Blind source separation method, device, equipment, medium and program product

Abstract

The invention discloses a blind source separation method, device and equipment, a medium and a program product, and relates to the technical field of fault diagnosis, and the method comprises the steps: obtaining a vibration signal; based on a preset embedding dimension range and time delay, performing phase-space reconstruction of different embedding dimensions on the vibration signal to obtain a first high-dimensional signal, and determining a target linear region, a target embedding dimension and target delay time for the first high-dimensional signal by using a K-center point clustering algorithm; performing phase-space reconstruction on the vibration signal based on the target embedded dimension and the target delay time to obtain a second high-dimensional signal, and performing rapid independent component analysis on the second high-dimensional signal to obtain a plurality of decomposed signals; and calculating the permutation entropy of the plurality of decomposition signals, judging and processing the permutation entropy and the corresponding decomposition signals by using a minimum permutation entropy judgment algorithm, and outputting a target decomposition signal. According to the invention, the K-center point clustering algorithm is adopted to determine the target linear region, and the effect that the target linear region does not need to be determined manually is achieved.

Description

Technical Field

[0001] The present application relates to the field of fault diagnosis technology, and in particular to a blind source separation method, device, equipment, medium and program product. Background Art

[0002] In rotating machinery, due to the noise caused by multi-stage transmission and external environment, the collected vibration signals are full of interference errors. The main functions of rotating machinery are completed by rotating parts. Whether it is the defects of the rotating machinery itself or the problems of the support system and connection caused by unreasonable installation, it will cause abnormal vibration during the operation of the rotating machinery. These abnormal vibrations and noises are the main characteristics of rotating machinery failures, and their vibration signals reflect the fault information of the machine from the amplitude domain, frequency domain and time domain. However, in practical applications, due to the complexity of the rotating machinery structure and the harshness of the operating environment, the vibration signals collected on site are often the aliasing of multiple vibration source signals, and carry strong background noise and measurement noise related to the data acquisition system. These noise interferences greatly increase the difficulty of extracting effective and sensitive features from the original signal. In order to effectively extract fault information and improve the accuracy of subsequent fault analysis and diagnosis, researchers have proposed a variety of noise reduction methods. Among them, blind source separation technology is an important means. Blind source separation technology can decompose multiple observation signals into several components that are as independent as possible according to the principle of statistical independence through optimization algorithms when both the source signal and the propagation characteristics are unknown, so as to obtain an estimate of the source signal.

[0003] In the fault diagnosis of rotating machinery, the blind source separation technology can be used to separate the useful signal and background noise in the vibration signal. Since the source signals are considered to be independent of each other, such as independent component analysis (ICA), the source signals are separated from their mixture. In addition to signal independence, another requirement of independent component analysis is that the number of mixed signals should not be less than the number of source signals, but in actual engineering applications, most monitoring points usually only have a single channel of monitoring data. For the underdetermined vibration signal with only one or fewer dimensions, the main problem is how to expand the single-channel observation mixed signal to a multi-channel mixed signal. Phase space reconstruction (PSR) can provide a method to project low-dimensional space signals to high-dimensional space signals. The phase space of the reconstructed time series can be used to describe the vibration state and characteristics of any mechanical system.

[0004] In the blind source separation method based on phase space reconstruction, it is necessary to determine the embedding dimension and correlation dimension according to the target linear region of the double logarithmic curve, so as to carry out the subsequent phase space reconstruction. However, the target linear region of the double logarithmic curve is currently mainly determined manually, which is easily affected by subjective judgment and produces errors, resulting in poor blind source separation effect.

[0005] Therefore, a new blind source separation method is urgently needed to solve the problem that the target linear region depends on artificial determination in the blind source separation method based on phase space reconstruction.

[0006] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0007] The main purpose of the present application is to provide a blind source separation method, device, equipment, medium and program product, aiming to solve the technical problem that the target linear region relies on artificial determination in the blind source separation method based on phase space reconstruction.

[0008] To achieve the above object, the present application proposes a blind source separation method, which comprises:

[0009] Obtain vibration signal;

[0010] A determination step, based on a preset embedding dimension range and time delay, reconstructing the phase space of the vibration signal with different embedding dimensions to obtain a first high-dimensional signal, and determining a target linear region for the first high-dimensional signal using a K-center point clustering algorithm, and determining a target embedding dimension and a target delay time based on the target linear region;

[0011] a decomposition step, reconstructing the vibration signal in phase space based on the target embedding dimension and the target delay time to obtain a second high-dimensional signal, and performing fast independent component analysis FastICA on the second high-dimensional signal to obtain multiple independent decomposed signals;

[0012] The permutation entropies of the plurality of decomposition signals are calculated, and the permutation entropies and the corresponding decomposition signals are judged and processed using a minimum permutation entropy judgment algorithm, and a target decomposition signal is output.

[0013] In one embodiment, the determining step includes:

[0014] Based on the preset embedding dimension range and the time delay, reconstructing the vibration signal in phase space with different embedding dimensions to obtain the first high-dimensional signal, and calculating the correlation dimension of the first high-dimensional signal;

[0015] Based on the correlation dimension, drawing a double logarithmic curve within the preset embedding dimension range;

[0016] Processing the double logarithmic curve using scatter point transformation combined with a K-center point clustering algorithm to obtain multiple target clusters;

[0017] Determine a cluster with the most stable change rate among the plurality of target clusters to obtain the target linear region;

[0018] Based on the target linear region, the target embedding dimension and the target delay time are determined.

[0019] In one embodiment, the step of processing the double logarithmic curve using scatter point transformation in combination with a K-center point clustering algorithm to obtain multiple target clusters includes:

[0020] Performing scatter point transformation on the double logarithmic curve to obtain a non-determined region;

[0021] Setting a plurality of initial center points in the non-fixed area;

[0022] A clustering step, performing clustering processing according to the distance between each discrete point in the non-fixed area and each of the initial center points, to obtain multiple clusters corresponding to multiple initial center points;

[0023] An updating step, updating the positions of the multiple initial center points so that the sum of distances from points in the same cluster to the corresponding new center point is minimized;

[0024] The clustering step and the updating step are repeated until the first iteration number is reached or the position of the new center point is fixed, thereby obtaining a plurality of target clusters.

[0025] In one embodiment, before the determining step, the following steps are included:

[0026] The calculation step is to calculate the time delay by using a cross-correlation method based on a preset embedding dimension range.

[0027] In one embodiment, the step of calculating the permutation entropy of the plurality of decomposition signals, judging and processing the permutation entropy and the corresponding decomposition signal using a minimum permutation entropy judgment algorithm, and outputting a target decomposition signal comprises:

[0028] a permutation entropy calculation step, performing permutation entropy calculation on a plurality of the decomposed signals to obtain a permutation entropy of each of the decomposed signals;

[0029] a sorting step, sorting and screening the plurality of permutation entropies to obtain a minimum permutation entropy;

[0030] If the minimum permutation entropy is less than or equal to the expected entropy value, confirming that the decomposition signal corresponding to the minimum permutation entropy is the target decomposition signal, and outputting the target decomposition signal;

[0031] If the minimum permutation entropy is greater than the expected entropy value, the calculation step, the determination step, the decomposition step, the permutation entropy calculation step and the sorting step are repeated until the number of repeated cycles reaches a preset number or the minimum permutation entropy is less than or equal to the expected entropy value, then the decomposition signal corresponding to the minimum permutation entropy is confirmed to be the target decomposition signal, and the target decomposition signal is output.

[0032] In one embodiment, the decomposition step comprises:

[0033] Based on the target embedding dimension and the target delay time, reconstructing the vibration signal in phase space to obtain a second high-dimensional signal;

[0034] Performing a decentralization process on the second high-dimensional signal to obtain a signal to be demixed;

[0035] Initialization step, randomly initializing the unmixing matrix of the signal to be unmixed;

[0036] an iterative step, iterating the unmixing matrix according to an unmixing matrix update formula and in combination with orthogonalization processing to obtain an updated unmixing matrix;

[0037] In the second iteration number, if the updated unmixing matrix does not converge, repeating the iterative step until the updated unmixing matrix converges; if the updated unmixing matrix converges, repeating the initialization step and the iterative step until all unmixing matrices are traversed;

[0038] Based on the entire unmixing matrix and the second high-dimensional signal, a plurality of independent decomposed signals are calculated.

[0039] In addition, to achieve the above purpose, the present application also proposes a blind source separation device, the blind source separation device comprising:

[0040] An acquisition module, used for acquiring vibration signals;

[0041] A determination module, used for determining the step, based on a preset embedding dimension range and time delay, reconstructing the phase space of the vibration signal with different embedding dimensions to obtain a first high-dimensional signal, and determining a target linear region for the first high-dimensional signal using a K-center point clustering algorithm, and determining a target embedding dimension and a target delay time based on the target linear region;

[0042] A decomposition module, used for the decomposition step, based on the target embedding dimension and the target delay time, reconstructing the vibration signal in phase space to obtain a second high-dimensional signal, and performing fast independent component analysis FastICA on the second high-dimensional signal to obtain multiple independent decomposed signals;

[0043] The output module is used to calculate the permutation entropy of the multiple decomposition signals, and use the minimum permutation entropy judgment algorithm to judge and process the permutation entropy and the corresponding decomposition signal, and output the target decomposition signal.

[0044] In addition, to achieve the above-mentioned purpose, the present application also proposes a blind source separation device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the blind source separation method described above.

[0045] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the blind source separation method described above are implemented.

[0046] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the blind source separation method described above are implemented.

[0047] One or more technical solutions proposed in this application have at least the following technical effects:

[0048] The technical solution of the present application adopts a blind source separation method, which includes: obtaining a vibration signal; a determination step, based on a preset embedding dimension range and time delay, reconstructing the vibration signal in phase space with different embedding dimensions to obtain a first high-dimensional signal, and determining a target linear region for the first high-dimensional signal using a K-center point clustering algorithm, and determining a target embedding dimension and a target delay time based on the target linear region; a decomposition step, based on the target embedding dimension and the target delay time, reconstructing the vibration signal in phase space to obtain a second high-dimensional signal, and performing fast independent component analysis FastICA on the second high-dimensional signal to obtain multiple independent decomposition signals; calculating the permutation entropy of the multiple decomposition signals, and using a minimum permutation entropy judgment algorithm to judge and process the permutation entropy and the corresponding decomposition signal, and outputting a target decomposition signal.

[0049] The technical solution of the present application performs preliminary phase space reconstruction of the vibration signal based on a preset embedding dimension range and time delay to obtain a first high-dimensional signal, and determines the target linear region by using the K-center point clustering algorithm for the first high-dimensional signal. Compared with the defect that the method of manually determining the target linear region is prone to errors due to subjective judgment, the K-center point clustering algorithm can accurately and efficiently determine the target linear region, thereby obtaining a more appropriate and accurate target embedding dimension and target delay time based on the target linear region. Based on the appropriate and accurate target embedding dimension and target delay time, the vibration signal is reconstructed in phase space to obtain a second high-dimensional signal, and the second high-dimensional signal is subjected to fast independent component analysis FastICA, thereby obtaining multiple independent decomposition signals, achieving a more accurate blind source separation effect. Afterwards, the permutation entropy of multiple decomposition signals is calculated. Since the entropy value of the permutation entropy can reflect the amount of noise contained in the decomposition signal, the minimum permutation entropy judgment algorithm is used to judge and process the permutation entropy and the corresponding decomposition signal. It is possible to automatically and quickly judge and output the target decomposition signal with the least noise content among multiple decomposition signals, without the need to manually select the target decomposition signal from multiple decomposition signals, thereby improving the efficiency of obtaining the target decomposition signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0051] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0052] Figure 1 A schematic diagram of a process flow provided for Embodiment 1 of the blind source separation method of the present application;

[0053] Figure 2 A schematic diagram of a flow chart provided for Embodiment 2 of the blind source separation method of the present application;

[0054] Figure 3 An example diagram of a double logarithmic curve of a first high-dimensional signal provided in Example 2 of the blind source separation method of the present application;

[0055] Figure 4 For Figure 3 The scatter diagram of the double logarithmic curve after scatter transformation;

[0056] Figure 5 For Figure 4 Schematic diagram of clustering after clustering the non-fixed regions in;

[0057] Figure 6 A relationship diagram between the correlation dimension and the embedding dimension provided in Example 2 of the blind source separation method of the present application;

[0058] Figure 7 A schematic diagram of a process flow provided for Embodiment 3 of the blind source separation method of the present application;

[0059] Figure 8 This is a schematic diagram of the module structure of the blind source separation device according to an embodiment of the present application;

[0060] Fig. 9 Schematic diagram of the device structure of the hardware operating environment involved in the blind source separation method in the embodiment of the present application.

[0061] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0062] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0063] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0064] The main solution of the embodiment of the present application is: obtaining a vibration signal; a determination step, based on a preset embedding dimension range and time delay, reconstructing the phase space of the vibration signal with different embedding dimensions to obtain a first high-dimensional signal, and using a K-center point clustering algorithm to determine a target linear region for the first high-dimensional signal, and determining a target embedding dimension and a target delay time based on the target linear region; a decomposition step, based on the target embedding dimension and the target delay time, reconstructing the phase space of the vibration signal to obtain a second high-dimensional signal, and performing fast independent component analysis FastICA on the second high-dimensional signal to obtain multiple independent decomposition signals; calculating the permutation entropy of the multiple decomposition signals, and using a minimum permutation entropy judgment algorithm to judge and process the permutation entropy and the corresponding decomposition signal, and outputting a target decomposition signal.

[0065] In this embodiment, for ease of description, the following description is made with the computing terminal as the execution subject.

[0066] In rotating machinery, due to the noise caused by multi-stage transmission and external environment, the collected vibration signals are full of interference errors. The main functions of rotating machinery are completed by rotating parts. Whether it is the defects of the rotating machinery itself or the problems of the support system and connection caused by unreasonable installation, it will cause abnormal vibration during the operation of the rotating machinery. These abnormal vibrations and noises are the main characteristics of rotating machinery failures, and their vibration signals reflect the fault information of the machine from the amplitude domain, frequency domain and time domain. However, in practical applications, due to the complexity of the rotating machinery structure and the harshness of the operating environment, the vibration signals collected on site are often the aliasing of multiple vibration source signals, and carry strong background noise and measurement noise related to the data acquisition system. These noise interferences greatly increase the difficulty of extracting effective and sensitive features from the original signal. In order to effectively extract fault information and improve the accuracy of subsequent fault analysis and diagnosis, researchers have proposed a variety of noise reduction methods. Among them, blind source separation technology is an important means. Blind source separation technology can decompose multiple observation signals into several components that are as independent as possible according to the principle of statistical independence through optimization algorithms when both the source signal and the propagation characteristics are unknown, so as to obtain an estimate of the source signal.

[0067] In the fault diagnosis of rotating machinery, the blind source separation technology can be used to separate the useful signal and background noise in the vibration signal. Since the source signals are considered to be independent of each other, such as independent component analysis (ICA), the source signals are separated from their mixture. In addition to signal independence, another requirement of independent component analysis is that the number of mixed signals should not be less than the number of source signals, but in actual engineering applications, most monitoring points usually only have a single channel of monitoring data. For the underdetermined vibration signal with only one or fewer dimensions, the main problem is how to expand the single-channel observation mixed signal to a multi-channel mixed signal. Phase space reconstruction (PSR) can provide a method to project low-dimensional space signals to high-dimensional space signals. The phase space of the reconstructed time series can be used to describe the vibration state and characteristics of any mechanical system.

[0068] In the blind source separation method based on phase space reconstruction, it is necessary to determine the embedding dimension and correlation dimension according to the target linear region of the double logarithmic curve, so as to carry out the subsequent phase space reconstruction. However, the target linear region of the double logarithmic curve is currently mainly determined manually, which is easily affected by subjective judgment and produces errors, resulting in poor blind source separation effect.

[0069] Therefore, a new blind source separation method is urgently needed to solve the problem that the target linear region depends on artificial determination in the blind source separation method based on phase space reconstruction.

[0070] Based on this, the present application provides a solution by adopting a blind source separation method, the blind source separation method comprising: obtaining a vibration signal; a determination step, based on a preset embedding dimension range and time delay, reconstructing the vibration signal in phase space with different embedding dimensions to obtain a first high-dimensional signal, and determining a target linear region for the first high-dimensional signal using a K-center point clustering algorithm, and determining a target embedding dimension and a target delay time based on the target linear region; a decomposition step, based on the target embedding dimension and the target delay time, reconstructing the vibration signal in phase space to obtain a second high-dimensional signal, and performing fast independent component analysis FastICA on the second high-dimensional signal to obtain multiple independent decomposition signals; calculating the permutation entropy of the multiple decomposition signals, and using a minimum permutation entropy judgment algorithm to judge and process the permutation entropy and the corresponding decomposition signal, and outputting a target decomposition signal.

[0071] The technical solution of the present application performs preliminary phase space reconstruction of the vibration signal based on a preset embedding dimension range and time delay to obtain a first high-dimensional signal, and determines the target linear region by using the K-center point clustering algorithm for the first high-dimensional signal. Compared with the defect that the method of manually determining the target linear region is prone to errors due to subjective judgment, the K-center point clustering algorithm can accurately and efficiently determine the target linear region, thereby obtaining a more appropriate and accurate target embedding dimension and target delay time based on the target linear region. Based on the appropriate and accurate target embedding dimension and target delay time, the vibration signal is reconstructed in phase space to obtain a second high-dimensional signal, and the second high-dimensional signal is subjected to fast independent component analysis FastICA, thereby obtaining multiple independent decomposition signals, achieving a more accurate blind source separation effect. Afterwards, the permutation entropy of multiple decomposition signals is calculated. Since the entropy value of the permutation entropy can reflect the amount of noise contained in the decomposition signal, the minimum permutation entropy judgment algorithm is used to judge and process the permutation entropy and the corresponding decomposition signal, so that the target decomposition signal with the least noise content can be automatically and quickly judged and output from multiple decomposition signals. There is no need to manually select the target decomposition signal from multiple decomposition signals, thereby improving the efficiency of obtaining the target decomposition signal; the target decomposition signal is the source signal with less noise content that is ultimately required.

[0072] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a computing terminal, etc. The following takes a computing terminal as an example to illustrate this embodiment and the following embodiments.

[0073] Based on this, the present application embodiment provides a blind source separation method, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the blind source separation method of the present application.

[0074] In this embodiment, the blind source separation method includes steps S10 to S50:

[0075] Step S10, obtaining a vibration signal;

[0076] It should be noted that the vibration signal can be a low-dimensional vibration signal, such as a one-dimensional vibration signal, a two-dimensional vibration signal, etc., or a high-dimensional vibration signal. Since this method is a blind source separation method based on phase space reconstruction, and phase space reconstruction can expand a single-channel observation mixed signal (i.e., a low-dimensional vibration signal) to a multi-channel mixed signal (i.e., a high-dimensional vibration signal), this method is particularly suitable for performing blind source separation processing on low-dimensional, mixed vibration signals.

[0077] Step S30, a determination step, based on a preset embedding dimension range and a time delay, reconstructing the phase space of the vibration signal with different embedding dimensions to obtain a first high-dimensional signal, and determining a target linear region for the first high-dimensional signal using a K-center point clustering algorithm, and determining a target embedding dimension and a target delay time based on the target linear region;

[0078] It should be noted that the delay time and embedding dimension need to be determined in phase space reconstruction, where the embedding dimension m should be greater than the system degree of freedom D, that is, m ≥ 2D + 1, so that the matrix dimension is large enough to obtain sufficient dynamic information. Given a time series x(t), the reconstructed phase space X is defined as:

[0079] X=[x(t),x(t+τ),…,x{t+(m-1)τ}] T

[0080] Where τ is the delay time and m is the embedding dimension.

[0081] In step S30, the vibration signal is first reconstructed in phase space with different embedding dimensions based on the preset embedding dimension range and time delay, and the first high-dimensional signal is preliminarily reconstructed. The target linear region of the first high-dimensional signal is then determined using the K-center point clustering algorithm, and the target embedding dimension and target delay time are obtained. The target embedding dimension and target delay time obtained at this time are more accurate than the previous preset embedding dimension range and time delay, which is helpful to obtain the second high-dimensional signal with better phase space reconstruction effect in the subsequent process.

[0082] In addition, it should be noted that the K-medoids clustering algorithm is a partition-based clustering method that uses actual data points as representatives of clusters to divide different data points into several clusters. Since the K-medoids clustering algorithm uses actual data points as center points, it is insensitive to noise and outliers and is suitable for noisy data sets, such as the first high-dimensional signal mixed with noise in this application.

[0083] Step S40, a decomposition step, based on the target embedding dimension and the target delay time, performing phase space reconstruction on the vibration signal to obtain a second high-dimensional signal, and performing fast independent component analysis FastICA on the second high-dimensional signal to obtain multiple independent decomposed signals;

[0084] It should be noted that Fast Independent Component Analysis, namely FastICA, is an efficient Independent Component Analysis (ICA) algorithm, which can separate independent source signals from multi-channel mixed signals, that is, separate and obtain multiple independent decomposition signals.

[0085] Step S50, calculating the permutation entropies of the plurality of decomposition signals, and using a minimum permutation entropy judgment algorithm to judge and process the permutation entropies and the corresponding decomposition signals, and outputting a target decomposition signal.

[0086] It should be noted that the permutation entropy can quantitatively evaluate the mutation relationship of the time series. The larger the permutation entropy value, the higher the complexity, which means that the sequence to be tested has a higher disorder, that is, it contains higher noise and needs further processing.

[0087] This embodiment provides a blind source separation method, which performs preliminary phase space reconstruction on the vibration signal based on a preset embedding dimension range and time delay to obtain a first high-dimensional signal, and determines the target linear region by using the K-center point clustering algorithm on the first high-dimensional signal. Compared with the defect that the method of manually determining the target linear region is prone to errors due to subjective judgment, the K-center point clustering algorithm can accurately and efficiently determine the target linear region, thereby obtaining a more appropriate and accurate target embedding dimension and target delay time based on the target linear region. Based on the appropriate and accurate target embedding dimension and target delay time, the vibration signal is reconstructed in phase space to obtain a second high-dimensional signal, and the second high-dimensional signal is subjected to fast independent component analysis FastICA, thereby obtaining multiple independent decomposition signals, and achieving a more accurate blind source separation effect. Afterwards, the permutation entropy of multiple decomposition signals is calculated. Since the entropy value of the permutation entropy can reflect the amount of noise contained in the decomposition signal, the minimum permutation entropy judgment algorithm is used to judge and process the permutation entropy and the corresponding decomposition signal, so that the target decomposition signal with the least noise content can be automatically and quickly judged and output from multiple decomposition signals. There is no need to manually select the target decomposition signal from multiple decomposition signals, thereby improving the efficiency of obtaining the target decomposition signal; the target decomposition signal is the source signal with less noise content that is ultimately required.

[0088] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can refer to the above introduction, and will not be repeated later. Figure 2 , step S30 includes steps S31 to S35:

[0089] Step S31, based on the preset embedding dimension range and the time delay, reconstructing the phase space of the vibration signal with different embedding dimensions to obtain the first high-dimensional signal, and calculating the correlation dimension of the first high-dimensional signal;

[0090] It should be noted that the calculation process of the correlation dimension can refer to the following method: calculate the correlation integral C(k) of the phase space reconstruction under different embedding dimensions, take the logarithm to get ln(C(k)). The relationship between the correlation dimension d and the correlation integral C(k) satisfies the logarithmic linear relationship:

[0091]

[0092] Among them, r is the interval step size.

[0093] Step S32, drawing a double logarithmic curve within the preset embedding dimension range based on the correlation dimension;

[0094] In step S32, according to the correlation dimension, that is, the corresponding relationship between ln(C(k)) and ln(r), a double logarithmic curve within the embedding dimension range is preset. The double logarithmic curve obtained by drawing can be referred to Figure 3 .

[0095] Step S33, processing the double logarithmic curve using scatter point conversion combined with a K-center point clustering algorithm to obtain a plurality of target clusters;

[0096] In an optional implementation, step S33 may include steps S331 to S335:

[0097] Step S331, performing scatter point conversion on the double logarithmic curve to obtain a non-fixed area;

[0098] In step S331, the double logarithmic curve is converted into discrete points by using scatter point conversion, so that the clustering algorithm can process the discrete points and divide each discrete point into a cluster of the corresponding category. Figure 4 .

[0099] Step S332, setting a plurality of initial center points in the non-fixed area;

[0100] It should be noted that the number of initial medoids needs to be set before clustering begins. For example, the number of initial medoids is set to 3, that is, there are 3 initial medoids.

[0101] Step S333, a clustering step, performing clustering processing according to the distance between each discrete point in the non-fixed area and each of the initial center points, to obtain multiple clusters corresponding to multiple initial center points;

[0102] The specific method of clustering processing is: calculating the distance from each discrete point to each initial center point medoids in the non-fixed area, and classifying each discrete point into the cluster to which the initial center point medoids with the smallest distance belongs according to the size of the distance.

[0103] Step S334, an updating step, updating the positions of the multiple initial center points so that the sum of the distances from the points in the same cluster to the corresponding new center point is minimized;

[0104] Step S335, repeating the clustering step and the updating step until the first iteration number is reached or the position of the new center point is fixed, thereby obtaining a plurality of target clusters.

[0105] For ease of understanding, please refer to Figure 5 , Figure 5 For Figure 4 The clustering diagram of the non-fixed area in the example is shown in Figure 5 In the figure, there are three target clusters. The discrete points of different target clusters are marked with different colors. The discrete points surrounded by red circles in each target cluster are the new center points obtained by the final iteration.

[0106] In this embodiment, multiple initial center points are set and clustering is performed to obtain multiple clusters, and the initial center points are updated, and the clustering steps and updating steps are repeated to obtain multiple target clusters. By adopting clustering processing and iteratively updating the center points, each discrete point can be classified into a suitable cluster to which it belongs, and multiple suitable target clusters are obtained. Since the K-center point clustering algorithm uses actual data points as center points, it has high robustness to noise and outliers in discrete points in non-fixed areas, and achieves effective clustering division.

[0107] Step S34, determining a cluster with the most stable change rate among the plurality of target clusters, and obtaining the target linear region;

[0108] The rate of change refers to the slope of the double logarithmic curve corresponding to the discrete points in the target cluster, that is, the correlation dimension.

[0109] In step S34, as an optional implementation, the GP algorithm (Grassberger-Procaccia algorithm) can be used to determine the cluster with the most stable change rate among multiple target clusters. The GP algorithm is a method for estimating the correlation dimension of a time series. Its basic principle is to estimate the dimension of the attractor by calculating the distance distribution between point pairs in the phase space at different scales. In step S34, the GP algorithm estimates the correlation dimension by calculating the distance distribution between discrete point pairs in a non-fixed region under different embedding dimensions to obtain the change rate of different target clusters. Figure 5 Taking the example in as an example, according to the GP algorithm, it can be obtained that the second type of target cluster in the figure (that is, the cluster area containing green discrete points) is the target cluster with the most stable change rate among the three target distances, so the target cluster is determined to be the target linear area.

[0110] Step S35: determining the target embedding dimension and target delay time based on the target linear region.

[0111] In step S35, based on the target linear region, the relationship between the correlation dimension and the embedding dimension is calculated, and a relationship diagram between the correlation dimension and the embedding dimension is drawn. According to the relationship diagram, the minimum embedding dimension corresponding to the stable value of the correlation dimension is selected. For example, please refer to Figure 6 ,exist Figure 6 In the case shown, the embedding dimension (i.e. Figure 6 When the horizontal axis in is greater than or equal to 10, the correlation dimension (i.e. Figure 6The vertical coordinate in ( ) tends to a stable value, so the target embedding dimension is taken as 10, and the corresponding target delay time is calculated to be 8. It should be noted that the values ​​given in the above example are only example values, and different values ​​from the above example may be obtained after performing the above steps on different vibration signals.

[0112] In this embodiment, the vibration signal is reconstructed in phase space, so that the low-dimensional vibration signal is projected into a high-dimensional space to obtain a first high-dimensional signal, and then the correlation dimension of the first high-dimensional signal is calculated to draw a double logarithmic curve required for determining the target linear region, thereby providing a data basis for the subsequent determination of the target linear region; the double logarithmic curve is converted into discrete points by utilizing scatter point transformation for processing by a clustering algorithm; the double logarithmic curve is processed by adopting a K-center point clustering algorithm to obtain multiple target clusters, and then a cluster with the most stable rate of change is determined from the multiple target clusters to obtain the required target linear region; in the target linear region, the target embedding dimension and the target delay time are determined based on the relationship between the correlation dimension and the embedding dimension.

[0113] Compared with manually judging that the area with the most stable rate of change in the double logarithmic curve is the target linear area, this embodiment achieves the effect of efficiently and accurately determining the target linear area by first using the K-center point clustering algorithm to convert discrete points in the double logarithmic curve into multiple target clusters, and then determining a cluster with the most stable rate of change from the multiple target clusters.

[0114] Based on the first embodiment of the present application, in the third embodiment of the present application, the same or similar contents as those in the first embodiment can be referred to the above description, and will not be described in detail later. Figure 7 Before step S30, the blind source separation method further includes step S20:

[0115] Step S20, a calculation step, based on a preset embedding dimension range, uses a cross-correlation method to calculate the time delay.

[0116] In this embodiment, the time delay is calculated by using the cross-correlation method by presetting the embedding dimension range; since the time delay is a key parameter in the phase space reconstruction, it determines the relative position between the points in the reconstructed phase space, and selecting a suitable time delay can ensure that the reconstructed phase space can fully reflect the dynamic characteristics of the original time series. The cross-correlation method can find a suitable time delay based on the preset embedding dimension range by calculating the cross-correlation function between the time series and its own lagged version, thereby providing the accuracy of subsequent phase space reconstruction using the preset embedding dimension range and time delay, and avoiding the generation of redundant information or information loss during the phase space reconstruction process.

[0117] Further, as an optional implementation, step S50 may include steps S51 to S54:

[0118] Step S51, a permutation entropy calculation step, performing permutation entropy calculation on a plurality of the decomposed signals to obtain a permutation entropy of each of the decomposed signals;

[0119] The calculation process of permutation entropy can be referred to as follows: Assume that the time series of the decomposed signal is Y = [y1, y2, ..., y n ], discretize the time series of the decomposed signal and sort each sub-element to obtain a sequence I=[i1,i2,…,i n ], that is, y(i1)≤y(i2)≤…≤y(i n ), and when there are equal elements, y(i1)=y(i2), sorted by index size, that is, sorted by the size of i1 and i2. For the time series of n-dimensional decomposed signals, there may be n! sorting methods, and the probability of occurrence of various arrangements is P1, P2, ..., P n , according to the calculation formula of permutation entropy:

[0120]

[0121] Step S52, a sorting step, sorting and screening a plurality of the permutation entropies to obtain a minimum permutation entropy;

[0122] Step S53, if the minimum permutation entropy is less than or equal to the expected entropy value, confirming that the decomposition signal corresponding to the minimum permutation entropy is a target decomposition signal, and outputting the target decomposition signal;

[0123] As an optional implementation, the expected entropy value may be set to 1.

[0124] Step S54, if the minimum permutation entropy is greater than the expected entropy value, repeat the calculation step, the determination step, the decomposition step, the permutation entropy calculation step and the sorting step until the number of repeated cycles reaches a preset number or the minimum permutation entropy is less than or equal to the expected entropy value, then confirm that the decomposition signal corresponding to the minimum permutation entropy is the target decomposition signal, and output the target decomposition signal.

[0125] In this embodiment, the minimum permutation entropy is obtained by executing the permutation entropy calculation step and the calculation step; since the smaller the entropy value of the minimum permutation entropy is, the more ordered it is, and the lower the noise contained in the corresponding decomposition signal is, the minimum permutation entropy is used as the judgment standard. If the minimum permutation entropy is less than or equal to the expected entropy value, it means that the decomposition signal corresponding to the minimum permutation entropy is well separated and can be confirmed as the required target decomposition signal; if the minimum permutation entropy is greater than the expected entropy value, it means that the signal with the lowest noise content in multiple decomposition signals is still not well separated, so it is necessary to repeat the calculation step, the determination step, the decomposition step, the permutation entropy calculation step and the sorting step until the number of repeated cycles reaches the preset number or the minimum permutation entropy is less than or equal to the expected entropy value, then the decomposition signal corresponding to the minimum permutation entropy is confirmed as the target decomposition signal, and the target decomposition signal is output. After being processed by the minimum permutation entropy algorithm, the obtained target decomposition signal has a lower noise content and can more accurately reflect the abnormal vibration characteristics in the original vibration signal.

[0126] Based on the first embodiment of the present application, in the fourth embodiment of the present application, the same or similar contents as any of the above embodiments can be referred to the above introduction, and will not be repeated later. On this basis, step S40 includes steps S41 to S46:

[0127] Step S41, based on the target embedding dimension and the target delay time, reconstructing the vibration signal in phase space to obtain a second high-dimensional signal;

[0128] Step S42, performing a decentralized process on the second high-dimensional signal to obtain a signal to be demixed;

[0129] Step S43, an initialization step, randomly initializing the unmixing matrix of the signal to be unmixed;

[0130] Step S44, an iteration step, iterating the unmixing matrix according to an unmixing matrix update formula and in combination with orthogonalization processing to obtain an updated unmixing matrix;

[0131] Among them, the update formula of the unmixing matrix is:

[0132] B=E{zg(B T z)}-E{g'(B T z)}w

[0133] Where B is the unmixing matrix, E(*) is the mean calculation process, Z is the whitened signal, and the whitened signal is obtained by whitening the unmixed signal, g(*) and g′(*) are the first-order derivative and second-order derivative of G(*), respectively, and G(s) = -exp(-s 2 / 2), w is a column vector of the unmixing matrix B.

[0134] Step S45, within the second iteration number, if the updated demixing matrix does not converge, repeating the iteration step until the updated demixing matrix converges; if the updated demixing matrix converges, repeating the initialization step and the iteration step until all demixing matrices are traversed;

[0135] Step S46: Based on the entire unmixing matrix and the second high-dimensional signal, a plurality of independent decomposed signals are calculated.

[0136] In this embodiment, by decentralizing the second high-dimensional signal, the mean of the second high-dimensional signal becomes zero, simplifying subsequent calculations and reducing computational complexity; by randomly initializing the unmixing matrix, the risk of the algorithm falling into a local optimum is reduced, which helps to find the global optimal solution of the unmixing matrix; by iterating the unmixing matrix according to the unmixing matrix update formula and combining orthogonalization processing, the unmixing matrix is ​​optimized, and the column vectors of the unmixing matrix are ensured to remain orthogonal, avoiding the correlation between the column vectors of the unmixing matrix and the overfitting phenomenon; after multiple iterations and judging the convergence of the unmixing matrix, it is ensured that the final unmixing matrix is ​​convergent and traverses all the required unmixing matrices. Based on all the unmixing matrices and the second high-dimensional signal, multiple independent decomposition signals are calculated. Since all the unmixing matrices converge to the optimal value after multiple iterations, the separation accuracy of all the unmixing matrices for the second high-dimensional signal is improved, so that the separated independent decomposition signals can more clearly reflect the characteristics of the original vibration signal, improve the signal quality, and facilitate subsequent analysis and processing.

[0137] This application also provides a blind source separation device, please refer to Figure 8 , the blind source separation device comprises:

[0138] An acquisition module 10, used to acquire a vibration signal;

[0139] A determination module 30, used for the determination step, based on a preset embedding dimension range and time delay, reconstructing the phase space of the vibration signal with different embedding dimensions to obtain a first high-dimensional signal, and determining a target linear region for the first high-dimensional signal using a K-center point clustering algorithm, and determining a target embedding dimension and a target delay time based on the target linear region;

[0140] A decomposition module 40, used for the decomposition step, based on the target embedding dimension and the target delay time, reconstructing the vibration signal in phase space to obtain a second high-dimensional signal, and performing fast independent component analysis FastICA on the second high-dimensional signal to obtain a plurality of independent decomposed signals;

[0141] The output module 50 is used to calculate the permutation entropy of the plurality of decomposition signals, and use a minimum permutation entropy judgment algorithm to judge and process the permutation entropy and the corresponding decomposition signal, and output a target decomposition signal.

[0142] The blind source separation device provided by the present application adopts the blind source separation method in the above embodiment, and can solve the technical problem that the target linear region in the blind source separation method based on phase space reconstruction relies on artificial determination. Compared with the prior art, the beneficial effects of the blind source separation device provided by the present application are the same as the beneficial effects of the blind source separation method provided by the above embodiment, and other technical features in the blind source separation device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0143] The present application provides a blind source separation device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the blind source separation method in the above-mentioned embodiment 1.

[0144] Reference below Fig. 9 , which shows a schematic diagram of the structure of a blind source separation device suitable for implementing the embodiment of the present application. The blind source separation device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Fig. 9 The blind source separation device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0145] like Fig. 9As shown, the blind source separation device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the blind source separation device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the blind source separation device to communicate with other devices wirelessly or wired to exchange data. Although the figure shows a blind source separation device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or provided alternatively.

[0146] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0147] The blind source separation device provided by the present application adopts the blind source separation method in the above embodiment, which can solve the technical problem that the target linear region depends on artificial determination in the blind source separation method based on phase space reconstruction. Compared with the prior art, the beneficial effects of the blind source separation device provided by the present application are the same as the beneficial effects of the blind source separation method provided by the above embodiment, and other technical features in the blind source separation device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0148] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0149] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0150] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the blind source separation method in the above-mentioned embodiment.

[0151] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0152] The computer-readable storage medium may be included in the blind source separation device; or may exist independently without being assembled into the blind source separation device.

[0153] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the blind source separation device, the blind source separation device: obtains a vibration signal; determines a step, based on a preset embedding dimension range and time delay, reconstructs the vibration signal in phase space with different embedding dimensions to obtain a first high-dimensional signal, and determines a target linear region for the first high-dimensional signal using a K-center point clustering algorithm, and determines a target embedding dimension and a target delay time based on the target linear region; decomposes the vibration signal in phase space based on the target embedding dimension and the target delay time to obtain a second high-dimensional signal, and performs fast independent component analysis FastICA on the second high-dimensional signal to obtain multiple independent decomposed signals; calculates the permutation entropy of the multiple decomposed signals, and uses a minimum permutation entropy judgment algorithm to judge and process the permutation entropy and the corresponding decomposed signals, and outputs a target decomposed signal.

[0154] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0155] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0156] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0157] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned blind source separation method, and can solve the technical problem that the target linear region depends on artificial determination in the blind source separation method based on phase space reconstruction. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the blind source separation method provided by the above-mentioned embodiment, and will not be repeated here.

[0158] The present application also provides a computer program product, including a computer program, which implements the steps of the blind source separation method as described above when executed by a processor.

[0159] The computer program product provided by the present application can solve the technical problem that the target linear region depends on artificial determination in the blind source separation method based on phase space reconstruction. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as the beneficial effects of the blind source separation method provided by the above embodiment, which will not be described in detail here.

[0160] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A blind source separation method, characterized in that: The method comprises: Obtain vibration signal; A determination step, based on a preset embedding dimension range and time delay, reconstructing the phase space of the vibration signal with different embedding dimensions to obtain a first high-dimensional signal, and determining a target linear region for the first high-dimensional signal using a K-center point clustering algorithm, and determining a target embedding dimension and a target delay time based on the target linear region; a decomposition step, reconstructing the vibration signal in phase space based on the target embedding dimension and the target delay time to obtain a second high-dimensional signal, and performing fast independent component analysis FastICA on the second high-dimensional signal to obtain multiple independent decomposed signals; The permutation entropies of the plurality of decomposition signals are calculated, and the permutation entropies and the corresponding decomposition signals are judged and processed using a minimum permutation entropy judgment algorithm, and a target decomposition signal is output.

2. The method according to claim 1, characterized in that The determining step comprises: Based on the preset embedding dimension range and the time delay, reconstructing the vibration signal in phase space with different embedding dimensions to obtain the first high-dimensional signal, and calculating the correlation dimension of the first high-dimensional signal; Based on the correlation dimension, drawing a double logarithmic curve within the preset embedding dimension range; Processing the double logarithmic curve using scatter point transformation combined with a K-center point clustering algorithm to obtain multiple target clusters; Determine a cluster with the most stable change rate among the plurality of target clusters to obtain the target linear region; Based on the target linear region, the target embedding dimension and the target delay time are determined.

3. The method according to claim 2, characterized in that The step of processing the double logarithmic curve using scatter point conversion in combination with a K-center point clustering algorithm to obtain multiple target clusters includes: Performing scatter point transformation on the double logarithmic curve to obtain a non-determined region; Setting a plurality of initial center points in the non-fixed area; A clustering step, performing clustering processing according to the distance between each discrete point in the non-fixed area and each of the initial center points, to obtain multiple clusters corresponding to multiple initial center points; An updating step, updating the positions of the multiple initial center points so that the sum of distances from points in the same cluster to the corresponding new center point is minimized; The clustering step and the updating step are repeated until the first iteration number is reached or the position of the new center point is fixed, thereby obtaining a plurality of target clusters.

4. The method according to claim 1, characterized in that Prior to the determining step, the method includes: The calculation step is to calculate the time delay by using a cross-correlation method based on a preset embedding dimension range.

5. The method according to claim 4, characterized in that The step of calculating the permutation entropy of the plurality of decomposition signals, judging and processing the permutation entropy and the corresponding decomposition signal using a minimum permutation entropy judgment algorithm, and outputting a target decomposition signal comprises: a permutation entropy calculation step, performing permutation entropy calculation on a plurality of the decomposed signals to obtain a permutation entropy of each of the decomposed signals; a sorting step, sorting and screening the plurality of permutation entropies to obtain a minimum permutation entropy; If the minimum permutation entropy is less than or equal to the expected entropy value, confirming that the decomposition signal corresponding to the minimum permutation entropy is the target decomposition signal, and outputting the target decomposition signal; If the minimum permutation entropy is greater than the expected entropy value, the calculation step, the determination step, the decomposition step, the permutation entropy calculation step and the sorting step are repeated until the number of repeated cycles reaches a preset number or the minimum permutation entropy is less than or equal to the expected entropy value, then the decomposition signal corresponding to the minimum permutation entropy is confirmed to be the target decomposition signal, and the target decomposition signal is output.

6. The method according to any one of claims 1 to 5, characterized in that The decomposition step comprises: Based on the target embedding dimension and the target delay time, reconstructing the vibration signal in phase space to obtain a second high-dimensional signal; Performing a decentralization process on the second high-dimensional signal to obtain a signal to be demixed; An initialization step of randomly initializing the unmixing matrix of the signal to be unmixed; an iterative step, iterating the unmixing matrix according to an unmixing matrix update formula and in combination with orthogonalization processing to obtain an updated unmixing matrix; In the second iteration number, if the updated unmixing matrix does not converge, repeating the iterative step until the updated unmixing matrix converges; if the updated unmixing matrix converges, repeating the initialization step and the iterative step until all unmixing matrices are traversed; Based on the entire unmixing matrix and the second high-dimensional signal, a plurality of independent decomposed signals are calculated.

7. A blind source separation device, characterized in that: The device comprises: An acquisition module, used for acquiring vibration signals; A determination module, used for determining the step, based on a preset embedding dimension range and time delay, reconstructing the phase space of the vibration signal with different embedding dimensions to obtain a first high-dimensional signal, and determining a target linear region for the first high-dimensional signal using a K-center point clustering algorithm, and determining a target embedding dimension and a target delay time based on the target linear region; A decomposition module, used for the decomposition step, based on the target embedding dimension and the target delay time, reconstructing the vibration signal in phase space to obtain a second high-dimensional signal, and performing fast independent component analysis FastICA on the second high-dimensional signal to obtain multiple independent decomposed signals; The output module is used to calculate the permutation entropy of the multiple decomposition signals, and use the minimum permutation entropy judgment algorithm to judge and process the permutation entropy and the corresponding decomposition signal, and output the target decomposition signal.

8. A blind source separation device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the blind source separation method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the blind source separation method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the blind source separation method according to any one of claims 1 to 6 are implemented.

Images