Abrasive particle feature extraction method based on differential signal band selection adaptive filtering

Abstract

The invention belongs to the technical field of oil abrasive particle monitoring, and particularly relates to an abrasive particle feature extraction method based on differential signal band selection adaptive filtering, which comprises the following steps: carrying out common-mode rejection on an oil abrasive particle signal to obtain a differential signal to be detected; performing wavelet packet decomposition on the differential signal to be detected through an orthogonal wavelet basis function to obtain a multi-decomposition-scale feature vector; calculating a similarity index on each decomposition scale, and extracting a target scale feature vector; processing the target scale feature vector by adopting an improved minimum root mean square adaptive filter to obtain a filtering vector; performing wavelet reconstruction on the zero vector and the filtering vector to obtain a noise reduction signal; dividing the noise reduction signal into a plurality of segments, and calculating a composite identification index of each segment; calculating the weight of each segment according to the composite identification index, and weighting all segments to obtain an oil abrasive particle characteristic signal extraction result; according to the invention, the signal-to-noise ratio of the abrasive particle signal is improved, the self-adaptive capability is stronger, and the accurate detection of the oil abrasive particle characteristics is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of oil wear particle monitoring, and in particular relates to a wear particle feature extraction method based on differential signal band-selective adaptive filtering. Background Art

[0002] When mechanical equipment is in working condition, changes in factors such as load and working environment will cause changes in the friction characteristics between the contact parts of the transmission device, thereby accelerating the wear of parts and causing particles to fall off. As the wear process continues, the degree of wear between parts gradually changes from mild to severe, resulting in an increase in the rate of abrasive particle generation. When the contact surface is in poor cleaning condition or the adhesion is strong, some abrasive particles will remain on the surface of the parts, and further aggravate the wear and induce lubrication failure in subsequent contacts, causing serious impact on the performance and service life of the mechanical equipment. According to data, wear is the main cause of 80% of mechanical equipment failures. Oil wear particle detection extracts rich wear particle information from the lubricating oil system through non-destructive detection and analysis, promptly discovers potential abnormal wear conditions and takes corresponding measures to adjust, so as to reduce maintenance costs, extend the service life of mechanical equipment, and solve potential hidden dangers of poor operation.

[0003] The inductive wear particle sensor can accurately detect wear particles by sensing and extracting the magnetic flux change-induced voltage signal characteristics induced by different material particles passing through the sensor magnetic field area. It is widely used in industries such as aerospace, shipping, and energy and power due to its convenience, timeliness, and interpretable wear particle physical property parameter mapping capabilities. However, the working environment of wear particle sensors is usually extremely harsh. In actual application scenarios, the output signal of the inductive sensor will inevitably be affected by various interferences and noises, resulting in the inability to directly and accurately extract the wear particle signal. Especially when the machine is subjected to vibration shock or strong electromagnetic interference, a large number of low-frequency and high-energy pulse signals will appear in the output signal, which will bring great errors to the signal parameter estimation, making it difficult for the wear particle signal identification algorithm to achieve balanced performance in terms of wear particle feature integrity protection and interference component elimination, which seriously misleads the accurate identification of wear particle signals. At present, the inductive sensor based on a single coil can achieve an improvement in signal-to-noise ratio through structural optimization and algorithm design, but the extraction of wear particle induced voltage characteristics under strong pulse interference still faces principle limitations. Summary of the invention

[0004] To solve the above problems, the present invention provides a wear particle feature extraction method based on differential signal band selection adaptive filtering. The working principle is based on the highly similar time domain morphology of the output signals of the mutual reference induction coils of the differential sensor. Considering the energy distribution characteristics of the wear signal frequency band, wavelet packet decomposition and correlation measurement are used to complete the band selection. An improved adaptive filtering algorithm with a variable step size is combined to suppress the intra-band noise. Finally, the wear particle feature signal extraction is realized through wavelet reconstruction and composite identification indicators.

[0005] The specific plan includes the following steps:

[0006] S1. Use a two-phase differential inductive sensor to build a lubricating oil wear particle monitoring system to collect differential oil wear particle signals p and q in real time;

[0007] S2. Low-pass filtering and periodic harmonic elimination are performed on the oil wear particle signals p and q to obtain differential signals p0 and q0 to be detected;

[0008] S3. Calculate the optimal decomposition scale J according to the broadband characteristics of the oil wear particle signal spectrum, and perform wavelet packet decomposition on each differential signal to be detected through an orthogonal wavelet basis function to obtain a multi-decomposition scale feature vector;

[0009] S4. Calculate the similarity index of the two multi-decomposition scale feature vectors at each decomposition scale, and extract the zero vector and the target scale feature vector p according to the similarity index. b and the target scale feature vector q b ; The zeroing vector includes a first zeroing vector and a second zeroing vector;

[0010] S5. Based on the target scale feature vector p b and q b , an improved minimum mean square adaptive filter with variable step size is used to suppress in-band noise to obtain a filter vector;

[0011] S6. Perform wavelet reconstruction on the zero vector and the filter vector to obtain the noise reduction signal

[0012] S7. The noise reduction signal Divide into multiple segments of equal length and calculate the composite identification index of each segment;

[0013] S8. The weight of each segment is obtained by binarizing the weight according to the composite identification index, and all segments are weighted to obtain the final oil wear particle characteristic signal extraction result.

[0014] Beneficial effects of the present invention:

[0015] The algorithm proposed in the present invention can effectively avoid the misidentification caused by low-frequency high-energy pulse signals while achieving background noise suppression, and realize accurate identification of wear induced voltage signals under the premise of fully protecting the integrity of wear particle characteristics. At present, traditional algorithms in this field will inevitably cause distortion of target signal characteristics. The most prominent problem is that it will change the amplitude of the wear voltage signal, thereby affecting the judgment of the health status of the machinery and failing to provide a reliable basis for the subsequent wear evolution analysis of the machinery. The wear particle feature extraction method based on differential signal band adaptive filtering proposed in the present invention can be based on a dual-channel differential induction wear particle sensor, without causing damage to the wear particle characteristics and retaining the original characteristics of the signal to the maximum extent, and quickly and accurately perform output signal noise reduction and wear particle feature identification. At the same time, it avoids excessive reliance on prior conditions in the threshold selection of complex traditional feature extraction algorithms, and also reduces the requirements for users' professional knowledge in the field of signal processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of the method of the present invention;

[0017] Figure 2 It is a two-phase differential signal to be detected after preprocessing by the present invention;

[0018] Figure 3 The present invention selects a two-phase differential signal after adaptive filtering;

[0019] Figure 4 The composite identification index and adaptive threshold based on energy estimation and Euclidean distance measurement of the present invention;

[0020] Figure 5 This is the final abrasive particle identification result of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] The present invention provides a wear particle feature extraction method based on differential signal band selection adaptive filtering, such as Figure 1 As shown, the following steps are included:

[0023] S1. A lubricating oil wear particle monitoring system is constructed using a two-phase differential inductive sensor to collect differential oil wear particle signals p = [p(0), p(1), ..., p(N-1)] and q = [q(0), q(1), ..., q(N-1)] in real time; p(n) and q(n) are the n = 0, 1, ..., N-1 elements in the differential oil wear particle signals p and q, respectively.

[0024] Specifically, the process of collecting raw signals by the lubricating oil wear particle monitoring system includes:

[0025] In the lubricating oil wear particle monitoring system, the acquisition card model is NI-9219, and the sampling frequency is set to f s =5000, sampling time is 4s, and the number of sampling points is N=20000. The inductive wear particle detection sensor (two-phase differential inductive sensor) uses DC excitation, driving current I=0.5A, the amplifier multiple of the preamplifier is 2000, and the peristaltic pump flow rate is set to 510ml / min. The wear particle size (equivalent diameter of the sphere) passes through the inductive wear particle detection sensor in order of 170mm and 130mm, and the inductive wear particle detection sensor will generate two corresponding induced voltage signals similar to a single-cycle sine waveform.

[0026] Specifically, the output signal model of the two coils of the two-phase differential inductive sensor is:

[0027]

[0028] q(n)=r2(n)

[0029] Among them, r1(n) and r2(n) are the interference signals output by the two coils, u(n) represents the wear particle characteristic signal, ε represents the wear particle signal width, and τ represents the center position of the wear particle signal.

[0030] S2. Perform low-pass filtering and periodic harmonic elimination on the oil wear particle signals p and q to obtain the differential signals p0 and q0 to be detected.

[0031] S3. Calculate the optimal decomposition scale J according to the broadband characteristics of the oil wear particle signal spectrum, and perform wavelet packet decomposition on each differential signal to be detected through orthogonal wavelet basis functions to obtain multi-decomposition scale feature vectors.

[0032] Specifically, the optimal decomposition scale J described in step S3 is calculated by the following formula:

[0033]

[0034] Among them, L c represents the axial width of the induction coil of the two-phase differential inductive sensor, v dRepresents the estimated flow velocity; γ represents the frequency band expansion index, which is generally 2.5-3.5.

[0035] S4. Calculate the similarity index of the two multi-decomposition scale feature vectors at each decomposition scale, and extract the zero vector and the target scale feature vector p according to the similarity index. b and the target scale feature vector q b ; The zeroing vector includes a first zeroing vector and a second zeroing vector.

[0036] Specifically, step S4 includes:

[0037] S41. The multi-decomposition scale feature vector obtained by performing wavelet packet decomposition on each differential signal to be detected through orthogonal wavelet basis functions (Haar, DB, etc.) is expressed as

[0038] X1={p1,p2,...,p M},p m ∈R 1×(N / M)

[0039] X2={q1,q2,...,q M},q m ∈R 1×(N / M)

[0040] Among them, X1 represents the multi-decomposition scale feature vector of the differential signal p0 to be detected, p m represents the feature vector of the mth decomposition scale of the differential signal to be detected p0; X2 represents the multi-decomposition scale feature vector of the differential signal to be detected q0, q m R represents the feature vector of the mth decomposition scale of the differential signal q0 to be detected; 1×(N / M) Represents a real number matrix. The length of each decomposed wavelet coefficient is equal to the number of sampling points N, M = 2 J Indicates the number of decomposition scales;

[0041] S42. Since the similarity of the non-wear particle components in the two differential signals to be detected is high, the energy of the wear particle characteristic component will be significantly greater than the energy of the interference signal component alone. Based on the frequency band concentration of the wear particle signal, a composite similarity index combining the Euclidean distance (ED) and similarity measurement is used to calculate the correlation between the decomposition scales of the differential signal to be detected. The calculation formula of the composite similarity index of two multi-decomposition scale feature vectors at each decomposition scale is expressed as

[0042]

[0043] Among them, k m represents the composite similarity index at the mth decomposition scale, <·> represents the vector inner product, and ||·||2 represents the L-2 norm;

[0044] S43. Define a band selection vector statistical characteristic condition according to the statistical characteristics, and select the band scale by using the band selection vector statistical characteristic condition. Specifically, for each composite similarity index, if it satisfies the band selection vector statistical characteristic condition, the decomposition scale corresponding to the composite similarity index is marked as the band selection scale; the band selection vector statistical characteristic condition is expressed as

[0045]

[0046] Among them, b represents the band selection scale;

[0047] S44. Extract the feature vectors corresponding to all selected scales from the multi-decomposition scale feature vector X1 to form the target scale feature vector p b , and at the same time extract all feature vectors corresponding to non-band-selective scales from the multi-decomposition scale feature vector X1 and set them to zero to obtain the first zero-set vector; extract all feature vectors corresponding to band-selective scales from the multi-decomposition scale feature vector X2 to form the target scale feature vector q b At the same time, all feature vectors corresponding to non-band-selective scales are extracted from the multi-decomposition scale feature vector X2 and set to zero to obtain the second zero-set vector.

[0048] Specifically, the out-of-band noise can be eliminated by directly setting the non-domain vector (the eigenvector corresponding to the non-band-selected scale) to zero.

[0049] S5. Based on the target scale feature vector p b and q b , an improved minimum mean square adaptive filter with variable step size is used to suppress in-band noise and obtain the filter vector.

[0050] Specifically, step S5 includes:

[0051] S51. Construct the adaptive filtering framework based on the least mean square root

[0052] z(n)=p b (n)-w(n)q b (n)

[0053] w(n+1)=w(n)+2μz(n)q b (n)

[0054] Among them, z(n) represents the filtering result, p b (n) represents the eigenvector p b The nth element in, w(n) represents the nth order coefficient of the filter, q b (n) represents the eigenvector q b The nth element in , w(n+1) represents the n+1th order coefficient of the filter, and μ represents the update step size.

[0055] S52. Calculate and estimate the fragment feature width Select the closest fragment feature width The average segment length N s ; where f s represents the sampling frequency, L c represents the axial width of the induction coil of the two-phase differential inductive sensor, v d Indicates the estimated flow rate.

[0056] S53. According to the average molecular segment length N s Divide and get G = N / N s sub-segments, and the cosine similarity function is used to calculate the correlation coefficient ρ of each sub-segment g , g=1,2,…,G. Since the noise in the long-term signal is highly correlated, the correlation coefficient value will drop sharply when the fragmentation feature appears. For the noise correlation sequence, directly eliminating any value has almost no effect on the average value of the sequence. Therefore, in order to avoid the influence of the fragmentation feature, the average value of all correlation coefficients except the maximum correlation coefficient is calculated as the initial weight ω0.

[0057] S54. A fixed step size may cause the filter weight to fluctuate too much in the final convergence stage, thereby reducing the fitting performance. Therefore, the present invention constructs an adaptive step size μ s (n) replaces the fixed update step size μ, and obtains the improved minimum mean square adaptive filter with variable step size, which is expressed as

[0058] w(n+1)=w(n)+2μ s (n)z(n)q b (n)

[0059]

[0060] Among them, w(0)=ω0.

[0061] S55. Use the improved least mean square adaptive filter to adjust the target scale feature vector p b ,q b In-band noise suppression is performed to obtain a filter vector.

[0062] S6. Perform wavelet reconstruction on the zero vector and the filter vector to obtain the noise reduction signal

[0063] Specifically, step S6 includes:

[0064] S61. Combining the first zero-setting vector and the filtering vector into a vector Perform wavelet reconstruction to obtain the denoised signal Restruct(·) represents the wavelet reconstruction process, Representation vector The eigenvector of the mth decomposition scale;

[0065] S62. Combining the second zero-setting vector and the filtering vector into a vector Perform wavelet reconstruction to obtain the denoised signal Representation vector The eigenvector of the mth decomposition scale.

[0066] Specifically, the zero vector and the filter vector are reconstructed by wavelet, and the reconstructed vector can also be expressed as

[0067]

[0068] Among them, LMS(·) represents the adaptive filtering operator, and zeros(·) represents the vector zeroing operator.

[0069] S7. Based on the stability of the morphological characteristics of the wear particle signal before and after filtering and the decay characteristics of the interference components, a composite identification index based on energy estimation and Euclidean distance measurement is established. Specifically, a sliding window is set to reduce the noise signal. Divide into multiple segments of equal length and calculate the composite recognition index of each segment.

[0070] Specifically, step S7 calculates the composite identification index according to the noise reduction signal, including:

[0071] S71. In the adaptive filtering process, due to the high similarity of signal components at the same time, the energy of the fragment containing the wear particle characteristics will be significantly different from the residual noise. The energy characteristic index calculation formula is defined as

[0072]

[0073] Among them, λ l Represents the noise reduction signal The energy characteristic index of the lth segment, η p,l ∈R 1×S Represents the noise reduction signal The lth segment, S represents the segment length, R 1×S represents a real matrix, Represents the square of the L-2 norm. After introducing the logarithm-based measurement method, it can not only highlight the energy characteristics, but also reduce the impact of the peak differences of different wear particle signals on the recognition results.

[0074] S72. The electromagnetic induction caused by the abrasive particles passing through the outside of the coil's closed magnetic field can be ignored, and the calculation formula for the matching index with the same timestamp is defined as:

[0075]

[0076] Among them, φ l represents the time stamp matching index of the lth segment, η q,l Represents the noise reduction signal The first segment;

[0077] S73. Calculation of composite identification index

[0078] ζ l =λ l ×φ l

[0079] Among them, l Represents the composite identification index of the lth fragment.

[0080] S8. The weight of each segment is obtained by binarizing the weight according to the composite identification index, and all segments are weighted to obtain the final oil wear particle characteristic signal extraction result.

[0081] Specifically, the weight of each composite identification index is binarized to obtain the weight of each fragment, which is expressed as

[0082]

[0083] in, represents the weight of the l=1,2,…,Lth fragment, L represents the number of fragments, represents a binary weight vector, N represents the number of signal sampling points, mean(·) represents the mean function, ζ=(ζ1,ζ2,...,ζ L ) represents a composite index vector, Represents the signal segment width coefficient.

[0084] The weighting operation is expressed as:

[0085]

[0086] Where ξ represents the final oil wear particle characteristic signal extraction result, Hadamard(·) represents the Hadamard product operator, Represents the binary weight vector after zero padding

[0087] In one embodiment, a differential oil wear particle signal is collected by a two-phase differential inductive sensor, and the signal is preprocessed by using common mode suppression to obtain the following: Figure 2 The signal to be detected is shown. Calculate the optimal decomposition scale J = 3. Use the orthogonal wavelet basis function to decompose the signal to be detected by wavelet packet, and obtain the wavelet packet tree M = 2 J= 8 multi-decomposition scale feature vectors. The composite similarity index is used to calculate the correlation between the features of each layer of the signal to be detected, and the band selection scale is obtained. Here, the composite similarity index of each decomposition scale is 0.0029, 0.0015, 0.0004, 0.0008, 0.0001, 9.2528e-05, 2.2610e-05, 3.4523e-06; according to the statistical characteristics of the band selection vector, the adaptive threshold method is set to extract the target scale feature vector, where the adaptive threshold The non-domain vectors are directly set to zero to eliminate out-of-band noise; for the differential vectors corresponding to the serial numbers, an improved minimum mean square root adaptive filter with a variable step size is used to suppress the in-band noise. The filter vector and the zeroed vector are reconstructed by wavelet to obtain the noise reduction signal, such as Figure 3 As shown in Figure 2. Based on the stability of the morphological characteristics of the wear particle signal before and after filtering and the decay characteristics of the interference components, a composite identification index based on energy estimation and Euclidean distance measurement is established, and a sliding window is set for weight calculation. Among them, the identification feature domain and the adaptive recognition threshold are shown in Figure 2. Figure 4 Finally, the adaptive threshold is used to perform weight binarization operation, and the wear particle feature recognition is further realized through signal weighting. Among them, the identified wear particle feature signal is as follows Figure 5 shown.

[0088] In the present invention, unless otherwise clearly stipulated and limited, the terms such as "installation", "setting", "connection", "fixation" and "rotation" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral one; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. Unless otherwise clearly defined, ordinary technicians in this field can understand the specific meanings of the above terms in the present invention according to the specific circumstances.

[0089] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A wear particle feature extraction method based on differential signal band selection adaptive filtering, characterized in that: The following steps are involved: S1. Use a two-phase differential inductive sensor to build a lubricating oil wear particle monitoring system to collect differential oil wear particle signals p and q in real time; S2. Low-pass filtering and periodic harmonic elimination are performed on the oil wear particle signals p and q to obtain differential signals p0 and q0 to be detected; S3. Calculate the optimal decomposition scale J according to the broadband characteristics of the oil wear particle signal spectrum, and perform wavelet packet decomposition on each differential signal to be detected through an orthogonal wavelet basis function to obtain a multi-decomposition scale feature vector; S4. Calculate the composite similarity index of the two multi-decomposition scale feature vectors at each decomposition scale, and extract the zero vector and the target scale feature vector p according to the composite similarity index. b and the target scale feature vector q b ; The zeroing vector includes a first zeroing vector and a second zeroing vector; S5. Based on the target scale feature vector p b and q b , an improved minimum mean square adaptive filter with variable step size is used to suppress in-band noise to obtain a filter vector; S6. Perform wavelet reconstruction on the zero vector and the filter vector to obtain the noise reduction signal S7. The noise reduction signal Divide into multiple segments of equal length and calculate the composite identification index of each segment; S8. The weight of each segment is obtained by binarizing the weight according to the composite identification index, and all segments are weighted to obtain the final oil wear particle characteristic signal extraction result.

2. The wear particle feature extraction method based on differential signal band selection adaptive filtering according to claim 1 is characterized in that: Step S4 specifically includes: S41. The multi-decomposition scale feature vector obtained by performing wavelet packet decomposition on each differential signal to be detected through orthogonal wavelet basis functions is expressed as X1={p1,p2,...,p M },p m ∈R 1×(N / M) X2={q1,q2,...,q M },q m ∈R 1×(N / M) Among them, X1 represents the multi-decomposition scale feature vector of the differential signal p0 to be detected, p m represents the feature vector of the mth decomposition scale of the differential signal to be detected p0; X2 represents the multi-decomposition scale feature vector of the differential signal to be detected q0, q m R represents the feature vector of the mth decomposition scale of the differential signal q0 to be detected; 1×(N / M) represents a real number matrix, N represents the number of sampling points, M = 2 J Indicates the number of decomposition scales; S42. Calculate the composite similarity index of two multi-decomposition scale feature vectors at each decomposition scale, expressed as Among them, k m represents the composite similarity index at the mth decomposition scale, <·> represents the vector inner product, and ||·||2 represents the L-2 norm; S43. For each composite similarity index, if it satisfies the band selection vector statistical characteristic condition, the decomposition scale corresponding to the composite similarity index is marked as the band selection scale; the band selection vector statistical characteristic condition is expressed as Among them, b represents the band selection scale; S44. Extract the feature vectors corresponding to all selected scales from the multi-decomposition scale feature vector X1 to form the target scale feature vector p b , and at the same time extract all feature vectors corresponding to non-band-selective scales from the multi-decomposition scale feature vector X1 and set them to zero to obtain the first zero-set vector; extract all feature vectors corresponding to band-selective scales from the multi-decomposition scale feature vector X2 to form the target scale feature vector q b At the same time, all feature vectors corresponding to non-band-selective scales are extracted from the multi-decomposition scale feature vector X2 and set to zero to obtain the second zero-set vector.

3. The wear particle feature extraction method based on differential signal band selection adaptive filtering according to claim 1 is characterized in that: Step S5 specifically includes: S51. Construct the adaptive filtering framework based on the least mean square root z(n)=p b (n)-w(n)q b (n) w(n+1)=w(n)+2μz(n)q b (n) Among them, z(n) represents the filtering result, p b (n) represents the eigenvector p b The nth element in, w(n) represents the nth order coefficient of the filter, q b (n) represents the eigenvector q b The nth element in , w(n+1) represents the n+1th order coefficient of the filter, and μ represents the update step size; S52. Calculate the fragment feature width According to the fragment feature width Determine the average segment length N s ; where f s represents the sampling frequency, L c represents the axial width of the induction coil of the two-phase differential inductive sensor, v d represents the estimated flow rate; S53. According to the average molecular segment length N s Divide and get G = N / N s sub-segments, calculate the correlation coefficient ρ of each sub-segment g , g=1,2,…,G; calculate the average value of all correlation coefficients except the maximum correlation coefficient as the initial weight ω0; S54. Constructing adaptive step size μ s (n) replaces the fixed update step size μ, and obtains the improved minimum mean square adaptive filter with variable step size, which is expressed as w(n+1)=w(n)+2μ s (n)z(n)q b (n) S55. Use the improved least mean square adaptive filter to adjust the target scale feature vector p b ,q b In-band noise suppression is performed to obtain a filter vector.

4. The wear particle feature extraction method based on differential signal band selection adaptive filtering according to claim 1 is characterized in that: Step S6 specifically includes: S61. Combining the first zero-setting vector and the filtering vector into a vector Perform wavelet reconstruction to obtain the denoised signal Restruct(·) represents the wavelet reconstruction process, Representation vector The eigenvector of the mth decomposition scale; S62. Combining the second zero-setting vector and the filtering vector into a vector Perform wavelet reconstruction to obtain the denoised signal Representation vector The eigenvector of the mth decomposition scale.

5. The wear particle feature extraction method based on differential signal band selection adaptive filtering according to claim 1 is characterized in that: Step S7 calculates the composite identification index according to the noise reduction signal, including: S71. Calculate energy characteristic index Among them, λ l Represents the noise reduction signal The energy characteristic index of the lth segment, η p,l Represents the noise reduction signal The first fragment, represents the square of the L-2 norm; S72. Calculate the timestamp matching index Among them, φ l represents the time stamp matching index of the lth segment, η q,l Represents the noise reduction signal The first segment; S73. Calculation of composite identification index g l =λ l ×φ l Among them, l Represents the composite identification index of the lth fragment.

6. The wear particle feature extraction method based on differential signal band selection adaptive filtering according to claim 1 is characterized in that: The weight of each composite identification index is binarized to obtain the weight of each fragment, which is expressed as in, represents the weight of the lth segment, N represents the number of signal sampling points, mean(·) represents the mean function, ζ=(ζ1,ζ2,...,ζ L ) represents the composite index vector, ζ l represents the composite identification index of the lth fragment, L represents the number of fragments, Represents the signal segment width coefficient.

Images