CN113537152A - Flow field state fault detection method based on deep neural network - Google Patents
Flow field state fault detection method based on deep neural network Download PDFInfo
- Publication number
- CN113537152A CN113537152A CN202110952366.9A CN202110952366A CN113537152A CN 113537152 A CN113537152 A CN 113537152A CN 202110952366 A CN202110952366 A CN 202110952366A CN 113537152 A CN113537152 A CN 113537152A
- Authority
- CN
- China
- Prior art keywords
- neural network
- time
- frequency
- layer
- convolutional neural
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 18
- 238000001514 detection method Methods 0.000 title abstract description 5
- 238000000034 method Methods 0.000 claims abstract description 46
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 36
- 238000011176 pooling Methods 0.000 claims abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000012360 testing method Methods 0.000 claims abstract description 19
- 238000003062 neural network model Methods 0.000 claims abstract description 13
- 230000009466 transformation Effects 0.000 claims abstract description 9
- 238000003745 diagnosis Methods 0.000 claims description 19
- 238000000605 extraction Methods 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 5
- 238000011478 gradient descent method Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000012800 visualization Methods 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 abstract description 7
- 239000010410 layer Substances 0.000 description 39
- 230000006870 function Effects 0.000 description 14
- 238000001228 spectrum Methods 0.000 description 5
- 230000007547 defect Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 2
- 239000002356 single layer Substances 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- 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
- G06F18/211—Selection of the most significant subset of features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Signal Processing (AREA)
- Probability & Statistics with Applications (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a flow field state fault detection method based on a deep neural network model, which comprises the following steps: (1) collecting gamma photon signal data, segmenting the data and carrying out trend item removing pretreatment; (2) carrying out short-time Fourier time-frequency transformation analysis on the signal data to obtain time-frequency representation of each vibration signal, and displaying the time-frequency representation by using a pseudo color chart; (3) reducing the image resolution by an interpolation method and superposing all images to form a training sample and a test sample which are used as the input of a convolutional neural network; (4) constructing a deep convolutional neural network model, which comprises an input layer, two convolutional layers, two pooling layers, a full-link layer, a softmax classification layer and an output layer; (5) and introducing the training sample into the model for training, obtaining the convolution characteristic, the pooling characteristic and the neural network structure parameter, and diagnosing unknown fault signals according to the constructed deep neural network model. Compared with the existing time domain or frequency domain method, the method has better accuracy and stability.
Description
Technical Field
The invention belongs to the technical field of flow field state fault diagnosis, and relates to a flow field state fault diagnosis method based on a deep neural network structure.
Background
Gamma photon signals in the rocket engine cavity can be timely collected through equipment such as a gamma photon detector arranged on the outer ring of the engine cavity and the like, so that the flow field state in the engine cavity can be accurately reflected. How to extract effective information capable of accurately reflecting flow field state fault characteristics in the rocket engine from massive gamma photon signals and determine fault types and working states are main research contents of flow field state fault diagnosis. The working environment of rocket engine equipment containing fault characteristics is usually very complex, more complex parts of metal and machinery are provided, the background noise is strong, and gamma photon signals measured on site are usually multi-component and non-stable complex signals under strong background noise. Therefore, the signal processing method for fault feature extraction and fault diagnosis has higher requirements on accuracy and diagnosis effect.
In the prior art, a time domain feature extraction method or a frequency domain feature extraction method is generally adopted for gamma photon signal processing. Because the time domain characteristics can not reflect information such as frequency, frequency spectrum and the like, and the frequency domain method can only reflect local characteristic information in the vibration signal and can not reflect time-varying characteristic information, the useful information in the vibration signal can not be accurately, effectively and completely expressed by directly adopting a single time domain or frequency domain method. The extraction and identification of fault characteristic information are seriously influenced by the cross term interference in the traditional time-frequency analysis of wiener-Weili distribution and the like, and the fault mode identification method based on the deep convolutional neural network can effectively avoid the defects of the cross term interference and the like. With the rise of artificial intelligence methods such as machine learning, neural networks are attracting more and more attention of many researchers. The deep neural network structure can automatically learn effective characteristic features from input data through a hidden layer by an optimization algorithm such as a gradient descent method. Deep learning algorithms such as sparse coding, a Boltzman machine, a convolutional neural network and the like are widely applied to researches such as image processing, audio processing, character recognition and the like. And also gradually introduced into gamma photon signal feature extraction.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a flow field state fault detection method based on a deep convolutional neural network structure.
In order to solve the technical problems, the invention adopts the following technical scheme.
The invention discloses a flow field state fault detection method based on a deep convolutional neural network structure, which is characterized by comprising the following steps of:
eliminating trend in each vibration signal by utilizing cubic polynomial fitting
In the formula, yk、xkIs a sample point, a0、a1Is a coefficient, n, k is 1, 2, 3 … … n, wherein
The process of the step 2 is as follows:
writing a set of data samples into a matrix form SN×MRespectively selecting the optimal sampling rate, the threshold value and the time-frequency center, and performing short-time Fourier transform (STFT) by adopting the following formula to obtain time-frequency transform characteristic information data corresponding to each section of signal of the data sample set:
in the formula, f is frequency, tau is time, omega (tau-t) is a time-frequency window function, and x (tau) omega (tau-t) is a time-frequency center; and then, drawing a pseudo color image by using the time-frequency transformation characteristic information data obtained by calculation, carrying out visualization processing on the data, and superposing to form a three-dimensional data sample set.
And 3, reducing the image resolution by using an interpolation method, and superposing all the images to form a training sample set and a testing sample set which are used as the input of the convolutional neural network.
The process of the step 3 is as follows:
reducing the resolution of the time-frequency image without distortion by using a nearest neighbor interpolation algorithm to reduce the size of the time-frequency image to 64 multiplied by 64; and superposing all the time-frequency images, adding corresponding fault labels, and randomly selecting 50% as a training sample set and 50% as a testing sample set.
And 4, constructing a deep convolutional neural network model, wherein the deep convolutional neural network model comprises an input layer, two convolutional layers, two pooling layers, a full connection layer, a softmax classification layer and an output layer. The two convolutional layers and the two pooling layers are respectively overlapped in a cross mode, then the output characteristics of the second pooling layer are unfolded according to the full-connection mode, and the single layer softmax classification layer is connected to the full-connection layer to form the deep convolutional neural network model.
The process of the step 4 is as follows:
respectively and crossly superposing the two convolution layers and the two pooling layers to form a network main body structure, connecting the softmax classification layer to the full-connection layer to serve as a fault characteristic identification layer, and forming a complete deep convolution neural network model;
in the convolutional layer, for any input x, the subsequence is: x is the number ofj＝ρwjxj-1
In the formula, W and rho are convolution operation and nonlinear activation function respectively; w is ajIs the weight of the filter mapping, and each layer is written as the sum of the convolutions of the previous layer:
where u is an element of x, j is 1, 2, 3, …, N is the convolution:
in the formula, h is a mapping function, and g is an activation function.
Calculating a gradient through a convex optimization algorithm of a random gradient descent method, and optimizing and solving a weight Wj;
and (3) dividing the target area into non-overlapping areas by adopting a pooling method, selecting a pooling dimension of 4 and a neuron number of 300. Extracting features by adopting a maximum pooling function and improving the calculation efficiency of calculating a neural network;
extracting convolution characteristics and pooling characteristics of input data by using a deep convolution neural network structure model, and trainingTraining the structure parameters of the network model; the activation function at the Softmax layer adopts a sigmoid function:
and 5, importing the training sample set into a deep convolutional neural network model for training, obtaining convolutional characteristics, pooling characteristics and neural network structure parameters, and realizing flow field state fault diagnosis based on time-frequency characteristic extraction on unknown fault signals according to the built deep neural network model.
Further, the optimal sampling rate and the threshold value in step 2 are 512% and 5%, respectively.
Compared with the prior art, the invention has the advantages and beneficial effects that:
1. the invention takes the time-frequency representation of the vibration signal as the input of the deep neural network, intelligently learns the fault characteristics in the time-frequency representation, and classifies and identifies the fault characteristics. The time-frequency joint analysis method combines the advantages of time-domain and frequency-domain analysis characteristics, avoids the defect that a time-domain signal cannot reflect frequency spectrum characteristics, also avoids the defects that frequency spectrum leakage and a frequency-domain signal cannot reflect phase characteristics and time-varying characteristics of a gamma photon signal, can accurately reflect time-varying characteristics and frequency spectrum characteristics of the gamma photon signal, reflects energy concentration positions and distribution characteristics of the gamma photon signal, and has strong information capture and characterization capabilities.
Therefore, the time-frequency representation of the gamma photon signals is used as the identification target, the time-varying frequency spectrum characteristic and the energy characteristic of the vibration signals can be accurately extracted, the fault characteristic can be more favorably represented and extracted, and the characteristic information in the gamma photon signals can be more completely and accurately reflected.
2. The convolutional neural network gridds the identification target, extracts the target characteristics in blocks, is very sensitive to the difference and small change between adjacent grids in the identification target, and has excellent performance in the aspect of two-dimensional data characteristic learning.
3. The invention provides a flow field fault characteristic information extraction and intelligent diagnosis method based on the combination of short-time Fourier transform and a deep convolutional neural network model, which can effectively carry out efficient identification and judgment on the flow field state and fault information of a rocket engine and carry out state monitoring and health management on mechanical equipment.
Drawings
FIG. 1 is a flow chart of one method of practicing the present invention.
Fig. 2 is a diagram of a set of fault signals for different types of gamma photons and their time-frequency representation.
FIG. 3 is a diagram of a deep convolutional neural network structure model in accordance with one embodiment of the present invention.
FIG. 4 is a schematic diagram of the bearing fault convolution and pooling features extracted by the deep convolutional neural network model of an embodiment method of the present invention.
Detailed Description
The invention discloses a rotary machine fault feature intelligent diagnosis method based on a deep convolutional neural network structure, which comprises the following steps of: (1) collecting gamma photon signal test data, reasonably segmenting the data and carrying out trend item removing pretreatment; (2) carrying out short-time Fourier time-frequency transformation analysis on the acquired signal data to obtain time-frequency characteristics of each gamma photon signal, and displaying the time-frequency characteristics by using a pseudo-color image; (3) reducing the image resolution by an interpolation method and superposing all images to form a training sample and a test sample which are used as the input of a convolutional neural network; (4) constructing a deep convolutional neural network model, which comprises an input layer, two convolutional layers, two pooling layers, a full-link layer, a classification layer and an output layer; (5) and importing the training sample into a deep convolution neural network model for training, obtaining convolution characteristics, pooling characteristics and neural network structure parameters, and realizing the rotary machine fault diagnosis based on time-frequency characteristic extraction on unknown fault signals according to the built deep convolution neural network model. The invention takes the time-frequency representation of the vibration signal as the input of the deep neural network, intelligently learns the fault characteristic information in the time-frequency representation, and classifies and identifies the fault characteristics. The time-frequency representation of the rotating mechanical fault signal is subjected to fault diagnosis through the deep convolutional neural network structure, and compared with the existing time domain or frequency domain method, the method has better accuracy and stability.
The present invention will be described in further detail with reference to the accompanying drawings.
FIG. 1 is a flow chart of one method of practicing the present invention. As shown in fig. 1, the implementation method includes the following steps.
Step 1: collecting a fault vibration test signal of a rotary machine, and preprocessing an original vibration signal, wherein the method comprises the following specific processes:
the method comprises the steps of carrying out segmentation processing on collected original vibration signals, extracting M sections of signals from the vibration signals by adopting a random overlapping method, wherein each section of signals has N sample points, and forming a sample data set { s } by the extracted signalsN×MWherein s isj∈sN×1Indicating that the jth segment contains N data points;
eliminating trend in each vibration signal by utilizing cubic polynomial fitting
In the formula, yk、xkIs a sample point, a0、a1Is a coefficient, n, k is 1, 2, 3 … … n, wherein
To perform polynomial trend term elimination processing on the sampled data. In CaseWestrevers
Bearing fault test data of a univorsit bearing test center is taken as an example, 200 groups of samples are selected for each fault, each sample has 1024 points, and 10 faults are in total.
Step 2: extracting time-frequency characteristic information in vibration signal by using short-time Fourier time-frequency transformation
(2) Short-time Fourier time-frequency transformation analysis: writing the sample data set into a matrix form SN×MSetting the sampling rate and the threshold value as 512 and 5 percent respectively, setting the time-frequency window function as omega (tau-t), setting the time-frequency center as x (tau) omega (tau-t), and calculating the time-frequency representation of each section of signal by using the following short-time Fourier transform equation, wherein f is frequency and tau is time. And overlapping the obtained time-frequency transformation data to form a three-dimensional data set. And simultaneously, a time-frequency contour map can be calculated, and data can be visualized.
And f is frequency, tau is time, a pseudo color image is drawn by using the time-frequency transformation data obtained by calculation, time-frequency characteristic information is displayed, the data is visualized and superposed to form a three-dimensional data set.
And step 3: establishing a training set and a testing set:
and reducing the image resolution by using a nearest neighbor interpolation algorithm without distortion, reducing the image size to 64 multiplied by 64, reducing the size of time-frequency representation data, superposing all images, adding corresponding fault feature labels, wherein the label serial numbers are from 1 to 10, randomly selecting 50% of samples as a training set, and using the rest 50% of samples as a test set.
And 4, step 4: establishing a deep convolutional neural network structure model: and (3) connecting the single-layer softmax model to the full-connection layer by overlapping the two convolutional layers and the pooling layer in a cross mode to form a deep convolutional neural network model.
The convolutional neural network comprises two basic steps of convolutional characteristic extraction and pooling characteristic extraction, the two convolutional layers and the pooling layer are overlapped in a crossed mode to form a network main body structure, and the single softmax layer is connected to the full-connection layer and serves as a fault characteristic identification layer to form a complete deep convolutional neural network model. In the convolutional layer, for any input x, the subsequence is:
xj＝ρwjxj-1
in the formula, W and ρ are convolution operation and nonlinear activation function, respectively. w is ajAre the weights of the filter map, and each layer can be written as the sum of the convolutions of the previous layer:
in the formula, the operation is convolution:
in the formula, h is a mapping function, and g is an activation function.
Calculating gradient by a convex optimization algorithm of a random gradient descent method, and optimally solving weight wj。
The pooling method is used as a down-sampling strategy, a target area can be divided into non-overlapping areas, the pooling dimension is selected to be 4, the number of neurons is 300, the features are extracted by adopting a maximum pooling function, and the calculation efficiency of calculating the neural network is improved.
And extracting the convolution characteristic and the pooling characteristic of the input data by using the neural network structure model, and training the network model structure parameters. The activation function at the Softmax layer adopts a sigmoid function:
and 5: importing input data into a network model for fault diagnosis test
(5) And importing the training set into the constructed deep convolution neural network model, training the structural parameters of the neural network model, and extracting convolution characteristics and pooling characteristics. The accuracy of the test set reaches 100%, the test set is led into the trained model, and the classification accuracy of the test set reaches 99.75%.
Claims (6)
1. A flow field state fault diagnosis method based on a deep convolutional neural network structure specifically comprises the following steps:
step 1, collecting gamma photon signals and preprocessing the gamma photon signals;
step 2, extracting time-frequency characteristics of the vibration signals by using a short-time Fourier transform method;
step 3, reducing the image resolution by using an interpolation method and superposing all images to form a training sample set and a test sample set which are used as the input of a convolutional neural network;
step 4, constructing a deep convolutional neural network model, wherein the deep convolutional neural network model comprises an input layer, two convolutional layers, two pooling layers, a full connection layer, a softmax classification layer and an output layer;
and 5, importing the training sample set into a deep convolutional neural network model for training, obtaining convolutional characteristics, pooling characteristics and neural network structure parameters, and realizing flow field state fault diagnosis based on time-frequency characteristic extraction on unknown fault signals according to the built deep neural network model.
2. The flow field state fault diagnosis method based on the deep convolutional neural network structure as claimed in claim 1, wherein: the specific method for collecting gamma photon signals and preprocessing comprises the following steps:
carrying out segmentation processing on the acquired original gamma photon signals, and extracting M sections of signals from the vibration signals by adopting a random overlapping method, wherein each section of signals has N sample points; the extracted signals are combined into a sample data set { s }N×MWherein s isj∈sN×1Indicating that the jth segment contains N data points;
eliminating trend in each vibration signal by utilizing cubic polynomial fitting
In the formula, yk、xkIs a sample point, a0、a1Is a coefficient, n, k is 1, 2, 3 … … n, wherein
。
3. The flow field state fault diagnosis method based on the deep convolutional neural network structure as claimed in claim 1, wherein: the specific method for extracting the time-frequency characteristics of the vibration signal by using the short-time Fourier transform method comprises the following steps:
writing a set of data samples into a matrix form SN×MRespectively selecting the optimal sampling rate, the threshold value and the time-frequency center, and performing short-time Fourier transform (STFT) by adopting the following formula to obtain time-frequency transform characteristic information data corresponding to each section of signal of the data sample set:
in the formula, f is frequency, tau is time, omega (tau-t) is a time-frequency window function, and x (tau) omega (tau-t) is a time-frequency center; and then, drawing a pseudo color image by using the time-frequency transformation characteristic information data obtained by calculation, carrying out visualization processing on the data, and superposing to form a three-dimensional data sample set.
4. The flow field state fault diagnosis method based on the deep convolutional neural network structure as claimed in claim 1, wherein: reducing the image resolution by using an interpolation method, superposing all images to form a training sample set and a test sample set, wherein the specific method used as the input of the convolutional neural network comprises the following steps:
reducing the resolution of the time-frequency image without distortion by using a nearest neighbor interpolation algorithm to reduce the size of the time-frequency image to 64 multiplied by 64; and superposing all the time-frequency images, adding corresponding fault labels, and randomly selecting 50% as a training sample set and 50% as a testing sample set.
5. The flow field state fault diagnosis method based on the deep convolutional neural network structure as claimed in claim 1, wherein: the specific method for constructing the deep convolutional neural network model comprises the following steps of:
respectively and crossly superposing the two convolution layers and the two pooling layers to form a network main body structure, connecting the softmax classification layer to the full-connection layer to serve as a fault characteristic identification layer, and forming a complete deep convolution neural network model;
in the convolutional layer, for any input x, the subsequence is: x is the number ofj＝ρwjxj-1
In the formula, W and rho are convolution operation and nonlinear activation function respectively; w is ajIs the weight of the filter mapping, and each layer is written as the sum of the convolutions of the previous layer:
where u is an element of x, j is 1, 2, 3.
In the formula, h is a mapping function, and g is an activation function.
Calculating a gradient through a convex optimization algorithm of a random gradient descent method, and optimizing and solving a weight Wj;
and (3) dividing the target area into non-overlapping areas by adopting a pooling method, selecting a pooling dimension of 4 and a neuron number of 300. Extracting features by adopting a maximum pooling function and improving the calculation efficiency of calculating a neural network;
extracting convolution characteristics and pooling characteristics of input data by using a deep convolution neural network structure model, and training network model structure parameters; the activation function at the Softmax layer adopts a sigmoid function:
。
6. the flow field state fault diagnosis method based on the deep convolutional neural network structure as claimed in claim 1, wherein: the specific method for guiding the training sample set into the deep convolutional neural network model for training and obtaining convolutional characteristics, pooling characteristics and neural network structure parameters and realizing flow field state fault diagnosis based on time-frequency characteristic extraction on unknown fault signals according to the built deep neural network model is as follows:
leading the training set into the constructed deep convolution neural network model, training the structural parameters of the neural network model, and extracting convolution characteristics and pooling characteristics; the accuracy of the test set reaches 100%, the test set is led into the trained model, and the classification accuracy of the test set reaches 99.75%.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110952366.9A CN113537152A (en) | 2021-08-19 | 2021-08-19 | Flow field state fault detection method based on deep neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110952366.9A CN113537152A (en) | 2021-08-19 | 2021-08-19 | Flow field state fault detection method based on deep neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113537152A true CN113537152A (en) | 2021-10-22 |
Family
ID=78091246
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110952366.9A Pending CN113537152A (en) | 2021-08-19 | 2021-08-19 | Flow field state fault detection method based on deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113537152A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114065822A (en) * | 2021-11-30 | 2022-02-18 | 中国海洋大学 | Electromagnetic identification method and system for ocean tide fluctuation |
| CN115311530A (en) * | 2022-07-15 | 2022-11-08 | 哈尔滨工业大学 | Flow field feature extraction method based on depth convolution self-encoder |
| CN115546498A (en) * | 2022-09-28 | 2022-12-30 | 大连海事大学 | Deep learning-based compression storage method for flow field time-varying data |
| CN118468192A (en) * | 2024-07-10 | 2024-08-09 | 山东大学 | Power distribution network abnormal data detection method and system based on lightweight neural network |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108830127A (en) * | 2018-03-22 | 2018-11-16 | 南京航空航天大学 | A kind of rotating machinery fault feature intelligent diagnostic method based on depth convolutional neural networks structure |
| US20210048487A1 (en) * | 2019-08-12 | 2021-02-18 | Wuhan University | Power transformer winding fault positioning method based on deep convolutional neural network integrated with visual identification |
-
2021
- 2021-08-19 CN CN202110952366.9A patent/CN113537152A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108830127A (en) * | 2018-03-22 | 2018-11-16 | 南京航空航天大学 | A kind of rotating machinery fault feature intelligent diagnostic method based on depth convolutional neural networks structure |
| US20210048487A1 (en) * | 2019-08-12 | 2021-02-18 | Wuhan University | Power transformer winding fault positioning method based on deep convolutional neural network integrated with visual identification |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114065822A (en) * | 2021-11-30 | 2022-02-18 | 中国海洋大学 | Electromagnetic identification method and system for ocean tide fluctuation |
| CN114065822B (en) * | 2021-11-30 | 2024-04-26 | 中国海洋大学 | Electromagnetic identification method and system for ocean tide fluctuation |
| CN115311530A (en) * | 2022-07-15 | 2022-11-08 | 哈尔滨工业大学 | Flow field feature extraction method based on depth convolution self-encoder |
| CN115311530B (en) * | 2022-07-15 | 2023-08-29 | 哈尔滨工业大学 | Flow field feature extraction method based on depth convolution self-encoder |
| CN115546498A (en) * | 2022-09-28 | 2022-12-30 | 大连海事大学 | Deep learning-based compression storage method for flow field time-varying data |
| CN115546498B (en) * | 2022-09-28 | 2023-10-17 | 大连海事大学 | Flow field time-varying data compression storage method based on deep learning |
| CN118468192A (en) * | 2024-07-10 | 2024-08-09 | 山东大学 | Power distribution network abnormal data detection method and system based on lightweight neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108830127B (en) | Rotary machine fault feature intelligent diagnosis method based on deep convolutional neural network structure | |
| CN113537152A (en) | Flow field state fault detection method based on deep neural network | |
| CN110059601B (en) | Intelligent fault diagnosis method for multi-feature extraction and fusion | |
| Li et al. | Multiscale local features learning based on BP neural network for rolling bearing intelligent fault diagnosis | |
| CN111898634B (en) | Intelligent fault diagnosis method based on depth-to-reactance-domain self-adaption | |
| CN109765053B (en) | Rolling bearing fault diagnosis method using convolutional neural network and kurtosis index | |
| Ma et al. | A novel bearing fault diagnosis method based on 2D image representation and transfer learning-convolutional neural network | |
| CN112418013B (en) | Complex working condition bearing fault diagnosis method based on meta-learning under small sample | |
| CN110751073A (en) | Pipeline early damage mode identification method based on distributed optical fiber sensing and deep learning | |
| CN112101085B (en) | Intelligent fault diagnosis method based on importance weighted domain antagonism self-adaptation | |
| CN106556781A (en) | Shelf depreciation defect image diagnostic method and system based on deep learning | |
| CN109782603A (en) | The detection method and monitoring system of rotating machinery coupling fault | |
| CN105841961A (en) | Bearing fault diagnosis method based on Morlet wavelet transformation and convolutional neural network | |
| CN111751133B (en) | Intelligent fault diagnosis method of deep convolutional neural network model based on non-local mean embedding | |
| CN111753891B (en) | Rolling bearing fault diagnosis method based on unsupervised feature learning | |
| CN111238814A (en) | Rolling bearing fault diagnosis method based on short-time Hilbert transform | |
| CN112906644A (en) | Mechanical fault intelligent diagnosis method based on deep migration learning | |
| CN108509701B (en) | Rotary machine fault direct intelligent diagnosis method based on vibration signal | |
| CN112364706A (en) | Small sample bearing fault diagnosis method based on class imbalance | |
| CN113076920B (en) | Intelligent fault diagnosis method based on asymmetric domain confrontation self-adaptive model | |
| CN114034481A (en) | Fault diagnosis system and method for rolling mill gearbox | |
| CN112116029A (en) | Intelligent fault diagnosis method for gearbox with multi-scale structure and characteristic fusion | |
| CN116011507A (en) | Rare fault diagnosis method for fusion element learning and graph neural network | |
| Xue et al. | A novel framework for motor bearing fault diagnosis based on multi-transformation domain and multi-source data | |
| CN116451133A (en) | Heavy-duty robot speed reducer fault diagnosis method based on graph structure information |
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 |