Truly Nonlinear Model-Order Reduction Techniques

Model-order reduction (MOR) aims at automatic creation of compact and sufficiently accurate approximations of large-scale simulation models for efficient system design and optimization. While MOR is reaching the maturity in the area of linear system, nonlinear MOR applications are still quite sparse. Most of the existing nonlinear MOR approaches employ polynomial approximation of the nonlinear model operator that limits the applicability of the resulting reduced models. The objective of this paper is to introduce a class of truly nonlinear MOR techniques that do not alter the original nonlinear model formulation in the process of MOR subspace projection. The existing and new techniques for the accurate subspace creation and efficient nonlinear projection are discussed separately.