An intelligent controller for aluminum smelter potlines

Abstract

The control parameters for the Hall-Heroult electrolytic process to industrially produce aluminum metals are cryolite ratio, current efficiency, and the heat balance which are seen to be dependent on the bath resistance. In this paper, we present an intelligent controller based on the bath resistance process model, using the multilevel intelligent control (MLIAC) architecture. The design is targeted to improve real-time performance of the intelligent control, i.e., through use of a simple model, a priori learning, adaptation based on asynchronous parametric estimation, and an inference mechanism based on structured query language constructs, in a structured data environment. The controllers are modular, distributed, hierarchical, and well coordinated. The simulation studies show that it is possible to achieve higher plant productivity and net energy savings with such a controller.