Model Based Diagnostics of an Aircraft Generator Using AAKR and SPRT

Electrical power generation is an important component in the Electrical Power System of an Aircraft (EPS). In this paper we present a model based diagnostic framework for early generator degradation detection and tracking within an Aircraft Generator. The nominal health state is modeled empirically using an Auto-associative Kernel Regression (AAKR) technique using signals extracted from a healthy generator. Later the health state is estimated by comparing sensor observations with the model predictions. Finally, a Sequential Probability Ratio Test (SPRT) is used to detect and track degradation. This model based framework showed excellent degradation tracking performance when it was tested on a unit that was run to failure.