Since nearly the conception of the finite element method, superelements, or substructures for linear static (static condensation) or linear dynamic (Craig-Bampton) analyses, have been a relevant technology for achieving computational tractability for models that exceed computation capability. This is done by reducing segments of an overall assembly model into reduced representations (e.g., stiffness matrices, mass matrices, etc.).
With modern computation capabilities, the role of superelements has shifted toward a method for finalizing part representations prior to inclusion in larger assembly models. These reduced models significantly limit the design flexibility in downstream analyses, which may be required to meet systems-level requirements. By leveraging machine learning techniques with ODYSSEE, Smart Superelements (SSE), introduced in MSC Nastran, can be generated, and include predetermined parameter vaariations.
This capability provides the part owner with the ability to embed select variations into the superelement prior to its inclusion in larger assembly models. Similarly, the assembly analyst can adjust these select parameters at solution time without the expertise of the part owner. As a result, the institutional knowledge of the part owner and prior what-if studies can be captured to allow for flexibility in meeting systems-level requirements. In this work, an example end-to-end workflow will be demonstrated for a drone model.