Intelligent real-time alert generation and operational interventions in a closed loop supply chain for repairable parts

Inspired by a case of a logistics service provider in aviation, we develop a model for alert generation and operational interventions within an operational service control tower. Our main contribution lies in using a probabilistic neural network exploiting real time information on the repair progress to obtain the repair lead time distribution. We integrate this distribution in an integer linear programming model that generates alerts, prioritises them, and determines interventions under a constraint on the finite capacity of operational planners. We apply our model to data from an aerospace logistics provider. Our results show that incorporating real time repair status improves lead time prediction accuracy by 7 to 12% compared to benchmark models. In our case study, the intervention model achieves 65% of the total cost benefits within a single planner's daily workload, corresponding to a more than threefold improvement in cost benefit per intervention compared to a non-prioritized first come, first served approach. Our findings show that job-specific lead time prediction significantly improves intervention outcomes, especially under resource constraints. Key drivers of uncertainty, such as external delays, can be proactively managed by embedding real time forecasts into decision making, enabling more effective and cost efficient planning.