Monitoring Jet Engine Characteristics by a Learning Algorithm

Abstract

Characteristic curves or surfaces of engines are important regarding the safety and the efficiency of operation. Deriving these functions automatically, and then monitoring them constantly against their longer term variations would provide the necessary information required to find the optimal operating points, to detect impending failures or to diagnose after a failure. This paper discusses the use of a learning algorithm for this purpose. The algorithm has originally been used for building heating automation, and in this paper the feasibility and the effects of the algorithm parameters for a jet engine fuel consumption application are investigated after presenting the algorithm and studying some of its fundamental properties. The engine characteristic used in simulations is that of a typical turbofan aircraft engine, and its Thrust—Specific—Fuel—Consumption is derived against the Mach number and the thrust as the independent variables, serving as an example and also illustrating some further properties of the algorithm.