CN112364991B - LSTM-E bearing fault recognition model training method - Google Patents
LSTM-E bearing fault recognition model training method Download PDFInfo
- Publication number
- CN112364991B CN112364991B CN202011190307.4A CN202011190307A CN112364991B CN 112364991 B CN112364991 B CN 112364991B CN 202011190307 A CN202011190307 A CN 202011190307A CN 112364991 B CN112364991 B CN 112364991B
- Authority
- CN
- China
- Prior art keywords
- lstm
- data
- layer
- bearing
- 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
- 238000000034 method Methods 0.000 title claims abstract description 22
- 238000012549 training Methods 0.000 title claims abstract description 21
- 238000003745 diagnosis Methods 0.000 claims abstract description 16
- 238000013528 artificial neural network Methods 0.000 claims abstract description 10
- 238000002372 labelling Methods 0.000 claims abstract description 7
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000013480 data collection Methods 0.000 claims abstract description 4
- 238000004519 manufacturing process Methods 0.000 claims abstract description 4
- 238000012545 processing Methods 0.000 claims abstract description 4
- 230000004913 activation Effects 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 15
- 210000002569 neuron Anatomy 0.000 claims description 12
- 210000004027 cell Anatomy 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 7
- 238000012795 verification Methods 0.000 claims description 7
- 238000005096 rolling process Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000009795 derivation Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 description 6
- 210000005056 cell body Anatomy 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 230000005764 inhibitory process Effects 0.000 description 2
- 230000003213 activating effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000009849 deactivation Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a training method for an LSTM-E bearing fault identification model, which comprises the following steps: s1, data collection: bearing frequency data under the condition of various parameters in a public bearing data set of Kassi university is collected in batches; s2, manufacturing and processing a data set: preprocessing and standardizing the collected bearing frequency data respectively, and labeling corresponding data labels; s3, establishing an LSTM-E diagnosis model: establishing a local diagnosis model for the processed multiple bearing frequency data by using an LSTM-E neural network; s4, model training discrimination: the data is fitted layer by layer based on the LSTM principle and the multi-layer perceptron model, a loss function and an optimizer are constructed, the weight is continuously updated, probability value estimation is carried out on the diagnosis result output by the local diagnosis model, and a final discrimination result is obtained; s5, verifying a data set: model performance is verified on CWRU bearing data sets, and the prediction accuracy reaches 100%, so that the problem that the existing LSTM bearing fault recognition model based on the neural network is low in prediction accuracy is solved.
Description
Technical Field
The invention relates to the technical field of vibration detection, in particular to an LSTM-E bearing fault identification model training method.
Background
Bearings are the core load-bearing parts of the machinery that carry the rotating parts, typically rolling bearings. During operation of the machine, the variable impact load may cause bearing failure problems, and as the machine is used, the severity of the aforementioned failure increases gradually and may cause cascading failures; because of this, it is important to discover bearing failure early.
Currently, methods for diagnosing bearing faults include a time domain analysis method, a frequency domain analysis method and a deep learning method; the time domain analysis method is used for carrying out statistical analysis on the time domain waveform of the measured vibration signal to obtain a starting statistical index; the time domain statistics indexes comprise a maximum value, a root mean square value, a kurtosis index, a peak value index and the like of the signal; because the vibration signal characteristics of the micro faults are very weak, fault information and fault characteristics cannot be effectively mined by adopting time domain statistics indexes; the frequency domain analysis method is to process the vibration signal by Fourier transform, hilbert transform and other methods to obtain a frequency domain signal, and then to identify the fault type by power spectrum analysis, cepstrum analysis or spectrum kurtosis analysis.
The existing deep learning method is used for predicting an LSTM bearing fault recognition model based on a neural network, but the accuracy of the prediction method is low, so that a more accurate bearing fault recognition model training method is provided.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: how to solve the problem that the existing LSTM bearing fault recognition model based on the neural network predicts, but the accuracy of the prediction method is lower, and provides a training method for the LSTM-E bearing fault recognition model.
The invention solves the technical problems through the following technical proposal, and the invention comprises the following steps:
s1, data collection: bearing frequency data under the condition of various parameters in a public bearing data set of Kassi university is collected in batches;
S2, manufacturing and processing a data set: preprocessing and standardizing the collected bearing frequency data respectively, and labeling corresponding data labels;
S3, establishing an LSTM-E diagnosis model: establishing a local diagnosis model for the processed multiple bearing frequency data by using an LSTM-E neural network;
S4, model training discrimination: the data is fitted layer by layer based on the LSTM principle and the multi-layer perceptron model, a loss function and an optimizer are constructed, the weight is continuously updated, probability value estimation is carried out on the diagnosis result output by the local diagnosis model, and a final discrimination result is obtained;
S5, verifying a data set: model performance was verified on CWRU bearing dataset.
In step S1, bearing frequency data under the condition of disclosing a plurality of different parameters in bearing data sets of kesixi university are collected in batches, wherein the conditions are respectively that under the fixed premise that the approximate rotating speed of a motor is 1797r/min and the load of the motor is 0, an inner ring fault is respectively taken when the fault diameter is 0.1778mm, 0.3556mm and 0.5334mm, and the relative positions of a rolling body fault and the outer ring meet the bearing frequency data of a region.
Further, in the step S2, the collected bearing frequency data is preprocessed and standardized, and corresponding data labels are labeled, which specifically includes the following steps:
s2-1, respectively setting 2 signal periods for the bearing frequency data collected in the S1 under different parameters, wherein 10 types of signals are generated in each signal period, and 1000 types of data are generated in each type;
S2-2, labeling a label of a corresponding signal;
s2-3, performing Z-score standardization on a data set based on the mean value and standard deviation of original data, compressing all data to be near 0, and facilitating reverse derivation self-updating calculation in a section with variance of 1;
s2-4, sampling the training set and the verification set and carrying out forward data enhancement;
S2-5, setting a training set, a verification set and a test set to be 70%, 20% and 10% respectively.
Further, in the steps S3 and S4, the LSTM-E bearing fault recognition model is built by keras frames, and the network structure is a 6-layer structure: the method comprises the steps of inputting layers, 16 stacked LSTM-e layers, an activation layer formed by relu activation functions, a full-connection layer formed by 32 perceptrons, a dropoff layer with the inhibition rate of 0.2 and a softmax classified activation layer, wherein each LSTM-e layer comprises 32 LSTM memory cells, the same data are received in parallel, each neuron outputs with different weights, each LSTM memory neuron output result of each layer is accumulated to solve a product-mean and activated by tanH functions to obtain an LSTM-e layer output result, the output result of the layer is used as the input of the next LSTM-e, the last LSTM-e layer is activated by relu activation functions in the activation layer, the dropoff layer enables 20% of perceptrons to be randomly deactivated through the extraction characteristics of the full-connection layer of the 32 perceptrons, the output result is the input of the softmax classified layer, and finally the model output result is judged by the softmax classified layer to be the maximum.
Further, 10% of the test set data already divided in S2-5 is input to the LSTM-E model.
Compared with the prior art, the invention has the following advantages:
1. The verification accuracy of the application under the open bearing data set of Kassi university can reach 100%.
2. The neural network layer number in the network model is refined, and the parameter quantity is small.
3. The data set is characterized in that limited sample data is subjected to data preprocessing, and beneficial noise (Gaussian white noise) is added, so that a trained network has certain generalization capability.
4. The Z-Score-lg converts sample data into a unitless Z-Score value, so that the data standards are unified, the data comparability is improved, and the data interpretation is weakened.
Drawings
Fig. 1 is an overall construction diagram of the present invention.
Detailed Description
The following describes in detail the examples of the present invention, which are implemented on the premise of the technical solution of the present invention, and detailed embodiments and specific operation procedures are given, but the scope of protection of the present invention is not limited to the following examples.
As shown in fig. 1, this embodiment provides a technical solution: a training method for an LSTM-E bearing fault recognition model comprises the following steps:
s1, data collection: bearing frequency data under the condition of various parameters in a public bearing data set of Kassi university is collected in batches;
S2, manufacturing and processing a data set: preprocessing and standardizing the collected bearing frequency data respectively, and labeling corresponding data labels;
S3, establishing an LSTM-E diagnosis model: establishing a local diagnosis model for the processed multiple bearing frequency data by using an LSTM-E neural network;
S4, model training discrimination: the data is fitted layer by layer based on the LSTM principle and the multi-layer perceptron model, a loss function and an optimizer are constructed, the weight is continuously updated, probability value estimation is carried out on the diagnosis result output by the local diagnosis model, and a final discrimination result is obtained;
s5, verifying a data set: the recognition accuracy can reach 100% on CWRU bearing data sets.
In the step S1, bearing frequency data under the condition that a plurality of different parameters are disclosed in bearing data sets of Kasixi university are collected in batches, namely, under the condition that the approximate rotating speed of a motor is 1797r/min and the load of the motor is (horsepower) 0, bearing frequency data of an inner ring fault, a rolling body fault and an outer ring relative position coincidence zone (the central position is in the 6-point direction) are respectively taken when the fault diameters are 0.1778mm, 0.3556mm and 0.5334 mm.
In the step S2, the collected bearing frequency data are preprocessed and standardized, and corresponding data labels are labeled, which specifically includes the following steps:
s2-1, respectively setting 2 signal periods for the bearing frequency data collected in the S1 under different parameters, wherein 10 types of signals are generated in each signal period, and 1000 types of data are generated in each type;
S2-2, labeling a label (one-hot coding) of a corresponding signal;
S2-3, performing Z-Score-lg standardization (Z-Score-lg: performing lg (base 10) operation on a sample data tensor twice, namely, performing x=lg (x) operation, and normalizing the data standard by using (x-mean)/variance as an output result, (wherein the Z-Score converts (x-mu)/sigma sample data into a unitless Z-Score value, so that the data standardization is unified, the data comparability is improved, and the data interpretation is weakened), wherein the processed data accords with standard normal distribution, namely, the mean value is 0, the standard deviation is 1, and the conversion function is:
wherein mu is the mean value of all sample data, sigma is the standard deviation of all sample data, all data are compressed near 0, and the interval with variance of 1 is convenient for reverse derivation self-updating calculation;
s2-4, sampling the training set and the verification set and carrying out forward data enhancement;
S2-5, setting a training set, a verification set and a test set to be 70%, 20% and 10% respectively.
In the steps S3 and S4, the LSTM-E bearing fault recognition model is built by keras frames, and the network structure is a 6-layer structure: the method comprises the steps of inputting layers, 16 stacked LSTM-e layers, an activation layer formed by relu activation functions, a full-connection layer formed by 32 perceptrons, a dropoff layer with the inhibition rate of 0.2 and a softmax classified activation layer, wherein each LSTM-e layer comprises 32 LSTM memory cells, the same data are received in parallel, each neuron outputs with different weights, each LSTM memory neuron output result of each layer is accumulated to solve a product-mean and activated by tanH functions to obtain an LSTM-e layer output result, the output result of the layer is used as the input of the next LSTM-e, the last LSTM-e layer is activated by relu activation functions in the activation layer, the dropoff layer enables 20% of perceptrons to be randomly deactivated through the extraction characteristics of the full-connection layer of the 32 perceptrons, the output result is the input of the softmax classified layer, and finally the model output result is judged by the softmax classified layer to be the maximum. The functional model expression of the part is:
An input door:
Forgetting the door:
Cell body:
Output door:
one cell output:
Wherein,
1. The input of the neuron is denoted a and the output is denoted b;
2.w ij denotes the connection weight from neuron i to j;
3. Subscripts iota, phi and omega respectively represent an input door, a forget door and an output door;
the c subscript indicates the cell body, and the peephole (structural variant) weights from the cell body to the input, forget, output gates are denoted as w cι, And w cω;
5.s c represents the state of cell c;
6. the activation function of the control gate is expressed by f, g and h respectively express the input and output activation functions of lstm cells;
7.I denotes the number of neurons in the input layer, K is the number of neurons in the output layer, and H is the number of hidden layer cells.
Parallel input/output of 32 neurons (each neuron multiplied by a corresponding random weight to get the output of one LSTM-e layer):
ki∈(0，1)
the y (output) of the LSTM-e of the first layer is used as the x (input) layer of the LSTM-e of the next layer, the y of the last layer is obtained through 16 LSTM-e layers and the relu (x), and then the y is used as the x of the full connection layer and the full connection layer is used;
y＝w*x+b
Obtaining y of the full connection layer, and passing through a dropout layer:
Dropout mathematical model (0.2 deactivation rate):
Activating through a softmax layer, and predicting the maximum probability as an output result;
y＝arg max(p(y＝c|x))；
And (3) inputting 10% of the test set data already divided in S2-5 (the input of the x sample is changed into 10% of the divided test set data) into the LSTM-E model, so that the accuracy of an output result reaches 100%.
Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include at least one such feature. In the description of the present invention, the meaning of "plurality" means at least two, for example, two, three, etc., unless specifically defined otherwise.
In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, the different embodiments or examples described in this specification and the features of the different embodiments or examples may be combined and combined by those skilled in the art without contradiction.
While embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the invention, and that variations, modifications, alternatives and variations may be made to the above embodiments by one of ordinary skill in the art within the scope of the invention.
Claims (4)
1. The LSTM-E bearing fault recognition model training method is characterized by comprising the following steps of:
s1, data collection: bearing frequency data under the condition of various parameters in a public bearing data set of Kassi university is collected in batches;
S2, manufacturing and processing a data set: preprocessing and standardizing the collected bearing frequency data respectively, and labeling corresponding data labels;
S3, establishing an LSTM-E diagnosis model: establishing a local diagnosis model for the processed multiple bearing frequency data by using an LSTM-E neural network;
S4, model training discrimination: the data is fitted layer by layer based on the LSTM principle and the multi-layer perceptron model, a loss function and an optimizer are constructed, the weight is continuously updated, probability value estimation is carried out on the diagnosis result output by the local diagnosis model, and a final discrimination result is obtained;
S5, verifying a data set: verifying model performance on CWRU bearing datasets;
In the steps S3 and S4, the LSTM-E bearing fault recognition model is built by keras frames, the network structure is a 6-layer structure, LSTM-E is a model name of the neural network, LSTM-E is a layer name of the neural network, and the network structure includes: the method comprises the steps of inputting layers, 32 stacked LSTM-e layers, an activation layer formed by relu activation functions, a fully-connected layer formed by 32 perceptrons, a dropoff layer with a rejection rate of 0.2 and a softmax classified activation layer, wherein each LSTM-e layer comprises 16 LSTM memory cells, the same data are received in parallel, each neuron outputs with different weights, each LSTM memory neuron output result of each layer is accumulated to solve a product-mean and activated by tanH functions to obtain an LSTM-e layer output result, the output result of the layer is used as the input of the next LSTM-e layer, the last LSTM-e layer is activated by relu activation functions in the activation layer, the dropoff layer enables 20% of perceptrons to be randomly deactivated through the extraction characteristics of the fully-connected layer of the 32 perceptrons, the output result is the input of the softmax classified layer, and finally the output result of the softmax classified layer is judged to be the maximum normalized output result of the model by the softmax classified layer.
2. The LSTM-E bearing failure recognition model training method of claim 1, wherein: in the step S1, bearing frequency data under the condition that a plurality of different parameters are disclosed in bearing data sets of Kasixi university are collected in batches, namely bearing frequency data of an inner ring fault, a rolling body fault and an outer ring relative position coincidence zone are respectively taken under the condition that the fault diameter is 0.1778mm, 0.3556mm and 0.5334mm on the premise that the approximate rotating speed of a motor is 1797r/min and the motor load is 0.
3. The LSTM-E bearing failure recognition model training method of claim 2, wherein: in the step S2, the collected bearing frequency data are preprocessed and standardized, and corresponding data labels are labeled, which specifically includes the following steps:
s2-1, respectively setting 2 signal periods for the bearing frequency data collected in the S1 under different parameters, wherein 10 types of signals are generated in each signal period, and 1000 types of data are generated in each type;
S2-2, labeling a label of a corresponding signal;
s2-3, performing Z-score standardization on a data set based on the mean value and standard deviation of original data, compressing all data to be near 0, and facilitating reverse derivation self-updating calculation in a section with variance of 1;
s2-4, sampling the training set and the verification set and carrying out forward data enhancement;
S2-5, setting a training set, a verification set and a test set to be 70%, 20% and 10% respectively.
4. A method for training a LSTM-E bearing failure recognition model according to claim 3, wherein: 10% of the test set data already partitioned in S2-5 is input to the LSTM-E model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011190307.4A CN112364991B (en) | 2020-10-30 | 2020-10-30 | LSTM-E bearing fault recognition model training method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011190307.4A CN112364991B (en) | 2020-10-30 | 2020-10-30 | LSTM-E bearing fault recognition model training method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112364991A CN112364991A (en) | 2021-02-12 |
| CN112364991B true CN112364991B (en) | 2024-05-10 |
Family
ID=74513311
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011190307.4A Active CN112364991B (en) | 2020-10-30 | 2020-10-30 | LSTM-E bearing fault recognition model training method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112364991B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112988714B (en) * | 2021-04-06 | 2023-07-07 | 中车青岛四方机车车辆股份有限公司 | Bearing fault database construction method, analysis method and system |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111832228A (en) * | 2020-07-17 | 2020-10-27 | 中国人民解放军海军工程大学 | Vibration transmission system based on CNN-LSTM |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108388860B (en) * | 2018-02-12 | 2020-04-28 | 大连理工大学 | Aero-engine rolling bearing fault diagnosis method based on power entropy spectrum-random forest |
-
2020
- 2020-10-30 CN CN202011190307.4A patent/CN112364991B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111832228A (en) * | 2020-07-17 | 2020-10-27 | 中国人民解放军海军工程大学 | Vibration transmission system based on CNN-LSTM |
Non-Patent Citations (5)
| Title |
|---|
| "Deep reinforcement learning and LSTM for optimal renewable energy accommodation in 5G internet of energy with bad data tolerant";Lin Lin 等;《Computer Communications》;全文 * |
| "基于信息融合的柴油机热工故障诊断技术研究";杨世雄;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;全文 * |
| "基于柔性曲轴多体动力学模型的曲轴优化设计";秦利明 等;《石河子大学学报(自然科学版)》;第38卷(第1期);全文 * |
| "基于注意力机制BiLSTM的设备智能故障诊断方法";王太勇 等;《天津大学学报(自然科学与工程技术版)》;第53卷(第6期);第601-607页 * |
| "基于深度学习的故障诊断方法综述";文成林 等;《电子与信息学报》;第42卷(第1期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112364991A (en) | 2021-02-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Han et al. | An adaptive spatiotemporal feature learning approach for fault diagnosis in complex systems | |
| Cheng et al. | A convolutional neural network based degradation indicator construction and health prognosis using bidirectional long short-term memory network for rolling bearings | |
| Qin et al. | Gated dual attention unit neural networks for remaining useful life prediction of rolling bearings | |
| Li et al. | Understanding and improving deep learning-based rolling bearing fault diagnosis with attention mechanism | |
| CN108614548B (en) | Intelligent fault diagnosis method based on multi-mode fusion deep learning | |
| CN109102005B (en) | Small sample deep learning method based on shallow model knowledge migration | |
| He et al. | Deep variational autoencoder classifier for intelligent fault diagnosis adaptive to unseen fault categories | |
| CN107657250B (en) | Bearing fault detection and positioning method and detection and positioning model implementation system and method | |
| CN109781411B (en) | Bearing fault diagnosis method combining improved sparse filter and KELM | |
| Wan et al. | Bearing remaining useful life prediction with convolutional long short-term memory fusion networks | |
| US20230351177A1 (en) | Enhanced discriminate feature learning deep residual cnn for multi-task rotating machinery fault diagnosis with information fusion | |
| Ding et al. | A novel remaining useful life prediction method of rolling bearings based on deep transfer auto-encoder | |
| CN110781592A (en) | Method and system for predicting residual life of equipment | |
| Liu et al. | LSTM-GAN-AE: A promising approach for fault diagnosis in machine health monitoring | |
| CN111459144A (en) | Airplane flight control system fault prediction method based on deep cycle neural network | |
| Jiang et al. | Dual-attention-based multiscale convolutional neural network with stage division for remaining useful life prediction of rolling bearings | |
| Ma et al. | Multi-vibration information fusion for detection of HVCB faults using CART and D–S evidence theory | |
| Li et al. | Parallel multi-fusion convolutional neural networks based fault diagnosis of rotating machinery under noisy environments | |
| Lu et al. | Joint training of a predictor network and a generative adversarial network for time series forecasting: A case study of bearing prognostics | |
| CN115859077A (en) | Multi-feature fusion motor small sample fault diagnosis method under variable working conditions | |
| Yao et al. | Improving rolling bearing fault diagnosis by DS evidence theory based fusion model | |
| Joshuva et al. | A study of various blade fault conditions on a wind turbine using vibration signals through histogram features | |
| CN112364991B (en) | LSTM-E bearing fault recognition model training method | |
| Mishra et al. | Fault detection of elevator systems using multilayer perceptron neural network | |
| CN110991471A (en) | Fault diagnosis method for high-speed train traction system |
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 |