RU2511773C1 - Method of diagnostics of oscillations of turbomachine impeller - Google Patents

Abstract

FIELD: machine building.

SUBSTANCE: in the method of diagnostics of oscillations of a turbomachine impeller, signals are measured simultaneously, at least, from two sensors, the second of which is a vibrator inverter, installed on stator parts of the turbomachine near the impeller, as a non-dimensional parameter, characterising loss of stability, they use an excess coefficient, previously they set threshold levels for signals from a sensor of pulsations of flow pressure and a vibrator inverter, and they determine their appropriate threshold values of the excess coefficients, signals are measured in a narrow frequency band, for each signal they determine values of excess coefficients and moments of time, when they reach their threshold values. At the same time, if the excess coefficient for a signal from a sensor of flow pressure pulsations reaches its threshold value earlier than the excess coefficient for a signal from the vibrator inverter, it means there are stalling oscillations in the impeller, if excess coefficients for signals from a sensor of pressure pulsations of the flow and the vibrator inverter simultaneously reach their threshold values, this means there are self-excited oscillations in the impeller.

EFFECT: higher efficiency and reliability of diagnostics of oscillations of a turbomachine impeller in real time mode.

2 cl, 3 dwg

Description

The present invention relates to aircraft engine building and power engineering and may find application in the development of gas turbine engines (GTE), as well as to create systems for diagnosing vibrations.

A known method for diagnosing oscillations of the impeller of a turbomachine (RF patent for the invention No. 2111469, IPC G01M 15/00, publ. 05/20/1998), in which the signals from the pressure pulsation sensor installed on the stator parts of the turbomachine near the impeller are measured during the tests, and judge the form of oscillation of the impeller. The method allows you to diagnose self-oscillations and a rotating stall by the presence and ratio of the levels of two diagnostic components in the spectrum of pulsations of the pressure of the stream.

When diagnosing this method, as shown by an experimental check, a false diagnosis is possible (a second kind error), because the appearance of diagnostic components that are different in level may not be associated with the vibrations of the blades or the formation of shear flow zones. The same manifestation and combination of diagnostic signs was observed in the presence of interference and noise from operating bench equipment. Due to the Doppler effect, the oscillations perceived by the sensor mounted on the stator parts of the turbomachine become frequency-modulated and have a discrete spectrum. A similar modulating effect can be exerted by components from noises of various origins, moreover, in the spectrum of the noise signal, the left and right side components characterizing its modulation will also be different in level. All this makes the method unreliable and low efficient. In addition, diagnostics are performed after testing according to already registered information.

A known method for diagnosing oscillations of the impeller of a turbomachine (RF patent for the invention No. 2395068, IPC G01M 15/14, publ. 07/20/2010), in which during the test the signals from a vibration transducer mounted on the stator parts of the turbomachine near the impeller are measured, they judge the form impeller vibrations.

The method is based on determining the propagation directions of the deformation waves. It gives more reliable results compared to the previous method, however, its use is limited only by the impellers of small-sized gas turbine engines, the values of the frequencies of the natural vibrations of the working blades of which exceed the product of the rotational speed by the number of nodal diameters.

Closest to the proposed one is a method for diagnosing oscillations of the impeller of a turbomachine (RF patent for the invention No. 2402751, IPC G01M 15/14, publ. 10/27/2010), in which the signals from sensors are measured during the tests, one of which is a flow pressure pulsation sensor mounted on the stator parts of the turbomachine near the impeller, determine the corresponding values of the dimensionless parameter characterizing the loss of stability, and the times at which they change in a predetermined manner, by which they judge the form of Impeller oscillations, wherein, if the change in dimensionless parameter values occurred earlier for a signal from the sensor flow pressure pulsations, it indicates the presence of shear vibrations in the impeller.

In this method, signals from load cells mounted on rotor blades and from a pressure pulsation sensor mounted on a turbomachine body are measured and recorded. Diagnostic blades oscillation frequencies are determined in the flow pulsation spectrum. As a dimensionless parameter characterizing the loss of stability, the excitation coefficient is used in an unsteady signal at the vibration frequencies of the working blades and at the diagnostic vibration frequencies in the spectrum of flow pulsations. The dependences of the values of the excitation coefficients on time for the signals recorded from the sensors are constructed. The time moments are determined at which the values of the excitation coefficients from negative become positive, and they are used to judge the type of vibrations of the working blades (self-oscillations or rotating stall).

The method allows diagnostics according to already registered information, i.e. not in real time, but after stopping the turbomachine.

However, when diagnosing fluctuations by this method, a false diagnosis is possible. This is due to the fact that when determining the excitation coefficient for a real signal with an amplitude that grows in time at certain points in time, the value of the signal amplitude can decrease while maintaining the general growth trend. When the signal amplitude decreases, the excitation coefficient in accordance with this method will change its sign to the opposite, which can lead to a false diagnosis.

When determining the excitation coefficient by this method, the amplitudes at adjacent signal half-waves are determined, therefore, with a relatively high frequency of the signal and its growing character, these amplitudes differ slightly from each other, as a result, their ratio is close to unity, and the natural logarithm of unity is zero. To offset the signal values relative to the zero level, a scale factor was introduced into the formula for determining the excitation coefficient, which is taken as the signal frequency. However, simultaneously with the amplification of the signal and its removal from the zero level, a proportional amplification of all errors occurs. As a result, there is a constant change of sign, which can lead to a false diagnosis. In radio electronics, the term “contact bounce” is used for a similar phenomenon. Therefore, this method is applicable only for idealized signals (smooth, artificially modeled curves).

The presence of self-oscillations is judged by the simultaneous passage through the zero level of the excitation coefficients of two signals. At the same time, the previous and subsequent amplitudes for these signals are determined at different points in time. Those. the method contains a methodological error that reduces the reliability of diagnostics caused by the need to perform analysis at two different frequencies, and therefore, at different points in time, therefore there is no reason to conclude that there is "simultaneity".

Due to the fact that, when approaching the boundary of self-oscillations, the process is obviously unsteady, in this method the excitation coefficients for such processes are determined by the Prony method (O. B. Balakshin, B. G. Kukharenko, A. A. Khorikov. Investigation of dynamic processes at blade flutter using the prony method). The complex technical implementation of the Proni method and the lack of proven algorithms make it difficult to use the diagnostic method in practice.

For diagnostics by this method, strain gauging is required, while the strain gauge mounted on the working blade of the turbomachine wheel, according to which the frequency and coefficient of excitation of the blade vibrations are determined during the development of vibrations, has low reliability. Dynamic strain gauging is associated with significant costs (strain gauge sticker, preparation withdrawal, etc.). In addition, where, according to the design features of the turbomachine, strain gauging is not possible, for example, for the high pressure cascade of a twin-shaft gas turbine engine, this method is not applicable.

An object of the invention is to develop a method that allows diagnosing oscillations only by sensors installed on the stator (non-rotating) parts of a turbomachine, such as a housing, and using the kurtosis coefficient as a dimensionless parameter, which is most effective specifically for non-stationary processes, while less laborious in computational terms.

The technical result of the invention is to increase the efficiency and reliability of diagnosis of oscillations of the impeller of a turbomachine in real time.

The technical result is achieved by the fact that in the method for diagnosing oscillations of the impeller of a turbomachine, in which the signals from the sensors are measured during the test, one of which is a flow pressure pulsation sensor mounted on the stator parts of the turbomachine near the impeller, the corresponding values of the dimensionless parameter characterizing them are determined the loss of stability, and the moments of time at which they change in a predetermined manner, by which they judge the form of oscillations of the impeller, while if When the dimensionless parameter was measured earlier for the signal from the flow pressure pulsation sensor, this indicates the presence of stall oscillations in the impeller, in contrast to the known one, the signals are measured simultaneously from at least two sensors, the second of which is a vibration transducer mounted on stator the details of the turbomachine near the impeller, the kurtosis coefficient is used as a dimensionless parameter characterizing the loss of stability, the threshold levels for signals with of the pressure pulsation pulsations and the vibration transducer and determine the corresponding threshold values of the kurtosis coefficients, measure the signals in a narrow frequency band, for each of the signals determine the values of the kurtosis coefficients and the times at which they reach their threshold values, and if the kurtosis coefficient for the signal from the flow pressure pulsation sensor reaches its threshold value earlier than the kurtosis coefficient for the signal from the vibration transducer, this indicates the presence of stall vibrations in the impeller, if the kurtosis coefficients for the signals from the flow pressure pulsation sensor and vibration transducer simultaneously reach their threshold values, this indicates the presence of self-oscillations in the impeller.

Signals are measured in the frequency band bounded by a low-pass filter.

The attached figures show:

figure 1 - two zones of dynamic amplification of signals from the vibration transducer (upper graph) and the sensor of the pulsation of the flow pressure (lower graph);

figure 2 - envelopes of the amplitudes of the signals from the vibration transducer (upper graph) and the sensor of the pulsations of the pressure of the flow (lower graph) in the areas of dynamic amplification of signals;

figure 3 - values of the coefficients of kurtosis for signals from the vibration transducer (upper graph) and the sensor of the pulsations of the pressure of the flow (lower graph) in the areas of dynamic amplification of signals.

The method is as follows.

Preliminarily, according to the results of experimental studies of rapidly changing processes, threshold levels are set for signals from a flow pressure pulsation sensor and a vibration transducer mounted on the stator parts of a turbomachine near the impeller, the achievement of which indicates the occurrence of oscillations. The corresponding threshold values of the dimensionless parameter characterizing the loss of stability are determined as the excess coefficient.

In the process of testing simultaneously (synchronously) measure signals from at least two sensors mounted on the stator parts of the turbomachine near the impeller, one of which is a flow pressure pulsation sensor, and the second is a vibration transducer. The measurement of signals, in order to increase the efficiency of use of the kurtosis coefficient, is carried out in a narrow frequency band, for example, limited by a low-pass filter.

For each of the measured signals from the vibration transducer and the flow pressure pulsation sensor, the kurtosis coefficients are determined and compared with the corresponding threshold values.

The time moments are determined at which the kurtosis coefficients for the signals from the flow pressure pulsation sensor and the vibration transducer reach their threshold values, which are used to judge the form of the non-stationary oscillations of the impeller.

If the kurtosis coefficient for the signal from the flow pressure pulsation sensor reaches its threshold value earlier than the kurtosis coefficient for the signal from the vibration transducer, this indicates the presence of stall vibrations in the impeller.

If the kurtosis coefficients for the signals from the flow pressure pulsation sensor and the vibration transducer reach their threshold values at the same time, this indicates the presence of self-oscillations in the impeller.

Additionally, during the test process, the measured signals can be recorded and converted into frequency spectra, which will allow for a detailed analysis of the results and confirm the diagnosis.

The method was implemented when testing a gas turbine engine.

In the process of conducting experimental studies of the compressor fan stage (investigation of the resistance to self-oscillations of fan blades in a wide range of operating modes, determining the boundary of gas-dynamic stability), the possibilities of diagnosing oscillations using the proposed method were evaluated using information from various sensors installed on the stator parts near the impeller.

The threshold levels were set for the signals from the flow pressure pulsation sensor - 0.025 kgf / cm ² and vibration transducer - 4 mm / s, which corresponded to the values of the threshold amplitudes of vibrational stresses in the blades. Exceeding these levels indicated the onset of dangerous vibrations.

We determined the corresponding threshold values of the excess coefficient, given that for the normal distribution the excess coefficient is three (Tyurin Yu.N., Makarov A.A. Data analysis on a computer. Edited by V.E. Figurnov. - M.: INFRA- M. 2002.S. 34).

From the course of mathematical statistics and probability theory it is known that the kurtosis coefficient is determined by the formula:

, $$E=\frac{\mathrm{one}}{N{\sigma }^{\mathrm{four}}}{\displaystyle \sum _{i=0}^{N-\mathrm{one}}{(V_i-MO)}^{\mathrm{four}}}$$

,

where σ is the standard deviation,

, $$\sigma =\sqrt{\frac{\mathrm{one}}{N-\mathrm{one}}{\displaystyle \sum _{i=0}^{N-\mathrm{one}}{(V_i-MO)}^2}}$$

,

MO - expectation (average value),

, $$MO=\frac{\mathrm{one}}{N}{\displaystyle \sum _{i=0}^{N-\mathrm{one}}{V}_i}$$

,

V _{i is the} amplitude of the signal from the sensor;

N is the number of values.

The threshold values of the kurtosis coefficients for the signal from the flow pressure pulsation sensor and the vibration transducer for both signals, determined by the formula, amounted to the same value E = 3.5. It should be noted that in general they can have different meanings.

In a number of sources, the formula for determining the excess coefficient is given taking into account its equality to zero under the normal distribution law. This is taken into account by a decrease in the determined value by 3. However, since the deviation from the normal probability distribution law is not a criterion for the onset of stability loss, there is no need to bring the formula for determining the excess coefficient to a form in which its value for the normal distribution law becomes equal to zero. Using this formula, the threshold level value instead of the value E = 3.5 would be taken equal to E = 0.5.

Thus, the formula for determining the kurtosis coefficient is chosen on the basis that a small close to zero value of the dimensionless parameter does not contribute to increasing the reliability of diagnostics (from the point of view of its implementation).

Before testing, the fan casing was prepared with a Kulite XTE-190 high-frequency flow pressure pulsation sensor and VTK-7 high-frequency three-component vibration transducer. To measure and analyze dynamic parameters, the MIC-300M digital dynamic signal recorders were used.

During testing, simultaneously (synchronously) measured signals from the pressure pulsation sensor of the flow and vibration transducer and converted them into narrow-band using a low-pass filter with a cutoff frequency of 300 Hz.

For clarity of the method, synchronous registration of signals from the vibration transducer and the flow pressure pulsation sensor was performed. Figure 1 shows two zones of dynamic amplification for the signal from the vibration transducer V, B and the pressure pulsation sensor of the flow P, B depending on time.

An analysis of the increase in amplitudes was performed for the first (10–20 s) and second (72.5–82.5 s) zones of dynamic amplification of signals from the vibration transducer (upper graph) and the flow pressure pulsation sensor (lower graph) of FIG. 2, where V , V 'are the signal values from the vibration transducer in the zones of dynamic vibration amplification, P, P' are the signal values from the flow pressure pulsation sensor in the zones of dynamic pulsation amplification.

For these zones, with an interval of 2.5 s, the kurtosis coefficients were determined for the signals measured by the sensors. Figure 3 shows Eν, Eν '- the kurtosis coefficient for the signal from the vibration transducer in the zones of dynamic amplification, Ep, Ep' - the kurtosis coefficient for the signal from the sensor of pressure pulsations of the flow in the dynamic amplification zones.

The values of the excess coefficients for signals from the pressure pulsation sensor and the vibration transducer were compared with their threshold values E = 3.5.

The time moments were determined at which the kurtosis coefficients took the value E = 3.5. The moment of fixation corresponded to the threshold value of the excess coefficient E = 3.5, which was uniform for vibrations and for pulsations of the flow pressure: for the first dynamic amplification zone 16.5 s, for the second 79 s. For both zones of dynamic amplification of signals at the indicated time points, the vibration level was 0.15 V, which corresponds to 4 mm / s (the maximum value of the amplitude in the zone of dynamic vibration amplification was 10 mm / s), and the ripple level was 0.05 V, which corresponds to 0.025 kgf / cm ² (the maximum value of the amplitude in the zone of dynamic amplification of the flow pressure pulsation was 0.05 kgf / cm ² ).

The kurtosis coefficients for the signals from the vibration transducer and the pulsation pressure transmitter simultaneously reached the threshold value. Consequently, the conclusion was drawn about the occurrence of self-oscillations.

The invention allowed to create a method for diagnosing oscillations of the impeller of a turbomachine in real time, with higher reliability and efficiency, not allowing, thanks to timely diagnostics, destruction of parts and components of the turbomachine and allowing to determine the type of dangerous vibrations and take measures to eliminate them, minimizing the risk destruction of a turbomachine.

Claims (2)

1. A method for diagnosing oscillations of the impeller of a turbomachine, in which signals from sensors are measured during the test, one of which is a flow pressure pulsation sensor mounted on the stator parts of the turbomachine near the impeller, and the corresponding values of the dimensionless parameter characterizing the loss of stability are determined and moments the time at which they change in a predetermined manner, by which they judge the type of impeller vibrations, while if a change in the dimensionless parameter occurs earlier for the signal from the flow pressure pulsation sensor, this indicates the presence of stall vibrations in the impeller, characterized in that the signals are measured simultaneously from at least two sensors, the second of which is a vibration transducer mounted on the stator parts of the turbomachine near the impeller , as the dimensionless parameter characterizing the loss of stability, the kurtosis coefficient is used, threshold levels for signals from the flow pressure pulsation sensor and vibration transducer are pre-set index and determine the corresponding threshold values of the kurtosis coefficients, measure signals in a narrow frequency band, for each of the signals determine the values of the kurtosis coefficients and the times at which they reach their threshold values, if the kurtosis coefficient for the signal from the pressure pulsation sensor flow reaches its threshold value earlier than the excess coefficient for the signal from the vibration transducer, this indicates the presence of stall oscillations in the impeller, if oeffitsienty kurtosis for the signals from the sensor and the flow pressure pulsations vibrator simultaneously reach their threshold values, then this indicates the presence of oscillations in the impeller. 2. The method according to claim 1, characterized in that the measurement of the signals is carried out in a frequency band limited by a low-pass filter.

Images