Adaptive geno-fuzzy control of process plants via perpetual evolution

This paper presents an adaptive mechanism based on genetic algorithms for basic fuzzy controllers. The concept of "perpetual evolution" is proposed as a novel strategy for using genetic search to tune fuzzy controllers in real-time. In this way, a population of potential fuzzy controllers evolves continuously over the time responding to changes in the environment. At a given time, the best fitting controller in the latest population is accepted as the solution. Adaptability for any real-time process changes is achieved by using an online neural network model for evaluating trial solutions. The proposed technique is demonstrated through simulations by applying to the control a chemical reactor.