CN115165363A - CNN-based light bearing fault diagnosis method and system - Google Patents
CNN-based light bearing fault diagnosis method and system Download PDFInfo
- Publication number
- CN115165363A CN115165363A CN202210737500.8A CN202210737500A CN115165363A CN 115165363 A CN115165363 A CN 115165363A CN 202210737500 A CN202210737500 A CN 202210737500A CN 115165363 A CN115165363 A CN 115165363A
- Authority
- CN
- China
- Prior art keywords
- data
- layer
- light
- cnn
- bearing
- 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.)
- Granted
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 59
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000010606 normalization Methods 0.000 claims abstract description 58
- 238000013501 data transformation Methods 0.000 claims abstract description 10
- 238000010586 diagram Methods 0.000 claims description 20
- 238000011176 pooling Methods 0.000 claims description 18
- 238000005070 sampling Methods 0.000 claims description 9
- 238000012952 Resampling Methods 0.000 claims description 6
- 230000001133 acceleration Effects 0.000 claims description 6
- 238000011550 data transformation method Methods 0.000 claims description 6
- 238000011478 gradient descent method Methods 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 4
- 238000002360 preparation method Methods 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 description 53
- 230000006870 function Effects 0.000 description 9
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
Images
Classifications
-
- 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
- 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
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Acoustics & Sound (AREA)
- Complex Calculations (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention provides a CNN-based light bearing fault diagnosis method and system, belonging to the technical field of fault diagnosis, and the method comprises the following steps: acquiring data; data transformation; making a data set; constructing a light CNN network by using instance normalization and grouping convolution; network training; storing and deploying; and diagnosing the fault of the bearing and outputting the fault type of the bearing. Through the design, the problems that the existing bearing fault diagnosis model occupies too much memory and limits the deployment and application of the model in most embedded systems are solved, and meanwhile, convenience is provided for fault diagnosis of small equipment based on the improved bearing fault diagnosis model and the deployment range of the improved bearing fault diagnosis model.
Description
Technical Field
The invention belongs to the technical field of fault diagnosis, and particularly relates to a light bearing fault diagnosis method and system based on CNN.
Background
The existing bearing fault diagnosis network based on deep learning is mainly divided into a bearing fault diagnosis network based on a full-connection neural network and a convolutional neural network. Because all the neurons between the layers are fully connected, the bearing fault diagnosis model based on full connection enables the model parameters to be huge and the over-fitting problem to easily occur, and reduces the fault classification precision. Due to the unique design of the convolutional neural network, the network has the characteristics of sparse connection and weight sharing, so that the overfitting problem is greatly improved, the occupied size of a network memory can be reduced, and the attention of more researchers is paid. Although the bearing fault diagnosis model based on the convolutional neural network can realize high diagnosis precision, the memory occupation of the model is from several to hundreds of megabytes, so that the application of the model in most small embedded systems is greatly limited, and small equipment cannot be deployed with the model for fault diagnosis.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides the CNN-based light bearing fault diagnosis method and the CNN-based light bearing fault diagnosis system, which solve the problems that the existing bearing fault diagnosis model occupies too much memory and limits the deployment and application of the model in most embedded systems, and provide convenience for fault diagnosis based on the improved bearing fault diagnosis model and the deployment range thereof.
In order to achieve the above purpose, the invention adopts the technical scheme that:
the scheme provides a CNN-based light bearing fault diagnosis method, which comprises the following steps:
s1, data acquisition: resampling a vibration signal collected by an acceleration sensor arranged on a bearing to obtain a time domain signal with the continuous length of 800 data points, and performing frequency domain transformation on the time domain signal to obtain a frequency domain signal with the length of 400 data points;
s2, data transformation: changing the one-dimensional vector of the frequency domain signal into a two-dimensional vector, and taking the two-dimensional vector as a data sample;
s3, data set preparation: repeating the step S1 and the step S2 to obtain a plurality of data samples of various faults, and respectively marking fault labels to obtain a training data set;
s4, constructing a light network: constructing a light CNN network by using instance normalization and packet convolution;
s5, network training: training parameters of the light CNN network by using a random gradient descent method according to the training data set;
s6, saving and deploying: storing the trained light CNN network parameters, and deploying the light CNN network to application equipment;
s7, bearing fault diagnosis: and making a data sample according to the bearing fault data to be diagnosed by a data acquisition and data transformation method, inputting the made data sample into deployed application equipment, and outputting the fault type of the bearing.
Further, the step S4 includes the steps of:
s401, reducing the input size of the light CNN network;
s402, eliminating the dependence of the light CNN network on batches by using an example normalization method;
s403, sequentially connecting an input layer, a first convolution layer, a first normalization layer, a first pooling layer, a second convolution layer, a second normalization layer, a second pooling layer, a full-connection layer and an output layer;
s404, using packet convolution to replace ordinary convolution to complete the construction of the light CNN network.
Still further, the example normalized process expression is as follows:
where y represents an example normalization process for some data x in the feature map, μ IN Denotes the mean value, σ IN Denotes the standard deviation,. Epsilon.denotes the minimum number,. Gamma.and.beta.denote the parameters to be trained, u IN (x) Representing the mean, σ, of a certain data x in a feature map IN (x) Representing the standard deviation of certain data x in the characteristic diagram, W and H respectively representing the width and height of the characteristic diagram to be normalized, b representing certain data x in the characteristic diagramOne batch, c represents a certain channel, w represents a certain coordinate in the feature map width, and h represents a certain coordinate in the feature map height.
Still further, the input layer is used for receiving a vibration signal to be input;
the first convolution layer is used for extracting the characteristics of the vibration signal;
the first normalization layer is used for carrying out example normalization processing on the extracted features;
the first pooling layer is used for performing down-sampling processing on the features subjected to the example normalization processing;
the second convolution layer is used for carrying out grouping convolution processing on the features subjected to the downsampling processing;
the second normalization layer is used for carrying out example normalization processing on the features subjected to the grouping convolution processing of the second convolution layer;
the second pooling layer is used for performing down-sampling processing on the features processed by the second normalization layer;
the full connection layer is used for flattening all the characteristics processed by the second pooling layer;
and the output layer is used for outputting the fault type of the bearing based on all the characteristics output by the full connection layer.
Still further, the expression of the loss function of the lightweight CNN network is as follows:
wherein Cross Encopy represents the cross entropy loss function of the lightweight CNN network, p i Representing the true distribution of the ith data sample, q i Represents the prediction distribution of the ith data sample, and n represents the number of data samples.
The invention also provides a CNN-based light bearing fault diagnosis system, which comprises:
the data acquisition module is used for resampling vibration signals acquired by the acceleration sensor arranged on the bearing to obtain time domain signals with the continuous length of 800 data points, and performing frequency domain transformation on the time domain signals to obtain frequency domain signals with the length of 400 data points;
the data transformation module is used for transforming the one-dimensional vector of the frequency domain signal into a two-dimensional vector and taking the two-dimensional vector as a data sample;
the data set making module is used for obtaining a plurality of data samples of various faults based on the data obtaining module and the data transformation module, and respectively marking fault labels to obtain a training data set;
constructing a light network module for constructing a light CNN network by using example normalization and packet convolution;
the network training module is used for training the parameters of the light CNN network by using a random gradient descent method according to the training data set;
the saving and deploying module is used for saving the trained light CNN network parameters and deploying the light CNN network to the application equipment;
and the bearing fault diagnosis module is used for making a data sample from the bearing fault data to be diagnosed according to a data acquisition and data transformation method, inputting the made data sample into deployed application equipment and outputting the fault type of the bearing.
The invention has the beneficial effects that:
(1) The invention solves the problems that the existing bearing fault diagnosis model occupies large memory and limits the application of the model in most embedded systems by constructing the improved light CNN network, thereby maximally improving the deployment range of the model and providing convenience for fault diagnosis of small equipment based on the improved bearing fault diagnosis model and the deployment thereof.
(2) The memory occupation of the light bearing fault diagnosis model provided by the invention is only 0.09M, which is far smaller than that of the existing bearing fault diagnosis model based on the convolutional neural network, and the extremely small memory occupation can enable the model to be more widely deployed in small equipment, thereby providing convenience for fault diagnosis of the small equipment.
(3) According to the invention, when the light CNN network is constructed, the input characteristic diagram and the output characteristic diagram are connected in a local mode, namely, the grouping connection is adopted, so that the parameter quantity of the model can be greatly reduced, and the memory occupation of the model is reduced.
(4) The method utilizes the example normalization to replace the existing batch normalization, and further improves the performance of the model.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
FIG. 2 is a schematic diagram of an example normalization process in accordance with the present invention.
Fig. 3 is a diagram illustrating a general convolution and a packet convolution according to the present invention.
Fig. 4 is a schematic diagram of a lightweight CNN network structure according to the present invention.
FIG. 5 is a schematic diagram of the system of the present invention.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined by the appended claims, and all changes that can be made by the invention using the inventive concept are intended to be protected.
Example 1
As shown in fig. 1, the invention provides a CNN-based light bearing fault diagnosis method, which is implemented as follows:
s1, data acquisition: resampling a vibration signal collected by an acceleration sensor arranged on a bearing to obtain a time domain signal with the continuous length of 800 data points, and performing frequency domain transformation on the time domain signal to obtain a frequency domain signal with the length of 400 data points;
s2, data transformation: changing the one-dimensional vector of the frequency domain signal into a two-dimensional vector, and taking the two-dimensional vector as a data sample;
s3, data set preparation: repeating the step S1 and the step S2 to obtain a plurality of data samples of various faults, and respectively marking fault labels to obtain a training data set, for example, 600 data samples of various faults are obtained;
s4, constructing a light network: and (3) constructing a light CNN network by using example normalization and packet convolution, wherein the method is realized as follows:
s401, reducing the input size of the light CNN network;
s402, eliminating the dependence of the light CNN network on batches by using an example normalization method;
s403, sequentially connecting an input layer, a first convolution layer, a first normalization layer, a first pooling layer, a second convolution layer, a second normalization layer, a second pooling layer, a full-connection layer and an output layer;
s404, using the grouping convolution to replace the common convolution to complete the construction of the light CNN network;
the input layer is used for receiving a vibration signal to be input; the first convolution layer is used for extracting the characteristics of the vibration signal; the first normalization layer is used for carrying out example normalization processing on the extracted features; the first pooling layer is used for performing down-sampling processing on the features subjected to the example normalization processing; the second convolution layer is used for performing grouping convolution processing on the features subjected to the down-sampling processing; the second normalization layer is used for carrying out example normalization processing on the features subjected to the grouping convolution processing of the second convolution layer; the second pooling layer is used for performing down-sampling processing on the features processed by the second normalization layer; the full connection layer is used for flattening all the characteristics processed by the second pooling layer; and the output layer is used for outputting the fault type of the bearing based on all the characteristics output by the full connection layer.
In the embodiment, the first convolution layer can automatically and effectively extract features, and the first normalization layer and the second normalization layer perform batch-size-independent instance normalization processing on data to accelerate model convergence and improve training speed, so that the method is more suitable for training models in small batches, the memory occupation of the models can be reduced, and the method can be deployed on small-sized equipment; the first pooling layer and the second pooling layer are subjected to down-sampling processing to reduce the parameter and the calculated amount of the model; the second convolution layer is a block convolution, and the calculated amount and parameters of the model can be greatly reduced through the block convolution, so that the key of light weight is realized.
S404, replacing the common convolution with the grouping convolution to complete the construction of the light CNN network;
the expression of the loss function of the lightweight CNN network is as follows:
wherein Cross Encopy represents the cross entropy loss function of the lightweight CNN network, p i Representing the true distribution of the ith data sample, q i Representing the prediction distribution of the ith data sample, and n represents the number of the data samples;
s5, network training: training parameters of the light CNN network by using a random gradient descent method according to the training data set;
s6, saving and deploying: storing the trained light CNN network parameters, and deploying the light CNN network to application equipment;
s7, bearing fault diagnosis: and making a data sample according to the bearing fault data to be diagnosed by a data acquisition and data transformation method, inputting the made data sample into deployed application equipment, and outputting the fault type of the bearing.
In this embodiment, a calculation formula of memory occupation m of a deep learning network is as follows:
m=B×(i+f+b)+p (1)
in the formula, B represents the batch size, i represents the model input memory occupation size, f is the forward propagation memory occupation size, B represents the backward propagation memory occupation size, and p represents the parameter memory occupation size of the model. The input memory occupation size of the model is determined by the input size of the model, and the memory occupation sizes of the forward propagation and backward propagation parameters are related by the depth of the model. Therefore, to reduce the memory usage of the model, the model needs to be adjusted from 4 parts, namely, the batch B during model training is reduced, the input size of the model is reduced, the depth of the network is reduced, and the parameter number of the model is reduced.
Most of the existing deep learning-based bearing fault diagnosis models adopt a batch normalization method, and have the advantages that gradient descent and gradient explosion can be avoided, and a certain regularization effect can replace Dropout; the disadvantage is that the batch normalization method is extremely dependent on batch size, since batch normalization uses the mean and standard deviation of the training samples to train the model as the mean and standard deviation of the overall distribution. If the batch size is too small, the mean and standard deviation of the training samples and the overall distribution are too different, so that the model performance is not good, therefore, the batch normalization network adopts large batch to improve the model performance, and as can be seen from formula (1), the large batch size inevitably increases the memory occupation of the model linearly, so the light network cannot use the batch normalization method.
The Normalization process of example Normalization (IN) was not performed on a batch and is therefore an excellent alternative to batch Normalization. Suppose a feature map to be normalized isWherein B, C, W and H respectively represent the batch number, the channel number, the width and the height of the characteristic diagram. Then an example normalization process for a certain value x in the feature map is as follows:
example normalization is calculated over the W and H dimensions, while preserving the B and c dimensions, so u IN And σ IN The calculation formula of (a) is as follows:
where y represents an example normalization process for some data x in the feature map, μ IN Denotes the mean value, σ IN Denotes the standard deviation,. Epsilon.denotes the minimum number,. Gamma.and.beta.denote the parameters to be trained, u IN (x) Means, σ, representing a certain data x in the characteristic map IN (x) The standard deviation of certain data x in the characteristic diagram is shown, W and H respectively show the width and height of the characteristic diagram to be normalized, b shows a certain batch, c shows a certain channel, W shows a certain coordinate in the width of the characteristic diagram, and H shows a certain coordinate in the height of the characteristic diagram.
To better understand the example normalization process, the normalization process is visualized, and as shown in fig. 2, the dark part of the graph is the part to be normalized. It can be seen from the figure that the normalization of the example normalized IN is independent of lot B, so the present invention uses the example normalized IN as the normalization method for the model, while fixing B to 1.
In this embodiment, the existing convolutional neural network-based bearing fault diagnosis model is a common convolution, that is, full connection is adopted between the input feature map and the output feature map. In order to reduce the parameter quantity of the network, the invention adopts the grouping connection between the input characteristic diagram and the output characteristic diagram, namely the local connection is adopted between the input characteristic diagram and the output characteristic diagram, thus greatly reducing the parameter quantity of the model and further reducing the memory occupation of the model. To better understand the difference between the two, examples of the normal convolution and the packet convolution are shown in fig. 3.
In this embodiment, in order to reduce the memory usage of the model to the maximum, a bearing fault diagnosis structure based on a convolutional neural network is designed. First, the model input size is reduced by reducing the input size of the model to 20 × 20; then, an example normalization method is adopted to eliminate the dependence on the batches; secondly, the designed network only has two convolution layers and 2 pooling layers to reduce the network depth so as to reduce the memory occupation in the forward and backward propagation processes; finally, packet convolution is employed to reduce the number of parameters of the network. The memory occupation of the model can be reduced to the limit through the 4 steps, so that the deployment range of the model is improved to the maximum extent. The specific structure of the network is shown in fig. 4, where in fig. 4, input represents an Input layer and Output represents an Output layer. Specific parameters of the network structure are shown in table 1, wherein the size format in the table is C × W × H, and N represents the number of types of faults to be classified.
TABLE 1
In this embodiment, when the lightweight CNN network is trained, the lightweight CNN network is trained according to the training parameters in table 2 below with the training data set.
TABLE 2
According to the invention, the memory occupation of the light bearing fault diagnosis model is only 0.09M and is far smaller than that of the existing bearing fault diagnosis model based on the convolutional neural network, and the extremely small memory occupation can enable the model to be more widely deployed in small equipment, so that the deployment range of the model is maximally improved, and meanwhile, convenience is provided for fault diagnosis based on the improved bearing fault diagnosis model and the deployment thereof.
Example 2
As shown in the figure, the present invention provides a CNN-based fault diagnosis system for a light bearing, comprising:
the data acquisition module is used for resampling vibration signals acquired by the acceleration sensor arranged on the bearing to obtain time domain signals with the continuous length of 800 data points, and performing frequency domain transformation on the time domain signals to obtain frequency domain signals with the length of 400 data points;
the data transformation module is used for transforming the one-dimensional vector of the frequency domain signal into a two-dimensional vector and taking the two-dimensional vector as a data sample;
the data set making module is used for obtaining a plurality of data samples of various faults based on the data obtaining module and the data transformation module, and respectively marking fault labels to obtain a training data set;
constructing a light network module for constructing a light CNN network by using example normalization and packet convolution;
the network training module is used for training the parameters of the light CNN network by using a random gradient descent method according to the training data set;
the saving and deploying module is used for saving the trained light CNN network parameters and deploying the light CNN network to the application equipment;
and the bearing fault diagnosis module is used for making a data sample from the bearing fault data to be diagnosed according to a data acquisition and data transformation method, inputting the made data sample into deployed application equipment and outputting the fault type of the bearing.
The CNN-based light bearing fault diagnosis system provided in the embodiment shown in fig. 5 can implement the technical solution shown in the CNN-based light bearing fault diagnosis method in the above-described method embodiment, and the implementation principle and beneficial effects thereof are similar, and are not described herein again.
In the embodiment of the invention, the functional units can be divided according to the CNN-based light bearing fault diagnosis method, for example, each function can be divided into each functional unit, and two or more functions can be integrated into one processing unit. The integrated unit may be implemented in the form of hardware, or may be implemented in the form of a software functional unit. It should be noted that the division of the cells in the present invention is schematic, and is only a logical division, and there may be another division manner in actual implementation.
In the embodiment of the invention, in order to realize the principle and the beneficial effect of the CNN-based light bearing fault diagnosis method, the CNN-based light bearing fault diagnosis system comprises a hardware structure and/or a software module which are corresponding to the execution of each function. Those of skill in the art will readily appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as hardware and/or combinations of hardware and computer software, where a function is performed in a hardware or computer software-driven manner, and that the function described may be implemented in any suitable manner for each particular application depending upon the particular application and design constraints imposed on the technology, but such implementation is not to be considered as beyond the scope of the present application.
According to the invention, the memory occupation of the light bearing fault diagnosis model is only 0.09M and is far smaller than that of the existing bearing fault diagnosis model based on the convolutional neural network, and the extremely small memory occupation can enable the model to be more widely deployed in small equipment, so that the deployment range of the model (such as the small equipment) is maximally improved, and meanwhile, convenience is provided for fault diagnosis based on the improved bearing fault diagnosis model and the deployment range thereof.
Claims (6)
1. A CNN-based light bearing fault diagnosis method is characterized by comprising the following steps:
s1, data acquisition: resampling a vibration signal collected by an acceleration sensor arranged on a bearing to obtain a time domain signal with the continuous length of 800 data points, and performing frequency domain transformation on the time domain signal to obtain a frequency domain signal with the length of 400 data points;
s2, data transformation: changing the one-dimensional vector of the frequency domain signal into a two-dimensional vector, and taking the two-dimensional vector as a data sample;
s3, data set preparation: repeating the step S1 and the step S2 to obtain a plurality of data samples of various faults, and respectively marking fault labels to obtain a training data set;
s4, constructing a light network: constructing a light CNN network by using instance normalization and packet convolution;
s5, network training: training parameters of the light CNN network by using a random gradient descent method according to the training data set;
s6, saving and deploying: storing the trained light CNN network parameters, and deploying the light CNN network to application equipment;
s7, bearing fault diagnosis: and making a data sample according to the bearing fault data to be diagnosed by a data acquisition and data transformation method, inputting the made data sample into deployed application equipment, and outputting the fault type of the bearing.
2. The CNN-based light bearing fault diagnosis method according to claim 1, wherein said step S4 comprises the steps of:
s401, reducing the input size of the light CNN network;
s402, eliminating the dependence of the light CNN network on batches by using an example normalization method;
s403, sequentially connecting an input layer, a first convolution layer, a first normalization layer, a first pooling layer, a second convolution layer, a second normalization layer, a second pooling layer, a full-connection layer and an output layer;
s404, using packet convolution to replace ordinary convolution to complete the construction of the light CNN network.
3. The CNN-based light bearing fault diagnosis method according to claim 2, wherein the example normalized process expression is as follows:
where y represents an example normalization process for some data x in the feature map, μ IN Denotes the mean value, σ IN Denotes the standard deviation,. Epsilon.denotes the minimum number,. Gamma.and.beta.denote the parameters to be trained, u IN (x) Representing the mean, σ, of a certain data x in a feature map IN (x) Representing the standard deviation of certain data x in the characteristic diagram, W and H respectively representing the width and height of the characteristic diagram to be normalized, b representing a certain batch, c representing a certain channel, W representing a certain characteristic diagram widthOne coordinate, h, represents a certain coordinate in the height of the feature map.
4. The CNN-based light bearing failure diagnosis method of claim 3, wherein the input layer is configured to receive a vibration signal to be input;
the first convolution layer is used for extracting the characteristics of the vibration signal;
the first normalization layer is used for carrying out example normalization processing on the extracted features;
the first pooling layer is used for performing down-sampling processing on the features subjected to the example normalization processing;
the second convolution layer is used for performing grouping convolution processing on the features subjected to the down-sampling processing;
the second normalization layer is used for carrying out example normalization processing on the features subjected to the grouping convolution processing of the second convolution layer;
the second pooling layer is used for performing down-sampling processing on the features processed by the second normalization layer;
the full connection layer is used for flattening all the characteristics processed by the second pooling layer;
and the output layer is used for outputting the fault type of the bearing based on all the characteristics output by the full connection layer.
5. The CNN-based lightweight bearing fault diagnosis method according to claim 3, wherein the expression of the loss function of said lightweight CNN network is as follows:
wherein Cross Encopy represents the cross entropy loss function of the lightweight CNN network, p i Representing the true distribution of the ith data sample, q i Represents the predicted distribution of the ith data sample, and n represents the number of data samples.
6. A CNN-based lightweight bearing fault diagnosis system, comprising:
the data acquisition module is used for resampling vibration signals acquired by the acceleration sensor arranged on the bearing to obtain time domain signals with the continuous length of 800 data points, and performing frequency domain transformation on the time domain signals to obtain frequency domain signals with the length of 400 data points;
the data transformation module is used for transforming the one-dimensional vector of the frequency domain signal into a two-dimensional vector and taking the two-dimensional vector as a data sample;
the data set making module is used for obtaining a plurality of data samples of various faults based on the data obtaining module and the data transformation module, and respectively marking fault labels to obtain a training data set;
constructing a light network module for constructing a light CNN network by using example normalization and packet convolution;
the network training module is used for training the parameters of the light CNN network by using a random gradient descent method according to the training data set;
the saving and deploying module is used for saving the trained light CNN network parameters and deploying the light CNN network to the application equipment;
and the bearing fault diagnosis module is used for making a data sample from the bearing fault data to be diagnosed according to a data acquisition and data transformation method, inputting the made data sample into deployed application equipment and outputting the fault type of the bearing.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210737500.8A CN115165363B (en) | 2022-06-27 | 2022-06-27 | Light bearing fault diagnosis method and system based on CNN |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210737500.8A CN115165363B (en) | 2022-06-27 | 2022-06-27 | Light bearing fault diagnosis method and system based on CNN |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115165363A true CN115165363A (en) | 2022-10-11 |
CN115165363B CN115165363B (en) | 2024-07-19 |
Family
ID=83486591
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210737500.8A Active CN115165363B (en) | 2022-06-27 | 2022-06-27 | Light bearing fault diagnosis method and system based on CNN |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115165363B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116704266A (en) * | 2023-07-28 | 2023-09-05 | 国网浙江省电力有限公司信息通信分公司 | Power equipment fault detection method, device, equipment and storage medium |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180046894A1 (en) * | 2016-08-12 | 2018-02-15 | DeePhi Technology Co., Ltd. | Method for optimizing an artificial neural network (ann) |
CN109863537A (en) * | 2016-10-21 | 2019-06-07 | 谷歌有限责任公司 | Stylized input picture |
WO2020118553A1 (en) * | 2018-12-12 | 2020-06-18 | 深圳鲲云信息科技有限公司 | Method and device for quantizing convolutional neural network, and electronic device |
US20210020360A1 (en) * | 2019-07-15 | 2021-01-21 | Wuhan University | Internal thermal fault diagnosis method of oil-immersed transformer based on deep convolutional neural network and image segmentation |
CN112488287A (en) * | 2020-02-17 | 2021-03-12 | 上海交通大学 | Convolutional neural network compression method, system, device and medium |
CN112597812A (en) * | 2020-12-03 | 2021-04-02 | 西安格威西联科技有限公司 | Finger vein identification method and system based on convolutional neural network and SIFT algorithm |
CN112613581A (en) * | 2020-12-31 | 2021-04-06 | 广州大学华软软件学院 | Image recognition method, system, computer equipment and storage medium |
CN112801989A (en) * | 2021-02-02 | 2021-05-14 | 西南交通大学 | Contact net diagonal line fault automatic detection method based on image enhancement processing |
CN113505830A (en) * | 2021-07-09 | 2021-10-15 | 西安交通大学 | Rotating machine fault diagnosis method, system, equipment and storage medium |
CN113919220A (en) * | 2021-10-08 | 2022-01-11 | 西安科技大学 | Intelligent fault diagnosis method for rolling bearing driven by vibration image |
CN114444582A (en) * | 2022-01-07 | 2022-05-06 | 太原理工大学 | Mechanical equipment fault diagnosis method based on convolutional neural network and Bayesian network |
-
2022
- 2022-06-27 CN CN202210737500.8A patent/CN115165363B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180046894A1 (en) * | 2016-08-12 | 2018-02-15 | DeePhi Technology Co., Ltd. | Method for optimizing an artificial neural network (ann) |
CN109863537A (en) * | 2016-10-21 | 2019-06-07 | 谷歌有限责任公司 | Stylized input picture |
WO2020118553A1 (en) * | 2018-12-12 | 2020-06-18 | 深圳鲲云信息科技有限公司 | Method and device for quantizing convolutional neural network, and electronic device |
US20210020360A1 (en) * | 2019-07-15 | 2021-01-21 | Wuhan University | Internal thermal fault diagnosis method of oil-immersed transformer based on deep convolutional neural network and image segmentation |
CN112488287A (en) * | 2020-02-17 | 2021-03-12 | 上海交通大学 | Convolutional neural network compression method, system, device and medium |
CN112597812A (en) * | 2020-12-03 | 2021-04-02 | 西安格威西联科技有限公司 | Finger vein identification method and system based on convolutional neural network and SIFT algorithm |
CN112613581A (en) * | 2020-12-31 | 2021-04-06 | 广州大学华软软件学院 | Image recognition method, system, computer equipment and storage medium |
CN112801989A (en) * | 2021-02-02 | 2021-05-14 | 西南交通大学 | Contact net diagonal line fault automatic detection method based on image enhancement processing |
CN113505830A (en) * | 2021-07-09 | 2021-10-15 | 西安交通大学 | Rotating machine fault diagnosis method, system, equipment and storage medium |
CN113919220A (en) * | 2021-10-08 | 2022-01-11 | 西安科技大学 | Intelligent fault diagnosis method for rolling bearing driven by vibration image |
CN114444582A (en) * | 2022-01-07 | 2022-05-06 | 太原理工大学 | Mechanical equipment fault diagnosis method based on convolutional neural network and Bayesian network |
Non-Patent Citations (5)
Title |
---|
张应军;江永全;杨燕;张卫华;陈锦雄;: "基于深度卷积神经网络的未知复合故障诊断", 中国科技论文, vol. 14, no. 02, 15 February 2019 (2019-02-15) * |
汪洋;郭利进;: "一种新型DSCNN-GRU结构的减速机轴承故障诊断方法", 机械科学与技术, no. 02, 2 June 2019 (2019-06-02) * |
田勇;董国贵: "基于轻量级CNN的电机轴承故障诊断研究", 齐齐哈尔大学学报(自然科学版), vol. 38, no. 01, 16 November 2021 (2021-11-16) * |
程越;刘志刚;: "基于轻量型卷积神经网络的交通标志识别方法", 计算机***应用, no. 02, 15 February 2020 (2020-02-15) * |
韩逍豫: "基于CNN的滚动轴承故障诊断与寿命预测方法研究", 《全国优秀硕士学位论文全文数据库》, 1 December 2020 (2020-12-01) * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116704266A (en) * | 2023-07-28 | 2023-09-05 | 国网浙江省电力有限公司信息通信分公司 | Power equipment fault detection method, device, equipment and storage medium |
CN116704266B (en) * | 2023-07-28 | 2023-10-31 | 国网浙江省电力有限公司信息通信分公司 | Power equipment fault detection method, device, equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN115165363B (en) | 2024-07-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111369563B (en) | Semantic segmentation method based on pyramid void convolutional network | |
CN112991354B (en) | High-resolution remote sensing image semantic segmentation method based on deep learning | |
CN109596326B (en) | Rotary machine fault diagnosis method based on convolution neural network with optimized structure | |
CN112381097A (en) | Scene semantic segmentation method based on deep learning | |
CN113033570A (en) | Image semantic segmentation method for improving fusion of void volume and multilevel characteristic information | |
CN110533591B (en) | Super-resolution image reconstruction method based on codec structure | |
CN116416497B (en) | Bearing fault diagnosis system and method | |
CN114283120B (en) | Domain-adaptive-based end-to-end multisource heterogeneous remote sensing image change detection method | |
CN111008726A (en) | Class image conversion method in power load prediction | |
CN113203566A (en) | Motor bearing fault diagnosis method based on one-dimensional data enhancement and CNN | |
CN112766283B (en) | Two-phase flow pattern identification method based on multi-scale convolution network | |
CN115165363B (en) | Light bearing fault diagnosis method and system based on CNN | |
CN115049941A (en) | Improved ShuffleNet convolutional neural network and remote sensing image classification method thereof | |
CN115452376A (en) | Bearing fault diagnosis method based on improved lightweight deep convolution neural network | |
CN113344110B (en) | Fuzzy image classification method based on super-resolution reconstruction | |
CN112634391B (en) | Gray image depth reconstruction and fault diagnosis system based on compressed sensing | |
CN113869451A (en) | Rolling bearing fault diagnosis method under variable working conditions based on improved JGSA algorithm | |
CN113192084A (en) | Machine vision-based highway slope micro-displacement deformation monitoring method | |
CN112149518A (en) | Pine cone detection method based on BEGAN and YOLOV3 models | |
CN114494284B (en) | Scene analysis model and method based on explicit supervision area relation | |
CN112990336B (en) | Deep three-dimensional point cloud classification network construction method based on competitive attention fusion | |
CN113536998A (en) | Motor fault diagnosis method and system based on improved stack type sparse self-encoder | |
CN114298101A (en) | Dynamic weighing method and system | |
CN114548153B (en) | Planetary gear box fault diagnosis method based on residual error-capsule network | |
CN117878928B (en) | Wind power prediction method and device based on deep learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant |