Learning Forum

[v.bhatnagar.1]

Hello,

I have been performing viscoelastic free surface extrusion simulations using Ansys Polyflow on my University's HPC. The model I am using is called the simplified viscoelastic model. I am currently running a sample simulation where I am attempting to optimise a weighting parameter, psi, that controls the extent of die swell. To do this, I am using a Bayesian optimisation framework (Ax) using a Python script that suggests the next psi value to try.

So, the whole workflow is as follows: 
1) Simulate with a psi_initial for a particular material
2) Extract the extrudate swell boundary (the downstream-most perimeter of the extrudate) from the last_csv.csv output file
3) Compare with an ideal boundary
4) Quantify the difference between the simulated and ideal boundary as an RMSE value
5) Feed the psi_initial and the RMSE value to the Bayesian optimiser (Ax)
6) Get the next psi value(s)
7) Repeat steps 1-6 for a fixed number of iterations or until the RMSE error reaches below a certain threshold, whichever is earlier

The HPC SLURM script uses the polyflow.lst and polyflow.dat files created by setting up the simulation on Ansys Workbench's Polyflow module. Once the simulation is done, it outputs a certain "last_csv.csv" file that contains the nodal coordinates of the FE mesh and other simulation output data. I am working in the .wbpj file directory in this whole workflow.

My problem is as follows: Though my Python script finds the next psi value and edits the polyflow.dat file accordingly, the simulation results are the same each time. Meaning, the psi value is changed in the .dat file, but the simulation results are the same as for the original psi_initial value. What can I do to fix this?

Any help would be greatly appreciated.