Mobility/impedance computation methods, largely developed at Cornell, are used daily in bearing design analysis calculations by engine manufacturers throughout the world.
The methods, inverses of one another, are extremely fast and robust and run quite readily on small personal computers.
The secret is their utilization of stored nonlinear dynamic bearing characteristics, which are themselves the results of extensive (and slow) numerical solutions of nonlinear partial differential equations using massive supercomputers.
The process of data storage and retrieval is the subject of the project.
The best data currently available have been fit using polynomial regression. The resulting fits are accurate and can be computed rapidly, but they are simply 'ugly' and obscure whatever physical insights more 'elegant' fits might provide.
A solution appears to be at hand in the form of a powerful data mining program, Eureqa/Formulize, also developed at Cornell. Applying this available tool to raw mobility/impedance data sets should allow the user BOTH to develop truly elegant fits AND to discover underlying physical insights.
Successful completion of this project should lead to shared publication in an appropriate engineering journal.
Prof. Booker, jfb5@cornell.edu