The article presents the basic principles of designing Mamdani-type fuzzy controllers. A general formula for a controller with a MIMO structure is presented. The selection of membership functions for individual linguistic variables is made. The relative positions and ranges of the membership functions are determined. The rule base of the fuzzy controller is defined. Two control surfaces of the controller for the center-of-sums defuzzification method are presented.The issue of designing Mamdani-type fuzzy controllers is complex. This results both from the diverse possibilities regarding the choice of specific membership functions for the control process variables, as well as from the very specificity of the linguistic rules describing the controller’s operation algorithm. An additional issue is the number of inputs and outputs, which in a way determines the size of the rule base, which for the Mamdani-type structure has fuzzy antecedents and consequents [2, 8].
Tuning a designed fuzzy controller requires modifying the shape of the membership functions and their relative positions for individual control process variables. Consequently, the tuning procedure allows for shaping the non-linear control surface of the controller. This surface is composed of spatial segments that can have varying degrees of convexity. The number of segments increases with the number of membership functions describing the control process variables. (…)
The article contains 13,196 characters. < /p>
Source: Logistyka Magazine
Design and tuning of Mamdani-type fuzzy controllers – PDF file for download