Large - Eddy Simulation (LES) has served as a powerful tool for turbulence research over the last two decades. Computers with ever increasing speed and memory have also now made it possible to compute some turbulent flows using Direct Numerical Simulation (DNS). The research on LES of separated turbulent flows is proposed using parallel computing techniques. The study will determine the turbulence statistics for these flows. The proposed computation is very intensive in terms of memory and CPU time and, therefore, implementation on a multi-processor computer should lead to development of an efficient algorithm. The study will provide valuable experience and knowledge in the development of numerical algorithms for LES of separated turbulent flows, which can then be effectively applied to even more intensive computations using DNS of similar flows. To approximate the spatial derivatives, a combination of fourth- order finite-differences and spectral discretization will be used. Time derivatives will be approximated by an implicit method. The proposed methods will be implemented on CRAY- YMP8/864.
