Simulation of a Billet Heating Furnace

Abstract

This work presents the method developed in the scope of the “Audit Furnace” project to support the manufacturing industry in understanding the energy efficiencies of its furnaces and to identify strategies for the continuous improvement of its processes. A digital representation to support the development, calibration, and training of a physical-based reduced-order model for industrial furnaces is sought by integrating experimental data obtained in energy audits performed at several industrial units with detailed numerical results from computational fluid dynamics simulations of the furnaces. Composite models with two blocks, a physics-based reduced-order block, and a machine learning model block, are proposed in order to simultaneously achieve performance and flexibility in its adaptation to different furnaces, while keeping the computational load in acceptable levels. In this paper, preliminary results of the application of the method to a billet heating furnace are presented, namely the results of the computational fluid dynamics simulations of the furnace and their comparison with the measurements performed in an energy audit. This is the first, essential step of the proposed method. The numerical results generated will allow calibrating and training the reduced-order model and will feed the machine learning model training process.