| |
By André Bakker, Fluent Inc.
View the pdf of this article
FloWizard has a unique, wizard-based interface. Its innovative
GUI is designed to make it easy to set up and run CFD problems,
and to shield users from some of the more difficult modeling
decisions by making choices based on other user inputs. One
area where this happens is in modeling turbulent flow. FLUENT
offers many different turbulence models, each with its own specific
sub-models, settings, and constants. FloWizard greatly simplifies the
choices that the user has to make for setting up a turbulent flow. A
Flow Type Guide panel is available that clearly explains the available
options, which are laminar, turbulent, turbulent with strong swirl,
and unknown. The objectives are to keep the user from having to
make a decision based on esoteric turbulence modeling issues, and
to ensure the best possible solution.
The Flow Type Guide panel in FloWizard illustrates different
flow regimes
When the user chooses “turbulent”, FloWizard enables the realizable
k-ε turbulence model, which provides an excellent balance
between computational speed and solution quality. For applications
with strong swirl, FloWizard activates the Reynolds stress model, the
only available steady-state turbulence model that gives correct predictions
for such flows. To make this choice, it asks if the equipment
being modeled is purposely designed to create strong swirl, as
would be the case for a cyclone or a swirl combustor.
When the user chooses “unknown” in the Flow Guide panel,
FloWizard will choose a turbulent or laminar flow model based on
the inlet Reynolds numbers. For pressure driven flows, where this
cannot be calculated in advance, it will periodically recalculate the
inlet Reynolds number during convergence, and switch between
laminar and turbulent flow as needed.
All turbulence models in FloWizard are based on the steady-state,
Reynolds averaged Navier-Stokes (RANS) equations because of the
much longer computation times associated with transient models,
such as large eddy simulation (LES).
The predicted evolution of typical grid cell counts and turbulence
modeling approaches, based on the approximate average time
spent per simulation, shown for design engineers, analysts, and
researchers
In design, timeframes are much shorter than in engineering
analysis and scientific research. Researchers may accept CPU times
on the order of thousands of hours, employing either large computing
systems or patient students. Many detailed CFD analyses at commercial
companies will be limited to approximately 50 hours, which
corresponds to an overnight run on a typical four CPU cluster. By
contrast, design projects often require a review of multiple design
permutations in a single day, and are therefore restricted to computing
times of an hour or less for each run.
Based on these different requirements, and on the historical progression
in computer speed, a forecast can be created for the turbulence
modeling methods that will most likely be favored in different
types of CFD work in the years to come. While three-dimensional,
steady-state RANS models are usable for the typical design project
today, it will take several years before transient RANS and LES
methodologies will be widely used. In the meantime, the more computationally
intensive turbulence models will be used primarily in
analysis and research to perform tasks such as validation and process
optimization. One day, however, these techniques will become commonplace
for engineering design projects as well.
|
|
|
