A signal sparse reconstruction method, device and application based on GI coefficient guidance

Abstract

The application relates to a signal sparse reconstruction method based on GI coefficient guidance, a device and application, and the method comprises the following steps: converting an original signal to be reconstructed which is collected into a Hankel matrix, and performing SVD decomposition on the Hankel matrix; calculating GI coefficient values of each component after SVD decomposition, screening the components according to a set GI coefficient threshold value, and obtaining a purified signal based on the screened components; setting algorithm parameters of a time-shifted K-SVD algorithm, reconstructing the purified signal by using the time-shifted K-SVD algorithm, and obtaining a reconstructed signal. Compared with the prior art, the application can effectively extract impact components in a vibration signal and inhibit noise and signal phase interference.

Description

TECHNICAL FIELD

[0001] The present application relates to the processing of mechanical fault signals, in particular to a signal sparse reconstruction method based on GI coefficient guidance, equipment and application. BACKGROUND

[0002] The mechanical equipment vibration signal contains rich fault information, but the effective fault impact component is often submerged by noise and interference, which is not conducive to the fault analysis of the equipment. Secondly, the vibration signal has the characteristics of non-stationary and non-linear. In order to extract the effective fault impact component from the original signal, improve the signal-to-noise ratio, and realize fault diagnosis, researchers have proposed many practical fault diagnosis methods. Sparse reconstruction, as a more concise and higher resolution algorithm for describing the inherent characteristics of the signal, is widely studied. The purpose of sparse representation is to represent the signal with as few atoms as possible in a given overcomplete dictionary, which facilitates further processing of the signal.

[0003] In the sparse representation model, the basis vector used to represent the signal, also known as the dictionary, plays a key role in the representation ability of the signal. The traditional dictionary construction method uses some signal transformation techniques to generate the dictionary, such as Fourier transform, discrete cosine transform, Gabor transform, wavelet transform, etc. The specific form of the transformation is selected artificially according to the characteristics of the signal. However, this method has many limitations and cannot obtain satisfactory sparse representation results. The dictionary learning method can be used to construct an empirical learning dictionary, providing more degrees of freedom for capturing the potential feature components of complex signals.

[0004] Among various dictionary learning methods such as optimal direction and generalized principal component analysis (GPCA), K-SVD has become the most commonly used feature extraction method due to its high efficiency. However, K-SVD still has some problems when used for fault diagnosis. First, in the sparse representation process based on K-SVD, the reconstructed signal is severely affected by the interference and noise components in the original fault signal. In addition, the traditional K-SVD method cannot produce a shift-invariant dictionary, which means that the learned atoms will be affected by the signal phase and background noise. Finally, K-SVD will obtain multiple atoms, but some of these atoms are useless for fault diagnosis and may even affect the results. SUMMARY

[0005] The purpose of the present application is to overcome the problems of being affected by the signal phase and background noise in the prior art, and to provide a signal sparse reconstruction method based on GI coefficient guidance, equipment and application, which can effectively extract the impact component in the vibration signal, suppress noise and signal phase interference, and facilitate subsequent signal analysis and equipment state evaluation.

[0006] The purpose of the present application can be achieved by the following technical solutions:

[0007] A signal sparse reconstruction method based on GI coefficient guidance, comprising the following steps:

[0008] The collected original signal to be reconstructed is converted into a Hankel matrix, and the Hankel matrix is subjected to SVD decomposition;

[0009] The GI coefficient values of each component after SVD decomposition are calculated, the components are screened according to a set GI coefficient threshold, and a purified signal is obtained based on the screened components;

[0010] The algorithm parameters of the time-shifted K-SVD algorithm are set, the purified signal is reconstructed by using the time-shifted K-SVD algorithm, and a reconstructed signal is obtained.

[0011] Further, the number of rows and columns of the Hankel matrix is determined according to the length of the original signal.

[0012] Further, the screening of the components according to the set GI coefficient threshold is specifically:

[0013] The components with GI coefficient values greater than the GI coefficient threshold are taken as the screened components.

[0014] Further, the obtaining of the purified signal based on the screened components is specifically:

[0015] The screened components are accumulated to form the purified signal.

[0016] Further, the GI coefficient threshold is determined according to the historical data richness, the equipment working condition and the result fault tolerance.

[0017] Further, the algorithm parameters include the number of non-zero sparse coefficients, the length of extracted atoms and the number of extracted atoms.

[0018] Further, the number of non-zero sparse coefficients is determined according to the fault characteristic frequency and the sampling time;

[0019] The length of the extracted atoms is determined according to the fault characteristic period and the sampling frequency;

[0020] The number of extracted atoms is determined according to the number of fault characteristic types.

[0021] The application also provides an electronic device, comprising:

[0022] One or more processors;

[0023] Memory; and

[0024] One or more programs stored in the memory, the one or more programs comprising instructions for performing the signal sparse reconstruction method as described above.

[0025] The application also provides a computer readable storage medium comprising one or more programs for execution by one or more processors of an electronic device, the one or more programs comprising instructions for performing the signal sparse reconstruction method as described above.

[0026] The application also provides a mechanical equipment fault diagnosis method, which extracts a signal of the mechanical equipment by using the signal sparse reconstruction method as described above, and realizes mechanical equipment fault diagnosis based on the signal.

[0027] Compared with the prior art, the application has the following beneficial effects:

[0028] 1. The application uses GI coefficient as a standard to purify the original signal, effectively reduces the influence of noise, and can be used in a worse signal-to-noise ratio environment, and can also extract effective fault features in a lower signal-to-noise ratio.

[0029] 2. The application first pre-processes the signal, converts the collected column signal into a Hankel matrix, and performs singular value decomposition on the converted Hankel matrix. After calculating the Gini coefficient of each component after singular value decomposition, the components meeting the requirements are screened out through a threshold value to form the purified column signal. Then, the purified signal is converted into a sparse representation problem, and the signal is reconstructed by using the time-shifted K-SVD algorithm. The characteristics of the signal itself can be accurately learned, the description ability of the dictionary atom to the signal is improved, and the signal characteristics are beneficial to be extracted.

[0030] 3. The method of the application can be applied to the reconstruction of vibration signals, encoder signals or other signals collected during the working process of bearings and gearboxes in an industrial scene to realize signal denoising and feature extraction, serve mechanical equipment fault prediction, and be beneficial to subsequent signal analysis and equipment state evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 is a flowchart of the application;

[0032] Figure 2 is a simulation signal diagram;

[0033] Figure 3 is a learning atom diagram using the proposed method;

[0034] Figure 4 is a signal reconstruction diagram using the proposed method;

[0035] Figure 5 is a learning atom diagram using time-shifted K-SVD;

[0036] Figure 6 is a signal reconstruction diagram using time-shifted K-SVD. DETAILED DESCRIPTION

[0037] The application will be described in detail below with reference to the drawings and specific embodiments. The embodiments are implemented on the premise of the technical solutions of the application, and detailed implementation modes and specific operation processes are given, but the protection scope of the application is not limited to the following embodiments.

[0038] Embodiment 1

[0039] As shown in the figure, the embodiment provides a signal sparse reconstruction method based on GI coefficient guidance, including the following steps: Figure 1

[0040] The collected original signal to be reconstructed is converted into a Hankel matrix, and the Hankel matrix is subjected to SVD decomposition;

[0041] The GI coefficient values of each component after SVD decomposition are calculated, the components are screened according to the set GI coefficient threshold, and a purified signal is obtained based on the screened components;

[0042] The algorithm parameters of the time-shifted K-SVD algorithm are set, including the number of non-zero sparse coefficients, the length of the extracted atom and the number of extracted atoms, the purified signal is reconstructed by using the time-shifted K-SVD algorithm, and a reconstructed signal is obtained.

[0043] In the above method, the number of rows and columns of the Hankel matrix is determined according to the length of the original signal; the GI coefficient threshold is determined according to the historical data richness, the equipment working condition and the result fault tolerance; the number of non-zero sparse coefficients is determined according to the fault feature frequency and the sampling time; the length of the extracted atom is determined according to the fault feature period and the sampling frequency; and the number of extracted atoms is determined according to the number of fault feature types.

[0044] In the above method, the purified signal is a column signal, and the acquisition method is specifically: the components with GI coefficient values greater than the GI coefficient threshold are taken as the screened components, and the screened components are accumulated to form the purified signal.

[0045] The above method uses GI coefficient as a standard to purify the original signal, effectively reduces the influence of noise, can extract effective fault features in the case of low signal-to-noise ratio, and improves the subsequent fault diagnosis accuracy.

[0046] The embodiment verifies the effectiveness of the method by using a constructed simulation signal.

[0047] In the embodiment, the simulation signal x(t) is composed of a periodic impact signal and a noise signal:

[0048] y(t)=x(t)+n(t)

[0049] ​

[0050] Wherein, y(t) is a constructed analog signal, x(t) is an analog fault feature component, n(t) is a Gaussian white noise, t is set to be in [0, 1.2]s, A = 0.75, f = 100Hz, θ = π / 2, T = 0.2, the signal-to-noise ratio is -8dB, and the sampling frequency is 8000Hz, as shown in FIG. 1, wherein the upper graph is an impact signal, and the lower graph is a synthesized signal. Figure 2 The process of reconstructing the above simulation signal includes:

[0051] Referring to FIG. 2, the process of reconstructing the above simulation signal includes: Figure 1

[0052] Step S1: constructing a Hankel matrix from the signal to be reconstructed, and performing SVD decomposition on the Hankel matrix.

[0053] The number of rows and the number of columns of the constructed Hankel matrix should not differ too much and should be determined according to the length of the signal. In this embodiment, the number of rows of the Hankel matrix is taken as 100 in consideration of the efficiency of SVD.

[0054] Step S2: calculating the GI coefficient values of each component after SVD decomposition of the Hankel matrix.

[0055] Gini Coefficient:

[0056]

[0057] Wherein, ||x||1 represents the l1 norm; and K represents the length of the signal.

[0058] Step S3: screening the signal according to the set GI coefficient threshold value and obtaining a purified signal.

[0059] In this embodiment, the GI coefficient threshold value is set to be 0.45. The components with GI coefficients greater than 0.45 are accumulated to form the purified signal.

[0060] Step S4: setting the number of non-zero sparse coefficients, the length of the extracted atom, and the number of extracted atoms.

[0061] In this embodiment, the number of non-zero sparse coefficients is set to be 14, the length of the extracted atom is set to be 300, and the number of extracted atoms is set to be 1.

[0062] Step S5: reconstructing the purified signal by using the time-shifted K-SVD algorithm to obtain a reconstructed signal sig_reconstruction.

[0063] In this embodiment, the time-shifted K-SVD method is compared with the method proposed in the present application. Figure 3 The atoms learned by the method proposed in the present application are shown in FIG. 3. Figure 5 ​Atoms learned by the moving-invariant K-SVD method. By comparison, the proposed method can effectively suppress noise interference, Figure 4 Signals reconstructed using the proposed method, Figure 6 Signals reconstructed using the moving-invariant K-SVD method. By comparison, the proposed method can more effectively and accurately extract fault features.

[0064] The above method, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present application, essentially or the part that contributes to the prior art, or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0065] Embodiment 2

[0066] The present embodiment provides a mechanical equipment fault diagnosis method, which applies the signal sparse reconstruction method as described in embodiment 1 to extract the vibration signal of the mechanical equipment, and realizes fault diagnosis based on the vibration signal.

[0067] The above describes the preferred embodiments of the present application in detail. It should be understood that those skilled in the art can make many modifications and changes to the present application without creative labor based on the concept of the present application. Therefore, any technical solutions obtained by logical analysis, reasoning or limited experiments based on the prior art according to the concept of the present application shall be within the protection scope determined by the claims.

Claims

1. A signal sparse reconstruction method based on GI coefficient guidance, characterized in that, The method comprises the following steps: transforming the collected original signal to be reconstructed into a Hankel matrix, and performing SVD decomposition on the Hankel matrix; calculating the GI coefficient values of each component after SVD decomposition, screening the components according to a set GI coefficient threshold, and accumulating the screened components to form a purified signal, wherein the GI coefficient threshold is determined according to the historical data richness, the equipment working condition and the result fault tolerance; setting algorithm parameters of a time-shifted K-SVD algorithm, and reconstructing the purified signal by using the time-shifted K-SVD algorithm to obtain a reconstructed signal; the algorithm parameters comprise the number of non-zero sparse coefficients, the extracted atom length and the number of extracted atoms; the number of non-zero sparse coefficients is determined according to the fault characteristic frequency and the sampling time; the extracted atom length is determined according to the fault characteristic period and the sampling frequency; the number of extracted atoms is determined according to the number of fault characteristic types.

2. The method of claim 1, wherein, The number of rows and columns of the Hankel matrix is determined according to the length of the original signal.

3. The method of claim 1, wherein, The screening of the components according to the set GI coefficient threshold is specifically: the components with GI coefficient values greater than the GI coefficient threshold are taken as the screened components.

4. An electronic device, comprising: comprise: one or more processors; a memory; and one or more programs stored in the memory, the one or more programs comprising instructions for performing the signal sparse reconstruction method according to any one of claims 1-3.

5. A computer readable storage medium, characterized in that, comprise one or more programs for one or more processors of an electronic device to execute, the one or more programs comprising instructions for performing the signal sparse reconstruction method according to any one of claims 1-3.

6. A mechanical equipment failure diagnosis method characterized by comprising: application of the signal sparse reconstruction method according to any one of claims 1-3 to extract the signal of the mechanical equipment, and implementation of the mechanical equipment fault diagnosis based on the signal.

Images