Reducing Uncertainty in Reheat Furnace Conditions Prerolling Using Artificial Intelligence

In reheat or tunnel furnaces, inefficiencies arise from uncertainty in observed conditions and multimachine scheduling, resulting in operators or hard-coded algorithms taking actions which lead to excessive energy consumption. The paper discusses the impact of using dynamic mathematical optimization and reinforcement learning to organize slab schedules and predict heating time considering uncertainties demonstrated in a flat products steel mill. The results show a 10% reduction in heating time, leading to a monthly energy savings of 300 MWh and a reduction of 180 metric tons in carbon emissions, thus contributing significantly to the decrease of the steel industry's environmental footprint.

Authors:

Dr Kwangkyu Alex Yoo | Deep.Meta
 

Dr Osas Omoigiade | Deep.Meta
 

Session Chairs:

Jorge Fernandez | AMI Automation
 

Nancy Hake | NLMK Indiana