CN115221958A - Mechanical equipment selective migration fault diagnosis method based on composite weight - Google Patents
Mechanical equipment selective migration fault diagnosis method based on composite weight Download PDFInfo
- Publication number
- CN115221958A CN115221958A CN202210841133.6A CN202210841133A CN115221958A CN 115221958 A CN115221958 A CN 115221958A CN 202210841133 A CN202210841133 A CN 202210841133A CN 115221958 A CN115221958 A CN 115221958A
- Authority
- CN
- China
- Prior art keywords
- domain
- weight
- source domain
- samples
- training
- 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
- 238000013508 migration Methods 0.000 title claims abstract description 37
- 230000005012 migration Effects 0.000 title claims abstract description 37
- 238000003745 diagnosis Methods 0.000 title claims abstract description 34
- 238000000034 method Methods 0.000 title claims abstract description 31
- 239000002131 composite material Substances 0.000 title claims abstract description 13
- 238000012549 training Methods 0.000 claims abstract description 29
- 230000006978 adaptation Effects 0.000 claims abstract description 17
- 150000001875 compounds Chemical class 0.000 claims abstract description 7
- 238000009826 distribution Methods 0.000 claims description 20
- 230000006870 function Effects 0.000 claims description 12
- 238000005457 optimization Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000012360 testing method Methods 0.000 description 6
- 230000036541 health Effects 0.000 description 5
- 230000003044 adaptive effect Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Images
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
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
A mechanical equipment selective migration fault diagnosis method based on compound weight comprises the steps of firstly, collecting vibration signals under various working conditions from mechanical equipment by using collection equipment, and establishing data sets of different working conditions; then constructing a selective migration network based on the composite weight, wherein the selective migration network comprises a feature extractor, a state classifier, a domain discriminator and a domain adaptation module based on Wasserstein distance; pre-training a feature extractor and a state recognizer using a back propagation algorithm; loading the pre-training weight into the selective migration network, and finishing the training by using a back propagation algorithm; finally, acquiring a selective migration fault diagnosis result of the mechanical equipment; according to the invention, a state classifier and a domain discriminator are started at the same time, class-level weight and sample-level weight are respectively obtained, a target function of the state classifier and a domain adaptation module based on Wasserstein distance is weighted by using a composite weighting mode, the reliability of selective migration is improved, and the selective migration fault diagnosis of cross-working-condition mechanical equipment is effectively realized.
Description
Technical Field
The invention belongs to the technical field of intelligent diagnosis of mechanical equipment, and particularly relates to a selective migration fault diagnosis method of mechanical equipment based on compound weight.
Background
With the rapid development of modern manufacturing, mechanical equipment has become an important pillar of modern industry, and then the mechanical equipment inevitably fails due to a complicated working environment and external disturbance. Once a fault occurs, there is a possibility of significant economic loss and casualties, and therefore, the fault diagnosis of mechanical equipment has attracted great attention in the industry.
With the development of deep learning, the intelligent fault diagnosis based on the depth model has been developed greatly, however, when the supervised depth model (application number CN201911155556.7, named as: the intelligent fault diagnosis method for mechanical equipment based on partial migration convolutional network) is applied to fault diagnosis for mechanical equipment, there are three disadvantages: 1) In an actual industrial environment, it is very difficult to obtain enough labeled data, so that the difficulty in training a supervised depth model is very high; 2) Due to factors such as working condition change, external interference and equipment loss, original data and target test data in practical application have different data distribution; 3) The label space of the original training data set and the target test data set are quite possibly different, and in the actual fault diagnosis task, the data label category of the target domain is inevitably smaller than that of the source domain. These disadvantages not only create the possibility of misdiagnosis and missed diagnosis, but also can cause significant losses to modern production.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a mechanical equipment selective migration fault diagnosis method based on compound weight, the adopted training method is simple, the diagnosis knowledge of a source domain can be selectively migrated to a target domain, and the intelligent cross-domain diagnosis of mechanical equipment is realized.
In order to achieve the purpose, the invention adopts the technical scheme that:
a mechanical equipment selective migration fault diagnosis method based on compound weight comprises the following steps:
the method comprises the following steps: collecting vibration data of mechanical equipment under different working conditions, wherein each working condition corresponds to different domains;
step two: constructing a selective migration network based on composite weight, wherein the selective migration network comprises a feature extractor F and a trainable parameter theta F A state classifier C and trainable parameters theta C A domain discriminator D and a trainable parameter theta D A domain adaptation module based on Wasserstein distance; the feature extractor comprises a source domain feature extractor and a target domain feature extractor;
for the intelligent fault diagnosis task of a mechanical device, the source domain data set is
Wherein
n s Respectively representing source domain samples, source domain sample labels and the number of the source domain samples; the target domain data set is
Wherein
n t Respectively representing the number of target domain samples and the number of target domain samples; training feature extractors and state classifiers under labeled source domain data, and utilizing cross entropy loss function L ce Reducing empirical risk loss on source domain and achieving a class separationThe optimization target of the above process is expressed as:
wherein L is c To classify the losses, F (x) i ) Extracting a sample x for a feature extractor i C (x) i ) Softmax output, y) as classifier i Is the corresponding source domain sample;
the domain discriminator is used to distinguish whether the sample is from the source domain or the target domain, the samples of the source domain and the target domain are labeled with the domain label d i Respectively 1 and 0, and training to obtain a domain discriminator for distinguishing a source domain from a target domain, wherein the optimization goal of the training process is defined as follows:
wherein L is bce Representing a two-class cross entropy loss function;
a Wasserstein distance-based domain adaptation module is used to obtain fine-grained class-level feature alignment, distributing P for the source domain s And target domain distribution P t Distance L of Wasserstein w The definition is as follows:
wherein the distribution P belongs to a combined distribution set Π (P) s ,P t )，h s And h t Indicating the location of the source domain distribution and the location of the target domain distribution; the domain adaptation module based on Wasserstein distance has no parameters needing to be updated in model training;
for a state classifier trained on a source domain, samples of a target domain are more classified into a shared class of the target domain and the source domain than a unique class of the source domain, so that a prediction result is used as a class-level weight to select diagnosis knowledge for migration; inputting the samples on the target domain into the state classifier trained by the source domain, the soft label y' of the source domain samples can be expressed as:
thus, the class-level weight α is calculated by:
similarly, for a trained domain discriminator, the pseudo domain label d' for all samples of the source domain is represented as:
thus, the sample level weight β i Calculated from the following formula:
β i ＝1-d′ i (i＝1,2…,n s )
source domain samples x i Composite weight w of i Calculated from the following formula:
w i ＝β i ×α(y i |x i )(i＝1,2,…,n s )
wherein, α (y) i |x i ) Is a source domain sample x i Is given by the label y i Weight of (1), beta i Is a source domain sample x i The weight of (c);
all weights are normalized using the maximum of the weights as follows:
w i ＝w i /max(w)(i＝1,2,…,n s )
wherein, the weight w is the weight of all samples in training, and the weight is the weighted value of the composite weight module CCR;
step three: pre-training the feature extractor and the state classifier by using a back propagation algorithm, and storing parameters of the feature extractor and the state classifier;
step four: and loading the weights of the pre-training into the selective migration network constructed in the second step, wherein the total training targets are as follows:
L＝L c +L d +γL w
wherein γ is a trade-off parameter of the domain adaptation module based on the Wasserstein distance;
using complex weight module CCR to L c And L w Weighting is carried out to obtain a total objective function L, the objective function is optimized through a random gradient descent Adam algorithm, and the specific parameter updating rule is as follows:
wherein,
and
learning rate and partial derivative respectively;
step five: and inputting the samples of the target domain into the trained feature learner and the trained state classifier to obtain a fault diagnosis result.
The invention has the advantages that:
(1) According to the invention, the shared health state is weighted by adopting the composite weight of the class-level weight and the sample-level weight, so that the influence of the singular health state is reduced, and the identification of the shared health state is promoted;
(2) According to the method, domain adaptation modules based on Wasserstein distance are weighted, so that the source domain distribution is weighted to be aligned with the target domain distribution, and the shared category diagnosis knowledge of the source domain is promoted to be migrated to the target domain;
(3) The invention uses the end-to-end deep convolutional network as the basic framework of the invention, forms an end-to-end fault diagnosis framework, can adaptively learn the shared characteristic space with separable categories and inseparable domains, and improves the accuracy of selective migration fault diagnosis.
Drawings
FIG. 1 is a platform for testing a conventional planetary gear box according to the present invention.
Fig. 2 is a diagram of a selective migration network of the present invention.
Fig. 3 shows the results of Kappa coefficient evaluation for the implementation of the routine planetary gear box selective migration tasks T4 and T6.
Detailed Description
The present invention will be described in detail below with reference to the drawings and examples.
A mechanical equipment selective migration fault diagnosis method based on compound weight comprises the following steps:
the method comprises the following steps: the method comprises the following steps that vibration data of mechanical equipment under different working conditions are collected through signal collection equipment, each working condition corresponds to different domains, a planetary gear box experiment platform in the embodiment is shown in figure 1, one end of a testing planetary gear box is connected with a driving motor through a sensor, the other end of the testing planetary gear box is connected with a load motor through a test planetary gear box, and the experiment platform has three operation working conditions, namely a load of 50Nm, a load of 150Nm and a load of 250Nm; the first experimental gear has four health states including normal (Nor), sun gear fault (SG), planet gear fault (PG) and inner gear ring fault (RG), each health state comprises 200 samples, and each sample comprises 1024 data points;
step two: as shown in FIG. 2, a selective migration network based on compound weight is built, and the selective migration network comprises a feature extractor F and trainable parameters theta F A state classifier C and trainable parameters θ C A domain discriminator D and a trainable parameter theta D A domain adaptation module based on Wasserstein distance; it is noted that the feature extractor comprises a sourceA domain feature extractor and a target domain feature extractor;
for the planetary gearbox selective migration fault diagnosis task, the source domain data set is
Wherein
n s Respectively representing the number of source domain samples, source domain sample labels and source domain samples; the target domain data set is
Wherein
n t Respectively representing the number of target domain samples and the number of target domain samples; training feature extractors and state classifiers under labeled source domain data, and utilizing cross entropy loss function L ce Reducing the experience risk loss on the source domain and obtaining a class-separated feature space, the optimization goal of the above process can be expressed as:
wherein L is c To classify the losses, F (x) i ) Extracting a sample x for a feature extractor i C (x) i ) Softmax output, y) as classifier i Is the corresponding source domain sample;
the domain discriminator is used for distinguishing whether the sample comes from the source domain or the target domain, and the samples of the source domain and the target domain are marked with domain labels d i Respectively 1 and 0, a domain discriminator for distinguishing a source domain from a target domain can be obtained through training, and the optimization goal of the training process is defined as follows:
wherein L is bce Representing a two-class cross entropy loss function;
a domain adaptation module based on Wasserstein distance is used for obtaining feature alignment of class levels of fine granularity, the Wasserstein distance is widely used for measuring difference of two distributions, and the domain adaptation module is quite suitable for intelligent fault diagnosis tasks of domain adaptation and aims at source domain distribution P s And target domain distribution P t Distance L of Wasserstein w The definition is as follows:
wherein the distribution P belongs to a combined distribution set pi (P) s ,P t )，h s And h t Indicating the location of the source domain distribution and the location of the target domain distribution; it should be noted that the Wasserstein distance-based domain adaptation module has no parameters to be updated in model training;
for a state classifier trained on a source domain, samples of a target domain are more classified into a shared class of the target domain and the source domain rather than a unique class of the source domain, so that a prediction result can be used as a class-level weight to select diagnosis knowledge for migration; inputting the samples on the target domain into the state classifier trained by the source domain, the soft label y' of the source domain samples can be expressed as:
thus, the class-level weight α can be calculated by:
similarly, for a trained domain discriminator, the source domain samples of the shared class are difficult to distinguish, while the source domain samples of the unshared class are easy to be discriminated from the source domain, and based on this, the pseudo domain labels d' of all the samples of the source domain can be expressed as:
thus, the sample level weight β i Can be calculated from the following formula:
β i ＝1-d′ i (i＝1,2…,n s )
source domain samples x i Composite weight w of i Can be calculated from the following formula:
w i ＝β i ×α(y i |x i )(i＝1,2,…,n s )
wherein, α (y) i |x i ) Is a source domain sample x i Is given by the label y i Weight of (1), beta i Is a source domain sample x i The weight of (c);
all weights are normalized using the maximum of the weights as follows:
w i ＝w i /max(w)(i＝1,2,…,n s )
wherein, the weight w is the weight of all samples in training, and the weight is the weighted value of the composite weight module CCR;
step three: pre-training the feature extractor and the state classifier by using a back propagation algorithm, and storing parameters of the feature extractor and the state classifier;
step four: and loading the weights of the pre-training into the selective migration network constructed in the second step, wherein the total training targets are as follows:
L＝L c +L d +γL w
wherein the trade-off parameter γ of the Wasserstein distance-based domain adaptation module is set to 10;
using complex weight module CCR to L c And L w Weighting is carried out to obtain a total objective function L, the objective function is optimized through a random gradient descent Adam algorithm, and the specific parameter updating rule is as follows:
wherein,
and
learning rate and partial derivative respectively;
step five: and inputting the samples of the target domain into the trained feature learner and the trained state classifier to obtain a fault diagnosis result.
To verify the effectiveness of the present invention, other methods were chosen as compared to other methods: convolutional Neural Network (CNN), maximum Mean Difference (MMD), domain Adaptive Neural Network (DANN), deep convolutional migration learning network (DCTLN), importance sampling countermeasure network (IWAN), class weighted countermeasure network (CWDA), weighted sub-domain adaptive network (WSAN), weighted adaptive migration network (WAtn) as a comparison method. All the selective migration tasks for this planetary gearbox were set as in table 1, for a total of 6 tasks.
TABLE 1
| Source domain | Target domain | Source domain label | Target domain label | Task name |
| 50Nm | 250Nm | Nor,SG,PG,RG | Nor,SG,PG | T1 |
| 150Nm | 50Nm | Nor,SG,PG,RG | Nor,SG,RG | T2 |
| 150Nm | 250Nm | Nor,SG,PG,RG | Nor,PG,RG | T3 |
| 250Nm | 150Nm | Nor,SG,PG,RG | Nor,SG,RG | T4 |
| 250Nm | 150Nm | Nor,SG,PG,RG | Nor,SG,PG | T5 |
| 250Nm | 150Nm | Nor,SG,PG,RG | Nor,RG | T6 |
To eliminate the chance of network training, each method was repeated ten times per task to obtain the average accuracy and variance. The cross-working-condition fault diagnosis results of all the 6 planetary gearbox selective migration tasks are shown in the table 2;
TABLE 2
| Task | T1 | T2 | T3 | T4 | T5 | T6 |
| CNN | 71.19±2.08 | 91.92±1.79 | 86.33±1.82 | 75.63±2.47 | 92.80±1.87 | 72.35±3.35 |
| MMD | 80.83±2.70 | 69.56±2.89 | 80.77±4.31 | 70.35±3.38 | 70.67±4.06 | 68.55±2.62 |
| DANN | 90.70±1.60 | 93.68±2.15 | 88.83±2.35 | 80.50±3.67 | 84.93±2.68 | 79.10±1.34 |
| DCTLN | 95.93±1.31 | 68.06±3.87 | 83.93±2.82 | 67.85±3.17 | 70.27±3.65 | 79.40±2.21 |
| IWAN | 98.30±1.17 | 93.32±2.61 | 89.75±1.69 | 95.53±1.37 | 96.60±1.43 | 95.43±1.81 |
| CWDA | 96.82±1.86 | 92.92±1.99 | 90.85±1.59 | 95.72±2.05 | 95.55±0.65 | 95.98±2.46 |
| WSAN | 96.50±2.34 | 92.33±2.56 | 91.00±1.74 | 97.15±1.43 | 96.00±1.41 | 91.73±2.42 |
| WATN | 97.88±1.77 | 90.51±1.67 | 91.50±1.84 | 93.34±1.70 | 96.23±1.50 | 93.65±2.37 |
| The invention | 99.58±0.51 | 94.55±2.61 | 92.48±1.93 | 99.72±0.31 | 99.15±0.96 | 98.93±0.72 |
By comparison, the method provided by the invention has the advantages that the fault diagnosis precision of various selective migration tasks of the planetary gear box is obviously improved, and the accuracy of all tasks shows the effectiveness of the method provided by the invention. As shown in FIG. 3, for the selective migration tasks T4 and T6, the Kappa coefficient is used to further measure the classification accuracy of each method, and the method of the present invention is significantly higher than the comparison method, which further illustrates the superiority of the method of the present invention.
Claims (1)
1. A mechanical equipment selective migration fault diagnosis method based on compound weight is characterized by comprising the following steps:
the method comprises the following steps: collecting vibration data of mechanical equipment under different working conditions, wherein each working condition corresponds to different domains;
step two: constructing a selective migration network based on composite weight, wherein the selective migration network comprises a feature extractor F and a trainable parameter theta F A state classifier C and trainable parameters theta C A domain discriminator D and a trainable parameter theta D A domain adaptation module based on Wasserstein distance; the feature extractor comprises a source domain feature extractor and a target domain feature extractor;
for the intelligent fault diagnosis task of a mechanical device, the source domain data set is
Wherein
n s Respectively representing source domain samples, source domain sample labels and the number of the source domain samples; the target domain data set is
Wherein
n t Respectively representing the number of target domain samples and the number of target domain samples; in the presence of a labelTraining a feature extractor and a state classifier under source domain data by using a cross entropy loss function L ce Reducing the experience risk loss on the source domain and acquiring a class-separated feature space, wherein the optimization goal of the process is expressed as:
wherein L is c To classify the losses, F (x) i ) Extracting a sample x for a feature extractor i C (x) i ) Softmax output, y) as classifier i Is the corresponding source domain sample;
the domain discriminator is used to distinguish whether the sample is from the source domain or the target domain, the samples of the source domain and the target domain are labeled with the domain label d i Respectively 1 and 0, and training to obtain a domain discriminator for distinguishing a source domain from a target domain, wherein the optimization goal of the training process is defined as follows:
wherein L is bce Representing a two-class cross entropy loss function;
a Wasserstein distance-based domain adaptation module is used to obtain fine-grained class-level feature alignment, distributing P for the source domain s And target domain distribution P t Distance L of Wasserstein w The definition is as follows:
wherein the distribution P belongs to a combined distribution set pi (P) s ,P t )，h s And h t Indicating the location of the source domain distribution and the location of the target domain distribution; the domain adaptation module based on Wasserstein distance has no parameters needing to be updated in model training;
for a state classifier trained on a source domain, samples of a target domain are more classified into a shared class of the target domain and the source domain than a unique class of the source domain, so that a prediction result is used as a class-level weight to select diagnosis knowledge for migration; inputting samples on the target domain into a state classifier trained by a source domain, wherein a soft label y' of the source domain samples is expressed as:
thus, the class-level weight α is calculated by:
similarly, for a trained domain discriminator, the pseudo domain label d' for all samples of the source domain is represented as:
thus, the sample level weight β i Calculated from the following formula:
β i ＝1-d′ i (i＝1,2…,n s )
source domain samples x i Composite weight w of i Calculated from the following formula:
w i ＝β i ×α(y i |x i )(i＝1,2,…,n s )
wherein, α (y) i |x i ) Is a source domain sample x i Is given by the label y i Weight of (1), beta i Is a source domain sample x i The weight of (c);
all weights are normalized using the maximum of the weights as follows:
w i ＝w i /max(w)(i＝1,2,…,n s )
wherein, the weight w is the weight of all samples in training, and the weight is the weighted value of the composite weight module CCR;
step three: pre-training the feature extractor and the state classifier by using a back propagation algorithm, and storing parameters of the feature extractor and the state classifier;
step four: and loading the pre-trained weight into the selective migration network constructed in the second step, wherein the total training target is as follows:
L＝L c +L d +γL w
wherein γ is a trade-off parameter of the domain adaptation module based on the Wasserstein distance;
CCR-pair L using a complex weight module c And L w Weighting is carried out to obtain a total objective function L, the objective function is optimized through a random gradient descent Adam algorithm, and the specific parameter updating rule is as follows:
wherein,
and
learning rate and partial derivative respectively;
step five: and inputting the samples of the target domain into the trained feature learning device and the trained state classifier to obtain a fault diagnosis result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210841133.6A CN115221958A (en) | 2022-07-18 | 2022-07-18 | Mechanical equipment selective migration fault diagnosis method based on composite weight |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210841133.6A CN115221958A (en) | 2022-07-18 | 2022-07-18 | Mechanical equipment selective migration fault diagnosis method based on composite weight |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115221958A true CN115221958A (en) | 2022-10-21 |
Family
ID=83612361
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210841133.6A Pending CN115221958A (en) | 2022-07-18 | 2022-07-18 | Mechanical equipment selective migration fault diagnosis method based on composite weight |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115221958A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117407698A (en) * | 2023-12-14 | 2024-01-16 | 青岛明思为科技有限公司 | Hybrid distance guiding field self-adaptive fault diagnosis method |
| CN117708656A (en) * | 2024-02-05 | 2024-03-15 | 东北大学 | Rolling bearing cross-domain fault diagnosis method for single source domain |
-
2022
- 2022-07-18 CN CN202210841133.6A patent/CN115221958A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117407698A (en) * | 2023-12-14 | 2024-01-16 | 青岛明思为科技有限公司 | Hybrid distance guiding field self-adaptive fault diagnosis method |
| CN117407698B (en) * | 2023-12-14 | 2024-03-08 | 青岛明思为科技有限公司 | Hybrid distance guiding field self-adaptive fault diagnosis method |
| CN117708656A (en) * | 2024-02-05 | 2024-03-15 | 东北大学 | Rolling bearing cross-domain fault diagnosis method for single source domain |
| CN117708656B (en) * | 2024-02-05 | 2024-05-10 | 东北大学 | Rolling bearing cross-domain fault diagnosis method for single source domain |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112161784B (en) | Mechanical fault diagnosis method based on multi-sensor information fusion migration network | |
| CN109918752B (en) | Mechanical fault diagnosis method, equipment and medium based on migration convolution neural network | |
| CN110334764B (en) | Rotary machine intelligent fault diagnosis method based on integrated depth self-encoder | |
| CN109186973B (en) | Mechanical fault diagnosis method of unsupervised deep learning network | |
| CN104792522B (en) | Intelligent gear defect analysis method based on fractional wavelet transform and BP neutral network | |
| CN115600150A (en) | Multi-mode gearbox fault diagnosis method based on deep migration learning | |
| CN112577747A (en) | Rolling bearing fault diagnosis method based on space pooling network | |
| CN108256556A (en) | Wind-driven generator group wheel box method for diagnosing faults based on depth belief network | |
| CN115221958A (en) | Mechanical equipment selective migration fault diagnosis method based on composite weight | |
| CN113837000A (en) | Small sample fault diagnosis method based on task sequencing meta-learning | |
| CN108228716A (en) | SMOTE_Bagging integrated sewage handling failure diagnostic methods based on weighting extreme learning machine | |
| CN115563536A (en) | Rolling bearing fault diagnosis method based on subdomain self-adaptation | |
| CN108875772B (en) | Fault classification model and method based on stacked sparse Gaussian Bernoulli limited Boltzmann machine and reinforcement learning | |
| CN108871762A (en) | A kind of gearbox of wind turbine method for diagnosing faults | |
| CN111983569B (en) | Radar interference suppression method based on neural network | |
| CN113505655A (en) | Bearing fault intelligent diagnosis method for digital twin system | |
| CN109190464A (en) | Mechanical breakdown intelligent diagnosing method based on transfer learning under a kind of variable working condition | |
| CN114295377A (en) | CNN-LSTM bearing fault diagnosis method based on genetic algorithm | |
| CN111753891A (en) | Rolling bearing fault diagnosis method based on unsupervised feature learning | |
| CN114358124A (en) | Rotary machine new fault diagnosis method based on deep-antithetical-convolution neural network | |
| CN114417913B (en) | Bearing life prediction method based on PC-TCN and migration learning | |
| CN115290326A (en) | Rolling bearing fault intelligent diagnosis method | |
| CN113919220A (en) | Intelligent fault diagnosis method for rolling bearing driven by vibration image | |
| CN110232402A (en) | A kind of adaptive depth confidence network Fault Diagnosis of Roller Bearings of optimization | |
| CN109444604A (en) | A kind of DC/DC converter method for diagnosing faults based on convolutional neural networks |
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 |