The idea of applying intelligent control using neural networks and fuzzy logic to actual plants, where modern control theory is not powerful enough, is becoming popular. This paper presents one example applied to a recovery boiler in paper pulp mill.
A recovery boiler aims to generate steam and to collect expensive soda, which is a waste from chemical reaction at the previous process. Experts still exist to operate the plant efficiently because of its process's complexity. Among them are the operators' actions based on the visual perception of the charbed. In order to reduce these operators' burden, control algorithm handling both conventional process data and image data is discussed in this paper.
The control algorithm,
visual feedback control, has two part. One is the perception of the charbed's shape organized by Neural Networks and the other is fuzzy control system based on operators' experience.
First we discuss the present plant operations done by operators and point out the possibility of automatic manipulation. Next, explain the detailed algorithm including how we implemented neural networks and fuzzy logic. Finally the result of control to the plant is discussed.
Due to the kindness support of a certain user, we have implemented this algorithm and evaluated to the actual plant.
View full abstract