Trains cross the railway bridge, what if there's no monitoring, and simple ways of cycle counting; would including the train's axles make correct sens !
If the influence line of the detail is small ( 4 m e.g. ) , the detail will see every axle causing one cycle. If it is long ( 100 m ) there will be only one cycle with each train.
This question is in the eurocodes treated with the lambda-1 factor. For road bridges, the eurocodes do not provide good rules for all cases to compute this factor, but only for usual cases.
I don't know if the eurocodes are all right for railway bridges.
In the general case, the solution is to cumule the fatigue damages : For the influence lines, the calculation leads to an history of stresses. This history of stress can be treated according to EN 1993-1-9 ( normative Annex A ) with
the Rainflow counting method. For this purpose you need as M. Suresh Bhalla wrote, records of the real traffic.
a way to address this is through the computation of an effective stress and using the rain flow counting method. The rain flow method is a convenient way of treating a complex stress history and reducing it to a simpler accounting of a number of smaller cycles of stress that can then be converted into an equivalent constant amplitude stress. This report http://onlinepubs.trb.org/onlinepubs/shrp2/SHRP2_S2-R19B-RW-1.pdf address the use of the rain flow method for the recent calibration of the AASHTO fatigue limit states. You can do it analytically by running a series of moving loads over an influence line and tracking the stress history (via moments).