CN114624027A - Bearing fault diagnosis method based on multi-input CNN - Google Patents

Bearing fault diagnosis method based on multi-input CNN Download PDF

Info

Publication number
CN114624027A
CN114624027A CN202210254905.6A CN202210254905A CN114624027A CN 114624027 A CN114624027 A CN 114624027A CN 202210254905 A CN202210254905 A CN 202210254905A CN 114624027 A CN114624027 A CN 114624027A
Authority
CN
China
Prior art keywords
feature
fault diagnosis
samples
bearing fault
diagnosis method
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
Application number
CN202210254905.6A
Other languages
Chinese (zh)
Inventor
吴起
凌六一
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Anhui University of Science and Technology
Original Assignee
Anhui University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Anhui University of Science and Technology filed Critical Anhui University of Science and Technology
Priority to CN202210254905.6A priority Critical patent/CN114624027A/en
Publication of CN114624027A publication Critical patent/CN114624027A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00—Testing of machine parts
    • G01M13/04—Bearings
    • G01M13/045—Acoustic or vibration analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/2431—Multiple classes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • 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)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Acoustics & Sound (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention discloses a bearing fault diagnosis method based on multi-input CNN. The method comprises the following steps: firstly, an accelerometer is used for collecting a bearing vibration signal; then converting the vibration signals into time-frequency graphs through short-time Fourier transform and wavelet transform respectively, labeling the 5 types of samples respectively, and dividing the samples into a data set and a test set; sending the training set into the established CNN model, continuously carrying out forward propagation and backward propagation to update parameters until a preset total iteration number is reached, and storing the model obtained at the moment; then, testing the diagnostic performance of the bearing fault diagnosis model by using the test set; and finally, carrying out fault analysis by using the model, thereby realizing fault diagnosis of the bearing. The bearing fault diagnosis method converts the original vibration data into time-frequency graphs through short-time Fourier transform and continuous wavelet transform respectively, and finally inputs the time-frequency graphs into a network simultaneously, so that the original signal characteristics are more fully extracted, and the accurate diagnosis of the bearing fault is realized.

Description

Bearing fault diagnosis method based on multi-input CNN
Technical Field
The invention relates to the technical field of rolling bearing fault detection, in particular to a bearing fault diagnosis method based on multi-input CNN.
Background
The rolling bearing is a core component in mechanical equipment, and the fault of the bearing can cause serious casualties and economic losses, so that the fault diagnosis of the bearing and the smooth operation guarantee of the bearing are essential steps for maintaining the safe and stable operation of modern mechanical equipment.
The traditional bearing fault diagnosis method focuses on extracting features manually, but the processes are too tedious and lack of intelligence, and the features are difficult to extract manually under the condition of extremely large data volume. With the rise of big data driven artificial intelligence technology, deep learning is widely applied in the fields of feature learning, pattern recognition, data mining and the like by virtue of strong learning ability. Fault diagnosis has also made new research advances by the application of deep learning techniques. As a typical technique of deep learning, the study of convolutional neural network on fault diagnosis has advanced to some extent.
Because the actually measured bearing fault signals usually contain strong noise, most models usually use a time-frequency conversion method to convert one-dimensional vibration signals into two-dimensional time sequence signals in order to better extract signal characteristics, however, most of the models cannot sufficiently extract the characteristics of original signals only by using a time-frequency conversion method, and have the risk of losing important information, which results in insufficient training precision.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a bearing fault diagnosis method based on multi-input CNN, which can accurately convert an original vibration signal into a time domain signal by using two time-frequency conversion methods of Continuous Wavelet Transform (CWT) and Short Time Fourier Transform (STFT), so that the original signal characteristics can be fully extracted, and the training precision of a network is improved.
In order to achieve the purpose, the invention adopts the technical scheme that:
a bearing fault diagnosis method based on the change of multiple input CNN comprises the following steps:
the method comprises the following steps:
the method comprises the following steps that vibration signals collected by an accelerometer are one-dimensional signals, the vibration signals are converted into a time-frequency diagram through short-time Fourier transform and continuous wavelet transform, a sample label is given, and a training set and a test set are divided;
step two:
building a convolutional neural network model, comprising: the system comprises a feature extraction layer, a feature superposition layer and a feature classification layer, wherein the feature extraction layer comprises two channels, convolution kernels with different sizes are respectively used for extracting time-frequency graph features obtained by different time-frequency conversion methods, the feature superposition layer superposes two output feature graphs of the feature extraction layer on channel dimensionality, and the feature classification layer classifies fault categories through a Softmax classifier;
step three:
initializing parameters of a neural network, initializing a convolution kernel, bias, weight and the like into a random number, and then setting hyper-parameters such as learning rate, iteration times and the like.
Step four:
training the constructed network by using samples in the training set, and updating network parameters through forward propagation and backward propagation until the model reaches a preset period;
step five:
and inputting the test set sample into the trained network to obtain a diagnosis result.
In the first step, data is preprocessed, and a data set and a test set are divided in a mode that:
the data is divided into the number of sample points acquired for one revolution and the amplitude of the data is normalized to between-1, 1. Since the data set used had 5 classes, each of which included 500 samples, there were 2500 samples. Training data and test data were randomly selected, in a ratio of 4: 1.
the second step is specifically as follows:
(2-1) the feature extraction layer is composed of two channels, the convolution kernel sizes and step sizes in the two channels are different, for a given sample
Figure BDA0003544605190000021
S represents a sample, L represents a corresponding label, and the working process of the convolutional layer is described by the following formula:
Fi＝CONV(S*γiconv)
where CONV (x) represents the convolution operation of CNNWork, gammaiconv denotes the parameter of the i-th convolutional layer, FiRepresenting the extracted spatial feature information.
(2-2): the feature superposition layer superposes the two output feature maps of the feature extraction layer on the channel dimension, and finally outputs a feature map with the channel number of 28 and the length and width of 32 multiplied by 32.
(2-3): the characteristic classification layer classifies fault categories through a Softmax classifier, and the process of the Softmax classifier is described by the following formula:
Figure BDA0003544605190000031
wherein denotes the Softmax classifier,
Figure BDA0003544605190000032
softmax parameter, SC, representing class jjRepresenting the predicted probability distributions of different classes.
The third step is specifically as follows:
the hyper-parameter settings of the model are as follows: the learning rate was set to 0.005; the number of training rounds is 50; randomly disordering 64 samples in each round and extracting the samples into a batch; the optimization algorithm selects an Adam optimization algorithm.
The invention has the beneficial effects that:
the invention discloses a bearing fault diagnosis method based on multi-input CNN, which converts a one-dimensional original vibration signal into a time-frequency signal through short-time Fourier transform (STFT) and Continuous Wavelet Transform (CWT) respectively and stores the time-frequency signal as an image as the input of a model, realizes bearing fault diagnosis by adjusting the hyper-parameters of the model, and can more fully extract the characteristics of the original vibration signal compared with other models using a time-frequency conversion method.
Drawings
FIG. 1 is a flow chart of the algorithm of the method.
FIG. 2 is a time-frequency diagram after a short-time Fourier transform.
Fig. 3 is a time-frequency diagram after continuous wavelet transform.
Fig. 4 is a network structure diagram of the method.
Detailed Description
The invention is described in detail below with reference to the attached drawing figures:
referring to fig. 1, a bearing fault diagnosis method based on multi-input CNN of the present invention includes:
(1) the vibration signal collected by the accelerometer is a one-dimensional signal, and the vibration signal is converted into a time-frequency diagram through short-time Fourier transform and continuous wavelet transform, wherein fig. 2 is the time-frequency diagram after short-time Fourier transform under a certain fault, and fig. 3 is the time-frequency diagram after continuous wavelet transform under a certain fault. Giving a time-frequency diagram label, and dividing a training set and a test set;
in this example, the southeast university gearbox dataset was used directly to describe the embodied steps of the invention.
The data was also from the southeast university gearbox dataset, bearing failure was tested by manually setting the constant speed stage. Data for 5 different fault conditions were selected.
Labeling the collected bearing vibration data according to the damage types of the bearing vibration data, and performing labeling treatment according to the following table I;
table sample data set
Figure BDA0003544605190000041
The table above divides the loss state into 5 categories, where each category includes 500 samples, thus there are 2500 samples in total. Training data and test data were randomly selected, in a ratio of 4: 1.
(2) building a neural network model based on the improved CNN, and using labeled bearing vibration data to train the model, wherein the model comprises the following steps: the device comprises a feature extraction layer, a feature superposition layer and a feature classification layer.
(2-1) the feature extraction layer consists of two channels, the convolution kernel sizes and step sizes in the two channels are different, for a given sample
Figure BDA0003544605190000042
S represents a sample, L represents a corresponding label, and the process is described by the following formula:
Fi＝CONV(S*γIconv)
where CONV (x) denotes the convolution operation of CNN, γiconv denotes the parameter of the ith convolution layer, FiRepresenting the extracted spatial feature information.
(2-2) the feature superposition layer superposes the two output feature maps of the feature extraction layer on the channel dimension, and finally outputs a feature map with the channel number of 28 and the length and width of 32 multiplied by 32.
(2-3): the characteristic classification layer classifies fault categories through a Softmax classifier, and the process of the Softmax classifier is described by the following formula:
Figure BDA0003544605190000051
wherein denotes the Softmax classifier,
Figure BDA0003544605190000052
softmax parameter, SC, representing class jjRepresenting the predicted probability distributions of different classes.
(3) The training samples are transmitted into the network according to the quantity set in advance for forward propagation, a predicted value is obtained after the training samples pass through a classification layer, the loss of the predicted value and the real value is calculated, the error is propagated reversely by utilizing an optimization algorithm, and the network parameters are updated. The hyper-parameter settings of the model are as follows: the learning rate was set to 0.005; the number of training rounds is 50; randomly disordering 64 samples in each round and extracting the samples into a batch; the optimization algorithm selects an Adam optimization algorithm.
(4) And inputting the test set into the trained model to obtain a diagnosis result, and completing bearing fault diagnosis.

Claims (6)

1. The bearing fault diagnosis method based on the multi-input CNN is characterized by comprising the following steps:
a. the method comprises the following steps that a vibration signal acquired by an accelerometer is a one-dimensional signal, the vibration signal is converted into a time-frequency graph through short-time Fourier transform and continuous wavelet transform, a sample label is given, and a training set and a testing set are divided;
b. building a convolutional neural network model, comprising: the system comprises a feature extraction layer, a feature superposition layer and a feature classification layer, wherein the feature extraction layer comprises two channels, convolution kernels with different sizes are respectively used for extracting time-frequency graph features obtained by different time-frequency conversion methods, the feature superposition layer superposes two output feature graphs of the feature extraction layer on channel dimensionality, and the feature classification layer classifies fault categories through a Softmax classifier;
c. initializing parameters of a neural network, initializing a convolution kernel, bias, weight and the like into a random number, and then setting hyper-parameters such as learning rate, iteration times and the like. (ii) a
d. Training the constructed network by using samples in the training set, and updating network parameters through forward propagation and backward propagation until the model reaches a preset period;
e. and inputting the test set sample into the trained network to obtain a diagnosis result.
2. The multi-input CNN-based bearing fault diagnosis method of claim 1, wherein in step a, the data is preprocessed, and the training set and the test set are divided as follows:
the data is divided into the number of sample points acquired for one revolution and the amplitude of the data is normalized to between-1, 1. Since the data set used had 5 classes, each of which included 500 samples, there were 2500 samples. Training data and test data were randomly selected, in a ratio of 4: 1.
3. the multi-input CNN-based bearing fault diagnosis method according to claim 1, wherein in step b, the feature extraction layer is specifically as follows:
for a given sample
Figure FDA0003544605180000011
S represents a sample, L represents a corresponding label, and the process is represented by the following formulaThe following steps are described: fi＝CONV(S*γIconv) (1)
Where CONV (x) denotes the convolution operation of CNN, γiconv denotes the parameter of the i-th convolutional layer, FiRepresenting the extracted spatial feature information.
4. The multi-input CNN-based bearing fault diagnosis method of claim 1, wherein in step b, the feature superposition layers are specifically as follows:
the feature superposition layer superposes the two output feature maps of the feature extraction layer on the channel dimension, and finally outputs a feature map with the channel number of 28 and the length and width of 32 multiplied by 32.
5. The multi-input CNN-based bearing fault diagnosis method as claimed in claim 1, wherein in step b, the feature classification layers are specifically as follows:
the characteristic classification layer classifies fault categories through a Softmax classifier, and the process of the Softmax classifier is described by the following formula:
Figure FDA0003544605180000012
wherein denotes the Softmax classifier,
Figure FDA0003544605180000013
softmax parameter, SC, representing class jjRepresenting the predicted probability distributions of different classes.
6. The multi-input CNN-based bearing fault diagnosis method of claim 1, wherein in step c, the constructed network is trained by using samples in a training set until a preset period of the model is reached, specifically as follows:
the learning rate is set to 0.005; the number of training rounds is 50; randomly disordering 64 samples in each round and extracting the samples into a batch; the optimization algorithm selects an Adam optimization algorithm.
CN202210254905.6A 2022-03-14 2022-03-14 Bearing fault diagnosis method based on multi-input CNN Pending CN114624027A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210254905.6A CN114624027A (en) 2022-03-14 2022-03-14 Bearing fault diagnosis method based on multi-input CNN

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210254905.6A CN114624027A (en) 2022-03-14 2022-03-14 Bearing fault diagnosis method based on multi-input CNN

Publications (1)

Publication Number Publication Date
CN114624027A true CN114624027A (en) 2022-06-14

Family

ID=81902079

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210254905.6A Pending CN114624027A (en) 2022-03-14 2022-03-14 Bearing fault diagnosis method based on multi-input CNN

Country Status (1)

Country Link
CN (1) CN114624027A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114593917A (en) * 2022-03-08 2022-06-07 安徽理工大学 Small sample bearing fault diagnosis method based on triple model

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114593917A (en) * 2022-03-08 2022-06-07 安徽理工大学 Small sample bearing fault diagnosis method based on triple model

Similar Documents

Publication Publication Date Title
CN104792522B (en) Intelligent gear defect analysis method based on fractional wavelet transform and BP neutral network
CN112084974B (en) Multi-label rolling bearing fault diagnosis method based on meta-learning
CN109946389B (en) Structural damage identification method based on ensemble empirical mode decomposition and convolutional neural network
CN109765053B (en) Rolling bearing fault diagnosis method using convolutional neural network and kurtosis index
CN112731309B (en) Active interference identification method based on bilinear efficient neural network
CN112418013B (en) Complex working condition bearing fault diagnosis method based on meta-learning under small sample
CN107702922B (en) Rolling bearing fault diagnosis method based on LCD and stacking automatic encoder
CN109190464B (en) Mechanical fault intelligent diagnosis method based on transfer learning under variable working conditions
CN107016241A (en) Based on the rotating machinery lifetime stage recognition methods for adding sample enhancing depth own coding learning network of making an uproar
CN115758212A (en) Mechanical equipment fault diagnosis method based on parallel network and transfer learning
CN112304614B (en) Intelligent fault diagnosis method for end-to-end rolling bearing by adopting multi-attention mechanism
CN112577748A (en) Reinforced lightweight multi-scale CNN-based rolling bearing fault diagnosis method
CN115965057B (en) Brain-like continuous learning fault diagnosis method for train transmission system
CN112990082B (en) Detection and identification method of underwater sound pulse signal
CN113567131A (en) Bearing fault diagnosis method based on S transformation and miniature convolution neural network model
CN116702076A (en) Small sample migration learning fault diagnosis method, system, computer and storage medium based on CNN feature fusion
CN113919220A (en) Intelligent fault diagnosis method for rolling bearing driven by vibration image
CN114624027A (en) Bearing fault diagnosis method based on multi-input CNN
CN113792685A (en) Microseism event detection method based on multi-scale convolution neural network
CN113780242A (en) Cross-scene underwater sound target classification method based on model transfer learning
Yang et al. Few-shot learning for rolling bearing fault diagnosis via siamese two-dimensional convolutional neural network
CN115859077A (en) Multi-feature fusion motor small sample fault diagnosis method under variable working conditions
CN115221958A (en) Mechanical equipment selective migration fault diagnosis method based on composite weight
CN109886097A (en) A kind of method for detecting fatigue driving based on artificial fish school optimization H-ELM
CN105823634A (en) Bearing damage identification method based on time frequency relevance vector convolution Boltzmann machine

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