CN112395968A - Mechanical rotating part fault diagnosis method and device based on neural network - Google Patents
Mechanical rotating part fault diagnosis method and device based on neural network Download PDFInfo
- Publication number
- CN112395968A CN112395968A CN202011253582.6A CN202011253582A CN112395968A CN 112395968 A CN112395968 A CN 112395968A CN 202011253582 A CN202011253582 A CN 202011253582A CN 112395968 A CN112395968 A CN 112395968A
- Authority
- CN
- China
- Prior art keywords
- neural network
- dimensional
- layer
- fault
- convolutional neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
- G06F17/141—Discrete Fourier transforms
- G06F17/142—Fast Fourier transforms, e.g. using a Cooley-Tukey type algorithm
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
- G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
-
- 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
- G06F2218/10—Feature extraction by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Mathematical Analysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- General Health & Medical Sciences (AREA)
- Computational Mathematics (AREA)
- Discrete Mathematics (AREA)
- Signal Processing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention relates to a method and a device for diagnosing faults of a mechanical rotating part based on a neural network, belongs to the technical field of fault diagnosis, and solves the problem that the existing method for diagnosing faults of the mechanical rotating part is poor in accuracy of diagnosis results. The method comprises the following steps: acquiring a historical vibration signal of the mechanical rotating part, and adding a label to the historical vibration signal; constructing a high-dimensional convolutional neural network model, and training the high-dimensional convolutional neural network model based on a historical vibration signal to obtain an optimal network structure of the high-dimensional convolutional neural network model; and inputting the vibration signal to be diagnosed into the optimal network structure of the high-dimensional convolutional neural network model to obtain a diagnosis result. The fault diagnosis of the mechanical rotating part is realized, and the diagnosis precision is improved.
Description
Technical Field
The invention relates to the technical field of fault diagnosis, in particular to a method and a device for diagnosing faults of a mechanical rotating part based on a neural network.
Background
The bearing, the gear, the shaft and the like are important components of mechanical equipment and play important roles in supporting and transmitting the rotating speed and the like for the mechanical equipment. In a high-speed rotation state, a mechanical rotating part fails to cause a major accident, and for complex machinery, the fault is very difficult to locate, so that the fault maintenance efficiency is low and the maintenance cost is high.
The traditional manual diagnosis method is difficult to realize accurate positioning and fault degree identification of the fault of the mechanical rotating part. Meanwhile, a large number of model parameters are generally required to be generated based on the conventional convolutional neural network deep learning fault diagnosis method, so that the network structure is very complex, the calculation amount is large, and the diagnosis result precision is poor.
Disclosure of Invention
In view of the foregoing analysis, embodiments of the present invention are directed to a method and an apparatus for diagnosing a fault of a mechanical rotating component based on a neural network, so as to solve the problem of poor accuracy of a diagnosis result obtained by a conventional method for diagnosing a fault of a mechanical rotating component.
In one aspect, an embodiment of the present invention provides a method for diagnosing a fault of a mechanical rotating component based on a neural network, including the following steps:
acquiring a historical vibration signal of a mechanical rotating part, and adding a label to the historical vibration signal, wherein the label comprises a fault type, a fault position or a fault degree;
constructing a high-dimensional convolutional neural network model, and training the high-dimensional convolutional neural network model based on the historical vibration signal to obtain an optimal network structure of the high-dimensional convolutional neural network model;
and inputting the vibration signal to be diagnosed into the optimal network structure of the high-dimensional convolutional neural network model to obtain a diagnosis result.
Further, the historical vibration signal comprises an original vibration signal of a fault position of the mechanical rotating part, or an average value, a root mean square value, a kurtosis and a peak value of the original vibration signal, or an amplitude frequency spectrum obtained by performing fast fourier transform, wavelet decomposition and empirical mode decomposition on the original vibration signal.
Further, the high-dimensional convolutional neural network model includes:
the first ReLU convolutional network is used for preprocessing an input original vibration signal to obtain a preprocessed signal;
the second ReLU convolutional network is used for convolving the preprocessed signals output by the first ReLU convolutional network to obtain convolutional signals;
the first high-dimensional neuron layer is used for converting the convolution signals output by the second ReLU convolution network into matrixes to obtain high-dimensional neuron signals;
the second high-dimensional neuron layer is used for performing high-dimensional convolution operation on the high-dimensional neuron signals output by the first high-dimensional neuron layer to obtain high-dimensional convolution neuron signals;
and the output layer is used for obtaining and outputting the fault type, the fault position or the fault degree according to the high-dimensional convolution neuron signals output by the second high-dimensional neuron layer.
Further, the first layer and the second layer respectively comprise a boundary filling layer, a convolution layer, a pooling layer, a normalization layer and an activation layer which are sequentially connected.
Further, the output layer is a Capsule full connection layer.
Further, training a high-dimensional convolutional neural network model based on the historical vibration signal to obtain an optimal network structure of the high-dimensional convolutional neural network model, comprising the following steps of:
dividing the historical vibration signal into a training signal set and a verification signal set;
inputting the training signal set into the high-dimensional convolutional neural network model to obtain a trained high-dimensional convolutional neural network;
and verifying the trained high-dimensional convolutional neural network based on the verification signal set to obtain the optimal network structure of the high-dimensional convolutional neural network.
In another aspect, an embodiment of the present invention provides a mechanical rotating component fault diagnosis apparatus based on a neural network, including:
the system comprises a historical vibration signal acquisition module, a fault detection module and a fault detection module, wherein the historical vibration signal acquisition module is used for acquiring a historical vibration signal of a mechanical rotating part and adding a label to the historical vibration signal, and the label comprises a fault type, a fault position or a fault degree;
the model training module is used for constructing a high-dimensional convolutional neural network model and training the high-dimensional convolutional neural network model based on the historical vibration signal to obtain an optimal network structure of the high-dimensional convolutional neural network model;
and the diagnosis result obtaining module is used for inputting the vibration signal to be diagnosed into the optimal network structure of the high-dimensional convolutional neural network model to obtain a diagnosis result.
Further, the historical vibration signal comprises an original vibration signal of a fault position of the mechanical rotating part, or an average value, a root mean square value, a kurtosis and a peak value of the original vibration signal, or an amplitude frequency spectrum obtained by performing fast fourier transform, wavelet decomposition and empirical mode decomposition on the original vibration signal.
Further, the high-dimensional convolutional neural network model includes:
the first ReLU convolutional network is used for preprocessing an input original vibration signal to obtain a preprocessed signal;
the second ReLU convolutional network is used for convolving the preprocessed signals output by the first ReLU convolutional network to obtain convolutional signals;
the first high-dimensional neuron layer is used for converting the convolution signals output by the second ReLU convolution network into matrixes to obtain high-dimensional neuron signals;
the second high-dimensional neuron layer is used for performing high-dimensional convolution operation on the high-dimensional neuron signals output by the first high-dimensional neuron layer to obtain high-dimensional convolution neuron signals;
and the output layer is used for obtaining and outputting the fault type, the fault position or the fault degree according to the high-dimensional convolution neuron signals output by the second high-dimensional neuron layer.
Further, the first layer and the second layer respectively comprise a boundary filling layer, a convolution layer, a pooling layer, a normalization layer and an activation layer which are sequentially connected.
Compared with the prior art, the invention can realize at least one of the following beneficial effects:
1. a fault diagnosis method and a fault diagnosis device for a mechanical rotating part based on a neural network are characterized in that a historical vibration signal of the mechanical rotating part is obtained, a high-dimensional convolutional neural network model constructed by training of the historical vibration signal is utilized to obtain an optimal network structure of the high-dimensional convolutional neural network model, finally, a vibration signal to be diagnosed is input into the optimal network structure of the high-dimensional convolutional neural network model to obtain a fault type, a fault position or a fault degree of the mechanical rotating part, and the fault diagnosis method and the fault diagnosis device are simple, easy to implement and easy to implement, and improve the precision of a diagnosis result.
2. Through the historical vibration signal of gathering mechanical rotating part to for historical vibration signal adds the label, provide technical support and basis for the later stage carries out high-dimensional convolution neural network model.
3. The high-dimensional convolutional neural network model adopts a high-dimensional neuron architecture, so that the number of network layers is reduced, network parameters are reduced, and the real-time performance and the precision of fault diagnosis are effectively improved.
4. After the optimal network structure of the high-dimensional convolutional neural network is obtained, the vibration signal to be diagnosed can be input, and a diagnosis result can be obtained. The optimal network structure of the high-dimensional convolutional neural network can be deployed on a LabVIEW, acquisition system and FPGA test system, so that the online real-time diagnosis of the fault of the mechanical rotating part is realized, and the real-time performance of the diagnosis result is improved.
In the invention, the technical schemes can be combined with each other to realize more preferable combination schemes. Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and drawings.
Drawings
The drawings are only for purposes of illustrating particular embodiments and are not to be construed as limiting the invention, wherein like reference numerals are used to designate like parts throughout.
FIG. 1 is a schematic diagram of a method for fault diagnosis of a mechanical rotating component based on a neural network in one embodiment;
FIG. 2 is a flow diagram of a method for fault diagnosis of a mechanical rotating component based on a neural network in one embodiment;
FIG. 3 is a block diagram of a fault diagnosis device for a mechanical rotating part based on a neural network in another embodiment;
reference numerals:
100-historical vibration signal acquisition module, 200-model training module and 300-diagnosis result acquisition module.
Detailed Description
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate preferred embodiments of the invention and together with the description, serve to explain the principles of the invention and not to limit the scope of the invention.
The traditional manual diagnosis method is difficult to realize accurate positioning and fault degree identification of the fault of the mechanical rotating part. Meanwhile, a large number of model parameters are generally required to be generated based on the conventional convolutional neural network deep learning fault diagnosis method, so that the network structure is very complex, the calculation amount is large, and the diagnosis result precision is poor. Therefore, as shown in fig. 1, the application provides a method and a device for diagnosing the fault of a mechanical rotating part based on a neural network, and the method and the device are simple, easy and easy to implement, and improve the precision of a diagnosis result.
In an embodiment of the present invention, a method for diagnosing a fault of a mechanical rotating component based on a neural network is disclosed, as shown in fig. 2, including the following steps S1-S3.
And step S1, acquiring historical vibration signals of the mechanical rotating part, and adding labels to the historical vibration signals, wherein the labels comprise fault types, fault positions or fault degrees. Specifically, the historical vibration signal of the mechanical rotating component may be obtained by collecting data at a relevant position of the mechanical rotating component with the fault, where the historical vibration signal may be an original vibration signal of the fault position of the mechanical rotating component, may also be an average value, a root mean square value, a kurtosis value, or a peak value of the original vibration signal obtained by processing the collected original vibration signal, and may also be an amplitude-frequency spectrum obtained by performing fast fourier transform, wavelet decomposition, or empirical mode decomposition on the collected original vibration signal. The label comprises three types of failure types, failure positions and failure degrees, wherein the failure types comprise abrasion, gluing and breakage; the degree of failure includes normal, minor and major damage. Any one or more of three labels of fault type, fault position and fault degree can be added to the historical vibration signal, and the specific situation is determined by the user.
Through the historical vibration signal of gathering mechanical rotating part to for historical vibration signal adds the label, provide technical support and basis for the later stage carries out high-dimensional convolution neural network model.
And S2, constructing a high-dimensional convolutional neural network model, and training the high-dimensional convolutional neural network model based on the historical vibration signal to obtain the optimal network structure of the high-dimensional convolutional neural network model. Considering that the accuracy of the fault diagnosis result obtained by the traditional neural network model is low, a high-dimensional convolutional neural network model is constructed, model training is carried out on the high-dimensional convolutional neural network model through the historical vibration signal obtained in the step S1, and finally the optimal network structure is obtained.
Preferably, the high-dimensional convolutional neural network model comprises:
the first ReLU convolutional network is used for preprocessing an input original vibration signal to obtain a preprocessed signal;
the second ReLU convolutional network is used for convolving the preprocessed signals output by the first ReLU convolutional network to obtain convolutional signals;
the first high-dimensional neuron layer is used for converting the convolution signals output by the second ReLU convolution network into matrixes to obtain high-dimensional neuron signals;
the second high-dimensional neuron layer is used for performing high-dimensional convolution operation on the high-dimensional neuron signals output by the first high-dimensional neuron layer to obtain high-dimensional convolution neuron signals;
and the output layer is used for obtaining and outputting the fault type, the fault position or the fault degree according to the high-dimensional convolution neuron signals output by the second high-dimensional neuron layer.
Specifically, the high-dimensional convolutional neural network model includes five layers, two ReLU convolutional networks, two high-dimensional neuron layers, and one output layer. The convolution kernel of the first ReLU convolution network is 64, the step size of the convolution kernel is 8, the number of channels is 16, and the convolution kernel is used for preprocessing an original signal, where the preprocessing refers to filtering processing to obtain a preprocessed signal. The convolution kernel of the second ReLU convolution network is set to be 1, the step length of the convolution kernel is 1, the number of channels is 32, and the convolution is mainly performed on the preprocessed signal output by the first ReLU convolution network to extract the fine feature of the signal and obtain the convolution signal. In detail, both ReLU convolutional networks include a boundary padding layer, a convolutional layer, a pooling layer, a normalization layer, and an activation layer, which are connected in sequence.
The first high-dimensional neuron layer can bind the convolution signals output by every four channels of the second ReLU convolution network into a matrix, the matrix has directionality and size, and the matrix obtained by binding is the high-dimensional neuron signals. The second high-dimensional neuron layer is a convolution layer and is mainly used for performing convolution on the high-dimensional neuron signals output by the first neuron layer to obtain high-dimensional convolution neuron signals, and the signals also have directionality and magnitude. The output layer is a Capsule full-connection layer, the size of the second high-dimensional neuron layer can be extracted, and the modular length is the fault type, the fault position or the fault degree corresponding to the mechanical rotating component.
The high-dimensional convolutional neural network model adopts a high-dimensional neuron architecture, so that the number of network layers is reduced, network parameters are reduced, and the real-time performance and the precision of fault diagnosis are effectively improved.
Preferably, training the high-dimensional convolutional neural network model based on the historical vibration signal to obtain an optimal network structure of the high-dimensional convolutional neural network model, comprising the following steps:
the historical vibration signal is divided into a training signal set and a verification signal set. Specifically, after the high-dimensional convolutional neural network model is built, the high-dimensional convolutional neural network model can be trained according to the collected historical vibration signals. Firstly, respectively training a signal set and a verification signal set for historical vibration signals, wherein the training signal set accounts for 70% of the total number of the historical vibration signals, is used for carrying out model training on a high-dimensional convolutional neural network model to obtain a trained high-dimensional convolutional neural network, and the verification signal set accounts for 30% of the total number of the historical vibration signals, and is used for verifying the trained high-dimensional convolutional neural network to obtain an optimal network structure of the high-dimensional convolutional neural network.
And inputting the training signal set into a high-dimensional convolutional neural network model to obtain a trained high-dimensional convolutional neural network. Before the model training, the initial convolution kernel number, the neuron number, the weight, the iteration number and the learning rate of the model can be set firstly, then the training signal set is input into the high-dimensional convolution neural network model for training, the weight, the convolution kernel number and the neuron number are adjusted, and when the loss function corresponding to the training signal set is smaller than a set threshold value, the trained high-dimensional convolution neural network is obtained.
And verifying the trained high-dimensional convolutional neural network based on a verification signal set to obtain an optimal network structure of the high-dimensional convolutional neural network. Similarly, inputting the verification signal set into the trained high-dimensional convolutional neural network to obtain a fault diagnosis result, comparing the fault diagnosis result with the label to obtain diagnosis precision, if the diagnosis precision is higher, the trained high-dimensional convolutional neural network is the optimal network structure of the high-dimensional convolutional neural network, if the diagnosis precision is lower, the weight, the number of convolutional kernels and the number of neurons are readjusted to enable the diagnosis precision to be optimal, and the network with the best precision is the optimal network structure of the high-dimensional convolutional neural network.
And step S3, inputting the vibration signal to be diagnosed into the optimal network structure of the high-dimensional convolutional neural network model to obtain a diagnosis result. After the optimal network structure of the high-dimensional convolutional neural network is obtained based on step S2, the vibration signal to be diagnosed may be input, and the diagnosis result may be obtained. The optimal network structure of the high-dimensional convolutional neural network can be deployed on a LabVIEW, acquisition system and FPGA test system, so that the online real-time diagnosis of the fault of the mechanical rotating part is realized, and the real-time performance of the diagnosis result is improved.
Compared with the prior art, the method and the device for diagnosing the fault of the mechanical rotating part based on the neural network provided by the embodiment have the advantages that the historical vibration signal of the mechanical rotating part is firstly obtained, the high-dimensional convolutional neural network model constructed by training the historical vibration signal is utilized to obtain the optimal network structure of the high-dimensional convolutional neural network model, finally, the vibration signal to be diagnosed is input into the optimal network structure of the high-dimensional convolutional neural network model to obtain the fault type, the fault position or the fault degree of the mechanical rotating part, and the method and the device are simple, easy to implement and easy to implement, improve the precision of the diagnosis result and have higher practical value.
In another embodiment of the present invention, a fault diagnosis device for a mechanical rotating component based on a neural network is disclosed, as shown in fig. 2. The method comprises the following steps: a historical vibration signal obtaining module 100, configured to obtain a historical vibration signal of the mechanical rotating component, and add a label to the historical vibration signal, where the historical vibration signal includes a fault type, a fault location, or a fault degree; the model training module 200 is used for constructing a high-dimensional convolutional neural network model and training the high-dimensional convolutional neural network model based on a historical vibration signal to obtain an optimal network structure of the high-dimensional convolutional neural network model; and a diagnostic result obtaining module 300, configured to input the vibration signal to be diagnosed into the optimal network structure of the high-dimensional convolutional neural network model, so as to obtain a diagnostic result.
Because the implementation principle of the neural network-based mechanical rotating part fault diagnosis device is similar to that of the neural network-based mechanical rotating part fault diagnosis device, the implementation principle is not repeated here.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention.
Claims (10)
1. A fault diagnosis method for a mechanical rotating part based on a neural network is characterized by comprising the following steps:
acquiring a historical vibration signal of a mechanical rotating part, and adding a label to the historical vibration signal, wherein the label comprises a fault type, a fault position or a fault degree;
constructing a high-dimensional convolutional neural network model, and training the high-dimensional convolutional neural network model based on the historical vibration signal to obtain an optimal network structure of the high-dimensional convolutional neural network model;
and inputting the vibration signal to be diagnosed into the optimal network structure of the high-dimensional convolutional neural network model to obtain a diagnosis result.
2. The neural network-based mechanical rotary component fault diagnosis method according to claim 1, wherein the historical vibration signals comprise raw vibration signals of collected mechanical rotary component fault locations, or mean values, root mean square values, kurtosis, peak values of the raw vibration signals, or magnitude frequency spectrums obtained by performing fast fourier transform, wavelet decomposition, empirical mode decomposition on the raw vibration signals.
3. The neural network-based fault diagnosis method for mechanical rotating components according to claim 1, wherein the high-dimensional convolutional neural network model comprises:
the first ReLU convolutional network is used for preprocessing an input original vibration signal to obtain a preprocessed signal;
the second ReLU convolutional network is used for convolving the preprocessed signals output by the first ReLU convolutional network to obtain convolutional signals;
the first high-dimensional neuron layer is used for converting the convolution signals output by the second ReLU convolution network into matrixes to obtain high-dimensional neuron signals;
the second high-dimensional neuron layer is used for performing high-dimensional convolution operation on the high-dimensional neuron signals output by the first high-dimensional neuron layer to obtain high-dimensional convolution neuron signals;
and the output layer is used for obtaining and outputting the fault type, the fault position or the fault degree according to the high-dimensional convolution neuron signals output by the second high-dimensional neuron layer.
4. The neural network-based mechanical rotary component fault diagnosis method of claim 3, wherein the first layer and the second layer each comprise a boundary filler layer, a convolutional layer, a pooling layer, a normalization layer, and an activation layer, which are connected in this order.
5. The neural network-based fault diagnosis method for mechanical rotating components, according to claim 4, wherein the output layer is a Capsule full connection layer.
6. The method for diagnosing the fault of the mechanical rotating component based on the neural network as claimed in claim 3 or 4, wherein a high-dimensional convolutional neural network model is trained based on the historical vibration signals to obtain an optimal network structure of the high-dimensional convolutional neural network model, and the method comprises the following steps:
dividing the historical vibration signal into a training signal set and a verification signal set;
inputting the training signal set into the high-dimensional convolutional neural network model to obtain a trained high-dimensional convolutional neural network;
and verifying the trained high-dimensional convolutional neural network based on the verification signal set to obtain the optimal network structure of the high-dimensional convolutional neural network.
7. A fault diagnosis device for a mechanical rotation component based on a neural network, comprising:
the system comprises a historical vibration signal acquisition module, a fault detection module and a fault detection module, wherein the historical vibration signal acquisition module is used for acquiring a historical vibration signal of a mechanical rotating part and adding a label to the historical vibration signal, and the label comprises a fault type, a fault position or a fault degree;
the model training module is used for constructing a high-dimensional convolutional neural network model and training the high-dimensional convolutional neural network model based on the historical vibration signal to obtain an optimal network structure of the high-dimensional convolutional neural network model;
and the diagnosis result obtaining module is used for inputting the vibration signal to be diagnosed into the optimal network structure of the high-dimensional convolutional neural network model to obtain a diagnosis result.
8. The neural network-based mechanical rotary component fault diagnosis device according to claim 7, wherein the historical vibration signals comprise raw vibration signals of collected fault positions of the mechanical rotary component, or mean values, root mean square values, kurtosis, peak values of the raw vibration signals, or amplitude frequency spectrums obtained by performing fast Fourier transform, wavelet decomposition, empirical mode decomposition on the raw vibration signals.
9. The neural network-based fault diagnosis device for mechanical rotating components according to claim 8, wherein the high-dimensional convolutional neural network model comprises:
the first ReLU convolutional network is used for preprocessing an input original vibration signal to obtain a preprocessed signal;
the second ReLU convolutional network is used for convolving the preprocessed signals output by the first ReLU convolutional network to obtain convolutional signals;
the first high-dimensional neuron layer is used for converting the convolution signals output by the second ReLU convolution network into matrixes to obtain high-dimensional neuron signals;
the second high-dimensional neuron layer is used for performing high-dimensional convolution operation on the high-dimensional neuron signals output by the first high-dimensional neuron layer to obtain high-dimensional convolution neuron signals;
and the output layer is used for obtaining and outputting the fault type, the fault position or the fault degree according to the high-dimensional convolution neuron signals output by the second high-dimensional neuron layer.
10. The neural network-based mechanical rotary component failure diagnostic apparatus of claim 9, wherein the first layer and the second layer each comprise a boundary filler layer, a convolutional layer, a pooling layer, a normalization layer, and an activation layer, which are connected in this order.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011253582.6A CN112395968B (en) | 2020-11-11 | 2020-11-11 | Mechanical rotating part fault diagnosis method and device based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011253582.6A CN112395968B (en) | 2020-11-11 | 2020-11-11 | Mechanical rotating part fault diagnosis method and device based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112395968A true CN112395968A (en) | 2021-02-23 |
| CN112395968B CN112395968B (en) | 2021-08-27 |
Family
ID=74600655
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011253582.6A Active CN112395968B (en) | 2020-11-11 | 2020-11-11 | Mechanical rotating part fault diagnosis method and device based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112395968B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115683687A (en) * | 2023-01-03 | 2023-02-03 | 成都大汇物联科技有限公司 | Dynamic and static rub-impact fault diagnosis method for hydroelectric mechanical equipment |
| WO2023206860A1 (en) * | 2022-04-27 | 2023-11-02 | 山东瑞美油气装备技术创新中心有限公司 | Method and apparatus for determining mechanical device fault |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170052060A1 (en) * | 2014-04-24 | 2017-02-23 | Alstom Transport Technologies | Method and system for automatically detecting faults in a rotating shaft |
| CN108106841A (en) * | 2017-12-21 | 2018-06-01 | 西安交通大学 | Epicyclic gearbox intelligent failure diagnosis method based on built-in encoder signal |
| CN108510153A (en) * | 2018-02-08 | 2018-09-07 | 同济大学 | A kind of multi-state rotary machinery fault diagnosis method |
| CN108681747A (en) * | 2018-05-11 | 2018-10-19 | 武汉理工大学 | Rotary machinery fault diagnosis based on deep learning and condition monitoring system and method |
| CN108830127A (en) * | 2018-03-22 | 2018-11-16 | 南京航空航天大学 | A kind of rotating machinery fault feature intelligent diagnostic method based on depth convolutional neural networks structure |
| CN109655259A (en) * | 2018-11-23 | 2019-04-19 | 华南理工大学 | Combined failure diagnostic method and device based on depth decoupling convolutional neural networks |
| CN110017991A (en) * | 2019-05-13 | 2019-07-16 | 山东大学 | Rolling bearing fault classification method and system based on spectrum kurtosis and neural network |
| CN111046583A (en) * | 2019-12-27 | 2020-04-21 | 中国铁道科学研究院集团有限公司通信信号研究所 | Switch machine fault diagnosis method based on DTW algorithm and ResNet network |
| CN111458142A (en) * | 2020-04-02 | 2020-07-28 | 苏州智传新自动化科技有限公司 | Sliding bearing fault diagnosis method based on generation of countermeasure network and convolutional neural network |
| CN111626345A (en) * | 2020-05-15 | 2020-09-04 | 北京航空航天大学 | Multi-stage deep convolution transfer learning fault diagnosis method between different bearing devices |
| CN111626994A (en) * | 2020-05-18 | 2020-09-04 | 江苏远望仪器集团有限公司 | Equipment fault defect diagnosis method based on improved U-Net neural network |
-
2020
- 2020-11-11 CN CN202011253582.6A patent/CN112395968B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170052060A1 (en) * | 2014-04-24 | 2017-02-23 | Alstom Transport Technologies | Method and system for automatically detecting faults in a rotating shaft |
| CN108106841A (en) * | 2017-12-21 | 2018-06-01 | 西安交通大学 | Epicyclic gearbox intelligent failure diagnosis method based on built-in encoder signal |
| CN108510153A (en) * | 2018-02-08 | 2018-09-07 | 同济大学 | A kind of multi-state rotary machinery fault diagnosis method |
| CN108830127A (en) * | 2018-03-22 | 2018-11-16 | 南京航空航天大学 | A kind of rotating machinery fault feature intelligent diagnostic method based on depth convolutional neural networks structure |
| CN108681747A (en) * | 2018-05-11 | 2018-10-19 | 武汉理工大学 | Rotary machinery fault diagnosis based on deep learning and condition monitoring system and method |
| CN109655259A (en) * | 2018-11-23 | 2019-04-19 | 华南理工大学 | Combined failure diagnostic method and device based on depth decoupling convolutional neural networks |
| CN110017991A (en) * | 2019-05-13 | 2019-07-16 | 山东大学 | Rolling bearing fault classification method and system based on spectrum kurtosis and neural network |
| CN111046583A (en) * | 2019-12-27 | 2020-04-21 | 中国铁道科学研究院集团有限公司通信信号研究所 | Switch machine fault diagnosis method based on DTW algorithm and ResNet network |
| CN111458142A (en) * | 2020-04-02 | 2020-07-28 | 苏州智传新自动化科技有限公司 | Sliding bearing fault diagnosis method based on generation of countermeasure network and convolutional neural network |
| CN111626345A (en) * | 2020-05-15 | 2020-09-04 | 北京航空航天大学 | Multi-stage deep convolution transfer learning fault diagnosis method between different bearing devices |
| CN111626994A (en) * | 2020-05-18 | 2020-09-04 | 江苏远望仪器集团有限公司 | Equipment fault defect diagnosis method based on improved U-Net neural network |
Non-Patent Citations (5)
| Title |
|---|
| SILE YANG ET.AL: "Bearing fault diagnosis of two-dimensional improved Att-CNN2D neural network based on Attention mechanism", 《2020 IEEE INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND INFORMATION SYSTEMS (ICAIIS)》 * |
| 王志坚: "《齿轮箱复合故障诊断方法研究》", 31 August 2017, 兵器工业出版社 * |
| 王新 等: "《机电设备故障诊断技术及应用》", 30 April 2014, 煤炭工业出版社 * |
| 邵思羽: "基于深度学习的旋转机械故障诊断方法研究", 《中国博士学位论文全文数据库 工程科技Ⅱ辑》 * |
| 黄文艺: "基于特征优化与自主学习的滚动轴承故障诊断与性能退化评估", 《中国博士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023206860A1 (en) * | 2022-04-27 | 2023-11-02 | 山东瑞美油气装备技术创新中心有限公司 | Method and apparatus for determining mechanical device fault |
| CN115683687A (en) * | 2023-01-03 | 2023-02-03 | 成都大汇物联科技有限公司 | Dynamic and static rub-impact fault diagnosis method for hydroelectric mechanical equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112395968B (en) | 2021-08-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112906473B (en) | Fault diagnosis method for rotary equipment | |
| CN112418013B (en) | Complex working condition bearing fault diagnosis method based on meta-learning under small sample | |
| CN110672343B (en) | Rotary machine fault diagnosis method based on multi-attention convolutional neural network | |
| Wang et al. | Intelligent rolling bearing fault diagnosis via vision ConvNet | |
| CN103033359B (en) | A kind of main transmission in wind generating set method for diagnosing faults of multiple features Multilateral Comprehensive Judge | |
| CN104729853B (en) | A kind of rolling bearing performance degradation assessment device and method | |
| CN107702922B (en) | Fault Diagnosis Method of Rolling Bearing Based on LCD and Stacked Autoencoder | |
| CN108426713A (en) | Rolling bearing Weak fault diagnostic method based on wavelet transformation and deep learning | |
| CN113255078A (en) | Bearing fault detection method and device under unbalanced sample condition | |
| CN110132554B (en) | A deep Laplacian self-encoding method for fault diagnosis of rotating machinery | |
| CN114235415B (en) | Wind driven generator variable pitch bearing fault diagnosis method and device based on neural network | |
| CN111160167A (en) | Spindle fault classification and identification method based on S-transform deep convolutional neural network | |
| CN112395968B (en) | Mechanical rotating part fault diagnosis method and device based on neural network | |
| CN109583386A (en) | A kind of intelligent rotating mechanical breakdown depth network characterization discrimination method | |
| CN104713728B (en) | Large slewing bearing residual life online prediction method based on multidimensional data drive | |
| CN106226074A (en) | Based on convolutional neural networks and the rotary machinery fault diagnosis method of small echo gray-scale map | |
| CN111855202B (en) | Gearbox fault diagnosis method and system | |
| CN114659790B (en) | A method for identifying faults of variable-speed wind power high-speed shaft bearings | |
| CN111878322B (en) | Wind Turbine Device | |
| CN114354184B (en) | A method and device for establishing a health warning model for a spindle of large rotary equipment based on deep learning | |
| CN111412114B (en) | Wind turbine generator impeller imbalance detection method based on stator current envelope spectrum | |
| CN114964780A (en) | Wind power bearing fault diagnosis method based on time-frequency domain convolutional network and deep forest | |
| CN116106009A (en) | Rolling bearing fault diagnosis method based on Wen Zhen information fusion | |
| CN212363649U (en) | A system for realizing a method for fault diagnosis of gearboxes | |
| CN113240022A (en) | Wind power gear box fault detection method of multi-scale single-classification convolutional network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |