CN109774740A - A kind of wheel tread damage fault diagnostic method based on deep learning - Google Patents
A kind of wheel tread damage fault diagnostic method based on deep learning Download PDFInfo
- Publication number
- CN109774740A CN109774740A CN201910109045.5A CN201910109045A CN109774740A CN 109774740 A CN109774740 A CN 109774740A CN 201910109045 A CN201910109045 A CN 201910109045A CN 109774740 A CN109774740 A CN 109774740A
- Authority
- CN
- China
- Prior art keywords
- layer
- wheel tread
- wheel
- neural networks
- deep learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of wheel tread damage fault diagnostic method based on deep learning: relatively strong non-stationary and easily by the wheel tread vibration signal of strong background noise jamming for having the characteristics that, propose the method for diagnosing faults based on Short Time Fourier Transform and convolutional neural networks, Short Time Fourier Transform is carried out to wheel tread vibration signal, time-frequency spectrum sample is obtained, training set and test set are divided into；Training set is inputted in convolutional neural networks and is learnt, network parameter is constantly updated；The convolutional neural networks model for having learnt parameter is applied to test set, exports fault identification result.
Description
Technical field
The invention belongs to invent to be related to fault diagnosis technology field, and in particular to a kind of wheel tread based on deep learning
Damage fault diagnostic method.
Background technique
Take turns to the critical piece for being bullet train operation, while easily damaging again, and be also cause wheel set bearing therefore
The one of the major reasons of barrier, principal mode show as flat sliding and removing.If breaking down, or even to entire train wheel
And track components do great damage.Due to from collection in worksite to wheel tread vibration signal, often with non-stationary
Feature, and noise jamming is serious, brings huge difficulty to Fault Pattern Recognition.In recent years, as the device cluster of monitoring is advised
Moding is big, and the measuring point that each equipment needs increases, and the sample frequency of each measuring point is high, the data from beginning one's duty end-of-life
Collection lasts length, and mechanical fault diagnosis field is caused also to enter " big data " epoch." big data " how is utilized to mention
The fault diagnosis precision of high very noisy, non-stationary signal, it has also become current research hotspot.
Summary of the invention
The purpose of the present invention is to overcome the shortcomings of prior art, a kind of wheel tread damage fault based on deep learning
Diagnostic method.This method, which is directed to, to be had the characteristics that relatively strong non-stationary and is easily believed by the wheel tread vibration of strong background noise jamming
Number, the method for diagnosing faults based on Short Time Fourier Transform and convolutional neural networks is proposed, failure mould end to end is realized
Formula identification.Short Time Fourier Transform is carried out to wheel tread vibration signal first, time-frequency spectrum sample is obtained, is divided into training set and survey
Examination collection；Then training set is inputted in convolutional neural networks and is learnt, constantly update network parameter；Parameter will finally have been learnt
Convolutional neural networks model be applied to test set, export fault identification result.
The wheel tread damage fault diagnostic method based on deep learning is carried out mainly to comprise the steps of:
1) wheel tread fault vibration signal mechanism model, is established；
2) Short Time Fourier Transform, is carried out to wheel tread vibration signal, time-frequency spectrum sample is obtained, is divided into training set and survey
Examination collection；
3), training set is inputted in convolutional neural networks and is learnt, after convolutional layer, pond layer and full articulamentum,
Output is the category of fault type, is divided into propagated forward process and reverse link parameter renewal process；
4) the convolutional neural networks model for having learnt parameter, is applied to test set, exports fault identification result；
Preferably, described to establish wheel tread fault vibration signal mechanism model, specifically wheel tread failure and vibration
Impulsive model, wheel tread is if there is failure, then wheel couples moment with rail when wheel rolling is at failure
Impact.This periodic impact makes vehicle and rail generate coupled vibrations, damages to it.The failure that one radius is R
The frequency of impact formula of wheel are as follows:
P=v/ (2 π R)
Wherein P is frequency of impact, and v is the speed of service, and R is radius of wheel.
Preferably, the window letter that Short Time Fourier Transform is carried out to wheel tread vibration signal, is using regular length
Several pairs of time-domain signals intercept, and carry out Fourier transformation to the signal that interception obtains, and obtain the very little period near moment t
On local spectrum.By translation of the window function on entire time shaft, final transformation obtains upper local spectrum of each period
Set, Short Time Fourier Transform (STFT) is the two-dimensional function about time and frequency.Basic operation formula is as follows:
Wherein f (t) is time-domain signal,Centered on be located atThe time window at moment, ω are frequency, and j is imaginary number
Unit；
Preferably, the implementation of the step 3) are as follows: propagated forward process is mainly three parts: convolutional layer, Chi Hua
Layer and full articulamentum；
On convolutional layer, convolution is carried out with multiple convolution kernels and input picture, in addition passing through an activation primitive after biasing
It can be obtained by series of features figure, convolution process mathematic(al) representation are as follows:
WhereinFor l j-th of element of layer, MjFor j-th of convolution region of l-1 layers of characteristic pattern,For member therein
Element,For the weight matrix of corresponding convolution kernel,For bias term, f (x) is activation primitive, mathematic(al) representation are as follows:
F (x)=max (0, lg (1+ex))
Pond layer often connects behind convolutional layer, carries out dimensionality reduction to characteristic pattern, while keeping characteristic dimension to a certain extent
Invariance.On the layer of pond, pond is carried out for the region n × n in each nonoverlapping size to the characteristic pattern of convolutional layer output
Operation, chooses the maximum value or average value on each region, finally makes to export the n that image all reduces on two dimensions
Times.
After input picture is propagated by the alternating of multiple convolutional layers and pond layer, it is directed to and extracts by full connection layer network
Feature classify.On full articulamentum, input is the weighted summation of one-dimensional characteristic vector of all characteristic pattern expansion and leads to
It crosses after activation primitive and to obtain:
yk=f (wkxk-1+bk)
Wherein k is the signal of network layer, ykFor the output of full articulamentum, xk-1It is the one-dimensional characteristic vector of expansion, wkFor power
Weight coefficient, bkFor bias term.
The process that reverse link parameter updates be it is layer-by-layer update convolutional neural networks can learning parameter, expression formula are as follows:
Wherein w ' and b ' is updated weight and biasing, and w and b are existing weight and biasing, and η is learning rate ginseng
Number, E are the loss for minimizing network.
Compared with prior art, the present invention its beneficial technical effect are as follows: this method has different types of faults very high
Accuracy of identification, and the robustness of the method can be improved by way of increasing fault data type and quantity, be a kind of
It is adapted to the method for diagnosing faults of processing " big data ", while considering time domain and frequency domain information, and utilizes convolutional neural networks
To the extracted in self-adaptive that fault signature carries out, realizes fault diagnosis end to end, i.e., mass data is capable of handling by building
Deep learning frame improves the fault identification precision of very noisy interference, non-stationary signal.
Detailed description of the invention
Fig. 1 is step flow chart of the invention.
Specific embodiment
This method, which is directed to, to be had the characteristics that relatively strong non-stationary and is easily believed by the wheel tread vibration of strong background noise jamming
Number, the method for diagnosing faults based on Short Time Fourier Transform and convolutional neural networks is proposed, failure mould end to end is realized
Formula identification.Short Time Fourier Transform is carried out to wheel tread vibration signal first, time-frequency spectrum sample is obtained, is divided into training set and survey
Examination collection；Then training set is inputted in convolutional neural networks and is learnt, constantly update network parameter；Parameter will finally have been learnt
Convolutional neural networks model be applied to test set, export fault identification result.
The wheel tread damage fault diagnostic method based on deep learning is carried out mainly to comprise the steps of:
1) wheel tread fault vibration signal mechanism model, is established；
Specifically wheel tread failure and vibratory impulse model, wheel tread is if there is failure, then working as wheel rolling
When at failure, wheel couples moment with rail and impacts.This periodic impact makes vehicle couple vibration with rail generation
It is dynamic, it is damaged.The frequency of impact formula for the failure wheel that one radius is R are as follows:
P=v/ (2 π R)
Wherein P is frequency of impact, and v is the speed of service, and R is radius of wheel.
2) Short Time Fourier Transform, is carried out to wheel tread vibration signal, time-frequency spectrum sample is obtained, is divided into training set and survey
Examination collection；
Specifically time-domain signal is intercepted using the window function of regular length, and Fu is carried out to the signal that interception obtains
In leaf transformation, obtain local spectrum of the moment t nearby on the very little period, by translation of the window function on entire time shaft,
Final transformation obtains the set of upper local spectrum of each period, and Short Time Fourier Transform (STFT) is about time and frequency
Two-dimensional function.Basic operation formula is as follows:
Wherein f (t) is time-domain signal,Centered on be located atThe time window at moment, ω are frequency, and j is imaginary number
Unit.
3), training set is inputted in convolutional neural networks and is learnt, after convolutional layer, pond layer and full articulamentum,
Output is the category of fault type, is divided into propagated forward process and reverse link parameter renewal process；
Specifically propagated forward process is mainly three parts: convolutional layer, pond layer and full articulamentum；
On convolutional layer, convolution is carried out with multiple convolution kernels and input picture, in addition passing through an activation primitive after biasing
It can be obtained by series of features figure, convolution process mathematic(al) representation are as follows:
WhereinFor l j-th of element of layer, MjFor j-th of convolution region of l-1 layers of characteristic pattern,For member therein
Element,For the weight matrix of corresponding convolution kernel,For bias term, f (x) is activation primitive, mathematic(al) representation are as follows:
F (x)=max (0, lg (1+ex))
Pond layer often connects behind convolutional layer, carries out dimensionality reduction to characteristic pattern, while keeping characteristic dimension to a certain extent
Invariance.On the layer of pond, pond is carried out for the region n × n in each nonoverlapping size to the characteristic pattern of convolutional layer output
Operation, chooses the maximum value or average value on each region, finally makes to export the n that image all reduces on two dimensions
Times.
After input picture is propagated by the alternating of multiple convolutional layers and pond layer, it is directed to and extracts by full connection layer network
Feature classify.On full articulamentum, input is the weighted summation of one-dimensional characteristic vector of all characteristic pattern expansion and leads to
It crosses after activation primitive and to obtain:
yk=f (wkxk-1+bk)
Wherein k is the signal of network layer, ykFor the output of full articulamentum, xk-1It is the one-dimensional characteristic vector of expansion, wkFor power
Weight coefficient, bkFor bias term.
The process that reverse link parameter updates be it is layer-by-layer update convolutional neural networks can learning parameter, expression formula are as follows:
Wherein w ' and b ' is updated weight and biasing, and w and b are existing weight and biasing, and η is learning rate ginseng
Number, E are the loss for minimizing network.
4) the convolutional neural networks model for having learnt parameter, is applied to test set, exports fault identification result；
The present invention is described in detail above, specific case used herein is to the principle of the present invention and embodiment party
Formula is expounded, and the above description of the embodiment is only used to help understand the method for the present invention and its core ideas；Meanwhile it is right
In those of ordinary skill in the art, according to the thought of the present invention, change is had in specific embodiments and applications
Place, in conclusion the contents of this specification are not to be construed as limiting the invention.
Claims (4)
1. a kind of wheel tread damage fault diagnostic method based on deep learning, it is characterised in that: this method be directed to have compared with
It is strong non-stationary and easily by the wheel tread vibration signal of strong background noise jamming feature, it proposes based on the change of Fourier in short-term
The method for diagnosing faults with convolutional neural networks is changed, Fault Pattern Recognition end to end is realized；It is carried out based on deep learning
Wheel tread damage fault diagnostic method mainly comprise the steps of:
1) wheel tread fault vibration signal mechanism model, is established；
2) Short Time Fourier Transform, is carried out to wheel tread vibration signal, time-frequency spectrum sample is obtained, is divided into training set and test
Collection；
3), training set is inputted in convolutional neural networks and is learnt, after convolutional layer, pond layer and full articulamentum, output
For the category of fault type, it is divided into propagated forward process and reverse link parameter renewal process；
4) the convolutional neural networks model for having learnt parameter, is applied to test set, exports fault identification result.
2. a kind of wheel tread damage fault diagnostic method based on deep learning according to claim 1, feature exist
In: it is described to establish wheel tread fault vibration signal mechanism model, specifically wheel tread failure and vibratory impulse model, wheel pair
Tyre tread is if there is failure, then when wheel rolling is at failure, wheel couples moment with rail and impacts, this period
Property impact so that vehicle and rail is generated coupled vibrations, it is damaged, the frequency of impact for the failure wheel that a radius is R
Formula are as follows:
P=v/ (2 π R)
Wherein P is frequency of impact, and v is the speed of service, and R is radius of wheel.
3. a kind of wheel tread damage fault diagnostic method based on deep learning according to claim 1, feature exist
In: it is described that Short Time Fourier Transform is carried out to wheel tread vibration signal, it is the window function using regular length to time-domain signal
It is intercepted, and Fourier transformation is carried out to the signal that interception obtains, obtain the part frequency near moment t on the very little period
Spectrum, by translation of the window function on entire time shaft, final transformation obtains the set of upper local spectrum of each period, in short-term
Fourier transformation (STFT) is the two-dimensional function about time and frequency.Basic operation formula is as follows:
Wherein f (t) is time-domain signal, the time window at τ moment is located at centered on g (t- τ), ω is frequency, and j is imaginary unit.
4. a kind of wheel tread damage fault diagnostic method based on deep learning according to claim 1, feature exist
In: the implementation of the step 3) are as follows: propagated forward process is mainly three parts: convolutional layer, pond layer and full articulamentum；
On convolutional layer, convolution is carried out with multiple convolution kernels and input picture, in addition can by an activation primitive after biasing
To obtain series of features figure, convolution process mathematic(al) representation are as follows:
WhereinFor l j-th of element of layer, MjFor j-th of convolution region of layer characteristic pattern,For element therein,For
The weight matrix of corresponding convolution kernel,For bias term, f (x) is activation primitive, mathematic(al) representation are as follows:
F (x)=max (0, lg (1+ex))
Pond layer often connects behind convolutional layer, carries out dimensionality reduction to characteristic pattern, while keeping characteristic dimension not to a certain extent
Denaturation carries out pondization operation in each nonoverlapping size to the characteristic pattern of convolutional layer output on the layer of pond for the region n × n,
The maximum value or flat l-1 mean value on each region are chosen, finally makes to export n times that image all reduces on two dimensions；
After input picture is propagated by the alternating of multiple convolutional layers and pond layer, by full connection layer network for the spy extracted
Sign is classified；On full articulamentum, input is the weighted summation of one-dimensional characteristic vector of all characteristic pattern expansion and passes through sharp
It is obtained after function living:
yk=f (wkxk-1+bk)
Wherein k is the signal of network layer, ykFor the output of full articulamentum, xk-1It is the one-dimensional characteristic vector of expansion, wkFor weight system
Number, bkFor bias term；
The process that reverse link parameter updates be it is layer-by-layer update convolutional neural networks can learning parameter, expression formula are as follows:
Wherein w ' and b ' is updated weight and biasing, and w and b are existing weight and biasing, and η is learning rate parameter, and E is
Minimize the loss of network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910109045.5A CN109774740A (en) | 2019-02-03 | 2019-02-03 | A kind of wheel tread damage fault diagnostic method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910109045.5A CN109774740A (en) | 2019-02-03 | 2019-02-03 | A kind of wheel tread damage fault diagnostic method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109774740A true CN109774740A (en) | 2019-05-21 |
Family
ID=66503181
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910109045.5A Pending CN109774740A (en) | 2019-02-03 | 2019-02-03 | A kind of wheel tread damage fault diagnostic method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109774740A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110261139A (en) * | 2019-06-12 | 2019-09-20 | 中国神华能源股份有限公司 | Wheel tread flat recognition methods and identification device |
| CN110597240A (en) * | 2019-10-24 | 2019-12-20 | 福州大学 | Hydroelectric generating set fault diagnosis method based on deep learning |
| CN111055881A (en) * | 2019-12-31 | 2020-04-24 | 南京工大桥隧与轨道交通研究院有限公司 | Wheel-rail interface damage evolution monitoring method based on noise signals |
| CN111413075A (en) * | 2020-04-02 | 2020-07-14 | 重庆交通大学 | Fan base bolt loosening diagnosis method of multi-scale one-dimensional convolution neural network |
| CN111680665A (en) * | 2020-06-28 | 2020-09-18 | 湖南大学 | Motor mechanical fault diagnosis method based on data driving and adopting current signals |
| CN112231975A (en) * | 2020-10-13 | 2021-01-15 | 中国铁路上海局集团有限公司南京供电段 | Data modeling method and system based on reliability analysis of railway power supply equipment |
| CN114056381A (en) * | 2021-11-24 | 2022-02-18 | 西南交通大学 | Railway vehicle wheel flat scar monitoring method |
| CN114233581A (en) * | 2021-12-13 | 2022-03-25 | 山东神戎电子股份有限公司 | Intelligent patrol alarm system for fan engine room |
| CN116204821A (en) * | 2023-04-27 | 2023-06-02 | 昆明轨道交通四号线土建项目建设管理有限公司 | Vibration evaluation method and system for rail transit vehicle |
| CN117708574A (en) * | 2024-02-02 | 2024-03-15 | 江苏南高智能装备创新中心有限公司 | CNN variable-speed rolling bearing fault diagnosis method embedded with physical information |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN201615872U (en) * | 2010-02-11 | 2010-10-27 | 广州市地下铁道总公司 | Wheel tread flaw detection device |
| CN105929025A (en) * | 2016-06-07 | 2016-09-07 | 同济大学 | Wheel tread and rail fault detection method based on time and space continuity |
| CN106845529A (en) * | 2016-12-30 | 2017-06-13 | 北京柏惠维康科技有限公司 | Image feature recognition methods based on many visual field convolutional neural networks |
| CN106874957A (en) * | 2017-02-27 | 2017-06-20 | 苏州大学 | A kind of Fault Diagnosis of Roller Bearings |
| CN106920224A (en) * | 2017-03-06 | 2017-07-04 | 长沙全度影像科技有限公司 | A kind of method for assessing stitching image definition |
| CN107560849A (en) * | 2017-08-04 | 2018-01-09 | 华北电力大学 | A kind of Wind turbines Method for Bearing Fault Diagnosis of multichannel depth convolutional neural networks |
| CN108613802A (en) * | 2018-05-10 | 2018-10-02 | 重庆大学 | A kind of mechanical failure diagnostic method based on depth mixed network structure |
| CN108830127A (en) * | 2018-03-22 | 2018-11-16 | 南京航空航天大学 | A kind of rotating machinery fault feature intelligent diagnostic method based on depth convolutional neural networks structure |
| US20180357542A1 (en) * | 2018-06-08 | 2018-12-13 | University Of Electronic Science And Technology Of China | 1D-CNN-Based Distributed Optical Fiber Sensing Signal Feature Learning and Classification Method |
-
2019
- 2019-02-03 CN CN201910109045.5A patent/CN109774740A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN201615872U (en) * | 2010-02-11 | 2010-10-27 | 广州市地下铁道总公司 | Wheel tread flaw detection device |
| CN105929025A (en) * | 2016-06-07 | 2016-09-07 | 同济大学 | Wheel tread and rail fault detection method based on time and space continuity |
| CN106845529A (en) * | 2016-12-30 | 2017-06-13 | 北京柏惠维康科技有限公司 | Image feature recognition methods based on many visual field convolutional neural networks |
| CN106874957A (en) * | 2017-02-27 | 2017-06-20 | 苏州大学 | A kind of Fault Diagnosis of Roller Bearings |
| CN106920224A (en) * | 2017-03-06 | 2017-07-04 | 长沙全度影像科技有限公司 | A kind of method for assessing stitching image definition |
| CN107560849A (en) * | 2017-08-04 | 2018-01-09 | 华北电力大学 | A kind of Wind turbines Method for Bearing Fault Diagnosis of multichannel depth convolutional neural networks |
| CN108830127A (en) * | 2018-03-22 | 2018-11-16 | 南京航空航天大学 | A kind of rotating machinery fault feature intelligent diagnostic method based on depth convolutional neural networks structure |
| CN108613802A (en) * | 2018-05-10 | 2018-10-02 | 重庆大学 | A kind of mechanical failure diagnostic method based on depth mixed network structure |
| US20180357542A1 (en) * | 2018-06-08 | 2018-12-13 | University Of Electronic Science And Technology Of China | 1D-CNN-Based Distributed Optical Fiber Sensing Signal Feature Learning and Classification Method |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110261139A (en) * | 2019-06-12 | 2019-09-20 | 中国神华能源股份有限公司 | Wheel tread flat recognition methods and identification device |
| CN110597240B (en) * | 2019-10-24 | 2021-03-30 | 福州大学 | Hydroelectric generating set fault diagnosis method based on deep learning |
| CN110597240A (en) * | 2019-10-24 | 2019-12-20 | 福州大学 | Hydroelectric generating set fault diagnosis method based on deep learning |
| CN111055881A (en) * | 2019-12-31 | 2020-04-24 | 南京工大桥隧与轨道交通研究院有限公司 | Wheel-rail interface damage evolution monitoring method based on noise signals |
| CN111413075A (en) * | 2020-04-02 | 2020-07-14 | 重庆交通大学 | Fan base bolt loosening diagnosis method of multi-scale one-dimensional convolution neural network |
| CN111680665A (en) * | 2020-06-28 | 2020-09-18 | 湖南大学 | Motor mechanical fault diagnosis method based on data driving and adopting current signals |
| CN112231975A (en) * | 2020-10-13 | 2021-01-15 | 中国铁路上海局集团有限公司南京供电段 | Data modeling method and system based on reliability analysis of railway power supply equipment |
| CN114056381A (en) * | 2021-11-24 | 2022-02-18 | 西南交通大学 | Railway vehicle wheel flat scar monitoring method |
| CN114233581A (en) * | 2021-12-13 | 2022-03-25 | 山东神戎电子股份有限公司 | Intelligent patrol alarm system for fan engine room |
| CN116204821A (en) * | 2023-04-27 | 2023-06-02 | 昆明轨道交通四号线土建项目建设管理有限公司 | Vibration evaluation method and system for rail transit vehicle |
| CN116204821B (en) * | 2023-04-27 | 2023-08-11 | 昆明轨道交通四号线土建项目建设管理有限公司 | Vibration evaluation method and system for rail transit vehicle |
| CN117708574A (en) * | 2024-02-02 | 2024-03-15 | 江苏南高智能装备创新中心有限公司 | CNN variable-speed rolling bearing fault diagnosis method embedded with physical information |
| CN117708574B (en) * | 2024-02-02 | 2024-04-12 | 江苏南高智能装备创新中心有限公司 | CNN variable-speed rolling bearing fault diagnosis method embedded with physical information |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109774740A (en) | A kind of wheel tread damage fault diagnostic method based on deep learning | |
| Sadoughi et al. | Physics-based convolutional neural network for fault diagnosis of rolling element bearings | |
| Zhou et al. | A novel method based on nonlinear auto-regression neural network and convolutional neural network for imbalanced fault diagnosis of rotating machinery | |
| Zou et al. | Fault diagnosis on the bearing of traction motor in high-speed trains based on deep learning | |
| Jia et al. | GTFE-Net: A gramian time frequency enhancement CNN for bearing fault diagnosis | |
| Wu et al. | Faulted gear identification of a rotating machinery based on wavelet transform and artificial neural network | |
| CN109632309A (en) | Based on the rolling bearing fault intelligent diagnosing method for improving S-transformation and deep learning | |
| CN108830127A (en) | A kind of rotating machinery fault feature intelligent diagnostic method based on depth convolutional neural networks structure | |
| CN106124212A (en) | Based on sparse coding device and the Fault Diagnosis of Roller Bearings of support vector machine | |
| CN110672343A (en) | Rotary machine fault diagnosis method based on multi-attention convolutional neural network | |
| CN108827605A (en) | A kind of mechanical breakdown characteristic automatic extraction method based on improvement sparseness filtering | |
| CN102721545A (en) | Rolling bearing failure diagnostic method based on multi-characteristic parameter | |
| Chen et al. | Fault feature extraction and diagnosis of gearbox based on EEMD and deep briefs network | |
| CN116226646B (en) | Method, system, equipment and medium for predicting health state and residual life of bearing | |
| CN107543722A (en) | The Rolling Bearing Fault Character extracting method of dictionary learning is stacked based on depth | |
| CN112364706A (en) | Small sample bearing fault diagnosis method based on class imbalance | |
| Duan et al. | A novel classification method for flutter signals based on the CNN and STFT | |
| CN112926728A (en) | Small sample turn-to-turn short circuit fault diagnosis method for permanent magnet synchronous motor | |
| Grezmak et al. | Multi-stream convolutional neural network-based fault diagnosis for variable frequency drives in sustainable manufacturing systems | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| Liu et al. | A Rolling Bearing Fault Diagnosis‐Optimized Scale‐Space Representation for the Empirical Wavelet Transform | |
| Han et al. | Data-enhanced stacked autoencoders for insufficient fault classification of machinery and its understanding via visualization | |
| Chen et al. | Gear compound fault detection method based on improved multiscale permutation entropy and local mean decomposition | |
| CN110222386A (en) | A kind of planetary gear degenerate state recognition methods | |
| Ding et al. | Deep time–frequency learning for interpretable weak signal enhancement of rotating machineries |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190521 |