[RKE]

I am running two versions of a transient Star-CCM+ simulation of a stirred tank, which represent the same system but have different spatial resolutions.

The first one (with the lower resolution) runs faster; the second one runs very slowly.

I would like to use the results from the first one, at a given time, as the initial velocity field for the second (higher resolution) simulation. This would allow me to shorten the time required to examine the operation of the stirred tank beyond the initial start-up.

Would someone kindly provide suggestions on how to import a flow field from the low-resolution simulation for use as the initial condition for the higher-resolution one?

Of course, I will need to make sure the position and speed of the impeller in both simulations match each other.

Thank you

[fluid23]

You should be able to run the low resolution simulation to whatever point you want, then re-mesh to give you the desired high-resolution discritization. The software will interpolate the coarse mesh results on to the fine mesh. Then clear results (but not fields so you retain all field function values, but start at iteration 1 or t=0) and start your simulation again.

[LuckyTran]

Re-meshing and letting Star interpolate automatically is easiest if you don't already have two distinct cases. Otherwise you have to...


Export all each variable as a csv table and then import these tables into your new case. Sometimes you can get away with using just x,y,z tables to assign the initial conditions. More often, this is not good enough and you need to use a data mapper, to map the data from the table onto your new mesh. This then creates field functions (with funny mapped names) that you can then use to assign to each field.


It's a pain.

[flotus1]

Just a few days ago, I was searching for a reasonable method to do this. Glad to hear that I was not just too stupid to find it.

[me3840]

Originally Posted by LuckyTran

Export all each variable as a csv table and then import these tables into your new case. Sometimes you can get away with using just x,y,z tables to assign the initial conditions. More often, this is not good enough and you need to use a data mapper, to map the data from the table onto your new mesh. This then creates field functions (with funny mapped names) that you can then use to assign to each field.

For future folks, I would strongly advise against using a table in this manner. The mapper is definitely the way to go unless your mesh is very small.


The reason is that tables are not partitioned. Say your mesh and solution consume 10GB of memory. At runtime one copy of the table is provided to each process. This means on a node with 16 processes, that table will consume 160GB of memory. Ouch.


The data mappers use field functions which are partitioned, so they are far more efficient.