Data Generation and Statistical Analysis

In 1941, Andrei N. Kolmogorov predicted that, within all highly turbulent flows, there is a universal energy-conserving cascade whereby the energy of the large-scale eddies is transferred to finer and finer scales, down to the scales at which the energy is finally dissipated to heat. It is difficult to measure such a cascade directly, but related benchmark predictions for the statistical behavior of turbulent flows can now be calculated and examined using advanced simulation and flow visualization tools. Los Alamos scientists have been able to simulate flows of Reynolds numbers up to 105, the largest of which needed of the order of terabytes of data storage and used the full power of the Advanced Simulation and Computing (ASC) Q machine for several weeks of computer time. Through clever analysis of single frames of the simulations, a great deal of information can be extracted to show that the original constraints for the Kolmogorov theory can be relaxed so that, in fact, his statistical predictions hold locally in time. Furthermore, scientists are able to measure new statistical quantities that demonstrate the conditions under which departures from Kolmogorov theory begin to occur. This type of statistical analysis of numerical data is setting the agenda for future research. Visualization of vorticity in a portion of a 2563 subdomain of the 20483 turbulence simulation performed on the ASC Q machine at Los Alamos.