A turbulent methane flame simulation. Flame simulations help build new
technologies that reduce air pollution and allow industry to meet clean
air requirements. In spite of their prevalence and technological importance,
our knowledge of combustion processes is surprisingly incomplete. Burners
are used in water heaters, HVAC systems, stoves, clothes dryers, as
well as many other industrial and commercial products.
The Scientist
Marc Day is a Staff Scientist at Lawrence Berkeley National Laboratory
(LBNL). He works in The Center for Computational Sciences and Engineering
(CCSE), led by Dr. John Bell. CCSE is an applied mathematics group
that focuses on large-scale parallel simulation of complex fluid flows.
CCSE's work is primarily sponsored by the Department of Energy's (DoE)
Mathematical, Information, and Computational Sciences Program. Over the
last three years much of Marc's work has been supported by the DoE's
SciDAC: Scientific Discovery through Advanced Computing program.
The Animation
The movie shows a ten centimeter nozzle emitting a turbulent mixture
of methane and air. Blue temperature iso-surfaces help visualize the
methane flame. A thin rod spans the nozzle and creates a small flame-stablilizing
recirculation zone. The animation represents the last 21 milliseconds
of simulation time, after the flame has reached statistical equilibrium.
The green and pink blocks in the animation are part of the adaptive
grid. The basic concept of adaptive meshes is to group cells into rectangular
regions that get further refined until the entire problem has adequate
resolution. The grid hierarchy moves with the flame as the simulation
integrates.
The green grid blocks are the first level of refinement, and are placed
in the flame's surface and turbulent pipe inflow regions. The pink blocks
represent the next level of refinement. One finer level of blocks exists,
but is not displayed in the animation.
The data was generated over several months using 128 processors of the
IBM SP3 at LBNL's National Energy Research Scientific Computing Center.
Approximately 200 microseconds of physical simulation time is computed
per wall-clock hour. The entire simulation, including a refinement study
performed to ensure numerical accuracy, generated approximately six terabytes
of data.
To make the animation, he processed the remote data and extracted a
sequence of 91 iso-surfaces, each at about 22 MB, and then brought them
back for local processing with Tecplot. He built a Tecplot layout, and
used Tecplot's animation tools to generate the movie.
The Simulation
Numerical simulation advances combustion science - closing the gap between
theory and experiment. Theoretical combustion science provides a foundation
for basic flame analysis, but is unable to address the complexity of
realistic flames. Laboratory measurements are difficult to interpret
and are limited in the level of detail.
Flame simulations provide specifics about 3-D flame morphology and statistics,
including information about combustion processes and the formation of
emissions (NOx and unburned hydrocarbons). With this data, designers
can construct better clean-burning combustion systems.
This simulation is performed using a dynamically adaptive low Mach number
reacting flow model. It incorporates detailed chemistry (20 species,
84 reactions) and preferential species diffusion using a low Mach number
approach that analytically filters acoustic waves from the system.
Sound waves have negligible effects on low-speed flows such as this,
but lead to severe time-step restrictions for numerical integration in
direct numerical simulation. Marc and his colleagues have developed a
particularly robust discretization for variable-density low Mach number
flows. They are implemented within a block-structured adaptive mesh refinement
framework that operates on distributed-memory parallel hardware.
A 2-D fuel consumption contour plot. The fuel consumption peaks at the
flame surface. The second image zooms-in on the flame surface where the
grid is most refined and the action takes place. The streamlines in the
zoomed plot show local advection near the surface. Three normals are
drawn perpendicular to the flame and extract fluid and chemical properties.
The flame consumption rate along the normals is plotted in the upper
XY plot. The goal is to understand exactly what the flame is doing, and
how turbulence affects the combustion process. Gathering these types
of statistics is a crucial aspect of Marc's investigation. "The
ability to easily make a wide variety of plots, from contours to vectors,
to streamlines, to XY-plots is invaluable," he noted.
Tecplot
Marc has been using Tecplot since Version 7.5. He uses Tecplot's full
range of capabilities including iso-surfaces, slices, contours, velocity
arrows, and streamtraces. He works with XY, 2- and 3-D data, as well
as derived data that grows to much larger dimensions.
According to Marc, Tecplot's three greatest strengths as it relates
to his work are:
1. Flexibility and painless data input/output.
2. Easily adjusting details like axes and fonts - then saving, reusing and
adjusting these attributes for quick viewing or publication-quality output.
3. Tecplot's continually evolving list of capabilities.
Prior to Tecplot, Marc relied on internal visualization tools. For 3-D,
this was limited to volume visualization and contour slices. For routine
plotting, he typically used PV-Wave, Mathematica, or free software like
xmgrace.
Marc adds, "There are a few tools that approach Tecplot for its
flexibility and ease of use. However, the ones I've used inevitably go
down the road of trying to be a programming or mathematical analysis
tools. The data visualization aspect takes back seat...Tecplot focuses
on visualization and data extraction."
"Tecplot software has provided an incredible boost to our productivity in terms of our research output and the insights we are able to obtain." Rajat Mittal, George Washington University
RSS Feeds
System Requirements
