Visual mining and statistics for a turbofan engine fleet

Snecma, as a turbofan manufacturer, needs to deal with a wide fleet of more than thousands of engines. Every day, data from aircraft engines are broadcasted to the ground. Some airlines companies rely on their engine manufacturer to control the engines´ behavior and help prepare for maintenance scheduling. The goal of the manufacturer is to detect abnormalities to help schedule maintenance operations. The advantage of the manufacturer as MRO operator is the registered memory of all past events that appears on its fleet of engines. If one opens the possibility to look in this huge amount of data for corresponding similar behaviors, which may have append in the past (for all engines of all customer companies), it becomes possible to make some targeted statistics of the future. This paper describes an algorithm to help engineers looking at some wear indicators and proposes a new full automatic method able to fetch information stored from many years of operations. This methodology is based on mathematic foundations. It uses a formalization of engine trajectories; creates a metric space where it becomes possible to compare time intervals of the evolution of engines. A generic maintenance application uses this methodology and finds in the fleet database engines that were similar in behavior to build future statistics.