Wrong definition of normals at edges of 3D problem

Numerical methods and mathematical models of Elmer
Post Reply
denis.cohen
Posts: 56
Joined: 15 Dec 2010, 13:50

Wrong definition of normals at edges of 3D problem

Post by denis.cohen »

Hello,

I am solving flow in a ring shear device (like a doughnut). Because of symmetry I can solve the flow over a pi/6 angle instead of the whole 2pi.
I am using perdiodic bc for in the inlet/outle and I am imposing the velocity at the top of the doughnut (see mesh in first attachment). My problem is that the normals at the edges of the 3d domain are not consistent with the flow field (see red circles in attachments 2 and 3) and thus the flow field is incorrect (slows down at the outer edge of the doughnut near the inlet and outlet).
How can I correct the definitions of the normals to get a consistent flow field?
Attached are the bc in my sif file. Mass Consistent Normals on or off does not seem to make any difference.

Thank you for your help.

Denis
Boundary Condition 1
Name = "Inner-Outer Walls"
Target Boundaries(4) = 6 7 9 10

Normal-Tangential Velocity = Logical True

Velocity 1 = Real 0.0

! Mass Consistent Normals = Logical True
End

!---------------------------------------------------
Boundary Condition 2
Name = "Inlet"
Target Boundaries = 5

!! Next line needed when periodic bc
Normal-Tangential Velocity = Logical True

! Mass Consistent Normals = Logical True
End

!---------------------------------------------------

Boundary Condition 3
Name = "Outlet"
Target Boundaries = 8

Normal-Tangential Velocity = Logical True

Periodic BC = 2
Periodic BC Rotate(3) = 0.0 0.0 $180/6
Periodic BC Velocity 1 = Logical True
Periodic BC Velocity 2 = Logical True
Periodic BC Velocity 3 = Logical True
Periodic BC Pressure = Logical True
Periodic BC Temperature = Logical True

! Mass Consistent Normals = Logical True
End

!---------------------------------------------------

Boundary Condition 4
Name = "Top"
Target Boundaries(2) = 3 4

Temperature = Real 10.0

Velocity 1 = Variable Coordinate
Real MATC "vtheta(0.0) * sqrt(pow(tx(0),2) + pow(tx(1),2)) * (-sin(atan(tx(1)/tx(0))))"
Velocity 2 = Variable Coordinate
Real MATC "vtheta(0.0) * sqrt(pow(tx(0),2) + pow(tx(1),2)) * (cos(atan(tx(1)/tx(0))))"
Velocity 3 = Real 0.0

End
!---------------------------------------------------

Boundary Condition 5
Name = "Bottom"
Target Boundaries(2) = 1 2

Temperature = Real 0.0

! Slip
Normal-Tangential Velocity = Logical True
Velocity 1 = 0.0

! Mass Consistent Normals = Logical True
End
Attachments
Red circles show where there is a problem with the definition of the normals at the edges of the domain
Red circles show where there is a problem with the definition of the normals at the edges of the domain
velo_edges_2.png (80.78 KiB) Viewed 1642 times
Red circles show where there is a problem with the definition of the normals at the edges of the domain
Red circles show where there is a problem with the definition of the normals at the edges of the domain
velo_edges_1.png (89.14 KiB) Viewed 1642 times
mesh + vector field
mesh + vector field
mesh.png (59.92 KiB) Viewed 1642 times
Post Reply