For approximate studies (estimation of mass flow rate and pressure loss), you could consider each bend as a pipe with a certain equivalent length; for example, for 90┬║ smooth bends, the equivalent pipe length is around 50 bend diameters (there is much literature available about this topic). Then, you could simulate the equivalent straight pipe with a 2D-axisymmetric model (or more simply, use the 1D momentum equation with Darcy-Weisbach + Colebrrok-White equations for the pressure loss, thus giving an analytic solution).
In any case, as points out, the error induced by the simplifications depends on the flow. In your particular case, there will be a value of Re from which the flow is separated in a region around the bends (and this flow is far from simple).