Technical Field

This invention relates to measurement and evaluation of vibration in solids.

Background Art

Manufacturers of turbomachinery are increasingly facing the risk of a contact between some rotor parts, including blades, and the stator, an occurrence which may lead to accidents. The roots of this situation lie in the manufacturers' effort to maximize the efficiency of machines by minimizing the clearance between the rotor and the stator.

Another source of problems might be the imperfect flow of the working medium around blade foils which results from changes in operating conditions and leads to increased excitation in rotor blades. As a consequence, the blades are under higher dynamic loads which may cause fatigue fractures in critical locations. For these reasons, severely-stressed locations in machines are often provided with measuring instruments that rely on various physical principles.

One of the earliest concepts involves strain gauge measurement and continuous telemetric transmission of data from the blade to a central measuring console. The weakness of such a system stems from its labour intensity and high cost due to which only a few blades are measured in each stage of the machine and the system thus cannot provide full-scale information about the condition of the entire blading. On the other hand, its advantage is that the information from the measured locations is supplied continuously.

Recently, measuring systems referred to as BTT (Blade Tip Timing) systems have been increasingly used which typically monitor the tip vibrations of all blades on the particular rotor in a discrete manner, The equipment is relatively simple in principle, despite the state- of-the-art devices used. It comprises a high-precision clock counter of very high frequency and two or more stator-mounted probes. The first of these is a reference probe which determines the start of each revolution, while the others monitor the passage of blade tips in front of them. As the machine rotates, the probes trigger pulses which govern the logging of instantaneous states of the counter to a memory device whose content is finally written to data files. Typically, the records from a single measurement run take a binary form where the data from all probes is coded and stored in either a single file or in as many files as the probes. The time histories in these files are ordinarily used for evaluating differences in blade passing times at the probes from which blade tip deflections are calculated, whose maxima are an indication of the vibration severity.

BTT systems are typically employed to monitor radial clearance between the rotor and the stator and to finding natural frequencies of blades in service. Where a single probe is used for blade tip timing, the resulting signals are typically undersampled which is why the discrete Fourier transform of the time series of blade tip position samples does not yield a true Fourier amplitude spectrum but a spectrum with aliasing distortion in the frequency domain. This may result in some complications to the signal analysis, as described below. Signal undersampling with all its consequences is known in this field from professional literature, e.g.: M. Balda: Lfvod do statisticki mechaniky. Zapadoceska univerzita v Plzni, 2001, ISBN 80-7082-820-X. A description of the state of the art closest to the present invention can be found in the GB2466817 patent file. The document describes not only the basic BTT system but also its upgrade for estimating the fatigue damage of blades. In the upgraded version, the ordinary BTT system includes an additional accelerometer which is attached to the machine's stator and transmits a composite signal that comprises rotational-frequency imbalance-induced signals and signals from the vibrating blades. Using appropriate filtering, rotational components are removed from this acceleration signal, leaving a signal which should contain blade vibration signals and, possibly, other still unidentified signals. Integer multiples of the sampling (rotational) frequency are added to or subtracted from the resonance frequencies of the aliased Fourier blade spectra, until a match is found of one of resonance frequencies from the accelerometer.

From this point onwards, both the amplitude of the harmonic signal from the aliased blade spectrum signal and the frequency from the frequency analysis of the acceleration signal are available. Both types of information will be used for calculating damage increment. This approach has the following drawbacks:

1. The system for generating damage estimates requires two simultaneous and independent measuring processes and an analysis of the discrete signal from the BTT system and the continuous accelerometer signal.

2. A signal filtering system is necessary for preparing the accelerometer signal for the subsequent analysis of stator vibrations.

3. The amplitude of the measured blade signal from the aliased Fourier spectrum can only be determined accurately if it is a line spectrum. However, this can only be carried out in exceptional cases where the resonance frequency lies in the middle of the frequency increment of df width in Fourier analysis. If that is not the case, the resonance frequency decomposes across several frequency bands and its value cannot be determined accurately.

4. An inaccurately determined amplitude translates into a large discrepancy in the damage increment estimate and, in turn, in the fatigue life of the blade.

5. In this manner, only one (average) amplitude is determined which remains constant in throughout the measuring.

The above-described drawbacks are eliminated by the present invention.

Disclosure of Invention

The invention relates to a method of monitoring residual life of turbomachinery rotor blades. The method uses records of sampled movement of blade tips obtained from clock counter data in a BTT system. The records are decoded and clock counter overflows, mean value nonstationarities and signal trends are removed from the data. The discrete Fourier transform is then applied to the data and its results for each rotating blade are translated into aliased amplitude spectra. The aliased amplitude spectra are linked to the natural frequencies of blade vibrations which are known from calculations or found by testing on test equipment.

The discrete Fourier transform produces aliased amplitude spectra in frequency bands (-/N, +fii), where f _N is Nyquist frequency, which is half the rotational frequency. Resonance bands of the aliased amplitude spectra are then translated, while respecting the polarity, into reconstructed spectra of individual blades within the (-f _& interval, where f _R is half the sampling frequency of the reconstructed spectrum and its value is an order of magnitude larger than fy.

The subsequent inverse Fourier transform of the reconstructed spectra yields reconstructed profiles of original signals of blade vibrations in the tangential direction. Using the results of the aforementioned calculations of natural modes of rotating blades and the frequency transfer functions derived from them which relate the differences in blade tip passing times at the probes and mechanical loads in critical locations of the blades, relevant stress tensors and, subsequently, effective damaging stresses are determined from the reconstructed signals. In the field of this invention, the matter of damaging stresses has been described in professional literature, for instance: M. Balda, J. Svoboda, V. Frohlich: An estimation of fatigue life under general stress.

Proc. Conf Engineering Mechanics 2007, (ed. I. Zolotarev), IT CAS, Svratka, 2007,

ISBN 978-80-870J 2-06-2.

Damage increments caused by sequences of extrerna in time series of damaging stresses decomposed by the rainflow counting technique are then determined from the effective damaging stress values with the aid of a chosen damage rule. These damage increments are accumulated separately for each blade.

The accumulated damage increments are stored in processing logs for further use. Where required, the accumulated damage increments are used for plotting diagrams which provide information on the damage growth in and the residual life of each rotating blade. The relative residual life of each blade is a complement of its accumulated relative damage to 1 , which corresponds to complete destruction of the blade.

Where multiple probes are used in the BTT system, the records are evaluated independently for each probe, up to the stage of damage increment calculation. From these results, the data from the probe which represents the largest damage increment will be used for evaluating the accumulated damage. This makes the estimate of accumulated damage a conservative result. The damage rule for calculating the damage increments is either Palmgren-Miner rule or another rule derived from it.

To account for the period of machine's operation for which no BTT measurement records are available, proportionate fractions of past, recent or average damage increments are added to the accumulated damage levels.

In an alternative and particularly preferred embodiment, the data can be displayed and analyzed during processing. This means that the operator may suspend the automatic processing sequence of measured BTT data, e.g. in order to transform the data in detail to a graphic form, and then return to the processing sequence without losing its results.

In a preferred embodiment, the turbomachine operator receives a warning at the time when a pre-defined blade damage extent is reached.

After damaged blades are replaced, the accumulation of the relative damage continues in the remaining blades which were not replaced. The initial damage level for the new replaced blades is set as zero.

To determine blade damage levels with greater accuracy, it is advantageous - after the resonance bands were transferred to the reconstructed Fourier spectrum - to distribute the remainder of the aliased Fourier spectrum across the reconstructed Fourier spectrum. The distribution can be either uniform or non-uniform.

It is advantageous if the value of half the sampling frequency of the reconstructed spectrum is 25 times higher than fy. The width of the frequency band of the reconstructed signal can be chosen by the user. The broader this band is, the better is the reconstruction of the extrema in the reconstructed signal, although it comes at the cost of higher computational demands. The extrema in the reconstructed signal govern the intensity of the fatigue damage in the rotating blades.

Brief Description of Drawings

An exemplary embodiment of the proposed invention is described with reference to the following figures

Fig. 1 - shows moduli of the aliased Fourier vibration spectrum of a single blade sparsely sampled at the operating speed, as obtained from the discrete Fourier transformation of the data file generated by the BTT system. The numbers at the top of the plot show the positions of aliased natural frequencies calculated using a formula from professional literature: M. Balda: Uvod do statisticke mechaniky. Zdpaduceskd univerzita v Plzni, 2001, ISBN 80-7082- 820-X;

Fig. 2 - shows moduli of reconstructed vibration spectrum of the same blade, in which natural frequency bands from the aliased spectrum shown in Fig. 1 were transferred to the correct frequencies within the reconstructed spectrum; Fig. 3 - shows a selected interval of the reconstructed blade tip movement signal derived from the reconstructed spectrum in Fig. 2 by means of inverse Fourier transform, and the points which represent the original sparse sampling of the original signal in the BTT file; Fig. 4 - illustrates how relative damage increased with time in individual blades over the course of a long single BTT measuring run. The x axis indicates blade numbers, the y axis shows numerical codes of measurement intervals for a single probe, which also represent the time scale, and the z axis represents relative damage increments. The x - z plane contains relative damage increments for individual blades accumulated from all measurement intervals from the relevant probe. The figure clearly shows that substantial damage only occurs in some blades. These are the blades whose vibration is extensive, causing higher dynamic loads which, in turn, lead to fatigue crack initiation in critical locations.

Fig. 5 - is a graphical representation of accumulated relative damage in blades of the entire rotor. This figure shows, as does Fig. 4, that relative damage does not increase equally in all blades. The relative residual life of a blade is a complement of the radius of the damage (dark) sector for the given blade to the radius of 1 of the circle. For a blade with very little damage, the circle with the radius of 1 which represents total relative damage can be outside the area captured in the graph.

Best Mode for Carrying Out the Invention

The above-described method of monitoring residual life of blades of turbomachinery (RFLB - Residual Fatigue Life of Blades) is linked to general BTT systems either directly or via their output files. This is provided by an input procedure which depends on the particular BTT system used. Two variants have been implemented: for BTT systems from Starman and Hood companies. The system is operated in an interactive manner.

Upon starting, the processing takes place in the following steps:

A. The BTT system type is entered and, the input procedure is selected accordingly.

B. Blade parameters are defined, including the blade material, methods to be used for file processing, determination of frequency transfer functions relating the time differences and the loads in critical locations of the blades, and output types. The subsequent processing of measurement data files takes place automatically without operator's intervention. The measurement processing cycle is as follows: The output file(s) from the BTT system is/are read and, if required, decoded; upon long measurement runs, the files are split into shorter intervals, if needed.

Trends and mean value nonstationarities which affect mean values are eliminated from the data described under point 1.

The results of the step described under point 2 are transformed using the discrete Fourier transform and an aliased Fourier spectrum is obtained for each probe and for each blade or blade measurement interval.

Using a formula, the aliased resonance frequencies are converted into real frequencies, as known in this field (see M, Balda: Uvod do statisticke mechaniky. ZdpadocesM univerzita v Plzni, 2001, ISBN 80-7082-820-X). A concrete embodiment of the invention offers the processing operator a value of fa which is equal to 25 times the /N. The operator may accept it or enter their own value. Currently, the said value appears to be optimal, as it leads to a sufficient amount of data without being extremely computationally demanding.

The resonance bands are transferred from the aliased Fourier spectra to the correct frequencies in the reconstructed spectra determined according to point 4, while respecting the polarity. The remainder of the aliased Fourier spectrum can be distributed across the reconstructed spectrum as desired by the user. The currently implemented variant distributes frequency components uniformly across all bands of the reconstructed spectrum up to and including the highest resonance frequency. The spectra reconstructed according to point 5 are converted using the inverse Fourier transformation to obtain reconstructed profiles which are close to the real movement of blade tips in units of time. If sparsely sampled signal data and densely sampled reconstructed signal data is plotted in diagrams, it becomes apparent that the original data is very close to the reconstructed signals and one can therefore expect a very good agreement between the evaluated and real damage levels.

From the signals reconstructed according to point 6, extremum values are extracted which, after frequency transfer functions under point B are applied, yield the stress tensor for the critical location of the blade which characterizes the combined stress components.

From the stress tensor according to point 7, effective damaging stress is derived, whose sequence is decomposed using the rainflow counting technique. The concept of effective damaging stresses is known to persons skilled in the art from, for instance, M. Balda, J. Svoboda, V. Frohlich: An estimation of fatigue life under general stress. Proc. Con Engineering Mechanics 2007, (ed. I. Zolotarev), IT CAS, Svratka, 2007, ISBN 978-80-87012-06-2.

9. Full cycles and half-cycles of the effective damaging stress obtained under point 8 form the basis for calculating damage increments using one of the rules derived from the Palmgren-Miner rule or using the Palmgren-Miner rule itself.

10. The relative damage increments are accumulated separately for each blade, including the periods during which no measuring took place. The accumulated relative damage levels are written into a file and, simultaneously, backups are created.

11. Once the processing of the last measurement file from the storage is completed, the PvFLB system waits for the BTT system to supply further measurement data.

12. The operator may interrupt the automatic processing cycle in order to deal with immediate tasks (e.g. in order to generate images from any portion of the processing sequence completed up to that point, to assess operating parameters with respect to the growth of relative damage, etc.) and, having completed these additional activities, return to the automatic operation cycle without losing any information.

13. End of measurement processing cycle

C. The present method allows sampled variable-amplitude signals to be processed as well, which represents a considerable strength because measuring runs involving constant amplitudes are very rare.

D. One of the extraordinary aspects of the method according to the present invention is that its processing time is shorter than the actual BTT measuring time, thanks to which earlier data can be processed without the new data building up.

E. If some of the blades have been replaced, the system enables the accumulation of relative damage to be continued for the remaining blades and to be started anew for the new blades.

Given the configuration variations of BTT measuring equipment from various manufacturers, the subject of the present invention will take various forms, differing in certain special aspects, most notably in the measurement data input. For this reason, the subject of the invention may take the following forms:

• RFLB for the BTT system from the Starman company,

• RFLB for the BTT system from the Hood company,

• Hybrid RFLB for the BTT systems from the Starman and Hood companies,

• RFLB for other BTT systems.

Other alternatives of the method may use different damage rules. Upon considering the situation in the Czech Republic, a hybrid RFLB system for BTT systems from Starman and Hood companies has been developed and combined with the use of Palmgren-Miner damage rule, which is frequently referred to as a superior one when used in conjunction with signal decomposition by means of the rainflow counting technique.

The exemplary embodiment is shown in Figs. 1 through 5.