Store results of optimization in system not in Substrate

ContactEngineering / ContactMechanics

Contact mechanics using elastic half-space methods

https://contactengineering.github.io/ContactMechanics/

MIT License

21 stars 5 forks source link

Store results of optimization in system not in Substrate #38

Closed sannant closed 2 years ago

sannant commented 3 years ago

At least that's what we decided for the Adhesion potentials in Adhesion.

Example: https://github.com/ContactEngineering/ContactMechanics/blob/47c100be6eef46711db70bcd6d15c1abba672277/ContactMechanics/Systems.py#L863

@pastewka what do you think ?

pastewka commented 3 years ago

Sounds good - I think this should go into system

pastewka commented 2 years ago

I think this is already the case and the line is a bug. substrate.force should contain just the force contribution from the substrate, self.force should contain the sum of substrate plus interaction.

sannant commented 2 years ago

I disagree. when I call system.force I expect to get the physical value I am interested in (either the substrate force or the interaction force), not the residual of the optimization problem.

Happy to create a variable called residual_force

sannant commented 2 years ago

I want to additionally point out that:

we made the code-style decision not to store internal states in the interaction potentials, and for consistency we should not store any force in the substrate.
If we change our mind and want to store a force in the substrate, this line is not a bug. In the context of dual minimization without adhesion, result.x is exactly the force applied on the substrate.