Adding a new two fluid solver for hydraulic application

[uxuech]

📝 Description
This PR introduces a new two-fluid solver with a hydraulic application, and includes several key differences from the previous "two_fluid_hydraulic_solver":

• The addition of an option to implement artificial viscosity based on the residual, which improves robustness and eliminates potential numerical noise.

• A correction is applied to the historical velocity variable before solving the N.S equations, achieved by solving a convection problem where the velocity variable from the previous step is convected (eulerian fm ale problem), thus preserving energy.

• The addition of an option to calculate the stabilization parameter tau for the convection element nodally, which prevents sudden jumps in the value of the tau between elements.

In addition, a new test has been added to verify all changes made in both 2D and 3D of the sloshing benchmark.

[ddiezrod]

@rubenzorrilla @uxuech Is there any plan to unify the developments here with the navier stokes two fluid solver? I understand that separating both solvers will accelerate the implementation but I would like to have only 1 solver at some point....

[rubenzorrilla]

@rubenzorrilla @uxuech Is there any plan to unify the developments here with the navier stokes two fluid solver? I understand that separating both solvers will accelerate the implementation but I would like to have only 1 solver at some point....

First, I'd like to note that this is what we tried first.

Answering to your question, not in the short term, an I foresee that not in the mid term either. The point is that while developing we realized that @uxuech 's target application needs various specifics (mass conservation algorithm, history convection, artificial viscosity, etc.) that would result in an (even more) cumbersome implementation in current solver.

Besides, I think that the defaults are to be rather different in both cases. Note that current 2-fluid solver is being used for quite different stuff (yours, microfluidics, etc.) while this one is intended to be applied large scale hydraulic applications.

[ddiezrod]

Thanks for the answer @rubenzorrilla I am afraid that the whole structure might be a little misleading (for example I think my casting filling examples will have more in common with this hydraulic solver than with some of the more recent developments in the two fluid solver, mostly related to bubble formation etc). Besides the code duplication will be a little of a headache. This being said, I understand that we have to be flexible until everything is properly tested and stabilized, so for me it is ok to add this new solver.

[rubenzorrilla]

Thanks for the answer @rubenzorrilla I am afraid that the whole structure might be a little misleading (for example I think my casting filling examples will have more in common with this hydraulic solver than with some of the more recent developments in the two fluid solver, mostly related to bubble formation etc). Besides the code duplication will be a little of a headache. This being said, I understand that we have to be flexible until everything is properly tested and stabilized, so for me it is ok to add this new solver.

Thinking in the long run, the best solution could be to have a base "do nothing" two-fluid solver with all the ingredients and different derived solvers re-implementing the defaults and key functions.

[rubenzorrilla]

There are some concerns to be amended.

Besides, from my understanding I think you should have one results file per test . However there are only two. Regarding the mdpa(s), do we need them to be that fine? Couldn't we use a coarser mesh?

[uxuech]

Answering to why the "time_scheme" is present in the default Parameters, it's here. This solver is derived from the fluid_solver, and therefore, we decided to implement this temporary solution to make it work. @rubenzorrilla , do you think we should modify this part of the code?

[rubenzorrilla]

Considering that this new solver is specifically tailored for your element, which hardcodes an alpha-scheme within the RHS and LHS implementation, I'd leave it as it is. In other words, this solver is only compatible with ResidualBasedIncrementalUpdateStaticSchemeSlip.

[uxuech]

Regarding the Python tests, I would like to know how much you suggest we should reduce the MDPA (Mesh Data and Parameter Archive). Thank you for your help @rubenzorrilla.

[rubenzorrilla]

Regarding the Python tests, I would like to know how much you suggest we should reduce the MDPA (Mesh Data and Parameter Archive). Thank you for your help @rubenzorrilla.

If it works, I'd do a mesh of 5 times 5 divisions (5 times 5 times 5 in 3D). At the end of the day you just want to check one of the very first steps.

[rubenzorrilla]

Ready 🚀