I am getting some streamlines normal to wall(postprocess :Matlab) and the normal velocity matrix from my Fortran code shows that in those gridpoints adjacent to wall the sense/sign of normal velocity is toward the wall ...

I wonder if the remedy is more resolution in grid toward wall or is something seriously wrong with my discretization and boundary conditions (e.g. violation of continuity in those regions)..

remedial suggestions will be of gr8 help....

If it is streamline in the 2D section of a 3D flow, then it is resonable, other wise there is error.

The streamlines are usually drawn as iso-lines of the stream-function field. Here, the property is that the flux between the two streamlines is always constant, i.e. they should not bump into walls. If you are running an FV-based code, you've got two representations of the velocity field: the cell-centred velocity and the face-centred flux. I know this makes it harder, but you should base the calculation of stream function beased on the flux. It is easy to check if your fluxes are OK: just calculate div(u) for an incompressible fluid - this should be zero to pressure solver tolerance.

There are several possibilities: 1) You are calculating the stream function based on the cell field and the error accumulates next to the wall. This will never work properly. 2) You are trying to calculate stream function on cell centres and not on vertices. In this case, interpolation from cell to vertex will mess you up no matter how good the cell values are. 3) You are using the fluxes, but you've messed up the algorithm (it is quote tricky, because you have to cover all points by marching through the faces and unless the mesh is nicely structured, you may need several passes) 4) The algorithm needs a reference - a point where the stream function is zero. You may have ended up with more than one reference.

Hope this was useful.

If you have flow impacting a wall and dividing, then one streamline will actually be (more-or-less) normal to the wall at the impact point. The value of the stream function on that line will be the same as the value of the stream function on the wall at that point.

Without knowing more about your problem, it's not possible to say that your solution is bad.

ThE flow is 2d flow of a dielectric fluid placed in a DC electric field setup in an enclusure and is driven by electrostatic forces. The wall in question is the wall of an electrode ...qualitatively I expect vortices close to the wall and may be dividing flow and stagnation points...

For simplicity I treated the wall during solution of equations by giving very big influence coefficients in discretization equation to the points within (and very small to neighbors)..the wall boundary are cv faces...

the storage of geometry is similar to finite-difference method though the formulation cv-based.

The grid is also much coarser than I will like it to be....

If I understand this, you should expect a steamline running into the stagnation point on the electrode.

So perhaps your solution has captured the essence of the physics. Hope so!

Thanks Mr. Jasak..your discussion was enlightening

...I did not calculate the streamfunctions: I am sweeping the solution domain up-down once and left-right once to provide starting points for the 'streamline' function in matlab (as I have no apriori 'feel' of the EHD flow field in the enclosure)..I need to be clearer about the matlab algo as well as my algo...

actually the storage in my program is much like simple finite difference methods....I was imposing no slip on a grid point half cv away from wall for simplicity ..now I have done it on face (uright=-uleft(imaginary))..it is showing more funny structures near wall(vortices and streamline landing at angles)...