Hybrid RANS LES simulation of non-equilibrium boundary layers

View/ Open
Date
2007-04-25Author
De Prisco, Giuseppe
Advisor
Piomelli, Ugo
Metadata
Show full item recordAbstract
Hybrid simulations that couple the solution of the Reynolds-Averaged
Navier-Stokes equations (RANS) to Large-Eddy Simulations (LES) have
the ability to apply the high accuracy of LES only in regions of the
flow that demand it, while using the less expensive RANS approach in
regions of the flow where standard turbulence models are expected to
be accurate, while LES is used in non-equilibrium flow regions. One
issue that arises in these applications is the behavior of the flow in
the transition zone between the RANS and LES regions. In the RANS zone
the flow solution is either steady, or only contains information on
the largest scales of motion; most or all of the Reynolds shear stress
is provided by the turbulence model. In the LES region the resolved
scales must supply most of the Reynolds shear stress. Typically, a
transition zone exists in which the resolved, energy-containing eddies
are gradually generated and grow.
In this work, methodologies for the improvement of hybrid LES/RANS are
studied, to shorten the transition from the smooth RANS field to the
LES, which requires energy- and momentum-supporting eddies. The method
tested is based on the generation of synthetic turbulence with a
realistic spectrum, and statistics obtained from the RANS. The eddies
thus generated are then selectively forced by a control method that
amplifies the bursts, and maintains a desired Reynolds shear stress
downstream of the RANS/LES interface. This method allows to match the
two techniques smoothly, and to minimize the extent of the region
required to develop the realistic turbulent eddies. It was found to
perform well in several non-equilibrium boundary layers achieved by
imposing either a variable freestream velocity or a spanwise pressure
gradient on a flat-plate boundary layer. A finely resolved LES of
the flow in a boundary layer subjected to strong acceleration was also
performed. The flow in this configuration reverts to a laminar state
and then retransitions to turbulence. Statistics and flow
visualizations of this flow indicate the presence of two of the
mechanisms that have been conjectured to cause the re-laminarization.