Laminar flow in converging diverging channel


I am trying to simulate a channel flow in a channel that has an expansion and contraction. The setup is as follows:
u_inlet =0.01
aspect ratio 1:50

The channel consists of three sections. The first section is given by 0<=iX<=0.1Nx && 0.3Ny<=iY<=0.7Ny. The second section is given by 0.1<=iX<=0.2Nx && 0.1Ny<=iY<=0.9Ny.The first section is given by 0.2Nx<=iX<=Nx && 0.4Ny<=iY<=0.6Ny. I am setting Poiseuille Velocity conditions as described in the tutorial 1.5.(while also taking into account the different channel heights in each section as well as max velocity). The velocity profile seems to come out correctly but the density profile is not looking good. At the one end of the channel(the region from 0 to 0.6Nx roughly) it starts to increase and on the other end it is decreasing(0.6*Nx to Nx roughly). I tried to take one of the sections out and see how the results change. It seems that if I only have an expansion(so the domain consists of either either section 1 and 2 or section 3 and 2 or section 1 and 3) then the density profile looks very reasonable. If I only have a contraction (so the domain consists of either section 2 and 1 or section 3 and 2 or section 3 and 1) then the density variations are quite large. Does any one have any recommendations on how to properly setup this case in order to get the correct density profile?