Artificial intelligence for process innovation: data-driven modeling, optimization and monitoring in plasma physics and materials processing.

The rapid expansion of Industry 4.0 and the increasing interconnectivity of devices and systems have enabled access to large and complex datasets, creating new opportunities as well as challenges for scientific and industrial innovation. Effectively extracting meaningful and actionable knowledge requires systematic data management - including data collection, modeling, and interpretation - combined with multidisciplinary expertise to identify relevant information and uncover significant patterns. Artificial Intelligence (AI) offers a strategic solution to these challenges, supporting data analysis, interpretation, and process optimization across a wide range of domains, including manufacturing, energy management and materials science. Among AI approaches, Machine Learning (ML), Deep Learning (DL), and Explainable AI (XAI) are particularly relevant for modeling complex physical and industrial processes while maintaining interpretability. This PhD thesis investigates the integration of AI-based tools into process design and monitoring, with the objective of improving product quality and favouring undesirable behaviours detection in non-thermal plasma, fusion, and material surface modification. The study combines a review of the theoretical foundation of experimental design strategies and AI methodologies with the development of a modular conceptual framework for scientific and industrial data management, providing a coherent methodological approach for managing datasets and process control. The tools here presented have been applied to several case studies, including laser ablation processes, automated Large Language Models (LLM)-based data extraction for plasma actuators, active learning for optimal experimental data selection in argon plasma discharge, and anomaly detection in plasma fusion systems. These applications demonstrate the versatility of the framework and highlight how AI can address the complexity of modern research and production systems, enhancing both efficiency and interpretability.