Vibration signal denoising and feature extraction method based on wavelet adaptive threshold

Abstract

The invention discloses a vibration signal denoising and feature extraction method based on a wavelet adaptive threshold, and relates to the technical field of vibration sensor signal processing. The method comprises the following steps of: 1, performing multi-layer decomposition on a signal based on a filter bank formed by discrete wavelet transform; step 2, adaptively calculating a noise threshold; step 3, realizing wavelet coefficient shrinkage denoising based on a SureShrink algorithm; 4, reconstructing a vibration signal based on quasi-discrete wavelet transform; and step 5, automatically extracting signal features based on the multilayer recurrent neural network. According to the invention, through the discrete wavelet transform-based signal decomposition and adaptive threshold denoising algorithm, the noise of the vibration sensor and the error influence caused by environmental disturbance are reduced, and most of the characteristics of the original vibration signal are retained; and a multi-layer recurrent neural network is established for the denoised vibration signals, so that automatic extraction of time features is realized, and an important basis is provided for subsequent fault diagnosis and construction of a health monitoring algorithm.

Description

Technical Field

[0001] The invention relates to the technical field of vibration sensor signal processing, in particular to a vibration signal denoising and feature extraction method based on wavelet adaptive threshold. Background Art

[0002] With the continuous development of the field of structural health monitoring and fault diagnosis, the importance of vibration sensors as core monitoring tools has become increasingly prominent. In a laboratory environment, vibration sensors combined with digital filters, signal amplifiers and other equipment can achieve high-precision, high-frequency measurement of structural vibration response, which can then be used for load inversion and health status assessment of the structure. However, the growing demand for structural fault diagnosis and health monitoring of aerospace equipment in recent years has posed new challenges to the application of vibration sensors. Unlike laboratory environments, vibration sensors need to monitor vibration while the equipment is in operation, and the working environment of the equipment may also change; therefore, traditional vibration signal preprocessing methods cannot accurately extract the vibration information of the structure.

[0003] With the vigorous development of artificial intelligence technology in recent years, deep neural networks can realize the automatic extraction of complex data features. The automatic extraction of vibration signal features based on deep neural network technology can not only greatly facilitate the construction and deployment of algorithms such as fault diagnosis, but also lay a solid foundation for efficient and accurate structural health detection. Summary of the invention

[0004] The purpose of the present invention is to propose a vibration signal denoising and feature extraction method based on wavelet adaptive threshold, which reduces the error influence caused by the vibration sensor's own noise and environmental disturbance, realizes the accurate extraction of structural vibration information and automatic extraction of signal features in complex environments, and provides a basis for the deployment and optimization of structural health monitoring in complex environments.

[0005] To achieve the above purpose, the present invention proposes a vibration signal denoising and feature extraction method based on wavelet adaptive threshold, and the specific steps are as follows:

[0006] S1. Decomposing the collected vibration signal into high and low frequency signals through a filter bank based on discrete wavelet transform and finally realizing multi-layer decomposition of the vibration signal;

[0007] S2, adaptive calculation of noise threshold;

[0008] S3, wavelet coefficient shrinkage denoising based on SureShrink algorithm, adaptive threshold determination of sub-waveband coefficients of different vibration intensities and frequency ranges;

[0009] S4, constructing a multi-layer filter bank based on inverse discrete wavelet transform to achieve vibration signal reconstruction;

[0010] S5. Automatic extraction of signal features based on multi-layer recurrent neural network.

[0011] Preferably, in S1, the vibration signal collected by the vibration sensor is subjected to a filter group composed of discrete wavelet transform to achieve decomposition of high-frequency and low-frequency signals; wherein the filter group is composed of a high-pass filter and a low-pass filter, and the vibration signal is subjected to the operation of the low-pass filter to obtain the low-frequency signal wavelet coefficient c(n), and the calculation formula is as follows:

[0012]

[0013] The vibration signal is processed by a high-pass filter to obtain the high-frequency signal wavelet coefficient d(n), which is calculated as follows:

[0014]

[0015] Among them, * represents the convolution operation, h k is a low-pass filter, g l is a high-pass filter, x(n) is the vibration signal, k and l are the lengths of the filter coefficients, and n is the length of the discrete signal;

[0016] The low-frequency signal wavelet coefficient c(n) and the high-frequency signal wavelet coefficient d(n) obtained by decomposition are calculated again through high-pass filter and low-pass filter to further obtain wavelet coefficients with finer frequency resolution. The multi-layer decomposition of the signal is realized by analogy. The calculation formula of multi-layer decomposition is as follows:

[0017]

[0018]

[0019] in, is the low-frequency wavelet coefficient of the next layer, is the high-frequency wavelet coefficient of the next layer, and j is the layer to be decomposed.

[0020] Preferably, the low-pass filter calculation formula is as follows:

[0021]

[0022]

[0023] The high-pass filter is calculated as follows:

[0024]

[0025]

[0026] Among them, φ(x) is the scaling function, ψ(x) is the wavelet function, and x is the function independent variable.

[0027] Preferably, in S2, the noise threshold is divided into a global threshold and a SURE threshold; the noise threshold takes into account all signal characteristics, and its calculation formula is as follows:

[0028]

[0029] Among them, T univ is the global threshold, σ n is the noise standard deviation, N is the signal length;

[0030] The SURE threshold only considers the wavelet coefficient of the current band, and its calculation formula is as follows:

[0031]

[0032] Among them, T sure is the SURE threshold, is the noise variance, N sub is the sub-band coefficient length, #{·} represents the number of elements that meet the conditions, and Y t is the current sub-band coefficient, T is the noise threshold to be optimized, and i is an integer.

[0033] Preferably, the noise standard deviation σ n It is estimated by the finest frequency wavelet coefficient, and the calculation formula is as follows:

[0034] σ n =median(|d N -median(d N )|) / 0.6745;

[0035] Among them, d N is the wavelet coefficient with the finest frequency.

[0036] Preferably, in S3, based on the SureShrink algorithm, adaptive threshold determination is performed on sub-band coefficients of different vibration intensities and different frequency ranges, and the noise threshold determination calculation formula is as follows:

[0037]

[0038] η N =log2(N sub ) 3 / 2 ;

[0039]

[0040] Among them, η N , is the noise threshold determination factor;

[0041] After the threshold is determined, the wavelet coefficient shrinkage denoising is implemented based on the soft threshold method. The coefficients with amplitudes less than or equal to the threshold are set to 0, and the coefficients greater than the threshold are shrunk to the difference between the coefficient and the threshold. The calculation formula is as follows:

[0042]

[0043] Among them, T(x) is the wavelet coefficient.

[0044] Preferably, in S4, based on the denoised wavelet coefficients of each layer obtained in S3, a multi-layer filter bank is constructed, and the signal is reconstructed layer by layer starting from the finest scale wavelet coefficient. The calculation formula of wavelet reconstruction is as follows:

[0045]

[0046] in, are the reconstructed wavelet coefficients.

[0047] Preferably, in S5, a sliding window input of a long signal is implemented based on a sliding window algorithm, and the time characteristics of the signal are automatically extracted through a multi-layer recurrent neural network to obtain a signal characteristic data output in a sequence form.

[0048] Therefore, the present invention proposes a vibration signal denoising and feature extraction method based on wavelet adaptive threshold, and its beneficial effects are as follows:

[0049] (1) The present invention discloses a vibration signal denoising and feature extraction method based on wavelet adaptive threshold, which reduces the error influence caused by the vibration sensor's own noise and environmental disturbance through signal decomposition based on discrete wavelet transform and adaptive threshold denoising algorithm, and retains most of the features of the original vibration signal.

[0050] (2) The present invention provides a vibration signal denoising and feature extraction method based on wavelet adaptive threshold. A multi-layer recurrent neural network is established for the denoised vibration signal to realize the automatic extraction of time features, which provides an important foundation for the construction of subsequent fault diagnosis and health monitoring algorithms.

[0051] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of the overall process of a vibration signal denoising and feature extraction method based on wavelet adaptive threshold in the present invention;

[0053] Figure 2 The invention discloses a feature automatic extraction flow chart of a vibration signal denoising and feature extraction method based on wavelet adaptive threshold. DETAILED DESCRIPTION

[0054] In order to make the technical solutions, advantages and purposes of the present invention clearer, the technical solutions of the embodiments of the present invention are clearly and completely described below. The described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the protection scope of this application.

[0055] Unless otherwise defined, technical or scientific terms used in the present invention shall have the common meanings understood by one having ordinary skills in the field to which the present invention belongs.

[0056] like Figure 1 As shown, the present invention provides a vibration signal denoising and feature extraction method based on wavelet adaptive threshold, and the specific steps are as follows:

[0057] Step 1: Multi-layer decomposition of vibration signals based on discrete wavelet transform.

[0058] In actual applications, affected by equipment working conditions and environmental conditions, the measured signals of vibration sensors often contain a lot of noise, and the collected signals cannot be directly used for vibration analysis and fault diagnosis. For the measured vibration signals, the high and low frequency signals can be decomposed through the filter group composed of discrete wavelet transform.

[0059] The filter group consists of a high-pass filter and a low-pass filter. The vibration signal is processed by the low-pass filter to obtain the low-frequency signal wavelet coefficient c(n). The calculation formula is as follows:

[0060]

[0061] The vibration signal is processed by a high-pass filter to obtain the high-frequency signal wavelet coefficient d(n), which is calculated as follows:

[0062]

[0063] Among them, * represents the convolution operation, h k is a low-pass filter, g l is a high-pass filter, x(n) is the vibration signal, k and l are the lengths of the filter coefficients, and n is the length of the discrete signal;

[0064] Low pass filter h k The calculation formula is as follows:

[0065]

[0066]

[0067] High pass filter g l The calculation formula is as follows:

[0068]

[0069]

[0070] Among them, φ(x) is the scaling function and ψ(x) is the wavelet function.

[0071] The low-frequency signal wavelet coefficient c(n) and the high-frequency signal wavelet coefficient d(n) obtained by decomposition are calculated again through high-pass filter and low-pass filter to further obtain wavelet coefficients with finer frequency resolution. The multi-layer decomposition of the signal is realized by analogy. The calculation formula of multi-layer decomposition is as follows:

[0072]

[0073]

[0074] in, is the low-frequency wavelet coefficient of the next layer, is the high-frequency wavelet coefficient of the next layer, and j is the layer to be decomposed.

[0075] Step 2: Adaptive calculation of noise threshold.

[0076] Based on the multi-layer decomposition coefficients obtained in step 1, the global signal noise threshold is first calculated. The calculation formula of the global threshold is as follows:

[0077]

[0078] Among them, T univ is the global threshold, σ n is the noise standard deviation, N is the signal length;

[0079] Since the global threshold takes into account the data characteristics of the entire signal acquisition process, the noise threshold is generally too large. Therefore, the local adaptive threshold SURE threshold is calculated based on the coefficient characteristics of each sub-band. The calculation formula is as follows:

[0080]

[0081] Among them, T sure is the SURE threshold, is the noise variance, N sub is the sub-band coefficient length, #{·} represents the number of elements that meet the conditions, and Y t is the current sub-band coefficient, T is the noise threshold to be optimized, and i is an integer.

[0082] Noise standard deviation σ n It is estimated by the finest frequency wavelet coefficient, and the calculation formula is as follows:

[0083] σ n=median(|d N -median(d N )|) / 0.6745;

[0084] Among them, d N is the wavelet coefficient with the finest frequency.

[0085] Step 3: Wavelet coefficient shrinkage denoising based on SureShrink algorithm.

[0086] Based on the two noise thresholds calculated in step 2, the sub-band threshold is first determined. The noise threshold determination formula is as follows:

[0087]

[0088] η N =log2(N) 3 / 2 ;

[0089]

[0090] After the threshold is determined, the wavelet coefficient shrinkage denoising is realized based on the soft threshold method. The coefficients with amplitudes less than or equal to the threshold are set to 0, and the coefficients greater than the threshold are shrunk to the difference between the coefficient and the threshold. The formula is as follows:

[0091]

[0092] Among them, T(x) is the wavelet coefficient.

[0093] Step 4: Reconstruct vibration signal based on inverse discrete wavelet transform.

[0094] Based on the denoised wavelet coefficients of each layer obtained in step 3, a multi-layer filter bank is constructed. Based on the inverse discrete wavelet transform, starting from the finest scale wavelet coefficient, the signal is reconstructed layer by layer. The wavelet reconstruction calculation formula is as follows:

[0095]

[0096] in, are the reconstructed wavelet coefficients.

[0097] Step 5: Automatically extract signal features based on multi-layer recurrent neural network.

[0098] like Figure 2 As shown, based on the denoised vibration signal obtained in step 4, the data features of the actual signal measured by the vibration sensor are automatically extracted through a multi-layer recurrent neural network; first, the recurrent neural network input is determined based on the sliding window algorithm, and a multi-layer recurrent neural network is constructed to extract the time features of the vibration signal, and finally the sequence feature output of the vibration signal is obtained.

[0099] Therefore, the present invention provides a vibration signal denoising and feature extraction method based on wavelet adaptive threshold. Through signal decomposition based on discrete wavelet transform and adaptive threshold denoising algorithm, the error influence caused by the vibration sensor's own noise and environmental disturbance is reduced, and most of the features of the original vibration signal are retained. At the same time, a multi-layer recurrent neural network is established for the denoised vibration signal to realize the automatic extraction of time features, which provides an important foundation for the construction of subsequent fault diagnosis and health monitoring algorithms.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A vibration signal denoising and feature extraction method based on wavelet adaptive threshold, characterized in that: The specific steps are as follows: S1. Decomposing the collected vibration signal into high and low frequency signals through a filter bank based on discrete wavelet transform and finally realizing multi-layer decomposition of the vibration signal; S2, adaptive calculation of noise threshold; S3, wavelet coefficient shrinkage denoising based on SureShrink algorithm, adaptive threshold determination of sub-waveband coefficients of different vibration intensities and frequency ranges; S4, constructing a multi-layer filter bank based on inverse discrete wavelet transform to achieve vibration signal reconstruction; S5. Automatic extraction of signal features based on multi-layer recurrent neural network.

2. The vibration signal denoising and feature extraction method based on wavelet adaptive threshold according to claim 1 is characterized in that: In S1, the vibration signal collected by the vibration sensor is decomposed into high-frequency and low-frequency signals through a filter group composed of discrete wavelet transform. The filter group consists of a high-pass filter and a low-pass filter. The vibration signal is processed by the low-pass filter to obtain the low-frequency signal wavelet coefficient c(n). The calculation formula is as follows: The vibration signal is processed by a high-pass filter to obtain the high-frequency signal wavelet coefficient d(n), which is calculated as follows: Among them, * represents the convolution operation, h k is a low-pass filter, g l is a high-pass filter, x(n) is the vibration signal, k and l are the lengths of the filter coefficients, and n is the length of the discrete signal; The low-frequency signal wavelet coefficient c(n) and the high-frequency signal wavelet coefficient d(n) obtained by decomposition are calculated again through high-pass filter and low-pass filter to further obtain wavelet coefficients with finer frequency resolution. The multi-layer decomposition of the signal is realized by analogy. The calculation formula of multi-layer decomposition is as follows: in, is the low-frequency wavelet coefficient of the next layer, is the high-frequency wavelet coefficient of the next layer, and j is the layer to be decomposed.

3. The vibration signal denoising and feature extraction method based on wavelet adaptive threshold according to claim 2 is characterized in that: The low-pass filter calculation formula is as follows: The high-pass filter is calculated as follows: Among them, φ(x) is the scaling function, ψ(x) is the wavelet function, and x is the function independent variable.

4. The vibration signal denoising and feature extraction method based on wavelet adaptive threshold according to claim 1 is characterized in that: In S2, the noise threshold is divided into a global threshold and a SURE threshold; the noise threshold takes into account all signal characteristics, and its calculation formula is as follows: Among them, T univ is the global threshold, σ n is the noise standard deviation, N is the signal length; The SURE threshold only considers the wavelet coefficient of the current band, and its calculation formula is as follows: Among them, T sure is the SURE threshold, is the noise variance, N sub is the length of the sub-band coefficient, #{·} represents the number of elements that meet the conditions, and Y t is the current sub-band coefficient, T is the noise threshold to be optimized, and i is an integer.

5. The vibration signal denoising and feature extraction method based on wavelet adaptive threshold according to claim 4 is characterized in that: The noise standard deviation σ n It is estimated by the finest frequency wavelet coefficient, and the calculation formula is as follows: σ n =median(|d N -median(d N )|) / 0.6745; Among them, d N is the wavelet coefficient with the finest frequency.

6. The vibration signal denoising and feature extraction method based on wavelet adaptive threshold according to claim 4 is characterized in that: In S3, based on the SureShrink algorithm, adaptive threshold determination is performed on sub-band coefficients of different vibration intensities and frequency ranges. The noise threshold determination calculation formula is as follows: or N =log2(N sub ) 3 / 2 ; Among them, η N , is the noise threshold determination factor; After the threshold is determined, the wavelet coefficient shrinkage denoising is implemented based on the soft threshold method. The coefficients with amplitudes less than or equal to the threshold are set to 0, and the coefficients greater than the threshold are shrunk to the difference between the coefficient and the threshold. The calculation formula is as follows: Among them, T(x) is the wavelet coefficient.

7. The vibration signal denoising and feature extraction method based on wavelet adaptive threshold according to claim 6 is characterized in that: In S4, based on the denoised wavelet coefficients of each layer obtained in S3, a multi-layer filter bank is constructed, starting from the finest scale wavelet coefficients, and the signal is reconstructed layer by layer. The calculation formula of wavelet reconstruction is as follows: in, are the reconstructed wavelet coefficients.

8. The vibration signal denoising and feature extraction method based on wavelet adaptive threshold according to claim 1 is characterized in that: In S5, the sliding window input of the long signal is realized based on the sliding window algorithm, and the time characteristics of the signal are automatically extracted through the multi-layer recurrent neural network to obtain the signal characteristic data output in the form of a sequence.

Images