Export Simulation Output to File

[sbrandstaeter]

Summary:

Add functionality to write the output of a MIRCO simulation to a file. This includes both the displacement field and the active contact set on the structured grid.

Motivation:

Currently, MIRCO runs simulations but does not provide a built-in way to persist simulation results for later use. Being able to export key output data would greatly enhance the workflow for post-processing, visualization, and potential continuation of simulations.

Proposed Feature:

• Implement a mechanism to write simulation results — particularly the displacement field and the active contact set — to a file.
• The output format should ideally be a standard binary format suitable for interoperability with visualization and post-processing tools (e.g., .bin, .h5, .vtk, etc.).
• The feature should include an option for selecting the output file path.

Use Cases:

1. Visualization and Analysis: Exported data can be imported into standard visualization tools to inspect the results of a simulation.

2. Post-processing: Enables further numerical or analytical processing of the displacement and contact data.

3. Future Extensions (Restart / Load Steps):
   While out of scope for this issue, the same mechanism could support future capabilities such as restarting simulations or running multiple load steps sequentially. In such cases, the active set from one load step could serve as an initial guess for the next, making this output mechanism foundational for later developments.

Benefits:

• Facilitates integration with external tools.
• Improves reproducibility and analysis capabilities.
• Lays groundwork for future restart and load-step functionality.

Related Issues and Merge Requests

• Related to Implement postprocess #84 Define desired input-output relationship and fix the code accordingly #14 Verbosity of MIRCO output #118

Additional Information

Interested Parties

[mayrmt]

@sbrandstaeter thanks for opening this issue! Visualization of results is for sure important when presenting results produced by MIRCO.

@RShaw026 do you know, how this has been done in the past?

[RShaw026]

@mayrmt In the past, I have used some scripts of my own to visualize results. I agree that we should implement a mechanism to write output results that can be visualized with standard tools.

I see one problem for the displacement field. MIRCO currently predicts the set of points that can possibly come into contact and calculates the displacements (and tractions) only at these points. This helps in keeping the system size small. We do this because we are only interested in the mean pressure. To get the displacement field for all the grid points in the half-space, we have to either have to get rid of the predictor step and consider all points in the system (not practical), or generate the full Green's function matrix considering all the points at the end and multiply it with the tractions field to recover the complete displacements field (more practical).

[PhilipOesterlePekrun]

@RShaw026 We have all of the pressures (because the pressures at inactive points is 0), we can calculate the full displacement field with Equation 1 of (Bemporad, Paggi, 2015).

This would also require computing the entire H matrix. The overall computational cost could be significant (so visualization will of course be optional), but it is a one time cost and would not increase for larger active set size--it would increase with the topology size and actually slightly decrease with increasing active set size, if my logic is correct.

Edit: Ah, right, I see that is what you meant with the second option. I agree this is practical and we should do this.