14:00
Wall Bounded Turbulence 5
14:00
15 mins
#397
|
Investigation of dynamics of secondary currents in marginally turbulent semi-filled pipe flow
Julian Brosda, Michael Manhart
Abstract: In order to investigate the generation and sustaining mechanism of secondary currents in semi-filled pipe flows, we examine the flow using Proper Orthogonal Decomposition (POD). The flow is simulated by a direct numerical simulation (DNS). There is still an ongoing debate on the generation mechanism of secondary flow of Prandtl's second kind. Explanations are offered on the one side by analysing averaged flow quantities: the budget of mean and turbulent kinetic energy and the mean vorticity equation \cite{Nikora2012}. On the other side by studying instantaneous flow fields: investigating the preferential locations and dynamics of coherent structures \cite{Sakai2016} and the different spatio-temporal energy modes of the flow by using POD \cite{Nawroth2015, Vanderwel2019}.
To investigate the POD modes, we carried out a DNS in a semi-filled pipe flow, with periodic boundaries in flow direction, no-slip condition at the wall and a slip condition at the top wall. The duct length is $L_x = 8 \pi H$. The Cartesian grid spacing corresponds to $\Delta x_1^+\approx 3.2$ in streamwise and to $\Delta x_2^+ = \Delta x_3^+ \approx 0.8$ in spanwise direction. The simulations were performed with a Reynolds number $Re_\tau = \frac{u_\tau \cdot H}{\nu} \approx 110$, with $u_\tau$ being the friction velocity and $H$ being the water depth. After reaching a statistically steady state the simulation was running for more than $5000 \cdot H /U_b$. So far $5000$ cross-sectional snapshots with all three instantaneous velocities were taken to perform the POD in order to investigate the spatio-temporal development of the flow.
The non-symmetric distribution of the mean streamwise velocity and the mean secondary streamlines (see Figure \ref{f:figure}, top left) show that there is a structure persisting for a long time in one side of the open duct. Nevertheless the instantaneous flow fields reveal a more complex flow field over the whole cross section. These slow dynamics we want to elaborate using the POD. The preliminary results show that the first to the sixth mode are covering together over 50 \% of the turbulent kinetic energy (see Figure \ref{f:figure}, top right). The zeroth spatial mode is basically showing the mean secondary flow (see Figure \ref{f:figure}, bottom left) with high energy on the left side, where a strong vortex pair with high velocities is located for the mean secondary flow. The first spatial mode shows three structures with a symmetric energy distribution (see Figure \ref{f:figure}, bottom right).
In our conference presentation we would like to discuss the mean secondary flow, Reynolds stress distributions and the dominant POD modes of the flow to better understand the flow dynamics in a semi-filled pipe flow.
|
14:15
15 mins
#413
|
Slip and transpiration velocity to model textured surfaces in turbulent channel flow
Simon Pasche, Ugis Lacis, Shervin Bagheri
Abstract: Rough surfaces induce not only slip velocity but also transpiration velocity. As one can expect, these conditions are specific to the roughness properties, and they can be described from a macroscopic point of view as feedback parameters from the boundary. It is an exciting property because first, if these feedback parameters can be predicted, the flow dynamics can be potentially controlled, and second these features can yield a natural way to characterise rough surfaces. We have developed a systematic approach to compute the slip and transpiration velocity from textured surface, and we want to investigate the potential of these feedback parameters to modify the friction drag of wall bounded flows.
Slip velocity is modelled by the common Navier slip boundary condition, which imposes a linear relation between the tangential velocity and the wall shear stress. This relationship is defined through a slip length L. Considering turbulent channel flow, Busse and Sandhamn 2012 show the potential of slip boundary condition to modify friction drag. It turns out that streamwise slip velocity is relevant for friction drag reduction, while spanwise slip increases the friction drag. Transpiration velocity, on contrary, was often neglegted because it represents a higher order correction of the boundary condition. Its impacts for rough surfaces has been highlighted by Gomez et al. 2018, while considering wall injection, normal velocity shows already its effectivness to control the boundary layer transition, Choi 1994. We will introduce a new boundary condition for the transpiration velocity associated with a transpiration length M, which is brought through physical arguments. We will show that it is connected to the slip velocity via mass conservation, Lacis 2018.
Preliminary results
We can evaluate the slip and transpiration length of a given microstructure like riblets or shark skin by solving a microscale problem. The results emphasize the flow sensitivity to the textured surfaces. The coefficients M and L computed for riblets, cube posts, cone posts and shark skin denticles are provided in the table 1. The flow field for the shark skin denticles is displayed in fig. 1, which has been used to compute the coefficients.
Streamwise riblets have the larger difference in slip lengths (L_xx-L_yy) reflecting the well known drag reduction effect from this microstructure. The largest coefficients are produced by the cone posts. These coefficients can vary significantly between different microstructures, cone posts have a larger slip length than a transpiration length, while cube posts have a larger transpiration length than a slip length, and approximately a factor four is observed between slip lengths. Finally, shark skin denticles have similar value for both slip and transpiration length in the streamwise direction and have a larger slip length than the cube posts.
We observe that the slip length is correlated with the surface area at the interface. Microstructures having a large surface touching the interface have a smaller slip length. A similar correlation for the transpiration velocity is not observed. Therefore, further investigations are required to understand the flow sensitivity with respect to transpiration length.
|
14:30
15 mins
#468
|
Predicting Particle Phase Velocity Statistics in a Sheared Turbulent suspension Using Fluctuating Force-Fluctuating Torque (F3TS) Model
Swagnik Ghosh, Partha Sarathi Goswami
Abstract: Particle-laden turbulent flows find application in wide range of industrial and natural processes, like fluidized bed reactors, pneumatic conveying of solids in chemical and pharmaceutical industries, drying operations, formation of aerosols and motion of pollutants in air and ocean. Though the advent of fast computing facility has enabled investigation of such flows through Direct Numerical Simulation (DNS) but still simulating such flows in case of practically applicable geometry is still far from the reality. Therefore modeling such flows are inevitable. Fluctuating Force Simulation (FFS) [1] is such a modeling methodology where the effect of fluid velocity fluctuations on the particle is modeled as anisotropic Gaussian White Noise. For dilute suspensions, strength of the noise is extracted from velocity-space diffusivity data of unladen fluid phase. In the Fluctuating Force Simulation, the inter-particle and wall-particle collisions were assumed to be perfectly smooth and elastic. Such an assumption is oversimplified. Here a more generalized collision rules are developed introducing co-efficient of restitution and roughness factor in hard sphere collision. Due to the introduction of roughness, rotational motion of the particles starts to play important role on the overall dynamics. Consequently the effect of fluid vorticity field on the particle phase becomes important. The fluctuations in turbulent fluid vorticity result in fluctuating torque acting on particles. Here in addition to the fluctuating force, a fluctuating torque has been included in Langevin type equation to describe the particle motion (F3TS model). Simulations are performed in horizontal turbulent shear flow in case of dilute suspensions for different roughness factor in the limit of high Stokes number. It is observed that roughness factor plays a very important role even in the dilute limit. The system size dependence of particle phase statistics have been investigated through simulations. Particle mean velocity and moments of the velocity fluctuations, and also the distribution functions have been predicted using F3TS model. The predicted results have been compared with DNS results with one-way coupling. The accuracy of the model was tested for different particle Stokes Number.
|
14:45
15 mins
#472
|
Structural effect on turbulent drag over porous media
Yuki Okazaki, Ayumi Shimizu, Yusuke Kuwata, Kazuhiko Suga
Abstract: PIV measurements of turbulent channel flow over isotropic porous media embedded acrylic plates are conducted at the bulk Reynolds number of 5000. With this arrangement, the overall wall normal permeability becomes a half of the streamwise permeability at the porous interface. It is found that turbulence becomes weak compared with that over the homogeneous porous wall, and the total wall friction becomes 13% smaller than average friction of the homogeneous porous walled channel and the smooth impermeable walled channel.
|
15:00
15 mins
#503
|
Investigation of the interaction between inner and outer region through the scaling of the streamwise turbulence intensity in a fully developed turbulent pipe flow
Lucia Mascotelli, Bharathram Ganapathisubramani, Gabriele Bellani, Jason Hearst, Eda Dogan, Spencer Zimmerman, Milad Samie, Xiao Bo Zheng, Alessandro Talamelli, Joseph Klewicki, Ivan Marusic, Nicholas Hutchins, Jason Monty
Abstract: Single-wire hot-wire measurements with full resolution were acquired in the Long Pipe at CICLoPE to investigate the interaction between the small and large scales in a fully developed turbulent pipe flow at high Reynolds numbers. The scale interaction, already noticed in [1] and [2], challenges the classical idea of scaling of the streamwise turbulence intensity, based on the full separation between motions at small and large scales. This possible connection introduces a Reynolds number dependency that was previously only ascribed to measurements uncertainty [3]. In order to shed light on this controversy the broadband turbulence intensity was filtered into the small-scale and large-scale contributions. The trend with Reynolds number is investigated to assess whether we can predict their contribution to skin friction. The cutoff frequency is chosen based on the recently proposed scaling law for the small-scale streamwise spectra for turbulent boundary layers [4] as the lower bound for the high-frequency regime where the law-of-the-wall is expected to hold. Furthermore, the effect of spatial filtering corrections is analysed and an empirical transfer function with filtering effect
is proposed. This was possible thanks to the exclusive characteristics of CICLoPE, that allowed in a previous experiment to acquire velocity fluctuations at high Reynolds numbers with full resolution. Here we present the results of wall-normal scans performed at the test section of the Long Pipe, located at x/D = 122.3m, spanning from the beneath the near-wall peak of the streamwise variance to y=0.9R (where R is the pipe radius), for friction Reynolds numbers ranging from 10000 to 40000.
|
15:15
15 mins
#455
|
QUANTIFYING AMPLITUDE MODULATION IN FLOWS WITHOUT CLEAR SCALE SEPARATION
Davide Gatti, Ricardo Vinuesa, Ramis Örlü, Philipp Schlatter
Abstract: The influence of large scales in the outer region of wall-bounded turbulent flows on the small-scale fluctuations in the near-wall region has been documented beyond doubt over the last decades \cite{Hutchins2007}. This top-down influence, better known under the term \emph{amplitude modulation} (AM), enhances with increasing Reynolds number ($Re$), and is thought to cause the inadequacy of inner scaling (or the $Re$-dependence) of the streamwise variance profile (among other quantities) in the inner region. In particular, the quantification of AM has caused some controversy, which is partially due to the lack of clear scale separation for the limited-$Re$-range numerical data sets and the associated ambiguity in decomposing the signal into large ($u_L$) and small-scale ($u_S$) components (see the review in Ref.~\cite{Dogan2019}). From a more physical point of view, the difficulty of relating the various AM measures to the physical phenomena of AM remains. A typical symptom of this inadequacy is the fact that we obtain nonzero AM measures in synthetic signals specifically constructed with a lack of large-scale coherence.
Here, on the other hand, we utilise a number of turbulent channel flows at a friction Reynolds number of $Re_\tau=1000$, in (periodic) narrow, short, wide and long domains to create different well-defined large-scale structures and address the aforementioned difficulties. The advantage of such numerical experiments becomes apparent in Figure~\ref{Fig1}, where the two-point covariance $C^*_{AM}$ of the small-scale energy ($u_S^2$) and the large-scale signal $u_L$ is depicted for a sufficiently long and wide domain (denoted \emph{full-size channel}) and a short and wide domain \emph{\`a la} Abe \emph{et al.}~\cite{Abe2018} (denoted \emph{minimal streamwise unit} (MSU)). The latter, in agreement with Ref.~\cite{Abe2018}, has a relatively larger streamwise energy, but above all depicts a clear and strong off-diagonal peak in the covariance map. This peak, contrary to its diagonal counterpart, implies a strong interaction of large scales in the overlap layer with the small scales in the near-wall region, and is non-existing in the reference channel.
|
15:30
15 mins
#467
|
Vertical coherence of turbulence in the atmospheric surface layer
Woutijn Baars, Dominik Krug, Nicholas Hutchins, Ivan Marusic
Abstract: Statistical descriptions of coherent flow motions in the atmospheric surface layer have many applications in wind engineering. For instance, the dynamical characteristics of large-scale motions in wall-turbulence play an important role in predicting the dynamical loads on buildings, or the fluctuations in the power distribution across wind farms. Davenport's seminal study on the subject (Davenport 1961) proposed a hypothesis that is still widely used to date. Here, we demonstrate how the empirical formulation of Davenport is consistent with the analysis of Baars et al. 2017 in the spirit of Townsend's attached eddy hypothesis in wall-turbulence. We further study stratification effects based on two-point measurements of the atmospheric boundary layer flow over the Utah salt flats. No self-similar scaling is observed in stable conditions, putting the application of both, Davenport's framework as well as the attached eddy hypothesis, in question for this case. Data obtained under unstable conditions exhibit clear self-similar scaling and our analysis reveals a strong sensitivity of the statistical aspect ratio of coherent structures (defined as the ratio of streamwise and wall-normal extent) to the magnitude of the stability parameter.
|
|