Explainable predictive analytics and decision support for real-time control room management

Korte inhoud:Control rooms serve as crucial centers for real-time monitoring and intervention in a wide range of industries, including rail and air traffic, nuclear power plants, chemical production, and healthcare services. The ongoing digitization over the past decades transformed control rooms into centers where decision-makers interact with technology to manage environments or processes in real-time. These digitized environments provide considerable opportunities for researching data-driven technologies. With the rise in data availability, this dissertation studies the untapped potential in harnessing artificial intelligence (AI) and advanced analytics tailored to the unique needs of decision-makers in these highly digitized environments. Literature indicates that data-driven technologies, which leverage a vast amount of data, can augment decision-making processes to a great extent, as they can cope with a large amount of data, analyze it in real-time, and provide actionable insights to operators and managers. This enables proactive responses to emerging challenges and opportunities. However, these technologies should be carefully developed and implemented, and must incorporate the human decision-maker at the core. The research efforts position itself within Industry 5.0, proposing a human-centric approach for the development of new technologies. Next to the research initiatives, a great deal of focus goes into the valorization of the research findings into a proof of concept (POC) for decision support in control rooms. The dissertation includes both quantitative and qualitative research initiatives to ensure the incorporation of control room employees' needs into the POC. The research is conducted at the Traffic Control Centers (TCCs) of Infrabel, Belgium’s railway infrastructure company, where TCC operators manage the railway traffic on a 24/7 basis. The real-world data has been purposefully captured for a realistic representation of the actual setting. For the research, the data is anonymized to respect the privacy of the employees. The quantitative studies propose explainable analytics and decision support for train delays, employee workload, and automation usage. The POC embeds the insights from the quantitative studies with a focus on the user and the potential operational impact. The end-user feedback on the POC sheds light on the predominantly positive user perception of the POC but indicates the need for further validation and refinement of the POC. Furthermore, this dissertation proposes a framework for user-centric implementation of business analytics to increase business value and operational impact.

Jaar van publicatie:2024

Toegankelijkheid:Closed