Accurate-interpretable based fuzzy modeling of a DC motor

The aim of this paper is to obtain a fuzzy model and transform it into one closer to a linguistic model, so as to achieve a better interpretability of the fuzzy rules that define the model. The fuzzy model employs Mamdani fuzzy rules that have been tuned from input/output data. The main objective is to achieve an accurate, compact and interpretable model. This is possible if the obtained set of fuzzy rules present properties such as compactness, distinguishability, etc... that are not usual when fuzzy modeling is involved but are in fuzzy linguistic models. The methodology used in this paper is based on the modification of a known algorithm for obtaining TSK fuzzy models and a mecanism based on a genetic algorithm in order to improve the fuzzy rule sets generated, avoiding some of the problems associated with ”fuzzy modeling”, improving its linguistic properties and tuning the input variables (rule antecedents) in order to weight their relevance. This procedure has been applied to a real DC motor obtaining a more compact, accurate and interpretable model based on a Mamdani fuzzy rule set, that finally explains the behavior of that device.