High-reliability gearbox signal denoising method, system, medium and equipment

Abstract

The invention discloses a high-reliability planetary gearbox signal denoising method, system, medium and equipment, and the method comprises the steps: obtaining original vibration signals of a nuclear power circulating pump planetary gearbox in different health states, and carrying out the detrending and demean preprocessing of the signals; performing modal decomposition on the original vibration signal by adopting an empirical mode decomposition (EMD) algorithm, an ensemble empirical mode decomposition (EEMD) algorithm and a variational mode decomposition (VMD) algorithm to obtain a plurality of different modal components; iteratively optimizing a hyper-parameter value in the variational mode decomposition algorithm VMD by adopting a sparrow search algorithm SSA so as to realize the self-adaptive decomposition of the variational mode decomposition algorithm VMD on the signal; and carrying out modal decomposition on the gearbox vibration signal by adopting a variational modal decomposition algorithm VMD after iterative optimization, removing noise components in modal components, and reconstructing the signal to realize gearbox vibration signal denoising.

Description

Technical Field

[0001] The present invention relates to the field of high-reliability equipment data processing technology, and more particularly to a high-reliability planetary gearbox signal denoising method, system, medium, and device. Specific applications include high-reliability planetary gearboxes and engineering machinery planetary gearboxes. Background Art

[0002] Planetary gearboxes operate constantly in low-speed, heavy-load environments, resulting in low failure rates, high maintenance costs, and high risks. Their raw signal components are extremely complex, encompassing multiple signals, including gear meshing and bearing operation. Furthermore, strong on-site electromagnetic interference, mechanical vibration, and other noise can contribute to significant non-stationary and nonlinear characteristics. Under complex operating conditions, such as sudden load changes or fluctuating water quality, when pump vibration is abnormal, the characteristic frequency components of the fault are easily overwhelmed by background noise, making it difficult to accurately extract them using traditional signal processing methods. This makes early fault detection difficult, hindering effective fault warnings and significantly increasing the risk of unplanned downtime. In the nuclear power sector, planetary gearboxes in circulating water pumps, for example, bear the heavy responsibility of driving circulating water pumps and must operate stably under harsh, long-term low-speed, heavy-load conditions. In the construction machinery sector, excavators, loaders, cranes, and other types of construction machinery operate under complex operating conditions, requiring planetary gearboxes to frequently change speeds and directions, while also bearing high load torques. This poses severe challenges to their reliability, durability, and maintainability.

[0003] There are many existing vibration signal denoising methods, which have been studied in both the time-frequency domain and deep learning. In the time-frequency domain, there are methods such as wavelet threshold denoising, singular value decomposition, and empirical mode decomposition; in machine learning, especially deep learning, there are methods such as artificial neural networks, convolutional neural networks, and autoencoders. However, when performing signal decomposition or feature extraction, using only one of the above methods may be limited by the single algorithm. For example, the choice of threshold and wavelet basis in the wavelet threshold denoising method has a significant impact on the signal denoising effect; the singular value decomposition method has a high computational complexity, and the gain effect depends on the choice of singular values; the empirical mode decomposition method has problems with modal aliasing and endpoint effects in signal denoising, which in turn affects the effective extraction of fault feature information; and the deep learning method has a strong dependence on training data and high computational complexity, which may lead to unsatisfactory denoising results.

[0004] In order to solve the problems of complex components of planetary gearbox vibration signals, strong noise interference, and poor adaptability of signal denoising methods, a highly reliable planetary gearbox signal denoising method is urgently needed to achieve effective noise removal and deep feature mining of planetary gearbox vibration signals.

[0005] The above information disclosed in the Background section is only for enhancing the understanding of the background of the present application, and therefore can include information that does not constitute prior art that is already known to those of ordinary skill in the art. SUMMARY

[0006] The present application provides a high-reliability planetary gearbox signal denoising method, system, medium and equipment, which solves the problems of complex planetary gearbox vibration signal components, strong noise interference, poor signal denoising method adaptability, etc., and realizes effective noise removal and feature deep mining of planetary gearbox vibration signals.

[0007] A high-reliability planetary gearbox signal denoising method comprises:

[0008] The first step is data acquisition and preprocessing, wherein the original vibration signals of the nuclear power pump planetary gearbox in different health states are acquired, the signals are de-trended, the data is fitted into a straight line, the linear trend represented by the straight line is subtracted from the data, and the mean value preprocessing is performed to eliminate the long-term trend and constant component in the signal, making the signal more stable.

[0009] The second step is to decompose the original vibration signal, wherein the empirical mode decomposition algorithm EMD, the ensemble empirical mode decomposition algorithm EEMD and the variational mode decomposition algorithm VMD are used to decompose the original vibration signal to obtain a plurality of different modal components.

[0010] The third step is to compare and optimize the decomposition algorithm, wherein the fast Fourier transform is performed on the IMF components obtained by decomposition, the frequency domain characteristics of the IMF components obtained after decomposition are analyzed and evaluated, the decomposition effects of the empirical mode decomposition algorithm EMD, the ensemble empirical mode decomposition algorithm EEMD and the variational mode decomposition algorithm VMD are compared, and the algorithm optimization is performed. The decomposition results of the empirical mode decomposition algorithm EMD and the ensemble empirical mode decomposition algorithm EEMD both have modal aliasing phenomenon, the signals between different IMF components have overlap, the signal decomposition effect is poorer than that of the variational mode decomposition algorithm VMD, and the variational mode decomposition algorithm VMD has better effect in the planetary gearbox signal decomposition.

[0011] The fourth step is algorithm parameter iterative optimization, which uses sparrow search algorithm SSA for iterative optimization to determine the hyperparameter value in the variational mode decomposition algorithm VMD, so as to realize the best signal decomposition effect.

[0012] The fifth step is vibration signal decomposition and reconstruction, which uses the algorithm optimized by iteration to decompose the gearbox vibration signal, removes the noise components in the modal components and reconstructs the signal, and realizes the denoising of the gearbox vibration signal.

[0013] The high-reliability planetary gearbox signal denoising method, in the first step, collects original vibration signals of different health states under the same working condition through high-reliability planetary gearbox health parts and fault parts, 1000 experimental samples are collected for each health state, the sample length is 32768, and the health states include health, sun gear spalling, sun gear pitting, planetary gear spalling and planetary gear pitting.

[0014] The high-reliability planetary gearbox signal denoising method, in the second step, the empirical mode decomposition algorithm EMD decomposes the original vibration signal into a superposition of a finite number of modal components, each of which represents the vibration characteristic form of the original vibration signal at a certain time scale, and the modal components satisfy the following conditions: ① In the entire time series, the number of zero-crossing points and the number of maximum points must be the same or differ by at most one; ② At any time, the average value of the lower envelope line formed by the local minimum points and the upper envelope line formed by the local maximum points is zero, that is, the local average value is zero, and the steps of the empirical mode decomposition algorithm EMD for decomposing the gear vibration signal are as follows:

[0015] Step 1, search for local extreme points of the original vibration signal s(t), determine all local extreme points of the original vibration signal, in order to obtain the upper envelope line u(t) and the lower envelope line l(t) of the original vibration signal, use cubic spline interpolation method to connect all maximum points and minimum points respectively, ensure that the modal components obtained by decomposition satisfy that in the entire time series, the number of zero-crossing points and the number of maximum points must be the same or differ by at most one;

[0016] Step 2, take the average value of the upper and lower envelope lines:

[0017] m(t)=(u(t)+l(t)) / 2

[0018] Where m(t) is the average value of the upper and lower envelope lines, t is the time,

[0019] And define the difference between the original vibration signal s(t) and the average value m(t) as:

[0020] h(t)=s(t)-m(t)

[0021] Where h(t) is the difference between the original vibration signal and the average value,

[0022] Step 3, judge whether h(t) is a modal component: if h(t) satisfies conditions ① and ②, h(t) is recorded as the first modal component IMF1; otherwise, go to step 4;

[0023] Step 4, taking h(t) as the original vibration signal to be decomposed, and repeating steps 1 to 3, in each decomposition process, it is necessary to judge whether the modal components obtained meet the conditions ① and ② of IMF, if not, it is necessary to loop k times, until a series of modal components IMF2, IMF3, …, IMF n and the residual signal c(t) meet the conditions ① and ②,

[0024] Step 5, adding IMF1, IMF2, …, IMF n and the last residual term c(t) to restore the original vibration signal s(t),

[0025]

[0026] wherein is the sum of the modal components.

[0027] In the high-reliability planetary gearbox signal denoising method, in the second step, the step of decomposing the original vibration signal by the ensemble empirical mode decomposition algorithm EEMD is as follows:

[0028] Step 2-1, setting the ensemble average number M according to the standard deviation of the original vibration signal, adding white noise n i (t) to the original vibration signal s(t) to obtain a new signal S i (t):

[0029] S i (t) = s(t) + a*n i (t),

[0030] wherein S i (t) is the new signal after adding white noise, a is the number of white noise signals, t is time,

[0031] Step 2-2, decomposing the new signal with added white noise by the empirical mode decomposition algorithm EMD to obtain n IMF components and a residual component R(t):

[0032]

[0033] wherein is the sum of the modal components of a single group, R i (t) is a single group of residual signals,

[0034] Step 2-3, repeating steps 2-1 to 2-2 for M times, wherein the amplitude of the noise signal added each time is different, and M groups of modal components IMFs

[0035] Step 2-4, averaging each component of all groups of IMFs and the residual component:

[0036]

[0037]

[0038] where is the average of all group modal components, M is the total number of groups, is the sum of all group modal components, is the average of all group residuals, is the sum of all group residuals,

[0039] Step 2-5, add the two averages to get the final modal component IMF decomposition result:

[0040]

[0041] where s(t) is the original vibration signal, is the average of all group modal components, is the average of all group residuals.

[0042] In the high-reliability planetary gearbox signal denoising method, in the second step, the VMD algorithm is used to decompose the original vibration signal, and the steps are as follows:

[0043] (1) The original vibration signal s(t) is decomposed into K modal functions u k (t), so that the sum of the estimated bandwidth of each modal function u k (t) is minimized, and the corresponding constraint variational model is expressed as:

[0044]

[0045] where {u k}={u1,...,u k} is the K modal functions obtained by decomposition; {w k}={w1,...,w k} is the center frequency of each modal; is the time derivative, i.e. the differential of time; δ(t) is the unit impulse function; is the analytic signal under Hilbert transform; is a complex modulation factor, which modulates the frequency spectrum of each modal function to the corresponding base frequency band.

[0046] In order to convert the constrained variational problem into an unconstrained variational problem, a quadratic penalty term factor α and a Lagrange multiplier λ(t) are introduced, and the expression is:

[0047]

[0048] where a is a penalty factor, l is a Lagrange multiplier, {u k}={u1,...,u k} are K modal functions obtained by decomposition.

[0049] The alternating direction multiplier method is used to solve the variational problem, and is updated until the k modal components are obtained, and the update of l n+1 The process is as follows:

[0050]

[0051] where l n+1 (w) is the new round of Lagrange multiplier frequency domain value, t is the step size of the Lagrange operator term, is the current modal reconstruction error.

[0052] Set the decision accuracy e>0, when the following conditions are met, stop iteration:

[0053]

[0054] where is the frequency domain representation of the kth modal component in the n+1th iteration, is the frequency domain representation of the kth modal component in the n-1th iteration, is the change ratio of the kth modal component relative to the last iteration, e is the set decision accuracy threshold, and the typical value is 10 -6 or 10 -7 .

[0055] After iteration, the signal is transformed, and the decomposed modal components are finally output.

[0056] In the fourth step of the high-reliability planetary gearbox signal denoising method, the variational modal decomposition algorithm VMD is optimized by comparing and analyzing the frequency domain characteristics of the modal components, and the superparameters in the variational modal decomposition algorithm VMD are iteratively optimized by using the sparrow search algorithm, the superparameters including the penalty factor a and the decomposition layer number K, and the steps of iteratively optimizing the superparameters in the variational modal decomposition algorithm VMD by using the sparrow search algorithm are as follows:

[0057] Initialize the candidate solution set and search algebra parameters;

[0058] According to each solution in the candidate solution set, the value of the objective function is calculated, and position updating and speed updating are performed according to the sparrow behavior law to obtain a new candidate solution set;

[0059] The new candidate solution set is screened and sorted, and according to the search algebra parameters, it is determined whether to end the optimization process.

[0060] The sparrow search algorithm is used for iterative optimization of hyperparameters in a variational mode decomposition algorithm VMD,

[0061] (1) Signal acquisition: acquire the original vibration signal of the gear box to be processed, and determine the modal number K and the penalty factor a of the variational mode decomposition algorithm VMD that need to be optimized. Since there is no scientific conclusion for the values of the two hyperparameters, an optimization algorithm is used to iteratively optimize them.

[0062] (2) Parameter initialization: set the initial population size, maximum iteration number and search parameter range of the sparrow search algorithm, and generate an initial population containing the modal number K and the penalty factor a by using a random initialization method.

[0063] (3) Fitness evaluation: for each parameter set in the population, use the variational mode decomposition algorithm VMD to decompose the original signal and obtain the corresponding modal component set; calculate the envelope entropy of each modal component, and take the sum of all modal envelope entropies as the fitness function value of the current parameter combination; evaluate the decomposition performance of the current parameter set with the goal of minimizing the total envelope entropy; when the total envelope entropy is minimized, the decomposition performance is optimal.

[0064] (4) Parameter updating: according to the search strategy of the sparrow search algorithm SSA, update the position of individuals in the population to obtain a new parameter combination population.

[0065] (5) Optimal parameter selection: repeat steps 3 to 4 until the maximum iteration number is reached or the fitness converges, stop the optimization process, and select the parameter set K, a corresponding to the current fitness as the optimal solution.

[0066] A system for implementing the method comprises:

[0067] A collection module is used to acquire original vibration signals of nuclear power pump planetary gearboxes in different health states, and to perform detrending and mean removal preprocessing on the signals;

[0068] A signal decomposition unit is used to perform modal decomposition on the original vibration signals by using empirical mode decomposition algorithm EMD, ensemble empirical mode decomposition algorithm EEMD and variational mode decomposition algorithm VMD, respectively, to obtain a plurality of different modal components;

[0069] A comparison unit is used to perform fast Fourier transform on the obtained IMF components, analyze and evaluate the frequency domain characteristics of each IMF component, and compare the decomposition effects of the empirical mode decomposition algorithm EMD, the ensemble empirical mode decomposition algorithm EEMD and the variational mode decomposition algorithm VMD to optimize the algorithms;

[0070] An optimization unit is used for algorithm parameter iterative optimization, adopts sparrow search algorithm SSA for iterative optimization, and adaptively determines the super parameter value in the preferred algorithm;

[0071] A reconstruction unit is used for vibration signal decomposition reconstruction, adopts the algorithm iteratively optimized to perform modal decomposition on the gearbox vibration signal, removes the noise components in the modal components, and reconstructs the signal, thereby realizing gearbox vibration signal denoising.

[0072] A computer storage medium comprises computer instructions which, when executed on a computer, cause the computer to perform the method.

[0073] An electronic device comprises:

[0074] A memory, a processor and a computer program stored on the memory and executable on the processor, wherein,

[0075] The processor implements the method when executing the program.

[0076] Compared with the prior art, the present application has the following advantages: the present application optimizes the decomposition effect of EMD, EEMD and VMD three decomposition algorithms, and optimizes the variational mode decomposition algorithm VMD as the most suitable decomposition algorithm for the planetary gearbox vibration signal, so that the effective features and non-stationary characteristics of the original signal can be better preserved, and the modal aliasing problem of other decomposition algorithms can be avoided. The super parameters of the variational mode decomposition algorithm VMD are iteratively optimized by the SSA algorithm, so that the variational mode decomposition algorithm VMD has adaptability when decomposing signals, so as to achieve the best decomposition effect, so that the noise components are easily removed, and the denoising precision of the gearbox vibration signal is improved. BRIEF DESCRIPTION OF DRAWINGS

[0077] Various other advantages and benefits of the present application will become apparent to those of ordinary skill in the art, reading the following detailed description of the preferred embodiment. The drawings accompanying the specification are simply illustrative of the preferred embodiment and are presented for the purpose of simplification and are not deemed to limit the present application. Obviously, the drawings described below are only some embodiments of the present application, and other drawings can be obtained from these drawings without creative labor for those of ordinary skill in the art. Moreover, the same reference numerals are used to represent the same components throughout the drawings.

[0078] In the drawings:

[0079] Figure 1 The modal component time domain graph obtained by the EMD algorithm of the high-reliability planetary gearbox signal denoising method in the present application;

[0080] Figure 2The modal component frequency domain local amplification diagram obtained by the EMD algorithm of the high-reliability planetary gearbox signal denoising method in the application, Figure 2 The middle (a) is an IMF1-2 frequency domain local amplification diagram, Figure 2 The middle (b) is an IMF3-4 frequency domain local amplification diagram, and the characteristic frequency of the gearbox is clearly visible, which indicates that the signal decomposition is effective, but the modal aliasing phenomenon occurs;

[0081] Figure 3 The modal component frequency domain local amplification diagram obtained by the EEMD algorithm of the high-reliability planetary gearbox signal denoising method in the application, Figure 3 The middle (a) is an IMF2-3 frequency domain local amplification diagram, Figure 3 The middle (b) is an IMF7 frequency domain local amplification diagram, and the meshing frequency and rotational frequency of the gearbox are clearly visible, which indicates that the signal decomposition is effective, but the modal aliasing phenomenon occurs;

[0082] Figure 4 The modal component frequency domain local amplification diagram obtained by the VMD algorithm of the high-reliability planetary gearbox signal denoising method in the application, Figure 4 The middle (a) is an IMF1 frequency domain local amplification diagram, and the meshing frequency of the gearbox is clearly visible, Figure 4 The middle (b) is an IMF1-3 frequency domain local amplification diagram, and there is no modal aliasing phenomenon, but the noise component is large, and the decomposition effect is poor;

[0083] Figure 5 The SSA algorithm iteration optimization function diagram of the high-reliability planetary gearbox signal denoising method in the application;

[0084] Figure 6 The modal component frequency domain local amplification diagram of the high-reliability planetary gearbox signal denoising method in the application before and after VMD algorithm optimization, Figure 6 The middle (a) is an IMF1-3 frequency domain amplification diagram before VMD optimization, Figure 6 The middle (b) is an IMF1-3 frequency domain amplification diagram after VMD optimization, and it can be seen that the decomposition result of the VMD algorithm after optimization is better, and the noise component is greatly reduced;

[0085] Figure 7 The flowchart of the high-reliability planetary gearbox signal denoising method in the application.

[0086] The application will be further explained in combination with the drawings and embodiments. DETAILED DESCRIPTION

[0087] Specific embodiments of the present application will be described herein below with reference to drawings. While specific embodiments of the application are shown in the drawings, it should be understood that the application can be implemented in various forms and should not be limited to the embodiments set forth in the description below. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the application to those skilled in the art.

[0088] It should be noted that certain terms are used throughout the present specification and claims which have particular meanings as set forth below. Those of ordinary skill in the art will understand that the terms can mean different and / or same thing to different people. Headings of sections provided in the specification and the claims should not be understood as limiting the scope of the claims. Absent a specific definition, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein was chosen to best explain the principles of the application, the practical applicability of which will be understood as set forth in the detailed description included herein below.

[0089] In order to make the present application more clearly understood, the following will further explain and describe specific embodiments of the present application with reference to the accompanying drawings, and each drawing does not constitute a limitation on the embodiments of the present application.

[0090] As shown in Figures 1 to 7 The high-reliability planetary gearbox signal denoising method includes the following steps:

[0091] The first step is data acquisition and preprocessing, wherein the original vibration signals of the nuclear power pump planetary gearbox in different health states are acquired, the signals are preprocessed by detrending (fitting the data into a straight line, and subtracting the linear trend represented by the straight line from the data) and mean removal (removing the average value of the waveform data), the long-term trend and constant component in the signal are eliminated, and the signal is more stable;

[0092] The second step is to decompose the original vibration signal, wherein the empirical mode decomposition algorithm EMD, the ensemble empirical mode decomposition algorithm EEMD and the variational mode decomposition algorithm VMD are used to respectively decompose the original vibration signal, and a plurality of different modal components are obtained;

[0093] In the third step, the decomposition algorithm comparison and optimization, the IMF components obtained by decomposition are subjected to fast Fourier transform, and the frequency domain characteristics of the IMF components obtained after decomposition are analyzed and evaluated. It can be seen that, considering that the effective frequency range of the actual gearbox signal is mainly the low frequency range below 3000 Hz, the IMF components in the high frequency part are not in the effective frequency range, and therefore only the IMF components in the low frequency range need to be analyzed. The characteristic frequencies of the gearbox can be seen in the low frequency IMF components obtained by the three algorithms, which indicates the effectiveness of signal decomposition. The decomposition results of the EMD algorithm and the EEMD algorithm both have modal aliasing phenomenon, and the signals between different IMF components overlap, and the signal decomposition effect is poorer than that of the VMD algorithm, and therefore the VMD algorithm is selected as the preferred algorithm for signal decomposition of the planetary gearbox.

[0094] In the fourth step, the algorithm parameter iterative optimization, the sparrow search algorithm SSA is used for iterative optimization to adaptively determine the super parameter value in the preferred algorithm.

[0095] In the fifth step, the vibration signal decomposition and reconstruction, the algorithm optimized by iteration is used for modal decomposition of the gearbox vibration signal, the noise components in the modal components are removed, and the signal is reconstructed to realize the denoising of the gearbox vibration signal.

[0096] In the preferred embodiment of the high-reliability planetary gearbox signal denoising method, in the first step, the original vibration signals in the healthy state of the high-reliability planetary gearbox under the same working condition are collected by collecting the healthy and fault parts of the high-reliability planetary gearbox, 1000 experimental samples are collected for each health state, the sample length is 32768, and the health states include health, sun gear spalling, sun gear pitting, planetary gear spalling and planetary gear pitting.

[0097] In the preferred embodiment of the high-reliability planetary gearbox signal denoising method, in the second step, the empirical mode decomposition algorithm EMD decomposes the original vibration signal into a stack of a finite number of modal components, each of which represents the vibration characteristic form of the original vibration signal at a certain time scale. The modal components satisfy the following conditions: ① In the entire time series, the number of zero-crossing points and the number of maximum points must be the same or differ by at most one; ② At any time, the average value of the lower envelope line formed by the local minimum points and the upper envelope line formed by the local maximum points is zero, i.e. the local average value is zero. The steps of the empirical mode decomposition algorithm EMD for decomposing the gear vibration signal are as follows:

[0098] Step 1: Search for local extreme points of the original vibration signal s(t) to determine all local extreme points of the original vibration signal. In order to obtain the upper envelope u(t) and lower envelope l(t) of the original vibration signal, use the cubic spline interpolation method to connect all the maximum points and minimum points respectively, ensuring that the modal components obtained by decomposition satisfy the requirement that the number of zero-crossing points and the number of maximum points in the entire time series must be the same or differ by at most one;

[0099] Step 2, take the average of the upper and lower envelopes:

[0100] m(t)=(u(t)+l(t)) / 2

[0101] Where m(t) is the average value of the upper and lower envelopes, t is time,

[0102] And define the difference between the original vibration signal s(t) and the average value m(t) as:

[0103] h(t)=s(t)-m(t)

[0104] Where h(t) is the difference between the original vibration signal and the average value,

[0105] Step 3: Determine whether h(t) is a modal component: If h(t) satisfies conditions ① and ②, then h(t) is recorded as the first modal component IMF1; otherwise, proceed to step 4.

[0106] Step 4: Take h(t) as the original vibration signal to be decomposed and repeat steps 1 to 3. In each decomposition process, it is necessary to determine whether the modal components obtained by decomposition meet the conditions ① and ② of IMF. If not, it is necessary to loop k times until a series of modal components IMF2, IMF3, ..., IMF are obtained. n And the residual signal c(t) satisfies conditions ① and ②,

[0107] Step 5: IMF1, IMF2, ..., IMF n Add it to the final residual term c(t) to restore the original vibration signal s(t),

[0108]

[0109] in is the sum of all modal components.

[0110] In a preferred embodiment of the high-reliability planetary gearbox signal denoising method, in the second step, the steps of decomposing the original vibration signal using the EEMD algorithm are as follows:

[0111] Step 2-1, set the number of set averages M according to the standard deviation of the original vibration signal, add white noise n to the original vibration signal s(t) to obtain a new signal S i (t) i (t):

[0112] S i (t)=s(t)+a*n i (t),

[0113] wherein S i (t) is the new signal after adding white noise, a is the number of white noise signals, t is time,

[0114] Step 2-2, the new signal added with white noise is decomposed by using empirical mode decomposition algorithm EMD to obtain n IMF components and residual component R(t):

[0115]

[0116] wherein is the sum of single group modal components, R i (t) is a single group residual signal,

[0117] Step 2-3, repeat steps 2-1 to 2-2 for M times, wherein the amplitude of the noise signal added each time is different, and M groups of modal components IMFs can be obtained,

[0118] Step 2-4, average all groups of IMFs and the residual component:

[0119]

[0120] wherein is the average of all groups of modal components, M is the total number of groups, is the sum of all groups of modal components, is the average of all groups of residuals, is the sum of all groups of residuals,

[0121] Step 2-5, add the two averages to obtain the final modal component IMF decomposition result:

[0122]

[0123] wherein s(t) is the original vibration signal, is the average of all groups of modal components, is the average of all groups of residuals.

[0124] In the preferred embodiment of the high-reliability planetary gear box signal denoising method, in the second step, the VMD algorithm is used to decompose the original vibration signal as follows:

[0125] (1) The original vibration signal s(t) is decomposed into K modal functions u k (t), so that the sum of the estimated bandwidths of each modal function u k (t) is minimized, and the corresponding constraint variational model is expressed as:

[0126]

[0127] where {u k} = {u1,...,u k} is the K modal functions obtained by decomposition; {w k} = {w1,...,w k} is the center frequency of each mode; is the time derivative, i.e., the differential with respect to time; δ(t) is the unit impulse function; is the analytic signal under Hilbert transform; is a complex modulation factor, which modulates the spectrum of each modal function to the corresponding base frequency band.

[0128] In order to convert the constraint variational problem into an unconstrained variational problem, a quadratic penalty term factor α and a Lagrange multiplier λ(t) are introduced, and the expression is:

[0129]

[0130] where α is the penalty factor, λ is the Lagrange multiplier, and {u k} = {u1,...,u k} is the K modal functions obtained by decomposition.

[0131] The alternating direction multiplier algorithm is used to solve the variational problem, and the update is performed until the k modal components are obtained. The update process of λ n+1 is as follows:

[0132]

[0133] where λ n+1 (w) is the frequency domain value of the new round of Lagrange multiplier, τ is the step size of the Lagrange operator term, is the current modal reconstruction error.

[0134] The determination accuracy ε>0 is set, and when the following conditions are met, the iteration is stopped:

[0135]

[0136] where is the frequency domain representation of the kth modal component in the n+1th iteration, is the frequency domain representation of the Kth modal component in the n-1th iteration, is the change ratio of the kth modal component relative to the last iteration, and ε is a set precision threshold, typically 10 -6 or 10 -7 .

[0137] After the iteration is stopped, the signal is transformed, and the decomposed modal components are finally output.

[0138] In the preferred embodiment of the high-reliability planetary gearbox signal denoising method, in the fourth step, by comparing and analyzing the frequency domain characteristics of the modal components, the variational mode decomposition algorithm VMD is preferably selected as the best in the decomposition of the original vibration signal. The sparrow search algorithm is used to iteratively optimize the hyperparameters in the variational mode decomposition algorithm VMD, including the penalty factor α and the number of decomposition layers K. The steps of iteratively optimizing the hyperparameters in the variational mode decomposition algorithm VMD by the sparrow search algorithm are as follows:

[0139] Initialize the candidate solution set and search algebraic parameters;

[0140] According to each solution in the candidate solution set, calculate the value of its objective function, and update the position and speed according to the sparrow behavior law to obtain a new candidate solution set;

[0141] The new candidate solution set is screened and sorted, and according to the search algebraic parameters, it is determined whether to end the optimization process.

[0142] In the preferred embodiment of the high-reliability planetary gearbox signal denoising method, in the step of iteratively optimizing the hyperparameters in the variational mode decomposition algorithm VMD by the sparrow search algorithm,

[0143] (1) Signal acquisition: acquire the original vibration signal of the gearbox to be processed, and determine the number of modal components K and the penalty factor α in the variational mode decomposition algorithm VMD that need to be optimized. Since there is no scientific conclusion for the values of these two hyperparameters, an optimization algorithm is needed to iteratively optimize them.

[0144] (2) Parameter initialization: set the initial population size, maximum number of iterations, and search parameter range of the sparrow search algorithm, and use random initialization to generate an initial population containing the number of modal components K and the penalty factor α.

[0145] (3) Fitness evaluation: for each set of parameters in the population, use the variational mode decomposition algorithm VMD to decompose the original signal and obtain the corresponding set of modal components; calculate the envelope entropy of each modal component and take the sum of all modal envelope entropies as the fitness function value of the current parameter combination; evaluate the decomposition performance of the current parameter combination with the goal of minimizing the total envelope entropy value; when the total envelope entropy value is minimized, the decomposition performance is optimal.

[0146] (4) Parameter update: according to the search strategy of the sparrow search algorithm SSA, the position of the individual in the population is updated to obtain a new parameter combination population.

[0147] (5) Optimal parameter selection: repeat steps 3 to 4 until the maximum iteration number or fitness convergence is reached, stop the optimization process, and select the parameter group K, a corresponding to the current fitness as the optimal solution. A system for implementing the method comprises:

[0148] The acquisition module is used to obtain original vibration signals of the nuclear power pump planetary gear box in different health states, and to perform detrending (fitting data into a straight line, and subtracting the linear trend represented by the straight line from the data) and mean removal (removing the average value of waveform data) preprocessing on the signals, so as to eliminate long-term trends and constant components in the signals and make the signals more stable.

[0149] The signal decomposition unit is used to perform modal decomposition on the original vibration signals by using empirical mode decomposition algorithm EMD, ensemble empirical mode decomposition algorithm EEMD and variational mode decomposition algorithm VMD respectively, to obtain a plurality of different modal components.

[0150] The comparison unit is used to perform fast Fourier transform on the decomposed IMF components, analyze and evaluate the frequency domain characteristics of each IMF component, and compare the decomposition effects of the empirical mode decomposition algorithm EMD, the ensemble empirical mode decomposition algorithm EEMD and the variational mode decomposition algorithm VMD to perform algorithm optimization.

[0151] The optimization unit is used for iterative optimization of algorithm parameters, and the sparrow search algorithm SSA is used for iterative optimization to adaptively determine the hyperparameter value in the optimized algorithm.

[0152] The reconstruction unit is used for vibration signal decomposition and reconstruction, and the algorithm optimized by iteration is used for modal decomposition of the gear box vibration signal, to remove noise components in the modal components and reconstruct the signal, so as to realize gear box vibration signal denoising.

[0153] In one embodiment, the high-reliability planetary gear box signal denoising method comprises:

[0154] The first step is data acquisition and preprocessing. The original vibration signals of the nuclear power pump planetary gear box in different health states (healthy, sun gear spalling, sun gear pitting, planetary gear spalling, and planetary gear pitting) are obtained, and the signals are subjected to detrending (fitting data into a straight line, and subtracting the linear trend represented by the straight line from the data) and mean removal (removing the average value of waveform data) preprocessing, to eliminate long-term trends and constant components in the signals and make the signals more stable.

[0155] Second step, original vibration signal decomposition. Three algorithms of EMD, EEMD and VMD are used to decompose the original vibration signal of the gearbox to obtain a plurality of different intrinsic mode components (IMF).

[0156] Third step, decomposition algorithm comparison and optimization. Fast Fourier transform is performed on the decomposed IMF components to analyze and evaluate the frequency domain characteristics of each IMF component. The decomposition effects of EMD, EEMD and VMD algorithms are compared to optimize the algorithm.

[0157] Fourth step, iterative optimization of algorithm parameters. The SSA algorithm is used for iterative optimization to adaptively determine the hyperparameter value in the VMD algorithm, so as to achieve the best decomposition effect of the signal.

[0158] Fifth step, vibration signal decomposition and reconstruction. The SSA-VMD algorithm is used to decompose the vibration signal of the gearbox, remove the noise components in the modal components and reconstruct the signal, and realize the denoising of the vibration signal of the gearbox.

[0159] Further, the vibration signal data of five health states (healthy, sun gear peeling, sun gear pitting, planetary gear peeling, and planetary gear pitting) of the gear under the same working condition are collected through the gearbox health and prefabricated fault experiments. Each health state collects 1000 experimental samples, and the sample length is 32768. In the second step, EMD algorithm is an adaptive signal decomposition method suitable for analysis of non-linear and non-stationary signals, which can decompose any complex signal into a set of finite IMFs, each of which represents the vibration characteristic form of the signal at a certain time scale. The modal function IMF has the advantages of self-similarity, self-adaptation and local characteristic analysis, and must satisfy two conditions: ① In the entire time series, the number of zero-crossing points and the number of maximum points must be the same or differ by at most one; ② At any time, the average value of the lower envelope line formed by the local minimum points and the upper envelope line formed by the local maximum points is zero, i.e. the local average is zero. The specific steps of using EMD algorithm to decompose the gear vibration signal are as follows:

[0160] (1) Search for local extreme points of the original signal s(t). First, determine all the local extreme points of the original signal. In order to obtain the upper envelope line u(t) and the lower envelope line l(t), use cubic spline interpolation method to connect all the maximum points and minimum points respectively. This step can ensure that the modal components obtained by decomposition satisfy condition ①.

[0161] (2) Take the average of the upper and lower envelope lines:

[0162] m(t) = (u(t) + l(t)) / 2

[0163] where m(t) is the mean value of the upper and lower envelopes, and t is time.

[0164] and define the difference between the original signal s(t) and the mean value m(t) as:

[0165] h(t) = s(t) - m(t)

[0166] where h(t) is the difference between the original signal and the mean value.

[0167] (3) Determine whether h(t) is an IMF: if h(t) satisfies the above conditions ① and ②, then h(t) is recorded as the first modal component IMF1; otherwise, proceed to step 4.

[0168] (4) Take h(t) as the original signal to be decomposed, and repeat steps 1 to 3. In each decomposition process, it is necessary to determine whether the modal components obtained satisfy the two conditions of IMF. If not, it needs to be looped k times until a series of modal components IMF2, IMF3, …, IMF n and a residual signal c(t) meet the requirements.

[0169] (5) Add IMF1, IMF2, …, IMF n and the last residual term c(t) to recover the original signal s(t).

[0170]

[0171] where is the sum of the modal components, and c(t) is the residual term.

[0172] In the second step, the EEMD algorithm is an improved method of the EMD algorithm. By introducing a white noise sequence into the original time domain signal and reconstructing it multiple times, the EEMD algorithm improves the modal aliasing phenomenon that may occur in the decomposition process of the EMD algorithm, while retaining the advantages of adaptability and completeness, and can more stably obtain multiple modal components. The specific steps of using the EEMD algorithm to decompose the gear vibration signal are as follows:

[0173] (1) According to the standard deviation of the original signal, set the ensemble average number M, add a white noise n i (t) to the original signal s(t) to obtain a new signal S i (t):

[0174] S i (t) = s(t) + a*n i (t)

[0175] where S i (t) is the new signal after adding white noise, a is the number of white noise signals, and t is time.

[0176] (2) The EMD algorithm is used to decompose the new signal with white noise to obtain a series of IMF components and residual component R(t):

[0177]

[0178] wherein is the sum of the single group modal components, R i (t) is the single group residual signal.

[0179] (3) The above two steps are repeated for M times, wherein the added noise signal amplitude is different each time, and M groups of modal components IMFs can be obtained.

[0180] (4) The average values of all groups of IMFs and the residual components are calculated:

[0181]

[0182] wherein is the average value of all groups of modal components, M is the number of groups, is the sum of all groups of modal components, is the average value of all groups of residuals, is the sum of all groups of residuals.

[0183] (5) Adding the two average values can obtain the final modal component IMF decomposition result:

[0184]

[0185] wherein s(t) is the original signal; is the average value of all groups of modal components; is the average value of all groups of residuals.

[0186] In the second step, the variational mode decomposition algorithm VMD is an adaptive signal decomposition method based on classical Wiener filtering and Hilbert transform. The basic idea is to solve the optimal solution of the variational problem through iteration, and decompose the original data into multiple IMF components, and the optimal center frequency and bandwidth of each IMF component can be adaptively updated. A penalty factor a and a decomposition layer number K need to be preset artificially. This algorithm can decompose a complex signal into vibration modes of different frequencies, and each vibration mode has a corresponding modulation function to describe the amplitude and frequency changes over time. On the basis of the EMD algorithm, a condition constraint is further added: to minimize the sum of the bandwidths of the center frequencies of each modal component. The basic idea of the variational mode decomposition algorithm VMD is to decompose the signal into a group of sub-functions, and then obtain the vibration frequency and amplitude modulation function of each sub-function through a variational optimization process. The specific steps of using the variational mode decomposition algorithm VMD to decompose the gear vibration signal are as follows:

[0187] (1) The original signal s(t) is decomposed into K modal functions u k (t) so that the sum of the estimated bandwidths of each modal function is minimized, and the corresponding constrained variational model can be expressed as:

[0188]

[0189] where {u k} = {u1,...,u k} are the K modal functions obtained by decomposition; {w k} = {w1,...,w k} are the center frequencies of the respective modes; is the time derivative, i.e., the differential with respect to time; δ(t) is the unit impulse function; is the analytic signal under the Hilbert transform; is the complex modulation factor, which modulates the spectrum of each modal function to the corresponding base frequency band.

[0190] (2) In order to convert the above constrained variational problem into an unconstrained variational problem, a quadratic penalty term factor α and a Lagrange multiplier λ(t) are introduced, and the expression is:

[0191]

[0192] where α is the penalty factor, λ is the Lagrange multiplier, and {u k} = {u1,...,u k} are the K modal functions obtained by decomposition.

[0193] (3) The alternating direction multiplier algorithm is used to solve the variational problem, and the process is updated until the K IMFs are obtained, and the update process of λ n+1 is as follows:

[0194]

[0195] where λ n+1 (w) is the frequency domain value of the new round of Lagrange multipliers, τ is the step size of the Lagrange operator term, is the current modal reconstruction error.

[0196] (4) Set the decision accuracy ε > 0, and stop the iteration when the following conditions are met:

[0197]

[0198] where is the frequency domain representation of the kth modal component in the n+1th iteration, is the frequency domain representation of the Kth modal component in the n-1th iteration, is the change ratio of the kth modal component relative to the last iteration, and ε is a set decision accuracy threshold, typically 10 -6 or 10 -7 .

[0199] (5) After the iteration is stopped, the signal is transformed, and the decomposed IMF components are finally output.

[0200] In the third step, the IMF components obtained by decomposing the gear vibration signal using the EMD, EEMD, and VMD algorithms are subjected to fast Fourier transform. By comparing and analyzing the frequency domain characteristics of the IMF components, the variational mode decomposition algorithm VMD is selected as the best algorithm for decomposing the pump gear vibration signal. In the fourth step, to achieve the best decomposition effect of the variational mode decomposition algorithm VMD on the signal, the SSA algorithm is used to iteratively optimize the hyperparameters, i.e., the penalty factor α and the decomposition layer number K, in the variational mode decomposition algorithm VMD. The SSA algorithm is a heuristic algorithm based on the foraging and anti-predation behavior of sparrows in nature, which is used to solve optimization problems. The algorithm was initially proposed based on the cooperation, adaptation, and learning ability of sparrows during flight and foraging, which is analogous to solving optimization problems. The specific steps of the SSA algorithm are as follows:

[0201] (1) Initialize the candidate solution set and search iteration parameter.

[0202] (2) According to each solution in the candidate solution set, calculate the value of the objective function, and update the position and speed according to the sparrow behavior rules to obtain a new candidate solution set.

[0203] (3) Screen and sort the new candidate solution set, and decide whether to end the optimization process according to the search iteration parameter.

[0204] In one embodiment, in the fourth step, the specific steps of optimizing the VMD hyperparameters using the SSA algorithm are as follows:

[0205] (1) Signal acquisition: acquire the original vibration signal of the gear box to be processed, and determine the hyperparameters of the variational mode decomposition algorithm VMD that need to be optimized, i.e., the modal number K and the penalty factor α. Since there is no scientific conclusion for the values of these two hyperparameters, an optimization algorithm is needed to iteratively optimize them.

[0206] (2) Parameter initialization: set the initial population size, maximum iteration number, and search parameter range of the sparrow search algorithm, and generate an initial population containing the modal number K and the penalty factor α using random initialization.

[0207] (3) Fitness evaluation: for each parameter set in the population, use the variational mode decomposition algorithm VMD to decompose the original signal, obtain the corresponding modal component set; calculate the envelope entropy of each modal component, and take the sum of all modal envelope entropies as the fitness function value of the current parameter combination; take the minimum envelope entropy total value as the target, evaluate the decomposition performance of the current parameter combination, and when the envelope entropy total value is the minimum, the decomposition performance is the best.

[0208] (4) Parameter update: according to the search strategy of sparrow search algorithm SSA, the position of the individual in the population is updated to obtain a new parameter combination population.

[0209] (5) Optimal parameter selection: repeat steps 3 to 4 until the maximum iteration number or fitness convergence is reached, stop the optimization process, and select the parameter set K, α corresponding to the current fitness as the optimal solution.

[0210] In the fifth step, the SSA-variational mode decomposition algorithm VMD is used to decompose the gear box vibration signal, remove the noise component in the modal component and reconstruct the signal, and realize the denoising of the gear box vibration signal.

[0211] In the present application, the data acquisition and preprocessing are carried out to obtain the original vibration signal (sample number 1000, length 32768) under different health states; the signal is detrended (linear trend is fitted and subtracted) and de-meaned (DC offset is removed). The signal stationarity is improved, the long-term trend component is removed, and the subsequent modal decomposition is avoided. The fixed offset component is eliminated, the signal center is symmetric to zero, and the spectral analysis accuracy is improved. The stability of the subsequent algorithm is enhanced, the EMD / VMD type algorithm is sensitive to the initial value of the signal, and the convergence problem caused by abnormal initial conditions can be avoided after de-meaning; the influence of low-frequency drift on modal aliasing is reduced, and the decomposition quality is improved.

[0212] Signal modal decomposition and algorithm comparison and optimization, EMD, EEMD, VMD three modal decomposition methods are used to process the signal; The IMFs are obtained by decomposition, and FFT frequency domain analysis is performed on them; By comparing the performance of each algorithm in the high / low frequency band, the modal aliasing phenomenon is identified, and in the effective frequency range below 3000 Hz, VMD shows the best frequency domain resolution and modal separation ability, VMD can more accurately locate the fault related frequency components of the gearbox (such as meshing frequency, sideband, modulation frequency, etc.), providing high-quality modal components for subsequent denoising; The modal aliasing missing problem is significantly better than EMD / EEMD, which shows that it is more suitable for precise signal processing of planetary gearboxes. The sparrow search algorithm SSA optimizes the VMD super parameters, taking the two key parameters of VMD, the number of modes K and the penalty factor a, as the optimization target; The sparrow search algorithm SSA is used to iteratively search for the optimal parameter combination; The sum of the envelope entropy of different modal components is used as the fitness function, and the minimum value is taken as the goal. Automatically determine the optimal parameter combination: avoid subjective errors caused by artificial experience setting; Improve the adaptability of VMD under different working conditions. Improve the purity and physical interpretability of the mode, the envelope entropy reflects the regularity and complexity of the signal, minimizing it means extracting more effective modal components, which helps to distinguish the real fault features from the noise components. Accelerate convergence and global optimization: SSA algorithm has good global search ability and fast convergence characteristics; Compared with traditional grid search or genetic algorithm, SSA is more efficient and stable in parameter space.

[0213] Signal modal decomposition and reconstruction denoising based on VMD, the VMD parameters optimized by SSA are used for modal decomposition of the signal; Determine whether each IMF component belongs to noise dominant component (can be judged by center frequency, energy, entropy, kurtosis, spectral distribution, etc.); After removing the IMF components containing noise, the remaining IMF components are reconstructed into a denoised signal. Effectively suppress noise interference: retain useful frequency components, remove random high-frequency noise or pseudo frequency; Significantly improve the signal-to-noise ratio (SNR), making the fault features more prominent. Unlike traditional filtering methods (such as wavelet threshold, Butterworth), VMD is self-adaptive decomposition, which can better match the nonlinear characteristics of the signal itself, ensure the integrity of the signal time-frequency structure, and facilitate subsequent deep learning model recognition. Improve the accuracy of subsequent fault identification: clear signal features help to extract more accurate fault features (such as envelope spectrum, resonance demodulation, time-frequency diagram, etc.); Support the construction of a more accurate intelligent diagnosis system.

[0214] Although the embodiments of the present application have been described above with reference to the accompanying drawings, the present application is not limited to the above-described specific embodiments and areas of application, and the above-described specific embodiments are merely illustrative and instructive, but are not restrictive. Many modifications can be made by those skilled in the art under the teachings of the present specification and without departing from the scope of the present application as defined by the claims.

Claims

1. A high-reliability planetary gearbox signal denoising method, characterized in that, The steps include: The first step is data acquisition and preprocessing. The raw vibration signals of the planetary gearbox of the nuclear power circulating pump in different health states are acquired, the signals are detrended, and the data are fitted into a straight line. The linear trend represented by the straight line is subtracted from the data, and the mean is removed to eliminate the long-term trend and constant component in the signal, making the signal more stable. The second step is to decompose the original vibration signal, wherein the empirical mode decomposition algorithm EMD, the ensemble empirical mode decomposition algorithm EEMD and the variational mode decomposition algorithm VMD are used to perform modal decomposition on the original vibration signal to obtain multiple different modal components; The third step is to compare and optimize the decomposition algorithms. In this step, the IMF components obtained by decomposition are subjected to fast Fourier transform, and the frequency domain characteristics of the IMF components obtained after decomposition are analyzed and evaluated. The decomposition effects of the empirical mode decomposition algorithm EMD, the ensemble empirical mode decomposition algorithm EEMD, and the variational mode decomposition algorithm VMD are compared to optimize the algorithm. The decomposition results of the empirical mode decomposition algorithm EMD and the ensemble empirical mode decomposition algorithm EEMD both have modal aliasing, and there is signal overlap between different IMF components. The signal decomposition effect is poorer than that of the variational mode decomposition algorithm VMD. The variational mode decomposition algorithm VMD is more effective in the decomposition of planetary gearbox signals. The fourth step is iterative optimization of algorithm parameters. The sparrow search algorithm (SSA) is used for iterative optimization to determine the hyperparameter values ​​in the variational mode decomposition algorithm (VMD) to achieve the best signal decomposition effect. The fifth step is to decompose and reconstruct the vibration signal. The gearbox vibration signal is modally decomposed using an iterative optimization algorithm to remove the noise components in the modal components and reconstruct the signal to achieve gearbox vibration signal denoising.

2. The high-reliability planetary gearbox signal denoising method according to claim 1, characterized in that, Preferably, in the first step, the original vibration signals of the healthy state under the same working conditions are collected through the healthy parts and faulty parts of the high-reliability planetary gearbox, 1000 experimental samples are collected for each healthy state, and the sample length is 32768. The healthy states include healthy, sun gear spalling, sun gear pitting, planet gear spalling and planet gear pitting.

3. The high reliable planetary gearbox signal denoising method of claim 1, wherein, In the second step, the empirical mode decomposition algorithm (EMD) decomposes the original vibration signal into a superposition of a finite number of modal components, where each modal component represents the vibration characteristic form of the original vibration signal at a certain time scale. The modal components meet the following conditions: ① In the entire time series, the number of zero-crossing points and the number of maximum points must be the same or differ by at most one; ② At any time, the average value of the lower envelope formed by the local minimum points and the upper envelope formed by the local maximum points is zero, that is, the local average value is zero. The steps of the empirical mode decomposition algorithm (EMD) to decompose the gear vibration signal are as follows: Step 1, search for local extreme points of the original vibration signal s(t), determine all local extreme points of the original vibration signal, in order to obtain the upper envelope u(t) and the lower envelope l(t) of the original vibration signal, use cubic spline interpolation method to connect all maximum points and minimum points respectively, ensure that the modal components obtained by decomposition meet the condition that the number of zero-crossing points and the number of maximum points must be the same or at most differ by one in the entire time sequence; Step 2, take the average value of the upper and lower envelope lines: m(t) = (u(t) + l(t)) / 2 Where m(t) is the average value of the upper and lower envelope lines, t is the time, And define the difference between the original vibration signal s(t) and the average value m(t) as: h(t) = s(t) - m(t) Where h(t) is the difference between the original vibration signal and the average value, Step 3, judge whether h(t) is a modal component: if h(t) satisfies condition ① and condition ②, then h(t) is recorded as the first modal component IMF1; otherwise, go to step 4; Step 4, taking h(t) as the original vibration signal to be decomposed, and repeating the execution of steps 1 to 3, in each decomposition process, it is necessary to judge whether the modal component obtained satisfies the condition 1 and the condition 2 of IMF, if not, it is necessary to loop k times until a series of modal components IMF2, IMF3, …, IMF n and the residual signal c(t) satisfy the condition 1 and the condition 2, Step 5: IMF1, IMF2, ..., IMF n Add it to the final residual term c(t) to restore the original vibration signal s(t), wherein is the sum of the components for each modality.

4. The high-reliability planetary gearbox signal denoising method of claim 1, wherein, In the second step, the steps of the ensemble empirical mode decomposition algorithm EEMD for decomposing the original vibration signal are as follows: Step 2-1, set the number of set averages M according to the standard deviation of the original vibration signal, add white noise n to the original vibration signal s(t) i (t), to obtain a new signal S i (t): S i (t) = s(t) + a * n i (t), where S i (t) is the new signal after adding white noise, a is the number of white noise signals, and t is time. Step 2-2, use the empirical mode decomposition algorithm EMD to decompose the new signal with added white noise to obtain n IMF components and a residual component R(t): wherein is the sum of the single group modal components, R i (t) is the single group residual signal, Step 2-3, repeat steps 2-1-2-2 for M times, where the amplitude of the noise signal added each time is different, and M groups of modal components IMFs can be obtained, Step 2-4, average all groups of IMFs and the residual component: wherein is the average of all group modal components, M is the total number of groups, is the sum of all group modal components, is the average of all group residuals, is the sum of all group residuals, Step 2-5, add the two average values to obtain the final modal component IMF decomposition result: where s(t) is the original vibration signal, is the average of all group modal components, is the average of all group residuals.

5. The high reliable planetary gearbox signal denoising method of claim 1, wherein, In the second step, the steps of the variational mode decomposition algorithm VMD for decomposing the original vibration signal are as follows: (1) The original vibration signal s(t) is decomposed into K modal functions u k (t), so that the sum of the estimated bandwidths of each modal function u k (t) is minimized, and the corresponding constraint variational model is expressed as: where {u k} = {u1,...,u k} are the K modal functions resulting from the decomposition; {w k} = {w1,...,w k} are the center frequencies of the respective modalities; is the time derivative, i.e. the differential with respect to time; δ(t) is the unit impulse function; is the analytic signal under Hilbert transform; is the complex modulation factor, which modulates the spectrum of each modal function to the corresponding base frequency band, In order to convert the constrained variational problem into an unconstrained variational problem, a quadratic penalty term factor α and a Lagrange multiplier λ(t) are introduced, and the expression is: Where α is the penalty factor, and λ is the Lagrange multiplier, The alternating direction multiplier algorithm is used to solve the variation problem, and is updated until k modal components are obtained n+1 The process is: where λ n+1 (w) is the new Lagrange multiplier frequency domain value, τ is the Lagrange operator term step size, is the current modal reconstruction error, Set the judgment accuracy ε>0, when the following conditions are met, stop iteration: wherein is the frequency domain representation of the kth modal component at the n+1th iteration, is the frequency domain representation of the kth modal component at the n-1th iteration, is the change ratio of the kth modal component relative to the last iteration, and ε is a set decision accuracy threshold. After the iteration is stopped, the signal is transformed, and the decomposed modal components are finally output.

6. The high-reliability planetary gearbox signal denoising method of claim 1, wherein, In the fourth step, by comparing and analyzing the frequency domain characteristics of the modal components, the variational mode decomposition algorithm VMD is optimized in the decomposition of the original vibration signal, and the sparrow search algorithm is used to iteratively optimize the hyperparameters in the variational mode decomposition algorithm VMD, including the penalty factor α and the decomposition layer number K, the steps of the sparrow search algorithm for iteratively optimizing the hyperparameters in the variational mode decomposition algorithm VMD are as follows: Initialize the candidate solution set and search iteration parameter; According to each solution in the candidate solution set, calculate the value of the objective function, and update the position and speed according to the sparrow behavior law to obtain a new candidate solution set; Screen and sort the new candidate solution set, and decide whether to end the optimization process according to the search iteration parameter.

7. A high-reliability planetary gearbox signal denoising method according to claim 6, characterized in that, In the sparrow search algorithm for iteratively optimizing the hyperparameters in the variational mode decomposition algorithm VMD, (1) Signal acquisition: acquire the original vibration signal of the gear box to be processed, and determine the hyperparameters to be optimized in the variational mode decomposition algorithm VMD as the modal number K of signal decomposition and the penalty factor α, (2) Parameter initialization: set the initial population size, maximum iteration number and search parameter range of the sparrow search algorithm, and generate the initial population containing the modal number K and the penalty factor a by random initialization, (3) Fitness evaluation: for each parameter group in the population, the original signal is decomposed using the variational modal decomposition algorithm VMD to obtain the corresponding modal component set; the envelope entropy of each modal component is calculated, and the sum of all modal envelope entropies is taken as the fitness function value of the current parameter combination; the decomposition performance of the current parameter group is evaluated by taking the minimum total envelope entropy as the target, and the decomposition performance is optimal when the total envelope entropy is minimum, (4) Parameter update: according to the search strategy of sparrow search algorithm SSA, the position of individual in the population is updated to obtain a new parameter combination population, (5) Optimal parameter selection: repeat steps 3 to 4 until the maximum iteration number or fitness convergence is reached, stop the optimization process, and select the parameter group K, a corresponding to the current fitness as the optimal solution.

8. A system for implementing the method of any one of claims 1-7, characterized by It comprises: The acquisition module is used for acquiring original vibration signals of nuclear power pump planetary gearboxes in different health states, and performing detrending and mean value removal preprocessing on the signals; The signal decomposition unit is used for performing modal decomposition on the original vibration signals by using empirical mode decomposition algorithm EMD, ensemble empirical mode decomposition algorithm EEMD and variational modal decomposition algorithm VMD, respectively, to obtain a plurality of different modal components; The comparison unit is used for performing fast Fourier transform on the obtained IMF components, analyzing and evaluating the frequency domain characteristics of each IMF component, and comparing the decomposition effects of empirical mode decomposition algorithm EMD, ensemble empirical mode decomposition algorithm EEMD and variational modal decomposition algorithm VMD to perform algorithm optimization; The optimization unit is used for iterative optimization of algorithm parameters, and adopts sparrow search algorithm SSA for iterative optimization to adaptively determine the hyperparameter value in the optimized algorithm; The reconstruction unit is used for vibration signal decomposition and reconstruction, and adopts the iteratively optimized algorithm to perform modal decomposition on the gearbox vibration signal, remove noise components in the modal components and reconstruct the signal, so as to realize gearbox vibration signal denoising.

9. A computer storage medium, characterized in that The storage medium comprises computer instructions which, when executed on a computer, cause the computer to perform the method of any one of claims 1-7.

10. An electronic device, comprising: The electronic device comprises: A memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any one of claims 1-7 when executing the program.

Images