Optimal control and scheduling of an experimental laboratory-scale hot-rolling mill using intelligent systems-based paradigms.

A novel mechanism for the optimal scheduling and the control of the hot-rolling of steel is presented in this work. Such a mechanism provides optimal rolling parameters to set-up an experimental laboratory-scale hot-rolling mill and thus produce metals with desired microstructural and mechanical characteristics. The proposed methodology combines physically-based models and those associated with 'intelligence' such as Neural Networks, Fuzzy Systems, and Genetic Algorithms, to systematically calculate the optimal rolling schedule so as to achieve a right-first-time production of steel alloys, a challenge for academia in general. Unlike current design methods, the scheduling problem is here
treated as an optimisation problem which aims at satisfying multiple process objectives and a set of user-defined requirements. Such objectives are expressed in terms of the quantitative elements of the steel microstructure and its mechanical properties' such as strength and toughness. Three main aspects are considered to define and approach the optimisation problem: (1) THE PROCESS MODEL, which includes integrated knowledge of the stock microstructure, the mechanical properties, and the processing route; (2) THE PHYSICAL CONSTRAINTS associated with the metal due to its chemical composition, as well as the mill operating limitations; and (3) a set of OPTIMALITY CRITERIA which is used as a performance index to evaluate the quality and feasibility of each solution. In order to show the efficacy of this methodology, extensive experimental studies, metallographic analyses, and laboratory mechanical tests, are presented using the commercial type C-Mn Steel alloy (Bright Mild Steel) grade 080A15. The results from such experimental studies showed that the final product was in good agreement with the desired design in terms of the microstructure and the mechanical properties. The experiment results also demonstrated the advantages of the proposed methodology over current methods which are
generally ad hoc and lack adequate capabilities for finding the optimal process parameters. The software SISSCOR is also introduced as a friendly graphical user interface for a fast experiment design and analysis of the dynamic performance associated with the rolling mill.
This work will also review the application of a modified model-based approach in the form of the Generalised Predictive Control to reflect a Fuzzy Model of the mill. Such a hybrid strategy was implemented in order to provide robustness and flexibility to the overall control system, and to guarantee an optimal control performance during real-time operations.