CN109765053A - Utilize the Fault Diagnosis of Roller Bearings of convolutional neural networks and kurtosis index - Google Patents
Utilize the Fault Diagnosis of Roller Bearings of convolutional neural networks and kurtosis index Download PDFInfo
- Publication number
- CN109765053A CN109765053A CN201910056906.8A CN201910056906A CN109765053A CN 109765053 A CN109765053 A CN 109765053A CN 201910056906 A CN201910056906 A CN 201910056906A CN 109765053 A CN109765053 A CN 109765053A
- Authority
- CN
- China
- Prior art keywords
- signal
- sample
- kurtosis
- convolutional neural
- neural networks
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses the Fault Diagnosis of Roller Bearings using convolutional neural networks and kurtosis index, include the following steps: the outer ring failure, inner ring failure, rolling element failure and the other vibration signal of four type of normal condition that obtain rolling bearing respectively by testing, specifically includes sample expansion;Generate kurtosis achievement data collection;Generate fault picture.It is complicated that the present invention overcomes existing method processes, the not strong disadvantage of specific aim.The present invention can extract the profound feature of data using depth model, have better data expression capability, while also avoiding dependence of the traditional characteristic extracting method to professional knowledge, reduce the complexity of diagnosis process.It is tested simultaneously by acquiring the bearing data of different faults type, the results showed that the diagnostic accuracy of the model is higher than conventional method, the Precise Diagnosis suitable for rolling bearing fault.
Description
Technical field
The present invention relates to a kind of Fault Diagnosis of Roller Bearings, more specifically say it is a kind of and utilize convolutional neural networks
With the Fault Diagnosis of Roller Bearings of kurtosis index.
Background technique
Rolling bearing is very widely used important equipment base parts and components and mechanical general original part, in equipment manufacture
In it is indispensable, it directly decides the performance of Grand Equipments and main computer boxes, q&r, is known as " industrial pass
Section ".But simultaneously as it is often in severe working environment, and have that the speed of service is high, structure is complicated and is easy to happen
The characteristics of failure, rolling bearing are also one of rotating machinery easily worn part.According to statistics, the failure of rotating machinery has 70% or more to be
It is related to bearing fault, and bearing once breaks down, it will cause serial cascading failure, it is serious to directly affect entirely
The operational safety of equipment.Therefore, the condition monitoring and fault diagnosis of rolling bearing has a very important significance, and is always machine
One of the direction given priority in tool fault diagnosis.
Along with the fast development of artificial intelligence (artificial intelligence, AI), machine learning method is being rolled
It is had been widely used in the condition monitoring and fault diagnosis of dynamic bearing.But traditional intelligent failure diagnosis method
There are problems that needing to carry out artificial selection feature and a large amount of label data is needed to be trained, in order to solve in fault diagnosis
The deep learning method of these existing problems, rise is gradually applied in fault diagnosis field by people.2006, by
Hinton etc., which is proposed using self-encoding encoder (autoencoder), reduces the dimension of data, and proposes the mode with pre-training
Quickly training depth confidence network, to inhibit gradient disappearance problem.As mark, a kind of mode of the deep learning as rise
The new method in identification field, image recognition, speech recognition and in terms of achieve breakthrough development.Together
When, due to the multilayered structure of deep learning, profound relationship can be extracted from mass data, in bearing failure diagnosis field
Also biggish concern and application have been obtained.Guo etc. propose it is a kind of using convolutional neural networks to bearing vibration signal
The new diagnostic method of continuous wavelet transform scalogram progress Direct Classification.Xie etc. proposes a kind of adaptive depth confidence net
Network model realization fault diagnosis end to end represents feature for extracting deep layer from rotating machinery, identifies bearing fault type
And fault degree.Shao etc. proposes a kind of new convolution depth confidence network (convolutional deep
Beliefnetwork) it is used for the bearing failure diagnosis of electric locomotive.
But the experimentation of the above method is very complicated, specific aim is not strong.
Summary of the invention
The purpose of the invention is to provide the shortcoming for overcoming above-mentioned existing background technique, and a kind of utilize is provided and is rolled up
The Fault Diagnosis of Roller Bearings of product neural network and kurtosis index, deep learning method is used for bearing failure diagnosis by it,
The experimental results showed that working well.
Object of the present invention is to what is reached by following measure: utilize the rolling bearing of convolutional neural networks and kurtosis index
Method for diagnosing faults, it is characterised in that include the following steps:
Obtain outer ring failure, inner ring failure, rolling element failure and the normal condition four of rolling bearing respectively by testing
The other vibration signal of type,
Step 1: sample expands, and for the fault diagnosis for realizing deep learning, a large amount of training sample is needed to be used as support,
In view of the impact signal of faulty bearings is lain in vibration signal, while between stick signal adjacent element as much as possible
Correlation, increase participate in deep learning model training sample size, using segmentation overlay intercept method carry out vibration letter
Number interception;Four kinds of outer ring failure, inner ring failure, rolling element failure and normal condition classifications of foundation rolling bearing, according to
The standard of overlapped signal 50% intercepts vibration signal sample;
Step 2: kurtosis achievement data collection is generated;The kurtosis index of four kinds of classification vibration signals of interception is calculated, composition is outer
Enclose failure, inner ring failure, four kinds of classification vibration signals of rolling element failure and normal condition kurtosis achievement data collection;
Step 3: fault picture is generated;Using the conversion method of data to image, kurtosis achievement data collection is converted into ash
Spend image;
Step 4: the training and test of convolutional neural networks model;The grayscale image of four seed types is sent into convolutional Neural net
Network model carries out the training and test of network model using TensorFlow deep learning frame, finally completes rolling bearing event
Barrier classification.
The present invention has the advantage that the axis of rolling of convolutional neural networks and vibration signal kurtosis index that the present invention establishes
Holding intelligent fault diagnosis model is a kind of hierarchical fault diagnosis model based on deep learning.
The results showed that (1) can extract the profound feature of data using depth model, there is better tables of data
Danone power, while dependence of the traditional characteristic extracting method to professional knowledge is also avoided, reduce the complexity of diagnosis process.
(2) it is tested by acquiring the bearing data of different faults type, the results showed that the diagnostic accuracy of the model is higher than tradition side
Method, the Precise Diagnosis suitable for rolling bearing fault.
Detailed description of the invention
Fig. 1 is the typical structure of convolutional neural networks;
Fig. 2 is the convolutional neural networks structure of building;
Fig. 3 is mechanical breakdown integrated simulation experiment bench;
Fig. 4 is that sample intercepts schematic diagram;
Fig. 5 is transformation result schematic diagram of the data to image;
Fig. 6-1, Fig. 6-2, Fig. 6-3, Fig. 6-4, Fig. 6-5 are training set and verifying ensemble average accuracy rate and the number of iterations relationship
Figure;
Fig. 7 is the accuracy rate and loss function value change curve of model training collection and verifying collection;
Fig. 8 is the test set confusion matrix figure of model output.
1. speed regulating motor in figure, 2. shaft couplings, 3. first bearing seats, 4. rotors, 5. acceleration transducers, 6. second bearings
Seat, 7. workbench.
Specific embodiment
The performance that the invention will now be described in detail with reference to the accompanying drawings, but they and do not constitute a limitation of the invention, only
It is for example.Keep advantages of the present invention more clear by explanation simultaneously and is readily appreciated that.
Find in scientific research: when rolling bearing breaks down, many statistical nature parameters in vibration signal are all
It can change with the property and size of failure, this can be used as the foundation of fault diagnosis.Wherein, kurtosis value refers to as dimensionless
Mark, suitable for the fault diagnosis of surface abrasion, the index is very sensitive to the impulse fault of signal, while it is dimensionless ginseng
Number, thus the revolving speed of the value and rolling bearing, size and load are all not related, are especially advantageous for the mathematical statistics to signal.
In conjunction with above-mentioned, the present invention propose it is a kind of based on convolutional neural networks (convolution neural network, abbreviation CNN) and
Deep learning method is used for bearing failure diagnosis, is by the rolling bearing fault intelligent diagnosing method of vibration signal kurtosis index
The validity of verifying this method, is tested using the rolling bearing with different faults type, and a large amount of vibration letters are being obtained
Failure modes are carried out using mentioned method after number, the experimental results showed that working well.
Refering to known to attached drawing: utilizing the Fault Diagnosis of Roller Bearings of convolutional neural networks and kurtosis index, feature
It is to include the following steps:
Obtain outer ring failure, inner ring failure, rolling element failure and the normal condition four of rolling bearing respectively by testing
The other vibration signal of type,
Step 1: sample expands, and for the fault diagnosis for realizing deep learning, a large amount of training sample is needed to be used as support,
In view of the impact signal of faulty bearings is lain in vibration signal, while between stick signal adjacent element as much as possible
Correlation, increase participate in deep learning model training sample size, using segmentation overlay intercept method carry out vibration letter
Number interception;Four kinds of outer ring failure, inner ring failure, rolling element failure and normal condition classifications of foundation rolling bearing, according to
The standard of overlapped signal 50% intercepts vibration signal sample;
Step 2: kurtosis achievement data collection is generated;The kurtosis index of four kinds of classification vibration signals of interception is calculated, composition is outer
Enclose failure, inner ring failure, four kinds of classification vibration signals of rolling element failure and normal condition kurtosis achievement data collection;
Step 3: fault picture is generated;Using the conversion method of data to image, kurtosis achievement data collection is converted into ash
Spend image;
Step 4: the training and test of convolutional neural networks model;The grayscale image of four seed types is sent into convolutional Neural net
Network model carries out the training and test of network model using TensorFlow deep learning frame, finally completes rolling bearing event
Barrier classification.
In the above-mentioned technical solutions, the kurtosis index mathematic(al) representation described in step 2 are as follows:
Wherein: K is the kurtosis index of signal x;N is the length of signal x;μ is the mean value of signal x;σ is the standard of signal x
Difference, E are to calculate mathematic expectaion, mathematic sign.
In the above-mentioned technical solutions, in step 3 gray level image data preprocessing method are as follows: data sample X (i),
Middle i indicates sample point number, is handled according to formula (2), so that data area transforms between [0,255].
In formula, P (m, n), m=1 ... j, n=1 ... j indicate the pixel of gray level image, and function round () is to round up
Function, xiIndicate i-th of sample, xminIndicate sample minimum, xmaxIndicate sample maximum.
One, typical convolutional neural networks structure
One typical CNN network is as shown in Figure 1, it includes input layer, convolutional layer, pond layer, full articulamentum and outputs
Layer.The digital picture size of input layer is 32 × 32, and by first layer convolutional layer, wherein convolution kernel size is 3 × 3, and number is
16,16 32 × 32 pictures are obtained, reduces by maximum pondization and samples, obtain 16 16 × 16 images, then convolution pond again
Change, finally forms 32 8 × 8 images;Flatness layer is established, 32 8 × 8 images are converted to one-dimensional vector, it is right respectively
Answer 2048 neurons;Next hidden layer and output layer are established, there are four neurons for output layer, respectively indicate four kinds of classifications.
Wherein, in order to mitigate model overfitting the case where, joined Dropout layers in a model, by random in the training process
The response of certain proportion neuron is ignored on ground, alleviates over-fitting, effectively improves the Generalization Capability of model.
Two, the Fault Diagnosis of Roller Bearings of convolutional neural networks and kurtosis index is utilized
Invention describes the Fault Diagnosis of Roller Bearings using convolutional neural networks and vibration signal kurtosis index.
Firstly, describing the kurtosis criterion in vibration signal.Then, in order to handle original signal, used one kind from data to image
Conversion method.Finally, illustrating the training process of convolutional neural networks model.
2.1 kurtosis criterion
Kurtosis (Kurtosis) is sensitive to impact signal, is used to the severe degree of measurement mechanical breakdown, evaluates bearing, tooth
The power of the vibratory impulses ingredients such as wheel.Its mathematic(al) representation are as follows:
Wherein: K is the kurtosis index of signal x;N is the length of signal x;μ is the mean value of signal x;σ is the standard of signal x
Difference, E are constant.It can be seen that kurtosis is substantially normalized fourth central away from, the value and rolling bearing sensitive to singular signal
Revolving speed, size and load it is all not related, be conducive to the mathematical statistics to signal.
Conversion method of 2.2 data to image
In the method for diagnosing faults of data-driven, the pretreatment of data is most important, and major function is exactly from a large amount of
The feature of original signal is extracted in historical data.And these features will have a great impact to final fault diagnosis result.At this
Wen Zhong has used a kind of effective data preprocessing method.Data sample X (i), wherein i indicates sample point number, according to formula
(2) it is handled, so that data area transforms between [0,255].
In formula, P (m, n), m=1 ... j, n=1 ... j indicate the pixel of gray level image, and function round () is to round up
Function, xiIndicate i-th of sample, xminIndicate sample minimum, xmaxIndicate sample maximum.This data processing method it is excellent
Point be that it provides it is a kind of explore signal two dimensional character representation method, can in the case where no any predefined parameter into
Row calculates, and does not need manually to be intervened, also without expertise knowledge.
2.3 method for diagnosing faults
Method proposed by the invention is the vibration letter measured in different faults type rolling bearing operational process first
Number, and kurtosis index will be obtained after vibration signal segmentation processing, then using the conversion method of data to image, by kurtosis index
Two dimensional gray figure is converted to, convolutional neural networks model is finally sent into and completes failure modes.The network structure of building such as Fig. 2 institute
Show, including four parts: data input layer, feature extraction layer, full articulamentum and output layer.The effect of data input layer be by
Original vibration signal carries out segmentation pretreatment, and extracts kurtosis index.Feature extraction layer includes multiple convolution sum pondization operation,
It realizes the feature extraction to data, obtains multiple characteristic vectors.The effect of full articulamentum is to complete data " flattening ", connects institute
Some characteristic vectors.Output layer completes the other classification of target output class using the classifier of definition.
3 engineering experiments
Rolling bearing mechanical fault diagnosis is carried out using the mechanical breakdown integrated simulation experiment bench of SpectraQuest company
Experiment, as shown in Figure 3.The experimental bench is mainly by parts groups such as rolling bearing, detachable bearing block, motor and speed regulation transposition
At.
Experiment be in experimental bench carry out on matched MBER-16K series bearing.The series bearing respectively inner ring,
The prefabricated pitting fault of electric spark is utilized on outer ring and rolling element.The revolving speed of rolling bearing by motor experimental field according to
It needs to be adjusted, picks up the vibration signal of rolling bearing in experiment using piezoelectric acceleration transducer, sample frequency is
12kHz is amplified, is filtered respectively in the case where revolving speed is tri- kinds of 1800r/min, 2400r/min and 2900r/min different operating conditions
Afterwards, signal acquisition is carried out by data collection system.
3.1 samples expand
For the fault diagnosis for realizing deep learning, need a large amount of training sample as support.In view of faulty bearings
Impact signal is lain in vibration signal, while the correlation between stick signal adjacent element as much as possible, increases ginseng
With the sample size of deep learning model training, the interception of vibration signal is carried out using the method for segmentation overlay interception herein, such as
Shown in Fig. 4.
Since the 1st point of vibration signal data, n-th point of interception is then next as first sample of signal
Sample of signal since (n-m) a point, interception (2n-m) a o'clock as second sample of signal, and so on realize signal
Segmentation overlay interception, intercept N number of sample of signal in total.In experiment, according to the outer ring failure of rolling bearing, inner ring failure, rolling
Four kinds of classifications of kinetoplast failure and normal condition have intercepted 32 × 32 × 500 sections of vibrations according to the standard of overlapped signal 50% respectively
Dynamic sample of signal.In this way, the gray level image that 32 × 32 pixels are opened to obtain every kind of fault type 500 below provide it is enough
Data.
3.2 generate fault picture
The other vibration signal of four types of above-mentioned interception is substituted into respectively in formula (1), kurtosis achievement data collection { K is obtainedi|i
=1,2 ..., 32 × 32 × 500, { Ko| O=1,2 ..., 32 × 32 × 500, { Kr| r=1,2 ..., 32 × 32 × 500 },
{Kn| n=1,2 ..., 32 × 32 × 500 }, wherein Ki、Ko、Kr、KnRespectively indicate outer ring failure, inner ring failure, rolling element failure
And the kurtosis achievement data collection of four kinds of classification vibration signals of normal condition.Use the conversion method of data to image, four types
Other data sample transformation result is as figure 5 illustrates.Obtaining the gray level image that each type 500 is opened after conversion, pixel size is 32 ×
32。
As can be seen that the gray level image different from of different faults type from the image after conversion.Wherein, outer ring failure
Obvious, inner ring fault picture and outer ring fault graph with the image and rolling element failure of inner ring failure and the image difference of normal condition
As compared to there is apparent streak feature, normal condition image is more darker than rolling element fault picture.Simultaneously, it was found that will vibrate
After signal is converted into image using kurtosis index, the relationship of fault type and rolling bearing revolving speed obviously dies down, can from table
Out, four seed types are under three kinds of working conditions, and image difference is unobvious, and above-mentioned characteristic is to establish the sample of convolutional neural networks model
This library provides a great convenience.
The training and test of 3.3 convolutional neural networks models
The present invention carries out the building of convolutional neural networks model using TensorFlow deep learning frame, and is trained
And test.This is the open source software developed and safeguarded by Google, artificial intelligence team, Google brain (Google Brain)
Library is absorbed in and provides the basic algorithm component of machine learning and deep learning neural network.
The gray level image that each type 500 of above-mentioned acquisition is opened carries out random division, wherein 450 are used to train,
50 for testing.1800 training set pictures are finally obtained and 200 test set pictures take in the training set of composition
20% data are used for cross validation.Due in the training process of model, the selection of hyper parameter to the training speed of model, point
Class accuracy rate has direct influence, considers convolutional neural networks model and its classification task constructed by this paper, begs for separately below
By convolutional layer neuron number, optimizer type and every batch of processing three kinds of factors of number of samples to category of model accuracy rate
It influences, with the Model Matching parameter that determination is optimal.
(1) convolutional layer neuron number
In convolutional neural networks model, the effect of convolutional layer neuron is the local feature extracted in training sample.For
The abundant characteristic information extracted in different classes of training sample with distinctiveness, can pass through and increase the convolution number of plies or increase
Neuron number purpose mode is realized in convolutional layer.In view of training samples number herein is relatively fewer, utmostly to subtract
Few influence of the over-fitting to classification results, model training are investigated in convolutional layer not under the premise of the fixed convolution number of plies is two layers
Influence with neuron number to category of model accuracy rate carries out 10 model trainings respectively, takes training set, verifying collection and test
The Average Accuracy of collection, the results are shown in Table 1.
The different convolutional layer neuron number purpose experimental results of table 1
Training set and verifying ensemble average accuracy rate and the number of iterations relational graph, as shown in Fig. 6-1- Fig. 6-5.
Model training result, it can be seen that with convolutional layer neuron number purpose increase, knowledge of the model for training sample
Other accuracy rate increases, but verifies collection accuracy rate and gradually reducing compared to corresponding training set accuracy rate, illustrates network model
Over-fitting is gradually aggravating, and takes into account the recognition accuracy in view of test set, the convolutional layer neuron of this paper institute climbing form type
Number uses the allocation plan of (16,32).
(2) optimizer type
Convolutional neural networks model and classification task constructed by the present invention are considered, in conjunction with TensorFlow deep learning frame
Algorithm assembly provided by frame, the optimizer of selection have totally three kinds of Adadelta, Rmsprop and Adam, every group of optimizer point
Not carry out 10 model trainings, using training set, verifying integrates and the Average Accuracy of test set is comprehensively considered as evaluation index,
The results are shown in Table 2.The accuracy rate of training set and test set is taken into account, the optimizer of institute's climbing form type of the present invention selects Adam.
The model experiment results of the selection Different Optimization device of table 2
(3) number of batch processed sample
The training sample number of convolutional neural networks is more, is carried out if parameter is updated with the iterative manner of single sample,
It not only will increase training time cost, but also network model is easily trapped into local optimum, its Generalization Capability caused to be deteriorated.Therefore,
When being trained, accelerate the convergence of model by using the mode of batch processed sample, but with batch processed sample number
Purpose increases, and also brings along that memory consumption is excessive, the excessive problem of the number of iterations.Herein in conjunction with actual sample quantity and classification
Task, be respectively set batch processing number of samples be 50,100,200,300 carry out 10 model trainings, take training set, verifying collection and
The Average Accuracy of test set.The results are shown in Table 3.Know as it can be seen that can achieve optimal test set when batch processing number is 100
Not rate, therefore it is 100 that model, which selects batch processing number of samples,.
The experimental result of the different batch processing numbers of samples of table 3
3.4 experimental result
According to discussion result above-mentioned, convolutional neural networks model parameter is as shown in table 4.
4 convolutional neural networks model parameter of table
When carrying out model training using TensorFlow deep learning frame, for the influence for mitigating over-fitting, setting
Dropout value is 0.25, and the accuracy rate and loss function value change curve of training set and verifying collection are as shown in Figure 7.
During model training, either training set still verifies collection, and accuracy rate is all higher and higher.After iteration
Phase, although the accuracy rate of training set is higher than the accuracy rate of verifying collection, the two gap becomes smaller, and the degree of over-fitting has subtracted
Gently, after 50 iteration, model output is 99.5% in the accuracy rate of test set.Intuitively to show model on test set
All types of recognition results, it is as shown in Figure 8 to establish confusion matrix.
It is the correct number of prediction on the diagonal line of confusion matrix, is the number of prediction error on off-diagonal.As a result it sends out
In 200 samples of present test set, there are 2 sample classifications wrong, be the one of bearing roller failure and normal condition respectively
A sample, which produces, to be obscured, and image is found in comparison diagram 5, and the image difference of rolling element failure and normal condition is really smaller.
Analysis of experimental results shows proposed disaggregated model fault recognition rate with higher herein.
4 conclusions
The rolling bearing fault intelligent diagnostics model of convolutional neural networks and vibration signal kurtosis index that the present invention establishes
It is a kind of hierarchical fault diagnosis model based on deep learning.The results showed that (1) can extract number using depth model
According to profound feature, there is better data expression capability, while also avoiding traditional characteristic extracting method to professional knowledge
Dependence, reduce the complexity of diagnosis process.(2) it is tested by acquiring the bearing data of different faults type, as a result
Show that the diagnostic accuracy of the model is higher than conventional method, the Precise Diagnosis suitable for rolling bearing fault.
It is other unspecified to belong to the prior art.
Claims (3)
1. utilizing the Fault Diagnosis of Roller Bearings of convolutional neural networks and kurtosis index, it is characterised in that including walking as follows
It is rapid:
Obtain outer ring failure, four type of inner ring failure, rolling element failure and normal condition of rolling bearing respectively by testing
Other vibration signal,
Step 1: sample expands, and for the fault diagnosis for realizing deep learning, needs a large amount of training sample as support, considers
Impact signal to faulty bearings is lain in vibration signal, while the phase between stick signal adjacent element as much as possible
Guan Xing increases the sample size for participating in deep learning model training, carries out vibration signal using the method for segmentation overlay interception
Interception;According to four kinds of outer ring failure, inner ring failure, rolling element failure and normal condition classifications of rolling bearing, according to overlapping
The standard of signal 50% intercepts vibration signal sample;
Step 2: kurtosis achievement data collection is generated;Calculate the kurtosis index of four kinds of classification vibration signals of interception, the event of composition outer ring
The kurtosis achievement data collection of barrier, inner ring failure, rolling element failure and four kinds of classification vibration signals of normal condition;
Step 3: fault picture is generated;Using the conversion method of data to image, kurtosis achievement data collection is converted into grayscale image
Picture;
Step 4: the training and test of convolutional neural networks model;The grayscale image of four seed types is sent into convolutional neural networks mould
Type carries out the training and test of network model using TensorFlow deep learning frame, finally completes rolling bearing fault point
Class.
2. the Fault Diagnosis of Roller Bearings according to claim 1 using convolutional neural networks and kurtosis index,
It is characterized in that, the kurtosis index mathematic(al) representation described in step 2 are as follows:
Wherein: K is the kurtosis index of signal x;N is the length of signal x;μ is the mean value of signal x;σ is the standard deviation of signal x, E
It is mathematic sign.
3. the Fault Diagnosis of Roller Bearings according to claim 1 using convolutional neural networks and kurtosis index,
It is characterized in that,
The data preprocessing method of gray level image in step 3 are as follows: data sample X (i), wherein i indicates sample point number, presses
Illuminated (2) is handled, so that data area transforms between [0,255],
In formula, P (m, n), m=1 ... j, n=1 ... j indicate the pixel of gray level image, and function round () is the letter that rounds up
Number, xiIndicate i-th of sample, xminIndicate sample minimum, xmaxIndicate sample maximum.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910056906.8A CN109765053B (en) | 2019-01-22 | 2019-01-22 | Rolling bearing fault diagnosis method using convolutional neural network and kurtosis index |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910056906.8A CN109765053B (en) | 2019-01-22 | 2019-01-22 | Rolling bearing fault diagnosis method using convolutional neural network and kurtosis index |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109765053A true CN109765053A (en) | 2019-05-17 |
CN109765053B CN109765053B (en) | 2022-04-29 |
Family
ID=66454925
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910056906.8A Active CN109765053B (en) | 2019-01-22 | 2019-01-22 | Rolling bearing fault diagnosis method using convolutional neural network and kurtosis index |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109765053B (en) |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110132626A (en) * | 2019-05-20 | 2019-08-16 | 西南石油大学 | A kind of Fault Diagnoses of Oil Pump method based on multiple dimensioned convolutional neural networks |
CN110296802A (en) * | 2019-06-26 | 2019-10-01 | 国网辽宁省电力有限公司葫芦岛供电公司 | Shaft tower bolt looseness method of discrimination based on vibrating speed sensors waveform acquisition |
CN110455530A (en) * | 2019-09-18 | 2019-11-15 | 福州大学 | Compose the gear case of blower combined failure diagnostic method of kurtosis combination convolutional neural networks |
CN110487549A (en) * | 2019-09-10 | 2019-11-22 | 哈工大机器人(山东)智能装备研究院 | Bearing fault recognition methods, device, computer equipment and storage medium |
CN110595780A (en) * | 2019-09-20 | 2019-12-20 | 西安科技大学 | Bearing fault identification method based on vibration gray level image and convolution neural network |
CN111080607A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Rolling bearing oil slinging fault detection method based on image recognition |
CN111104887A (en) * | 2019-12-11 | 2020-05-05 | 北京化工大学 | Whole-period keyless phase monitoring method based on vibration mechanism and deep learning technology |
CN111122161A (en) * | 2019-12-24 | 2020-05-08 | 北京交通大学 | Variable working condition bearing fault diagnosis method based on fast kurtogram and deep residual learning |
CN111289251A (en) * | 2020-02-27 | 2020-06-16 | 湖北工业大学 | Rolling bearing fine-grained fault identification method |
CN111539152A (en) * | 2020-01-20 | 2020-08-14 | 内蒙古工业大学 | Rolling bearing fault self-learning method based on two-stage twin convolutional neural network |
CN111539381A (en) * | 2020-05-18 | 2020-08-14 | 中车永济电机有限公司 | Construction method of wind turbine bearing fault classification diagnosis model |
CN111597948A (en) * | 2020-05-11 | 2020-08-28 | 苏州求臻智能科技有限公司 | Fault detection and classification method for bearing vibration signals |
CN111832428A (en) * | 2020-06-23 | 2020-10-27 | 北京科技大学 | Data enhancement method applied to strip breakage fault diagnosis of cold rolling mill |
CN111832617A (en) * | 2020-06-05 | 2020-10-27 | 上海交通大学 | Engine cold state test fault diagnosis method |
CN112113755A (en) * | 2020-09-02 | 2020-12-22 | 西安交通大学 | Mechanical fault intelligent diagnosis method based on deep convolution-kurtosis neural network |
CN112304614A (en) * | 2020-11-27 | 2021-02-02 | 中国人民解放军海军工程大学 | End-to-end rolling bearing intelligent fault diagnosis method adopting multi-attention machine system |
CN112577745A (en) * | 2020-12-02 | 2021-03-30 | 上海应用技术大学 | Rolling bearing fault diagnosis method based on 1D-CNN |
CN112633132A (en) * | 2020-12-18 | 2021-04-09 | 合肥工业大学 | Bearing fault diagnosis method and system based on two-dimensional vibration image enhancement |
CN112651167A (en) * | 2020-12-02 | 2021-04-13 | 杭州电子科技大学 | Semi-supervised rolling bearing fault diagnosis method based on graph neural network |
WO2021243838A1 (en) * | 2020-06-03 | 2021-12-09 | 苏州大学 | Fault diagnosis method for intra-class self-adaptive bearing under variable working conditions |
CN114383844A (en) * | 2021-12-15 | 2022-04-22 | 太原理工大学 | Rolling bearing fault diagnosis method based on distributed deep neural network |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106874957A (en) * | 2017-02-27 | 2017-06-20 | 苏州大学 | A kind of Fault Diagnosis of Roller Bearings |
CN107657088A (en) * | 2017-09-07 | 2018-02-02 | 南京工业大学 | Fault Diagnosis of Roller Bearings based on MCKD algorithms and SVMs |
CN108757426A (en) * | 2018-07-04 | 2018-11-06 | 温州大学 | Oilfield water filling plunger pump trouble diagnostic method |
-
2019
- 2019-01-22 CN CN201910056906.8A patent/CN109765053B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106874957A (en) * | 2017-02-27 | 2017-06-20 | 苏州大学 | A kind of Fault Diagnosis of Roller Bearings |
CN107657088A (en) * | 2017-09-07 | 2018-02-02 | 南京工业大学 | Fault Diagnosis of Roller Bearings based on MCKD algorithms and SVMs |
CN108757426A (en) * | 2018-07-04 | 2018-11-06 | 温州大学 | Oilfield water filling plunger pump trouble diagnostic method |
Cited By (31)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110132626A (en) * | 2019-05-20 | 2019-08-16 | 西南石油大学 | A kind of Fault Diagnoses of Oil Pump method based on multiple dimensioned convolutional neural networks |
CN110296802A (en) * | 2019-06-26 | 2019-10-01 | 国网辽宁省电力有限公司葫芦岛供电公司 | Shaft tower bolt looseness method of discrimination based on vibrating speed sensors waveform acquisition |
CN110487549A (en) * | 2019-09-10 | 2019-11-22 | 哈工大机器人(山东)智能装备研究院 | Bearing fault recognition methods, device, computer equipment and storage medium |
CN110455530B (en) * | 2019-09-18 | 2021-08-31 | 福州大学 | Fan gear box composite fault diagnosis method combining spectral kurtosis with convolutional neural network |
CN110455530A (en) * | 2019-09-18 | 2019-11-15 | 福州大学 | Compose the gear case of blower combined failure diagnostic method of kurtosis combination convolutional neural networks |
CN110595780A (en) * | 2019-09-20 | 2019-12-20 | 西安科技大学 | Bearing fault identification method based on vibration gray level image and convolution neural network |
CN110595780B (en) * | 2019-09-20 | 2021-12-14 | 西安科技大学 | Bearing fault identification method based on vibration gray level image and convolution neural network |
CN111104887A (en) * | 2019-12-11 | 2020-05-05 | 北京化工大学 | Whole-period keyless phase monitoring method based on vibration mechanism and deep learning technology |
CN111104887B (en) * | 2019-12-11 | 2024-03-29 | 北京化工大学 | Full-period keyless phase monitoring method based on vibration mechanism and deep learning technology |
CN111080607B (en) * | 2019-12-12 | 2020-10-09 | 哈尔滨市科佳通用机电股份有限公司 | Rolling bearing oil slinging fault detection method based on image recognition |
CN111080607A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Rolling bearing oil slinging fault detection method based on image recognition |
CN111122161A (en) * | 2019-12-24 | 2020-05-08 | 北京交通大学 | Variable working condition bearing fault diagnosis method based on fast kurtogram and deep residual learning |
CN111539152A (en) * | 2020-01-20 | 2020-08-14 | 内蒙古工业大学 | Rolling bearing fault self-learning method based on two-stage twin convolutional neural network |
CN111539152B (en) * | 2020-01-20 | 2022-08-26 | 内蒙古工业大学 | Rolling bearing fault self-learning method based on two-stage twin convolutional neural network |
CN111289251A (en) * | 2020-02-27 | 2020-06-16 | 湖北工业大学 | Rolling bearing fine-grained fault identification method |
CN111597948A (en) * | 2020-05-11 | 2020-08-28 | 苏州求臻智能科技有限公司 | Fault detection and classification method for bearing vibration signals |
CN111539381A (en) * | 2020-05-18 | 2020-08-14 | 中车永济电机有限公司 | Construction method of wind turbine bearing fault classification diagnosis model |
CN111539381B (en) * | 2020-05-18 | 2022-04-05 | 中车永济电机有限公司 | Construction method of wind turbine bearing fault classification diagnosis model |
WO2021243838A1 (en) * | 2020-06-03 | 2021-12-09 | 苏州大学 | Fault diagnosis method for intra-class self-adaptive bearing under variable working conditions |
US11886311B2 (en) | 2020-06-03 | 2024-01-30 | Soochow University | Intra-class adaptation fault diagnosis method for bearing under variable working conditions |
CN111832617A (en) * | 2020-06-05 | 2020-10-27 | 上海交通大学 | Engine cold state test fault diagnosis method |
CN111832617B (en) * | 2020-06-05 | 2022-11-08 | 上海交通大学 | Engine cold state test fault diagnosis method |
CN111832428B (en) * | 2020-06-23 | 2024-02-23 | 北京科技大学 | Data enhancement method applied to cold rolling mill broken belt fault diagnosis |
CN111832428A (en) * | 2020-06-23 | 2020-10-27 | 北京科技大学 | Data enhancement method applied to strip breakage fault diagnosis of cold rolling mill |
CN112113755A (en) * | 2020-09-02 | 2020-12-22 | 西安交通大学 | Mechanical fault intelligent diagnosis method based on deep convolution-kurtosis neural network |
CN112304614A (en) * | 2020-11-27 | 2021-02-02 | 中国人民解放军海军工程大学 | End-to-end rolling bearing intelligent fault diagnosis method adopting multi-attention machine system |
CN112651167A (en) * | 2020-12-02 | 2021-04-13 | 杭州电子科技大学 | Semi-supervised rolling bearing fault diagnosis method based on graph neural network |
CN112577745A (en) * | 2020-12-02 | 2021-03-30 | 上海应用技术大学 | Rolling bearing fault diagnosis method based on 1D-CNN |
CN112633132A (en) * | 2020-12-18 | 2021-04-09 | 合肥工业大学 | Bearing fault diagnosis method and system based on two-dimensional vibration image enhancement |
CN114383844A (en) * | 2021-12-15 | 2022-04-22 | 太原理工大学 | Rolling bearing fault diagnosis method based on distributed deep neural network |
CN114383844B (en) * | 2021-12-15 | 2023-06-23 | 太原理工大学 | Rolling bearing fault diagnosis method based on distributed deep neural network |
Also Published As
Publication number | Publication date |
---|---|
CN109765053B (en) | 2022-04-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109765053A (en) | Utilize the Fault Diagnosis of Roller Bearings of convolutional neural networks and kurtosis index | |
Wang et al. | A deep learning method for bearing fault diagnosis based on time-frequency image | |
Chen et al. | Bearing fault diagnosis base on multi-scale CNN and LSTM model | |
Shao et al. | Highly accurate machine fault diagnosis using deep transfer learning | |
Wang et al. | A new intelligent bearing fault diagnosis method using SDP representation and SE-CNN | |
Souza et al. | Deep learning for diagnosis and classification of faults in industrial rotating machinery | |
Hu et al. | Data-driven fault diagnosis method based on compressed sensing and improved multiscale network | |
CN110361176A (en) | A kind of intelligent failure diagnosis method for sharing neural network based on multitask feature | |
CN109781411A (en) | A kind of combination improves the Method for Bearing Fault Diagnosis of sparse filter and KELM | |
CN113128338B (en) | Intelligent diagnosis method for printer roller faults under small sample | |
CN112304614B (en) | Intelligent fault diagnosis method for end-to-end rolling bearing by adopting multi-attention mechanism | |
Kumar et al. | Least square fitting for adaptive wavelet generation and automatic prediction of defect size in the bearing using Levenberg–Marquardt backpropagation | |
Zhang et al. | Deep convolutional neural network using transfer learning for fault diagnosis | |
CN112883994A (en) | Rotating machinery variable working condition fault diagnosis method with balanced distribution adaptation | |
CN115563553B (en) | Aviation bearing fault diagnosis method of deep migration map convolution network under fluctuation working condition | |
CN112257530A (en) | Rolling bearing fault diagnosis method based on blind signal separation and support vector machine | |
Irfan et al. | A Scalo gram-based CNN ensemble method with density-aware smote oversampling for improving bearing fault diagnosis | |
Shen et al. | A new penalty domain selection machine enabled transfer learning for gearbox fault recognition | |
CN111368649B (en) | Emotion perception method operated on raspberry pie | |
Remeseiro et al. | Automatic detection of defective crankshafts by image analysis and supervised classification | |
Peng et al. | Research on fault diagnosis method of rolling bearing based on 2DCNN | |
CN115952408A (en) | Multi-channel cross-domain few-sample stamping production line bearing fault diagnosis method | |
Zhang | Application of artificial intelligence recognition technology in digital image processing | |
Cui et al. | A self-attention based contrastive learning method for bearing fault diagnosis | |
Peng et al. | A visual vibration characterization method for intelligent fault diagnosis of rotating machinery |
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 |