Disputation i data- och informationsvetenskap: Mehdi Saman Azari

Abstract

As Industry 4.0 transforms manufacturing through smart, interconnected systems, the complexity of these environments increases vulnerability to faults, posing safety and financial risks. This thesis addresses these challenges by enhancing fault diagnosis using advanced machine learning, self-adaptive models, and digital twin technologies.
Three main objectives guide the study: (1) developing a robust diagnostic framework resilient to fault pattern variability and data imbalance, (2) enabling knowledge transfer across domains using digital twins and transfer learning, and (3) building a self-adaptive system to handle unseen and evolving working conditions.
The proposed solutions significantly improve fault diagnosis accuracy and adaptability, bridging the gap between research and real-world industrial deployment.