Digital Twin Models and Novel Electromagnetic Sensors for Zero Waste Steel Production

The pursuit of sustainable steel production is crucial for reducing environmental impact and advancing process efficiency. Achieving these goals requires innovative approaches such as virtual models for optimization and waste reduction during design phases, and digital twins combined with predictive maintenance for real-time process monitoring and optimization. This thesis contributes to these objectives by exploring novel numerical methods for faster electromagnetic simulations and developing advanced electromagnetic sensors tailored for steel production processes. Building on the existing VINCO framework, which leverages integral methods for solving eddy current problems, this research investigates replacing the Fast Multipole Method (FMM) with Adaptive Cross Approximation (ACA) and Fast Fourier Transform (FFT). These alternative techniques are evaluated in terms of computational performance and accuracy, demonstrating the versatility of the framework for tackling different problems with the adequate compression approach. In parallel, this thesis addresses the need for reliable process monitoring by designing novel electromagnetic sensors for use in steel production environments. Inductive Position Sensors (IPS) are developed for measurement of distance and tilt angles, incorporating Machine Learning (ML) techniques for robust position estimation. These sensors are designed to operate effectively in the harsh conditions of steel manufacturing, ensuring accuracy and reliability. Furthermore, Magnetic Inductive Tomography (MIT) is explored as a contactless technique for imaging the internal state of steel billets during production. MIT provides valuable insights into critical process parameters, such as the temperature distribution and phase transitions within the billet, enabling control over the production process.