A method for extracting propulsion pump modulation frequency based on weighted enhanced envelope spectrum

Abstract

The present invention discloses a method for extracting the modulation frequency of a propulsion pump based on a weighted enhanced envelope spectrum, comprising: (1) collecting a hydroacoustic signal of a propulsion pump, removing the mean value to obtain a centered signal; (2) estimating a first-order cyclostationary signal component in the centered signal, removing the component and obtaining a de-noised signal; (3) calculating a spectral correlation function of the de-noised signal; (4) calculating an enhanced envelope spectrum and a spectral coherence function of the de-noised signal; (5) preliminarily locking the modulation fundamental frequency of the propulsion pump from the enhanced envelope spectrum; (6) performing square slicing processing on the spectral coherence function at the modulation fundamental frequency to obtain an adaptive weighting function; (7) performing weighted processing on the spectral correlation function using the adaptive weighting function to obtain a weighted spectral correlation function; (8) calculating a weighted enhanced envelope spectrum based on the weighted spectral correlation function. By using the present invention, the modulation frequency component in the hydroacoustic signal of the propulsion pump can be effectively extracted under low signal-to-noise ratio and cyclostationary noise interference.

Description

Technical Field

[0001] The invention belongs to the field of signal processing, and in particular relates to a method for extracting modulation frequency of a propulsion pump based on a weighted enhanced envelope spectrum. Background Art

[0002] Propulsion pumps are widely used in fields such as ship propulsion. The extraction of their characteristic frequencies can support the status monitoring of our equipment and help to detect and identify unknown targets. Traditional characteristic frequency extraction methods mainly include spectrum analysis and demodulation analysis.

[0003] The main way to implement spectrum analysis is Fourier transform, which decomposes the time domain signal into the sum of trigonometric functions of different frequencies, thereby depicting the amplitude and phase differences of signal components of different frequencies.

[0004] For example, a Chinese patent document with publication number CN107677362A discloses an improved method for extracting the modulation frequency of rotating machinery based on real cepstrum analysis. Spectrum analysis can better describe the frequency distribution of deterministic signals, but is not suitable for the analysis of random signals. In practice, the propulsion pump water acoustic signal is essentially a random signal, which limits the applicability of spectrum analysis.

[0005] Demodulation analysis is an effective way to extract characteristic frequency information, mainly including signal processing methods such as narrowband envelope demodulation and cyclostationary analysis. For example, the Chinese patent document with publication number CN114166507A discloses a harmonic identification method based on fast spectral kurtosis. Narrowband envelope demodulation locates the frequency band that is most likely to contain the modulated signal through indicators such as spectral kurtosis, filters the frequency band, and then obtains the envelope curve of the filtered signal, and finally performs Fourier transform to obtain the envelope spectrum. This method is suitable for local defect detection and weak fault diagnosis of general parts such as gears and bearings, but this method is sensitive to non-Gaussian noise and is prone to failure under low signal-to-noise ratio conditions containing multiple types of noise, so it is subject to certain limitations.

[0006] Cyclostationary analysis can make full use of the modulation components in the time domain signal, achieve asymptotic immunity to Gaussian white noise, and effectively extract the modulation frequency components. However, this method is also subject to certain limitations in the application of propulsion pump modulation frequency extraction. This is mainly because in addition to characteristic modulation frequencies such as shaft frequency, blade frequency and its harmonics, the collected hydroacoustic signals usually also include cyclostationary noise interference from other mechanical, electrical and communication equipment, and traditional cyclostationary analysis cannot effectively remove these noise interferences. Summary of the invention

[0007] The present invention provides a method for extracting the modulation frequency of a propulsion pump based on a weighted enhanced envelope spectrum, which can effectively extract the modulation frequency components in the propulsion pump hydroacoustic signal under low signal-to-noise ratio and cyclostationary noise interference, and can be applied to the fields of detection and identification, status monitoring and fault diagnosis of various propulsion pumps.

[0008] A method for extracting propulsion pump modulation frequency based on weighted enhanced envelope spectrum comprises the following steps:

[0009] (1) Collecting the hydroacoustic signal of the propulsion pump as a monitoring signal, and de-averaging the monitoring signal to obtain a centralized signal;

[0010] (2) Using a linear predictor to estimate the first-order cyclostationary signal component in the centralized signal, and removing it to obtain a denoised signal;

[0011] (3) Calculate the spectral correlation function of the noise reduction signal;

[0012] (4) Based on the spectral correlation function, the enhanced envelope spectrum and spectral coherence function of the noise reduction signal are calculated;

[0013] (5) Based on the prior information of the propulsion pump shaft frequency, the modulation fundamental frequency of the propulsion pump is preliminarily locked from the enhanced envelope spectrum;

[0014] (6) performing square slicing processing on the spectral coherence function at the modulation base frequency to obtain an adaptive weighting function;

[0015] (7) performing weighted processing on the spectral correlation function using an adaptive weighting function to obtain a weighted spectral correlation function;

[0016] (8) Based on the weighted spectral correlation function, the weighted enhanced envelope spectrum is calculated.

[0017] The method of the present invention can overcome the defects of existing spectrum analysis and demodulation methods that cannot fully and accurately extract the modulation frequency of the propulsion pump, or that the demodulation results contain too much noise interference. The method is simple and easy to implement, and can effectively realize the extraction and characterization of the modulation frequency of the propulsion pump under low signal-to-noise ratio conditions and cyclostationary noise interference.

[0018] Furthermore, in step (1), the specific steps of obtaining the centralized signal are:

[0019] (1-1) Calculate the average value of the monitoring signal;

[0020] (1-2) Subtract the average value from the monitoring signal to obtain the centralized signal.

[0021] The specific steps of step (2) are:

[0022] (2-1) According to the structural characteristics and operating parameters of the propulsion pump, the expression of the first-order cyclostationary signal is preset;

[0023] (2-2) The time domain synchronous averaging method is used to estimate the first-order cyclostationary signal component in the centralized signal;

[0024] (2-3) Subtract the estimated first-order cyclostationary signal from the centered signal to obtain the denoised signal.

[0025] The specific process of step (3) is as follows:

[0026] (3-1) Calculate the time-varying autocorrelation function R of the noise reduction signal x(t) x (t,τ), the calculation formula is as follows:

[0027]

[0028] In the formula, t represents time, τ represents time delay, * is the conjugate symbol, and E[·] represents the statistical mean;

[0029] (3-2) Calculate the cyclic autocorrelation function R of the noise reduction signal x(t) x (α,τ), the calculation formula is as follows:

[0030]

[0031] Where -T / 2 and T / 2 are the lower and upper limits of time, e is a natural constant, j is the symbol of an imaginary number, and α is the modulation frequency;

[0032] (3-3) Calculate the spectral correlation function S of the noise reduction signal x(t) x (α,f), the calculation formula is as follows:

[0033]

[0034] Where f is the carrier frequency.

[0035] The specific process of step (4) is as follows:

[0036] (4-1) Calculate the enhanced envelope spectrum ESC of the noise reduction signal x (α), the calculation formula is as follows:

[0037]

[0038] Where f1 and f2 are the lower and upper limits of the carrier frequency integral, and their default values ​​are 0 and F s / 2,F s is the sampling frequency of the monitoring signal;

[0039] (4-2) Calculate the spectral coherence function γ of the noise reduction signal x (α,f), the calculation formula is as follows:

[0040]

[0041] Where P x (f) represents the power spectral density of the noise reduction signal x(t), which is equal to the result when the modulation frequency of the spectrum correlation is zero, that is, S x (0,f)=P x (f).

[0042] The specific process of step (5) is as follows:

[0043] (5-1) Based on the prior information, the estimated value of the shaft frequency of the propulsion pump is obtained;

[0044] (5-2) Since the modulation frequency of the enhanced envelope spectrum is an integer multiple of the modulation frequency resolution Δα, the enhanced envelope spectrum ESC x (α) to obtain the modulation frequency value α that is closest to the axis frequency estimation value opt =KΔα;

[0045] (5-3) The modulation frequency range [α opt -5Δα,α opt +5Δα] is determined as the modulation base frequency candidate range, Δα is the modulation frequency resolution; the modulation frequency corresponding to the maximum value of the enhanced envelope spectrum amplitude is selected from it and determined as the modulation base frequency α b .

[0046] In step (6), the specific process of obtaining the adaptive weighting function is:

[0047] (6-1) Calculate the overall weight function w x (α,f), the calculation formula is as follows:

[0048] w x (α,f)=|γ x (α,f)| 2

[0049] (6-2) Let α = α b , then we can obtain the adaptive weighting function w x (f) = w x (α b ,f).

[0050] In step (7), the calculation process of obtaining the weighted spectral correlation function is:

[0051] (7-1) The spectral correlation function S x (α,f) is expanded into the matrix form S

[0052]

[0053] Where Δα represents the modulation frequency resolution, NΔα represents the modulation frequency upper limit, Δf represents the carrier frequency resolution, MΔf represents the carrier frequency upper limit, and the spectral correlation matrix S∈C (M+1)×(N+1) ;

[0054] (7-2) The adaptive weighting function is replicated and expanded along the modulation frequency axis to form a weighting matrix W

[0055]

[0056] In the formula, the weight matrix W∈C (M+1)×(N+1) ;

[0057] (7-3) Calculate the weighted spectral correlation function WSC(α,f), the calculation formula is as follows:

[0058]

[0059] In the formula, Z∈C (M+1)×(N+1) is the matrix form of the weighted spectral correlation function.

[0060] In step (8), the specific process of calculating the weighted enhanced envelope spectrum is:

[0061] (8-1) Calculate the weighted spectral correlation function |WSC(α,f)| after modulo calculation, the calculation formula is as follows:

[0062]

[0063] (8-2) The modulated weighted spectrum correlation function is accumulated along the carrier frequency axis and averaged to obtain the weighted enhanced envelope spectrum WESC(α), which is calculated as follows:

[0064]

[0065] Where E∈R 1×(N+1) is the matrix form of the weighted enhanced envelope spectrum.

[0066] Compared with the prior art, the present invention has the following beneficial effects:

[0067] (1) The present invention adopts an enhanced envelope spectrum method that first modulates, then integrates, and finally calculates the average of the spectrum correlation function. This method can effectively extract the modulation frequency component in the time domain signal.

[0068] (2) The present invention proposes a method for evaluating the signal-to-noise ratio of a monitoring signal using a spectral coherence function, which can reliably capture the carrier frequency range corresponding to the target modulation signal.

[0069] (3) The method of the present invention performs signal demodulation based on cyclostationary analysis and signal-to-noise ratio weighted processing, and can comprehensively and accurately extract characteristic modulation frequency components such as the shaft frequency of the propulsion pump. The relevant information can be used for detection and identification, status monitoring and fault diagnosis of the propulsion pump. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is a flow chart of a method for extracting modulation frequency of a propulsion pump based on weighted enhanced envelope spectrum of the present invention;

[0071] Figure 2 is the spectral coherence function result containing Gaussian noise and impulse noise in the embodiment of the present invention;

[0072] Figure 3 is the weighted enhanced envelope spectrum result containing Gaussian noise and impulse noise in the embodiment of the present invention;

[0073] Figure 4 is the spectral coherence function result containing cyclostationary noise in the embodiment of the present invention;

[0074] Figure 5 is the result of the adaptive weighting function containing cyclostationary noise in the embodiment of the present invention;

[0075] Figure 6 is the weighted enhanced envelope spectrum result containing cyclostationary noise in the embodiment of the present invention;

[0076] Figure 7 is the narrow-band demodulation result of the kurtosis spectrum of the propulsion pump water acoustic signal in an embodiment of the present invention;

[0077] Figure 8 The demodulation result of the cyclostationary analysis of the propulsion pump water acoustic signal in the embodiment of the present invention;

[0078] Fig. 9 This is the result of the weighted enhanced envelope spectrum of the propulsion pump water acoustic signal in the embodiment of the present invention. DETAILED DESCRIPTION

[0079] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be pointed out that the embodiments described below are intended to facilitate the understanding of the present invention and do not have any limiting effect on the present invention.

[0080] like Figure 1 As shown, a method for extracting propulsion pump modulation frequency based on weighted enhanced envelope spectrum includes the following steps:

[0081] S01, collecting the hydroacoustic signal of the propulsion pump as a monitoring signal, and de-averaging the monitoring signal to obtain a centralized signal.

[0082] S02, using a linear predictor to estimate the first-order cyclostationary signal component in the centralized signal, and removing it to obtain a denoised signal.

[0083] S03, calculating the spectral correlation function of the noise reduction signal.

[0084] S04, based on the spectral correlation function, calculating the enhanced envelope spectrum and the spectral coherence function of the noise reduction signal.

[0085] S05, based on the prior information of the propulsion pump shaft frequency, preliminarily lock the modulation fundamental frequency of the propulsion pump from the enhanced envelope spectrum.

[0086] S06, performing square slicing processing on the spectral coherence function at the modulation base frequency to obtain an adaptive weighting function.

[0087] S07, performing weighted processing on the spectral correlation function using an adaptive weighting function to obtain a weighted spectral correlation function.

[0088] S08, based on the weighted spectrum correlation function, a weighted enhanced envelope spectrum is calculated.

[0089] In order to verify the effectiveness of the present invention, weighted enhanced envelope spectrum calculations are performed on the modulated signal containing Gaussian noise and impulse noise, the modulated signal containing cyclostationary noise, and the propulsion pump water acoustic signal, as follows:

[0090] Use the above method to analyze the modulated signal x1(t) containing Gaussian noise and impulse noise:

[0091]

[0092] In the formula, the number of modulation frequency components is p = 4, the modulation amplitude is A1 = A2 = A3 = A4 = 1, and the modulation frequencies are α1 = 50 Hz, α2 = 100 Hz, α3 = 150 Hz, and α4 = 200 Hz. The number of carrier signals is q = 2, c1(t) and c2(t) are broadband carrier signals, and their corresponding frequency ranges are 2500-7500 Hz and 12500-17500 Hz, respectively. Gaussian white noise n is added. s (t) makes the signal-to-noise ratio 0dB; then, 5 impulse signals with a duration of 0.1s, a signal power of 50dB, and a frequency range of 5-10kHz are randomly added. These impulse signals together constitute n i (t). The corresponding spectral coherence function results and weighted enhanced envelope spectrum results are as follows: Figure 2 and Figure 3As shown. It can be seen that the spectral coherence function has a significantly higher value in the frequency range corresponding to the broadband carrier, which shows that the spectral coherence function can effectively select multiple appropriate demodulation bands. In the weighted enhanced envelope spectrum, the four modulation frequencies α1, α2, α3, α4 and their double frequencies are clearly highlighted, which shows that the modulation frequency is effectively extracted. The above results prove that under the interference of Gaussian noise and impulse noise, the method proposed in the present invention can effectively and accurately extract and characterize the modulation frequency components.

[0093] The above method is used to analyze the modulated signal x2(t) containing cyclostationary noise:

[0094]

[0095] In the formula, the parameter settings of the modulation signal, Gaussian noise, and impulse noise are consistent with the previous example. On this basis, cyclostationary noise x is further added. s (t), the number of modulation frequency components of cyclostationary noise is 4, the frequency values ​​are 70, 140, 210, 280Hz, the modulation amplitude is 1, the broadband carrier frequency range is set to 6500-13500Hz, and the carrier signal power is 40dB. The corresponding spectral coherence function results, adaptive weighting function results and weighted enhanced envelope spectrum results are as follows: Figure 4-6 As shown. It can be seen that the spectral coherence function takes a higher value in the wideband carrier range corresponding to the modulated signal and the cyclostationary noise, which means that the spectral coherence function can locate the carrier frequency bands of all cyclostationary signals, but cannot distinguish between the modulated signal and the cyclostationary noise interference. The adaptive weighting function can further effectively locate the two carrier frequency bands corresponding to the modulated signal, thereby removing the cyclostationary noise interference. In the weighted enhanced envelope spectrum, the four modulation frequencies α1, α2, α3, α4 and their double frequencies are clearly prominent, and basically do not contain other noise interference components, indicating that the modulation frequency is effectively extracted. The above results prove that under the joint interference of Gaussian noise, impulse noise and cyclostationary noise, the method proposed in the present invention can effectively and accurately extract and characterize the modulation frequency components.

[0096] The above method is used to analyze the propulsion pump water acoustic signal. The traditional kurtosis spectrum demodulation results, the cyclostationary analysis demodulation results and the weighted enhanced envelope spectrum results proposed by the present invention are as follows: Figure 7-9 As shown. Due to the diverse excitation sources, large interference noise and complex transmission path of the propulsion pump water acoustic signal, its signal-to-noise ratio is relatively low, and the modulation frequency is difficult to be extracted by traditional methods such as kurtosis spectrum analysis and cyclostationary analysis. In the weighted enhanced envelope spectrum, certain interference components are generated due to the strong electromagnetic noise of the motor, but the shaft frequency and blade frequency are still clearly distinguishable. The above results prove that when processing the propulsion pump water acoustic signal containing strong and complex noise, the method proposed in the present invention can effectively and accurately extract and characterize the modulation frequency component.

[0097] The embodiments described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for extracting propulsion pump modulation frequency based on weighted enhanced envelope spectrum, characterized in that: The following steps are involved: (1) Collecting the hydroacoustic signal of the propulsion pump as a monitoring signal, and de-averaging the monitoring signal to obtain a centralized signal; (2) Using a linear predictor to estimate the first-order cyclostationary signal component in the centralized signal, and removing it to obtain a denoised signal; (3) Calculate the spectral correlation function of the noise reduction signal; (4) Based on the spectral correlation function, the enhanced envelope spectrum and spectral coherence function of the noise reduction signal are calculated; (5) Based on the prior information of the propulsion pump shaft frequency, the modulation fundamental frequency of the propulsion pump is initially locked from the enhanced envelope spectrum; the specific process is as follows: (5-1) Based on the prior information, the estimated value of the shaft frequency of the propulsion pump is obtained; (5-2) Since the modulation frequency of the enhanced envelope spectrum is an integer multiple of the modulation frequency resolution Δα, the enhanced envelope spectrum ESC x (α) to obtain the modulation frequency value α that is closest to the axis frequency estimation value opt =KΔα; (5-3) The modulation frequency range [α opt -5Δα,α opt +5Δα] is determined as the alternative range of modulation base frequency, and Δα is the modulation frequency resolution; Select the modulation frequency corresponding to the maximum value of the enhanced envelope spectrum amplitude, and determine it as the modulation base frequency α b ; (6) performing square slicing processing on the spectral coherence function at the modulation base frequency to obtain an adaptive weighting function; (7) The spectral correlation function is weighted by using an adaptive weighting function to obtain a weighted spectral correlation function. The calculation process is: (7-1) Expand the spectral correlation function into a matrix form; (7-2) copying and extending the adaptive weighting function along the modulation frequency axis to form a weighting matrix; (7-3) Calculating weighted spectral correlation function; (8) Based on the weighted spectrum correlation function, the weighted enhanced envelope spectrum is calculated. The specific process is as follows: (8-1) Calculate the weighted spectral correlation function after modulo; (8-2) The modulated weighted spectral correlation function is accumulated along the carrier frequency axis and averaged to obtain the weighted enhanced envelope spectrum.

2. The method for extracting propulsion pump modulation frequency based on weighted enhanced envelope spectrum according to claim 1, characterized in that: In step (1), the specific steps to obtain the centralized signal are: (1-1) Calculate the average value of the monitoring signal; (1-2) Subtract the average value from the monitoring signal to obtain the centralized signal.

3. The method for extracting propulsion pump modulation frequency based on weighted enhanced envelope spectrum according to claim 1, characterized in that: The specific steps of step (2) are: (2-1) According to the structural characteristics and operating parameters of the propulsion pump, the expression of the first-order cyclostationary signal is preset; (2-2) The time domain synchronous averaging method is used to estimate the first-order cyclostationary signal component in the centralized signal; (2-3) Subtract the estimated first-order cyclostationary signal from the centered signal to obtain the denoised signal.

4. The method for extracting propulsion pump modulation frequency based on weighted enhanced envelope spectrum according to claim 1, characterized in that: The specific process of step (3) is as follows: (3-1) Calculate the time-varying autocorrelation function R of the noise reduction signal x(t) x (t,τ), the calculation formula is as follows: In the formula, t represents time, τ represents delay, * is a conjugate symbol, E[·] represents the statistical mean; (3-2) Calculate the cyclic autocorrelation function R of the noise reduction signal x(t) x (α,τ), the calculation formula is as follows: Where -T / 2 and T / 2 are the lower and upper limits of time, e is a natural constant, j is the symbol of an imaginary number, and α is the modulation frequency; (3-3) Calculate the spectral correlation function S of the noise reduction signal x(t) x (α,f), the calculation formula is as follows: Where f is the carrier frequency.

5. The method for extracting propulsion pump modulation frequency based on weighted enhanced envelope spectrum according to claim 1, characterized in that: The specific process of step (4) is as follows: (4-1) Calculate the enhanced envelope spectrum ESC of the noise reduction signal x (α), the calculation formula is as follows: Where f1 and f2 are the lower and upper limits of the carrier frequency integral, and their default values ​​are 0 and F s / 2,F s is the sampling frequency of the monitoring signal; (4-2) Calculate the spectral coherence function γ of the noise reduction signal x (α,f), the calculation formula is as follows: In the formula, S x (α,f) represents the spectral correlation function, P x (f) represents the power spectral density of the noise reduction signal x(t), which is equal to the result when the modulation frequency of the spectrum correlation is zero, that is, S x (0,f)=P x (f).

6. The method for extracting propulsion pump modulation frequency based on weighted enhanced envelope spectrum according to claim 1, characterized in that: The specific process of step (6) is as follows: (6-1) Calculate the overall weight function w x (α,f), the calculation formula is as follows: w x (a,f)=|γ x (a,f)| 2 (6-2) Let α = modulation base frequency α b , then we can obtain the adaptive weighting function w x (f) = w x (α b ,f).

7. The method for extracting propulsion pump modulation frequency based on weighted enhanced envelope spectrum according to claim 1, characterized in that: In step (7-1), the spectral correlation function S x (α,f) is expanded into the matrix form S, and the formula is as follows: Where Δα represents the modulation frequency resolution, NΔα represents the modulation frequency upper limit, Δf represents the carrier frequency resolution, MΔf represents the carrier frequency upper limit, and the spectral correlation matrix S∈C (M+1)×(N+1) ; In step (7-2), the adaptive weighting function is copied and extended along the modulation frequency axis to form a weighting matrix W, which is as follows: In the formula, the weight matrix W∈C (M+1)×(N+1) ; In step (7-3), the weighted spectral correlation function WSC(α,f) is calculated, and the formula is as follows: Where Z∈C (M+1)×(N+1) is the matrix form of the weighted spectral correlation function.

8. The method for extracting propulsion pump modulation frequency based on weighted enhanced envelope spectrum according to claim 1, characterized in that: In step (8-1), the weighted spectral correlation function |WSC(α,f)| after modulo calculation is as follows: In step (8-2), the modulized weighted spectrum correlation function is accumulated along the carrier frequency axis and averaged to obtain the weighted enhanced envelope spectrum WESC(α), which is as follows: Where E∈R 1×(N+1) is the matrix form of the weighted enhanced envelope spectrum.

Images