The main thing to keep in mind is that the results are given at the nodes. So you want enough divisions so that there is a node close to where maximum result will occur.
Beam elements are the most accurate element type. (Okay, truss elements are also accurate.) For example, a cantilever beam (fixed on one end, nodal force on the other end) can give accurate results with just one element: the maximum displacement is at the free end, and the maximum stress is at the fixed end. The results will match the hand calculations. (I know this is true for linear stress, but I do not know for certain if MES gives the same accuracy.) Of course, with only one element, the displaced shape will look a little bit odd. So in a case where the maximum stress is expected to occur "somewhere" between the joints or supports, you will want to have enough divisions to capture a reasonable result.
I have not tested the following for a long time, but I think that gravity load is approximated by putting half of the element weight at each end of the element. So if gravity were a significant load, then more elements would be recommended to get a more accurate representation.
Usually, shorter beams are better according to the general rule on mesh size.
May I know why you ask this? Is your model sensitive to beam length?
To clarify one thing, which might relate to what John mentioned here.
In ASIM2013, we fixed a bug about beam reaction force calculation in MES when gravity is applied. Right now, the reaction force should be right, no matter how many divisions the whole beam structure is cut into.
I am working an cabin for aerial ropeway. I am model them with beam elements. dimensions width 2000 x depht 1400 x height 2300 mm. add picture
I guess:
20 divisions for long beams, 5 for middle ones, and 2 for short ones, roughly.