Bearing fault diagnosis method based on single channel time-space intrinsic mode decomposition

Abstract

The present invention provides a bearing fault diagnosis method based on single-channel spatiotemporal intrinsic modal decomposition, including collecting one-dimensional signals through a single-channel signal collector for analysis to obtain modal variables; screening the modal variables to form a new multi-channel signal; analyzing the multi-channel signal to obtain a mixed matrix B; selecting an initial phase function to decompose a part of the restored source signal in the mixed matrix B to obtain an unmixed signal; and performing frequency domain analysis and envelope spectrum analysis on the unmixed signal to obtain mixed features and fault impact features. The method can decompose the frequency conversion features and impact features from the single-channel signal. When processing high-noise simulation signals, compared with traditional variational modal decomposition, it can separate source signal components with larger correlation coefficients and impact components with larger kurtosis.

Description

Technical Field

[0001] The invention relates to the field of mechanical fault diagnosis, and in particular to a bearing fault diagnosis method based on single-channel spatiotemporal intrinsic mode decomposition. Background Art

[0002] With the development of intelligent manufacturing, industrial robots, etc., research on mechanical fault diagnosis is more urgent. Rotating parts faults account for the vast majority of mechanical faults. In recent years, many fault diagnosis methods have been applied to actual production, but bearing motion signals are generally multi-source signals and are mixed with a certain degree of noise, resulting in insignificant decomposition effects. Traditional modal decomposition fault diagnosis methods are often applied to one-dimensional signals to decompose modal components, and cannot effectively decompose multidimensional signals into their vibration signal components, that is, it is difficult to handle the single-channel blind source separation problem, and the analysis of complex single-channel multi-component noisy non-stationary bearing motion signals needs to be further studied. Hou and Shi proposed data-driven time-frequency analysis (DDTFA) based on adaptive time-frequency analysis (ATFA). This method uses Newton's iteration method to solve nonlinear problems and complete the sparsest representation of the signal in a dictionary composed of intrinsic mode functions. In 2020, Seth M. Hirsh et al. proposed the concept of spatiotemporal intrinsic mode decomposition based on data-driven time-frequency domain analysis combined with blind source separation methods. This method can separate the modulated signal component in the mixed signal. For the development of underdetermined blind source separation, that is, the number of observed signals is less than the number of source signals, Gou et al. proposed a solution to the single-channel blind source separation problem by combining local mean decomposition (LMD) and independent component analysis (ICA). This method needs to be strengthened in the face of high noise and other situations. Summary of the invention

[0003] The purpose of the present invention is to provide a bearing fault diagnosis method based on single-channel spatiotemporal intrinsic mode decomposition, which can decompose the rotation frequency characteristics and impact characteristics from the single-channel signal. When processing high-noise simulation signals, compared with traditional variational mode decomposition, it can separate source signal components with larger correlation coefficients and impact components with larger kurtosis.

[0004] To achieve the purpose of the present invention, the present invention adopts the following technical solutions:

[0005] According to one aspect of the present invention, a single-channel spatiotemporal intrinsic modal decomposition method is provided, comprising collecting a one-dimensional signal through a single-channel signal collector for analysis to obtain modal variables; screening the modal variables to form a new multi-channel signal; analyzing the multi-channel signal to obtain a mixing matrix B; selecting an initial phase function θ0 to decompose a portion of the restored source signal in the mixing matrix B to obtain an unmixed signal; and performing frequency domain analysis and envelope spectrum analysis on the unmixed signal to obtain mixing characteristics and fault impact characteristics.

[0006] According to one embodiment of the present invention, collecting a one-dimensional signal by a single-channel signal collector for analysis to obtain modal variables includes: collecting a one-dimensional signal by a single-channel signal collector, and analyzing the one-dimensional signal by an integrated empirical mode decomposition method to obtain multiple modal components.

[0007] According to an embodiment of the present invention, the one-dimensional signal is a signal of only one channel collector, including a time domain signal of a bearing vibration signal.

[0008] According to an embodiment of the present invention, screening modal variables to form a new multi-channel signal includes: finding a signal with a correlation coefficient close to the original one-dimensional signal among multiple modal variables to form a new multi-channel signal.

[0009] According to one embodiment of the present invention, the correlation coefficient is:

[0010]

[0011] Wherein, Cov(X,Y) is the covariance of X and Y, Var(X) is the variance of X, Var(Y) is the variance of Y, and the correlation coefficient is used to measure the correlation between two signals. The larger the correlation coefficient, the more correlated the signals are.

[0012] According to an embodiment of the present invention, analyzing the multi-channel signal to obtain the mixing matrix B includes: performing a fast independent component algorithm analysis on the multi-channel signal to obtain the mixing matrix B.

[0013] According to one embodiment of the present invention, selecting an initial phase function θ0 to decompose a partially restored source signal in a mixing matrix B to obtain an unmixed signal includes: restoring the mixed components of the source signal, selecting a suitable initial phase function θ0 to perform a nonlinear matching pursuit decomposition on the partially restored signal in the mixing matrix B, and obtaining each unmixed signal in a single-channel mixed signal.

[0014] According to one embodiment of the present invention, frequency domain analysis and envelope spectrum analysis are performed on the unmixed signal to obtain mixed characteristics and fault impact characteristics, including: taking the extreme value after Hilbert transforming the decomposed unmixed signal, then taking the envelope of the one-dimensional data obtained after taking the extreme value, connecting the extreme values ​​to obtain an envelope signal, and performing Fourier transform on the envelope signal to obtain corresponding data.

[0015] According to one embodiment of the present invention, the amplitude of the fault characteristic frequency in the spectrum diagram is small, and there is a fault frequency multiplication feature in the envelope spectrum diagram, that is, it can be seen that there are peaks at the fault frequency and its multiples at the same time. Compared with spectrum analysis, envelope spectrum analysis eliminates unnecessary frequency interference, highlights the fault characteristic frequency, and is easier to judge the fault type of the rolling bearing.

[0016] According to an embodiment of the present invention, if there is a frequency doubling feature in the envelope spectrum, there is an impact in the bearing motion signal.

[0017] An embodiment of the present invention has the following advantages or beneficial effects:

[0018] The single-channel spatiotemporal intrinsic mode decomposition method of the present invention can decompose the frequency conversion characteristics and the impact characteristics from the single-channel signal. When processing high-noise simulation signals, compared with the traditional variational mode decomposition, it can separate the source signal component with a larger correlation coefficient and the impact component with a larger kurtosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and other features and advantages of the present invention will become more apparent by describing in detail exemplary embodiments thereof with reference to the attached drawings.

[0020] Figure 1 The figure is a flow chart showing a single-channel spatiotemporal intrinsic mode decomposition method according to an exemplary embodiment.

[0021] Figure 2 FIG. 4 is a schematic diagram of a simulated signal mixing single-channel time domain according to an exemplary embodiment.

[0022] Figure 3 is a schematic diagram according to an exemplary embodiment.

[0023] Figure 4 is a schematic diagram showing how a second modal component is obtained by decomposing a simulation signal according to an exemplary embodiment.

[0024] Figure 5 is a schematic diagram showing a first modal component obtained by decomposing a simulation signal according to an exemplary embodiment.

[0025] Figure 6FIG. 1 is a schematic diagram showing a third modal component obtained by decomposing a simulation signal according to an exemplary embodiment.

[0026] Figure 7 It is a schematic diagram showing a frequency domain diagram of a first modal component obtained by decomposing a simulation signal according to an exemplary embodiment.

[0027] Figure 8 is a schematic diagram of a frequency domain diagram of a second modal component obtained by decomposing a simulation signal according to an exemplary embodiment.

[0028] Fig. 9 is a schematic diagram of an envelope diagram of a third modal component obtained by decomposing a simulation signal according to an exemplary embodiment.

[0029] Fig.10 is a schematic diagram of a curve fitting diagram showing the relationship between the signal-to-noise ratio and the average correlation coefficient according to an exemplary embodiment.

[0030] Fig.11 is a schematic diagram of a curve fitting diagram showing the relationship between signal-to-noise ratio and kurtosis according to an exemplary embodiment.

[0031] Fig.12 is a schematic diagram of a single-channel signal source function of experimental data according to an exemplary embodiment.

[0032] Fig.13 FIG. 4 is a schematic diagram showing the decomposition of the first modal component of test data according to an exemplary embodiment.

[0033] Fig.14 FIG. 4 is a schematic diagram showing decomposition of the second modal component of test data according to an exemplary embodiment.

[0034] Fig.15 FIG. 4 is a schematic diagram showing test data decomposed into a third modal component according to an exemplary embodiment.

[0035] Fig.16 FIG. 4 is a schematic diagram showing a frequency domain diagram of a first modal component of decomposed test data according to an exemplary embodiment.

[0036] Fig.17 FIG. 4 is a schematic diagram showing an envelope diagram of a third modal component of decomposed test data according to an exemplary embodiment. DETAILED DESCRIPTION

[0037] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that the present invention will be comprehensive and complete and fully convey the concepts of the example embodiments to those skilled in the art. The same reference numerals in the figures represent the same or similar structures, and thus their detailed description will be omitted.

[0038] The terms "a", "an", "the", and "said" are used to indicate the presence of one or more elements / components / etc.; the terms "comprising" and "having" are used to express an open-ended inclusive meaning and mean that additional elements / components / etc. may be present in addition to the listed elements / components / etc.

[0039] like Figure 1 As shown, Figure 1 A flow chart of a single-channel spatiotemporal intrinsic mode decomposition method provided by the present invention is shown.

[0040] A single-channel spatiotemporal intrinsic modal decomposition method according to an embodiment of the present invention comprises collecting a one-dimensional signal through a single-channel signal collector for analysis to obtain modal variables; screening the modal variables to form a new multi-channel signal; analyzing the multi-channel signal to obtain a mixing matrix B; selecting an initial phase function to decompose a part of the restored source signal in the mixing matrix B to obtain an unmixed signal; and performing frequency domain analysis and envelope spectrum analysis on the unmixed signal to obtain mixing characteristics and fault impact characteristics.

[0041] Among them, firstly, the single-channel bearing signal, that is, the signal with only one channel collector is a one-dimensional signal: the time domain signal of the bearing vibration signal, is subjected to integrated empirical mode decomposition EEMD, and the signal with the closest correlation coefficient to the single-channel mixed signal is found from the modal decomposition IMF obtained by decomposition to form a new multi-channel signal. The new mixed signal is processed by the time-space inherent mode decomposition method, that is, FastICA is used to calculate the mixing matrix and matching pursuit is used to find the fault frequency and impact characteristics. After decomposition, the modal components can be obtained, and the envelope spectrum of the modal components can be drawn to find the frequency doubling characteristics of the bearing fault frequency.

[0042] In a preferred embodiment of the present invention, collecting a one-dimensional signal by a single-channel signal collector for analysis to obtain a modal variable includes: using a single-channel signal collector to collect a one-dimensional signal, and analyzing the one-dimensional signal by an integrated empirical mode decomposition method to obtain multiple modal components. The one-dimensional signal is a signal of only one channel collector, including a time domain signal of a bearing vibration signal.

[0043] Among them, the EEMD algorithm adds Gaussian white noise on the basis of the EMD algorithm, which can alleviate the problem of modal aliasing in EMD. Its basic principles are as follows:

[0044] 1) Add normally distributed white noise n to the original signal x(t) m (t), denoted as

[0045] x m (t) = x(t) + n m (t)

[0046] 2) Perform EMD decomposition on the signal with white noise added to obtain the intrinsic mode component, denoted as IMFC n,m (n=1, 2, ..., N), where C n,m , is the nth IMF obtained after the mth decomposition.

[0047] 3) Repeat the above two steps and add normal white noise of different amplitudes.

[0048] 4) Calculate the average IMF obtained each time, which is the nth IMF, denoted as

[0049]

[0050] In a preferred embodiment of the present invention, screening the modal variables to form a new multi-channel signal includes: finding a signal with a correlation coefficient close to the original one-dimensional signal from multiple modal variables to form a new multi-channel signal. The correlation coefficient is:

[0051]

[0052] Among them, Cov(X,Y) is the covariance of X and Y, Var(X) is the variance of X, Var(Y) is the variance of Y, and the correlation coefficient is used to measure the degree of correlation between two signals. The larger the correlation coefficient, the more correlated the signals are.

[0053] like Figure 1 As shown, for mixed signal decomposition, the larger the correlation coefficient between the decomposed signal and the unmixed signal, the more accurate the decomposition effect of this method is.

[0054] In a preferred embodiment of the present invention, analyzing the multi-channel signal to obtain the mixing matrix B includes: performing a fast independent component algorithm analysis on the multi-channel signal to obtain the mixing matrix B.

[0055] Among them, fast independent component analysis, namely FastICA. The FastICA algorithm is as follows:

[0056] The observed signal x(t) is centered to set its mean to zero, and then whitened to obtain whitened data z.

[0057] Note: Centering refers to subtracting the mean of the signal, and whitening is used to remove redundancy and reflect the differences between signal components.

[0058] 2) Randomly select an initialization vector w, update w iteratively and normalize it, w←w / ‖w‖.

[0059] 3) Convergence judgment is performed on w. If it does not converge, the iteration continues. If it converges or meets the maximum number of iterations, the separation signal is output.

[0060] In a preferred embodiment of the present invention, selecting an initial phase function θ0 to decompose a partially restored source signal in the mixing matrix B to obtain an unmixed signal includes: restoring the mixed components of the source signal, selecting a suitable initial phase function θ0 to perform a nonlinear matching pursuit decomposition on the partially restored signal in the mixing matrix B, and obtaining each unmixed signal in the single-channel mixed signal.

[0061] like Figure 1 As shown, nonlinear matching pursuit (NMP)

[0062] In order to reduce the complexity of sparse solution, the principle of maximum projection energy is adopted to decompose the signal. This method is called matching pursuit. The calculation process of nonlinear matching pursuit is as follows:

[0063] The initial condition f is the initial signal, the initial phase function θ0, r0 = f, k = 1, θ k 0 =θ0.

[0064] 1) The nonlinear least squares algorithm uses the Gauss-Newton iteration method:

[0065]

[0066] 2) Update the residual:

[0067]

[0068] 3) Judgment conditions:

[0069]

[0070] If the condition is met, the iteration stops. If the condition is not met, k=k+1 and returns to step 1. Nonlinear matching pursuit uses a greedy algorithm to continuously approach the component value of the signal and the high-order nonlinear expression of its phase function through cyclic iteration. Since the Gauss-Newton iteration method is greatly affected by the initial trial value, the initial phase function greatly affects its convergence, resolution accuracy and efficiency, but the relative function error within a certain degree can still ensure the accuracy of this method.

[0071] In a preferred embodiment of the present invention, frequency domain analysis and envelope spectrum analysis are performed on the unmixed signal to obtain the mixed characteristics and fault impact characteristics, including: taking the extreme value after Hilbert transforming the decomposed unmixed signal, then taking the envelope of the one-dimensional data obtained after taking the extreme value, connecting the extreme values ​​to obtain the envelope signal, and performing Fourier transform on the envelope signal to obtain the corresponding data. The amplitude of the fault characteristic frequency in the spectrum diagram is small, and the amplitude of the fault characteristic frequency in the envelope spectrum diagram is large. Compared with the spectrum analysis, the envelope spectrum analysis eliminates unnecessary frequency interference, highlights the fault characteristic frequency, and is easier to judge the type of rolling bearing fault. If there is a frequency doubling feature in the envelope spectrum, there is an impact in the bearing motion signal.

[0072] like Figure 1 As shown in the figure, frequency domain analysis is performed using Fourier transform. Fourier transform:

[0073]

[0074] Fourier transform can convert time domain signals into frequency domain signals, which can better reflect the frequency components in the bearing motion signal. Envelope spectrum: After the signal is Hilbert transformed, the extreme value is taken, and then the envelope of the one-dimensional data obtained after taking the extreme value is taken, the extreme values ​​are connected to obtain the envelope signal, and the envelope signal is Fourier transformed to obtain the data. (The horizontal axis is frequency, and the vertical axis is amplitude) The envelope spectrum is sensitive to the fault of the impact event. The amplitude of the fault characteristic frequency in the spectrum diagram is small, and the amplitude of the fault characteristic frequency in the envelope spectrum diagram is very high and easy to identify. Therefore, compared with spectrum analysis, envelope spectrum analysis eliminates unnecessary frequency interference and can better highlight the fault characteristic frequency. According to the envelope spectrum diagram, it is easier to judge the type of rolling bearing fault. If there is a frequency doubling feature in the envelope spectrum, it means that there is an impact in the bearing motion signal, that is, there is a fault.

[0075] Variational mode decomposition (VMD), variational mode decomposition decomposes the original signal into modal components with different center frequencies and bandwidths through iteration. The sum of the estimated bandwidths of each modal component is minimized, and the sum of the modal components is equal to the input signal as a constraint. Its variational model is as follows:

[0076]

[0077] In the formula, is the partial derivative of t, δ(t) is the impulse function, μ k is the kth mode function, ω k is the center frequency of each mode, and f is the original signal. The quadratic penalty factor α and the Lagrange multiplication operator λ(t) are introduced to transform it into an unconstrained problem. The expression is as follows:

[0078]

[0079] In the formula, α can ensure the accuracy of the reconstructed signal, and λ(t) makes the constraint condition more stringent. Introduce the alternating power operator to solve and alternately update μ k (n+1) ,ω k (n+1) and λ k Solve the Lagrangian expression for saddle points.

[0080] The process is as follows:

[0081] 1) Initialization λ k and n

[0082] 2) Set the loop condition n=n+1

[0083] 3) When ω ≥ 0, update

[0084]

[0085] 4) Update k

[0086]

[0087] 5) Update λ

[0088]

[0089] 6) When the loop steps 2) to 5) satisfy the judgment expression, the iteration is terminated. The expression is as follows:

[0090]

[0091] Where e is the judgment accuracy.

[0092] Variational mode decomposition can decompose the signal f into K narrowband modal components μ_k, where k∈1,2,…,k.

[0093] Signal-to-Noise Ratio (SNR)

[0094]

[0095] Ps refers to signal power, and Pn refers to noise power. The signal-to-noise ratio can show the ratio of the effective signal to the noise signal in the signal component. The smaller the signal-to-noise ratio, the greater the noise.

[0096] Kurtosis (K)

[0097]

[0098] Kurtosis is sensitive to shock characteristics, and it is generally believed that a kurtosis exceeding 3.5 indicates a large shock and a fault.

[0099] The simulation is performed through the above formula. The details are as follows:

[0100] Assume there is a set of source signals s j , linearly mixed into the observed signal x j ,s j Both are IMF with inter-wave frequency modulation. The characteristic frequencies of the two source signals are f1 = 50 Hz and f2 = 120 Hz. Among them, the attenuation coefficient C = 700 and the frequency conversion f r =30Hz, natural frequency f n =4000Hz. Sampling data length N = 4096.

[0101] The following example creates three modes:

[0102] s1(t)=(cos(30πt)+1)sin(2πf1t))

[0103] s2(t)=cos(60πt+1)sin(2πf2t))

[0104]

[0105] Mixing matrix B = [0.3, 0.3, 0.4]. The time domain waveform and spectrum of the mixed simulation signal are obtained as follows Figure 1 As shown, there is no obvious impact feature in the time domain waveform, and there are frequency peaks of 50Hz and 120Hz in the frequency domain diagram. The signal will be decomposed to obtain its three source signals.

[0106] Reference Figures 2 to 3 It can be seen that Figure 1 It is the time domain diagram of the simulation signal, that is, the time domain diagram of the single-channel signal. Figure 2 This is the frequency domain diagram of the simulated signal, which is obtained by Fourier transform. It can be seen that the mixed signal contains 50Hz and 120Hz frequency components.

[0107] like Figures 4 to 6 As shown, the single-channel spatiotemporal intrinsic mode decomposition method is used for analysis, and the initial phase function is θ0(t) = 2πft, that is, θ1(t) = 100πt, θ2(t) = 240πt, θ3(t) = 3000πt. The three modal components obtained by the channel spatiotemporal intrinsic mode decomposition method, that is, the decomposition results, are analyzed. The number of sampling points on the horizontal axis of the picture is 4096 points, and the amplitude on the vertical axis is A, which represents the vibration characteristics of its motion signal.

[0108] Perform frequency domain analysis and envelope spectrum transformation on the decomposed results. Figures 7 to 9As shown, the first and second figures are frequency domain diagrams. It can be seen from the figures that the frequency components of 50Hz and 120Hz are consistent with the simulation signal components before mixing, which means that the decomposition is very accurate. The third figure is the envelope diagram, from which it can be seen that the fault characteristic frequency is the impact characteristic of 30Hz, which means that it can reflect the impact characteristics, that is, it can reflect the fault characteristics.

[0109] Compared with the traditional method, the advantages of this method are shown. Random noise is introduced into the original simulation signal. Figure 10 to Figure 11 As shown in the first figure, the horizontal axis is the signal-to-noise ratio, the vertical axis is the correlation coefficient, the solid line is the single-channel spatiotemporal inherent mode decomposition method, the dotted line is the variational mode decomposition method, and the picture is a scatter plot fitted to the fifth power.

[0110] The picture shows that in the case of high noise (negative signal-to-noise ratio), it is still possible to effectively separate the various mixed components in the mixed signal, and it is more accurate than traditional decomposition methods.

[0111] like Fig.11 As shown in the figure, the horizontal axis is the signal-to-noise ratio, the vertical axis is the kurtosis, the solid line is the single-channel spatiotemporal intrinsic mode decomposition method, the dotted line is the traditional variational mode decomposition method, and the third line is the method without any decomposition.

[0112] The picture shows that in the case of high noise, the impact characteristics can be masked by the noise, but the kurtosis value of the impact characteristics of the single-channel spatiotemporal intrinsic mode decomposition method has been maintained above 3.5, indicating that the impact characteristics can be reflected through the kurtosis index.

[0113] The simulation was verified by experiments. It includes a drive motor to provide power, and the entire equipment also includes a manual adjuster, speed monitor, flexible coupling, acceleration sensor, 2 experimental bearings and load gearbox.

[0114] The test bench diagram shows the test bearing dimensions as follows:

[0115]

[0116] like Figures 12 to 17 As shown, the motor speed f r =29.87Hz, sampling frequency f s =25.6kHz, select the number of points as 25600, and select the duration as 1s. The calculated characteristic frequency of the inner ring fault is f i =147.81Hz, the outer ring fault frequency is f o =91.15Hz. The horizontal coordinate of the test bench data collection is time, and the vertical coordinate is amplitude. The modal components are obtained by the single-channel time-space inherent modal decomposition method. Fig.16 As shown, it means that its frequency conversion component is 30Hz. Fig.17As shown, it has obvious fault frequency doubling characteristics, and one of the doubling frequencies is the outer ring fault frequency of 91.15Hz.

[0117] The example data proves that the single-channel spatiotemporal intrinsic mode decomposition method can accurately separate the mixed components in the bearing vibration signal, namely the rotation frequency component and the impact component. This method can accurately find the fault characteristics in the bearing vibration signal.

[0118] In the embodiments of the present invention, the term "plurality" refers to two or more than two, unless otherwise clearly defined. The terms "installation", "connection", "fixation" and the like should be understood in a broad sense. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection. For those skilled in the art, the specific meanings of the above terms in the embodiments of the present invention can be understood according to the specific circumstances.

[0119] In the description of the embodiments of the present invention, it needs to be understood that the directions or positional relationships indicated by the terms "upper" and "lower" are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the embodiments of the present invention and simplifying the description, rather than indicating or implying that the device or unit referred to must have a specific direction, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the embodiments of the present invention.

[0120] In the description of this specification, the description of the terms "one embodiment", "a preferred embodiment", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the embodiment of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0121] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A bearing fault diagnosis method based on single-channel spatiotemporal intrinsic mode decomposition, characterized in that: include: The one-dimensional signal is collected by a single-channel signal collector for analysis to obtain the modal variables; Screening modal variables to form new multi-channel signals; Analyze the multi-channel signal to obtain a mixing matrix B; Selecting an initial phase function θ0 to decompose the partially restored source signal in the mixing matrix B to obtain an unmixed signal; as well as The unmixed signal is subjected to frequency domain analysis and envelope spectrum analysis to obtain mixed features and fault impact features. The specific steps for performing this step are: taking the extreme value after Hilbert transforming the decomposed unmixed signal, then taking the envelope of the one-dimensional data obtained after taking the extreme value, connecting the extreme values ​​to obtain the envelope signal, and performing Fourier transform on the envelope signal to obtain the corresponding data. The amplitude of the fault characteristic frequency in the spectrum diagram is small, and there is a fault frequency multiplication feature in the envelope spectrum diagram. It is concluded that there are peaks at the fault frequency and its multiples. Compared with spectrum analysis, envelope spectrum analysis eliminates unnecessary frequency interference, highlights the fault characteristic frequency, and is easier to judge the type of rolling bearing fault.

2. The bearing fault diagnosis method based on single-channel spatiotemporal intrinsic mode decomposition according to claim 1 is characterized in that: Collecting a one-dimensional signal through a single-channel signal collector for analysis to obtain modal variables includes: using a single-channel signal collector to collect a one-dimensional signal, and analyzing the one-dimensional signal using an integrated empirical mode decomposition method to obtain multiple modal components.

3. The bearing fault diagnosis method based on single-channel spatiotemporal intrinsic mode decomposition according to claim 2 is characterized in that: The one-dimensional signal is a signal from only one channel collector, including a bearing vibration signal time domain signal.

4. The bearing fault diagnosis method based on single-channel spatiotemporal intrinsic mode decomposition according to claim 1 is characterized in that: Screening the modal variables to form a new multi-channel signal includes: finding a signal with a correlation coefficient close to the original one-dimensional signal in multiple modal variables to form a new multi-channel signal.

5. The bearing fault diagnosis method based on single-channel spatiotemporal intrinsic mode decomposition according to claim 4 is characterized in that: The correlation coefficient is: ； Wherein, Cov(X,Y) is the covariance of X and Y, Var(X) is the variance of X, Var(Y) is the variance of Y, and the correlation coefficient is used to measure the correlation between two signals. The larger the correlation coefficient, the more correlated the signals are.

6. The bearing fault diagnosis method based on single-channel spatiotemporal intrinsic mode decomposition according to claim 1 is characterized in that: Analyzing the multi-channel signal to obtain the mixing matrix B includes: performing a fast independent component algorithm analysis on the multi-channel signal to obtain the mixing matrix B.

7. The bearing fault diagnosis method based on single-channel spatiotemporal intrinsic mode decomposition according to claim 1 is characterized in that: Selecting an initial phase function θ0 to decompose the partially restored source signal in the mixing matrix B to obtain an unmixed signal includes: restoring the mixed components of the source signal, selecting a suitable initial phase function θ0 to decompose the partially restored signal in the mixing matrix B using a nonlinear matching pursuit method, and obtaining each unmixed signal in the single-channel mixed signal.

8. The bearing fault diagnosis method based on single-channel spatiotemporal intrinsic mode decomposition according to claim 1 is characterized in that: If there are frequency doubling features in the envelope spectrum, there is impact in the bearing motion signal.

Images