Modelagem térmica de um forno panela utilizando redes neurais artificiais

Abstract

The national steel industries are investing in construction of hydroelectric or thermo electrical power stations in partnership with the energy companies with the objective of energy cost reduction in their business. Then, actions that search alternatives of energy consumption reduction and increase the productivity became priority for research and development. The ladle furnace of V&M is one of the largest consuming units of energy in the steel plant, consuming 2.400 MWh on average a month. Due to the process complexity, system optimization became difficult to be implemented using conventional techniques. However, applications of computational intelligence have been used as auxiliary tools or main tools to modeling process with difficult approach. Due to the low knowledge of ladle furnace dynamics and the high variability of specific energy consumption, the use of neural networks was defined as a modeling tool. This work demonstrates the use of neural networks in complex industrial problems through the steel temperature prediction during the ladle furnace process. This work proves the generalization capability of the neural networks, obtaining smaller medium error than the medium error specified by the instrument of measurement of steel temperature. Besides, this work demonstrates that use of the thermal model will result in productivity increase and operational and energy cost reduction.