CN113656919B - Asymmetric rotor displacement field reconstruction method based on deep convolutional neural network - Google Patents
Asymmetric rotor displacement field reconstruction method based on deep convolutional neural network Download PDFInfo
- Publication number
- CN113656919B CN113656919B CN202111094165.6A CN202111094165A CN113656919B CN 113656919 B CN113656919 B CN 113656919B CN 202111094165 A CN202111094165 A CN 202111094165A CN 113656919 B CN113656919 B CN 113656919B
- Authority
- CN
- China
- Prior art keywords
- neural network
- convolutional neural
- displacement field
- asymmetric rotor
- dimensional
- 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 46
- 238000006073 displacement reaction Methods 0.000 title claims abstract description 45
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 37
- 238000005070 sampling Methods 0.000 claims abstract description 26
- 238000013528 artificial neural network Methods 0.000 claims abstract description 20
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000010606 normalization Methods 0.000 claims abstract description 7
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims abstract description 4
- 238000012795 verification Methods 0.000 claims abstract description 4
- 230000010354 integration Effects 0.000 claims abstract description 3
- 239000011159 matrix material Substances 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 10
- 238000010200 validation analysis Methods 0.000 claims description 6
- 238000005520 cutting process Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 5
- 230000003321 amplification Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000004513 sizing Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 230000015556 catabolic process Effects 0.000 claims description 2
- 238000006731 degradation reaction Methods 0.000 claims description 2
- 238000012546 transfer Methods 0.000 claims description 2
- 230000007547 defect Effects 0.000 abstract description 5
- 230000009467 reduction Effects 0.000 abstract description 3
- 238000012544 monitoring process Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- 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
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Medical Informatics (AREA)
- Pure & Applied Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Optimization (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses an asymmetric rotor displacement field reconstruction method based on a deep convolutional neural network, which comprises the following steps: collecting vibration signals at all measuring points based on a finite element method and a numerical integration method; establishing an asymmetric rotor displacement field reconstruction database based on equal length overlapping sampling; dividing a training set and a verification set based on data normalization; establishing a feature extractor and a signal generator based on a convolutional neural network, bicubic interpolation up-sampling and a fully-connected neural network; training a model based on the mean square error, the Adam algorithm and a learning rate step-down method to obtain a final reconstruction model. The reconstruction method of the invention can overcome the defect that the traditional method is difficult to apply to an asymmetric rotor system with nonlinearity, is suitable for any input signal size, has high prediction speed, can overcome the defect that the reconstruction error of the traditional method is obviously increased along with the reduction of the number of measuring points, and is a method very suitable for industrial sites where more measuring points are difficult to arrange.
Description
Technical Field
The invention belongs to the technical field of rotary machine operation monitoring, and particularly relates to an asymmetric rotor displacement field reconstruction method based on a deep convolutional neural network.
Background
In practical industrial production, there is a class of rotor systems with asymmetric structures, such as generators, motors, crankshafts, fans, and cracked rotors. The asymmetric rotor has complex structure and severe working environment, and is easier to break down than the symmetric rotor under the condition of long-time operation. If the failure can not be detected in time, the machine is stopped in an unplanned mode due to the fact that the vibration quantity exceeds the limit value, and the machine is damaged and casualties are caused in a heavy mode. The health monitoring technology has a great propulsion effect on the detection and inhibition of the vibration of the asymmetric rotor, and the displacement field real-time reconstruction technology is a key technology for health monitoring and intelligent vibration control. Therefore, the research on the real-time reconstruction method of the asymmetric rotor displacement field has very important significance for guaranteeing the safe operation of the asymmetric rotor, reducing great economic loss and avoiding the occurrence of disastrous accidents.
The dynamic equation of the asymmetric rotor system has periodic time-varying parameters and is nonlinear, so that the traditional displacement field reconstruction method, such as a modal expansion method, is difficult to apply to the asymmetric rotor system. In addition, the modal expansion method has the defect that reconstruction errors are obviously increased along with the reduction of the number of the measuring points, and is not in line with the actual situation that more measuring points are difficult to arrange in an industrial field. In summary, there is a need to develop a new real-time reconstruction method for asymmetric rotor displacement fields.
Disclosure of Invention
The invention aims to provide an asymmetric rotor displacement field reconstruction method based on a deep convolutional neural network. The invention establishes a feature extractor and a signal generator based on a deep learning method, takes displacement signals of a small number of measuring points as input and displacement signals of all nodes as output aiming at an asymmetric rotor, so as to complete the real-time reconstruction of the displacement field of the asymmetric rotor system.
In order to achieve the above purpose, the invention adopts the following technical scheme:
an asymmetric rotor displacement field reconstruction method based on a deep convolutional neural network comprises the following steps:
1) Based on a finite element method, establishing a dynamic equation of an asymmetric rotor system, and obtaining initial vibration displacement signals at all nodes of the rotor through a numerical integration method;
2) Performing equal length cutting on the initial vibration displacement signals, respectively processing signals of all nodes and signals of a small number of measuring points into labels and input signals, and establishing an asymmetric rotor displacement field reconstruction database;
3) Carrying out normalization processing on an asymmetric rotor displacement field reconstruction database to form a sample set, and dividing the sample set into a training set and a verification set according to a set proportion;
4) Based on a convolutional neural network, bicubic interpolation up-sampling and a fully connected neural network, establishing a characteristic extractor and a signal generator which are applicable to any input signal size and used for reconstructing an asymmetric rotor displacement field, enabling an input signal to sequentially pass through the characteristic extractor and the signal generator to obtain an output signal, and defining a residual error of the network based on the output signal;
5) And the final feature extractor and the signal generator are obtained through the reverse transfer of the gradient after training, so that an asymmetric rotor displacement field real-time reconstruction model is formed.
The invention is further improved in that in the step 1), latin hypercube sampling is adopted to sample the fault parameters of the asymmetric rotor system, and vibration displacement signals under various different fault parameters are obtained.
The invention further improves that the data is expanded by adopting an overlapping sampling method, and the sample lengths after overlapping sampling are ensured to be equal in the expansion process.
In the step 3), all samples are mapped to between 0 and 1 by adopting a maximum value and minimum value normalization method, and the processing formula is X= (X-min (X))/(max (X) -min (X)), wherein X represents input signals of all nodes or a small number of measuring points; 80% of the samples were treated as training set (X less,t ,X all,t ) An additional 20% of the samples were treated as validation set and validation set (X less,v ,X all,v )。
In the step 4), considering that the number of measuring points is far smaller than the number of time steps, the feature extractor consists of a bicubic interpolation up-sampling layer, a plurality of one-dimensional convolutional neural networks and a fully-connected neural network; in order to increase the depth of the network, the maximum pooling is performed once every two one-dimensional convolutional neural networks; after the one-dimensional convolution, stopping the one-dimensional convolution operation when the length of the time dimension becomes 1, and mapping the output of the one-dimensional convolution neural network into a feature vector through the fully connected neural network to prevent the feature extractor from losing the feature, wherein the dimension of the feature vector is not smaller than the number of independent variables of the displacement field reconstruction problem.
The invention is further improved in that in the step 4), the signal generator is composed of a fully connected neural network, a plurality of two-dimensional convolution neural networks and a plurality of bicubic interpolation up-sampling layers; the feature vector output by the feature extractor is changed into a one-dimensional long vector through a fully connected neural network, and is changed into a two-dimensional matrix through dimension transformation; the two-dimensional matrix is changed into a two-dimensional matrix with the size enlarged twice through a two-dimensional convolutional neural network with the unchanged size and a bicubic interpolation up-sampling layer with the amplification factor of two; after passing through a plurality of two-dimensional convolutional neural networks and bicubic interpolation upsampling layers, finally passing through a two-dimensional convolutional neural network and a sizing bicubic interpolation upsampling layer, so that the size of a finally output two-dimensional matrix is equal to the matrix size of signals of all nodes.
A further improvement of the invention is that in step 4), in order to prevent network degradation problems, in the signal generator, a residual connection is used after each completion of the two-dimensional convolutional neural network and bicubic interpolation upsampling layer.
In step 5), the error is selected as the mean square error, the optimization algorithm is selected as the Adam algorithm, the initial learning rate is 0.001, the learning rate of the network can be stepped down in the training process, and when the error is continuously updated for many times, iteration is stopped, and the calculation convergence is achieved.
Compared with the prior art, the invention has at least the following beneficial technical effects:
the invention provides an asymmetric rotor displacement field reconstruction method based on a deep convolutional neural network. The method has the advantages that: firstly, the defect that the traditional method is difficult to apply to an asymmetric rotor system with nonlinearity can be overcome; secondly, the method is suitable for any input signal size, has high prediction speed, can realize real-time reconstruction, and has important significance for real-time state monitoring of asymmetric rotor vibration; finally, the method can overcome the defect that the reconstruction error of the traditional method is obviously increased along with the reduction of the number of the measuring points, and is suitable for industrial sites where more measuring points are difficult to arrange.
Drawings
FIG. 1 is a schematic block diagram of a method for reconstructing an asymmetric rotor displacement field based on a deep convolutional neural network in accordance with an embodiment of the present invention;
fig. 2 is a schematic diagram of a network architecture of a feature extractor and signal generator constructed in accordance with the present invention.
Detailed Description
The invention is explained in further detail below with reference to the drawings and the specific embodiments.
Referring to fig. 1, the method for reconstructing an asymmetric rotor displacement field based on a deep convolutional neural network provided by the invention comprises the following steps:
1. based on a finite element method, a dynamic equation of an asymmetric rotor system is established, and initial vibration displacement signals at all nodes of the rotor are obtained through a numerical analysis method.
The method comprises the steps of carrying out finite element meshing and boundary condition application on an asymmetric rotor, generating different fault parameter samples by using Latin hypercube sampling, and further obtaining initial vibration displacement signals of all nodes under different fault parameters based on a Dragon-Gregorian tower method.
2. And performing equal-length cutting on the initial vibration displacement signals, respectively processing signals of all nodes and signals of a small number of measuring points into labels and input signals, and establishing an asymmetric rotor displacement field reconstruction database.
Cutting initial vibration data into data with the length of 64, ensuring the cutting length to be unchanged, and completing data capacity expansion through overlapping sampling; and respectively processing the signals of all the nodes and the signals of a small number of measuring points into input signals of the tag and the deep learning network to form an asymmetric rotor displacement field reconstruction database.
3. And carrying out normalization processing on the asymmetric rotor displacement field reconstruction database to form a sample set, and dividing the sample set into a training set and a verification set according to a set proportion.
Specifically, a maximum value and minimum value normalization method is adopted to map all samples to between 0 and 1, and a processing formula is X= (X-min (X))/(max (X) -min (X)), wherein X represents input signals of all nodes or a small number of measuring points; 80% of the samples were treated as training set (X less,t ,X all,t ) An additional 20% of the samples were treated as validation set and validation set (X less,v ,X all,v )。
4. Based on a convolutional neural network, bicubic interpolation up-sampling and a fully connected neural network, a feature extractor and a signal generator which can be suitable for any input signal size and used for reconstructing an asymmetric rotor displacement field are established, input signals sequentially pass through the feature extractor and the signal generator to obtain output signals, and residual errors of the network are defined based on the output signals.
Specifically, the network structure of the created feature extractor and signal generator is shown in fig. 2, and specifically includes: the feature extractor consists of a bicubic interpolation up-sampling layer, a plurality of one-dimensional convolutional neural networks and a fully-connected neural network; in order to increase the depth of the network, the maximum pooling is performed once every two one-dimensional convolutional neural networks; after the one-dimensional convolution, stopping the one-dimensional convolution operation when the length of the time dimension becomes 1, and mapping the output of the one-dimensional convolution neural network into a feature vector through the fully connected neural network. The input of the feature extractor is vibration signal X of a small number of measuring points less The output is the vibration signature Y. The following table is a network parameter table of the feature extractor, in which the size of the input parameters is (4×m less )×N T Wherein M is less The number of the measuring points is a small number of the measuring points, N T For the number of time domain sampling points, the number of convolution blocks n=ceil (log 8 N T ) Where ceil represents a round up.
The signal generator is then formed by a full connectionThe system comprises a neural network, a plurality of two-dimensional convolutional neural networks and a plurality of bicubic interpolation up-sampling layers; the feature vector output by the feature extractor is changed into a one-dimensional long vector through a fully connected neural network, and is changed into a two-dimensional matrix through dimension transformation; the two-dimensional matrix is changed into a two-dimensional matrix with the size enlarged twice through a two-dimensional convolutional neural network with the unchanged size and a bicubic interpolation up-sampling layer with the amplification factor of two; after passing through a plurality of two-dimensional convolutional neural networks and bicubic interpolation upsampling layers, finally passing through a two-dimensional convolutional neural network and a sizing bicubic interpolation upsampling layer, so that the size of a finally output two-dimensional matrix is equal to the matrix size of signals of all nodes. The input of the signal generator is the characteristic vector Y output by the characteristic extractor, and the vibration signal X is output as all nodes all . The following table is a network parameter table of the signal generator, wherein the size of the input parameter is N Y The size of the output parameter is C×M all ×N T Wherein C is the number of channels, i.e. the number of degrees of freedom, M all N is the number of nodes of all the nodes T For the number of time-domain sampling points, the number of up-sampling blocks n=floor (log 2 M all ) Wherein floor represents a rounding down.
The error is selected as the mean square error, the optimization algorithm is selected as the Adam algorithm, the initial learning rate is 0.001, the learning rate of the network can be reduced stepwise in the training process, the learning rate is reduced by 10 times every 100 steps, and when the error is continuously updated 10 times without obvious update, iteration is stopped, and convergence is calculated.
While the invention has been described in connection with what is presently considered to be the most practical and preferred embodiment, it is to be understood that the invention is not limited to the disclosed embodiment, but on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.
Claims (5)
1. The asymmetric rotor displacement field reconstruction method based on the deep convolutional neural network is characterized by comprising the following steps of:
1) Based on a finite element method, establishing a dynamic equation of an asymmetric rotor system, and obtaining initial vibration displacement signals at all nodes of the rotor through a numerical integration method;
2) Performing equal length cutting on the initial vibration displacement signals, respectively processing signals of all nodes and signals of a small number of measuring points into labels and input signals, and establishing an asymmetric rotor displacement field reconstruction database;
3) Carrying out normalization processing on an asymmetric rotor displacement field reconstruction database to form a sample set, and dividing the sample set into a training set and a verification set according to a set proportion;
4) Based on a convolutional neural network, bicubic interpolation up-sampling and a fully connected neural network, establishing a characteristic extractor and a signal generator which are applicable to any input signal size and used for reconstructing an asymmetric rotor displacement field, enabling an input signal to sequentially pass through the characteristic extractor and the signal generator to obtain an output signal, and defining a residual error of the network based on the output signal; considering that the number of measuring points is far smaller than the time step number, the feature extractor consists of a bicubic interpolation up-sampling layer, a plurality of one-dimensional convolutional neural networks and a fully-connected neural network; in order to increase the depth of the network, the maximum pooling is performed once every two one-dimensional convolutional neural networks; after a plurality of one-dimensional convolutions, stopping one-dimensional convolution operation when the length of the time dimension becomes 1, and mapping the output of the one-dimensional convolution neural network into a feature vector through a fully connected neural network, wherein the dimension of the feature vector is not less than the number of independent variables of the displacement field reconstruction problem in order to prevent the feature extractor from losing the feature;
the signal generator consists of a fully connected neural network, a plurality of two-dimensional convolutional neural networks and a plurality of bicubic interpolation up-sampling layers; the feature vector output by the feature extractor is changed into a one-dimensional long vector through a fully connected neural network, and is changed into a two-dimensional matrix through dimension transformation; the two-dimensional matrix is changed into a two-dimensional matrix with the size enlarged twice through a two-dimensional convolutional neural network with the unchanged size and a bicubic interpolation up-sampling layer with the amplification factor of two; after passing through a plurality of two-dimensional convolutional neural networks and bicubic interpolation upsampling layers, finally passing through a two-dimensional convolutional neural network and a sizing bicubic interpolation upsampling layer, so that the size of a finally output two-dimensional matrix is equal to the matrix size of signals of all nodes;
in order to prevent the network degradation problem, in the signal generator, residual connection is adopted after a two-dimensional convolutional neural network and a bicubic interpolation up-sampling layer are completed each time;
5) And the final feature extractor and the signal generator are obtained through the reverse transfer of the gradient after training, so that an asymmetric rotor displacement field real-time reconstruction model is formed.
2. The method for reconstructing the asymmetric rotor displacement field based on the deep convolutional neural network according to claim 1, wherein in the step 1), the Latin hypercube sampling is adopted to sample the fault parameters of the asymmetric rotor system, so as to obtain vibration displacement signals under a plurality of different fault parameters.
3. The asymmetric rotor displacement field reconstruction method based on the deep convolutional neural network according to claim 1, wherein the data is expanded by adopting an overlap sampling method, and the sample lengths after overlap sampling are ensured to be equal in the expansion process.
4. The asymmetric rotor displacement field reconstruction method based on the deep convolutional neural network according to claim 1, wherein in the step 3), all samples are mapped to between 0 and 1 by adopting a maximum-minimum normalization method, and the processing formula is x= (X-min (X))/(max (X) -min (X)), wherein X represents input signals of all nodes or a small number of measuring points; 80% of the samples were treated as training set (X less,t ,X all,t ) An additional 20% of the samples were treated as validation set and validation set (X less,v ,X all,v )。
5. The method for reconstructing the asymmetric rotor displacement field based on the deep convolutional neural network according to claim 1, wherein in the step 5), the error is selected as a mean square error, the optimization algorithm is selected as an Adam algorithm, the initial learning rate is 0.001, the learning rate of the network can be stepped down in the training process, and when the error is continuously updated for a plurality of times, iteration is stopped, and calculation convergence is achieved.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111094165.6A CN113656919B (en) | 2021-09-17 | 2021-09-17 | Asymmetric rotor displacement field reconstruction method based on deep convolutional neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111094165.6A CN113656919B (en) | 2021-09-17 | 2021-09-17 | Asymmetric rotor displacement field reconstruction method based on deep convolutional neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113656919A CN113656919A (en) | 2021-11-16 |
CN113656919B true CN113656919B (en) | 2024-04-02 |
Family
ID=78483832
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111094165.6A Active CN113656919B (en) | 2021-09-17 | 2021-09-17 | Asymmetric rotor displacement field reconstruction method based on deep convolutional neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113656919B (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109612708A (en) * | 2018-12-28 | 2019-04-12 | 东北大学 | Based on the power transformer on-line detecting system and method for improving convolutional neural networks |
CN111707458A (en) * | 2020-05-18 | 2020-09-25 | 西安交通大学 | Rotor monitoring method based on deep learning signal reconstruction |
WO2020244134A1 (en) * | 2019-06-05 | 2020-12-10 | 华南理工大学 | Multi-task feature sharing neural network-based intelligent fault diagnosis method |
-
2021
- 2021-09-17 CN CN202111094165.6A patent/CN113656919B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109612708A (en) * | 2018-12-28 | 2019-04-12 | 东北大学 | Based on the power transformer on-line detecting system and method for improving convolutional neural networks |
WO2020244134A1 (en) * | 2019-06-05 | 2020-12-10 | 华南理工大学 | Multi-task feature sharing neural network-based intelligent fault diagnosis method |
CN111707458A (en) * | 2020-05-18 | 2020-09-25 | 西安交通大学 | Rotor monitoring method based on deep learning signal reconstruction |
Non-Patent Citations (1)
Title |
---|
基于无监督学习卷积神经网络的振动信号模态参数识别;方宁;周宇;叶庆卫;李玉刚;;计算机应用(03);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN113656919A (en) | 2021-11-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Shao et al. | Rolling bearing fault diagnosis using adaptive deep belief network with dual-tree complex wavelet packet | |
CN111596604B (en) | Intelligent fault diagnosis and self-healing control system and method for engineering equipment based on digital twinning | |
He et al. | Deep transfer multi-wavelet auto-encoder for intelligent fault diagnosis of gearbox with few target training samples | |
CN112684379A (en) | Transformer fault diagnosis system and method based on digital twinning | |
CN108197648A (en) | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models | |
CN114429153A (en) | Lifetime learning-based gearbox increment fault diagnosis method and system | |
CN116010900A (en) | Multi-scale feature fusion gearbox fault diagnosis method based on self-attention mechanism | |
CN107025341A (en) | A kind of photovoltaic DC-to-AC converter method for diagnosing faults | |
CN108428023A (en) | Trend forecasting method based on quantum Weighted Threshold repetitive unit neural network | |
CN115809596A (en) | Digital twin fault diagnosis method and device | |
CN115983374A (en) | Cable partial discharge database sample expansion method based on optimized SA-CACGAN | |
Miao et al. | Fault diagnosis of wheeled robot based on prior knowledge and spatial-temporal difference graph convolutional network | |
CN113656919B (en) | Asymmetric rotor displacement field reconstruction method based on deep convolutional neural network | |
CN112348158B (en) | Industrial equipment state evaluation method based on multi-parameter deep distribution learning | |
CN106682312A (en) | Industrial process soft-measurement modeling method of local weighing extreme learning machine model | |
Yan et al. | Reliability prediction of CNC machine tool spindle based on optimized cascade feedforward neural network | |
CN117407675A (en) | Lightning arrester leakage current prediction method based on multi-variable reconstruction combined dynamic weight | |
CN117744859A (en) | Wind turbine generator fault early warning method based on self-adaptive double control strategy | |
Tao et al. | A digital twin-based fault diagnostic method for subsea control systems | |
CN115659844B (en) | Simulation method and device based on wind power plant dynamic model and computer equipment | |
CN117034169A (en) | Power grid main transformer equipment abnormal state prediction method based on time sequence causality network | |
CN116432359A (en) | Variable topology network tide calculation method based on meta transfer learning | |
Ning et al. | An intelligent device fault diagnosis method in industrial internet of things | |
CN116881686A (en) | Nuclear pipeline fault diagnosis method of quantum BP neural network | |
CN115239989A (en) | Distributed fault diagnosis method for oil immersed transformer based on data privacy protection |
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 |