Industrial furnace control: an experiment in iterative identification and design

A set of linear multivariable models for a 20 MW, large industrial, gas-fired furnace, used in the production of welded steel pipes, has been developed and identified on the grounds of experimental step and PRBS tests. Uncertainty within models is represented in terms of delays and steady-state gains. Iterative identification and control design investigation has been applied to arrive at a feasible and realistic solution for this multivariable process control problem. Extensive simulation experiments were carried out using the convolution-based simulation language, and a comparison analysis of several control system design strategies has led to a couple of implementable alternatives. A robust multivariable computer process control has been synthesised that gives satisfactory transient responses and minimal fuel consumption during furnace load changes. Stability property or ultimate boundedness of outputs has been preserved within all designs, both the ideal nonfeasible and the realistic feasible ones.