Modified Simulink interface for hybrid system

[camirmas]

Hi OpenFAST Team,

I'm new to OpenFAST and wind turbine engineering in general, so thanks in advance for your patience! I'm investigating the possibility of modifying OpenFAST to apply platform force/moment inputs from Simulink to ElastoDyn, and decoupling HydroDyn. For context, these changes are intended to be part of a hybrid simulation in which the wave loading and platform are modeled physically, and OpenFAST will be used to model aero-elastic effects and output platform/tower displacements back to the physical system. This seems somewhat similar to #282, though I've hit some issues in implementation. I've modified the Simulink + S-Function interface to be able to pass forces to OpenFAST, and am attempting to validate my implementation using the following process:

• Peform a baseline semi-sub run (.fst file) with no Simulink inputs applied. In ED_CalcOutput, add PlatformPtmesh inputs to the output parameters
• Using this run's output file, extract forces/moments at each time step to .mat files
• Run the same simulation, but with HydroDyn turned off, feeding the forces/moments from the .mat files
  • Apply Simulink-provided forces/moments at the end of ED_InputSolve at each time step
• Observe ED platform displacements for the base case vs. those of the hybrid Simulink model

The result of this process is that the turbine slowly sinks in the modified case. I was wondering if this is related to the implicit hydro-elastic solver issues discussed in #282, and that there is some added force that is not being accounted for. Here's my branch, and I'm happy to provide more results if needed.

Here's what the outputs look like for now:

[jjonkman]

Dear @camirmas,

I'm guessing that the problem is related to how Simulink interfaces with OpenFAST. The implicit nature of the hydrodynamic-structural coupling could certainly be an issue. Another issue is that OpenFAST is implemented as a discrete time block in Simulink. That is, Simulink will call OpenFAST with inputs at time step "n", OpenFAST will then advance it states from time step "n" to "n+1", and calculate its outputs at time step "n+1", the outputs of which are then returned to Simulink at time step "n+1". Because time is advanced within OpenFAST during the call, the inputs passed from Simulink to OpenFAST get extrapolated within OpenFAST. I would guess this extrapolation is causing an issue in the coupling you have implemented.

But I'm not sure I know what to suggest to get around these issues given the way OpenFAST is interfaced with Simulink. You'd likely have more flexibility in how you develop the interface if you choose not to use Simulink.

Best regards,