Rolling bearing signal reconstruction method fusing 5G LDPC and compressed sensing

Abstract

The invention discloses a rolling bearing signal reconstruction method fusing 5G LDPC and compressed sensing. The method comprises the steps of obtaining an original vibration signal of a rolling bearing; performing DCT on the original vibration signal to obtain a frequency domain vibration signal; performing weighting processing on the frequency domain vibration signal; setting a threshold value and a weight based on a DCT coefficient of the original vibration signal to generate a weighting matrix, and performing calculation according to the frequency domain vibration signal and the weighting matrix to obtain a weighted vibration signal; compressed sensing processing is carried out on the weighted vibration signal through an observation matrix, a weighted observation signal is obtained, and the observation matrix is realized through a 5G-NR-LDPC code; and after channel transmission is carried out on the weighted observation signal, a reconstructed vibration signal is obtained through signal reconstruction. According to the method, signal reconstruction with a better effect is realized on the basis of ensuring high compression of the vibration signal of the original rolling bearing.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical fault detection, and in particular to a rolling bearing signal reconstruction method integrating 5G LDPC and compressed sensing. Background Art

[0002] Rolling bearings are core components in rotating machinery, and their working status is crucial to the overall machine performance. The status information of rolling bearings is often contained in vibration signals. Therefore, if the working status of rolling bearings is to be monitored and fault diagnosis and prediction are to be achieved, the collection and transmission of vibration signals becomes an extremely critical link. The traditional Nyquist sampling theorem first samples and then compresses the signal processing method. After the data compression is completed, most of the redundant data is discarded, resulting in data waste. To solve these problems, the prior art discloses the theory of compressed sensing (CS). This method obtains the global information of the signal through a column-full rank measurement matrix for sparse or sparsely representable signals, thereby greatly reducing the pressure on the sampling end in data collection, storage, and transmission. Finally, the optimized algorithm is used at the receiving end with sufficient computing resources to reconstruct a reconstructed signal that is close to the original signal. Many studies have verified the feasibility of compressed sensing technology in rolling bearing vibration signal processing. However, the accurate reconstruction of rolling bearing vibration signals still relies on the high compression ratio of the signal or the complex reconstruction algorithm. That is, the traditional compressed sensing technology cannot achieve a good compromise between compression ratio and reconstruction accuracy.

[0003] Compressed sensing also has the problem of difficulty in hardware implementation. To address this problem, existing theoretical analysis proves that if a given binary parity code has good performance in channel coding linear programming decoding, then the matrix as a compressed sensing observation matrix also has good performance in compressed reconstruction, proving that LDPC codes are feasible for compressed sensing. However, the existing technology still has the limitation of being difficult to construct a check matrix of any size according to demand, which limits the flexible application of compressed sensing in rolling bearing vibration signal compression. Summary of the invention

[0004] Technical purpose: In view of the defects in the prior art, the present invention discloses a rolling bearing signal reconstruction method integrating 5G LDPC and compressed sensing, which achieves better signal reconstruction while ensuring the high compression of the original rolling bearing vibration signal.

[0005] Technical solution: In order to achieve the above technical objectives, the present invention adopts the following technical solution.

[0006] A rolling bearing signal reconstruction method integrating 5G LDPC and compressed sensing, the method comprising:

[0007] Obtain the original vibration signal of the rolling bearing;

[0008] Perform DCT transformation on the original vibration signal to obtain a frequency domain vibration signal;

[0009] Perform weighted processing on the frequency domain vibration signal: set the threshold and weight based on the DCT coefficient of the original vibration signal to generate a weighted matrix, and calculate the weighted vibration signal based on the frequency domain vibration signal and the weighted matrix;

[0010] The weighted vibration signal is subjected to compressed sensing processing through an observation matrix to obtain a weighted observation signal, wherein the observation matrix is ​​implemented through a 5G-NR-LDPC code;

[0011] After the weighted observation signal is transmitted through the channel, the reconstructed vibration signal is obtained through signal reconstruction.

[0012] Preferably, the calculation formula of the frequency domain vibration signal is:

[0013] ξ=Ψx

[0014] Among them, Ψ is a sparse matrix and also the DCT basis matrix.

[0015] Preferably, the process of generating the weighting matrix includes:

[0016] Generate weight factor q according to DCT coefficient distribution and preset threshold T i , weight factor q i The calculation formula is:

[0017] q i =ExT(|ξ i |-T)

[0018] Among them, ξ i is the i-th DCT coefficient of the original vibration signal, i = 1, 2, ..., N, N is the length of the original vibration signal; the function ExT is used to determine |ξ i |Is it greater than T?

[0019] Based on the weight factor q i Construct the weighted matrix Q, the calculation formula is:

[0020] Q = diag(q1, q2, ..., q i , ..., q N ).

[0021] Preferably, the calculation formula of function ExT is:

[0022]

[0023] Among them, z is the variable of function ExT.

[0024] Preferably, the calculation formula of the weighted vibration signal is:

[0025] Ξ=QΨΨθ

[0026] Among them, Ξ is the weighted vibration signal, Q is the weighting matrix, Ψ is the DCT basis matrix, and θ is the sparse coefficient.

[0027] Preferably, the calculation formula of the weighted observation signal is:

[0028] Y=ΦΞ=A Q θ

[0029] Where Y is the weighted observation signal, Φ is the 5G-NR-LDPC code check matrix, which is also the observation matrix, Ξ is the weighted vibration signal, and A Q is the weighted perception matrix, and θ is the sparse coefficient.

[0030] Preferably, the number of columns of the observation matrix is ​​a power of 2.

[0031] Beneficial effects: The present invention uses the sparse coefficients of the original time domain signal in the discrete domain to construct a weighted matrix, and improves the ability of compressed sensing to capture important information in the original signal through the weighted matrix; then the 5G-NR-LDPC code check matrix is ​​used to implement the sampling and compression of the original signal; finally, the signal is reconstructed through the compressed sensing reconstruction algorithm. The present invention can achieve better signal reconstruction on the basis of ensuring the high compression of the original rolling bearing vibration signal. In addition, the hardware implementation of the method described in the present invention is relatively easy, and has broad application prospects in the fields of industrial signal processing, mechanical fault diagnosis, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is a flow chart of the method of the present invention;

[0033] Figure 2 Schematic diagram of the reconstructed Pearson correlation coefficient for different CRs in the simulation;

[0034] Figure 3 Schematic diagram of reconstructed NMSE for different CRs in simulation;

[0035] Figure 4 Schematic diagram of the reconstructed NMSE of the 5G-WCS algorithm and the 5G-CS algorithm under different CRs in the simulation;

[0036] Figure 5 Schematic diagram of the reconstructed Pearson correlation coefficient of the 5G-WCS algorithm and the 5G-CS algorithm under different CRs in the simulation;

[0037] Figure 6 Schematic diagram of the original signal of the normal signal of the rolling bearing in the simulation, the reconstructed signal using the 5G-WCS algorithm, and the reconstructed signal using the 5G-CS algorithm;

[0038] Figure 7 Schematic diagram of the original signal of the rolling bearing inner ring fault in the simulation, the reconstructed signal of the 5G-WCS algorithm, the reconstructed signal of the 5G-CS algorithm, and the reconstructed signal of the traditional CS algorithm;

[0039] Figure 8 Schematic diagram of the original signal of the rolling bearing outer ring fault in the simulation, the reconstructed signal of the 5G-WCS algorithm, the reconstructed signal of the 5G-CS algorithm, and the reconstructed signal of the traditional CS algorithm;

[0040] Fig. 9 Schematic diagram of the original signal of rolling element failure of rolling bearing in simulation, the reconstructed signal using 5G-WCS algorithm, the reconstructed signal using 5G-CS algorithm and the reconstructed signal using traditional CS algorithm. DETAILED DESCRIPTION

[0041] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0042] A rolling bearing signal reconstruction method integrating 5G LDPC and compressed sensing in this embodiment introduces a weighted strategy and a 5G-NR-LDPC code check matrix to realize signal reconstruction, performs weighted processing in the signal preprocessing stage, and uses the weighted matrix to improve the ability of compressed sensing to capture important information; and introduces the 5G-NR-LDPC code check matrix in the signal observation stage to realize efficient signal compression and hardware implementation of compressed sensing.

[0043] As attached Figure 1 As shown, a rolling bearing signal reconstruction method integrating 5G LDPC and compressed sensing in this embodiment includes the following steps:

[0044] S1. Obtain the original vibration signal of the rolling bearing; the original vibration signal is a time domain signal, represented by x, and has a length of N;

[0045] S2. Perform DCT transformation on the original vibration signal to obtain a frequency domain vibration signal. The calculation formula of the frequency domain vibration signal ξ is:

[0046] ξ=Ψx

[0047] Among them, Ψ is a sparse matrix, which is also the DCT basis matrix. The DCT basis matrix can be obtained by performing DCT transformation on the unit matrix;

[0048] S3, performing weighted processing on the frequency domain vibration signal: setting a threshold and a weight based on the DCT coefficient of the original vibration signal to generate a weighted matrix, and obtaining a weighted vibration signal according to the frequency domain vibration signal and the weighted matrix;

[0049] In order to better realize weighted processing, DCT transform is used to realize sparse representation of signals. The core of weighted processing lies in the reasonable setting of weight matrix. By assigning differentiated weights to the original vibration signal components, the priority capture of important information is realized, thereby reducing the sampling points while maintaining the reconstruction accuracy of the signal. Since the representation of the original vibration signal on the DCT basis is usually sparse, that is, most of the energy is concentrated on a few low-frequency coefficients, this energy concentration is used to generate the weight factor q according to the DCT coefficient distribution and the preset threshold T i , weight factor q i The calculation formula is:

[0050] q i =ExT(|ξ i |-T)

[0051] where ξ i is the i-th component of the frequency domain vibration signal, which is also the DCT coefficient, i = 1, 2, ..., N, N is the length of the original vibration signal, which is also the number of columns of the observation matrix; the function ExT is used to determine |ξ i |Is it greater than T? The detailed calculation formula of function ExT is as follows:

[0052]

[0053] Among them, z is the variable of function ExT.

[0054] Based on the weight factor q i Construct the weighted matrix Q. The weighted matrix Q is a diagonal matrix that reflects the importance of each DCT coefficient and can be expressed as:

[0055] Q = diag(q1, q2, ..., q i , ..., q N )

[0056] In the weighted sampling process of signal processing, each component of the frequency domain vibration signal is weighted by the weighting matrix Q, that is, ξ i This adjustment is intended to perform corresponding weighted processing on the components of the frequency domain vibration signal according to the importance of each DCT coefficient. The calculation formula is:

[0057] Ξ=Qξ=QΨx=QΨΨθ

[0058] Wherein, Ξ is the weighted vibration signal, and θ is the sparse coefficient, which is calculated according to x=Ψθ.

[0059] S4, performing compressed sensing processing on the weighted vibration signal through an observation matrix to obtain a weighted observation signal, wherein the observation matrix is ​​implemented by a 5G-NR-LDPC code;

[0060] 5G-NR-LDPC code improves the flexibility and efficiency of LDPC code by optimizing the structure of QC-LDPC code. 5G-NR-LDPC code adopts Raptor-like structure, and by gradually expanding the core matrix of high code rate to low code rate, it allows LDPC code to use a fixed-size base matrix B to generate parity check matrices (PCM) of different code rates and sizes, thereby achieving greater flexibility and efficiency in the encoding and decoding process. The size of base matrix B is M×(K+M), where K is the number of information bits and M is the number of check bits. The structure of base matrix B is as follows:

[0061]

[0062] Among them B H and B U Represent the maximum code rate and incremental redundancy, 0 and I represent the zero matrix and the unit matrix, respectively. The first K columns and the last M columns of the base matrix B become the information columns and the parity check columns, respectively. Each element in the base matrix B specifies the generation rule of the submatrix in the check matrix. Specifically, a non-negative integer indicates that the submatrix is ​​a cyclic shift matrix whose shift steps are determined by the integer value, while "-1" indicates that the submatrix is ​​an all-zero matrix. This structure not only ensures the flexibility of LDPC codes, but also guarantees efficient performance in the actual encoding and decoding process.

[0063] Compressed sensing and LDPC codes have significant similarities in structure and function, and both have commonalities in data processing and reconstruction. The sampling and reconstruction of compressed sensing correspond to the encoding and decoding of LDPC codes respectively. 5G-NR-LDPC codes have superior performance compared to general LDPC codes, especially 5G-NR-LDPC codes can construct a check matrix with richer signal compression ratio. This flexibility makes the 5G-NR-LDPC code check matrix more suitable for compressed sensing applications. Using the 5G-NR-LDPC code check matrix as a compressed sensing observation matrix, the weighted vibration signal is observed and obtained:

[0064] Y=ΦΞ=ΦQΨΨθ=A Q θ

[0065] Wherein, Y is the weighted observation signal with a length of M, Φ is the 5G-NR-LDPC code check matrix, which is also the observation matrix. The number of rows of the observation matrix is ​​the length of the weighted observation signal, the number of columns of the observation matrix is ​​N, and the number of columns of the observation matrix is ​​a power of 2; A Qis the weighted sensing matrix. After the compressed sensing processing of the weighted strategy, the weighted observation signal not only contains the key information of the original signal, but also emphasizes the more important part of the sparse coefficients, highlighting the reconstruction priority of important information.

[0066] S5. After the weighted observation signal is transmitted through the channel, a reconstructed vibration signal is obtained by signal reconstruction; the reconstructed vibration signal can be subjected to envelope spectrum analysis to further identify the type of rolling bearing fault. The reconstructed vibration signal is a frequency domain signal. Signal reconstruction can adopt the OMP (Orthogonal Matching Pursuit) algorithm, the BP (Basis Pursuit) algorithm, and the BSBL (Block Sparse Bayesian Learning) algorithm.

[0067] Simulation Description:

[0068] The simulation of the present invention includes two parts: one is the simulation performance analysis of the observation matrix, and the other is the simulation performance analysis of the method of the present invention.

[0069] For the first part, the number of columns N of the measurement matrix is ​​the same as the length of the original vibration signal. For a signal of a certain length, if the value of N in the measurement matrix is ​​too large, the time for a single compression reconstruction will be too long; if the measurement matrix is ​​too small, the amount of data at a certain compression ratio will be too small, which will cause the loss of important feature data and lead to a decrease in the quality of the reconstructed signal. Therefore, it is necessary to determine the appropriate value of the number of columns N of the measurement matrix.

[0070] In addition, the signal compression ratio represents the ratio between the number of data points in the original signal and the number of measurement values ​​retained for reconstructing the signal. The calculation formula of the signal compression ratio CR is:

[0071]

[0072] Where, M is the weighted observation signal length;

[0073] The normal signal in the rolling bearing experimental data set of the University of Western Reserve University in the United States is used for the experiment. The value of the observation matrix N is taken in turn as N={256, 512, 1024, 2048, 3072, 4096}, and the signal compression ratio CR is fixed at 0.25. The corresponding value of M is M={64, 128, 256, 512, 768, 1024}. The sparse matrix uses the DCT basis matrix, and the observation matrix uses the 5G-NR-LDPC code check matrix. The more general OMP algorithm is used to reconstruct the observation signal. The influence of different sizes of observation matrices on the signal reconstruction quality and efficiency is calculated. The experimental results are shown in Table 1.

[0074] Table 1 The influence of the measurement matrix size on the signal reconstruction effect

[0075]

[0076] As shown in Table 1, under the same signal compression ratio, as the number of observation matrix columns N increases, the Pearson correlation coefficient and NMSE gradually increase, indicating that the reconstruction quality of the observation signal of the reconstruction algorithm gradually improves, but the corresponding reconstruction time also increases. When N = 1024, the quality of the reconstructed signal is high and the reconstruction time is reasonable. Thereafter, as the N value continues to increase, the reconstruction time becomes longer but the reconstruction quality of the signal is relatively stable. Therefore, considering the signal reconstruction quality and reconstruction time, it is more reasonable to choose N = 1024 for signal processing.

[0077] In order to verify the good observation performance of the 5G-NR-LDPC code check matrix on the vibration signal, the effects of Gaussian random matrix, Bernoulli random matrix, sparse random matrix, partial Hadamard matrix, Toeplitz matrix and the 5G-NR-LDPC code check matrix used in the present invention on the reconstruction quality of the vibration signal are compared. The DCT basis matrix is ​​used as the sparse matrix, and the OMP algorithm is used as the reconstruction algorithm. The experimental conditions are set as follows: a normal rolling bearing signal with a length of 1024 is selected, and the signal compression ratio CR = M / N = {0.25, 0.375, 0.4375, 0.5, 0.625, 0.75, 0.8125} is used to compare the reconstruction effects of different observation matrices on the signal at different compression ratios. The experimental results are as follows: Figure 2 and Figure 3 As shown. Figure 2 and Figure 3 It can be seen that when the signal compression ratio CR is only 0.25, the reconstruction Pearson correlation coefficient of the 5G-NR-LDPC code check matrix is ​​0.88, and the NMSE is about 0.21, which is the best performance compared with other matrices. This shows that when the compression ratio is low, the 5G-NR-LDPC code check matrix is ​​more stable in signal reconstruction, and can better capture and reconstruct important information in the signal, ensuring a high reconstruction accuracy. As the compression ratio decreases, the NMSE of other observation matrices increases rapidly, and the Pearson correlation coefficient also decreases significantly, while the NMSE of the 5G-NR-LDPC code check matrix increases more slowly, and the correlation coefficient remains at a high level. This result shows that the 5G-NR-LDPC code check matrix can still effectively reconstruct the signal when the signal is highly compressed, while the performance of other matrices degrades rapidly.

[0078] For the second part, in order to further verify the effectiveness of the method of the present invention, hereinafter referred to as the 5G-WCS algorithm, the 5G-WCS algorithm is compared with the traditional CS algorithm and the compressed sensing (5G-CS algorithm) that only introduces the 5G-NR-LDPC code check matrix. The data set uses the rolling bearing experimental data set of the University of the Western Reserve in the United States, which is widely used in the research field of fault diagnosis. The data set comes from a motor-driven mechanical system. Single-point faults are created on the inner ring, outer ring and rolling element of the bearing through electric spark technology to simulate the bearing fault state. The fault diameter is set to 0.007, 0.014 and 0.021 inches (1 inch = 25.4mm). The data set covers four types of data: normal data, inner ring fault data, outer ring fault data and rolling element fault data. The present invention conducts simulation experiments and detailed analysis on these four types of data, and comprehensively evaluates the reconstruction performance of the proposed 5G-WCS algorithm under different signals. The specific characteristics of the signal selected in this article are as follows: the fault diameter is 0.021 inches (0.5334mm), the load is 2 horsepower, and the speed is 1750r / min.

[0079] The experimental conditions are set as follows: a normal rolling bearing signal with a length of 1024 is selected, the signal compression ratio CR = M / N = {0.25, 0.375, 0.4375, 0.5, 0.625, 0.75, 0.8125}, the observation matrix uses the 5G-NR-LDPC code check matrix, the sparse matrix uses the DCT basis matrix, and the reconstruction algorithm uses the OMP algorithm. The reconstruction effects of the 5G-WCS algorithm and the 5G-CS algorithm are compared. The experimental results are as follows: Figure 4 and Figure 5 The optimization percentage of reconstruction indicators is shown in Table 2.

[0080] Table 2 Optimization of reconstruction effect of 5G-WCS algorithm compared with 5G-CS algorithm

[0081] <![CDATA[CR s ]]> 0.25 0.375 0.4375 0.5 0.625 0.75 0.8125 <![CDATA[Δ NMSE ]]> 63.62％ 97.52％ 97.42％ 90.22％ 74.40％ 53.45％ 65.30％ <![CDATA[Δ Pearson ]]> 7.83％ 4.28％ 4.12％ 1.17％ 0.30％ 0.12％ 0.04％

[0082] from Figure 4 and Figure 5It can be seen that when the signal compression ratio is greater than 0.375, the NMSE and Pearson correlation coefficient of the signal reconstruction of the 5G-WCS algorithm change less, showing a more stable reconstruction performance. Compared with the 5G-CS algorithm, the 5G-WCS algorithm has significant advantages based on the high compression of the signal. By introducing a weighted strategy, the sparsity and structural characteristics in the signal are better utilized to effectively enhance the reconstruction accuracy of the signal. Especially in scenarios with large data volumes and high information complexity, the 5G-WCS algorithm can give priority to reconstructing important information, thereby improving the quality of overall signal reconstruction. This optimization effect is particularly important under low compression ratios, because low compression ratios mean that more information is compressed and discarded, and the effectiveness of the weighted strategy directly affects the success rate and accuracy of reconstruction. The 5G-WCS algorithm achieves that less signal can be transmitted when the reconstruction effect is equivalent, further reducing the amount of data that needs to be transmitted. This advantage enables it to more effectively reduce the amount of data transmission and improve channel utilization under conditions of limited channel bandwidth.

[0083] As can be seen from Table 2, under different compression ratios, the optimization of the 5G-WCS algorithm relative to the 5G-CS algorithm in reconstruction effect varies with the compression ratio. In particular, when the signal compression ratio is 0.375, the optimization effect of the 5G-WCS algorithm is most significant, with an increase in NMSE of 97.5159% and an increase in Pearson correlation coefficient of 4.2810%. Therefore, considering the optimization effect and reconstruction effect, it is more reasonable to choose CR = 0.375 for signal processing.

[0084] In order to further analyze the performance of the 5G-WCS algorithm proposed in the present invention, the reconstructed frequency domain signal waveforms obtained by using the algorithm and the 5G-CS algorithm are also given. The simulation experiment uses a normal rolling bearing signal with a length of 1024, a compression ratio CR=0.375, a DCT basis matrix for the sparse matrix, and an OMP algorithm for the reconstruction algorithm, such as Figure 6 As shown. Figure 6 It can be seen that the NMSE of the 5G-WCS algorithm is 0.0025, which is about 97.36% less than that of the 5G-CS algorithm. In addition, the Pearson correlation coefficient of the 5G-WCS algorithm is 0.9988, which is about 4.85% higher than that of the 5G-CS algorithm. These data show that the reconstructed signal of the 5G-WCS algorithm is highly consistent with the original signal in the frequency domain, retaining the main frequency components of the signal, while greatly reducing the reconstruction error and providing higher reconstruction accuracy.

[0085] In actual industrial environments, bearing fault vibration signals often show short-term transient features, such as mechanical shock, sudden vibration or abnormal noise. These transient features appear as spikes or pulses in the time domain and cause broadband disturbances in the frequency domain. Compared with normal bearing vibration signals, fault signals contain more complex frequency components. Different types of faults will produce different frequency components and transient features. Coupled with the complex noise environment, this places higher demands on the 5G-WCS algorithm. The rolling bearing experimental data set of the University of the Western Reserve in the United States includes three types of fault signals, including inner ring fault, outer ring fault and rolling element fault. The 5G-WCS algorithm, 5G-CS algorithm and traditional CS algorithm are used to reconstruct the three fault signals respectively. In the experiment, 1024 sampling points of the relevant data are selected for the experiment, and the signal compression ratio CR is set to 0.375. The frequency spectra of the original signals and reconstructed signals of the three types of faults are shown in the figure. Figure 7-9 As shown in the figure, for these three types of faults, the 5G-WCS algorithm of the present invention reconstructs the signal with the highest similarity to the original signal, that is, the method of the present invention has the best effect.

[0086] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A rolling bearing signal reconstruction method integrating 5G LDPC and compressed sensing, characterized in that the method include: Obtain the original vibration signal of the rolling bearing; Perform DCT transformation on the original vibration signal to obtain a frequency domain vibration signal; Perform weighted processing on the frequency domain vibration signal: set the threshold and weight based on the DCT coefficient of the original vibration signal to generate a weighted matrix, and calculate the weighted vibration signal based on the frequency domain vibration signal and the weighted matrix; The weighted vibration signal is subjected to compressed sensing processing through an observation matrix to obtain a weighted observation signal, wherein the observation matrix is ​​implemented through a 5G-NR-LDPC code; After the weighted observation signal is transmitted through the channel, the reconstructed vibration signal is obtained through signal reconstruction.

2. According to claim 1, a rolling bearing signal reconstruction method integrating 5G LDPC and compressed sensing is characterized in that: The calculation formula of the frequency domain vibration signal is: ξ=Ψx Among them, Ψ is a sparse matrix and also the DCT basis matrix.

3. According to claim 1, a rolling bearing signal reconstruction method integrating 5G LDPC and compressed sensing is characterized in that: The process of generating the weighting matrix includes: Generate weight factor q according to DCT coefficient distribution and preset threshold T i , weight factor q i The calculation formula is: q i =ExT(|ξ i |-T) Among them, ξ i is the i-th DCT coefficient of the original vibration signal, i = 1, 2, ..., N, N is the length of the original vibration signal; the function ExT is used to determine |ξ i |Is it greater than T? Based on the weight factor q i Construct the weighted matrix Q, the calculation formula is: Q=diag(q1,q2,...,q i ,...,q N )。 4. According to claim 3, a rolling bearing signal reconstruction method integrating 5G LDPC and compressed sensing is characterized in that: The calculation formula of function ExT is: Among them, z is the variable of function ExT.

5. According to claim 1, a rolling bearing signal reconstruction method integrating 5G LDPC and compressed sensing is characterized in that: The calculation formula of the weighted vibration signal is: Ξ=OΨΨθ Among them, Ξ is the weighted vibration signal, Q is the weighting matrix, θ is the sparse coefficient, and Ψ is the DCT basis matrix.

6. According to claim 1, a rolling bearing signal reconstruction method integrating 5G LDPC and compressed sensing is characterized in that: The calculation formula of the weighted observation signal is: Y=ΦΞ=A Q i Where Y is the weighted observation signal, Φ is the 5G-NR-LDPC code check matrix, which is also the observation matrix, Ξ is the weighted vibration signal, and A Q is the weighted perception matrix, and θ is the sparse coefficient.

7. The rolling bearing signal reconstruction method integrating 5G LDPC and compressed sensing according to claim 6 is characterized in that: The number of columns in the observation matrix is ​​a power of 2.

Images