.MultiPhaseMixture.Information.Limitations.ModelicaSpecification

Information

Following issues have been observed that is related to the Modelica specification:

Inconvenient to calculate parameters

Currently it is inconvenient to use a model or block based structure to calculate parameters as illustrated below. It would be more user friendly if a model or block could be used in a similar way as a function to calculate parameters.

parameter  SpecificEnthalpy h_start (fixed=false)  annotation(Evaluate=true);
SpecificEnthalpy h(start=h_start);


Medium.MultiPhaseProperties
    flash_init(Z=Z_start,p=p_start,T=T_start,
    presentPhases=presentPhases,
    presentPhasesStatus=presentPhasesStatus,
    init(p=p_start, x=fill(Z_start, Medium.nP)),
    inputs=MultiPhaseMixture.Interfaces.Inputs.pTX)
initial equation 
  h_start=flash.h;

Solver callback interface

The external interface ExternalMixtureMedia has been designed with a structure that supports caching. The idea is to cache result from a calculation and use it as start values in a next coming calculation to decrease the number of internal iterations and increase robustness. A problem with this approach is that it is not possible to distinguish a function call during normal continuous simulation from one where the steady-state solver desperately tries to find a solution.

During continous simulation a good strategy would be to use values from the last accepted step.  For the steady-state case it might be an idea to let the non-linear solver update the starting values of the iteration variables hidden in these algorithms, when the solver makes good progress.

A solution would be to have the possibility to register a solver callback interface, which could be used to update iteration start values in a controlled way.

A suggestion

•   onSolverAcceptedStep() - called by solver/simulation environment when an accepted step has occurred. Place to implement updates of iteration start variables.

•   onSolverSteadyStateProgress() - called by solver/simulation environment when progress in steady-state solver. Place to update iteration start values variables.

Advantages with introducing callback methods are that the iteration start values can be updated in a controlled way and thereby avoiding the risk of an update during a bad steady-state iteration step.