CN116296338A - Rotary machine fault diagnosis method - Google Patents
Rotary machine fault diagnosis method Download PDFInfo
- Publication number
- CN116296338A CN116296338A CN202310329968.8A CN202310329968A CN116296338A CN 116296338 A CN116296338 A CN 116296338A CN 202310329968 A CN202310329968 A CN 202310329968A CN 116296338 A CN116296338 A CN 116296338A
- Authority
- CN
- China
- Prior art keywords
- layer
- wavelet
- convolution
- relu activation
- neural network
- 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 53
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000012549 training Methods 0.000 claims abstract description 41
- 238000002474 experimental method Methods 0.000 claims abstract description 26
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 20
- 238000013528 artificial neural network Methods 0.000 claims abstract description 9
- 230000004913 activation Effects 0.000 claims description 37
- 238000010606 normalization Methods 0.000 claims description 30
- 238000012360 testing method Methods 0.000 claims description 28
- 238000011176 pooling Methods 0.000 claims description 24
- 238000002405 diagnostic procedure Methods 0.000 claims description 12
- 238000005457 optimization Methods 0.000 claims description 10
- 230000002265 prevention Effects 0.000 claims description 7
- 238000003062 neural network model Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 abstract description 3
- 238000012216 screening Methods 0.000 abstract description 3
- 240000007651 Rubus glaucus Species 0.000 abstract description 2
- 235000011034 Rubus glaucus Nutrition 0.000 abstract description 2
- 235000009122 Rubus idaeus Nutrition 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 abstract description 2
- 230000007812 deficiency Effects 0.000 abstract description 2
- 238000010801 machine learning Methods 0.000 abstract description 2
- 238000001994 activation Methods 0.000 description 26
- 230000006870 function Effects 0.000 description 14
- 238000004458 analytical method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 108010076504 Protein Sorting Signals Proteins 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000001360 synchronised 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
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a rotary machine fault diagnosis method, which comprises the following steps: adopting collected vibration signals on bearings under different working conditions to construct a sample data set; classifying the bearing states by using three different diagnosis methods according to the sample data set to obtain classification accuracy; and determining a diagnosis result according to the classification accuracy. On one hand, the rotary machine fault type can be judged more accurately by screening the same data set by adopting three different diagnosis methods; on the other hand, the effectiveness of a wavelet convolutional neural network model obtained based on VGG11 neural network improvement as a basic framework in a third diagnosis method is verified, the feature extraction is more sensitive and effective through the optimized wavelet basis, the training model accuracy is higher, the method is a novel method for introducing deep learning into the field of fault diagnosis, and a novel method is provided for deploying machine learning models in the resource deficiency equipment such as raspberry pie, embedded system and the like; meanwhile, experimental data are the results of multiple experiments, so that the accidental of the results is avoided.
Description
Technical Field
The invention relates to a fault diagnosis method for rotary machinery, and belongs to the field of fault diagnosis of mechanical systems.
Background
With the development of modern industrial manufacturing, mechanical equipment is moving towards large-scale, automatic and complicated, and enterprises place higher demands on high performance, safety and reliability, so that fault diagnosis of rotating machinery is becoming more and more important and difficult. The rotating mechanical equipment has the complex characteristics of variable working condition operation, reversing and the like, and is required to perform preprocessing on signals, namely, perform stable working condition extraction on the signals, so that synchronous acquisition on different types of signals is required. The fault of the mechanical equipment is not obvious, and in addition, the noisy equipment working environment, the collected signals are very weak in fault characteristics and contain noise signal interference, so that the method is particularly important for effectively extracting the fault characteristics in the signals to the fault diagnosis of the mechanical equipment.
Disclosure of Invention
The invention provides a rotary machine fault diagnosis method, which can more accurately judge the type of rotary machine fault by screening the same data set by adopting three different diagnosis methods, and further verifies the effectiveness of a wavelet convolution neural network model obtained based on VGG11 neural network improvement as a basic framework in a third diagnosis method.
The technical scheme of the invention is as follows: a rotary machine fault diagnosis method, comprising:
adopting collected vibration signals on bearings under different working conditions to construct a sample data set;
classifying the bearing states by using three different diagnosis methods according to the sample data set to obtain classification accuracy;
and determining a diagnosis result according to the classification accuracy.
The method for constructing a sample data set by adopting collected vibration signals on bearings under different working conditions comprises the following steps: and carrying out sectional processing on signals under different working conditions, and marking.
The method for classifying the bearing states by using three different diagnosis methods to obtain classification accuracy comprises the following steps:
first diagnostic method: dividing the sample data set into a first training set and a first test set; extracting features from samples in the sample data set, and constructing the features and the labels into feature matrixes; classifying by adopting an SVM model according to the feature matrixes of the first training set and the first testing set, carrying out n times of experiments, and taking the average value of the n times of experiments as the first classification accuracy; wherein the characteristics include mean square value, effective value, skewness, kurtosis, peak value, peak-to-peak value, waveform factor, pulse factor and peak factor;
a second diagnostic method: dividing the sample data set into a second training set and a second test set; the second training set is fed into the wavelet convolutional neural network model in batches for training, and the parameters of the first wavelet convolutional layer are fixed during training; testing the wavelet convolutional neural network model by using a second test set, carrying out n times of experiments, and taking the average value of the n times of experiments as a second classification accuracy;
third diagnostic method: the second training set is fed into the wavelet convolution neural network model in batches for training, and any layer is not fixed during training, so that a wavelet convolution kernel based on input data self-adaptive optimization is obtained; and testing the wavelet convolutional neural network model subjected to the self-adaptive optimization by using a second test set, carrying out n times of experiments, and taking the average value of the n times of experiments as a third classification accuracy.
The determining the diagnosis result according to the classification accuracy comprises the following steps: and using a model corresponding to the diagnosis method with highest classification accuracy for diagnosing the data to be tested to obtain a final diagnosis result.
The wavelet convolutional neural network model framework is based on a VGG11 neural network and comprises 8 convolutional layers, 8 normalization layers, 10 Relu activation layers, 2 pooling layers, 3 linear full-connection layers, 2 Dropout overfitting prevention layers and a last Softmax; the wavelet convolution layer is constructed by using Daubechies wavelet function structure as a new wavelet convolution kernel to replace the traditional convolution kernel, and the pooling adopts a maximum pooling mode.
The connection sequence of the 8-layer convolution layer, the 8-layer normalization layer, the 10-time Relu activation, the 2-layer pooling layer, the 3-layer linear full connection layer, the 2-layer Dropout overfitting prevention layer and the final Softmax layer is as follows: wavelet convolution layer 1, normalization layer 1, relu activation 1, max pooling layer 1, convolution layer 2, normalization layer 2, relu activation 2, max pooling layer 2, convolution layer 3, normalization layer 3, relu activation 3, dropout1, convolution layer 4, normalization layer 4, relu activation 4, convolution layer 5, normalization layer 5, relu activation 5, convolution layer 6, normalization layer 6, relu activation 6, dropout2, convolution layer 7, normalization layer 7, relu activation 7, convolution layer 8, normalization layer 8, relu activation 8, linear full connection layer 1, relu activation 9, linear full connection layer 2, relu activation 10, linear full connection layer 3, softmax function.
The beneficial effects of the invention are as follows: on one hand, the rotary machine fault type can be judged more accurately by screening the same data set by adopting three different diagnosis methods; on the other hand, the effectiveness of a wavelet convolutional neural network model obtained based on VGG11 neural network improvement as a basic framework in a third diagnosis method is verified, the feature extraction is more sensitive and effective through the optimized wavelet basis, the training model accuracy is higher, the method is a novel method for introducing deep learning into the field of fault diagnosis, and a novel method is provided for deploying machine learning models in the resource deficiency equipment such as raspberry pie, embedded system and the like; meanwhile, experimental data are the results of multiple experiments, so that the accidental of the results is avoided.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a graph comparing the before and after fine tuning of the custom wavelet convolution kernel;
FIG. 3 is a graph of failure recognition accuracy versus effectiveness.
Detailed Description
The invention will be further described with reference to the drawings and examples, but the invention is not limited to the scope.
Example 1: as shown in fig. 1 to 3, a fault diagnosis method for a rotary machine includes: adopting collected vibration signals on bearings under different working conditions to construct a sample data set; classifying the bearing states by using three different diagnosis methods according to the sample data set to obtain classification accuracy; and determining a diagnosis result according to the classification accuracy.
Further, the construction of the sample data set by adopting the collected vibration signals on the bearings under different working conditions comprises the following steps: and carrying out sectional processing on signals under different working conditions, and marking.
Further, the method for classifying the bearing state by using three different diagnostic methods to obtain the classification accuracy comprises the following steps:
first diagnostic method: dividing the sample data set into a first training set and a first test set; extracting features from samples in the sample data set, and constructing the features and the labels into feature matrixes; classifying by adopting SVM according to the feature matrixes of the first training set and the first testing set, carrying out n times of experiments, and taking the average value of the n times of experiments as the first classification accuracy; wherein the characteristics include mean square value, effective value, skewness, kurtosis, peak value, peak-to-peak value, waveform factor, pulse factor and peak factor;
a second diagnostic method: dividing the sample data set into a second training set and a second test set; the second training set is fed into the wavelet convolution neural network model in batches for training, and the parameters of the first layer of wavelet convolution layer are fixed (namely, the input and output parameter values of the wavelet convolution layer are the same) during training, so that the self-adaptive optimization of the wavelet convolution kernel is not carried out; testing the wavelet convolutional neural network model by using a second test set, carrying out n times of experiments, and taking the average value of the n times of experiments as a second classification accuracy;
third diagnostic method: the second training set is fed into the wavelet convolution neural network model in batches for training, and any layer is not fixed during training, namely, the wavelet convolution kernel is subjected to self-adaptive fine adjustment on the input data of the wavelet convolution kernel, so that the wavelet convolution kernel based on self-adaptive optimization of the input data is obtained; and testing the wavelet convolutional neural network model subjected to the self-adaptive optimization by using a second test set, carrying out n times of experiments, and taking the average value of the n times of experiments as a third classification accuracy.
It should be noted that, for the second diagnostic method and the third diagnostic method, model training parameters are set, the input data length is 2048, the batch size training size is 128, the learning rate is 0.001, the attenuation rate is 0.99, all parameters are updated through back propagation and Adam optimization algorithm, and the cross entropy loss function is used for calculating the loss function.
Further, determining the diagnosis result according to the classification accuracy comprises: and using the diagnosis method with highest classification accuracy for diagnosing the data to be tested to obtain a final diagnosis result.
Further, the wavelet convolutional neural network model framework is based on a VGG11 neural network and comprises 8 convolutional layers, 8 normalization layers, 10 Relu activations, 2 pooling layers, 3 linear full-connection layers, 2 Dropout overfitting prevention layers and a last Softmax; the wavelet convolution layer is constructed by using a Daubechies wavelet function structure as a new wavelet convolution kernel to replace the traditional convolution kernel, and a maximum pooling mode is adopted for pooling; and using the third diagnosis method for diagnosing the data to be tested to obtain a final diagnosis result.
Further, the 8-layer convolution layer, the 8-layer normalization layer, the 10-time Relu activation, the 2-layer pooling layer, the 3-layer linear full connection layer, the 2-layer Dropout overfitting prevention and the final Softmax layer are connected in the following sequence: wavelet convolution layer 1, normalization layer 1, relu activation 1, max pooling layer 1, convolution layer 2, normalization layer 2, relu activation 2, max pooling layer 2, convolution layer 3, normalization layer 3, relu activation 3, dropout1, convolution layer 4, normalization layer 4, relu activation 4, convolution layer 5, normalization layer 5, relu activation 5, convolution layer 6, normalization layer 6, relu activation 6, dropout2, convolution layer 7, normalization layer 7, relu activation 7, convolution layer 8, normalization layer 8, relu activation 8, linear full connection layer 1, relu activation 9, linear full connection layer 2, relu activation 10, linear full connection layer 3, softmax function.
Further, wavelet convolutional layers, classical convolutional layers, max-pooling layers, and fully-connected layers in a wavelet convolutional neural network model are presented. The method comprises the following steps:
the output calculation of the input signal sequence x and the convolution kernel w in the convolution layer is shown in formula (1):
where k and i represent the ith convolution kernel of the kth layer of the convolutional neural network, z (i) represents the feature map learned by the ith convolution kernel,
representing the input signal>
Representing convolution kernel +.>
Representing the bias of the convolution kernel.
The wavelet transform is as in formula (2):
S(a,τ)＝f(t)*ψ(f,t)＝∫f(t)ψ(τ-t,a)dt (2)
wherein psi (f, t) is a scale base function, and f (t) is a vibration signal of a time domain;
thus, according to the two formulas above, the convolution kernel w can be constructed using the Daubechies wavelet function, and wavelet analysis of the convolution kernel can be achieved. The dbN wavelet function is shown as formula (3), N is the order of wavelet, has the characteristic of orthogonality, and is assumed
Wherein->
Coefficients that are binomial are:
in the method, in the process of the invention,
the purpose of the pooling layer is to reduce the dimension of the features, so that the model is more concerned about whether certain features exist rather than the specific positions of the features, and a small displacement of some features is tolerated. Maximum pooling is a common pooling approach for models, which receives only the largest elements in the field of view, and which may overfit training data.
The full-connection layer is fitted by a Softmax function, and probability distribution prediction is carried out on input data. The Softmax function formula is as follows:
wherein: assuming that the output layer has n neurons in total, y k For the output of the kth neuron, a k For input data.
Still further, alternative embodiments of the invention are described below:
(1) The method comprises the steps of testing rolling bearing vibration signal data of a rotary mechanical fault simulation test bed, wherein the vibration signals comprise normal bearing signals, bearing inner ring fault signals, bearing outer ring fault signals and bearing rolling body fault signals; the four groups of signals are segmented, the length of each segment is 2048, four types of labels are marked, 800 samples are obtained, 80% of the samples are used as training samples, and 20% of the samples are used as test samples.
(2) Parameter table of the wavelet convolutional neural network model:
TABLE 1 network model size parameter table
| Sequence number | Network layer | Size of the device | Activation | Output size | |
| 1 | |
1×27×55 | ReLU | 2048×27 | |
| 2 | |
16×1 | |||
| 3 | |
27×27×55 | ReLU | 128×27 | |
| 4 | |
16×1 | |||
| 5 | |
27×27×55 | ReLU | 128×27 | |
| 6 | |
||||
| 7 | |
27×27×55 | |
8×27 | |
| 8 | |
27×27×55 | ReLU | 8×27 | |
| 9 | |
27×27×55 | ReLU | 8×27 | |
| 10 | Dropout | ||||
| 11 | |
27×27×55 | ReLU | 8×27 | |
| 12 | |
27×27×55 | ReLU | 8×27 | |
| 13 | Flatten | 216 | |||
| 14 | Full1 | 216×432 | ReLU | 216×1 | |
| 15 | Full2 | 432×16 | ReLU | 64×1 | |
| 16 | Full3 | 16×4 | Softmax | 4×1 |
The wavelet convolution layer is a wavelet convolution layer 1, the size of the wavelet convolution layer 1 is [1,27,55], the size means that 27 one-dimensional wavelet convolutions with the length of 55 are shared by the wavelet convolution layer 1 to convolve signals, the whole frequency domain is divided into 27 frequency bands, the center frequency of each frequency band corresponds to a wavelet function, parameters of the 27 wavelet functions are assigned to 27 convolution kernels respectively, each convolution kernel convolves with the signals to obtain frequency domain information of the corresponding frequency band, the subsequent network structure can further conduct feature mining on the frequency band information obtained by the wavelet convolution layer 1, the gradient descent back propagation algorithm is used for fine tuning the wavelet convolution kernels based on input data, and therefore better wavelet convolution kernels are obtained, and the analysis effect on the data is more advantageous. As shown in fig. 2, a comparison graph before and after the optimization and fine tuning of one of the custom wavelet convolution kernels (the abscissa refers to the size 55 of the wavelet convolution kernel, and the ordinate refers to the amplitude), a curve without a sign in the graph is an input wavelet convolution kernel without self-adaption, a curve with a sign is a self-adaption optimized wavelet curve, and the self-adaption fine tuning is performed on the input data of the original wavelet convolution kernel, so that the optimal convolution kernel is achieved, and the characteristic of the input data is adapted to the input data so as to deeply excavate the input data, thereby improving the accuracy of fault diagnosis.
(3) The experimental analysis steps are as follows:
step1: marking each working condition data of four groups of data, and respectively extracting characteristics from each sample, wherein the characteristics comprise 9 characteristics including a mean square value, an effective value, a skewness, a kurtosis, a peak value, a peak-to-peak value, a waveform factor, a pulse factor and a peak factor, and all the characteristics are constructed into a characteristic matrix. The ratio of the divided training set to the test set is 8: and 2, sending the feature matrix into an SVM for classification. To ensure accuracy of the experimental results, 10 experiments were performed, and the obtained results are shown in table 2.
Step2: the ratio of the divided training set to the test set is 8:2, obtaining a divided data set; and sending the training data set into the wavelet convolution neural network in batches for training, wherein the first layer of convolution layer, namely the wavelet convolution layer, is fixed during training, and the wavelet convolution kernel is not subjected to self-adaptive optimization. And testing the wavelet convolutional neural network model to obtain the fault recognition classification accuracy of the fixed wavelet method. To ensure the accuracy of the experimental results, 10 experiments were performed in total, and the obtained results are shown in table 2.
The wavelet convolution neural network model is
Step3: and sending the divided training data set described in step2 into the wavelet convolution neural network in batches for training, wherein any layer is not fixed during training, namely, the wavelet convolution check is subjected to self-adaptive fine tuning on the input data so as to obtain a wavelet convolution kernel which is self-adaptively optimized based on the input data. And testing based on the self-adaptive optimized wavelet convolutional neural network model to obtain the fault classification recognition accuracy of the self-adaptive optimized wavelet method. To ensure the accuracy of the experimental results, 10 experiments were performed in total, and the obtained results are shown in table 2.
Table 2 comparative table of classification accuracy effect for 10 experiments
As shown in the comparison table of the fault recognition classification accuracy rate effect of 10 times of experiments, the average classification accuracy rate of the third diagnosis method is obviously improved compared with that of the first diagnosis method and the second diagnosis method, and the effectiveness of the wavelet convolutional neural network model obtained based on VGG11 neural network improvement as a basic framework in the third diagnosis method is demonstrated. Meanwhile, in order to more intuitively embody the results of the three methods, a fault recognition accuracy comparison chart shown in fig. 3 is provided.
Further, an optional diagnostic step is further given in accordance with the above-described experiment: adopting collected vibration signals on bearings under different working conditions to construct a sample data set; according to the sample data set, using a model corresponding to the third diagnosis method for diagnosing the data to be tested to obtain a final diagnosis result; the model adopted in the third diagnosis method is a wavelet convolutional neural network model comprising 8 layers of convolutional layers, 8 layers of normalization layers, 10 layers of Relu activation, 2 layers of pooling layers, 3 layers of linear full-connection layers, 2 layers of Dropout overfitting prevention and a last layer of Softmax.
While the present invention has been described in detail with reference to the drawings, the present invention is not limited to the above 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 (6)
1. A rotary machine fault diagnosis method, characterized by comprising:
adopting collected vibration signals on bearings under different working conditions to construct a sample data set;
classifying the bearing states by using three different diagnosis methods according to the sample data set to obtain classification accuracy;
and determining a diagnosis result according to the classification accuracy.
2. The rotary machine fault diagnosis method according to claim 1, wherein the constructing a sample data set using collected vibration signals on bearings under different conditions comprises: and carrying out sectional processing on signals under different working conditions, and marking.
3. The rotary machine fault diagnosis method according to claim 1, wherein the classification of the bearing state by three different diagnosis methods to obtain the classification accuracy comprises:
first diagnostic method: dividing the sample data set into a first training set and a first test set; extracting features from samples in the sample data set, and constructing the features and the labels into feature matrixes; classifying by adopting an SVM model according to the feature matrixes of the first training set and the first testing set, carrying out n times of experiments, and taking the average value of the n times of experiments as the first classification accuracy; wherein the characteristics include mean square value, effective value, skewness, kurtosis, peak value, peak-to-peak value, waveform factor, pulse factor and peak factor;
a second diagnostic method: dividing the sample data set into a second training set and a second test set; the second training set is fed into the wavelet convolutional neural network model in batches for training, and the parameters of the first wavelet convolutional layer are fixed during training; testing the wavelet convolutional neural network model by using a second test set, carrying out n times of experiments, and taking the average value of the n times of experiments as a second classification accuracy;
third diagnostic method: the second training set is fed into the wavelet convolution neural network model in batches for training, and any layer is not fixed during training, so that a wavelet convolution kernel based on input data self-adaptive optimization is obtained; and testing the wavelet convolutional neural network model subjected to the self-adaptive optimization by using a second test set, carrying out n times of experiments, and taking the average value of the n times of experiments as a third classification accuracy.
4. The rotary machine fault diagnosis method according to claim 1, wherein the determining the diagnosis result according to the classification accuracy comprises: and using a model corresponding to the diagnosis method with highest classification accuracy for diagnosing the data to be tested to obtain a final diagnosis result.
5. The rotary machine fault diagnosis method according to claim 3, wherein the wavelet convolutional neural network model framework is based on VGG11 neural network, and comprises 8 convolutional layers, 8 normalized layers, 10 Relu activated layers, 2 pooled layers, 3 linear full-connection layers, 2 Dropout prevention overfitting, and the last Softmax; the wavelet convolution layer is constructed by using Daubechies wavelet function structure as a new wavelet convolution kernel to replace the traditional convolution kernel, and the pooling adopts a maximum pooling mode.
6. The rotary machine fault diagnosis method according to claim 5, wherein the 8-layer convolution layer, 8-layer normalization layer, 10-time Relu activation, 2-layer pooling layer, 3-layer linear full-connection layer, 2-layer Dropout overfitting prevention and last Softmax layer are connected in the following order: wavelet convolution layer 1, normalization layer 1, relu activation 1, max pooling layer 1, convolution layer 2, normalization layer 2, relu activation 2, max pooling layer 2, convolution layer 3, normalization layer 3, relu activation 3, dropout1, convolution layer 4, normalization layer 4, relu activation 4, convolution layer 5, normalization layer 5, relu activation 5, convolution layer 6, normalization layer 6, relu activation 6, dropout2, convolution layer 7, normalization layer 7, relu activation 7, convolution layer 8, normalization layer 8, relu activation 8, linear full connection layer 1, relu activation 9, linear full connection layer 2, relu activation 10, linear full connection layer 3, softmax function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310329968.8A CN116296338A (en) | 2023-03-30 | 2023-03-30 | Rotary machine fault diagnosis method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310329968.8A CN116296338A (en) | 2023-03-30 | 2023-03-30 | Rotary machine fault diagnosis method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116296338A true CN116296338A (en) | 2023-06-23 |
Family
ID=86777895
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310329968.8A Pending CN116296338A (en) | 2023-03-30 | 2023-03-30 | Rotary machine fault diagnosis method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116296338A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116499748A (en) * | 2023-06-27 | 2023-07-28 | 昆明理工大学 | Bearing fault diagnosis method and system based on improved SMOTE and classifier |
-
2023
- 2023-03-30 CN CN202310329968.8A patent/CN116296338A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116499748A (en) * | 2023-06-27 | 2023-07-28 | 昆明理工大学 | Bearing fault diagnosis method and system based on improved SMOTE and classifier |
| CN116499748B (en) * | 2023-06-27 | 2023-08-29 | 昆明理工大学 | Bearing fault diagnosis method and system based on improved SMOTE and classifier |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112084974B (en) | Multi-label rolling bearing fault diagnosis method based on meta-learning | |
| CN108319962B (en) | Tool wear monitoring method based on convolutional neural network | |
| Li et al. | Fault diagnosis of rotating machinery based on combination of deep belief network and one-dimensional convolutional neural network | |
| CN111964908A (en) | MWDCNN-based bearing fault diagnosis method under variable working conditions | |
| CN112418013B (en) | Complex working condition bearing fault diagnosis method based on meta-learning under small sample | |
| CN112577747A (en) | Rolling bearing fault diagnosis method based on space pooling network | |
| Liang et al. | Rolling bearing fault diagnosis based on one-dimensional dilated convolution network with residual connection | |
| CN109186973A (en) | A kind of mechanical failure diagnostic method of unsupervised deep learning network | |
| CN109186964B (en) | rotary machine fault diagnosis method based on angle resampling and ROC-SVM | |
| CN112257530B (en) | Rolling bearing fault diagnosis method based on blind signal separation and support vector machine | |
| CN110672343A (en) | Rotary machine fault diagnosis method based on multi-attention convolutional neural network | |
| CN113567130A (en) | Bearing fault diagnosis method based on multiple working conditions of equipment | |
| CN113111820B (en) | Rotary part fault diagnosis method and device based on improved CNN and relation module | |
| CN113435322B (en) | Method, system, equipment and readable storage medium for detecting faults of spindle bearing | |
| CN111753891B (en) | Rolling bearing fault diagnosis method based on unsupervised feature learning | |
| CN112378660A (en) | Intelligent fault diagnosis method for aero-engine bearing based on data driving | |
| CN113203566A (en) | Motor bearing fault diagnosis method based on one-dimensional data enhancement and CNN | |
| CN116296338A (en) | Rotary machine fault diagnosis method | |
| CN113392881A (en) | Rotary machine fault diagnosis method based on transfer learning | |
| CN114563186A (en) | Rolling bearing unbalance fault diagnosis method and system | |
| CN112633098A (en) | Fault diagnosis method and system for rotary machine and storage medium | |
| CN112633132A (en) | Bearing fault diagnosis method and system based on two-dimensional vibration image enhancement | |
| CN111983569B (en) | Radar interference suppression method based on neural network | |
| CN115290326A (en) | Rolling bearing fault intelligent diagnosis method | |
| CN114861740B (en) | Self-adaptive mechanical fault diagnosis method and system based on multi-head attention mechanism |
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 |