CN115795275A - Bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising - Google Patents
Bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising Download PDFInfo
- Publication number
- CN115795275A CN115795275A CN202211398056.8A CN202211398056A CN115795275A CN 115795275 A CN115795275 A CN 115795275A CN 202211398056 A CN202211398056 A CN 202211398056A CN 115795275 A CN115795275 A CN 115795275A
- Authority
- CN
- China
- Prior art keywords
- wavelet
- decomposition
- frequency component
- signal
- mode
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention provides a bearing fault diagnosis method based on variational modal decomposition and wavelet joint denoising, which is used for carrying out multi-scale characteristic entropy extraction on a signal subjected to joint denoising; obtaining a bearing fault diagnosis result according to the extracted multi-scale entropy and the pre-trained neural network model; wherein, the joint denoising comprises: determining the optimal decomposition modal layer number K by adopting an energy difference rule best Performing variation modal decomposition on the vibration signal to obtain K best The mode components calculate mutual information entropy of each adjacent mode; according to the mutual information entropy obtained by calculation, determining the boundary point of the high-frequency component mode and the low-frequency component mode of the signal, and performing wavelet threshold function on the high-frequency component modeDenoising, namely reconstructing a low-frequency component mode and a denoised high-frequency component mode to obtain a combined denoised signal; the invention can eliminate the interference of noise while keeping the original signal characteristics to a greater extent, and improves the fault diagnosis effect.
Description
Technical Field
The invention relates to the technical field of bearing fault diagnosis, in particular to a bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising.
Background
The statements in this section merely provide background information related to the present disclosure and may not constitute prior art.
The bearing plays an important role of a joint in industry, and the safe and efficient operation of rotary mechanical equipment can be ensured under good health condition. However, the bearing is very easy to be damaged to greatly reduce the service life and even fail under the complex working condition of variable speed and variable load after long-term operation, and is one of the parts with higher failure rate in mechanical equipment. Related reports have shown that 45% -55% of rotating machine failures are caused by bearing failures.
The amplitude of a pulse signal generated by a local defect is weak when a bearing runs, and the pulse signal is inevitably interfered by harmonic waves, random pulses and background noise in the running process, so that the characteristic of the pulse signal in bearing vibration data measured by an acceleration vibration sensor is not obvious, and the classification accuracy of the classifier is low. Therefore, the method effectively removes noise, improves the accuracy of bearing fault diagnosis, and has important significance for reducing enterprise loss and ensuring personnel life safety.
The inventor finds that in the past practice, a method for denoising a vibration signal by signal decomposition is common. The method simply considers that the high-frequency parts of the signals after decomposition are all noise components and can be directly removed and not considered; in fact, the high-frequency part of the decomposed signal not only has noise components, but also contains effective characteristic components; if the decomposed high-frequency part of the signal is directly removed, it is difficult to retain the effective characteristics in the signal to the maximum extent, resulting in partial effective information loss and higher misjudgment rate in the diagnosis result.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising, which can eliminate the noise interference while keeping the original signal characteristics to a greater extent and improve the fault diagnosis effect.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides a bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising.
A bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising comprises the following processes:
acquiring a bearing vibration signal;
carrying out combined denoising on the obtained bearing vibration signals;
performing multi-scale entropy feature extraction on the combined denoised signal;
obtaining a bearing fault diagnosis result according to the extracted multi-scale entropy and the pre-trained neural network model;
wherein, the joint denoising comprises:
determining the optimal decomposition modal layer number K by adopting an energy difference rule best Performing variation modal decomposition on the vibration signal to obtain K best The mode components calculate mutual information entropy of each adjacent mode;
according to the mutual information entropy obtained through calculation, a boundary point of a high-frequency component mode and a low-frequency component mode of the signal is determined, the high-frequency component mode and the low-frequency component mode are further obtained, wavelet threshold function denoising is conducted on the high-frequency component mode, and the low-frequency component mode and the denoised high-frequency component mode are reconstructed to obtain a combined denoised signal.
As an optional implementation manner of the first aspect of the present invention, the optimal decomposition modal layer number is determined by using an energy difference ruleK best The method comprises the following steps:
performing k and k +1 layer variation modal decomposition on the signal, and calculating the difference value theta of total energy of each decomposition modal when the number of decomposition layers is k and k +1 k,k+1 When theta is increased as the number of decomposition layers is increased k,k+1 When the increase occurs, the signal is over-decomposed, and K at the moment is the optimal decomposition modal layer number K best 。
As an implementation manner of the first aspect of the present invention, the mutual information entropy of adjacent modalities includes:
wherein p (c, d) is two adjacent modal vectors u i And u j P (c) and p (d) are two adjacent modal vectors u i And u j Edge probability distribution of (2).
As a further limitation of the first aspect of the present invention, determining a boundary point m between a high frequency component mode and a low frequency component mode of a signal includes:
where K is the total number of eigenmode functions, MI (u) k ,u k+1 ) Is of mode u k And u k+1 Entropy of mutual information between.
As an implementation manner that can be selected in the first aspect of the present invention, the denoising of the wavelet threshold function is performed on the high-frequency component mode, and includes:
performing wavelet transformation on each high-frequency component mode to obtain a corresponding wavelet coefficient;
setting the adaptive threshold of the wavelet threshold function, performing relevant processing on the wavelet coefficients of the part larger than the set threshold, removing the wavelet coefficients of the part smaller than the set threshold, and performing wavelet reconstruction to obtain a denoised high-frequency component mode.
As a further limitation of the first aspect of the present invention, setting the adaptive threshold λ of the wavelet threshold function comprises:
wherein, σ is the noise standard deviation, n is the signal length, and J is the wavelet decomposition layer number.
As a further limitation of the first aspect of the present invention, the correlating process is performed on wavelet coefficients that are greater than a set threshold, and the wavelet coefficients that are less than the set threshold are removed, and the method includes:
wherein, ω is x The wavelet coefficients before denoising, lambda is the adaptive threshold of the wavelet threshold function, J is the wavelet decomposition coefficient, and sgn is the sign function.
As a further limitation of the first aspect of the present invention, performing multi-scale entropy feature extraction on the jointly denoised signal includes:
carrying out multi-scale processing on the combined denoised signal;
reconstructing a group of vector matrixes with dimension m x = [ x (i), x (i + 1), …, x (i + m-1) ] for the signals subjected to joint denoising under each scale, wherein i is more than or equal to 1 and less than or equal to n-m +1;
calculating the distance d between x (i) and x (j) ij (j is more than or equal to 1 and less than or equal to n-m, j is not equal to i), calculating
Calculating sample entropy SampEn (m, r, n) = lim (-ln (B) m+1 (r)/B m (r)));
And combining the sample entropies of all scales to obtain a multi-scale entropy extraction result.
The invention provides a bearing fault diagnosis system based on variational modal decomposition and wavelet combined denoising.
A bearing fault diagnosis system based on variational modal decomposition and wavelet combined denoising comprises:
a vibration signal acquisition module configured to: acquiring a bearing vibration signal;
a signal joint denoising module configured to: carrying out combined denoising on the obtained bearing vibration signals;
a multi-scale entropy feature extraction module configured to: carrying out multi-scale entropy feature extraction on the combined denoised signal;
a bearing fault diagnostic module configured to: obtaining a bearing fault diagnosis result according to the extracted multi-scale entropy and the pre-trained neural network model;
the joint denoising comprises the following steps:
determining the optimal decomposition modal layer number K by adopting an energy difference rule best Performing variation modal decomposition on the vibration signal to obtain K best Calculating mutual information entropy of each adjacent mode according to the mode components;
according to the mutual information entropy obtained through calculation, a boundary point of a high-frequency component mode and a low-frequency component mode of the signal is determined, the high-frequency component mode and the low-frequency component mode are further obtained, wavelet threshold function denoising is conducted on the high-frequency component mode, and the low-frequency component mode and the denoised high-frequency component mode are reconstructed to obtain a combined denoised signal.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention innovatively provides a bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising, and considers that after a signal is subjected to variational modal decomposition, a high-frequency modal part still has effective information while still having a large amount of noise, and further denoising is carried out by adopting a wavelet threshold function, so that the original signal characteristics can be retained to a greater extent, noise interference is eliminated, and the fault diagnosis effect is improved.
2. The invention innovatively provides a bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising, and due to the high complexity and irregularity of bearing vibration signals under the field complex service conditions of alternating load, strong electromagnetic interference and the like, the traditional time domain and frequency domain indexes are difficult to effectively represent the characteristics of the signals.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
Fig. 1 is an overall flowchart of a bearing fault diagnosis method based on variational modal decomposition and wavelet joint denoising according to embodiment 1 of the present invention;
fig. 2 is a schematic diagram of a HFZZ-II rotating machine fault simulation platform according to embodiment 1 of the present invention;
fig. 3 is a detailed process diagram of a signal joint denoising processing part provided in embodiment 1 of the present invention;
FIG. 4 is a vibration signal under a normal bearing operating condition provided in embodiment 1 of the present invention;
fig. 5 shows a low-frequency component mode of a signal before noise reduction provided in embodiment 1 of the present invention;
fig. 6 shows a high-frequency component mode of a signal before noise reduction according to embodiment 1 of the present invention;
fig. 7 is a high-frequency component mode after noise reduction according to embodiment 1 of the present invention;
FIG. 8 is a signal after the joint noise reduction provided in embodiment 1 of the present invention;
fig. 9 is a network structure diagram of an extreme learning machine provided in embodiment 1 of the present invention.
Detailed Description
The invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, 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 invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
The embodiments and features of the embodiments of the present invention may be combined with each other without conflict.
Example 1:
as shown in fig. 1, embodiment 1 of the present invention provides a bearing fault diagnosis method based on variational modal decomposition and wavelet joint denoising, including the following processes:
s1: experimental System establishment and data preparation
A HFZZ-II rotating machinery fault simulation platform is built, as shown in figure 2, the platform consists of a variable frequency speed regulation three-phase alternating current asynchronous motor, a motor control system, a shaft, a bearing seat (containing a bearing), a radial loading device, an acceleration sensor (two in the horizontal direction and the vertical direction) and the like; the method comprises the steps of utilizing an acceleration sensor to collect vibration signals of a normal bearing, an inner ring fault bearing, an outer ring fault bearing, a rolling body fault bearing and a retainer fault bearing in the operating state, collecting the vibration signals by the sensor at the frequency of 12.8kHz, collecting 60 signals by each bearing, wherein each signal can be obtained by each bearingExpressed as x = (x) 1 ,x 2 ,…,x N ),x i Representing the ith sampling point of the vibration signal x, wherein i is more than or equal to 1 and less than or equal to N, N is the total sampling point number of the signal and is set to 2000, and 300 groups of collected signals divide a training set and a testing set according to the proportion of 7:3 to train and evaluate a subsequent fault diagnosis model.
S2: joint de-noising processing of signals
The signal joint denoising process is composed of three parts, namely signal decomposition, demarcation point discrimination and wavelet threshold function denoising and reconstruction, and the specific flow is shown in fig. 3.
S2.1: signal decomposition
The signal decomposition adopts a Variational Modal Decomposition (VMD) algorithm, and the algorithm can decompose a signal into Intrinsic Mode Function (IMF) components [ u ] at K different frequency scales under the assumption that the number of decomposition layers is K 1 ,u 2 ,…,u K ]However, the decomposition layer number K is often determined according to previous experience and cannot be changed adaptively according to the characteristics of the signal, and the invention provides an energy difference rule to determine the optimal decomposition layer number Kb est :
Firstly, k and k +1 layer variation modal decomposition is carried out on the signal, and theta k is calculated according to a formula (1) ,k+1 Wherein K =4,5, …, K max . At the corresponding theta k,k+1 Then, the number of decomposition layers K was determined according to the following rule: when theta is k,k+1 When smaller, the signal is under-decomposed, and as the number of decomposed layers increases, when theta k,k+1 When the growth occurs, the signal is over-decomposed, and K is the optimal decomposition modal layer number K best 。
Wherein the content of the first and second substances,is the energy of the first mode in k-layer decomposition, E k Is the sum of all modal energies at k-layer decomposition, E k+1 Is the sum of the energies of the respective modes of the k +1 layer decomposition, theta k,k+1 RepresentThe difference of the total energy of each decomposition mode when the decomposition layer number is k and k + 1.
In some optional implementation manners, a signal acquired in a normal bearing operation state is selected for analysis, the signal includes 2000 sampling points, the sampling frequency is 12.8kHz, S2.1 is performed, and K is set max =10, θ for each number of decomposed layers of the signal k,k+1 The values are shown in Table 1.
Table 1: theta at each number of decomposition layers k,k+1 Value of
k | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
θ kk+1 | 1.049 | 0.426 | 0.247 | 0.155 | 0.101 | 0.122 | 0.086 |
As can be seen from Table 1,. Theta. k,k+1 If there is a significant increase in the value of (c) at k =9, then k =8 is the optimal number of decomposition levels for the signal.
S2.2: demarcation point discrimination
According to the information theory, the boundary point of the high-frequency part and the low-frequency part must exist between the modal components after the signal decomposition. Thus obtaining the optimal number of decomposition layers K best Then, the vibration signal is subjected to variation modal decomposition to obtain K best Individual modal componentMutual information entropy of adjacent modes is calculated according to the formula (2):
wherein p (c, d) is a dyad u i And u j Combined probability distribution, p (c) and p (d) each being u i And u j Edge probability distribution of (2).
After mutual information entropy of adjacent modes is obtained, a boundary point m of a high-frequency component mode and a low-frequency component mode of the signal is determined according to a formula (3), namely the boundary point m is a first minimum value point of the mutual information entropy value obtained through calculation.
Where K is the total number of eigenmode functions, MI (u) k ,u k+1 ) Is of mode u k And u k+1 Entropy of mutual information between.
In some optional implementation manners, the signal after S2.1 is selected and subjected to the above steps, and an entropy value of adjacent mode mutual information under 8-layer variation mode decomposition is obtained as shown in table 2, and it can be known from table 2 that a boundary point m =3 between a high-frequency part and a low-frequency part of the signal.
Table 2: entropy of mutual information between adjacent modalities
Mutual information entropy | MI(u 1 ,u 2 ) | MI(u 2 ,u 3 ) | MI(u 3 ,u 4 ) | MI(u 4 ,u 5 ) | MI(u5,u 6 ) | MI(u 6 , 7 ) | MI(u 7 ,u 8 ) |
Numerical value | 0.7874 | 0.8024 | 0.7983 | 0.8073 | 0.7964 | 0.8044 | 0.7913 |
S2.3: wavelet threshold function denoising and reconstruction
After S2.2, the high frequency part of the signal can be obtainedAnd low frequency part L = [ u ] 1 ,u 2 ,…,u m ]Carrying out wavelet threshold function denoising on the high-frequency part, firstly carrying out wavelet transformation on each high-frequency component mode according to a formula (4) to obtain a corresponding wavelet coefficient:
wherein, a is a scale factor, b is a positioning factor, a -1/2 The energy before and after transformation is ensured to be the same, psi is a wavelet basis function, and is set as a 'sym 6' wavelet, and can be set as other wavelet basis functions according to requirements.
Setting adaptive threshold of wavelet threshold function according to formula (5), performing correlation processing on wavelet coefficients larger than the set threshold part according to formula (6), removing wavelet coefficients smaller than the set threshold part, and performing wavelet reconstruction according to formula (7) to obtain denoised high-frequency component modal
Wherein, sigma is noise standard deviation, n is signal length, J is wavelet decomposition layer number, and lambda isAn adaptive threshold, sgn stands for sign function,representing wavelet coefficients, omega, after noise reduction x Representing wavelet coefficients before denoising, a being a scale factor, b being a localization factor, a -1/2 Ensuring the same energy before and after transformation, C ψ The constraint imposed to make the wavelet transform have an inverse transform.
Finally, the high-frequency component mode after noise reductionAnd the original low-frequency component mode L = [ u = [ u ] ] 1 ,u 2 ,…,u m ]Reconstruction, i.e. addition of components, to obtain a jointly de-noised signal
In some optional implementation manners, the signal after S2.2 is selected to perform the above steps, and a low-frequency component mode and a high-frequency component mode after the signal is decomposed are obtained first, where the low-frequency component mode is shown in fig. 5, and the high-frequency component mode is shown in fig. 6. The wavelet threshold function denoising is then performed on the high-frequency component modes, so that the denoised high-frequency component modes are shown in fig. 7, where σ =0.000045, j =8, λ =0.000136, n =2000 are calculated, and the signal after the joint denoising is shown in fig. 8.
S3: signal feature extraction
The multi-scale entropy has the characteristic of high complexity in each scale, the signal is unstable and irregular and high when the bearing vibrates, and after the combined denoising processing of the signal is completed, the multi-scale entropy of the denoised signal is extracted, so that the classification performance of a subsequent model can be obviously enhanced.
Defining algorithm related parameter r, where r is a real number and represents a similarity metric, and is generally 0.1 × std-0st, std represent signalsStandard deviation of (d).
According to the noise-reduced signalReconstructing a set of vector matrices of dimension m x = [ x (i), x (i + 1), …, x (i + m-1) ]]Wherein i is more than or equal to 1 and less than or equal to n-m + 1;
calculating the distance d between x (i) and x (j) ij (j is more than or equal to 1 and less than or equal to n-m, j is not equal to i), and d is satisfied by statistics ij Number num (d) of vectors x (j) under the condition of < r ij R) and is obtained from the formula (8)
Its sample entropy is calculated according to equation (10):
SampEn(m,r,n)=lim(-ln(B m+1 (r)/B m (r))) (10)
the multi-scale entropy calculation principle is to replace a single-scale time sequence of sample entropy by a multi-scale time sequence, wherein the time sequence can be obtained in a coarse graining mode, and a signal after noise reduction is carried outMultiscale is performed and a coarse grained vector is established according to equation (11).
Wherein τ is a time scale.
Calculating at each time scale according to the method for calculating sample entropyThe finally calculated multiple sample entropies are combined together to form a multi-scale entropy which is used as the characteristic of the signal of the rolling bearing in each state.
In some optional implementations, the signal after S2.3 processing is selected and subjected to the above steps, the time scale τ is set to 20, and the extracted multi-scale entropy of the signal is shown in table 3.
Table 3: multi-scale entropy
0.5476 | 0.4087 | 0.3951 | 0.3367 | 0.3035 |
0.3275 | 0.2699 | 0.2133 | 0.1658 | 0.1754 |
0.1547 | 0.1287 | 0.1058 | 0.0883 | 0.0878 |
0.0714 | 0.0734 | 0.0498 | 0.0605 | 0.0371 |
S4: bearing fault diagnosis model building and training
And (4) building an extreme learning machine model, inputting the multi-scale entropy extracted in the step (S3) into the extreme learning machine as fault characteristics, and performing model training.
The extreme learning machine has three structures of an input layer, a hidden layer and an output layer, and the number of neurons in the input layer, the number of neurons in the middle layer and the number of neurons in the output layer are assumed to be n, L and m; hidden layer weight and bias set to ω j 、b j (ii) a Output layer weight is set to beta j Then the network structure is as shown in fig. 9, and the model outputs are as follows:
Y=Hβ (12)
where H represents the output matrix, passing through ω j And b j And (3) performing joint calculation, wherein beta is output weight, Y is a classification result predicted by the model, and expressions of each part of the formula (12) are as follows:
marking the extracted multi-scale entropy values according to fault types, leading the extracted multi-scale entropy values into an extreme learning machine for catenary suspension, continuously reducing the difference between target labels and predicted label values in the final target of the training process, continuously reading in input samples, and continuously updating a solving equation to obtain an output weight matrix capable of being accurately classified, so that the predicted label values can be directly output when a test set passes through, wherein the specific formula is as follows:
minE=min β ||Hβ-T|| (16)
wherein E represents the deviation between the predicted value and the true value, H represents an output matrix, beta is an output weight, T is an actual tag value of the data, and Y is a data tag value predicted by the model.
In the training process, the number of hidden layer nodes in the network and the selection of the activation functions have great influence on the classification effect, and three activation functions, namely Sigmoid, sine and Hardlimit, are selected from the number [0, 400] of the network nodes for analysis.
In some optional implementations, the number of network nodes is 254, and the classification effect is best when the activation function is Hardlimit.
S5: bearing fault diagnosis
After the rolling bearing fault diagnosis model training based on the extreme learning machine is completed, inputting the data extraction characteristics of the test set into the model, and outputting the predicted rolling bearing fault type.
Example 2:
the embodiment 2 of the invention provides a bearing fault diagnosis system based on variational modal decomposition and wavelet combined denoising, which comprises:
a vibration signal acquisition module configured to: acquiring a bearing vibration signal;
a signal joint denoising module configured to: carrying out combined denoising on the obtained bearing vibration signals;
a multi-scale entropy feature extraction module configured to: performing multi-scale entropy feature extraction on the combined denoised signal;
a bearing fault diagnostic module configured to: obtaining a bearing fault diagnosis result according to the extracted multi-scale entropy and the pre-trained neural network model;
wherein, the joint denoising comprises:
determining the optimal decomposition modal layer number K by adopting an energy difference rule best Performing variation modal decomposition on the vibration signal to obtain K best The mode components calculate mutual information entropy of each adjacent mode;
according to the mutual information entropy obtained through calculation, a boundary point of a high-frequency component mode and a low-frequency component mode of the signal is determined, the high-frequency component mode and the low-frequency component mode are further obtained, wavelet threshold function denoising is conducted on the high-frequency component mode, and the low-frequency component mode and the denoised high-frequency component mode are reconstructed to obtain a combined denoised signal.
The working method of the system is the same as the bearing fault diagnosis method based on the variational modal decomposition and wavelet joint denoising in embodiment 1, and is not described herein again.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention has been described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising is characterized by comprising the following processes:
acquiring a bearing vibration signal;
carrying out combined denoising on the obtained bearing vibration signals;
performing multi-scale entropy feature extraction on the combined denoised signal;
obtaining a bearing fault diagnosis result according to the extracted multi-scale entropy and the pre-trained neural network model;
wherein, the joint denoising comprises:
determining the optimal decomposition modal layer number K by adopting an energy difference rule best Performing variation modal decomposition on the vibration signal to obtain K best The mode components calculate mutual information entropy of each adjacent mode;
according to the mutual information entropy obtained through calculation, a boundary point of a high-frequency component mode and a low-frequency component mode of the signal is determined, the high-frequency component mode and the low-frequency component mode are further obtained, wavelet threshold function denoising is conducted on the high-frequency component mode, and the low-frequency component mode and the denoised high-frequency component mode are reconstructed to obtain a combined denoised signal.
2. The bearing fault diagnosis method based on the variational modal decomposition and wavelet joint de-noising as claimed in claim 1,
determining the optimal decomposition modal layer number K by adopting an energy difference rule best The method comprises the following steps:
carrying out k and k +1 layer variation modal decomposition on the signal, and calculating the difference value theta of total energy of each decomposition mode when the number of decomposition layers is k and k +1 k,k+1 When theta is increased as the number of decomposition layers is increased k,k+1 When the increase occurs, the signal is over-decomposed, and K at the moment is the optimal decomposition modal layer number K best 。
3. The bearing fault diagnosis method based on the variational modal decomposition and wavelet joint de-noising as claimed in claim 1,
mutual information entropy of adjacent modalities, including:
wherein p (c, d) is two adjacent modal vectors u i And u j P (c) and p (d) are two adjacent modal vectors u i And u j Edge probability distribution of (2).
4. The bearing fault diagnosis method based on the variational modal decomposition and wavelet joint de-noising as claimed in claim 3,
determining a demarcation point m of a high-frequency component mode and a low-frequency component mode of the signal, which comprises the following steps:
where K is the total number of eigenmode functions, MI (u) k ,u k+1 ) Is of mode u k And u k+1 Entropy of mutual information between.
5. The bearing fault diagnosis method based on the variational modal decomposition and wavelet joint de-noising as claimed in claim 1,
carrying out wavelet threshold function denoising on a high-frequency component mode, wherein the wavelet threshold function denoising comprises the following steps:
performing wavelet transformation on each high-frequency component mode to obtain a corresponding wavelet coefficient;
setting the self-adaptive threshold of the wavelet threshold function, performing relevant processing on the wavelet coefficients of the part larger than the set threshold, removing the wavelet coefficients of the part smaller than the set threshold, and performing wavelet reconstruction to obtain a denoised high-frequency component mode.
6. The bearing fault diagnosis method based on the variational modal decomposition and wavelet joint de-noising as claimed in claim 5,
setting an adaptive threshold λ of a wavelet threshold function, comprising:
wherein, σ is the noise standard deviation, n is the signal length, and J is the wavelet decomposition layer number.
7. The bearing fault diagnosis method based on the variational modal decomposition and wavelet joint de-noising as claimed in claim 6,
the wavelet coefficient which is larger than the set threshold value part is subjected to correlation processing, and the wavelet coefficient which is smaller than the set threshold value part is removed, and the method comprises the following steps:
wherein, ω is x The wavelet coefficients before denoising, lambda is the adaptive threshold of the wavelet threshold function, J is the wavelet decomposition coefficient, and sgn is the sign function.
8. The bearing fault diagnosis method based on the variational modal decomposition and wavelet joint de-noising as claimed in claim 6,
carrying out multi-scale entropy feature extraction on the combined denoised signal, comprising:
carrying out multi-scale on the jointly denoised signal;
reconstructing a group of vector matrixes with dimension m x = [ x (i), x (i + 1), …, x (i + m-1) ] for the signals subjected to joint denoising under each scale, wherein i is more than or equal to 1 and less than or equal to n-m +1;
calculating the distance d between x (i) and x (j) ij (j is more than or equal to 1 and less than or equal to n-m, j is not equal to i), calculating
Calculating sample entropy SampEn (m, r, n) = lim (-ln (B) m+1 (r)/B m (r)));
And combining the sample entropies of all scales to obtain a multi-scale entropy feature extraction result.
9. A bearing fault diagnosis system based on variational modal decomposition and wavelet combined denoising is characterized in that,
the method comprises the following steps:
a vibration signal acquisition module configured to: acquiring a bearing vibration signal;
a signal joint denoising module configured to: carrying out combined denoising on the obtained bearing vibration signals;
a multi-scale entropy feature extraction module configured to: performing multi-scale entropy feature extraction on the combined denoised signal;
a bearing fault diagnostic module configured to: obtaining a bearing fault diagnosis result according to the extracted multi-scale entropy and the pre-trained neural network model;
wherein, the joint denoising comprises:
determining the optimal decomposition modal layer number K by adopting an energy difference rule best Performing variation modal decomposition on the vibration signal to obtain K best The mode components calculate mutual information entropy of each adjacent mode;
according to the mutual information entropy obtained through calculation, a boundary point of a high-frequency component mode and a low-frequency component mode of the signal is determined, the high-frequency component mode and the low-frequency component mode are further obtained, wavelet threshold function denoising is conducted on the high-frequency component mode, and the low-frequency component mode and the denoised high-frequency component mode are reconstructed to obtain a combined denoised signal.
10. The system for diagnosing bearing failure based on the joint de-noising of the variational modal decomposition and the wavelet as claimed in claim 9,
determining the optimal decomposition modal layer number K by adopting an energy difference rule best The method comprises the following steps:
carrying out k and k +1 layer variation modal decomposition on the signal, and calculating the difference value theta of total energy of each decomposition mode when the number of decomposition layers is k and k +1 k,k+1 When theta is increased as the number of decomposition layers is increased k,k+1 When the increase occurs, the signal is over-decomposed, and K at the moment is the optimal decomposition modal layer number K best 。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211398056.8A CN115795275A (en) | 2022-11-09 | 2022-11-09 | Bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211398056.8A CN115795275A (en) | 2022-11-09 | 2022-11-09 | Bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115795275A true CN115795275A (en) | 2023-03-14 |
Family
ID=85436305
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211398056.8A Pending CN115795275A (en) | 2022-11-09 | 2022-11-09 | Bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115795275A (en) |
-
2022
- 2022-11-09 CN CN202211398056.8A patent/CN115795275A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111323220B (en) | Fault diagnosis method and system for gearbox of wind driven generator | |
Chen et al. | Fault feature extraction and diagnosis of rolling bearings based on wavelet thresholding denoising with CEEMDAN energy entropy and PSO-LSSVM | |
Al-Bugharbee et al. | A fault diagnosis methodology for rolling element bearings based on advanced signal pretreatment and autoregressive modelling | |
CN113780056B (en) | Rolling bearing fault diagnosis method based on vibration signals | |
CN106682688B (en) | Particle swarm optimization-based stacked noise reduction self-coding network bearing fault diagnosis method | |
CN112304613B (en) | Wind turbine generator bearing early warning method based on feature fusion | |
CN110795843B (en) | Method and device for identifying faults of rolling bearing | |
CN104792530B (en) | Deep-learning rolling bearing fault diagnosis method based on SDA (stacked denoising autoencoder) and Softmax regression | |
CN110617966A (en) | Bearing fault diagnosis method based on semi-supervised generation countermeasure network | |
Guo et al. | Deep convolution feature learning for health indicator construction of bearings | |
CN109443752B (en) | Gear vibration signal noise reduction and fault diagnosis method based on VMD | |
CN113176092B (en) | Motor bearing fault diagnosis method based on data fusion and improved experience wavelet transform | |
Gharehbaghi et al. | Supervised damage and deterioration detection in building structures using an enhanced autoregressive time-series approach | |
Huang et al. | Deep residual networks-based intelligent fault diagnosis method of planetary gearboxes in cloud environments | |
Che et al. | Intelligent fault diagnosis method of rolling bearing based on stacked denoising autoencoder and convolutional neural network | |
CN113627375A (en) | Planetary gear fault diagnosis method and system, storage medium and computing device | |
CN114118138A (en) | Bearing composite fault diagnosis method based on multi-label field self-adaptive model | |
CN114662386A (en) | Bearing fault diagnosis method and system | |
Wang et al. | A novel rolling bearing fault diagnosis method based on adaptive denoising convolutional neural network under noise background | |
CN114371009A (en) | High-speed train bearing fault diagnosis method based on improved random forest | |
CN115795275A (en) | Bearing fault diagnosis method based on variational modal decomposition and wavelet combined denoising | |
CN112132069A (en) | Rolling bearing weak fault intelligent diagnosis method based on deep learning | |
CN115452377B (en) | Rolling bearing fault diagnosis method based on CEEMDAN and adaptive denoising | |
CN111721528A (en) | Wind generating set gear box fault early warning method based on CMS system big data | |
CN115901255A (en) | Bearing fault diagnosis method based on parameter optimization VMD-SVM |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |