data

for steady flow
Figure 7 presents the measured and computed static pressure distributions
at the bottom wall of the nozzle for all ten positions of the shock wave (within
the range of 0.15m). The chosen locations are in the region of numerically sim-
ulated shock motion. A qualitative and quantitative comparison gives acceptable
coincidence, although small discrepancies can be observed. These are due to the
inability of the 2D numerics to capture the effect of boundary layer develop-
ment, which includes the side walls in the real channel. It was assumed that
the shock wave position should be the same for the calculations and the exper-
iment. This caused different values of static pressure at the outlet in relation to
the wind tunnel measurements. Moreover, the artificial dissipation model, SLIP,
lacked weak boundary layer separation at the upper wall of the nozzle, which was
present in the experiment. The visible differences in the pressure minima upstream
of the shock are directly linked to the difference in the maximum Mach number
at about ∆Ma = 0.03.
The static pressure distribution in the middle of the channel presented in
Figure 8 corresponds to Figure 7. It is worth mentioning that a region of local
acceleration (expansion) of the flow exists just downstream of the shock wave, which
is marked by a circle in Figure 8. This effect is due to the thickening of the boundary
layers downstream of the region of shock-boundary layer interaction.
Figures 9, 10 and 11 present an example of flow with a Mach number of
Ma = 1.34 upstream of the shock. A contour plot of the Mach number is shown
in Figure 9. The experimental schlieren image in Figure 10 is compared with its

tq109m-e/122 11 V 2005 BOP s.c., http://www.bop.com.pl

 Unsteady Shock Wave –Turbulent Boundary Layer Interaction. . . 123

Figure 7. Static pressure distribution at the wall Figure 8. Static pressure distribution
of the nozzle in the middle of the nozzle

Figure 9. Isolines of Mach number, Ma = 1.34

Figure 10. Experimental schlieren picture, Figure 11. Numerical schlieren picture,
Ma = 1.34 Ma = 1.34

tq109m-e/123 11 V 2005 BOP s.c., http://www.bop.com.pl

124 P. Doerffer et al.

numerical simulation in Figure 11. The coincidence of the obtained results is very
good and positively verifies the numerical methods used.

6. Results of the unsteady calculations

The oscillations of back pressure at the outlet of the nozzle caused shock wave
movement in the range of 0.1m, with shock upstream maximum Mach numbers rang-
ing from Ma = 1.18 to Ma = 1.35. Figure 12 presents the static pressure distribution
in the middle of the channel for three steady-state cases with pressures at the outlet
given as p(t) = p̄ + Ap /2, p(t) = p̄ and p(t) = p̄ − Ap /2.

Figure 12. Static pressure distributions in the middle of the nozzle for average
and limiting shock positions

The time-dependent shock wave position, based on the pressure minimum in the
middle of the channel and at the wall, for at least one period of oscillations, is presented
in Figure 13 for the frequencies of 0.5, 8, 64, 128, and 256Hz. Apparently, the shock
wave oscillates harmonically with the frequency of imposed pressure fluctuation, with
an amplitude decreasing with increasing frequency. The mean position of the shock
wave does not correspond to the steady state solution with Pext = p̄ and depends on
the frequency as well.
The relative difference in shock wave position between the middle of the channel
and the wall, due to the shock-boundary layer interaction, is changing during one
period of oscillations, although no phase lag between these two motions is observed.
The amplitude of the shock motion at the wall is smaller than in the middle
of the channel and this trend decreases with increasing frequency. The λ-foot for
Ma = 1.35 is much greater than for Ma = 1.18: the higher the Mach number, the
larger the distance between the main shock and the beginning of interaction at
the wall. For low frequencies, the shock motion sweeps the whole range of shock
location and the increase of the λ-foot plays a significant role. At high frequencies of
excitation the forced shock movement is small; hence the size of the λ-foot does not
change much.
The most important conclusion is that for intermediate frequencies (from 50Hz
to 512Hz) the amplitude of the shock wave motion is inversely proportional to
the imposed frequency (Figure 14), which is in accordance with the literature.
A local maximum of amplitude has been found in the region of very low frequencies

tq109m-e/124 11 V 2005 BOP s.c., http://www.bop.com.pl

 Unsteady Shock Wave –Turbulent Boundary Layer Interaction. . . 125

Figure 13. Instantaneous shock wave position in the middle of the channel (solid line)
and at the wall (dashed line)

Figure 14. Relation between imposed frequency and amplitude of the shock wave motion in the
middle of the channel (solid line) and at the wall (dashed line)

tq109m-e/125 11 V 2005 BOP s.c., http://www.bop.com.pl

126 P. Doerffer et al.

Figure 15. Isolines of the Mach number, f = 8Hz

(f < 50Hz). Such behavior was previously found during measurements with a similar
configuration and flow parameters by only one of the authors (Liou et al. [18]). The
confirmation of these results suggests the necessity of further investigation into this
phenomenon.
Our results have also shown that for calculations with frequencies greater than
512Hz the position of the shock wave are unaffected by the pressure changes at the
outlet (within the accuracy of mesh resolution). Nevertheless, natural oscillation of
low frequency occurs.
The case with shock wave motion of the largest amplitude and the frequency
of 8Hz is presented in Figure 15 for illustration. Time-dependent isolines of the
Mach number in the region of shock wave movement are shown together with the
corresponding static pressure distribution in the middle of the channel (Figure 16).
The Mach number varies from Ma = 1.18 (the most upstream shock) to Ma = 1.35
(the most downstream shock). The arrows in Figure 15 indicate the direction of shock
movement.
Instantaneous static pressure signals for four representative points in the middle
and at the wall of the nozzle are presented in Figure 17. The first point remains in

tq109m-e/126 11 V 2005 BOP s.c., http://www.bop.com.pl

 Unsteady Shock Wave –Turbulent Boundary Layer Interaction. . . 127

Figure 16. Pressure distribution in the middle of the nozzle, f = 8Hz

the supersonic region (X = 0), where the pressure signal is constant at all times due
to the disturbances’ inability to propagate upstream of the shock wave.
The second characteristic point corresponds to the x position of the shock wave
taken from the steady state computation with exit pressure Pext = p̄; X = 0.15m for
the middle of the channel and X = 0.13m for the wall. The shock wave always passes
this point during a period of oscillation and produces a strongly non-harmonic signal.
It is especially clearly visible for the middle of the channel, where a sudden pressure
jump is connected with the shock’s passage. The effect is much weaker at the wall.
The next point is at X = 0.3m, which remains in the subsonic region down-
stream of the shock. The shock wave never reaches or passes this point. At this point,
pressure is only subject to mild, periodic changes of amplitude decreasing with in-
creasing frequency.
The last monitored point is at X = 0.9m, which corresponds to the outlet of
the nozzle. The amplitude and frequency of pressure changes at this plane are strictly
connected with the boundary condition imposed. Stagnation pressure downstream the
oscillating shock wave changes in accordance with the strength of the shock wave.

tq109m-e/127 11 V 2005 BOP s.c., http://www.bop.com.pl

128 P. Doerffer et al.

Figure 17. Pressure signals in the middle (left) and at the wall (right) of the nozzle:
x = 0.0; x = 0.15 (left) and x = 0.13 (right); x = 0.3; x = 0.9

A comparison of the oscillation phase at the outlet of the nozzle and at any
distant point reveals the lag due to the finite speed of disturbance propagation. It is
especially evident for higher frequencies, when the characteristic time of oscillation is
short compared with the time taken by the pressure disturbance to reach the shock.
Globally, the amplitude of the pressure signal decreases with increasing frequency.
Moreover, it can be observed that the signal at the wall decreases to much greater
extent than in the middle of the channel. This is due to the shock-boundary layer
interaction and the movement of the local acceleration region just downstream
of the shock.

tq109m-e/128 11 V 2005 BOP s.c., http://www.bop.com.pl

 Unsteady Shock Wave –Turbulent Boundary Layer Interaction. . . 129

Figure 18. Pressure in the middle Figure 19. Pressure at the wall
of the nozzle, steady state and the unsteady of the nozzle, steady state and the unsteady
case with 8Hz case with 8Hz

In Figures 18 and 19 pressure distributions are presented for the steady case
(dashed line) and the 8Hz unsteady case (solid line), where the shock wave is
at exactly the same position while moving upstream or downstream the channel.
Although the shock wave is at the same location in the nozzle for all cases, the steady
state computation differs from the unsteady one. Unsteady pressure downstream of
the shock bears little resemblance to the steady state distribution. The potential
appearance of shock wave oscillations on an aircraft’s wing may introduce large
unsteady loads very different from the predicted steady-state values.

7. The flow at high frequencies of excitation

When pressure variation at the outlet is of a high frequency, the shock wave
ceases to respond to forced excitations. In such cases it has been observed that
a new natural oscillation of the shock appears. This oscillation is of considerably
lower frequency and insensitive to forced oscillation. The observed shock oscillation
is connected with the formation and shedding of vortices from the separation zone
downstream of the shock wave. The effect is illustrated in Figures 20 and 21, which
show pressure changes in time at four locations at the wall with and without forced
excitation, respectively. The time interval shown in both plots is the same.

Figure 20. Pressure signals at the nozzle’s Figure 21. Pressure signals at the nozzle’s
wall without forced oscillation wall for the 512Hz case

tq109m-e/129 11 V 2005 BOP s.c., http://www.bop.com.pl

130 P. Doerffer et al.

The topmost curve (1) in both diagrams shows the exit pressure (the highest
value). It is periodic in Figure 21 and constant in Figure 20. The immediately lower
curve (2) shows the pressure change at the location where vortices are formed and
shed. The character of these two curves is very similar in both of the figures. Without
excitation (Figure 20) the vortices are stronger and exhibit more pronounced pressure
extrema. It may be concluded that pressure excitation at the outlet diminishes the
strength of the shed vortices. The next lower curve (3) shows the pressure under the
front branch of the λ-foot. In fact, it indicates the movement of the shock wave. Outlet
pressure excitation makes the distribution line in Figure 21 faintly wavy. The line in
Figure 20 has the same amplitude and frequency as the corresponding line in Figure 21
but it is smooth, as there is no excitation. The bottom curve (4) shows the pressure
upstream of the shock wave and does not exhibit any effect of the investigated process
because it is located in the supersonic part of the flow.
Disappearance of the pressure fluctuations’ effect on shock oscillations does not
mean that a quasi-steady flow may be obtained. Natural oscillation of the shock wave
remains independently from the excitation.
Along the lower wall downstream the shock, the streamlines deviate from the
wall making room for a separation (Figure 22). There are local pressure minima at
some distance from the shock, which exhibit significant variations in time. These
pressure minima indicate the formation of vortices in the separated flow area and
prove that this is an unsteady process.

8. Conclusions
The present investigation has provided an insight into the nature of transonic
flow in a two-dimensional Laval nozzle for steady and unsteady flow conditions. The
general features of such flows were studied by means of the SPARC computer code,
which solves the time-dependent, Reynolds-averaged, compressible Navier-Stokes
equations with a linear and an adaptive, non-linear, k-τ turbulence closure.
The comparison of numerical predictions and the experiment was satisfactory
in the case of steady flows as the Mach numbers were low (1.2–1.4). However, the
unavoidable presence of three-dimensionality in the experimental flow decreased the
applicability of two-dimensional simulations.
In the unsteady flow cases, the variation of the shock’s position and structure as
well as the behavior of the flow upstream and downstream of the shock were analyzed
numerically for various excitation frequencies ranging from 0Hz to 1024Hz. Compu-
tations of the flow predicted most of the essential features observed experimentally
by other researchers for similar geometrical configurations and flow conditions.
All the observations revealed differences in the shock wave behavior between
the bottom wall and the middle of the channel. The amplitude of the shock wave
movement decreased with increasing excitation frequency for both locations. A closer
study of low frequencies (f < 50Hz) revealed the maximum amplitude of the shock
wave movement. For moderate frequencies, the amplitude of the shock motion was
inversely proportional to the forcing oscillations.
At 500Hz, the forced shock wave displacement disappeared. Natural oscillations
were observed for higher excitations.

tq109m-e/130 11 V 2005 BOP s.c., http://www.bop.com.pl

 Unsteady Shock Wave –Turbulent Boundary Layer Interaction. . . 131

Figure 22. Static pressure maps and streamlines: natural vortex shedding in the separation zone

The unsteady flow appeared to be very different from the steady-state flow
field. The pressure jump and the unsteady pressure downstream of the shock were
different for identical shock positions during one cycle of fluctuations.
Further experimental and numerical investigations are necessary to validate
the local amplitude maximum in the low frequency range. Taking into account the
three-dimensional effects of the real flow may significantly improve coincidence of the
results.

tq109m-e/131 11 V 2005 BOP s.c., http://www.bop.com.pl

132 P. Doerffer et al.

References
[1] Szulc O 2001 Unsteady Transonic Flow in the Laval Nozzle, MSc Thesis, Faculty of Technical
Physics and Applied Mathematics, Technical University of Gdansk, Poland (in Polish)
[2] Magagnato F 1998 TASK Quart. 2 (2) 215
[3] Magagnato F 2000 Proc. 8 th Int. Symp. on Transport Phenomena and Dynamics of Rotating
Machinery, Honolulu, Hawaii, C557/139/99
[4] Magagnato F 2001 Proc. CFD for Turbomachinery Applications, Gdansk, Poland, CFDTurbo-
2001-K05
[5] Craft T J, Launder B E and Suga K 1993 Proc. 5 th IAHR Conf. on Refined-Flow Modeling
and Turbulence Measurement, Paris, France
[6] Adamson T C 1972 J. Fluid Mech. 52 (3) 437
[7] Adamson T C and Richey G K 1973 J. Fluid Mech. 60 (2) 363
[8] Richey G K and Adamson T C 1976 AIAA Journal 14 (8) 1054
[9] Chan J S-K and Adamson T C 1978 AIAA Journal 16 (4) 377
[10] Adamson T C Jr, Messiter A F and Liou M S 1978 AIAA Journal 16 (12) 1240
[11] Messiter A F and Adamson T C Jr 1984 AIAA Journal 22 (11) 1590
[12] Rizetta D P and Chin W C 1979 AIAA Journal 17 (7) 779
[13] Bölcs A, Fransson T H and Platzer M F 1989 J. Turbomachinery 111 (July) 169
[14] Ott P 1992 Oszillierender senkrechter verdichtungsstoss in einer ebenen düse, Lausanne EPFL
[15] Ott P, Bölcs A and Fransson T H 1995 J. Turbomachinery 117 (January) 106
[16] Seegmiller H L, Marvin J G and Levy L L Jr 1978 AIAA Journal 16 (12) 1262
[17] Chen C P, Sajben M and Kroutil J C 1979 AIAA Journal 17 (10) 1076
[18] Liou M S and Coakley T J 1984 AIAA Journal 22 (8) 1139
[19] Arnone A, Liou M S and Povinelli L A 1995 AIAA Journal 33 (6) 985
[20] Gerolymos G A, Vallet I, Bölcs A and Ott P 1996 AIAA Journal 34 (7) 1331
[21] Bron O, Ferrand P and Fransson T 2002 Proc. 9 th Int. Symp. on Transport Phenomena and
Dynamics of Rotating Machinery, Honolulu, Hawaii, pp. 1–9
[22] Setoguchi T, Matsuo S and Kim H D 2003 Proc. Symp. on Shock Waves, Japan, pp. 49–52
[23] Rieger H and Jameson A 1987 AIAA paper 87-0619 1
[24] Tatsumi S, Martinelli L and Jameson A 1995 AIAA Journal 33 (2) 252
[25] Swanson R C, Radespiel R and Turkel E 1997 NASA, CR-201726 ICASE Report No. 97-40
[26] Carpenter M H, Gottlieb D, Abarbanel S and Don W S 1993 NASA, CR-191561 ICASE
Report No. 93-83

tq109m-e/132 11 V 2005 BOP s.c., http://www.bop.com.pl