Hello!

I am modelling flow over a sphere with a Re(p)=0.5 and have used many different grids producing roughly the same erronuous result. The flow is in a pipe and the sphere is in the center with no surrounding boundry closer than 200r. The sphere has approx 6000 mesh faces and is scaled to 1mm in diameter. Picture a sphere within a sphere within a pipe within a pipe. The cells between the pipes were generated with Cooper, while all other meshes where tetrahedral. The inner walls are declared as internal. The mesh which I am using now, consists of 14e-6 cells (mill.), 3e-6 faces, 355e-3 nodes. The press. force is 1.79e-9 and viscous force 3.81e-9. The total force should be 1/3 pressure and 2/3 viscous with Cd equal to about 52.1 (with Oseen) but I have 57. I have used meshes with many fewer cells with the same result; about 6-9% in error. I have used both segregated, which has shown to raise the viscous force proportionally to the pressure, and the coupled solver. In previous meshes, I have adapted the mesh after velocity gradient and pressure gradient and have ensured that the area normal to the flow is correct by projecting the surface on the plane with high resolution. The outer pipe wall was declared a moving boundry with 0.0005 m/s and the inlet 0.0005 m/s. Density 1000 viscosity 0.00100.

Does anyone have any suggestions?

Question 2. How should the flow be initialized, with 0.005 m/s in the entire field etc...?

Should have read:

1.4e6 cells, 2.9e6 faces, 3.5e5 nodes

(1). I think you have done a good job. (2).You could select other numerical schemes and check the result again. (3). You could run another case at Re=1.0 to see the trend, whether it is consistently higher or not.

Do you use double precision solver ?

Since your forces values are very small this discrepancy can be a round off error