Model comparison and simplification

In this paper we consider comparison and simplification of dynamical models. These models may contain non-linearities as well as uncertainty, where both are described using Integral Quadratic Constraints (IQCs). The proposed method includes simplification by truncation and singular perturbation approximation as special cases. The simplification error is defined in terms of the L₂-induced gain. It is shown that each non-linear or uncertain system component can be assigned a positive value, computable by convex optimization, such that the simplification error is always bounded by the sum of these values corresponding to the simplified components.