A method and device for extracting time-frequency features of an engine vibration signal

Abstract

The application provides a time-frequency feature extraction method and device for an engine vibration signal. The method comprises the following steps: dividing the vibration acceleration signal of the engine into a noise sub-signal and a vibration sub-signal, filtering the noise sub-signal by using a non-local mean method, and recombining the filtered signal and the vibration sub-signal in the time domain, processing the recombined signal by using a TVF-EMD method, further suppressing the noise signal by constructing a plurality of different filters, and obtaining a vibration acceleration signal composed of a plurality of intrinsic mode function sub-signals. The application removes the noise interference information of the vibration signal from two directions of time and frequency domains, strengthens the denoising effect, and realizes accurate extraction of vibration characteristic information.

Description

Technical Field

[0001] This invention belongs to the field of vibration signal processing technology, specifically relating to a method and apparatus for extracting time-frequency features of engine vibration signals. Background Technology

[0002] As the core of a vehicle's power unit, the engine's vibration and noise issues severely restrict the development of power systems towards higher power density. To accurately pinpoint the vibration sources, in-depth analysis of the measured vibration signal characteristics is necessary. However, the engine's operating environment is extremely harsh, and the measured vibration signals may contain various vibration excitation sources or noise sources unrelated to the operating process, making it difficult to clearly identify the characteristics of the vibration source of interest. Therefore, engine vibration signal denoising technology is receiving increasing attention.

[0003] Currently, commonly used signal denoising methods mainly include Empirical Mode Decomposition (EMD), Variational Mode Decomposition (VMD), and Empirical Wavelet Transform (EWT). However, these methods have the following problems when processing engine vibration signals: First, engine vibration signals are non-steady-state time-varying signals, and the above-mentioned signal denoising methods are generally not suitable for processing time-varying signals; second, the above methods only remove interference noise signals in the frequency domain, without processing noise in the time domain. Therefore, the denoising effects of the above methods are not ideal. Summary of the Invention

[0004] To address the aforementioned problems in the prior art, this invention provides a method and apparatus for extracting time-frequency features of engine vibration signals.

[0005] To achieve the above objectives, the present invention adopts the following technical solution.

[0006] In a first aspect, the present invention provides a method for extracting time-frequency features of engine vibration signals, comprising the following steps:

[0007] The vibration acceleration signal of the engine is acquired in real time, the instantaneous energy signal of the vibration acceleration signal is calculated based on the envelope derivative operator, and the short-time energy signal is obtained by integration calculation;

[0008] Based on the energy difference between noise signals and vibration signals, the short-time energy signal is divided into noise energy sub-signals and vibration energy sub-signals, and the vibration acceleration signal is divided into noise sub-signals and vibration sub-signals by time alignment;

[0009] The noise sub-signal is filtered using the nonlocal mean method, and the filtered signal is reconstructed in the time domain with the vibration sub-signal.

[0010] The reconstructed signal is processed by TVF-EMD, and noise signals are further suppressed by constructing multiple different filters to obtain a vibration acceleration signal composed of multiple intrinsic mode function sub-signals.

[0011] Furthermore, the method for calculating the instantaneous energy signal of the vibration acceleration signal includes:

[0012] The envelope of the vibration acceleration signal is represented as:

[0013] S[x(t)]=|x(t)+jH[x(t)]| 2 (1)

[0014] In the formula, x(t) is the vibration acceleration signal at the t-th data acquisition time, S[x(t)] is the envelope signal of x(t), and H[*] represents the Hilbert transform;

[0015] The instantaneous energy signal of x(t) is obtained by calculating the envelope derivative signal based on frequency weighting:

[0016] Γ[x(t)]=|diff(x(t))+jH[diff(x(t))]| 2 =[diff(x(t))] 2 +{H[diff(x(t))]} 2 (2)

[0017] In the formula, Γ[x(t)] is the instantaneous energy signal of x(t), and diff(x(t)) is the derivative of x(t) in frequency-weighted form. The expression for diff(x(t)) is:

[0018] diff(x(t))=[x(t+1)-x(t-1)] / 2 (3)

[0019] Substituting equation (3) into equation (2), we get:

[0020]

[0021] In the formula, h(t) = H(x(t)).

[0022] Furthermore, the method for calculating the short-time energy signal of the vibration acceleration signal includes:

[0023] Divide Γ[x(t)] into N data segments of length L according to the time sequence, with p data points overlapping between two adjacent data segments;

[0024] The short-time energy signal for each data segment is obtained by summing the data within each segment, using the following formula:

[0025]

[0026] In the formula, STE k Let be the short-time energy signal in the k-th data segment, where k = 1, 2, ..., N.

[0027] Furthermore, the method for dividing the short-time energy signal into noise energy sub-signals and vibration energy sub-signals includes:

[0028] If STE k If it is less than the set threshold, then STE k For noise energy sub-signals; otherwise, STE k Let be the vibrational energy sub-signal, where k = 1, 2, ..., N.

[0029] Furthermore, the threshold is:

[0030]

[0031] In the formula, LT is the threshold.

[0032] Furthermore, the method for filtering the noise sub-signal using the nonlocal mean method includes:

[0033] The weighting coefficients of the noise sub-signal x1(i) are calculated using the following formula:

[0034]

[0035] In the formula, x1(i) is the noise sub-signal at the i-th data acquisition time, ω(i,j) is the weighting coefficient, i.e. the similarity of similar blocks centered at i and j, δ is the Gaussian scaling coefficient, P is the influence coefficient related to the number of similar blocks, λ is the search step size, and Δ is the search block centered at i.

[0036] Calculate the filtered signal x after filtering the noise sub-signal x1(i) using the following formula. 1S (i):

[0037]

[0038]

[0039] In the formula, Ω i This is a search window centered on 'i'.

[0040] Furthermore, δ = 8.

[0041] Furthermore, the TVF-EMD processing of the reconstructed signal also includes adaptive adjustment of the Gaussian scaling coefficients of multiple filters, as follows:

[0042] Calculate x2(t) and IMF i The difference entropy F(x2(t)) and F(IMF) of x2(t) i x2(t) is the recombined signal, and IMF is the signal after recombination. i (t) represents the i-th intrinsic mode function sub-signal obtained after TVF-EMD processing, where i = 1, 2, ..., K, and K is the number of intrinsic mode function sub-signals. The formula for calculating the differential entropy is:

[0043]

[0044]

[0045] In the formula, z(t) is x2(t) or IMF. i (t), M is x2(t) or IMF i The data length of (t);

[0046] The adaptive Gaussian scaling coefficients are calculated using the following formula:

[0047] δ i =0.25*exp(F(x2(t)) / F(IMF) i (t))) (12)

[0048] In the formula, δ i For adaptive Gaussian scaling coefficients, i = 1, 2, ..., K.

[0049] In a second aspect, the present invention provides a time-frequency feature extraction device for engine vibration signals, comprising:

[0050] The energy calculation module is used to acquire the vibration acceleration signal of the engine measured in real time, calculate the instantaneous energy signal of the vibration acceleration signal based on the envelope derivative operator, and obtain the short-time energy signal through integration calculation;

[0051] The signal segmentation module is used to divide the short-time energy signal into noise energy sub-signals and vibration energy sub-signals based on the energy difference between the noise signal and the vibration signal, and to divide the vibration acceleration signal into noise sub-signals and vibration sub-signals through time alignment;

[0052] The first filtering module is used to filter the noise sub-signal using the nonlocal mean method, and to reconstruct the filtered signal with the vibration sub-signal in the time domain.

[0053] The second filtering module is used to perform TVF-EMD processing on the reconstructed signal. By constructing multiple different filters, the noise signal is further suppressed to obtain a vibration acceleration signal composed of multiple intrinsic mode function sub-signals.

[0054] Furthermore, the method for calculating the instantaneous energy signal of the vibration acceleration signal includes:

[0055] The envelope of the vibration acceleration signal is represented as:

[0056] S[x(t)]=|x(t)+jH[x(t)]| 2 (1)

[0057] In the formula, x(t) is the vibration acceleration signal at the t-th data acquisition time, S[x(t)] is the envelope signal of x(t), and H[*] represents the Hilbert transform;

[0058] The instantaneous energy signal of x(t) is obtained by calculating the envelope derivative signal based on frequency weighting:

[0059] Γ[x(t)]=|diff(x(t))+jH[diff(x(t))]| 2 =[diff(x(t))] 2 +{H[diff(x(t))]} 2 (2)

[0060] In the formula, Γ[x(t)] is the instantaneous energy signal of x(t), and diff(x(t)) is the derivative of x(t) in frequency-weighted form. The expression for diff(x(t)) is:

[0061] diff(x(t))=[x(t+1)-x(t-1)] / 2 (3)

[0062] Substituting equation (3) into equation (2), we get:

[0063]

[0064] In the formula, h(t) = H(x(t)).

[0065] Compared with the prior art, the present invention has the following beneficial effects.

[0066] This invention divides the engine vibration acceleration signal into a noise sub-signal and a vibration sub-signal. It uses a nonlocal mean method to filter the noise sub-signal, then reconstructs the filtered signal with the vibration sub-signal in the time domain. The reconstructed signal is then processed by TVF-EMD, and multiple different filters are constructed to further suppress the noise signal, resulting in a vibration acceleration signal composed of multiple intrinsic mode function sub-signals. This achieves effective noise suppression. Furthermore, by stripping noise interference information from the vibration signal in both the time and frequency domains, this invention enhances the denoising effect and achieves accurate extraction of vibration characteristic information. Attached Figure Description

[0067] Figure 1 This is a flowchart of a method for extracting time-frequency features of engine vibration signals according to an embodiment of the present invention.

[0068] Figure 2 This is an overall flowchart of an embodiment of the present invention.

[0069] Figure 3 This is a schematic diagram of the engine vibration acceleration signal waveform.

[0070] Figure 4 This is a schematic diagram of the instantaneous vibration energy waveform.

[0071] Figure 5 This is a schematic diagram of short-time vibration energy waveform.

[0072] Figure 6 This is a schematic diagram of the waveforms of the vibration sub-signal and the noise sub-signal.

[0073] Figure 7 This is a schematic diagram of the waveform of the TVF-EMD processing result. The vertical axis is in mm / s. 2 .

[0074] Figure 8 This is a block diagram of a time-frequency feature extraction device for engine vibration signals according to an embodiment of the present invention. Detailed Implementation

[0075] To make the objectives, technical solutions, and advantages of this invention clearer and more understandable, the invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0076] Figure 1 This is a flowchart of a method for extracting time-frequency features of engine vibration signals according to an embodiment of the present invention, including the following steps:

[0077] Step 101: Obtain the vibration acceleration signal of the engine measured in real time, calculate the instantaneous energy signal of the vibration acceleration signal based on the envelope derivative operator, and obtain the short-time energy signal through integration calculation;

[0078] Step 102: Based on the energy difference between the noise signal and the vibration signal, the short-time energy signal is divided into a noise energy sub-signal and a vibration energy sub-signal, and the vibration acceleration signal is divided into a noise sub-signal and a vibration sub-signal by time alignment;

[0079] Step 103: The noise sub-signal is filtered using the nonlocal mean method, and the filtered signal is reconstructed in the time domain with the vibration sub-signal;

[0080] Step 104: Perform TVF-EMD processing on the recombined signal, and further suppress the noise signal by constructing multiple different filters to obtain a vibration acceleration signal composed of multiple intrinsic mode function sub-signals.

[0081] In this embodiment, step 101 is mainly used to calculate the instantaneous energy signal and short-time energy signal of the vibration signal. The vibration signal specifically refers to the vibration acceleration signal of the engine obtained through experiments, or simply the vibration signal. The waveform of the vibration acceleration signal is shown below. Figure 3 As shown. The engine vibration signal measured on-site includes not only the signal generated by vibration during engine operation but also noise signal. The noise component has relatively low energy, while the actual vibration signal component has higher energy; furthermore, their frequencies differ significantly. To accurately distinguish between these two signal components, this embodiment uses the envelope derivative operator to process the engine vibration signal to obtain the instantaneous energy signal. The advantage of this method is that the envelope derivative operator not only uses Hilbert transform to highlight the differences in amplitude between different signal components but also utilizes frequency weighting to statistically analyze the differences in signal energy. Frequency weighting refers to assigning greater weight to high-frequency components and less weight to low-frequency components. This embodiment obtains the short-time energy signal by performing short-time integration (integration over a very short time interval). A waveform diagram of the instantaneous vibration energy is shown below. Figure 4 As shown in the diagram, the waveform of short-time vibrational energy is as follows: Figure 5 As shown.

[0082] In this embodiment, step 102 is mainly used to divide the vibration acceleration signal into a noise sub-signal and a vibration sub-signal. Based on the energy difference between the noise signal and the vibration signal—that is, the energy of the actual vibration signal is significantly higher than the energy of the noise signal—this embodiment divides the obtained short-time energy signal into a noise energy sub-signal and a vibration energy sub-signal. Then, by aligning the vibration acceleration signal with the short-time energy signal in time, the vibration acceleration signal is further divided into a noise sub-signal and a vibration sub-signal. Specifically, vibration acceleration signal data points with the same time as the vibration energy sub-signal data points are designated as vibration sub-signal data points, and vice versa. The waveform diagrams of the vibration sub-signal and the noise sub-signal are shown below. Figure 6 As shown. It is worth noting that although the vibration acceleration signal is divided into a noise sub-signal and a vibration sub-signal in the time domain, the resulting noise sub-signal is not 100% noise; it still contains some genuine vibration signal. Similarly, the resulting vibration sub-signal also contains some noise. Therefore, subsequent steps require filtering the noise sub-signal and further processing the reconstructed signal.

[0083] In this embodiment, step 103 is mainly used for filtering and reconstructing the noise sub-signal. This embodiment uses the Non-Local Mean (NLM) method to filter the noise sub-signal in the time domain. The NLM method can effectively reduce noise while reducing signal distortion. To significantly suppress noise components, the Gaussian scaling coefficient in the NLM can be set to a large value. After NLM denoising is completed, the denoised noise sub-signal and the vibration sub-signal are reconstructed in the time domain.

[0084] In this embodiment, step 104 is mainly used to perform TVF-EMD (Time Varying Filtering-Empirical Mode Decomposition) processing on the reconstructed signal. Although the reconstructed signal has undergone noise suppression in the time domain, noise interference still exists in the frequency domain. Therefore, TVF-EMD is used to process the time-domain reconstructed signal. TVF-EMD filters the input signal by constructing multiple time-varying filters to further suppress noise signals. Each filter outputs a sub-signal with different characteristics, also known as an intrinsic mode function sub-signal. The waveform diagram of the signal output after TVF-EMD processing is shown below. Figure 7 As shown.

[0085] As an optional embodiment, the method for calculating the instantaneous energy signal of the vibration acceleration signal includes:

[0086] The envelope of the vibration acceleration signal is represented as:

[0087] S[x(t)]=|x(t)+jH[x(t)]| 2 (1)

[0088] In the formula, x(t) is the vibration acceleration signal at the t-th data acquisition time, S[x(t)] is the envelope signal of x(t), and H[*] represents the Hilbert transform;

[0089] The instantaneous energy signal of x(t) is obtained by calculating the envelope derivative signal based on frequency weighting:

[0090] Γ[x(t)]=|diff(x(t))+jH[diff(x(t))]| 2 =[diff(x(t))] 2 +{H[diff(x(t))]} 2 (2)

[0091] In the formula, Γ[x(t)] is the instantaneous energy signal of x(t), and diff(x(t)) is the derivative of x(t) in frequency-weighted form. The expression for diff(x(t)) is:

[0092] diff(x(t))=[x(t+1)-x(t-1)] / 2 (3)

[0093] Substituting equation (3) into equation (2), we get:

[0094]

[0095] In the formula, h(t) = H(x(t)).

[0096] This embodiment provides a technical solution for calculating the instantaneous energy signal of a vibration acceleration signal. This embodiment obtains the instantaneous energy signal by processing the vibration acceleration signal using the envelope derivative operator. First, the envelope signal of the vibration acceleration signal x(t) is calculated, as shown in equation (1). Then, the envelope derivative signal is calculated based on frequency weighting to obtain the instantaneous energy signal of x(t), as shown in equation (2). The derivative diff(x(t)) of x(t) in equation (2) using frequency weighting is represented by x(t+1) and x(t-1), as shown in equation (3), and finally, the instantaneous energy signal shown in equation (4) is obtained.

[0097] As an optional embodiment, the method for calculating the short-time energy signal of the vibration acceleration signal includes:

[0098] Divide Γ[x(t)] into N data segments of length L according to the time sequence, with p data points overlapping between two adjacent data segments;

[0099] The short-time energy signal for each data segment is obtained by summing the data within each segment, using the following formula:

[0100]

[0101] In the formula, STE k Let be the short-time energy signal in the k-th data segment, where k = 1, 2, ..., N.

[0102] This embodiment provides a technical solution for calculating the short-time energy signal of a vibration acceleration signal. This embodiment obtains the short-time energy signal by performing short-time integration on the instantaneous energy signal. The specific method is as follows: First, the instantaneous energy signal is divided into N data segments in chronological order, with N data points in each data segment and p data points overlapping between adjacent data segments; then, the sum of all data in each data segment is calculated to obtain the short-time energy in each data segment. Equation (5) is a general expression for calculating the short-time energy in any data segment. For example, when calculating the short-time energy STE1 in the first data segment, k=1 is substituted into equation (5), and STE1 is equal to the sum of Γ[x(1)]~Γ[x(L)]; the short-time energy STE2 in the second data segment is equal to the sum of Γ[x(L-p+1)]~Γ[x(2L-p)].

[0103] As an optional embodiment, the method for dividing the short-time energy signal into noise energy sub-signals and vibration energy sub-signals includes:

[0104] If STE k If it is less than the set threshold, then STE k For noise energy sub-signals; otherwise, STE k Let be the vibrational energy sub-signal, where k = 1, 2, ..., N.

[0105] This embodiment provides a technical solution for dividing short-time energy signals into noise energy sub-signals and vibration energy sub-signals. Based on the characteristic that vibration signal energy is significantly higher than noise signal energy, this embodiment sets a judgment threshold and uses the STE obtained in the previous embodiment... k Compared with the threshold, if STE k If the value is less than the set threshold, it is considered STE. k If it belongs to the noise energy sub-signal; otherwise, it is considered STE. k It belongs to the category of vibrational energy quantum signals.

[0106] As an optional embodiment, the threshold is:

[0107]

[0108] In the formula, LT is the threshold.

[0109] This embodiment provides a technical solution for determining the threshold value set in the previous embodiment. The threshold value can be set to a fixed value based on experience. However, since the vibration signal energy varies greatly under different operating conditions of different engines, setting a uniform threshold value will bring a large error. Therefore, the threshold value set in this embodiment adopts a dynamic threshold value, that is, the value of the threshold value is proportional to the average value of the short-time energy signal value. The specific calculation formula is shown in equation (6).

[0110] As an optional embodiment, the method for filtering the noise sub-signal using the non-local mean method includes:

[0111] The weighting coefficients of the noise sub-signal x1(i) are calculated using the following formula:

[0112]

[0113] In the formula, x1(i) is the noise sub-signal at the i-th data acquisition time, ω(i,j) is the weighting coefficient, i.e. the similarity of similar blocks centered at i and j, δ is the Gaussian scaling coefficient, P is the influence coefficient related to the number of similar blocks, λ is the search step size, and Δ is the search block centered at i.

[0114] Calculate the filtered signal x after filtering the noise sub-signal x1(i) using the following formula. 1S (i):

[0115]

[0116]

[0117] In the formula, Ω i This is a search window centered on 'i'.

[0118] This embodiment presents a technical solution for filtering noisy sub-signals. This embodiment calculates Ω for each search window. i The weighted average of the internal noise sub-signal x1(i) is used to filter the noise sub-signal, and the calculation formulas are shown in equations (8) and (9). Search window Ω i The wider the width, the better the denoising effect, but the longer the computation time. ω(i,j) is the weighting coefficient, and the calculation formula is shown in equation (7). The calculation process of ω(i,j) can be understood as: comparing the neighborhoods around similar blocks centered on i and j, the higher the similarity, the larger ω(i,j). δ is the Gaussian scaling coefficient, which affects the smoothness of the denoised signal. P is the coefficient variable of ω(i,j), which affects the number of similar blocks. λ represents the search step size, that is, the step size of the search window movement. Δ is the search block centered on i. Ω iBoth δ and P will affect the results of NLM, but it can be clearly seen from equation (7) that the Gaussian scaling coefficient δ has a greater influence than the other two parameters. To obtain a good filtering effect, δ should be taken as a larger value.

[0119] As an optional embodiment, δ = 8.

[0120] This embodiment limits the Gaussian scaling coefficient δ in NLM processing. As mentioned above, to obtain a good filtering effect, δ should be a larger value; in this embodiment, δ is set to 8. It is worth noting that this embodiment only provides a preferred implementation method and does not deny or exclude other feasible implementation methods, such as other settings different from δ=8.

[0121] As an optional embodiment, the TVF-EMD processing of the reconstructed signal further includes adaptive adjustment of the Gaussian scaling coefficients of multiple filters, as follows:

[0122] Calculate x2(t) and IMF i The difference entropy F(x2(t)) and F(IMF) of x2(t) i x2(t) is the recombined signal, and IMF is the signal after recombination. i (t) represents the i-th intrinsic mode function sub-signal obtained after TVF-EMD processing, where i = 1, 2, ..., K, and K is the number of intrinsic mode function sub-signals. The formula for calculating the differential entropy is:

[0123]

[0124]

[0125] In the formula, z(t) is x2(t) or IMF. i (t), M is x2(t) or IMF i The data length of (t);

[0126] The adaptive Gaussian scaling coefficients are calculated using the following formula:

[0127] δ i =0.25*exp(F(x2(t)) / F(IMF) i (t))) (12)

[0128] In the formula, δ i For adaptive Gaussian scaling coefficients, i = 1, 2, ..., K.

[0129] This embodiment presents a technical solution for adaptively adjusting the Gaussian scaling coefficients of multiple filters in TVF-EMD processing. The intrinsic mode function (IMF) sub-signals obtained from TVF-EMD processing contain three signal components: a noise sub-signal, a mixed signal of noise and vibration sub-signals, and a vibration sub-signal. To effectively remove the noise component, different Gaussian scaling coefficients need to be assigned to these three signal components: a smaller Gaussian scaling coefficient is assigned to the vibration sub-signal to prevent signal distortion caused by the NLM denoising process; while a larger Gaussian scaling coefficient is assigned to the noise sub-signal to achieve a higher level of denoising. Therefore, this embodiment proposes an adaptive NLM Gaussian scaling coefficient calculation formula based on differential entropy. That is, by calculating the differential entropy of each IMF sub-signal obtained from TVF-EMD processing, the Gaussian scaling coefficient δ corresponding to each IMF sub-signal is obtained. i The input is then fed into the corresponding filter to achieve adaptive denoising. The calculation formulas for differential entropy and adaptive Gaussian scaling coefficients are given in equations (10) to (12), and will not be explained in detail here.

[0130] Figure 8 This is a schematic diagram of the composition of a time-frequency feature extraction device for engine vibration signals according to an embodiment of the present invention. The device includes:

[0131] The energy calculation module 11 is used to acquire the vibration acceleration signal of the engine measured in real time, calculate the instantaneous energy signal of the vibration acceleration signal based on the envelope derivative operator, and obtain the short-time energy signal through integration calculation.

[0132] The signal segmentation module 12 is used to divide the short-time energy signal into noise energy sub-signals and vibration energy sub-signals based on the energy difference between the noise signal and the vibration signal, and to divide the vibration acceleration signal into noise sub-signals and vibration sub-signals through time alignment;

[0133] The first filtering module 13 is used to filter the noise sub-signal using the nonlocal mean method, and to reconstruct the filtered signal with the vibration sub-signal in the time domain.

[0134] The second filtering module 14 is used to perform TVF-EMD processing on the recombined signal. By constructing multiple different filters, the noise signal is further suppressed to obtain a vibration acceleration signal composed of multiple intrinsic mode function sub-signals.

[0135] The apparatus of this embodiment can be used to perform Figure 1 The technical solutions of the illustrated method embodiments are similar in principle and technical effect, and will not be described again here. The same applies to the subsequent embodiments, which will not be elaborated upon further.

[0136] As an optional embodiment, the method for calculating the instantaneous energy signal of the vibration acceleration signal includes:

[0137] The envelope of the vibration acceleration signal is represented as:

[0138] S[x(t)]=|x(t)+jH[x(t)]| 2 (1)

[0139] In the formula, x(t) is the vibration acceleration signal at the t-th data acquisition time, S[x(t)] is the envelope signal of x(t), and H[*] represents the Hilbert transform;

[0140] The instantaneous energy signal of x(t) is obtained by calculating the envelope derivative signal based on frequency weighting:

[0141] Γ[x(t)]=|diff(x(t))+jH[diff(x(t))]| 2 =[diff(x(t))] 2 +{H[diff(x(t))]} 2 (2)

[0142] In the formula, Γ[x(t)] is the instantaneous energy signal of x(t), and diff(x(t)) is the derivative of x(t) in frequency-weighted form. The expression for diff(x(t)) is:

[0143] diff(x(t))=[x(t+1)-x(t-1)] / 2 (3)

[0144] Substituting equation (3) into equation (2), we get:

[0145]

[0146] In the formula, h(t) = H(x(t)).

[0147] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for extracting time-frequency features of engine vibration signals, characterized in that, Includes the following steps: The vibration acceleration signal of the engine is acquired in real time, the instantaneous energy signal of the vibration acceleration signal is calculated based on the envelope derivative operator, and the short-time energy signal is obtained by integration calculation; Based on the energy difference between noise signals and vibration signals, the short-time energy signal is divided into noise energy sub-signals and vibration energy sub-signals, and the vibration acceleration signal is divided into noise sub-signals and vibration sub-signals by time alignment; The noise sub-signal is filtered using the nonlocal mean method, and the filtered signal is reconstructed in the time domain with the vibration sub-signal. The reconstructed signal is processed by TVF-EMD, and noise signals are further suppressed by constructing multiple different filters to obtain a vibration acceleration signal composed of multiple intrinsic mode function sub-signals. The method for calculating the instantaneous energy signal of the vibration acceleration signal includes: The envelope of the vibration acceleration signal is represented as: (1) In the formula, x(t) is the vibration acceleration signal at the t-th data acquisition time. Let H be the envelope signal of x(t). ] represents the Hilbert transform; The instantaneous energy signal of x(t) is obtained by calculating the envelope derivative signal based on frequency weighting: (2) In the formula, Let x(t) be the instantaneous energy signal. Let x(t) be the derivative in frequency-weighted form. The expression is: (3) Substituting equation (3) into equation (2), we get: (4) In the formula, .

2. The method for extracting time-frequency features of engine vibration signals according to claim 1, characterized in that, The method for calculating the short-time energy signal of the vibration acceleration signal includes: Will The data is divided into N segments of length L according to the time sequence, and there are p data overlaps between two adjacent data segments. The short-time energy signal for each data segment is obtained by summing the data within each segment, using the following formula: (5) In the formula, STE k Let k be the short-time energy signal in the k-th data segment, where k = 1, 2, ..., N.

3. The method for extracting time-frequency features of engine vibration signals according to claim 2, characterized in that, The method for dividing the short-time energy signal into noise energy sub-signals and vibration energy sub-signals includes: If STE k If it is less than the set threshold, then STE k For noise energy sub-signals; otherwise, STE k Let be the vibrational energy sub-signal, where k = 1, 2, ..., N.

4. The method for extracting time-frequency features of engine vibration signals according to claim 3, characterized in that, The threshold is: (6) In the formula, LT is the threshold.

5. The method for extracting time-frequency features of engine vibration signals according to claim 1, characterized in that, The method for filtering the noise sub-signal using the nonlocal mean method includes: The weighting coefficients of the noise sub-signal x1(i) are calculated using the following formula: (7) In the formula, x1(i) is the noise sub-signal at the i-th data acquisition time, ω(i, j) is the weighting coefficient, i.e. the similarity of similar blocks centered at i and j, δ is the Gaussian scaling coefficient, P is the influence coefficient related to the number of similar blocks, λ is the search step size, and Δ is the search block centered at i. Calculate the filtered signal x after filtering the noise sub-signal x1(i) using the following formula. 1S (i): (8) (9) In the formula, Ω i This is a search window centered on 'i'.

6. The method for extracting time-frequency features of engine vibration signals according to claim 5, characterized in that, δ=8.

7. The method for extracting time-frequency features of engine vibration signals according to claim 1, characterized in that, The TVF-EMD processing of the reconstructed signal also includes adaptive adjustment of the Gaussian scaling coefficients of multiple filters, as follows: Calculate x2(t) and IMF i The difference entropy F(x2(t)) and F(IMF) of x2(t) i x2(t) is the recombined signal, and IMF is the signal after recombination. i (t) represents the i-th intrinsic mode function sub-signal obtained after TVF-EMD processing, i=1,2,...,K, where K is the number of intrinsic mode function sub-signals. The formula for calculating the differential entropy is: (10) (11) In the formula, z(t) is x2(t) or IMF. i (t), M is x2(t) or IMF i The data length of (t); The adaptive Gaussian scaling coefficients are calculated using the following formula: (12) In the formula, δ i For adaptive Gaussian scaling coefficients, i=1,2,...,K.

8. A device for extracting the time-frequency features of an engine vibration signal, characterized in that, include: The energy calculation module is used to acquire the vibration acceleration signal of the engine measured in real time, calculate the instantaneous energy signal of the vibration acceleration signal based on the envelope derivative operator, and obtain the short-time energy signal through integration calculation; The signal segmentation module is used to divide the short-time energy signal into noise energy sub-signals and vibration energy sub-signals based on the energy difference between the noise signal and the vibration signal, and to divide the vibration acceleration signal into noise sub-signals and vibration sub-signals through time alignment; The first filtering module is used to filter the noise sub-signal using the nonlocal mean method, and to reconstruct the filtered signal with the vibration sub-signal in the time domain. The second filtering module is used to perform TVF-EMD processing on the reconstructed signal. By constructing multiple different filters, the noise signal is further suppressed to obtain a vibration acceleration signal composed of multiple intrinsic mode function sub-signals. The method for calculating the instantaneous energy signal of the vibration acceleration signal includes: The envelope of the vibration acceleration signal is represented as: (1) In the formula, x(t) is the vibration acceleration signal at the t-th data acquisition time. Let H be the envelope signal of x(t). ] represents the Hilbert transform; The instantaneous energy signal of x(t) is obtained by calculating the envelope derivative signal based on frequency weighting: (2) In the formula, Let x(t) be the instantaneous energy signal. Let x(t) be the derivative in frequency-weighted form. The expression is: (3) Substituting equation (3) into equation (2), we get: (4) In the formula, .

Images