rectangular natural convection loop
 

I was trying to do 2d modelling of a rectangular natural convection loop. It has two vertical and two horizontal limbs connected togetehr to form a rectangular loop. The heating section on left vertical limb has a constant heat flux maintained along a length of 75 cm and the cooling section on right vertical limb is assumed to be maintained at a constant temperature of around 15 deg. C. All tubes are of 16 mm diameter. The heating and cooling section is made of copper tube and at other places glass tubes are used. An adiabatic boundary is assumed at those places. The flow is expected to be in the clockwise direction.
We tried to model this system using steady state 2d laminar pressure driven flow with boussinesq approximation in fluent. Rectangular coordinate system was used. Coupled solution of Momentum and energy were employed. But we were not able to get a proper solution.

We are thinking that this is because of the rectangular coordinate system. When we used rectangular coordinate system we were able to mention initial values of ‘u’ velocity as zero and a value obtained from measurements was assigned for ‘v’ velocity. But we found that this resulted in a vertical component of velocity in the horizontal section.
Is it possible to solve this using the cylindrical coordinate system with one coordinate axis always along the axis of the tube at vertical and horizontal section?

Also do we need to consider the source term?