Fuzzy identification of dynamic systems with adaptive structure

This paper deals with two approaches for on line structure identification of fuzzy models. In the first one, a constructive algorithm is adopted to generate the fuzzy rules: it starts with a single pattern and a single fuzzy rule and grows progressively to reduce the system error within the specified tolerance. In the second one, an evolutionary algorithm is applied based on an alternation between incremental and pruning criteria. Indeed, the rule base is expanded when the model can not reduce the system error and one rule is removed if it has a petty contribution in the model output along some patterns. The presented approaches have been applied for two examples of dynamic systems to compare the identification performance.