A method based on Bayesian neural networks and genetic algorithm is proposed
to optimally control complex systems. The proposed methodology is
applied to control fermentation process where sugar is converted into alcohol
using yeast, a living organism, as a catalyst. The relationship between
input variables and output variables is modelled using Bayesian neural
networks trained using hybrid Monte Carlo method. A feedback loop
based on genetic algorithm is used to change input variables so that the
output variables are as close to the desired target as possible without
the loss of confidence level on the prediction the network gives.
The objective function constructed in this regard, is a weighted sum of
square of errors between the target and the neural networks output as well
as the confidence levels given by the Bayesian networks. The proposed
procedure is found to reduce the distance between the desired target and
measured outputs significantly.
