Each day, planners at SNCF Voyageurs face a complex challenge: scheduling urgent maintenance for passenger trains while minimizing service disruption. Modern rolling stock units are equipped with a signalling system that monitors the health of its components and transmits status signals. These signals are used to generate a list of predictive and corrective maintenance operations. The planner's task is to assign these operations to limited depot tracks within short time windows determined by the rolling stock units schedules. This process is currently done manually and under significant time pressure. Scheduling any maintenance operation involves allocating it to a specific track at a maintenance depot for a specific period of time. The core challenge is to efficiently schedule these maintenance operations as early as possible within a short-term horizon, respecting all routes and resource constraints. 

This work is the result of a collaboration between the LIFAT (Laboratoire d'Informatique Fondamentale et Appliquée de Tours) and SNCF Voyageurs. We present a decision-support tool built around a Recovering Beam Search (RBS) heuristic designed specifically to tackle this challenge. The tool finds high-quality schedules and suggests specific actions such as exchanging trips between rolling stock units to create earlier maintenance opportunities. Initial feedback indicates the tool can reduce planning time by up to 30\% compared to manual methods while improving the efficiency of the planning.