Search for information on the Davis formula. This is the most commonly used one.
Factors include type of bearing, flange contact, weight on axles, and wind resistance. Different factors are used for leading and trailing cars. It is mostly experimentally derived from train data.
We make rr maintenance equipment. I have found davis to be optimistic (lower than actual rollng resistance) for lower speeds, smaller wheels, and for equipment with lots of stuff hanging out all over (ie. no body or shrouds). It has as expected worked well for higher speeds, larger wheels, and shapes and surfaces similar to car bodies.
kcj