CFX4.2 Pressure contour kink at symmetry wall.

I am trying to model Newtonian flow at approx. zero Reynolds number past a periodic array of right circular cylinders. Symmetry conditions are used so that only one half of one cylinder geometry is modelled. The no-slip velocity boundary condition is applied on the cylinder surface and pressure boundary conditions are applied at inlet and outlet of flow to specify a pressure drop.

A plot of the steady state results indicate a kink in the pressure contour near the wall of the symmetry boundary condition above the cylinder, when the pressure contour should be normal to the symmetry wall. This occurs for a range of cylinder area fractions and is not removed by mesh refinement. Default settings in CFX version 4.2 are being used for the equation solver, pressure correction algorithm, differencing scheme etc.

The mean velocity result at the inlet is used to determine the non-dimensional drag past the cylinder. The results compare with analytical findings and therefore appear to be unaffected by this pressure kink.

Does anyone know why this has occurred and what changes I could make to the model setup to remove it?

Thank you in advance.

Stuart

(1). If you compute only a half of the flow field due to symmetry, then you have a symmetry boundary condition connecting from the hi-light point of the cylinder to the inlet and another symmetry boundary condition connecting the trailing edge of the cylinder (rear end) to the exit. (2). What is the type of boundary condition opposite to the symmetry condition? I mean on the opposite side of the half cylinder? (3). Where is this symmetry wall? Is this your upper boundary condition? (4). Since the pressure field is related to the velocity field, you can vary the Reynolds number to change the velocity, and then check the pressure contour plot. (do not change the mesh). (5). The other one you can do is to extend the inlet and the exit boundary further away from the half cylinder, and check the result. (you will have to change the geometry and the mesh). (6). Then, you can keep on running the code, until it is absolutely converged. Depending upon the method, the convergence of the pressure field sometimes can be very very slow. (not just slow).

Thank you for your reply.

The boundary condition opposite to the cylinder half is also a symmetry condition (symmetry wall).

I have increased the Re. number by a factor of 10 but the pressure contour kink still occurs.

I intend to perform additional studies with similar boundary conditions varying a number of parameters other than Re. number.

Do you think the kink could be related to the way the contour plotting packing interpolates data at the symmetry boundary?

Stuart

(1). It is possible. But you have already said that it doesn't go away with fine mesh. (2). If the problem is related to the mesh due to the contour plot program, then the problem area will change as you refine the mesh. (3). So, you can double check whether the trouble area actually reduces in size when the fine mesh in that area is used. (4). In most cases, the contour plot scheme is linear within a cell. It also will try to do different things if the cell is around the corner. A good contour plot code must do a lot of checking in order to arrive at the right result. Well, it is one area you can check.