Method for the Synthesis of Fuzzy Controllers from Experimental Data Based on Clustering

The purpose of this work is to develop a method for the synthesis of fuzzy controllers based on experimental data based on clustering. To achieve this goal, it is proposed to take experimental data on the input and output signals of the control system of a technical object with a classical controller. On the basis of experimental data, a data clustering method is developed that allows determining the term-sets of input and output linguistic variables of a fuzzy controller that implements the Mamdani fuzzy inference algorithm. Clustering is performed by estimating the boundaries of the intervals of change in experimental data, evenly dividing them into clusters depending on the required power of the linguistic variables term-sets, and determining whether the data belongs to one or another cluster. Since the experimental data are related, that is, for each moment of time, data on both the input and output signals of the classical controller is stored and their belonging to clusters is determined, then the rule base of a fuzzy controller is easily compiled. According to the results of experiments, it was found that in most cases the rule base is formed with duplicate rules, that is, there are rules with the same antecedent, but different consequent, so the authors conducted additional research related to the reduction of the rule base and assigning redundant rules of different weights. To simplify the research, the authors developed software in the MatLab environment that allows both to obtain experimental data, and to synthesize a fuzzy controller and check its performance. The research results will be useful for developers of fuzzy control models. #CSOC1120.