Abstract
Different fuzzy logic approaches are applied for speed control of a 1.44 kW cooling blast engine. Selftuning of a classical PI controller by fuzzy rules provides the most simple solution. Higher control performance across the whole working range of this nonlinear process can only be achieved by using some knowledge about process behaviour. These informations in the form of a qualitative dynamic fuzzy model can be obtained from measured input/output data by fuzzy identification. The learning CoS (Center of Singletons) fuzzy system performing this identification is shown to be mathematically equivalent to a neural network with radial basis functions and is therefore called a “neuro=fuzzy-structure”. Finally three different approaches for synthesis of a nonlinear control law are compared with respect to performance and design effort: imitation of human operators by fuzzy rules, imitation of classical controllers by fuzzy rules and fuzzy feedforward adaptation as qualitative feedback linearization.
Get full access to this article
View all access options for this article.