PetaApps Visualization: Fluid dynamics simuation visualization

People: Rohith MV and Chandra Kambhamettu

Collaborators: Lian Ping Wang

Numerical simulation of turbulence with suspended droplets using Petascale computing produces large amounts of output data in the form of fluid velocities, droplet positions. This data is difficult to analyze as numbers and hence an interactive visualization engine is required to display the data in a user-friendly manner enabling easy identification of physical phenomena. Our current focus is on development of visualizations which can aid in verification of four main phenomenon of interest: particle clustering, differential sedimentation, local shear and inertial/sling effects. The challenges encountered mainly stem from the large amounts of time-varying particle and fluid data generated from simulations. Various fluid properties are available for grid ranging from 256^3 to 512^3, with the goal of 1024 cubic volumes. The number of particles is in the order of 10^6 to be increased in the 1024 cubic simulations. Typically, 3000 or more time steps of data is available for visualization. Interactive visualizations with dynamic change of parameters (for e.g., isovalues, volume of interest) specifically targeted towards understanding of underlying physical phenomenon are currently being developed. Special event detection routines (collisions and other specified particle behaviors) will also be combined into the visualization for verification, dissemination and understanding of simulated effects.

Data such as fluid vorticity, velocity and acceleration are volumetric in nature and cannot be displayed in their entirety. Hence, to observe the interaction between fluid structures ad droplets, we need to select relevant portions of these volumetric data. We are currently working on methods to automatically identify such portions in the data adaptively as the particle moves through the fluid. Identification of structures corresponding to various phenomena and creation of a smooth transition between their rendering will lead to a coherent visualization of the simulation data.

Isosurface of fluid vorticity

Particle trajectory and particle-fluid interactions