Abstract:
The flow past a right square cylinder in a duct at a Reynolds
number of 22 x 103 has been employed to validate the use of
second order upwinding, instead of a subgrid model in a largeeddy
simulation. In this extensively studied problem, all the
numerical work has been based on a simplifying assumption that
the square cylinder is infinite, which resulted in all previous
workers using cyclic boundary conditions so as to reduce the
required domain size. It is not clear how the size of the domain
had been established and, therefore, whether it was sufficiently
large to adequately represent the experimental flow in a duct.
The integral quantities of the drag and lift coefficient and the
Strouhal number, converged towards the experimental values as
the grid resolution is increased. However, the cyclic boundary
condition assumption leads to a flow width that provides too
small a region of uncorrelated flow. A model of the full duct
case, identical to experimental domain, was used to contrast the
cyclic domain results. Surprisingly the second order upwind
model generates power spectra that appear to correctly capture
the energy cascade down to the inertial and viscous ranges.
