Designing predictive controllers for a hot strip mill

A complete simulation has been developed to model the behaviour of a hot strip rolling process and the interaction with the mill machine dynamics and controller actions. The numerical solution method for the coupled interstand, rollgap and looper equations, borrowed from fluid dynamics, has been found to be very robust allowing coarse space discretisation. The simulation for six stands runs in about twice real time on a 5OOMHz PC, using a coarse discretisation. A start has been made towards control improvements by using linearised models, derived from the simulation, as a basis for exploring optimal PI tuning (Edmund´s algorithm). This has been done using the Multivariable Frequency Domain toolbox. Further model changes, to gain a better correspondence between real world and simulated results, are possible, the main one being a change from Sims´s roll gap equation to Orowan´s solution. Model based control appears to offer potential benefits for further improvements by making more use of upstream data