ABSTRACT In this paper is presented ANFIS (Adaptive Network based Fuzzy Inference System) method of modeling of pH neutralisator. Data that has been used for training of ANFIS are obtained by performing experiments on a real pH neutralisator. The mathematical model of the pH neutralisator is strongly nonlinear so it is advisable to use ANFIS, since ANFIS is an universal approximator. Inputs of the process (control variables) are flow of acid and hydroxide. The working temperature of the process is 128C, with instaled pH meter of operating range 2-12 pH with combined glass electrode in the upper part of the neutralizer, 20 cm bellow the surface. Then using Genetic Algorithms (GA) parameters of PID controller of pH value has been adjusted in optimal way. Since GA is method of search of global optimum, achieved solution is optimal according to integral criterion of performance. Simulation results show utility of the proposed methodology.

In this paper a two-step design methodology of the near optimal Mamdani type fuzzy logic controller (FLC) has been applied to three types of systems: a linear time invariant system (LTI), a LTI system with time delay and an nonlinear system. In the first step, the tuning/learning procedure of a data base/knowledge base of the FLC is based on the model of the system using genetic algorithms (GA). The achieved solution is the optimal one with regard to the model of the system. In the second step experiments are performed on the real system at the vicinity of the optimum (achieved by GA) and using response surface methodology (RSM) control parameters are readjusted in order to achieve the near optimal solution for the real system. The proposed two step methodology gives a systematic way of the near optimal FLC tuning/learning when confronted with the real system and a very efficient combination of off-line and online part of design procedure.