16:15
Wall Bounded Turbulence 3
16:15
15 mins
#203
|
Bifurcations of turbulent patterns in channel flow
Masaki Shimizu, Manneville Paul
Abstract: On its way to/from turbulence, channel flow displays a fluctuating pattern of laminar domains and turbulent bands. By direct numerical simulation we study the observed inconsistency between turbulence decay according a two-dimensional directed-percolation scenario and the presence of localized turbulent bands below its critical point. We point out a bifurcation restoring the statistical spanwise propagation symmetry of these turbulent bands, and show that the percolation-like transitional behavior is retrieved only somewhat above symmetry restoration.
|
16:30
15 mins
#351
|
SECONDARY FLOW GENERATION MECHANISMS IN TURBULENT SINUSOIDAL CHANNELS
Alvaro Vidal, Hassan Nagib, Philipp Schlatter, Ricardo Vinuesa
Abstract: In this contribution we study the secondary flow of Prandtl’s second kind present in spanwise-periodic channels with in-phase sinusoidal walls by means of direct numerical simulation (DNS). This work aims at extending our previous assessment of this type of secondary flow in ducted geometries. By analyzing cases in which the wavelength is much smaller than the half-height of the channel, we show that the cross-flow distribution depends almost entirely on the separation between the scales of the instantaneous vortices, where the upper and lower bounds are determined by lambda and lambda+, respectively. Therefore, the distribution of the secondary flow relative to the size of the wave at a given Reh can be replicated at higher Reynolds number by decreasing lambda and keeping lambda+ constant. The sweeps direct the instantaneous cross-flow from the core of the channel towards the wall, turning in the wall-tangent direction towards the peaks. The ejections drive the instantaneous cross flow from the near-wall region towards the core. This preferential behavior is identified as one of the main contributors to the secondary flow.
|
16:45
15 mins
#98
|
Structure of the skin-friction drag fluctuations in turbulent channel flows
Cheng Cheng, Weipeng Li, Lozano Duran Adrian, Hong Liu
Abstract: The ability to understand and predict skin-friction drag in turbulent
wall-bounded flows is fundamental for most engineering
applications. Skin-friction drag can be decomposed in mean and
fluctuating components. The latter, (Cf’), is of obvious importance
for noise radiation, structural vibration, drag properties, and wall
heat transfer mechanisms . Previous studies have reported that Cf’ is not only associated with the near-wall structures, but also linked to the large-scale energy-containing motions in the logarithmic and outer layers of the flow, especially at
high Reynolds number . In the present study,
the spatial structure of Cf’ and its relation to Townsend's
attached-eddy model are investigated following a
clustering methodology for turbulent channel
flows at $Re_{\tau}$ = 550, 950, and 2000 provided by Jim\'enez and
co-workers.
Depending on the sign of Cf’, the structures of intense Cf’ can
be classified as positive-friction events (PFs) and
negative-friction events (NFs). Detailed analysis of PFs and
NFs provides the following remarkable findings:
The length (lx) and width (lz) of the identified PFs
and NFs follow $lx^+=0.64(lz^{+})^{1.46}$, and NFs are more
likely to be organized as large-scale structures, as shown in figure
1(a,b).
The mean-squared skin-friction fluctuations,
$\overline{Cf'^{2}(lz)}$, carried by PFs and NFs are shown
in figure 1(c) conditioned to different widths. For $lz^+<600$,
$\overline{Cf'^{2}(lz)}$ contained in PFs is notably larger
than that of NFs, and for $lz^+>700$, $\overline{Cf'^{2}(lz)}$
carried by NFs tends to a constant value.
The present study suggests that NFs are the footprint of
wall-attached eddies and, as such, they may serve as indicators for
identifying Townsend's attached eddies in high-Reynolds-number wall
turbulence.
|
17:00
15 mins
#241
|
Mechanism of quasi-linear Orr burst in turbulent channel flows
Yongseok Kwon, Javier Jimenez
Abstract: Please find the attached file
|
17:15
15 mins
#380
|
DATA-DRIVEN QUANTIFICATION OF NONLINEAR INTERACTIONS IN THE RESOLVENT ANALYSIS OF TURBULENT CHANNEL FLOW
Ryan McMullen, Kevin Rosenberg, Beverley McKeon
Abstract: Tools from linear systems theory have enjoyed much recent success in modeling wall-bounded turbulence and other
complex flows. However, a complete model must be able to make predictions about the nonlinear interactions of various
modes. In the absence of a complete theoretical description, researchers have focused on various ways of augmenting
linear models with data, usually using some subset of the available statistics [1, 2, 5, 7, 8]. Such informed simple models
may serve as valuable tools for providing insight into complex nonlinear behavior. We extend the work of [5], who used
the resolvent analysis framework [4] and convex optimization to compute the resolvent mode weights that minimize the
deviation between a low-order model of the time-integrated velocity spectra and those computed from a direct numerical
simulation (DNS) of Re = 2003 channel flow [3]. Recently, it was shown that additionally decomposing the resolvent
operator into operators related to the classical Orr-Sommerfeld and Squire operators allows for particularly compact
representations of exact coherent states [6]. In the present work, we utilize this approach and show that it results in
drastic improvements in the matching of DNS statistics over the standard resolvent decomposition [5]. The results of
the optimization using three Orr-Sommerfeld and three Squire modes per wavenumber-frequency triplet are compared to
the DNS in Figure 1 in the form of premultiplied one-dimensional energy spectra. Even with relatively few modes, the
agreement between the model and DNS spectra is very good, and in particular, the peaks are captured almost exactly.
Integrating the spectra over all wavenumbers yields deviation errors of 4.3%, 0.95%, 0.66%, and 3.8% for hu2i+, hv2i+,
hw2i+, and huvi+, respectively. These should be compared with errors of 20%, 17%, 6%, and 25% using twice as
many modes per wavenumber-frequency triplet in [5]. Additionally, although matching with the DNS data is performed
in a time-integrated sense, access to the full three-dimensional model spectra is possible, providing valuable information
that can guide further model reduction efforts by, e.g. identifying regions of spectral space where energy content is sparse
in wavespeed. Finally, the decomposition into Orr-Sommerfeld and Squire modes has the potential to shed light on modal
interactions in the setting of turbulent channel flow.
The support of a VBFF Fellowship under US ONR grant #N00014-17-1-3022 is gratefully acknowledged.
|
17:30
15 mins
#99
|
Analysis of the skin-friction line structure in channel turbulence
Lipo Wang, Weipeng Li
Abstract: Wall turbulence has been among the key topics in fluid dynamics research. The near wall flow structure has significant influences on the wall turbulence dynamics. In the present work, we focus on the structure of the skin-friction line, i.e. the vectorline of the surface shear-stress vector, in channel turbulence. Such flat-plate configuration subjected to a preset pressure gradient was chosen in order to exclude any complex geometry effects. The vectorline segment method~\cite{streamline} is adopted here to investigate the nonlocal flow structure, quantitatively rather than illustratively. In a specified vector field, tracing from any nonsingular point, either along the vector direction or its inverse, one will reach a local extremum of the vector magnitude $\phi$ or other predefined scalar quantity. The part of the vectorline between the two local extrema is defined as the corresponding vectorline segment of the given spatial point. Each segment can be characterized with the arclength $l$ and $\Delta \phi$, the difference between two extrema of $\phi$ along the vector direction. Depending of the sign of $\Delta\phi$, segments can be positive ($\Delta\phi>0$) or negative ($\Delta\phi<0$). For the velocity and vorticity, the two vectors of essential importance, segments are then respectively named streamline segment and vorticity line segment.
\\A snapshot of the skin-friction line is shown in Fig.~\ref{total} (a), colored with the magnitude of the wall shear stress vector. Because of turbulence, the skin-friction lines appear quite irregular. Structural properties, e.g. the curvature, stretching ratio and the vector magnitude inherit important imprints of wall turbulence. Fig.~\ref{total} (b) shows the joint probability density function (PDF) between $\phi$ and $l$. Clearly the positive $\Delta \phi$ branch is smaller than the negative $\Delta\phi$ one, which means on average the positive segments have less velocity change and shorter arclength, and vice versa. In contrast, in isotropic turbulence statistically the positive streamline segments have larger velocity difference and larger length scale. More detailed analysis indicate that such asymmetry in Fig.~\ref{total} (b) is related to the turbulent eddy sweeping in the near wall region. Fig.~\ref{total} (c) presents the marginal PDF of $l$. The fact that at $l=0$ the PDF approaches to zero results from the molecular diffusion effect~\cite{DE}. For all simulated cases with different Reynolds numbers, these segment length PDFs overlap in the normalized scale $\tilde{l}=l/l_m$, where $l_m$ is the mean of $l$. Physically the extremal points are jointly determined by the turbulent perturbation for production and the molecular diffusion for removal. Based on a dynamic balance scenario between these two counteracting mechanisms, Wang and Peters~\cite{DE} first introduced a model equation of $\tilde{l}$. It proves that this equation's solution (solid line in Fig.~\ref{total} (c)) fits the vectorline segment result as well.
|
|