Oscillatory solution at pipe inlet
 

Hi,

I'm simulating the development of a initially uniform velocity profile into fully developed flow in a pipe. But depending on the length of the pipe, (if it is too short or long), I'm getting oscillatory solution near the inlet as shown in the attached image. Should I let the simulation run for more time steps and let it correct itself(which I'm not hopeful about) or how do I choose what length to simulate for a given Reynolds number?

Right now I have used the analytical relation for entrance length Le/D = 0.06*Re and multiplied it by a factor of 1.5 and run the simulations.

Why am I getting such oscillations? However long the pipe is, I should get the correct output right?

Thanks & Regards,
Qrie

P.S. Attached image is contour of streamwise velocity 'U'. X is the pipe's axial (streamwise) direction, y is the radial direction. y=0 is the pipe centreline.
(Only part of the pipe length is shown in the pic. )

Hi Qrie,

Although you did not supply enough information for answering (Laminar or turbulent flow? Steady or transient? What numerical method is used? etc.) my guess is that running for more steps won't help. I also guess you experience stability issues. Therefore, the first thing to check (assuming your problem is correctly set with appropriate BCs etc) is to "slow down" the solution by a more aggressive relaxation (either smaller under relaxation factor or smaller false time steps, depending on your software). Another guess is that your domain is too short, but this again depends on the numerical method and the BCs you hide from us http://www.cfd-online.com/Forums/ima...ons/icon12.gif.

Hi,

Did not realise I was hiding anything :D at the inlet uniform velocity profile was imposed. This is steady,incompressible, laminar pipe flow. Uses finite differences. Maybe it has to do with the grid? Because the domain I'm sure is long enough.

Thanks & Regards,
Qrie

Alright, some details revealed, others (e.g, the numerical method, the coulping method of the mass and momentum, the linear equation solver, the BCs, ...) are still under the carper.

Nevertheless, I still suggest to use more relaxation.