Monitoring the blast furnace hearth Detecting the end of the cast and wear of the hearth

The coming decade, the blast furnace remains a cornerstone of global steel production, yet its operation must be optimized to improve safety, efficiency, and environmental performance during the transition towards low carbon steelmaking. This thesis addresses two critical challenges in blast furnace hearth management: the reliable detection of the end of the casting cycle and the accurate estimation of refractory wear.
For end-of-cast detection, acoustic monitoring techniques are developed to identify the onset of gas release through the taphole, a phenomenon associated with unstable flow. Two approaches are proposed: a spectral analysis method targeting high-frequency acoustic signatures, and a data-driven method based on machine learning for the classification of transient acoustic events associated with bursting gas bubbles. The methods are validated using measurements from industrial blast furnaces and demonstrate robust and early detection capabilities. In addition, a visual monitoring approach based on image analysis is implemented to provide a comparative benchmark.
For refractory wear estimation, limitations inherent to existing acousto-ultrasonic echo techniques and thermal inverse models are addressed. A two-dimensional, multi-sensor inversion framework for elastic wave measurements is introduced, enabling spatially resolved and reproducible estimates of hearth lining thickness while reducing operator dependency.
The results demonstrate that the integration of acoustic sensing, data-driven methodologies, and advanced physical modeling enhances the monitoring of blast furnace hearth conditions, thereby supporting improved operational safety, reduced environmental impact, and extended campaign lifetimes.