Development of Compressible Mixing Layer Instability Simulated Using the Direct Simulation Monte Carlo Method

The Kelvin–Helmholtz instability developing in the mixing layer between two supersonic streams is simulated with the Direct Simulation Monte Carlo (DSMC) method using the SMILE-GPU software. No initial perturbations are introduced into the flow so that the disturbances are excited by and develop from the statistical fluctuations inherent in the DSMC method because of its stochastic nature. Multiple graphics processing units (GPUs) are employed for numerical simulations and efficient parallelization strategies for DSMC implementation on GPU clusters are presented. Between 0.4 and 1.6 billion of test particles are used to reproduce the development of the flow instability. The influence of the number of particles on mean flow properties and pulsation characteristics is investigated and discussed. It is shown that the pulsation characteristics are substantially affected by the number of particles because of a delay in the instability onset and vortex formation at a lower level of statistical fluctuations. A new algorithm for identification and analysis of vortex motion in noisy flow data is considered and applied to flowfields resulted from the unsteady DSMC simulations.