14:00
Transport and Mixing 4
14:00
15 mins
#497
|
Nearfield flow establishment in a Pure Coherent Shear source turbulent flow tunnel
Anoop Mohan Vijaya, Panchapakesan N. R.
Abstract: We describe a flow device - Pure Coherent Shear Source Turbulent flow tunnel, to study turbulent flows. The schematic diagram of the device is shown in Figure 1. The flow domain is a tunnel that has a square cross section of side $T$. A number of jets (Re=24000) are used to inject fluid into the domain which serve as the source of momentum and kinetic energy for establishing the flow. Fluid is removed from the domain by a number of openings at the same plane where the jets are situated. This plane is named as the Momentum source plane. The outlet flow conditions are adjusted such that the net mass efflux at the momentum source plane is negligible. This configuration enables the establishment of a turbulent flow with little mean advection. This is similar to a stirred tank approach for producing high Reynolds number controlled turbulent flow. The flow field in a preliminary configuration of the device was investigated by performing RANS simulations.
The flow behaviour near the momentum source plane was studied experimentally. Experiments were performed using stationary hotwires in the nearfield region. For the experimental mapping of the nearfield, the non-dimensional parameters pertaining to the geometry - $\alpha\equiv d/T$, the ratio of the effective jet-inlet diameter to the side of the cross section and $\beta\equiv \pi d^2/4T^2 $, the ratio of total jet-inlet area to the total cross section area, were unaltered.
The results identify a high intensity turbulence region in the bulk of the domain that evolves into a near homogeneous turbulent flow regime as we move downstream. Here, we focus on the flow structure in the advection dominated region near the momentum source plane. The evolution of individual jets, their interaction and entrainment pattern are studied.
|
14:15
15 mins
#363
|
Passive scalar dispersion and mixing in turbulence: Direct numerical simulations and modelling
Michel Orsi, Fabio Feraco, Massimo Marro, Maurizio Boffadossi, Duane Rosenberg, Raffaele Marino, Pietro Salizzoni
Abstract: Environmental concerns due to the spread of urban areas and human activity are increasingly demanding a better prediction of the impacts of hazardous gases released from a ground level or an elevated source. Detailed information of the statistics of the concentration are required to estimate the risks associated with dispersion of atmospheric contaminants. This need motivates studies devoted to the characterization and modelling of concentration statistics and mixing processes of pollutants in atmospheric flows.
We investigate the phenomena using state-of-the-art pseudo-spectral direct numerical simulations (DNS) [1] on isotropic grids of 5123 points of a decaying puff (Figure 1) in homogeneous isotropic turbulence (HIT). Then, we show that the statistics obtained from the simulations do seem to match those based on the measurements performed in the wind-tunnel with a continuous source in a turbulent boundary layer (i.e., an anisotropic inhomogeneous velocity field), the former being a proxy of the passive scalar carried in the tunnel as seen in a Lagrangian reference frame. Namely, we are comparing fixed-time Lagrangian statistics with fixed-point Eulerian ones (Figure 2).
In previous studies [2, 3] it has already been shown that a one-parameter Gamma distribution is a suitable model for the probability density function (PDF) of the passive scalar concentration disregarding the source conditions, appropriately modelling also its temporal behaviour through the different phases of the mixing processes: the meandering and the relative dispersion. Then, we test this distribution with the numerical results and we come to a model for the evolution in
time of the concentration fluctuations intensity ic = σc/c related to the mixing timescale τm (i.e., the characteristic time of the mixing processes), giving the opportunity to easily calculate the latter directly from the profile of ic. Finally, we show the validity of the model presented.
|
14:30
15 mins
#444
|
Lagrangian mixing in wall-bounded turbulence: a network perspective
Giovanni Iacobello, Stefania Scarsoglio, Hans Kuerten, Luca Ridolfi
Abstract: Turbulent mixing is undoubtedly a crucial mechanism in many natural phenomena and industrial applications involving turbulent flows. Due to its ability to enhance transport, turbulent mixing is a fundamental process to understand – among others – atmospheric dispersion, geophysical phenomena, as well as combustion mechanisms [1, 7]. In this work, to take advantage of the powerful capabilities of recent developments in network science, we propose a complex network-based approach to study turbulent mixing from a Lagrangian viewpoint [3]. In the last two decades, the application of complex networks has revealed several important insights in many research fields, including fluid flows [2, 4, 6, 8]. Here, we exploited a direct numerical simulation (DNS) of a fully-developed turbulent channel flow as a paradigm of possible applications. The DNS was run at Re_τ = H u_τ/ν = 950 for T = 15200ν/u_τ^2, where u_τ is the friction velocity, H is the half-channel height and ν is the kinematic viscosity [5]. A set of 100 × 100 fluid particles was initially released in the domain as a uniformly distributed grid in the plane (y^+, z^+) at x^+ = 0, where the streamwise, wall-normal and spanwise coordinates are (x^+, y^+, z^+), respectively. To build the network, particles were grouped into N_y = 100 wall-normal levels corresponding to the network nodes, each one comprising a subset of N_z = 100 particles. At any time, connections between particle pairs are active based on their spatial proximity: a particle i is connected to a particle j if i lies inside a reference ellipsoid centred in j, and vice versa (by symmetry). The ellipsoid was selected to consider the anisotropy of the flow: its semi-axes were set proportional to the average pairwise Euclidean distance between all particles, so that the effect of the streamwise dispersion on particle positions is captured. As a result, we obtained a time-varying network, in which nodes represent the y^+−levels, while link activation depends on the extent to which turbulent mixing affects particle dynamics. Since each node represents a set of N_z particles, each link is weighted, thus quantifying the interaction strength, E, as the total number of connections between particles in each node (see Fig.1a). The results show that the networks are fully able to capture the intensity of wall-normal mixing on particle dynamics, by highlighting characteristic mixing-advection regimes as well as the appearance of peculiar events (see Fig.1). Based on present findings, a promising tool is proposed, which can be exploited to extend the level of information of classical statistics for turbulence analysis.
|
14:45
15 mins
#591
|
Towards a simple mixing model for passive scalar transport using Hierarchical Parcel Swapping (HiPS)
Tommy Starick, David O. Lignell, Heiko Schmidt
Abstract: The accurate resolution of all scales in turbulent flows with passive scalar transport is a challenging task for most simulation approaches. The wide range of length and time scales and associated high computational costs often only allows simulations at low Schmidt numbers. The Hierarchical Parcel Swapping (HiPS) model [Kerstein:2013, Kerstein:2014} is a relatively novel and computationally efficient mixing model that belongs to the family of map-based stochastic models. HiPS uses a binary tree structure, where the state variables associated with fluid parcels reside at the base of the tree, as shown in Fig. 1 (node 7 to 14). Every level of the tree corresponds to a turbulent length and time scale. The turbulent advection is modeled by randomly selected swaps of sub-trees, which are sampled from a Poisson process at rates determined by the corresponding turbulent time scale. Adjacent fluid parcels are mixed either instantaneously or at rates consistent with the prevailing diffusion time scales. In Fig. 1, two swaps and a mixing event are shown. The first swap (blue nodes) is a permutation at node 0. The second swap (red node numbers) changes the proximity of parcel pairs and requires a subsequent mixing event. The simplicity and computational efficiency combined with a high resolution makes HiPS a particularly attractive model for flow simulations [Kerstein:2014] and even more for turbulent mixing [Kerstein:2013] of multiple scalars. In Lignell et al. [Lignell:2019], an alternative to instantaneous mixing is described and first qualitative results for the application of HiPS to turbulent reactive flows are provided.
In the talk, we present an overview of the HiPS model and discuss its application to mixing and transport of passive scalars. In Fig. 2, the initial distribution for the mixing of one scalar (top) and the temporal evolution of the RMS value (bottom) is shown for Schmidt numbers smaller than one. We provide a new model extension for Schmidt numbers greater than one and compare scalar statistics against existing reference data [Warhaft:2000].
|
15:00
15 mins
#415
|
Entrainment in non-Boussinesq jets
Mathieu Creyssels, Samuel Vaux, John Craske, Maarten van Reeuwijk, Pietro Salizzoni
Abstract: We investigate the dynamics of jets of light and heavy fluids within an unstratified still environment. The density of the jet is significantly smaller or significantly greater than the density of the ambient fluid. This kind of jets releases are commonly identified in the literature as "non-Boussinesq" jets. A main issue is to determine how the entrainment of the surrounding fluid by the jet is modified when the density ratio between the two fluids is highly increased or highly decreased. The entrainment is usually measured by the ratio (α) between the horizontal entrainment velocity and the vertical velocity of the jet averaged over the jet horizontal sections.
Early experiments from Ricou and Spalding have been interpreted as follows: the entrainment rate α might rescale on the square roots of the density ratio of the two fluids. Hence, for a light jet, α should be smaller than in a iso-density jet whereas for a heavy jet α should be increased compared to a "Boussinesq" jet. However, the dynamical effects that are responsible for this power-law dependence of the entrainment rate with the density ratio are still not elucidated.
To investigate this we have performed Large Eddy Simulations of light and heavy jets and the corresponding results are analysed following the method recently presented by Craske and Van Reeuwijk and Ezzamel et al.
Two main results have been obtained. First, the numerical simulations show that the entrainment coefficient varies as a power law of the density ratio but with a slope -1/3 which is completely different from the slope +1/2 traditionally given in the literature. Secondly, we present the effects of a decreasing or increasing density ratio on the kinetic energy budget within the jet and its relation to the entrainment coefficient.
|
15:15
15 mins
#610
|
Closure theory for particle clustering in turbulence
Taketo Ariki, Kyo Yoshida, Keigo Matsuda, Katsunori Yoshimatsu
Abstract: A self-consistent moment closure theory is applied to particle clustering in turbulence, where autocorrelation of the number-density, e.g. pair-correlation function and radial distribution function, in the inertial range can be solved in a self-consistent manner. The obtained closure model explains an interscale transfer of the number-density variance in a similar manner to that of Kolmogorov-Obukhov-Corrsin theory for passive scalar turbulence, implying an invariant scale-similarity in particle's statistics.
|
15:30
15 mins
#177
|
Effects of grid resolution, source size and source elevation on large eddy simulation of plume dispersion in an infinite-Re neutral boundary layer
Hamidreza Ardeshiri, Massimo Cassiani, Soon-Young Park, Andreas Stohl, Kerstin Stebel, Ignacio Pisso, Anna Solvejg Dinger, Arve Kylling Arve Kylling, Norbert Schmidbauer
Abstract: A set of large eddy simulation (LES) experiments has been performed using the open source code
PALM. Firstly with a constant source size a wide range of grid refinements has been used to
investigate the effects of grid resolution on statistical characteristics of velocity and scalar fields for
a plume dispersing in the neutral boundary layer. For the velocity, first and second order moments
and turbulent structures are examined. Despite the good consistency of one point second order
velocity statistics, we found that the mean velocity profiles slightly changed with grid resolution,
with the highest resolution case showing the highest mean velocity. The two-point correlation
analysis showed also that the highest resolution simulation developed larger turbulent structures,
characterized by longer length and time scales. Our study also underlined the importance of
accounting for numerical dissipation for properly understanding the LES results. For the scalar, the
statistical moments up to the fourth order and the shape of the concentration probability density
function (PDF) are examined. Long averaging time has been used to reduce statistical errors. The
LES results are analyzed with the aid of published experimental wind tunnel measurements [1,2],
the phenomenological meandering plume model of Gifford [3], and published theoretical results on
the nature of mixing for a confined mixture [4]. Our analysis shows that the mean concentration is
influenced by grid resolution, with the highest resolution simulation showing lower mean
concentration, which is linked to larger turbulent structures. However, a clear tendency to
convergence of concentration variance is observed at the two higher resolutions. This behavior is
explained in detail by showing that the mechanisms driving the evolution of the mean and the
variance of the concentration are differently influenced by the grid resolution. Skewness and
kurtosis show that, irrespective of grid resolution, a family of Gamma PDFs represents the shape of
the concentration PDF in the dissipative phase of plume evolution. The most resolved simulation
demonstrates the highest level of realism and most closely match the wind tunnel observations.
Based on the highest resolution setting the effects of source size and source elevation on the plume
dispersion are also investigated. For an elevated sources, as expected, the mean concentration is
very weakly affected by changing the source size while higher level of scalar fluctuations are
observed for the smaller source. However, we found that for a ground level source the concentration
variance is weakly influenced by the source size. The Gamma PDF confirmed to be a valid model
for both elevated and ground level sources, irrespective of source size.
References
[1] C. Nironi, P. Salizzoni, M. Marro, P. Mejean, N. Grosjean and L. Soulhac. Dispersion of a passive scalar fluctuating plume in a turbulent
boundary layer. part I: Velocity and concentration measurements. Boundary-Layer Meteorol. 156, 415-446, 2015.
[2] J. E. Fackrell, and A. G. Robins. Concentration fluctuations and fluxes in plumes from point sources in a turbulent boundary layer. J. Fluid
Mech. 117, 1-26, 1982.
[3] F. A. Gifford. Statistical properties of a fluctuating plume dispersion model. Adv. Geophys. 6, 117-137, 1959.
[4] J. Duplat, and E. Villermaux. Mixing by random stirring in confined mixtures. J. Fluid Mech. 617, 51-86, 2008.
|
15:45
15 mins
#174
|
Preferential sampling of elastic chains in turbulent flows
Dario Vincenzi, Jason R. Picardo, Nairita Pal, Samriddhi Sankar Ray
Abstract: It is shown that an elastic chain in a turbulent flow preferential samples the vortical regions. The degree of preferential sampling and its dependence on the elasticity of the chain is quantified via the Okubo-Weiss parameter. The effect of modifying the deformability of the chain via the number of links that form it is also examined.
|
|