Development of AI-based diagnostic systems applied to muon radiography for monitoring applications

In recent years, the application of muon radiography has gained considerable traction across various fields due to its non-invasive and highly effective imaging capabilities. Originating from the study of cosmic rays, muon radiography leverages naturally occurring muons generated in the upper atmosphere to penetrate dense materials. This imaging technique has proven to be particularly advantageous in contexts where conventional methods struggle due to environmental constraints or the need for non-destructive analysis. Building on this premise, this thesis explores and advances the implementation of muon radiography, applying it to both industrial and geological settings. This doctoral research focuses on two primary objectives. The first is the development and deployment of a scintillating bar detector system for the BLEMAB project. BLEMAB (Blast Furnace Stack Density Estimation through Muon Absorption Measurements) represents a significant advancement in the use of muon radiography for real-time monitoring of density variations in blast furnaces. The thesis delves into the design, construction, and installation of this innovative detector, specially engineered to endure the harsh conditions within an industrial setting. Through muon absorption, the detector enables precise density mapping, a capability that holds the potential to revolutionize blast furnace management by offering insights into internal processes that were previously difficult to observe. The second objective centers on applying AI-based diagnostic systems to muon radiography applications for monitoring of geological structures. By focusing on tailing dams and underground mining environments, this thesis demonstrates that the use o machine learning techniques, combined with muon radiography, can lead to promising results in high-density environments where traditional methods may fall short. Through the detection of water infiltration in tailing dams and the segmentation of underground cavities in mining applications, this research employs muon transmission data coupled with advanced machine learning algorithms, such as Random Forest and UNet, to enhance imaging accuracy and segmentation. Given the need for a large amount of data to train the aforementioned models and the limited availability of databases with muographic measurements, a fundamental part of this thesis was the development of software capable of replicating a muographic measurement in a short time. The work addresses critical issues of structural integrity and environmental safety, offering a framework for early detection of potential hazards. In both industrial and geological settings, the thesis aims to contribute with innovative methodologies and practical solutions, expanding the scope of muon imaging and establishing it as a critical tool for structural and environmental monitoring. This thesis work is organized as following. Chapter 1 introduces the principles of cosmic rays and muon interactions with matter, providing the foundational knowledge for understanding muon radiography. Chapter 2 discusses the BLEMAB project, outlining the design and capabilities of the scintillating bar detectors developed for blast furnace monitoring. Chapter 3 and 4 addresses the challenges of muon radiography in geological structures, illustrating the novel use of AI and machine learning for water infiltration detection and cavity mapping in mines. Lastly, Chapter 5 summarizes the contributions of this research, presents conclusions, and suggests directions for future exploration in the field of muon radiography.