Support export of FMUs for cosimulation

[ufechner7]

I would like to export my model https://github.com/ufechner7/KiteModels.jl to an FMU. How should I start?

[ThummeTo]

Dear @ufechner7,

export of CS-FMUs is work in progesss, one of our students is already on it. A nice tutorial is also WIP.

If model-exchange (ME) FMUs is also an option (almost any FMI-compliant modelling environement can handle ME- as well as CS-FMUs), you can have a look on the BouncingBall-Example in the "example"-folder - which exports a ME-FMU.

All you have to do for ME is basically to fill the four functions in the example with your Kite-Model's state space equations.

Please note, that this package is very fresh and we only tested it on some specific models from our use-cases.

Regards!

[ufechner7]

Any update on this?

[ThummeTo]

Dear @ufechner7,
from a technical perspective, there is not that much to do, the most "complex" missing functions to be implemented is fmi2DoStep that must only trigger a fmiSimulateME (CS can be seen as ME with some extra steps). The only reason we haven't implemented it yet is that we have no application for it now. Do you have an urgent use case? Any way, this is on the todo list, but priority is low until it is needed somewhere.

[ufechner7]

Well, I created a new model, a tether model, which is not possible to solve using Simulink solvers, we need the Julia solvers, they are at least 1000 times faster... Have a look at: https://github.com/ufechner7/Tethers.jl

I really would like to make this model available to my colleagues from the IEA workgroup "airborne wind energy" who still use Python or Simulink... Exporting this as FMU would be perfect...

[ufechner7]

Perhaps you could give me some hints how to start this work... What would be the first step? Define the inputs, outputs, configuration struct and states of my model?

[ThummeTo]

Sure :)
From a technical perspective, one only needs to extend the template package for FMI2 FMUs in FMIBuild by the mandatory FMI2 functions for CS (see the FMI-spec), the most "complicated" function would be fmi2DoStep that would use the existing template functions (like FMU_FCT_EVALUATE defined for example here) to setup a DifferentialEquations.jl ODEProblem and do one simulation step. The biggest show stopper could be: I don't know if we can compile the entire DifferentialEquations.jl with PackageCompiler.jl - I did not test this yet. Also, binary sizes might be large, but there is (and was) a lot of progress in PackageCompiler regarding binary sizes.

[Xiuchanhun]

From test-0: stepping - Error test_fmi2DoStep: Current communication point ( 5.999999865889549255e-02 s) does not match last evaluation time ( 0.000000000000000000e+00 s) within tolerance 9.999999999999999547e-08. Time rollback is not allowed how to SLove this？

[ThummeTo]

Please always provide a (minimum) example, it's really hard to guess what is going on there. Also, I am not sure if this is related to this topic (see title) and the package FMIExport.jl - if this is a simulation issue it might be an issue for FMI.jl.