Design and analysis of adaptive fuzzy logic controller with dual outputs and its application to plants with time-varying dead-zone

Abstract

In this dissertation, we study on the design and analysis of the adaptive fuzzy logic controller with dual outputs and its application to the plants with time-varying dead-zone. A self-organizing fuzzy logic controller (SOC or SOFLC) is a sort of the adaptive fuzzy logic controller (AFLC) which can construct the control rules. Even though the SOC can obtain the control rules, it has some limitations because the number of elements of the performance decision table is increased to $m^n$ rules if the fuzzy logic controller (FLC) handles n physical input variables with m attached linguistic fuzzy sets. This limits the basic controller as the PI or PD-like FLC. To give the PID-like FLC, we present the self-organizing fuzzy logic controller with dual outputs (SOCDO). Since the SOCDO has only two input variables, it does not increase the number of elements of the performance decision table. In addition, we proposed the generalized performance decision table based on gradient descent which is similar to the fuzzy sliding mode control. To analyze the FLC, we represent the defuzzified outputs as controller functions when the FLC adopts the isosceles input MF````s and singleton output MF````s. For this purpose, the control rule table is divided into four cases when it is assumed to have the same one as the generalized performance decision table. As a result, defuzzified outputs can be represented as fuzzified linguistic variables. To represent them as controller functions, the relationships between fuzzified linguistic variables are determined from the analysis of the firing region and the individual fuzzified variables are represented as functions of the firing blocks. The acquired controller functions are used to obtain the sufficient conditions to guarantee the bounded-input bounded-output stability using the well-known small gain theorem. Based on these sufficient conditions, we propose the parameter-selection-procedures. To apply the plants with time-varying dead-z...