CN113792770A - Zero-sample rolling bearing fault diagnosis method and system based on attribute description - Google Patents
Zero-sample rolling bearing fault diagnosis method and system based on attribute description Download PDFInfo
- Publication number
- CN113792770A CN113792770A CN202111010900.0A CN202111010900A CN113792770A CN 113792770 A CN113792770 A CN 113792770A CN 202111010900 A CN202111010900 A CN 202111010900A CN 113792770 A CN113792770 A CN 113792770A
- Authority
- CN
- China
- Prior art keywords
- fault
- attribute
- preset
- rolling bearing
- time
- 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
- 238000005096 rolling process Methods 0.000 title claims abstract description 63
- 238000003745 diagnosis Methods 0.000 title claims abstract description 48
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000012549 training Methods 0.000 claims abstract description 37
- 238000012360 testing method Methods 0.000 claims abstract description 29
- 238000000605 extraction Methods 0.000 claims description 40
- 238000013527 convolutional neural network Methods 0.000 claims description 20
- 238000010586 diagram Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 230000003862 health status Effects 0.000 claims 1
- 238000012545 processing Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 11
- 230000000694 effects Effects 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 238000007781 pre-processing Methods 0.000 description 4
- 238000012800 visualization Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000009776 industrial production Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000000547 structure data Methods 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a zero-sample rolling bearing fault diagnosis method and system based on attribute description, and relates to the field of equipment fault diagnosis and the field of signal processing. According to the technical scheme, the fault description consisting of attributes is provided for each fault to serve as auxiliary information, meanwhile, the attribute prediction of the unseen class is completed through the attribute learning of the seen class, and the accurate classification of the test set is completed under the condition that no test set sample can be used for training.
Description
Technical Field
The invention relates to the field of equipment fault diagnosis and the field of signal processing, in particular to a zero-sample rolling bearing fault diagnosis method and system based on attribute description.
Background
The fault diagnosis of the rolling bearing is an indispensable task in the current industrial production process, and the mechanical fault diagnosis is the best way for maintaining industrial production equipment. The rolling bearing is in a continuous working state for a long time and the working environment is extremely severe, so the rolling bearing is easy to damage. Once the rolling bearing is damaged, the machine is slightly broken down to delay production, and serious accidents and even casualties are caused. Therefore, the rolling bearing fault detection method has important significance in monitoring the working condition of the rolling bearing, confirming the fault type and timely replacing corresponding parts.
In recent years, deep learning techniques have been rapidly developed and widely used in image recognition and segmentation, automatic driving, target style conversion, and the like. As a class of pattern analysis method in deep learning, the convolutional neural network is applied to the field of fault diagnosis by more and more scholars and experts due to strong adaptive feature extraction capability and learning capability. Zhao Xiaoping and the like provide a diagnosis method based on a multi-task deep learning model, and a multi-label system is introduced aiming at the condition that a single label system ignores the relation between compound faults, so that the multi-fault condition is accurately classified. The method has the advantages that the problem that vibration signals of all parts in the gear box are overlapped seriously is considered, a multi-resonance-component fusion convolution neural network is provided, and a good classification effect is obtained. Aiming at the problems that the traditional rotating machinery fault algorithm is poor in anti-interference performance and cannot accurately extract fault characteristics, Shao provides an anti-domain self-adaption method based on transfer learning, and inter-domain self-adaption is carried out on the fault characteristics by using a maximum mean square error and a domain confusion function, so that cross-domain fault diagnosis is realized. However, most of deep learning fault diagnosis models are trained in a data-driven manner, and model training is performed on data acquired through a laboratory bench, which often results in poor mobility of the trained fault diagnosis models. In the actual production process, due to different working conditions, complex production environment and the like, the types of the faults which occur are unpredictable, and therefore, no available test sample can be used for training the fault diagnosis model.
Disclosure of Invention
The invention aims to provide a zero-sample rolling bearing fault diagnosis method combining an Xception network and a convolutional neural network, namely an X-CNN fault diagnosis model. The model adopts an Xconvergence network to extract the characteristics of a fault signal time-frequency diagram; constructing an attribute matrix according to the attribute description of the fault category, and performing attribute learning on the extracted features by using a convolutional neural network; and finally, completing fault diagnosis work through similarity comparison of the attribute matrix so as to solve the problems in the prior art.
In order to achieve the purpose, the invention provides the following technical scheme:
a zero-sample rolling bearing fault diagnosis method based on attribute description is characterized by collecting fault signal data of a rolling bearing corresponding to each fault type, converting the fault signal data into corresponding time-frequency images through data conversion to obtain a rolling bearing time-frequency image set, applying the following steps A to E based on the rolling bearing time-frequency image set, training and obtaining a feature extraction network and an attribute learning device, and applying the feature extraction network and the attribute learning device to carry out fault prediction on the rolling bearing to be diagnosed to obtain a fault prediction result:
step A, dividing each time-frequency image in a time-frequency image set of the rolling bearing into a training set for obtaining a feature extraction network and an attribute learning device and a test set for testing the feature extraction network and the attribute learning device according to a preset proportion, and then entering step B;
b, training a first network to be trained according to preset iteration times by taking each fault type in the time-frequency image as input and each preset characteristic value corresponding to each fault type as output based on the time-frequency image in the training set to obtain a characteristic extraction network corresponding to each fault type in the time-frequency image in the training set, and then entering the step C;
step C, aiming at each fault type in the training set, obtaining each preset characteristic value corresponding to the fault type by using a characteristic extraction network, marking each preset characteristic value by using a preset attribute label, obtaining each fault attribute characteristic value corresponding to each preset characteristic value, namely obtaining each fault attribute characteristic value corresponding to each fault type in the training set, taking each preset characteristic value corresponding to the fault type as input, taking each fault attribute characteristic value corresponding to the fault type as output, training a second network to be trained according to preset iteration times, obtaining an attribute learner, and then entering the step D;
d, respectively aiming at each time-frequency image in the test set, sequentially utilizing a feature extraction network and an attribute learning device to obtain each fault attribute characteristic value corresponding to the time-frequency image, utilizing the fault attribute characteristic values to predict the fault type corresponding to each time-frequency image in the test set to obtain a fault prediction result, and then entering the step E;
and E, verifying the accuracy of the fault prediction result by using a preset attribute table containing preset attribute labels respectively aiming at each time-frequency image in the test set, if the preset accuracy cannot be met, returning to the step B and updating the preset iteration times of the step B, and if the preset accuracy is met, outputting the fault prediction result obtained in the step D.
Further, the preset fault types included in the time-frequency diagram set of the rolling bearing include an inner ring fault, an outer ring fault, a rolling element fault and a normal health state.
Further, the preset proportion in the step a is that the time-frequency images in the time-frequency diagram set of the rolling bearing are arranged according to a ratio of 22: the proportion of 5 is divided into a training set and a testing set.
Furthermore, the rolling bearing time-frequency graph set comprises preset fault types, and fault loads, fault positions and fault sizes are respectively marked through each preset attribute label in the preset attribute table.
Further, before the step B is executed, each time-frequency image in the time-frequency image set of the rolling bearing is processed into time-frequency images with preset sizes according to a preset proportion.
Further, the structure of the first network to be trained is a network structure in which an Xception network structure and a convolutional neural network are parallel, the Xception network structure is used for extracting preset feature values of each fault type in the time-frequency image, and the convolutional neural network structure is used for predicting the fault type corresponding to the extracted preset feature values.
Further, the second network to be trained is a convolutional neural network.
The second aspect of the present invention provides a zero-sample rolling bearing fault diagnosis system based on attribute description, including:
one or more processors;
a memory storing instructions operable, when executed by the one or more processors, to cause a process of a zero sample rolling bearing fault diagnosis method based on attribute descriptions.
A third aspect of the present invention proposes a computer-readable medium storing software comprising instructions executable by one or more computers, which when executed by the one or more computers perform the operations of any of the attribute-based description zero-sample rolling bearing fault diagnosis methods as described above.
Compared with the prior art, the zero-sample rolling bearing fault diagnosis method and system based on attribute description have the following technical effects by adopting the technical scheme:
the invention provides a zero-sample rolling bearing fault diagnosis method based on attribute description, and researches a fault diagnosis system for model training of samples without target faults. The method does not adopt the learning mode of the traditional fault diagnosis system, but provides fault description composed of attributes as auxiliary information for each fault. The method comprises the steps that attribute description is embedded between a fault sample layer and a fault category layer, fine-grained shared attributes of the attribute description layer are used for constructing a cascade diagnosis system for migrating attribute knowledge of training faults to target faults to carry out zero-sample fault diagnosis, meanwhile, attribute prediction of unseen classes can be finished through learning of attributes of the seen classes, and accurate classification of a test set can be finished under the condition that no test set sample can be used for training.
Drawings
FIG. 1 is a flow chart of a zero sample rolling bearing fault diagnostic method of an exemplary embodiment of the present invention;
fig. 2(a) is a schematic diagram of a network structure data preprocessing stage of an X-CNN fault diagnosis model according to an exemplary embodiment of the present invention, fig. 2(b) is a schematic diagram of a network structure feature extraction stage of an X-CNN fault diagnosis model according to an exemplary embodiment of the present invention, and fig. 2(c) is a schematic diagram of a network structure attribute learning and classification stage of an X-CNN fault diagnosis model according to an exemplary embodiment of the present invention
Fig. 3 is a schematic structural diagram of a CNN-based attribute learning period according to an exemplary embodiment of the present invention;
FIG. 4 is a dimension reduction visualization scatter plot of Xconcept feature extraction network extracted features in an exemplary embodiment of the present invention;
FIG. 5 illustrates the attribute learning accuracy of an Xcaption feature-based extraction network in accordance with an exemplary embodiment of the present invention.
Detailed Description
In order to better understand the technical content of the present invention, specific embodiments are described below with reference to the accompanying drawings.
In this disclosure, aspects of the present invention are described with reference to the accompanying drawings, in which a number of illustrative embodiments are shown. Embodiments of the present disclosure are not necessarily defined to include all aspects of the invention. It should be appreciated that the various concepts and embodiments described above, as well as those described in greater detail below, may be implemented in any of numerous ways, as the disclosed concepts and embodiments are not limited to any one implementation. In addition, some aspects of the present disclosure may be used alone, or in any suitable combination with other aspects of the present disclosure.
With reference to the exemplary embodiments of the present invention shown in fig. 1 to 5, the present invention provides a zero-sample rolling bearing fault diagnosis method based on attribute description, which completes a fault diagnosis task by constructing an X-CNN model, where the X-CNN fault diagnosis model is composed of a data preprocessing stage, a feature extraction stage, and an attribute learning stage, and specifically, acquires each fault signal data corresponding to each fault type of a rolling bearing, converts each fault signal data into a corresponding time-frequency image through data conversion, obtains a rolling bearing time-frequency diagram set, and applies the following steps a to E based on the rolling bearing time-frequency diagram set, trains and obtains a feature extraction network and an attribute learner, and applies the feature extraction network and the attribute learner to perform fault prediction on the rolling bearing to be diagnosed, so as to obtain a fault prediction result:
step A, dividing each time-frequency image in a time-frequency image set of the rolling bearing into a training set for obtaining a feature extraction network and an attribute learning device and a test set for testing the feature extraction network and the attribute learning device according to a preset proportion, and then entering step B;
b, training a first network to be trained according to preset iteration times by taking each fault type in the time-frequency image as input and each preset characteristic value corresponding to each fault type as output based on the time-frequency image in the training set to obtain a characteristic extraction network corresponding to each fault type in the time-frequency image in the training set, and then entering the step C;
step C, aiming at each fault type in the training set, obtaining each preset characteristic value corresponding to the fault type by using a characteristic extraction network, marking each preset characteristic value by using a preset attribute label, obtaining each fault attribute characteristic value corresponding to each preset characteristic value, namely obtaining each fault attribute characteristic value corresponding to each fault type in the training set, taking each preset characteristic value corresponding to the fault type as input, taking each fault attribute characteristic value corresponding to the fault type as output, training a second network to be trained according to preset iteration times, obtaining an attribute learner, and then entering the step D;
d, respectively aiming at each time-frequency image in the test set, sequentially utilizing a feature extraction network and an attribute learning device to obtain each fault attribute characteristic value corresponding to the time-frequency image, utilizing the fault attribute characteristic values to predict the fault type corresponding to each time-frequency image in the test set to obtain a fault prediction result, and then entering the step E;
and E, verifying the accuracy of the fault prediction result by using a preset attribute table containing preset attribute labels respectively aiming at each time-frequency image in the test set, if the preset accuracy cannot be met, returning to the step B and updating the preset iteration times of the step B, and if the preset accuracy is met, outputting the fault prediction result obtained in the step D.
Detailed description of the preferred embodiment
With reference to fig. 1, a flow of a zero-sample rolling bearing fault diagnosis method according to an exemplary embodiment of the present invention is shown, the present invention mainly includes a data preprocessing stage, a feature extraction stage, and an attribute learning stage, and a specific implementation process is described with reference to the processes shown in fig. 2(a) to 2 (c);
(1) data preprocessing stage
The CWRU data set is widely applied to a rolling bearing fault signal data set of the university of Kersi reservoir, and by utilizing the rolling bearing data set provided by the CWRU, the CWRU bearing data set provides three bearing faults, namely an inner ring fault (InnerRace), an outer ring fault (OuterRace) and a rolling element fault (Ball), and selects different fault volume levels of 7 mils, 14 mils, 21 mils and 28 mils, besides, the outer ring fault is divided into three damage points of 3 o ' clock of an outer ring, 6 o ' clock of the outer ring and 12 o ' clock of the outer ring.
Because the zero sample rolling bearing fault diagnosis is realized by an attribute description method, and attribute description is to judge whether a certain fault attribute appears in an experiment, the attribute is defined according to information such as fault positions of different fault types, motor load, fault magnitude and the like, wherein the fault positions comprise an inner ring fault, an outer ring fault in a 3 o 'clock direction, an outer ring fault in a 6 o' clock direction and a rolling body fault; the motor load comprises a load 0, a load 2 and a load 3; the fault sizes include 7 mils, 14 mils, 21 mils and 28 mils, 11 different attributes are defined according to the selected data set, and an attribute description correspondence table is shown in table 1:
table 1 attribute description correspondence table
After the fault type is selected, defining the data code of the fault type, wherein the data code represents the occurrence position, the fault size and the operation load of the machine: for the fault position, the normal type is N, the fault of the rolling body is B, the fault of the inner ring is IR, and the fault of the outer ring is OR; the three digit number after the fault type represents the fault size. If "OR 007_0@ 3" represents the outer ring fault with a load of 0, the fault size is 7 mils, and the fault location is in the outer ring 3 o' clock direction.
When the vibration signal provided by the CWRU is preprocessed, the selected feature extraction network has excellent performance on image processing, so that the fault signal data is converted into a time-frequency graph by short-time Fourier transform, and the frequency spectrum information of the fault signal data changing along with time is obtained. The short-time fourier transform multiplies a function by a window function and then performs a one-dimensional fourier transform. And a series of frequency spectrum functions are obtained through the sliding of the window function, and a two-dimensional time-frequency graph is obtained by sequentially opening the results. A hamming window is used as a window function in the short-time fourier transform and a preset window function length of 256, a window overlap of 50% and a sampling frequency of 120khz are set.
(2) Feature extraction stage
And inputting the time-frequency diagram of the vibration signal of the training set as a sample into an Xception feature extraction network for training and carrying out feature extraction, and training until the network is converged, thereby obtaining the features of the time-frequency diagram and storing the trained Xception feature extraction network. And transmitting the time-frequency graph characteristics to an attribute learning device, training the attribute learning device by using the time-frequency graph characteristics, and finally storing the trained attribute learning device for use in a test stage. Referring to fig. 3, the X-CNN fault diagnosis network finishes gradual extraction of features of a time-frequency graph through Entry flow and middlleflow, a Softmax layer is added at the end of the feature extraction network, and a loss function of the network is set as follows:
where y is the true sample label,
for the predicted value of the label, minimizing by the loss function
The difference from y to complete the classification test.
In order to verify whether the Xception network has excellent feature extraction capability, the method of the embodiment of the application uses a t-SNE method to perform dimensionality reduction on the extracted multidimensional features and then projects the multidimensional features into a two-dimensional space for visualization analysis, and if the projection result has good separability, the method can be described to have strong feature extraction capability. The method comprises the steps that 5 faults are randomly selected to carry out feature extraction and dimension reduction visualization, the visualization result is shown in fig. 4, through observation, the fact that similarity misjudgment occurs on the extracted features only under the conditions that the faults are the same in inner ring, the loads are the same, and the sizes of the faults are different can be found, confusion is generated on a part of the features which are yellow and green in fig. 4, and it can be seen that the Xception network has extremely excellent feature extraction capability on a time-frequency diagram of a vibration signal.
(3) Attribute learning phase
After the feature extraction of the vibration signal is completed, CNN is used as an attribute learning device to complete attribute prediction of the feature, Relu is used as an activation function in a convolution layer in the attribute learning device, Adam is used as an optimizer in a network, a cross entropy loss function is selected as a loss function, in the training process, the feature attribute of the fault of the rolling bearing is learned after conversion, namely, 11 different attributes corresponding to the original single fault are converted into 11 different attributes to test whether the single attribute exists in 20 different faults, so that 11 new arrays are obtained, after splicing, transposition operation is performed to restore the corresponding attribute matrix, and the final fault type is determined by adopting a nearest neighbor search method according to the finally predicted attribute matrix. The accuracy of attribute learning after feature extraction using Xception is shown in fig. 5.
Observing fig. 5, it can be found that the CNN-based attribute learner has a better learning effect on most attributes, and the learning accuracy of some attributes can reach 100%, in addition, the same as the machine learning method is used, the learning accuracy of the attributes 5 and 6 is still very low, and the attributes corresponding to the attributes 5 and 6 are different loads, it can be seen that the load recognition capability of the method is weak, in order to ensure the rationality and the authenticity of the experimental process, the selected data set is randomly selected from the test set and the training set, and five times of experiments are performed, and the final classification accuracy of the test set is shown in table 2:
TABLE 2 Classification accuracy
As can be seen from table 2, the attribute learning and the fault diagnosis are performed after the features are extracted by using Xception, and compared with the PCA method, the method can obtain a higher classification accuracy rate up to approximately 80%, which is approximately 14% higher than the probabilistic naive bayes method with the best effect after the features are extracted by using the PCA method, and the average accuracy of 5 experiments is higher than 77%, which are far higher than the experimental results obtained after the features are extracted by using the PCA. In addition, compared with the method of using a machine learning method as an attribute learning device, the method of using the CNN as the attribute learning device can obtain higher classification accuracy, the accuracy of single classification can reach 95.2% at most, and the average accuracy reaches 92.66%. In summary, the zero-sample rolling bearing fault diagnosis method based on attribute description can finish accurate classification of the test class under the condition that no test class sample is trained.
Although the present invention has been described with reference to the preferred embodiments, it is not intended to be limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, the protection scope of the present invention should be determined by the appended claims.
Claims (9)
1. A zero-sample rolling bearing fault diagnosis method based on attribute description is characterized by comprising the steps of collecting fault signal data of a rolling bearing corresponding to each fault type, converting the fault signal data into corresponding time-frequency images through data conversion to obtain a rolling bearing time-frequency image set, applying the following steps A to E based on the rolling bearing time-frequency image set, training and obtaining a feature extraction network and an attribute learning device, and applying the feature extraction network and the attribute learning device to carry out fault prediction on the rolling bearing to be diagnosed to obtain a fault prediction result:
step A, dividing each time-frequency image in a time-frequency image set of the rolling bearing into a training set for obtaining a feature extraction network and an attribute learning device and a test set for testing the feature extraction network and the attribute learning device according to a preset proportion, and then entering step B;
b, training a first network to be trained according to preset iteration times by taking each fault type in the time-frequency image as input and each preset characteristic value corresponding to each fault type as output based on the time-frequency image in the training set to obtain a characteristic extraction network corresponding to each fault type in the time-frequency image in the training set, and then entering the step C;
step C, aiming at each fault type in the training set, obtaining each preset characteristic value corresponding to the fault type by using a characteristic extraction network, marking each preset characteristic value by using a preset attribute label, obtaining each fault attribute characteristic value corresponding to each preset characteristic value, namely obtaining each fault attribute characteristic value corresponding to each fault type in the training set, taking each preset characteristic value corresponding to the fault type as input, taking each fault attribute characteristic value corresponding to the fault type as output, training a second network to be trained according to preset iteration times, obtaining an attribute learner, and then entering the step D;
d, respectively aiming at each time-frequency image in the test set, sequentially utilizing a feature extraction network and an attribute learning device to obtain each fault attribute characteristic value corresponding to the time-frequency image, utilizing the fault attribute characteristic values to predict the fault type corresponding to each time-frequency image in the test set to obtain a fault prediction result, and then entering the step E;
and E, verifying the accuracy of the fault prediction result by using a preset attribute table containing preset attribute labels respectively aiming at each time-frequency image in the test set, if the preset accuracy cannot be met, returning to the step B and updating the preset iteration times of the step B, and if the preset accuracy is met, outputting the fault prediction result obtained in the step D.
2. The zero-sample rolling bearing fault diagnosis method based on attribute description of claim 1, wherein the preset fault types included in the rolling bearing time-frequency diagram set include inner ring fault, outer ring fault, rolling element fault, and normal health status.
3. The zero-sample rolling bearing fault diagnosis method based on attribute description as claimed in claim 2, wherein the preset proportion in step a is that the time-frequency images in the time-frequency image set of the rolling bearing are processed according to a ratio of 22: the proportion of 5 is divided into a training set and a testing set.
4. The zero-sample rolling bearing fault diagnosis method based on attribute description according to any one of claims 1 or 2, characterized in that the rolling bearing time-frequency diagram set comprises preset fault types, and fault loads, fault positions and fault sizes are respectively marked through preset attribute labels in a preset attribute table.
5. The zero-sample rolling bearing fault diagnosis method based on attribute description as claimed in claim 1, wherein before step B is executed, each time-frequency image in the time-frequency image set of the rolling bearing is processed into time-frequency images with preset sizes according to a preset proportion.
6. The zero-sample rolling bearing fault diagnosis method based on attribute description as claimed in claim 5, wherein the structure of the first network to be trained is a network structure in which an Xception network structure and a convolutional neural network are parallel, the Xception network structure is used for extracting preset feature values of fault types in a time-frequency image, and the convolutional neural network structure is used for predicting fault types corresponding to the extracted preset feature values.
7. The zero-sample rolling bearing fault diagnosis method based on attribute description of claim 1, wherein the second network to be trained is a convolutional neural network.
8. A zero-sample rolling bearing fault diagnosis system based on attribute description is characterized by comprising:
one or more processors;
a memory storing instructions operable, when executed by the one or more processors, to cause a process of a zero sample rolling bearing fault diagnosis method based on attribute descriptions.
9. A computer-readable medium storing software, the software comprising instructions executable by one or more computers which, when executed by the one or more computers, perform the operations of the attribute description based zero-sample rolling bearing fault diagnosis method of any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111010900.0A CN113792770A (en) | 2021-08-31 | 2021-08-31 | Zero-sample rolling bearing fault diagnosis method and system based on attribute description |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111010900.0A CN113792770A (en) | 2021-08-31 | 2021-08-31 | Zero-sample rolling bearing fault diagnosis method and system based on attribute description |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113792770A true CN113792770A (en) | 2021-12-14 |
Family
ID=78876718
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111010900.0A Pending CN113792770A (en) | 2021-08-31 | 2021-08-31 | Zero-sample rolling bearing fault diagnosis method and system based on attribute description |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113792770A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114626420A (en) * | 2022-03-25 | 2022-06-14 | 合肥工业大学 | phi-OTDR vibration event classification method based on CWT-Xceptance-RF |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110889489A (en) * | 2019-11-08 | 2020-03-17 | 北京小米移动软件有限公司 | Neural network training method, image recognition method and device |
| US20200210843A1 (en) * | 2018-12-29 | 2020-07-02 | Canon Kabushiki Kaisha | Training and application method of a multi-layer neural network model, apparatus and storage medium |
| CN112084974A (en) * | 2020-09-14 | 2020-12-15 | 北京工商大学 | Multi-label rolling bearing fault diagnosis method based on meta-learning |
| CN112434732A (en) * | 2020-11-17 | 2021-03-02 | 西安交通大学 | Deep learning classification method based on feature screening |
| CN112529020A (en) * | 2020-12-24 | 2021-03-19 | 携程旅游信息技术（上海）有限公司 | Animal identification method, system, equipment and storage medium based on neural network |
| CN112763214A (en) * | 2020-12-31 | 2021-05-07 | 南京信息工程大学 | Rolling bearing fault diagnosis method based on multi-label zero-sample learning |
-
2021
- 2021-08-31 CN CN202111010900.0A patent/CN113792770A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200210843A1 (en) * | 2018-12-29 | 2020-07-02 | Canon Kabushiki Kaisha | Training and application method of a multi-layer neural network model, apparatus and storage medium |
| CN110889489A (en) * | 2019-11-08 | 2020-03-17 | 北京小米移动软件有限公司 | Neural network training method, image recognition method and device |
| CN112084974A (en) * | 2020-09-14 | 2020-12-15 | 北京工商大学 | Multi-label rolling bearing fault diagnosis method based on meta-learning |
| CN112434732A (en) * | 2020-11-17 | 2021-03-02 | 西安交通大学 | Deep learning classification method based on feature screening |
| CN112529020A (en) * | 2020-12-24 | 2021-03-19 | 携程旅游信息技术（上海）有限公司 | Animal identification method, system, equipment and storage medium based on neural network |
| CN112763214A (en) * | 2020-12-31 | 2021-05-07 | 南京信息工程大学 | Rolling bearing fault diagnosis method based on multi-label zero-sample learning |
Non-Patent Citations (4)
| Title |
|---|
| ZHAO X 第: "A multi-fault diagnosis method of gear-box running on edge equipment", 《JOURNAL OF CLOUD COMPUTING》, pages 1 - 13 * |
| 吕凯扬: "基于迁移学习与属性描述的齿轮箱故障诊断研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》, no. 1, pages 029 - 479 * |
| 谭雯雯: "基于神经网络的轴承故障诊断研究及其系统实现", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》, no. 2, pages 029 - 250 * |
| 赵晓平 等: "基于属性描述的零样本滚动轴承故障诊断", 《振动与冲击》, vol. 41, no. 15, pages 105 - 115 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114626420A (en) * | 2022-03-25 | 2022-06-14 | 合肥工业大学 | phi-OTDR vibration event classification method based on CWT-Xceptance-RF |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Convolutional neural network-based hidden Markov models for rolling element bearing fault identification | |
| CN113935406B (en) | Mechanical equipment unsupervised fault diagnosis method based on countermeasure flow model | |
| CN105466693B (en) | The pre- diagnostic method of Fault of Diesel Fuel System based on gray model | |
| CN112964469B (en) | Online fault diagnosis method for rolling bearing under variable load of transfer learning | |
| CN111562108A (en) | Rolling bearing intelligent fault diagnosis method based on CNN and FCMC | |
| Jiang et al. | Fault diagnosis for rolling bearing using a hybrid hierarchical method based on scale-variable dispersion entropy and parametric t-SNE algorithm | |
| CN115600150A (en) | Multi-mode gearbox fault diagnosis method based on deep migration learning | |
| CN111966076A (en) | Fault positioning method based on finite-state machine and graph neural network | |
| Wang et al. | Construction of the efficient attention prototypical net based on the time–frequency characterization of vibration signals under noisy small sample | |
| CN114170184A (en) | Product image anomaly detection method and device based on embedded feature vector | |
| CN116702076A (en) | Small sample migration learning fault diagnosis method, system, computer and storage medium based on CNN feature fusion | |
| CN116610993A (en) | Bearing fault diagnosis method, system, equipment and storage medium | |
| CN114462446A (en) | Rolling bearing fault diagnosis method based on vibration signal and electronic equipment | |
| Han et al. | A novel intelligent fault diagnosis method based on dual convolutional neural network with multi-level information fusion | |
| CN113792770A (en) | Zero-sample rolling bearing fault diagnosis method and system based on attribute description | |
| CN116595313A (en) | Fault diagnosis method and system based on multi-scale residual error attention network | |
| CN114740726B (en) | Industrial process small sample fault diagnosis method based on deep optimal feature transmission | |
| CN117056865B (en) | Method and device for diagnosing operation faults of machine pump equipment based on feature fusion | |
| CN117849193A (en) | Online crack damage monitoring method for neodymium iron boron sintering | |
| CN117949209A (en) | Zero sample rolling bearing composite fault diagnosis system and method based on envelope spectrum | |
| CN116721097A (en) | Bearing fault diagnosis method and device and electronic equipment | |
| Hao et al. | New fusion features convolutional neural network with high generalization ability on rolling bearing fault diagnosis | |
| Zhang et al. | Multi‐attribute quantitative bearing fault diagnosis based on convolutional neural network | |
| Zhu et al. | Bidirectional Current WP and CBAR Neural Network Model Based Bearing Fault Diagnosis | |
| Wang et al. | Adversarial based unsupervised domain adaptation for bearing fault diagnosis |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211214 |
|
| RJ01 | Rejection of invention patent application after publication |