CN114459760A - Rolling bearing fault diagnosis method and system under strong noise environment - Google Patents
Rolling bearing fault diagnosis method and system under strong noise environment Download PDFInfo
- Publication number
- CN114459760A CN114459760A CN202111672334.XA CN202111672334A CN114459760A CN 114459760 A CN114459760 A CN 114459760A CN 202111672334 A CN202111672334 A CN 202111672334A CN 114459760 A CN114459760 A CN 114459760A
- Authority
- CN
- China
- Prior art keywords
- layer
- cnn
- feature extractor
- fault
- senet
- 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 36
- 238000005096 rolling process Methods 0.000 title claims abstract description 34
- 238000003745 diagnosis Methods 0.000 title claims abstract description 30
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 75
- 238000012549 training Methods 0.000 claims abstract description 53
- 238000005070 sampling Methods 0.000 claims abstract description 38
- 238000003062 neural network model Methods 0.000 claims abstract description 12
- 238000013135 deep learning Methods 0.000 claims abstract description 11
- 230000001133 acceleration Effects 0.000 claims description 33
- 238000012360 testing method Methods 0.000 claims description 33
- 238000012545 processing Methods 0.000 claims description 20
- 238000010606 normalization Methods 0.000 claims description 15
- 230000004913 activation Effects 0.000 claims description 9
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000009826 distribution Methods 0.000 claims description 8
- 238000011176 pooling Methods 0.000 claims description 6
- 230000002708 enhancing effect Effects 0.000 abstract 1
- 230000006870 function Effects 0.000 description 8
- 230000000694 effects Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 238000002474 experimental method Methods 0.000 description 4
- 239000002131 composite material Substances 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 239000000654 additive Substances 0.000 description 2
- 230000000996 additive effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 102100033620 Calponin-1 Human genes 0.000 description 1
- 102100033592 Calponin-3 Human genes 0.000 description 1
- 101000945318 Homo sapiens Calponin-1 Proteins 0.000 description 1
- 101000945410 Homo sapiens Calponin-3 Proteins 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009760 electrical discharge machining Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000036039 immunity Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012847 principal component analysis method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 230000000007 visual effect Effects 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Abstract
The invention discloses a method and a system for diagnosing faults of a rolling bearing in a strong noise environment, wherein deep learning is adopted for bearing fault diagnosis, and a neural network model consists of a random sampling layer, a SENet combined CNN global type feature extractor, a CNN-based high-dimensional feature extractor, a GRU-based feature classifier, a full connection layer and a softmax layer. Inputting the normalized training data into a neural network model, randomly sampling the training data by a random sampling layer to increase the anti-interference performance of the model, extracting global fault features by adopting a CNN global feature extractor combined with SEnet, abstracting the global fault features into higher-dimensional features by a CNN-based high-dimensional feature extractor, enhancing feature expression capability by adopting a GRU-based feature classifier, and obtaining a fault category diagnosis result by a full connection layer and a Softmax layer. The method can overcome noise interference and diagnose the rolling bearing signals under various loads.
Description
Technical Field
The invention belongs to the technical field of fault diagnosis of rolling bearings, and particularly relates to a fault diagnosis method and system for a rolling bearing in a strong noise environment.
Background
Rolling bearings are one of the most important parts of a rotating machine, and a great economic loss is generated once serious failure occurs. With the rapid development of the sensing technology, people can well record the running condition of the bearing, but how to dig out fault information in the mass data becomes a research hotspot in the field of bearing fault diagnosis.
In recent years, with the rapid development of deep learning technology, the feature representation and learning ability of the deep learning technology are greatly improved, and even an end-to-end use mode appears, namely, an output result can be obtained by inputting an original signal to a neural network model or only adopting the original signal processed by a simple preprocessing and data expansion method. In addition to convenience, it is important that the end-to-end approach exhibit very robust performance, far more practical than most methods of manual feature extraction.
Although most of the current bearing fault diagnosis methods based on deep learning have high accuracy, most of the bearing fault diagnosis methods do not consider the compound factors of strong noise and multiple loads. Although some factors consider strong noise, the bearing fault diagnosis is only carried out under a certain load; the invention aims to provide a rolling bearing fault diagnosis method and system suitable for the high-noise environment, which can overcome noise interference and diagnose rolling bearing signals under various loads.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a method and a system for diagnosing the fault of a rolling bearing in a strong noise environment, which can improve the fault diagnosis performance of the rolling bearing in a strong noise and multi-load composite environment.
The technical solution for realizing the purpose of the invention is as follows:
a fault diagnosis method for a rolling bearing under a strong noise environment comprises the following steps:
s1 at sampling frequency fsCollecting vibration acceleration signals x of bearings with no fault, different fault types and different fault degrees under different loadsi,j；
S2: for the collected vibration acceleration signal xi,jCarrying out normalization processing to obtain normalized vibration acceleration signal
Setting a deep learning training label for the normalized vibration acceleration signal according to the fault type and the fault degree;
s3: constructing a neural network model, comprising: the device comprises a random sampling layer, a CNN global type feature extractor combined with SENet, a CNN-based high-dimensional feature extractor, a GRU-based feature classifier, a full connection layer and a Softmax layer; the normalized sample is sequentially processed by a random sampling layer, a CNN global type feature extractor combined with SENet, a CNN-based high-dimensional feature extractor, a GRU-based feature classifier and a full connection layer, and finally output by a Softmax layer to obtain a result label; the random sampling layer is used for adding training data interfered by analog noise; the CNN global type feature extractor combined with SENEt is used for extracting global type fault features; the CNN-based high-dimensional feature extractor is used for abstracting high-dimensional fault features from the extracted global fault features; the GRU-based feature classifier is used for clustering features extracted by the CNN-based high-dimensional feature extractor;
s4: the normalized vibration acceleration signal processed in the step S2 is
Inputting the corresponding training labels into the neural network model in the step S3 for training;
s5: using the same sampling frequency f as in step S1sAcquiring vibration data of a rolling bearing of the current equipment to obtain a vibration acceleration signal X to be measuredi,j；
S6: according to the step S2, the vibration acceleration signal X to be measuredi,jNormalization processing is carried out to obtain normalized test data
Test data after normalization
Inputting the current bearing fault state into the trained model in step S4.
A rolling bearing fault diagnosis system in a strong noise environment includes:
signal acquisition module for sampling frequency fsCollecting vibration acceleration signals x of bearings with no fault, different fault types and different fault degrees under different loadsi,j；
A normalization processing module for processing the collected vibration acceleration signal xi,jCarrying out normalization processing to obtain normalized vibration acceleration signal
The training label setting module is used for setting a deep learning training label for the normalized vibration acceleration signal according to the fault type and the fault degree;
a neural network training module comprising: the device comprises a random sampling layer, a CNN global type feature extractor combined with SENet, a CNN-based high-dimensional feature extractor, a GRU-based feature classifier, a full connection layer and a Softmax layer; the normalized sample is sequentially processed by a random sampling layer, a CNN global type feature extractor combined with SENet, a CNN-based high-dimensional feature extractor, a GRU-based feature classifier and a full connection layer, and finally output by a Softmax layer to obtain a result label; the random sampling layer is used for adding training data interfered by analog noise; the CNN global type feature extractor combined with SENEt is used for extracting global type fault features; the CNN-based high-dimensional feature extractor is used for abstracting high-dimensional fault features from the extracted global fault features; the GRU-based feature classifier is used to cluster features extracted by the CNN-based high-dimensional feature extractor.
Compared with the prior art, the invention has the following remarkable advantages:
1. the invention can automatically learn the fault information from the original vibration signal without carrying out complicated denoising and artificial characteristic selection processes.
2. The invention carries out random sampling on the training data, and can greatly improve the anti-noise interference capability.
3. The invention adopts the CNN global type feature extractor combined with SEnet to obtain important global features efficiently.
4. The invention adopts the GRU-based feature classifier to process the features extracted by the CNN feature extractor, thereby preventing the loss of the associated information of the original vibration signal.
5. The invention can realize stable and accurate rolling bearing fault diagnosis under strong noise and multi-load composite environment by automatically adjusting the distribution of test data.
Drawings
Fig. 1 is a flowchart of an embodiment of a method and a system for diagnosing a rolling bearing fault in a strong noise environment according to the present invention.
FIG. 2 is a diagram of a neural network model architecture of the present invention.
Fig. 3 is a manifold representation of the random sampling layer of the present invention.
Fig. 4 is a specific configuration diagram of the SENet in fig. 2.
Fig. 5 is an external view of the rolling bearing test stand according to the present embodiment.
Fig. 6 is a comparison graph of the diagnosis results of the present invention and five comparison methods in the present embodiment under the complex environment of strong noise and multiple loads.
Fig. 7 is a graph showing the result of the visual experiment in this example, and the results of mapping the original signal, the CNN1 layer, the SENet layer, the CNN3 layer, the GRU layer, and the FC layer are shown in order from 7(a to f).
Detailed Description
The following description of the embodiments of the present invention is provided in order to better understand the present invention for those skilled in the art with reference to the accompanying drawings. It is to be expressly noted that in the following description, a detailed description of known functions and designs will be omitted when it may obscure the subject matter of the present invention.
Fig. 1 is a flowchart of an embodiment of a method and a system for diagnosing a rolling bearing fault in a strong noise environment according to the present invention. As shown in fig. 1, the method for diagnosing the fault of the rolling bearing in the strong noise and multi-load composite environment comprises the following specific steps:
s1 at sampling frequency fsCollecting vibration acceleration signals x of bearings with no fault, different fault types and different fault degrees under different loadsi,j；
S2: for the collected vibration acceleration signal xi,jCarrying out normalization processing to obtain normalized vibration acceleration signal
Setting a deep learning training label for the normalized vibration acceleration signal according to the fault type and the fault degree;
s3: constructing a neural network model;
the structure diagram of the neural network model of the invention is shown in fig. 2, and comprises six parts: the system comprises a random sampling layer, a SENEt-combined CNN global feature extractor, a CNN-based high-dimensional feature extractor, a GRU-based feature classifier, a full connection layer and a Softmax layer, wherein all the components are explained below.
1. The random sampling layer is a random sampling that follows a bernoulli distribution with a probability p, where p takes a random value between 0.5 and 1 per training period. To preserve data length, the data bits that are not sampled are set to 0. The step is provided based on two aspects, on one hand, partial information is actively lost during training, so that the effect of simulating noise interference is achieved, and the anti-noise interference capability of the whole network is greatly improved. On the other hand, manifold theory holds that if data of low dimensionality can be mapped to a number of high dimensionality by manifoldAccordingly, it can represent the characteristics of the data itself. Similarly, the bearing fault information is abstractly represented by the low-dimensional manifold shown in fig. 3, the point x on the solid line is the original training data, and the neural network can predict the test data x' completely. But if it is desired to predict the test data to which the noise is added
It is desirable to broaden the range of the original training data x to cover the noisy test data
New training data x-mapped from the original training data can be obtained by disruption of the random sampling of the original data
The extent of the mapping is in the dashed circle as shown in fig. 3. Continuously trained by neural networks, with noise-added test data within the manifold represented by the dashed line
Can be predicted, thereby greatly improving the robustness of the whole neural network.
2. The enhanced CNN feature extractor combined with SENEt is composed of a CNN feature extractor and a SENEt in turn.
2.1: the CNN feature extractor is composed of a convolution layer, an AdaBN layer and an ELU activation layer in sequence. A large-sized convolution kernel is employed to increase the field of view of the network to extract periodic fault features and suppress high frequency noise. The characteristics extracted from the convolutional layer are processed by adopting an AdaBN method so as to reduce the difference between the distribution of test data and training data, accelerate the convergence speed of the model and reduce the training time of the model. And finally, an ELU activation function is used, so that the input negative half-axis information is saved while gradient disappearance is avoided, and therefore, more comprehensive information can be acquired in the subsequent steps.
2.1.1: the AdaBN layer can carry out standardization operation on the convolution layer result, and the standardization formula is as follows:
in the formula: x is the number of(k)As input to the AdaBN layer, γ(k)、β(k)Scaling and biasing parameters for AdaBN layer, y(k)Is the output of AdaBN layer, where gamma(k)、β(k)In the training mode is a parameter of the autonomous learning training, and in the test mode the test data is updated with the parameter
So that the test data is
Even if it is affected by noise, it is associated with training data
Good prediction results can also be obtained in different distributions.
2.1.2: the ELU function is:
in the formula: x is the input of the ELU function, a is a positive over-parameter, and the saturation of a negative input value can be adjusted, so that the network has stronger robustness. It can solve the problem of gradient disappearance to some extent like the ReLU function in the non-negative region of the input, but it is more remarkable than the ReLU function in that it can utilize the information of the negative half axis of the input x of the ELU function and make the overall output of the function more toward 0, thereby speeding up the convergence of the network.
2.2: the SEnet structure is shown in FIG. 4 and comprises a global average pooling layer, a full connection layer, an ELU active layer, a full connection layer and a Sigmoid active layer in sequence. And the SENEt block is adopted to carry out channel feature enhancement on the features extracted by the CNN feature extractor, so that beneficial features are automatically enhanced, useless features are inhibited, and the effect of global feature extraction is achieved. The input of SEnet is the result of the output of the CNN feature extractor, and the output of the enhanced CNN feature extractor in combination with SEnet is the result of the multiplication of the output of SEnet and the output of CNN feature extractor.
3. The CNN-based high-dimensional feature extractor consists of a plurality of CNN feature extractors, and the following example employs two CNN feature extractors. Each CNN feature extractor consists of a convolution layer, an AdaBN layer, an ELU activation layer and a maximum pooling layer in sequence.
4. The GRU-based feature classifier is formed by stacking a plurality of GRU layers, and the last time step of the last GRU layer is used as the output of the GRU-based feature classifier. Since the GRU can extract the information associated with each other, it can be used to process the CNN generated features, and can better capture the associated information of the original data. Compared with a network consisting of single CNN, the network formed by combining the CNN and the GRU can stably work under the interference of strong noise better.
5. The fully-connected layer is a linear neuron layer, and high-dimensional information is mapped to the probability of each fault state.
6. And the Softmax layer takes out the maximum value of the fault state probabilities output by the full connection layer as a fault diagnosis result.
S4: the normalized vibration acceleration signal processed in the step S2 is
Inputting the corresponding training labels into the neural network model in the step S3 for training;
s5: using the same sampling frequency f as in step S1sAcquiring vibration data of a rolling bearing of the current equipment to obtain a vibration acceleration signal X to be measuredi,j；
S6: vibration to be measured according to step S2Dynamic acceleration signal Xi,jNormalization processing is carried out to obtain normalized test data
Test data after normalization
Inputting the current bearing fault state into the trained model in step S4.
Based on the above method, the present embodiment further provides a system for diagnosing a fault of a rolling bearing in a strong noise environment, including:
signal acquisition module for sampling frequency fsCollecting vibration acceleration signals x of bearings with no fault, different fault types and different fault degrees under different loadsi,j；
A normalization processing module for processing the collected vibration acceleration signal xi,jCarrying out normalization processing to obtain normalized vibration acceleration signal
The training label setting module is used for setting a deep learning training label for the normalized vibration acceleration signal according to the fault type and the fault degree;
a neural network training module comprising: the device comprises a random sampling layer, a CNN global type feature extractor combined with SENet, a CNN-based high-dimensional feature extractor, a GRU-based feature classifier, a full connection layer and a Softmax layer; the normalized sample is sequentially processed by a random sampling layer, a CNN global type feature extractor combined with SENet, a CNN-based high-dimensional feature extractor, a GRU-based feature classifier and a full connection layer, and finally output by a Softmax layer to obtain a result label; the random sampling layer is used for adding training data interfered by analog noise; the CNN global type feature extractor combined with SENEt is used for extracting global type fault features; the CNN-based high-dimensional feature extractor is used for abstracting high-dimensional fault features from the extracted global fault features; the GRU-based feature classifier is used to cluster features extracted by the CNN-based high-dimensional feature extractor.
The arrangement and method of each layer of the neural network are as above, and are not described herein again.
In order to better illustrate the technical effects of the present invention, the present invention was experimentally verified by using a specific embodiment. This experiment verified that using bearing experimental data from the university of Keyssient storage (CWRU), using electrical discharge machining to provide a fault for the motor bearing, introducing a fault of 0.007 inch to 0.040 inch diameter at the inner raceway, rolling elements (i.e., balls) and outer raceway, respectively, reinstalling the faulty bearing into the test motor, and recording vibration data for motor loads of 0 to 3 horsepower (motor speeds of 1797 to 1720 RPM). Fig. 5 is an external view of a rolling bearing test stand according to the present embodiment, including a 1.5KW motor (left side of the figure), a torque sensor/encoder (middle connection of the figure) and a power meter (right side of the figure).
This example uses data collected at a frequency of 12Hz, one sample per 2048 points. In order to make the training samples more sufficient, the training data is expanded by adopting a data enhancement technology, so that more fault characteristics can be learned by deep learning, and the overall performance is improved. The data enhancement method adopted by the embodiment is an overlapping sampling method, so that not only are training samples increased, but also the fault information of the sample edge can be better learned. It should be noted that the data set must be divided into training data and test data before data enhancement can be performed on the training data, otherwise, the training data and the test data have partially repeated data. To verify the noise immunity of the proposed model, additive white gaussian noise was added to the test data to simulate random noise in an industrial environment. In order to simulate a real environment as much as possible, in the embodiment, in addition to adding Gaussian white noise of-10 dB into the data set, data under different loads of 1-3 hp are mixed with one another. The final training data is data of a multi-load working condition without noise, and the test data is data of a multi-load working condition with-10 dB of Gaussian white noise. The number of training data and test data and the type of failure are shown in table 1.
TABLE 1
Adam was selected as the optimizer for training in this experiment, with an initial learning rate of 0.005. In order to quickly improve the training precision of the model, a learning rate step-down strategy is adopted: the learning rate is reduced by 20% every 50 cycles, and 500 cycles are trained. The specific structural parameters of AnNet are shown in Table 2.
TABLE 2
TABLE 3
This example is to demonstrate the superiority of the proposed method, compared to the existing method, which is somewhat similar to the present invention. The method comprises the following steps: AAnNet based on CNN-GRU (see Jin G, Zhu T, Akram M W, et al. A Adaptive Anti-Noise Neural Network for Bearing Fault Diagnosis Under Noise and Bearing Conditions [ J ]. IEEE Access,2020,8: 74793-), WDCNN based on AdaBN (see Wei Z, Peng, Li C, et al. A New Learning Model for Bearing Diagnosis with Good Anti-Noise and Domain Adaptation on fiber optics channels [ J ]. Sensors,2017,17(3):425.) and CAT-GRU based on Attention channel (see Z, Wang H.Wang H.Wang.and testing H.F.H.F.F.H.and testing J. Noise, and Bearing Diagnosis, and "weight H.S. J.S. and testing J." weight-Noise and Bearing Conditions "balance 1 H.S. and testing J." weight 1 and weight correction ". The experiment was run independently 10 times and the mean and variance of the 10 experimental results were calculated.
The experimental result under the mixed environment of strong noise and multiple loads is shown in fig. 6, the neural network model provided by the invention has the highest prediction accuracy, and meanwhile, for tiny gaussian white noise interference, most methods can perform high-accuracy prediction, and with the increase of gaussian white noise, only the method provided by the invention has the slowest attenuation until the gaussian white noise is-6 dB, the gaussian white noise has the accuracy of 94.0%, and the ten independent test variance is 0.020, which also indicates that the stability is very good.
In order to intuitively know the processing effect of each network layer, the processing result of each layer of the bearing fault signal can be mapped by the method through the t-SNE dimension reduction technology. For fast processing, the embodiment reduces the high-dimensional feature dimension from the high-dimensional feature to 100 dimensions by using a principal component analysis method, and then maps the 100-dimensional feature onto a two-dimensional plane by using a t-SNE algorithm, so as to fully show the feature processing effect of each network layer.
The outputs of the various layers of the neural network were visualized using the t-SNE dimensionality reduction technique with the test data processed with-4 dB additive white gaussian noise as the input to the network, with the results shown in fig. 7 (a-f). There are 10 different types of data points in the graph, the values of which correspond one-to-one to table 1. It can be seen from the figure that the original signals are mixed together in the feature space, and then, with the processing of each network layer of AnNET, the output features of each layer start to be similar to each other and separated from each other. This in turn means that the proposed network is able to extract useful features from the raw signal, being able to distinguish between different types of bearing faults under high noise and varying load conditions.
Although illustrative embodiments of the present invention have been described above to facilitate the understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, and various changes may be made apparent to those skilled in the art as long as they are within the spirit and scope of the present invention as defined and defined by the appended claims, and all matters of the invention which utilize the inventive concepts are protected.
Claims (10)
1. A fault diagnosis method for a rolling bearing under a strong noise environment is characterized by comprising the following steps:
s1, at sampling frequency fsCollecting vibration acceleration signals x of bearings with no fault, different fault types and different fault degrees under different loadsi,j；
S2: for the collected vibration acceleration signal xi,jCarrying out normalization processing to obtain normalized vibration acceleration signal
Setting a deep learning training label for the normalized vibration acceleration signal according to the fault type and the fault degree;
s3: constructing a neural network model, comprising: the device comprises a random sampling layer, a CNN global type feature extractor combined with SENet, a CNN-based high-dimensional feature extractor, a GRU-based feature classifier, a full connection layer and a Softmax layer; the normalized sample is sequentially processed by a random sampling layer, a CNN global type feature extractor combined with SENet, a CNN-based high-dimensional feature extractor, a GRU-based feature classifier and a full connection layer, and finally output by a Softmax layer to obtain a result label; the random sampling layer is used for adding training data interfered by analog noise; the CNN global type feature extractor combined with SENEt is used for extracting global type fault features; the CNN-based high-dimensional feature extractor is used for abstracting high-dimensional fault features from the extracted global fault features; the GRU-based feature classifier is used for clustering features extracted by the CNN-based high-dimensional feature extractor;
s4: the normalized vibration acceleration signal processed in the step S2 is
Inputting the corresponding training labels into the neural network model in the step S3 for training;
s5: using the same sampling frequency f as in step S1sAcquiring vibration data of a rolling bearing of the current equipment to obtain a vibration acceleration signal X to be measuredi,j；
S6: according to the step S2, the vibration acceleration signal X to be measuredi,jNormalization processing is carried out to obtain normalized test data
Test data after normalization
Inputting the current bearing fault state into the trained model in step S4.
2. The rolling bearing fault diagnosis method according to claim 1, wherein the random sampling layer in step S3 is random sampling with a bernoulli distribution subject to a probability p, wherein p takes a random value between 0.5 and 1 in each training period, and the non-sampled data is set to 0 in order to maintain the data length.
3. The rolling bearing fault diagnosis method according to claim 1, wherein the CNN global type feature extractor in combination with SENet in step S3 is composed of one CNN feature extractor and one SENet; the CNN extractor comprises a convolution layer, an AdaBN layer and an ELU active layer in sequence, wherein the AdaBN layer can carry out standardization operation on convolution layer results, and the standardization formula is as follows:
in the formula: x is the number of(k)As AdaBN layerInput of gamma(k)、β(k)Scaling and biasing parameters for AdaBN layer, y(k)Is the output of AdaBN layer, where gamma(k)、β(k)In the training mode is a parameter of the autonomous learning training, and in the test mode the test data is updated with the parameter
The distribution of (a); the SENEt is composed of a global average pooling layer, a full connection layer, an ELU activation layer, a full connection layer and a Sigmoid activation layer in sequence; the input of SEnet is the output of the CNN feature extractor, and the output of the CNN global-type feature extractor combined with SEnet is the product of SEnet and the output of the CNN feature extractor.
4. The rolling bearing fault diagnosis method according to claim 1, wherein the CNN-based high-dimensional feature extractor in step S3 is composed of a plurality of CNN feature extractors, each CNN feature extractor being composed of a convolution layer, an AdaBN layer, an ELU activation layer, and a maximum pooling layer in this order.
5. The rolling bearing fault diagnosis method according to claim 1, wherein the GRU-based feature classifier is stacked of a plurality of GRU layers in step S3, and the last time step of the last GRU layer is taken as an output of the GRU-based feature classifier.
6. A rolling bearing fault diagnosis system under a strong noise environment is characterized by comprising:
signal acquisition module for sampling frequency fsCollecting vibration acceleration signals x of bearings with no fault, different fault types and different fault degrees under different loadsi,j；
A normalization processing module for processing the collected vibration acceleration signal xi,jCarrying out normalization processing to obtain normalized vibration acceleration signal
The training label setting module is used for setting a deep learning training label for the normalized vibration acceleration signal according to the fault type and the fault degree;
a neural network training module comprising: the device comprises a random sampling layer, a CNN global type feature extractor combined with SENet, a CNN-based high-dimensional feature extractor, a GRU-based feature classifier, a full connection layer and a Softmax layer; the normalized sample is sequentially processed by a random sampling layer, a CNN global type feature extractor combined with SENet, a CNN-based high-dimensional feature extractor, a GRU-based feature classifier and a full connection layer, and finally output by a Softmax layer to obtain a result label; the random sampling layer is used for adding training data interfered by analog noise; the CNN global type feature extractor combined with SENEt is used for extracting global type fault features; the CNN-based high-dimensional feature extractor is used for abstracting high-dimensional fault features from the extracted global fault features; the GRU-based feature classifier is used to cluster features extracted by the CNN-based high-dimensional feature extractor.
7. The rolling bearing fault diagnosis system under strong noise environment according to claim 6, wherein the random sampling layer is a random sampling with Bernoulli distribution obeying a probability p, wherein p takes a random value between 0.5 and 1 in each training period, and the data not sampled is set to 0 in place for maintaining the data length.
8. The rolling bearing fault diagnosis system under strong noise environment according to claim 6, wherein the CNN global type feature extractor combined with SENet is composed of one CNN feature extractor and one SENet; the CNN extractor comprises a convolution layer, an AdaBN layer and an ELU active layer in sequence, wherein the AdaBN layer can carry out standardization operation on convolution layer results, and the standardization formula is as follows:
in the formula: x is the number of(k)As input to the AdaBN layer, γ(k)、β(k)Scaling and bias parameters, y, for AdaBN layer(k)Is the output of AdaBN layer, where gamma(k)、β(k)In the training mode is a parameter of the autonomous learning training, and in the test mode the test data is updated with the parameter
The distribution of (a); the SENEt is composed of a global average pooling layer, a full connection layer, an ELU activation layer, a full connection layer and a Sigmoid activation layer in sequence; the input of SEnet is the output of the CNN feature extractor, and the output of the CNN global-type feature extractor combined with SEnet is the product of the SEnet and the output of the CNN feature extractor.
9. The rolling bearing fault diagnosis system under strong noise environment of claim 6, wherein the CNN-based high-dimensional feature extractor is composed of a plurality of CNN feature extractors, each CNN feature extractor is composed of a convolution layer, an AdaBN layer, an ELU activation layer and a maximum pooling layer in turn.
10. The rolling bearing fault diagnosis system under strong noise environment according to claim 6, wherein the GRU based feature classifier is formed by stacking a plurality of GRU layers, and the last time step of the last GRU layer is taken as an output of the GRU based feature classifier.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111672334.XA CN114459760A (en) | 2021-12-31 | 2021-12-31 | Rolling bearing fault diagnosis method and system under strong noise environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111672334.XA CN114459760A (en) | 2021-12-31 | 2021-12-31 | Rolling bearing fault diagnosis method and system under strong noise environment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114459760A true CN114459760A (en) | 2022-05-10 |
Family
ID=81407593
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111672334.XA Pending CN114459760A (en) | 2021-12-31 | 2021-12-31 | Rolling bearing fault diagnosis method and system under strong noise environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114459760A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2622124A (en) * | 2022-09-01 | 2024-03-06 | Univ Chongqing | Intelligent mechanical fault diagnosis method based on relationship transfer domain generalization network (RTDGN) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109029999A (en) * | 2018-09-19 | 2018-12-18 | 河北工业大学 | Fault Diagnosis of Roller Bearings based on enhanced modulation double-spectrum analysis |
| CN110595775A (en) * | 2019-09-19 | 2019-12-20 | 电子科技大学 | Rolling bearing fault diagnosis method based on multi-branch multi-scale convolutional neural network |
| CN110647867A (en) * | 2019-10-09 | 2020-01-03 | 中国科学技术大学 | Bearing fault diagnosis method and system based on self-adaptive anti-noise neural network |
| CN110896381A (en) * | 2019-11-25 | 2020-03-20 | 中国科学院深圳先进技术研究院 | Deep neural network-based traffic classification method and system and electronic equipment |
| CN112001314A (en) * | 2020-08-25 | 2020-11-27 | 江苏师范大学 | Early fault detection method for variable speed hoist |
| KR102189269B1 (en) * | 2019-10-22 | 2020-12-09 | 경북대학교 산학협력단 | Fault Diagnosis method and system for induction motor using convolutional neural network |
| CN112284735A (en) * | 2020-10-21 | 2021-01-29 | 兰州理工大学 | Multi-sensor rolling bearing fault diagnosis based on one-dimensional convolution and dynamic routing |
| CN112577748A (en) * | 2020-12-07 | 2021-03-30 | 东南大学 | Reinforced lightweight multi-scale CNN-based rolling bearing fault diagnosis method |
| CN113776834A (en) * | 2021-10-11 | 2021-12-10 | 山东大学 | Rolling bearing fault diagnosis method and system based on discrete cosine cycle spectrum coherence |
-
2021
- 2021-12-31 CN CN202111672334.XA patent/CN114459760A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109029999A (en) * | 2018-09-19 | 2018-12-18 | 河北工业大学 | Fault Diagnosis of Roller Bearings based on enhanced modulation double-spectrum analysis |
| CN110595775A (en) * | 2019-09-19 | 2019-12-20 | 电子科技大学 | Rolling bearing fault diagnosis method based on multi-branch multi-scale convolutional neural network |
| CN110647867A (en) * | 2019-10-09 | 2020-01-03 | 中国科学技术大学 | Bearing fault diagnosis method and system based on self-adaptive anti-noise neural network |
| KR102189269B1 (en) * | 2019-10-22 | 2020-12-09 | 경북대학교 산학협력단 | Fault Diagnosis method and system for induction motor using convolutional neural network |
| CN110896381A (en) * | 2019-11-25 | 2020-03-20 | 中国科学院深圳先进技术研究院 | Deep neural network-based traffic classification method and system and electronic equipment |
| CN112001314A (en) * | 2020-08-25 | 2020-11-27 | 江苏师范大学 | Early fault detection method for variable speed hoist |
| CN112284735A (en) * | 2020-10-21 | 2021-01-29 | 兰州理工大学 | Multi-sensor rolling bearing fault diagnosis based on one-dimensional convolution and dynamic routing |
| CN112577748A (en) * | 2020-12-07 | 2021-03-30 | 东南大学 | Reinforced lightweight multi-scale CNN-based rolling bearing fault diagnosis method |
| CN113776834A (en) * | 2021-10-11 | 2021-12-10 | 山东大学 | Rolling bearing fault diagnosis method and system based on discrete cosine cycle spectrum coherence |
Non-Patent Citations (2)
| Title |
|---|
| 陈晓雷等: "基于卷积神经网络和双向长短期记忆的稳定抗 噪声滚动轴承故障诊断", 吉林大学学报（ 工学版）, vol. 52, no. 2, 28 February 2022 (2022-02-28), pages 296 - 309 * |
| 韩东颖等: "基于自适应变尺度频移带通随机共振降噪的EMD 多频微弱信号检测", 机械工程学报, vol. 49, no. 8, 30 April 2013 (2013-04-30), pages 10 - 18 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2622124A (en) * | 2022-09-01 | 2024-03-06 | Univ Chongqing | Intelligent mechanical fault diagnosis method based on relationship transfer domain generalization network (RTDGN) |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110516305B (en) | Intelligent fault diagnosis method under small sample based on attention mechanism meta-learning model | |
| CN112257530B (en) | Rolling bearing fault diagnosis method based on blind signal separation and support vector machine | |
| CN113008559B (en) | Bearing fault diagnosis method and system based on sparse self-encoder and Softmax | |
| CN111506862A (en) | Rolling bearing fault diagnosis method based on multi-source weighted integration transfer learning | |
| CN114548190B (en) | Wind turbine fault diagnosis method and device based on self-adaptive residual error neural network | |
| CN116304861A (en) | Time-frequency characteristic fusion fault diagnosis method based on self-attention | |
| CN114169377A (en) | G-MSCNN-based fault diagnosis method for rolling bearing in noisy environment | |
| CN114048688A (en) | Method for predicting service life of bearing of wind power generator | |
| CN112651426A (en) | Fault diagnosis method for rolling bearing of wind turbine generator | |
| CN114441173B (en) | Rolling bearing fault diagnosis method based on improved depth residual error shrinkage network | |
| CN116340859A (en) | Marine wind turbine generator gearbox fault diagnosis method based on vibration signals under noise background | |
| CN114459760A (en) | Rolling bearing fault diagnosis method and system under strong noise environment | |
| Wang et al. | Fault diagnosis of industrial robots based on multi-sensor information fusion and 1D convolutional neural network | |
| Liu et al. | Rolling bearing fault diagnosis of train running gear based on optimized deep residual network | |
| Peng et al. | Research on fault diagnosis method of rolling bearing based on 2DCNN | |
| Wang et al. | Rolling bearing fault diagnosis method based on stacked denoising autoencoder and convolutional neural network | |
| CN116735206A (en) | Bearing friction state identification method based on attractor and convolutional neural network | |
| Lu et al. | An automated bearing fault diagnosis using a self-normalizing convolutional neural network | |
| CN116380466A (en) | Rolling bearing intelligent fault diagnosis method and system based on enhanced event visual data | |
| CN115587289A (en) | Multi-source domain adaptive network-based aeroengine fault identification method | |
| He et al. | Fault Identification of Rotating Machinery Based on Dynamic Feature Reconstruction Signal Graph | |
| Xing et al. | Intelligent diagnosis of rolling bearings with imbalanced data based on the ConVAE-CNN model | |
| CN116625689B (en) | Rolling bearing fault diagnosis method and system based on SMDER | |
| Wang et al. | Bearing Remaining Life Prediction Method Based on Wavelet Transform Denoising and Feature Preservation CycleGAN under Unbalanced Samples | |
| Zhang et al. | Fault Diagnosis of Rolling Bearing Based on CNN with Attention Mechanism and Dynamic Learning Rate |
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 |