Iterative Learning Identification and Design of Industrial Furnace Process Control

Abstract

An applications oriented control systems engineering approach for a class of well- posed thermal systems, e.g. industrial furnaces and ovens, that is consistent with most of theoretical results in systems and control sciences has been elaborated and tested in designing controls for several industrial-scale furnaces. It provides a methodology for iterative learning and resolving process identification and control design for multi-variable systems within a discrete convolution framework and using truncated k-time sequence matrices of characteristic input-output modes as well as their characteristic patterns and singular characteristic patterns, starting with standard non-parametric process time-domain models identified under operating conditions. Within computer process control environment and for practical engineering and maintenance reasons, digital implementations are sought in terms of partial steady-state decoupling and two-term laws or combination of certain MIMO and SISO controls. The pusher furnace in Skopje Steelworks is used to illustrate this methodological approach