A probabilistic approach to analyse Blade Tip Timing data of non-synchronous vibrations under constant rotor speed

Blades are among the most critical components of turbomachines, their monitoring and characterization undergoing working conditions are fundamental for the insiders, both for preventing eventual breakage and for optimising future development. Two approaches are possible for monitoring rotor blade vibrations: a traditional one based on the use of strain gauges and another one called Blade Tip Timing (BTT). BTT is an indirect, non-intrusive simple and robust measurement method, but the processing of such data is not easy because they are often subsampled with respect to the Nyquist limit and the ordering of the samples is not unique. In this work the focus is on multi component non-synchronous vibrations, typical for example of flutter, measured at constant rotor speed by a BTT system. These data are organized into batches of fixed length called snapshots and they are interpreted as members of a random vector. When the signal contains only one harmonic component the frequency can be determined using a method here described and called Harmonic Matching (HM). While for the analyses of multi harmonic component vibrations a probabilistic approach capable of separating and identify the components using Principal Component Analysis (PCA) and Independent Component Analysis (ICA) is proposed. For the development of data processing methods, the possibility of having controllable and repeatable data is fundamental, for this reason two test rigs of increasing complexity have been developed and are here described.