Is it possible to calculate $B^*$ term in TLE data using only a few or more velocity and position vectors for a satellite?
Dwight E. Andersen claims that its an impossible task in his thesis "Computing Norad Mean Elements From a State Vector" (1994)
An extension of this [calculating other TLE elements than drag and mean motion] would be to compute, from just the state, the mean elements and the drag terms and B* in SGP4, SDP4, SGP8, and SDP8. Future research on satellite drag and methods of estimating the drag would be valuable. Since the drag term is a function of the physical geometry as well as atmospheric conditions, some aspects of the satellites' physical characteristics would be needed.