CN114964782A - Rolling bearing fault diagnosis acoustic emission signal processing method and system based on compressed sensing - Google Patents
Rolling bearing fault diagnosis acoustic emission signal processing method and system based on compressed sensing Download PDFInfo
- Publication number
- CN114964782A CN114964782A CN202210624919.2A CN202210624919A CN114964782A CN 114964782 A CN114964782 A CN 114964782A CN 202210624919 A CN202210624919 A CN 202210624919A CN 114964782 A CN114964782 A CN 114964782A
- Authority
- CN
- China
- Prior art keywords
- layer
- acoustic emission
- compression
- emission signal
- rolling 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.)
- Pending
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 39
- 238000005096 rolling process Methods 0.000 title claims abstract description 38
- 238000003672 processing method Methods 0.000 title claims abstract description 19
- 238000007906 compression Methods 0.000 claims abstract description 95
- 230000006835 compression Effects 0.000 claims abstract description 93
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 41
- 238000012549 training Methods 0.000 claims abstract description 36
- 239000011159 matrix material Substances 0.000 claims abstract description 24
- 238000000034 method Methods 0.000 claims abstract description 18
- 238000012360 testing method Methods 0.000 claims abstract description 14
- 230000009467 reduction Effects 0.000 claims abstract description 9
- 239000010410 layer Substances 0.000 claims description 118
- 238000011176 pooling Methods 0.000 claims description 28
- 230000004913 activation Effects 0.000 claims description 19
- 239000002356 single layer Substances 0.000 claims description 16
- 238000012545 processing Methods 0.000 claims description 10
- 238000003860 storage Methods 0.000 claims description 10
- 210000002569 neuron Anatomy 0.000 claims description 8
- 238000000605 extraction Methods 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 19
- 238000001994 activation Methods 0.000 description 16
- 238000005070 sampling Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 230000000875 corresponding effect Effects 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 238000013144 data compression Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000003754 machining Methods 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010892 electric spark Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011160 research Methods 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
- G06N3/04—Architecture, e.g. interconnection topology
-
- 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
Abstract
The invention discloses a rolling bearing fault diagnosis acoustic emission signal processing method and system based on compressed sensing, wherein the method comprises the following steps: collecting original acoustic emission signals of different rolling bearing fault states; calculating the reconstruction error of the original acoustic emission signal in the compression ratio setting range by using an orthogonal matching pursuit algorithm, and selecting the compression ratio for participating in compression from the compression ratio setting range according to the reconstruction error; constructing a Gaussian random matrix as a projection matrix to perform dimension reduction projection on the original acoustic emission signal to obtain a compressed signal; dividing the obtained compressed signal adding label into a training sample set and a testing sample set; establishing a convolutional neural network model, inputting a training sample set into the convolutional neural network model to continuously carry out iterative training until the training is finished, and obtaining a trained convolutional neural network model; and inputting the test sample set into the trained convolutional neural network model to diagnose the fault of the rolling bearing. The invention realizes the extraction and accurate classification of the fault characteristics of the compressed data.
Description
Technical Field
The invention relates to a rolling bearing fault diagnosis acoustic emission signal processing method and system based on compressed sensing, and belongs to the field of acoustic emission signal processing.
Background
The rolling bearing is a common and important component part in mechanical equipment, so that the failure of the rolling bearing has an extremely important influence on the operation of the equipment and even the whole system, causes great economic loss and possibly causes casualties if serious. Therefore, in order to avoid the occurrence of the fault as much as possible and ensure the normal operation of the equipment, the early fault of the rolling bearing needs to be intelligently diagnosed and predicted. The vibration signal is difficult to capture the weak signal of early failure, and the acoustic emission signal is more sensitive to early failure and is less influenced by mechanical background noise.
The signal acquisition mode based on the traditional nyquist (Nyqusit) sampling theorem requires that the sampling frequency is at least 2 times of the highest frequency of the signal. The frequency of the acoustic emission signals can reach more than a few kilohertz, and the sampling frequency of a multi-channel data high-speed acquisition system adopted in the practical situation is 5-10 times of the highest frequency of the signals, so that huge data volume is necessarily generated, the hardware implementation cost is high, and huge pressure is caused for the transmission, storage and processing of subsequent data, which is unfavorable for the long-term monitoring of the signals.
In order to deal with the ultra-large amount of data caused by the high sampling frequency and the high sampling point number of the acoustic emission signal, a compressed sensing processing frame is introduced to carry out data compression, feature extraction and diagnosis evaluation research, however, the traditional fault diagnosis method has limitations, the diagnosis performance of the traditional fault diagnosis method depends on expert experience and priori knowledge to a great extent, and the network model is usually a shallow structure and is difficult to dig out the features of the compressed data, so that the improvement of the classification accuracy of the compressed data is still to be solved.
Disclosure of Invention
The invention provides a rolling bearing fault diagnosis acoustic emission signal processing method and system based on compressed sensing, which are used for rolling bearing fault diagnosis acoustic emission signal processing by utilizing a mode of common cooperation of compressed sensing and a convolutional neural network.
The technical scheme of the invention is as follows: a rolling bearing fault diagnosis acoustic emission signal processing method based on compressed sensing comprises the following steps:
collecting original acoustic emission signals of different rolling bearing fault states;
calculating the reconstruction error of the original acoustic emission signal in the compression ratio setting range by using an orthogonal matching pursuit algorithm, and selecting the compression ratio for participating in compression from the compression ratio setting range according to the reconstruction error;
according to the dimension M of the original acoustic emission signal and the dimension N of a compressed signal determined by selecting a compression rate participating in compression, constructing an MXN Gaussian random matrix as a projection matrix to perform dimension reduction projection on the original acoustic emission signal to obtain a compressed signal;
dividing the obtained compressed signal adding label into a training sample set and a testing sample set;
and establishing a convolutional neural network model, inputting the training sample set into the convolutional neural network model to continuously carry out iterative training until the training is finished, and obtaining the trained convolutional neural network model.
The compression ratio is set in the range of 1-20.
The selecting, according to the reconstruction error, a compression ratio for participating in compression from a compression ratio setting range specifically includes: and selecting the compression rate with the maximum compression rate as the compression rate participating in compression when the reconstruction error is less than or equal to 25 percent according to the reconstruction errors under different compression rates in the compression rate setting range.
The convolutional neural network model is constructed by 4 convolutional layers, 3 maximum pooling layers, 1 full-link layer and 1 softmax layer.
The convolutional neural network model is divided into 9 layers, specifically:
layer 5 is a convolutional layer: the size of convolution kernels is 1 multiplied by 55, the number of the convolution kernels is 64, the edges are filled, the filling mode is zero filling, the step length is 1, the output is 51 multiplied by 64, each convolution kernel uses a ReLU activation function, and a single-layer convolution network is realized;
layer 7 is a convolutional layer: the convolution kernel size is 1 x 55, the number of convolution kernels is 64, the edge is filled, the filling mode is zero filling, the step length is 1, the output is 12 x 64, each convolution kernel uses a ReLU activation function, and a single-layer convolution network is realized;
layer 9 is softmax layer: the output of the softmax layer is the value number of the label, which is 3 multiplied by 1.
The fault diagnosis result comprises three fault types, namely normal, roller fault and race fault.
According to another aspect of the embodiment of the invention, a rolling bearing fault diagnosis acoustic emission signal processing system based on compressed sensing is provided, and comprises:
the acquisition module is used for acquiring original acoustic emission signals of different rolling bearing fault states;
the selection module is used for calculating the reconstruction error of the original acoustic emission signal in the compression ratio setting range by using an orthogonal matching pursuit algorithm, and selecting the compression ratio for participating in compression from the compression ratio setting range according to the reconstruction error;
the first obtaining module is used for constructing an MxN Gaussian random matrix as a projection matrix to perform dimension reduction projection on the original acoustic emission signal according to the dimension M of the original acoustic emission signal and the dimension N of a compressed signal determined by selecting a compression ratio participating in compression, and obtaining the compressed signal;
the dividing module is used for dividing the obtained compressed signal adding label into a training sample set and a testing sample set;
and the second obtaining module is used for establishing a convolutional neural network model, inputting the training sample set into the convolutional neural network model for continuous iterative training until the training is finished, and obtaining the trained convolutional neural network model.
According to another aspect of the embodiments of the present invention, there is provided a processor for executing a program, wherein the program is executed to execute the rolling bearing fault diagnosis acoustic emission signal processing method based on compressive sensing.
According to another aspect of the embodiments of the present invention, a computer-readable storage medium is provided, where the computer-readable storage medium includes a stored program, where the program is executed to control an apparatus where the computer-readable storage medium is located to execute any one of the above rolling bearing fault diagnosis acoustic emission signal processing methods based on compressive sensing.
The invention has the beneficial effects that:
1. the invention adopts the compressed sensing technology to compress the acoustic emission signal sample, is not limited by the Nyquist sampling theorem any more, reduces the data acquisition cost, and can simultaneously reserve useful information from the original sample signal in the data preprocessing process, thereby improving the data analysis efficiency.
2. The invention utilizes the property that the signal energy of the Gaussian random projection matrix is approximately kept unchanged in the random projection process to compress data, selects the most appropriate compression multiple according to the relation between the compression error and the compression ratio, compresses the data to the greatest extent possible on the basis of keeping the original information, and reduces the pressure for the transmission, storage and analysis of the subsequent data.
3. In the process of fault diagnosis through compressed data, the convolutional neural network in deep learning is introduced, deep feature mining is performed through multilayer convolution, features of different layers are spliced into a whole through a full connection layer, and a diagnosis result is output through a softmax layer, so that the fault feature extraction and accurate classification of the compressed data are realized.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a diagram illustrating compression of signal samples according to the compressive sensing theory of the present invention;
FIG. 3 is a diagram illustrating the relationship between reconstruction error and compression ratio provided by the present invention;
FIG. 4 is a diagram illustrating an example of the reconstruction obtained by the OMP reconstruction algorithm according to the present invention;
FIG. 5 is an enlarged view of a portion of FIG. 4;
FIG. 6 is an error of an original signal and a reconstructed signal;
FIG. 7 is a schematic diagram of a Convolutional Neural Network (CNN) training model structure according to the present invention;
FIG. 8 is a diagram illustrating the classification result according to the present invention.
Detailed Description
The invention will be further described with reference to the following figures and examples, but the scope of the invention is not limited thereto.
Example 1: as shown in fig. 1 to 8, a rolling bearing fault diagnosis acoustic emission signal processing method based on compressed sensing includes:
collecting original acoustic emission signals of different rolling bearing fault states;
calculating a reconstruction error of the original acoustic emission signal in a compression rate setting range by using an Orthogonal Matching Pursuit (OMP) algorithm, and selecting a compression rate for participating in compression from the compression rate setting range according to the reconstruction error;
according to the dimension M of the original acoustic emission signal and the dimension N of a compressed signal determined by selecting a compression rate participating in compression, an MXN Gaussian random matrix is constructed to be used as a projection matrix to perform dimension reduction projection on the original acoustic emission signal, the signal obtained after projection is a compressed signal, and the length of the obtained compressed signal is far smaller than that of the original acoustic emission signal;
dividing the obtained compressed signal adding label into a training sample set and a testing sample set;
and establishing a convolutional neural network model, inputting the training sample set into the convolutional neural network model to continuously carry out iterative training until the training is finished, and obtaining the trained convolutional neural network model.
Alternatively, the compression ratio is set in the range of 1 to 20.
Optionally, the selecting, according to the reconstruction error, a compression ratio for participating in compression from a compression ratio setting range specifically includes: and selecting the compression rate with the maximum compression rate as the compression rate participating in compression when the reconstruction error is less than or equal to 25 percent according to the reconstruction errors under different compression rates in the compression rate setting range.
Optionally, the convolutional neural network model is constructed of 4 convolutional layers, 3 max pooling layers, 1 fully-connected layer, and 1 softmax layer.
Optionally, the convolutional neural network model is divided into 9 layers, specifically:
layer 5 is a convolutional layer: the size of convolution kernels is 1 multiplied by 55, the number of the convolution kernels is 64, the edges are filled, the filling mode is zero filling, the step length is 1, the output is 51 multiplied by 64, each convolution kernel uses a ReLU activation function, and a single-layer convolution network is realized;
layer 7 is a convolutional layer: the size of convolution kernels is 1 multiplied by 55, the number of the convolution kernels is 64, the edges are filled, the filling mode is zero filling, the step length is 1, the output is 12 multiplied by 64, each convolution kernel uses a ReLU activation function, and a single-layer convolution network is realized;
layer 9 is softmax layer: the output of the softmax layer is the value number of the label, which is 3 multiplied by 1.
Optionally, the fault diagnosis result includes three fault types, normal, roller fault, race fault.
Example 2: an alternative embodiment of the invention is described in detail below, as shown in fig. 1-8.
S1, acquiring original acoustic emission signals of different rolling bearing fault states by using an acoustic emission sensor;
s2, calculating the reconstruction error of the original acoustic emission signal in the compression ratio setting range by using an orthogonal matching pursuit algorithm, and selecting the compression ratio for participating in compression from the compression ratio setting range according to the reconstruction error;
s3, constructing an MXN Gaussian random matrix as a projection matrix to perform dimension reduction projection on the original acoustic emission signal according to the dimension M of the original acoustic emission signal and the dimension N of the compressed signal determined by selecting the compression rate participating in the compression, wherein the signal obtained after the projection is a compressed signal, and the length of the obtained compressed signal is far less than that of the original acoustic emission signal;
s4, dividing the obtained compressed signal adding labels into a training sample set and a test sample set according to the ratio of 8: 2;
s5, establishing a convolutional neural network model, inputting the training sample set into the convolutional neural network model to continuously carry out iterative training until the training is finished, and obtaining the trained convolutional neural network model;
and S6, inputting the test sample set or the compressed data set to be tested into the trained convolutional neural network model to diagnose the fault of the rolling bearing.
In step S3, the original acoustic emission signal is sparsely represented by a DCT (discrete cosine transform) method, and the signal is reconstructed by an OMP method.
In an embodiment of the present invention, the acoustic emission signal needs to be compressed, and fig. 2 is a schematic diagram of compressing the signal sampling by the compressed sensing provided by the present invention.
As shown in FIG. 2, the original acoustic emission sample signal comes from the acoustic emission sensor acquisition device to obtain the acoustic emission signal
Presence of orthogonal base Ψ ∈ R N×N The base psi is composed of a set of orthogonal base functions psi ═ psi 1 ,ψ 2 ,···,ψ N ]The transform coefficients θ in the formula are sparse, i.e., there are only a few non-zero terms, and most of the rest are zero values, and N represents the dimension of the original signal. Original acoustic emission signal x, compressed signal y ═ Φ x, where Φ M×N A gaussian random projection matrix representing the compressed sensing. In compressed sensing, a signal x is projected onto an M × N dimensional gaussian random matrix Φ that is not correlated with a transform matrix, and the dimension of the obtained measurement value y is M × 1, the sample dimension is reduced.
As shown in fig. 3, the compression ratio R of the signal is M/N, where M represents the dimension of the signal before compression and N represents the dimension of the signal after compression. Defining a reconstruction ERROR ERROR | | | | x-y | |/| | x | |, and calculating the variation trend of the reconstruction ERROR under different compression rates. The compression ratio ranges from 1 to 20, and reconstruction error curves under different compression ratios are obtained, as shown in fig. 3, the reconstruction error increases with the increase of the compression ratio. According to practical experience, when the reconstruction error is about 25%, the practical requirement can be satisfied, and thus a proper compression rate is selected.
As shown in fig. 4-6, the data is compressed at a suitable compression rate and then reconstructed, and the reconstruction of the signal at the compression rate is checked, fig. 4 is an example of reconstruction based on an OMP algorithm, fig. 5 is a partial enlarged view of a reconstructed signal curve, fig. 6 is a partial enlarged view of an original signal and a reconstructed signal, and the errors of the original signal and the reconstructed signal are analyzed, and the total error is kept within a range of less than 0.7, thereby proving that the original acoustic emission data can be recovered by signal reconstruction. And under the condition that the observation projection matrix phi, the compressed data y and the sparsity theta are known, reconstructing a signal by utilizing the OMP (object-oriented programming) to obtain an approximate value of the signal x.
According to the principle of compressed sensing, in order to realize data compression, firstly, selecting a projection matrix meeting approximate equidistant projection properties in the compression process is a key, the approximate equidistant projection properties ensure that the structure of signals before and after compressed projection is approximately unchanged, then, carrying out dimension reduction projection on the acquired original acoustic emission signals by using the projection matrix, and obtaining compressed data after projection; meanwhile, the method can participate in deep learning more efficiently, and accurate classification is realized.
The more the number of layers of the convolutional neural network is, the larger the calculation amount is, and the problems such as overfitting are caused, so that the training effect is not ideal, and the network with too few layers cannot be well learned. Therefore, experiments verify that the convolutional neural network established in the application is divided into 9 layers, as shown in fig. 7, the convolutional layers and the maximum pooling function to perform feature extraction on different layers; and connecting the obtained features of different layers into a whole to obtain a connecting layer, fusing the features through the full connecting layer, and finally realizing fault diagnosis through the softmax classification layer. The training compressed data samples generated in step S4 are input at the first layer, the n +1 th layer uses the output of the nth layer as input, and the softmax layer outputs the fault classification.
layer 5 is a convolutional layer: the size of convolution kernels is 1 multiplied by 55, the number of the convolution kernels is 64, the edges are filled, the filling mode is zero filling, the step length is 1, the output is 51 multiplied by 64, each convolution kernel uses a ReLU activation function, and a single-layer convolution network is realized;
layer 7 is a convolutional layer: the size of convolution kernels is 1 multiplied by 55, the number of the convolution kernels is 64, the edges are filled, the filling mode is zero filling, the step length is 1, the output is 12 multiplied by 64, each convolution kernel uses a ReLU activation function, and a single-layer convolution network is realized;
layer 9 is softmax layer: the output of the softmax layer is the value number of the label, which is 3 multiplied by 1.
The loss function of the convolutional neural network is a cross entropy cost function.
At this point, the compressed signal of the test sample set is used as model input, the probability of occupation of each possible fault reason can be calculated, and the highest probability is taken as a prediction label.
In the standard one-dimensional convolution layer, each randomly initialized kernel is convolved with the input width and height, and the process of convolution operation can be expressed as:
in the formula:
weight of the kth convolution kernel represented as the l-th layer, x (t) represents the outputThe input signal is sent to the signal processing device,
represents the deviation, is the convolution operator. After the convolution operation, an activation function is needed, and the activation process can be expressed as:
f (x) is the activation function of the first convolutional layer,
representing the output of the activated feature map.
The process of performing feature fusion by full connection can be expressed as follows:
wherein k ∈ (1, n) is totally n neurons, l represents the number of layers,
representing the weight of the kth neuron at layer l,
the threshold for the kth neuron represented as layer l,
an output representing a kth neuron of the l layer; f (x) is an activation function.
As shown in fig. 8, the sample to be tested is sent to the trained model, and the corresponding fault diagnosis result is obtained.
And (3) verifying the validity:
the rotating speed of a main shaft of the test bed is 400rpm, the sampling frequency is 1MHz, single-point pit machining is carried out on the central positions of a bearing race and a rolling element by adopting an electric spark machining technology, the sizes of machined faults are respectively 0.5mm roller faults and 0.5mm race fault depth is 0.65mm, 2000 samples are collected under each bearing state, and each data sample comprises 8192 data points.
TABLE 1
| Bearing condition | Diameter of failure | Label (R) | |
| Is | 0mm | 0 | |
| Roller failure | 0.5 |
1 | |
| Race failure | 0.5 |
2 |
The specific steps are shown in figure 1 and are described as follows:
(1) data acquisition and compression
And preprocessing the acquired original sensor data by adopting compressed sensing. From the 2000 data samples collected for each bearing state type, 1600 data samples were randomly drawn as training samples and an additional 400 data samples were drawn as test samples. Since each data sample contains 8192 data points, when the compression ratio R is selected to be 10, the data length of the compressed signal y is 819. The signal compression process using compressed sensing is shown in fig. 2.
(2) The convolutional neural network parameter settings are shown in table 2:
TABLE 2
| Parameter(s) | Parameter value |
| Epoch (iteration number) | 100 |
| Batchsize (batch size) | 100 |
| Learningate (learning rate) | 0.001 |
| Deacy (attenuation rate) | 0.9 |
(3) Model fault diagnosis classification result
And substituting the test sample into the diagnostic model trained in the step S5 to obtain a corresponding diagnostic result, wherein the fault diagnostic result is a confusion matrix diagram shown in fig. 8. Fig. 8 shows that the classification accuracy of the rolling bearing fault diagnosis acoustic emission signal processing method based on compressed sensing is 98% for the first bearing state type, 99% for the second bearing state type and 100% for the second bearing state type, which indicates that the classification accuracy for each fault type can reach more than 95%, and the method has good diagnosis performance.
The invention combines the compressed sensing theory and the convolutional neural network characteristics in deep learning, realizes that effective characteristics are still mined for fault diagnosis while the data volume is reduced, effectively solves the problems that large data volume is obtained and data transmission and data analysis are difficult to perform due to the fact that acoustic emission signals are collected at the early stage of fault diagnosis, and can also be used for mining the compressed data characteristics by using a deep learning method. The method provided by the invention is verified by taking the rolling bearing as an example, can be popularized to various aspects such as rotating machinery, processing and manufacturing, equipment maintenance and the like in the practical application process, and has good engineering practicability.
Example 2: a rolling bearing fault diagnosis acoustic emission signal processing system based on compressed sensing comprises:
the acquisition module is used for acquiring original acoustic emission signals of different rolling bearing fault states;
the selection module is used for calculating the reconstruction error of the original acoustic emission signal in the compression ratio setting range by using an orthogonal matching pursuit algorithm, and selecting the compression ratio for participating in compression from the compression ratio setting range according to the reconstruction error;
the first obtaining module is used for constructing an MXN Gaussian random matrix as a projection matrix to perform dimension reduction projection on the original acoustic emission signal according to the dimension M of the original acoustic emission signal and the dimension N of the compressed signal determined by the compression rate selected to participate in compression, so as to obtain the compressed signal;
the dividing module is used for dividing the obtained compressed signal adding label into a training sample set and a testing sample set;
and the second obtaining module is used for establishing a convolutional neural network model, inputting the training sample set into the convolutional neural network model for continuous iterative training until the training is finished, and obtaining the trained convolutional neural network model.
Example 3: a processor for executing a program, wherein the program executes the method for processing the acoustic emission signal based on rolling bearing fault diagnosis based on compressive sensing.
Example 4: a computer-readable storage medium, which includes a stored program, and when the program runs, the apparatus where the computer-readable storage medium is located is controlled to execute any one of the above rolling bearing fault diagnosis acoustic emission signal processing methods based on compressive sensing.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
While the present invention has been described in detail with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.
Claims (9)
1. A rolling bearing fault diagnosis acoustic emission signal processing method based on compressed sensing is characterized in that: the method comprises the following steps:
collecting original acoustic emission signals of different rolling bearing fault states;
calculating the reconstruction error of the original acoustic emission signal in the compression ratio setting range by using an orthogonal matching pursuit algorithm, and selecting the compression ratio for participating in compression from the compression ratio setting range according to the reconstruction error;
according to the dimension M of the original acoustic emission signal and the dimension N of a compressed signal determined by selecting a compression rate participating in compression, constructing an MXN Gaussian random matrix as a projection matrix to perform dimension reduction projection on the original acoustic emission signal to obtain a compressed signal;
dividing the obtained compressed signal adding label into a training sample set and a testing sample set;
and establishing a convolutional neural network model, inputting the training sample set into the convolutional neural network model to continuously carry out iterative training until the training is finished, and obtaining the trained convolutional neural network model.
2. The rolling bearing fault diagnosis acoustic emission signal processing method based on compressed sensing according to claim 1, characterized in that: the compression ratio is set in the range of 1-20.
3. The rolling bearing fault diagnosis acoustic emission signal processing method based on compressed sensing according to claim 1, characterized in that: the selecting, according to the reconstruction error, a compression ratio for participating in compression from a compression ratio setting range specifically includes: and selecting the compression rate with the maximum compression rate as the compression rate participating in compression when the reconstruction error is less than or equal to 25 percent according to the reconstruction errors under different compression rates in the compression rate setting range.
4. The rolling bearing fault diagnosis acoustic emission signal processing method based on compressed sensing according to claim 1, characterized in that: the convolutional neural network model is constructed by 4 convolutional layers, 3 maximum pooling layers, 1 full-link layer and 1 softmax layer.
5. The rolling bearing fault diagnosis acoustic emission signal processing method based on compressed sensing according to claim 4, characterized in that: the convolutional neural network model is divided into 9 layers, specifically:
layer 1 is a convolutional layer: the input is a compressed signal, the convolution kernel adopts a 1 × 55 structure, the number of the convolution kernels is 81, the step length is 1, the edge filling is performed, the filling mode is zero filling, the output is 819 × 81, and each convolution kernel uses a ReLU activation function to realize a single-layer convolution network;
layer 2 is the maximum pooling layer: the height and width of the pooling window are 4, the step length is 1, and the output is 204 multiplied by 81;
layer 3 is a convolutional layer: the size of convolution kernels is 1 multiplied by 55, the number of the convolution kernels is 81, the edges are filled, the filling mode is zero filling, the step length is 1, the output is 204 multiplied by 81, each convolution kernel uses a ReLU activation function, and a single-layer convolution network is realized;
layer 4 is the maximum pooling layer: the height and width of the pooling window are 4, the step length is 1, and the output is 51 multiplied by 81;
layer 5 is a convolutional layer: the size of convolution kernels is 1 multiplied by 55, the number of the convolution kernels is 64, the edges are filled, the filling mode is zero filling, the step length is 1, the output is 51 multiplied by 64, each convolution kernel uses a ReLU activation function, and a single-layer convolution network is realized;
layer 6 is the largest pooling layer: the height and width of the pooling window are 4, the step length is 1, and the output is 12 multiplied by 64;
layer 7 is a convolutional layer: the size of convolution kernels is 1 multiplied by 55, the number of the convolution kernels is 64, the edges are filled, the filling mode is zero filling, the step length is 1, the output is 12 multiplied by 64, each convolution kernel uses a ReLU activation function, and a single-layer convolution network is realized;
layer 8 is a fully connected layer: the output is 768 eigenvalues for a full connection layer containing 768 neurons;
layer 9 is softmax layer: the output of the softmax layer is the value number of the label, which is 3 multiplied by 1.
6. The rolling bearing fault diagnosis acoustic emission signal processing method based on compressed sensing of claim 1, characterized in that: the fault diagnosis result comprises three fault types, namely normal, roller fault and race fault.
7. The utility model provides a antifriction bearing fault diagnosis acoustic emission signal processing system based on compressed sensing which characterized in that: the method comprises the following steps:
the acquisition module is used for acquiring original acoustic emission signals of different rolling bearing fault states;
the selection module is used for calculating the reconstruction error of the original acoustic emission signal in the compression ratio setting range by using an orthogonal matching pursuit algorithm, and selecting the compression ratio for participating in compression from the compression ratio setting range according to the reconstruction error;
the first obtaining module is used for constructing an MXN Gaussian random matrix as a projection matrix to perform dimension reduction projection on the original acoustic emission signal according to the dimension M of the original acoustic emission signal and the dimension N of the compressed signal determined by the compression rate selected to participate in compression, so as to obtain the compressed signal;
the dividing module is used for dividing the obtained compressed signal adding label into a training sample set and a testing sample set;
and the second obtaining module is used for establishing a convolutional neural network model, inputting the training sample set into the convolutional neural network model for continuous iterative training until the training is finished, and obtaining the trained convolutional neural network model.
8. A processor, characterized in that: the processor is used for running a program, wherein the program is run to execute the rolling bearing fault diagnosis acoustic emission signal processing method based on compressed sensing according to any one of claims 1 to 6.
9. A computer-readable storage medium characterized by: the computer-readable storage medium comprises a stored program, wherein when the program runs, the computer-readable storage medium is controlled to execute the rolling bearing fault diagnosis acoustic emission signal processing method based on compressed sensing according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210624919.2A CN114964782A (en) | 2022-06-02 | 2022-06-02 | Rolling bearing fault diagnosis acoustic emission signal processing method and system based on compressed sensing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210624919.2A CN114964782A (en) | 2022-06-02 | 2022-06-02 | Rolling bearing fault diagnosis acoustic emission signal processing method and system based on compressed sensing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114964782A true CN114964782A (en) | 2022-08-30 |
Family
ID=82959431
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210624919.2A Pending CN114964782A (en) | 2022-06-02 | 2022-06-02 | Rolling bearing fault diagnosis acoustic emission signal processing method and system based on compressed sensing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114964782A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115290761A (en) * | 2022-09-27 | 2022-11-04 | 中国科学院地质与地球物理研究所 | Acoustic emission data reconstruction method and device based on deep learning |
| CN116010805A (en) * | 2023-03-24 | 2023-04-25 | 昆明理工大学 | Rolling bearing fault feature extraction method and device based on convolutional neural network |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109858352A (en) * | 2018-12-26 | 2019-06-07 | 华中科技大学 | A kind of method for diagnosing faults based on compressed sensing and the multiple dimensioned network of improvement |
| CN111458146A (en) * | 2020-04-02 | 2020-07-28 | 兰州理工大学 | Rolling bearing multi-measuring-point vibration signal compression sampling and synchronous reconstruction method |
| CN112257747A (en) * | 2020-09-16 | 2021-01-22 | 北京工业大学 | Diagnostic method based on compressed data and supervised global-local/non-local analysis |
| US20210382120A1 (en) * | 2020-06-08 | 2021-12-09 | Wuhan University | Failure diagnosis method for power transformer winding based on gsmallat-nin-cnn network |
| CN113919395A (en) * | 2021-10-12 | 2022-01-11 | 大连理工大学 | Water supply pipe network leakage accident diagnosis method based on one-dimensional convolutional neural network |
-
2022
- 2022-06-02 CN CN202210624919.2A patent/CN114964782A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109858352A (en) * | 2018-12-26 | 2019-06-07 | 华中科技大学 | A kind of method for diagnosing faults based on compressed sensing and the multiple dimensioned network of improvement |
| CN111458146A (en) * | 2020-04-02 | 2020-07-28 | 兰州理工大学 | Rolling bearing multi-measuring-point vibration signal compression sampling and synchronous reconstruction method |
| US20210382120A1 (en) * | 2020-06-08 | 2021-12-09 | Wuhan University | Failure diagnosis method for power transformer winding based on gsmallat-nin-cnn network |
| CN112257747A (en) * | 2020-09-16 | 2021-01-22 | 北京工业大学 | Diagnostic method based on compressed data and supervised global-local/non-local analysis |
| CN113919395A (en) * | 2021-10-12 | 2022-01-11 | 大连理工大学 | Water supply pipe network leakage accident diagnosis method based on one-dimensional convolutional neural network |
Non-Patent Citations (2)
| Title |
|---|
| 刘畅: "基于压缩感知的滚动轴承特征提取与特征约简方法研究", 《中国博士学位论文全文数据库工程科技Ⅱ辑》 * |
| 陈志刚等: "IEWT-CS和LCNN在轴承故障诊断中的应用", 《哈尔滨工程大学学报》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115290761A (en) * | 2022-09-27 | 2022-11-04 | 中国科学院地质与地球物理研究所 | Acoustic emission data reconstruction method and device based on deep learning |
| CN115290761B (en) * | 2022-09-27 | 2023-01-17 | 中国科学院地质与地球物理研究所 | Acoustic emission data reconstruction method and device based on deep learning |
| CN116010805A (en) * | 2023-03-24 | 2023-04-25 | 昆明理工大学 | Rolling bearing fault feature extraction method and device based on convolutional neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109858352B (en) | Fault diagnosis method based on compressed sensing and improved multi-scale network | |
| CN111964908B (en) | Bearing fault diagnosis method under variable working condition based on multi-scale weight distribution convolutional neural network model | |
| CN114964782A (en) | Rolling bearing fault diagnosis acoustic emission signal processing method and system based on compressed sensing | |
| CN110595780B (en) | Bearing fault identification method based on vibration gray level image and convolution neural network | |
| CN110132554B (en) | Rotary machine fault diagnosis method based on deep Laplace self-coding | |
| CN109993105A (en) | A kind of improved adaptive sparse sampling Fault Classification | |
| CN109782603A (en) | The detection method and monitoring system of rotating machinery coupling fault | |
| CN113567130A (en) | Bearing fault diagnosis method based on multiple working conditions of equipment | |
| CN109946080B (en) | Mechanical equipment health state identification method based on embedded circulation network | |
| CN111401136B (en) | Plunger pump cavitation degree detection method and device and terminal | |
| CN111122161B (en) | Variable working condition bearing fault diagnosis method based on fast kurtogram and deep residual learning | |
| CN112461543B (en) | Rotary machine fault diagnosis method based on multi-classification support vector data description | |
| CN114169377A (en) | G-MSCNN-based fault diagnosis method for rolling bearing in noisy environment | |
| CN114004135A (en) | Agricultural machinery bearing fault type diagnosis method and system based on Transformer neural network | |
| CN115687973A (en) | Mechanical equipment fault diagnosis method based on multi-signal fusion | |
| CN114741945A (en) | Diagnosis method for online faults of aircraft engine | |
| CN113409213B (en) | Method and system for enhancing noise reduction of time-frequency diagram of fault signal of plunger pump | |
| CN116296338A (en) | Rotary machine fault diagnosis method | |
| Lu et al. | An automated bearing fault diagnosis using a self-normalizing convolutional neural network | |
| CN115452376A (en) | Bearing fault diagnosis method based on improved lightweight deep convolution neural network | |
| Li et al. | A robust fault diagnosis method for rolling bearings based on deep convolutional neural network | |
| CN114861349A (en) | Rolling bearing RUL prediction method based on model migration and wiener process | |
| CN114186593A (en) | Bearing fault identification method, device, equipment and storage medium under noise condition | |
| CN112834194B (en) | Fault intelligent detection method based on soft target measurement under fault-free sample | |
| Wang et al. | Bearing Remaining Life Prediction Method Based on Wavelet Transform Denoising and Feature Preservation CycleGAN under Unbalanced Samples |
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 |