CN112580588B - Intelligent flutter signal identification method based on empirical mode decomposition - Google Patents
Intelligent flutter signal identification method based on empirical mode decomposition Download PDFInfo
- Publication number
- CN112580588B CN112580588B CN202011587834.9A CN202011587834A CN112580588B CN 112580588 B CN112580588 B CN 112580588B CN 202011587834 A CN202011587834 A CN 202011587834A CN 112580588 B CN112580588 B CN 112580588B
- Authority
- CN
- China
- Prior art keywords
- signal
- flutter
- network
- dimensional data
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 16
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 34
- 238000012549 training Methods 0.000 claims abstract description 22
- 238000012360 testing method Methods 0.000 claims abstract description 17
- 230000000694 effects Effects 0.000 claims abstract description 14
- 230000006870 function Effects 0.000 claims description 18
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000011176 pooling Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 abstract description 9
- 238000004458 analytical method Methods 0.000 abstract description 5
- 238000013461 design Methods 0.000 abstract description 5
- 238000004140 cleaning Methods 0.000 abstract description 2
- 238000010276 construction Methods 0.000 abstract description 2
- 210000002569 neuron Anatomy 0.000 abstract description 2
- 238000002360 preparation method Methods 0.000 abstract description 2
- 238000011160 research Methods 0.000 abstract description 2
- 230000004044 response Effects 0.000 description 7
- 238000013016 damping Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001687 destabilization Effects 0.000 description 1
- 230000001066 destructive effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Mathematical Optimization (AREA)
- Molecular Biology (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Databases & Information Systems (AREA)
- Algebra (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Complex Calculations (AREA)
Abstract
The invention relates to an intelligent flutter signal identification method based on empirical mode decomposition, which uses an EMD algorithm to acquire the vibration mode of a signal and converts the vibration mode into two-dimensional data to be used as the input of a CNN network. The time-frequency diagram can also be obtained by carrying out short-time Fourier transform on the flutter signal, and the time-frequency diagram comprises the time-frequency characteristics of the signal. The time-frequency diagram is directly used as the input of the CNN network, and good effect can be obtained. The invention obtains the eigenmode function of the flutter signal by using the EMD algorithm, converts the eigenmode function into two-dimensional data, inputs the two-dimensional data into the convolutional neural network for training, and realizes intelligent identification of the flutter phenomenon. And performing preliminary classification, data preparation and data cleaning on the measured multi-dimensional signals of the flutter to generate training sets and test sets required by network research. Structural design of convolutional neural network facing flutter analysis, construction of network frame, and adjustment of super parameters such as punishment factors, hidden layer neurons, stopping points of back propagation algorithm, optimal network depth and the like.
Description
Technical Field
The invention belongs to a flutter signal identification method, and relates to an intelligent flutter signal identification method based on empirical mode decomposition.
Background
Flutter is a self-excited vibration that is a destructive aeroelastic unstable state due to the coupling of the elastic structure with aerodynamic, inertial and elastic forces. It is therefore important how the flutter characteristics of an aircraft can be determined from the structural response signals.
When the vibration occurs, the damping of the vibration signal is zero or negative, the vibration signal is simple harmonic or divergent in the time domain, and the vibration signal is single in mode and has increased energy in the frequency domain. The main current flutter signal processing method comprises a damping method and a system stability-based prediction method, wherein the system stability analysis method takes flutter as a destabilization phenomenon of a structural system for analysis, and the basic thinking is as follows: and establishing a proper dynamic data model for subcritical test signals, constructing corresponding stability criteria by combining model parameters with certain criteria, calculating stability parameters, fitting the parameters through a curve, and extrapolating the flutter critical speed.
Disadvantages of the prior art:
1. for burst type flutter, the damping method and the system stability analysis method can not accurately predict the flutter boundary under the condition of being far away from the flutter critical point, and the prediction results obtained by the time domain method for the same group of states have larger difference;
2. the quality of the actually measured structural response signal is low due to the limitations of the test environment and test conditions, so that the prediction result of the traditional method is not ideal.
Disclosure of Invention
Technical problem to be solved
In order to avoid the defects of the prior art, the invention provides an intelligent flutter signal identification method based on empirical mode decomposition, which is a method for intelligently classifying structural response signals of a pneumatic elastic model so as to achieve the effect of accurately identifying flutter signals.
Technical proposal
An intelligent flutter signal identification method based on empirical mode decomposition is characterized by comprising the following steps:
step 1: performing zero-mean normalization processing on the acquired signal x (n) to obtain y (n)
Wherein: mean (x (n)) is the mean of x (n), std (x (n)) is the standard deviation of x (n);
step 2: the EMD method is adopted to decompose the signal into a plurality of IMFs components and a remainder, wherein the remainder is a monotonous and smooth trend of the original signal, and each IMF component meets the following two conditions:
1. the number of extreme points of the sequence is equal to the number of zero crossing points or differs by at most 1 in the whole time course of the signal;
2. at any time point, the upper envelope line formed by the local maximum value and the lower envelope average value formed by the local minimum value of the signal are zero;
step 3: selecting the first n IMFs and converting the IMFs into two-dimensional data;
step 4: dividing the two-dimensional data set into a training set and a test set, wherein the training set is used as the input of the convolutional neural network
The convolutional neural network CNN network: the two groups of four convolutional layers and two pooling layers are used for extracting features, and further carrying out flutter signal classification through a three-layer full-connection network;
optimizing parameters by using an Adam optimizer, and adding a dropout layer behind the first two full-connection layers to prevent overfitting;
step 5: after training, checking the classification effect through a test set, using the loss function and the accuracy as training and testing standards, judging the effect of parameter optimization and the result of a model to be effective when the loss function value reaches the minimum and the accuracy reaches the maximum, otherwise, adjusting the iteration times, the learning rate, the dropout layer parameters and the delta super parameters to retrain the CNN network
1. Loss function:
wherein y is i Representing the score of the signal on each class (only two classes of flutter and non-flutter), wherein the true result of the signal is the j-th class, and delta is a threshold value;
2. accuracy rate of
Wherein acc represents the accuracy, N represents the number of neural network result predictions that are correct, and N represents the dimension of the data amount;
after training, sequentially performing step 2 and step 3 on new unknown signals with flutter or not, and inputting the signals into a trained CNN network, wherein the output learning signals of the CNN network are flutter or not.
When the signal length is 2048 and the two-dimensional data is converted in the step 3, the first eight IMFs are selected, and the eight IMFs with the length of 2048 are converted into 128@128 two-dimensional data.
Advantageous effects
According to the intelligent flutter signal identification method based on empirical mode decomposition, an EMD algorithm is used for obtaining the vibration mode of signals, and the vibration mode is converted into two-dimensional data to be used as the input of a CNN network. The time-frequency diagram can also be obtained by carrying out short-time Fourier transform on the flutter signal, and the time-frequency diagram comprises the time-frequency characteristics of the signal. The time-frequency diagram is directly used as the input of the CNN network, and good effect can be obtained. In the invention, the following components are added: 1. and obtaining an eigenmode function of the flutter signal by using an EMD algorithm, converting the eigenmode function into two-dimensional data, inputting the two-dimensional data into a convolutional neural network for training so as to realize intelligent identification of the flutter phenomenon. 2. And performing preliminary classification, data preparation and data cleaning on the measured multi-dimensional signals of the flutter to generate training sets and test sets required by network research. 3. Structural design of convolutional neural network facing flutter analysis, construction of network frame, and adjustment of super parameters such as punishment factors, hidden layer neurons, stopping points of back propagation algorithm, optimal network depth and the like.
The EMD algorithm identifies all vibration modes contained in the signal through the characteristic time scale, and each IMF component contains local characteristic signals of different time scales of the original signal, so that the method has obvious advantages in processing non-stationary and non-linear data such as flutter signals. The CNN network has the characteristic of automatically extracting the characteristics, and the combination of the CNN network and the CNN network can effectively reduce the influence of lower quality of actually measured structural response signals caused by the limitations of test environments and test conditions, thereby realizing the flutter signal identification effect with the accuracy rate up to 100%.
Drawings
Fig. 1: EMD decomposition flow chart
Fig. 2: random signal EMD decomposition results
Fig. 3: time series of occurrence of dither signal
Fig. 4: IMFs generating flutter signals
Fig. 5: time series of non-occurrence of chatter signals
Fig. 6: IMFs without dither signal
Fig. 7: CNN convolutional neural network structure
Fig. 8: loss curve
Fig. 9: correction rate curve
Fig. 10: integral thought flow chart
Detailed Description
The invention will now be further described with reference to examples, figures:
the scheme provides a method for intelligently classifying structural response signals of a pneumatic elastic model, so that the effect of accurately identifying flutter signals is achieved, and the overall design is as follows:
1. the signal data set is obtained through experiments, wherein the signal data set comprises flutter signals and non-flutter signals, in order to reduce the influence caused by human errors in the actual signal acquisition process, zero-mean normalization processing pretreatment is carried out on the signal x (n) to obtain y (n), and the pretreatment method is as follows:
where mean (x (n)) is the mean of x (n), standard deviation of std (x (n)).
2. The signal is decomposed by EMD (empirical mode decomposition) into a number of eigenmode function (Intrinsic Mode Function, IMF) components and a remainder, which is also a trend term of the original signal, which is monotonic and smooth. Each IMF component satisfies the following two conditions:
1) The number of extreme points of the sequence is equal to the number of zero crossing points or differs by at most 1 in the whole time course of the signal;
2) At any one point in time, the upper envelope of local maxima and the lower envelope of local minima of the signal have an average value of zero.
An EMD decomposition flow chart is shown in fig. 1.
The IMF components of each order arranged according to the scale size respectively comprise components with different signals, and IMFs and residuals after EMD decomposition of a random signal are shown in fig. 2:
3. the known wind tunnel test signals are artificially classified into signals in a flutter state and signals without flutter according to the flutter phenomenon as data sets. The IMFs of the data set signals are obtained by processing the above described method, and the first 8 IMFs of each signal are selected and converted (reshape) into two-dimensional data.
The two-dimensional data set is divided into a training set and a test set, the training set being an input to a convolutional neural network (Convolutional Neural Networks, CNN). The CNN network design of the scheme uses two groups of four layers of convolution layers and two layers of pooling layers to extract characteristics, and further carries out flutter signal classification through a three-layer full-connection network. The whole network utilizes an Adam optimizer to optimize parameters, and a dropout layer is added behind the first two fully connected layers to prevent overfitting. Fig. 7 is a CNN network structure according to the present scheme:
when the signal length is 2048, the first 8 IMF components are selected after EMD decomposition, and are converted into two-dimensional data of 128×128@1. The first group of convolution layers is adopted, the filter size of the convolution layer C1 is 5*5, the output is 6 channels, the step size is 1, the padding is 2, and the feature layer with the size of 128 x 128@6 is obtained through the convolution layer C1. The filter size of the convolution layer C2 is 5*5, the output is 10 channels, the step size is 1, the padding is 2, and the characteristic layer with the size of 128 x 128@10 is obtained through the convolution layer C2. And obtaining a 64 x 64@10 characteristic layer through a maximum pooling layer P1 with a pooling core size of 2 x 2. Then the second group of convolution layers is entered, the filter size of the convolution layer C3 is 3*3, the input is 10 channels, the output is 14 channels, the step length is 1, the padding is 1, and the characteristic layer with the size of 64 x 64@14 is obtained through the convolution layer C3. The filter size of the convolution layer C4 is 3*3, the input is 14 channels, the output is 20 channels, the step size is 1, the padding is 1, and the characteristic layer with the size of 64 x 64@20 is obtained through the convolution layer C4. And then the maximum pooling layer P2 of 2 x 2 is entered to obtain a characteristic layer of 32 x 32@20. The three full connection layers are 20480×1024,1024×512,512×2, respectively, and finally obtain the two classification results (flutter or non-flutter) of the structural response signal.
4. After training, the classifying effect is checked through the test set, the loss function and the correct rate are used as the training and testing standard, and whether the loss function value reaches the minimum and the correct rate reaches the maximum is judged through the loss function and the correct rate of the data set in the training process, so that the effect of parameter optimization and the result of the model are judged whether to be effective or not.
1) Loss function
The convolution neural network adopts a range loss function (formula 2) for calculating the error between the forward feedback result of the neural network and the label, and the value participates in parameter updating of the whole neural network through gradient descent and optimization functions.
Wherein y is i The score of the signal on each category is represented, the true result of the signal is the j-th category, and delta is the threshold.
2) Accuracy rate of
The calculation result of a neural network is obtained after each training set is fed back forward, and parameters of the whole model 0 are obtained at the same time, and the accuracy is obtained by calculating the ratio between the same number of the results of the neural network and the results of the labels and the total data volume, wherein the value is used as a standard for measuring the model effect, and the formula is as follows:
where acc represents the accuracy, N represents the number of neural network results predicted correctly, and N represents the dimension of the data amount.
After multiple training, when the iteration number of the super parameter is set to 1000, the learning rate is 0.00001, and the two dropout layer parameters are 0.5, the delta=10 loss function is well reduced and the accuracy is highest, and the loss curve and the accuracy curve are shown in fig. 8 and 9:
table 1 results of 1000 th iteration
From table 1, it can be seen that after 1000 iterations, the convolutional neural network prediction result can correctly classify 100% of the signals, that is, the identification function of the flutter signals is realized.
5. After training is finished, whether the new structural response signal is a flutter signal is identified, and only after the signal is subjected to EMD decomposition, the first 8 IMF components are taken and converted into two-dimensional data, and the two-dimensional data is input into a trained CNN neural network, so that whether the signal is the flutter signal can be identified.
6. The whole design thought flow chart of the scheme is shown in fig. 10:
the scheme obtains the vibration mode of the signal by using an EMD algorithm and converts the vibration mode into two-dimensional data to be used as the input of a CNN network. The time-frequency diagram can also be obtained by carrying out short-time Fourier transform on the flutter signal, and the time-frequency diagram comprises the time-frequency characteristics of the signal. The time-frequency diagram is directly used as the input of the CNN network, and good effect can be obtained.
Claims (2)
1. An intelligent flutter signal identification method based on empirical mode decomposition is characterized by comprising the following steps:
step 1: performing zero-mean normalization processing on the acquired signal x (n) to obtain y (n)
Wherein: mean (x (n)) is the mean of x (n), std (x (n)) is the standard deviation of x (n);
step 2: the EMD method is adopted to decompose the signal into a plurality of IMFs components and a remainder, wherein the remainder is a monotonous and smooth trend of the original signal, and each IMF component meets the following two conditions:
1. the number of extreme points of the sequence is equal to the number of zero crossing points or differs by at most 1 in the whole time course of the signal;
2. at any time point, the upper envelope line formed by the local maximum value and the lower envelope average value formed by the local minimum value of the signal are zero;
step 3: selecting the first n IMFs and converting the IMFs into two-dimensional data;
step 4: dividing the two-dimensional data set into a training set and a test set, wherein the training set is used as the input of the convolutional neural network
The convolutional neural network CNN network: the two groups of four convolutional layers and two pooling layers are used for extracting features, and further carrying out flutter signal classification through a three-layer full-connection network;
optimizing parameters by using an Adam optimizer, and adding a dropout layer behind the first two full-connection layers to prevent overfitting;
step 5: after training, checking the classification effect through a test set, using the loss function and the accuracy as training and testing standards, judging the effect of parameter optimization and the result of a model to be effective when the loss function value reaches the minimum and the accuracy reaches the maximum, otherwise, adjusting the iteration times, the learning rate, the dropout layer parameters and the delta super parameters to retrain the CNN network
1. Loss function:
wherein y is i Representing the score of the signal in two categories of flutter and non-flutter, wherein the true result of the signal is j-th category, and delta is a threshold value;
2. accuracy rate of
Wherein acc represents the accuracy, N represents the number of neural network result predictions that are correct, and N represents the dimension of the data amount;
after training, sequentially performing step 2 and step 3 on new unknown signals with flutter or not, and inputting the signals into a trained CNN network, wherein the output learning signals of the CNN network are flutter or not.
2. The intelligent flutter signal identification method based on empirical mode decomposition according to claim 1, wherein the method is characterized by comprising the following steps of: when the signal length is 2048 and the two-dimensional data is converted in the step 3, the first eight IMFs are selected, and the eight IMFs with the length of 2048 are converted into 128@128 two-dimensional data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011587834.9A CN112580588B (en) | 2020-12-29 | 2020-12-29 | Intelligent flutter signal identification method based on empirical mode decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011587834.9A CN112580588B (en) | 2020-12-29 | 2020-12-29 | Intelligent flutter signal identification method based on empirical mode decomposition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112580588A CN112580588A (en) | 2021-03-30 |
| CN112580588B true CN112580588B (en) | 2024-01-12 |
Family
ID=75143827
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011587834.9A Active CN112580588B (en) | 2020-12-29 | 2020-12-29 | Intelligent flutter signal identification method based on empirical mode decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112580588B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113408357B (en) * | 2021-05-21 | 2024-02-13 | 西北工业大学 | Method for generating flutter turbulence response signal to impulse response signal |
| CN113465732B (en) * | 2021-08-13 | 2022-04-05 | 重庆大学 | Vibration table structure displacement response prediction method and device based on EMD-DNN |
| CN113705649B (en) * | 2021-08-20 | 2024-01-12 | 哈尔滨医科大学 | EMD-SVD feature extraction-based hand tremor detection method and system |
| CN113822170B (en) * | 2021-08-31 | 2024-02-06 | 西安理工大学 | Method for identifying tremor of hands of body in non-stationary state |
| CN116466667A (en) * | 2023-04-20 | 2023-07-21 | 成都工业职业技术学院 | Intelligent control method, system and storage medium for part machining |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103786069A (en) * | 2014-01-24 | 2014-05-14 | 华中科技大学 | Flutter online monitoring method for machining equipment |
| CN106052854A (en) * | 2016-06-13 | 2016-10-26 | 浙江理工大学 | Grinding machine grinding chatter fault on-line diagnosis method |
| CN107330375A (en) * | 2017-06-05 | 2017-11-07 | 天津大学 | Lathe flutter on-line monitoring method based on empirical mode decomposition and Sample Entropy feature |
| CN110502991A (en) * | 2019-07-18 | 2019-11-26 | 武汉理工大学 | Internal combustion engine health monitor method and system based on random convolutional neural networks structure |
| CN110866448A (en) * | 2019-10-21 | 2020-03-06 | 西北工业大学 | Flutter signal analysis method based on convolutional neural network and short-time Fourier transform |
| CN111222458A (en) * | 2020-01-06 | 2020-06-02 | 浙江工业大学 | Rolling bearing fault diagnosis method based on ensemble empirical mode decomposition and convolutional neural network |
| CN111811617A (en) * | 2020-07-10 | 2020-10-23 | 杭州电子科技大学 | Liquid level prediction method based on short-time Fourier transform and convolutional neural network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7346461B2 (en) * | 2005-09-30 | 2008-03-18 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | System and method of analyzing vibrations and identifying failure signatures in the vibrations |
-
2020
- 2020-12-29 CN CN202011587834.9A patent/CN112580588B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103786069A (en) * | 2014-01-24 | 2014-05-14 | 华中科技大学 | Flutter online monitoring method for machining equipment |
| CN106052854A (en) * | 2016-06-13 | 2016-10-26 | 浙江理工大学 | Grinding machine grinding chatter fault on-line diagnosis method |
| CN107330375A (en) * | 2017-06-05 | 2017-11-07 | 天津大学 | Lathe flutter on-line monitoring method based on empirical mode decomposition and Sample Entropy feature |
| CN110502991A (en) * | 2019-07-18 | 2019-11-26 | 武汉理工大学 | Internal combustion engine health monitor method and system based on random convolutional neural networks structure |
| CN110866448A (en) * | 2019-10-21 | 2020-03-06 | 西北工业大学 | Flutter signal analysis method based on convolutional neural network and short-time Fourier transform |
| CN111222458A (en) * | 2020-01-06 | 2020-06-02 | 浙江工业大学 | Rolling bearing fault diagnosis method based on ensemble empirical mode decomposition and convolutional neural network |
| CN111811617A (en) * | 2020-07-10 | 2020-10-23 | 杭州电子科技大学 | Liquid level prediction method based on short-time Fourier transform and convolutional neural network |
Non-Patent Citations (1)
| Title |
|---|
| 一种基于 EM- KS 算法的连续变速颤振边界预测方法;刘俊豪 等;《西北工业大学学报》;第37卷(第6期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112580588A (en) | 2021-03-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112580588B (en) | Intelligent flutter signal identification method based on empirical mode decomposition | |
| CN110334580A (en) | The equipment fault classification method of changeable weight combination based on integrated increment | |
| CN110826791A (en) | Hybrid wind power prediction method based on long-time and short-time memory neural network | |
| CN110070102B (en) | Method for establishing sequence-to-sequence model for identifying power quality disturbance type | |
| CN112926265A (en) | Atmospheric porous probe measurement calibration method based on genetic algorithm optimization neural network | |
| CN105893669A (en) | Global simulation performance predication method based on data digging | |
| CN112597702B (en) | Pneumatic modeling generation type confrontation network model training method based on radial basis function | |
| CN112329339B (en) | Short-term wind speed prediction method for wind power plant | |
| CN110956312A (en) | Photovoltaic power distribution network voltage prediction method based on EMD-CNN deep neural network | |
| CN112149883A (en) | Photovoltaic power prediction method based on FWA-BP neural network | |
| CN111506868B (en) | Ultra-short-term wind speed prediction method based on HHT weight optimization | |
| CN113762625A (en) | Power distribution network state evaluation method and system based on graph convolution network | |
| CN113468817A (en) | Ultra-short-term wind power prediction method based on IGOA (optimized El-electric field model) | |
| CN110991689B (en) | Distributed photovoltaic power generation system short-term prediction method based on LSTM-Morlet model | |
| CN110879927A (en) | Sea clutter amplitude statistical distribution field modeling method for sea target detection | |
| CN114912666A (en) | Short-time passenger flow volume prediction method based on CEEMDAN algorithm and attention mechanism | |
| CN110675278A (en) | Photovoltaic power short-term prediction method based on RBF neural network | |
| CN106055761B (en) | A kind of load model parameters method of calibration based on decision tree classification | |
| Danilin et al. | Algorithm for determining optimum operation tolerances of memristor-based artificial neural networks | |
| CN112529328B (en) | Product performance prediction method and system | |
| CN113722951B (en) | Scatterer three-dimensional finite element grid optimization method based on neural network | |
| Li et al. | A short-term wind power forecasting method based on NWP wind speed fluctuation division and clustering | |
| CN114860931A (en) | Relay protection defect text grading method based on Voting Classifier model | |
| CN113536685A (en) | Neural network extrapolation-based wind speed probability model modeling method | |
| CN113111588A (en) | NO of gas turbineXEmission concentration prediction method and device |
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 |