Turbofan engines generate a lot of data for maintenance purpose. During each flight the aircraft send information to the ground using small messages. On those messages one can find a description of the engine behavior (shaft speed, oil temperature, pressures, etc.) and the observation context (air temperature, aircraft attitude, altitude, etc.). Mainly those signals are used for trend analysis which goal is wear detection and scheduling for shop visits. But the temporal curves obtained this way also hide some very interesting fleeting events that may be connected to sudden changes in the turbofan configuration. The changes of our interest are buried in the signal noise and are often hidden by the flight context variation. To reveal those events one uses two successive original algorithms. The first one resolves the context dependency and the second filters the signal using change detection. The filtered signal with change information is compared to the flight log-book for validation purpose. One detects almost all known problems that were cause of maintenance operation but the algorithm also finds some unsuspected changes now under investigation.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below