Online Normalization Algorithm for Engine Turbofan Monitoring

To understand the behavior of a turbofan engine, one first needs to deal with the variety of data acquisition contexts. Each time a set of measurements is acquired, and such set may account for tens of parameters, the aircraft evolves in a specific flight mode. A diagnostic of the engine behavior models the observations and tests if anything appears as expected. A model of the engine measurement vector may be very complex to produce and even more to deploy on board. The idea is to solve the problem locally on recurrent phases on which each single problem may be easier to answer. Civil flight missions are straightforward to decompose as they are very recurrent. It is more difficult with military missions and bench tests. Once a set of phases is defined, local regression models may be built. To solve nonlinearities a selection of computed variables is a good approach but such algorithm needs the definition of a stable set of recurrent phases and a very complex learning procedure that uses a huge amount of memory to deal with the high dimensionality of the problem. Such algorithm is very powerful but is not adapted for an online use. Our new solution does not require the a priori knowledge of recurrent phases; it learns recurrent contexts on the fly and adapts a small local regression model on a selected optimal subspace. The application of this algorithm seems to be efficient on long term flight trend monitoring and on real time test bench measurements. It solves the memory problem for calibration by an iterative autoadaptive procedure and suppress the need of preliminary computations of specific parameter as it auto-adapts itself with piecewise linear models.