A method for heuristic fuzzy modeling in noisy environment

This paper presents a fully automatic algorithm for fuzzy model identification that pays attention to the interpretability and reliability of the model and is particularly suitable for working in difficult conditions where data may be both noisy and corrupted. The working principles and essential characteristics of the algorithm are explained on the basis of simple examples, its approximation properties are tested on Box-Jenkins data set and its application to fed-batch fermentation process demonstrates that in conditions resembling real life it can take full responsibility for the modeling task in modeling-for-control methodology.