Visualization of a jet in turbulent crossflow?

G. Araya1 , G. Marin2 , F. Cucchietti2 , I. Meta2 , and R. Grima2

1
Dept. of Mechanical Eng., University of Puerto Rico at Mayagüez, PR 00681, USA.
2
Barcelona Supercomputing Center (BSC), Barcelona 08034, Spain.
araya@mailaps.org
Abstract. Direct Numerical Simulation (DNS) with high spatial and
temporal resolution of a jet transversely issuing into a turbulent bound-
ary layer subject to very strong favorable pressure gradient (FPG) has
been performed. The analysis is done by prescribing accurate turbulent
information (instantaneous velocity and temperature) at the inlet of a
computational domain for simulations of spatially-developing turbulent
boundary layers based on the Dynamic Multiscale Approach (JFM, 670,
pp. 581-605, 2011). Scientific visualization of flow parameters is carried
out with the main purpose of gaining a better insight into the complex set
of vortical structures that emerge from the jet-crossflow interaction. An
interface has been created to convert the original binary output files by
the parallel flow solver PHASTA into readable input documents for Au-
todesk Maya software. Specifically, a set of scripts that create customized
Maya nCache files from structured datasets. Inside Maya, standard tools
and techniques, commonly utilized in feature film production, are used
to produce high-end renderings of the converted files. The major effect
of strong FPG on crossflow jets has been identified as a damping process
of the counter-rotating vortex pair system (CVP).
Keywords: Fluid dynamics · DNS · HPC · Data visualization.

1 Introduction
Incompressible jets transversely issuing into a spatially-developing turbulent
boundary layer is one of the most challenging types of three dimensional flows
due to its thermal-fluid complexity and technological applications; for instance,
film cooling of turbine blades, chimney plumes, fuel injection, etc. The capability
to control a flow field in such a way to enhance thermal efficiency is of crucial
relevance in aerospace and other engineering applications. A classical example
of active flow control by three-dimensional local blowing perturbations is the jet
in crossflow. A complicated set of flow structures and vortex systems is gener-
ated by the interaction of the jet with the crossflow: the shear-layer vortices, the
counter-rotating vortex pair (CVP), the wake vortices and the horseshoe vortex.
These coherent structures have been the motivation of several studies by many
researchers. A recent comprehensive review was performed by Karagozian [1].
Coherent structures in such a complex environment and their interactions (tur-
bulent events) are better identified and visualized by DNS. At the beginning,
flow visualization by smoke and dye injection was the only technique available to
describe these coherent structures, which can be considered the building-blocks
of turbulent flows, [2]. Generally speaking and based on the premise “seeing
?
Supported by GECAT-NCSA
2 G. Araya, G. Marin, F. Cucchietti, I. Meta, and R. Grima

is believing”, visualization techniques have substantially evolved in the last few

decades spanning all disciplines. According to Friendly [3], scientific visualization
“is primarily concerned with the visualization of 3D+ phenomena (architectural,
meteorological, medical, biological, etc.), where the emphasis is on realistic ren-
derings of volumes, surfaces, illumination sources, and so forth, perhaps with a
dynamic (time) component.” In this regard, it is important to stress the relevance
of identifying the target audience: the constituent parts, or formal attributes, of
a visual product will change in relation to its purpose and intended audience. For
example, a visualization for a scientific publication may incorporate very tech-
nical annotations. On the contrary, a visualization intended as a dissemination
product for stakeholders or general public will not need as many technical de-
tails, or the color schemes would be chosen based on aesthetics rather than field
conventions. In this short article, we intend to create visual displays of crossflow
jet simulations oriented to the scientific community.
2 Approach and Outcome Discussion
The numerical tools to be employed in the present study are briefly discussed
below.
Turbulent Inflow Generation: Computationally speaking, it is very challenging
to capture the physics of unsteady spatially-developing turbulent boundary lay-
ers (SDTBL), for the following reasons: (i) the high resolution required to resolve
both large and small scales (Kolmogorov/Batchelor scales), (ii) the computa-
tional box must be large enough to appropriately capture the influence of the
large scale motions, and (iii) realistic time-dependent inflow turbulent condi-
tions must be prescribed. Therefore, we propose to use the inflow generation
method devised by Araya et al. [4], which is an improvement to the original
rescaling-recycling method by Lund et al.[5]. The seminal idea of the rescaling-
recycling method is to extract the flow solution (mean and fluctuating compo-
nents of the velocity and thermal fields) from a downstream plane (called “re-
cycle”) and after performing a transformation by means of scaling functions,
the transformed profiles are re-injected at the inlet plane, as seen in fig. 1.

The Flow Solver : To successfully

perform the proposed DNS, a
highly accurate, very efficient,
and highly scalable flow solver is
required. PHASTA is an open-
source, parallel, hierarchic (2nd
to 5th order accurate), adaptive,
stabilized (finite-element) tran-
sient analysis tool for the solution
of compressible or incompressible Fig. 1. Schematic of the computational domain.
flows (Jansen [6]). It has been extensively validated in a suite of DNS (velocity
and thermal boundary layers) [7, 4]. PHASTA has been carefully constructed for
parallel performance and scaling to 786,432 cores in Mira supercomputer.
Boundary Conditions: At the wall, the classical no-slip condition is imposed for
 Visualization of a jet in turbulent crossflow 3

velocities. An isothermal condition is prescribed for the temperature field at the

wall, which is assumed a passive scalar. The lateral boundary conditions are
handled via periodicity. The pressure is weakly prescribed at the outlet plane.
At the top inclined surface, the normal component of the velocity is prescribed
a zero value (streamline) and freestream value for temperature. The jet is at
a lower temperature than the wall and is modeled by imposing a wall-normal
parabolic laminar profile velocity at the surface, in a circle with a radius R. In
the present investigation, the radius is approximately half of the inlet boundary
layer thickness, i.e. R ≈ δinl /2. Therefore, the wall-normal velocity profile, V (r),
within the jet is as follows; V (r) = Vmax [1−(r/R)2 ], where r is the local distance
to the jet center and Vmax is the vertical velocity at the jet center. The velocity
ratio (VR) is defined as the ratio between Vmax and the incoming freestream ve-
locity, U∞ . We are considering a low velocity ratio of 0.5. The Reynolds number
based on the pipe diameter (2R) and Vmax is 1520, which demonstrates that the
pipe flow is laminar and the parabolic velocity profile is therefore quite appropri-
ate. The temperature of the jet (coolant) is prescribed as 60% of the freestream
temperature, T∞ . The number of grid points is roughly 4M. The case was run
in 192 processors, consuming about 24,000 CPU hours.
Scientific Visualization: We are using Autodesk Maya and Paraview as the main
computer graphics programs for 3D animations. The Paraview toolkit is utilized
in principal domain cut and data interpolation in order to visualize a particu-
lar zone of interest, for instance, the vertical jet region. Fig. 2 shows a lateral
view of the computational domain and the blue volume contains the jet, whose
flow parameters (velocity and temperature) have been dynamically extracted.
In our pipeline, the data is first converted to a structured grid inside Paraview.
The outcome is then parsed by a proprietary Maya plug-in developed at the BSC
that reads structured-grid based simulation datasets, and converts them to Maya
nCache binary files, the standard simulation cache files of Autodesk Maya. This
tool provides an easy and consistent way to load three dimensional simulation
data in Maya, where one can apply advanced animation, shading, and rendering
techniques to create high-end visualizations. The color scales are variations of
standards in the field, which are also perceptually correct: a divergent ice-fire
scale, and the magma color scale. The high-end renderings obtained in Maya
were composited, edited, and color-corrected in post-production using the soft-
ware Adobe After Effects. For example, the titles, color scales, and annotations
were added in this final stage.
Major Outcomes: A snapshot of the video can be seen in fig. 3, which is avail-
able at https://vimeo.com/268976317. During the start-up process of the jet-
crossflow problem, it can be observed a weak penetration and low thermal mix-
ing of the vertical jet into the boundary layer due to the low velocity ratio
prescribed. Furthermore, visualization of instantaneous velocity (colored by in-
stantaneous temperature) reveals that the jet-crossflow interaction generates a
pulsating wake (the counter-rotating vortex pair system or CVP) downstream
of the hole. This evident coherent structure exhibits a quick attenuation due to
the strong flow acceleration.
4 G. Araya, G. Marin, F. Cucchietti, I. Meta, and R. Grima

Fig. 2. Domain cut and variable interpolation in Paraview.

Fig. 3. Start-up of a crossflow jet.

Conclusions: Flow animation of a cold jet interacting with a turbulent crossflow
is carried out. The data is first transformed to a structured grid inside Paraview,
and later converted to Maya nCache binary files via a Maya plug-in developed
at the BSC. The jet wake is weakened downstream, which is attributed to the
strong favorable pressure gradient imposed. Future work involves flow visualiza-
tion of crossflow jet simulations in large scale systems at much larger Reynolds
numbers.
Acknowledgment: This project is supported by subaward #074984-16663
(GECAT - University of Illinois). GA acknowledges XSEDE (Project Number:
CTS170006) and NSF-CBET grant #1512393.
References
1. A. Karagozian, “The jet in crossflow,” Physics of Fluids, vol. 26, no. 101303, pp.
1–17, 2014.
2. D. J. Dennis, “Coherent structures in wall-bounded turbulence,” Anais da Academia
Brasileira de Ciencias, vol. 87, pp. 1161 – 1193, 06 2015.
3. M. Friendly, “Milestones in the history of thematic cartography, statistical graphics,
and data visualization,” York University, Dept. of Mathematics and Statistics, 2009.
4. G. Araya, L. Castillo, C. Meneveau, and K. Jansen, “ A dynamic multi-scale ap-
proach for turbulent inflow boundary conditions in spatially evolving flows,” Journal
of Fluid Mechanics, vol. 670, pp. 518–605, 2011.
5. T. Lund, X. Wu, and K. Squires, “Generation of turbulent inflow data for spatially-
developing boundary layer simulations,” J. of Comp. Physics, vol. 140, no. 2, pp.
233–258, 1998.
6. K. E. Jansen, “A stabilized finite element method for computing turbulence,” Comp.
Meth. Appl. Mech. Engng., vol. 174, pp. 299–317, 1999.
7. G. Araya, C. Castillo, and F. Hussain, “The log behaviour of the Reynolds shear
stress in accelerating turbulent boundary layers,” Journal of Fluid Mechanics, vol.
775, pp. 189–200, 2015.