[oschi]

Hello,
I am asking for help with unexpected wiggles in the Ux solution of a horizontally moving piston (pimpleDyMFoam solver).

The attached case simulates the motion of a piston on the right side of the model domain with a velocity ramp from (0 0 0) to (1 0 0) over the duration of 3ms.
The left side is set to pressureInletOutletVelocity/totalPressure. (For other details please ask or check the attached case.)

I am expecting a constant velocity profile, but results show strong oscillations of Ux:


Is it possible to change boundary conditions or solver to get rid of the wiggles? I have tried a compressible solver (rhoPimpleDyMFoam) as well with the same problem. The aim is to solve the flow inside a closed volume which may or may not have inlets and outlets. Without inlets/outlets, of course a compressible solver will be used, otherwise both compressible or incompressible solvers can be tried.

Greetings
Olaf

[jakob15]

Hi Olaf,
you have fixed the velocity at the outlet. Better fix the velocity at the inlet, which is known in this case, and use outlet conditions on the right side.

Other solutions to remove the oscillations are to run potentialFoam to start with a converged solution or to use "internalField (1 0 0)" as initial condition for U or to increase the fluid viscosity and solve for a longer time period until the oscillations decay.

Jakob

[oschi]

Hello Jakob,
I have tried to follow your suggestions.
1) The constant zero velocity field chosen as U internalField actually is the correct solution at time t=0, because I apply a velocity ramp from 0...1 m/s to the piston on the right side. For this reason the results are worse, if I initialise with U=(1 0 0), and potentialFoam changes nothing.
2) Applying the known velocity at the inlet and setting p=0, zerogradient U at the outlet shifts the wiggles from the right side to the left side. I have attached the case and Ux profile below.
3) Calculating with a higher viscosity helps a little bit, and the relative size of wiggles do shrink if I use a larger endTime, but I urgently need a solution for the given viscosity and given integration time.

Why can't OpenFOAM converge to the correct, known velocity profile (U=const)? What is the numerical reason for the problem?

Greetings
Olaf