The steel industry is changing and demands for cleaner raw materials is pressuring suppliers to
develop newer, more effective and sustainable methods for iron ore beneficiation.
This thesis investigates the possibility of developing a new capability for automating process control
of gravity spirals not only for better product quality but also for balancing between yield and
quantity, as those two have a somewhat inverse relationship.
The research begins by exploring methods of computer vision and machine learning in one of
the main beneficiation processes of the plant, combined with time series data and a collaborative
interdisciplinary effort to empower machine learning models to deliver real-time decision support.
While still providing operators the illusion of control.
Additionally, the thesis investigates the benefits of remote operations, leveraging cloud-based
computing and data analysis by migrating computing and storage resources to the cloud, the
plant gains agility and scalability, essential for adjusting to the iterative nature of capability
development. Remote operations not only streamline processes, but also offer environmental
advantages by reducing on-site infrastructure.
Sustainability remains a core theme and recognizes the importance of aligning production goals
with environmental regulations.