Design of a pH controller using a neural network optimized by genetic algorithm

In new design of chemical processes, it is required to design both a reactor and a control system concurrently based on a few process informations. Applications of neural networks (NN) are very attractive in control of the non-linear chemical reactors. Here it is considered that genetic algorithm (GA) is a powerful tool in optimizing NN controllers without teacher data. The purpose of this study is to investigate on a GANN (NN optimized by GA) pH control system from the viewpoint of both control performance and flexibility. In results, there is found an appropriate value as to the number of input time series data and hidden units. GA can automatically optimize the number of hidden units, and therefore GANN is effective in design of powerful NN controllers.