Integrating Augmented Reality in Knowledge-Based Maintenance: A Layered Architecture for Informed Decision Support in Metro Systems

Current railway maintenance faces increasing demands to ensure reliability, efficiency, and safety, creating a need for informed decision-making. Knowledge-Based Maintenance (KBM) improves maintenance decision quality by exploiting existing knowledge and uncovering new knowledge through diagnostics and prediction, thereby integrating predictive and prescriptive maintenance into a coherent decision-support framework. Augmented Reality (AR) facilitates intelligent information provision to human operators and knowledge transfer in socio-technical systems. Focusing on the literature of asset management, the potentials for integration of KBM and AR remains unexhausted. A significant research gap is how to effectively deliver and capture real-time, contextualised knowledge through AR interfaces for frontline operators. This paper addresses this gap by presenting PriM-AR, a layered architecture that integrates AR into KBM. The architecture comprises five layers: Data (data acquisition and pre-processing), Knowledge (data analytics and semantification), Decision (prescription and reasoning), Visualisation (AR environment), and Feedback (adaptive learning). PriM-AR addresses the lack of systematic data-to-AR mapping and closes the gap of integrating user–system feedback into KBM, enabling informed and adaptive decision-making. The architecture is conceptually applied to a metro maintenance case study and evaluated by domain experts.