Variable-scale evolutionary adaptive noise cancellation method, noise cancellation system and fault diagnosis system

Abstract

The present invention belongs to the technical field of eliminating complex and strong background noise in gearboxes. It specifically discloses a variable-scale evolutionary adaptive denoising method, denoising system, and fault diagnosis system. The denoising method includes the following steps: S1, obtaining and initializing the original vibration signal; S2, establishing a variational model, solving the variational problem, and updating the modal parameters until convergence conditions are met; S3, using the updated modal parameters, calculating the kurtosis coefficient under each K and α condition to determine the center frequency and bandwidth of the bandpass filter; S4, constructing a signal model of the adaptive filter to obtain optimal filtering parameters; S5, based on the optimal filtering parameters, obtaining two sets of decomposed signals at different scales; S6, adaptively filtering the decomposed signals at different scales to obtain a characteristic signal after denoising. Using this technical solution, a composite fault signal separation algorithm based on variable-scale evolutionary adaptive decomposition is used to denoise the vibration signal and obtain a characteristic signal with better denoising effect.

Description

Technical Field

[0001] The present invention belongs to the technical field of eliminating complex strong background noise of a gearbox, and relates to a variable-scale evolutionary adaptive noise elimination method, a noise elimination system and a fault diagnosis system. Background Art

[0002] Mechanical equipment is a vital component of the manufacturing industry and can be found everywhere in production and daily life. It generally carries out important tasks such as transmitting power and bearing loads. The working environment of mechanical equipment is often very harsh, such as high temperature, impact, heavy loads, and high speeds. Ensuring the safety and reliability of mechanical equipment in such harsh environments has become an unavoidable issue, which is of great significance to protecting people's lives and property.

[0003] Expensive mechanical equipment such as transportation vehicles, machine tools, and wind turbines all have crucial transmission components, and most often utilize gear transmission systems. This is due to their advantages, including a constant transmission ratio, high operational stability, a wide range of transmission ratio variations, and adaptability to both deceleration and acceleration. Bearings and gears are crucial components in gear transmission systems, yet they are also vulnerable links prone to failure. Promptly diagnosing potentially faulty mechanical parts has become a hotly debated issue.

[0004] Currently, vibration detection technology is the most widely used in the field of mechanical fault diagnosis. This technology traditionally uses vibration sensors to extract mechanical vibration signal data, analyzes the data in the time and frequency domains, and identifies typical fault characteristics and frequencies. However, due to the increasing complexity of the mechanical industry's production environment and some relatively extreme working environments in recent years, the process of collecting mechanical vibration signals is often accompanied by a large amount of noise signals, which can sometimes even mask the fault signal itself. This background noise is generally colored noise, which is more complex than white noise.

[0005] In addition, mechanical failures generally do not occur in a single form, but in a composite form with multiple failures occurring simultaneously, which makes the extraction of fault features more difficult. Therefore, how to effectively remove noise and amplify fault features when multiple failures occur simultaneously is a difficulty that needs to be solved. Commonly used signal denoising methods include: Empirical Mode Decomposition (EMD), Discrete Wavelet Transform and Variational Mode Decomposition (VMD). These traditional methods all have some shortcomings, such as: the EMD algorithm has problems with intrinsic function selection criteria, endpoint effects and modal aliasing, and the matching degree between the wavelet basis function and the fault characteristic waveform in the wavelet transform still needs to be improved. Summary of the Invention

[0006] The object of the present invention is to provide a scale-variable evolutionary adaptive denoising method, a denoising system and a fault diagnosis system to obtain a characteristic signal with better denoising effect.

[0007] In order to achieve the above object, the basic scheme of the present invention is: a variable scale evolutionary adaptive denoising method, comprising the following steps:

[0008] S1, obtain the original vibration signal and initialize it. The initialization parameters include a set of decomposition layer number K and penalty factor α value, as well as convergence conditions;

[0009] S2, establish a variational model, solve the variational problem, and update the modal parameters until the convergence conditions are met;

[0010] S3, using the updated modal parameters, calculate the kurtosis coefficient under each K and α, and determine the center frequency and bandwidth of the bandpass filter according to the maximum kurtosis index principle;

[0011] S4, constructing a signal model of an adaptive filter, inputting the center frequency and bandwidth of the bandpass filter into the signal model, and obtaining the optimal filtering parameters through a filtering algorithm;

[0012] S5, performing band-pass filtering on the main input signal and the reference signal of the filter signal model according to the optimal filtering parameters to obtain two sets of decomposed signals of different scales;

[0013] S6, adaptively filtering the decomposed signals of different scales to obtain the characteristic signals after noise reduction.

[0014] The working principle and beneficial effects of this basic scheme are as follows: This scheme iteratively selects the optimal decomposition scale by maximizing the kurtosis index principle, that is, finding the optimal number of decomposition layers and penalty factor. It also determines the center frequency and bandwidth based on the maximum kurtosis index principle. These optimal filter parameters are aggregated into a set of filter parameters, and this filter set is used to simultaneously decompose the reference signal and the input signal to the corresponding scales. At different scales, the signal is subjected to biological evolutionary adaptive filtering (EDM evolutionary filtering) to obtain the denoised characteristic signal. This overcomes the limitation of the traditional VMD algorithm, where the number of decomposition layers and penalty factors must be manually set and cannot be easily determined. It is more flexible and adaptable, and can obtain characteristic signals with better denoising effects.

[0015] Furthermore, the modal parameters in the original vibration signal include modal function, center frequency, Lagrange multiplier, convergence accuracy and number of iterations.

[0016] The original vibration signal contains the required parameters, which is convenient for subsequent use.

[0017] Furthermore, the method of establishing the variational model is as follows:

[0018] The Hilbert transform is used to calculate the analytical signal of each mode and obtain the single-sided spectrum of each analytical signal. The analytical signal of the mode function is:

[0019]

[0020] Where t represents time, δ(t) represents the impulse function, and u k (t) represents the kth mode at time t, j is the imaginary unit;

[0021] The spectrum of each mode is modulated to the corresponding baseband by mixing the predicted center frequencies of each spectrum:

[0022]

[0023] Where, ω k represents the k-th predicted center frequency;

[0024] Calculate the squared norm of the gradient of the above formula to obtain the variational model:

[0025]

[0026]

[0027] In the formula, {u k (t)}:={u1(t),u2(t),…,u k (t)} and {ω k}:={ω1,…,ω k} represent all mode functions and center frequencies respectively.

[0028] Establishing a variational model is simple and easy to use.

[0029] Furthermore, the method for updating the modal parameters is as follows:

[0030] Solve the quadratic optimization problem in the frequency domain using the Parseval / Plancherel-Fourier isometry method:

[0031]

[0032] in, represents the fast Fourier transform of the kth mode, refers to the fast Fourier transform of the input signal, represents the fast Fourier transform of the i-th mode; ω k represents the kth predicted center frequency, u k is the modal function, α is the penalty factor, represents the fast Fourier transform of the Lagrange multiplier, j is the imaginary unit;

[0033] Replace the variable ω in the first term with ω-ω k ,get:

[0034]

[0035] Using the Hermitian symmetry of real signals, we can rewrite the two terms as 0-∞ integrals over non-negative frequencies:

[0036]

[0037] By eliminating the first variation of the positive frequency, the solution to the quadratic optimization problem is obtained, that is, the modal components are decomposed:

[0038]

[0039] Similarly, we get the center frequency ω k :

[0040]

[0041] Among them, the center frequency ω k Located at the center of gravity of the power spectrum of this modal component, represents the modal component of the last iteration, represents the modal component of the previous iteration;

[0042] The update formula of Lagrange multiplier is:

[0043]

[0044] Update modal parameters for subsequent use.

[0045] Further, iteratively update the modal function Center frequency and Lagrange multiplier λ(t) until the convergence condition is met, and the time domain waveforms of k modes are obtained by inverse Fourier transform:

[0046]

[0047] Among them, ε represents the convergence accuracy, represents the modal component of the last iteration, represents the modal component of the previous iteration.

[0048] Simple operation and easy to use.

[0049] Furthermore, the method for determining the center frequency and bandwidth of the bandpass filter is as follows:

[0050] For a sequence x=[x1,x2,…,x N The kurt coefficient of ] is defined as follows:

[0051]

[0052] In the formula, μ represents the mean of the sequence, σ represents the standard deviation of the sequence, and E represents the expectation;

[0053] Select the maximum value from all kurtosis values and determine the optimal center frequency ω according to the corresponding decomposition layer number K and penalty factor α k ;

[0054] Similarly, for the decomposition parameter α corresponding to the maximum kurtosis, after obtaining a set of modal component IMFs and corresponding center frequencies through VMD decomposition, the upper and lower frequency cutoff points corresponding to 0.1 times the center frequency amplitude of each modal component spectrum [f 下 ,f 上 ], the required bandwidth is:

[0055] B=f 上 -f 下 .

[0056] Determine the center frequency and bandwidth of the bandpass filter for subsequent use.

[0057] Furthermore, the method for obtaining the optimal filtering parameters is as follows:

[0058] The input of the adaptive filter is two signals, namely the main input signal and the reference signal, and its signal model is as follows:

[0059]

[0060] Where x represents the main input signal, x' represents the reference signal, and N represents the number of data points of the signal;

[0061] By using the variable scale filtering characteristics of the signal using the VMD algorithm, a set of bandpass filter parameters are obtained:

[0062]

[0063] Where V (K,α) *{·} indicates that under the conditions of the optimal number of decomposition layers and penalty factors (K, α), the optimal bandpass filter parameter set for signal · is obtained according to the VMD algorithm. Represents the center frequency of the bandpass filter parameter group, T is the transpose; B K =[(f down1 ,f up1 ),(f down2 ,f up2 ),…,(fdownK ,f upK )] T represents the bandwidth of the bandpass filter parameter group, where the bandwidth is determined by the lower cutoff frequency f down and upper cutoff frequency f up It consists of two parameters;

[0064] After the determined center frequency and bandwidth are brought into the filter signal model, the main input signal and the reference signal are filtered respectively to obtain a set of signal pairs consisting of the components of the main input signal and the reference signal, which are:

[0065]

[0066] The symbol · represents a filtering operation, the left side of the symbol represents a set of Butterworth bandpass filter parameters, and the right side of the symbol represents the signal to be analyzed; i1 with y i2 The signal pair obtained after filtering is formed, namely:

[0067]

[0068] Among them, y i1 with y i2 They represent the signals obtained after the main input signal and the reference signal are filtered by the i-th band-pass filter;

[0069] For each signal pair, the EDF algorithm is used for filtering to obtain the filtered output of the i-th component signal of the main input signal x, that is:

[0070]

[0071] Among them, the symbol Represents the biological evolution theory adaptive filtering operation. The right side of the symbol represents the processed signal pair, and the left side of the symbol represents the filtering parameters of the EDF adaptive filter:

[0072]

[0073] Where w(a,b) represents the optimal IIRF after optimization by the biological evolution algorithm.

[0074] Obtain the optimal filtering parameters for subsequent use.

[0075] Furthermore, the method for obtaining the characteristic signal after noise reduction is as follows:

[0076] Perform biological evolutionary filtering on the decomposed signals of the corresponding scale to obtain a set of filtered component signals, and calculate the Pearson correlation coefficient between them and the original main input signal;

[0077] Using the Pearson correlation coefficient as the weight of the component signal, the component signal after EDF filtering is weighted and reconstructed to obtain the denoised characteristic signal Y i , the expression of Pearson correlation coefficient is as follows:

[0078]

[0079] Among them, x and y represent two variables. represent the means of the two variables.

[0080] At different scales, the signal is subjected to biological evolution adaptive filtering to obtain the characteristic signal after noise reduction. The operation is simple and easy to use.

[0081] The present invention also provides a variable-scale evolutionary adaptive noise cancellation system, which includes a data acquisition module and a processing module. The data acquisition module is used to collect original vibration signals. The output end of the data acquisition module is connected to the input end of the processing module. The processing module executes the method described in the present invention to perform signal noise cancellation.

[0082] By using this system, we can break the limitations of the traditional VMD algorithm in which the number of decomposition layers K and the penalty factor α need to be manually set and are difficult to determine, and achieve a more flexible and adaptable noise reduction operation.

[0083] The present invention also provides a gearbox fault diagnosis system, including the adaptive noise reduction system described in the present invention, a feature extraction and dimensionality reduction module, and a fault diagnosis module. The feature extraction and dimensionality reduction module is used to receive the feature signal after noise reduction, and extract and reduce the fault features. The fault diagnosis module is used to receive the fault features after dimensionality reduction and input them into the fault diagnosis model preset therein to obtain the fault diagnosis results.

[0084] The diagnostic system is used to complete the fault diagnosis of the bearing, which is convenient for use. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 Schematic diagram of the process of the variable-scale evolutionary adaptive denoising method of the present invention;

[0086] Figure 2 It is a kurtosis diagram of the VMD decomposition and reconstruction signal of a preferred embodiment of the present invention;

[0087] Figure 3 It is a schematic diagram of decomposition signals of a preferred embodiment of the present invention;

[0088] Figure 4 This is a time domain waveform diagram after noise reduction of a preferred solution of the present invention. DETAILED DESCRIPTION

[0089] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0090] In the description of the present invention, it should be understood that the terms "longitudinal", "transverse", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention.

[0091] In the description of the present invention, unless otherwise specified and limited, it should be noted that the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a mechanical connection or an electrical connection, or it can be the internal communication between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0092] The present invention aims at the problem of signal decoupling and denoising (multi-scale information enhancement), and discloses a variable scale evolutionary adaptive denoising method based on the VMD algorithm, which has greater flexibility and adaptability. Figure 1 As shown, the method includes the following steps:

[0093] S1, obtain the original vibration signal and initialize it. The initialization parameters include a set of decomposition layer number K and penalty factor α value, as well as convergence conditions. Given a set of decomposition layer number K and penalty factor α value, that is, an initial range and step size of K and α are given respectively; the modal parameters in the original vibration signal include modal function, center frequency, Lagrange multiplier, convergence accuracy and number of iterations, and the modal function u k , center frequency ω k , Lagrange multiplier λ, convergence accuracy ε, number of iterations, etc. are initialized.

[0094] S2, establish a variational model, solve the variational problem, and update the modal parameters until the convergence conditions are met;

[0095] S3, using the updated modal parameters, calculate the kurtosis coefficient under each K and α, and determine the center frequency and bandwidth of the bandpass filter according to the maximum kurtosis index principle;

[0096] S4, constructing a signal model of an adaptive filter, inputting the center frequency and bandwidth of the bandpass filter into the signal model, and obtaining the optimal filtering parameters through a filtering algorithm;

[0097] S5, based on the optimal filtering parameters, bandpass filtering is performed on the main input signal and reference signal of the filter signal model to obtain two sets of decomposed signals of different scales; the main input signal refers to a signal with a large signal-to-noise ratio, and the reference signal refers to a signal with a small signal-to-noise ratio and a large amount of noise, such as the bearing vibration signal (main input signal) and the test bench base vibration signal (reference signal);

[0098] S6, adaptively filtering the decomposed signals of different scales to obtain the characteristic signals after noise reduction.

[0099] In a preferred embodiment of the present invention, the method for establishing the variational model is as follows:

[0100] The Hilbert transform is used to calculate the analytical signal of each mode and obtain the single-sided spectrum of each analytical signal. The analytical signal of the mode function is:

[0101]

[0102] Where t represents time, δ(t) represents the impulse function, and u k (t) represents the kth mode at time t, j is the imaginary unit;

[0103] The spectrum of each mode is modulated to the corresponding baseband by mixing the predicted center frequencies of each spectrum:

[0104]

[0105] Where, ω k represents the k-th predicted center frequency;

[0106] Calculate the squared norm of the gradient of the above formula to obtain the variational model for evaluating the bandwidth:

[0107]

[0108]

[0109] In the formula, {u k (t)}:={u1(t),u2(t),…,u k (t)} and {ω k}:={ω1,…,ω k} represent all mode functions and center frequencies respectively.

[0110] In a preferred embodiment of the present invention, the method for updating the modal parameters is as follows:

[0111] By introducing the penalty term and the Lagrange multiplier λ, the problem is transformed into an unconstrained optimization problem, and the augmented Lagrangian expression is obtained:

[0112]

[0113] Where, the penalty factor α ensures the reconstruction accuracy, while the Lagrange multiplier λ makes the problem unconstrained; ω k represents the center frequency, u k is the modal function; δ(t) represents the impulse function, and f(t) represents the input signal;

[0114] Apply the alternating direction method of multipliers (ADMM) to solve the variational problem and update the mode function Center frequency and the Lagrange multiplier λ(t), computing the saddle points of the augmented Lagrangian expression;

[0115] Solve the equivalent minimization problem and get the decomposition pattern

[0116]

[0117] Among them, u i (t) is the i-th mode;

[0118] Solve the quadratic optimization problem in the frequency domain using the Parseval / Plancherel-Fourier isometry method:

[0119]

[0120] in, represents the fast Fourier transform of the kth mode, refers to the fast Fourier transform of the input signal, represents the fast Fourier transform of the i-th mode; ω k represents the kth predicted center frequency, u k is the modal function, α is the penalty factor, represents the fast Fourier transform of the Lagrange multiplier, j is the imaginary unit;

[0121] Replace the variable ω in the first term of the above formula with ω-ω k ,get:

[0122]

[0123] Using the Hermitian symmetry of real signals, the two terms in the above equation can be rewritten as 0-∞ integrals over non-negative frequencies:

[0124]

[0125] By eliminating the first variation of the positive frequency, the solution to the quadratic optimization problem is obtained, that is, the modal components are decomposed:

[0126]

[0127] Similarly, we get the center frequency ω k :

[0128]

[0129] Among them, the center frequency ω k Located at the center of gravity of the power spectrum of this modal component, represents the modal component of the last iteration, represents the modal component of the previous iteration;

[0130] The update formula of Lagrange multiplier is:

[0131]

[0132] Iteratively update the modal function Center frequency and Lagrange multiplier λ(t) until the convergence condition is met, and the time domain waveforms of k modes are obtained by inverse Fourier transform:

[0133]

[0134] Among them, ε represents the convergence accuracy, represents the modal component of the last iteration, represents the modal component of the previous iteration.

[0135] In a preferred embodiment of the present invention, the method for determining the center frequency and bandwidth of the bandpass filter is as follows:

[0136] For a sequence x=[x1,x2,…,x N The kurt coefficient of ] is defined as follows:

[0137]

[0138] In the formula, μ represents the mean of the sequence, σ represents the standard deviation of the sequence, and E represents the expectation;

[0139] Select the maximum value from all kurtosis values and determine the optimal center frequency ω according to the corresponding decomposition layer number K and penalty factor α k ;

[0140] Similarly, for the optimal parameter decomposition α (i.e., the one corresponding to the maximum kurtosis), after obtaining a set of modal component IMFs and corresponding center frequencies through VMD decomposition, the upper and lower frequency cutoff points corresponding to 0.1 times the center frequency amplitude of each modal component spectrum [f 下 ,f 上 ], the required bandwidth is:

[0141] B=f 上 -f 下 .

[0142] In a preferred embodiment of the present invention, the method for obtaining the optimal filtering parameters is as follows:

[0143] The input of the adaptive filter is two signals, namely the main input signal and the reference signal, and its signal model is as follows:

[0144]

[0145] Where x represents the main input signal, x' represents the reference signal, and N represents the number of data points of the signal;

[0146] By using the variable scale filtering characteristics of the signal using the VMD algorithm, a set of bandpass filter parameters are obtained:

[0147]

[0148] Where V (K,α) *{·} indicates that under the conditions of the optimal number of decomposition layers and penalty factors (K, α), the optimal bandpass filter parameter set for signal · is obtained according to the VMD algorithm. Represents the center frequency of the bandpass filter parameter group, T is the transpose; B K =[(f down1 ,f up1 ),(f down2 ,f up2 ),…,(f downK ,f upK )] T represents the bandwidth of the bandpass filter parameter group, where the bandwidth is determined by the lower cutoff frequency f down and upper cutoff frequency f up It consists of two parameters;

[0149] After the determined center frequency and bandwidth are brought into the filter signal model, the main input signal and the reference signal are filtered respectively to obtain a set of signal pairs consisting of the components of the main input signal and the reference signal, which are:

[0150]

[0151] The symbol · represents a filtering operation, the left side of the symbol represents a set of Butterworth bandpass filter parameters, and the right side of the symbol represents the signal to be analyzed; i1 with y i2 The signal pair obtained after filtering is formed, namely:

[0152]

[0153] Among them, y i1 with y i2 They represent the signals obtained after the main input signal and the reference signal are filtered by the i-th band-pass filter;

[0154] For each signal pair, the EDF algorithm is used for filtering to obtain the filtered output of the i-th component signal of the main input signal x, that is:

[0155]

[0156] Among them, the symbol It represents the evolutionary adaptive filtering operation (EDF algorithm). The right side of the symbol represents the processed signal pair, and the left side of the symbol represents the filtering parameters of the EDF adaptive filter:

[0157]

[0158] Where w(a,b) represents the optimal IIRF (filter) after optimization by the biological evolution algorithm.

[0159] In a preferred embodiment of the present invention, the method for obtaining the characteristic signal after noise reduction is as follows:

[0160] According to the maximum kurtosis principle, the optimal K and α are selected, and the decomposition signal of the corresponding scale is subjected to biological evolutionary filtering to obtain a set of filtered component signals, and the Pearson correlation coefficient between the component signals and the original main input signal is calculated.

[0161] Using the Pearson correlation coefficient as the weight of the component signal, the component signal after EDF filtering is weighted and reconstructed to obtain the denoised characteristic signal Y i , the expression of Pearson correlation coefficient is as follows:

[0162]

[0163] Among them, x and y represent two variables. They represent the means of the two variables, and N represents the number of data points of the signal.

[0164] This approach overcomes the limitations of traditional VMD algorithms, where the number of decomposition levels K and the penalty factor α must be manually set and are difficult to determine. Instead, it proposes an iterative approach to select the optimal decomposition scale by maximizing the kurtosis index, thereby finding the optimal number of decomposition levels K and penalty factor α. The center frequency and bandwidth are also determined based on the kurtosis index maximization principle. These optimal filter parameters are then aggregated into a set of filter parameters, which are then used to simultaneously decompose the reference and input signals into corresponding scales. Finally, at different scales, the signal is subjected to evolutionary adaptive filtering (EDM) to obtain the de-noised characteristic signal.

[0165] In a preferred embodiment of the present invention, a signal segment is selected as the analysis object, and the signal-to-noise ratio of the input signal is set to -2dB. The main input signal is subjected to VMD traversal decomposition, wherein the decomposition level K is traversed from 2 to 10, and the penalty coefficient α is traversed in the range of 1000 to 15000. The kurtosis graph of the VMD decomposition and reconstructed signal is calculated as follows: Figure 2 As shown in the figure, it can be seen that when the number of decomposition layers is 2 and the penalty coefficient is 6000, the maximum kurtosis value is 4.779. According to the optimal parameters, the main input signal is band-pass filtered to obtain the decomposition signal, as shown in Figure 3 As shown in the figure, the center frequencies of the two component signals are 3000Hz and 8000Hz respectively. The population size of the biological evolution algorithm involved in this solution is set to 100, the maximum number of iterations is 100, and the signal length is 0.1s for analysis. The time domain waveform after noise reduction is very clean, the periodic impact phenomenon is obvious, and the noise signal is basically suppressed. Figure 4 The method proposed in the present invention has obvious noise reduction advantages compared with traditional noise reduction methods (such as EMD, VMD, etc.) or EDF adaptive noise reduction algorithm.

[0166] The present invention also provides a variable-scale evolutionary adaptive noise cancellation system, which includes a data acquisition module and a processing module. The data acquisition module is used to collect original vibration signals. The output end of the data acquisition module is connected to the input end of the processing module. The processing module executes the method described in the present invention to perform signal noise cancellation.

[0167] By using this system, we can break the limitations of the traditional VMD algorithm in which the number of decomposition layers K and the penalty factor α need to be manually set and are difficult to determine, and achieve a more flexible and adaptable noise reduction operation.

[0168] The present invention also provides a gearbox fault diagnosis system, including the adaptive noise reduction system described in the present invention, a feature extraction and dimensionality reduction module, and a fault diagnosis module. The feature extraction and dimensionality reduction module is used to receive the feature signal after noise reduction, and extract and reduce the fault features. The fault diagnosis module is used to receive the fault features after dimensionality reduction and input them into the fault diagnosis model preset therein to obtain the fault diagnosis results.

[0169] The diagnostic system is used to complete the fault diagnosis of the bearing, which is convenient for use.

[0170] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0171] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A scale-variable evolutionary adaptive denoising method, characterized in that: The steps include: S1, obtain the original vibration signal and initialize it. The initialization parameters include a set of decomposition layer number K and penalty factor α value, as well as convergence conditions; S2, establish a variational model, solve the variational problem, and update the modal parameters until the convergence conditions are met; S3, using the updated modal parameters, calculate the kurtosis coefficient under each K and α, and determine the center frequency and bandwidth of the bandpass filter according to the maximum kurtosis index principle; S4, constructing a signal model of an adaptive filter, inputting the center frequency and bandwidth of the bandpass filter into the signal model, and obtaining the optimal filtering parameters through a filtering algorithm; S5, performing band-pass filtering on the main input signal and the reference signal of the filter signal model according to the optimal filtering parameters to obtain two sets of decomposed signals of different scales; S6, adaptively filtering the decomposed signals at different scales to obtain the characteristic signals after noise reduction; The method to obtain the optimal filtering parameters is as follows: The input of the adaptive filter is two signals, namely the main input signal and the reference signal, and its signal model is as follows: Where x represents the main input signal, x' represents the reference signal, and N represents the number of data points of the signal; By using the variable scale filtering characteristics of the signal using the VMD algorithm, a set of bandpass filter parameters are obtained: Where V (K,α) *{·} indicates that under the conditions of the optimal number of decomposition layers and penalty factors (K, α), the optimal bandpass filter parameter set for signal · is obtained according to the VMD algorithm. Represents the center frequency of the bandpass filter parameter group, T is the transpose; B K =[(f down1 ,f up1 ),(f down2 ,f up2 ),…,(f downK ,f upK )] T represents the bandwidth of the bandpass filter parameter group, where the bandwidth is determined by the lower cutoff frequency f down and upper cutoff frequency f up It consists of two parameters; After the determined center frequency and bandwidth are brought into the filter signal model, the main input signal and the reference signal are filtered respectively to obtain a set of signal pairs consisting of the components of the main input signal and the reference signal, which are: The symbol · represents a filtering operation, the left side of the symbol represents a set of Butterworth bandpass filter parameters, and the right side of the symbol represents the signal to be analyzed; i1 with y i2 The signal pair obtained after filtering is formed, namely: Among them, y i1 with y i2 They represent the signals obtained after the main input signal and the reference signal are filtered by the i-th band-pass filter; For each signal pair, the EDF algorithm is used for filtering to obtain the filtered output of the i-th component signal of the main input signal x, that is: Among them, the symbol Represents the biological evolution theory adaptive filtering operation. The right side of the symbol represents the processed signal pair, and the left side of the symbol represents the filtering parameters of the EDF adaptive filter: Where w(a,b) represents the optimal IIRF after optimization by the biological evolution algorithm.

2. The variable-scale evolutionary adaptive denoising method according to claim 1, characterized in that: The modal parameters in the original vibration signal include modal function, center frequency, Lagrange multiplier, convergence accuracy and number of iterations.

3. The variable-scale evolutionary adaptive denoising method according to claim 1, wherein: The method to build a variational model is as follows: The Hilbert transform is used to calculate the analytical signal of each mode and obtain the single-sided spectrum of each analytical signal. The analytical signal of the mode function is: Where t represents time, δ(t) represents the impulse function, and u k (t) represents the kth mode at time t, j is the imaginary unit; The spectrum of each mode is modulated to the corresponding baseband by mixing the predicted center frequencies of each spectrum: Where, ω k represents the k-th predicted center frequency; Calculate the squared norm of the gradient of the above formula to obtain the variational model: In the formula, {u k (t)}:={u1(t),u2(t),…,u k (t)} and {ω k }:={ω1,…,ω k } represent all mode functions and center frequencies respectively.

4. The variable-scale evolutionary adaptive denoising method according to claim 1, wherein: The method to update the modal parameters is as follows: Solve the quadratic optimization problem in the frequency domain using the Parseval / Plancherel-Fourier isometry method: in, represents the fast Fourier transform of the kth mode, refers to the fast Fourier transform of the input signal, represents the fast Fourier transform of the i-th mode; ω k represents the kth predicted center frequency, u k is the modal function, α is the penalty factor, represents the fast Fourier transform of the Lagrange multiplier, j is the imaginary unit; Replace the variable ω in the first term with ω-ω k ,get: Using the Hermitian symmetry of real signals, we can rewrite the two terms as 0-∞ integrals over non-negative frequencies: By eliminating the first variation of the positive frequency, the solution to the quadratic optimization problem is obtained, that is, the modal components are decomposed: Similarly, we get the center frequency ω k : Among them, the center frequency ω k Located at the center of gravity of the power spectrum of this modal component, represents the modal component of the last iteration, represents the modal component of the previous iteration; The update formula of Lagrange multiplier is:

5. The variable-scale evolutionary adaptive denoising method according to claim 1 or 4, characterized in that: Iteratively update the modal function u k n (t), center frequency ω k n and Lagrange multiplier λ(t) until the convergence condition is met, and the time domain waveforms of k modes are obtained by inverse Fourier transform: Among them, ε represents the convergence accuracy, represents the modal component of the last iteration, represents the modal component of the previous iteration.

6. The variable-scale evolutionary adaptive denoising method according to claim 1, wherein: The method for determining the center frequency and bandwidth of the bandpass filter is as follows: For a sequence x=[x1,x2,…,x N The kurt coefficient of ] is defined as follows: In the formula, μ represents the mean of the sequence, σ represents the standard deviation of the sequence, and E represents the expectation; Select the maximum value from all kurtosis values and determine the optimal center frequency ω according to the corresponding decomposition layer number K and penalty factor α k ; Similarly, for the decomposition parameter α corresponding to the maximum kurtosis, after obtaining a set of modal component IMFs and corresponding center frequencies through VMD decomposition, the upper and lower frequency cutoff points corresponding to 0.1 times the center frequency amplitude of each modal component spectrum [f 下 ,f 上 ], the required bandwidth is: B=f 上 -f 下 。 7. The variable-scale evolutionary adaptive denoising method according to claim 1, wherein: The method for obtaining the characteristic signal after noise reduction is as follows: Perform biological evolutionary filtering on the decomposed signals of the corresponding scale to obtain a set of filtered component signals, and calculate the Pearson correlation coefficient between them and the original main input signal; Using the Pearson correlation coefficient as the weight of the component signal, the component signal after EDF filtering is weighted and reconstructed to obtain the denoised characteristic signal Y i , the expression of Pearson correlation coefficient is as follows: Among them, x and y represent two variables. They represent the means of the two variables, and N represents the number of data points of the signal.

8. A scale-variable evolutionary adaptive noise cancellation system, characterized in that: It includes a data acquisition module and a processing module. The data acquisition module is used to collect original vibration signals. The output end of the data acquisition module is connected to the input end of the processing module. The processing module executes the method according to one of claims 1 to 7 to perform signal noise reduction.

9. A gearbox fault diagnosis system, characterized in that: It includes the adaptive noise reduction system according to claim 8, a feature extraction and dimensionality reduction module, and a fault diagnosis module. The feature extraction and dimensionality reduction module is used to receive the feature signal after noise reduction, and extract and reduce the fault features. The fault diagnosis module is used to receive the fault features after dimensionality reduction and input them into the fault diagnosis model preset therein to obtain the fault diagnosis result.

Images