CN112149726B - Totally-enclosed compressor fault diagnosis method based on knowledge sharing and model migration - Google Patents
Totally-enclosed compressor fault diagnosis method based on knowledge sharing and model migration Download PDFInfo
- Publication number
- CN112149726B CN112149726B CN202010996926.6A CN202010996926A CN112149726B CN 112149726 B CN112149726 B CN 112149726B CN 202010996926 A CN202010996926 A CN 202010996926A CN 112149726 B CN112149726 B CN 112149726B
- Authority
- CN
- China
- Prior art keywords
- data
- model
- target domain
- training
- migration model
- 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.)
- Active
Links
- 238000013508 migration Methods 0.000 title claims abstract description 91
- 230000005012 migration Effects 0.000 title claims abstract description 91
- 238000003745 diagnosis Methods 0.000 title claims abstract description 25
- 238000000034 method Methods 0.000 title claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 46
- 239000013598 vector Substances 0.000 claims abstract description 29
- 238000013459 approach Methods 0.000 claims abstract description 6
- 238000002360 preparation method Methods 0.000 claims abstract description 5
- 230000008014 freezing Effects 0.000 claims description 6
- 238000007710 freezing Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 5
- 230000003213 activating effect Effects 0.000 claims description 3
- 238000005452 bending Methods 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 2
- 230000001537 neural effect Effects 0.000 claims description 2
- 238000009826 distribution Methods 0.000 abstract description 4
- 238000005057 refrigeration Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 230000005611 electricity Effects 0.000 description 3
- 238000004134 energy conservation Methods 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 235000004257 Cordia myxa Nutrition 0.000 description 2
- 244000157795 Cordia myxa Species 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000003550 marker Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- CBENFWSGALASAD-UHFFFAOYSA-N Ozone Chemical compound [O-][O+]=O CBENFWSGALASAD-UHFFFAOYSA-N 0.000 description 1
- 229910000831 Steel Inorganic materials 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 239000003507 refrigerant Substances 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 238000010792 warming Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
A totally-enclosed compressor fault diagnosis method based on knowledge sharing and model migration belongs to the technical field of compressor fault diagnosis. It comprises the following steps: 1. data preparation and pretreatment; 2. preparing a migration model and independently training the model; 3. constructing a discriminator, training the discriminator and retraining a target migration model; 4. training a classifier; and fifthly, fault diagnosis. According to the method, the target domain equipment is not required to provide the data label, only part of the label-free data is provided, and the data dependence degree is greatly reduced; the approach of the source domain feature vector of the target domain feature vector distribution item is realized by using a gradient descent algorithm, so that model migration is realized; the depth dense connection self-encoder structure is utilized to realize unsupervised feature learning, and the data dependence degree is reduced again; aiming at the ultra-long time sequence data such as vibration signals, the deep dense connection network structure can realize the efficient training of the ultra-long time sequence data, and the data use efficiency can be greatly improved by folding the one-dimensional ultra-long data into a two-dimensional data structure.
Description
Technical Field
The invention belongs to the technical field of compressor fault diagnosis, and particularly relates to a totally-enclosed compressor fault diagnosis method based on knowledge sharing and model migration.
Background
The production of domestic air conditioner and refrigerator is over 70% of the global. The refrigerating equipment has high energy consumption, the electricity consumption of the refrigerating equipment accounts for more than 15 percent of the electricity consumption of the whole society of China, the annual average speed is increased by more than 20 percent, and the electricity consumption load of the large and medium city air conditioner accounts for about 60 percent of the peak load in summer. At the same time, refrigerants are also major contributors to the atmospheric environment, being major contributors to Ozone Depletion Potential (ODP) and Global Warming Potential (GWP). As the second major economy and 14 hundred million people of the world, energy conservation and emission reduction have become important national strategies, and energy utilization and environmental protection have been put in unprecedented important positions. China performs the national climate change frame convention and Paris protocol on International society commitment, and the emission reduction target and schedule are defined. The refrigeration equipment has the target tasks of energy conservation and emission reduction, so that the refrigeration equipment becomes a main battlefield and an important object of high efficiency and energy conservation.
The compressor, as a core component of the refrigeration equipment, like the heart of the equipment, has a direct impact on the overall performance of the refrigeration equipment. Because the compressor has a complex structure and a severe operating environment, various faults are easy to generate during long-term operation, and the whole refrigeration equipment is greatly influenced. Therefore, timely and reliable diagnosis of the fault of the compressor is of great importance.
The compressor in refrigeration equipment is usually totally enclosed, and the surface is 0.2mm thick steel plate, so that direct fault diagnosis is very difficult. An indirect diagnosis is generally performed using an external signal of the compressor, such as a vibration signal. With the development of deep learning technology, a large number of data driving algorithms emerge, and the fault diagnosis precision of the totally-enclosed compressor based on external signals is extremely improved. However, these algorithms are severely limited by the tag data, and in fact, in most cases the industry cannot provide sufficient tag data, and generally only no tag data. Therefore, practical application of the intelligent diagnosis method based on deep learning is greatly limited.
This lack of data is not present in all conditions, and in some cases the marker data is very readily available, as in laboratory conditions. However, the data acquired in such a case is different from the actual system, that is, the data distribution domain is different, so that the effect of training the diagnostic model by directly using such data is poor.
Disclosure of Invention
In view of the above problems in the prior art, an object of the present invention is to provide a knowledge sharing and model migration method for fault diagnosis of a totally enclosed compressor, which includes structural design of migration model, knowledge learning of feature vectors, and feature vector knowledge migration method based on countermeasure training, solving the dependence on the target compressor marker data, and using only a small amount of vibration signals of the target compressor, so-called migration training data.
The invention provides the following technical scheme: the fault diagnosis method of the totally-enclosed compressor based on knowledge sharing and model migration is provided; the method comprises the following steps:
step one, data preparation and pretreatment: collecting marked source domain data from a compressor with a known fault type, collecting unmarked target domain data from a compressor to be detected, training a source domain migration model and a target domain migration model, carrying out normalization processing on the source domain data and the target domain data, and bending one-dimensional data into two-dimensional data for input of a convolution layer;
step two, preparing a source domain migration model and a target domain migration model and training the source domain migration model and the target domain migration model independently: adopting a dense connection network as a backbone network of a source domain migration model and a target domain migration model, and encoding the data set acquired in the first step into a feature vector with a fixed length based on an encoder-decoder structure; respectively training a source domain migration model and a target domain migration model by utilizing the source domain data and the target domain data;
step three, constructing a discriminator, training the discriminator and retraining a target domain migration model:
301. constructing a discriminator: constructing a discriminator based on a convolutional neural network;
302. training a discriminator: loading a pre-trained source domain encoder and a pre-trained target domain encoder, connecting the pre-trained source domain encoder and the pre-trained target domain encoder to a discriminator, freezing parameters of a source domain migration model and a target domain migration model, training the discriminator by using source domain data and target domain data respectively, and optimizing the discriminator to accurately judge the data source;
303. retraining a target domain migration model: connecting the target domain migration model to a discriminator, freezing the parameters of the discriminator, activating the parameters of the target domain migration model, training the target domain migration model to enable the output of the discriminator to approach the source domain value, and forming an countermeasure training strategy with the step 302;
training a classifier: constructing a fault classifier of a full-connection network structure, loading a source domain migration model, and training the classifier by using the marked source domain data;
step five, fault diagnosis: connecting the trained classifier to a target domain migration model for fault diagnosis;
the data set collected by the target equipment is used as an input signal to be input into an encoder, the encoder encodes the data set into a feature vector and inputs the feature vector into a classifier, and the classifier directly outputs the fault type.
In the second step, a source domain migration model and a target domain migration model are constructed by using a dense connection network, wherein the input and output of the source domain migration model and the target domain migration model are data sets, a decoder reconstructs the data sets through feature vectors, and the feature vectors are feature knowledge representations of the data sets.
Further, in the third step, the discriminant and the encoder are trained by using the generated countermeasure training strategy, the model input of the discriminant model is a feature vector, the model is output as whether the model belongs to the source domain data, the final output layer of the model comprises a neural unit, the activation function uses softmax, and 0 or 1 indicates that the data belongs to the source domain or the target domain.
In the third step, the training process uses a gradient descent algorithm to gradually optimize network parameters, so that the target domain feature vector is finally close to the source domain diagnosis vector, and domain alignment is realized in the countermeasure strategy.
In the fifth step, the data set of the target compressor is collected in real time and input into the encoder, the encoder encodes the data set into a feature vector and inputs into the classifier, and the classifier directly outputs the fault type.
By adopting the technology, compared with the prior art, the invention has the following beneficial effects:
1) According to the method, the target domain equipment is not required to provide the data label, only part of the unmarked data is provided, and the data dependence degree is greatly reduced. The label of the source domain data can be acquired by the fault compressors obtained by implementation, and the label data of one type of compressors can be used for fault diagnosis of various compressors;
2) According to the invention, the parameter self-adjustment of the target domain migration model is realized by using the generated countermeasure training model, the approach of the source domain feature vector of the target domain feature vector distribution item is realized by using a gradient descent algorithm, and the model migration is realized;
3) According to the invention, the unsupervised feature learning is realized by using the deep-dense-connection self-encoder structure, so that the source domain migration model and the target domain migration model do not need to be marked with data in the training process, the requirement on the number of source domain labels is not great, and the data dependence degree is reduced again;
4) Aiming at the ultra-long time sequence data such as the vibration signal, the sequence length is usually more than 1000, the deep dense connection network structure can realize the efficient training of the ultra-long sequence data, and the data use efficiency can be greatly improved by folding the one-dimensional ultra-long data into the two-dimensional data structure.
Drawings
FIG. 1 is a schematic diagram of a source domain migration model and a target domain migration model according to the present invention;
FIG. 2 is a schematic diagram of the countermeasure training of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings and examples of the present invention. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
On the contrary, the invention is intended to cover any alternatives, modifications, equivalents, and variations as may be included within the spirit and scope of the invention as defined by the appended claims. Further, in the following detailed description of the present invention, certain specific details are set forth in order to provide a better understanding of the present invention. The present invention will be fully understood by those skilled in the art without the details described herein.
Referring to fig. 1-2, a fully-enclosed compressor fault diagnosis method based on knowledge sharing and model migration takes two types of compressors (R600A and R134A) of known fault types to collect vibration data as an example. There are three types of faults for compressors, including: the crank contacts the inner exhaust pipe, the supporting spring falls off, the air suction valve plate has foreign matters, the vibration signal at the top end of the compressor shell is collected at the sampling frequency of 10KHz, and every 1 second of data is used as a group of samples, namely one-dimensional time sequence data with the length of 10000. The R600A type compressor is used as a source domain data acquisition device, and the R134A type compressor is used as a target domain data acquisition device.
The method comprises the following steps:
step 101) data preparation: collecting vibration signals from equipment with known fault types as source domain data; and collecting vibration signals from the target equipment to be tested as target domain data. The difference is that the device types of the source domain and the target domain can be different, one type of data training model is realized, and a plurality of devices are applicable to the model.
Step 102) data preprocessing: and carrying out normalization processing on the vibration signal, and bending the one-dimensional data into two-dimensional data by utilizing a reshape function for input of a convolution layer.
Step 201) source domain migration model and target domain migration model preparation: the source domain migration model and the target domain migration model encode the vibration signal to a fixed length feature vector based on an encoder-decoder structure using a novel dense connection network as a backbone network. The decoder reconstructs the vibration signal from the feature vectors. The structure realizes the unsupervised learning of the vibration signal and reduces the dependence on the data label. Knowledge representation of the data is obtained by an unsupervised learning method.
Step 202) independently training a source domain migration model and a target domain migration model: training a source domain migration model and a target domain migration model by using source domain data and target domain data respectively, wherein 10000-length samples are used as a group of data, and reshape is formed into 100X100 two-dimensional data, 7500 groups of samples are used as a training data set, and 2500 groups of samples are used as a verification data set. Training round 300 rounds the optimizer uses selu for Adam and activation functions.
Step 301) constructing a discriminator: and constructing a discriminator based on the convolutional neural network, inputting the feature vector reshape into the discriminator, wherein the output layer is a single neural network, and 0 or 1 is used for representing input data as a source domain or a target domain. The arbiter uses the SGD optimizer, the activation function is selu, and the last layer activation function is softmax.
Step 302) training a arbiter: loading the pre-trained source domain encoder and the target domain encoder, connecting the pre-trained source domain encoder and the target domain encoder to the discriminator, freezing parameters of the source domain migration model and the target domain migration model, training the discriminator by utilizing the source domain data and the target domain data respectively, and optimizing the discriminator to accurately judge the data source.
Step 303) retraining the target domain migration model: connecting the target domain migration model to the discriminator, freezing the parameters of the discriminator, activating the parameters of the target domain migration model, training the target domain migration model, and enabling the output of the discriminator to gradually approach the source domain value, namely enabling the input data of the discriminator task to approach the source domain data. Steps 302 and 303 are repeated, and finally, the feature vector of the target domain migration model is sufficiently close to the source domain migration model.
Step 40) training a classifier: and constructing a fault classifier of the fully-connected network structure, wherein the input is a feature vector. And loading a source domain migration model, and training a classifier by using the marked source domain data.
Step 50) fault diagnosis: and connecting the trained classifier to a target domain migration model, and directly diagnosing the fault type by using a vibration signal acquired by target equipment as an input signal. Because the target domain migration model is subjected to countermeasure training, the distribution of the target domain migration model is close to that of the source domain migration model, and therefore fault diagnosis can be directly achieved.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (5)
1. The fault diagnosis method of the totally-enclosed compressor based on knowledge sharing and model migration is provided; the method is characterized by comprising the following steps of:
step one, data preparation and pretreatment: collecting marked source domain data from a compressor with a known fault type, collecting unmarked target domain data from a compressor to be detected, training a source domain migration model and a target domain migration model, carrying out normalization processing on the source domain data and the target domain data, and bending one-dimensional data into two-dimensional data for input of a convolution layer;
step two, preparing a source domain migration model and a target domain migration model and training the source domain migration model and the target domain migration model independently: adopting a dense connection network as a backbone network of a source domain migration model and a target domain migration model, and encoding the data set acquired in the first step into a feature vector with a fixed length based on an encoder-decoder structure; respectively training a source domain migration model and a target domain migration model by utilizing the source domain data and the target domain data;
step three, constructing a discriminator, training the discriminator and retraining a target domain migration model:
301. constructing a discriminator: constructing a discriminator based on a convolutional neural network;
302. training a discriminator: loading a pre-trained source domain encoder and a pre-trained target domain encoder, connecting the pre-trained source domain encoder and the pre-trained target domain encoder to a discriminator, freezing parameters of a source domain migration model and a target domain migration model, training the discriminator by using source domain data and target domain data respectively, and optimizing the discriminator to accurately judge the data source;
303. retraining a target domain migration model: connecting the target domain migration model to a discriminator, freezing the parameters of the discriminator, activating the parameters of the target domain migration model, training the target domain migration model to enable the output of the discriminator to approach the source domain value, and forming an countermeasure training strategy with the step 302;
training a classifier: constructing a fault classifier of a full-connection network structure, loading a source domain migration model, and training the classifier by using the marked source domain data;
step five, fault diagnosis: connecting the trained classifier to a target domain migration model for fault diagnosis;
the data set collected by the target equipment is used as an input signal to be input into an encoder, the encoder encodes the data set into a feature vector and inputs the feature vector into a classifier, and the classifier directly outputs the fault type.
2. The method for diagnosing faults of the totally-enclosed compressor based on knowledge sharing and model migration as claimed in claim 1, wherein in the second step, a source domain migration model and a target domain migration model are constructed by using a dense connection network of a high-efficiency neural network structure, the input and the output of the source domain migration model and the target domain migration model are data sets, a decoder reconstructs the data sets through feature vectors, and the feature vectors are feature knowledge representations of the data sets.
3. The method for diagnosing faults of a totally enclosed compressor based on knowledge sharing and model migration as claimed in claim 1, wherein in the third step, a discriminant and an encoder are trained by generating an countermeasure training strategy, model inputs of the discriminant model are feature vectors, outputs are whether the data belong to a source domain, a final output layer of the model contains a neural unit, and an activation function uses softmax to indicate that the data belong to the source domain or a target domain with 0 or 1.
4. The method for diagnosing faults of the totally-enclosed compressor based on knowledge sharing and model migration as claimed in claim 1, wherein in the third step, the training process uses a gradient descent algorithm to gradually optimize network parameters, finally the target domain feature vector is close to the source domain feature vector, and domain alignment is realized in an countermeasure strategy.
5. The fault diagnosis method for the totally-enclosed compressor based on knowledge sharing and model migration as claimed in claim 1, wherein in the fifth step, a data set of the target compressor is collected in real time and input into an encoder, the data set is encoded into a feature vector by the encoder and input into a classifier, and the classifier directly outputs the fault type.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010996926.6A CN112149726B (en) | 2020-09-21 | 2020-09-21 | Totally-enclosed compressor fault diagnosis method based on knowledge sharing and model migration |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010996926.6A CN112149726B (en) | 2020-09-21 | 2020-09-21 | Totally-enclosed compressor fault diagnosis method based on knowledge sharing and model migration |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112149726A CN112149726A (en) | 2020-12-29 |
| CN112149726B true CN112149726B (en) | 2024-02-09 |
Family
ID=73893640
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010996926.6A Active CN112149726B (en) | 2020-09-21 | 2020-09-21 | Totally-enclosed compressor fault diagnosis method based on knowledge sharing and model migration |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112149726B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113935406B (en) * | 2021-09-27 | 2023-05-16 | 苏州大学 | Mechanical equipment unsupervised fault diagnosis method based on countermeasure flow model |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109947086A (en) * | 2019-04-11 | 2019-06-28 | 清华大学 | Mechanical breakdown migration diagnostic method and system based on confrontation study |
| CN111060318A (en) * | 2020-01-09 | 2020-04-24 | 山东科技大学 | Bearing fault diagnosis method based on deep countermeasure migration network |
| CN111223553A (en) * | 2020-01-03 | 2020-06-02 | 大连理工大学 | Two-stage deep migration learning traditional Chinese medicine tongue diagnosis model |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109710636B (en) * | 2018-11-13 | 2022-10-21 | 广东工业大学 | Unsupervised industrial system anomaly detection method based on deep transfer learning |
-
2020
- 2020-09-21 CN CN202010996926.6A patent/CN112149726B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109947086A (en) * | 2019-04-11 | 2019-06-28 | 清华大学 | Mechanical breakdown migration diagnostic method and system based on confrontation study |
| CN111223553A (en) * | 2020-01-03 | 2020-06-02 | 大连理工大学 | Two-stage deep migration learning traditional Chinese medicine tongue diagnosis model |
| CN111060318A (en) * | 2020-01-09 | 2020-04-24 | 山东科技大学 | Bearing fault diagnosis method based on deep countermeasure migration network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112149726A (en) | 2020-12-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yan et al. | Semi-supervised learning for early detection and diagnosis of various air handling unit faults | |
| Li et al. | Self-attention ConvLSTM and its application in RUL prediction of rolling bearings | |
| CN111275288B (en) | XGBoost-based multidimensional data anomaly detection method and device | |
| CN111709448A (en) | Mechanical fault diagnosis method based on migration relation network | |
| Li et al. | Joint attention feature transfer network for gearbox fault diagnosis with imbalanced data | |
| CN112149726B (en) | Totally-enclosed compressor fault diagnosis method based on knowledge sharing and model migration | |
| Zhang et al. | Imbalanced data based fault diagnosis of the chiller via integrating a new resampling technique with an improved ensemble extreme learning machine | |
| CN113902104A (en) | Non-invasive load monitoring method combining unsupervised domain self-adaptive strategy and attention mechanism | |
| Lv et al. | Deep transfer network with multi-kernel dynamic distribution adaptation for cross-machine fault diagnosis | |
| Zucker et al. | Improving energy efficiency of buildings using data mining technologies | |
| Liu et al. | Wind turbine anomaly detection based on SCADA: A deep autoencoder enhanced by fault instances | |
| CN107403244A (en) | Short-term wind speed forecasting method based on artificial neural network updated gray correlation analysis | |
| Pang et al. | Discrete cosine transformation and temporal adjacent convolutional neural network-based remaining useful life estimation of bearings | |
| CN113076920B (en) | Intelligent fault diagnosis method based on asymmetric domain confrontation self-adaptive model | |
| CN110850717B (en) | Neural network heat pump defrosting control device and control method by utilizing fan current | |
| CN115791174B (en) | Rolling bearing abnormality diagnosis method, system, electronic equipment and storage medium | |
| Cheng et al. | Fault detection and isolation for chiller system based on deep autoencoder | |
| CN116412162A (en) | Digital twin model-based magnetic suspension blower fault diagnosis method and system | |
| CN115965156A (en) | Scheduling method and scheduling device of energy system | |
| CN112413814B (en) | Online renewable heating ventilation air conditioner sensor fault detection method based on comprehensive distance | |
| CN114781507B (en) | 1 DCNN-DS-based water chilling unit fault diagnosis method | |
| CN111814396A (en) | Centrifugal fan fault early warning method based on transfer learning | |
| Jiang et al. | A hybrid degradation tendency measurement method for mechanical equipment based on moving window and Grey–Markov model | |
| Yu et al. | Intelligent fault diagnosis of chemical process based on twin convolution deep neural network of wavelet transform | |
| Tian et al. | Few-shot condition diagnosis of rolling bearing using adversarial transfer network with class aggregation-guided |
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 |