Visualization is a key component of the process of obtaining quantitative and mathematical knowledge regarding the numerical models investigated in CFD (Computational Fluid Dynamics) by sharpening the intuition of engineers and scientists [12]. The approach we take to visualization includes presenting data in a form where measurement (quantification) of relevant magnitudes or diagnostics is emphasized. This approach, called visiometrics [2], consists of several steps: identification, quantification and understanding/mathematization [3]. Although this process presents a natural progression, it is an iterative refinement procedure, whose end produces the appropriate mathematical formulation [4, 3].
In this work we are illustrating the use of diagnostics and
visiometrics to obtain insight into complex physical processes. This
involves analyzing both direct numerical simulations
(e.g. Navier-Stokes turbulence, 
mesh resolution), and also
simpler reduced models (mathematical abstractions with fewer
degrees of freedom that behave like the full equations in a restricted
parameter range or regime of validity, e.g. vortex dynamics via
Biot-Savart, single- and multi-filament simulations). In our
particular problem, observation of the turbulence simulations revealed
the presence of tube-like vortex structures. This motivated the study
of vortex interactions via vortex-filament models. After examining
these reduced models we proceeded to search the turbulence field for
the fundamental effects observed in the vortex-filament studies. In
the case of the reduced models, visiometrics helped us to characterize
the vorticity amplification process observed in the vortex collapse
phenomenon [5]. For investigating the dominant vortex
structures in isotropic 3D turbulence [6, 7], we
developed feature extraction/object identification algorithms, which
had to be efficient in dealing with the large amount of data generated
by the direct simulations.
We explore the use of both a massive parallel computer as well as high end workstations communicating in the frame of CMAVS/AVS networks and the use of simple feature extraction techniques and interactive post-processing on the workstation with our own Data, Visualization and Diagnostics (DAVID) environment. The examples reported below were obtained on the CM5 (1024 nodes) and the SGI Onyx at the Advanced Computing Laboratory at Los Alamos National Laboratory (ACL-LANL) and the CM5 (512 nodes) at the National Center for Supercomputing Applications (NCSA) and the SGI Onyx at the VIZLAB in Rutgers University.