Comparison Research of Feature Fusion Methods in Characterizing Performance Degradation of Aeroengine

摘要

Advanced characterization techniques of aero-engine performance degradation play an important role in the aero-engine health management system, which remains to be a great challenge. Monitoring parameter from one signal source usually contains limited health information of the operation conditions, which is probably not suitable for describing the system-level complex equipment. Feature fusion methods can well fuse multisource monitoring parameters, and they are usually applied to overcome the shortcomings of single-source parameters. However, how to select the sensitive single-source parameters of aero-engine for effective fusion has not been fully studied. In this paper, comparison research of six types of feature fusion methods are performed to characterize the performance degradation of aero-engine in this paper, including kernel principal component analysis (KPCA), locality preserving projections (LPP), locally linear embedding (LLE), linear local tangent space alignment (LLTSA) and so on. The effectiveness and differences of these six feature fusion methods are preliminarily investigated with the simulated multi-source performance parameters of aero-engines provided by NASA.