Add KRUSTY 15 Cents Reactivity Insertion Model

[miaoyinb]

closes #477

[miaoyinb]

@ycao-anl FYI

[moosebuild]

Job VTB Documentation, step Sync Docs on a02dd53 wanted to post the following:

View the site here

This comment will be updated on new commits.

[miaoyinb]

@GiudGiud This is to add a transient model to the existing KRUSTY model. Thanks in advance for your review.

[GiudGiud]

you ll want to add a few tags to the main krusty page, mainly RIA ?

any ideas for testing? can we regenerate the checkpoint file in a testing run (ideal) ?

[miaoyinb]

you ll want to add a few tags to the main krusty page, mainly RIA ?

any ideas for testing? can we regenerate the checkpoint file in a testing run (ideal) ?

RIA might not be the ideal term as this is a warm critical test instead of an accident test. Maybe just use "warm critical test"?

Testing could be tricky as it will rely on the steady-state simulation checkpoint/SolutionVector output files, which also depend on the mpi ranks used. Any suggestion?

[GiudGiud]

Running the steady case then running the transient is the easiest.

which also depend on the mpi ranks used. Any suggestion?

alternatively, if we LFS checked in the steady checkpoint we would have to be mindful of that

[GiudGiud]

there s a model under "Reactivity Insertion" already, for MR drum. could go with that one?

[miaoyinb]

Running the steady case then running the transient is the easiest.

It takes ~4 hours on 10 sawtooth nodes to finish SN(2,3) NA=3 steady state.

Maybe we can try SN(1,3) NA=1 and force it to finish after one Richardson iteration and start the transient to run one timestep?

[GiudGiud]

we can try something like that though we wont see a steady state in the beginning of the transient

[miaoyinb]

Let's try to see how long it takes to run such a case.

[miaoyinb]

Good news: SN(1,3) NA=1 steady state with one Richardson iteration takes ~1000 seconds to run using 40 cores on a sawtooth node.
Bad news: the corresponding transient case using SN(1,3) NA=1 does not have enough memory to run on a single node. (it needs more nodes + distributed mesh to relieve the memory burden on each node.)

Maybe we can run the "simplified" steady-state case and then do syntax check only for the transient?

[GiudGiud]

can we reduce the mesh maybe?

[miaoyinb]

Let me try

[miaoyinb]

Reducing axial mesh density works. Using SN(1,3) NA=1 and allowing Richardson iteration to converge, the steady-state run takes ~2 hours and one time step transient takes about 500 seconds. I guess, this is okay for a "heavy" test.

[miaoyinb]

Added the tests. It seems the syntax override+!include does not like cli_args. I had to add an input file for testing.

[miaoyinb]

@GiudGiud Let me know if these tests are okay or if you have any additional comments. Thanks.

[GiudGiud]

Will do.
Need to get a submodule update done first