Nuclear operating procedures govern operator actions across normal, abnormal, and emergency conditions, forming the backbone of safe and reliable plant operation. While recent advances in artificial intelligence have enabled the transformation of these procedures into structured knowledge graphs (KGs), their potential as dynamic, data-generating models for analysis and decision support remains largely unexplored. This work introduces a novel framework that leverages procedure-derived KGs to enable scenario-driven simulation and reduced-order modeling of procedural execution. Within this framework, the KGs are not simply static representations of procedural logic but generative structures from which diverse execution trajectories can be simulated under varying plant conditions. These trajectories capture the combinatorial space of operator actions, including conditional branching, response-not-obtained pathways, and interprocedure transitions. To extract dominant behavioral patterns from the KGs, we construct a data-driven reduced-order model by embedding execution trajectories into a high-dimensional step space and applying singular value decomposition. Modeling results show that procedural executions exhibit strong low-dimensional structure, with a small number of modes capturing most of the variability across scenarios. These modes correspond to interpretable operational pathways and enable compact representation and efficient analysis of complex procedural behavior. The framework does more than analyze a single procedure; it also preserves and exploits connections across multiple procedures, allowing for coordinated simulation and system-level reasoning across the procedure network. This capability enables researchers to study cross-procedure interactions, transition dynamics, and emergent behaviors that are not observable within isolated procedure models.