Abstract
The sustainability and decarbonization of processes in the steel industry are enhanced with the valorization of the gas generated during the chemical reactions produced in blast furnaces. However, the combustion of blast furnace gas (BFG) faces the drawback of lower flame stability, which increases the chance of operation shifts towards abnormal conditions and even the flashback or extinction of the flame. Thus, early detection and correction of regime deviations are needed to increase combustion efficiency, for which image-based systems have a high potential. This work focuses on monitoring an industrial furnace for steelmaking processes based on estimating O2 concentration in flue gases using color images captured inside the combustion chamber. An experimental campaign was performed in a 1.2-MW burner to develop the supervision system, using three fuel blends of BFG and natural gas. Images were processed to extract intensity and textural features, which were used to train predictive models based on machine learning algorithms: logistic regression, support vector machines, and artificial neural networks (multilayer perceptron). O2 concentration in flue gases was correctly estimated for at least 97 % of all the test samples and fuel blends. This study shows the potential of image-based systems for the automated control of BFG combustion at the industrial scale.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call