Institute of Fundamental Technological Research

Blade Tip Timing (BTT) is a non-intrusive method to measure blade vibration in turbomachinery. Time of Arrival (TOA) is recorded when a blade is passing a stationary sensor. The measurement data, in form of undersampled (aliased) tip-deflection signal, are difficult to analyze with standard signal processing methods like digital filters or Fourier Transform. Several indirect methods are applied to process TOA sequences, such as reconstruction of aliased spectrum and Least-Squares Fitting to harmonic oscillator model. We used standard sine fitting algorithms provided by IEEE-STD-1057 to estimate blade vibration parameters. Blade-tip displacement was simulated in time domain using SDOF model, sampled by stationary sensors and then processed by the sinefit.m toolkit. We evaluated several configurations of different sensor placement, noise level and number of data. Results of the linear sine fitting, performed with the frequency known a priori, were compared with the non-linear ones. Some of non-linear iterations were not convergent. The algorithms and testing results are aimed to be used in analysis of asynchronous blade vibration.

turbomachinery, blade vibration, tip-timing, sine fitting, least squares, harmonic oscillator

The uncertainty of non-contact blade vibration measurements depends on the precision of time-of-arrival (TOA) or phase measurement. Various TOA estimation methods are used in the existing tip-timing measurement systems, which contribute to differences in vibration amplitude results. The known hardware triggering circuits have structure related to the sensor type and limitations specific to analogue circuits. The paper reviews the current hardware and software methods for conditioning of tip sensors and estimation of time of arrival, including customizable triggering circuits and processing of the sampled sensor output. It presents some measurement solutions developed at ITWL and example results. Selected algorithms of phase estimation, their possible implementation and performance issues are discussed. An exemplary software phase detector, implemented as a state machine, able to adapt to the mean pulse amplitude and ignore noise peaks is demonstrated. Numerical techniques, such as linear and polynomial interpolation and curve fitting, are proposed to process sampled sensor signal in order to increase resolution in time and measure characteristics of the blade-related pulse, including zero-crossing, maximum amplitude, rise time and pulse width. The selection of optimum clock or sampling rate in relation to the blade passing frequency and pulse rise time is discussed.