A rolling bearing fault diagnosis method, system and device based on acoustic signals

Abstract

The present invention belongs to the field of physical data measurement, and provides a rolling bearing fault diagnosis method, system and equipment based on acoustic signals. The method can calculate a signal aliasing matrix according to the acoustic signal of the rolling bearing, and obtain a hybrid matrix of the signal aliasing matrix and perform an inverse transformation on the hybrid matrix by setting a plurality of virtual individuals in the signal aliasing matrix and making the plurality of virtual individuals perform hybridization as an estimated source signal; the estimated source signal is separated to obtain a source signal separation matrix, and each order of modal components is obtained from the source signal separation matrix and a modal filter component is screened out therefrom; after signal reconstruction of the modal filter component, its frequency domain characteristics are extracted for use in rolling bearing fault diagnosis, which significantly improves the accuracy and stability of signal separation and provides more reliable technical support for rolling bearing fault diagnosis.

Description

Technical Field

[0001] The invention belongs to the field of physical data measurement, and in particular relates to a rolling bearing fault diagnosis method, system and equipment based on acoustic signals. Background Art

[0002] Rolling bearings are key components in mechanical equipment, and their fault diagnosis is of great significance for ensuring the normal operation of equipment, reducing maintenance costs, and avoiding sudden failures. As a non-contact, real-time diagnostic method, acoustic signal analysis has received extensive attention and application in the field of rolling bearing fault detection in recent years. By analyzing the acoustic signals generated during the operation of rolling bearings, early fault characteristics of components such as rolling elements, inner rings, outer rings, and cages can be identified.

[0003] The current industry situation shows that acoustic signal analysis mainly relies on core technologies such as signal separation, noise reduction and frequency domain feature extraction. However, there are a lot of background noise and aliasing signals in complex industrial environments, which makes it a technical challenge to accurately extract rolling bearing fault features. Therefore, how to effectively separate fault signals, reduce noise interference and improve the accuracy of feature extraction is a key issue that needs to be solved in the current industry.

[0004] Existing rolling bearing acoustic signal fault diagnosis technology usually includes the following main steps: signal acquisition, preprocessing, signal separation, noise reduction processing, and feature extraction and analysis. Independent component analysis (ICA), as a common signal separation method, is widely used in the field of fault diagnosis to separate aliased signals and extract target source signals. However, traditional signal processing methods, such as ICA, fast FFT, and empirical EMD, have obvious limitations. The traditional ICA method is based on linear assumptions and non-Gaussian optimization, and separates the linear independent components of the signal through centering and whitening. The main advantages of this method are fast processing speed and mature algorithms, but the separation accuracy of ICA will be significantly reduced in complex noise environments. Especially when the aliased signal is serious, the sensitivity of ICA to noise leads to unstable separation effect and difficulty in extracting accurate fault feature signals. FFT is used to convert time domain signals into frequency domain signals and extract frequency components in the signal. It is a commonly used frequency domain analysis method in fault diagnosis. However, the drawback of FFT is that it cannot process non-stationary signals, has a poor response to fault characteristics whose frequency varies with time, and in the presence of noise interference, the spectrum information is easily masked, resulting in misjudgment and omission of frequency components. As an adaptive decomposition method, EMD extracts intrinsic mode functions (IMFs) of different frequencies by decomposing the signal step by step. Although EMD has certain advantages in nonlinear and non-stationary signal processing, it faces problems such as modal aliasing and endpoint effects in practical applications, which affect the accuracy and robustness of the decomposition, especially when the noise is large, the decomposition results often contain irrelevant components.

[0005] Traditional ICA methods perform poorly in high-noise environments, and are not effective in separating background noise and interference signals, which can easily lead to mis-separation and fail to ensure high signal fidelity. As the complexity of industrial environmental noise increases, the quality of signals after ICA separation decreases significantly, affecting the accuracy of fault feature extraction and diagnosis.

[0006] Traditional methods are mostly based on linear assumptions and cannot fully process complex nonlinear aliasing signals. For the nonlinear and non-stationary characteristics of rolling bearing fault signals, the processing capabilities of traditional methods are significantly insufficient, resulting in poor signal feature extraction and difficulty in identifying fault types.

[0007] Traditional signal processing methods mostly use fixed processing parameters and steps, and lack the ability to adaptively optimize signal characteristics. When faced with diverse industrial signals and complex noise interference, fixed parameter processing methods lack flexible adjustment capabilities and are difficult to cope with real-time and variable diagnostic requirements.

[0008] Traditional methods such as EMD are prone to modal aliasing and endpoint effects when processing rolling bearing acoustic signals, resulting in the decomposed signal being unable to accurately reflect the characteristics of the original signal. Especially under high noise conditions, the modal components often contain irrelevant information, which affects the results of signal reconstruction and fault diagnosis. Summary of the invention

[0009] The purpose of the present invention is to propose a rolling bearing fault diagnosis method, system and equipment based on acoustic signals to solve one or more technical problems existing in the prior art and at least provide a beneficial choice or create conditions.

[0010] The application of traditional rolling bearing fault diagnosis technology in complex industrial environments is limited by noise interference, insufficient nonlinear processing capabilities, and low separation accuracy. With the increasing demand for equipment reliability and maintenance cost control in industrial production, traditional methods have been unable to meet the requirements of efficient and accurate fault diagnosis. In response to the above problems, the technical solution described in the present invention significantly improves the accuracy and stability of signal separation through innovative signal hybrid optimization, nonlinear separation, and multimodal noise reduction, providing more reliable technical support for rolling bearing fault diagnosis.

[0011] In order to achieve the above object, according to one aspect of the present invention, a rolling bearing fault diagnosis method based on acoustic signals is provided, the method comprising the following steps:

[0012] Calculate a signal aliasing matrix (Aliasing, which may be an overlapping frequency in the signal spectrum) according to the acoustic signal of the rolling bearing, set a plurality of virtual individuals in the signal aliasing matrix and make the plurality of virtual individuals perform mixing, so as to obtain a mixing matrix of the signal aliasing matrix and perform an inverse transformation on the mixing matrix as an estimated source signal;

[0013] Separating and processing the estimated source signal to obtain a source signal separation matrix;

[0014] Separate the source signal matrix to obtain the modal components of each order and filter out the modal filter components;

[0015] After signal reconstruction of the modal filter component, its frequency domain features are extracted for rolling bearing fault diagnosis.

[0016] Furthermore, in the process of obtaining each order modal component of the source signal separation matrix, a modal decomposition method is used to obtain each order modal component of the source signal separation matrix. The modal decomposition method disturbs the extreme points of the source signal separation matrix by adding symmetrical noise to the source signal separation matrix to reduce the modal aliasing phenomenon and improve the decomposition accuracy.

[0017] Furthermore, in the step of reconstructing the signal of the modal filter components, the cross-correlation coefficients of the modal components are first calculated, and only the modal components with cross-correlation coefficients higher than a preset threshold are retained for signal reconstruction.

[0018] Furthermore, the reconstructed signal is subjected to Fourier transform to extract frequency domain features, which are used to identify the fault type of the rolling bearing.

[0019] Furthermore, a signal aliasing matrix is ​​calculated according to the acoustic signal of the rolling bearing, and a plurality of virtual individuals are set in the signal aliasing matrix and the plurality of virtual individuals are mixed to obtain a mixed matrix of the signal aliasing matrix and perform an inverse transformation on the mixed matrix as an estimated source signal, specifically:

[0020] A plurality of virtual individuals are set in the signal aliasing matrix, each virtual individual represents a candidate aliasing matrix solution, and a candidate aliasing matrix solution may be one of the eigenvalues ​​of the signal aliasing matrix, the positions of the virtual individuals are initialized, and an initial weight value is assigned to each agent to represent its fitness in the signal aliasing matrix;

[0021] Each agent performs mixing in the signal aliasing matrix according to nonlinear dynamic rules. In each round of mixing, each virtual individual dynamically adjusts its weight according to its current position and fitness value: a virtual individual is placed in the signal aliasing matrix and moves once in each round of mixing, and the number of virtual individuals in each direction and the fitness value of each virtual individual are compared. The movement direction of the virtual individual is selected according to the number of virtual individuals in each direction and the fitness value of each virtual individual. When there are at least two virtual individuals in adjacent positions, the fitness values ​​of the virtual individuals in the adjacent positions are mixed;

[0022] When the mathematical expectation of the quadratic partial derivative values ​​of the position of each virtual individual and its corresponding fitness value tends to zero, the values ​​of the corresponding positions in the signal aliasing matrix at the positions of each virtual individual are point-to-point multiplied, and the obtained matrix is ​​used as the hybrid matrix of the signal aliasing matrix, and the hybrid matrix is ​​inversely transformed to obtain the estimated source signal.

[0023] Furthermore, the method of separating the estimated source signal to obtain a source signal separation matrix is:

[0024] An independent component analysis is performed on the estimated source signal, and independent components in the estimated source signal are separated by an independent component analysis algorithm for optimizing the signal, so as to separate the source signal separation matrix.

[0025] Further, the method of separating the estimated source signal to obtain a source signal separation matrix is ​​either:

[0026] Performing centering and whitening processing on the estimated source signal to eliminate linear correlation between signal features and standardize signal energy;

[0027] The independence matrix is ​​a tensor that can calculate the non-Gaussian measurement value of the acoustic signal based on the acoustic signal of the rolling bearing and can be used to measure the independence of the acoustic signal of the rolling bearing: the estimated source signal after centering and whitening is processed with a nonlinear characteristic function matrix to obtain an independence matrix, wherein the matrix size of the nonlinear characteristic function matrix is ​​consistent with the estimated source signal, each row and column position in the nonlinear characteristic function matrix uses the same type of nonlinear characteristic function, but the weights of the variables in the nonlinear characteristic function used are not completely the same in each row and column position, and the nonlinear characteristic function matrix can be used to maximize the non-Gaussian measurement value in the estimated source signal after centering and whitening by transforming the weights of the variables in the nonlinear characteristic function used in each row and column position;

[0028] The independence matrix is ​​orthogonalized to obtain a source signal separation matrix.

[0029] The present invention also provides a rolling bearing fault diagnosis system based on acoustic signals, the rolling bearing fault diagnosis system based on acoustic signals comprising: a processor, a memory and a computer program stored in the memory and executable on the processor, the processor implementing the steps in the rolling bearing fault diagnosis method based on acoustic signals when executing the computer program, the rolling bearing fault diagnosis system based on acoustic signals can be run in computing devices such as desktop computers, laptop computers, PDAs and cloud data centers, the executable system may include, but not limited to, processors, memories, server clusters, the processor executing the computer program runs in the following system units:

[0030] A signal processing unit, configured to calculate a signal aliasing matrix according to an acoustic signal of a rolling bearing, and obtain a hybrid matrix of the signal aliasing matrix by setting a plurality of virtual individuals in the signal aliasing matrix and causing the plurality of virtual individuals to perform hybridization, and perform an inverse transformation on the hybrid matrix as an estimated source signal;

[0031] A signal separation unit, used for separating and processing the estimated source signal to produce a source signal separation matrix;

[0032] A signal simulation unit is used to obtain modal components of various orders from the source signal separation matrix and filter out modal filter components therefrom;

[0033] The signal detection unit is used to reconstruct the signal of the modal filter component and then extract its frequency domain characteristics for the fault diagnosis of the rolling bearing.

[0034] Correspondingly, the present invention also provides an electronic device, a readable storage medium and a computer program product:

[0035] An electronic device comprises: 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 rolling bearing fault diagnosis method based on acoustic signals and the steps therein.

[0036] A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the rolling bearing fault diagnosis method based on acoustic signals and the steps therein.

[0037] A computer program product comprises a computer program. When the computer program is executed by a processor, the method for diagnosing rolling bearing faults based on acoustic signals and the steps thereof are implemented.

[0038] The beneficial effects of the present invention are as follows: the present invention provides a rolling bearing fault diagnosis method, system and equipment based on acoustic signals, calculates a signal aliasing matrix according to the acoustic signal of the rolling bearing, sets multiple virtual individuals in the signal aliasing matrix and makes the multiple virtual individuals perform hybridization to obtain a hybrid matrix of the signal aliasing matrix and performs an inverse transformation on it as an estimated source signal; separates and processes the estimated source signal to obtain a source signal separation matrix, obtains modal components of each order from the source signal separation matrix and screens out modal filter components therefrom; reconstructs the modal filter components, and then extracts their frequency domain features for use in rolling bearing fault diagnosis, which significantly improves the accuracy and stability of signal separation and provides more reliable technical support for rolling bearing fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The above and other features of the present invention will become more obvious by describing in detail the embodiments shown in the accompanying drawings. The same reference numerals in the accompanying drawings of the present invention represent the same or similar elements. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other accompanying drawings can be obtained based on these accompanying drawings without creative work. In the accompanying drawings:

[0040] Figure 1 Shown is a flow chart of a rolling bearing fault diagnosis method based on acoustic signals;

[0041] Figure 2 Shown is a system structure diagram of a rolling bearing fault diagnosis system based on acoustic signals. DETAILED DESCRIPTION

[0042] The following will be combined with the embodiments and drawings to clearly and completely describe the concept, specific structure and technical effects of the present invention, so as to fully understand the purpose, scheme and effect of the present invention. It should be noted that the embodiments and features in the embodiments of this application can be combined with each other without conflict.

[0043] In the description of the present invention, "several" means one or more, "more" means more than two, "greater than", "less than", "exceed" etc. are understood as not including the number itself, and "above", "below", "within" etc. are understood as including the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing the technical features, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0044] like Figure 1 The flowchart of a rolling bearing fault diagnosis method based on acoustic signals according to the present invention is shown below. Figure 1A rolling bearing fault diagnosis method, system and device based on acoustic signals according to an embodiment of the present invention are described.

[0045] The present invention proposes a rolling bearing fault diagnosis method based on acoustic signals, as shown in Example 1, wherein the method specifically comprises the following steps:

[0046] Calculating a signal aliasing matrix according to the acoustic signal of the rolling bearing, setting a plurality of virtual individuals in the signal aliasing matrix and causing the plurality of virtual individuals to perform mixing, thereby obtaining a mixing matrix of the signal aliasing matrix and performing an inverse transformation on the mixing matrix as an estimated source signal;

[0047] Separating and processing the estimated source signal to obtain a source signal separation matrix;

[0048] Separate the source signal matrix to obtain the modal components of each order and filter out the modal filter components;

[0049] After signal reconstruction of the modal filter component, its frequency domain features are extracted for rolling bearing fault diagnosis.

[0050] Furthermore, a signal aliasing matrix is ​​calculated according to the acoustic signal of the rolling bearing, and a plurality of virtual individuals are set in the signal aliasing matrix and the plurality of virtual individuals are mixed to obtain a mixed matrix of the signal aliasing matrix and perform an inverse transformation on the mixed matrix as an estimated source signal, specifically:

[0051] In Example 2, multiple virtual individuals are first set in the signal aliasing matrix, each virtual individual represents a candidate aliasing matrix solution, and a candidate aliasing matrix solution can be one of the eigenvalues ​​of the signal aliasing matrix. The positions of these virtual individuals are initialized by a random function (the number of the virtual individuals can be equal to the number of eigenvalues ​​in the eigenvector, which can be a unit eigenvector), and each agent is assigned an initial weight value (one of the eigenvalues ​​of the signal aliasing matrix) to represent its fitness in the signal aliasing matrix.

[0052] Then each agent moves in the signal aliasing matrix according to nonlinear dynamic rules, and performs mixing during the movement. In each round of mixing, each virtual individual dynamically adjusts its weight according to its current position and fitness value: a virtual individual is placed in the signal aliasing matrix and needs to move once in each round of mixing. Virtual individuals with higher fitness values ​​have an attraction effect on other virtual individuals. For example, when a virtual individual needs to move once, it searches in the directions of up to 8 adjacent positions around it (including up, down, left, right, upper left, lower left, upper right, lower right, and only part of the edge positions), and compares the number of virtual individuals in each direction with the fitness value of each virtual individual. The value of the fitness is determined, and the direction of movement of the virtual individual is selected according to the number of virtual individuals in each direction and the fitness value of each virtual individual. When there are at least two virtual individuals in adjacent positions, the fitness values ​​of the virtual individuals in adjacent positions are mixed. Preferably, the fitness value of the virtual individual with a larger value is reduced, and the fitness value of the virtual individual with a smaller value is increased. For example, the fitness value of the virtual individual with a larger value is calculated from the fitness value of the virtual individual with a smaller value, and the fitness value of the virtual individual with a larger value is reduced by the square root of the absolute difference, and the fitness value of the virtual individual with a smaller value is increased. If there is no virtual individual with a lower fitness around it, it is mixed at its own position.

[0053] After multiple mixing, when the mathematical expectation of the second partial derivative value of the position of each virtual individual and its corresponding fitness value tends to zero, or the second partial derivative value begins to turn negative, the value of the corresponding position in the signal aliasing matrix corresponding to the position of each virtual individual is point-to-point multiplied, and the obtained matrix is ​​used as the mixing matrix of the signal aliasing matrix, and the mixing matrix is ​​inversely transformed to obtain the estimated source signal.

[0054] Among them, the reason why multiple virtual individuals are set in the signal aliasing matrix, each virtual individual represents a candidate aliasing matrix solution. This is because treating the elements in the signal aliasing matrix as virtual individuals is an innovative signal optimization strategy. Each virtual individual represents a candidate solution in the aliasing matrix, and the initial position is randomly distributed in the matrix space. This setting makes the search process highly random and diverse, and can prevent the traditional optimization algorithm from falling into the local optimum. This setting effectively improves the global search ability of the algorithm by randomly initializing the position of the virtual individual and assigning the initial fitness, which is particularly suitable for finding the optimal solution for signal separation in a complex noise environment. Compared with the traditional fixed parameter optimization method, this method can flexibly adapt to different signal characteristics and improve the accuracy and robustness of signal separation.

[0055] Each agent moves in the signal aliasing matrix and adjusts the weight according to nonlinear dynamic rules. The movement of virtual individuals (class Rotatr) in the signal aliasing matrix can be adjusted through nonlinear dynamic rules to simulate an adaptive optimization process. The movement of virtual individuals not only considers their own fitness, but also refers to the state and relative position of surrounding virtual individuals to achieve a dynamic balance between global and local optimization. This dynamic adjustment mechanism is equivalent to adaptive optimization of the signal, which can gradually approach the optimal aliasing matrix solution while maintaining a diverse search. Compared with traditional optimization methods, this mechanism avoids the fixation and simplification of parameters, effectively improves the processing capability of complex signals, and enhances the stability and accuracy of signal separation.

[0056] As for virtual individuals with higher fitness, they attract other virtual individuals, while virtual individuals with lower fitness mix at their own positions. In this way, the fitness value reflects the current optimization effect of the virtual individual. Virtual individuals with higher fitness attract other individuals around them, which is equivalent to a behavior of guiding the optimization path. This setting imitates the group behavior in nature and can guide the separation matrix to gradually converge to the global optimal solution during the signal optimization process. Therefore, the fitness-driven attraction mechanism enhances the adaptive characteristics of the algorithm and enables the optimization process to have the ability to adjust flexibly. Virtual individuals with low fitness remain in their original positions and mix, which helps to retain the ability to explore local optimization and prevent the algorithm from converging prematurely. This setting effectively balances the relationship between global search and local optimization, ensuring the accuracy of signal separation and optimization efficiency.

[0057] After multiple hybrids of the virtual individuals, the mathematical expectation of the quadratic partial derivative of the fitness value approaches zero, forming a stable hybrid matrix. Here, when the fitness values ​​of each virtual individual gradually stabilize and tend to zero, it means that the optimization process has reached a state of equilibrium. At this time, the hybrid matrix has been formed, representing a global optimal or suboptimal solution. By calculating the mathematical expectation of the quadratic partial derivative of the fitness value to be zero, the convergence and stability of the matrix are further verified. Therefore, this dynamic convergence judgment mechanism ensures the reliability of the hybrid matrix, which is more adaptable and accurate than the traditional fixed number of hybrids or threshold judgments. This method can flexibly deal with signal complexity and noise influences, and provides a more accurate initial signal estimation for subsequent signal inverse transformation.

[0058] Then, the stable hybrid matrix is ​​inversely transformed to obtain the estimated source signal. In this way, the inverse transformation operation on the stable hybrid matrix can effectively eliminate the influence of signal aliasing and obtain a preliminary estimated source signal. This process is equivalent to unmixing and optimizing the separation of the signal to reconstruct the signal characteristics close to the original state. The estimated source signal after inverse transformation provides high-quality input data for subsequent fault diagnosis. Compared with directly separating the unoptimized aliased signal, this method significantly reduces the noise and aliasing errors in the signal, and enhances the identifiability of the fault characteristics and the diagnostic accuracy.

[0059] In summary, the above steps combine virtual individual simulation, dynamic adjustment, fitness guidance, hybrid convergence and inverse transformation and other innovative means, providing strong technical support for acoustic signal optimization and fault diagnosis of rolling bearings. These steps not only improve the flexibility and accuracy of signal processing, but also greatly improve the practicality in complex industrial environments, which is a major advancement in traditional signal processing technology.

[0060] Furthermore, the method of separating the estimated source signal to obtain a source signal separation matrix is:

[0061] In Example 3, independent component analysis is performed on the estimated source signal, and independent components in the estimated source signal are separated by an independent component analysis algorithm for optimizing signals to separate the source signal separation matrix.

[0062] Further, the method of separating and processing the estimated source signal to obtain a source signal separation matrix may also be:

[0063] In Embodiment 4, the estimated source signal is centered and whitened to eliminate the linear correlation between the signal features and to standardize the signal energy;

[0064] The independence matrix is ​​a tensor that can calculate the non-Gaussian measurement value of the acoustic signal according to the acoustic signal of the rolling bearing and can be used to evaluate the independence of the acoustic signal of the rolling bearing. This embodiment provides a calculation method of the independence matrix: the estimated source signal after centering and whitening is processed with a nonlinear characteristic function matrix to obtain the independence matrix;

[0065] Among them, the matrix size of the nonlinear characteristic function matrix is ​​consistent with the estimated source signal, and each row and column position in the nonlinear characteristic function matrix uses the same type of nonlinear characteristic function, but the weights of the variables in the nonlinear characteristic function used are not exactly the same in each row and column position, and the nonlinear characteristic function matrix can be used to maximize the non-Gaussian measure value in the estimated source signal after centering and whitening by transforming the weights of the variables in the nonlinear characteristic function used in each row and column position; in one embodiment, an activation function with different controllable weights on the variables can be used for each row and column position in the nonlinear characteristic function matrix.

[0066] The independence matrix is ​​orthogonalized to maintain a stable separation effect, thereby obtaining a source signal separation matrix.

[0067] The technical solution in Example 3 is to perform independent component analysis on the estimated source signal and directly use the existing independent component analysis algorithm to separate the signal. This method usually involves centralization, whitening and linear separation, and mainly relies on the non-Gaussianity of the signal to achieve independent separation. However, the technical solution in Example 3 is sensitive to noise and the separation accuracy is limited. Traditional independent component analysis is highly sensitive to noise during the separation process, especially in complex industrial environments, where the signal contains a large amount of background noise and interference, and direct separation is prone to misseparation or incomplete separation.

[0068] The technical solution in Example 3 has the limitation of linear processing. The traditional independent component analysis method is based on the assumption of a linear model and has limited signal processing capabilities. When the complexity of the signal increases (such as nonlinear aliasing signals), the traditional linear method cannot fully optimize the separation effect, resulting in the inability to fully extract the feature information. The technical solution in Example 3 has the problem of lack of flexibility in fixed parameter optimization. The separation algorithm therein usually uses fixed parameters and processing steps, lacks the ability to adaptively adjust to signal changes, and is difficult to cope with diverse signal characteristics and noise environments.

[0069] Therefore, an improved signal separation technical solution is proposed in the technical solution in Example 4, which significantly improves the effect and reliability of signal separation through multi-step innovative design. The centralization process adjusts the mean of the estimated source signal to zero, and the whitening process standardizes each component of the signal to equal variance, while eliminating the linear correlation between different signal features. This step effectively provides a unified data basis for subsequent nonlinear separation. Compared with traditional linear separation, centralization and whitening operations can reduce the impact of noise on the separation results and lay the foundation for nonlinear feature optimization. This data preprocessing enhances the robustness of the signal and improves the accuracy of separation.

[0070] The signal is processed using a nonlinear characteristic function matrix to obtain an independence matrix. By processing the signal through a nonlinear characteristic function matrix, the non-Gaussian characteristics of the signal can be amplified, making the independence of the signal more obvious. The elements of each function matrix are weighted according to the different signal characteristics to maximize the non-Gaussian measurement value of the signal. This method breaks the limitations of traditional linear separation and greatly enhances the adaptability to complex signals by introducing a nonlinear function matrix. Unlike fixed linear separation, nonlinear processing can handle more complex signal aliasing situations and effectively improve the effect of signal separation. By dynamically adjusting the weights of each row and column of the nonlinear characteristic function matrix, the non-Gaussianity of the signal can be flexibly optimized, and the characteristics of the signal can be fully utilized for adaptive separation. This innovative step significantly improves the limitations of traditional methods and has higher flexibility and processing capabilities.

[0071] Among them, the independence matrix is ​​orthogonalized to obtain the source signal separation matrix. The orthogonalization process stabilizes the separation result by maintaining the orthogonality of each component of the matrix, avoiding mutual interference between signal components during the separation process. This step ensures that the separated signal components remain independent and avoids the aliasing of repeated information. The orthogonalization operation can effectively improve the stability and accuracy of the separation and reduce the error accumulation during signal reconstruction. Compared with the traditional independent component analysis, this step ensures that the signal has clear independence after separation and effectively improves the reliability of signal processing. By orthogonalizing the independence matrix, the separation result is optimized, so that the signal separation is not only high-precision but also highly robust. This step breaks the linearization limitations of traditional technology and provides a wider adaptability for signal separation.

[0072] The improved solution in the technical solution in Example 4 makes up for the defects of the traditional method by introducing centralization and whitening preprocessing, dynamic optimization of the nonlinear characteristic function matrix and orthogonalization processing, especially showing significant advantages in noise processing and complex signal separation. Compared with the traditional solution, each step of Example 4 is innovative, ensuring that the accuracy, stability and adaptability of signal separation are greatly improved. These improvements make the technical solution in Example 4 a more superior choice for rolling bearing fault diagnosis in complex industrial environments.

[0073] Furthermore, in the process of obtaining each order modal component of the source signal separation matrix, a modal decomposition method is used to obtain each order modal component of the source signal separation matrix. The modal decomposition method disturbs the extreme points of the source signal separation matrix by adding symmetrical noise to the source signal separation matrix to reduce the modal aliasing phenomenon and improve the decomposition accuracy.

[0074] In Example 5, the source signal separation matrix Ssep is read, and the matrix is ​​obtained by the aforementioned signal separation step (such as independence matrix optimization processing). The complementary set empirical mode decomposition is selected as the modal decomposition method. This method realizes multi-modal decomposition of the signal by adding symmetrical noise multiple times, and effectively reduces the modal aliasing phenomenon. When the source signal separation matrix is ​​modally decomposed, in order to suppress the modal aliasing phenomenon, positive and negative symmetrical white noise are added to the signal to perform perturbation processing on the modal components of each order. The standard deviation σ of the symmetrical noise should be optimized according to the experiment to ensure the rationality of the disturbance (such as the noise amplitude accounts for 0.1% to 0.2% of the signal amplitude) to avoid excessive disturbance causing distortion of signal characteristics.

[0075] Through multiple noise disturbances, the source signal is subjected to multiple hybrid decompositions. After each hybrid, the average value of each modal component is calculated to reduce the random error of the decomposition. The acquired modal components are used to represent the intrinsic modes of the signal, arranged in sequence from low frequency to high frequency. Each modal component reflects the vibration characteristics of the signal in different frequency ranges, which helps to refine the analysis of fault characteristics in the signal. In each round of modal decomposition, the extreme points of the signal are adjusted by adding symmetrical noise to avoid spectrum overlap between different modes. The addition of noise guides the change of the decomposition path of the signal, making the frequency characteristics of each modal component more independent. The positive and negative symmetry of the noise ensures that no additional offset or trend is introduced during the decomposition process, maintaining the overall balance of the signal.

[0076] Comparing the decomposition results without noise and with symmetrical noise, the cross-correlation of adjacent modal components is used as a measure of the degree of aliasing. The decomposition method with noise can significantly reduce the correlation of adjacent modes and improve the independence of each mode. The mean of the cross-correlation coefficient ρ of adjacent modes is used as the basis for judgment. After adding noise, the mean of ρ is significantly reduced (usually less than 0.2), effectively avoiding modal aliasing.

[0077] The mutual correlation coefficient is calculated for each modal component and the original signal separation matrix to measure the contribution of each modal component to signal reconstruction. A correlation screening threshold (such as 0.1) is set to retain only the modal components with a mutual correlation coefficient higher than this threshold for signal reconstruction, ensuring that the reconstructed signal retains the main fault characteristics.

[0078] The filtered modal components are reconstructed to form a denoised signal, which retains the most important fault information in each order mode and removes irrelevant noise components. The frequency domain characteristics of the reconstructed signal show clearer characteristic frequency peaks and energy distribution compared to the original signal separation matrix, which improves the accuracy of fault diagnosis.

[0079] Furthermore, in the step of reconstructing the signal of the modal filter components, the cross-correlation coefficients of the modal components are first calculated, and only the modal components with cross-correlation coefficients higher than a preset threshold are retained for signal reconstruction.

[0080] In Embodiment 6, the preset threshold of the cross-correlation coefficient may be calculated based on the median, mean and / or mode of the cross-correlation coefficients of the modal components.

[0081] Furthermore, the reconstructed signal is subjected to Fourier transform to extract frequency domain features, which are used to identify the fault type of the rolling bearing.

[0082] In Example 7, a high-sensitivity acoustic sensor array arranged near the bearing can be used to collect the acoustic signals generated by the rolling bearing during operation in real time; the sampling frequency should be higher than twice the bearing fault frequency to ensure complete capture of the fault characteristic frequency;

[0083] The collected original sound signal is processed with low noise to filter out the environmental background noise, and the signal is subjected to blind source separation and noise reduction reconstruction using the aforementioned method to obtain a reconstructed clear signal Srec; the reconstructed signal Srec is processed in segments, and a reasonable time window (such as 0.1 second to 0.5 second) is selected to avoid spectrum leakage, and a windowing process (such as Hanning window) is performed to reduce spectrum fluctuations; a fast Fourier transform (FFT) can also be performed on each segmented signal to obtain a spectrum graph F(ω), where ω represents the frequency; in the spectrum graph F(ω), key frequency components are extracted, such as the fundamental frequency, multiple frequencies and their harmonics, with a focus on common bearing fault frequencies (such as the characteristic frequencies of the outer ring, inner ring, rolling element and cage);

[0084] Calculate the amplitude, energy and harmonic distribution of the frequency components, and record the peak amplitude and frequency band energy corresponding to each characteristic frequency; analyze the extracted frequency domain features and identify whether there are abnormal frequency components and their amplitude abnormalities by comparing them with the standard fault spectrum library.

[0085] Common fault types include: outer race fault (manifested as a specific frequency-multiplied signal), inner race fault (accompanied by a high-amplitude non-fundamental frequency signal), rolling element fault (random broadband peak characteristics) and cage fault (enhanced low-frequency components).

[0086] Intelligent diagnostic algorithms (such as rule bases and machine learning models) can also be used to further analyze the spectrum characteristics, and the results of feature matching can be used to determine the specific fault type and its severity.

[0087] The identification results are output to the monitoring system to realize real-time fault alarm, display the specific fault type, location and recommended maintenance measures. In industrial production sites, the results can be used to guide equipment maintenance personnel to promptly inspect or replace bearings to prevent the expansion of faults and cause equipment downtime. This implementation method specifically explains how to use fast Fourier transform to extract frequency domain features of reconstructed acoustic signals in industrial sites, which is used for the actual production needs of rolling bearing fault diagnosis, and can ensure the accuracy and practicality of diagnosis.

[0088] In the test experiment, a rolling bearing experimental platform that meets the actual industrial site was used. The acoustic signals of the rolling bearing during operation were collected through a high-sensitivity acoustic sensor array. The sensor parameters include: the number of pickups is 64, the sampling frequency is 192KHz, and the frequency range is 20Hz~96KHz. The experiment simulated different types of bearing faults, including rolling element, inner ring, outer ring and cage faults.

[0089] In the original signal spectrum, the fault characteristic frequency is seriously interfered by noise and cannot be clearly identified. For example, the characteristic frequency of the rolling element fault is 3.98Hz, which is submerged in the unprocessed signal. However, through the combined processing of Example 1 and Example 7, the optimized signal separation matrix and its frequency domain feature extraction can effectively separate the key fault frequency and significantly improve the accuracy of fault identification.

[0090] The experiment verified the strategy of obtaining the aliasing matrix hybrid matrix by setting virtual individuals and performing hybrid optimization in Example 2. This method achieves accurate optimization of the signal aliasing matrix through hybridization and fitness adjustment of virtual individuals. In the experiment, after multiple hybridizations, the expectations of the virtual individual positions and their fitness values ​​tend to be stable, and the hybrid matrix finally obtained can accurately reflect the actual fault characteristics. Experimental data show that after processing in Example 2, the signal of the rolling element fault successfully extracted the key frequencies in the range of 102Hz~220Hz, which were completely submerged in the unprocessed original signal. The data proves the superiority of the hybrid optimization strategy in complex noise environments.

[0091] The method for obtaining the signal separation matrix in Example 3 and Example 4 enhances the separation effect. Example 3 describes the separation processing of the estimated source signal by the independent component analysis method, optimizing the non-Gaussian characteristics of the signal, and thus separating the source signal separation matrix. In the experiment, this processing method significantly improved the separation accuracy, making each fault frequency more prominent and easy to identify in the spectrum. Example 4 supplements another way of obtaining independent components, which optimizes the signal independence through centering, whitening and nonlinear characteristic function matrix, and performs orthogonalization processing, which further improves the stability and effectiveness of the separation. This diverse signal processing method provides a more flexible and reliable technical path for rolling bearing fault diagnosis.

[0092] The noise reduction effect of modal decomposition and signal reconstruction in Examples 5 and 6 is significant. According to the experimental test of the technical solution in Example 5, the source signal separation matrix is ​​processed using the modal decomposition method, and the modal aliasing phenomenon is reduced by adding symmetrical noise perturbation extreme points, thereby improving the decomposition accuracy. The experimental results show that this processing can retain the key fault frequency components and improve the recognition of the signal. In the steps of Example 6, by calculating the mutual correlation coefficient of the modal components, only the components with correlation coefficients higher than the preset threshold are retained for signal reconstruction. This screening process significantly reduces the impact of noise on the diagnostic results. For example, experimental data show that after the screened signal reconstruction, the characteristic frequency of the inner ring fault, 6.02Hz, becomes more obvious and the amplitude increases, verifying the effectiveness of this processing.

[0093] The frequency domain feature extraction method described in Implementation Plan 7 uses fast Fourier transform to perform spectrum analysis on the reconstructed signal to accurately capture the typical frequencies of each fault. In the experiment, the processed signal spectrum can clearly show the characteristic frequency of the outer ring fault 2.97Hz and its related harmonics, demonstrating efficient diagnostic capabilities. Combined with the standard fault spectrum library for feature matching, the experiment successfully identified a variety of common fault types, including outer ring, inner ring, rolling element and cage faults, which greatly improved the accuracy and reliability of rolling bearing diagnosis and provided effective technical support for industrial sites.

[0094] In summary, the multi-step joint processing method combines innovative technical means such as virtual individual hybrid optimization, signal separation and modal noise reduction, and frequency domain feature extraction to significantly improve the effect of rolling bearing fault diagnosis. The experimental data fully verifies the practicality and superiority of the technical solution described in the present invention in a complex noise environment, which is an important technological innovation in the field of fault diagnosis.

[0095] The rolling bearing fault diagnosis system based on acoustic signals runs in any computing device such as a desktop computer, a laptop computer, a PDA or a cloud data center. The computing device includes: a processor, a memory and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps in the rolling bearing fault diagnosis method based on acoustic signals are implemented. The executable system may include, but is not limited to, a processor, a memory, and a server cluster.

[0096] An embodiment of the present invention provides a rolling bearing fault diagnosis system based on acoustic signals, such as Figure 2As shown, a rolling bearing fault diagnosis system based on acoustic signals in this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned rolling bearing fault diagnosis method based on acoustic signals are implemented. The processor executes the computer program to run in the following system units:

[0097] A signal processing unit, configured to calculate a signal aliasing matrix according to an acoustic signal of a rolling bearing, and obtain a hybrid matrix of the signal aliasing matrix by setting a plurality of virtual individuals in the signal aliasing matrix and causing the plurality of virtual individuals to perform hybridization, and perform an inverse transformation on the hybrid matrix as an estimated source signal;

[0098] A signal separation unit, used for separating and processing the estimated source signal to produce a source signal separation matrix;

[0099] A signal simulation unit is used to obtain modal components of various orders from the source signal separation matrix and filter out modal filter components therefrom;

[0100] The signal detection unit is used to reconstruct the signal of the modal filter component and then extract its frequency domain characteristics for the fault diagnosis of the rolling bearing.

[0101] Among them, in order to better unify the numerical linear relationship and probabilistic connection between physical quantities of different units of measurement, different physical quantities can be dimensionally processed.

[0102] Among them, preferably, all undefined variables in the present invention, if not clearly defined, can be manually set thresholds.

[0103] The rolling bearing fault diagnosis system based on acoustic signals can be run in computing devices such as desktop computers, laptops, PDAs, and cloud data centers. The rolling bearing fault diagnosis system based on acoustic signals includes, but is not limited to, processors and memories. Those skilled in the art can understand that the example is only an example of a rolling bearing fault diagnosis method, system, and device based on acoustic signals, and does not constitute a limitation on a rolling bearing fault diagnosis method, system, and device based on acoustic signals. It can include more or fewer components than the example, or a combination of certain components, or different components. For example, the rolling bearing fault diagnosis system based on acoustic signals can also include input and output devices, network access devices, buses, etc.

[0104] The present invention also provides an electronic device, a readable storage medium and a computer program product:

[0105] An electronic device comprises: 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 rolling bearing fault diagnosis method based on acoustic signals and the steps therein.

[0106] A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the rolling bearing fault diagnosis method based on acoustic signals and the steps therein.

[0107] A computer program product comprises a computer program. When the computer program is executed by a processor, the method for diagnosing rolling bearing faults based on acoustic signals and the steps thereof are implemented.

[0108] Wherein, electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0109] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0110] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0111] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, 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 foregoing.

[0112] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0113] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0114] A computer system may include clients and servers. Clients and servers are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship to each other.

[0115] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete component gate circuits or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the rolling bearing fault diagnosis system based on acoustic signals, and uses various interfaces and lines to connect the various sub-areas of the entire rolling bearing fault diagnosis system based on acoustic signals.

[0116] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the rolling bearing fault diagnosis method, system and device based on acoustic signals by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0117] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in the present disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in the present invention can be achieved, and this document is not limited here.

[0118] The present invention provides a rolling bearing fault diagnosis method, system and equipment based on acoustic signals. The method calculates a signal aliasing matrix according to the acoustic signal of the rolling bearing, sets a plurality of virtual individuals in the signal aliasing matrix and makes the plurality of virtual individuals perform aliasing to obtain a aliasing matrix of the signal aliasing matrix and performs an inverse transformation to estimate the source signal; separates and processes the estimated source signal to obtain a source signal separation matrix, obtains modal components of each order from the source signal separation matrix and selects modal filter components therefrom; reconstructs the modal filter components and extracts their frequency domain features for use in rolling bearing fault diagnosis, thereby significantly improving the accuracy and stability of signal separation and providing more reliable technical support for rolling bearing fault diagnosis.

[0119] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A rolling bearing fault diagnosis method based on acoustic signals, characterized in that: The method comprises: A signal aliasing matrix is ​​calculated according to the acoustic signal of the rolling bearing, and a plurality of virtual individuals are set in the signal aliasing matrix and the plurality of virtual individuals are mixed to obtain a mixing matrix of the signal aliasing matrix and perform an inverse transformation on the mixing matrix as an estimated source signal; the estimated source signal is separated to obtain a source signal separation matrix; each order modal component is obtained from the source signal separation matrix and a modal filter component is screened out therefrom; after signal reconstruction of the modal filter component, its frequency domain feature is extracted for use in rolling bearing fault diagnosis; Among them, the signal aliasing matrix is ​​calculated according to the acoustic signal of the rolling bearing, and a plurality of virtual individuals are set in the signal aliasing matrix and the plurality of virtual individuals are mixed to obtain the mixed matrix of the signal aliasing matrix and perform an inverse transformation on it as the estimated source signal, specifically: a plurality of virtual individuals are set in the signal aliasing matrix, each virtual individual represents a candidate aliasing matrix solution, and a candidate aliasing matrix solution is one of the eigenvalues ​​of the signal aliasing matrix, and the position of the virtual individual is initialized to represent its fitness in the signal aliasing matrix; in each round of mixing, each virtual individual is placed in the signal aliasing matrix and moves once in each round of mixing, and each direction is compared. The number of virtual individuals existing on the signal aliasing matrix and the fitness value of each virtual individual are selected for the movement direction of the virtual individuals according to the number of virtual individuals existing in each direction and the fitness value of each virtual individual. When there are at least two virtual individuals in adjacent positions, the fitness values ​​of the virtual individuals in adjacent positions are mixed. When the mathematical expectation of the second partial derivative values ​​of the positions of each virtual individual and their corresponding fitness values ​​tends to zero, point-to-point multiplication is performed with the values ​​of the corresponding positions in the signal aliasing matrix corresponding to the positions of each virtual individual, and the obtained matrix is ​​used as the hybrid matrix of the signal aliasing matrix, and the hybrid matrix is ​​inversely transformed to obtain the estimated source signal.

2. A rolling bearing fault diagnosis method based on acoustic signals according to claim 1, characterized in that: in, In the process of obtaining each order modal component from the source signal separation matrix, a modal decomposition method is used to obtain each order modal component from the source signal separation matrix. The modal decomposition method disturbs the extreme points of the source signal separation matrix by adding symmetrical noise to the source signal separation matrix.

3. A rolling bearing fault diagnosis method based on acoustic signals according to claim 2, characterized in that: in, In the step of reconstructing the signal of the modal filter components, the cross-correlation coefficients of the modal components are first calculated, and only the modal components with cross-correlation coefficients higher than a preset threshold are retained for signal reconstruction.

4. A rolling bearing fault diagnosis method based on acoustic signals according to claim 3, characterized in that: in, The reconstructed signal is transformed into frequency domain features to identify the fault type of the rolling bearing.

5. The rolling bearing fault diagnosis method based on acoustic signals according to claim 1 is characterized in that: in, The method for separating the estimated source signal to obtain the source signal separation matrix is: Independent component analysis is performed on the estimated source signal, and the independent components in the estimated source signal are separated by an independent component analysis algorithm for optimizing the signal to separate the source signal separation matrix.

6. A rolling bearing fault diagnosis method based on acoustic signals according to claim 1, characterized in that: in, The method for separating the estimated source signal to obtain the source signal separation matrix is: After centering and whitening the estimated source signal, an independence matrix is ​​calculated; The independence matrix is ​​a tensor that can calculate the non-Gaussian measurement value of the acoustic signal based on the acoustic signal of the rolling bearing and can be used to measure the independence of the acoustic signal of the rolling bearing: the estimated source signal after centering and whitening is processed with a nonlinear characteristic function matrix to obtain an independence matrix, wherein the matrix size of the nonlinear characteristic function matrix is ​​consistent with the estimated source signal, each row and column position in the nonlinear characteristic function matrix uses the same type of nonlinear characteristic function, but the weights of the variables in the nonlinear characteristic function used are not completely the same in each row and column position, and the nonlinear characteristic function matrix can be used to maximize the non-Gaussian measurement value in the estimated source signal after centering and whitening by transforming the weights of the variables in the nonlinear characteristic function used in each row and column position; The independence matrix is ​​orthogonalized to obtain a source signal separation matrix.

7. A rolling bearing fault diagnosis system based on acoustic signals, characterized in that: The rolling bearing fault diagnosis system based on acoustic signals runs in any computing device such as a desktop computer, a laptop computer or a cloud data center. The computing device includes: a processor, a memory and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps in the rolling bearing fault diagnosis method based on acoustic signals as described in any one of claims 1 to 6 are implemented.

8. An electronic device, comprising: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 6.

Images