Posted by Richard Phillips.
Brian presented work of modelling of a multi-physics phenomena. (i.e. When say thermal, mechanical, chemical, and electrical effects all important to some extent.)
An application was claimed with the physics field of tribology, but I don't know enough about that to comment. The main theme was really modeling methodology.
He came up with a hybrid model using traditional finite difference and cellular automata.
He described it. One key element is seperating each of the processes and then applying them sequentially. He said that's easier to do in dicrete time than for continuous time. That seemed like a good practical modelling advantage of NKS type rules.
He gave criteria for the rules to satisfy, of physical realism, computational explicitness, and numerical stability.
He covered each type of process in turn: advection, reaction kinetics, ..., giving CA type rules for each. The diffusion rule was the known swap-neighbors rule.
Each of his rules was just one or two lines of Mathematica code, so this certainly qualifies as NKS rules.
For diffusion, he claimed roughly that for diffusivity contants below .2 the diffusion model works well for his purposes.
Then he gave results for combined advection/diffusion and external interactions.
When he added in reaction kinetics the results were interesting.
Then he discussed how changing the order of each type of process can affect the results, but he implied that with care it doesn't matter.
His summary was that he took a practical approach using techniques that were easy to work with and cascading various rules for each type of effect.
He now has a library of mix-and-match rules that he can use for new phenomenon.
He said a key thing he got from the NKS book was to free up his thinking from only considering differential equations to considering more general rules.
Both Brian and the previous speaker Vallorie had obviously gone through a process of realization that these new approach to making models could work.