Time-curvature and Time-torsion of Virtual Bubbles as Fluid Mixing Indicators

Massive data sets describing vector fields from computational fluid dynamic simulations on high-performance computers can no longer be visualized directly by displaying the data values at each point in time and space, but require data reduction to analyze the essential properties. We describe early results of work in progress to visualize a dataset of 500GB of raw data, consisting out of multiple time steps from a CFD simulation describing the mixing of fluids within a Stirred tank. The objective is to assess the quality of the mixing of the fluids after some time. We compute pathlines with out-of-core memory management to handle the massive dataset on a single desktop machine and use these to trace the evolution of user-specified initial geometric structures. On top of these structures we compute quantities indicating the mixture, such as curvature and torsion of the pathlines as information complementary to the intrinsic curvature of the evolving time surfaces.