Robust Data Reconciliation to Determine Basic Oxygen Furnace Set-points

In the steel industry, the determination of the control system set-points of batch processes is a common problem. It consists in adjusting the set-points in order to reach the given product specifications thanks to a process model. Small changes in operating conditions may impact final product quality. This is particularly true for the Basic Oxygen Furnace (BOF) where the information collected during a specific batch serves to adjust the set-points of the next batch. For being able to control that type of process, measurements must be made coherent and it may be convenient to use data reconciliation procedure. The proposed paper describes a method allowing simultaneous data reconciliation and model parameter estimation. Parameter estimation results can either be used to update the process model or to detect abnormal parameter variations due, e.g. to fouling, corrosion, degradation of parts of the process.