Processing Multicomponent Blade Tip Timing Experimental Data by Independent Component Analysis

Abstract

Blade tip timing (BTT) is a simple, robust, and nonintrusive method for measuring the rotor blade vibrations in turbomachinery. Using this method, the analysis of vibrations characterized by multiple spectral components typical, for example, of flutter is challenging. This chapter proposes a probabilistic method able to separate and identify multiple harmonic components sampled according to a BTT-like schema. The data are divided into batches of fixed length called snapshots, which are interpreted as realization of random vectors. The statistical properties of the subspace spanned by such random vectors is used to identify the number of components present in the signal (i.e., number of active modes), to separate the components and estimate their frequency and amplitude. These results are obtained by applying sequentially the principal component analysis (PCA), the independent component analysis (ICA), and the harmonic matching (HM). The proposed technique is applied to experimental data obtained from a test rig with fixed disk and traveling load. The time-resolved measurements are resampled according to the sampling pattern induced by specified sensor spacing.