Hello,
I would like to apply for advice concerning the velocity boundary conditions in case of solidifying/melting problem with convection in melt. On the phase change interface, boundary conditions v \cdot t = 0 should be satisfied since the melting front is a no-slip surface. (with v the convection velocity vector and t unit tangent vector). Thus the tangent velocity to the interface should be zero and the normal velocity equal to the normal velocity of the interface (continuity of velocity).
So, in elmer, supposing that the front is horizontal, at the phase change interface velocity 1 should be equal to zero and velocity 2 should be equal to mesh velocity 2 (since mesh velocity 2 = velocity 2 of the melting front). However, when I impose these BC, navier-stokes solver doesn't converge.
Zero velocity boundary conditions (velocity 1 = velocity 2 = 0 ) work and give reasonable results, but I am still struggling around since it's not clear to me why velocity 2 = mesh velocity 2 BC doesn't work. Is there something obvious I am missing? Does it concern the ALE formulation of N-S?
best regards, Martina.
phase change w/ fluid flow
-
raback
- Site Admin
- Posts: 4832
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: phase change w/ fluid flow
Hi Martina
I don't have an immediate answer. However, how does the situation look from continuity equation point of view? Is the mass conserved if you integrate over BCs? Impossible imcompressibility constraints result easily to a divergent solution. Assuming a closed domain with one phase change boundary one would come to conclude that setting normal velocity would not be possible. Maybe for phase change the no-slip is ok after all... What is obviously needed is the ALE velocity because otherwise mesh deformation in the fluid would lead to wrong mapping of previous timestep in the bulk. However, at the interface no-slip velocity would mean that also ALE velocity vanishes. So the explicit coupling of phase change & Navier-Stokes would step in here.
-Peter
I don't have an immediate answer. However, how does the situation look from continuity equation point of view? Is the mass conserved if you integrate over BCs? Impossible imcompressibility constraints result easily to a divergent solution. Assuming a closed domain with one phase change boundary one would come to conclude that setting normal velocity would not be possible. Maybe for phase change the no-slip is ok after all... What is obviously needed is the ALE velocity because otherwise mesh deformation in the fluid would lead to wrong mapping of previous timestep in the bulk. However, at the interface no-slip velocity would mean that also ALE velocity vanishes. So the explicit coupling of phase change & Navier-Stokes would step in here.
-Peter
Re: phase change w/ fluid flow
Hello Peter,
thank you for your reply.
looking at the mass conservation, the difference between fluxes at the top and bottom of the domain (the vertical walls are insulated) is 1.6 %. If I look at loads over bottom and top boundaries, their difference is 0.07 %.
you don't set up normal velocities, but velocity 1 = velocity 2 = 0. Since the melting front is not planar, velocity 2 doesn't correspond to normal velocity.
I am sorry, it's still not clear to me what's happening.
BR, Martina.
thank you for your reply.
looking at the mass conservation, the difference between fluxes at the top and bottom of the domain (the vertical walls are insulated) is 1.6 %. If I look at loads over bottom and top boundaries, their difference is 0.07 %.
you don't set up normal velocities, but velocity 1 = velocity 2 = 0. Since the melting front is not planar, velocity 2 doesn't correspond to normal velocity.
I am sorry, it's still not clear to me what's happening.
BR, Martina.