Refueling outages are complex, high-consequence operations involving tightly coupled tasks, shared resources, regulatory constraints, and significant uncertainty. Unplanned schedule extensions incur substantial financial penalties and introduce additional operational and safety risks. This paper presents a data analytics framework for risk-informed outage management, integrating scheduling, uncertainty quantification, and decision support.
First, outage scheduling is formulated as a Resource-Constrained Project Scheduling Problem (RCPSP) that explicitly represents workforce skill constraints, equipment availability, system configurations, radiation dose limits, and spatial interdependencies. This formulation captures the combinatorial complexity and operational coupling inherent in outage execution, enabling more realistic schedule generation than deterministic critical path methods.
Second, schedule uncertainty is modeled through probabilistic activity durations and Monte Carlo propagation. Historical work order data are retrieved via semantic similarity and used to construct empirical duration distributions, enabling estimation of completion risk metrics, criticality indices, and path sensitivity. These outputs support risk-informed contingency allocation and prioritization.
Third, emergent work is addressed through a decision-support pipeline combining knowledge graph–based timeline analysis, temporal causal reasoning, and schedule impact evaluation. The framework generates evidence-traceable recommendations (ESCALATE/PROCEED/DEFER), supporting decision-making under uncertainty. A companion module proactively identifies high-risk components from maintenance and condition report data.
This integrated approach advances outage management from state of practice deterministic scheduling toward probabilistic, evidence-based decision-making, improving robustness, transparency, and flexibly.