You are on page 1of 27

J. Fluid Mech. (2018), vol. 844, pp. 878–904.

c Cambridge University Press 2018 878


doi:10.1017/jfm.2018.211

Experiments on critical Reynolds number and


global instability in roughness-induced
laminar–turbulent transition

Dominik K. Puckert1, † and Ulrich Rist1


1 Institut für Aerodynamik und Gasdynamik, Universität Stuttgart, Pfaffenwaldring 21,
D-70569 Stuttgart, Germany

(Received 2 August 2017; revised 22 December 2017; accepted 6 March 2018;


first published online 13 April 2018)

The effects of isolated, cylindrical roughness elements on laminar–turbulent transition


in a flat-plate boundary layer are investigated in a laminar water channel. Our
experiments aim at providing a comparison to global linear stability theory (LST)
by means of hot-film anemometry and particle image velocimetry. Although the
critical Reynolds number from theory does not match the transition Reynolds number
observed in experiments, there are distinct experimental observations indicating a
changeover from purely convective to absolute/global instability very close to the
critical Reynolds number predicted by theory. Forcing with a vibrating wire reveals
the evolution of the system dynamics from an amplifier to a wavemaker when the
critical Reynolds number is exceeded. The mode symmetry is varicose for thick
roughness elements and a changeover from varicose to sinuous modes is observed at
the critical Reynolds number for thin roughness elements. Therefore, most predictions
by global LST can be confirmed, but additional observations in the physical flow
demonstrate that not all features can be captured adequately by global LST.

Key words: absolute/convective instability, boundary layer stability, transition to turbulence


https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

1. Introduction
The effects of surface roughness on laminar boundary layers is a long-lasting
problem and particularly driven by the increasing demand for economical and
sustainable flight. Early observations on isolated, three-dimensional (3-D) roughness
elements are summarized in § 1.1 to provide an overview of known experimental and
numerical observations. Some of the physical mechanisms, however, are not fully
understood, and as a consequence, the definition of the ‘critical’ roughness Reynolds
number has been rather ambiguous up to now. Global linear stability theory is a
relatively new and promising tool to predict laminar–turbulent transition of certain
flows unambiguously. The objective of this experimental investigation is to provide a
comparison to recent theoretical advances described in § 1.2.

† Email address for correspondence: puckert@iag.uni-stuttgart.de


Experiments on global instability in roughness-induced transition 879
1.1. Roughness-induced instability
One of the first investigations of 3-D isolated roughness elements was done by
Gregory & Walker (1956), who visualized a steady horseshoe vortex that wraps
around the front side of the roughness and then travels downstream as a pair of
counter-rotating, streamwise vortices. They also observed a pair of vortices in the aft
recirculation zone which develops into periodic vortex shedding with the shape of a
hairpin if the roughness is high enough. This observation was also made by Mochizuki
(1961a,b) and linked to the rapid laminar–turbulent transition in the roughness wake.
Later, hairpin vortices were studied in detail by Acarlar & Smith (1987) who found
strong similarity to the turbulence production process in the near-wall transitional
boundary layer. Klebanoff, Tidstrom & Sargent (1962) attributed this observation to a
Kelvin–Helmholtz (KH) type of instability, which was later substantiated by Klebanoff,
Cleveland & Tidstrom (1992) and more recently by Ergin & White (2006).
Another characteristic feature of 3-D roughness elements is streamwise elongated,
steady streaks of alternating high- and low-speed fluid. Such streaks arise from an
exchange of momentum of the near- and far-wall boundary layer through streamwise
vorticity and can be depicted as a scar being ripped into the boundary layer
(Andersson et al. 2001). The mechanism has been known since early visualizations by
Mochizuki (1961a) and was later denoted the ‘lift-up effect’ by Landahl (1990). From
a theoretical point of view, boundary layer streaks can be linked to transient growth
(Reshotko 2001), which relies on the non-orthogonality of eigenmodes resulting in
short-term growth in contrast to the classic modal growth of Tollmien–Schlichting (TS)
waves. Andersson, Berggren & Henningson (2005) found that the so-called ‘optimal’
disturbances, leading to the maximum transient growth, are steady, streamwise vortices
for a flat-plate boundary layer. If the amplitude of the resulting optimal streaks
grows larger than 26 % of the free-stream velocity, secondary instabilities can set
in (Andersson et al. 2001). However, White (2002) found that neither wall-normal
disturbance profiles nor the streamwise stations of maximum disturbance growth
match between roughness streaks and transient growth theory. Denissen & White
(2013) analysed this problem theoretically and with direct numerical simulation
(DNS) and confirmed that optimal growth is more present in the mid-wake where
exponential growth may have already caused transition. The DNS of Rizzetta & Visbal
(2007) confirmed earlier findings of strong hairpin vortex shedding downstream of the
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

roughness. Bi-local, or ‘biglobal’ (Theofilis 2011) stability computations of streamwise


planes in the roughness wake have been done by Denissen & White (2013) and Shin,
Rist & Krämer (2015). In the work by Shin et al. (2015) the two unstable 2-D
modes were either of sinuous or varicose shape, whereas the varicose mode could
be associated with hairpin vortex shedding and was thus much more distinct in the
water channel experiment than the sinuous mode. Theoretical predictions by bi-local
stability theory are promising, but a major shortcoming is the restriction to parallel
flows.
Besides understanding the flow topology, many early workers were particularly
motivated to identify criteria for laminar–turbulent transition. As already suggested
by Schiller (1932) and advanced by Klanfer & Owen (1953), the roughness Reynolds
number Rekk = uk k/ν, where k is the roughness height, uk the streamwise velocity
at the tip of the roughness of the undisturbed boundary layer and ν the kinematic
viscosity, is deemed best as a measure for roughness-induced transition. Sometimes
the control parameter is also the Reynolds number Rek = Ue k/ν based on the boundary
layer edge velocity Ue , but the majority of the literature prefers to use Rekk . Based
on this parameter, Klanfer & Owen (1953) reanalysed the data of Gregory & Walker
880 D. K. Puckert and U. Rist
(1956) and found transition occurring at Rekk,tr = 440 for conical roughness, and
similarly, Tani (1961) found 500 < Rekk,tr < 800 for hemispherical roughness elements
with the data of Klebanoff, Schubauer & Tidstrom (1955). In another study, Tani
et al. (1962) detected transition in a range of 600 < Rekk,tr < 1000. In their attempt to
organize early data, von Doenhoff & Braslow (1961) found that Rekk,tr is lower
for increasing aspect ratios η = d/k, where d is the roughness diameter. They
presented a famous transition diagram which is still used in many engineering
applications to date. Nevertheless, the spread between published data remains large,
most authors accounting differences in roughness shape and experimental detection
methods for the observed deviations, but also pressure gradients and free-stream
turbulence (FST) (Klebanoff et al. 1992). In a more recent investigation at the low
turbulence wind tunnel at KTH Sweden, Fransson et al. (2005) reported a stabilizing
effect of cylindrical roughness arrays with respect to TS instability. The upper limit
of their technique was determined to be Rekk,tr = 422 with rapid transition when this
value was exceeded. Although the published data from the last and present century are
extensive and helpful for many purposes, they neither give an unambiguous transition
criterion nor explain basic processes that lead to roughness-induced laminar–turbulent
transition.

1.2. Global stability theory


Three-dimensional global, or ‘triglobal’ (Theofilis 2011) linear stability theory (LST)
provides global instability modes by solving the linearized Navier–Stokes equations
as an eigenvalue problem with a disturbance ansatz in all three spatial directions.
This is an advancement from bi-local stability theory, which is restricted to the
parallel-flow assumption. Therefore, 3-D global LST is particularly beneficial for
highly non-parallel flows such as roughness wakes. One drawback of 3-D global LST
is the enormous memory requirement, which often means that numerical techniques
such as Krylov subspaces must be used for computability (Bagheri et al. 2009). The
changeover from a globally stable to unstable system typically occurs when increasing
a control parameter such as Rekk beyond a certain threshold. Mathematically speaking,
this event is a supercritical Hopf bifurcation from a stable fixed point to a limit
cycle oscillation. In a physical system the latter state reveals itself as self-sustained,
intrinsic system oscillations, independent of external noise. Such a system is often
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

called an oscillator or wavemaker and should be distinguished from an amplifier


which only amplifies external noise. Descriptive introductions to nonlinear dynamics
and linear stability theory can be found in the book of Manneville (2004) and the
review article by Theofilis (2011), respectively.
Closely related classifications are local absolute and convective instabilities.
Whereas local absolute instabilities grow temporally at a fixed location, convective
instabilities can only grow if they travel in the downstream direction. In this sense,
local absolute and global instabilities are similar except for their different spatial
extent. Huerre & Monkewitz (1990) showed that 3-D global instability can ignite
within a reasonably large streamwise extent of local absolute instability and Chomaz,
Huerre & Redekopp (1991) and Monkewitz, Huerre & Chomaz (1993) concluded that
within the linear and weakly non-parallel approximation, local absolute instability
is in general a necessary condition for the existence of 3-D global instability. The
restriction of linearity and weak non-parallelism was later abolished for more general
conditions as described in the review article by Chomaz (2005).
Three-dimensional global stability analyses on roughness elements have been done
by Loiseau et al. (2014) who found that the first unstable global mode is varicose
Experiments on global instability in roughness-induced transition 881
(symmetric) for large aspect ratios and sinuous (antisymmetric) for small aspect
ratios. Apparently the two modes compete against each other as in the observations
by Asai, Minagawa & Nishioka (2002) for pieces of screens inside the boundary
layer. However, to our knowledge, there has been no physical experiment on
this observation for roughness elements yet. Further, the practical significance of
global modes in roughness-induced transition is still unclear, particularly due to the
convective nature of the flow (Asai et al. 2002; Brandt et al. 2003) and transiently
growing perturbations (Cherubini et al. 2013). As shown by Chomaz (2005), transient
growth can be considered as a convective instability from a local point of view and
as non-normality of the evolution operator from a global point of view. In this work,
instead of immersing ourselves in a theoretical analysis, we aim at directly clarifying
the role of 3-D global LST and its critical Reynolds number in a physical experiment.
In theory, the significance of the critical Reynolds number is clear: when
the growth rate of a mode becomes positive for the first time, the system is
unstable. Experimentally, it requires innovative strategies to detect a globally or
absolutely unstable flow unambiguously and therefore some successful examples are
summarized here. One prominent example is the vortex shedding behind 2-D cylinders.
Sreenivasan, Strykowski & Olinger (1987) determined the constants of the Landau
equation for this flow in the vicinity of the critical Reynolds number and thus showed
experimentally that the vortex shedding is due to a Hopf bifurcation as typically
encountered when global modes become unstable. For the problem of rotating disc
flows, Lingwood (1996) detected absolute instability by inducing a perturbation in
the form of a wavepacket that spreads in time as it travels outside towards higher
Reynolds numbers: the front velocity of the wavepacket remains constant whereas
the rear velocity asymptotically approaches zero. Inspired by this work, a similar
procedure will be applied here. For the plane sheet flow de Luca (1999) observed
the angle at which disturbances spread from a small wire spanned perpendicular to
the plane sheet flow. His criterion for absolute instability was when the disturbance
angle approached 90◦ . A changeover from convective to absolute instability has
also been found experimentally in a rotating Hagen–Poiseuille flow by observing
the inlet region of the pipe in Shrestha et al. (2012) and later more quantitatively
by measuring wavepacket fronts travelling upstream and downstream during pipe
spin-down (Miranda-Barea et al. 2016). Gondret et al. (1999) found global instability
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

in a Hele-Shaw cell; however, their technique can also be hardly applied here. In
the case of 3-D roughness in a boundary layer, Kurz & Kloker (2016) showed with
DNS that after exposing a roughness of a certain height to controlled disturbances,
turbulence appears behind the roughness and even remains after switching off the
disturbance source. This indicates the existence of a self-sustained mode; however,
this study has been done for a roughness bump in cross-flow and for artificial base
flows with selective frequency damping. What has been missing to date is a physical
experiment detecting global instability in a 3-D isolated roughness case. Preliminary
work has been done by Puckert & Rist (2018), where time traces and spectral data
strongly support the 3-D global instability from theory for one case. Motivated by this
observation, a more extensive study is provided here to extend the predominant belief
in the literature that roughness instabilities are of a convective, not absolute nature.
Earlier results by Puckert & Rist (2016) reveal good agreement of the frequencies
of the first unstable global mode and vortex shedding in the experiment. Also, the
mode shape for larger aspect ratios seems to be in good agreement with the physical
observation. However, vortex shedding typically occurs earlier than predicted by
theory. Bucci et al. (2018) explained this phenomenon by quasi-resonance of a global
882 D. K. Puckert and U. Rist
mode, meaning that background noise can be selectively amplified by the roughness
at a frequency identical to the global mode. Pseudospectra analysis revealed that the
effect of non-modal growth is rather small. The contradiction in the previous literature
is that the roughness seems to act both as an amplifier and as an oscillator. This
issue is still to be resolved and addressed in the present work. To be more specific
we aim at the following main questions.
(i) Can the critical Reynolds numbers from global LST be confirmed experimentally?
(ii) When is the roughness an amplifier or wavemaker?
(iii) Can both varicose and sinuous modes be found in an experiment?
(iv) What is the physical relevance of global instability?
This paper is structured as follows. In § 2 the facility, experimental methods
and procedures are explained, in § 3 a comparison between numerical base flow
and experimental mean flow is provided and § 4 introduces a key problem on the
transition Reynolds number versus the critical Reynolds number in the literature.
This problem is resolved in § 5 with the present experimental data. The difference
between amplifier and wavemaker in regard to roughness wakes is presented in § 6.
Further, the difficult detection of the sinuous mode in experiments is resolved and
also discussed in § 6.

2. Experimental methods
The experimental work of this investigation has been done in the laminar water
channel (Laminarwasserkanal, LaWaKa) at the Institute of Aerodynamics and
Gasdynamics (IAG), University of Stuttgart. It is a closed-loop facility with a
8 × 1.2 × 0.2 m3 test section. Wiegand (1996) reported a low turbulence intensity of
0.05 % of the free-stream velocity in the frequency range 0.1–10 Hz at 0.145 ms−1 .
At low velocities, low-frequency disturbances start to rise due to thermal stratification
in the contraction and can be reduced by additional screens in front of the test section:
see Puckert, Dieterle & Rist (2017). The advantages of this water channel are the
great boundary layer length and time scales compared to similar wind tunnels, which
allows one to observe phenomena in greater detail both visually and quantitatively.
As a drawback, the total measurement time can become huge and thus requires a
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

high level of automation. To deal with this requirement, a 3-axes traverse system
has been installed to traverse hot-film probes over the full extent of the test section.
Good optical accessibility through the free surface and the glass walls further enable
high-quality optical methods such as visualizations (Puckert et al. 2015) and particle
image velocimetry (PIV) (Lang, Rist & Wagner 2004). A permanently installed flat
plate with elliptical leading edge creates a 2-D laminar boundary layer of Blasius-type
with a shape factor H12 deviation of only ±3 % from theory (Wiegand 1996). A wire
mesh at the rear end of the flat plate provides a minor adverse pressure gradient
for edge boundary layer suction between the flat plate and side walls. However, this
pressure gradient is reduced by the growing boundary layer in the first part of the
test section. Therefore, the Blasius equation can be used to describe the experimental
velocity profile. This fact is illustrated by figure 1(a), where experimental profiles of
different streamwise stations and free-stream velocities are compared to the theoretical
profile. The measurement time was 60 s at each position and the experimental data
show good agreement with the theoretical Blasius profile. Further, all velocity profiles
in this figure are normalized by the theoretical displacement thickness δ1 based on the
exact physical station, thus revealing if virtual origins are necessary or not. For the
Experiments on global instability in roughness-induced transition 883

(a) 4 (b)

2 Blasius (theory)

0 0.2 0.4 0.6 0.8 1.0 0 1 2

F IGURE 1. Normalized wall distance y/δ1 versus (a) experimental mean velocity profiles
ū/Ue in comparison with theoretical Blasius profile and (b) disturbance velocity profiles
u0rms /Ue at various stations from the leading edge.

cases shown here and for the vast majority of experiments such corrections are not
necessary; however, the flow quality was regularly monitored to ensure the Blasius
assumption was valid. In rare cases virtual origins have been applied as described
in Kachanov & Michalke (1994), but even those are of the order of less than 2 %
of the test section length. To enable a reproduction of our experiments in follow-up
works, the wall-normal distribution of disturbance velocity root mean square (r.m.s.)
is provided in figure 1(b). It can be seen that the disturbance intensity in the free
stream is on average approximately 0.08 % for the cases considered here. The spread
of measurements is due to the absence of external forcing and the influence of
low-frequency disturbances. Spectral data of the turbulence intensity are provided for
instance in Bucci et al. (2018).
The roughness elements in this investigation are isolated cylinders with height
k, diameter d and aspect ratio η = d/k. They are made of stainless steel or brass
and remain at any station xk solely by their own weight. Except for xk , which is
measured from the leading edge, the right-handed coordinate system xyz originates at
the roughness element as shown in figure 2 and all length scales are normalized with
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

respect to k. The boundary layer edge velocity is denoted Ue and the streamwise
velocity at the roughness height is uk . In some experiments a classical horizontal
vibrating wire is installed upstream of the roughness to force the flow with 2-D
TS-like disturbances. The disturbance source is a 0.1 mm stainless steel wire
which is moved sinusoidally up and down by two stepper motors connected to one
eccentric wheel each. Another type of disturbance generator with spanwise mechanical
excitation has been developed and improved according to the design of Shin et al.
(2015). Its vertical wire of diameter 1 mm is placed 1–2 cm upstream of the leading
edge and is traversed sinusoidally by a linear actuator with an amplitude of 3 cm.
In contrast to the horizontal wire, this construction does not introduce disturbances
directly into the boundary layer but into the free stream where higher disturbance
amplitudes are necessary. Data acquisition is done with a Dantec 55R15 hot-film
probe, ‘Streamline’ bridge and an NI 16-bit A/D converter. The measurements can be
triggered from one of the two disturbance generators to lock the measurement to the
phase of the disturbance. King’s law is used to convert voltage into velocity signals
which are decomposed into mean (ū) and fluctuating component (u0 ). Digital filtering
is applied between 0.1 and 10 Hz, and physical frequencies f (Hz) are presented
884 D. K. Puckert and U. Rist

Camera

x
be
Pro
Horizontal wire d

k
Laser

Vertical wire z

F IGURE 2. Sketch of the experimental set-up.

as circular frequencies ω = 2πfkUe−1 when deemed advantageous for a comparison


to theory. The particle image velocimetry (PIV) system comprises a Q-switched
dual-pulse Nd:YAG laser (Quantel Brilliant Twin W) with a wavelength of 532 nm,
an external sequencer and a PCO Sensicam for image acquisition. Seeding is done
with 4.2 µm polyamide particles.

3. Base flow comparison


Three-dimensional global linear stability analysis solves the linearized Navier–
Stokes equations as an eigenvalue problem pertinent to a base flow of the particular
flow field (Taira et al. 2017). Often, the computation of a steady numerical base flow
requires techniques such as selective frequency damping (SFD) to artificially settle
the flow towards a steady equilibrium as described in Åkervik et al. (2006). This can
be done by an extension of the Navier–Stokes equations with an additional feedback
term and is therefore difficult to compare with physical flows for which there is no
such technique. Further, physical flows may always behave nonlinearly if Reynolds
numbers are high enough, thus invalidating the assumptions of LST. In this section, a
comparison is done between a numerical base flow with SFD provided by the authors
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

of Loiseau et al. (2014) and our experimental mean flow for the Reynolds number
Rek = 550. In Loiseau et al. (2014), this case is called the ‘Fransson experiment’
according to the intermittency experiment by Fransson et al. (2005), and this case is
slightly subcritical with respect to global modes.
Figure 3 shows isolines of the normalized streamwise velocity from the numerical
computation (Loiseau et al. 2014) on the left versus the normalized streamwise
mean velocity from our experiments on the right at three streamwise slices x = 10,
20 and 30. Root mean squared (r.m.s.) velocity fluctuations are available from the
experiment and are shown as red pseudocolours on the experimental side of the
plots. To obtain these graphs, the hot-film has been traversed to 61 × 15 equidistant
points in z and y, respectively, each with 30 s measurement time. The first slice
at x = 10 demonstrates great qualitative agreement of both strength and location
of velocity streaks resulting from streamwise vorticity induced by the roughness.
High-speed streaks can be found where the isolines are closer to the wall, e.g. at
approximately z = ±1. Vice versa, low-speed streaks are regions of lifted isolines,
e.g. at the spanwise centre and at approximately z = ±2–3. In this slice, velocity
disturbances are particularly located around the centre low-speed streak which can be
Experiments on global instability in roughness-induced transition 885

CFD Exp.
(a) 2.5 0.20

2.0 1.0 1.0 0.16

1.5 0.9 0.12


y 0.9
1.0 0.08

0.04
0.5
0
–5 –4 –3 –2 –1 0 1 2 3 4 5

CFD Exp.
(b) 2.5 0.20
1.0
2.0 1.0 0.16
0.9 0.12
1.5 0.9
y
1.0 0.08

0.04
0.5
0
–5 –4 –3 –2 –1 0 1 2 3 4 5

CFD Exp.
(c) 2.5 0.20
1.0
1.0 0.16
2.0
0.9
1.5 0.9 0.12
y
1.0 0.08

0.04
0.5
0
–5 –4 –3 –2 –1 0 1 2 3 4 5
z

F IGURE 3. (Colour online) Comparison of numerical base flow kindly provided by


Loiseau et al. (2014) on the left versus mean flow from experiments on the right at (a)
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

x = 10, (b) x = 20 and (c) x = 30. Contour levels show u/Ue from 0 to 1 in steps of 0.1.
Rek = 550, η = 3 and xk = 57.1 (red online). Dashed lines (blue online) indicate roughness
dimensions.

attributed to inflection point instabilities of the KH type as explained by Klebanoff


et al. (1992). In the slice x = 20, the disturbances become significantly stronger and
even appear in regions of high shear around the outside low-speed streak at z = 2–3.
In the slice at x = 30, the intensity of the disturbances in the centre slightly decreases
and increases at z = 2–3 due to high spanwise shear. A quantitative analysis of the
data has been done by calculating the mean relative deviation of the experimental
flow from the numerical base flow. This yields 8 %, 10 % and 17 % deviation for
x = 10, 20 and 30, respectively. In particular at x = 10 there is good agreement of
the velocity isolines whereas the deviations at x = 20 and 30 are due to the absence
of SFD in the experiment. Nonlinear interactions account for the distorted mean flow
isolines and a comparison at these stations would thus be misleading. The agreement
at x = 10, where nonlinearities are not yet strong enough, is satisfactory, considering
886 D. K. Puckert and U. Rist
the necessary but artificial selective frequency damping in numerical simulations
and also the complexity of the experiment regarding facility-specific free-stream
turbulence, measurement difficulties and tolerances of geometric and flow parameters.
For lower Reynolds numbers the agreement would thus be as good as in figure 3(a).
The question to be answered in the following is why there are obvious fluctuations
in the experiment although theory predicts global stability!

4. Transition Reynolds number


The set-up of Fransson et al. (2005) is reproduced at the LaWaKa to detect
the transition Reynolds numbers and to compare with existing results. This case is
particularly important because it has more recently been used by Loiseau et al. (2014)
to compare to the critical Reynolds number obtained from a global stability analysis.
The purpose of this section is to distinguish between global instability in theory and
transition in a physical flow, which have been strongly linked in Loiseau et al. (2014).
In contrast to these authors, we believe that this link is rather weak. To support our
point of view, the transition detection method has to be explained first, followed by
a brief verification study on the basis of the Fransson et al. (2005) experiment and
a comparison to theory. The results are then substantiated with a parameter study
for different aspect ratios. The primary parameter in this section is the roughness
Reynolds number Rekk to be in compliance with many other experimental works
in the literature, and in particular with Fransson et al. (2005) who used the same
parameter but with different nomenclature.

4.1. Intermittency detection


Regarding the transition detection method, an intermittency function γ has been
computed to detect turbulence as a function of Rekk . In contrast to methods using
integral quantities like cf or H12 , intermittency does not require y-profiles and is thus
very suitable if measurement times are long. The computation follows the work of
Zhang et al. (2013) and is demonstrated in figure 4. First, the signal u0 in figure 4(a)
is multiplied by its derivative du0 /dt to enhance the sensitivity to turbulence. The
derivative is determined numerically based on wavelet transforms to smooth noisy
experimental data (Luo, Bai & Shao 2006). The detector function is then averaged
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

into a criterion function every 1/100 times of the signal length and compared to
a threshold value which originates from a reference in laminar flow. The reference
signal is multiplied by a constant factor in the range 1–10 to allow manual control
of the method. These manual adjustments only become necessary when the set-up
changes significantly. It has been found that the results do not change significantly as
long as reasonable inputs are given. The intermittency function γ is the percentage
of turbulence to the total signal length and can be estimated by summing up the
parts of the signal where the criterion function is greater than the threshold value.
An intermittency of 0 % or 100 % represents laminar or turbulent flow, respectively,
and an intermittency of 50 % is the criterion to define the transition Reynolds number
Rekk,tr . This value is considered as the ‘midpoint’ of transition (Halstead et al. 1995;
Stotz et al. 2014) and the same criterion has also been used by Fransson et al. (2005).
To detect this value for a specific configuration, Rekk was increased by increasing
the free-stream velocity in steps of 7.5 × 10−4 m s−1 and evaluating the signal
for each velocity at a fixed downstream station (x, y, z) = (200, 1, 0), similarly to
Fransson et al. (2005). Doing this gives illustrations as shown in figure 4(c) where
the transition Reynolds number Rekk,tr is at γ = 0.5. The results from this method
Experiments on global instability in roughness-induced transition 887

(a)
0.02

–0.02

(b) 1.0
d

0.5
d

Threshold
0
0 20 40 60 80 100

(c) 1.0

0.5

0
360 380 400 420 440

F IGURE 4. (Colour online) Intermittency computation: (a) streamwise disturbance signal


u0 , (b) detector function u0 du0 /dt (yellow online), criterion function (solid black line) and
threshold value (dashed line, blue online) and (c) intermittency factor γ .

are confirmed by dye streakline visualizations in a visual check. Further, y-profiles


have been measured to confirm the Blasius-flow assumption which is necessary to
compute Rekk as a function of uk .

4.2. Transition in the Fransson et al. (2005) experiment


The intermittency method was verified for the Fransson et al. (2005) experiment
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

with different roughness heights to see if Rekk is a general quantity which can be
compared to the literature independent of dimensional quantities such as k or Ue .
Three different roughness heights k = 3.5, 5 and 10 mm have been installed with
constant xk = 57.1 and η = 3 to match the set-up of Fransson et al. (2005). Each
experiment was done with both isolated elements and one periodic spanwise array
of seven elements of the same type with spacing 1z = 10 to investigate if arrays in
numerical simulations with symmetry conditions or in experiments such as Fransson
et al. (2005) behave similarly. Moreover, the effect of hysteresis was investigated
by determining the transition Reynolds number for both increasing and decreasing
free-stream velocities.
The results of this parameter study are summarized in figure 5 and table 1. Figure 5
shows transition Reynolds numbers of three different roughness heights, isolated
elements versus arrays and increasing (‘UP’) versus decreasing (‘DN’) free-stream
velocities. Figure 5 shows that none of these variations change the transition Reynolds
number significantly. In particular the size of the roughness has no significant impact
on transition, and for practical purposes Rekk,tr can therefore be considered as a
general quantity if other parameters such as k/δ1 and η are not changed. The
888 D. K. Puckert and U. Rist
500
Isolated UP
Array UP

450 Isolated DN
Array DN

Fransson et al. (2005)

Loiseau et al. (2014)


400

2 4 6 8 10 12

F IGURE 5. Comparison of transition Reynolds numbers Rekk,tr for different roughness


heights k with results from the literature.

k k/δ1 Rek,tr Rekk,tr Reδ1 Medium Reference


3.5 mm 1.69 483 401 286 water —
5 mm 1.69 486 404 287 water —
10 mm 1.69 484 402 286 water —
1 1.83 564 490 309 theory Loiseau et al. (2014)
1.4 mm 1.72 502 422 291 air Fransson et al. (2005)
TABLE 1. Verification study with different roughness heights (first three rows) and
comparison with the literature (last two rows). In all cases roughness is isolated with η = 3
and xk = 57.1.

differences between ‘UP’ versus ‘DN’ are negligible and are in agreement with
Klebanoff et al. (1992) who investigated quasi-steady oscillating roughness heights
and found no noteworthy hysteresis. Interestingly, the greatest difference seems to be
in isolated versus array of roughness elements. Arrays are less prone to transition
than isolated elements in four out of six cases. The spanwise periodicity of streaks
seems slightly stabilizing, but this effect is within the given measurement accuracy
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

and has not been investigated further. It can be concluded that the transition Reynolds
number for the Fransson et al. (2005) experiment in this work is Rekk,tr ≈ 402 with
an accuracy of ±5 % independent of the choice of k, Ue , isolated element or array.
Fransson et al. (2005) themselves found Rekk,tr = 422, which is only 5 % higher than
our result. The higher Reynolds number is probably due to the smaller turbulence
intensity of the MTL wind tunnel at KTH Sweden and also within the given accuracy
of the measurement. Unfortunately, the same cannot be said about the agreement
with Loiseau et al. (2014), who found a theoretical critical Reynolds number of
Rek,c,t = 564, equivalent to Rekk,c,t = 489. This discrepancy is graphically illustrated
by the dashed line in figure 5 and is too high to be explained by measurement
error. Apparently, transition does not coincide with global instability for this case.
Note that Loiseau et al. (2014) reported good agreement with Fransson et al. (2005),
which may result from a comparison with Reδ1 = Ue δ1 /ν instead of Rekk . As already
explained in the introduction, we prefer the more established parameter Rekk for such
a comparison because it is widely used in the literature and incorporates the effect
of ‘blockage’ by the roughness more than any other boundary layer parameter.
Experiments on global instability in roughness-induced transition 889

η xk Rek,c,t Rek,tr → dev. Rek,c,e → dev.


3 105.0 656 505 → −26 % 650 → −1 %
2 128.8 805 630 → −24 % 817 →1%
1 166.4 1040 860 → −19 % 1025 → −1 %
0.85 145.6 910 845 → −7 % 945 →4%
3 (F) 57.1 564 484 → −15 % 556 → −1 %
TABLE 2. Transitional and critical Reynolds numbers with deviations from theory
(Loiseau et al. 2014).

4.3. Parameter study


The same comparison has been repeated for the parameter study of Loiseau et al.
(2014) with different aspect ratios η = 3, 2, 1 and 0.85. In contrast to the previous
comparison, the streamwise station is now variable to obtain a constant height with
respect to the displacement thickness k/δ1 = 1.83. Experimentally, this requires us to
reposition the roughness element many times manually. Such a procedure is practically
difficult when experiments are long and not necessarily required if the boundary
layer grows relatively slowly with x. Similarly to the well-known parallel-flow
assumption in theory, the experimental parameter study has been adapted in order
for the experimental circumstances to retain the required level of automation. First, a
fixed station xk has been determined at which theory predicts the Reynolds number
to be critical. Then, only the free-stream velocity has been changed. The fixed station
xk at which the roughness was placed in the experiment is provided in table 2. To
distinguish η = 3 of the parameter study from the Fransson et al. (2005) experiment,
the latter case is marked by ‘(F)’ in case of confusion.
Comparing the transition Reynolds numbers determined as shown in figure 6
with the critical Reynolds numbers from theory confirms that there is again a
discrepancy, as in the Fransson et al. (2005) experiment. The relative deviation (dev.)
between Rek,tr and Rek,c,t is given in table 2 and is computed with the formula dev.
= 2 × (Rek,tr − Rek,c,t )/(Rek,tr + Rek,c,t ) × 100. Particularly for large aspect ratios, the
relative error is large and reduces with smaller aspect ratio. A possible explanation for
this phenomenon is related to the smaller receptivity to free-stream turbulence. More
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

importantly, there is an obvious mismatch between theory and experiments for all
cases in this parameter study despite the good agreement between the experimental
mean flow and theoretical base flow (see § 3). Therefore, global instability may
not be the decisive mechanism for transition at moderate Reynolds numbers and
globally critical Reynolds numbers can hardly be compared to transition Reynolds
numbers from experiments, contrary to what has been attempted by Loiseau et al.
(2014). This attempt has also been done by the present authors in a visual attempt
to find transition in Puckert et al. (2015) but without the accuracy of a quantitative
measurement technique. The relevance of global instability in roughness-induced
transition must be different and the next section aims at shedding light on this
apparent discrepancy.

5. Critical Reynolds number


The previous detection of transition Reynolds numbers has been done with the
probe located in the far wake to measure turbulence without knowing which type
of instability was responsible for laminar–turbulent breakdown. However, the critical
890 D. K. Puckert and U. Rist

1.0

0.5

0
–100 –50 0 50 100

F IGURE 6. Intermittency of different aspect ratios η. Rek,tr is provided in table 2. Probe


position (x, y, z) = (200, 1, 0) with 100 s measurement time for each Reynolds number.

Reynolds number is defined by the occurrence of the first (exponentially) growing


disturbance, here with respect to global instability. It is therefore to be shown in
this section if such a mechanism exists for all aspect ratios considered in this work.
Measurements have been done closer to the roughness element in the range 0 < x < 40,
where instability develops first. The probe is traversed within this range in equally
spaced increments of 1x = 2 at y = 1 to measure the time traces of u0 /Ue at different
free-stream velocities (which is equivalent to changing the Reynolds number). The
definition of the coordinate system origin at the roughness centre leads to different
starting points for the probe, depending on aspect ratio; however, the spatial resolution
is the same in all cases. The Reynolds number Rek is used instead of Rekk to enable
a more direct comparison with Loiseau et al. (2014).
First, the r.m.s. streamwise velocity fluctuations are shown in figure 7(a,c,e,g)
for different aspect ratios. In all plots it can be seen that the fluctuation r.m.s.
grows with both increasing x and Rek . When Rek is increased beyond the solid
lines, the contours become more independent of Rek and remain at a constant
streamwise station. These changes occur at one particular Reynolds number which
will be denoted as the experimental critical Reynolds number Rek,c,e , analogous to the
theoretical critical Reynolds number Rek,c,t . In contrast to Rek,c,t , which is the result of
numerical–theoretical computations, the experimental critical Reynolds number Rek,c,e
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

is a result of physical observation. It may be difficult to find Rek,c,t visually from


these graphs, and other phenomena which are capable of detecting Rek,c,e have been
summarized in Puckert & Rist (2018). One such phenomenon has been developed to
a robust detection method and is applied to the cases considered here. We will return
to figure 7 and its right-hand side after a detection of Rek,c,e .
To determine the threshold Rek,c,e quantitatively, time-dependent fluctuation signals
are evaluated. Figure 8(a,b) shows exemplary time traces for the case η = 3 at
Rek = 600 and 700, respectively. Interestingly, the signal at Rek = 600 fluctuates with
somewhat randomly changing amplitudes, whereas the signal at Rek = 700 fluctuates
with almost steady amplitude. This observation is exactly what can be expected from
the difference between convective and global instability: the signal at Rek = 600
shows a typical shape of convective beatings from amplified unsteady noise that
may later develop into intermittent turbulence, whereas the fluctuation amplitudes at
Rek = 700 have grown to nonlinear saturation without depending on external noise. In
other words, convective and global instability could be distinguished by the amount
of envelope fluctuations, which allows for a simple detection of a threshold value.
A procedure to determine the envelope functions requires smooth signals. Therefore,
Experiments on global instability in roughness-induced transition 891

(a) (b) 0.2


0 0.02 0.04 0.06 0.08 0.10 0.12

0.1

700 0
500 600 700
600

500
(d)
0 10 20 30 40 0.2

(c)
0.1
900

800
0
650 700 750 800 850 900
700

0 10 20 30 40 ( f)
0.2

(e)
1100
0.1
1000
900
0
800 800 900 1000 1100
0 10 20 30 40

(h) 0.2
(g) 1200

1000 0.1

800
0
700 800 900 1000 1100 1200
0 10 20 30 40
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

F IGURE 7. (Colour online) (a,c,e,g) Contour plots showing u0rms /Ue for various η (red
online) and (b,d,f,h) line plots at particular streamwise stations indicated by the vertical
lines in the left-hand side plots.

low-frequency oscillations, higher harmonics and noise are removed from the signal
by bandpass filtering around the fundamental frequency f0 , which can be determined
by fast Fourier transformation. The filter setting was chosen to be 0.1–1.7 × f0 , but
a different choice does not alter the results significantly as long as higher harmonics
are removed. Local extrema of the filtered signals can then be easily determined to
form an envelope function: see figure 8(c,d). As a last step, the envelope r.m.s. is
computed for both upper and lower envelopes and averaged to obtain one measure
for the amplitude fluctuation. A high envelope r.m.s. is an indicator for convective
instability and when crossing Rek,c,e , the envelope r.m.s. drops due to nonlinearly
saturated signal amplitudes, most likely from a global instability.
892 D. K. Puckert and U. Rist

(a) Subcritical (b) Supercritical


0.2

–0.2

(c) 0.2 (d)

–0.2
0 20 40 60 0 20 40 60

F IGURE 8. (Colour online) Determination of critical Reynolds number based on envelope


fluctuations: (a,b) time trace of u0 /Ue and (c,d) filtered time trace with envelopes (blue
solid lines). Rek = 600 for (a,c) and 700 for (b,d).

6
Envelope r.m.s.

8
4 6
4
2
2
0 0
450 500 550 600 650 700 750 800 900 1000 1100

6 6
Envelope r.m.s.

4 4

2 2

0 0
600 700 800 900 700 800 900 1000 1100
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

F IGURE 9. Envelope fluctuation r.m.s. for different aspect ratios.

This procedure has been applied to time traces close to the roughness (x = 4–7,
depending on the aspect ratio) at increasing Reynolds numbers. Results are presented
in figure 9, where the envelope r.m.s. is plotted as a function of Rek for different
aspect ratios. In all cases, the envelope r.m.s. rises initially with increasing Rek and
then suddenly drops. As already explained, this steep drop can be explained by a
changeover of instability mechanism. Similarly to the intermittency threshold, we
define the threshold between convective and global instability in the middle of the
envelope r.m.s. drop. The experimental critical Reynolds numbers in the contour
plots of figure 7 are determined with this procedure. The uncertainty (one standard
deviation) is estimated to be σRek = ±15 and mainly due to measurement errors of Ue .
Minor sources of error are imprecise temperature detection, which has an impact on
the kinematic viscosity, manufacturing tolerances of the roughness or local deviations
from the Blasius profile.
Experiments on global instability in roughness-induced transition 893
We now return to figure 7. To illustrate the changeover between convective and
global instability further, the fluctuation r.m.s. of the time traces at certain streamwise
stations are shown in figure 7(b,d,f,h). The respective streamwise stations are indicated
in figure 7(a,c,e,g) with dashed, dash-dotted and dotted lines. For all aspect ratios,
the slope of u0 /Ue clearly settles to a more constant level beyond Rek,c,e , which is
again indicated by a solid line in figure 7(b,d,f,h). It can be seen that Rek,c,e separates
the plots in figure 7 into two distinct regions which are, following nomenclature from
theory, denoted by the subcritical and supercritical regime. Plots similar to figure 7
with skewness, kurtosis and intermittency have also been created (not shown), but
the r.m.s. remains the most revealing quantity. Rek,c,e is not to be confused with
the transition Reynolds number, which was introduced in the previous section. In
the subcritical regime, background noises from the facility are amplified around and
downstream of the roughness, leading to convective disturbance growth, whereas
the disturbances in the supercritical regime grow temporally due to a region of
self-sustained oscillations. These findings are strong evidence of a global instability
in the supercritical regime, in contrast to convective amplification in the subcritical
regime.
The experimental critical Reynolds numbers obtained through analysis of figures 7
and 9 are compared to the theoretical critical Reynolds numbers in table 2. The
case η = 3 (F) is from Puckert & Rist (2018) and is added for completeness. In all
cases, the deviation between experimental and theoretical critical Reynolds number is
surprisingly small, particularly when considering all possible errors in the experiment
and possible influences of the domain size in the numerical computation. This
successful comparison is very strong evidence of a global instability mechanism as
predicted theoretically. The experimental detection method with envelopes is fairly
simple but powerful to confirm a possible changeover from convective to global
instability. Transition detection algorithms as in § 4 or Fransson et al. (2005) do not
reveal this type of changeover.
In order to relate the present results to other existing data in the literature, the
transition diagram of von Doenhoff & Braslow (1961) has been reproduced in
figure 10. The data in this diagram originate from various experimental investigations
with different kinds of shapes, isolated elements and arrays of roughness elements
from different facilities using different methods. This explains the huge spread of
the data but has not reduced the diagram’s popularity in many engineering and
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

scientific applications. Regarding our results, both critical and transition roughness
Reynolds numbers from our experiments fit well into this diagram and confirm that
our results are in quantitative agreement with the previous literature. The critical
Reynolds number is typically higher than the transition Reynolds number and can
possibly provide an upper limit when transition occurs at the latest.

6. Instability mechanism
It has been found that globally unstable oscillations can ignite at the theoretically
predicted Reynolds numbers. On the other hand, transition may occur earlier in an
experiment before global instability sets in. To understand these two mechanisms, it
is helpful to perceive the roughness as an amplifier in the subcritical regime and
wavemaker in the supercritical regime. This section will clarify (i) if a linear relation
exists between disturbance input and output in the subcritical regime and (ii) if self-
sustained modes can cause transition in the supercritical regime. The roughness in the
following study has been set up according to the Fransson et al. (2005) experiment
with η = 3 and xk = 57.1.
894 D. K. Puckert and U. Rist

60

40

20

10 Loftin (1946)
8 Gregory & Walker (1956)
6 Schwartzberg & Braslow (1952)
Klebanoff, Schubauer & Tidstrom (1955)
4 Smith & Clutter (1957)
Carmichael (1958)
Peterson & Horton (1959)
2 Puckert & Rist: Transitional
Puckert & Rist: Critical
1
0 1 2 3 4 5

F IGURE 10. (Colour online) Reproduction of the transition diagram by von Doenhoff &
Braslow (1961) with critical and transition roughness Reynolds numbers of the present
study.

6.1. Amplifier
A low subcritical Reynolds number of Rek = 430 was chosen in this experiment to
reduce the influence of uncontrolled background noise. The horizontal vibrating wire
has been installed at (x, y) = (−7, 0.9) upstream of the roughness element to introduce
controlled disturbances with a forcing frequency of ω = 0.73 (2 Hz), close to the
expected frequency of the stable global mode ωt = 0.8 at Rek = 550: see Loiseau
et al. (2014). This frequency is damped in an undisturbed Blasius boundary layer, but
convective amplification may happen in the presence of roughness. Fine mechanical
adjustment of the forcing generator’s eccentric wheels has been done to increase
the wire oscillation amplitude linearly from 0 to 0.2 mm in steps of 0.025 mm.
Figure 11(a) shows 5 s excerpts of the time signals measured at (x, y, z) = (12, 1, 0)
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

for increasing wire oscillation amplitudes. The x-position is chosen far enough
from the recirculation zone and also far enough from possible transition. From
their envelopes it is obvious that the oscillation amplitudes are well controllable
by the amplitude of the forcing wire. A more quantitative impression is given in
figure 11(b), which shows the fluctuation r.m.s. as a function of wire amplitude for
both undisturbed flow and roughness-disturbed flow. In the undisturbed flow, the
fluctuation increases linearly with increasing forcing amplitude. This is exactly what
is expected from a convective instability that evolves as a wave-like disturbance
in the linear regime of the flow: u0rms /Ue ∼ Awire where the control parameter is
the wire oscillation amplitude Awire . The same is true for the roughness-disturbed
flow, but linear dependency is only valid up to ∼3.5 % of Ue . A flow is typically
considered linear up to 1 % fluctuation amplitude. However, in practical applications
the nonlinear terms can often be neglected even for higher fluctuation amplitudes.
The linear relation between forcing amplitude and response amplitude proves that the
roughness acts as an amplifier of convective disturbances in the subcritical regime
and it can be assumed that the amplification rate is a function of Reynolds number,
which has been fixed in this experiment. Here, the controlled disturbances have been
Experiments on global instability in roughness-induced transition 895

(a) (b)
0.05 0.04
No roughness
0
0.02 Roughness

–0.05
0
0 10 20 30 0 0.05 0.10 0.15 0.20

F IGURE 11. Forced roughness for (η, Rek ) = (3, 430) with horizontal wire amplitude Awire
being the control parameter. (a) Time signals with increasing Awire every 5 s, (b) r.m.s. of
streamwise fluctuation velocities.

amplified by a factor of approximately 12 by the roughness. When the oscillating


wire amplitude is further increased and the disturbance amplitude grows beyond 3.5 %
of Ue , nonlinear saturation sets in and the disturbance amplitude increases similarly
to a square-root function. Depending on the stability of the downstream boundary
layer, this disturbance may develop into turbulence. Variations of frequency lead to
similar observations but with different amplification factors.

6.2. Wavemaker
At higher Reynolds numbers, the flow dynamics may change from amplifier to
wavemaker due to an unstable self-sustained global mode. If the global instability is
purely hydrodynamic without a far-field feedback loop, it necessarily requires a region
of local absolute instability within the WKBJ (Wentzel–Kramers–Brillouin–Jeffreys)
approximation as shown by Chomaz et al. (1991) and Monkewitz et al. (1993) for
general conditions. The region of local absolute instability nurtures a globally unstable
mode by shedding upstream and downstream travelling waves, in contrast to local
convective instabilities which shed only downstream travelling waves. The temporal
evolution of disturbances is investigated here by introducing a wavepacket to the same
geometric set-up as previously, but this time the vibrating wire is activated only for
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

the first 6 s and remains off for the remaining 44 s at each measurement point. The
probe has been placed at (y, z) = (1, 0) and four different streamwise stations at x = 10,
20, 30, 40. The time signals have been recorded five times and ensemble-averaged
to reduce random influences. The excitation frequency has been chosen constant at
3 Hz, which is still close to the most receptive frequency expected in the vicinity
of the critical Reynolds number, as in the previous section. The higher frequency is
due to the higher Reynolds numbers in this experiment. Repeating the acquisition
of ensemble-averaged time signals for different Reynolds numbers gives the contour
plots of figure 12, where greyscale shows the normalized disturbance magnitudes
of ensemble-averaged time signals at different Reynolds numbers. Disturbances in
figure 12(a) have peak magnitudes at approximately t = 2 s and 8 s due to transient
effects when the wire is switched on and off. The time difference from t = 0 s to
2 s is the time required by the wavepacket to travel from its source to the probe
location for the case x = 10 in figure 12(a). At more downstream probe stations, the
front and trailing edges of the wavepackets appear at later times, as can be seen in
figure 12(a–d), because the wavepackets need to travel over longer distances from
their source to the probe. In contrast, the wavepacket front and trailing edges appear
896 D. K. Puckert and U. Rist

(a) (b)
600 0.01 600 0.01

0.008 0.008
550 550
0.006 0.006
500 500
0.004 0.004
450 450 0.002
0.002

400 0 400 0
0 10 20 30 40 50 0 10 20 30 40 50

(c) (d )
600 0.01 600 0.01

0.008 0.008
550 550
0.006 0.006
500 500
0.004 0.004
450 450
0.002 0.002

400 0 400 0
0 10 20 30 40 50 0 10 20 30 40 50

F IGURE 12. Ensemble-averaged disturbance evolution versus Rek at (a) x = 10, (b) x = 20,
(c) x = 30 and (d) x = 40.

earlier when increasing the Reynolds number due to higher convection velocities.
Remarkably, the wavepacket trailing edge becomes slower in all plots when the
Reynolds number is increased. A similar phenomenon is reported in experiments
on rotating disc flows by Lingwood (1996) and has been explained by the rise of
an absolute instability. The hydrodynamic feedback of upstream and downstream
travelling waves is similar here and can be explained by a global instability which is
most likely based on a sufficiently large region of absolute instability. The threshold
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

at which this happens has already been detected and the graphs of figure 12 show
interesting features at the experimental critical Reynolds number Rek,c,e = 556. In
figure 12(b) the light grey disturbance magnitudes approach this value asymptotically,
in (c) there is even a strip of low disturbance magnitude below Rek,c,e = 556 and
in (d) this strip becomes thicker, separating the sub- from the supercritical regime
most obviously. A precise detection of a threshold is difficult with figure 12, because
convective instabilities partially overshadow the changeover to global instability. With
this experiment, however, it is shown that the wavepacket trailing edge velocity
becomes slower with increasing Reynolds number and is eventually zero or negative.
This confirms the existence of a global instability in the dark regions at high Reynolds
numbers and convective instability at low Reynolds numbers.
The coexistence of both convective and global instabilities is schematically drawn
in figure 13. In the subcritical regime, convective instabilities (CI) prevail and
background noise can be amplified to large amplitudes due to the influence of the
roughness. The supercritical regime begins at Rek,c,e with a region of global instability
(GI). The dashed line symbolizes the evolution of wavepacket leading and trailing
edges.
Experiments on global instability in roughness-induced transition 897

600
GI

550

CI
500

450 Trailing edge

Leading edge
400
0 10 20 30 40 50

F IGURE 13. (Colour online) Schematic of instability regions derived from figure 12.

6.3. Varicose versus sinuous mode


It is known from both early and more recent experimental literature that hairpin
vortex shedding begins when the Reynolds number is high enough. Hairpin vortices
are in general symmetric (varicose) with respect to the streamwise centreline whereas
other famous examples of vortex shedding such as the von Kármán street behind 2-D
cylinders are antisymmetric (sinuous). In the work of Loiseau et al. (2014), the first
unstable global mode is of varicose nature if the roughness aspect ratio is large (>2)
but sinuous if the aspect ratio is small (1 or below). The agreement for large aspect
ratios is good as already shown by Puckert et al. (2015); however, to the authors’
knowledge there is no reported experimental observation of sinuous von Kármán-type
vortex shedding behind 3-D roughness elements embedded into a boundary layer.
This apparent disagreement with the work of Loiseau et al. (2014) is investigated
here by means of PIV.
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

PIV snapshots in the xz-plane at the height of the roughness are shown in figures 14
and 15. Greyscale represents the magnitude of the instantaneous in-plane velocity,
showing the unsteady streaky wake of the roughness element with the roughness
being highlighted light grey for clarity. In the case of a large aspect ratio η = 3
(F) (see figure 14), the disturbance symmetry is clearly symmetric with respect to
the spanwise centreline. Here, the Reynolds number Rek = 700 is supercritical and
thus supports the observation of a leading unstable varicose mode from global LST.
However, the symmetric unsteadiness can also be seen for lower Reynolds numbers in
the subcritical regime due to amplification of background noise (Puckert & Rist 2016;
Bucci et al. 2018). Therefore, global and convective instability cannot be clearly
distinguished based on mode symmetry for this kind of configuration.
Interestingly, this is different for small aspect ratios. Figure 15 shows three
snapshots for η = 1 at increasing Reynolds numbers. On the one hand, the
leading disturbance in figure 15 is clearly symmetric for Rek = 990 (figure 15a),
but on the other hand it becomes antisymmetric for Rek = 1230 (figure 15c).
At an intermediate Reynolds number Rek = 1140 (figure 15b) the unsteadiness
can be understood as a mixture between symmetric and weakly antisymmetric
898 D. K. Puckert and U. Rist

4 1.0

0.8
2
0.6
z 0
0.4
–2
0.2

–4 0
0 2 4 6 8 10 12 14 16 18 20 22
x

F IGURE 14. PIV snapshot of the xz-plane with η = 3, xk = 57.1 and Rek = 700.

(a) 2 1.0
0.8
0.6
z 0
0.4
0.2
–2 0
0 2 4 6 8 10 12 14 16 18

(b) 2 1.0
0.8
0.6
z 0
0.4
0.2
–2 0
0 2 4 6 8 10 12 14 16 18

(c) 2 1.0
0.8
0.6
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

z 0
0.4
0.2
–2 0
0 2 4 6 8 10 12 14 16 18
x

F IGURE 15. PIV snapshots of the xz-plane with η = 1, xk = 166.4 for (a) Rek = 990,
(b) Rek = 1130 and (c) Rek = 1230.

disturbances. Supplementary movie 1, available at https://doi.org/10.1017/jfm.2018.211,


demonstrates this changeover even more descriptively. In this experiment, the water
channel velocity has been continuously increased to increase the Reynolds number.
The free-stream velocity has been measured simultaneously to the PIV measurements
with a hot-film probe and with this information, the instantaneous Reynolds number
can be provided. The movie shows clearly how the patches of velocity excess and
deficit alternate symmetrically in the subcritical regime and antisymmetrically in the
Experiments on global instability in roughness-induced transition 899

(a)

(b)

F IGURE 16. Potassium permanganate visualization of η = 1 with (a) Rek = 820 and
(b) Rek = 1080.

supercritical regime. Obviously there must be a changeover in the competition between


varicose and sinuous modes when the Reynolds number is increased. Moreover, this
changeover occurs close to the theoretical critical Reynolds number Rek,c,e = 1040 and
it can be deduced that this observation is due to the unstable global sinuous mode
found by Loiseau et al. (2014) for thin aspect ratios. To our knowledge, there has not
been any experimental observation of a sinuous disturbance behind isolated roughness
elements before, probably due to the strong influence of convective hairpin shedding.
At subcritical Reynolds numbers the varicose disturbance dominates because the
sinuous global mode is stable. Note that in a physical flow, the two modes coexist
and interact nonlinearly with each other. As a result, the sinuous mode can remain
hidden in visual observations.
This is demonstrated in figure 16. The pictures show dye streakline visualizations
with potassium permanganate around and on top of the roughness with η = 1 for two
different Reynolds numbers. Hairpin vortices can be seen downstream of the roughness
in both the subcritical (figure 16a) and supercritical regimes (figure 16b) . Apparently,
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

the varicose hairpin shedding is still active in the supercritical regime despite the
globally unstable sinuous mode. Although the basic mechanism of the convective
varicose disturbance is different from the global sinuous mode, both disturbances
grow to nonlinear saturation at high Reynolds numbers. When the critical Reynolds
number is exceeded, the roughness is both amplifier and wavemaker. As a result, there
is competition between globally unstable modes from the intrinsic system dynamics
and highly amplified external noise. Strong convective growth can outweigh the
temporal growth of global modes, and the sinuous oscillation predicted by Loiseau
et al. (2014) may thus only dominate in extremely low turbulence environments.
When looking closely at the legs of the hairpin vortices in figure 16(b), one can
see antisymmetric wiggling submerged in the overall varicose shedding. Similar
visualizations have been done for many different roughness elements of various
shapes, some of them available in Puckert et al. (2015), and in none of these cases
could a clear von Kármán-type vortex shedding be observed. Instead, the shedding
seems rather influenced by different modes.
The remaining question is whether the stable sinuous mode in the subcritical
regime can also be forced similarly to the varicose mode. This is demonstrated
900 D. K. Puckert and U. Rist

0.05
1 0.03

0.01
z 0
–0.01

–1 –0.03

–0.05
5 10 15 20 25 30
x

F IGURE 17. (Colour online) Varicose forcing with ff = 0.7 Hz (ω = 0.68) at Rek = 620,
η = 1.

1 0.05
0.03
0.01
z 0
–0.01
–0.03
–1 –0.05
5 10 15 20 25 30
x

F IGURE 18. (Colour online) Sinuous forcing with ff = 0.65 Hz (ω = 0.56) at Rek = 694,
η = 1.

in an experiment with symmetric and antisymmetric disturbance sources to force


both varicose and sinuous modes. The Reynolds numbers for this experiment are in
the subcritical regime and forcing frequencies were chosen according to expected
maximum amplification in theory. Either the horizontal or vertical disturbance
generator is installed at a time and the goal of this experiment is to demonstrate
the sensitivity of the subcritical flow to both varicose and sinuous modes. The
roughness with η = 1 is again placed at xk = 57.1 and both Reynolds numbers
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

and forcing amplitudes are carefully adjusted to obtain detectable disturbances. The
signal-to-noise ratio needs to be as high as possible, but too large Reynolds numbers
or forcing amplitudes can lead to uncontrolled vortex shedding which can disturb
phase-locked hot-film measurements. It turned out that antisymmetric forcing needs
to be much stronger than symmetric forcing due to shear sheltering at the edge of
the boundary layer (Hernon, Walsh & McEligot 2007). Higher excitation amplitudes
are thus required.
Figures 17 and 18 show the phase-locked measurements in xz-slices for the
symmetric and antisymmetric forcing, respectively. The spatial resolutions in the
x and z directions are 0.5 and 0.25, respectively, and the measurement time was
30 s at each position, adding up to a total measurement time of 14 h. Figures 17
and 18 show remarkably well that both sinuous and varicose modes can be forced if
parameters are chosen correctly. For the varicose mode this was expected; however,
forcing of the sinuous mode is a relatively new technique (see Shin et al. 2015).
With this method it could be demonstrated that von Kármán-type instabilities grow
when the respective forcing is strong enough. Sinuous modes can thus also develop
convectively, similarly to the varicose disturbances which account for hairpin vortex
Experiments on global instability in roughness-induced transition 901
shedding. Although theoretical data of these cases are not at hand, there is qualitative
agreement with global mode shapes in Loiseau et al. (2014). Both forced convective
instabilities can (quasi-)resonate (Bucci et al. 2018) in the roughness wake. The
reason for the hairpin vortex shedding in most experimental facilities is therefore due
to (i) rather varicose background noise and (ii) higher growth rates of the varicose
mode behind the roughness. Nevertheless, the predicted sinuous mode can also be
found experimentally with the right environmental conditions.

7. Conclusions
The effects of 3-D cylindrical roughness elements embedded into a laminar
boundary layer have been investigated experimentally with particular focus on 3-D
global instability. Hot-film measurements revealed that the critical Reynolds number
from 3-D global stability theory can also be found experimentally. However, transition
does not occur at this critical Reynolds number. Instead, it already begins at lower
Reynolds numbers. This is in contrast to what has been reported by Loiseau et al.
(2014). It was shown that transition in the subcritical regime is due to convective
instabilities whereas the supercritical regime, i.e. beyond the critical Reynolds number,
is due to global instability. Further, it could be shown that both sinuous and varicose
disturbance modes exist and occur as predicted by theory. For thick roughness
elements, the natural disturbances of the unforced flow are symmetric with respect to
the spanwise coordinate for both sub- and supercritical conditions, being in agreement
with Loiseau et al. (2014). For thin roughness elements, competition between varicose
and sinuous modes was found in the proximity of the critical Reynolds number. In
the subcritical regime, the unforced flow is disturbed symmetrically by the roughness
whereas both antisymmetric wiggling and symmetric hairpin vortices occur in the
supercritical regime. This confirms that the roughness is solely an amplifier in the
subcritical regime and both amplifier and wavemaker in the supercritical regime. The
results from 3-D global LST could be successfully verified in a physical experiment:
the mode symmetry and critical Reynolds numbers show good agreement for all the
cases studied in this work. As a drawback, the stability of global modes itself is not
sufficient yet to fully predict the flow dynamics in highly convective configurations
such as in roughness wakes where transient effects can be significant. The critical
Reynolds number from theory can thus be understood as a parameter at which
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

instability occurs at the latest.

Acknowledgements
We thank J.-Ch. Robinet, S. Cherubini, J.-Ch. Loiseau and M. A. Bucci for their
fruitful cooperation and the anonymous reviewers for their comments. The financial
support of the Deutsche Forschungsgemeinschaft (DFG) under reference number RI
680/32-1 is gratefully acknowledged.

Supplementary movie
A supplementary movie is available at https://doi.org/10.1017/jfm.2018.211.

Appendix A. Parameters
Details of the parameter study in § 4.3 are given in table 3 (adapted from the
theoretical investigation of Loiseau et al. (2014)). Table 4 provides the corresponding
experimental parameters.
902 D. K. Puckert and U. Rist

η xk k/δ1 Rek,c,t Rekk,c,t


3 105.0 1.45 656 493
2 128.8 1.45 805 605
1 166.4 1.45 1040 781
0.85 145.6 1.45 910 683
3 (F) 57.1 1.83 564 489
TABLE 3. Parameters at the critical state in theory (Loiseau et al. 2014).

η xk k/δ1 Rek,c,e Rekk,c,e


3 105.0 1.46 660 497
2 128.8 1.47 820 620
1 166.4 1.42 1035 764
0.85 145.6 1.49 963 738
3 (F) 57.1 1.81 556 481
TABLE 4. Parameters at the critical state in the experiment.

REFERENCES

ACARLAR , M. & S MITH , C. 1987 A study of hairpin vortices in a laminar boundary-layer. Part 1.
Hairpin vortices generated by a hemisphere protuberance. J. Fluid Mech. 175, 1–41.
Å KERVIK , E., B RANDT, L., H ENNINGSON , D. S., H ŒPFFNER , J., M ARXEN , O. & S CHLATTER , P.
2006 Steady solutions of the Navier–Stokes equations by selective frequency damping. Phys.
Fluids 18 (6), 068102.
A NDERSSON , P., B ERGGREN , M. & H ENNINGSON , D. S. 2005 Optimal disturbances and bypass
transition in boundary layers. Phys. Fluids 11 (1), 134–150.
A NDERSSON , P., B RANDT, L., B OTTARO , A. & H ENNINGSON , D. S. 2001 On the breakdown of
boundary layer streaks. J. Fluid Mech. 428, 29–60.
A SAI , M., M INAGAWA , M. & N ISHIOKA , M. 2002 The instability and breakdown of a near-wall
low-speed streak. J. Fluid Mech. 455, 289–314.
BAGHERI , S., A KERVIK , E., B RANDT, L. & H ENNINGSON , D. S. 2009 Matrix-free methods for the
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

stability and control of boundary layers. AIAA J. 47 (5), 1057–1068.


B RANDT, L., C OSSU , C., C HOMAZ , J.-M. & H UERRE , P. 2003 On the convectively unstable nature
of optimal streaks in boundary layers. J. Fluid Mech. 485, 221–242.
B UCCI , M. A., P UCKERT, D. K., A NDRIANO , C., L OISEAU , J.-C., C HERUBINI , S., ROBINET, J.-C.
& R IST, U. 2018 Roughness-induced transition by quasi-resonance of a varicose global mode.
J. Fluid Mech. 836, 167–191.
C ARMICHAEL , B. H. 1958 Prediction of critical Reynolds numbers for single three-dimensional
roughness elements. Tech. Rep. N.A.I.-58-412 (BLC109), Northrop Aircraft Inc.
C HERUBINI , S., D E T ULLIO , M. D., D E PALMA , P. & PASCAZIO , G. 2013 Transient growth in the
flow past a three-dimensional smooth roughness element. J. Fluid Mech. 724, 642–670.
C HOMAZ , J.-M. 2005 Global instabilities in spatially developing flows: non-normality and nonlinearity.
Annu. Rev. Fluid Mech. 37, 357–392.
C HOMAZ , J.-M., H UERRE , P. & R EDEKOPP, L. G. 1991 A frequency selection criterion in spatially
developing flows. Stud. Appl. Maths 84, 119–144.
D ENISSEN , N. A. & W HITE , E. B. 2013 Secondary instability of roughness-induced transient growth.
Phys. Fluids 25 (11), 114108.
VON D OENHOFF , A. E. & B RASLOW, A. L. 1961 The effect of distributed surface roughness on
laminar flow. Boundary Layer Control 2, 657–681.
Experiments on global instability in roughness-induced transition 903
E RGIN , F. G. & W HITE , E. B. 2006 Unsteady and transitional flows behind roughness elements.
AIAA J. 44 (11), 2504–2514.
F RANSSON , J. H. M., B RANDT, L., TALAMELLI , A. & C OSSU , C. 2005 Experimental study of the
stabilization of Tollmien–Schlichting waves by finite amplitude streaks. Phys. Fluids 17 (5),
054110.
G ONDRET, P., E RN , P., M EIGNIN , L. & R ABAUD , M. 1999 Experimental evidence of a nonlinear
transition from convective to absolute instability. Phys. Rev. Lett. 82, 1442–1445.
G REGORY, N. & WALKER , W. S. 1956 The effect on transition of isolated surface excrescence in
the boundary layer. ARC Reports & Memoranda no. 2779, Aeronautical Research Council.
HMSO, London.
H ALSTEAD , D. E., W ISLER , D. C., O KIISHI , T. H., WALKER , G. J., H ODSON , H. P. & S HIN , H.
W. 1995 Boundary layer development in axial compressors and turbines. Part 2. Compressors.
In ASME 1995 International Gas Turbine and Aeroengine Congress and Exposition. American
Society of Mechanical Engineers.
H ERNON , D., WALSH , E. J. & M C E LIGOT, D. M. 2007 Experimental investigation into the routes
to bypass transition and the shear-sheltering phenomenon. J. Fluid Mech. 591, 461–479.
H UERRE , P. & M ONKEWITZ , P. 1990 Local and global instabilities in spatially developing flows.
Annu. Rev. Fluid Mech. 22, 473–537.
K ACHANOV, Y. S. & M ICHALKE , A. 1994 Three-dimensional instability of flat-plate boundary layers:
theory and experiment. Eur. J. Mech. (B/Fluids) 13, 401–422.
K LANFER , L. & OWEN , P. R. 1953 The effect of isolated roughness on boundary layer transition.
RAE Tech. Memo. 355, Royal Aircraft Establishment.
K LEBANOFF , P. S., C LEVELAND , W. G. & T IDSTROM , K. D. 1992 On the evolution of a
turbulent boundary layer induced by a three-dimensional roughness element. J. Fluid Mech.
237, 101–187.
K LEBANOFF , P. S., S CHUBAUER , G. B. & T IDSTROM , K. D. 1955 Measurements of the effect
of two-dimensional and three-dimensional roughness elements on boundary-layer transition.
J. Aero. Sci. 22.
K LEBANOFF , P. S., T IDSTROM , K. D. & S ARGENT, L. M. 1962 The three-dimensional nature of
boundary-layer instability. J. Fluid Mech. 12, 1–34.
K URZ , H. B. E. & K LOKER , M. J. 2016 Mechanisms of flow tripping by discrete roughness elements
in a swept-wing boundary layer. J. Fluid Mech. 796, 158–194.
L ANDAHL , M. T. 1990 On sublayer streaks. J. Fluid Mech. 212, 593–614.
L ANG , M., R IST, U. & WAGNER , S. 2004 Investigations on controlled transition development in a
laminar separation bubble by means of LDA and PIV. Exp. Fluids 36 (1), 43–52.
L INGWOOD , R. J. 1996 An experimental study of absolute instability of the rotating-disk boundary-
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

layer flow. J. Fluid Mech. 314, 373–405.


L OFTIN , K. L. 1946 Effects of specific types of surface roughness on boundary-layer transition.
Tech. Rep. A.C.R. L5J29a, National Aeronautics and Space Administration.
L OISEAU , J. C., C HERUBINI , S., ROBINET, J. C. & L ERICHE , E. 2014 Investigation of the roughness-
induced transition: global stability analyses and direct numerical simulations. J. Fluid Mech.
760, 175–211.
DE L UCA , L. 1999 Experimental investigation of the global instability of plane sheet flows. J. Fluid
Mech. 399, 355–376.
L UO , J. W., B AI , J. & S HAO , J. H. 2006 Application of the wavelet transforms on axial strain
calculation in ultrasound elastography. Prog. Nat. Sci. 16 (9), 942–947.
M ANNEVILLE , P. 2004 Instabilities, Chaos and Turbulence: An Introduction to Nonlinear Dynamics
and Complex Systems. Imperial College Press.
M IRANDA -BAREA , A., FABRELLAS -G ARCIA , C., PARRAS , L. & DEL P INO , C. 2016 Spin-down in
rotating Hagen–Poiseuille flow: a simple criterion to detect the onset of absolute instabilities.
J. Fluid Mech. 793, 316–334.
M OCHIZUKI , M. 1961a Hot wire investigations of smoke patterns caused by a spherical roughness
element. Nat. Sci. Rep. Ochanomizu Univ. 12 (2), 87–101.
904 D. K. Puckert and U. Rist
M OCHIZUKI , M. 1961b Smoke observation on boundary layer transition caused by a spherical
roughness element. J. Phy. Soc. Japan 16, 995–1008.
M ONKEWITZ , P., H UERRE , P. & C HOMAZ , J.-M. 1993 Global linear stability analysis of weakly
non-parallel shear flows. J. Fluid Mech. 251, 1–20.
P ETERSON , J. B. J R & H ORTON , E. A. 1959 An investigation of the effect of a highly favorable
pressure gradient on boundary-layer transition as caused by various types of roughnesses on
a 10-foot-diameter hemisphere at subsonic speeds. Memo 2–8-59L, National Aeronautics and
Space Administration.
P UCKERT, D. K., D IETERLE , M. & R IST, U. 2017 Reduction of freestream turbulence at low velocities.
Exp. Fluids 58 (5), 45.
P UCKERT, D., S UBASI , A., R IST, U. & G UNES , H. 2015 Experimental investigations of critical
roughness heights in a laminar boundary layer. In The 13th International Symposium on Fluid
Control, Measurement and Visualization, FLUCOME2015, Doha, Qatar.
P UCKERT, D. K. & R IST, U. 2016 Transition downstream of an isolated cylindrical roughness element
on a flat plate boundary layer. In New Results in Numerical and Experimental Fluid Mechanics
XI (ed. A. Dillmann, G. Heller, E. Krämer, R. Radespiel & C. Wagner), Springer.
P UCKERT, D. K. & R IST, U. 2018 Global instability in a laminar boundary layer perturbed by an
isolated roughness element. Exp. Fluids 59 (3), 48.
R ESHOTKO , E. 2001 Transient growth: a factor in bypass transition. Phy. Fluids 13 (5), 1067–1075.
R IZZETTA , D. P. & V ISBAL , M. R. 2007 Direct numerical simulations of flow past an array of
distributed roughness elements. AIAA J. 45 (8), 1967–1976.
S CHILLER , L. 1932 Strömung in Rohren. Handbuch Experiment. 4.
S CHWARTZBERG , M. A. & B RASLOW, A. L. 1952 Experimental study of the effects of finite surface
disturbances and angle of attack on the laminar boundary layer of an NACA 64A010 airfoil
with area suction. Tech. Rep. 2796, National Aeronautics and Space Administration.
S HIN , Y. S., R IST, U. & K RÄMER , E. 2015 Stability of the laminar boundary-layer flow behind a
roughness element. Exp. Fluids 56 (1), 11.
S HRESTHA , K., PARRAS , L., D EL P INO , C. & S ANMIGUEL -ROJAS , E. 2012 Experimental evidence
of convective and absolute instabilities in rotating Hagen–Poiseuille flow. J. Fluid Mech. 716,
103–126.
S MITH , A. M. O. & C LUTTER , D. W. 1957 The smallest height of roughness capable of affecting
boundary-layer transition in low speed flow. Tech. Rep. ES 26803, Douglas Aircraft Co.
S REENIVASAN , K. R., S TRYKOWSKI , P. J. & O LINGER , D. J. 1987 Hopf bifurcation, Landau
equation and vortex shedding behind circular cylinders. In Proc. Forum Unsteady Flow Sep.
(ed. K. N. Ghia), vol. 52, pp. 1–13. American Society for Mechanical Engineers, Fluids
Engineering Division.
https://doi.org/10.1017/jfm.2018.211 Published online by Cambridge University Press

S TOTZ , S., WAKELAM , C. T., G UENDOGDU , Y. & N IEHUIS , R. 2014 Detection of boundary layer
transition on a low pressure turbine airfoil without separation using a Preston tube and hot
film anemometry. In 15th International Symposium of Transport Phenomena and Dynamics of
Rotating Machinery: ISROMAC-15.
TAIRA , K., B RUNTON , S. L., DAWSON , S., ROWLEY, C. W., C OLONIUS , T., M C K EON , B. J.,
S CHMIDT, O. T., G ORDEYEV, S., T HEOFILIS , V. & U KEILEY, L. S. 2017 Modal analysis of
fluid flows: an overview. arXiv:1702.01453.
TANI , I. 1961 Effect of two-dimensional and isolated roughness on laminar flow. Boundary Layer
Flow Control 2, 637–656.
TANI , I., K OMODA , H., K OMATSU , Y. & I UCHI , M. 1962 Boundary-layer transition by isolated
roughness. Aeron. Res. Inst. Univ. Tokyo Rep. 375, 129–143.
T HEOFILIS , V. 2011 Global linear instability. Annu. Rev. Fluid Mech. 43, 319–352.
W HITE , E. B. 2002 Transient growth of stationary disturbances in a flat plate boundary layer. Phys.
Fluids 14 (12), 4429–4439.
W IEGAND , T. 1996 Experimentelle Untersuchungen zum laminar-turbulenten Transitionsprozess eines
Wellenzugs in einer Plattengrenzschicht. PhD thesis, Universität Stuttgart.
Z HANG , J., X U , M., P OLLARD , A. & M I , J. 2013 Effects of external intermittency and mean shear
on the spectral inertial-range exponent in a turbulent square jet. Phys. Rev. E 87 (5), 053009.

You might also like