Cylindrical lens in the right orientation should do the job... in case more complex distirtions are there - then one needs more complex prisms...

I want to design an interferometric system, it work with elliptical beam also but i want to know can we covert it in circular beam for getting good circular fringes.

I too suggest the coupling the light into a single mode fiber to get a Gaussian like beam profile. but , using two appropriate cylindrical lenses is also a good option. 

To achieve a symmetrical beam shape, the ratio of the focal length of the two lenses should be approximately equivalent to the ratio of the beam divergences:

θ1/θ2 = 10°/40° = f1/f2.

If the circular fringes are only the ultimate goal, in the  mach zehnder setup, in one arm ,  just expand the beam using cylindrical lens  of focal length f1, in the other arm, use another cylindrical lens of focal length f2, in perpendicular orientation and expand the beam. select a plane of interference so as the curvatures from both the beams coincides, in the overlapping region, one can see circular fringes.

@Peter Kapusta: May be, but one can look at for what purpose a circular profile beam is needed.

please look in to the method given at this link..

https://www.newport.com/n/beam-shaping-with-cylindrical-lenses.

http://www.techbriefs.com/component/content/article/ntb/features/feature-articles/12683

Yes using a single mode fiber for system is good option but it increase complexity of system also. So I think using a combination cylindrical lens separately or a laser diode with cylindrical lens. If somebody know cheaper laser diode with cylindrical lens mount than inform me and some other solution also.