Hi,

To validate a code, I'm performing an explicit, finite volume inviscid compressible flow simulation of an internal flow through a duct with area changes. I was expecting the total pressures to be uniform throughout. However, it turned out that there was large amounts of total pressure loss at the outlet region.

Appreciate any advice on how this matter may be addressed.

Thanks very much.

Hi Chan,

"inviscid compressible" flow does not mean "no losses" flow. May be a shock wave is in the duct, near the exit of the duct (you should check for the critical value of the outlet pressure, if it's greater of the exit pressure you have set). If not, you should check for the computational method you are using. First order upwind schemes usually generates high level of numerical viscosity and central schemes, for example, need artificial dissipation function to properly work: this dissipation can be the origin of the total pressure loss (i don't think this is your case). Which computational scheme are you using? Is a 2D, 1D test? Are you sure boundary conditions are correctly handled? Good luck,

Nicola

Hi Nicola,

Thanks very much for your insight.

The outlet pressure is below critical, all Mach numbers are subsonic. The numerical scheme is Jameson's central differenced Runge-Kutta multi-stage time-stepping. The test case is axisymmetric and the boundary conditions are characteristic-type boundary conditions.

However, loss of total pressure occurs, beginning at about halfway down the duct and by the time the outlet is reached, about 15-20% of the total pressure has been lost.

Anyone have further advice on this?

Cheers

I use this same numerical scheme, with implicit residual smoothing and Brandt multigrid scheme to speed up convergency.

I think you are using Jameson's artificial dissipation scheme with pressure sensor to switch off dissipative terms. Some scaling constants must be added to artificial dissipation to avoid unphysical losses. Some authors use some other scaling factors with body fitted meshes.

Where losses origin is located? Near the wall or in the main flow region?

How boundary conditions are imposed at outlet and at wall? Phantom cells?

Check for conservation of other integral properties(mass flow, total enthalpy, ecc.) to be sure no errors are in the main discretization routines. Does Mach number in isentropic region agree with 1-d approach results?

Change outlet pressure value and see what happens. Best regards

Nicola

Losses are located in the main flow region.

Boundary conditions at the wall use phantom cells while at the outlet, characteristic type boundary conditions are used.

Thanks very much.

Hi Chan,
Could you provide me the references of this Jamenson's differencing scheme? I want to take a look at his paper.
Thanks in advance, cfd guy

Hi cfd guy,

The reference is AIAA Paper 81-1259.

Regards