Here's a set of messages regar
Here's a set of messages regarding the ddt scheme update:
Attached is an update for ddt schemes. As I've indicated, once you start switching ddt schemes on a moving mesh, the mesh motion flux is calculated differently for each ddt scheme. Therefore, for anything other than Euler implicit mesh.phi() will not do the right thing.
Zeljko has introduced a new fvc::meshPhi(...) which takes a field as an argument. The field is there to find which ddtScheme is currently switched on and it will return the appropriate mesh motion flux.
In solvers (I did not pack those, do you want them?) fvc::meshPhi(U) (typically, you want U) should be used instead of mesh.phi().
You will notice some strangemess in handling mesh motion data in the fvMesh.C - that's because mesh.phi() now carries old-tome motion fluxes as well and needs to be handled with care - I suspect the same kind of problem will show up when we get dimensionedFields which carry oldTime() for mesh.V().
I need to consider these changes carefully because they may be incompatible with my plans to introduce an extended substantive derivative to avoid the algorithm differences for moving-mesh problems. I also want to create a super-class for U which incorporates phi so that all the relative/absolute phi crap is handled automatically. All this is in progress but will not be in the next release. The other issue is if there changes are compatible with the new phi predictor step I implemented for the pressure equation to avoid spacial-staggering on refined meshes.
Me again :-)
I will carry on with the current stuff for the moment - if someone needs higher order temporal discretisation on moving meshes any time soon, please give me a shout.
|All times are GMT -4. The time now is 22:38.|