CN108304927A - Bearing fault modality diagnostic method and system based on deep learning - Google Patents
Bearing fault modality diagnostic method and system based on deep learning Download PDFInfo
- Publication number
- CN108304927A CN108304927A CN201810072846.4A CN201810072846A CN108304927A CN 108304927 A CN108304927 A CN 108304927A CN 201810072846 A CN201810072846 A CN 201810072846A CN 108304927 A CN108304927 A CN 108304927A
- Authority
- CN
- China
- Prior art keywords
- bearing
- learning
- training
- bearing fault
- deep learning
- 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
Classifications
-
- 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
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The bearing fault modality diagnostic method and system based on deep learning that the invention discloses a kind of, wherein method includes the following steps：It is used as activation primitive by automatically adjusting learning rate, introducing noise, introducing linear amending unit, to improve depth belief network according to activation primitive；Generated data collection is obtained by migrating data collection and shot and long term memory network, to train improved depth belief network according to the data set of generated data collection spread training, and by training dataset, to obtain bearing failure diagnosis model；The vibration signal of bearing is acquired, and bearing fault mode is gone out according to the vibration signal of bearing and bearing failure diagnosis Model Diagnosis.This method combination semi-supervised learning and transfer learning algorithm improve diagnostic accuracy in the case that data volume is insufficient.
Description
Technical field
The present invention relates to hoisting equipment structure failure diagnostic techniques field, more particularly to a kind of bearings based on deep learning
Fault mode diagnostic method and system.
Background technology
With the development of social production, the growth of complicated production equipment and data scale, the equipment fault based on state
Diagnosis is effective method to improve production efficiency, is strengthened safety in production.Using machine learning and other intelligent means, fully dig
The intrinsic information for digging creation data realizes the important research direction of the fault diagnosis based on data.Machine learning has become in recent years
The mainstream of intelligent trouble diagnosis.Equipment fault diagnosis is substantially the classification problem of the equipment method of operation, is broadly divided into two steps：It is special
Sign extraction and classification.Typical health monitoring and fault diagnosis system generally include signal acquisition, feature extraction and fault identification
Module.
And the problem of for bearing failure diagnosis, it is generally the case that primary monitoring data is the vibration about transmission mechanism
Acceleration or vibration displacement.The performance of the characteristic largely affects the result of fault diagnosis.In traditional bearing failure
In diagnosis, due to being difficult to abundant faults information for time-domain signal, signal processing becomes the necessary condition of extraction fault signature.
And feature extraction is then difficult point and emphasis in signal processing.When peak value, kurtosis, the nargin factor, root-mean-square value, peak factor are
The common trait in domain, but the ability to express of fault message is insufficient, and Immunity Performance is limited, therefore diagnostic result is not ideal enough.Scholar
The method for proposing to convert a signal into frequency domain, by quick FFT, (Fast Fourier Transformation, Fourier become
Change) analyze amplitude spectrum and power spectrum.However, the Non-stationary vibration signal in actual scene does not often meet the stationarity of FFT
It is assumed that influence diagnostic result.On this basis, STFT (short-time Fourier transform, Short-time Fourier
Transformation), WT (wavelet transform, wavelet transformation), EMD (Empirical Mode Decomposition, Empirical Mode
Decompose) the methods of successively be suggested.New method for diagnosing faults can automatically extract fault signature, break away to expert diagnosis and answer
The dependence of miscellaneous signal processing.
With the rise of deep learning, the above problem knows about new way certainly.2006, Hinton et al. successes
It is extracted the characteristics of image using Autoencoder (AE), and proposes depth belief network.Overcome gradient decaying and part
Extreme-value problem.As deep learning flourishes.In computer vision field, LeNet, AlexNet, GoogleNet, Resnet
The mainstream of field of image recognition is had become Deng the model based on convolutional neural networks.In natural language processing field, RNN
(Recurrent neural Network, recurrent neural network), LSTM (Long Short-Term Memory, shot and long term note
Recall network), the model based on RNN such as GRU (gated recurrent unit, thresholding recurrent neural network) be sequence analysis
Also excellent achievement is achieved.However, deep learning in plant maintenance and the application of fault diagnosis field still in the exploratory stage,
Data, which are carried out, using depth digs worth explore.
Invention content
The present invention is directed to solve at least some of the technical problems in related technologies.
For this purpose, an object of the present invention is to provide a kind of bearing fault modality diagnostic method based on deep learning,
This method combination semi-supervised learning and transfer learning algorithm improve diagnostic accuracy in the case that data volume is insufficient.
It is another object of the present invention to propose a kind of bearing fault modality diagnostic system based on deep learning.
In order to achieve the above objectives, one aspect of the present invention embodiment proposes a kind of bearing fault pattern based on deep learning
Diagnostic method includes the following steps：It is used as activation letter by automatically adjusting learning rate, introducing noise, introducing linear amending unit
Number, to improve depth belief network according to the activation primitive；It is synthesized by migrating data collection and shot and long term memory network
Data set, to train improved depth according to the data set of the generated data collection spread training, and by the training dataset
Belief network is spent, to obtain bearing failure diagnosis model；The vibration signal of bearing is acquired, and according to the vibration signal of the bearing
Go out the bearing fault pattern with the bearing failure diagnosis Model Diagnosis.
The bearing fault modality diagnostic method based on deep learning of the embodiment of the present invention, can believe according to improved depth
The Method for Bearing Fault Diagnosis model for reading network obtains more accurate fault diagnosis precision, and expanding with other field
Property, the method extension of transfer learning can be used to can be used for the database of training pattern, solve the problems, such as data deficiencies, and
Other neck rates have expansibility, can directly be carried out to the vibration signal of bearing in conjunction with the method for deep learning and transfer learning
Processing, obtains its fault mode, has important practical performance and industrial value.
In addition, the bearing fault modality diagnostic method according to the above embodiment of the present invention based on deep learning can also have
There is following additional technical characteristic：
Further, in one embodiment of the invention, described by automatically adjusting learning rate, introducing noise, introducing
Linear amending unit is further comprised as activation primitive with improving depth belief network according to the activation primitive：According to weight
Under reconstructed error before and after structure error adjusts the learning rate, and the partial derivative limit for defining distribution is distributed with reconstruction model
The difference of the limit of partial derivative, and according to the positive and negative regularized learning algorithm efficiency of difference；In the input of limitation Boltzmann machine
The noise is introduced, minimizes output error after unsupervised training, with Optimal Parameters；Using the linear amending unit as
Activation primitive, output is unsaturated, to accelerate the training speed of network.
Further, in one embodiment of the invention, the limitation Boltzman machine be one two layers, by two parts
The undirected graphical model constituted one group of visible element and connects between the two layers it includes one group of binaryzation hidden unit
Weight matrix.
Further, in one embodiment of the invention, the improvement depth belief network passes through adaptive semi-supervised
The training meta structure of method improves performance, and each meta structure of adaptive regularized learning algorithm rate, input sample in the training process
This and noise enter the identical sample of member, after minimizing the error between output, with Optimal Parameters.
Further, in one embodiment of the invention, the transport number is obtained by knowledge migration or transfer learning
According to collection.
In order to achieve the above objectives, another aspect of the present invention embodiment proposes a kind of bearing fault mould based on deep learning
Formula diagnostic system, including：Sort module, for being used as by automatically adjusting learning rate, introducing noise, introducing linear amending unit
Activation primitive, to improve depth belief network according to the activation primitive；Data Migration module, for by migrating data collection with
Shot and long term memory network obtains generated data collection, with according to the data set of the generated data collection spread training, and by described
Training dataset trains improved depth belief network, to obtain bearing failure diagnosis model；Diagnostic module, for acquiring bearing
Vibration signal, and the bearing fault mould is gone out according to the vibration signal of the bearing and the bearing failure diagnosis Model Diagnosis
Formula.
The bearing fault modality diagnostic system based on deep learning of the embodiment of the present invention, can believe according to improved depth
The Method for Bearing Fault Diagnosis model for reading network obtains more accurate fault diagnosis precision, and expanding with other field
Property, the method extension of transfer learning can be used to can be used for the database of training pattern, solve the problems, such as data deficiencies, and
Other neck rates have expansibility, can directly be carried out to the vibration signal of bearing in conjunction with the method for deep learning and transfer learning
Processing, obtains its fault mode, has important practical performance and industrial value.
In addition, the bearing fault modality diagnostic system according to the above embodiment of the present invention based on deep learning can also have
There is following additional technical characteristic：
Further, in one embodiment of the invention, the sort module is additionally operable to according to before and after reconstructed error
Reconstructed error adjusts the learning rate, and defines the pole of the partial derivative limit and the lower partial derivative of reconstruction model distribution of distribution
The difference of limit, and according to the positive and negative regularized learning algorithm efficiency of difference, the noise is introduced in the input of limitation Boltzmann machine,
Output error is minimized after unsupervised training, with Optimal Parameters, and using the linear amending unit as activation primitive,
Output will not be saturated, and better performance be provided in zero, to accelerate the training speed of network.
Further, in one embodiment of the invention, the limitation Boltzman machine be one two layers, by two parts
The undirected graphical model constituted one group of visible element and connects between the two layers it includes one group of binaryzation hidden unit
Weight matrix.
Further, in one embodiment of the invention, the improvement depth belief network passes through adaptive semi-supervised
The training meta structure of method improves performance, and each meta structure of adaptive regularized learning algorithm rate, input sample in the training process
This and noise enter the identical sample of member, after minimizing the error between output, with Optimal Parameters.
Further, in one embodiment of the invention, the transport number is obtained by knowledge migration or transfer learning
According to collection.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partly become from the following description
Obviously, or practice through the invention is recognized.
Description of the drawings
Above-mentioned and/or additional aspect and advantage of the invention will become from the following description of the accompanying drawings of embodiments
Obviously and it is readily appreciated that, wherein：
Fig. 1 is the flow according to the bearing fault modality diagnostic method based on deep learning of one embodiment of the invention
Figure；
Fig. 2 is the linear amending unit figure line schematic diagram according to one embodiment of the invention；
Fig. 3 is the schematic diagram according to the improvement DBN model frame of one embodiment of the invention；
Fig. 4 is the schematic diagram that total frame is handled according to the improved DBN model of one embodiment of the invention；
Fig. 5 is the schematic diagram according to the DBN model of the introducing noise of one embodiment of the invention；
Fig. 6 is the transfer learning process chart according to one embodiment of the invention；
Fig. 7 is to be shown according to the structure of the bearing fault modality diagnostic system based on deep learning of one embodiment of the invention
It is intended to.
Specific implementation mode
The embodiment of the present invention is described below in detail, examples of the embodiments are shown in the accompanying drawings, wherein from beginning to end
Same or similar label indicates same or similar element or element with the same or similar functions.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to for explaining the present invention, and is not considered as limiting the invention.
The bearing fault modality diagnostic based on deep learning proposed according to embodiments of the present invention is described with reference to the accompanying drawings
Method and system describe the bearing fault pattern based on deep learning proposed according to embodiments of the present invention with reference to the accompanying drawings first
Diagnostic method.
Fig. 1 is the flow chart of the bearing fault modality diagnostic method based on deep learning of one embodiment of the invention.
As shown in Figure 1, the bearing fault modality diagnostic method based on deep learning of being somebody's turn to do includes the following steps：
In step S101, it is used as activation letter by automatically adjusting learning rate, introducing noise, introducing linear amending unit
Number, to improve depth belief network according to activation primitive.
It is understood that the embodiment of the present invention can be by automatically adjusting learning rate, introducing noise, the linear amendment of introducing
Unit improves traditional depth belief network as activation primitive, makes it have better accuracy and Generalization Capability.
In one embodiment of the invention, by automatically adjusting learning rate, introducing noise, the linear amending unit work of introducing
Further comprised with improving depth belief network according to activation primitive for activation primitive：It is missed according to the reconstruct before and after reconstructed error
Poor regularized learning algorithm rate, and the difference of the partial derivative limit and the limit of the lower partial derivative of reconstruction model distribution of distribution is defined,
And according to the positive and negative regularized learning algorithm efficiency of difference；Noise is introduced in the input of limitation Boltzmann machine, in unsupervised training
Output error is minimized afterwards, with Optimal Parameters；Using linear amending unit as activation primitive, output is unsaturated, to accelerate net
The training speed of network.
It is understood that the learning rate adjustment algorithm based on adaptive approach, is missed according to the reconstruct before and after reconstructed error
Poor regularized learning algorithm rate accelerates training with more excellent regularized learning algorithm rate；Based on the unsupervised training method that noise introduces, in RMB
Noise is introduced in the input of (Restricted Boltzmann Machine limit Boltzmann machine), after unsupervised training
Output error is minimized, with Optimal Parameters；Based on the Activiation method of linear amending unit, traditional Sigmoid is activated into letter
Number replaces with the rectification linear unit (Relu) in hidden layer, to improve the performance of network.
That is, according to the reconstructed error regularized learning algorithm rate before and after reconstructed error, the partial derivative function of distribution is defined
The difference of the limit and the limit of the lower partial derivative function of reconstruction model distribution, according to the positive and negative regularized learning algorithm rate of interpolation.In the defeated of RMB
Enter middle introducing noise, minimizes output error after unsupervised training, with Optimal Parameters.Optimized using back-propagation algorithm
RBM parameters.Using linear amending unit as activation primitive, output will not be saturated, and better performance is provided in zero, be helped
Accelerate the training speed of network.
Specifically, as shown in Fig. 2, the embodiment of the present invention replaces with traditional Sigmoid activation primitives in hidden layer
Rectification linear unit (Relu), to improve the performance of network.One side Relu will not be saturated, and when x is positive increasing, this contributes to
Deep learning multitiered network is improved, another aspect Relu obtains hidden layer more rarefaction representation and goes to zero when x is negative value, it is carried in zero
Better performance has been supplied, the training speed for accelerating network is helped.
Further, in one embodiment of the invention, limitation Boltzman machine be one two layers, be made of two parts
Undirected graphical model one group of visible element and connect the power between the two layers it includes one group of binaryzation hidden unit
Weight matrix.
It is a kind of to continuous data specifically, deeply convince that network is made of many RBMs (Boltzmann code machine)
Handle effective deep learning model.
Limitation Boltzman machine (member) be one two layers, bipartite, undirected graphical model, it is comprising one group
Binaryzation hidden unit h, one group of (or real value) visible element v and connect the weight matrix w between the two layers, energy function
It is defined as follows：
P (v, h)=exp (- E (v, h))/Z (θ), (1)
Energy function is：
It can be seen that the conditional probability distribution of unit and hiding unit is：
bjBayes to imply unit biases ciIt is biased for the Bayes of visible element, λ (x)=1/ (1+exp (x)).Such as
The visible unit real value of fruit, then can define energy function is：
From, as can be seen that in visible layer, hidden unit conditional sampling, vice versa in energy function.Particularly, binary
The unit of layer is independent Bernoulli random variable (using another layer as condition).If visible layer is real value, it is seen that unit
(using hidden layer as condition) is the Gaussian function with diagonal covariance.Therefore, the embodiment of the present invention can be by each layer
Unit (parallel) alternating sampling to execute effective gibbs sampler, and can refer to desired value the swashing as it of a unit
It is living.
Further, in one embodiment of the invention, it improves depth belief network and passes through adaptive semi-supervised method
Training meta structure improve performance, and in the training process each meta structure of adaptive regularized learning algorithm rate, input sample and
Noise enters the identical sample of member, after minimizing the error between output, with Optimal Parameters.
Specifically, when RBMs is piled into the belief network of a depth, the model of generation is made of many layers, really
Strength occur.As shown in figure 3, in DBN, each layer includes one group of binary system or real value unit.Two adjacent layers it
Between there are one complete connections to collect, but the unit in same layer does not connect.Propose a kind of trained depth belief network
Efficient algorithm carries out greedy training, using the activation of preceding layer as input to each layer (from most as low as highest).This program exists
In practice very effectively.It includes positive study from bottom to top, and the study of trimming backward from the top to the bottom, simultaneously
The details of Treated Base data set, and handle the attribute of data set, category information etc..It is level of abstraction in layer, from low layer
It is secondary to arrive high level, go deep into the substantive characteristics of mining data.
The study of positive storehouse is unsupervised learning, and parameter is that the maximum likelihood method obtained according to function 6 obtains, and probability
The maximum value of function is obtained by stochastic gradient rise method.However, be difficult to obtain unbiased sample in practical applications, therefore,
Preferably estimated usually using CD (Contrastive Divergence, to sdpecific dispersion) fast algorithms.First, according to
Function 3 calculates the conditional probability for implying layer unit, and the probability of hidden unit is determined using Gibbs model.Next, according to
Conditional probability of 4 computing unit of function in visual layers, and determine using Gibbs model the state of visual unit.Parameter
Renewal function is formula 7-9：
L(θ；V)=∏vL (θ | v)=∏vP (v), (6)
Δwij=ε (<vihj>P(h|v)-<vihj>recon), (7)
Δbi=ε (<vi>P(h|v)-<vi>recon), (8)
Δaj=ε (<hj>P(h|v)-<hj>recon), (9)
ε is learning rate,<·>P(h|v)It is that the lower local derviation of P (v | h) distributions is marginal,<·>reconIt is the local derviation under reconstruction model
Limit.
It needs to finely tune study backward after forward direction stacks RBM study, model parameter learns from unsupervised learning, as prison
The initial value that educational inspector practises, provides priori.From last of DBN (Deep Belief Network, depth belief network)
Layer, model parameter is finely adjusted according to from height to high label, and Softmax graders is finally used to carry out the information of extraction
Classification.Assuming that DBN networks include l RBM layers, initial sample is x, and it is u that then can define final output vectorl(x)：
ul(x)=1/ [1+exp) (bl+wlul-1(x))], (10)
Decision Classes are the classes for having most probable value, and loss function is：
α is learning rate, and parameter can be adjusted to the last one by the embodiment of the present invention from first RBM layers.
Further, as shown in figure 4, the difficulty of network depth network is embodied in the training process of network, training parameter
There is conclusive influence to model.The embodiment of the present invention proposes adaptive semi-supervised method training meta structure to improve the property of DBN
Can, and each meta structure of adaptive regularized learning algorithm rate and input sample and noise enter the identical sample of member in the training process
This, the Optimal Parameters for minimizing the error rear unsupervised training between output.In traditional DBN, network parameter is with depth
The propagation of the increase of degree, error will reduce, and lack anti-noise ability, therefore can improve the preceding performance to the stage of stacking.
First, according to the reconstructed error regularized learning algorithm rate before and after reconstructed error, the partial derivative function pole of distribution is defined
The difference of limit and the limit of the lower partial derivative function of reconstruction model distribution：
Ω=<·>P(h|v)-<·>recon, (15)
Learning rate is updated to：
λ is the regulation coefficient of learning rate, and parameter more new formula is:
Δwij t=(1- γ) εt(<vihj>t P(h|v)-<vihj>t recon)+γεt-1Δwij t-1, (17)
Δbi t=(1- γ) εt(<vi>t P(h|v)-<vi>t recon)+γεt-1Δbi t-1, (18)
Δaj t=(1- γ) εt(<hj>t P(h|v)-<hj>t recon)+γεt-1Δaj t-1, (19)
γ is the inertia coeffeicent of renewal process, can reduce the impact of learning process.
Secondly, as shown in figure 5, introducing noise in the input of RMB, output error is minimized after unsupervised training,
With Optimal Parameters.If input is x, the input for introducing noise isOutput is divided into yiWithThe two difference is defined asOptimize RBM parameters using back-propagation algorithm.
In step s 102, generated data collection is obtained by migrating data collection and shot and long term memory network, with according to synthesis
The data set of data set extension training, and improved depth belief network is trained by training dataset, to obtain bearing fault
Diagnostic model.
It is understood that the embodiment of the present invention solves the problems, such as data volume deficiency by migrating data collection, come using LSTM
The data set that generated data collection carrys out spread training is obtained, as training data.That is, the embodiment of the present invention can be based on moving
Learning theory method is moved, the data set that generated data collection carrys out spread training is obtained using LSTM, as training data, input
The model mentioned to training.
Further, in one embodiment of the invention, migrating data collection is obtained by knowledge migration or transfer learning.
Specifically, data mining and machine learning techniques many knowledge engineering fields achieve significantly at
Work(, including classification, recurrence and cluster.However, many machine learning methods could only be run under a common hypothesis it is good
It is good：Training and test data come from identical feature space and identical distribution.When changes in distribution, most of statistical models need
The training data newly collected is used to rebuild from the beginning.In many real-life programs, required training data and reconstruction
Model is costly or impossible.Demand and effort of the reduction to training data are a kind of modes of worth exploration.
In this case, transfer learning is desirable to the knowledge migration between task domain in other words.
Given original Problem Areas Ds and target problem domain Dt, D={ X, P (X) }, former problem Ts target problems Tt, T={ Y, P
(Y | X) }, transfer learning is suitable for following four scene：
The feature space of 1.XS ≠ XT, source domain and aiming field is different.
2.P (Xs) ≠ P (Xt) is different in the distribution of the marginal probability of source domain and aiming field.
The Label space of 3.YS ≠ YT, two tasks are different.
4.P (Ys | Xs) ≠ P (Yt | Xt), conditional probability distribution originating task and target are different.
Shift learning can be applied in many cases.For example, in real world, it is generally difficult to collect data or shortage
Data, and learn then to be easier to collect data from simulation.And shift learning can help model to adapt to new field, this regarding
It is meaningful in feel or text-processing.
As shown in fig. 6, due to lacking data herein, transfer learning and DBN are combined together by the embodiment of the present invention.Due to
Vibration signal follows certain rule, has very strong correlation with front and back data, can obtain generated data using LSTM
Collection carrys out the data set of spread training.Growth data collection belongs to same feature space with initial data, but does not follow identical limit
Probability distribution, although prediction model is better, gap is with regard to smaller.With the help of shift learning, can allow DBN networks more fully
Training.
In step s 103, the vibration signal of bearing is acquired, and according to the vibration signal of bearing and bearing failure diagnosis mould
Type is diagnosed to be bearing fault pattern.
The bearing fault modality diagnostic method based on deep learning proposed according to embodiments of the present invention, can be according to improvement
The Method for Bearing Fault Diagnosis model of depth belief network obtain more accurate fault diagnosis precision, and there is other field
Expansibility, the method extension of transfer learning can be used to can be used for the database of training pattern, solve data deficiencies
Problem, and have expansibility in other neck rates, it can be in conjunction with the method for deep learning and transfer learning, directly to the vibration of bearing
Signal is handled, and obtains its fault mode, has important practical performance and industrial value.
The bearing fault modality diagnostic based on deep learning for describing to propose according to embodiments of the present invention referring next to attached drawing
System.
Fig. 7 is the structural representation of the bearing fault modality diagnostic system based on deep learning of one embodiment of the invention
Figure.
As shown in fig. 7, the bearing fault modality diagnostic system 10 based on deep learning of being somebody's turn to do includes：Sort module 100, data
Transferring module 200 and diagnostic module 300.
Wherein, sort module 100 is used to be used as by automatically adjusting learning rate, introducing noise, introducing linear amending unit
Activation primitive, to improve depth belief network according to activation primitive.Data Migration module 200 is used to pass through migrating data collection and length
Short-term memory network obtains generated data collection, with according to the data set of generated data collection spread training, and passes through training dataset
The improved depth belief network of training, to obtain bearing failure diagnosis model.Diagnostic module 300 is used to acquire the vibration letter of bearing
Number, and bearing fault mode is gone out according to the vibration signal of bearing and bearing failure diagnosis Model Diagnosis.The embodiment of the present invention is
System 10 combines semi-supervised learning and transfer learning algorithms, improves diagnostic accuracy in the case that data volume is insufficient.
Further, in one embodiment of the invention, sort module 100 is additionally operable to according to the weight before and after reconstructed error
Structure error transfer factor learning rate, and define the difference of the partial derivative limit and the limit of the lower partial derivative of reconstruction model distribution of distribution
Value, and according to the positive and negative regularized learning algorithm efficiency of difference, noise is introduced in the input of limitation Boltzmann machine, in non-supervisory instruction
Output error is minimized after white silk, with Optimal Parameters, and using linear amending unit as activation primitive, output will not be saturated,
Better performance is provided in zero, to accelerate the training speed of network.
Further, in one embodiment of the invention, limitation Boltzman machine be one two layers, be made of two parts
Undirected graphical model one group of visible element and connect the power between the two layers it includes one group of binaryzation hidden unit
Weight matrix.
Further, in one embodiment of the invention, it improves depth belief network and passes through adaptive semi-supervised method
Training meta structure improve performance, and in the training process each meta structure of adaptive regularized learning algorithm rate, input sample and
Noise enters the identical sample of member, after minimizing the error between output, with Optimal Parameters.
Further, in one embodiment of the invention, migrating data collection is obtained by knowledge migration or transfer learning.
It should be noted that the aforementioned explanation to the bearing fault modality diagnostic embodiment of the method based on deep learning
It is also applied for the bearing fault modality diagnostic system based on deep learning of the embodiment, details are not described herein again.
The bearing fault modality diagnostic system based on deep learning proposed according to embodiments of the present invention, can be according to improvement
The Method for Bearing Fault Diagnosis model of depth belief network obtain more accurate fault diagnosis precision, and there is other field
Expansibility, the method extension of transfer learning can be used to can be used for the database of training pattern, solve data deficiencies
Problem, and have expansibility in other neck rates, it can be in conjunction with the method for deep learning and transfer learning, directly to the vibration of bearing
Signal is handled, and obtains its fault mode, has important practical performance and industrial value.
In the description of the present invention, it is to be understood that, term "center", " longitudinal direction ", " transverse direction ", " length ", " width ",
" thickness ", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom" "inner", "outside", " up time
The orientation or positional relationship of the instructions such as needle ", " counterclockwise ", " axial direction ", " radial direction ", " circumferential direction " be orientation based on ... shown in the drawings or
Position relationship is merely for convenience of description of the present invention and simplification of the description, and does not indicate or imply the indicated device or element must
There must be specific orientation, with specific azimuth configuration and operation, therefore be not considered as limiting the invention.
In addition, term " first ", " second " are used for description purposes only, it is not understood to indicate or imply relative importance
Or implicitly indicate the quantity of indicated technical characteristic.Define " first " as a result, the feature of " second " can be expressed or
Implicitly include at least one this feature.In the description of the present invention, the meaning of " plurality " is at least two, such as two, three
It is a etc., unless otherwise specifically defined.
In the present invention unless specifically defined or limited otherwise, term " installation ", " connected ", " connection ", " fixation " etc.
Term shall be understood in a broad sense, for example, it may be being fixedly connected, may be a detachable connection, or integral；Can be that machinery connects
It connects, can also be electrical connection；It can be directly connected, can also can be indirectly connected through an intermediary in two elements
The interaction relationship of the connection in portion or two elements, unless otherwise restricted clearly.For those of ordinary skill in the art
For, the specific meanings of the above terms in the present invention can be understood according to specific conditions.
In the present invention unless specifically defined or limited otherwise, fisrt feature can be with "above" or "below" second feature
It is that the first and second features are in direct contact or the first and second features pass through intermediary mediate contact.Moreover, fisrt feature exists
Second feature " on ", " top " and " above " but fisrt feature be directly above or diagonally above the second feature, or be merely representative of
Fisrt feature level height is higher than second feature.Fisrt feature second feature " under ", " lower section " and " below " can be
One feature is directly under or diagonally below the second feature, or is merely representative of fisrt feature level height and is less than second feature.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is included at least one embodiment or example of the invention.In the present specification, schematic expression of the above terms are not
It must be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be in office
It can be combined in any suitable manner in one or more embodiments or example.In addition, without conflicting with each other, the skill of this field
Art personnel can tie the feature of different embodiments or examples described in this specification and different embodiments or examples
It closes and combines.
Although the embodiments of the present invention has been shown and described above, it is to be understood that above-described embodiment is example
Property, it is not considered as limiting the invention, those skilled in the art within the scope of the invention can be to above-mentioned
Embodiment is changed, changes, replacing and modification.
Claims (10)
1. a kind of bearing fault modality diagnostic method based on deep learning, which is characterized in that include the following steps：
It is used as activation primitive by automatically adjusting learning rate, introducing noise, introducing linear amending unit, with according to the activation letter
Number improves depth belief network；
Generated data collection is obtained by migrating data collection and shot and long term memory network, with according to the generated data collection spread training
Data set, and improved depth belief network is trained by the training dataset, to obtain bearing failure diagnosis model；With
And
The vibration signal of bearing is acquired, and institute is gone out according to the vibration signal of the bearing and the bearing failure diagnosis Model Diagnosis
State bearing fault pattern.
2. the bearing fault modality diagnostic method according to claim 1 based on deep learning, which is characterized in that described logical
Automatic adjustment learning rate is crossed, noise is introduced, introduces linear amending unit as activation primitive, to be improved according to the activation primitive
Depth belief network, further comprises：
The learning rate is adjusted according to the reconstructed error before and after reconstructed error, and defines the partial derivative limit and the reconstruct of distribution
The difference of the limit of partial derivative under model profile, and according to the positive and negative regularized learning algorithm efficiency of difference；
The noise is introduced in the input of limitation Boltzmann machine, is minimized output error after unsupervised training, with excellent
Change parameter；
Using the linear amending unit as activation primitive, output is unsaturated, to accelerate the training speed of network.
3. the bearing fault modality diagnostic method according to claim 2 based on deep learning, which is characterized in that the limit
Boltzman machine processed be one two layers, bipartite undirected graphical model, it includes one group of binaryzation hidden unit,
One group of visible element and connect the weight matrix between the two layers.
4. the bearing fault modality diagnostic method according to claim 2 based on deep learning, which is characterized in that described to change
Performance is improved by the training meta structure of adaptive semi-supervised method into depth belief network, and adaptive in the training process
The each meta structure of regularized learning algorithm rate, input sample and noise enter the identical sample of member, after minimizing the error between output,
With Optimal Parameters.
5. the bearing fault modality diagnostic method according to claim 1 based on deep learning, which is characterized in that by knowing
Know migration or transfer learning obtains the migrating data collection.
6. a kind of bearing fault modality diagnostic system based on deep learning, which is characterized in that including：
Sort module, for being used as activation primitive by automatically adjusting learning rate, introducing noise, introducing linear amending unit, with
Depth belief network is improved according to the activation primitive；
Data Migration module, for obtaining generated data collection by migrating data collection and shot and long term memory network, with according to
The data set of generated data collection spread training, and improved depth belief network is trained by the training dataset, to obtain
Bearing failure diagnosis model；And
Diagnostic module, the vibration signal for acquiring bearing, and examined according to the vibration signal of the bearing and the bearing fault
Disconnected Model Diagnosis goes out the bearing fault pattern.
7. the bearing fault modality diagnostic system according to claim 6 based on deep learning, which is characterized in that described point
Generic module is additionally operable to adjust the learning rate according to the reconstructed error before and after reconstructed error, and defines the partial derivative pole of distribution
The difference of limit and the limit of the lower partial derivative of reconstruction model distribution, and according to the positive and negative regularized learning algorithm efficiency of difference, limiting
The noise is introduced in the input of Boltzmann machine, minimizes output error after unsupervised training, with Optimal Parameters, and
Using the linear amending unit as activation primitive, output will not be saturated, and better performance be provided in zero, to accelerate network
Training speed.
8. the bearing fault modality diagnostic system according to claim 7 based on deep learning, which is characterized in that the limit
Boltzman machine processed be one two layers, bipartite undirected graphical model, it includes one group of binaryzation hidden unit,
One group of visible element and connect the weight matrix between the two layers.
9. the bearing fault modality diagnostic system according to claim 7 based on deep learning, which is characterized in that described to change
Performance is improved by the training meta structure of adaptive semi-supervised method into depth belief network, and adaptive in the training process
The each meta structure of regularized learning algorithm rate, input sample and noise enter the identical sample of member, after minimizing the error between output,
With Optimal Parameters.
10. the bearing fault modality diagnostic system according to claim 6 based on deep learning, which is characterized in that pass through
Knowledge migration or transfer learning obtain the migrating data collection.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810072846.4A CN108304927A (en) | 2018-01-25 | 2018-01-25 | Bearing fault modality diagnostic method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810072846.4A CN108304927A (en) | 2018-01-25 | 2018-01-25 | Bearing fault modality diagnostic method and system based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108304927A true CN108304927A (en) | 2018-07-20 |
Family
ID=62866252
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810072846.4A Pending CN108304927A (en) | 2018-01-25 | 2018-01-25 | Bearing fault modality diagnostic method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108304927A (en) |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108875918A (en) * | 2018-08-14 | 2018-11-23 | 西安交通大学 | It is a kind of that diagnostic method is migrated based on the mechanical breakdown for being adapted to shared depth residual error network |
| CN108900358A (en) * | 2018-08-01 | 2018-11-27 | 重庆邮电大学 | Virtual network function dynamic migration method based on deepness belief network resource requirement prediction |
| CN109084980A (en) * | 2018-10-10 | 2018-12-25 | 北京交通大学 | Bearing fault prediction technique and device based on equalization segmentation |
| CN109102126A (en) * | 2018-08-30 | 2018-12-28 | 燕山大学 | One kind being based on depth migration learning theory line loss per unit prediction model |
| CN109142976A (en) * | 2018-09-10 | 2019-01-04 | 国网辽宁省电力有限公司电力科学研究院 | Cable fault examination method and device |
| CN109343505A (en) * | 2018-09-19 | 2019-02-15 | 太原科技大学 | Gear method for predicting residual useful life based on shot and long term memory network |
| CN109407007A (en) * | 2018-12-24 | 2019-03-01 | 广东省智能制造研究所 | A kind of battery charge state detection method and device |
| CN109975702A (en) * | 2019-03-22 | 2019-07-05 | 华南理工大学 | A kind of DC gear decelerating motor product examine method based on recirculating network disaggregated model |
| CN110162018A (en) * | 2019-05-31 | 2019-08-23 | 天津开发区精诺瀚海数据科技有限公司 | The increment type equipment fault diagnosis method that knowledge based distillation is shared with hidden layer |
| CN110276189A (en) * | 2019-06-27 | 2019-09-24 | 电子科技大学 | A kind of method for authenticating user identity based on gait information |
| CN110333076A (en) * | 2019-06-19 | 2019-10-15 | 电子科技大学 | Method for Bearing Fault Diagnosis based on CNN-Stacking |
| CN110398369A (en) * | 2019-08-15 | 2019-11-01 | 贵州大学 | A kind of Fault Diagnosis of Roller Bearings merged based on 1-DCNN and LSTM |
| CN110763997A (en) * | 2019-11-04 | 2020-02-07 | 华北电力大学（保定） | Early fault early warning method for synchronous motor stator |
| CN110823574A (en) * | 2019-09-30 | 2020-02-21 | 安徽富煌科技股份有限公司 | Fault diagnosis method based on semi-supervised learning deep countermeasure network |
| CN111044285A (en) * | 2019-11-22 | 2020-04-21 | 军事科学院系统工程研究院军用标准研究中心 | Method for diagnosing faults of mechanical equipment under complex conditions |
| RU2727470C2 (en) * | 2018-11-12 | 2020-07-21 | Ринат Габдулхакович Кудояров | Method for rapid diagnosis of metal processing machine modules |
| CN111579243A (en) * | 2020-06-17 | 2020-08-25 | 大连理工大学 | Rolling bearing intelligent diagnosis system based on deep migration learning |
| CN111595582A (en) * | 2020-05-26 | 2020-08-28 | 中国人民解放军陆军装甲兵学院 | Fault diagnosis method for rolling bearing |
| CN111783945A (en) * | 2020-06-22 | 2020-10-16 | 东北石油大学 | Long oil transmission pipeline leakage real-time detection system based on transfer learning LSTM |
| CN111982514A (en) * | 2020-08-12 | 2020-11-24 | 河北工业大学 | Bearing fault diagnosis method based on semi-supervised deep belief network |
| CN113033773A (en) * | 2021-03-03 | 2021-06-25 | 北京航空航天大学 | Efficient search method for layered multi-branch network structure for fault diagnosis of rotating machinery |
| CN113221277A (en) * | 2021-05-14 | 2021-08-06 | 西安交通大学 | Bearing performance degradation evaluation method and system based on digital twin model |
| CN113405799A (en) * | 2021-05-20 | 2021-09-17 | 新疆大学 | Bearing early fault detection method based on health state index construction and fault early warning limit self-learning |
| CN113505655A (en) * | 2021-06-17 | 2021-10-15 | 电子科技大学 | Bearing fault intelligent diagnosis method for digital twin system |
| CN113723649A (en) * | 2020-05-25 | 2021-11-30 | 中国石油化工股份有限公司 | Chemical process abnormity monitoring method based on semi-supervised model and model optimization method |
| CN114417924A (en) * | 2022-01-17 | 2022-04-29 | 辽宁石油化工大学 | Rolling bearing fault diagnosis method based on undirected graph adjacency matrix of mixed features |
| US11443137B2 (en) | 2019-07-31 | 2022-09-13 | Rohde & Schwarz Gmbh & Co. Kg | Method and apparatus for detecting signal features |
| US11544557B2 (en) | 2019-11-04 | 2023-01-03 | Cisco Technology, Inc. | IoT-based network architecture for detecting faults using vibration measurement data |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104616033A (en) * | 2015-02-13 | 2015-05-13 | 重庆大学 | Fault diagnosis method for rolling bearing based on deep learning and SVM (Support Vector Machine) |
| CN107192951A (en) * | 2017-07-31 | 2017-09-22 | 湖南科技大学 | A kind of micro- method for diagnosing faults of wind-driven generator three-phase rotor current |
| CN107229269A (en) * | 2017-05-26 | 2017-10-03 | 重庆工商大学 | A kind of wind-driven generator wheel-box method for diagnosing faults of depth belief network |
-
2018
- 2018-01-25 CN CN201810072846.4A patent/CN108304927A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104616033A (en) * | 2015-02-13 | 2015-05-13 | 重庆大学 | Fault diagnosis method for rolling bearing based on deep learning and SVM (Support Vector Machine) |
| CN107229269A (en) * | 2017-05-26 | 2017-10-03 | 重庆工商大学 | A kind of wind-driven generator wheel-box method for diagnosing faults of depth belief network |
| CN107192951A (en) * | 2017-07-31 | 2017-09-22 | 湖南科技大学 | A kind of micro- method for diagnosing faults of wind-driven generator three-phase rotor current |
Non-Patent Citations (2)
| Title |
|---|
| TIANQI YANG, SHUANGXI HUANG: "Fault Diagnosis Based on improved Deep Belief Network", 《2017 5TH INTERNATIONAL CONFERENCE ON ENTERPRISE SYSTEMS》 * |
| 李巍华 等: "基于深度信念网络的轴承故障分类识别", 《振动工程学报》 * |
Cited By (40)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108900358A (en) * | 2018-08-01 | 2018-11-27 | 重庆邮电大学 | Virtual network function dynamic migration method based on deepness belief network resource requirement prediction |
| CN108900358B (en) * | 2018-08-01 | 2021-05-04 | 重庆邮电大学 | Virtual network function dynamic migration method based on deep belief network resource demand prediction |
| CN108875918A (en) * | 2018-08-14 | 2018-11-23 | 西安交通大学 | It is a kind of that diagnostic method is migrated based on the mechanical breakdown for being adapted to shared depth residual error network |
| CN108875918B (en) * | 2018-08-14 | 2021-05-04 | 西安交通大学 | Mechanical fault migration diagnosis method based on adaptive shared depth residual error network |
| CN109102126B (en) * | 2018-08-30 | 2021-12-10 | 燕山大学 | Theoretical line loss rate prediction model based on deep migration learning |
| CN109102126A (en) * | 2018-08-30 | 2018-12-28 | 燕山大学 | One kind being based on depth migration learning theory line loss per unit prediction model |
| CN109142976A (en) * | 2018-09-10 | 2019-01-04 | 国网辽宁省电力有限公司电力科学研究院 | Cable fault examination method and device |
| CN109343505A (en) * | 2018-09-19 | 2019-02-15 | 太原科技大学 | Gear method for predicting residual useful life based on shot and long term memory network |
| CN109084980B (en) * | 2018-10-10 | 2019-11-05 | 北京交通大学 | Bearing fault prediction technique and device based on equalization segmentation |
| CN109084980A (en) * | 2018-10-10 | 2018-12-25 | 北京交通大学 | Bearing fault prediction technique and device based on equalization segmentation |
| RU2727470C2 (en) * | 2018-11-12 | 2020-07-21 | Ринат Габдулхакович Кудояров | Method for rapid diagnosis of metal processing machine modules |
| CN109407007A (en) * | 2018-12-24 | 2019-03-01 | 广东省智能制造研究所 | A kind of battery charge state detection method and device |
| CN109975702A (en) * | 2019-03-22 | 2019-07-05 | 华南理工大学 | A kind of DC gear decelerating motor product examine method based on recirculating network disaggregated model |
| CN109975702B (en) * | 2019-03-22 | 2021-08-10 | 华南理工大学 | Direct-current gear reduction motor quality inspection method based on circulation network classification model |
| CN110162018A (en) * | 2019-05-31 | 2019-08-23 | 天津开发区精诺瀚海数据科技有限公司 | The increment type equipment fault diagnosis method that knowledge based distillation is shared with hidden layer |
| CN110333076A (en) * | 2019-06-19 | 2019-10-15 | 电子科技大学 | Method for Bearing Fault Diagnosis based on CNN-Stacking |
| CN110333076B (en) * | 2019-06-19 | 2021-01-26 | 电子科技大学 | Bearing fault diagnosis method based on CNN-Stacking |
| CN110276189B (en) * | 2019-06-27 | 2022-02-11 | 电子科技大学 | User identity authentication method based on gait information |
| CN110276189A (en) * | 2019-06-27 | 2019-09-24 | 电子科技大学 | A kind of method for authenticating user identity based on gait information |
| US11443137B2 (en) | 2019-07-31 | 2022-09-13 | Rohde & Schwarz Gmbh & Co. Kg | Method and apparatus for detecting signal features |
| CN110398369A (en) * | 2019-08-15 | 2019-11-01 | 贵州大学 | A kind of Fault Diagnosis of Roller Bearings merged based on 1-DCNN and LSTM |
| CN110823574A (en) * | 2019-09-30 | 2020-02-21 | 安徽富煌科技股份有限公司 | Fault diagnosis method based on semi-supervised learning deep countermeasure network |
| CN110763997A (en) * | 2019-11-04 | 2020-02-07 | 华北电力大学（保定） | Early fault early warning method for synchronous motor stator |
| US11544557B2 (en) | 2019-11-04 | 2023-01-03 | Cisco Technology, Inc. | IoT-based network architecture for detecting faults using vibration measurement data |
| CN111044285A (en) * | 2019-11-22 | 2020-04-21 | 军事科学院系统工程研究院军用标准研究中心 | Method for diagnosing faults of mechanical equipment under complex conditions |
| CN113723649A (en) * | 2020-05-25 | 2021-11-30 | 中国石油化工股份有限公司 | Chemical process abnormity monitoring method based on semi-supervised model and model optimization method |
| CN111595582A (en) * | 2020-05-26 | 2020-08-28 | 中国人民解放军陆军装甲兵学院 | Fault diagnosis method for rolling bearing |
| CN111579243A (en) * | 2020-06-17 | 2020-08-25 | 大连理工大学 | Rolling bearing intelligent diagnosis system based on deep migration learning |
| CN111783945A (en) * | 2020-06-22 | 2020-10-16 | 东北石油大学 | Long oil transmission pipeline leakage real-time detection system based on transfer learning LSTM |
| CN111783945B (en) * | 2020-06-22 | 2022-08-16 | 东北石油大学 | Long oil pipeline leakage real-time detection system based on transfer learning LSTM |
| CN111982514A (en) * | 2020-08-12 | 2020-11-24 | 河北工业大学 | Bearing fault diagnosis method based on semi-supervised deep belief network |
| CN113033773A (en) * | 2021-03-03 | 2021-06-25 | 北京航空航天大学 | Efficient search method for layered multi-branch network structure for fault diagnosis of rotating machinery |
| CN113033773B (en) * | 2021-03-03 | 2023-01-06 | 北京航空航天大学 | Efficient search method for layered multi-branch network structure for fault diagnosis of rotating machinery |
| CN113221277A (en) * | 2021-05-14 | 2021-08-06 | 西安交通大学 | Bearing performance degradation evaluation method and system based on digital twin model |
| CN113221277B (en) * | 2021-05-14 | 2022-12-09 | 西安交通大学 | Bearing performance degradation evaluation method and system based on digital twin model |
| CN113405799A (en) * | 2021-05-20 | 2021-09-17 | 新疆大学 | Bearing early fault detection method based on health state index construction and fault early warning limit self-learning |
| CN113505655A (en) * | 2021-06-17 | 2021-10-15 | 电子科技大学 | Bearing fault intelligent diagnosis method for digital twin system |
| CN113505655B (en) * | 2021-06-17 | 2023-10-13 | 电子科技大学 | Intelligent bearing fault diagnosis method for digital twin system |
| CN114417924A (en) * | 2022-01-17 | 2022-04-29 | 辽宁石油化工大学 | Rolling bearing fault diagnosis method based on undirected graph adjacency matrix of mixed features |
| CN114417924B (en) * | 2022-01-17 | 2024-03-29 | 辽宁石油化工大学 | Rolling bearing fault diagnosis method based on undirected graph adjacent matrix of mixed features |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108304927A (en) | Bearing fault modality diagnostic method and system based on deep learning | |
| Xie et al. | An end-to-end model based on improved adaptive deep belief network and its application to bearing fault diagnosis | |
| Zhang et al. | A hybrid attention improved ResNet based fault diagnosis method of wind turbines gearbox | |
| Shao et al. | Highly accurate machine fault diagnosis using deep transfer learning | |
| Han et al. | An intelligent diagnosis framework for roller bearing fault under speed fluctuation condition | |
| CN103728551B (en) | A kind of analog-circuit fault diagnosis method based on cascade integrated classifier | |
| CN103440495B (en) | A kind of composite insulator hydrophobic grade automatic identifying method | |
| CN112149316A (en) | Aero-engine residual life prediction method based on improved CNN model | |
| CN108647614A (en) | The recognition methods of electrocardiogram beat classification and system | |
| CN104616033A (en) | Fault diagnosis method for rolling bearing based on deep learning and SVM (Support Vector Machine) | |
| CN109949929A (en) | A kind of assistant diagnosis system based on the extensive case history of deep learning | |
| CN109389171B (en) | Medical image classification method based on multi-granularity convolution noise reduction automatic encoder technology | |
| CN108805167A (en) | L aplace function constraint-based sparse depth confidence network image classification method | |
| Ma et al. | Cross-domain meta learning fault diagnosis based on multi-scale dilated convolution and adaptive relation module | |
| CN110110809A (en) | The construction method of fuzzy automata based on Machine Fault Diagnosis | |
| CN109033632A (en) | A kind of trend forecasting method based on depth quantum nerve network | |
| CN109406118A (en) | A kind of mechanical failure prediction method based on level convolutional neural networks | |
| Ma et al. | An unsupervised domain adaptation approach with enhanced transferability and discriminability for bearing fault diagnosis under few-shot samples | |
| CN106682089A (en) | RNNs-based method for automatic safety checking of short message | |
| CN116451150A (en) | Equipment fault diagnosis method based on semi-supervised small sample | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| CN115618267A (en) | Unsupervised domain self-adaption and entropy optimization equipment sensing diagnosis method and system | |
| Liang et al. | 1d convolutional neural networks for fault diagnosis of high-speed train bogie | |
| Jin et al. | Ordinal few-shot learning with applications to fault diagnosis of offshore wind turbines | |
| Zhao et al. | A capsnet-based fault diagnosis method for a digital twin of a wind turbine gearbox |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180720 |