I am dealing with FSI calculations with Code_Saturne through an internal code coupling. The code uses as mentioned in the documentation a Newmark-HHT algorithm to determine the structure displacement. However, it seems to have no information about the numerical scheme to calculate the structure acceleration and velocity. Moreover, the frequency of structure displacement in my case depends on the dtstr value, which is not acceptable. What would you propose regarding this issue please ?

I don't understand very well your question, but did you check the file newmrk.f90. In this subroutine are computed the accelerations, velocities and displacements of the structure.

I meant that there only seems to have the details of the numerical scheme for the predicted displacement and forces driven by the coefficients aexxst, bexst and cfopre (on the file example "cs_user_fluid_structure_interaction.f90").

Then, how to handle the coefficients of the Newmark algorithm (alpnmk, betnmk, gamnmk) within a FSI calculation (done by means of the internal coupling of Code_Saturne) ?

The parameter alpnmk is initialized in iniini.f90 and the paremeters betnmk and gamnmk a defined in modini.f90. I think that you can copy the newmrk.f90 or strdep.f90 to the SRC directory and change the values.

I don't known if the values of the alpnmk, betnmk, gamnmk can be accessed from another user function like cs_user_paremeters. If it is possible will be better solution

I have included the use of alstru module in the "cs_user_paramters.f90" file. However, the structure displacement depends on the structure time step used within the internal coupling. Would you require a specific set of values to get accurate results ?

The set of values (aexxst, bexxst, cfopre, dtstr, alpnmk, betnmk, gamnmk). By changing dtstr for fixed values of the other parameters, I indeed get different frequencies of the structures displacement from the internal coupling.

Thanks in advance.

Regards,

So_nuc

Last edited by So_nuc on Mon May 20, 2019 5:56 pm, edited 1 time in total.

I don't known what kind of FSI problem are you solving. The time integration is implemented like in the paper"Improved Numerical Dissipation for Time Integration Algorithms in Structural Dynamics, Hilber, H.M, Hughes,T.J.R and Talor, R.L.". The change in the parameters will modify the behavior of the algorithm and the numerical damping added. Sometime some numerical damping should be added (alpha>0) in order to stabilized the simulation.

Did you try to increase the number of sub-iterations for implicit coupling (NALIMX)?

Thank you for these details. I am actually dealing with the simulation of structures in a fluid. I should observe the added mass effect through a reduced frequency of structures displacement but the result does not exhibit any effect (the structure oscillates with its natural frequency). I have also tried with the "nalimx" and "epalim" values of the Fluid-structure-interaction tutorial.

I think I get the issue relating to the mass structure: I indeed simulate a 2D-like simulation (height of 1e-6m) and I entered before the linear mass instead of the real mass within the height 1e-6m. Nevertheless the structure mass is now very low and leads to calculation instability despite of the parameters of the Newmark algorithm which should ensure calculation unconditional stability.