CN115186692A - Intelligent fault diagnosis method based on empirical mode decomposition and image conversion - Google Patents
Intelligent fault diagnosis method based on empirical mode decomposition and image conversion Download PDFInfo
- Publication number
- CN115186692A CN115186692A CN202111561899.0A CN202111561899A CN115186692A CN 115186692 A CN115186692 A CN 115186692A CN 202111561899 A CN202111561899 A CN 202111561899A CN 115186692 A CN115186692 A CN 115186692A
- Authority
- CN
- China
- Prior art keywords
- mode decomposition
- empirical mode
- fault
- data
- image conversion
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses an intelligent fault diagnosis method based on empirical mode decomposition and image conversion, which comprises the following steps: s1, pretreatment: acquiring vibration data of petrochemical rotating machinery faults through an EMT490 sensor, wherein each 1024 vibration data is a group of original vibration signal data; s2, complementary integration empirical mode decomposition: decomposing the collected original vibration signals into high-efficiency intrinsic mode functions for further data processing and bearing fault diagnosis; s3, image conversion: acquiring a data segment of the processed original vibration signal, and constructing a gray image to be used as two-dimensional input of a convolutional neural network; s4, identifying a convolutional neural network: and extracting depth features by the convolutional neural network for fault evaluation and detection. The invention can improve more new and various means for detecting mechanical faults in the petrochemical production process for practitioners in the petrochemical industry, and avoids major accidents caused by faults of petrochemical unit equipment.
Description
Technical Field
The invention belongs to the technical field of fault diagnosis, and particularly relates to an intelligent fault diagnosis method based on empirical mode decomposition and image conversion.
Background
In the times of rapid industrial development nowadays, the diagnosis and prediction method of the bearing health condition is always extremely emphasized by the modern petrochemical industry. The rotary machine is an important component of a petrochemical unit, and works under complex and diverse environments such as high temperature, high pressure and the like, so that a vibration monitoring signal has ambiguity, instability, nonlinearity and multi-coupling, and particularly weak impact characteristics of early damage of the rotary machine are difficult to extract, so that bearing faults are difficult to identify, and a challenging task is brought to fault signal analysis. The existing bearing fault diagnosis method based on time-frequency analysis has the defects of dependence on expert knowledge, insensitivity in fault feature extraction, low diagnosis precision on nonlinear and non-stable signals and the like, and cannot meet the fault diagnosis requirement, so that safety accidents of petrochemical enterprises frequently occur. Therefore, a new petrochemical unit fault diagnosis method is urgently needed to be explored, and the fault diagnosis problem of the petrochemical rotating machinery can be really solved so as to maintain the normal operation and safety of the petrochemical unit. To this end, we propose an intelligent fault diagnosis method based on empirical mode decomposition and image transformation to solve the above mentioned problems in the background art.
Disclosure of Invention
The invention aims to provide an intelligent fault diagnosis method based on empirical mode decomposition and image conversion, so as to solve the problems in the background technology.
In order to achieve the purpose, the invention provides the following technical scheme: an intelligent fault diagnosis method based on empirical mode decomposition and image conversion is characterized in that in preprocessing, bearing fault vibration signals are processed by complementary integrated empirical mode decomposition and image conversion and are combined with a convolutional neural network, and the method comprises the following steps:
s1, pretreatment: acquiring vibration data of petrochemical rotating machinery faults through an EMT490 sensor, wherein each 1024 vibration data is a group of original vibration signal data;
s2, complementary integration empirical mode decomposition: decomposing the collected original vibration signals into efficient intrinsic mode functions for further data processing and bearing fault diagnosis;
s3, image conversion: acquiring a data segment of the processed original vibration signal, and constructing a gray image to be used as two-dimensional input of a convolutional neural network;
s4, identifying a convolutional neural network: and extracting depth features by the convolutional neural network for fault evaluation and detection.
The complementary integration empirical mode decomposition of the step S2 is to add white noise and reduce white noise on the basis of an original vibration signal and complete empirical mode decomposition, and calculate the average value of two groups of decomposed intrinsic modes;
after the two groups of decomposed natural modes are averaged, the i groups of natural modes are averaged to obtain the final natural mode component, so that the reconstructed signal noise is reduced.
The image conversion in step S3 is to fill the time domain signals obtained after the complementary integrated empirical mode decomposition processing into the image pixels in sequence, where the length of data in the time domain signals is L, and each sampling is performed by a length of T 2 The data segment of (2) constructing a gray scale image having a pixel size of T x T, wherein each image is normalized from 0 to 255, respectively.
Dividing the processed original vibration signal into a normal sample, a fault sample 2, a fault sample 3 and a fault sample 4 … … as a sampling data segment, wherein the data length is in the time domain of L, and sampling T each time 2 Thereby constructing a gray scale image having a pixel size of T x T.
And taking the processed original oscillation signal as a sampling data segment and dividing the sampling data segment into a normal sample, a fault sample 2, a fault sample 3 and a fault sample 4 … … fault sample n, normalizing the gray level images from 0 to 255 respectively, and finally obtaining the pixel intensity of the image to eliminate expert experience.
The model of the convolutional neural network is subjected to parameter adjustment on the basis of LeNet-5, convolution and pooling operation are carried out, and finally a softmax function is used for outputting an identification result;
the convolutional neural network contains 2 alternating convolutional and pooling layers, and 3 fully-connected layers, and chooses ReLU as the activation function of the model.
Compared with the prior art, the invention has the beneficial effects that: the invention provides a method for processing insensitive and nonlinear non-stationary signal fault signals for the first time, which has the characteristics of accurately and quickly identifying different fault types under different conditions, and has excellent performances such as effectiveness and superiority. The invention firstly provides an intelligent fault diagnosis method based on the combination of empirical mode decomposition and image conversion, and compared with the traditional deep learning method, the intelligent fault diagnosis method comprises a recurrent neural network. In a word, the invention can improve more new and various means for detecting mechanical faults in the petrochemical production process for petrochemical industry practitioners, improve the discrimination and prevention capability of the petrochemical industry practitioners on accidents occurring in the petrochemical production process, avoid major accidents caused by faults of petrochemical unit equipment, and ensure the safety of normal production and working in the petrochemical industry.
Drawings
FIG. 1 is a schematic flow chart of an intelligent fault diagnosis method based on empirical mode decomposition and image transformation;
FIG. 2 is a schematic diagram of complementary integrated empirical mode decomposition (CEEMD) according to the present invention;
FIG. 3 is a schematic diagram of convolutional neural network identification according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention provides an intelligent fault diagnosis method based on empirical mode decomposition and image conversion, which adopts complementary integrated empirical mode decomposition and image conversion to process a bearing fault vibration signal in preprocessing and combines with a convolutional neural network, and comprises the following steps:
s1, pretreatment: acquiring vibration data of petrochemical rotating machinery faults through an EMT490 sensor, wherein each 1024 vibration data is a group of original vibration signal data;
s2, complementary integration empirical mode decomposition: decomposing the collected original vibration signals into efficient intrinsic mode functions for further data processing and bearing fault diagnosis;
s3, image conversion: acquiring a data segment of the processed original vibration signal, and constructing a gray image to be used as two-dimensional input of a convolutional neural network;
s4, identifying a convolutional neural network: and extracting depth features by the convolutional neural network for fault evaluation and detection.
The complementary integrated empirical mode decomposition in the step S2 is to add white noise and reduce white noise on the basis of an original vibration signal and complete the empirical mode decomposition, and calculate the average value of two groups of decomposed intrinsic modes;
after the two groups of decomposed natural modes are averaged, the i groups of natural modes are averaged to obtain the final natural mode component, so that the reconstructed signal noise is reduced.
The image conversion in step S3 is to fill the time domain signals obtained after the complementary integrated empirical mode decomposition processing into the image pixels in sequence, where the length of the data in the time domain signals is L, and each sampling has a length of T 2 The data segment of (2) constructing a gray scale image of pixel size T x T, wherein each image is divided intoRespectively normalized from 0 to 255.
Dividing the processed original vibration signal into a normal sample, a fault sample 2, a fault sample 3 and a fault sample 4 … … as a sampling data segment, wherein the data length is in the time domain of L, and sampling T each time 2 Thereby constructing a gray scale image having a pixel size of T x T.
And taking the processed original oscillation signal as a sampling data segment and dividing the sampling data segment into a normal sample, a fault sample 2, a fault sample 3 and a fault sample 4 … … fault sample n, normalizing the gray level images from 0 to 255 respectively, and finally obtaining the pixel intensity of the image to eliminate expert experience.
The model of the convolutional neural network is subjected to parameter adjustment on the basis of LeNet-5, convolution and pooling operation are carried out, and finally a softmax function is used for outputting an identification result;
the convolutional neural network contains 2 alternating convolutional and pooling layers, and 3 fully-connected layers, and chooses ReLU as the activation function of the model.
Step S2, complementary integrated empirical mode decomposition is performed according to data acquisition in the preprocessing in step S1, and is used for converting the original vibration signal into different modal functions imf.1, imf.2, imf.3, and imf.4.... Imf.n through the complementary integrated empirical mode decomposition, and then performing image conversion on the processed vibration signal, and dividing the time domain signal with the length L in the time domain into a normal sample, a fault sample 2, a fault sample 3, and a fault sample 4 … … fault sample n, thereby constructing a gray level image and performing normalization.
And finally, training and testing the gray level image in a convolutional neural network to obtain an output result. The data set acquisition 1 is combined with the complementary integrated empirical mode decomposition 2 and is used for measuring and processing different fault types acquired at the same position of a machine, and each group of data obtained contains 1024 fault acquisition points. The convolutional neural network identifies 4 and finally constructed gray level images, wherein the gray level images comprise 4 alternating convolutional and pooling layers and 3 complete connection layers, and ReLU is selected as an activation function of the model and is used for fault identification and classification.
In the embodiment of the invention, the data acquisition 1 acquires the vibration data of the fault through the EMT490 sensor, and each 1024 vibration data is a group of data. And processing the acquired data set by adopting complementary integrated empirical mode decomposition, wherein the complementary integrated empirical mode decomposition is optimization of the integrated empirical mode decomposition, the lumped average times can be reduced from hundreds of orders of magnitude of the integrated empirical mode decomposition to tens of orders of magnitude of the complementary integrated empirical mode decomposition, and the reconstructed signal noise is obviously reduced. Furthermore, the empirical mode decomposition does not need a basis function, and the complexity of setting the basis function is avoided. The signal may be decomposed into several natural modal functions using empirical mode decomposition, each of which may be decomposed to describe the dynamic characteristics of the original signal.
And taking the processed original vibration signal as a sampling data segment, dividing the sampling data segment into a normal sample, a fault sample 2, a fault sample 3 and a fault sample 4 … … fault sample n, and establishing a gray level image through image conversion.
And finally, selecting a convolutional neural network with ReLU as an activation function of the model, and finally identifying the result of 4 diagnosis outputs through a softmax function.
As an optional implementation mode, the intelligent fault diagnosis method based on the combination of empirical mode decomposition and image conversion has the advantages that the petrochemical rotating machinery can be accurately and quickly detected under various complex environments, fault categories can be accurately identified under various complex environments, original signals are effectively decomposed into inherent modal functions by complementary integrated empirical mode decomposition, each inherent modal function can be decomposed to describe the dynamic characteristics of the original signals, the modal aliasing problem does not exist, and the problem that diagnosis results are uncertain is solved. And by adopting an optimization model of integrated empirical mode decomposition, namely complementary integrated empirical mode decomposition, the lumped average frequency is reduced from hundreds of orders of magnitude of integrated empirical mode decomposition to tens of orders of magnitude of complementary integrated empirical mode decomposition, and the signal noise after reconstruction is obviously reduced. And the empirical mode decomposition is suitable for any data and is decomposed based on the data, and a basis function is not needed. Therefore, the invention can diagnose the equipment faults of various petrochemical industries in an all-around and all-weather manner.
Obtaining a natural mode function signal: the signal data set is obtained through complementary integrated empirical mode decomposition, the optimization of the integrated empirical mode decomposition is complementary integrated empirical mode decomposition, the lumped average times of the integrated empirical mode decomposition are reduced, the integrated empirical mode decomposition is reduced from hundreds of orders of magnitude to dozens of orders of magnitude of the complementary integrated empirical mode decomposition, and the noise of the reconstructed signal is obviously reduced. Empirical mode decomposition is applicable to any data, and does not require basis functions, only decomposition based on the data itself. The signal may be decomposed into natural modal functions using empirical mode decomposition, each of which may be decomposed to describe the dynamics of the original signal. The complementary integrated empirical mode decomposition algorithm is to decompose the original signal plus white noise and the original signal minus white noise through empirical modes at the same time, and calculate the average value, which is used for offsetting the noise added in the signal, and is different from the integrated empirical mode decomposition which only adds white noise each time.
And S (k) is an original signal, different white noises are added, and the sum of the times is i, namely the lumped average time is i. And respectively carrying out empirical mode decomposition on the signals with the noise added and the noise reduced, averaging two groups of decomposed intrinsic mode functions, and averaging the intrinsic mode functions of the i groups to obtain the final intrinsic mode function component. The invention provides a method for solving the problem of mode aliasing by effectively decomposing a fault vibration signal into an inherent mode function.
Complementary integrated empirical mode decomposition is added to reconstruct an original vibration signal and decompose the original vibration signal into a plurality of inherent modal functions, so that the noise of the reconstructed signal is obviously reduced, the decomposition efficiency is effectively improved, and the problems of large error and poor decomposition completeness of the integrated empirical mode decomposition reconstruction are solved.
The natural mode functions may be chosen to yield different numbers depending on the parameters. The invention performs experiments on different decomposition numbers of the inherent mode functions. As the number of decompositions of the natural mode function increases, the average accuracy of the fault does not increase all the time, but decreases significantly by a certain value. When the number of the decompositions of the inherent mode function is 3, the average diagnosis precision is the highest and reaches 99.17 percent. When the number of the natural mode function decompositions is more than 3, the average diagnosis precision is remarkably reduced, and the time consumption is longer and longer. In order to balance average precision and time consumption, the invention selects a scheme with 3 inherent mode function decompositions. The average diagnosis precision of the scheme is 99.17%, the consumption time of a training model is 297s, and the method has good performance.
Image conversion: receiving a vibration signal processed by the complementary integration empirical mode decomposition 2, and then classifying the sample data set into a sample data set classification 31, wherein the sample data set classification 31 is divided into a normal sample, a fault sample 2, a fault sample 3 and a fault sample 4 … … fault sample n. The method aims to convert an original signal into data of a gray image for preprocessing and eliminate expert experience.
The invention uses a method for converting a time sequence into a picture, firstly, data with the length of L in a time domain is sampled, and the length of T is 2 Then a gray scale image with pixels of T x T is constructed.
The convolutional neural network identification outputs a diagnosis result by using improved LeNet-5 identification.
The neural network model of the invention is improved in original LeNet-5, and a training model based on LeNet-5 is designed so as to identify the output result of the model. In LeNet-5, the model contains two alternating convolutional and pooling layers, respectively, and one fully connected layer. In the convolutional neural network model proposed in the present invention, 4 alternating convolutional and pooling layers are included, and 3 fully-connected layers are used for the image.
The invention finely adjusts the parameters on the basis of LeNet-5: 1) Number of layers of each convolutional layer and pooling layer: the number of the first layer and the third layer of the convolution layer is respectively as follows: 40 60, the number of the layers of the pooling layers of the second layer and the fourth layer is respectively as follows: 60 80, 80; 2) The number of the complete connection layers is 3, and the number of each layer is 2650, 1024 and 128; 3) The Dropout function is added after the second fully connected layer, setting the parameter Dropout to 0.5, i.e., discarding fifty percent of the neurons.
The present invention selects ReLU as the activation function of the model. Compared with the traditional activation function, the ReLU activation function can effectively avoid the problem of gradient disappearance in the aspect of gradient avoidance. In the definition of the optimizer, sparse class cross entropy is used as a loss function of the model herein. Overall, the fitting performance of the training model is better.
In summary, compared with the prior art, the invention provides a method for processing insensitive and nonlinear non-stationary signal fault signals for the first time, and the method has the characteristics of accurately and quickly identifying different fault types under different conditions, and has excellent performances such as effectiveness and superiority. The invention firstly provides an intelligent fault diagnosis method based on the combination of empirical mode decomposition and image conversion, and compared with the traditional deep learning method, the intelligent fault diagnosis method comprises a recurrent neural network. In a word, the invention can improve more new and various means for detecting mechanical faults in the petrochemical production process for petrochemical industry practitioners, improve the discrimination and prevention capability of the petrochemical industry practitioners on accidents occurring in the petrochemical production process, avoid major accidents caused by faults of petrochemical unit equipment, and ensure the safety of normal production and working in the petrochemical industry.
Finally, it should be noted that: although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that modifications may be made to the embodiments or portions thereof without departing from the spirit and scope of the invention.
Claims (6)
1. An intelligent fault diagnosis method based on empirical mode decomposition and image conversion is characterized in that in preprocessing, bearing fault vibration signals are processed by complementary integrated empirical mode decomposition and image conversion and are combined with a convolutional neural network, and the method comprises the following steps: the method comprises the following steps:
s1, pretreatment: acquiring vibration data of petrochemical rotating machinery faults through an EMT490 sensor, wherein each 1024 vibration data is a group of original vibration signal data;
s2, complementary integration empirical mode decomposition: decomposing the collected original vibration signals into efficient intrinsic mode functions for further data processing and bearing fault diagnosis;
s3, image conversion: acquiring a data segment of the processed original vibration signal, and constructing a gray image to be used as two-dimensional input of a convolutional neural network;
s4, identifying a convolutional neural network: and extracting depth features by the convolutional neural network for fault evaluation and detection.
2. The intelligent fault diagnosis method based on empirical mode decomposition and image conversion according to claim 1, characterized in that: the complementary integration empirical mode decomposition of the step S2 is to add white noise and reduce white noise on the basis of an original vibration signal and complete empirical mode decomposition, and calculate the average value of two groups of decomposed intrinsic modes; after the two groups of decomposed natural modes are averaged, the i groups of natural modes are averaged to obtain the final natural mode component, so that the reconstructed signal noise is reduced.
3. The intelligent fault diagnosis method based on empirical mode decomposition and image conversion according to claim 2, characterized in that: the image conversion in step S3 is to fill the time domain signals obtained after the complementary integrated empirical mode decomposition processing into the image pixels in sequence, where the length of the data in the time domain signals is L, and each sampling has a length of T 2 The data segment of (2) constructing a gray scale image having a pixel size of T x T, wherein each image is normalized from 0 to 255, respectively.
4. The intelligent fault diagnosis method based on empirical mode decomposition and image conversion according to claim 3, characterized in that: using the processed original vibration signal as the samplingThe sample data segment is divided into a normal sample, a fault sample 2, a fault sample 3 and a fault sample 4 … …, the fault sample n has the data length of L in the time domain, and T is sampled every time 2 The data segments of (a) thus construct a gray scale image of pixel size T x T.
5. The intelligent fault diagnosis method based on empirical mode decomposition and image conversion according to claim 4, characterized in that: and taking the processed original oscillation signal as a sampling data segment and dividing the sampling data segment into a normal sample, a fault sample 2, a fault sample 3 and a fault sample 4 … … fault sample n, normalizing the gray level image from 0 to 255 respectively, and finally obtaining the pixel intensity of the image to eliminate expert experience.
6. The intelligent fault diagnosis method based on empirical mode decomposition and image conversion according to claim 1, characterized in that: the model of the convolutional neural network is subjected to parameter adjustment on the basis of LeNet-5, convolution and pooling operation are carried out, and finally a softmax function is used for outputting an identification result; the convolutional neural network contains 2 alternating convolutional and pooled layers, and 3 fully-connected layers, and selects ReLU as the activation function of the model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111561899.0A CN115186692A (en) | 2021-12-16 | 2021-12-16 | Intelligent fault diagnosis method based on empirical mode decomposition and image conversion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111561899.0A CN115186692A (en) | 2021-12-16 | 2021-12-16 | Intelligent fault diagnosis method based on empirical mode decomposition and image conversion |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115186692A true CN115186692A (en) | 2022-10-14 |
Family
ID=83511357
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111561899.0A Pending CN115186692A (en) | 2021-12-16 | 2021-12-16 | Intelligent fault diagnosis method based on empirical mode decomposition and image conversion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115186692A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115712065A (en) * | 2023-01-05 | 2023-02-24 | 湖南大学 | Motor fault diagnosis method and system with perception matching of time-frequency revolving door and convolution kernel |
-
2021
- 2021-12-16 CN CN202111561899.0A patent/CN115186692A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115712065A (en) * | 2023-01-05 | 2023-02-24 | 湖南大学 | Motor fault diagnosis method and system with perception matching of time-frequency revolving door and convolution kernel |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | A method for rapidly evaluating reliability and predicting remaining useful life using two-dimensional convolutional neural network with signal conversion | |
| Jiao et al. | Deep coupled dense convolutional network with complementary data for intelligent fault diagnosis | |
| Cai et al. | Fault detection and diagnostic method of diesel engine by combining rule-based algorithm and BNs/BPNNs | |
| Kong et al. | An enhanced sparse representation-based intelligent recognition method for planet bearing fault diagnosis in wind turbines | |
| Zhou et al. | Mechanical equipment fault diagnosis based on redundant second generation wavelet packet transform | |
| Yuan et al. | Intelligent fault diagnosis of rolling element bearings based on HHT and CNN | |
| Tang et al. | A novel convolutional neural network for low-speed structural fault diagnosis under different operating condition and its understanding via visualization | |
| Wen et al. | Graph modeling of singular values for early fault detection and diagnosis of rolling element bearings | |
| Dou et al. | A rule-based intelligent method for fault diagnosis of rotating machinery | |
| Martinez-Garcia et al. | Visually interpretable profile extraction with an autoencoder for health monitoring of industrial systems | |
| CN111678699B (en) | Early fault monitoring and diagnosing method and system for rolling bearing | |
| CN102155988A (en) | Equipment monitoring and diagnosing method | |
| CN114818774A (en) | Intelligent gearbox fault diagnosis method based on multi-channel self-calibration convolutional neural network | |
| CN114034481A (en) | Fault diagnosis system and method for rolling mill gearbox | |
| CN110632484A (en) | ELM-based GIS partial discharge defect diagnosis and classification system and method | |
| Neupane et al. | Deep learning-based bearing fault detection using 2-D illustration of time sequence | |
| Lu et al. | Bearing fault diagnosis based on clustering and sparse representation in frequency domain | |
| KR102240775B1 (en) | Deep learning-based apparatus and method for determining breakdown of power transfer device using noise data | |
| CN115186692A (en) | Intelligent fault diagnosis method based on empirical mode decomposition and image conversion | |
| Lu et al. | Early fault warning and identification in condition monitoring of bearing via wavelet packet decomposition coupled with graph | |
| Ning et al. | An intelligent device fault diagnosis method in industrial internet of things | |
| CN116842379A (en) | Mechanical bearing residual service life prediction method based on DRSN-CS and BiGRU+MLP models | |
| CN113758708B (en) | Rolling bearing signal frequency domain fault diagnosis method based on L1 norm and group norm constraint | |
| CN115828088A (en) | High-voltage shunt reactor vibration abnormity fault diagnosis method based on deep learning | |
| Yanqiang et al. | Fault diagnosis of bearing based on LMD and MSE |
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 |