Anticipation models for on-line control in steel industry: Methodologies and case study

This paper describes a simulation system according to improve steelmaking's efficiency and to monitor its performances by anticipating the next period workload. Usually the production planning in those cases is made by the use of Gantt diagrams, based on operator's work. This means that if an accident occurs, the operator himself has to change in few minutes the production plan with a lower performance than the original one. The first consideration is obviously that the operator's experience itself it's not sufficient to re-plan a performing steelmaking chain. Hence the necessity of simulation as problem-solving technique in this complex situation. A brief introduction on this paper is devoted to identify the common problems in most plants about production planning, and this is indeed needed to define the boundary conditions and the framework of the problem. Then, a description of steelmaking processes and the general features of critical aspects about steelmaking planning (Paragraph 2) is given in order to understand the bonds, features, criticalities to be analyzed and implemented in the simulation model. In paragraph 3 a detailed analysis of proposed methodology and system architecture is given in order to make the reader understand the complexity that the Authors had to face in modeling the system and the solutions they found with approximations, considerations, techniques and algorithms that were the most suitable to be used in this particular situation. A short description of the likely steelmaking plant modeled and Verification and Validation (V&V) results are carried in paragraph 4. It was in fact very important in such a complex system, to define the acceptability of results in terms of verification of the correctness, validation of the results, and accreditation to the users. This is a generally valid principle in simulation, but moreover in a complex system modeling such a steelmaking process, where an error can cost millions. At last, in paragraph 5, the Authors present the conclusions of this work, with a particular attention to the savings that the simulation can bring to this real system, explaining the results of the case study to which it was applied, and the possible future developments of this research are discussed. This model can anyway be applied to other realities than the one exemplified in this paper. © 2010 American Institute of Physics.