This thesis focuses on fuzzy tuning systems for process control applications. Schemes are developed to tune the multicriteria fuzzy control (MFC) parameters, the scaling factors of the fuzzy logic controllers and the proportional-integral-derivative (PID) controllers.
A fuzzy tuning system is designed for the MFC, in which the λ-fuzzy measure used in the MFC is optimized by the evolutionary programming (EP) scheme. The basic idea behind the MFC is to analyze the attributes associated with the system output response and, to apply the fuzzy measure and integral theory to the existing fuzzy logic controller. Time-domain specifications such as rise time, overshoot, and settling time are used as attributes in the fuzzy integral evaluation. Through MFC we can obtain desired output responses by tuning the fuzzy measures associated with the attributes. The MFC scheme is easy to implement in practice by adding weight terms to the existing fuzzy rules. The fuzzy integral can be used to assign a weight to each rule. The weights are associated with an aggregation of partial evaluations of the attributes and a set of fuzzy measures of the attributes denoting the degrees of importance. EP can optimize the λ-fuzzy measures used in the MFC. EP belongs to the class of stochastic optimization techniques commonly known as evolutionary algorithms. The λ-fuzzy measures provide a linear tuning property for the fuzzy tuning system. Literally, it is very hard to find the best set of λ values for the λ-fuzzy measures which can guarantee the MFC tuning scheme to be linear in nature. As described above, the output response with the MFC can be tuned by a set of fuzzy measures operating on the elements of the time-domain specifications. The fuzzy tuning system developed can examine the features of each transient response and can calculate the corresponding fuzzy measures. The tuning system can suggest a new set of fuzzy measures which is more appropriate for the given time-domain specifications. The developed scheme can simplify the tuning process and can reduce the time taken for tuning. The viability and superiority of the proposed schemes are demonstrated through computer simulations and are verified experimentally.
Fuzzy logic control has been applied successfully to many industrial processes. However, the selection of a set of suitable scaling factors for the controller (before the fuzzification stage and after the defuzzification stage) still remains as a trial and error approach. The fuzzy tuning system described in this thesis determines a new set of scaling factors which can provide a more desirable time response. The scheme followed makes use of the features of each transient response and the corresponding controller parameters. EP is employed to optimize the rules used in the fuzzy tuning system. The developed scheme can determine and implement a different set of scaling factors for the operating regime. This system simplifies the tuning process and brings down the tuning time associated with the learning phase of the fuzzy logic controller. On a similar way, a fuzzy tuning system is proposed for tuning the PID controllers. The performance of these schemes are verified through computer simulation and experimental studies.