Fuzzy Neural Network for State Estimation and Optimal Control in L-Isoleucine Fermentation

Abstract

A state pre-estimation model of L-isoleucine fermentation is established by using neural network. The chemotaxis BP algorithm that can effectively overcome the local minimum is adopted for training the net. A neural network that equals a fuzzy system is applied for getting air flow to control dissolved oxygen.the membership functions are determined by the net learning. A new control method based on neural networks and fuzzy logic is proposed. This scheme can fully make use of expert knowledge and is robust for uncertainty in model and accurate with noise immunity.