[3E1] Distinguished overview speaker: Compare engines objectively
J Lacaille
Safran Aircraft Engines, France
In practice, we have very little data with anomalies. Direct methods based on learning are therefore fairly difficult to implement. In general, we model normal behaviour and try to detect degradation by analysing deviations from expected behaviour. Our idea is to monitor the current state of the motor or machine tool in order to evaluate either quality indicators or to simulate the operation of the system in a specific context. These simulators, developed by PhD students at DalaLab, use generative artificial intelligence (AI) techniques. They offer a unique opportunity to compare different engines or machines under identical operating conditions.
This technique has already been tried and tested in micro-fabrication, where fabs have workshops with several similar machines. The wafers (discs on which the components are etched) pass through one of these machines, randomly on their manufacturing route according to their availability. It was possible to score each machine according to its efficiency measured in terms of production output. The scores presented in the form of bar charts by workshop compare the machine tools and inform operators about the optimal changes to the recipes of each machine in a workshop to bring them up to the level of the best.
Conclusions: by analogy, we can optimise the engine maintenance schedule by identifying the most worn engines; with an objective measure of quality, it is also possible to compare different aircraft fleets; detecting anomalies becomes very simple, as all you have to do is compare the actual signal with the simulation result and it can even be made more reliable using statistically validated confidence tubes; and after each maintenance operation, we can also measure its effectiveness on virtual test flights.
This technique has already been tried and tested in micro-fabrication, where fabs have workshops with several similar machines. The wafers (discs on which the components are etched) pass through one of these machines, randomly on their manufacturing route according to their availability. It was possible to score each machine according to its efficiency measured in terms of production output. The scores presented in the form of bar charts by workshop compare the machine tools and inform operators about the optimal changes to the recipes of each machine in a workshop to bring them up to the level of the best.
Conclusions: by analogy, we can optimise the engine maintenance schedule by identifying the most worn engines; with an objective measure of quality, it is also possible to compare different aircraft fleets; detecting anomalies becomes very simple, as all you have to do is compare the actual signal with the simulation result and it can even be made more reliable using statistically validated confidence tubes; and after each maintenance operation, we can also measure its effectiveness on virtual test flights.