CN109918752A - Mechanical failure diagnostic method, equipment and medium based on migration convolutional neural networks - Google Patents
Mechanical failure diagnostic method, equipment and medium based on migration convolutional neural networks Download PDFInfo
- Publication number
- CN109918752A CN109918752A CN201910142412.1A CN201910142412A CN109918752A CN 109918752 A CN109918752 A CN 109918752A CN 201910142412 A CN201910142412 A CN 201910142412A CN 109918752 A CN109918752 A CN 109918752A
- Authority
- CN
- China
- Prior art keywords
- convolutional neural
- neural networks
- network
- migration
- layer
- 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.)
- Granted
Links
Abstract
The invention discloses a kind of mechanical failure diagnostic method, equipment and media based on migration convolutional neural networks, comprising steps of (1) is obtained from multiple experimental assemblies first largely has the historical data of label, and do simple normalized；(2) one-dimensional convolutional neural networks are constructed, obtain initial transferable convolutional neural networks model by successively stacking；(3) network model pre-training is updated the transportable convolutional neural networks model that optimization obtains optimization to the weight of convolutional neural networks and biasing using source domain data set；(4) adaptive successively tuning migration is carried out to network weight and biasing on the depth network of pre-training using small sample set on aiming field data set, to obtain the transportable convolutional neural networks of tuning；(5) example sample to be predicted is input to the classification output that fault category is obtained in the neural network of tuning.The present invention can be realized the diagnosis of the migration between different operating conditions and different tests equipment, and significantly improve nicety of grading and training speed.
Description
Technical field
The invention belongs to machinery manufacturing technology fields, are related to a kind of mechanical breakdown Intelligent Diagnosis Technology, and in particular to a kind of
Mechanical breakdown intelligent diagnosing method, equipment and medium based on depth migration convolutional neural networks.
Background technique
The mechanical equipment function of modern production is more and more, and structure becomes increasingly complex, and the degree of automation also increasingly improves,
And equipment is normally operated in the environment of high load capacity, high corrosion and high operating rate, operating condition is also increasingly harsher.Therefore it develops
Effective fault detection and diagnosis technology, the health status of reliable monitoring mechanical equipment in real time, to extending service life of equipment,
Ensure that normal production and human life are of great significance safely, in recent years gradually by attention both domestic and external.
Intelligent diagnosing method based on data-driven, such as decision tree (DT), support vector machine (SVM), artificial neural network
(ANN) by establishing using a large amount of historical datas and Optimal Parameters, and then intelligent diagnostics model is established, be widely applied
In the fault diagnosis of rotating machinery, and achieve preferable effect.
Deep learning method such as stacks autoencoder network (SAE) in recent years, depth confidence network (DBN) and convolutional Neural net
Network (CNN) effectively can carry out approximation to complicated function by building multilayer depth structure, to have compared to shallow-layer learning algorithm
There are more powerful feature learning and ability to express and data-handling capacity, be increasingly becoming under mechanical big data background, carries out
The effective tool of intelligent trouble diagnosis.
However the intelligent diagnostics algorithm based on deep neural network, it usually needs a large amount of sample trains effective diagnosis
Model, on the one hand, when the algorithm of building is applied to new diagnostic task, since operating condition or monitoring device are different, lead
Cause the data distribution of acquisition inconsistent, thus model needs the re -training under the task.On the other hand, it is answered in actual engineering
In, usual mechanical equipment is operated under varying load variable speed operating condition, equipment different working condition acquirings data usually along with
The problems such as imbalanced training sets, sample missing, typical fault sample is insufficient.In this case, deep learning model is engaged in predecessor
It does well, may result in big performance loss under new diagnostic task.
Summary of the invention
Of the existing technology in order to solve the problems, such as, the present invention designs a kind of machine based on depth migration convolutional neural networks
Tool intelligent fault diagnosis method, equipment and medium, the method using convolutional neural networks in a large amount of historical datas by being gone to school
The diagnostic knowledge of habit establishes reflecting between source domain diagnostic knowledge and aiming field diagnostic knowledge using the layer-by-layer tuning of transfer learning
It penetrates, reduces data distribution difference between the two, and then obtain transportable diagnostic model, the method for proposition is more suitable for practical work
The intelligent diagnostics example of condition.
The present invention is realized using following technical scheme:
A kind of mechanical failure diagnostic method based on migration convolutional neural networks, comprising steps of
Step 1, data acquisition and calibration, design a variety of rig for testing malfunction tests, acquire the data under various working, press
Certain data point length interception obtains a large amount of sample set X, and the sample of known fault type is marked according to fault category
It is fixed, class label Y is set, and source domain data set is marked off according to operating condition and experimental facilitiesAnd mesh
Mark numeric field data collectionWherein m represents the number of samples in source domain data, and n represents sample in target numeric field data
Number；
The characteristics of step 2, the one-dimensional convolutional neural networks model of building for vibration signal are one-dimensional time-domain signal, uses
TensorFlow deep learning frame, builds one-dimensional convolutional neural networks, and the one-dimensional convolutional neural networks structure includes convolution
Layer (Conv), maximum pond layer (Max-pooling), batch normalization layer (BN layers), Dropout layers, full articulamentum and
Softmax classification layer, obtains initial transferable convolutional neural networks model by successively stacking；
Step 3, network model pre-training, for l layers of convolutional neural networks, the source domain data set constructed using step 1
{Xs,Ys, using back-propagation algorithm and gradient descent method to the weight and biasing { W of convolutional neural networks1,W2,…,WlCarry out
Update optimization, W1~WlThe 1st layer to l layers of convolutional neural networks of initial parameter, including network weight and biasing are respectively corresponded,
And corresponding hyper parameter, the final transportable convolutional neural networks model for obtaining optimization are chosen using grid-search algorithms；
Step 4, knowledge migration, using aiming field data set, to the convolutional neural networks model parameter { W of pre-training1,
W2,…,WlAccording to new diagnostic task, adaptively layer-by-layer tuning is migrated；
Step 5, fault diagnosis are input to moving for tuning migration to sample to be predicted for the optimal models of acquisition
It moves in convolutional neural networks, to obtain the probability value of every class failure, realizes that the classification output of fault category, last diagnostic go out machine
Tool fault type.
Further, in step 1, the source domain data set and aiming field data set derive from laboratory experiment platform or reality
Border industrial environment equipment monitoring platform.
Further, it in step 1, is further comprised the steps of: when acquiring the data under various working and data collected is carried out
The data set of acquisition, i.e., be segmented, and resulting sample set is normalized between [- 1,1], be suitable for by pretreatment
The input of network.
Further, convolutional neural networks constructed by step 2 normalize layer (BN using convolutional layer (Conv), batch
Layer), maximum pond layer (Max-pooling), Dropout layers stack gradually arrangement, using the linear amending unit activation letter of ReLU
Number, output use Softmax classifier.
Further, the step 3 specifically includes:
For l layers of convolutional neural networks, the parameter that network needs to optimize is { W1,W2,…,Wl},W1~WlIt is right respectively
The initial parameter for answering l layers of convolutional neural networks, the source domain data set { X constructed using step 1s,Ys, using Adam algorithm and
Cross entropy loss function carries out convolution operation to input；
It include the input batch { x of m sample to network again using batch standardization1,x2,…,xm, calculate final batch
Secondary normalized outputAre as follows:
The mean μ of each small batch is first found out according to equation (1)B, the side of batch is then further found out in equation (2)
Poor σB, normalized next is done to input in equation (3), obtains normalized valueWherein ε is smoothing factor, is prevented
It when variance is 0, exports as infinity, and introduces zooming parameter γ and translation in the final batch normalized output of equation (4)
Parameter beta further increases numerical value output stability, and the zooming parameter γ and translation parameters β parameter use in network training
Back-propagation algorithm is updated；
It is normalized by batch, network uses pondization processing further to reduce characteristic size, while obtaining certain put down
Invariant features are moved, and then Dropout technology is used to the Feature Mapping X of acquisition, add input nonlinearities, final output are as follows:
X=XBernoulli (p) (5)
P~Uniform (0,1) (6)
Wherein, Bernoulli (p) be Bernoulli Jacob's bi-distribution based on Probability p, p by equation (6) from be uniformly distributed with
Machine sampling obtains；Equation (5) adds input nonlinearities for carrying out zero setting to input signal at random, to depth network, forces network
Learn part and global useful feature from missing values, reduce the over-fitting of network, improves classification performance；
Feedforward by being successively input to output is propagated, and in the training stage, is calculated using backpropagation convolutional neural networks
Method and gradient descent method are updated optimization to weight and biasing, and choose corresponding hyper parameter using grid-search algorithms, finally
Obtain the transportable convolutional neural networks model of optimization.
Further, the optimization algorithm that adaptive layer-by-layer tuning moves use described in step 4 includes SGD algorithm.
Further, the step 4 specifically includes:
K layer parameter W before keeping first1~WkWeight and bias it is constant, Direct Transfer is applied in target network, in turn
Only to remaining network layer parameter Wk+1~WlWeight and biasing carry out tuning, by set the number of iterations and nicety of grading as
Stopping criterion, finally obtains optimal migration number of plies k for different classifications task, and migration formula can state are as follows:
Ts=F (Xs,Ys；Ws) (7)
Tt=F (Xt,Yt；Wt) (8)
Wherein, F () represents the layer-by-layer mapping function of convolutional neural networks, WsAnd WtRespectively represent former migration network and mesh
Mark the weight and offset parameter of domain network, TsAnd TtThe output of source domain and aiming field is respectively represented, the purpose of transfer learning is to lead to
Cross Ws→Wt, realize that source domain diagnostic knowledge is examined to aiming field using Direct Transfer weight and biasing or tuning weight and biasing
The multiplexing of disconnected knowledge.
Further, the probability that every kind of failure belongs to a different category is exported by Softmax classifier in the step 5
Value obtains i-th of sample to be tested using Softmax formula for the collected data of monitoring device, and corresponding failure is general
Rate value P (yi|xi) calculation formula are as follows:
Wherein, x in molecule itemiRepresent the input of one layer of ith feature mapping, yiRepresent the output probability of corresponding classification
Value, denominator term ∑j exp(xj) represent the summations of all sample output valves.
A kind of electronic equipment including memory, processor, stores the calculating that can be run on a memory and on a processor
Machine program when the processor runs described program, is realized as described in any item of the claim 1 to 8 based on migration convolution
The mechanical failure diagnostic method of neural network.
A kind of computer readable storage medium, is stored thereon with computer program, and the computer program is held by processor
It is realized when row such as the mechanical failure diagnostic method described in any item of the claim 1 to 8 based on migration convolutional neural networks.
Compared with prior art, the beneficial effects of the present invention are:
1, the method for the present invention uses one-dimensional convolutional neural networks, can learn complicated failure from original vibration signal
Discriminant information does not need additional characteristic extraction procedure, while constructing depth net using Dropout and BN batch normalization technology
Network model improves the intelligence and accuracy of diagnosis.
2, the method for the present invention can be by the historical data Learner diagnosis knowledge of laboratory data or industrial equipment, and can
It moves to different operating conditions, in the fault diagnosis example of different experiments equipment, is particularly suitable in fault sample deficiency, sample is uneven
Fault diagnosis engineer application in the case of weighing apparatus.
3, the present invention gives target network model one good initialization value by pre-training strategy, to be quickly and to have
The training deep neural network of effect provides a potential tool and technology, while reducing the risk of over-fitting.Furthermore should
Strategy can also further be saved the training time, and classification performance is improved.
4, from the perspective of migration models, the present invention is applicable not only to convolutional neural networks model (CNN), while may be used also
To expand to other deep learning models, such as depth confidence network (DBN), autoencoder network (SAE) and shot and long term are stacked
Memory network (LSTM) etc..
Detailed description of the invention
Fig. 1 is the Troubleshooting Flowchart of the method for the present invention.
Fig. 2 is the schematic network structure of the method for the present invention.
Fig. 3 is the method for the present invention depth convolutional neural networks migration scheme figure.
Fig. 4 is the method for the present invention figure compared with conventional exercises method loss function.
Fig. 5 is the method for the present invention figure compared with conventional depth learning method precision.
Specific embodiment
In order to keep technical solution of the present invention and purpose more clear, with reference to the accompanying drawing and specific implementation step
The present invention is described in detail, it should be understood that specific implementation step described herein is served only for preferably illustrating the present invention
Application, but technical characteristic involved by embodiments of the present invention is without being limited thereto.
A kind of mechanical failure diagnostic method based on migration convolutional neural networks, algorithm flow is as shown in Figure 1, the side
Method is learnt in historical data using convolutional neural networks, and is moved to target network to improve diagnosis performance.It should
Method comprising steps of
Step 1: the acquisition and calibration of data set design a variety of rig for testing malfunction tests, acquire the number under various working
According to, obtain a large amount of sample set X by certain data point length interception, by the sample of known fault type according to fault category into
Rower is fixed, sets class label Y, and mark off source domain data set according to operating condition and experimental facilitiesIt (uses
In the transportable convolutional neural networks of building), and aiming field data set(for testing the net proposed
Network performance), wherein m represents the number of samples in source domain data, and n represents number of samples in target numeric field data；
Step 2: one-dimensional convolutional neural networks model is constructed, the schematic diagram of model as shown in Fig. 2, be for vibration signal
The characteristics of one-dimensional time-domain signal, builds one-dimensional convolutional neural networks (CNN) model using TensorFlow deep learning frame,
The model includes 6 structure blocks, convolutional layer (Conv), maximum pond layer of each structure block by sequence stacked multilayer altogether
(Max-pooling), BN layers, Dropout layers, full articulamentum and Softmax classify layer, by successively stack obtain it is initial can
Migrate convolutional neural networks.
Step 3: network model pre-training, using the data set training depth convolutional neural networks of step 1 building, network
Input is the vibration signal after original normalization, and the parameter that network needs to optimize is { W1,W2,…,W6, it is each to respectively correspond network
The weight and offset parameter of a structure block.Network uses the linear amending unit activation primitive of ReLU, and used input vector is
2000 × 1 vector, but other input vector length are applicable to, output uses Softmax classifier, using Dropout
To apply regularization to network with BN batch normalization technology.
Dropout by way of inputting random zero setting to preceding layer, is added additionally in the training process to network
Noise jamming, to force e-learning to the feature of more robust property.It is in convolutional neural networks that BN batch, which normalizes technology,
Output pressure in training process to each input batch in each layer of neural network normalizes to standardized normal distribution, thus
So that each layer of output Feature Mapping all keeps identical data to be distributed, is conducive to accelerate network convergence rate, it is general to improve network
Change performance.
Step 4: on the basis of step 3, using aiming field data set to network parameter { W1,W2,…,W6Weight
Layer-by-layer tuning migration is carried out with biasing, as shown in figure 3, then keeping in structure block B1~B5 first such as one layer of neural network of tuning
Network parameter { W1,W2,…,W6Weight and bias constant, directly migrate in aiming field network, and parameter W6Weight
Small data set is used to be trained fine tuning to adapt to new task, by the way that suitable the number of iterations and classification essence is arranged with biasing
Spend criterion, to different diagnostic tasks can obtain respectively optimal migration number of plies k (migration models) its migrate formula can state are as follows:
Ts=F (Xs,Ys；Ws) (7)
Tt=F (Xt,Yt；Wt) (8)
Wherein, F () represents the layer-by-layer mapping function of convolutional neural networks, WsAnd WtRespectively represent former migration network and mesh
Mark the weight and offset parameter of domain network, TsAnd TtThe output of source domain and aiming field is respectively represented, the purpose of transfer learning is to lead to
Cross Ws→Wt, realize that source domain diagnostic knowledge is examined to aiming field using Direct Transfer weight and biasing or tuning weight and biasing
The multiplexing of disconnected knowledge.
Step 5: fault diagnosis is input to moving for tuning migration to sample to be predicted for the optimal models of acquisition
It moves in convolutional neural networks, to obtain the probability value of every class failure, realizes that the classification output of fault category, last diagnostic go out machine
Tool fault type.Using Softmax formula, i-th of sample to be tested, corresponding probability of malfunction value P (y are obtainedi|xi) calculating
Formula are as follows:
Wherein, x in molecule itemiRepresent the input of one layer of ith feature mapping, yiRepresent the output probability of corresponding classification
Value, denominator term ∑jexp(xj) represent the summations of all sample output valves.
Specifically, the source domain data set and aiming field data set derive from laboratory experiment platform or reality in step 1
Border industrial environment equipment monitoring platform, and it is not limited to specific data acquisition environment.
Specifically, further comprising the steps of: when acquiring the data under various working and being carried out to data collected in step 1
The data set of acquisition, i.e., be segmented, and resulting sample set is normalized between [- 1,1], be suitable for by pretreatment
The input of network.
Specifically, the step 3 specifically includes:
For 6 layers of convolutional neural networks, the parameter that network needs to optimize is { W1,W2,…,W6},W1~W6It is right respectively
The initial parameter for answering l layers of convolutional neural networks, the source domain data set { X constructed using step 1s,Ys, using Adam algorithm and
Cross entropy loss function carries out convolution operation to input；
It include the input batch { x of m sample to network again using batch standardization1,x2,…,xm, calculate final batch
Secondary normalized outputAre as follows:
The mean μ of each small batch is first found out according to equation (1)B, the side of batch is then further found out in equation (2)
Poor σB, normalized next is done to input in equation (3), obtains normalized valueWherein ε is smoothing factor, is prevented
When variance is 0, export as infinity.And zooming parameter γ and translation are introduced in the final batch normalized output of equation (4)
Parameter beta further increases numerical value output stability, and the zooming parameter γ and translation parameters β parameter use in network training
Back-propagation algorithm is updated；
It is normalized by batch, network uses pondization processing further to reduce characteristic size, while obtaining certain put down
Invariant features are moved, and then Dropout technology is used to the Feature Mapping X of acquisition, add input nonlinearities, final output are as follows:
X=XBernoulli (p) (5)
P~Uniform (0,1) (6)
Wherein, Bernoulli (p) be Bernoulli Jacob's bi-distribution based on Probability p, p by equation (6) from be uniformly distributed with
Machine sampling obtains；Equation (5) adds input nonlinearities for carrying out zero setting to input signal at random, to depth network, forces network
Learn part and global useful feature from missing values, reduce the over-fitting of network, improves classification performance；
Feedforward by being successively input to output is propagated, and in the training stage, is calculated using backpropagation convolutional neural networks
Method and gradient descent method are updated optimization to weight and biasing, and choose corresponding hyper parameter using grid-search algorithms, finally
Obtain the transportable convolutional neural networks model (TCNN) of optimization.
Below in conjunction with attached drawing and experiment case study, the present invention will be further described.
Experiment case study:
1, experimental data
Data set includes source domain data set and aiming field data set.Source domain data set by gearbox drive experimental bench data
Collection and Case Western Reserve University (CWRU) bearing data center's motor experimental bench data set composition.Gear-box data set shares ten kinds and is good for
Health state, including fault-free gear and bearing, the slight broken teeth of gear, moderate broken teeth, complete broken teeth, bearing outer ring 0.2mm failure,
Outer ring 2mm failure and combined failure type (including inner ring 0.2mm failure and three kinds of combined failures of gear and inner ring 2mm with
Gear moderate and complete broken teeth combined failure).Every kind of health status acquires two kinds of running speeds 500rpm and 750rpm.Every class is strong
Health situation has 500 training samples and 400 test samples under single revolving speed.Motor test data set shares ten kinds of healthy shapes
Condition, including fault-free bearing, bearing inner race failure, outer ring failure, rolling element failure, every kind of failure include three kinds different serious
Horizontal (0.007inch, 0.014inch and 0.021inch) acquires three kinds of revolving speeds 1772rpm, 1750rpm and 1730rpm. altogether
Every class health status has 500 training samples and 300 test samples under single revolving speed.Therefore source domain data set shares 20 kinds of events
Hinder type, altogether includes 25000 training samples and 17000 test samples, training sample is for training up one-dimensional convolutional Neural net
Network, test sample are used to verify the convolutional network model performance of construction.
Aiming field data set includes four kinds of diagnosis examples, and example C1 and C2 pick up from source domain gearbox drive testing stand, have
Ten kinds of health status, but collected from different rotating speeds operating condition (C1 acquisition from 1250rpm, C2 acquisition from 1000rpm and 1250rpm),
Every class health status has 200 training samples and 100 test samples under single revolving speed.Example C3 and C4 pick up from an armature spindle
Experiment porch is held, 5 kinds of health status, including fault-free bearing, bearing inner race 0.5mm failure, inner ring 2mm failure, outer ring are shared
0.5mm failure, outer ring 2mm failure.Every kind of health status acquires two kinds of revolving speeds 800rpm and 1100rpm.Wherein C3 example acquires
1100rpm, C4 example are acquired from 800rpm and 1100rpm.C1 and C2 example is used to simulate the network model migration under variable working condition,
C3 and C4 example is used to simulate the network model migration under different test platforms.Every class health status has under single revolving speed
200 training samples and 100 test samples.
2, method validation
The convolutional neural networks parameter of building is as shown in table 1:
Table 1: the convolutional neural networks parameter of building
The convolutional neural networks put forward include six structure block B1~B6, and each structure block is made of different layers,
Middle B1~B5 contains a convolutional layer (Convolution), and one BN layers, one Max-pooling layers, one
Dropout layers.B6 is by two convolutional layers (Convolution), two BN layers, two Dropout layers and a Softmax layers of group
At other specific network parameters are as shown in table 1.Convolutional neural networks are filled on source domain gear and motor data collection first
The pre-training divided, training algorithm are Adam algorithm, and training the number of iterations is 200 times.After network pre-training is good, by last structure
It builds block B6 and adaptive change is carried out according to the difference of diagnostic task, it is strong that the classification of output layer corresponds to the example for needing to diagnose
Health condition category, in network tuning transfer training, using SGD algorithm, learning rate is set as 0.01, and momentum coefficient is set as
0.97, total iteration 100 times.
(1) the method for the present invention (TCNN) is made comparisons with the convolutional neural networks model (CNN) not migrated
CNN and TCNN network structure having the same, CNN is weight and the biasing of random initializtion, then in aiming field
Training in example, and TCNN is the network model of the abundant pre-training on source domain data set according to the present invention, thus obtained
One good initial value, the then training on object instance.Fig. 4 is the loss curve graph of CNN and TCNN, as seen from the figure, CNN
Still there is biggish fluctuation in 100 iteration, and TCNN is in 20 the number of iterations, it is good that loss curve has just had reached
Convergence, only small fluctuation.Table 2 is corresponding nicety of grading and training time.TCNN-20 represents TCNN under 20 iteration
The result of acquisition.Compared to the CNN not migrated it can be seen from table, the migration neural network TCNN put forward is in four data
Respectively reach 99.9%, 99.3%, 97.9% on collection, 96.5% nicety of grading, while there is smaller standard deviation and more
Few training time.Even if special TCNN-20 still achieves the result of competitiveness in the case where training 20 times.
Table 2:TCNN is compared with the CNN not migrated
(2) method of the invention (TCNN) is made comparisons with other deep learning methods
The TCNN model and other deep learning methods put forward according to the present invention is compared (to be used including DNN, 2DCNN
Two-dimensional convolution neural network structure), WDCNN and TWDCNN (invention put forward is used for the transfer learning of WDCNN model).Its
Comparison result such as table 3 and Fig. 5.From the results, it was seen that comparing other methods, TCNN is obtained in nicety of grading and standard deviation
Best effect.Special TWDCNN is using the inventive method proposed, hence it is evident that the performance for improving WDCNN network is shown
Superiority of the invention.
Table 3:TCNN is compared with other deep learning methods
In order to realize above-described embodiment, the embodiment of the invention also provides a kind of electronic equipment, including memory, processing
Device stores the computer program that can be run on a memory and on a processor, when the processor runs described program, realizes
Mechanical failure diagnostic method as mentioned based on migration convolutional neural networks.
In order to realize above-described embodiment, the embodiment of the invention also provides a kind of computer readable storage mediums, deposit thereon
Computer program is contained, is realized when the computer program is executed by processor as mentioned based on migration convolutional neural networks
Mechanical failure diagnostic method.
The present invention is for data distribution is inconsistent in mechanical breakdown intelligent diagnostics, fault sample is insufficient, generalization ability of network energy
The problem of power difference, using step-by-step variable gear case and motor as research object, by utilizing convolutional neural networks in a large amount of history
Fault knowledge is moved to new diagnostic task using transfer learning method, proposes to be suitable for by the diagnostic knowledge learnt in data
The intelligent diagnosing method of actual condition.
It should be noted that, although detailed elaboration carried out to present invention implementation referring to example, but this field
Technical staff is readily appreciated that, made within the spirit and principles in the present invention illustrated in without departing from appended claims to appoint
What modification, replacement and improvement etc., should all be included in the protection scope of the present invention.
Claims (10)
1. a kind of mechanical failure diagnostic method based on migration convolutional neural networks, which is characterized in that comprising steps of
Step 1, data acquisition and calibration, design a variety of rig for testing malfunction tests, acquire the data under various working, by certain
The interception of data point length obtain a large amount of sample set X, the sample of known fault type is demarcated according to fault category, if
Determine class label Y, and source domain data set is marked off according to operating condition and experimental facilitiesAnd aiming field number
According to collectionWherein m represents the number of samples in source domain data, and n represents number of samples in target numeric field data；
The characteristics of step 2, the one-dimensional convolutional neural networks model of building for vibration signal are one-dimensional time-domain signal, uses
TensorFlow deep learning frame, builds one-dimensional convolutional neural networks, and the one-dimensional convolutional neural networks structure includes convolution
Layer, maximum pond layer, batch normalization layer, Dropout layers, full articulamentum and Softmax classification layer, by successively stacking acquisition
Initial transferable convolutional neural networks model；
Step 3, network model pre-training, for l layers of convolutional neural networks, the source domain data set { X constructed using step 1s,
Ys, using back-propagation algorithm and gradient descent method to the weight and biasing { W of convolutional neural networks1,W2,…,WlCarry out more
New optimization, W1~WlThe 1st layer to l layers of convolutional neural networks of initial parameter, including network weight and biasing are respectively corresponded, and
Corresponding hyper parameter, the final transportable convolutional neural networks model for obtaining optimization are chosen using grid-search algorithms；
Step 4, knowledge migration, using aiming field data set, to the convolutional neural networks model parameter { W of pre-training1,W2,…,
WlAccording to new diagnostic task, adaptively layer-by-layer tuning is migrated；
Step 5, fault diagnosis are input to the transportable volume of tuning migration to sample to be predicted for the optimal models of acquisition
In product neural network, to obtain the probability value of every class failure, realize that the classification output of fault category, last diagnostic go out mechanical event
Hinder type.
2. the mechanical failure diagnostic method according to claim 1 based on migration convolutional neural networks, which is characterized in that step
In rapid 1, the source domain data set and aiming field data set are monitored from laboratory experiment platform or actual industrial environmental unit
Platform.
3. the mechanical failure diagnostic method according to claim 1 based on migration convolutional neural networks, which is characterized in that step
In rapid 1, is further comprised the steps of: when acquiring the data under various working and data collected are pre-processed, i.e., by the number of acquisition
It is segmented according to collection, and resulting sample set is normalized between [- 1,1], the input suitable for network.
4. the mechanical failure diagnostic method according to claim 1 based on migration convolutional neural networks, which is characterized in that step
Convolutional neural networks constructed by rapid 2 are using convolutional layer, batch normalization layer, maximum pond layer, the Dropout layers of row of stacking gradually
Column, using the linear amending unit activation primitive of ReLU, output uses Softmax classifier.
5. the mechanical failure diagnostic method according to claim 1 based on migration convolutional neural networks, which is characterized in that institute
The step 3 stated specifically includes:
For l layers of convolutional neural networks, the parameter that network needs to optimize is { W1,W2,…,Wl},W1~WlRespectively correspond volume
The initial parameter of l layers of neural network of product, the source domain data set { X constructed using step 1s,Ys, using Adam algorithm and intersection
Entropy loss function carries out convolution operation to input；
It include the input batch { x of m sample to network again using batch standardization1,x2,…,xm, it calculates final batch and returns
One changes outputAre as follows:
The mean μ of each small batch is first found out according to equation (1)B, the variances sigma of batch is then further found out in equation (2)B,
Next normalized is done to input in equation (3), obtains normalized valueWherein ε is smoothing factor, prevents variance
It when being 0, exports as infinity, and introduces zooming parameter γ and translation parameters β in the final batch normalized output of equation (4)
Further increase numerical value output stability, the zooming parameter γ and translation parameters β parameter are in network training using reversed
Propagation algorithm is updated；
It is normalized by batch, network further uses pondization processing to reduce characteristic size, while obtaining certain translation not
Become feature, and then Dropout technology used to the Feature Mapping X of acquisition, adds input nonlinearities, final output are as follows:
X=XBernoulli (p) (5)
P~Uniform (0,1) (6)
Wherein, Bernoulli (p) is Bernoulli Jacob's bi-distribution based on Probability p, and p is adopted at random by equation (6) from being uniformly distributed
Sample obtains；Equation (5) at random to input signal carry out zero setting, to depth network add input nonlinearities, force network from lack
Learn part and global useful feature in mistake value, reduce the over-fitting of network, improves classification performance；
Feedforward by being successively input to output is propagated, in the training stage, to convolutional neural networks using back-propagation algorithm and
Gradient descent method is updated optimization to weight and biasing, and chooses corresponding hyper parameter using grid-search algorithms, final to obtain
The transportable convolutional neural networks model of optimization.
6. the mechanical failure diagnostic method according to claim 1 based on migration convolutional neural networks, which is characterized in that step
The optimization algorithm that adaptive layer-by-layer tuning moves use described in rapid 4 includes SGD algorithm.
7. the mechanical failure diagnostic method according to claim 6 based on migration convolutional neural networks, which is characterized in that institute
Step 4 is stated to specifically include:
K layer parameter W before keeping first1~WkWeight and bias it is constant, Direct Transfer is applied in target network, and then only right
Remaining network layer parameter Wk+1~WlWeight and biasing carry out tuning, by set the number of iterations and nicety of grading as stopping
Criterion, finally obtains optimal migration number of plies k for different classifications task, and migration formula can state are as follows:
Ts=F (Xs,Ys；Ws) (7)
Tt=F (Xt,Yt；Wt) (8)
Wherein, F () represents the layer-by-layer mapping function of convolutional neural networks, WsAnd WtRespectively represent former migration network and aiming field
The weight and offset parameter of network, TsAnd TtThe output of source domain and aiming field is respectively represented, the purpose of transfer learning is to pass through Ws
→Wt, realize that source domain diagnostic knowledge is known to aiming field diagnosis using Direct Transfer weight and biasing or tuning weight and biasing
The multiplexing of knowledge.
8. the mechanical failure diagnostic method according to claim 1 based on migration convolutional neural networks, which is characterized in that institute
It states in step 5 and the probability value that every kind of failure belongs to a different category is exported by Softmax classifier, collected for monitoring device
Data i-th of sample to be tested, corresponding probability of malfunction value P (y are obtained using Softmax formulai|xi) calculation formula
Are as follows:
Wherein, x in molecule itemiRepresent the input of one layer of ith feature mapping, yiThe output probability value of corresponding classification is represented, point
Female item ∑jexp(xj) represent the summations of all sample output valves.
9. a kind of electronic equipment, it is characterised in that: on a memory and can be on a processor including memory, processor, storage
The computer program of operation when the processor runs described program, realizes such as base described in any item of the claim 1 to 8
In the mechanical failure diagnostic method of migration convolutional neural networks.
10. a kind of computer readable storage medium, is stored thereon with computer program, it is characterised in that: the computer program
It is realized when being executed by processor such as the mechanical breakdown described in any item of the claim 1 to 8 based on migration convolutional neural networks
Diagnostic method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910142412.1A CN109918752B (en) | 2019-02-26 | 2019-02-26 | Mechanical fault diagnosis method, equipment and medium based on migration convolution neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910142412.1A CN109918752B (en) | 2019-02-26 | 2019-02-26 | Mechanical fault diagnosis method, equipment and medium based on migration convolution neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109918752A true CN109918752A (en) | 2019-06-21 |
| CN109918752B CN109918752B (en) | 2022-12-16 |
Family
ID=66962382
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910142412.1A Active CN109918752B (en) | 2019-02-26 | 2019-02-26 | Mechanical fault diagnosis method, equipment and medium based on migration convolution neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109918752B (en) |
Cited By (82)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110225453A (en) * | 2019-06-24 | 2019-09-10 | 鲸数科技（北京）有限公司 | Mobile terminal locating method, device, electronic equipment and storage medium |
| CN110378431A (en) * | 2019-07-24 | 2019-10-25 | 中国人民解放军国防科技大学 | Convolutional neural network-based supersonic combustion chamber combustion mode detection method |
| CN110464611A (en) * | 2019-07-23 | 2019-11-19 | 苏州国科视清医疗科技有限公司 | A kind of digitlization amblyopia enhancing training device and system and its related algorithm |
| CN110502991A (en) * | 2019-07-18 | 2019-11-26 | 武汉理工大学 | Internal combustion engine health monitor method and system based on random convolutional neural networks structure |
| CN110533070A (en) * | 2019-07-26 | 2019-12-03 | 西安交通大学 | A kind of mechanical breakdown intelligent diagnosing method based on migration original shape network under small sample |
| CN110647923A (en) * | 2019-09-04 | 2020-01-03 | 西安交通大学 | Variable working condition mechanical fault intelligent diagnosis method based on self-learning under small sample |
| CN110705029A (en) * | 2019-09-05 | 2020-01-17 | 西安交通大学 | Flow field prediction method of oscillating flapping wing energy acquisition system based on transfer learning |
| CN110705456A (en) * | 2019-09-29 | 2020-01-17 | 华南理工大学 | Micro motor abnormity detection method based on transfer learning |
| CN110726813A (en) * | 2019-10-12 | 2020-01-24 | 浙江大学 | Electronic nose prediction method based on double-layer integrated neural network |
| CN110866365A (en) * | 2019-11-22 | 2020-03-06 | 北京航空航天大学 | Mechanical equipment intelligent fault diagnosis method based on partial migration convolutional network |
| CN110995475A (en) * | 2019-11-20 | 2020-04-10 | 国网湖北省电力有限公司信息通信公司 | Power communication network fault detection method based on transfer learning |
| CN110991295A (en) * | 2019-11-26 | 2020-04-10 | 电子科技大学 | Self-adaptive fault diagnosis method based on one-dimensional convolutional neural network |
| CN111007035A (en) * | 2019-12-12 | 2020-04-14 | 重庆科技学院 | Method, system and device for detecting concentration of high-temperature gas in secondary combustion chamber based on convolutional neural network and storage medium |
| CN111027678A (en) * | 2019-12-04 | 2020-04-17 | 湃方科技(北京)有限责任公司 | Data migration method and device |
| CN111178378A (en) * | 2019-11-07 | 2020-05-19 | 腾讯科技（深圳）有限公司 | Equipment fault prediction method and device, electronic equipment and storage medium |
| CN111191693A (en) * | 2019-12-18 | 2020-05-22 | 广西电网有限责任公司电力科学研究院 | Method for identifying thermal fault state of high-voltage switch cabinet based on convolutional neural network |
| CN111275164A (en) * | 2020-01-13 | 2020-06-12 | 浙江大学 | Underwater robot propulsion system fault diagnosis method |
| CN111291814A (en) * | 2020-02-15 | 2020-06-16 | 河北工业大学 | Crack identification algorithm based on convolution neural network and information entropy data fusion strategy |
| CN111442926A (en) * | 2020-01-11 | 2020-07-24 | 哈尔滨理工大学 | Fault diagnosis method for rolling bearings of different models under variable load based on deep characteristic migration |
| CN111476299A (en) * | 2020-04-07 | 2020-07-31 | 国家电网有限公司华东分部 | Improved convolutional neural network and power grid intelligent alarm system based on same |
| CN111497868A (en) * | 2020-04-09 | 2020-08-07 | 余姚市浙江大学机器人研究中心 | Automobile sensor fault classification method based on BN-L STM network |
| CN111562110A (en) * | 2020-05-25 | 2020-08-21 | 南京理工大学 | Fault diagnosis model based on convolutional neural network and cross-component fault diagnosis method |
| CN111582477A (en) * | 2020-05-09 | 2020-08-25 | 北京百度网讯科技有限公司 | Training method and device of neural network model |
| CN111582396A (en) * | 2020-05-13 | 2020-08-25 | 江南大学 | Fault diagnosis method based on improved convolutional neural network |
| CN111598161A (en) * | 2020-05-14 | 2020-08-28 | 哈尔滨工业大学（威海） | Engine gas circuit state diagnosis system based on CNN transfer learning |
| CN111638465A (en) * | 2020-05-29 | 2020-09-08 | 浙大宁波理工学院 | Lithium battery health state estimation method based on convolutional neural network and transfer learning |
| CN111651937A (en) * | 2020-06-03 | 2020-09-11 | 苏州大学 | Method for diagnosing similar self-adaptive bearing fault under variable working conditions |
| CN111667000A (en) * | 2020-06-05 | 2020-09-15 | 成都理工大学 | Earthquake early warning method of adaptive field deep neural network |
| CN111680788A (en) * | 2020-07-14 | 2020-09-18 | 同济大学 | Equipment fault diagnosis method based on deep learning |
| CN111751650A (en) * | 2020-07-06 | 2020-10-09 | 重庆大学 | Non-invasive household electric equipment on-line monitoring system and fault identification method |
| CN111766513A (en) * | 2020-06-18 | 2020-10-13 | 电子科技大学 | Capsule network-based variable-working-condition multi-fault diagnosis method for three-phase induction motor |
| CN111814396A (en) * | 2020-07-02 | 2020-10-23 | 重庆大学 | Centrifugal fan fault early warning method based on transfer learning |
| CN111881627A (en) * | 2020-08-05 | 2020-11-03 | 哈尔滨工程大学 | Nuclear power device fault diagnosis method and system |
| CN111899905A (en) * | 2020-08-05 | 2020-11-06 | 哈尔滨工程大学 | Fault diagnosis method and system based on nuclear power device |
| CN111897310A (en) * | 2020-07-24 | 2020-11-06 | 华中科技大学 | Industrial process fault classification method and system based on one-dimensional multi-head convolutional network |
| CN111931826A (en) * | 2020-07-20 | 2020-11-13 | 华南理工大学 | Rolling bearing fault diagnosis method and system based on multi-scale convolution migration model |
| CN112036496A (en) * | 2020-09-02 | 2020-12-04 | 哈尔滨工程大学 | Nuclear power device fault diagnosis method and system |
| CN112051062A (en) * | 2020-08-10 | 2020-12-08 | 华中科技大学 | Fault diagnosis method, device, equipment and storage medium for rolling bearing |
| CN112052628A (en) * | 2020-09-06 | 2020-12-08 | 上海电力大学 | Gas turbine gas circuit fault prediction and diagnosis method based on knowledge and data |
| CN112101220A (en) * | 2020-09-15 | 2020-12-18 | 哈尔滨理工大学 | Rolling bearing service life prediction method based on unsupervised model parameter migration |
| CN112113755A (en) * | 2020-09-02 | 2020-12-22 | 西安交通大学 | Mechanical fault intelligent diagnosis method based on deep convolution-kurtosis neural network |
| CN112146879A (en) * | 2020-08-21 | 2020-12-29 | 江苏大学 | Rolling bearing fault intelligent diagnosis method and system |
| CN112149866A (en) * | 2020-08-17 | 2020-12-29 | 江苏大学 | Intelligent manufacturing workshop anomaly prediction and control method based on edge cloud cooperation |
| CN112152201A (en) * | 2020-08-25 | 2020-12-29 | 国网山东综合能源服务有限公司 | Electricity load prediction method and system based on convolution length time memory neural network |
| CN112163640A (en) * | 2020-10-30 | 2021-01-01 | 北京信息科技大学 | Planetary gearbox fault diagnosis method and system based on deep convolution migration learning |
| CN112161784A (en) * | 2020-09-07 | 2021-01-01 | 华南理工大学 | Mechanical fault diagnosis method based on multi-sensor information fusion migration network |
| WO2021022852A1 (en) * | 2019-08-02 | 2021-02-11 | 华为技术有限公司 | Access request processing method, apparatus, device, and storage medium |
| CN112444748A (en) * | 2020-10-12 | 2021-03-05 | 武汉蔚来能源有限公司 | Battery abnormality detection method, battery abnormality detection device, electronic apparatus, and storage medium |
| CN112465030A (en) * | 2020-11-28 | 2021-03-09 | 河南大学 | Multi-source heterogeneous information fusion fault diagnosis method based on two-stage transfer learning |
| CN112541511A (en) * | 2019-09-20 | 2021-03-23 | 宫文峰 | Multi-channel time series data fault diagnosis method based on convolutional neural network |
| CN112667791A (en) * | 2020-12-23 | 2021-04-16 | 深圳壹账通智能科技有限公司 | Latent event prediction method, device, equipment and storage medium |
| CN112766200A (en) * | 2021-01-26 | 2021-05-07 | 福州大学 | Planetary gear box small sample fault diagnosis method based on one-dimensional convolutional neural network |
| CN112763214A (en) * | 2020-12-31 | 2021-05-07 | 南京信息工程大学 | Rolling bearing fault diagnosis method based on multi-label zero-sample learning |
| CN112836896A (en) * | 2021-03-03 | 2021-05-25 | 西门子工厂自动化工程有限公司 | Method for maintaining equipment and system for maintaining equipment |
| CN112884070A (en) * | 2021-03-17 | 2021-06-01 | 云南电网有限责任公司电力科学研究院 | High-voltage switch fault diagnosis method based on transfer learning |
| CN112926728A (en) * | 2021-03-22 | 2021-06-08 | 南京理工大学 | Small sample turn-to-turn short circuit fault diagnosis method for permanent magnet synchronous motor |
| CN112989702A (en) * | 2021-03-25 | 2021-06-18 | 河北工业大学 | Self-learning method for equipment performance analysis and prediction |
| US20210203900A1 (en) * | 2020-04-17 | 2021-07-01 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Image processing method and apparatus, electronic device and computer-readable storage medium |
| CN113063595A (en) * | 2021-03-27 | 2021-07-02 | 温州大学 | Full convolution neural network rolling bearing fault identification method based on one-dimensional convolution |
| CN113203566A (en) * | 2021-04-06 | 2021-08-03 | 上海吞山智能科技有限公司 | Motor bearing fault diagnosis method based on one-dimensional data enhancement and CNN |
| CN113343633A (en) * | 2021-06-10 | 2021-09-03 | 上海交通大学 | Thermal runaway fault classification and risk prediction method and system for power lithium battery |
| CN113392881A (en) * | 2021-05-27 | 2021-09-14 | 重庆大学 | Rotary machine fault diagnosis method based on transfer learning |
| CN113469219A (en) * | 2021-06-08 | 2021-10-01 | 中国地质大学（武汉） | Rotary machine fault diagnosis method under complex working condition based on element transfer learning |
| CN113469230A (en) * | 2021-06-17 | 2021-10-01 | 北京信息科技大学 | Method, system and medium for diagnosing deep migration fault of rotor system |
| CN113468805A (en) * | 2021-06-21 | 2021-10-01 | 国网黑龙江省电力有限公司电力科学研究院 | Method for evaluating reliability of sensor in extremely cold environment, electronic device and storage medium |
| CN113541985A (en) * | 2020-04-14 | 2021-10-22 | 中国移动通信集团浙江有限公司 | Internet of things fault diagnosis method, training method of model and related device |
| CN113627518A (en) * | 2021-08-07 | 2021-11-09 | 福州大学 | Method for realizing multichannel convolution-recurrent neural network electroencephalogram emotion recognition model by utilizing transfer learning |
| CN113642591A (en) * | 2020-04-27 | 2021-11-12 | 中国科学院声学研究所 | Migration learning-based multi-beam seabed sediment layer type estimation method and system |
| CN113867307A (en) * | 2021-08-20 | 2021-12-31 | 天津大学 | Spacecraft intelligent fault diagnosis method based on deep neural network |
| CN114057053A (en) * | 2022-01-18 | 2022-02-18 | 杭州浅水数字技术有限公司 | Method for monitoring fatigue degree of component of special machine |
| CN114152442A (en) * | 2021-12-24 | 2022-03-08 | 杭州电子科技大学 | Rolling bearing cross-working condition fault detection method based on migration convolutional neural network |
| CN114190897A (en) * | 2021-12-15 | 2022-03-18 | 中国科学院空天信息创新研究院 | Training method of sleep staging model, sleep staging method and device |
| CN114510968A (en) * | 2022-01-21 | 2022-05-17 | 石家庄铁道大学 | Fault diagnosis method based on Transformer |
| CN114548301A (en) * | 2022-02-25 | 2022-05-27 | 西安交通大学 | Compressor fault diagnosis method and system based on transfer learning |
| CN114638060A (en) * | 2022-03-10 | 2022-06-17 | 重庆英科铸数网络科技有限公司 | Fault prediction method, system, electronic device and computer program product |
| CN115062648A (en) * | 2022-06-02 | 2022-09-16 | 华南理工大学 | Fault diagnosis method, system and device for rotary machine and storage medium |
| CN115270997A (en) * | 2022-09-20 | 2022-11-01 | 中国人民解放军32035部队 | Rocket target attitude stability discrimination method based on transfer learning and related device |
| CN116681124A (en) * | 2023-08-03 | 2023-09-01 | 北京工业大学 | Transfer learning method for fault diagnosis of rotary machinery |
| CN116704266A (en) * | 2023-07-28 | 2023-09-05 | 国网浙江省电力有限公司信息通信分公司 | Power equipment fault detection method, device, equipment and storage medium |
| CN117218602A (en) * | 2023-10-20 | 2023-12-12 | 青岛理工大学 | Structural health monitoring data abnormality diagnosis method and system |
| CN117435441A (en) * | 2023-12-21 | 2024-01-23 | 国网山东省电力公司信息通信公司 | Log data-based fault diagnosis method and device |
| CN117435441B (en) * | 2023-12-21 | 2024-04-26 | 国网山东省电力公司信息通信公司 | Log data-based fault diagnosis method and device |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060178918A1 (en) * | 1999-11-22 | 2006-08-10 | Accenture Llp | Technology sharing during demand and supply planning in a network-based supply chain environment |
| JP2010182287A (en) * | 2008-07-17 | 2010-08-19 | Steven C Kays | Intelligent adaptive design |
| US20150052092A1 (en) * | 2013-08-16 | 2015-02-19 | Transoft (Shanghai), Inc. | Methods and systems of brain-like computing virtualization |
| US20170116497A1 (en) * | 2015-09-16 | 2017-04-27 | Siemens Healthcare Gmbh | Intelligent Multi-scale Medical Image Landmark Detection |
| CN106844904A (en) * | 2017-01-04 | 2017-06-13 | 滨州东瑞机械有限公司 | A kind of gas turbine vacuum pump impeller Optimization Design |
| US20180165575A1 (en) * | 2016-12-08 | 2018-06-14 | Via Alliance Semiconductor Co., Ltd. | Neural network unit with mixed data and weight size computation capability |
| CN108414226A (en) * | 2017-12-25 | 2018-08-17 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under the variable working condition of feature based transfer learning |
| CN108647723A (en) * | 2018-05-11 | 2018-10-12 | 湖北工业大学 | A kind of image classification method based on deep learning network |
-
2019
- 2019-02-26 CN CN201910142412.1A patent/CN109918752B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060178918A1 (en) * | 1999-11-22 | 2006-08-10 | Accenture Llp | Technology sharing during demand and supply planning in a network-based supply chain environment |
| JP2010182287A (en) * | 2008-07-17 | 2010-08-19 | Steven C Kays | Intelligent adaptive design |
| US20150052092A1 (en) * | 2013-08-16 | 2015-02-19 | Transoft (Shanghai), Inc. | Methods and systems of brain-like computing virtualization |
| US20170116497A1 (en) * | 2015-09-16 | 2017-04-27 | Siemens Healthcare Gmbh | Intelligent Multi-scale Medical Image Landmark Detection |
| US20180165575A1 (en) * | 2016-12-08 | 2018-06-14 | Via Alliance Semiconductor Co., Ltd. | Neural network unit with mixed data and weight size computation capability |
| CN106844904A (en) * | 2017-01-04 | 2017-06-13 | 滨州东瑞机械有限公司 | A kind of gas turbine vacuum pump impeller Optimization Design |
| CN108414226A (en) * | 2017-12-25 | 2018-08-17 | 哈尔滨理工大学 | Fault Diagnosis of Roller Bearings under the variable working condition of feature based transfer learning |
| CN108647723A (en) * | 2018-05-11 | 2018-10-12 | 湖北工业大学 | A kind of image classification method based on deep learning network |
Non-Patent Citations (2)
| Title |
|---|
| PHILLIP M. CHENG 等: "Transfer Learning with Convolutional Neural Networks for Classification of Abdominal Ultrasound Images", 《JOURNAL OF DIGITAL IMAGING》 * |
| 许敏等: "一种具有迁移学习能力的RBF-NN算法及其应用", 《智能系统学报》 * |
Cited By (121)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110225453A (en) * | 2019-06-24 | 2019-09-10 | 鲸数科技（北京）有限公司 | Mobile terminal locating method, device, electronic equipment and storage medium |
| CN110502991A (en) * | 2019-07-18 | 2019-11-26 | 武汉理工大学 | Internal combustion engine health monitor method and system based on random convolutional neural networks structure |
| CN110502991B (en) * | 2019-07-18 | 2022-12-02 | 武汉理工大学 | Internal combustion engine health monitoring method and system based on random convolutional neural network structure |
| CN110464611A (en) * | 2019-07-23 | 2019-11-19 | 苏州国科视清医疗科技有限公司 | A kind of digitlization amblyopia enhancing training device and system and its related algorithm |
| CN110378431A (en) * | 2019-07-24 | 2019-10-25 | 中国人民解放军国防科技大学 | Convolutional neural network-based supersonic combustion chamber combustion mode detection method |
| CN110533070A (en) * | 2019-07-26 | 2019-12-03 | 西安交通大学 | A kind of mechanical breakdown intelligent diagnosing method based on migration original shape network under small sample |
| WO2021022852A1 (en) * | 2019-08-02 | 2021-02-11 | 华为技术有限公司 | Access request processing method, apparatus, device, and storage medium |
| CN110647923B (en) * | 2019-09-04 | 2022-08-16 | 西安交通大学 | Variable working condition mechanical fault intelligent diagnosis method based on self-learning under small sample |
| CN110647923A (en) * | 2019-09-04 | 2020-01-03 | 西安交通大学 | Variable working condition mechanical fault intelligent diagnosis method based on self-learning under small sample |
| CN110705029A (en) * | 2019-09-05 | 2020-01-17 | 西安交通大学 | Flow field prediction method of oscillating flapping wing energy acquisition system based on transfer learning |
| CN112541511A (en) * | 2019-09-20 | 2021-03-23 | 宫文峰 | Multi-channel time series data fault diagnosis method based on convolutional neural network |
| CN110705456A (en) * | 2019-09-29 | 2020-01-17 | 华南理工大学 | Micro motor abnormity detection method based on transfer learning |
| CN110726813A (en) * | 2019-10-12 | 2020-01-24 | 浙江大学 | Electronic nose prediction method based on double-layer integrated neural network |
| CN110726813B (en) * | 2019-10-12 | 2021-04-27 | 浙江大学 | Electronic nose prediction method based on double-layer integrated neural network |
| CN111178378A (en) * | 2019-11-07 | 2020-05-19 | 腾讯科技（深圳）有限公司 | Equipment fault prediction method and device, electronic equipment and storage medium |
| CN111178378B (en) * | 2019-11-07 | 2023-05-16 | 腾讯科技（深圳）有限公司 | Equipment fault prediction method and device, electronic equipment and storage medium |
| CN110995475A (en) * | 2019-11-20 | 2020-04-10 | 国网湖北省电力有限公司信息通信公司 | Power communication network fault detection method based on transfer learning |
| CN110866365A (en) * | 2019-11-22 | 2020-03-06 | 北京航空航天大学 | Mechanical equipment intelligent fault diagnosis method based on partial migration convolutional network |
| CN110866365B (en) * | 2019-11-22 | 2021-06-01 | 北京航空航天大学 | Mechanical equipment intelligent fault diagnosis method based on partial migration convolutional network |
| CN110991295B (en) * | 2019-11-26 | 2022-05-06 | 电子科技大学 | Self-adaptive fault diagnosis method based on one-dimensional convolutional neural network |
| CN110991295A (en) * | 2019-11-26 | 2020-04-10 | 电子科技大学 | Self-adaptive fault diagnosis method based on one-dimensional convolutional neural network |
| CN111027678A (en) * | 2019-12-04 | 2020-04-17 | 湃方科技(北京)有限责任公司 | Data migration method and device |
| CN111027678B (en) * | 2019-12-04 | 2023-08-04 | 湃方科技(北京)有限责任公司 | Data migration method and device |
| CN111007035A (en) * | 2019-12-12 | 2020-04-14 | 重庆科技学院 | Method, system and device for detecting concentration of high-temperature gas in secondary combustion chamber based on convolutional neural network and storage medium |
| CN111007035B (en) * | 2019-12-12 | 2022-03-25 | 重庆科技学院 | Method, system and device for detecting concentration of high-temperature gas in secondary combustion chamber based on convolutional neural network and storage medium |
| CN111191693A (en) * | 2019-12-18 | 2020-05-22 | 广西电网有限责任公司电力科学研究院 | Method for identifying thermal fault state of high-voltage switch cabinet based on convolutional neural network |
| CN111442926A (en) * | 2020-01-11 | 2020-07-24 | 哈尔滨理工大学 | Fault diagnosis method for rolling bearings of different models under variable load based on deep characteristic migration |
| CN111275164A (en) * | 2020-01-13 | 2020-06-12 | 浙江大学 | Underwater robot propulsion system fault diagnosis method |
| CN111291814A (en) * | 2020-02-15 | 2020-06-16 | 河北工业大学 | Crack identification algorithm based on convolution neural network and information entropy data fusion strategy |
| CN111291814B (en) * | 2020-02-15 | 2023-06-02 | 河北工业大学 | Crack identification algorithm based on convolutional neural network and information entropy data fusion strategy |
| CN111476299A (en) * | 2020-04-07 | 2020-07-31 | 国家电网有限公司华东分部 | Improved convolutional neural network and power grid intelligent alarm system based on same |
| CN111497868A (en) * | 2020-04-09 | 2020-08-07 | 余姚市浙江大学机器人研究中心 | Automobile sensor fault classification method based on BN-L STM network |
| CN113541985A (en) * | 2020-04-14 | 2021-10-22 | 中国移动通信集团浙江有限公司 | Internet of things fault diagnosis method, training method of model and related device |
| US11930307B2 (en) * | 2020-04-17 | 2024-03-12 | Beijing Baidu Netcom Science Technology Co., Ltd. | Image processing method and apparatus, electronic device and computer-readable storage medium |
| US20210203900A1 (en) * | 2020-04-17 | 2021-07-01 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Image processing method and apparatus, electronic device and computer-readable storage medium |
| CN113642591A (en) * | 2020-04-27 | 2021-11-12 | 中国科学院声学研究所 | Migration learning-based multi-beam seabed sediment layer type estimation method and system |
| CN113642591B (en) * | 2020-04-27 | 2023-07-18 | 中国科学院声学研究所 | Multi-beam submarine sediment layer type estimation method and system based on transfer learning |
| CN111582477B (en) * | 2020-05-09 | 2023-08-29 | 北京百度网讯科技有限公司 | Training method and device for neural network model |
| CN111582477A (en) * | 2020-05-09 | 2020-08-25 | 北京百度网讯科技有限公司 | Training method and device of neural network model |
| CN111582396B (en) * | 2020-05-13 | 2023-05-02 | 江南大学 | Fault diagnosis method based on improved convolutional neural network |
| CN111582396A (en) * | 2020-05-13 | 2020-08-25 | 江南大学 | Fault diagnosis method based on improved convolutional neural network |
| CN111598161A (en) * | 2020-05-14 | 2020-08-28 | 哈尔滨工业大学（威海） | Engine gas circuit state diagnosis system based on CNN transfer learning |
| CN111562110A (en) * | 2020-05-25 | 2020-08-21 | 南京理工大学 | Fault diagnosis model based on convolutional neural network and cross-component fault diagnosis method |
| CN111638465B (en) * | 2020-05-29 | 2023-02-28 | 浙大宁波理工学院 | Lithium battery health state estimation method based on convolutional neural network and transfer learning |
| CN111638465A (en) * | 2020-05-29 | 2020-09-08 | 浙大宁波理工学院 | Lithium battery health state estimation method based on convolutional neural network and transfer learning |
| CN111651937B (en) * | 2020-06-03 | 2023-07-25 | 苏州大学 | Method for diagnosing faults of in-class self-adaptive bearing under variable working conditions |
| CN111651937A (en) * | 2020-06-03 | 2020-09-11 | 苏州大学 | Method for diagnosing similar self-adaptive bearing fault under variable working conditions |
| CN111667000B (en) * | 2020-06-05 | 2023-02-07 | 成都理工大学 | Earthquake early warning method of adaptive field deep neural network |
| CN111667000A (en) * | 2020-06-05 | 2020-09-15 | 成都理工大学 | Earthquake early warning method of adaptive field deep neural network |
| CN111766513A (en) * | 2020-06-18 | 2020-10-13 | 电子科技大学 | Capsule network-based variable-working-condition multi-fault diagnosis method for three-phase induction motor |
| CN111766513B (en) * | 2020-06-18 | 2021-08-20 | 电子科技大学 | Capsule network-based variable-working-condition multi-fault diagnosis method for three-phase induction motor |
| CN111814396A (en) * | 2020-07-02 | 2020-10-23 | 重庆大学 | Centrifugal fan fault early warning method based on transfer learning |
| CN111814396B (en) * | 2020-07-02 | 2024-02-20 | 重庆大学 | Centrifugal fan fault early warning method based on transfer learning |
| CN111751650B (en) * | 2020-07-06 | 2021-06-22 | 重庆大学 | Non-invasive household electric equipment on-line monitoring system and fault identification method |
| CN111751650A (en) * | 2020-07-06 | 2020-10-09 | 重庆大学 | Non-invasive household electric equipment on-line monitoring system and fault identification method |
| CN111680788A (en) * | 2020-07-14 | 2020-09-18 | 同济大学 | Equipment fault diagnosis method based on deep learning |
| CN111680788B (en) * | 2020-07-14 | 2024-03-01 | 同济大学 | Equipment fault diagnosis method based on deep learning |
| CN111931826A (en) * | 2020-07-20 | 2020-11-13 | 华南理工大学 | Rolling bearing fault diagnosis method and system based on multi-scale convolution migration model |
| CN111897310A (en) * | 2020-07-24 | 2020-11-06 | 华中科技大学 | Industrial process fault classification method and system based on one-dimensional multi-head convolutional network |
| CN111881627A (en) * | 2020-08-05 | 2020-11-03 | 哈尔滨工程大学 | Nuclear power device fault diagnosis method and system |
| CN111899905A (en) * | 2020-08-05 | 2020-11-06 | 哈尔滨工程大学 | Fault diagnosis method and system based on nuclear power device |
| CN111899905B (en) * | 2020-08-05 | 2022-11-01 | 哈尔滨工程大学 | Fault diagnosis method and system based on nuclear power device |
| CN112051062B (en) * | 2020-08-10 | 2021-07-27 | 华中科技大学 | Fault diagnosis method, device, equipment and storage medium for rolling bearing |
| CN112051062A (en) * | 2020-08-10 | 2020-12-08 | 华中科技大学 | Fault diagnosis method, device, equipment and storage medium for rolling bearing |
| CN112149866A (en) * | 2020-08-17 | 2020-12-29 | 江苏大学 | Intelligent manufacturing workshop anomaly prediction and control method based on edge cloud cooperation |
| CN112146879A (en) * | 2020-08-21 | 2020-12-29 | 江苏大学 | Rolling bearing fault intelligent diagnosis method and system |
| CN112152201A (en) * | 2020-08-25 | 2020-12-29 | 国网山东综合能源服务有限公司 | Electricity load prediction method and system based on convolution length time memory neural network |
| CN112152201B (en) * | 2020-08-25 | 2021-09-17 | 国网山东综合能源服务有限公司 | Electricity load prediction method and system based on convolution length time memory neural network |
| CN112113755A (en) * | 2020-09-02 | 2020-12-22 | 西安交通大学 | Mechanical fault intelligent diagnosis method based on deep convolution-kurtosis neural network |
| CN112036496A (en) * | 2020-09-02 | 2020-12-04 | 哈尔滨工程大学 | Nuclear power device fault diagnosis method and system |
| CN112052628A (en) * | 2020-09-06 | 2020-12-08 | 上海电力大学 | Gas turbine gas circuit fault prediction and diagnosis method based on knowledge and data |
| CN112052628B (en) * | 2020-09-06 | 2022-09-16 | 上海电力大学 | Gas turbine gas circuit fault prediction and diagnosis method based on knowledge and data |
| CN112161784A (en) * | 2020-09-07 | 2021-01-01 | 华南理工大学 | Mechanical fault diagnosis method based on multi-sensor information fusion migration network |
| CN112101220B (en) * | 2020-09-15 | 2023-03-03 | 哈尔滨理工大学 | Rolling bearing service life prediction method based on unsupervised model parameter migration |
| CN112101220A (en) * | 2020-09-15 | 2020-12-18 | 哈尔滨理工大学 | Rolling bearing service life prediction method based on unsupervised model parameter migration |
| CN112444748A (en) * | 2020-10-12 | 2021-03-05 | 武汉蔚来能源有限公司 | Battery abnormality detection method, battery abnormality detection device, electronic apparatus, and storage medium |
| CN112163640A (en) * | 2020-10-30 | 2021-01-01 | 北京信息科技大学 | Planetary gearbox fault diagnosis method and system based on deep convolution migration learning |
| CN112465030A (en) * | 2020-11-28 | 2021-03-09 | 河南大学 | Multi-source heterogeneous information fusion fault diagnosis method based on two-stage transfer learning |
| CN112465030B (en) * | 2020-11-28 | 2022-06-07 | 河南财政金融学院 | Multi-source heterogeneous information fusion fault diagnosis method based on two-stage transfer learning |
| CN112667791A (en) * | 2020-12-23 | 2021-04-16 | 深圳壹账通智能科技有限公司 | Latent event prediction method, device, equipment and storage medium |
| WO2022134834A1 (en) * | 2020-12-23 | 2022-06-30 | 深圳壹账通智能科技有限公司 | Potential event predicting method, apparatus and device, and storage medium |
| CN112763214B (en) * | 2020-12-31 | 2022-07-01 | 南京信息工程大学 | Rolling bearing fault diagnosis method based on multi-label zero-sample learning |
| CN112763214A (en) * | 2020-12-31 | 2021-05-07 | 南京信息工程大学 | Rolling bearing fault diagnosis method based on multi-label zero-sample learning |
| CN112766200A (en) * | 2021-01-26 | 2021-05-07 | 福州大学 | Planetary gear box small sample fault diagnosis method based on one-dimensional convolutional neural network |
| CN112836896A (en) * | 2021-03-03 | 2021-05-25 | 西门子工厂自动化工程有限公司 | Method for maintaining equipment and system for maintaining equipment |
| CN112884070A (en) * | 2021-03-17 | 2021-06-01 | 云南电网有限责任公司电力科学研究院 | High-voltage switch fault diagnosis method based on transfer learning |
| CN112926728A (en) * | 2021-03-22 | 2021-06-08 | 南京理工大学 | Small sample turn-to-turn short circuit fault diagnosis method for permanent magnet synchronous motor |
| CN112989702A (en) * | 2021-03-25 | 2021-06-18 | 河北工业大学 | Self-learning method for equipment performance analysis and prediction |
| CN113063595A (en) * | 2021-03-27 | 2021-07-02 | 温州大学 | Full convolution neural network rolling bearing fault identification method based on one-dimensional convolution |
| CN113203566A (en) * | 2021-04-06 | 2021-08-03 | 上海吞山智能科技有限公司 | Motor bearing fault diagnosis method based on one-dimensional data enhancement and CNN |
| CN113392881A (en) * | 2021-05-27 | 2021-09-14 | 重庆大学 | Rotary machine fault diagnosis method based on transfer learning |
| CN113469219A (en) * | 2021-06-08 | 2021-10-01 | 中国地质大学（武汉） | Rotary machine fault diagnosis method under complex working condition based on element transfer learning |
| CN113343633B (en) * | 2021-06-10 | 2022-04-26 | 上海交通大学 | Thermal runaway fault classification and risk prediction method and system for power lithium battery |
| CN113343633A (en) * | 2021-06-10 | 2021-09-03 | 上海交通大学 | Thermal runaway fault classification and risk prediction method and system for power lithium battery |
| CN113469230A (en) * | 2021-06-17 | 2021-10-01 | 北京信息科技大学 | Method, system and medium for diagnosing deep migration fault of rotor system |
| CN113469230B (en) * | 2021-06-17 | 2023-07-07 | 北京信息科技大学 | Rotor system deep migration fault diagnosis method, system and medium |
| CN113468805B (en) * | 2021-06-21 | 2022-10-14 | 国网黑龙江省电力有限公司电力科学研究院 | Method for evaluating reliability of sensor in extremely cold environment, electronic device and storage medium |
| CN113468805A (en) * | 2021-06-21 | 2021-10-01 | 国网黑龙江省电力有限公司电力科学研究院 | Method for evaluating reliability of sensor in extremely cold environment, electronic device and storage medium |
| CN113627518A (en) * | 2021-08-07 | 2021-11-09 | 福州大学 | Method for realizing multichannel convolution-recurrent neural network electroencephalogram emotion recognition model by utilizing transfer learning |
| CN113627518B (en) * | 2021-08-07 | 2023-08-08 | 福州大学 | Method for realizing neural network brain electricity emotion recognition model by utilizing transfer learning |
| CN113867307A (en) * | 2021-08-20 | 2021-12-31 | 天津大学 | Spacecraft intelligent fault diagnosis method based on deep neural network |
| CN114190897A (en) * | 2021-12-15 | 2022-03-18 | 中国科学院空天信息创新研究院 | Training method of sleep staging model, sleep staging method and device |
| CN114190897B (en) * | 2021-12-15 | 2024-04-05 | 中国科学院空天信息创新研究院 | Training method of sleep stage model, sleep stage method and device |
| CN114152442A (en) * | 2021-12-24 | 2022-03-08 | 杭州电子科技大学 | Rolling bearing cross-working condition fault detection method based on migration convolutional neural network |
| CN114057053A (en) * | 2022-01-18 | 2022-02-18 | 杭州浅水数字技术有限公司 | Method for monitoring fatigue degree of component of special machine |
| CN114057053B (en) * | 2022-01-18 | 2022-04-26 | 杭州浅水数字技术有限公司 | Method for monitoring fatigue degree of component of special machine |
| CN114510968B (en) * | 2022-01-21 | 2022-09-16 | 石家庄铁道大学 | Fault diagnosis method based on Transformer |
| CN114510968A (en) * | 2022-01-21 | 2022-05-17 | 石家庄铁道大学 | Fault diagnosis method based on Transformer |
| CN114548301A (en) * | 2022-02-25 | 2022-05-27 | 西安交通大学 | Compressor fault diagnosis method and system based on transfer learning |
| CN114548301B (en) * | 2022-02-25 | 2024-03-01 | 西安交通大学 | Compressor fault diagnosis method and system based on transfer learning |
| CN114638060B (en) * | 2022-03-10 | 2023-02-17 | 重庆英科铸数网络科技有限公司 | Fault prediction method, system and electronic equipment |
| CN114638060A (en) * | 2022-03-10 | 2022-06-17 | 重庆英科铸数网络科技有限公司 | Fault prediction method, system, electronic device and computer program product |
| CN115062648A (en) * | 2022-06-02 | 2022-09-16 | 华南理工大学 | Fault diagnosis method, system and device for rotary machine and storage medium |
| CN115270997A (en) * | 2022-09-20 | 2022-11-01 | 中国人民解放军32035部队 | Rocket target attitude stability discrimination method based on transfer learning and related device |
| CN115270997B (en) * | 2022-09-20 | 2022-12-27 | 中国人民解放军32035部队 | Rocket target attitude stability discrimination method based on transfer learning and related device |
| CN116704266B (en) * | 2023-07-28 | 2023-10-31 | 国网浙江省电力有限公司信息通信分公司 | Power equipment fault detection method, device, equipment and storage medium |
| CN116704266A (en) * | 2023-07-28 | 2023-09-05 | 国网浙江省电力有限公司信息通信分公司 | Power equipment fault detection method, device, equipment and storage medium |
| CN116681124A (en) * | 2023-08-03 | 2023-09-01 | 北京工业大学 | Transfer learning method for fault diagnosis of rotary machinery |
| CN117218602A (en) * | 2023-10-20 | 2023-12-12 | 青岛理工大学 | Structural health monitoring data abnormality diagnosis method and system |
| CN117435441A (en) * | 2023-12-21 | 2024-01-23 | 国网山东省电力公司信息通信公司 | Log data-based fault diagnosis method and device |
| CN117435441B (en) * | 2023-12-21 | 2024-04-26 | 国网山东省电力公司信息通信公司 | Log data-based fault diagnosis method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109918752B (en) | 2022-12-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109918752A (en) | Mechanical failure diagnostic method, equipment and medium based on migration convolutional neural networks | |
| CN106555788B (en) | Application based on the deep learning of Fuzzy Processing in hydraulic equipment fault diagnosis | |
| WO2020244134A1 (en) | Multi-task feature sharing neural network-based intelligent fault diagnosis method | |
| Luo et al. | Convolutional neural network based on attention mechanism and Bi-LSTM for bearing remaining life prediction | |
| CN106980822B (en) | A kind of rotary machinery fault diagnosis method based on selective ensemble study | |
| CN112577747A (en) | Rolling bearing fault diagnosis method based on space pooling network | |
| CN107066759A (en) | A kind of Vibration Fault Diagnosis of Turbine Rotor method and device | |
| CN110132554A (en) | A kind of rotary machinery fault diagnosis method that depth Laplce encodes certainly | |
| CN107688825A (en) | A kind of follow-on integrated weighting extreme learning machine sewage disposal failure examines method | |
| CN114429153A (en) | Lifetime learning-based gearbox increment fault diagnosis method and system | |
| CN111289250A (en) | Method for predicting residual service life of rolling bearing of servo motor | |
| CN106610584A (en) | Remanufacturing process quality control method based on neural network and expert system | |
| Zhang et al. | A deep learning-based multi-sensor data fusion method for degradation monitoring of ball screws | |
| Saufi et al. | Machinery fault diagnosis based on a modified hybrid deep sparse autoencoder using a raw vibration time-series signal | |
| CN112763215B (en) | Multi-working-condition online fault diagnosis method based on modular federal deep learning | |
| CN107016440B (en) | The multiresolution deep neural network intelligent diagnosing method of machine driving failure | |
| CN111562110B (en) | Fault diagnosis model based on convolutional neural network and cross-component fault diagnosis method | |
| CN112990601B (en) | Worm wheel machining precision self-healing system and method based on data mining | |
| CN113591625B (en) | Gearbox fault diagnosis method based on unbalanced sample importance weighting | |
| Kopparti et al. | Weight priors for learning identity relations | |
| CN113971489A (en) | Method and system for predicting remaining service life based on hybrid neural network | |
| Shuai | Research on Fault Diagnosis Technology of Industrial Robot Operation Based on Deep Belief Network | |
| Ge et al. | Remaining useful life prediction using deep multi-scale convolution neural networks | |
| CN108763728B (en) | Mechanical fault diagnosis method using parallel deep neural network hierarchical feature extraction | |
| Saxena et al. | Genetic algorithms for artificial neural net-based condition monitoring system design for rotating mechanical systems |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |