A Heterogeneous Time-Tracking Fusion and Application to Health Evaluation of Aerospace Engines

Abstract

Heterogeneous information fusion has long been a difficult problem due to the differences in the representation and feature of various physical information. Besides, the multi-sensor signals of large mechanical equipment such as aerospace engines often change in a complicated way during the start-up stage and long-term operation, which makes the multi-sensor fusion based health assessment research impending. To explore a suitable fusion method for the multi-physical signals with different change rates, as well as to monitor the health state of large mechanical equipment based on multi-sensor information, this paper proposes a heterogeneous time-tracking fusion algorithm. Firstly, the time-domain indexes and instantaneous frequencies of the fast-varying harmonic-like signals are obtained by employing index extraction and second-order synchrosqueezing transform respectively, by which the overall and detailed characteristics of the signals are thus obtained. Secondly, after structuring a dynamic time-tracking function consisting of the hyperbolic tangent function and modified arctangent function, the time-dynamic confidence upper limit for fast-varying signals and confidence interval for slow-varying signals are obtained creatively. Finally, the different varying-rate signals are fused into a dynamic normalized time-varying index representing the health state through the aforementioned functions. By applying the proposed method to the health evaluation for ignition start-up stage of gas generators and the long-term performance of the turbopump, its effectiveness and practicability in the aerospace engine health analysis have been validated.