CN112507785A - Rolling bearing fault analysis based on CNN and LSTM - Google Patents
Rolling bearing fault analysis based on CNN and LSTM Download PDFInfo
- Publication number
- CN112507785A CN112507785A CN202011205524.6A CN202011205524A CN112507785A CN 112507785 A CN112507785 A CN 112507785A CN 202011205524 A CN202011205524 A CN 202011205524A CN 112507785 A CN112507785 A CN 112507785A
- Authority
- CN
- China
- Prior art keywords
- time
- cnn
- layer
- input
- 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
- 238000005096 rolling process Methods 0.000 title claims abstract description 23
- 238000004458 analytical method Methods 0.000 title abstract description 15
- 238000000034 method Methods 0.000 claims abstract description 24
- 230000006870 function Effects 0.000 claims abstract description 16
- 230000004927 fusion Effects 0.000 claims abstract description 14
- 238000012360 testing method Methods 0.000 claims abstract description 5
- 230000015654 memory Effects 0.000 claims description 18
- 238000011176 pooling Methods 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000001228 spectrum Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000002474 experimental method Methods 0.000 claims description 4
- 238000012886 linear function Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 101000709114 Homo sapiens SAFB-like transcription modulator Proteins 0.000 claims 1
- 102100032664 SAFB-like transcription modulator Human genes 0.000 claims 1
- 238000007781 pre-processing Methods 0.000 claims 1
- 230000029058 respiratory gaseous exchange Effects 0.000 claims 1
- 230000000717 retained effect Effects 0.000 claims 1
- 238000001514 detection method Methods 0.000 abstract description 2
- 238000013527 convolutional neural network Methods 0.000 description 30
- 238000004590 computer program Methods 0.000 description 6
- 238000003745 diagnosis Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000013136 deep learning model Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000036391 respiratory frequency Effects 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000009760 electrical discharge machining Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000007943 implant Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
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
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—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
- G06F2218/08—Feature extraction
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Signal Processing (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
本申请公开了一种基于CNN和LSTM的滚动轴承故障分类方法。获取轴承振动数据并预处理，构建数据集。对处理后的振动数据使用巴特沃斯滤波器(Butterworth Filter)去噪声，并进行快速傅里叶变换(FFT)将预处理的时域信号转化为频域信号。利用CNN网络进行学习来获取时域图、频域图的图像特征，通过Add层进行图像特征融合。将CNN学习获得的图像特征输入到LSTM网络中，通过LSTM网络学习特征中包含的时序特征，通过一个全连接层和Softmax函数实现分类功能。用训练好的网络对测试样本进行故障分类，对滚动轴承微弱故障的早期检测分析具有重要的现实意义和实用价值。
The present application discloses a rolling bearing fault classification method based on CNN and LSTM. Acquire bearing vibration data and preprocess it to construct a dataset. Use Butterworth Filter to remove noise on the processed vibration data, and perform Fast Fourier Transform (FFT) to convert the preprocessed time domain signal into frequency domain signal. The CNN network is used for learning to obtain the image features of the time domain map and the frequency domain map, and the image feature fusion is performed through the Add layer. The image features learned by CNN are input into the LSTM network, the time series features contained in the features are learned through the LSTM network, and the classification function is realized through a fully connected layer and the Softmax function. Using the trained network to classify the test samples has important practical significance and practical value for the early detection and analysis of weak faults of rolling bearings.
Description
Technical Field
The application belongs to the technical field of machine learning and fault identification, and particularly relates to a rolling bearing fault analysis method of a deep learning model CNN and a deep learning model LSTM.
Background
The rolling bearing fault diagnosis technology is an important research subject in the field of mechanical fault diagnosis. The train is widely applied to public trips as a complex mechanical device. The mechanical system of the train breaks down, the operation of a traffic system is interfered, and the safety of personnel can be endangered in serious cases, so that huge economic loss is caused. The rolling bearing is a core part of the train. According to statistics, 30% of mechanical equipment faults are caused by bearing faults, and the method has important practical significance and practical value for early detection and analysis of weak faults of the rolling bearing. Different fault types have different signal characteristics, and the vibration signal can intuitively reflect the health state of the bearing, so the application is wide. The signal characteristic main packet expression forms are divided into two types: time domain and frequency domain, however, the failure characteristics in the vibration signal cannot be fully represented by the time domain analysis or the frequency domain analysis alone. In recent years, a Convolutional Neural Network (CNN) is widely used in the fields of image processing, voice recognition, failure analysis, and the like. In bearing fault diagnosis, one-dimensional time sequence data or a two-dimensional time-frequency graph is used as input, training is carried out through a neural network, and fault characteristics are extracted, so that the fault can be analyzed in a noisy working condition scene.
And Jing and the like learn characteristics from the frequency spectrum of the vibration signal by using CNN (CNN), and realize the health state monitoring of the gearbox. Qu and the like provide a one-dimensional convolutional neural network fault diagnosis algorithm, and self-adaptive feature extraction and fault diagnosis based on a deep network are realized. SE, etc. trains CNN from the frequency spectrum of the motor using frequency components through fast fourier transforms. Zhuang et al uses a Long Short-Term Memory (LSTM) network to extract time characteristics of time series data through a gate structure, thereby enhancing the generalization capability of the model. The integrated deep learning method for multi-bearing residual service life collaborative prediction is provided by combining time domain and frequency domain characteristics of Ren and the like, and a good prediction effect is obtained. Liu uses a CNN + LSTM model and is mapped into a service life index through an activation layer, and then the service life of the bearing is predicted.
Although the CNN model can sufficiently extract spatial features of data, it cannot extract temporal features of data. Under the condition of large data quantity, the LSTM network is difficult to extract the nonlinear features of the data, the data features are not extracted sufficiently, and the convergence speed is slow. And compared with a model taking the time domain features or the frequency domain features as input, the model combining the time domain features and the frequency domain features can show more excellent classification effect.
Disclosure of Invention
The invention aims to provide a rolling bearing fault analysis method based on CNN and LSTM. The method fully utilizes the spatial feature extraction capability of CNN and the time sequence feature learning capability of LSTM, fully extracts the relation between the vibration image and time dependence, and carries out classification recognition and fault judgment on the vibration feature of the rolling bearing through the full-connection layer and the softmax layer, thereby enhancing the classification precision. And compared with the traditional RNN, the LSTM solves the problem of gradient disappearance and can reduce the difficulty of model training.
According to an aspect of the present application, there is provided a CNN and LSTM-based bearing fault analysis method, including:
and acquiring bearing vibration data, and processing the respiratory frequency signal of the vibration data containing noise by adopting a Butterworth filter.
Wherein the Butterworth filter has the formula:
where n is the order of the filter, ωcThe frequency at which the amplitude drops to-3 db is taken for the cut-off frequency. The default order of a filter of the filter is 2, although a high-order Butterworth filter can realize clearer roll-off near a cut-off frequency, the high-order Butterworth filter can also cause serious signal distortion and influence the precision of a result, and experiments find that the performance of the filter with the order of 1 is better, namely a first-order Butterworth filter is used for processing vibration data containing noise;
converting the description of the preprocessed signal from the time domain to the frequency domain by a Fast Fourier Transform (FFT);
the FFT calculates the spectrum of the signal x (k) using discrete signals. Wherein the formula of the FFT is:
namely, the FFT algorithm can reduce the calculation amount of frequency domain conversion and improve the conversion speed, and the time complexity is o (nolg)2n)。
And taking the time domain graph and the frequency domain graph obtained after processing as input, respectively training through a preset CNN network, carrying out convolution operation on the data on the one-dimensional time axis by the CNN, and moving along the time t axis of the time domain signal and the time frequency graph to extract image characteristics. CNN performs feature extraction by 3 convolutional layers. The pooling layer is positioned behind each convolution layer to reduce the dimension of the feature map, wherein the time feature of the data is reserved by reducing the complexity of output and preventing overfitting of the data by using the maximum pooling operation;
wherein the maximum pooling layer calculation formula is:
w and d are the preset length and width of the maximum pooling filter.
The time domain graph and the frequency domain graph obtained by CNN learning are subjected to feature fusion through the add layer, the dimensionality of the image is not increased, the information amount under each dimension is increased, and the time feature can be reserved by performing feature fusion through the add layer. Taking the fusion characteristics as the input of a long-time memory network layer;
inputting the fusion characteristics obtained after the add layer characteristics are fused into a long-time memory network layer, connecting the long-time memory network layer and the short-time memory network layer in a serial mode, and taking data containing time sequence as input to obtain space-time characteristics; the long-time memory network layer is internally provided with 3 gates: forget gate, input gate, output gate, the update formula of every moment t gate is as follows:
forget door ftThe purpose of this is to let the LSTM network forget information that was previously useless:
ft＝σ(Wf·[ht-1,xt]+bf)
input door itThe role of (2) is to determine the input information of the LSTM network:
ct′＝tanh(Wc·[ht-1,xt]+bc)
ct＝ft*ct-1+iict′
output gate otThe role of (a) is to determine the outcome of the neuron:
ot＝σ(Wo[ht-1,xt]+bo)
ht＝ot*tanh(ct)
wherein, Wf、Wi、Wc、WoWeight matrix b of forgetting gate, input gate and output gatef、bi、bc、boIs its bias term; h ist-1The state of the hidden layer at the time t-1; σ is a logistic function with an output of (0, 1); x is the number oftIs the input vector at the time t; h ist-1The state of the previous moment; tanh is the activation function.
And taking the acquired space-time characteristics as input, mapping the space-time characteristics to a sample mark space in a full connection layer, and obtaining a classification probability result through subsequent softmax layer operation to identify and classify the rolling bearing faults.
Wherein the formula of the softmax function is:
denotes the sample vector x when there are K linear functionsTProbability of belonging to class j.
Drawings
Some specific embodiments of the present application will be described in detail hereinafter by way of illustration and not limitation with reference to the accompanying drawings. The same reference numbers in the drawings identify the same or similar elements or components. Those skilled in the art will appreciate that the drawings are not necessarily drawn to scale. In the drawings:
FIG. 1 is a schematic flow chart of a fault analysis method of a rolling bearing based on CNN and LSTM in the method of the present application;
fig. 2 is a schematic structural block diagram of a rolling bearing fault analysis method based on CNN and LSTM according to an embodiment of the present application.
Fig. 3 is a computing device provided in an embodiment of the present application.
Fig. 4 is a computer-readable storage medium according to an embodiment of the present application.
Detailed Description
Fig. 1 is a schematic flow chart of a CNN and LSTM-based rolling bearing fault analysis method according to an embodiment of the present application. Referring to fig. 1, a method and a system for analyzing a fault of a rolling bearing based on CNN and LSTM provided in an embodiment of the present application may include:
step S1: denoising the processed vibration data by using a Butterworth Filter (Butterworth Filter), and converting the preprocessed time domain signals into frequency domain signals by using Fast Fourier Transform (FFT);
step S2: learning by using a CNN (CNN network) to obtain image characteristics of a time domain graph and a frequency domain graph;
step S3: performing image feature fusion through the add layer;
step S4: inputting the fusion image characteristics obtained by the add layer into an LSTM network, and further learning the time sequence characteristics contained in the characteristics through the LSTM network;
step S5: the classification function is realized through a full connection layer and a Softmax function, and the trained network is used for carrying out fault classification on the test sample;
the invention aims to provide a rolling bearing fault analysis method based on CNN and LSTM. Firstly, noise filtering is carried out on a data set, vibration data are represented by a time domain graph and a time frequency graph, and feature extraction is better carried out. The vibration characteristics of the time-frequency graph and the time-domain graph are respectively extracted by utilizing the space characteristic extraction capability of the CNN, the characteristics are fused, the relation between the vibration characteristics and the time dependence is fully extracted by utilizing the time sequence characteristic learning capability of the LSTM, the vibration characteristics of the rolling bearing are classified and identified and the fault is judged through the full connection layer and the Softmax layer, and the classification precision is enhanced.
The source of an experimental data set adopted by the method is CWRU (Kaiser Sichu university bearing data center), and the data set is a rolling bearing fault data set which is most widely used internationally at present. The data set records actual test conditions for the motor and bearing fault conditions, using Electrical Discharge Machining (EDM) techniques to implant faults into the motor bearings. Faults ranging from 0.007 inches to 0.040 inches in diameter were introduced on the bearing inner race, the rolling elements and the bearing outer race, respectively. The failed bearing was reinstalled into the test motor and the bearing experiment recorded vibration data of 0 to 3 horsepower (motor speed 1797 to 1720RPM) at 12,000 samples/second and 48,000 samples/second.
And S1, acquiring bearing vibration data, and processing the respiratory frequency signal of the vibration data containing noise by adopting a Butterworth filter.
Wherein the Butterworth filter has the formula:
where n is the order of the filter, ωcThe frequency at which the amplitude drops to-3 db is taken for the cut-off frequency. The default order of the filter is 2, although the high-order Butterworth filter can realize clearer roll-off near the cut-off frequency, the high-order Butterworth filter can cause serious signal distortion and influence the precision of the result, and experiments find that the performance of the filter with the order of 1 is better, namely the filter with the first orderThe Butterworth filter processes the vibration data containing noise;
converting the description of the preprocessed signal from the time domain to the frequency domain by a Fast Fourier Transform (FFT);
the FFT calculates the spectrum of the signal x (k) using discrete signals. Where the FFT can be expressed as:
namely, the FFT algorithm can reduce the calculation amount of frequency domain conversion and improve the conversion speed, and the time complexity is o (nlog)2n)。
And S2, taking the processed time domain graph and frequency domain graph as input, respectively training through a preset CNN network, carrying out convolution operation on the data on the one-dimensional time axis by the CNN, and moving along the time t axis of the time domain signal and the time frequency graph to extract image characteristics. CNN performs feature extraction by 3 convolutional layers. The pooling layer is positioned behind each convolution layer to reduce the dimension of the feature map, wherein the time feature of the data is reserved by reducing the complexity of output and preventing overfitting of the data by using the maximum pooling operation;
wherein the maximum pooling layer calculation formula is:
w and d are the preset length and width of the maximum pooling filter.
S3, carrying out feature fusion on the time domain graph and the frequency domain graph obtained by CNN learning through the add layer, wherein the dimensionality of the image is not increased, the information content under each dimension is increased, and the time feature can be reserved by carrying out feature fusion through the add layer. Taking the fusion characteristics as the input of a long-time memory network layer;
s4, inputting the fusion characteristics obtained by fusing the characteristics of the Add layers into a long-short time memory network layer, connecting the layers of the long-short time memory network in a series mode, and taking data containing time sequence as input to obtain space-time characteristics; the long-time memory network layer is internally provided with 3 gates: forget gate, input gate, output gate, the update formula of every moment t gate is as follows:
forget door ftThe purpose of this is to let the LSTM network forget information that was previously useless:
ft＝σ(Wf·[ht-1,xt]+bf)
input door itThe role of (2) is to determine the input information of the LSTM network:
ct′＝tanh(Wc·[ht-1,xt]+bc)
ct＝ft*ct-1+iict′
output gate otThe role of (a) is to determine the outcome of the neuron:
ot＝σ(Wo[ht-1,xt]+bo)
ht＝ot*tanh(ct)
wherein, Wf、Wi、Wc、WoWeight matrix b of forgetting gate, input gate and output gatef、bi、bc、boIs its bias term; h ist-1The state of the hidden layer at the time t-1; σ is a logistic function with an output of (0, 1); x is the number oftIs the input vector at the time t; h ist-1The state of the previous moment; tanh is the activation function.
And S5, taking the acquired space-time characteristics as input, mapping the space-time characteristics to a sample mark space in a full connection layer, and obtaining a classification probability result through subsequent softmax layer operation to identify and classify the rolling bearing faults.
Wherein the formula of the softmax function is:
denotes the sample vector x when there are K linear functionsTProbability of belonging to class j.
Fig. 2 is a schematic structural block diagram of a rolling bearing fault analysis method based on CNN and LSTM according to an embodiment of the present application.
The embodiment of the present application in fig. 3 also provides a computing device comprising a memory 320, a processor 310 and a computer program stored in said memory 320 and executable by said processor 310, the computer program being stored in a space 330 for program code in the memory 320, the computer program, when executed by the processor 310, implementing the method steps 331 for performing any of the methods according to the present invention.
The embodiment of the application in fig. 4 also provides a computer-readable storage medium. The computer readable storage medium comprises a storage unit for program code provided with a program 331' for performing the steps of the method according to the invention, which program is executed by a processor.
The embodiment of the application also provides a computer program product containing instructions. Which, when run on a computer, causes the computer to carry out the steps of the method according to the invention.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed by a computer, cause the computer to perform, in whole or in part, the procedures or functions described in accordance with the embodiments of the application. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
Those of skill would further appreciate that the various illustrative components and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
It will be understood by those skilled in the art that all or part of the steps in the method for implementing the above embodiments may be implemented by a program, and the program may be stored in a computer-readable storage medium, where the storage medium is a non-transitory medium, such as a random access memory, a read only memory, a flash memory, a hard disk, a solid state disk, a magnetic tape (magnetic tape), a floppy disk (floppy disk), an optical disk (optical disk), and any combination thereof.
The above description is only for the preferred embodiment of the present application, but the scope of the present application 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 application should be covered within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (8)
1. A rolling bearing fault classification method based on CNN and SLTM comprises the following steps:
acquiring and preprocessing bearing vibration data to construct a data set;
denoising the processed vibration data by using a Butterworth Filter (Butterworth Filter), and converting the preprocessed time domain signals into frequency domain signals by using Fast Fourier Transform (FFT);
the CNN network is used for learning to obtain the image characteristics of a time domain graph and a frequency domain graph, and the image characteristics are fused through an Add layer;
inputting image features obtained by CNN learning into an LSTM network, learning time sequence features contained in the features through the LSTM network, and realizing a classification function through a full connection layer and a Softmax function;
and carrying out fault classification on the test sample by using the trained network.
2. The method of claim 1, wherein the breathing rate signal is processed using a Butterworth filter (Butterworth filter) on noisy vibration data.
Wherein the Butterworth filter has the formula:
where n is the order of the filter, ωcThe frequency at which the amplitude drops to-3 db is taken for the cut-off frequency. The default order of the filter is 2, although a high-order Butterworth filter can realize clearer roll-off near a cut-off frequency, the high-order Butterworth filter can cause serious signal distortion and influence the precision of a result, and experiments find that the performance of the filter with the order of 1 is better, namely the first-order Butterworth filter is used for processing vibration data containing noise.
3. The method of claim 1, wherein the description of the preprocessed signal is converted from the time domain to the frequency domain by a Fast Fourier Transform (FFT).
The FFT calculates the spectrum of the signal x (k) using discrete signals. Wherein the formula of the FFT is:
namely, the FFT algorithm can reduce the calculation amount of frequency domain conversion and improve the conversion speed, and the time complexity is o (nlog)2n)。
4. The method according to claim 1, wherein the time domain graph and the frequency domain graph obtained after the processing are used as input and are respectively trained through a preset CNN network, and the CNN only performs convolution operation on data on a one-dimensional time axis and moves along the time t axis of the time domain signal and the time frequency graph to extract image features. CNN performs feature extraction by 3 convolutional layers. The pooling layer is positioned behind each convolution layer to reduce the dimension of the feature map, wherein the time feature of the data is reserved by reducing the complexity of output and preventing overfitting of the data by using the maximum pooling operation;
wherein the maximum pooling layer calculation formula is:
w and d are the preset length and width of the maximum pooling filter.
5. The method according to claim 1 and claim 4, wherein the time domain graph and the frequency domain graph obtained by CNN learning are subjected to feature fusion through an add layer, the dimensionality of the image is not increased, the information amount under each dimension is increased, and the time feature can be retained by performing feature fusion through the add layer. And taking the fusion characteristics as the input of the long-time and short-time memory network layer.
6. The method according to claim 1 and claim 5, characterized in that the fused feature obtained by fusing the add layer features is input to the long-short term memory network layer, the layers of the long-short term memory network are connected in series, and the time-space feature is obtained by using the data containing time sequence as input; the long-time memory network layer is internally provided with 3 gates: forget gate, input gate, output gate, the update formula of every moment t gate is as follows:
forget door ftThe purpose of this is to let the LSTM network forget information that was previously useless:
ft＝σ(Wf·[ht-1,xt]+bf)
input door itThe role of (2) is to determine the input information of the LSTM network:
ct′＝tanh(Wc·[ht-1,xt]+bc)
ct＝ft*ct-1+iict′
output gate otThe role of (a) is to determine the outcome of the neuron:
ot＝σ(Wo[ht-1,xt]+bo)
ht＝ot*tanh(ct)
wherein, Wf、Wi、Wc、WoWeight matrix b of forgetting gate, input gate and output gatef、bi、bc、boIs its bias term; h ist-1The state of the hidden layer at the time t-1; σ is a logistic function with an output of (0, 1); x is the number oftIs the input vector at the time t; h ist-1The state of the previous moment; tanh is the activation function.
7. The method according to claim 1, characterized in that the acquired spatiotemporal features are used as input, mapped to a sample mark space in a full connection layer, and a classification probability result is obtained through subsequent softmax layer operation, so as to identify and classify the rolling bearing faults.
Wherein the formula of the softmax function is:
denotes the sample vector x when there are k linear functionsTProbability of belonging to class j.
8. The method of any one of claims 1 to 7, wherein the data set used in the method is a CWRU (university of Kaiser storage bearing data centre) data set in a laboratory environment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011205524.6A CN112507785A (en) | 2020-11-02 | 2020-11-02 | Rolling bearing fault analysis based on CNN and LSTM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011205524.6A CN112507785A (en) | 2020-11-02 | 2020-11-02 | Rolling bearing fault analysis based on CNN and LSTM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112507785A true CN112507785A (en) | 2021-03-16 |
Family
ID=74954857
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011205524.6A Pending CN112507785A (en) | 2020-11-02 | 2020-11-02 | Rolling bearing fault analysis based on CNN and LSTM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112507785A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113074940A (en) * | 2021-03-18 | 2021-07-06 | 昆明理工大学 | Rolling bearing health state estimation system and method based on OS-ELM |
| CN113469281A (en) * | 2021-07-22 | 2021-10-01 | 西北工业大学 | Industrial gear box multi-source information fusion fault diagnosis method |
| CN113536989A (en) * | 2021-06-29 | 2021-10-22 | 广州博通信息技术有限公司 | Refrigerator frosting monitoring method and system based on camera video frame-by-frame analysis |
| CN113887702A (en) * | 2021-09-10 | 2022-01-04 | 哈尔滨工业大学 | Early fault detection method of industrial robot harmonic reducer based on WLCTD and CNN-LSTM |
| CN113899809A (en) * | 2021-08-20 | 2022-01-07 | 中海石油技术检测有限公司 | In-pipeline detector positioning method based on CNN classification and RNN prediction |
| CN114037834A (en) * | 2021-12-01 | 2022-02-11 | 清华大学 | A semantic segmentation method and device based on vibration signal and RGB image fusion |
| CN114266269A (en) * | 2021-11-12 | 2022-04-01 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | Bearing fault diagnosis method and system, storage medium and equipment |
| CN114881086A (en) * | 2022-05-21 | 2022-08-09 | 重庆大学 | Intelligent quality identification method of paired bearing based on attention LSTM |
| CN114943256A (en) * | 2022-05-31 | 2022-08-26 | 常州大学 | Partial discharge identification method and device based on time-frequency characteristics and improved CNN |
| CN115392315A (en) * | 2022-08-31 | 2022-11-25 | 济南永信新材料科技有限公司 | Gearbox fault detection method based on transferable features |
| WO2023103268A1 (en) * | 2021-12-10 | 2023-06-15 | 烟台杰瑞石油服务集团股份有限公司 | Pump valve fault detection method |
| US11815087B2 (en) | 2021-09-27 | 2023-11-14 | Yantai Jereh Petroleum Equipment & Technologies Co., Ltd. | Automatic system and method for disassembly and assembly of plunger pumps |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108344564A (en) * | 2017-12-25 | 2018-07-31 | 北京信息科技大学 | A kind of state recognition of main shaft features Testbed and prediction technique based on deep learning |
| CN108764601A (en) * | 2018-04-03 | 2018-11-06 | 哈尔滨工业大学 | A kind of monitoring structural health conditions abnormal data diagnostic method based on computer vision and depth learning technology |
| CN109001557A (en) * | 2018-06-11 | 2018-12-14 | 西北工业大学 | A kind of aircraft utilities system fault recognition method based on random convolutional neural networks |
| CN109726524A (en) * | 2019-03-01 | 2019-05-07 | 哈尔滨理工大学 | A prediction method for the remaining service life of rolling bearings based on CNN and LSTM |
| CN110136745A (en) * | 2019-05-08 | 2019-08-16 | 西北工业大学 | A car horn recognition method based on convolutional neural network |
| CN110297479A (en) * | 2019-05-13 | 2019-10-01 | 国网浙江省电力有限公司紧水滩水力发电厂 | A kind of Fault Diagnosis Method of Hydro-generating Unit based on the fusion of convolutional neural networks information |
| CN110398369A (en) * | 2019-08-15 | 2019-11-01 | 贵州大学 | A Rolling Bearing Fault Diagnosis Method Based on Fusion of 1-DCNN and LSTM |
| CN111310589A (en) * | 2020-01-20 | 2020-06-19 | 河北科技大学 | A fault diagnosis method, fault diagnosis device and terminal for a mechanical system |
| CN111651937A (en) * | 2020-06-03 | 2020-09-11 | 苏州大学 | Intra-class adaptive bearing fault diagnosis method under variable operating conditions |
-
2020
- 2020-11-02 CN CN202011205524.6A patent/CN112507785A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108344564A (en) * | 2017-12-25 | 2018-07-31 | 北京信息科技大学 | A kind of state recognition of main shaft features Testbed and prediction technique based on deep learning |
| CN108764601A (en) * | 2018-04-03 | 2018-11-06 | 哈尔滨工业大学 | A kind of monitoring structural health conditions abnormal data diagnostic method based on computer vision and depth learning technology |
| CN109001557A (en) * | 2018-06-11 | 2018-12-14 | 西北工业大学 | A kind of aircraft utilities system fault recognition method based on random convolutional neural networks |
| CN109726524A (en) * | 2019-03-01 | 2019-05-07 | 哈尔滨理工大学 | A prediction method for the remaining service life of rolling bearings based on CNN and LSTM |
| CN110136745A (en) * | 2019-05-08 | 2019-08-16 | 西北工业大学 | A car horn recognition method based on convolutional neural network |
| CN110297479A (en) * | 2019-05-13 | 2019-10-01 | 国网浙江省电力有限公司紧水滩水力发电厂 | A kind of Fault Diagnosis Method of Hydro-generating Unit based on the fusion of convolutional neural networks information |
| CN110398369A (en) * | 2019-08-15 | 2019-11-01 | 贵州大学 | A Rolling Bearing Fault Diagnosis Method Based on Fusion of 1-DCNN and LSTM |
| CN111310589A (en) * | 2020-01-20 | 2020-06-19 | 河北科技大学 | A fault diagnosis method, fault diagnosis device and terminal for a mechanical system |
| CN111651937A (en) * | 2020-06-03 | 2020-09-11 | 苏州大学 | Intra-class adaptive bearing fault diagnosis method under variable operating conditions |
Non-Patent Citations (1)
| Title |
|---|
| 曾大懿, 杨基宏等: ""基于并行多通道卷积长短时记忆网络的轴承寿命 预测方法"", 《中国机械工程》, vol. 31, no. 20, 29 June 2020 (2020-06-29), pages 2454 - 2461 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113074940A (en) * | 2021-03-18 | 2021-07-06 | 昆明理工大学 | Rolling bearing health state estimation system and method based on OS-ELM |
| CN113536989A (en) * | 2021-06-29 | 2021-10-22 | 广州博通信息技术有限公司 | Refrigerator frosting monitoring method and system based on camera video frame-by-frame analysis |
| CN113469281A (en) * | 2021-07-22 | 2021-10-01 | 西北工业大学 | Industrial gear box multi-source information fusion fault diagnosis method |
| CN113469281B (en) * | 2021-07-22 | 2023-11-24 | 西北工业大学 | Industrial gearbox multisource information fusion fault diagnosis method |
| CN113899809A (en) * | 2021-08-20 | 2022-01-07 | 中海石油技术检测有限公司 | In-pipeline detector positioning method based on CNN classification and RNN prediction |
| CN113899809B (en) * | 2021-08-20 | 2024-02-27 | 中海石油技术检测有限公司 | In-pipeline detector positioning method based on CNN classification and RNN prediction |
| CN113887702A (en) * | 2021-09-10 | 2022-01-04 | 哈尔滨工业大学 | Early fault detection method of industrial robot harmonic reducer based on WLCTD and CNN-LSTM |
| CN113887702B (en) * | 2021-09-10 | 2024-06-04 | 哈尔滨工业大学 | Early fault detection method for industrial robot harmonic reducer based on WLCTD and CNN-LSTM |
| US11815087B2 (en) | 2021-09-27 | 2023-11-14 | Yantai Jereh Petroleum Equipment & Technologies Co., Ltd. | Automatic system and method for disassembly and assembly of plunger pumps |
| CN114266269A (en) * | 2021-11-12 | 2022-04-01 | 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) | Bearing fault diagnosis method and system, storage medium and equipment |
| CN114037834A (en) * | 2021-12-01 | 2022-02-11 | 清华大学 | A semantic segmentation method and device based on vibration signal and RGB image fusion |
| WO2023103268A1 (en) * | 2021-12-10 | 2023-06-15 | 烟台杰瑞石油服务集团股份有限公司 | Pump valve fault detection method |
| CN114881086B (en) * | 2022-05-21 | 2023-08-11 | 重庆大学 | Intelligent Quality Recognition Method for Paired Bearings Based on Attention LSTM |
| CN114881086A (en) * | 2022-05-21 | 2022-08-09 | 重庆大学 | Intelligent quality identification method of paired bearing based on attention LSTM |
| CN114943256A (en) * | 2022-05-31 | 2022-08-26 | 常州大学 | Partial discharge identification method and device based on time-frequency characteristics and improved CNN |
| CN115392315A (en) * | 2022-08-31 | 2022-11-25 | 济南永信新材料科技有限公司 | Gearbox fault detection method based on transferable features |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112507785A (en) | Rolling bearing fault analysis based on CNN and LSTM | |
| CN113344295B (en) | Method, system and medium for predicting residual life of equipment based on industrial big data | |
| CN108648748B (en) | Acoustic event detection method in hospital noise environment | |
| Wang et al. | Bearing fault diagnosis method based on Hilbert envelope spectrum and deep belief network | |
| CN110307982B (en) | Bearing fault classification method based on CNN and Adaboost | |
| CN113176092A (en) | Motor bearing fault diagnosis method based on data fusion and improved experience wavelet transform | |
| CN112926644A (en) | Method and system for predicting remaining service life of rolling bearing | |
| CN115798516B (en) | Migratable end-to-end acoustic signal diagnosis method and system | |
| Hong et al. | Supervised-learning-based intelligent fault diagnosis for mechanical equipment | |
| Wang et al. | Multi-scale attention mechanism residual neural network for fault diagnosis of rolling bearings | |
| Zhao et al. | A novel deep fuzzy clustering neural network model and its application in rolling bearing fault recognition | |
| CN116977708B (en) | Bearing intelligent diagnosis method and system based on self-adaptive aggregation visual view | |
| CN114638256A (en) | Transformer fault detection method and system based on sound wave signals and attention network | |
| CN116861343A (en) | Bearing fault diagnosis method | |
| Zhang et al. | Neural network in sports cluster analysis | |
| CN116484176A (en) | A bearing fault diagnosis method, system and storage medium based on ultrawavelet | |
| Du | Application of improved smote and xgboost algorithm in the analysis of psychological stress test for college students | |
| CN113990303B (en) | Environmental sound identification method based on multi-resolution cavity depth separable convolution network | |
| CN113569989B (en) | TI-TSDCN model construction method for stage equipment fault diagnosis | |
| CN114781450A (en) | A state identification method of train rolling bearing based on MOMEDA-MIA-CNN with parameter optimization | |
| He et al. | TFA-CLSTMNN: Novel convolutional network for sound-based diagnosis of COVID-19 | |
| CN115436057A (en) | A Contrastive Capsule Network Method for Intelligent Diagnosis of Bearings in High Noise Downstream Wheelsets | |
| CN113865870A (en) | Bearing fault edge diagnosis method and system based on wavelet improved MobileNet network | |
| CN108763728B (en) | Mechanical fault diagnosis method using parallel deep neural network hierarchical feature extraction | |
| Meng et al. | Rolling bearing fault diagnosis method based on MCMF and SAIMFE |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20210316 |
|
| WD01 | Invention patent application deemed withdrawn after publication |