ORIGINAL RESEARCH

published: 07 March 2022

doi: 10.3389/fenrg.2022.854019

Unsteady Motion of Shock Wave for a

Supersonic Compression Ramp Flow
Based on Large Eddy Simulation
Xing Huang 1, Li-Xu Wang 2, Dong-Dong Zhong 2 and Ning Ge 2*
1
AECC HUNAN Aviation Powerplant Research Institute, Zhuzhou, China, 2Nanjing University of Aeronautics and Astronautics,
Nanjing, China

A large eddy simulation (LES) is conducted to investigate shock wave/turbulent boundary

layer interaction in a 24° compression ramp at a high inlet Mach number of Ma
2.9. The
recycling/rescaling Method is used as the inflow turbulence generation technique. The
shock wave structure and boundary layer flow in the interaction region are studied by flow
visualization methods, such as vortex recognition and numerical schlieren. The
distributions of turbulent kinetic energy and Reynolds normal stresses at different
streamwise locations are analyzed. The results show that a strong anisotropy turbulent
flow appears in the reattached boundary layer after the interaction of the shock wave. The
large-scale unsteady motion of the separation shock wave is analyzed by using the
intermittent factor and power spectrum. It is found that the shock moves around the
Edited by: averaged separation position, and the length scale is equal to 72% of the inlet boundary
Lei Luo,
Harbin Institute of Technology, China
layer thickness. The power spectrum analysis reveals the existence of low-frequency
Reviewed by:
instability in the separation region.
Fangfei Ning,
Keywords: shock wave/turbulent boundary layer interaction, large eddy simulation (LES), low-frequency instability,
Beihang University, China
Görtler-like vortex, power spectrum analysis (PSA)
Hu Wu,
Northwestern Polytechnical
University, China
*Correspondence:
1 INTRODUCTION
Ning Ge
[email protected] The shock wave/turbulent boundary layer interaction (SWBLI) is one of the important physical
phenomena in the supersonic flow, which contains complex aerodynamic and thermodynamic
Specialty section: problems. The interaction can significantly alter the heat conduction characteristics and produce a
This article was submitted to strong pressure fluctuation. Although the 2D compression ramp is a simple geometric model in the
Advanced Clean Fuel Technologies, SWBLI, however the flow phenomena contain boundary layer unsteady flow, separation, reattached
a section of the journal flow, and turbulent fluctuation enhancement caused by a strong adverse pressure gradient. The
Frontiers in Energy Research transient flow field presents a highly three-dimensional state (Lee and Wang 1995), and the
Received: 13 January 2022 multiscale interaction produces complex flow structures (Wang 2015). Thus, the simple 2D
Accepted: 10 February 2022 model is the best model for the validation of the computational fluid dynamics (CFD) method.
Published: 07 March 2022
In the past, extensive research studies have been conducted on the SWBLI by experimental and
Citation: numerical methods, but there are still some issues to be explored further. Dolling (Dolling 2001)
Huang X, Wang L-X, Zhong D-D and considered that the low-frequency unsteady flow should be studied for the future in detail. Settles
Ge N (2022) Unsteady Motion of Shock
et al. (Settles et al., 1979) conducted a surface oil flow experiment on a compression ramp with an
Wave for a Supersonic Compression
Ramp Flow Based on Large
angle of 24°. It was found that the pattern of the “node-saddle point” is alternately arranged near the
Eddy Simulation. reattachment line downstream of the ramp, which proves the existence of the streamwise vortex.
Front. Energy Res. 10:854019. Smits and Muck (Smits and Muck 1987) used a hot wire anemometer to measure the compression
doi: 10.3389/fenrg.2022.854019 ramp flow and studied the effects of turbulent fluctuation enhancement with three different angles of

Frontiers in Energy Research | www.frontiersin.org 1 March 2022 | Volume 10 | Article 854019

Huang et al. Shock Wave–Induced Flow Separation

the ramp. It was found that when the intensity of the shock wave
is sufficiently strong, the unsteady motion of the shock wave will
become important due to turbulent fluctuation enhancement.
Ganapathisubramani et al. (Ganapathisubramani et al., 2007)
used high-speed particle image velocimetry (PIV) technology to
measure the compression ramp flow. It was found that the strip-
like structures with different momentum exist in the upstream
boundary layer, and these strip structures cause a low-frequency
pulsation of the separation bubble. Wu et al. (Wu et al., 2013)
used nano-tracer planar laser scattering (NPLS) technology to
study the laminar/turbulent SWBLI on the supersonic
compression ramp; the overall flow field was analyzed and
local fine structures were identified. FIGURE 1 | Computational domain for the LES.
Yi Zhuang et al. (Zhuang et al., 2018a) have performed an
experimental investigation on a compression ramp shock wave/
turbulent boundary layer interaction at Ma
2.83. The ice dominant frequency in the vicinity of the shock was exhibited
cluster–based planar laser scattering technique was applied to by the streamwise evolution of the pre-multiplied spectrum of
acquire high spatiotemporal resolution images at the center plane. pressure fluctuations. In the upstream boundary layer, the
Two-dimensional slices of the coherent vertical structure (CVS) spectrum presents mainly high-frequency content (St > 1)
were acquired and extracted from these images with a machine linked to the incoming turbulent eddies. However, in the
learning–based method. By comparing the features CVS acquired vicinity of the shock position, a low-frequency broadband
before and after the interaction, the evolution of CVSs in the range emerges and is centered approximately at St
0.03. As
SWBLI flow was analyzed. the author pointed out, the large gap between these frequency
In fact, a large number of experimental studies were carried scales has been reported in previous investigations by Touber and
out at higher Reynolds numbers. But Bookey et al. (Bookey et al., Sandham (Touber and Sandham 2011) and Dupont et al.
2005) selected a low-density gas as the working medium to reduce (Dupont et al., 2006).
the Reynolds number in the experiment and achieve the In this study, a large eddy simulation is used to study the
numerical simulation (DNS, LES) comparison with SWBLI phenomenon of the 24° compression ramp; detailed
experiments. Adams et al. (Adams 2000) used the DNS analysis and discussion are carried out to understand the
method to study the supersonic compression ramp flow and unsteady features of the shock wave in the SWBLI. The
found that the Reynolds shear stress after the SWBLI increases remainder of this manuscript is organized as follows. Section
more greatly than the Reynolds normal stress. Loginov et al. 2 gives a description of the numerical methods and techniques
(Loginov et al., 2006) used the LES method to study the statistical used in this study, including the generation of the inlet
parameter distributions and fluctuation characteristics of the turbulent boundary layer, geometric model, and mesh
turbulent boundary layer in the compression ramp and distribution. Section 3 compares the simulation results with
compared numerical results with the experimental data. The experiments and numerical results in the literature to verify the
flow field analysis showed that the large-scale three- reliability of the program, and the details of the flow field are
dimensional flow structure downstream of the ramp is the discussed and analyzed. The last section gives some conclusions
principal reason for the spanwise unevenness of the flow field. of this study.
Wu et al. (Wu and Martin 2008) used the DNS method to study
the SWBLI in the compression ramp, with the inlet flow
parameters consistent with the experimental conditions given 2 COMPRESSION RAMP AND NUMERICAL
by Bookey et al. (Bookey et al., 2005). By observing the METHOD
spatiotemporal evolution and using the correlation analysis of
the flow field, it was found that the shock wave motion always lags The work in this study is based on the NUAA-Turbo CFD solver
behind the pulsation of the separation bubble. Therefore, a developed by our research group. In the large eddy simulation
“feedback loop” model consisting of the separation bubble, (LES), the finite volume method is used and the ROE scheme is
shear layer, and shock wave system is proposed to explain the used for the evaluation of convective fluxes, the WENO_ZQ
mechanism of the low-frequency unsteady motion of the scheme (Zhu and Qiu 2017) with fifth-order precision for the
shock wave. interface reconstruction, and the sixth-order central difference
Kenzo S.et al. (Sasaki et al., 2021) have investigated the scheme for the spatial discretization of viscous fluxes. Time
mechanisms of low-frequency unsteadiness in an impinging discretization uses the Runge–Kutta method with a total
shock wave/turbulent boundary layer interaction at a Mach variation reduction property of third-order accuracy (Shu and
number of Ma
2. The Strouhal number St
fL/U∞ was Osher 1989). The dynamic sub-grid scale model is considered,
used for the space-time spectral analysis to identify the key and the sub-grid scale viscosity coefficient is determined with the
features of the shock motion, where L is defined as the two consecutive filtering by the method of Germano et al.
interaction length and f is the frequency of fluctuation. The (Germano 1991).

Frontiers in Energy Research | www.frontiersin.org 2 March 2022 | Volume 10 | Article 854019

Huang et al. Shock Wave–Induced Flow Separation

FIGURE 2 | Distribution of averaged wall pressure and skin friction coefficient. (A) Averaged wall pressure. (B) Averaged skin friction coefficient.

FIGURE 4 | Vortex structure in the interaction region.

FIGURE 3 | Shock surface wrinkling.

near the wall. The flow field of the recycled plane in the auxiliary
computation domain is extracted as the inlet boundary condition
The inlet boundary layer thickness δ of the compression ramp for the primary computation. It is specially noted that when the
is used to make the length scale dimensionless. The computation primary/auxiliary computation domain uses grids of different
domain consists of two parts: the flat plane computation domain resolutions, the process of flow field extraction needs to
(auxiliary computation domain) and the 24° compression ramp interpolate the variables. For details, this method can refer to
computation domain (primary computation domain). The Zhong et al., (2021).
“recycling/rescaling” method was used to generate the For the primary computation domain, the upper boundary of
turbulent boundary layer in the auxiliary computation domain the computational domain and the outlet are set to the subsonic
and the computed turbulence information as the inlet boundary outlet boundary condition. The wall condition of the non-slip
condition of the primary computation domain. In the auxiliary isothermal is adopted to the wall, and the wall isothermal
computation, the distance from the inlet plane “7.3δ” is set as the temperature is 307 K. The spanwise boundary adopts periodic
recycled plane. The schematic of the computation domain is boundary condition, and the inlet turbulent boundary conditions
shown in Figure 1. The coordinate system origin is located at the are dynamically given by the auxiliary computation. For the
ramp, and the coordinate axes “x, y, and z” indicate the auxiliary computation, the boundary layer conditions are
streamwise, spanwise, and normal directions respectively. The consistent with the main computation, and the flow Mach
upstream and downstream lengths of the ramp are both 7.73δ, the number is Ma
2.9, the flow static temperature is 108.1 K,
spanwise width is 2.15δ, and the normal height along the wall is and the flow density is 0.074kg/m3 .
5.23δ. The number of grid points in the three directions is “505 × In order to verify the reliability of the LES software, the
89 × 112”. The grid is evenly distributed in the spanwise and numerical results of the compression ramp will be compared
refined along the streamwise direction at the ramp and in the with the experimental results under the same inlet flow
normal direction to guarantee z+ ≈ 1 in the first layer of the grid conditions.

Frontiers in Energy Research | www.frontiersin.org 3 March 2022 | Volume 10 | Article 854019

Huang et al. Shock Wave–Induced Flow Separation

spanwise directions show the numerical schlieren diagram, which

is used to show the shock wave, turbulent boundary layer, and
flow structure in the separation bubble. It can be seen that the
large-scale structure in the interaction region breaks into small-
scale structures, which is well agreed with Zhuang Y.et al.
(Zhuang et al., 2018b) experimental observation. But, on the
slope downstream of the ramp, a large-scale quasi-order structure
in the boundary layer is re-established. The instantaneous
streamwise direction velocity contour (y/δ
0.02, approach
laminar sublayer) is shown on the slice parallel to the wall. It
can be seen that a velocity strip structure alternately arranged
along the spanwise direction is observed both upstream and
downstream of the ramp in this region. In addition, the flow
scale in the separation bubble along the spanwise direction is very
FIGURE 5 | Contour of time-spanwise averaged turbulent kinetic different, indicating that the separation bubble is a multiscale flow
energy. structure.

3.1 Turbulent Fluctuation

3 RESULT ANALYSIS AND DISCUSSION When the flow passes through the interaction region, the
turbulent fluctuation is significantly enhanced due to the
Figure 2 shows the distribution of the averaged wall pressure and strong inverse gradient pressure in this region. Figure 5 is a
skin friction coefficient along the streamwise direction in the contour of the time-space averaged turbulent kinetic energy. The
primary computation domain, where the DNS results of Wu et al. definition of turbulent kinetic energy is as follows:
(Wu and Martin 2008) and Tong et al. (Tong et al., 2017) and the k
(ρu′u′ + ρv′v′ + ρw′w′)/2ρ∞ U2∞ ). It can be seen from the
experimental data of Bookey et al. (Bookey et al., 2005) are also figure that the intensity of the turbulent kinetic energy after
presented for the purpose of comparison. By solving the position the interaction region rapidly increases and reaches its peak value
where the averaged skin friction coefficient Cf is zero, the downstream of the ramp. The increase of the turbulent kinetic
averaged separation point xsep
−2.7δ and the re-attachment energy means significantly that a large number of small-scale
point xrea
0.8δ in the ramp flow are obtained. Overall, the structures are generated after the SWBLI.
computed pressure distribution and friction coefficient Figure 6 presents the variation of Reynolds normal stress Rii at
distribution are in good agreement with the experimental different locations in the interaction region. It can be seen that
values. At upstream of the ramp, it agrees well with the near the ramp (x
0.66δ), the three Reynolds stress components
experimental values and DNS results. But it is slightly higher produce a sharp peak value. As the separated boundary layer
downstream of the ramp. The reason is that the predicted reattaches, the Reynolds stress continues to decrease and tends to
separation region size is slightly smaller and the separated return to the pre-interaction state at (x
2.68δ). It is noted that
shear layer completes the reattachment process in advance the distribution of Reynolds stress R22 in the boundary layer is
downstream of the ramp; thus, the wall pressure increases relatively smooth and appears as a bimodal distribution near
rapidly and the range of the pressure platform is shortened. the ramp.
Figure 3 illustrates the instantaneous flow structure in the
compression ramp computation domain. Among them, the quasi-
order vortex structure of the turbulent boundary layer is displayed 3.2 Unsteady Motion of Shock Waves
using the Q criterion, and the streamwise direction velocity is used In order to study the shock wave unsteady motion, intermittent
for coloring. The translucent gray surface in the figure is the factors are usually used to measure the range of shock wave
dimensionless pressure isosurface p/(ρ∞ U2∞ )
1.7, which is used motion along the streamwise direction. The intermittent factor is
to represent the three-dimensional structure of the separation shock defined as the time when the instantaneous wall pressure
wave. As can be seen in the figure, the flow field in the interaction somewhere in the streamwise direction is greater than the
region has significant three-dimensional characteristics. When a given threshold takes up the proportion of the total flow field
large vortex structure passes through the root of the shock wave, time. The computation formula is as follows (Dolling and Or
it breaks under the effect of turbulent fluctuation enhancement at the 1985):
ramp. At the same time, the shock wave deforms at the root, and the
Time
pw > pw + 3σpw 
shock surface wrinkles along the spanwise direction. While away λ
,
Totaltime
from the interaction region, the shock wave keeps still the typical
two-dimensional structure characteristic. where pw represents the instantaneous wall pressure; pw and
In order to further illustrate the interaction process between σ(pw ) represent the average wall pressure of the inlet boundary
the shock wave and turbulent boundary layer, a slice contour in layer and the standard deviation of the wall pressure, respectively,
three directions is used to show the flow field details inside the and the sum of the two is set as the threshold for the computation
interaction region, as shown in Figure 4. The streamwise and of the intermittent factor.

Frontiers in Energy Research | www.frontiersin.org 4 March 2022 | Volume 10 | Article 854019

Huang et al. Shock Wave–Induced Flow Separation

FIGURE 6 | Distribution of Reynolds normal stresses at different streamwise locations. (A) Streamwise component. (B) Spanwise component. (C) Wall-normal
component.

FIGURE 8 | Contour of the time-averaged skin friction coefficient.

3.3 Görtler Flow Vortex

From the abovementioned analysis, the supersonic compression
ramp flow has significant three-dimensional characteristics.
Especially on the slope downstream of the ramp, the flow field
shows a strong unevenness in the spanwise direction. This
FIGURE 7 | Intermittency factor of the separation shock. phenomenon is closely related to the presence of Görtler flow
vortex (Dolling and Or 1985; Grilli et al., 2013).
Figure 8 shows a time-averaged skin friction coefficient contour.
Figure 7 exhibits the local distribution of the intermittent The white dashed line represents the location of the ramp. The solid
factor in the compression ramp. It is known from the definition of black line in the figure is used to show the separation and
the intermittent factor that in the undisturbed upstream reattachment positions of the boundary layer near the ramp
boundary layer and downstream region of the ramp, the defined by the time-averaged friction coefficient equal to 0. It can
intermittent factor is always equal to 0 (or 1). However, in the be observed that the skin friction coefficient of the inlet boundary
vicinity of the average separation point, the separation shock layer represents a strip pattern alternately arranged which is
wave shows a strong unsteady flow characteristic, which is consistent with the spanwise distribution of the time-averaged
represented by the intermittent factor: 0 < λ < 1. Wu et al. (Wu separation line. This indicates that the distribution of the
and Martin 2008) considered that the motion of the shock wave is separation line is, in a great measure, determined by the skin
closely related to the pulsation of the separation bubble. For friction coefficient of the inlet boundary layer. Due to the low-
comparison, the upper and lower limitations of the given frequency pulsation of the separation bubble, the average
intermittent factor are 98 and 4%, respectively, to estimate the reattachment still exhibits strong transient flow characteristics.
streamwise direction range of the shock wave motion as follows: When the separated boundary layer reattaches, the skin friction
−3.2δ ≤ x ≤ − 2.48δ . It can be seen that the flow movement of the coefficient distribution appears as an obvious “V” structure, which is
separation shock wave in the intermittent region is carried out caused by the presence of Görtler vortex pairs. The existence of this
around the average separation point(x
−2.7δ). By “V” structure was also confirmed by Fu-Lin et al., (2016).
computation, the length of the intermittent region in this The formation of Görtler vortices is related to the curvature of
simulation is Li
0.72δ which is slightly smaller than that in the concave surface. In order to describe the spatial development
the DNS result of the study by Tong et al. (Tong et al., of the Görtler vortex in detail, Figure 9 shows the flow field details
2017Li
0.72δ) (Li
0.72δ) on six different sections downstream of the average separation

Frontiers in Energy Research | www.frontiersin.org 5 March 2022 | Volume 10 | Article 854019

Huang et al. Shock Wave–Induced Flow Separation

FIGURE 9 | Spatial evolution of Görtler-like vortices in the compression ramp. (A) Streamwise position of the six planes downstream of the separation point. (B)
Distribution of streamwise vorticity and streamlines in (“y”,η) planes.

point. These six sections are named A ~ F, corresponding to the given in Figure 9A, in which the background is space-
streamwise direction positions: −1.7δ, −1.4δ, −0.6δ, 0.7δ, 3.3δ, time–averaged numerical schlieren. Figure 9B shows the
and 5.5δ, respectively. A schematic diagram of six positions is streamwise direction vorticity ωx contour, and streamline

Frontiers in Energy Research | www.frontiersin.org 6 March 2022 | Volume 10 | Article 854019

Huang et al. Shock Wave–Induced Flow Separation

Figure 10 shows the time-averaged skin friction coefficient along

the span in different sections, including the inlet section of the
ramp computation domain and the A, C, D, E, and F sections in
Figure 9 so that the variation of the spanwise distribution of the
skin friction coefficient with the development of the flow field can
be presented. The dotted line in the figure indicates the time-
span–averaged result at this section. At the inlet section, it can be
found that the skin friction coefficient along the spanwise direction
is not evenly distributed, and a low skin friction coefficient region
appears near the spanwise center. Sections A, C, and D are all
located in the separation region, so the friction coefficients of these
three locations are close to 0. In addition, due to the shock wave/
boundary layer interaction and Görtler instability, the skin friction
coefficient of these three sections along the spanwise direction
shows strong fluctuation. Especially, an obvious low friction
coefficient region is formed between y/δ
1 ~ 1.5. In sections
E and F, the amplitude of this low-friction coefficient region is
further increased, and a V-shaped distribution is formed. The fluid
near the wall converges between the pair of Görtler vortexes and
moves toward the outer layer of the boundary layer, locally
resulting in a lower friction coefficient. In section F, the skin
friction coefficient at the spanwise position y
1.2δ reaches a
minimum, which is exactly the same as the center position of the
Görtler vortex pair. In fact, by observing the friction coefficient
distribution at the six different sections, it can be found that the
low-friction coefficient region of the inlet section has a high degree
of coincidence with the downstream section in the spanwise
position. This phenomenon indicates that the spanwise position
of the Görtler vortex downstream of the ramp is likely to be affected
by the spanwise friction coefficient distribution at the inlet section.

3.4 Low-Frequency Instability

In order to study the low-frequency motion of the separated shock
wave, it is necessary to collect the signal of numerical pressure for
spectrum analysis. Therefore, 528 wall pressure probes are arranged
along the flow direction in the middle section of the calculation
domain. The arrangement of these probes is consistent with the grid
FIGURE 10 | Distribution of the averaged skin friction coefficient in the distribution, that is, one probe is arranged in the center of each wall
spanwise direction. grid cell. It is worth noting that the signal acquisition process is
carried out after the SWBLI flow field is fully established. In order to
ensure sufficient time resolution, data recording is carried out every
distribution on six sections, where the vorticity contour has a ten time steps in the large eddy simulation, and the physical time
range −0.5 ≤ ωx ≤ 0.5. In sections A and B, the streamlines in the interval between two adjacent recording points is Δt
0.01δ/U∞ .
near wall region are gradually bent, and a smaller streamwise According to the Nyquist sampling theorem, the highest physical
vortex is formed. When the flow develops to section C, it can be frequency that can be captured is fs
1/(2Δt)
50U∞ /δ, which is
clearly seen in this figure that there is a large pair of vortexes in the far greater than the characteristic frequency U∞ /δ of the turbulent
streamwise direction (B) and a small pair of vortexes in the boundary layer. Because the low-frequency instability of the SWBLI
streamwise direction (A), and the entire boundary layer has a has a wide frequency range, in order to accurately capture the low-
large amount of vortexes in this direction. With the separated frequency characteristics, the signal acquisition time should cover at
boundary layer reattach in the sections D, E and F, the larger pair least one low-frequency instability period. Table 1 shows the
of the streamwise vortexes gradually occupies almost the entire relevant information of pressure signal acquisition in this large
span width (about 2δ) and appears as a classical kidney eddy eddy simulation.
structure. Figure 11 presents the variation of wall pressure with time at
Figure 9B illustrates that it can be estimated that the center of two different positions upstream of the corner (the corner
the Görtler vortex pair is located at y
1.2δ. In general, the position is x
0 ) in the center plane, and the wall pressure is
position of the Görtler vortex in the spanwise direction changes nondimensionlized by using the inlet pressure P∞ . The blue solid
constantly with time due to the turbulent flow (Floryan 1991), but line represents the pressure signal graph collected at the inlet

Frontiers in Energy Research | www.frontiersin.org 7 March 2022 | Volume 10 | Article 854019

Huang et al. Shock Wave–Induced Flow Separation

TABLE 1 | Relevant information of pressure signal acquisition.

Number of data Time span Sampling frequency/(U∞ /δ) Minimum frequency/(U∞ /δ)
/N /(δ/U∞ )

40,000 400 50 0.0025

FIGURE 11 | Wall pressure signal (A) and power spectrum (B).

turbulent boundary layer, where the pressure fluctuation is not weighted power spectrum density (WPSD) expressed as
disturbed by the downstream interference area flow. It can be WPSD(f)
fpPSD(f)df. From this figure, we can see more
seen that the pressure signal around Pw /P∞ presents a random clearly the frequency components of pressure signals at different flow
fluctuation, and the amplitude of fluctuation is very small. The direction positions. For the upstream boundary layer, the fluctuating
red solid line indicates the pressure fluctuation with time at the energy is mainly distributed between 100 ~ 101 , and the peak
average separation point (xsep
−2.7δ). Compared with the inlet, frequency is about St
3, which is consistent with the
the pressure signal at the average separation point has larger characteristic frequency of the fully developed turbulent boundary
amplitude of fluctuation due to the interaction between the shock layer. For the signal at the average separation point, there is also a
wave and turbulent boundary layer. In addition, from the red wide small peak area in the high-frequency region, which is similar
solid line, the pressure signal at the separation point not only to the WPSD curve obtained in the upstream turbulent boundary
includes the high-frequency fluctuation but also has the low- layer. At the same time, the red curve shows a more obvious energy
frequency fluctuation component as shown in Figure 11. peak appearing in the low-frequency region, and the corresponding
Next, we will make a strict quantitative analysis of the wall characteristic frequency is St
0.04, which is completely consistent
pressure fluctuation signal from the perspective of spectral analysis. with the result (0.02–0.05) summarized by Dussauge et al. (Dussauge
The Welch method (Barbe et al., 2010) is used to segment the et al., 2006). In addition, Wang Bo et al. (Wang 2015), Kenzo S. et al.
discrete pressure signal, aiming to obtain a smoother and less (Sasaki et al., 2021), Tong et al. (Tong et al., 2017), Touber and
variance power spectrum density (PSD) curve. In this study, the Sanham (Touber and Sandham 2009), and Pasquariello et al.
collected pressure signals are divided into three segments, and the (Pasquariello et al., 2017) obtained similar results in their
coincidence rate between the segments is 50%. The Hanning respective numerical simulations.
window is used to add windows for each segment to improve the In order to study the distribution of low-frequency instability
variance performance. It should be noted that in the Welch method, characteristics in the whole field of the compression corner,
the more segmented the data, the smoother the power spectrum Figure 13 shows the distribution of WPSD on all pressure
curve and the smaller the noise, but at the same time, the resolution probes. The white solid line is used to represent the position of
of the power spectrum will be affected. Therefore, in the process of the average separation point xsep and the average reattachment point
data segmentation, we must consider the balance of noise and xrea and the white dotted line is used to represent the position of the
resolution in the power spectrum curve. Figure 11B shows the corner. Once again, it can be clearly seen that the turbulence
PSD distribution curve corresponding to the abovementioned two generation technology in this study does not introduce any low-
wall pressure signals, where the abscissa is the dimensionless frequency energy at the inlet. By observing the whole flow field, it is
frequency proposed by Dussauge et al. (Dussauge et al., 2006), found that this low-frequency instability mainly exists near the
which is defined as St
fLsep /Ue . It can be seen from the figure that average separation point, while in other locations of the flow
the peak value of pressure fluctuation energy in the upstream field, the energy of pressure fluctuation almost exists in the high-
boundary layer is located in the high-frequency region, while the frequency range. Especially when the boundary layer is reattached in
peak value of pressure fluctuation energy at the average separation the downstream, its power spectrum distribution is almost the same
point is located in the low-frequency region. Figure 12 shows the as that in the upstream, and the reattached boundary layer shows a

Frontiers in Energy Research | www.frontiersin.org 8 March 2022 | Volume 10 | Article 854019

Huang et al. Shock Wave–Induced Flow Separation

FIGURE 12 | (A) Frequency weighted power spectrum distribution and (B) collected positions.

FIGURE 13 | (A) WPSD contour map of the wall pressure signal (B) peak map of wall pressure signal.

process of flow recovery. This study considers that this low- The LES simulation represents a pair of streamwise vortices,
frequency motion is caused by the unsteady motion of large-scale called Görtler vortex, occupying almost the entire span width
separation bubbles. Obviously, the scale of the separation bubble is about 2δ on the compression ramp. It is found that the spanwise
larger than that of any flow in the boundary layer. skin friction coefficient distribution at the inlet of the turbulent
boundary layer not only determines the spanwise separation
location of the turbulent boundary but also makes the
4 CONCLUSION downstream Görtler vortex to be fixed at a certain spanwise
position, thus forming a time stable flow structure.
In this study, a large eddy simulation (LES) is conducted to investigate The low-frequency instability in the SWBLI is successfully
the shock wave and turbulent boundary layer interaction in a 24° captured by using the power spectrum analysis method. The
compression ramp with an inflow Mach number of Ma
2.9 and corresponding characteristic frequency is St
0.04 near the
compared with experimental and numerical simulation results in the average separation point. The spectrum analysis indicates that the
literature. The shock wave/turbulent boundary layer interaction shock wave motion is related to the pulsation of the separation
(SWBLI) was analyzed from two aspects: time domain and bubble, while in other locations of flow field, the energy of pressure
frequency domain. The main conclusions are as follows: fluctuation represents almost exclusively the high-frequency
The shock wave/turbulent boundary layer interaction represents characteristics of the fully developed turbulent boundary layer.
some flow structure characteristics. When the large scale vortex
passes the root of the shock wave, the shock wave surface wrinkles DATA AVAILABILITY STATEMENT
due to the intermittence of the large-scale vortex in the spanwise
direction. Under the strong inverse gradient pressure of the The raw data supporting the conclusions of this article will be
compression ramp, the large-scale vortex breaks into a small-scale made available by the authors, without undue reservation.
vortex, the turbulent boundary layer manifests a strong anisotropy
characteristic, and the turbulent kinetic energy rapidly increases in
the outer layer of the boundary layer downstream of the ramp. AUTHOR CONTRIBUTIONS
In the region of interaction, the separated shock wave
represents an unsteady motion along the streamwise direction. XH, L-XW, and D-DZ contributed to the conception and design
This phenomenon is successfully captured by the intermittent of the study. XH provided an idea scheme; XH, L-XW, and D-DZ
factor. The range of the shock wave motion is coded; XH analyzed; and NG and l wrote the first draft. All
−3.2δ ≤ x ≤ − 2.48δ, indicating that the unsteady shock wave authors contributed to manuscript revision and read and
motion is around the average separation point. approved the submitted version.

Frontiers in Energy Research | www.frontiersin.org 9 March 2022 | Volume 10 | Article 854019

Huang et al. Shock Wave–Induced Flow Separation

REFERENCES Smits, A. J., and Muck, K.-C. (1987). Experimental Study of Three Shock Wave/
Turbulent Boundary Layer Interactions. J. Fluid Mech. 182 (182), 291. doi:10.
1017/S0022112087002349
Adams, N. A. (2000). Direct Simulation of the Turbulent Boundary Layer along a Tong, F., Yu, C., and Tang, Z. (2017). Numerical Studies of Shock Wave
Compression Ramp at M = 3 and Re=1685. J. J. Fluid Mech. 420 (1), 47–83. Interactions with a Supersonic Turbulent Boundary Layer in Compression
doi:10.1017/S0022112000001257 Corner: Turning Angle Effects. J. Comput. Fluids 149, 56–69. doi:10.1016/j.
Barbe, K., Pintelon, R., and Schoukens, J. (2010). Welch Method Revisited: compfluid.2017.03.009
Nonparametric Power Spectrum Estimation via Circular Overlap. J. IEEE Touber, E., and Sandham, N. D. (2009). Large-Eddy Simulation of Low-Frequency
Trans. Signal Process. 58 (2), 553–565. doi:10.1109/TSP.2009.2031724 Unsteadiness in a Turbulent Shock-Induced Separation Bubble. J. Theor.
Bookey, P., Wyckham, C., and Smits, A. (2005). New Experimental Data of STBLI Comput. Fluid Dyn. 23 (2), 79–107. doi:10.1007/s00162-009-0103-z
at DNS/LES Accessible Reynolds Numbers. Reno: 43rd AIAA Aerospace Touber, E., and Sandham, N. D. (2011). Low-order Stochastic Modelling of Low-
Sciences Meeting and Exhibit. doi:10.2514/6.2005-309 Frequency Motions in Reflected Shockwave/boundary-Layer Interactions.
Dolling, D. S. (2001). Fifty Years of Shock-Wave/boundary-Layer Interaction J. Fluid Mech. 671, 417. doi:10.1017/S0022112010005811
Research - what Next? AIAA J. 39 (8), 1517–1531. doi:10.2514/3.14896 Wang, Bo. (2015). The Investigation into the Shock Wave/boundary-Layer
Dolling, D. S., and Or, C. T. (1985). Unsteadiness of the Shock Wave Structure in Interaction Flow Field Organization. Changsha: Graduate School of National
Attached and Separated Compression Ramp Flows. J. Experiments Fluids 3 (1), University of Defense Technology.
24–32. doi:10.1007/BF00285267 Wu, M., and Martin, M. P. (2008). Analysis of Shock Motion in Shockwave
Dupont, P., Haddad, C., and Debieve, J. F. (2006). Space and Time Organization in and Turbulent Boundary Layer Interaction Using Direct Numerical
a Shock-Induced Separated Boundary Layer. J. Fluid Mech. 559, 255. doi:10. Simulation Data. J. J. Fluid Mech. 594 (594), 71–83. doi:10.1017/
1017/S0022112006000267 S0022112007009044
Dussauge, J. P., Dupont, P., and Debiève, J. F. (2006). Unsteadiness in Shock Wave Wu, Yu., Shi-He, Yi., and Chen, Zhi. (2013). Experimental Investigations on
Boundary Layer Interactions with Separation. J. Aerospace Sci. Tech. 10 (2), Structures of Supersonic Laminar/turbulent Flow over a Compression Ramp.
85–91. doi:10.1016/j.ast.2005.09.006 J. Acta Phys. 62 (18), 308–319. doi:10.7498/aps.62.184702
Floryan, J. M. (1991). On the Görtler Instability of Boundary Layers. J. Prog. Zhong, Dong-Dong., WangXu, Li., and Ge, Ning. (2021). Reflected Shock/
Aerospace Sci. 28 (3), 235–271. doi:10.1016/0376-0421(91)90006-P Boundary Layer Interaction Structure Analysis Based on Large Eddy
Fu-Lin, Ton., Tang, Zhi-Gong., and Xin-Liang, Li. (2016). Direct Numerical Simulation. Chin. J. aeronautics 34 (5), 364–372. doi:10.1016/j.cja.2020.12.009
Simulation of Shock-Wave and Transitional Boundary Layer Interaction in Zhu, J., and Qiu, J. (2017). A New Type of Finite Volume WENO Schemes for
a Supersonic Compression Ramp. J. Chin. J. Aeronautics 37 (12), 3588–3604. Hyperbolic Conservation Laws. J. Sci. Comput. 73 (2-3), 1338–1359. doi:10.
doi:10.7527/S1000-6893.2016.0096 1007/s10915-017-0486-8
Ganapathisubramani, B., Clemens, N., and Dolling, D. (2007). Effects of Upstream Zhuang, Yi., Tan, Hui-Jun., Li, Xin., Guo, Yun-Jie., and Sheng, Fa-Jia. (2018).
Coherent Structures on Low-Frequency Motion of Shock-Induced Turbulent Evolution of Coherent Vertical Structures in a Shock Wave/turbulent
Separation. Reno: 45th AIAA Aerospace Sciences Meeting and Exhibit. doi:10. Boundary-Layer Interaction Flow, Phys. Fluids 30, 111702. doi:10.1063/1.
2514/6.2007-1141 5058278
Germano, M. (1991). A Dynamic Subgrid-Scale Eddy Viscosity Model. Phys. Fluids Zhuang, Yi., Tan, Hui-Jun., Li, Xin., Sheng, Fa-Jia., and Zhang, Yu-chao. (2018).
A 3, 1760–1765. doi:10.1063/1.857955 Görtler-like Vortices in an Impingingshock Wave/turbulent Boundary
Grilli, M., Hickel, S., and Adams, N. A. (2013). Large-Eddy Simulation of a Supersonic Layerinteraction Flow. Phys. Fluids 30, 061702. doi:10.1063/1.5034242
Turbulent Boundary Layer over a Compression–Expansion Ramp. J. Int. J. Heat
Fluid Flow 42 (8), 79–93. doi:10.1016/j.ijheatfluidflow.2012.12.006 Conflict of Interest: Author XH was employed by the company AECC HUNAN
Lee, C. B., and Wang, S. (1995). Study of the Shock Motion in a Hypersonic Shock Aviation Powerplant Research Institute.
System/turbulent Boundary Layer Interaction. Experiments in Fluids 19 (3),
143–149. doi:10.1007/BF00189702 The remaining authors declare that the research was conducted in the absence of
Loginov, M. S., Adams, N. A., and Zheltovodov, A. A. (2006). Large-eddy any commercial or financial relationships that could be construed as a potential
Simulation of Shock-Wave/turbulent-Boundary-Layer Interaction. J. J. Fluid conflict of interest.
Mech. 565 (1), 135–169. doi:10.1017/S0022112006000930
Pasquariello, V., Hickel, S., and Adams, N. A. (2017). Unsteady Effects of Strong Publisher’s Note: All claims expressed in this article are solely those of the authors
Shock-Wave/Boundary-Layer Interaction at High Reynolds Number. J. J. Fluid and do not necessarily represent those of their affiliated organizations, or those of
Mech. 823 (1), 617–657. doi:10.1017/jfm.2017.308 the publisher, the editors and the reviewers. Any product that may be evaluated in
Sasaki, K., Barros, D. C., Andre, V., and Cavalieri, G. (2021). Causality in the Shock this article, or claim that may be made by its manufacturer, is not guaranteed or
Wave/turbulent Boundary Layer Interaction. J. Phys. Rev. Fluids 6 (6), 064609. endorsed by the publisher.
doi:10.1103/PhysRevFluids.6.064609
Settles, G. S., Fitzpatrick, T. J., and Bogdonoff, S. M. (1979). Detailed Study Copyright © 2022 Huang, Wang, Zhong and Ge. This is an open-access article
of Attached and Separated Compression Corner Flowfields in High distributed under the terms of the Creative Commons Attribution License (CC BY).
Reynolds Number Supersonic Flow. AIAA J. 17 (6), 579–585. doi:10. The use, distribution or reproduction in other forums is permitted, provided the
2514/3.61180 original author(s) and the copyright owner(s) are credited and that the original
Shu, C. W., and Osher, S. (1989). Efficient Implementation of Essentially Non- publication in this journal is cited, in accordance with accepted academic practice.
oscillatory Shock-Capturing Schemes. J. J. Comput. Phys. 77 (2), 439–471. No use, distribution or reproduction is permitted which does not comply with
doi:10.1016/0021-9991(88)90177-5 these terms.

Frontiers in Energy Research | www.frontiersin.org 10 March 2022 | Volume 10 | Article 854019