CN114088401A - Fault analysis method and device for rolling bearing of wind driven generator - Google Patents
Fault analysis method and device for rolling bearing of wind driven generator Download PDFInfo
- Publication number
- CN114088401A CN114088401A CN202111294441.3A CN202111294441A CN114088401A CN 114088401 A CN114088401 A CN 114088401A CN 202111294441 A CN202111294441 A CN 202111294441A CN 114088401 A CN114088401 A CN 114088401A
- Authority
- CN
- China
- Prior art keywords
- signal
- fault
- fault analysis
- rolling bearing
- bearing
- 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
- 238000004458 analytical method Methods 0.000 title claims abstract description 166
- 238000005096 rolling process Methods 0.000 title claims abstract description 107
- 239000011159 matrix material Substances 0.000 claims abstract description 113
- 238000000034 method Methods 0.000 claims abstract description 17
- 238000012549 training Methods 0.000 claims description 119
- 238000012360 testing method Methods 0.000 claims description 51
- 238000000354 decomposition reaction Methods 0.000 claims description 18
- 230000009466 transformation Effects 0.000 claims description 15
- 238000001228 spectrum Methods 0.000 claims description 12
- 230000006378 damage Effects 0.000 description 19
- 238000010586 diagram Methods 0.000 description 18
- 238000005299 abrasion Methods 0.000 description 12
- 238000011176 pooling Methods 0.000 description 9
- 238000003745 diagnosis Methods 0.000 description 7
- 238000004445 quantitative analysis Methods 0.000 description 5
- 238000000605 extraction Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000009528 severe injury Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010892 electric spark Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention provides a fault analysis method and a fault analysis device for a rolling bearing of a wind driven generator, wherein the method comprises the following steps: acquiring a signal to be detected corresponding to a bearing to be detected; determining a first signal point matrix according to a signal to be detected; determining a first characteristic map corresponding to the signal to be detected according to the first signal point matrix; and analyzing according to the first characteristic map to obtain the fault grade information of the bearing to be tested corresponding to the first characteristic map.
Description
Technical Field
The disclosure relates to the field of fault analysis, and in particular relates to a fault analysis method and device for a rolling bearing of a wind driven generator.
Background
The rolling bearing is an important part of the wind driven generator, and is easy to damage due to long-term continuous operation under heavy load and variable rotating speed environments in the using process of the wind driven generator; therefore, the damage condition of the rolling bearing needs to be judged by acquiring the vibration signal of the bearing; the vibration signal of the wind driven generator is usually submerged in a strong background noise environment, is easily influenced by the vibration signal of an excitation source of other equipment, and is difficult to realize quantitative analysis of the damage degree.
Disclosure of Invention
The present disclosure provides a fault analysis method and apparatus for a rolling bearing of a wind turbine to at least solve the above technical problems in the prior art.
According to a first aspect of the present disclosure, there is provided a fault analysis method for a rolling bearing, the method including: acquiring a signal to be detected corresponding to a bearing to be detected; determining a first signal point matrix according to the signal to be detected; determining a first characteristic map corresponding to the signal to be detected according to the first signal point matrix; and analyzing according to the first characteristic map to obtain the fault grade information of the bearing to be detected corresponding to the first characteristic map.
In an implementation manner, the analyzing according to the first feature map to obtain the fault level information of the bearing to be tested corresponding to the first feature map includes: obtaining a deep groove ball bearing training sample, and performing model training through the deep groove ball bearing training sample to obtain a first fault analysis model, wherein the first fault analysis model is used for performing fault analysis on the deep groove ball bearing; acquiring a rolling bearing training sample, and training the first fault analysis model through the rolling bearing training sample to obtain a second fault analysis model, wherein the second fault analysis model is used for carrying out fault analysis on the rolling bearing; and inputting the first characteristic map into the second fault analysis model to obtain the fault grade information of the bearing to be tested corresponding to the first characteristic map.
In an implementation manner, obtaining a rolling bearing training sample, and training the first fault analysis model through the rolling bearing training sample to obtain a second fault analysis model includes: obtaining a test sample, and testing the second fault analysis model through the test sample to obtain a test result, wherein the test result at least comprises an analysis accuracy rate corresponding to fault grade information; when the analysis accuracy rate meets a specified threshold value, determining the fault grade information of the bearing to be detected corresponding to the first characteristic map according to the second fault analysis model; and when the analysis accuracy does not meet a specified threshold value, acquiring the rolling bearing training sample, and retraining the second fault analysis model through the rolling bearing training sample.
In an embodiment, the obtaining a rolling bearing training sample includes: acquiring a training signal corresponding to a training bearing; determining a second signal point matrix according to the training signal; obtaining a second feature map corresponding to the training signal according to the second signal point matrix; acquiring fault grade information corresponding to each second feature map; and assembling the rolling bearing training sample according to a plurality of second feature maps and fault level information corresponding to the second feature maps.
In one embodiment, determining a first signal point matrix according to the signal to be measured includes: randomly intercepting a plurality of first signal points in a signal to be detected; constructing the first signal point matrix from the hank's matrix and the plurality of first signal points.
In an embodiment, the determining a first feature map corresponding to the signal to be measured according to the first signal point matrix includes: singular value decomposition is carried out on the first signal point matrix to obtain a plurality of effective singular values; performing signal reconstruction according to the effective singular value to obtain a vibration signal; and carrying out S transformation on the vibration signal to obtain a first characteristic map corresponding to the vibration signal.
In an embodiment, the performing singular value decomposition on the first signal point matrix to obtain a plurality of effective singular values includes: performing singular value decomposition on the first signal point matrix to obtain a first singular value matrix, wherein the first singular value matrix comprises a plurality of singular values; determining a first number of valid singular values from a singular value curvature spectrum; and determining a first number of singular values in the first singular value matrix as valid singular values according to the sequence in the first singular value matrix.
In one embodiment, the remaining singular values in the first singular value matrix are determined as invalid singular values; correspondingly, the reconstructing the signal according to the effective singular value to obtain the vibration signal includes: setting invalid singular values in the first singular value matrix to zero to obtain a second singular value matrix; and performing signal reconstruction according to the second singular value rectangle to obtain a vibration signal.
According to a second aspect of the present disclosure, there is provided a failure analysis device for a rolling bearing, the device including: the acquisition module is used for acquiring a signal to be detected corresponding to the bearing to be detected; the determining module is used for determining a first signal point matrix according to the signal to be detected; the determining module is used for determining a first characteristic map corresponding to the signal to be detected according to the first signal point matrix; and the analysis module is used for analyzing according to the first characteristic map to obtain the fault grade information of the bearing to be tested corresponding to the first characteristic map.
In one embodiment, the analysis module includes: the acquisition submodule is used for acquiring a deep groove ball bearing training sample, performing model training through the deep groove ball bearing training sample and acquiring a first fault analysis model, wherein the first fault analysis model is used for performing fault analysis on the deep groove ball bearing; the acquisition submodule is also used for acquiring a rolling bearing training sample, training the first fault analysis model through the rolling bearing training sample and acquiring a second fault analysis model, and the second fault analysis model is used for carrying out fault analysis on the rolling bearing; and the input submodule is used for inputting the first characteristic map into the second fault analysis model to obtain the fault grade information of the bearing to be tested corresponding to the first characteristic map.
In an implementation manner, the obtaining sub-module is further configured to obtain a test sample, and test the second fault analysis model through the test sample to obtain a test result, where the test result at least includes an analysis accuracy corresponding to the fault level information; when the analysis accuracy rate meets a specified threshold value, determining the fault grade information of the bearing to be detected corresponding to the first characteristic map according to the second fault analysis model; and when the analysis accuracy does not meet a specified threshold value, acquiring the rolling bearing training sample, and retraining the second fault analysis model through the rolling bearing training sample.
In an embodiment, the obtaining sub-module is further configured to obtain a training signal corresponding to a training bearing; determining a second signal point matrix according to the training signal; obtaining a second feature map corresponding to the training signal according to the second signal point matrix; acquiring fault grade information corresponding to each second feature map; and assembling the rolling bearing training sample according to a plurality of second feature maps and fault level information corresponding to the second feature maps.
In an implementation manner, the determining module is further configured to randomly intercept a plurality of first signal points in the signal to be measured; constructing the first signal point matrix from the hank's matrix and the plurality of first signal points.
In an embodiment, the determining module includes: the decomposition submodule is used for carrying out singular value decomposition on the first signal point matrix to obtain a plurality of effective singular values; the reconstruction submodule is used for carrying out signal reconstruction according to the effective singular value to obtain a vibration signal; and the transformation submodule is used for carrying out S transformation on the vibration signal to obtain a first characteristic map corresponding to the vibration signal.
In an implementation manner, the decomposition submodule is further configured to perform singular value decomposition on the first signal point matrix to obtain a first singular value matrix, where the first singular value matrix includes a plurality of singular values; determining a first number of valid singular values from a singular value curvature spectrum; and determining a first number of singular values in the first singular value matrix as valid singular values according to the sequence in the first singular value matrix.
In one embodiment, the remaining singular values in the first singular value matrix are determined as invalid singular values; correspondingly, the reconstruction submodule 5022 is further configured to set invalid singular values in the first singular value matrix to zero to obtain a second singular value matrix; and performing signal reconstruction according to the second singular value rectangle to obtain a vibration signal.
According to a third aspect of the present disclosure, there is provided a non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method of the present disclosure.
The invention discloses a fault analysis method and a fault analysis device for a rolling bearing of a wind driven generator, wherein a first signal point matrix is constructed through signal points in a signal to be detected through the signal to be detected of the bearing to be detected; the first characteristic map corresponding to the signal to be detected is determined through the first signal point matrix, so that vibration information of the bearing to be detected is displayed, the first characteristic map is analyzed, fault grade information of the bearing to be detected is obtained, quantitative analysis of the bearing to be detected is achieved, and the damage degree of the bearing to be detected is determined.
It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present disclosure, nor do they limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.
Drawings
The above and other objects, features and advantages of exemplary embodiments of the present disclosure will become readily apparent from the following detailed description read in conjunction with the accompanying drawings. Several embodiments of the present disclosure are illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which:
in the drawings, the same or corresponding reference numerals indicate the same or corresponding parts.
Fig. 1 is a schematic implementation flow diagram of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure;
FIG. 2 is a schematic diagram of a model training flow of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure;
FIG. 3 is a schematic implementation flow chart of a fault analysis method for a rolling bearing according to an embodiment of the disclosure;
FIG. 4 is a schematic diagram of a sample acquisition flow of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure;
FIG. 5 is a first signal diagram of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure;
FIG. 6 is a diagram illustrating a singular value curve of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure;
FIG. 7 is a diagram illustrating a spectrum of singular value curves of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure;
FIG. 8 is a schematic reconstructed signal diagram of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure;
FIG. 9 is a schematic diagram of a characteristic map of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure;
fig. 10 is a schematic view of a bottleneck structure of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure;
FIG. 11a is a time domain signal diagram of a normal state of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure;
FIG. 11b is a time domain signal diagram of a slight damage state of the fault analysis method for the rolling bearing according to the embodiment of the present disclosure;
FIG. 11c is a time domain signal diagram of a severe damage state of the fault analysis method for a rolling bearing according to the embodiment of the present disclosure;
fig. 12a is a time-frequency map diagram in a normal state of the fault analysis method for a rolling bearing according to the embodiment of the disclosure;
FIG. 12b is a schematic time-frequency diagram of a slight damage condition of the fault analysis method for a rolling bearing according to the embodiment of the disclosure;
fig. 12c is a schematic time-frequency map of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure in case of severe damage;
FIG. 13 is a schematic diagram of a fault model adaptive extraction feature of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure;
fig. 14 is a schematic view of a failure analysis device for a rolling bearing according to an embodiment of the present disclosure.
Detailed Description
In order to make the objects, features and advantages of the present disclosure more apparent and understandable, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present disclosure, and it is apparent that the described embodiments are only a part of the embodiments of the present disclosure, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments disclosed herein without making any creative effort, shall fall within the protection scope of the present disclosure.
Fig. 1 is a schematic implementation flow diagram of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure; with reference to figure 1 of the drawings,
according to a first aspect of the present disclosure, there is provided a fault analysis method for a rolling bearing, the method including: step 101, acquiring a signal to be detected corresponding to a bearing to be detected; step 102, determining a first signal point matrix according to a signal to be detected; 103, determining a first characteristic map corresponding to the signal to be detected according to the first signal point matrix; and 104, analyzing according to the first characteristic map to obtain the fault grade information of the bearing to be tested corresponding to the first characteristic map.
The invention discloses a fault analysis method and a fault analysis device for a rolling bearing of a wind driven generator, wherein a first signal point matrix is constructed through signal points in a signal to be detected through the signal to be detected of the bearing to be detected; the first characteristic map corresponding to the signal to be detected is determined through the first signal point matrix, so that vibration information of the bearing to be detected is displayed, the first characteristic map is analyzed, fault grade information of the bearing to be detected is obtained, quantitative analysis of the bearing to be detected is achieved, and the damage degree of the bearing to be detected is determined.
In step 101, the bearing to be tested refers to a rolling bearing, and further refers to a rolling bearing applied to a wind driven generator; acquiring vibration signals in multiple directions of a rolling bearing through a sensor to obtain multiple vibration signals, and determining the vibration signals with noise smaller than a specified threshold value in the multiple vibration signals as signals to be detected; or determining the vibration signal with the lowest noise value in the plurality of vibration signals as the signal to be detected; wherein, the sensor can be an ICP piezoelectric acceleration sensor; the plurality of directions may be both horizontal and vertical directions. The number of the bearings to be measured is not limited.
In step 102, randomly intercepting a plurality of first signal points in a signal to be detected; a first signal point matrix is constructed from the Hankel matrix and the plurality of first signal points. The random interception is to intercept signal points by randomly intercepting a starting point in a signal to be detected, and intercept a plurality of first signal points; specifically, the number of the first signal points may be determined to be 2048; 2048 signal points are constructed by means of a hankel matrix into a first signal point matrix of 1024 × 1025 orders.
In step 103, determining a first characteristic spectrum through a first signal point matrix, and first, performing singular value decomposition on the first signal point matrix to reduce noise of the first signal point matrix; specifically, a first singular value matrix is obtained by performing singular value decomposition on a first signal point matrix, wherein the first singular value matrix comprises a plurality of singular values; determining a first number of effective singular values in the singular values through a curve and a numerical value of a maximum peak value of a singular value curvature spectrum; determining a first number of singular values in the first singular value matrix as valid singular values according to the sequence in the first singular value matrix; the orders in the first singular value matrix are arranged from big to small. Determining the residual singular values in the first singular value matrix as invalid singular values according to the valid singular values, and setting the invalid singular values to zero, further determining non-zero singular values in the residual singular values as invalid singular values, and obtaining a second singular value matrix; performing signal reconstruction on the second singular value rectangle to obtain a vibration signal; and S transformation is carried out on the vibration signal to obtain a first characteristic map corresponding to the vibration signal. The first characteristic map is used for characterizing the vibrations emitted by the rolling bearing. The S transformation is time-frequency transformation, can show time domain and frequency domain characteristics in the vibration signal, and is beneficial to quantitative analysis of fault damage degree. According to the method, effective singular values are selected through the singular value curvature spectrum to carry out signal reconstruction, the fault damage degree information is quantified, the effective information is prevented from being lost, the fault damage degree is quantitatively analyzed, and meanwhile the problem that fault characteristics are difficult to extract when a rolling bearing of the wind driven generator is in an environment with complex working conditions, more signals of other equipment sources and strong background noise is solved. And then, the reconstructed signal is subjected to S transformation by utilizing the inverse transformation nondestructive property of the S transformation and the uniqueness of mapping on the time sequence to generate a feature map containing obvious fault features, the feature map is input into a neural network, and the accuracy of identifying the damage degree of the fault analysis model is improved by self-adapting the rest fault indexes and the damage degree indexes of the features.
In step 104, a fault analysis model may be obtained through training, and the first feature map is analyzed to obtain fault level information of the bearing to be tested corresponding to the first feature map. The fault grade information at least comprises information of whether the bearing to be detected is damaged and the degree grade of the damage, and specifically, the output result can be normal state, slight abrasion, moderate abrasion and severe abrasion. And recording one step, and comparing the characteristic map with the reference model to judge the fault grade information of the bearing to be detected. Preferably, model analysis is used.
FIG. 2 is a schematic diagram of a model training flow of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure; refer to fig. 2;
in an implementation manner, in step 104, performing analysis according to the first feature map to obtain fault level information of the bearing to be tested corresponding to the first feature map, including: step 201, obtaining a deep groove ball bearing training sample, and performing model training through the deep groove ball bearing training sample to obtain a first fault analysis model, wherein the first fault analysis model is used for performing fault analysis on the deep groove ball bearing; 202, acquiring a rolling bearing training sample, training a first fault analysis model through the rolling bearing training sample, and acquiring a second fault analysis model, wherein the second fault analysis model is used for carrying out fault analysis on the rolling bearing; and 203, inputting the first characteristic map into the second fault analysis model to obtain the fault grade information of the bearing to be detected corresponding to the first characteristic map.
In steps 201-203, obtaining a deep groove ball bearing training sample, performing model training through the deep groove ball bearing training sample, and obtaining a first fault analysis model, wherein data of the deep groove ball bearing is easier to obtain than data of a rolling bearing, so that training of the first fault analysis model is performed through obtaining the deep groove ball bearing, based on the first fault analysis model, the first fault analysis model is trained through obtaining the rolling bearing training model, and a second training model is obtained, wherein the training sample of the rolling bearing comprises a plurality of characteristic maps in a plurality of fault states under different loads; the load at least comprises three loads of 0Hp, 1Hp and 2 Hp; the plurality of fault states at least comprise characteristic maps corresponding to light abrasion, medium abrasion and heavy abrasion of the rolling body, the inner ring and the outer ring of the rolling bearing and the characteristic maps of the rolling bearing in a normal state. The second fault analysis model is used for carrying out fault analysis on the rolling bearing; and inputting the first characteristic map to be detected into the second fault analysis model to obtain the fault grade information of the bearing to be detected corresponding to the first characteristic map. The number of the deep groove ball bearing training samples is far more than that of the rolling bearing training samples; therefore, the first fault analysis model with sufficient width is obtained through the deep groove ball bearing training sample, the accuracy of the model is improved through the training sample of the rolling bearing, and the second fault analysis model with sufficient depth is obtained. Thereby the model can be suitable for production and use.
FIG. 3 is a schematic test flow chart of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure; refer to fig. 3;
in one embodiment, in step 202, obtaining a rolling bearing training sample, training the first fault analysis model through the rolling bearing training sample, and obtaining a second fault analysis model includes: step 301, obtaining a test sample, and testing the second fault analysis model through the test sample to obtain a test result, wherein the test result at least comprises an analysis accuracy corresponding to the fault level information; step 302, when the analysis accuracy rate meets a specified threshold value, determining the fault grade information of the bearing to be detected corresponding to the first characteristic map according to the second fault analysis model; and 303, when the analysis accuracy does not meet the specified threshold value, acquiring a rolling bearing training sample, and retraining the second fault analysis model through the rolling bearing training sample.
In step 301, a test sample is obtained, where the test sample is used to test the second fault analysis model, and the second fault analysis model is tested through the test sample to obtain a test result, where the test result at least includes an analysis accuracy corresponding to the fault level information; specifically, a plurality of test samples are input into the second fault analysis model, and an analysis result is obtained, namely the fault grade information obtained by analysis is compared with the fault grade information corresponding to the test samples, so that the analysis accuracy is determined; specifically, the test sample includes a third feature map and failure level information corresponding to the third feature map, where the failure level information may be labeled on the third feature map in the form of a label.
In steps 302-303, when the analysis accuracy meets a specified threshold, wherein the specified threshold may be 90%; when the accuracy of fault grade information obtained by testing the test sample according to the second fault analysis model reaches 90%, determining that the model training is finished; and when the analysis accuracy rate does not meet the specified threshold value, acquiring a rolling bearing training sample, retraining the second fault analysis model through the rolling bearing training sample, and determining that the model training is finished until the analysis accuracy rate meets the specified threshold value. And analyzing the first characteristic map through the completed second fault analysis model.
FIG. 4 is a schematic diagram of a sample acquisition flow of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure; refer to fig. 4;
in one embodiment, the step 202 of obtaining a rolling bearing training sample includes: step 401, acquiring a training signal corresponding to a training bearing; step 402, determining a second signal point matrix according to the training signal; obtaining a second characteristic map corresponding to the training signal according to the second signal point matrix; step 403, acquiring fault grade information corresponding to each second feature map; and step 404, assembling rolling bearing training samples according to the plurality of second characteristic maps and the fault level information corresponding to the second characteristic maps.
In steps 401-404, acquiring a signal of a rolling bearing for training by a sensor to obtain a training signal, determining a second signal point matrix for the training signal, performing singular value decomposition on the second signal point matrix, and performing S transformation to obtain a second characteristic map corresponding to the training signal; and acquiring the fault grade information corresponding to each second feature map, namely acquiring the fault grade information corresponding to the training bearing, specifically, marking the fault grade information as a label on the second feature maps, thereby acquiring the rolling bearing training sample. Further, 20% of the training samples may be used as test samples.
A specific embodiment is provided:
the 2048 signal points of the first signal are cut in a random mode according to a starting point, and the 2048 signal points are constructed into a first matrix of 1024 × 1025 orders according to a Hankel matrix.
Wherein the hank-kerr matrix is:
where Em is an m × N-order matrix, N is a data length, N is m + N-1, and m is less than or equal to N. N of the first matrix is 2048.
Performing singular value decomposition on the first matrix; the singular value curve is shown in fig. 6, where the abscissa in fig. 6 is used to characterize the sequence of singular values and the ordinate is used to characterize the singular value values.
According to the singular value formula: em＝UWVTWhere T is the transpose of the matrix; w is a singular value matrix, U is a matrix of m × m orders, and V is a matrix of n × n orders. I.e. the first matrix is decomposed into three matrices with singular value matrices.
Front gamma order diagonal element gamma of singular value matrix Wi(i ═ 1,2, …, r) is the singular value of the matrix W, which mainly reflects the energy concentration of the elements, and is arranged in the matrix in descending order, where the sequence of diag singular values is shown below.
The formula for the curvature is as follows:
the singular value sequence B is: b ═ γ 1, γ 2, …, γ n }; calculating to obtain a singular value curvature spectrum as shown in FIG. 7, wherein the abscissa singular value sequence and the ordinate singular value curvature are obtained; determining a first number of effective singular values according to the singular value curvature spectrum shown in the figure 7, wherein the singular value curve is convex at the position coordinate K of the maximum peak value of the curvature spectrum, and the number of the effective singular values is K; if the singular value curve is concave at K, then the number of valid singular values is K-1. For a hankel matrix constructed with noisy signals, there is no correlation between two adjacent rows of elements. Therefore, the effective values of the noise signal and the vibration signal can be effectively distinguished through the turning points, and the distinguishing can be realized according to the singular value curvature spectrum. The non-zero singular values except the effective singular value represent noise or other interference information, and the noise is set to zero, so that the purpose of reducing noise is achieved.
According to the first number of the effective singular values, the effective singular values in the singular value matrix are determined, the rest singular values are set to zero, then signal reconstruction is carried out according to the matrix W after zero setting, and a reconstructed second signal is obtained, wherein the second signal is shown in fig. 8.
Performing S-transform on the second signal to obtain a feature map as shown in fig. 9; specifically, the theory of S transformation is:
wherein: f is frequency, tau is the center point of the time window function, and t is the time variable.
Wherein is
The training set is a set of training data,
training set labels, AtrainThe number of samples in the training set;
the test set is a set of tests that,
test set tag, AtestTesting the number of samples in the collection; specifically, the sample configuration is shown in the following table:
adopting a rolling bearing experimental data set, namely an SKF6205-2RS deep groove ball bearing, preferably, the inner diameter of the bearing is 25mm, the outer diameter is 52mm, the thickness is 15mm, the pitch circle diameter is 39mm, the diameter of a rolling body is 7.94mm, the number of rollers is 9, the power of a driving motor of an experimental table is 1.5Kw, the sampling frequency is 12KHz, and the bearing is damaged by adopting an electric spark single point; and performing model training as a deep groove ball bearing training sample to obtain a first fault analysis model.
Inputting the training set samples into a first fault analysis model for learning and training to obtain a second fault analysis model, which can be specifically an ICNN model; inputting the test sample into the trained ICNN model to carry out fault diagnosis and damage degree judgment, judging whether the diagnosis of the test set is accurate, if so, storing the trained ICNN model, otherwise, returning to retrain the ICNN model until the diagnosis effect of the test set is ideal, and forming a final ICNN model structure. The ICNN network architecture model is shown in the following table.
The final ICNN model structure comprises an input layer, a first volume layer, a second volume layer, a BN layer, a first pooling layer, a third volume layer, a fourth volume layer, a BN layer, a second pooling layer, a Bottleneck structure, a BN layer, a third pooling layer, a Bottleneck structure, a BN layer, a fourth pooling layer, a Bottleneck structure, a BN layer, a fifth pooling layer, a first full-link layer, a Dropout layer, a second full-link layer, a Dropout layer and a Softmax layer; specifically, the first convolution layer is a wide convolution layer, and the convolution kernel width is 32;
the inputs to the first layer of convolutional layers are: s (τ, f); s is a matrix obtained after S transformation, tau is the central point of the time window function, and t is a time variable.
The first layer convolutional layer output xl is:
xl＝σ(Wl*xl-1+bl)；
wherein is the convolution operation symbol, WlIs the weight of the first convolutional layer, blBiased for layer l, xl-1For the output of the upper network, xlFor the current l-th layer output, σ is the activation function, and a rectifying linear unit (Relu) is used in the convolutional layer.
The improved CNN model adopts a Bottleneck structure, 27 small-scale convolution kernels are adopted to replace original large-scale convolution kernels for feature extraction, a 3 x 256 convolution layer is changed to pass through 1 x 1 convolution kernels, then pass through 3 x 3 convolution kernels and finally pass through 1 x 1 convolution kernels, fault features are effectively extracted, model parameters are reduced, and model training speed is improved; the bottle neck structure is shown in fig. 10:
the maximum pooling is adopted in the pooling layer:
sl＝fdown(xl-1)；
wherein f isdownFor pooling operations, xlThe output of the current l-th layer pooling layer;
dropout is introduced into the fully connected layer to randomly inactivate part of neurons and prevent overfitting, and the expression is as follows, and the structure is shown as the following:
wherein: p (Pi ═ 1) ═ P, belonging to the bernoulli random variable probability distribution; p is the probability of generating 1 for sample i; ba is the number of samples a in class i neurons.
The full connecting layer is as follows:
hl＝σ(Wl*hl-1+bl)；
wherein h isl-1Is the output of the upper network; h islIs the output of the current full link layer; wlIs the weight; blIs an offset; sigma is an activation function;
adopting an ADAM algorithm to adjust the network weight and deviation, and adopting a gradient descent method to optimize an ICNN network model;
and connecting a Softmax classifier after the second full connection layer:
wherein Z is the result obtained by weight mapping of the output of the last neuron, and K is the number of samples.
Furthermore, as the iteration times of the model are increased, the accuracy of fault diagnosis and damage degree judgment of the model is higher and higher, the loss value is lower and lower, when the iteration times are larger than 30, the accuracy of fault identification and damage degree judgment is 100%, the convergence speed of the network structure model is high, the identification accuracy is high, the method has practical significance under the actual variable load working condition, and the rolling bearing fault can be found in time for maintenance and replacement.
FIG. 13 is a schematic diagram of a fault model adaptive extraction feature of a fault analysis method for a rolling bearing according to an embodiment of the present disclosure; carrying out self-adaptive feature extraction on the training sample through the finally obtained ICNN model to obtain a feature distribution map; the training sample characteristics corresponding to 0-9 are completely separated without staggered areas, so that the model obtained by training is proved to have higher accuracy, fault characteristics can be accurately distinguished, and the method can be suitable for providing accurate fault analysis results under the complex working condition of the rolling bearing of the wind driven generator.
The method adopts a time-frequency analysis method, overcomes the problem of local information loss of time domain and frequency domain independent research, adopts an improved convolutional neural network to adaptively extract characteristics, introduces a Bottleneck structure, adds fine tuning, does not need to retrain a model aiming at different working conditions, and improves the accuracy of fault diagnosis and damage degree quantitative analysis.
Fig. 14 is a schematic view of a failure analysis device for a rolling bearing according to an embodiment of the present disclosure; refer to fig. 14;
according to a second aspect disclosed in an embodiment of the present invention, there is provided a failure analysis device for a rolling bearing, the device including: an obtaining module 501, configured to obtain a to-be-detected signal corresponding to a to-be-detected bearing; a determining module 502, configured to determine a first signal point matrix according to a signal to be detected; a determining module 502, configured to determine a first feature map corresponding to the signal to be detected according to the first signal point matrix; and the analysis module 503 is configured to perform analysis according to the first feature map to obtain fault level information of the bearing to be tested, which corresponds to the first feature map.
In one embodiment, the analysis module 503 includes: the obtaining submodule 5031 is used for obtaining a deep groove ball bearing training sample, performing model training through the deep groove ball bearing training sample, and obtaining a first fault analysis model, wherein the first fault analysis model is used for performing fault analysis on the deep groove ball bearing; the obtaining submodule 5031 is further configured to obtain a rolling bearing training sample, train the first fault analysis model through the rolling bearing training sample, and obtain a second fault analysis model, where the second fault analysis model is used for performing fault analysis on the rolling bearing; the input sub-module 5032 is configured to input the first feature map into the second fault analysis model, and obtain fault level information of the bearing to be tested, which corresponds to the first feature map.
In an implementation manner, the obtaining sub-module 5031 is further configured to obtain a test sample, and test the second fault analysis model through the test sample to obtain a test result, where the test result at least includes an analysis accuracy corresponding to the fault level information; when the analysis accuracy rate meets a specified threshold value, determining the fault grade information of the bearing to be detected corresponding to the first characteristic map according to the second fault analysis model; and when the analysis accuracy does not meet the specified threshold value, acquiring a rolling bearing training sample, and retraining the second fault analysis model through the rolling bearing training sample.
In one possible embodiment, the acquisition submodule 5031 is further configured to acquire a training signal corresponding to a training bearing; determining a second signal point matrix according to the training signal; obtaining a second characteristic map corresponding to the training signal according to the second signal point matrix; acquiring fault grade information corresponding to each second feature map; and collecting the rolling bearing training samples according to the plurality of second characteristic maps and the fault grade information corresponding to the second characteristic maps.
In an implementation, the determining module 502 is further configured to randomly intercept a plurality of first signal points in the signal to be measured; a first signal point matrix is constructed from the Hankel matrix and the plurality of first signal points.
In one embodiment, the determining module 502 includes: the decomposition submodule 5021 is used for performing singular value decomposition on the first signal point matrix to obtain a plurality of effective singular values; the reconstruction submodule 5022 is used for reconstructing signals according to the effective singular values to obtain vibration signals; the transformation submodule 5023 is used for performing S transformation on the vibration signal to obtain a first feature map corresponding to the vibration signal.
In an implementation manner, the decomposition submodule 5021 is further configured to perform singular value decomposition on the first signal point matrix to obtain a first singular value matrix, where the first singular value matrix includes a plurality of singular values; determining a first number of valid singular values from a singular value curvature spectrum; and determining a first number of singular values in the first singular value matrix as valid singular values according to the sequence in the first singular value matrix.
In one embodiment, the remaining singular values in the first singular value matrix are determined as invalid singular values; correspondingly, the reconstruction submodule 5022 is also used for setting invalid singular values in the first singular value matrix to zero to obtain a second singular value matrix; and performing signal reconstruction according to the second singular value rectangle to obtain a vibration signal.
It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present disclosure may be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved, and the present disclosure is not limited herein.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present disclosure, "a plurality" means two or more unless specifically limited otherwise.
The above description is only for the specific embodiments of the present disclosure, but the scope of the present disclosure is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present disclosure, and all the changes or substitutions should be covered within the scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims.
Claims (10)
1. A method for failure analysis of a rolling bearing for a wind turbine, characterized in that it comprises:
acquiring a signal to be detected corresponding to a bearing to be detected;
determining a first signal point matrix according to the signal to be detected;
determining a first characteristic map corresponding to the signal to be detected according to the first signal point matrix;
and analyzing according to the first characteristic map to obtain the fault grade information of the bearing to be detected corresponding to the first characteristic map.
2. The method according to claim 1, wherein the analyzing according to the first feature map to obtain the fault grade information of the bearing to be tested corresponding to the first feature map comprises:
obtaining a deep groove ball bearing training sample, and performing model training through the deep groove ball bearing training sample to obtain a first fault analysis model, wherein the first fault analysis model is used for performing fault analysis on the deep groove ball bearing;
acquiring a rolling bearing training sample, and training the first fault analysis model through the rolling bearing training sample to obtain a second fault analysis model, wherein the second fault analysis model is used for carrying out fault analysis on the rolling bearing;
and inputting the first characteristic map into the second fault analysis model to obtain the fault grade information of the bearing to be tested corresponding to the first characteristic map.
3. The method of claim 2, wherein obtaining rolling bearing training samples through which the first fault analysis model is trained to obtain a second fault analysis model comprises:
obtaining a test sample, and testing the second fault analysis model through the test sample to obtain a test result, wherein the test result at least comprises an analysis accuracy rate corresponding to fault grade information;
when the analysis accuracy rate meets a specified threshold value, determining the fault grade information of the bearing to be detected corresponding to the first characteristic map according to the second fault analysis model;
and when the analysis accuracy does not meet a specified threshold value, acquiring the rolling bearing training sample, and retraining the second fault analysis model through the rolling bearing training sample.
4. The method of claim 2, wherein the obtaining rolling bearing training samples comprises:
acquiring a training signal corresponding to a training bearing;
determining a second signal point matrix according to the training signal;
obtaining a second feature map corresponding to the training signal according to the second signal point matrix;
acquiring fault grade information corresponding to each second feature map;
and assembling the rolling bearing training sample according to a plurality of second feature maps and fault level information corresponding to the second feature maps.
5. The method of claim 1, wherein determining a first signal point matrix from the signal under test comprises:
randomly intercepting a plurality of first signal points in a signal to be detected;
constructing the first signal point matrix from the hank's matrix and the plurality of first signal points.
6. The method of claim 1, wherein determining the first feature map corresponding to the signal under test from the first signal point matrix comprises:
singular value decomposition is carried out on the first signal point matrix to obtain a plurality of effective singular values;
performing signal reconstruction according to the effective singular value to obtain a vibration signal;
and carrying out S transformation on the vibration signal to obtain a first characteristic map corresponding to the vibration signal.
7. The method of claim 6, wherein the performing a singular value decomposition on the first signal point matrix to obtain a plurality of valid singular values comprises:
performing singular value decomposition on the first signal point matrix to obtain a first singular value matrix, wherein the first singular value matrix comprises a plurality of singular values;
determining a first number of valid singular values from a singular value curvature spectrum;
and determining a first number of singular values in the first singular value matrix as valid singular values according to the sequence in the first singular value matrix.
8. The method of claim 7, wherein the singular values remaining in the first matrix of singular values are determined to be invalid singular values;
correspondingly, the reconstructing the signal according to the effective singular value to obtain the vibration signal includes:
setting invalid singular values in the first singular value matrix to zero to obtain a second singular value matrix;
and performing signal reconstruction according to the second singular value rectangle to obtain a vibration signal.
9. A fault analysis device for a rolling bearing of a wind generator, characterized in that it comprises:
the acquisition module is used for acquiring a signal to be detected corresponding to the bearing to be detected;
the determining module is used for determining a first signal point matrix according to the signal to be detected;
the determining module is used for determining a first characteristic map corresponding to the signal to be detected according to the first signal point matrix;
and the analysis module is used for analyzing according to the first characteristic map to obtain the fault grade information of the bearing to be tested corresponding to the first characteristic map.
10. The fault analysis device of claim 9, wherein the analysis module comprises:
the acquisition submodule is used for acquiring a deep groove ball bearing training sample, performing model training through the deep groove ball bearing training sample and acquiring a first fault analysis model, wherein the first fault analysis model is used for performing fault analysis on the deep groove ball bearing;
the acquisition submodule is also used for acquiring a rolling bearing training sample, training the first fault analysis model through the rolling bearing training sample and acquiring a second fault analysis model, and the second fault analysis model is used for carrying out fault analysis on the rolling bearing;
and the input submodule is used for inputting the first characteristic map into the second fault analysis model to obtain the fault grade information of the bearing to be tested corresponding to the first characteristic map.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111294441.3A CN114088401A (en) | 2021-11-03 | 2021-11-03 | Fault analysis method and device for rolling bearing of wind driven generator |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111294441.3A CN114088401A (en) | 2021-11-03 | 2021-11-03 | Fault analysis method and device for rolling bearing of wind driven generator |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114088401A true CN114088401A (en) | 2022-02-25 |
Family
ID=80298722
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111294441.3A Pending CN114088401A (en) | 2021-11-03 | 2021-11-03 | Fault analysis method and device for rolling bearing of wind driven generator |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114088401A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090028455A1 (en) * | 2004-06-03 | 2009-01-29 | Japan Science And Technology Agency | High-speed high-accuracy matrix singular value decomposition method, program, and device |
| CN107037327A (en) * | 2016-10-09 | 2017-08-11 | 中国电力科学研究院 | Partial discharges fault judges feature extracting method and decision method |
| CN109977920A (en) * | 2019-04-11 | 2019-07-05 | 福州大学 | Fault Diagnosis of Hydro-generator Set method based on time-frequency spectrum and convolutional neural networks |
| CN110705525A (en) * | 2019-10-24 | 2020-01-17 | 北京建筑大学 | Method and device for diagnosing rolling bearing fault |
| CN110826017A (en) * | 2019-09-25 | 2020-02-21 | 中国地质大学（武汉） | Signal denoising method based on parameter optimization Hankel matrix and singular value decomposition |
| CN111094927A (en) * | 2017-09-26 | 2020-05-01 | 舍弗勒技术股份两合公司 | Bearing fault diagnosis method and device, readable storage medium and electronic equipment |
| CN112304611A (en) * | 2020-09-14 | 2021-02-02 | 江苏大学 | Deep learning-based bearing fault diagnosis method |
| CN113567131A (en) * | 2021-07-29 | 2021-10-29 | 沈阳建筑大学 | Bearing fault diagnosis method based on S transformation and miniature convolution neural network model |
-
2021
- 2021-11-03 CN CN202111294441.3A patent/CN114088401A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090028455A1 (en) * | 2004-06-03 | 2009-01-29 | Japan Science And Technology Agency | High-speed high-accuracy matrix singular value decomposition method, program, and device |
| CN107037327A (en) * | 2016-10-09 | 2017-08-11 | 中国电力科学研究院 | Partial discharges fault judges feature extracting method and decision method |
| CN111094927A (en) * | 2017-09-26 | 2020-05-01 | 舍弗勒技术股份两合公司 | Bearing fault diagnosis method and device, readable storage medium and electronic equipment |
| CN109977920A (en) * | 2019-04-11 | 2019-07-05 | 福州大学 | Fault Diagnosis of Hydro-generator Set method based on time-frequency spectrum and convolutional neural networks |
| CN110826017A (en) * | 2019-09-25 | 2020-02-21 | 中国地质大学（武汉） | Signal denoising method based on parameter optimization Hankel matrix and singular value decomposition |
| CN110705525A (en) * | 2019-10-24 | 2020-01-17 | 北京建筑大学 | Method and device for diagnosing rolling bearing fault |
| CN112304611A (en) * | 2020-09-14 | 2021-02-02 | 江苏大学 | Deep learning-based bearing fault diagnosis method |
| CN113567131A (en) * | 2021-07-29 | 2021-10-29 | 沈阳建筑大学 | Bearing fault diagnosis method based on S transformation and miniature convolution neural network model |
Non-Patent Citations (1)
| Title |
|---|
| 张梅军 等: "《EEMD方法及其在机械故障诊断中的应用》", vol. 1, 31 October 2015, 国防工业出版社, pages: 35 - 38 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| He et al. | Deep learning based approach for bearing fault diagnosis | |
| CN112036301B (en) | Driving motor fault diagnosis model construction method based on intra-class feature transfer learning and multi-source information fusion | |
| CN110110768B (en) | Rolling bearing fault diagnosis method based on parallel feature learning and multiple classifiers | |
| CN112257530B (en) | Rolling bearing fault diagnosis method based on blind signal separation and support vector machine | |
| CN111562108A (en) | Rolling bearing intelligent fault diagnosis method based on CNN and FCMC | |
| CN109187025A (en) | A kind of rolling bearing remaining life prediction technique of integrated KELM | |
| CN105760839A (en) | Bearing fault diagnosis method based on multi-feature manifold learning and support vector machine | |
| CN109033719A (en) | A kind of wind turbine Method for Bearing Fault Diagnosis | |
| CN114091504B (en) | Rotary machine small sample fault diagnosis method based on generation countermeasure network | |
| CN111898644B (en) | Intelligent identification method for health state of aerospace liquid engine under fault-free sample | |
| CN107451760A (en) | Based on when the limited Boltzmann machine of window sliding Fault Diagnosis of Roller Bearings | |
| CN113865872B (en) | Bearing fault diagnosis method based on wavelet packet reconstruction imaging and CNN | |
| CN111595541A (en) | Multi-dimensional structure damage identification method based on mass vibration transmissibility data convolutional neural network processing | |
| CN110608884A (en) | Rolling bearing state diagnosis method based on self-attention neural network | |
| CN112541524A (en) | BP-Adaboost multi-source information motor fault diagnosis method based on attention mechanism improvement | |
| CN109612728A (en) | Method for Bearing Fault Diagnosis based on vibration spectrogram and depth convolutional neural networks | |
| CN114997218A (en) | Recognition and detection method for polygonal abrasion of wheels of railway vehicle | |
| CN110455512A (en) | The mostly integrated method for diagnosing faults of rotating machinery based on depth self-encoding encoder DAE | |
| CN114861349B (en) | Rolling bearing RUL prediction method based on model migration and wiener process | |
| CN110222386A (en) | A kind of planetary gear degenerate state recognition methods | |
| Jianhui et al. | New fault diagnosis method for rolling bearing based on PCA | |
| CN112880726A (en) | Sensor fault diagnosis method and device based on variational modal decomposition sample entropy | |
| CN115935187B (en) | Nuclear sensitivity alignment network-based mechanical fault diagnosis method under variable working conditions | |
| CN117516939A (en) | Bearing cross-working condition fault detection method and system based on improved EfficientNetV2 | |
| Zhang et al. | Gearbox health condition identification by neuro-fuzzy ensemble |
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 |