In 2003, I worked on a UT/ORNL project (with Stuart Daw, Charles Finney and Sreekanth Pannala to characterize fluidization based on videos. I designed and implemented algorithms to extract velocity fields from high-speed videos of a small section of the bed by tracking individual particles through successive frames. This approach has uncovered patterns of particle movement other than monolithic bulk motion. The toolbox of routines comes with a GUI that is intuitive and easy to use.

The concept of extracting information about dynamical properties of the bed from direct observation is quite powerful and with proper computing facilities, quite feasible. It might provide a better understanding of fluidization (and perhaps other processes) by relying on experimental observations rather than on empirical or semi-empirical correlations.

I also worked with a Computational Fluid Dynamics (CFD) package called MFIX and used it to model fluidized bed behavior. After some trial and error, we were able to come up with the 'correct' parameters for MFIX that produced an output commensurate with experimentation.

Earlier, I had used the MFIX output of a fluidized bed to develop a regime detection algorithm that extracted essential information from the simulation frames to form a feature vector which was later classified by neural networks and clustering algorithms. I presented a paper based on my findings.