I am trying to repeat the results in a paper by Gartling D.K(1990). The problem is a backward facing step problem. During the simulation, it seems that the velocity blows up after a certain time. I was wondering if the time step I have used is correct. My spatial step is 0.05m(for all three directions, x, y and z) and my time step is 0.0025 sec. the maximum velocity at the inlet is 1.5 m/s while the average velocity at the inlet is 1 m/s.

Another question I have is regarding the adjustment of pressure field,after a solution is obtain, such that the pressure at the step is zero. Do I have to solve the pressure poisson equation again or do I just set the pressure at the step corner to be zero?

Can anyone advice me on these issues? I really appreciate any help available.

Dear Sir,

I've computed the Gartling-case myself. It is quite more difficult than people believe due to the low Reynolds number (800). It is semi-laminar, and hence it is rather important that you use a good discretization scheme. I found significant different result when using Hybrid compared to van Leer-scheme. You should opt for as high order as possible, with and an odd-order scheme to gain some stabilizing numerical diffusion. If you use a central scheme you may be in trouble. Furthermore you need to resolve the geometry, and at first disregard the third-direction. As an example I used 200 by 200 nodes (stretched). If you read the paper, it was focused as a code-validation case, and it works ok in 2D. Using an incompressible code the pressure could be fixed whereever in the domain, but prefererly in location where nothing happens with the flow, set it to zero at the outlet is may suggestion. If you need it to be zero at the step-corner, shift the pressure field in the post-routine.

/Jonas

Jonas Bredberg,

thanks for the reply. I really appreciate your reply. I am still not sure which pressure we have to set to zero at the outlet. Is it the hydrostatic pressure, dynamic pressure or both? Again, thank you very much to have answered my previous post.

Hi again,

For an incompressible flow, you may set the solved pressure (delta static) to zero at some location. The correct pressure is then calculated (post-op) by adding the ambient pressure. The dynamic pressure is, as you know, a result of the flowfield (rho*U^2/2) - and should not be set to zero, as that implies zero velocities - maybe you introduced a bug yourself here.

/Jonas