Sabena Technics scheduled aircraft line maintenance with MILP and a picture

Aircraft line maintenance needs a costly workforce whose staffing and scheduling is hard to do well by hand at Sabena Technics.

KU Leuven and Vlerick built a mixed-integer linear programming model with a visualization layer that produced many cheaper, alternative schedules.

The visualization let managers evaluate scenarios and raise objections, which the team then incorporated into the model.

Make the optimizer explainable enough to argue with

MILP finds cheaper schedules than manual planning across many scenarios.

Visualization makes the optimizer legible to non-analysts.

Iterating with managers in meetings improved the model's real usefulness.

It produced considerable savings and more efficient use of human resources.

An optimization model works in operations when people can see and contest its output, so iterate on it in the room, not after rollout.

The model resulted in considerable savings and more efficient use of human resources, and the experience showed that an optimization tool's value depends on letting stakeholders challenge its output before it is deployed.