ENGINEERING AND CFD

The Boltzmann equation for the density distribution of particles in a fluid is the most fundamental equation of non-equilibrium gas kinetics and thus fluid motion. Due to its generality however, the numerical solution of the Boltzmann equation is only feasible for systems with low total particle count, e.g. rarefied gases.

For cases with higher particle counts, the equilibrium assumption is often valid, which allows the use of the well known equations of gasdynamics, the Navier-Stokes-Fourier system, which is the current state of the art model for turbulence fluid motion. In between these two regimes and overlapping them lies a physical regime where the validity of the equilibrium assumption is not clear-cut. In here, the numerical solution of the Boltzmann equation is still too computationally demanding, while the Navier-Stokes-Fourier system loses its validity. One way of tackling this problem is finding better mathematical descriptions of this regime, i.e. different sets of equations, often termed the "extended gasdynamics models''.

The goal of this work is to take a first significant step towards the evaluation of those extended gasdynamic models, such as e.g. the Grad 13 and the regularized Grad 13 equations, for the simulation and modeling of turbulent fluid motion.

The difficulty lies in the nature of turbulence: it is a highly chaotic phenomenon, which features a broad range of spatial and temporal scales, see Figure 1 for an example.

The approach in this project is to use direct numerical simulation of those model equations as a virtual experimental environment. To factor out numerical artifacts in the assessment of the results, a high resolution method was developed with a special emphasis on high performance computing. The multiscale nature of turbulence demands the largest available computing resources and simulations with O(10⁵) processors were performed to assess the difference on the JSC Blue-Gene system JUQUEEN.