CN118228174A - Pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV2 - Google Patents
Pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV2 Download PDFInfo
- Publication number
- CN118228174A CN118228174A CN202410339202.2A CN202410339202A CN118228174A CN 118228174 A CN118228174 A CN 118228174A CN 202410339202 A CN202410339202 A CN 202410339202A CN 118228174 A CN118228174 A CN 118228174A
- Authority
- CN
- China
- Prior art keywords
- data
- gafs
- pneumatic valve
- time sequence
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000003745 diagnosis Methods 0.000 title claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 9
- 238000012545 processing Methods 0.000 claims abstract description 4
- 238000005070 sampling Methods 0.000 claims abstract description 4
- 230000008569 process Effects 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 10
- 238000001514 detection method Methods 0.000 claims description 6
- 238000011176 pooling Methods 0.000 claims description 6
- 230000004927 fusion Effects 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 238000013145 classification model Methods 0.000 claims description 3
- 230000004044 response Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 7
- 238000010801 machine learning Methods 0.000 abstract description 7
- 230000006870 function Effects 0.000 abstract description 6
- 238000000605 extraction Methods 0.000 description 7
- 230000007246 mechanism Effects 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention belongs to the technical field of fault diagnosis based on machine learning, and discloses a pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV2, which comprises the following steps of firstly, sampling a plurality of signals on a data time domain of pneumatic valve faults; step two, converting the acquired data into a two-dimensional image in a manner of a gram angle field GAFs according to the time sequence data of each single variable; step three, fusing each image into one image through a cv2.add () image processing function so as to determine a sample; inputting the fused image into a ConvNeXt V model for training; and fifthly, predicting by using the trained model to obtain a pneumatic valve fault diagnosis result. The invention obtains better diagnosis effect when training through ConvNeXt V network structure.
Description
Technical Field
The invention belongs to the technical field of fault diagnosis based on machine learning, and particularly relates to a pneumatic valve fault diagnosis method based on GAFs multiple time sequence-to-two-dimensional image and ConvNeXtV.
Background
For pneumatic valve fault diagnosis, traditional modes of 'post maintenance' and 'periodic maintenance' cannot immediately obtain the real-time running state of mechanical equipment in a large industrial place. This results in a large amount of human resources being consumed in the daily maintenance work, easily causing safety accidents, and failing to cope with gradual or sudden failure situations in time. At present, by monitoring the running state of the equipment and predicting the possible fault type and time point of the equipment according to the monitoring information, on-site technicians can make judgment in time, so that the occurrence rate of safety accidents is reduced. Under the push of rapid development of industrial Internet, computer technology and artificial intelligence, human fault diagnosis is gradually replaced by intelligent fault diagnosis. The fault diagnosis method can be mainly classified into a method based on a mechanism model and a method based on data driving.
The method based on the mechanism model is to build a dynamic model for the pneumatic adjusting valve through a system identification model. When faults occur, whether faults occur or not is judged according to the input and output parameter changes of an observation model, the model is built on a process system capable of being quantified digitally, when the process is linear, modeling is easy, but modeling is difficult for a nonlinear process, so that the defect that the nonlinearity of a physical pneumatic valve and boundary conditions are difficult to consider exists in a diagnosis method based on a mechanism model.
The data driving-based method is to collect a large amount of physical valve fault data in the early stage and diagnose fault types by a machine learning method. At present, in the early stage of machine learning, after feature extraction is mainly performed on original data, valve faults are diagnosed through methods such as a support vector machine, a multi-layer perceptron and a decision tree. However, this method requires extraction of features from the raw data, which is time consuming and the different features have a significant impact on the final result.
At present, most of the pneumatic valve fault diagnosis methods adopt a traditional machine learning algorithm, firstly, the original data is subjected to feature extraction, the data is subjected to dimension reduction treatment, and then the fault classification is carried out through a classification algorithm to realize the fault classification diagnosis. Besides the fact that the prior data is subjected to feature extraction, a great deal of time is required to be spent for carrying out extraction mode comparison and searching for an optimal extraction method, and the phenomenon of losing the features of the original data is easily caused in the process of reducing the high-dimensional data to the low-dimensional data. And how to measure the effective dimensionality reduction of the original data is also a problem. For a small number of combinations of deep learning, the deep learning mode is realized by modifying the original model to realize that the model supports diagnosis of data of one-dimensional data, and for multi-element time sequence data, the deep learning mode is difficult to be adopted.
Disclosure of Invention
The invention aims to provide a pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV2, which aims to solve the technical problems.
In order to solve the technical problems, the pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV2 provided by the invention has the following specific technical scheme:
A pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV2 comprises the following steps:
step one, sampling a plurality of signals in a data time domain of a pneumatic valve fault;
Step two, converting the acquired data into a two-dimensional image in a manner of a gram angle field GAFs according to the time sequence data of each single variable;
Step three, fusing each image into one image through a cv2.add () image processing function so as to determine a sample;
inputting the fused image into a ConvNeXt V model for training;
And fifthly, predicting by using the trained model to obtain a pneumatic valve fault diagnosis result.
Further, the first step uses a microprocessor to collect data of 5 variables, wherein the 5 variables are respectively valve position feedback data, air chamber air pressure detection sensor data, air source air pressure detection sensor data, valve positioner control current data and valve differential signal reference voltage data.
Further, in the step one, the microprocessor is an MSP430 single-chip microcomputer.
Furthermore, the first acquisition frequency is 50Hz, each sample comprises 1000 sample points, 2000 samples are acquired in total, wherein the samples comprise healthy samples, air source leakage, air chamber leakage and 500 samples with loose valve positions, and 2000 groups of multi-element time sequence data are converted into two-dimensional images and input into a model for training.
Further, the second step is to scale the original signals x=x1, X2, …, xn to [ -1,1] by the normalization process of the formula (1) as follows:
converting the scaled data into polar coordinates through a formula (2), converting the data into an included angle cosine value, and taking time as a radius, wherein the converted angle range is [0, pi ] and the formula thereof is as follows:
In the special inner product equation (3) defined by GAF: < x1, x2> = cos (Φ1+Φ2)
And converting the acquired time sequence data with the length of 1000 of the 5 signal data into two-dimensional images with the size of 224 x 224 through GAFs.
Further, in the third step, the generated 5 GAFs pictures are respectively fused into a two-dimensional image through a cv2.add () image fusion function.
Further, the classification model adopted in the fourth step is ConvNeXt V < 2 > model, which is a full convolution mask self-composer framework and a global response normalization layer added on the basis of ConvNeXt, and the model framework:
a) Firstly, performing two-dimensional convolution output dimension of which the parameter is kernel_size=4 and stride=4 on 224×224×3 images is 128;
b) Then the normalized layers are overlapped and then respectively passed through 3 layers of parameters to be
Kernel_size=2, stride=2, and the two-dimensional convolution output dimension is 128;
c) Secondly, the parameters of the 3 layers are kernel_size=2, stride=2, and the two-dimensional convolution output dimension is 256;
d) Then, the two-dimensional convolution output dimension is 512 through 27 layers of parameters of kernel_size=2 and stride=2;
e) Then, the two-dimensional convolution output dimension is 1024 through the 3-layer parameter of kernel_size=2 and stride=2;
f) Finally, through global averaging pooling (Global Average Pooling), the normalization process in the feature dimension outputs faults for 5 different labels through the fully connected layer.
The pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV2 has the following advantages:
the invention directly converts the time sequence data into a two-dimensional image through GAFs algorithm, the middle part does not carry out dimension reduction treatment on the data, the conversion process is essentially a dimension-increasing process, the relation characterization among data values at different moments is correspondingly increased, and the main characteristics of the data are highlighted. And finally, the multi-element data are fused into a two-dimensional image in an image fusion mode, and training is carried out through a ConvNeXt V network structure, so that a good diagnosis effect is achieved, and a feature extraction mechanism in the field of computer vision is better utilized, so that the pneumatic valve fault features are extracted and classified more efficiently.
Drawings
FIG. 1 is a flow chart of the overall design of the present invention;
FIG. 2 is a graph of the effect of the inventive Glam angle field implementation;
FIG. 3 is a diagram showing the effect of the third implementation of the present invention;
FIG. 4 is a diagram of a model network structure of step four of the present invention;
Fig. 5 is a graph comparing the fault diagnosis effects of the conventional machine learning and the method of the present invention.
Detailed Description
For a better understanding of the objects, structures and functions of the present invention, a more detailed description of a pneumatic valve fault diagnosis method based on GAFs multiple sequential two-dimensional images and ConvNeXtV is provided below, with reference to the accompanying drawings.
As shown in fig. 1, the pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV in this aspect includes the following steps:
step one, sampling a plurality of signals in a data time domain of a pneumatic valve fault;
The MSP430 singlechip is used for collecting data of 5 variables, wherein the 5 variables are valve position feedback data, air chamber air pressure detection sensor data, air source air pressure detection sensor data, valve positioner control current data and valve differential signal reference voltage data respectively. The acquisition frequency was 50Hz, and each sample contained 1000 sample points. A total of 2000 samples were collected, including 500 samples of healthy samples, air-source leaks, air-chamber leaks, and valve looseness. Converting the 2000 groups of multi-element time sequence data into two-dimensional images, and inputting the two-dimensional images into a model for training.
Step two, converting the acquired data into a two-dimensional image in a manner of a Gram Angle Field (GAFs) according to the time sequence data of each single variable;
the original signal x=x1, X2, …, xn is first scaled to [ -1,1] by the normalization process of equation (1) as follows:
converting the scaled data into polar coordinates through a formula (2), converting the data into an included angle cosine value, and taking time as a radius, wherein the converted angle range is [0, pi ] and the formula thereof is as follows:
In the special inner product equation (3) defined by GAF: < x1, x2> = cos (Φ1+Φ2)
As shown in fig. 2, the implementation effect is that the acquired time series data with the length of 1000 of 5 signal data are respectively converted into two-dimensional images with the size of 224×224 through GAFs.
Step three, fusing each image into one image through a cv2.add () image processing function so as to determine a sample;
as shown in fig. 3, the generated 5 GAFs pictures are respectively fused into one two-dimensional image by a cv2.add () image fusion function.
Inputting the fused image into a ConvNeXt V model for training;
as shown in FIG. 4, the classification model adopted by the invention is ConvNeXt V < 2 > model, and the model is characterized in that a full convolution mask self-encoder frame and a global response normalization layer are added on the basis of ConvNeXt, so that feature competition among channels is increased, and identification in various fault diagnosis can be remarkably improved. The model frame:
g) Firstly, performing two-dimensional convolution output dimension of which the parameter is kernel_size=4 and stride=4 on 224×224×3 images is 128;
h) Then the normalized layers are overlapped and then respectively passed through 3 layers of parameters to be
Kernel_size=2, stride=2, and the two-dimensional convolution output dimension is 128;
i) Secondly, the parameters of the 3 layers are kernel_size=2, stride=2, and the two-dimensional convolution output dimension is 256;
j) Then, the two-dimensional convolution output dimension is 512 through 27 layers of parameters of kernel_size=2 and stride=2;
k) Then, the two-dimensional convolution output dimension is 1024 through the 3-layer parameter of kernel_size=2 and stride=2;
l) finally, through global averaging pooling (Global Average Pooling), a normalization process is performed in the feature dimension to output failures for 5 different labels through the fully connected layer.
And fifthly, predicting by using the trained model to obtain a pneumatic valve fault diagnosis result.
In order to verify the effectiveness of the invention, the original data is subjected to fault diagnosis in a traditional machine learning mode, and specific diagnosis effects are shown in fig. 5, so that compared with classification modes such as 1DCNN, KNN, SVM and DT, the accuracy of the invention is obviously improved.
It will be understood that the application has been described in terms of several embodiments, and that various changes and equivalents may be made to these features and embodiments by those skilled in the art without departing from the spirit and scope of the application. In addition, many modifications may be made to adapt a particular situation or material to the teachings of the application without departing from the essential scope thereof. Therefore, it is intended that the application not be limited to the particular embodiment disclosed, but that the application will include all embodiments falling within the scope of the appended claims.
Claims (7)
1. A pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV2 is characterized by comprising the following steps:
step one, sampling a plurality of signals in a data time domain of a pneumatic valve fault;
Step two, converting the acquired data into a two-dimensional image in a manner of a gram angle field GAFs according to the time sequence data of each single variable;
Step three, fusing each image into one image through a cv2.add () image processing function so as to determine a sample;
inputting the fused image into a ConvNeXt V model for training;
And fifthly, predicting by using the trained model to obtain a pneumatic valve fault diagnosis result.
2. The method for diagnosing a pneumatic valve failure based on GAFs multiple time sequence two-dimensional image and ConvNeXtV as set forth in claim 1, wherein the first step uses a microprocessor to collect data of 5 variables, wherein the 5 variables are valve position feedback data, air chamber air pressure detection sensor data, air source air pressure detection sensor data, valve positioner control current data and valve differential signal reference voltage data, respectively.
3. The method for diagnosing a pneumatic valve fault based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV according to claim 2, wherein the microprocessor in the first step is an MSP430 single-chip microcomputer.
4. The method for diagnosing a pneumatic valve failure based on GAFs multiple time sequence two-dimensional image and ConvNeXtV2 as claimed in claim 2, wherein the first acquisition frequency is 50Hz, each sample comprises 1000 sample points, and a total of 2000 samples are acquired, wherein the samples comprise healthy samples, air source leakage, air chamber leakage and valve loosening 500 samples, and 2000 groups of multiple time sequence data are converted into two-dimensional images and are input into a model for training.
5. The method for diagnosing a pneumatic valve fault based on GAFs multiple time sequential two-dimensional image and ConvNeXtV2 as claimed in claim 1, wherein said step two first scales the original signal x=x1, X2, …, xn to [ -1,1] by the normalization process of equation (1) as follows:
converting the scaled data into polar coordinates through a formula (2), converting the data into an included angle cosine value, and taking time as a radius, wherein the converted angle range is [0, pi ] and the formula thereof is as follows:
In the special inner product equation (3) defined by GAF: < x1, x2> = cos (Φ1+Φ2)
And converting the acquired time sequence data with the length of 1000 of the 5 signal data into two-dimensional images with the size of 224 x 224 through GAFs.
6. The method for diagnosing a pneumatic valve fault based on GAFs multiple time sequence two-dimensional image and ConvNeXtV2 according to claim 1, wherein the step three fuses the generated 5 GAFs images into one two-dimensional image through a cv2.add () image fusion function.
7. The pneumatic valve fault diagnosis method based on GAFs multiple time sequence two-dimensional image and ConvNeXtV2 according to claim 1, wherein the classification model adopted in the fourth step is a ConvNeXt V model, the model is formed by adding a full convolution mask self-encoder framework and a global response normalization layer on the basis of ConvNeXt, and the model framework is as follows:
a) Firstly, performing two-dimensional convolution output dimension of which the parameter is kernel_size=4 and stride=4 on 224×224×3 images is 128;
b) Then, after the normalization layers are overlapped, the two-dimensional convolution output dimension is 128 through a 3-layer parameter of kernel_size=2 and stride=2;
c) Secondly, the parameters of the 3 layers are kernel_size=2, stride=2, and the two-dimensional convolution output dimension is 256;
d) Then, the two-dimensional convolution output dimension is 512 through 27 layers of parameters of kernel_size=2 and stride=2;
e) Then, the two-dimensional convolution output dimension is 1024 through the 3-layer parameter of kernel_size=2 and stride=2;
f) Finally, through global averaging pooling (Global Average Pooling), the normalization process in the feature dimension outputs faults for 5 different labels through the fully connected layer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410339202.2A CN118228174A (en) | 2024-03-25 | 2024-03-25 | Pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV2 |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410339202.2A CN118228174A (en) | 2024-03-25 | 2024-03-25 | Pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV2 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118228174A true CN118228174A (en) | 2024-06-21 |
Family
ID=91497494
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410339202.2A Pending CN118228174A (en) | 2024-03-25 | 2024-03-25 | Pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV2 |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118228174A (en) |
-
2024
- 2024-03-25 CN CN202410339202.2A patent/CN118228174A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113128561A (en) | Machine tool bearing fault diagnosis method | |
| CN114818774A (en) | Intelligent gearbox fault diagnosis method based on multi-channel self-calibration convolutional neural network | |
| CN113128338A (en) | Intelligent diagnosis method for printing machine roller fault under small sample | |
| CN113807444A (en) | Fault diagnosis method based on constraint-confrontation convolutional self-coding memory fusion network | |
| CN112305388B (en) | On-line monitoring and diagnosing method for insulation partial discharge faults of generator stator winding | |
| CN116593157A (en) | Complex working condition gear fault diagnosis method based on matching element learning under small sample | |
| CN116842379A (en) | Mechanical bearing residual service life prediction method based on DRSN-CS and BiGRU+MLP models | |
| CN114062812B (en) | Metering cabinet fault diagnosis method and system | |
| CN115452376A (en) | Bearing fault diagnosis method based on improved lightweight deep convolution neural network | |
| CN116778272A (en) | Hydropower unit small sample working condition identification method based on multichannel information fusion | |
| Xue et al. | A novel framework for motor bearing fault diagnosis based on multi-transformation domain and multi-source data | |
| CN111783941B (en) | Mechanical equipment diagnosis classification method based on probability confidence convolutional neural network | |
| CN113703422A (en) | Gas turbine pneumatic actuator fault diagnosis method based on characteristic analysis processing | |
| CN118228174A (en) | Pneumatic valve fault diagnosis method based on GAFs multiple time sequence to two-dimensional image and ConvNeXtV2 | |
| Chen et al. | Integrated learning method of gramian angular field and optimal feature channel adaptive selection for bearing fault diagnosis | |
| CN112257773B (en) | Mechanical equipment fault diagnosis method based on multiple measurement vectors of wireless sensor network | |
| CN115144211A (en) | Fault detection method and device based on discrete wavelet transform and gated cyclic unit | |
| CN115186584A (en) | Width learning semi-supervised soft measurement modeling method integrating attention mechanism and adaptive composition | |
| CN115217534A (en) | Method and system for monitoring service quality state of steam turbine | |
| CN113723592A (en) | Fault diagnosis method based on wind power gear box monitoring system | |
| Chen et al. | Bearing fault diagnosis based on improved denoising auto-encoders | |
| CN117708574B (en) | CNN variable-speed rolling bearing fault diagnosis method embedded with physical information | |
| CN117274903B (en) | Intelligent early warning device and method for electric power inspection based on intelligent AI chip | |
| CN114782441B (en) | Electronic component production line fault information identification method | |
| CN114004360B (en) | Intelligent equipment diagnosis equipment and method based on fuzzy expert model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination |