Residence Time Regulation in Chemical Processes: Local Optimal Control Realization by Differential Neural Networks

摘要

A new method to design local optimal controller for uncertain system governed by continuous flow transformations (CFT) is presented. The on-line solution of the adaptive gains adjusting a linear control form yields the calculus of the sub-optimal controller. A special performance index, oriented to solve the transient evolution of CFT systems, is proposed. The class of systems considered in this study is highly uncertain: some components of chemical reactions are no measurable on line and then, they cannot be used in the controller realization. The recovering of this information was executed by a differential neural network (DNN) structure. The ozonation process of a single contaminant (as the particular example of CFT) is evaluated in detail using the control design proposed here.