CN115081468A - Multi-task convolutional neural network fault diagnosis method based on knowledge migration - Google Patents
Multi-task convolutional neural network fault diagnosis method based on knowledge migration Download PDFInfo
- Publication number
- CN115081468A CN115081468A CN202110276577.5A CN202110276577A CN115081468A CN 115081468 A CN115081468 A CN 115081468A CN 202110276577 A CN202110276577 A CN 202110276577A CN 115081468 A CN115081468 A CN 115081468A
- Authority
- CN
- China
- Prior art keywords
- knowledge
- grained
- coarse
- fine
- neural network
- 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
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000003745 diagnosis Methods 0.000 title claims abstract description 27
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 24
- 238000013508 migration Methods 0.000 title claims abstract description 16
- 230000005012 migration Effects 0.000 title claims abstract description 16
- 238000012546 transfer Methods 0.000 claims abstract description 9
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 23
- 239000011159 matrix material Substances 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 230000000750 progressive effect Effects 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 3
- 239000002904 solvent Substances 0.000 claims description 3
- 230000003595 spectral effect Effects 0.000 claims description 3
- 238000013526 transfer learning Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 2
- 238000013135 deep learning Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 4
- 238000005034 decoration Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000002405 diagnostic procedure Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- ULXYQAJWJGLCNR-YUMQZZPRSA-N Leu-Asp-Gly Chemical compound CC(C)C[C@H](N)C(=O)N[C@@H](CC(O)=O)C(=O)NCC(O)=O ULXYQAJWJGLCNR-YUMQZZPRSA-N 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 238000007405 data analysis Methods 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
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000001228 spectrum 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
- G06N3/084—Backpropagation, e.g. using gradient descent
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)
- Image Analysis (AREA)
Abstract
The invention discloses a multi-task convolutional neural network fault diagnosis method based on knowledge migration, which comprises the following steps of: step 1: preprocessing data; step 2: learning from a coarse structure to a fine structure; and 3, step 3: multitasking is migrated from coarse to fine knowledge. The invention shows significant advantages in large scale fault diagnosis. Since a good initialization of the CNN parameters obtained from the coarse-grained task can effectively avoid poor local minima. Meanwhile, effective judgment information is reserved and transmitted to a fine-grained task to realize effective fault identification, and compared with a flat CNN, the PKT-MCNN provided by the application converges to a better local minimum value, so that the remarkable influence of gradual knowledge transfer on the learning of the CNN is verified.
Description
Technical Field
The invention relates to the technical field of intelligent fault diagnosis, in particular to a multi-task convolutional neural network fault diagnosis method based on knowledge migration.
Background
Fault diagnosis is becoming increasingly important in modern society as an effective tool to maintain safe operation of industrial systems and reduce maintenance costs of unnecessary routine shutdowns. Therefore, various diagnostic methods have been proposed to detect faults early and accurately. In recent years, with the development of sensors and information technology, industry data is rapidly accumulated, and Deep Learning (DL) based diagnostic methods are being promoted. Based on a deep framework and a plurality of nonlinear layers, the DL algorithm can adaptively learn high-level representation characteristics, and the inherent defects of the traditional diagnosis method are overcome.
Most deep learning based methods work well for small numbers of fault diagnoses, but fail to converge to satisfactory results when dealing with large scale fault diagnoses, because the large number of fault types can lead to inter-class distance imbalances and local minima in the neural network. For example, in a large-scale diagnosis task for handling a large number of fault types, the DL-based diagnosis method has a drawback that, first, the DL-based method tends to fall into a local minimum when random initialization is performed, and those methods that need to handle a large number of fault types may aggravate the problem due to an increase in parameters and a complication of the structure. Furthermore, as the label space increases, the upper bound of the generalization error increases, resulting in a decrease in the final diagnostic performance, and, from the perspective of algorithm design, an increase in the fault category may result in the problem of imbalance of the intra-class/inter-class distances, i.e., large intra-class distances and small inter-class distances of faults. For example, the distance between similar failures of one component is small and difficult to distinguish, while the distance between failures of different subsystems is large and can be easily classified. Therefore, in such large-scale fault diagnosis tasks, the inter-class distance of some similar faults may be even smaller than the intra-class distance of other faults, which makes the conventional deep learning method difficult to apply. Therefore, large-scale fault diagnosis of complex systems with multiple fault types has been a difficult problem to crack.
Disclosure of Invention
The application provides a multi-task convolutional neural network fault diagnosis method based on knowledge migration, which can overcome the current defects of fault diagnosis of a large complex system, and learns fault information from coarse to fine into a network by using known fault information in a migration learning mode, so that better performance is obtained.
The invention provides a multi-task convolutional neural network fault diagnosis method based on knowledge migration,
the method comprises the following steps:
step 1: preprocessing data;
step 2: learning from a coarse structure to a fine structure;
and step 3: multitasking is migrated from coarse to fine knowledge.
Preferably, step 1 specifically comprises: and processing the input sample signal to obtain an input sample matrix.
Preferably, step 2 specifically comprises:
step 21: similar graph construction for fault types;
step 22: spectral clustering of fault types;
step 23: and outputting the knowledge structure.
Preferably, steps 21 and 22 specifically include: and clustering similar fault types to form coarse-grained knowledge.
Preferably, step 23 specifically includes:
step 231: training a neural network for extracting coarse-grained knowledge;
step 232: extracting coarse grain knowledge;
step 233: obtaining a network with coarse-grained knowledge extracted;
step 244: and transferring the knowledge of the network to a fine-grained network through transfer learning.
Preferably, step 3 specifically comprises:
step 31: constructing a multitask convolution neural network based on progressive knowledge transfer;
step 32: training PKT-MCNN model transfer by using progressive knowledge;
step 33: testing a new sample by using the trained PKT-MCNN model;
step 34: and outputting the fault type of the test sample.
Preferably, in step 3, the PKT-MCNN training process includes three stages:
(1) training a coarse-grained task to master coarse-grained knowledge;
(2) simultaneously training coarse-grained and fine-grained tasks in a multi-task mode;
(4) fine-grained tasks are trained individually by fine-tuning the parameters updated in stage (2).
Preferably, in step 3, the PKT algorithm is embedded on top of the multitasking CNN,
the PKT adjusts the different attentions for the two tasks, the training coarse-grained task and the training fine-grained task by a weight parameter lambda,
in particular, the amount of the solvent to be used,
in the first phase, λ is set to 1, learning only coarse-grained tasks, since the weight of the fine-grained tasks is 0, the loss stops from back propagation;
in the second stage, the lambda is gradually reduced to 0 according to the number of training epochs, so that the two tasks are learned simultaneously, and coarse-grained knowledge is gradually transferred to fine-grained knowledge; second stage λ, noted λ 2 Calculated from the following equation:
b1 and B3 are the training epochs of the first stage and the second stage respectively; b is the current epoch number, Bmax is the maximum epoch number;
finally, the value of λ at the third stage is opposite to the value of λ at the first stage, and λ is set to 0 to fine tune the fine-grained task without updating the parameters of the coarse-grained task.
Compared with the prior art, the invention has the advantages that,
the invention shows significant advantages in large scale fault diagnosis. Since a good initialization of the CNN parameters obtained from the coarse-grained task can effectively avoid poor local minima. Meanwhile, effective judgment information is reserved and transmitted to a fine-grained task to realize effective fault identification, and compared with a flat CNN, the PKT-MCNN provided by the application converges to a better local minimum value, so that the remarkable influence of gradual knowledge transfer on the learning of the CNN is verified.
The method can intelligently learn a reasonable knowledge structure consistent with the physical composition of the nuclear power system, is very good at handling large-scale fault diagnosis under a special physical background, and extracts and transfers coarse-grained knowledge to a final fine-grained diagnosis task. The method provided by the application can become a promising tool in future industrial big data analysis research.
Drawings
FIG. 1 is a flow chart of a method of the present application;
FIG. 2 is a flow chart of a coarse-to-fine structure learning method of the present application;
FIG. 3 is a diagram of the PKT-MCNN model architecture of the present application.
Detailed Description
It should be noted that, in the present application, the embodiments and features of the embodiments may be combined with each other without conflict.
The invention is described in further detail below with reference to the figures and specific examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As shown in fig. 1, a method for diagnosing a fault of a multitask convolutional neural network based on knowledge migration according to an embodiment of the present invention includes the following steps:
step 1: preprocessing data;
the input signal refers to an original signal of an input device, and is usually vibration data, and the input signal is processed by means of noise reduction, spectrum extraction, and the like to obtain a matrix, which is input to a subsequent algorithm and is called an input sample matrix.
And 2, step: learning from a coarse structure to a fine structure;
as shown in fig. 2, the method specifically comprises the following steps:
step 21: similar graph construction for fault types;
in the present embodiment, the construction of the failure type similarity graph adopts the following method:
to automatically extract the coarse-grained knowledge structure, the information in the data is used to group similar fault types into one common superior node of the structure, which represents the coarse-grained concept of a fault.
Give the jth sample of the ith fault
Similarity graph G ═ V, E for the error types contains similarity information for each of the two fault types, where each vertex V is a vertex i e.V is a failure type, and each edge e ij E is two connected vertices v i And v j The similarity between them. To construct this graph, the similarity of each pair of fault types should be calculated. Unlike instance-level clustering, each fault type needs to be represented before calculating the similarity. In this work, each fault type is vectorized and represented by the centroid of all its samples. Mathematically, the vertex v in the similarity graph i Is shown as
Wherein M is i The number of samples of the ith fault. This representation has two advantages for large scale fault diagnosis. On the one hand, for many fault samples, it is efficient to compute the first order statistics, i.e. the mean vector. On the other hand, representing the failure type with a mean vector may mitigate the adverse effects of noise samples that are typically present in large data sets. Through the expression of the vertex, the similarity information e between the two fault types i and j can be calculated through Gaussian similarity ij . The formula is as follows:
wherein σ is the sum of e ij Value of (A) to (B)To be [0,1]The scaling factor of (c).
Step 22: spectral clustering of fault types;
using the similarity graph G, the fault types are assigned to different coarse-grained fault concepts by clustering techniques. The method adopts a normalized cut (NCut) algorithm to cut the graph into a plurality of subgraphs, and can formally optimize the following objective functions:
s·t·H'DH＝I
in the formula, k is the number of the coarse-grained fault concepts to be clustered. If vi ∈ Cj, then h ij ＝1/|C j I, otherwise, h ij ＝0，h ij Are elements in the matrix H. i ∈ {1,2, ·, N }; j ∈ {1,2, ·, k }; L-D-G is a laplacian matrix; d is a secondary matrix of G; i is the identity matrix, |, is the cardinality of the set, and Shi and Malik demonstrate that this goal can be approximated with the eigenvector of L associated with the second smallest eigenvalue. Thus, k coarse-grained fault concepts C 1 ,C 2 ，..，C k And constructing a two-level knowledge structure T, wherein each coarse-grained concept comprises a plurality of fine-grained fault types.
And clustering similar fault types through the two steps to form coarse-grained knowledge.
Step 23: and outputting the knowledge structure.
Wherein, step 23 specifically comprises the following steps:
step 231: training a neural network for extracting coarse-grained knowledge;
step 232: extracting coarse grain knowledge;
step 233: obtaining a network with coarse-grained knowledge extracted;
step 244: and transferring the knowledge of the network to a fine-grained network through transfer learning.
And step 3: multitasking is migrated from coarse to fine knowledge.
The method specifically comprises the following steps:
step 31: constructing a multitask convolution neural network based on progressive knowledge transfer;
step 32: training PKT-MCNN model transfer by using progressive knowledge;
step 33: testing a new sample by using the trained PKT-MCNN model;
step 34: and outputting the fault type of the test sample.
In a preferred embodiment, as shown in fig. 3, for the architecture diagram of the PKT-MCNN model of the present application, in step 3, the training process of the PKT-MCNN includes three stages:
(1) training a coarse-grained task to master coarse-grained knowledge;
(2) simultaneously training coarse-grained and fine-grained tasks in a multi-task mode;
(3) fine-grained tasks are trained individually by fine-tuning the parameters updated in stage (2).
The PKT controls and switches the stage through the weight of the loss layer, and the weight represents the current attention degree of the corresponding task. The loss function used here is the cross entropy loss, noted
Wherein r is i Is 0 or 1, corresponding to a true tag; p is a radical of i (. cndot.) produces a predicted score, which is a learnable parameter W f As a function of (c). Therefore, the loss function L of PKT-MCNN M Defined as a weighted combination of coarse and fine grained tasks:
preferably, in step 3, the PKT algorithm is embedded on top of the multitasking CNN,
the PKT adjusts the different attentions to the two tasks, the coarse-grained task and the fine-grained task,
in particular, the amount of the solvent to be used,
in the first phase, λ is set to 1, learning only coarse-grained tasks, since the weight of the fine-grained tasks is 0, the loss stops from back propagation;
in the second stage, the lambda is gradually reduced to 0 according to the number of training epochs, so that the two tasks are learned simultaneously, and coarse-grained knowledge is gradually transferred to fine-grained knowledge; second stage λ, noted λ 2 Calculated from the following equation:
wherein, B1 and B3 are respectively the training epoch numbers of the first stage and the second stage; b is the current epoch number, Bmax is the maximum epoch number;
finally, the value of λ at the third stage is opposite to the value of λ at the first stage, and λ is set to 0 to fine tune the fine-grained task without updating the parameters of the coarse-grained task.
The technical means not described in detail in the present application are known techniques.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.
Claims (8)
1. A multi-task convolutional neural network fault diagnosis method based on knowledge migration is characterized in that,
the method comprises the following steps:
step 1: preprocessing data;
step 2: learning from a coarse structure to a fine structure;
and step 3: multitasking is migrated from coarse to fine knowledge.
2. The knowledge migration based multitask convolutional neural network fault diagnosis method of claim 1,
the step 1 specifically comprises the following steps: and processing the input sample signal to obtain an input sample matrix.
3. The knowledge migration based multitask convolutional neural network fault diagnosis method of claim 1,
the step 2 specifically comprises the following steps:
step 21: similar graph construction for fault types;
step 22: spectral clustering of fault types;
step 23: and outputting the knowledge structure.
4. The knowledge migration based multitask convolutional neural network fault diagnosis method of claim 3,
the steps 21 and 22 specifically include: and clustering similar fault types to form coarse-grained knowledge.
5. The knowledge migration based multitask convolutional neural network fault diagnosis method of claim 3,
step 23 specifically includes:
step 231: training a neural network for extracting coarse-grained knowledge;
step 232: extracting coarse grain knowledge;
step 233: obtaining a network with coarse-grained knowledge extracted;
step 244: and transferring the knowledge of the network to a fine-grained network through transfer learning.
6. The knowledge migration based multitask convolutional neural network fault diagnosis method of claim 1,
the step 3 specifically comprises the following steps:
step 31: constructing a multitask convolution neural network based on progressive knowledge transfer;
step 32: training PKT-MCNN model transfer by using progressive knowledge;
step 33: testing a new sample by using the trained PKT-MCNN model;
step 34: and outputting the fault type of the test sample.
7. The knowledge migration based multitask convolutional neural network fault diagnosis method of claim 6,
in step 3, the PKT-MCNN training process includes three stages:
(1) training a coarse-grained task to master coarse-grained knowledge;
(2) simultaneously training coarse-grained and fine-grained tasks in a multi-task mode;
(3) fine-grained tasks are trained individually by fine-tuning the parameters updated in stage (2).
8. The knowledge migration based multitask convolutional neural network fault diagnosis method of claim 6,
in step 3, the PKT algorithm is embedded on top of the multitasking CNN,
the PKT adjusts the different attentions to the two tasks, the coarse-grained task and the fine-grained task,
in particular, the amount of the solvent to be used,
in the first phase, λ is set to 1, learning only coarse-grained tasks, since the weight of the fine-grained tasks is 0, the loss stops from back propagation;
in the second stage, the lambda is gradually reduced to 0 according to the number of training epochs, so that the two tasks are learned simultaneously, and coarse-grained knowledge is gradually transferred to fine-grained knowledge; second stage λ, denoted λ 2 Calculated from the following equation:
b1 and B3 are the training epochs of the first stage and the second stage respectively; b is the current epoch number, Bmax is the maximum epoch number;
finally, the value of λ at the third stage is opposite to the value of λ at the first stage, and λ is set to 0 to fine tune the fine-grained task without updating the parameters of the coarse-grained task.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110276577.5A CN115081468A (en) | 2021-03-15 | 2021-03-15 | Multi-task convolutional neural network fault diagnosis method based on knowledge migration |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110276577.5A CN115081468A (en) | 2021-03-15 | 2021-03-15 | Multi-task convolutional neural network fault diagnosis method based on knowledge migration |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115081468A true CN115081468A (en) | 2022-09-20 |
Family
ID=83240950
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110276577.5A Pending CN115081468A (en) | 2021-03-15 | 2021-03-15 | Multi-task convolutional neural network fault diagnosis method based on knowledge migration |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115081468A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120137367A1 (en) * | 2009-11-06 | 2012-05-31 | Cataphora, Inc. | Continuous anomaly detection based on behavior modeling and heterogeneous information analysis |
| CN110031227A (en) * | 2019-05-23 | 2019-07-19 | 桂林电子科技大学 | A kind of Rolling Bearing Status diagnostic method based on binary channels convolutional neural networks |
| US20200167659A1 (en) * | 2018-11-27 | 2020-05-28 | Electronics And Telecommunications Research Institute | Device and method for training neural network |
| CN111221340A (en) * | 2020-02-10 | 2020-06-02 | 电子科技大学 | Design method of migratable visual navigation based on coarse-grained features |
| CN111444889A (en) * | 2020-04-30 | 2020-07-24 | 南京大学 | Fine-grained action detection method of convolutional neural network based on multi-stage condition influence |
| CN112488241A (en) * | 2020-12-18 | 2021-03-12 | 贵州大学 | Zero sample picture identification method based on multi-granularity fusion network |
-
2021
- 2021-03-15 CN CN202110276577.5A patent/CN115081468A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120137367A1 (en) * | 2009-11-06 | 2012-05-31 | Cataphora, Inc. | Continuous anomaly detection based on behavior modeling and heterogeneous information analysis |
| US20200167659A1 (en) * | 2018-11-27 | 2020-05-28 | Electronics And Telecommunications Research Institute | Device and method for training neural network |
| CN110031227A (en) * | 2019-05-23 | 2019-07-19 | 桂林电子科技大学 | A kind of Rolling Bearing Status diagnostic method based on binary channels convolutional neural networks |
| CN111221340A (en) * | 2020-02-10 | 2020-06-02 | 电子科技大学 | Design method of migratable visual navigation based on coarse-grained features |
| CN111444889A (en) * | 2020-04-30 | 2020-07-24 | 南京大学 | Fine-grained action detection method of convolutional neural network based on multi-stage condition influence |
| CN112488241A (en) * | 2020-12-18 | 2021-03-12 | 贵州大学 | Zero sample picture identification method based on multi-granularity fusion network |
Non-Patent Citations (2)
| Title |
|---|
| ZHOUQI MA ET AL.: "Learning via Social Preference A Coarse-to-Fine Training Strategy for Style Transfer Systems", 《2018 IEEE INTERNATIONAL CONFERENCE ON DATA MINING WORKSHOPS (ICDMW)》, 31 December 2018 (2018-12-31) * |
| 陶启生 等: "用于轴承故障诊断的两步迁移学习法", 《计算机工程与应用》, 14 January 2021 (2021-01-14) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108830296B (en) | Improved high-resolution remote sensing image classification method based on deep learning | |
| CN110132598B (en) | Fault noise diagnosis algorithm for rolling bearing of rotating equipment | |
| CN107505133B (en) | The probability intelligent diagnosing method of rolling bearing fault based on adaptive M RVM | |
| CN106895975B (en) | Bearing fault diagnosis method based on Stacked SAE deep neural network | |
| CN112926641B (en) | Three-stage feature fusion rotating machine fault diagnosis method based on multi-mode data | |
| CN111898095A (en) | Deep migration learning intelligent fault diagnosis method and device, storage medium and equipment | |
| CN111914883B (en) | Spindle bearing state evaluation method and device based on deep fusion network | |
| CN114048568B (en) | Rotary machine fault diagnosis method based on multisource migration fusion shrinkage framework | |
| CN110609524B (en) | Industrial equipment residual life prediction model and construction method and application thereof | |
| CN109299741B (en) | Network attack type identification method based on multi-layer detection | |
| CN110188047B (en) | Double-channel convolutional neural network-based repeated defect report detection method | |
| CN106557782B (en) | Hyperspectral image classification method and device based on class dictionary | |
| CN107957946B (en) | Software defect prediction method based on neighborhood embedding protection algorithm support vector machine | |
| CN109409425B (en) | Fault type identification method based on neighbor component analysis | |
| CN108537257B (en) | Zero sample image classification method based on discriminant dictionary matrix pair | |
| CN105678343A (en) | Adaptive-weighted-group-sparse-representation-based diagnosis method for noise abnormity of hydroelectric generating set | |
| CN114004252A (en) | Bearing fault diagnosis method, device and equipment | |
| CN114355240A (en) | Power distribution network ground fault diagnosis method and device | |
| CN110363230A (en) | Stacking integrated sewage handling failure diagnostic method based on weighting base classifier | |
| CN110414587A (en) | Depth convolutional neural networks training method and system based on progressive learning | |
| CN115112372A (en) | Bearing fault diagnosis method and device, electronic equipment and storage medium | |
| Gu et al. | Identification of concurrent control chart patterns with singular spectrum analysis and learning vector quantization | |
| CN112464990A (en) | Method and device for sensing vibration data based on current and voltage sensor | |
| CN116992953B (en) | Model training method, fault diagnosis method and device | |
| CN114357372A (en) | Aircraft fault diagnosis model generation method based on multi-sensor data driving |
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 |