[5D2] Simulation of the behaviour of engines in their current state of wear
A Madane¹,² and J Lacaille²
¹University of Versailles Saint-Quentin-en-Yvelines, France
²Safran Aircraft Engines, France
Aeronautical data is increasingly available thanks to new ways of downloading and storing. In addition, the computer clusters on which we capitalise this data become effective for the implementation of machine learning and artificial intelligence calculations. However, our companies, which focus on aeronautical technical solutions, find it difficult to exploit this data and to take advantage of the academic skills of our university laboratories. Indeed, this same data remains the property of the airlines and is under contract, unfortunately preventing it from being exchanged. Indeed, whenever we present the results of algorithms developed in-house, the laboratories insist that we open the data that would enable them to demonstrate the effectiveness of their methods. At Safran Aircraft Engines, we were able to demonstrate the effectiveness of using time-series downloaded from the engines after each flight to build a representative model of the engine using a conditional generative neural algorithm (CGAN). Subject to the flight conditions and controls, this model simulates the behaviour of the engine in its current state as if it had performed the simulated flight. It is therefore possible for us to provide virtual flights performed by our engines in their actual state of wear. These simulations pave the way for sharing open datasets, which we hope will influence research in the discovery of new techniques for monitoring our engines.
