In the past years, several large dynamical models of cellular metabolism have been published. Even more are going to come on the way towards models of the complete cellular metabolism. As the models are nonlinear, analyzing and understanding them is a not trivial task. This paper proposes a method to finding approximate models that explicitly take into account the switching behavior inherently present in several parts of the cellular metabolism. First, the states are separated into functional units based on similar “activity” along suitable trajectories. Important are the choices of sufficiently stimulating trajectories and of when a state is called active. Secondly, an automata model is build using the result of the first step. A model of carbohydrate uptake and the glycolysis is used to demonstrate the feasibility of the proposed method.