CN110307982B - Bearing fault classification method based on CNN and Adaboost - Google Patents

Bearing fault classification method based on CNN and Adaboost Download PDF

Info

Publication number
CN110307982B
CN110307982B CN201910530344.6A CN201910530344A CN110307982B CN 110307982 B CN110307982 B CN 110307982B CN 201910530344 A CN201910530344 A CN 201910530344A CN 110307982 B CN110307982 B CN 110307982B
Authority
CN
China
Prior art keywords
time domain
signal
layer
time
classification model
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.)
Active
Application number
CN201910530344.6A
Other languages
Chinese (zh)
Other versions
CN110307982A (en
Inventor
殷春
程玉华
彭威
陈凯
黄雪刚
马浩鹏
周静
杨晓
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.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
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 University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN201910530344.6A priority Critical patent/CN110307982B/en
Publication of CN110307982A publication Critical patent/CN110307982A/en
Application granted granted Critical
Publication of CN110307982B publication Critical patent/CN110307982B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/20—Ensemble learning
    • 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

Abstract

The invention discloses a bearing fault classification method based on CNN and Adaboost, which comprises the steps of firstly collecting bearing signals, then preprocessing the bearing signals, and extracting time domain signals and time-frequency domain signals; and then constructing a time domain weak classification model and a time frequency domain weak classification model based on the time domain signal and the time frequency domain signal respectively, integrating the time domain weak classification model and the time frequency domain weak classification model, and predicting the membership probability value of the bearing signal of the unmanned aerial vehicle to be detected by using the integrated classification model, thereby realizing the fault classification of the unmanned aerial vehicle bearing.

Description

Bearing fault classification method based on CNN and Adaboost
Technical Field
The invention belongs to the technical field of fault diagnosis of unmanned aerial vehicle systems, and particularly relates to an unmanned aerial vehicle bearing fault classification method based on CNN and Adaboost ensemble learning.
Background
The unmanned aerial vehicle technology is different day by day, and various unmanned aerial vehicles play a huge role in the military field. And the bearing failure of the aircraft engine is the main factor causing the failure of the unmanned aerial vehicle, and the reliability and the health condition of the engine can be directly influenced. Therefore, bearing fault diagnosis of the unmanned aerial vehicle is an important research topic. The fault modes of the unmanned aerial vehicle bearing are various, and how to identify the fault type of the bearing with high precision has important significance on the stability and reliability of the unmanned aerial vehicle system. In addition, the bearing stress environment that the space gesture of unmanned aerial vehicle flight often leads to is various, consequently has higher requirement to diagnostic system's generalization ability. The fault diagnosis system with high precision and strong generalization capability has important significance for unmanned aerial vehicle maintenance.
The fault diagnosis system often needs to preprocess the initial bearing fault signal, and the preprocessing of the signal is the basis of the analysis of subsequent fault data, so that the research of the proper fault signal preprocessing has important significance. The existing information preprocessing method is as follows: empirical decomposition EMD, LMD, wavelet analysis, variational modal analysis, and the like. The EMD and the LMD are recursive screening modes, the recursive screening method has general denoising robustness, and signal convergence control is not easy. Too many wavelet analysis denoising parameters exist, and the denoising performance is easily influenced by the parameters; the variational modal decomposition VMD is a method for signal decomposition and weighted fusion reconstruction, and has obvious denoising effect on signals with non-stationarity and low signal-to-noise ratio, so that the VMD is finally selected as a signal preprocessing algorithm.
For a fault diagnosis model of a bearing, most of traditional methods use time domain features or time-frequency domain features, and are combined with traditional machine learning algorithms such as a support vector machine and a Bayesian classification algorithm, but the methods are only suitable for small-scale data sets, and the model has limited learning capability, is sensitive to samples, and is easy to overfit. However, the monitoring data of the mechanical equipment of the unmanned aerial vehicle is usually large-scale mass data, so that researchers gradually introduce deep learning to perform fault diagnosis, such as ANN, RNN and CNN. The bearing vibration signals generally show certain structuredness, periodicity and large-scale property, the ANN and RNN models have no scale invariance, and the problem of low bearing fault identification precision caused by the fact that weight sharing cannot be carried out exists.
At present, the characteristic forms capable of representing faults are many, such as amplitude, phase, frequency, time domain signals, time frequency signals and the like, and because the time domain characteristics contain a large amount of fault information in the signals, the time frequency characteristics can better distinguish different fault types through the time frequency relation, the time domain characteristics and the time frequency characteristics are mainly used for fault classification.
Researchers usually input one-dimensional vibration time domain signals into a CNN model for fault diagnosis, and the input form does not consider the relevance inside signal faults, so that the model training efficiency and the fault diagnosis precision are low. In order to solve the problem, the invention provides a method for carrying out structural conversion on signals based on a certain arrangement cardinality to form a grid input form, but the problem is that the accuracy of a model is easily influenced by the cardinality. The method considers that different fault types show different periodicities in vibration signals, and performs time sequence conversion on the signals by taking the fault periods as the arrangement base number to obtain internal information of a time sequence matrix, which can better represent various faults, and can improve the training speed of a CNN model. On the other hand, in the use of time-frequency characteristics, the traditional method adopts offline Fourier transform to extract the time-frequency characteristics, so that the problem of inaccurate time and frequency positioning exists.
Considering that most of the current CNN fault diagnosis models use single time domain features or time-frequency features, and the time domain features and the time-frequency features have the phenomenon of complementary advantages, the time domain features have the advantages of small calculation complexity and benefit for the real-time performance of the algorithm, and the time-frequency features have the disadvantages of poor robustness, can represent the change frequency information of different time positions of signals, are very suitable for the analysis of non-stationary signals, have better robustness, but have higher calculation complexity, and therefore, the method has important significance on how to effectively realize the complementary advantages. In recent years, the application of Boosting ensemble learning is a large research hotspot in the processing of feature fusion and model optimization, and the method is based on feature sample serial training and has the advantages of flexible construction, difficulty in overfitting and high precision improvement. Adaboost is the most representative of the Boosting family of algorithms, and at present, in the aspect of fault diagnosis of an unmanned aerial vehicle, a diagnosis method combining deep learning and Adaboost integrated learning does not have published relevant literature data, so that the research of the fault diagnosis algorithm with high precision and strong generalization capability is of great significance.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a bearing fault classification method based on CNN and Adaboost.
In order to achieve the above object, the present invention provides a bearing fault classification method based on CNN and Adaboost, which is characterized by comprising the following steps:
(1) acquiring a signal data set
Acquiring all bearing signals in the unmanned aerial vehicle to form a signal data set F ═ F(i)|i∈[1,m]}，f(i)Representing a signal generated by the ith bearing, wherein m is the total number of the bearings in the unmanned aerial vehicle; wherein f is(i)Generating a normal signal or an inner ring fault signal or a ball fault signal or an outer ring fault signal for the ith bearing;
(2) signal preprocessing
(2.1) decomposing f by using variation mode(i)Dividing into n decomposed signals
Figure GDA0002701295150000031
Wherein, the k decomposition signal after decomposition is:
Figure GDA0002701295150000032
wherein, sigma is constant, alpha is secondary punishment factor, omega(i)Is f(i)The center frequency of (a) is,
Figure GDA0002701295150000033
is the center frequency of the kth decomposed signal;
center frequency
Figure GDA0002701295150000034
The calculation formula of (2) is as follows:
Figure GDA0002701295150000035
(2.2) filtering the n decomposed signals and then superposing the n decomposed signals to form a signal U(i)，
Figure GDA0002701295150000036
(2.3) decomposition of the signal U on the basis of the wavelet transform(i)Forming a time-frequency domain signal F(i)，；
(2.4) taking the period as a breakpoint, and converting the one-dimensional time domain signal U into a one-dimensional time domain signal U(i)Reconstructing the data into two-dimensional time domain data to obtain a time domain signal S(i)；
(2.5) respectively carrying out time and frequency domain signals F by utilizing self-service sampling method(i)And a time domain signal S(i)Sampling is carried out, and lambda groups of time-frequency domain characteristic data sets and time-domain characteristic data sets are obtained respectively;
(3) constructing a time domain weak classification model by using lambda group time domain feature data set
(3.1) Using the time-domain signal S(i)Building a CNN network model;
(3.2) initializing network weight value to
Figure GDA0002701295150000041
Training a CNN network model by using the first group of time domain characteristic data to obtain a first time domain weak classification model which is recorded as
Figure GDA0002701295150000042
(3.3) calculation of
Figure GDA0002701295150000043
Classification error rate of
Figure GDA0002701295150000044
(3.4) error Rate according to classification
Figure GDA00027012951500000420
Obtaining coefficients of a first time domain weak classification model
Figure GDA0002701295150000045
Figure GDA0002701295150000046
(3.5) model coefficients according to weak classification
Figure GDA0002701295150000047
Updating the weight distribution of the time domain feature data set;
if the first set of time domain feature data is correctly classified, the weight of the set of data is updated as:
Figure GDA0002701295150000048
if the first group of time domain feature data is not correctly classified, the weight of the group of data is kept unchanged;
(3.6) based on the weight value obtained in the step (3.5), returning to the step (3.2), training the CNN network model by using a second group of time domain feature data to obtain a second time domain weak classification model
Figure GDA0002701295150000049
And then, analogizing in sequence to obtain lambda group weak classification models and corresponding weak classification model coefficients, wherein the time domain weak classification models are recorded as:
Figure GDA00027012951500000410
the time domain weak classification model coefficients are recorded as:
Figure GDA00027012951500000411
(4) establishing a time-frequency domain weak classification model by using lambda group time-frequency domain characteristic data
(4.1) Using the time-frequency domain signal F(i)Building a CNN network model;
(4.2) initializing network weight value to
Figure GDA00027012951500000412
Training a CNN network model by using the first group of time domain characteristic data to obtain a first time domain weak classification model which is recorded as
Figure GDA00027012951500000413
(4.3) calculation of
Figure GDA00027012951500000414
Classification error rate of
Figure GDA00027012951500000415
(4.4) error Rate according to classification
Figure GDA00027012951500000416
Obtaining coefficients of a first time domain weak classification model
Figure GDA00027012951500000417
Figure GDA00027012951500000418
(4.5) model coefficients according to weak classification
Figure GDA00027012951500000419
Updating the weight distribution of the time domain feature data set;
if the first set of time domain feature data is correctly classified, the weight of the set of data is updated as:
Figure GDA0002701295150000051
if the first group of time domain feature data is not correctly classified, the weight of the group of data is kept unchanged;
(4.6) based on the weight value obtained in the step (4.5), returning to the step (4.2), training the CNN network model by using a second group of time domain feature data to obtain a second time domain weak classification model
Figure GDA0002701295150000052
And then, analogizing in sequence to obtain lambda group weak classification models and corresponding weak classification model coefficients, wherein the time domain weak classification models are recorded as:
Figure GDA0002701295150000053
the time domain weak classification model coefficients are recorded as:
Figure GDA0002701295150000054
(5) weak classification model integration
Based on Adaboost technology, fusing 2 lambda weak classification models by using weak classification model coefficients to form an integrated classification model Adaboost + CNN;
Figure GDA0002701295150000055
(6) classifying bearing faults of the unmanned aerial vehicle by utilizing the integrated classification model
The bearing signals of the unmanned aerial vehicle to be detected are input into the integrated classification model, and the integrated classification model predicts the membership probability value of the bearing signals of the unmanned aerial vehicle, so that the fault classification of the unmanned aerial vehicle bearing is realized.
The invention aims to realize the following steps:
the invention relates to a bearing fault classification method based on CNN and Adaboost, which comprises the steps of firstly collecting bearing signals, then preprocessing the bearing signals, and extracting time domain signals and time-frequency domain signals; and then constructing a time domain weak classification model and a time frequency domain weak classification model based on the time domain signal and the time frequency domain signal respectively, integrating the time domain weak classification model and the time frequency domain weak classification model, and predicting the membership probability value of the bearing signal of the unmanned aerial vehicle to be detected by using the integrated classification model, thereby realizing the fault classification of the unmanned aerial vehicle bearing.
Meanwhile, the bearing fault classification method based on CNN and Adaboost of the invention also has the following beneficial effects:
(1) the mechanical equipment monitoring data of the unmanned aerial vehicle is large-scale mass data generally, and the method using deep learning CNN has the advantages of scale invariance, strong feature learning capability and the like, and can improve the data processing capability;
(2) the time domain characteristics and the time frequency characteristics have the advantage of complementary advantages, the time domain characteristics have the advantages that the calculation complexity is low, the algorithm real-time performance is facilitated, the time frequency characteristics can represent the change frequency information of the signals at different time positions, and the method is very suitable for analyzing non-stationary signals. The two are combined to facilitate more accurate fault diagnosis;
(3) adaboost ensemble learning is based on serial training of feature samples, and has the advantages of flexible construction, difficulty in overfitting and high precision improvement. The recognition precision and the flourishing ability can be effectively improved.
Drawings
FIG. 1 is a flow chart of a bearing fault classification method based on CNN and Adaboost according to the present invention;
FIG. 2 is a flow chart of the preprocessing of the bearing signal;
FIG. 3 is a schematic diagram of time domain signal reconstruction;
FIG. 4 is a schematic diagram of a CNN network model established by time domain signals;
FIG. 5 is a schematic diagram of a CNN network model established by time-frequency domain signals;
FIG. 6 is a schematic diagram of the Adaboost + CNN model;
FIG. 7 is a bearing signal diagnostic flow chart;
fig. 8 is a comparison graph of classification accuracy of the time domain CNN model, the time-frequency domain CNN model, and the time domain + time-frequency domain integrated CNN model (CNN + adaboost).
Detailed Description
The following description of the embodiments of the present invention is provided in order to better understand the present invention for those skilled in the art with reference to the accompanying drawings. It is to be expressly noted that in the following description, a detailed description of known functions and designs will be omitted when it may obscure the subject matter of the present invention.
Examples
For convenience of description, the related terms appearing in the detailed description are explained:
EMD (Empical Mode composition): empirical mode decomposition;
VMD (spatial mode decomposition) variational modal decomposition;
adaboost (adaptive boosting): one of ensemble learning boosting;
CNN: a convolutional neural network;
ann (artificial Neutral network): an artificial neural network;
rnn (current Neural networks): a recurrent neural network;
FIG. 1 is a flow chart of a bearing fault classification method based on CNN and Adaboost in the invention.
In this embodiment, as shown in fig. 1, the method for classifying bearing faults based on CNN and Adaboost of the present invention includes the following steps:
s1, acquiring a signal data set
Acquiring all bearing signals in the unmanned aerial vehicle to form a signal data set F ═ F(i)|i∈[1,m]}，f(i)The signal generated by the ith bearing is represented, m is the total number of the bearings in the unmanned aerial vehicle, wherein the bearing signal can be a normal signal, an inner ring fault signal, a ball fault signal and an outer ring fault signal;
s2, as shown in FIG. 2, bearing signal preprocessing
S2.1, decomposing f by adopting variation mode(i)Dividing into n decomposed signals
Figure GDA0002701295150000071
Wherein, the k decomposition signal after decomposition is:
Figure GDA0002701295150000072
wherein, sigma is constant, alpha is secondary punishment factor, omega(i)Is f(i)The center frequency of (a) is,
Figure GDA0002701295150000073
is the center frequency of the kth decomposed signal;
center frequency
Figure GDA0002701295150000074
The calculation formula of (2) is as follows:
Figure GDA0002701295150000075
s2.2, filtering the n decomposed signals and then superposing the n decomposed signals to form a signal U(i)，
Figure GDA0002701295150000076
S2.3, decomposing signal U based on wavelet transformation(i)Forming a time-frequency domain signal F(i)The time-frequency domain signal is a spectrogram and is two-dimensional information, so reconstruction is not needed;
s2.4, as shown in FIG. 3, taking the period as a break point, and converting the one-dimensional time domain signal U into a one-dimensional time domain signal U(i)Reconstructing the data into two-dimensional time domain data to obtain a time domain signal S(i)；
S2.5, respectively setting time and frequency domain signals F by utilizing self-service sampling method(i)And a time domain signal S(i)Sampling is carried out, and lambda groups of time-frequency domain characteristic data sets and time-domain characteristic data sets are obtained respectively;
s3, constructing a time domain weak classification model by using lambda group time domain feature data sets
S3.1, using time-domain signal S(i)Building a CNN network model, and building the CNN network model according to the time-frequency domain data characteristicsAs shown in fig. 4, the multilayer chip mainly includes a convolution layer, a pooling layer, a full-link layer, and a discrimination layer.
An input layer: the size of an input time domain sample is 64 multiplied by 16, and the number of channels is 1;
convolutional layer C1: the convolution kernel size is set to be 3 x 3, the step size Stride is set to be 1, the zero padding Pad is set to be 1, the feature map size is 64 x 16, the feature map depth is 6, and the relu function is selected as the activation function;
pooling layer S1: stride is set to 2, Pad is set to 0, the feature map size is 32 x 8, and the depth of the feature map is not changed by the pooling layer;
convolutional layer C2: the convolution kernel size is set to be 3 x 3, Stride is set to be 1, Pad is set to be 1, the feature map size is 32 x 8, the feature map depth is 24, and the relu function is selected as the activation function;
pooling layer S2: stride is set to 2, Pad is set to 0, feature map size is 16 × 4, and pooling layers do not change feature map depth.
Full connection layer: setting the number of the neurons as 64, setting the Dropout parameter as 0.5, and enabling the probability of the layer neuron inactivation to be 0.5;
output layer Softmax: the output category number is 4, which respectively corresponds to a normal signal, an inner ring fault signal, a ball fault signal and an outer ring fault signal.
S3.2, initializing network weighted value to be
Figure GDA0002701295150000081
Training Adaboost + CNN network model by using the first group of time domain characteristic data to obtain a first time domain weak classification model which is recorded as
Figure GDA0002701295150000082
S3.3, calculating
Figure GDA0002701295150000083
Classification error rate of
Figure GDA0002701295150000084
S3.4, according to the classification error rate
Figure GDA0002701295150000085
Obtaining coefficients of a first time domain weak classification model
Figure GDA0002701295150000086
Figure GDA0002701295150000087
S3.5, according to weak classification model coefficient
Figure GDA0002701295150000088
Updating the weight distribution of the time domain feature data set;
if the first set of time domain feature data is correctly classified, the weight of the set of data is updated as:
Figure GDA0002701295150000089
if the first group of time domain feature data is not correctly classified, the weight of the group of data is kept unchanged;
s3.6, based on the weight value obtained in the step S3.5, returning to the step S3.2, training an Adaboost + CNN network model by using a second group of time domain feature data to obtain a second time domain weak classification model
Figure GDA00027012951500000810
And then, analogizing in sequence to obtain lambda group weak classification models and corresponding weak classification model coefficients, wherein the time domain weak classification models are recorded as:
Figure GDA0002701295150000091
the time domain weak classification model coefficients are recorded as:
Figure GDA0002701295150000092
s4, constructing a time-frequency domain weak classification model by using lambda group time-frequency domain characteristic data
S4.1. Using time-frequency domain signals F(i)A CNN network model is built, and the CNN network model is built according to the characteristics of the time-frequency domain data, as shown in fig. 5, and mainly includes a convolutional layer, a pooling layer, a full-link layer, and a discrimination layer.
The time-frequency domain model is as a graph and mainly comprises a convolution layer, a pooling layer, a full-connection layer and a discrimination layer similar to the time domain model.
An input layer: the input time domain sample size is 28 x 28, and the number of channels is 1.
Convolutional layer C1: the convolution kernel size is set to 5 × 5, Stride is set to 1, Pad is set to 0, feature map size is 24 × 24, feature map depth is 6, and the activation function selects the relu function.
Pooling layer S1: stride is set to 2, Pad is set to 0, feature map size is 12 × 12, and pooling layers do not change feature map depth.
Convolutional layer C2: the convolution kernel size is set to 5 × 5, Stride is set to 1, Pad is set to 0, feature map size is 8 × 8, feature map depth is 24, and the activation function selects the relu function.
Pooling layer S2: stride is set to 2, Pad is set to 0, feature map size is 4 x 4, and pooling layers do not change feature map depth.
Full connection layer FC: the neuron number was 336, and the Dropout parameter was 0.5, so that the probability of the layer neuron inactivation was 0.5.
Output layer Softmax: the output category number is 4, which respectively corresponds to a normal signal, an inner ring fault signal, a ball fault signal and an outer ring fault signal.
S4.2, initializing network weighted value to be
Figure GDA0002701295150000093
Training a CNN network model by using the first group of time domain characteristic data to obtain a first time domain weak classification model which is recorded as
Figure GDA0002701295150000094
S4.3, calculating
Figure GDA0002701295150000095
Classification error rate of
Figure GDA0002701295150000096
S4.4, according to the classification error rate
Figure GDA0002701295150000097
Obtaining coefficients of a first time domain weak classification model
Figure GDA0002701295150000098
Figure GDA0002701295150000099
S4.5, according to weak classification model coefficient
Figure GDA00027012951500000910
Updating the weight distribution of the time domain feature data set;
if the first set of time domain feature data is correctly classified, the weight of the set of data is updated as:
Figure GDA0002701295150000108
if the first group of time domain feature data is not correctly classified, the weight of the group of data is kept unchanged;
s4.6, based on the weight value obtained in the step S4.5, returning to the step S4.2, training the CNN network model by using a second group of time domain feature data to obtain a second time domain weak classification model
Figure GDA0002701295150000101
And then, analogizing in sequence to obtain lambda group weak classification models and corresponding weak classification model coefficients, wherein the time domain weak classification models are recorded as:
Figure GDA0002701295150000102
the time domain weak classification model coefficients are recorded as:
Figure GDA0002701295150000103
s5 integration of weak classification models
Based on Adaboost technology, fusing 2 lambda weak classification models by using weak classification model coefficients, and forming an integrated classification model Adaboost + CNN as shown in FIG. 6;
Figure GDA0002701295150000104
s6, classifying bearing faults of the unmanned aerial vehicle by utilizing the integrated classification model
The bearing signals of the unmanned aerial vehicle to be detected are input into the integrated classification model, and the integrated classification model predicts the membership probability value of the bearing signals of the unmanned aerial vehicle, so that the fault classification of the unmanned aerial vehicle bearing is realized.
Examples of the invention
Suppose an drone has n bearings, f1, f2, …, fn. For F1, firstly, decomposing the F1 bearing signal into a time domain signal S1 and a time frequency signal F1 based on Variational Modal Decomposition (VMD), and then carrying out fault diagnosis on F1 through an integrated model CNN + Adaboost. Then, fault diagnosis is performed on the bearings f2 … fn based on the above process, the diagnosis process is shown in fig. 7, and finally, the fault condition of the bearings is judged according to the diagnosis result of each bearing.
The model judges the indexes such as parameter accuracy Acc, precision P, recall R, F1 and operation rate.
Figure GDA0002701295150000105
Figure GDA0002701295150000106
Figure GDA0002701295150000107
Figure GDA0002701295150000111
Let SNAnd FNThe number of time domain models and the number of time-frequency domain models are respectively represented, and the final test result is shown in table 1.
Index parameter SN FN ACC(％) P(％) R(％) F1
20 20 97.625 99.739 95.324 0.975
TABLE 1
As shown in fig. 8, the accuracy rate of the time domain model training after 25 training periods is gradually flat and finally reaches 89.520%; the accuracy of the time-frequency domain model training is gradually smooth after 37 training periods, and finally 92.732% is achieved. The accuracy of the CNN model trained by the time-frequency domain data is higher than that of the model trained by the time-domain data, but still lower than 97.625% of the CNN + Adaboost integrated model trained by the time-domain and time-frequency domain data, so that the detection accuracy can be effectively improved by integrating the time-frequency domain model with the time-domain model.
Although illustrative embodiments of the present invention have been described above to facilitate the understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, and various changes may be made apparent to those skilled in the art as long as they are within the spirit and scope of the present invention as defined and defined by the appended claims, and all matters of the invention which utilize the inventive concepts are protected.

Claims (3)

1. A bearing fault classification method based on CNN and Adaboost is characterized by comprising the following steps:
(1) acquiring a signal data set
Acquiring all bearing signals in the unmanned aerial vehicle to form a signal data set F ═ F(i)|i∈[1,m]}，f(i)Representing a signal generated by the ith bearing, wherein m is the total number of the bearings in the unmanned aerial vehicle; wherein f is(i)Generating a normal signal or an inner ring fault signal or a ball fault signal or an outer ring fault signal for the ith bearing;
(2) signal preprocessing
(2.1) decomposing f by using variation mode(i)Dividing into n decomposed signals
Figure FDA0002735032600000011
Wherein, the k decomposition signal after decomposition is:
Figure FDA0002735032600000012
wherein x is ∈ [1, n ]]σ is a constant, α is a secondary penalty factor, ω(i)Is f(i)The center frequency of (a) is,
Figure FDA0002735032600000013
is the center frequency of the kth decomposed signal;
center frequency
Figure FDA0002735032600000014
The calculation formula of (2) is as follows:
Figure FDA0002735032600000015
(2.2) filtering the n decomposed signals and then superposing the n decomposed signals to form a signal U(i)，
Figure FDA0002735032600000016
(2.3) decomposition of the signal U on the basis of the wavelet transform(i)Forming a time-frequency domain signal F(i)；
(2.4) taking the period as a breakpoint, and converting the one-dimensional time domain signal U into a one-dimensional time domain signal U(i)Reconstructing the data into two-dimensional time domain data to obtain a time domain signal S(i)；
(2.5) respectively carrying out time and frequency domain signals F by utilizing self-service sampling method(i)And a time domain signal S(i)Sampling is carried out, and lambda groups of time-frequency domain characteristic data sets and time-domain characteristic data sets are obtained respectively;
(3) constructing a time domain weak classification model by using lambda group time domain feature data set
(3.1) Using the time-domain signal S(i)Building a CNN network model;
(3.2) initializing network weight value to
Figure FDA0002735032600000017
Training a CNN network model by using the first group of time domain characteristic data to obtain a first time domain weak classification model which is recorded as
Figure FDA0002735032600000021
(3.3) calculation of
Figure FDA0002735032600000022
Classification error rate of
Figure FDA0002735032600000023
(3.4) error Rate according to classification
Figure FDA0002735032600000024
Obtaining coefficients of a first time domain weak classification model
Figure FDA0002735032600000025
Figure FDA0002735032600000026
(3.5) model coefficients according to weak classification
Figure FDA0002735032600000027
Updating the weight distribution of the time domain feature data set;
if the first set of time domain feature data is correctly classified, the weight of the set of data is updated as:
Figure FDA0002735032600000028
if the first group of time domain feature data is not correctly classified, the weight of the group of data is kept unchanged;
(3.6) based on the weight value obtained in the step (3.5), returning to the step (3.2), training the CNN network model by using a second group of time domain feature data to obtain a second time domain weak classification model
Figure FDA0002735032600000029
And then, analogizing in sequence to obtain lambda group weak classification models and corresponding weak classification model coefficients, wherein the time domain weak classification models are recorded as:
Figure FDA00027350326000000210
the time domain weak classification model coefficients are recorded as:
Figure FDA00027350326000000211
(4) establishing a time-frequency domain weak classification model by using lambda group time-frequency domain characteristic data
(4.1) Using the time-frequency domain signal F(i)Building a CNN network model;
(4.2) initializing network weight value to
Figure FDA00027350326000000212
Training a CNN network model by using the first group of time domain characteristic data to obtain a first time domain weak classification model which is recorded as
Figure FDA00027350326000000213
(4.3) calculation of
Figure FDA00027350326000000214
Classification error rate of
Figure FDA00027350326000000215
(4.4) error Rate according to classification
Figure FDA00027350326000000216
Obtaining coefficients of a first time domain weak classification model
Figure FDA00027350326000000217
Figure FDA00027350326000000218
(4.5) model coefficients according to weak classification
Figure FDA00027350326000000219
Updating the weight distribution of the time domain feature data set;
if the first set of time domain feature data is correctly classified, the weight of the set of data is updated as:
Figure FDA00027350326000000220
if the first group of time domain feature data is not correctly classified, the weight of the group of data is kept unchanged;
(4.6) based on the weight value obtained in the step (4.5), returning to the step (4.2), training the CNN network model by using a second group of time domain feature data to obtain a second time domain weak classification model
Figure FDA0002735032600000031
And then, analogizing in sequence to obtain lambda group weak classification models and corresponding weak classification model coefficients, wherein the time domain weak classification models are recorded as:
Figure FDA0002735032600000032
the time domain weak classification model coefficients are recorded as:
Figure FDA0002735032600000033
(5) weak classification model integration
Based on Adaboost technology, fusing 2 lambda weak classification models by using weak classification model coefficients to form an integrated classification model Adaboost + CNN;
Figure FDA0002735032600000034
(6) classifying bearing faults of the unmanned aerial vehicle by utilizing the integrated classification model
The bearing signals of the unmanned aerial vehicle to be detected are input into the integrated classification model, and the integrated classification model predicts the membership probability value of the bearing signals of the unmanned aerial vehicle, so that the fault classification of the unmanned aerial vehicle bearing is realized.
2. The CNN and Adaboost-based bearing fault classification method according to claim 1, characterized in that in said step (3.1), time domain signal S is used(i)The CNN network model is built as follows:
the CNN model mainly comprises an input layer, a convolution layer, a pooling layer, a full-link layer and an output layer Softmax;
an input layer: the size of an input time domain sample is 64 multiplied by 16, and the number of channels is 1;
convolutional layer C1: the convolution kernel size is set to be 3 x 3, the step size Stride is set to be 1, the zero padding Pad is set to be 1, the feature map size is 64 x 16, the feature map depth is 6, and the relu function is selected as the activation function;
pooling layer S1: stride is set to 2, Pad is set to 0, the feature map size is 32 x 8, and the depth of the feature map is not changed by the pooling layer;
convolutional layer C2: the convolution kernel size is set to be 3 x 3, Stride is set to be 1, Pad is set to be 1, the feature map size is 32 x 8, the feature map depth is 24, and the relu function is selected as the activation function;
pooling layer S2: stride is set to 2, Pad is set to 0, the feature map size is 16 x 4, and the pooling layer does not change the feature map depth;
full connection layer: setting the number of the neurons as 64, setting the Dropout parameter as 0.5, and enabling the probability of the layer neuron inactivation to be 0.5;
output layer Softmax: the output category number is 4, which respectively corresponds to a normal signal, an inner ring fault signal, a ball fault signal and an outer ring fault signal.
3. The CNN and Adaboost-based bearing fault classification method according to claim 1, characterized in that in said step (4.1), a time-frequency domain signal F is used(i)Build CThe NN network model is as follows:
the CNN model mainly comprises an input layer, a convolution layer, a pooling layer, a full-link layer and an output layer Softmax; (ii) a
An input layer: the input time domain sample size is 28 multiplied by 28, and the number of channels is 1;
convolutional layer C1: setting the size of a convolution kernel to be 5 multiplied by 5, setting the step size Stride to be 1, setting the zero padding Pad to be 0, setting the size of a feature map to be 24 multiplied by 24, setting the depth of the feature map to be 6, and selecting a relu function by an activation function;
pooling layer S1: stride is set to 2, Pad is set to 0, the feature map size is 12 x 12, and the depth of the feature map is not changed by the pooling layer;
convolutional layer C2: setting the size of a convolution kernel to be 5 multiplied by 5, setting Stride to be 1, setting Pad to be 0, setting the size of a feature map to be 8 multiplied by 8, setting the depth of the feature map to be 24, and selecting a relu function by an activation function;
pooling layer S2: stride is set to 2, Pad is set to 0, the feature map size is 4 x 4, and the pooling layer does not change the feature map depth;
full connection layer FC: the number of neurons is 336, the Dropout parameter is 0.5, and the probability of the layer of neurons being inactivated is 0.5;
output layer Softmax: the output category number is 4, which respectively corresponds to a normal signal, an inner ring fault signal, a ball fault signal and an outer ring fault signal.
CN201910530344.6A 2019-06-19 2019-06-19 Bearing fault classification method based on CNN and Adaboost Active CN110307982B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910530344.6A CN110307982B (en) 2019-06-19 2019-06-19 Bearing fault classification method based on CNN and Adaboost

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910530344.6A CN110307982B (en) 2019-06-19 2019-06-19 Bearing fault classification method based on CNN and Adaboost

Publications (2)

Publication Number Publication Date
CN110307982A CN110307982A (en) 2019-10-08
CN110307982B true CN110307982B (en) 2021-01-26

Family

ID=68076961

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910530344.6A Active CN110307982B (en) 2019-06-19 2019-06-19 Bearing fault classification method based on CNN and Adaboost

Country Status (1)

Country Link
CN (1) CN110307982B (en)

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110751087B (en) * 2019-10-17 2020-09-22 湖南波尔坤雷信息科技有限公司 EOF-based unmanned aerial vehicle signal identification system and method
CN110763466A (en) * 2019-11-07 2020-02-07 江苏航空职业技术学院 Adaboost algorithm combined GABP rolling bearing diagnosis method
CN111060316A (en) * 2019-12-10 2020-04-24 华南理工大学 Rolling bearing state monitoring method and system based on convolutional neural network model
CN111259949A (en) * 2020-01-13 2020-06-09 山东超越数控电子股份有限公司 Fault identification model construction method, model and identification method for aircraft environmental control system
CN111289251A (en) * 2020-02-27 2020-06-16 湖北工业大学 Rolling bearing fine-grained fault identification method
CN112001450A (en) * 2020-08-27 2020-11-27 山东超越数控电子股份有限公司 Ship inertial navigation system fault identification method
CN112016473B (en) * 2020-08-31 2022-09-13 国网安徽省电力有限公司电力科学研究院 Power distribution network high-resistance grounding fault diagnosis method based on semi-supervised learning and attention mechanism
CN112069998A (en) * 2020-09-07 2020-12-11 中国人民解放军陆军工程大学 GBDT fault diagnosis integration method based on convolutional neural network
WO2022119426A1 (en) * 2020-12-01 2022-06-09 Université Internationale de RABAT Intelligent system for immediate detection and notification of disturbances in electrical signal quality
CN113971670B (en) * 2021-12-23 2022-04-15 武汉市利隆捷精密螺丝制造有限公司 Thread defect analysis method and system based on computer vision
CN114903308A (en) * 2022-05-20 2022-08-16 哈尔滨理工大学 Chair based on artificial intelligence

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105841961A (en) * 2016-03-29 2016-08-10 中国石油大学(华东) Bearing fault diagnosis method based on Morlet wavelet transformation and convolutional neural network
CN106650647A (en) * 2016-12-09 2017-05-10 开易（深圳）科技有限公司 Vehicle detection method and system based on cascading of traditional algorithm and deep learning algorithm
CN106845440A (en) * 2017-02-13 2017-06-13 山东万腾电子科技有限公司 A kind of augmented reality image processing method and system
CN107526853A (en) * 2016-06-22 2017-12-29 北京航空航天大学 Rolling bearing fault mode identification method and device based on stacking convolutional network
CN109406118A (en) * 2018-10-15 2019-03-01 华中科技大学 A kind of mechanical failure prediction method based on level convolutional neural networks
CN109596326A (en) * 2018-11-30 2019-04-09 电子科技大学 Rotary machinery fault diagnosis method based on optimization structure convolutional neural networks
CN109685118A (en) * 2018-12-07 2019-04-26 中国人民解放军陆军工程大学 A kind of Weak Classifier Adaboost vehicle checking method based on convolutional neural networks feature

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105841961A (en) * 2016-03-29 2016-08-10 中国石油大学(华东) Bearing fault diagnosis method based on Morlet wavelet transformation and convolutional neural network
CN107526853A (en) * 2016-06-22 2017-12-29 北京航空航天大学 Rolling bearing fault mode identification method and device based on stacking convolutional network
CN106650647A (en) * 2016-12-09 2017-05-10 开易（深圳）科技有限公司 Vehicle detection method and system based on cascading of traditional algorithm and deep learning algorithm
CN106845440A (en) * 2017-02-13 2017-06-13 山东万腾电子科技有限公司 A kind of augmented reality image processing method and system
CN109406118A (en) * 2018-10-15 2019-03-01 华中科技大学 A kind of mechanical failure prediction method based on level convolutional neural networks
CN109596326A (en) * 2018-11-30 2019-04-09 电子科技大学 Rotary machinery fault diagnosis method based on optimization structure convolutional neural networks
CN109685118A (en) * 2018-12-07 2019-04-26 中国人民解放军陆军工程大学 A kind of Weak Classifier Adaboost vehicle checking method based on convolutional neural networks feature

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
基于AdaBoost回归树的多目标预测算法;张晶;《计算机与现代化》;20171231(第9期);第89-95,105页 *
深度卷积神经网络的发展及其在计算机视觉领域的应用;张顺等;《计算机学报》;20190331;第42卷(第3期);第453-482页 *

Also Published As

Publication number Publication date
CN110307982A (en) 2019-10-08

Similar Documents

Publication Publication Date Title
CN110307982B (en) Bearing fault classification method based on CNN and Adaboost
CN110307983B (en) CNN-Bagging-based unmanned aerial vehicle bearing fault diagnosis method
CN110333076B (en) Bearing fault diagnosis method based on CNN-Stacking
CN110361176B (en) Intelligent fault diagnosis method based on multitask feature sharing neural network
CN109635677B (en) Compound fault diagnosis method and device based on multi-label classification convolutional neural network
CN112084974B (en) Multi-label rolling bearing fault diagnosis method based on meta-learning
Chen et al. Semi-supervised multiresolution classification using adaptive graph filtering with application to indirect bridge structural health monitoring
CN108734208B (en) Multi-source heterogeneous data fusion system based on multi-mode deep migration learning mechanism
CN103728551B (en) A kind of analog-circuit fault diagnosis method based on cascade integrated classifier
CN110609524B (en) Industrial equipment residual life prediction model and construction method and application thereof
CN111368885B (en) Gas circuit fault diagnosis method for aircraft engine
CN112001306A (en) Electroencephalogram signal decoding method for generating neural network based on deep convolution countermeasure
CN110657984A (en) Planetary gearbox fault diagnosis method based on reinforced capsule network
CN112557826A (en) Ship electric power system fault diagnosis method
CN115919330A (en) EEG Emotional State Classification Method Based on Multi-level SE Attention and Graph Convolution
Zhang et al. Intelligent machine fault diagnosis using convolutional neural networks and transfer learning
CN105046323A (en) Regularization-based RBF network multi-label classification method
CN108805206A (en) A kind of modified LSSVM method for building up for analog circuit fault classification
CN116756225B (en) Situation data information processing method based on computer network security
CN115801152B (en) WiFi action recognition method based on hierarchical transformer model
Jin et al. Multisource data fusion diagnosis method of rolling bearings based on improved multiscale CNN
CN111783941A (en) Mechanical equipment diagnosis and classification method based on probability confidence degree convolutional neural network
Liu et al. MRD-NETS: multi-scale residual networks with dilated convolutions for classification and clustering analysis of spacecraft electrical signal
CN116343016A (en) Multi-angle sonar image target classification method based on lightweight convolution network
Wang et al. Research on Fault Detection of Rolling Bearing Based on CWTDCCNN-LSTM

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