Capabilities and limitations (FMU import)

Common limitations for both Model Exchange and Co-simulation import, in addition to the rules of the FMI 2.0/3.0 specs:

• Scalar variables of type "String" with variability other than "Parameter" are unsupported — Amesim does not support string input/output/discrete-state variables.
• Enumeration, Boolean, and integer-type variables are partially supported via cast to/from double each time the import submodel accesses them (Amesim only supports double-type variables). This restriction does not apply to parameters — enumeration, Boolean, and integer parameters are fully supported natively.
• Source-only FMUs are not supported; import stops with an error.

FMI 3.0-specific limitations

• Variable aliases are ignored.
• Optional terminals are ignored — each input/output-causality variable is wrapped in one signal-type port.
• Configuration/re-configuration modes are unsupported — structural parameters can't change, so array variable sizes are fixed.
• Binary-type variables are unsupported and ignored.

Ignored features (warning only, import continues)

• Display units declared in the FMU are ignored (unit choice happens at model/application level, not submodel level).
• Vendor annotations are ignored.
• Default experiments are ignored (only meaningful for a fully assembled model).

Model Exchange related capabilities

• Step events are ignored during simulation — an FMU cannot require Amesim to process an event not identified by an event-indicator change or scheduled time event. FMUs relying on dynamic state selection are unsupported (warning reported, simulation continues).
• If the FMU's Model Description explicitly enables it, several instances of the same FMU for Model Exchange can be used on the same Amesim sketch.

Co-simulation related capabilities

• Only fixed-size co-simulation steps are supported — even if the slave FMU allows variable step size, Amesim's co-simulation masters don't use it.
• Serial explicit co-simulation (staggered sequential scheme) is not implemented.
• With a fixed-step numerical solver, the co-simulation step size must be a multiple of the numerical step size (the multi-rate factor must be an integer).
• Event iterations are not supported — the user-provided co-simulation step size must be strictly greater than 0, regardless of the capability flag in the FMU.
• Rejection of a co-simulation step is not allowed — if the slave FMU only computes a sub-interval successfully, co-simulation aborts with an error.
• Higher-order extrapolation of inputs is not implemented — only zero-order hold is used, regardless of the FMU's capability flag. On output, higher-order polynomial representations of output variables are neither retrieved nor used.
• If explicitly enabled in the FMU's Model Description, several instances of the same FMU for Co-simulation can be used on the same sketch.

FMI 3.0-specific co-simulation limitations

• Event mode and intermediate update modes are unsupported — FMUs with clock variables cannot be simulated.
• Early return (partial stepping) is not supported.