Parameterized reduced order models with guaranteed passivity using matrix interpolation

We present a novel parameterized model order reduction method based on matrix interpolation. The design space is sampled over an estimation grid and for each estimation point a Krylov subspace is computed. A common projection matrix is generated by the truncation of the singular values of the merged Krylov subspaces of all estimation points from the design space. The reduced matrices are then interpolated using positive interpolation schemes to build guaranteed passive parameterized reduced order models. The technique is validated by means of a pertinent numerical simulation.