Typically several hundred million data points arise from a comprehensive measurement campaign carried out in a centrifugal compressor test rig with the FRAP® system (see Part 1). In order to obtain a maximum of information about the unsteady flow at any position in this turbo machine the time-resolved data processing method has to be optimized. In contrast to the standard time-averaged flow measurements with pneumatic probes, the objective of FRAP® measurements and of data processing is to extract novel information about crucial unsteady phenomena like turbulence, row-to-row interaction, modal or rotating stall, leakage flow effects etc. In such cases the simultaneous measurement of static and total pressures and flow vectors is of particular interest. Novel information means the analysis of averaged and time-resolved (wavelet) spectra, autocorrelations or time averages properly conserving physical fluxes, etc..
Different averaging methods are applied to compress the time dependent data measured by a 1-sensor-probe (see Part 2) in a centrifugal compressor. Such results could be used for comparison with pneumatic sensor measurements and CFD calculations. The comparison of averaging methods includes the averaging theories by Traupel and by Dzung which are compared to simple arithmetic time averaging. From there the specific stage work is calculated.
In analysing the time dependency several ensemble-averaging procedures for flow pressure and velocity are utilized for separating deterministic from stochastic fluctuations, extracting blade row finger prints or investigating low-frequency surge type fluctuations.
With respect to the selection and overall optimization of data processing methods an overview of generic tools is given and the modularity of the processing procedures is discussed.
