Introduce the behavioral model from eFMI as layered standard to FMI

[hubertus65]

Motivation

The motivation for the introduction of the behavioral model in eFMI holds for many use cases in FMI as well. For traceability and model exchange it is useful to include validation trajectories also in an FMU. More use cases will be listed below. The implementation effort should be minimal since a reference to the eFMI-specification would do most of the work. In addition to that, only little "glue" is necessary to make it available with an FMU. It would be possible to store the behavioral model at a standardized and reserved location in the "extra" directory. The increased traceability and better verification of FMUs used for model exchange will make it easier to create a "credible simulation process" as in the UPSIM project with FMUs.

Use Cases

1. A ME-FMU requires specific tolerances or solvers or solver settings. If the same ones are not available from the importer as used by the exporter, verification trajectories allow us to find out what works in the importing tool.
2. Several FMUs are composed for co-simulation, and possibly re-exported as a composite CS-FMU. The behavioral model for one of the sub-system FMUs can be used for checking at least that the given sub-system trajectories don't deviate too much from the stored ones.

I'd be glad if you guys could do a poll on this during the f2f meeting in Berlin!

[chrbertsch]

For reference here is a link to the behavorial model definiotion in the eFMI standard: https://emphysis.github.io/pages/downloads/efmi_specification_1.0.0-alpha.4.html#BehavioralModel

[chrbertsch]

Design meeting:

Andreas: this is something like a "smoke test". Perhaps another name than "behavorial model" would be preferable. Markus F: similar idea as in the cross check, might create overhead. But should be easy to adapt from eFMI

Poll: who considers this important: in favour: Andreas, Christian B, Otto, Torsten S. Abstain: rest

[chrbertsch]

FMI Steering Committee:

Pierre: Name "behavorial model" name is misleading. Similar requests were addressed to SSP, could be a more genera concept in MA Andreas: Could be a layered standard Pierre: Could be a common layered standard. Hubertus: even for a Modelica model