Steam turbine blade vibration signal fault detection method based on HVD and spectral kurtosis analysis

Abstract

The invention discloses a steam turbine blade vibration signal fault detection method based on HVD and spectral kurtosis analysis, and the method comprises the steps: decomposing an original vibration signal of a steam turbine blade through Hilbert vibration decomposition HVD, and obtaining a plurality of intrinsic mode function IMF components; calculating mutual information of each IMF and an original vibration signal of the turbine blade; screening the IMF component corresponding to the maximum mutual information as a real component; performing spectral kurtosis analysis on the selected real component to obtain a fast spectral kurtosis map, determining a frequency band at the maximum spectral kurtosis on the fast spectral kurtosis map as a target frequency band, and determining a center frequency and a bandwidth corresponding to the target frequency band; designing a band-pass filter according to the center frequency and the bandwidth, and carrying out filtering processing on a real component; and calculating an envelope spectrum of the filtered signal, and extracting a fault characteristic frequency in the original vibration signal of the turbine blade from the envelope spectrum. The method can achieve the fault detection of the vibration signal of the turbine blade, and is high in flexibility and precision.

Description

Technical Field

[0001] This application belongs to the technical field of fault detection, and specifically relates to a fault detection method for turbine blade vibration signals based on HVD and spectral kurtosis analysis. Background Art

[0002] In recent years, with the continuous improvement of the requirements for turbine efficiency, major design and manufacturing enterprises have taken optimizing the performance of the low-pressure cylinder as the key research direction. By increasing the effective work done in the low-pressure cylinder to improve the unit efficiency, the design of the blades has gradually developed towards longer sizes. However, the increase in blade size makes the working environment during operation more complex. Especially under deep peak shaving conditions, the dynamic characteristics of the blades are more unstable, so higher requirements are put forward for the condition monitoring and fault diagnosis of the last-stage turbine blades. The current fault diagnosis technology mainly discriminates the fault type by collecting real-time vibration signals and combining methods such as time-frequency domain analysis.

[0003] Currently, the commonly used vibration signal analysis methods include wavelet analysis and empirical mode decomposition (EMD). Wavelet analysis depends on the empirical selection of wavelet bases and is difficult to adapt to the changes in different signal characteristics, so its flexibility in application is poor. As an adaptive signal decomposition method, EMD can decompose complex vibration signals into multiple modal components. However, EMD is sensitive to noise and is prone to being interfered, resulting in the occurrence of modal aliasing, thus affecting the accurate extraction of fault characteristics.

[0004] Therefore, it is necessary to provide a fault detection method for turbine blade vibration signals with stronger flexibility and higher accuracy. Summary of the Invention

[0005] The purpose of this application is to provide a fault detection method for turbine blade vibration signals based on HVD and spectral kurtosis analysis. This application can realize the fault detection of turbine blade vibration signals, with strong flexibility and high accuracy.

[0006] The technical solution provided by this application is as follows:

[0007] In the first aspect, this application provides a fault detection method for turbine blade vibration signals based on HVD and spectral kurtosis analysis, including:

[0008] Decompose the original vibration signal of the turbine blade through Hilbert vibration decomposition HVD to obtain several intrinsic mode function IMF components;

[0009] Calculate the mutual information between each IMF and the original vibration signal of the turbine blade;

[0010] Select the IMF component corresponding to the maximum mutual information as the true component;

[0011] For the selected true component, perform spectral kurtosis analysis to obtain a fast spectral kurtosis diagram, determine the frequency band at the maximum spectral kurtosis on the fast spectral kurtosis diagram as the target frequency band, and determine the corresponding center frequency and bandwidth of the target frequency band; design a band-pass filter according to the center frequency and bandwidth, and perform filtering processing on the true component;

[0012] Calculate the envelope spectrum of the filtered signal, and extract the fault characteristic frequency in the original vibration signal of the steam turbine blade from the envelope spectrum.

[0013] In a possible implementation, decompose the original vibration signal of the steam turbine blade by HVD to obtain several intrinsic mode function IMF components, including:

[0014] 1) Set the iteration termination condition; use the original vibration signal of the steam turbine blade as the signal to be analyzed ;

[0015] 2) Perform Hilbert transform on the signal to be analyzed to obtain its corresponding instantaneous frequency ;

[0016] 3) Based on the instantaneous frequency , estimate the frequency of the component with the largest amplitude in the signal to be analyzed ;

[0017] 4) Using as the frequency, construct two orthogonal signals using sine and cosine functions, multiply the two orthogonal signals by respectively, and synchronously detect the instantaneous amplitude and phase of the periodic wave signal corresponding to the instantaneous frequency, and thus separate and obtain the component with the largest amplitude in :

[0018] ;

[0019] Take as the component extracted in this iteration;

[0020] 5) Calculate the difference between and :

[0021] ;

[0022] Judge whether the iteration termination condition is satisfied. If not, take as the signal to be analyzed in the next round of iteration ; return to step 2) for iteration;

[0023] Otherwise, end the iteration, and denote the components extracted in each iteration as , which are the IMF components of the original vibration signal of the steam turbine blade.

[0024] In a possible implementation, the iteration termination condition is set as: the number of iterations reaches the maximum number of iterations N or the normalized standard deviation in each iteration is less than the set threshold.

[0025] In a possible implementation, the formula for calculating the normalized standard deviation is:

[0026] ;

[0027] ;

[0028] where is the standard deviation, represents the value of calculated in the th round of iteration, is the number of iterations; is the th iteration average value of is th iteration maximum value of the absolute value of

[0029] In a possible implementation, calculating the mutual information between each IMF and the original vibration signal of the steam turbine blade includes:

[0030] Calculate the mutual information between and the original vibration signal of the steam turbine blade respectively;

[0031] and the mutual information between The calculation formula is:

[0032] ;

[0033] where is the unconditional entropy of the original vibration signal of the steam turbine blade, is the conditional entropy of the original vibration signal of the steam turbine blade when the value of the known decomposition signal is taken.

[0034] In a possible implementation, performing spectral kurtosis analysis on the selected true components includes:

[0035] Let the selected true component be , and perform spectral kurtosis analysis on , including:

[0036] Let , where is a noise signal independent of the signal , then the spectral kurtosis of the signal is:

[0037] ;

[0038] where: is the spectral kurtosis of the signal , is the spectral kurtosis of the noise , and the calculation formula is as follows:

[0039] ;

[0040] is the fourth-order spectral cumulant of ;

[0041] ;

[0042] is the fourth-order time-averaged moment of , and is the second-order time-averaged moment of ;

[0043] ;

[0044] is the fourth-order spectral cumulant of ;

[0045] ;

[0046] is the fourth-order time-averaged moment of , and is the second-order time-averaged moment of ;

[0047] is the noise-to-signal ratio, and its calculation formula is:

[0048] .

[0049] In a possible implementation, when is stationary white noise, .

[0050] In a second aspect, the present application provides an electronic device, including: a memory and a processor;

[0051] The memory is used for storing a computer program;

[0052] The processor is used for calling the computer program to execute the method as described above.

[0053] In a third aspect, the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer program runs on an electronic device, the electronic device is enabled to implement the method as described above.

[0054] In a fourth aspect, the present application provides a computer program product, including a computer program. When the computer program runs on an electronic device, the electronic device is enabled to implement the method as described above.

[0055] For the specific implementation manners of the second to fourth aspects of the present application, reference may be made to the implementation manner of the first aspect above, and details are not described herein again.

[0056] Beneficial effects:

[0057] For the fault detection of the vibration signal of the last-stage blade of a steam turbine low-pressure cylinder, the present invention proposes a signal processing method based on Hilbert vibration decomposition (HVD) and spectral kurtosis analysis.

[0058] (1) Enhance the accuracy of signal decomposition. This method uses spectral kurtosis technology to automatically screen the target frequency band, thereby improving the accuracy of signal decomposition. Compared with directly decomposing the full-frequency band signal, it can more accurately focus on the fault-related components and reduce the influence of interference noise.

[0059] (2) Reduce the influence of modal aliasing. By combining spectral kurtosis and HVD decomposition strategies, the common modal aliasing problem in traditional methods is effectively avoided. Frequency band focusing makes the separation of each modal signal clearer, reduces the interference of irrelevant frequency components on the fault characteristics, and improves the reliability of the decomposition result.

[0060] (3) Realize the automatic selection of the target frequency band. Spectral kurtosis analysis provides an automatic frequency band positioning tool, which can directly identify the key frequency band containing impact characteristics in the signal and reduce the need for manual intervention.

[0061] (4) Wide applicability and flexibility. This method is suitable for processing the vibration signal analysis of the last-stage blade of a steam turbine, and is also applicable to the feature extraction of complex non-linear signals in other rotating machinery systems. It has strong generality and adaptability and can cope with various engineering application scenarios.

[0062] (5) Improve the efficiency and accuracy of fault diagnosis. This method can quickly detect the fault characteristics of the blade, significantly improving the efficiency and accuracy of equipment operation status monitoring, and providing reliable support for fault identification and status monitoring in engineering practical applications. Description of the Drawings

[0063] Figure 1 It is a flowchart of an embodiment of this application. Specific Embodiments

[0064] In order to enable those skilled in the art to better understand the solution of this application, the technical solutions in the embodiments of this application will be further described clearly and completely below in conjunction with the drawings in the embodiments of this application.

[0065] Embodiment 1

[0066] As Figure 1 shown, the embodiment of this application provides a method for detecting faults in vibration signals of steam turbine blades based on HVD and spectral kurtosis analysis. First, the original vibration signal of the steam turbine blade is decomposed by Hilbert vibration decomposition (HVD) to obtain several intrinsic mode function (IMF) components; then the mutual information between each IMF and the original vibration signal of the steam turbine blade is calculated; the IMF component corresponding to the maximum mutual information is selected as the true component, that is, the mutual information analysis is used to quantify the correlation between each IMF component and the original vibration signal of the steam turbine blade, and the IMF component most relevant to the original vibration signal of the steam turbine blade is selected as the true component, and other IMF components are used as false components. For the selected true component, spectral kurtosis analysis is performed to obtain a fast spectral kurtosis map, the frequency band at the maximum spectral kurtosis on the fast spectral kurtosis map is determined as the target frequency band, and the corresponding center frequency and bandwidth of the target frequency band are determined; a band-pass filter is designed according to the center frequency and bandwidth, and the true component is filtered; finally, the envelope spectrum of the filtered signal is calculated, and the fault characteristic frequency in the original vibration signal of the steam turbine blade is extracted from the envelope spectrum.

[0067] By using the Hilbert transform to generate an analytic signal and combining low-pass filtering and synchronous detection techniques, HVD can adaptively decompose the signal and extract the main modal components. HVD effectively overcomes the modal aliasing problem in traditional methods and at the same time retains good adaptive decomposition ability.

[0068] Spectral Kurtosis (SK) analysis, as a spectral analysis method based on statistical characteristics, can identify the impact characteristics in a signal and locate the target frequency band by evaluating the degree of non-Gaussianity of the signal spectrum. Different from traditional time-frequency analysis methods, spectral kurtosis can quickly determine the frequency region related to faults, thus more accurately focusing on the signal components containing characteristic information. Its superiority in target frequency band selection provides an efficient and reliable means for feature extraction of complex vibration signals.

[0069] This application effectively combines the adaptive decomposition advantage of HVD and the frequency band focusing characteristic of spectral kurtosis, achieving remarkable results in the fault feature extraction of the last-stage blade of the low-pressure cylinder of a steam turbine, and providing an efficient and reliable solution for the vibration signal analysis of rotating machinery.

[0070] In one embodiment, the original vibration signal of the steam turbine blade is decomposed by Hilbert vibration decomposition (HVD) including:

[0071] 1) Set the iteration termination condition; take the original vibration signal of the steam turbine blade as the signal to be analyzed ;

[0072] 2) Perform Hilbert transform on the signal to be analyzed to obtain its corresponding instantaneous frequency ;

[0073] For any non-linear and non-stationary signal , its Hilbert transform can be expressed as:

[0074] ;

[0075] where and are conjugate complex numbers, and are both time variables, represents the Hilbert transform function;

[0076] Based on the Hilbert vibration decomposition result, the analytic signal of the signal is obtained:

[0077] = ;

[0078] where is the imaginary unit; represents the instantaneous amplitude, , represents the instantaneous phase, representing the analytic signal The phase information is used to describe the phase change of the signal at moment; ;

[0079] is a complex number, is its real part, is its imaginary part. The imaginary part perpendicular to the real number is represented by .

[0080] The instantaneous frequency corresponding to the analytic signal of the signal is calculated as:

[0081] ;

[0082] 3) Estimate the frequency of the component with the largest amplitude of the signal to be analyzed ;

[0083] The multi-component non-stationary signal can be decomposed. According to Equation (2), the instantaneous frequency of the signal can also be expressed as:

[0084] ;

[0085] In the formula: The first half is the instantaneous frequency corresponding to the component with the largest amplitude of the signal ; After the time-domain signal is converted into a frequency-domain signal, each specific frequency of the converted frequency-domain signal has an energy (value) size, which is called the amplitude; The frequency-domain signal is represented by a spectrogram. The abscissa of the spectrogram is the frequency, and the ordinate is the amplitude; is the frequency (abscissa) corresponding to the point with the largest amplitude (the largest ordinate) in the spectrogram; The latter part of the above formula represents the oscillating frequency part that changes rapidly compared with the former. Among them, represents the component number, that is, the th component after the signal decomposition; represents the total number of components; represents the instantaneous amplitude; represents the th instantaneous amplitude of the component. The latter can be filtered out by using the low-pass filter of HVD to extract the instantaneous frequency corresponding to the component with the largest amplitude.

[0086] 4) Using as the frequency, two orthogonal signals are constructed using the sine and cosine functions, and the two orthogonal signals are multiplied by respectively to synchronously detect the instantaneous amplitude of the periodic wave signal corresponding to the instantaneous frequency and the phase , from which the component with the largest amplitude is separated in :

[0087] ;

[0088] In the formula, represents the corresponding instantaneous amplitude, ; represents the corresponding instantaneous phase, ;

[0089] Take as the component extracted in this iteration;

[0090] 5) Calculate the difference between and :

[0091] ;

[0092] Judge whether the iteration termination condition is satisfied. If not, take as the signal to be analyzed in the next round of iteration ; Return to step 2) for iteration;

[0093] Otherwise, end the iteration, and record the components extracted in each iteration as respectively, as the IMF components of the original vibration signal of the steam turbine blade.

[0094] In some embodiments, the iteration termination condition is set as: the number of iterations reaches the maximum number of iterations N or the normalized standard deviation of in each iteration process is less than the set threshold.

[0095] In some embodiments, < 0.001 can be used as the termination condition of the iteration.

[0096] The standard deviation represents the degree of deviation of the data from the mean value, and the calculation formula is:

[0097] ;

[0098] In the formula, represents the value of calculated in the round of iteration, is the number of iterations; is the average value of calculated in the

[0099] The normalized standard deviation is to make the standard deviation comparable to the overall amplitude range of the signal, usually achieved by normalizing the signal, and the formula is:

[0100] ;

[0101] In the formula, is the maximum value of the absolute value of obtained by the

[0102] In some embodiments, calculate the mutual information between each IMF and the original vibration signal of the steam turbine blade, that is, calculate the mutual information between and the original vibration signal of the steam turbine blade respectively.

[0103] Mutual Information (MI) originates from the concept of entropy and is a method to measure the average amount of information contained in a signal. Mutual information is used to evaluate the similarity degree between the probability density functions of multiple random variables. For two correlated decomposed signals and the original vibration signal of the steam turbine blade , when the value of the decomposed signal is known, the conditional entropy is usually less than or equal to the unconditional entropy of the original vibration signal of the steam turbine blade . Therefore, after the decomposed signal is known, the reduction in the uncertainty of the original vibration signal of the steam turbine blade can be expressed as , and this value is defined as the mutual information between and , and its formula is:

[0104] ;

[0105] The above formula shows that when the mutual information between two signals is large, their correlation is stronger; on the contrary, if the mutual information is zero, it means that the two signals are completely independent and irrelevant.

[0106] In some embodiments, let the selected true component be , and perform spectral kurtosis analysis on , including:

[0107] For a non-linear and non-stationary signal , where is a noise signal independent of the signal , then the spectral kurtosis of the signal is:

[0108] ;

[0109] Wherein: is the spectral kurtosis of the signal ; is the spectral kurtosis of the noise , and the calculation formula is as follows:

[0110] ;

[0111] is 's fourth-order spectral cumulant;

[0112] ;

[0113] is 's fourth-order time-averaged moment, is 's second-order time-averaged moment;

[0114] ;

[0115] is 's fourth-order spectral cumulant;

[0116] ;

[0117] is 's fourth-order time-averaged moment, is 's second-order time-averaged moment;

[0118] is the signal-to-noise ratio, and its calculation formula is:

[0119] ;

[0120] When is stationary white noise, the spectral kurtosis of the signal is simplified to:

[0121] ;

[0122] As can be seen from the above formula, at the frequency where the signal-to-noise ratio of the signal is very high, then is close to zero; at the frequency where the signal-to-noise ratio of the signal is very small, then is approximately equal to . Therefore, by calculating the spectral kurtosis of the entire frequency domain, the frequency band with the largest spectral kurtosis can be found, which is the target frequency band.

[0123] By performing HVD decomposition on the vibration signals of the last-stage blades of the steam turbine low-pressure cylinder, after extracting each modal component, the mutual information is used to analyze the correlation between each component and the original signal, and it is determined that the mutual information value between the true component and the original signal is significantly higher than that of the false component, indicating that there is a higher similarity between the true component and the original signal. Subsequently, spectral kurtosis analysis is performed on the selected true components to accurately locate the target frequency band containing fault characteristics.

[0124] In the verification stage, a band-pass filter is designed to filter the target frequency band, the envelope spectrum of the filtered signal is calculated, and the fault characteristic frequencies in the original vibration signals of the steam turbine blades are extracted based on the envelope spectrum. The results show that this method can effectively eliminate interference components and accurately extract blade fault characteristic information, with high accuracy and applicability.

[0125] Embodiment 2

[0126] This embodiment provides an electronic device, including: a memory and a processor;

[0127] The memory is used to store a computer program;

[0128] The processor is used to call the computer program to execute the method as described in Embodiment 1.

[0129] Embodiment 3

[0130] This embodiment provides a computer-readable storage medium, in which a computer program is stored. When the computer program runs on an electronic device, the electronic device is enabled to implement the method as described in Embodiment 1.

[0131] Embodiment 4

[0132] This embodiment provides a computer program product, including a computer program. When the computer program runs on an electronic device, the electronic device is enabled to implement the method as described in Embodiment 1.

[0133] The specific implementation manners of a system, an electronic device, a computer-readable storage medium, and a computer program product provided in the embodiments of the present application may refer to the specific embodiments of the above method, and will not be elaborated herein.

[0134] Obviously, those skilled in the art should understand that the various units or steps of the present application described above can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed over a network composed of multiple computing devices. Optionally, they can be implemented by program code executable by the computing device. Thus, they can be stored in a storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module to implement. In this way, the present application is not limited to any specific combination of hardware and software.

[0135] The above are only the preferred embodiments of the present application and are not used to limit the present application. For those skilled in the art, the present application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A method for detecting faults in turbine blade vibration signals based on HVD and spectral kurtosis analysis, characterized in that: include: The original vibration signal of the turbine blade is decomposed by Hilbert vibration decomposition (HVD) to obtain several intrinsic mode function (IMF) components. Calculate the mutual information between each IMF and the original vibration signal of the turbine blade; Filter the IMF component corresponding to the maximum mutual information as the true component; Perform spectral kurtosis analysis on the selected real component to obtain a fast spectral kurtosis diagram, determine the frequency band at the maximum spectral kurtosis on the fast spectral kurtosis diagram as the target frequency band, and determine the center frequency and bandwidth corresponding to the target frequency band; design a bandpass filter according to the center frequency and bandwidth, and perform filtering processing on the real component; The envelope spectrum of the filtered signal is calculated, and the fault characteristic frequency in the original vibration signal of the turbine blade is extracted from the envelope spectrum.

2. The method according to claim 1, characterized in that The original vibration signal of the turbine blade is decomposed by HVD to obtain several intrinsic mode function (IMF) components, including: 1) Set the iteration termination condition; take the original vibration signal of the turbine blade as the signal to be analyzed ; 2) Signals to be analyzed Perform Hilbert transform to obtain the corresponding instantaneous frequency ; 3) Based on instantaneous frequency , estimate the signal to be analyzed The frequency of the maximum amplitude component ; 4) As frequency, two orthogonal signals are constructed using sine and cosine functions. Multiply and synchronously detect the instantaneous frequency The instantaneous amplitude of the corresponding periodic wave signal and Phase , thus separated The component with the largest amplitude : ； Will As the component extracted in this iteration; 5) Calculation and The difference: ； Determine whether the iteration termination condition is met. If not, As the signal to be analyzed in the next iteration ; Return to step 2) for iteration; Otherwise, the iteration ends and the components extracted in each iteration are recorded as , as the IMF components of the original vibration signal of the turbine blade.

3. The method according to claim 2, characterized in that The iteration termination condition is set as follows: the number of iterations reaches the maximum number of iterations N or during each iteration The normalized standard deviation of Less than the set threshold.

4. The method according to claim 2, characterized in that: The normalized standard deviation The calculation formula is: ； ； In the formula, is the standard deviation, Indicates The iterative calculation results are The value of is the number of iterations; for The iterative calculation The average value of for The iterative calculation The maximum absolute value of .

5. The method according to claim 2, characterized in that: Calculate the mutual information between each IMF and the original vibration signal of the turbine blade, including: Calculate separately Mutual information with the original vibration signal of the steam turbine blade; and Mutual information between The calculation formula is: ； in, is the original vibration signal of the turbine blade The unconditional entropy of For a known decomposition signal When the value of The conditional entropy of .

6. The method according to claim 1, characterized in that For the selected real components, spectral kurtosis analysis is performed, including: Suppose the selected true component is ,right Perform spectral kurtosis analysis, including: make ,in Is a signal independent The noise signal, then the signal The spectral kurtosis for: ； in: For signal The spectral kurtosis of For noise The spectral kurtosis is calculated as follows: ； for The fourth-order spectral cumulant of ； for The fourth-order time-averaged moment of for The second-order time-averaged moment of ； for The fourth-order spectral cumulant of ； for The fourth-order time-averaged moment of for The second-order time-averaged moment of is the signal-to-noise ratio, and its calculation formula is: 。 7. The method according to claim 6, characterized in that when When it is a stationary white noise, .

8. An electronic device, characterized in that: include: Memory and processor; The memory is used to store computer programs; The processor is configured to call the computer program to execute the method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed on an electronic device, the electronic device implements the method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed on an electronic device, the electronic device implements the method according to any one of claims 1 to 7.

Images