Extraction of operating know-how of experienced operators using neural networks and its application to PID and neuro-fuzzy hierarchical controller

The paper discusses an extraction of operating know-how of experienced operators in the framework of data mining. A neural network is used to extract the know-how of controlling a receiver level of a petroleum plant. This neural network is installed into a neuro-fuzzy and PID hierarchical controller. The neuro-fuzzy system works as a supervisor of the conventional PID controller by generating the control target of the controller. This neuro-fuzzy and PID hierarchical controller system is successfully applied to the control of receiver level of a desulfurizing plant in the transition stage of feed oil switching. The results of this application are shown. The reboiler energy of the plant is also significantly reduced.