Moving blade asynchronous vibration parameter identification method based on order search

Abstract

The invention discloses a moving blade asynchronous vibration parameter identification method based on order search, and the method comprises the steps: carrying out the fast Fourier transform of a moving blade asynchronous vibration under-sampling signal; according to the method, an existing multi-signal classification algorithm is optimized, frequency search is converted into order search, and the integer order of the asynchronous vibration frequency of the moving blade is estimated; and fusing an FFT algorithm analysis result and an order search result to obtain asynchronous vibration parameters of the moving blade. According to the method, frequency search is converted into order search, and the calculation efficiency of frequency identification is greatly improved. The FFT algorithm can accurately obtain the non-integer order of asynchronous vibration of the moving blade, and the frequency identification precision of the method can be improved. According to the method, the asynchronous vibration parameter identification of the moving blade can be carried out according to the blade end timing signal under the background of no blade priori knowledge, and effective estimation of the blade vibration frequency and high-precision reconstruction of the vibration amplitude are realized.

Description

Technical Field

[0001] The invention relates to a non-contact measurement method for a rotating mechanical moving blade, and in particular to an asynchronous vibration parameter identification method for a moving blade based on order search. Background Art

[0002] Moving blades are the core components of rotating machinery to complete energy conversion and are widely used in major equipment fields such as aircraft engines, wind tunnel compressors, and air compressors. The safe and stable operation of moving blades directly affects the operating status and work efficiency of the unit. Therefore, carrying out moving blade status monitoring and operating parameter identification is of great value to the design and innovation of major equipment.

[0003] At present, Blade Tip Timing (BTT) technology has become a trend in the study of vibration problems of moving blades, and provides technical support for the status monitoring and health management of moving blades of major equipment. BTT technology installs a timing sensor on the equipment casing. When the blade passes the timing sensor, BTT technology can obtain the actual arrival time of the blade. Subsequently, the theoretical time for the blade to reach the timing sensor can be calculated through the blade synchronization signal, and then the vibration information of the blade can be obtained by calculating the time difference of the blade reaching the timing sensor. BTT technology can continuously sense the vibration information of all blades and is widely used in the field of vibration status monitoring of rotating blades. However, in practical applications, the vibration signal obtained by BTT technology has the characteristics of non-uniformity and under-sampling, and its sampling frequency cannot meet the Chanel theorem for blade vibration frequency identification. Therefore, how to carry out blade vibration parameter identification based on the under-sampling signal obtained by BTT technology has become a practical problem in studying blade vibration problems.

[0004] According to the numerical relationship between vibration frequency and rotation frequency, the vibration frequency of moving blades can be divided into synchronous vibration and asynchronous vibration. Synchronous vibration means that the vibration frequency of the blade is an integer multiple of the rotation frequency, and asynchronous vibration means that the vibration frequency of the blade is a non-integer multiple of the rotation frequency. Domestic and foreign scholars have conducted a lot of research on the problem of synchronous vibration parameter identification of moving blades. The existing algorithms mainly include single parameter method, dual parameter method, autoregressive method, and sine fitting method. The above algorithms show certain effects in solving the problem of synchronous vibration parameter identification of moving blades.

[0005] The existing algorithms for identifying asynchronous vibration parameters of moving blades mainly include improved Fourier transform algorithms, compressed sensing algorithms, and spatial defuzzification algorithms. The improved Fourier transform algorithm estimates the blade vibration frequency through multi-sampling average theory, overcoming the limitation of BTT signal non-uniformity, but the robustness of this method is poor. In practical engineering applications, the blade vibration parameter identification accuracy is low. The compressed sensing algorithm sparsely represents the blade vibration under-sampled signal and reconstructs the original signal through optimization problems. The compressed sensing algorithm further improves the blade vibration parameter identification accuracy. However, this type of algorithm has specific requirements for the timing sensor layout. When the sensor layout does not meet the algorithm requirements, its estimation result deviates from the true value. The spatial defuzzification algorithm realizes blade vibration parameter identification by decomposing the correlation matrix of the blade vibration under-sampled signal and using the relationship between the complex sinusoidal signal, the correlation matrix eigenvalue, and the eigenvector. The spatial defuzzification algorithm mainly includes the Estimating Signal Parameter via Rotational Invariance Techniques (ESPRIT) algorithm and the Multiple Signal Classification (MUSIC) algorithm. The ESPRIT algorithm uses two timing sensors to form a delay channel, and uses a spatial algorithm to deambiguate and estimate the blade vibration frequency based on the phase difference information of the vibration frequency carried by the signal. The ESPRIT algorithm has poor robustness and is suitable for blade vibration signals with low signal-to-noise ratios. Based on the orthogonality of complex sinusoidal signals and noise subspaces, the MUSIC algorithm searches for spectral peaks in the blade vibration frequency and estimates the blade vibration parameters. The MUSIC algorithm estimates blade vibration parameters through frequency search, and its search period is related to the frequency resolution and frequency band range. When the resolution of the vibration frequency estimation of the moving blade is high and the frequency band range is large, the search period of the MUSCI algorithm will increase.

[0006] The existing asynchronous vibration parameter identification algorithm of moving blades has strict requirements on the blade vibration signal and timing probe layout. When the vibration test parameters cannot meet the requirements of the algorithm, it will lead to the problem of "inaccurate measurement" of blade vibration. The asynchronous vibration parameter identification of moving blades is the basis of key technologies such as state monitoring, fault diagnosis, and life prediction of rotating machinery moving blades.

[0007] Therefore, studying a high-precision and high-efficiency method for identifying the asynchronous vibration parameters of moving blades is of great significance for the safe operation of moving blades of major equipment. Summary of the invention

[0008] Purpose of the invention: In view of the shortcomings of the prior art, the present invention provides a method for identifying the asynchronous vibration parameters of moving blades based on order search with high accuracy and high computational efficiency. Based on the FFT results, the present invention identifies the asynchronous vibration parameters of moving blades by order search, and on the basis of ensuring the test accuracy, further improves the operating efficiency of the blade vibration parameter identification algorithm.

[0009] Technical solution: The method for identifying the asynchronous vibration parameters of moving blades based on order search described in the present invention is based on the principle that the existing MUSIC algorithm estimates the vibration frequency of the blades by frequency search, and optimizes the existing MUSIC algorithm. First, a fast Fourier transform (FFT) is performed on the under-sampled signal of the asynchronous vibration of the moving blades to obtain the amplitude corresponding to the asynchronous vibration frequency of the blades and the non-integer order of the asynchronous vibration frequency. Subsequently, the existing MUSIC algorithm is optimized, the frequency search is converted into an order search, and the integer order of the asynchronous vibration frequency of the moving blades is searched to estimate the integer order of the asynchronous vibration frequency of the moving blades. Finally, the FFT analysis results and the order search results are integrated to obtain the asynchronous vibration parameters of the moving blades, including the vibration amplitude and the vibration frequency.

[0010] Further, the following steps are included:

[0011] Step 1: Perform FFT processing on the under-sampled signal of asynchronous vibration of moving blades obtained by BTT technology to obtain the amplitude and non-integer order of the asynchronous vibration frequency of the blades. Here, the asynchronous vibration frequency of the blades can be expressed as:

[0012] f=(N+ΔN)f r (1)

[0013] Where f represents the asynchronous vibration frequency of the blade; N represents the integer order of the asynchronous vibration frequency of the blade; ΔN represents the non-integer order of the asynchronous vibration frequency of the blade; f r Indicates the rotation frequency of the moving blade rotor.

[0014] Step 2: Establish the blade vibration autocorrelation matrix based on the under-sampling signal of asynchronous vibration of the moving blade obtained by BTT technology. M blade vibration signals are continuously collected by timing sensors, and the signal vector of the qth sampling is defined as:

[0015]

[0016] Among them, y(t q ) indicates t q The blade vibration displacement collected at the moment. The autocorrelation matrix calculation formula of the blade vibration signal is:

[0017] R = E{y(q)yH (q)} (3)

[0018] Among them, y H (q) represents the transpose of the signal vector y(q).

[0019] Step 3: Perform eigenvalue decomposition on the autocorrelation matrix obtained in the second step. Theoretically, the autocorrelation matrix has K (K represents the number of blade vibration frequency components) non-zero eigenvalues. In actual engineering applications, due to the interference of test noise, the values ​​of other eigenvalues ​​are close to 0. K+1 , K+2 , , M The corresponding eigenvector is defined as the noise subspace.

[0020] Step 4: Establish an estimation vector of the integer order of blade vibration frequency:

[0021]

[0022] Wherein, ΔN represents the non-integer order of the blade vibration frequency obtained in the first step; β represents the installation angle of the timing sensor.

[0023] Step 5: Set the search range of integer orders of blade asynchronous vibration frequency to identify the integer orders of asynchronous blade vibration frequency. The search formula for integer orders of blade asynchronous vibration frequency is:

[0024]

[0025] Among them, E n Represents the noise subspace obtained in the third step. Function P m-R (N) and P m-L (N) The N corresponding to the maximum value is the integer order of the asynchronous vibration frequency of the blade.

[0026] Step 6: According to the non-integer order of the blade asynchronous vibration frequency obtained in the first step and the integer order of the blade asynchronous vibration frequency obtained in the fifth step, the real frequency of the blade asynchronous vibration can be obtained, and the calculation formula is:

[0027]

[0028] Among them, max(g) means taking the maximum value.

[0029] Further, the rotation frequency of the moving blade rotor is the average rotation frequency of the blade passing through different sensors;

[0030] Further, the installation angle of the timing sensor is a relative value;

[0031] Furthermore, the timing sensor continuously collects M blade vibration signals. In order to ensure that all blade asynchronous vibration frequencies can be identified, the present invention requires that M>K, and M is an integer multiple of the number of timing sensors.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] (1) The present invention uses the FFT algorithm to identify the non-integer orders of the amplitude and frequency corresponding to the asynchronous vibration frequency of the moving blade, and then identifies the integer order of the asynchronous vibration frequency of the blade through the order search method, thereby realizing the accurate identification of the asynchronous vibration parameters of the moving blade.

[0034] (2) The present invention converts frequency search into order search, which greatly improves the computational efficiency of frequency identification. In addition, the FFT algorithm can accurately obtain the non-integer order of the asynchronous vibration of the moving blades. On this basis, an integer order search of the asynchronous vibration frequency of the moving blades is carried out, which can improve the frequency identification accuracy of the present invention.

[0035] (3) The present invention can identify the asynchronous vibration parameters of moving blades based on BTT signals without prior knowledge of the blades, thereby achieving effective estimation of the blade vibration frequency and high-precision reconstruction of the vibration amplitude. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is the vibration displacement of blade 7 under constant speed condition;

[0037] Figure 2 This is the spectrum analysis result of the vibration displacement of blade No. 7;

[0038] Figure 3 This is the search result of integer order of asynchronous vibration frequency of blade 7;

[0039] Figure 4 The time domain signal and spectrum analysis results of blade 7 strain. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be further described below.

[0041] The method for identifying parameters of asynchronous vibration of moving blades based on order search in this embodiment includes the following steps:

[0042] Step 1: Perform FFT processing on the under-sampled signal of asynchronous vibration of moving blades obtained by BTT technology to obtain the amplitude and non-integer order of the asynchronous vibration frequency of the blades. Here, the asynchronous vibration frequency of the blades can be expressed as:

[0043] f=(N+ΔN)f r (1)

[0044] Where f represents the asynchronous vibration frequency of the blade; N represents the integer order of the asynchronous vibration frequency of the blade; ΔN represents the non-integer order of the asynchronous vibration frequency of the blade; f r Indicates the rotation frequency of the moving blade rotor.

[0045] Step 2: Establish the blade vibration autocorrelation matrix based on the under-sampling signal of asynchronous vibration of the moving blade obtained by BTT technology. M blade vibration signals are continuously collected by timing sensors, and the signal vector of the qth sampling is defined as:

[0046]

[0047] Among them, y(r q ) indicates t q The blade vibration displacement collected at the moment. The autocorrelation matrix calculation formula of the blade vibration signal is:

[0048] R = E{y(q)u H (q)} (3)

[0049] Among them, y H (q) represents the transpose of the signal vector y(q).

[0050] Step 3: Perform eigenvalue decomposition on the autocorrelation matrix obtained in the second step. Theoretically, the autocorrelation matrix has K (K represents the number of blade vibration frequency components) non-zero eigenvalues. In actual engineering applications, due to the interference of test noise, the values ​​of other eigenvalues ​​are close to 0. K+1 , K+2 , , M The corresponding eigenvector is defined as the noise subspace.

[0051] Step 4: Establish an estimation vector of the integer order of blade vibration frequency:

[0052]

[0053] Wherein, ΔN represents the non-integer order of the blade vibration frequency obtained in the first step; β represents the installation angle of the timing sensor.

[0054] Step 5: Set the search range of integer orders of blade asynchronous vibration frequency to identify the integer orders of asynchronous blade vibration frequency. The search formula for integer orders of blade asynchronous vibration frequency is:

[0055]

[0056] Among them, E n Represents the noise subspace obtained in the third step. Function Pm-R (N) and P m-L (N) The N corresponding to the maximum value is the integer order of the asynchronous vibration frequency of the blade.

[0057] Step 6: According to the non-integer order of the blade asynchronous vibration frequency obtained in the first step and the integer order of the blade asynchronous vibration frequency obtained in the fifth step, the real frequency of the blade asynchronous vibration can be obtained, and the calculation formula is:

[0058]

[0059] Among them, max(g) means taking the maximum value.

[0060] The rotation frequency of the moving blade rotor is the average rotation frequency of the blade passing through different sensors; the installation angle of the timing sensor is a relative value; the timing sensor continuously collects M blade vibration signals. In order to ensure that all blade asynchronous vibration frequencies can be identified, this embodiment requires M>K, and M is an integer multiple of the number of timing sensors.

[0061] Specifically,

[0062] The data of this embodiment comes from a model-level centrifugal compressor impeller vibration test bench. The test bench structure mainly includes a drive motor, a gear transmission, a centrifugal compressor, etc. The rated speed of the centrifugal impeller is 9000rpm, and the number of impeller blades is 19. The impeller vibration test is carried out by BTT technology and strain gauge method, and the strain gauge measurement results can verify the accuracy of the blade vibration frequency estimation results. Four timing sensors are arranged at equal intervals at the inlet of the centrifugal compressor, and the angle between adjacent sensors is 10°. Strain gauges are pasted on the surface of the impeller blades. The BTT measurement system consists of a fiber optic timing sensor, a laser generator, a counter, and a test software developed based on LabVIEW. The strain measurement system includes strain gauges, telemetry, a collector, and DASP dynamic data acquisition software.

[0063] Step 1: Set the centrifugal compressor rotor speed to 5028 rpm. When the compressor is running stably, collect the blade vibration signal with a data length of 60 seconds. The vibration displacement of blade 7 obtained by the four timing sensors is as follows: Figure 1 shown.

[0064] Step 2: Perform FFT processing on the blade vibration signal. The frequency spectrum of the vibration displacement of blade 7 is as follows: Figure 2 It can be concluded that the amplitude of the asynchronous vibration of the moving blade is 0.006 mm, and the non-integer order of the asynchronous vibration frequency of the moving blade is ΔN=34.12 / 83.8≈0.41.

[0065] Step 3: Set the search range of the integer order of the asynchronous vibration frequency of the moving blade to [0, 10], and obtain the search results of the integer order of the asynchronous vibration frequency of the blade according to formula (5), as shown in Figure (3). Figure (3) shows that the integer order of the asynchronous vibration frequency of blade No. 7 is N = 8. Further, according to formula (6), it can be obtained that the asynchronous vibration frequency of blade No. 7 is (8-0.41) × 83.8 = 636.04 Hz.

[0066] Step 4: Analyze the strain signal of blade 7. Figure (4) shows the strain signal of blade 7 and the spectrum analysis results of the strain signal. It can be concluded that the asynchronous vibration frequency of the blade is 635Hz. Compared with the strain gauge method, it can be concluded that the absolute error of the frequency identification of the present invention is 1.04Hz, and the relative error is only 0.16%, which can prove the accuracy of the present invention.

[0067] The above is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any technician in the relevant technical field, without departing from the scope of the technical solution of the present invention, makes any form of equivalent replacement or modification to the technical solution and technical content disclosed in the present invention, which does not depart from the content of the technical solution of the present invention and still falls within the protection scope of the present invention.

Claims

1. A method for identifying asynchronous vibration parameters of moving blades based on order search, characterized in that: Firstly, the under-sampled signal of asynchronous vibration of moving blades is transformed by fast Fourier transform to obtain the amplitude corresponding to the asynchronous vibration frequency of blades and the non-integer order of asynchronous vibration frequency. Then, the existing MUSIC algorithm is optimized to convert the frequency search into order search to realize the estimation of integer order of asynchronous vibration frequency of moving blades. Finally, the asynchronous vibration parameters of moving blades are obtained by integrating the FFT analysis results and the order search results.

2. The method for identifying parameters of asynchronous vibration of moving blades based on order search according to claim 1 is characterized in that: The steps include: Step 1: Perform FFT processing on the under-sampled signal of asynchronous vibration of moving blades obtained by BTT technology to obtain the amplitude and non-integer order of the asynchronous vibration frequency of blades. The asynchronous vibration frequency of blades is expressed as: f=(N+ΔN)f r (1) Where f represents the asynchronous vibration frequency of the blade; N represents the integer order of the asynchronous vibration frequency of the blade; ΔN represents the non-integer order of the asynchronous vibration frequency of the blade; f r Indicates the rotation frequency of the moving blade rotor; Step 2: Based on the under-sampling signal of asynchronous vibration of moving blades obtained by BTT technology, the blade vibration autocorrelation matrix is ​​established. M blade vibration signals are continuously collected through timing sensors, and the signal vector of the qth sampling is defined as: Among them, y(r q ) indicates t q The blade vibration displacement collected at the moment, the autocorrelation matrix calculation formula of the blade vibration signal is: R=E{y(q)y H (q)} (3) Among them, y H (q) represents the transpose of the signal vector y(q); Step 3: Perform eigenvalue decomposition on the autocorrelation matrix obtained in the second step. Theoretically, the autocorrelation matrix has K non-zero eigenvalues, where K represents the number of blade vibration frequency components. In actual engineering applications, due to the interference of test noise, the values ​​of other eigenvalues ​​are close to 0. The eigenvalue λ close to 0 is K+1 , K+2 , λ M The corresponding eigenvector Defined as the noise subspace; Step 4: Establish an estimation vector of the integer order of blade vibration frequency: Wherein, ΔN represents the non-integer order of the blade vibration frequency obtained in the first step; β represents the installation angle of the timing sensor; Step 5: Set the search range of integer orders of blade asynchronous vibration frequency, and identify the integer orders of asynchronous blade vibration frequency. The search formula for integer orders of blade asynchronous vibration frequency is: Among them, E n Represents the noise subspace obtained in the third step, function P m-R (N) and P m-L (N) The N corresponding to the maximum value is the integer order of the blade asynchronous vibration frequency; Step 6: According to the non-integer order of the blade asynchronous vibration frequency obtained in the first step and the integer order of the blade asynchronous vibration frequency obtained in the fifth step, the real frequency of the blade asynchronous vibration can be obtained, and the calculation formula is: Among them, max(g) means taking the maximum value.

3. The method for identifying parameters of asynchronous vibration of moving blades based on order search according to claim 2 is characterized in that: The rotation frequency of the moving blade rotor is the average rotation frequency of the blades passing through different sensors.

4. The method for identifying parameters of asynchronous vibration of moving blades based on order search according to claim 2 is characterized in that: The installation angle of the timing sensor is a relative value.

5. The method for identifying parameters of asynchronous vibration of moving blades based on order search according to claim 2 is characterized in that: The timing sensor continuously collects M blade vibration signals, where M>K, and M is an integer multiple of the number of timing sensors.

Images