A neural network controller for a biochemical process

The authors consider control of a nonlinear dynamic process in biochemical engineering. Three state variables considered for the process are the microbial concentration, substrate concentration and product concentration. Product concentration is the controlled variable, and dilution rate is the manipulated variable. The Levenberg-Marquardt method is used to train feedforward neural networks by minimizing the sum of squares of the residuals. The output of each node is calculated by the logistic (sigmoid) or symmetric logarithmoid activation functions on the weighted sum of inputs to that node. Initially all the variables are assumed to be measurable, and all of them are fed in as inputs. Later only the product concentration is fed in as input. The feasibility of using neural networks for controlling a process is demonstrated. Knowledge of the process model is not required