Robust Passivity Preserving Parametric Model Order Reduction using Matrix Interpolation

Abstract

We present a novel robust parametric model order reduction method using matrix interpolation and common projection matrix. The design space is sampled over an estimation grid and transformation matrices are computed at the estimation points using a projection-based passivity preserving truncated balanced realization. A common projection matrix is obtained by the truncation of the singular values of the merged transformation matrices obtained from the estimation points in the design space. Finally, the reduced system matrices are interpolated using positive interpolation schemes to obtain a guaranteed passive parametric reduced model. The proposed technique is validated using pertinent numerical results.