Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-16T06:12:13.602Z Has data issue: false hasContentIssue false

Periodic forcing of a large turbulent separation bubble

Published online by Cambridge University Press:  11 March 2021

Abdelouahab Mohammed-Taifour
Affiliation:
Laboratoire de thermo-fluide pour le transport, École de technologie supérieure, Montréal, Québec H3C 1K3, Canada
Julien Weiss*
Affiliation:
Laboratoire de thermo-fluide pour le transport, École de technologie supérieure, Montréal, Québec H3C 1K3, CanadaInstitute of Aeronautics and Astronautics, Technical University of Berlin, 10587 Berlin, Germany
*
†Email address for correspondence: julien.weiss@tu-berlin.de

Abstract

The response of a pressure-induced turbulent separation bubble (TSB) to periodic forcing by pulsed-jet actuators (PJAs) positioned in the upstream boundary layer is investigated experimentally in an attempt to elucidate the mechanism of low-frequency contraction and expansion (‘breathing’) already documented in this flow by Mohammed-Taifour & Weiss (J. Fluid Mech., vol. 799, 2016, pp. 383–412). The TSB is generated on a flat test surface by a combination of adverse and favourable pressure gradients and the free-stream velocity is $25\ \textrm {m}\,\textrm {s}^{-1}$. The results indicate that periodic forcing artificially reduces the size of the TSB by moving separation downstream and reattachment upstream. The smaller TSB is associated with narrower streamwise distributions of average pressure and forward-flow fraction, as well as smaller turbulent stresses in the shear layer bounding the recirculation region. Transient forcing experiments further demonstrate that the TSB responds to upstream forcing with a characteristic time scale that is of the same order of magnitude as that of the breathing motion. Overall, the results of this study support a mechanism whereby the low-frequency breathing motion is a response of the TSB to upstream perturbations that affect the position of separation first and, indirectly, the position of reattachment through a global redistribution of the pressure and velocity fields. The low-frequency behaviour of the TSB appears to be well illustrated by a first-order low-pass filter model that converts the broadband fluctuations of the incoming turbulent boundary layer into a low-frequency, large-scale oscillation of the separation and reattachment fronts, thus leading to a contraction and expansion of the TSB. The results of the continuous forcing experiments also offer new insights into active separation control with PJAs by showing that such actuators generate strong starting vortices that, when convected within an adverse pressure gradient, are associated with a downstream shift of the separation front.

Information

Type
JFM Papers
Copyright
© The Author(s), 2021. Published by Cambridge University Press
Figure 0

Figure 1.(a) Average wall-pressure coefficient $c_p$ measured on the test-section centreline ($z=0\ \textrm {m}$) and at two symmetric spanwise positions ($z={\pm }0.15\ \textrm {m}$). (b) Profile sketch of the test section with contour plot of the longitudinal velocity (the numbers 1 to 5 refer to the particle image velocimetry measurement stations). Solid white and dashed blue lines denote the dividing streamline and the $\delta _{99}$ boundary-layer thickness, respectively. Inset: settling chamber and contraction of the PJA system (§ 2.2).

Figure 1

Figure 2.Exemplary time traces of the jet velocity for $F_j = 100\ \textrm {Hz}$.

Figure 2

Table 1.PJA parameters for $F_j=100\ \textrm {Hz}$.

Figure 3

Table 2.Non-dimensional forcing frequency $F^+=F_j L_b/U_{ref}$ and ratio of several forcing frequencies to the characteristic frequencies of the breathing and shedding modes of the unforced flow: $f_{breath} \simeq 0.6\ \textrm {Hz}$ and $f_{shed} \simeq 15\ \textrm {Hz}$, respectively.

Figure 4

Figure 3.Sample of oil-film visualizations. $(a)$ Unforced flow, $(b)$ forcing at $x=1.45\ \textrm {m}$ and $(c)$ forcing at $x=1.55\ \textrm {m}$ for a supply pressure of 4 bar and a forcing frequency of 20 Hz. In $(c)$ the upstream part of the test surface is blocked by the actuator set-up.

Figure 5

Figure 4.Averaged velocity field along the test-section centreline. The white solid line delimits $\bar {U}=0$.

Figure 6

Figure 5.Averaged and fluctuating pressure coefficients and forward-flow fraction along the test-section centreline.

Figure 7

Figure 6.The p.s.d. ($a$) and pre-multiplied p.s.d. ($b$) of wall-pressure fluctuations at the local and global maxima of $c_{p'}$.

Figure 8

Figure 7.Magnitude-squared coherence $C$ and phase angle $\theta$ of the low-frequency component of the wall-pressure fluctuations measured at $x_1=1.60\ \text {m} \simeq x_{ID}$ and $x_2=x_{CR}=2.30\ \textrm {m}$. The blue line is a linear fit on the phase angle between $0$ and $4\ \textrm {Hz}$.

Figure 9

Figure 8.($a$) Correlation coefficient between the low-pass filtered pressure fluctuations measured at reference positions $x=1.65\ \textrm {m}$ and $x=2.30\ \textrm {m}$ and a moving point on the test-section centreline. ($b$) Conceptual model of the contraction and expansion of the average pressure distribution.

Figure 10

Figure 9.Forward-flow fraction $\gamma$ for the unforced and forced flows with $P_j=4 \ \textrm {bar}$ ($C_\mu =1.4\,\% \pm 0.06\,\%)$. Forcing position at $x=1.45\ \textrm {m}$ ($a$) and at $x=1.55\ \textrm {m}$ ($b$). Vertical dashed lines indicate the streamwise position of actuation.

Figure 11

Figure 10.Average wall-pressure coefficient $c_p$ for the unforced and forced flows at varying forcing frequencies and amplitudes. Forcing position at $x=1.45\ \textrm {m}$ ($a$) and at $x=1.55\ \textrm {m}$ ($b$). Vertical dashed lines indicate the streamwise position of actuation.

Figure 12

Figure 11.Forward-flow fraction $\gamma$ for the unforced and forced flows with $F_j=100\ \textrm {Hz}$ and varying jet amplitudes $C_\mu =1.24\,\%$, $1.43\,\%$, $1.75\,\%$ and $1.79\,\%$. Forcing position at $x=1.55\ \textrm {m}$ (vertical dashed line).

Figure 13

Figure 12.($a$) Positions of ID/CR ($\gamma =99\,\%$) and TD/TR ($\gamma =50\,\%$) for unforced (dashed lines) and forced (solid lines) flows with $C_\mu =1.4\,\% \pm 0.06\,\%$. ($b$) Characteristic lengths $L_{50}$ and $L_{99}$ as function of $C_\mu$ and $F_j$. Forcing position at $x=1.55\ \textrm {m}$.

Figure 14

Figure 13.Contour plots of $-\overline {\omega _z}$ in s$^{-1}$, $\overline {u'u'}/U_{ref}^2$, $\overline {v'v'}/U_{ref}^2$ and $-\overline {u'v'}/U_{ref}^2$ (from (a,e,i,m,q) to (d,h,l,p,t), respectively). Unforced (a,b,c,d) and forced flow with $C_\mu =1.4\,\% \pm 0.06\,\%$ and $F_j=0$, $15$, $40$ and $100\ \textrm {Hz}$ (from (e,f,g,h) to (q,r,s,t), respectively). Reynolds stresses are multiplied by $10^3$. The white line depicts the iso-line $U=0$. Forcing position at $x=1.55\ \textrm {m}$.

Figure 15

Figure 14.The p.s.d.s of wall-pressure fluctuations for the natural and forced flows, $x=2.35\ \textrm {m}$: $(a)$ standard representation and $(b)$ pre-multiplied representation.

Figure 16

Figure 15.Illustration of vortical flow dynamics in the first PIV station with contour maps of $\varGamma _2 \leq -2/{\rm \pi}$ (counter-clockwise) and $\varGamma _2 \geq 2/{\rm \pi}$ (clockwise). Panels (a) to (d) show $F_j=15\ \textrm {Hz}$, $F_j=40\ \textrm {Hz}$, $F_j=100\ \textrm {Hz}$ and $F_j=200\ \textrm {Hz}$, respectively, with $C_\mu =1.4\,\%$ for all frequencies; $\varGamma _2$ was computed after subtracting the mean velocity field from the phase-averaged field.

Figure 17

Figure 16.Illustration of the flow dynamics for $F_j=15\ \textrm {Hz}$ and $C_\mu =1.4\,\%$ in PIV station 2 with contours of phase-averaged streamwise velocity (a,c,e,g) and $\varGamma _2$ (b,d,f,h). Time delay of 15.3 ms (approximately one quarter of a period) between the snapshots from (a,b) to (g,h). See also supplementary movie 1.

Figure 18

Figure 17.Illustration of the flow dynamics for $F_j=100\ \textrm {Hz}$ and $C_\mu =1.4\,\%$ in PIV station 3 with contours of phase-averaged streamwise velocity (a,c,e,g) and $\varGamma _2$ (b,d,f,h). Time delay of 2.5 ms (one quarter of a period) between the snapshots from (a,b) to (g,h). See also supplementary movie 2.

Figure 19

Figure 18.Illustration of the methodology used for transient forcing: ($a$) forcing signal; ($b$) output of calorimetric sensor at $x=1.75\ \textrm {m}$; ($c$) forward-flow fraction computed by ensemble averaging the calorimetric sensor output over 90 square-wave periods.

Figure 20

Figure 19.Comparison of forward-flow fraction distribution between unforced flow, continuous forcing, and transient forcing ($F_j=100\ \textrm {Hz}$). Dashed line indicates position of forcing.

Figure 21

Figure 20.Transient forward-flow fraction measured at ($a$) $x=1.75\ \textrm {m}$ for $F_j=15\ \textrm {Hz}$, ($b$) $F_j=40\ \textrm {Hz}$, ($c$) $F_j=100\ \textrm {Hz}$ (the blue and green curves are fitted to (5.1)) and ($d$) variation of forcing amplitude at $F_j=100\ \textrm {Hz}$.

Figure 22

Table 3.Time constants of ensemble-averaged $\gamma (t)$ at $x=1.75\ \textrm {m}$ according to (5.1).

Figure 23

Figure 21.($a$) Illustration of the definition of the time delay $\delta t$ between $t=1\ \textrm {s}$ and the vertical dashed-dotted line at $x=1.85\ \textrm {m}$ and $x=2.05\ \textrm {m}$, respectively; ($b$) variation of $\delta t$ as a function of the streamwise position.

Figure 24

Figure 22.Illustration of the flow dynamics in PIV station 1 for a start of forcing at $F_j=100\ \textrm {Hz}$ and $C_{\mu }=1.4\,\%$ with contours of phase-averaged streamwise velocity (a,c,e,g,i) and $\varGamma _2$ (b,d,f,h,j). Forcing is started at $t_0$. See also supplementary movie 3.

Figure 25

Figure 23.Illustration of the flow dynamics in PIV station 1 for a stop of forcing at $F_j=100\ \textrm {Hz}$ and $C_{\mu }=1.4\,\%$ with contours of phase-averaged streamwise velocity (a,c,e,g,i) and $\varGamma _2$ (b,d,f,h,j). Forcing is stopped at $t_0$. See also supplementary movie 4.

Figure 26

Figure 24.($a$) Solutions of (6.1) and (6.3) for $\gamma =50\,\%$ at $t=1\ \textrm {s}$. ($b$) Standard (blue) and pre-multiplied (red) autospectral density functions of $x(t)$ ((6.2)) for $G_{uu}=1$.

Figure 27

Figure 25.The p.s.d.s of streamwise velocity fluctuations ($G_{uu}$) measured at $x=1.45\ \textrm {m}$ and $x=1.60\ \textrm {m}$ ($y \simeq \delta$) and scaled output of first-order model $G_{xx}$ applied to the $x=1.45\ \textrm {m}$ data with $\tau = 0.2\ \textrm {s}$.

Mohammed-Taifour and Weiss supplementary movie 1

See word file for movie caption

Download Mohammed-Taifour and Weiss supplementary movie 1(Video)
Video 4.8 MB

Mohammed-Taifour and Weiss supplementary movie 2

See word file for movie caption

Download Mohammed-Taifour and Weiss supplementary movie 2(Video)
Video 4.6 MB

Mohammed-Taifour and Weiss supplementary movie 3

See word file for movie caption

Download Mohammed-Taifour and Weiss supplementary movie 3(Video)
Video 8.1 MB

Mohammed-Taifour and Weiss supplementary movie 4

See word file for movie caption

Download Mohammed-Taifour and Weiss supplementary movie 4(Video)
Video 19.7 MB
Supplementary material: File

Mohammed-Taifour and Weiss supplementary material

Captions for movies 1-4

Download Mohammed-Taifour and Weiss supplementary material(File)
File 12.6 KB