CN110108456A - A kind of rotating machinery health evaluating method of depth convolutional neural networks - Google Patents

A kind of rotating machinery health evaluating method of depth convolutional neural networks Download PDF

Info

Publication number
CN110108456A
CN110108456A CN201910304811.3A CN201910304811A CN110108456A CN 110108456 A CN110108456 A CN 110108456A CN 201910304811 A CN201910304811 A CN 201910304811A CN 110108456 A CN110108456 A CN 110108456A
Authority
CN
China
Prior art keywords
network
depth convolutional
neural networks
rotating machinery
training
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201910304811.3A
Other languages
Chinese (zh)
Inventor
贾民平
佘道明
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.)
Southeast University
Original Assignee
Southeast University
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 Southeast University filed Critical Southeast University
Priority to CN201910304811.3A priority Critical patent/CN110108456A/en
Publication of CN110108456A publication Critical patent/CN110108456A/en
Pending legal-status Critical Current

Links

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
    • 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
    • G01M13/02—Gearings; Transmission mechanisms
    • G01M13/028—Acoustic or vibration analysis
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00—Testing of machine parts
    • G01M13/04—Bearings
    • G01M13/045—Acoustic or vibration analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/04—Manufacturing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/04—Ageing analysis or optimisation against ageing
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Abstract

The invention discloses a kind of rotating machinery health evaluating methods of depth convolutional neural networks, including vibration signals collecting；The building of network training collection；The building of depth convolutional network；The training of depth convolutional network；The building of network test and health indicator；Health indicator evaluation.The advantage of the powerful ability in feature extraction of present invention combination deep learning, training label setting consider piecewise linearity degeneration.Original vibration signal is input in depth convolutional neural networks by the present invention, and the feature that depth convolutional neural networks extract is input in deep neural network and constructs health indicator, utilizes polynomial decay learning efficiency efficient training neural network.Energy accurate evaluation rotating machinery health status of the present invention can be widely applied to the fields such as chemical industry, metallurgy, electric power, aviation rotating machinery health evaluating, the dynamic process of these components performance degradations of energy accurate description, moreover it is possible to carry out predicting residual useful life.

Description

A kind of rotating machinery health evaluating method of depth convolutional neural networks
Technical field
The present invention relates to rotating machinery health assessment technology, especially a kind of rotating machinery of depth convolutional neural networks is strong Health appraisal procedure.
Background technique
Due to the development of advanced sensor and computer technology, a large amount of status monitoring number is had accumulated in industrial production According to data-driven method is widely used in bearing prediction, because they can be using Condition Monitoring Data come amount Change degenerative process, rather than establishes the accurate system's model being not readily available.
In general, the prediction technique of data-driven is usually made of following three steps: data acquisition, health indicator building and Predicting residual useful life.Health indicator attempts to identify and quantify history and by extracting characteristic information from the data of acquisition The degenerative process of progress.Therefore, the quality of constructed health indicator largely directly affects the pre- of data-driven The validity of survey method.From this angle, it is most important to construct the health indicator that effecting reaction mechanical equipment is degenerated.Industry Common rotating parts in scene, such as bearing, gear, rotor are the important composition components in rotating machinery, it Can health status directly affect rotating machinery and run well.These critical component well damages will lead to production and stop work, and bring Therefore huge economic losses are of great significance for equipment safety reliability service to its health status accurate evaluation.
Existing health indicator construction method is primarily present following problems: (1) extracting building health based on manual features and refer to Calibration method needs a large amount of Heuristics, and the feature rule of thumb screened is designed both for specific task； (2) deep learning method building health indicator is using fixed learning efficiency training network, and efficiency is lower；(3) health indicator structure It builds and does not account for segmentation degeneration；(4) failure threshold is difficult to determine in life prediction, and failure threshold is usually also rule of thumb to obtain , there is certain blindness.
Summary of the invention
In view of the above-mentioned deficiencies of the prior art, the technical problem to be solved by the present invention is to provide a kind of depth convolutional Neurals The rotating machinery health evaluating method of the rotating machinery health evaluating method of network, the depth convolutional neural networks can be improved institute The quality of the health indicator of building effectively assesses rotating machinery health status, and then improves rotating machinery predicting residual useful life Accuracy.
In order to solve the above technical problems, the technical solution adopted by the present invention is that: a kind of rotation of depth convolutional neural networks Mechanical health appraisal procedure, includes the following steps:
Step 1, it acquires vibration signal: the vibration signal of the critical component of rotating machinery is acquired；
Step 2, construct network training collection: the vibration signal acquired to step 1 constructs network training collection, by rotating machinery Input of the original vibration signal of critical component as convolutional neural networks, splits data into two parts, training set And test setM is the number of training sample, and T is the number of test sample.xiIt is i-th of training sample, yiIt is i-th The corresponding label of training sample.x'jIt is j-th of test sample.xi, x'jIt is all original vibration signal；
Step 3, framework depth convolutional network: network proposed by the present invention include 9 layers: 2 convolutional layers, 2 pond layers, 1 developer layer, 4 non-linear conversion layers.The input and output of model are exactly the data that step (1) obtains；
Step 4, training network: the training of entire model is instructed by Feedback error to minimize loss function Practice, the present invention proposes to train depth convolutional neural networks with polynomial decay learning efficiency；
Step 5, network test and building health indicator: by test setData be input to step (2) and train Network in, test set it is corresponding output be exactly we want evaluate rolling bearing degenerative process health indicator；
Step 6, evaluate the health indicator of building: the present invention evaluates the health indicator of building using two indices: dull Property and tendency.
In step 2, according to variation tendency of the health indicator before predicting residual useful life, mechanical degradation process is divided into not Same health status.In many industrial applications, if none is accurately based on the model of physics, it is then not possible to each The accurate health status and health indicator of time step assessment system.Under normal circumstances, engine components are worked normally in early stage, Then gradually degenerate.Current invention assumes that being worked normally during the entire process of bearing degradation in early stage, then gradually degenerate.And then it sets Surely trained label constructs the training set of deep learning network.
In step 3, X is allowedi,m-1Indicate m-1 layers of i-th of channel, the quantity of N (m-1) representation signal group.I-th channel is first It is divided intoL is that segmentation window is long.The Operation Definition of convolution is as follows:
Wherein, * indicates one-dimensional convolution.wk,i,m-1Indicate the weight in the channel i in the channel k and m-1 layers in m layers of connection, bi,m-1 It is m-1 layers of biasing.The output of convolutional layer activates to obtain by activation primitive:
F () indicates activation primitive.
After obtaining feature by the operation of convolutional layer, it is necessary to the spy obtained with down-sampled layer to a upper convolutional layer Sign is sampled, and realizes the reduction of dimension, reduces complexity.The Image Segmentation Methods Based on Features of input at nonoverlapping rectangular block, to each A rectangular block does corresponding operation, this process is called pond.What the present invention chose is maximum pond method, that is, selects image-region most Big value is as the value after the pool area:
N indicates the length in pond region,Indicate the output of m layers of j, the channel k neuron.
The whole features obtained through Chi Huahou are successively unfolded, form a line, and then constitute a feature vector.It allowsIndicate m-1 layers of output, full articulamentum FCmIt can indicate are as follows:
FCm=[O1,m-1,...,ON(m-1),m-1] (4)
N (m-1) is the number of m-1 layers of neuron, and full articulamentum output is exactly that convolutional neural networks learn the spy extracted Sign.
Finally, it is by a deep neural network that feature and health indicator that CNN is extracted, which constitute connection,.The application answers Deep neural network is made of 5 layers of neural network: 1 input layer, 3 hidden layers and 1 output layer.FCmIt is exactly deep The input of neural network is spent, hidden layer and output layer can be calculate by the following formula:
Ym+1=f (WmFCm+bm) (5)
WmThe weight that m layers and m+1 layers of connection, bmIt is biasing.F () is activation primitive.
In order to allow the output of the last layer between [0,1], the activation primitive of the last layer activates letter using sigmoid Number.Output be exactly it is desirable that label.It is now assumed that there is training setThere is M training sample, defines loss function J (W, b) are as follows:
YW,b(xi) be entire model output, yiIt is the label of setting.
In step 4, the original intention of polynomial decay, which is that network later period learning efficiency is too small in order to prevent, to be caused always at some It is vibrated in local minimum, other regions of area research for being doomed will not continue to increase can be jumped out by increasing learning efficiency suddenly. Polynomial decay learning efficiency calculation method is as follows:
Wherein, lr: current learning rate, lr0: initial learning rate, lrend: last learning rate, Sglobal: Current global study step number, Sdecay: the step number of every wheel study.
In step 6, tonality and tendency calculation formula are as follows:
The length of K expression time series.WithRespectively { Y (tk)}1:K{ T (tk)}1:KMean value.DY indicates that health refers to It is marked on the derivative at k time point.Spearman coefficient described in formula (8) can commonly used to the relationship between two variables of assessment To be used to describe monotonic function.Formula (9) assesses the growth trend of the characteristic.The evaluation index of both excellent degenerative characters is all It is confined in the range of [0,1], and is positively correlated with the performance of feature to be assessed, this makes them be highly suitable as health The measurement of index.
The utility model has the advantages that
1. considering that piecewise linearity degeneration more meets the actual degenerative process of bearing；
2. the depth convolutional neural networks that the present invention announces can effectively construct from original multi-channel vibration signal strong Kang Zhibiao, constructed health indicator are better than common convolutional neural networks method in monotonicity and tendency and are based on artificial The method that feature extraction constructs health indicator, greatly reduces demand of the expert to priori knowledge and human resources.
3. multinomial degeneration learning efficiency energy efficient training deep neural network, is greatly saved the time.
4. health indicator value constructed by solves failure threshold in life prediction and is difficult to determining ask between [0,1] Topic.
5. it is of the invention can accurate evaluation rotating machinery health status, and it is simple and easy, can be widely applied to chemical industry, metallurgy, The fields such as electric power, aviation rotating machinery health evaluating and predicting residual useful life.
Detailed description of the invention
Fig. 1 is a kind of flow chart of the rotating machinery health evaluating method of depth convolutional neural networks of the present invention.
Fig. 2 is bearing vibration signal time domain waveform；
Fig. 3 is the figure of changing of learning efficiency in training process；
Fig. 4 is the figure of changing of loss function in training process；
Fig. 5 is the bearing health indicator curve graph of the method for the present invention building.
Specific embodiment
Xia Mianjiehefutuhejuti compare Jia Shishifangshiduibenfamingzuojinyibuxiangxishuoming.
As shown in Figure 1, a kind of rotating machinery health evaluating method of depth convolutional neural networks, includes the following steps.
Step 1, vibration signals collecting: the vibration signal of the critical component of rotating machinery is acquired.Rotating machinery Critical component includes bearing, gear or rotor etc..
The acquisition method of critical component vibration signal is the prior art, in the application, by taking bearing as an example, using IEEE PHM The bearing accelerated life test data that 2012 challenge matches provide verify this method.Experimental data is real from PRONOSTIA Platform is tested, structure is as shown in Figure 2.The experiment is specially designed for the research of the fault detection of bearing, diagnosis and prediction technique.The reality Check system can carry out accelerated degradation test to bearing within a few hours.Under three kinds of different operating conditions, to 17 bearings It is tested.It is trained under each operating condition using two bearings, other bearings are for testing.Accelerometer is fixed On the outer ring of bearing, and capture vibration signal.The revolving speed for testing bearing inner race is 1800rpm, load 4kN.Sample frequency For 25.6kHz.0.1s is persistently sampled every time, every 10s repeated sampling is primary.
Step 2, network training collection and test set are constructed.
Step 21, construct network training collection: the vibration signal acquired to step 1 constructs network training collection.Choose 16 axis The data held are as training set.
Step 22, using remaining data as test set.For example, if select from bearing 1 to the data of bearing 16 as Training set then uses the data from bearing 17 as test set.Then, enabling training set is Dtrain={ (x1,y1),..., (x16,y16)}。
Step 23, training label is set, and the application considers that mechanical equipment piecewise linearity is degenerated, each life-cycle data are arranged Preceding 5% be it is normal, gradually degenerate with rear bearing.Such as: a bearing has been running for 14000s, the life-cycle of this bearing Time span is 52000s, then degrading scale is yi=ti/ (1-0.05) T=14000/ (1-0.05) * 52000=0.2834. tiIt is bearing runing time, T is time bearing life-cycle.Test set is Dtest={ (x17)}。
Step 3, framework depth convolutional network.
Step 31, convolution sum pond, network proposed by the present invention include 9 layers: 2 convolutional layers, 2 pond layers, 1 exhibition Layers apart, 4 non-linear conversion layers.Using the original vibration signal of rolling bearing as the input of convolutional neural networks, size is inputted It is 2560, convolution kernel and Chi Huahe size are respectively 60,20.
Step 32, feature connects entirely, after two layers of convolution sum, two layers of pond layer, be unfolded with flatten layers be exactly The feature learnt is input in deep neural network by the feature that convolutional neural networks learn, every layer of number of network node For [2000,500,200,50,1].
It step 33, is exactly the health indicator constructed after the output smoothing of last network.
The input and output of model are exactly the data that step (2) obtain；
Step 4, training network: the training of entire model is instructed by Feedback error to minimize loss function Practice, the present invention proposes to train depth convolutional neural networks with polynomial decay learning efficiency；By formula 7, learning efficiency is set, Initial learning efficiency is 0.01, and final learning efficiency is 0.0001, SdecayFor 500 steps, SglobalFor 3000 steps, entirely trained The variation of learning efficiency such as Fig. 4 in journey, the loss function of network training such as Fig. 5；
Step 5, network test and building health indicator: by test setData be input to step (2) and train Network in, test set it is corresponding output be exactly we want evaluate rolling bearing degenerative process health indicator.
Step 6, the health indicator of building is evaluated:
1 bearing 1-6 health indicator Monotonicity value of table and Trend value
The health indicator Monotonicity value and Trend value for the method building that the present invention announces are all bigger than its excess-three kind method, more can React the practical process degenerated of bearing.The learning efficiency of polynomial decay can effectively train network, and then learn vibration signal In feature.Health indicator based on the building of minimum quantization error method needs the help of the priori knowledge of expert and has certain Blindness.Often designed for specific task.
In short, the advantage of the powerful ability in feature extraction of present invention combination deep learning, discloses a kind of polynomial decay The rolling bearing health evaluating method of the depth convolutional neural networks of efficiency is practised, training label setting considers piecewise linearity and moves back Change.Original vibration signal is input in depth convolutional neural networks by the present invention, the spy that depth convolutional neural networks are extracted Sign, which is input in deep neural network, constructs health indicator, utilizes polynomial decay learning efficiency efficient training neural network.This Invention energy accurate evaluation rotating machinery health status, can be widely applied to the fields such as chemical industry, metallurgy, electric power, aviation rotating machinery Health evaluating, the dynamic process of these components performance degradations of energy accurate description, moreover it is possible to carry out predicting residual useful life.
The preferred embodiment of the present invention has been described above in detail, still, during present invention is not limited to the embodiments described above Detail a variety of equivalents can be carried out to technical solution of the present invention within the scope of the technical concept of the present invention, this A little equivalents all belong to the scope of protection of the present invention.

Claims (3)

1. a kind of rotating machinery health evaluating method of depth convolutional neural networks, which comprises the steps of:
Step 1, it acquires vibration signal: the vibration signal of the critical component of rotating machinery is acquired；
Step 2, construct network training collection: the vibration signal acquired to step 1 constructs network training collection, by rotating machinery key Input of the original vibration signal of component as convolutional neural networks, splits data into two parts, training setAnd survey Examination collectionWherein M is the number of training sample, and T is the number of test sample, xiIt is i-th of training sample, yiIt is i-th The corresponding label of training sample, x'jIt is j-th of test sample, xi, x'jIt is all original vibration signal；
Step 3, framework depth convolutional network: building includes the depth convolutional network of 9 layers, including 2 convolutional layers, 2 ponds Layer, 1 developer layer, 4 non-linear conversion layers, the input and output of model are exactly the data that step (1) obtains；
Step 4, training network: loss function is minimized by Feedback error to train, learns to imitate with polynomial decay Rate trains depth convolutional neural networks；
Step 5, network test and building health indicator: by test setData be input to step (2) trained network In, the corresponding output of test set is exactly the health indicator for evaluating rolling bearing degenerative process；
Step 6, it evaluates the health indicator of building: evaluating the health indicator of building using monotonicity and tendency two indices.
2. a kind of rotating machinery health evaluating method of depth convolutional neural networks according to claim 1, feature exist In: polynomial decay learning efficiency calculation method is as follows in step 4:
Wherein, lr: current learning rate, lr0: initial learning rate, lrend: last learning rate, Sglobal: current Overall situation study step number, Sdecay: the step number of every wheel study.
3. a kind of rotating machinery health evaluating method of depth convolutional neural networks according to claim 1, feature exist In: in step 6, tonality and tendency calculation formula are as follows:
K indicates the length of time series,WithRespectively { Y (tk)}1:K{ T (tk)}1:KMean value.DY indicates that health indicator exists The derivative at k time point.
CN201910304811.3A 2019-04-16 2019-04-16 A kind of rotating machinery health evaluating method of depth convolutional neural networks Pending CN110108456A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910304811.3A CN110108456A (en) 2019-04-16 2019-04-16 A kind of rotating machinery health evaluating method of depth convolutional neural networks

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910304811.3A CN110108456A (en) 2019-04-16 2019-04-16 A kind of rotating machinery health evaluating method of depth convolutional neural networks

Publications (1)

Publication Number Publication Date
CN110108456A true CN110108456A (en) 2019-08-09

Family

ID=67485563

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910304811.3A Pending CN110108456A (en) 2019-04-16 2019-04-16 A kind of rotating machinery health evaluating method of depth convolutional neural networks

Country Status (1)

Country Link
CN (1) CN110108456A (en)

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110398370A (en) * 2019-08-20 2019-11-01 贵州大学 A kind of Method for Bearing Fault Diagnosis based on HTS-CNN model
CN110633792A (en) * 2019-10-22 2019-12-31 西安交通大学 End-to-end bearing health index construction method based on convolution cyclic neural network
CN110866314A (en) * 2019-10-22 2020-03-06 东南大学 Method for predicting residual life of rotating machinery of multilayer bidirectional gate control circulation unit network
CN111562108A (en) * 2020-05-09 2020-08-21 浙江工业大学 Rolling bearing intelligent fault diagnosis method based on CNN and FCMC
CN111695209A (en) * 2020-05-13 2020-09-22 东南大学 Rotary machine small sample health assessment method driven by meta-depth learning
CN111898644A (en) * 2020-07-02 2020-11-06 西安交通大学 Intelligent identification method for health state of aerospace liquid engine under fault-free sample
CN111914883A (en) * 2020-06-18 2020-11-10 西安交通大学 Spindle bearing state evaluation method and device based on deep fusion network
CN112033463A (en) * 2020-09-02 2020-12-04 哈尔滨工程大学 Nuclear power equipment state evaluation and prediction integrated method and system
CN112347898A (en) * 2020-11-03 2021-02-09 重庆大学 Rolling bearing health index construction method based on DCAE neural network
CN112345956A (en) * 2020-11-24 2021-02-09 广州橙行智动汽车科技有限公司 Battery pack charge state detection method and device
CN112365186A (en) * 2020-11-27 2021-02-12 中国电建集团海外投资有限公司 Health degree evaluation method and system for electric power information system
CN112418277A (en) * 2020-11-03 2021-02-26 西安电子科技大学 Method, system, medium, and apparatus for predicting remaining life of rotating machine component
CN112528548A (en) * 2020-11-27 2021-03-19 东莞市汇林包装有限公司 Self-adaptive depth coupling convolution self-coding multi-mode data fusion method
CN113553903A (en) * 2021-06-11 2021-10-26 东南大学 Rotary machine health assessment method of deep time convolution network
CN113642465A (en) * 2021-08-13 2021-11-12 石家庄铁道大学 Bearing health assessment method based on relational network
CN114354184A (en) * 2021-12-28 2022-04-15 哈尔滨工业大学 Deep learning-based method and device for establishing health early warning model of main shaft of large-scale rotating equipment
CN114897007A (en) * 2022-04-26 2022-08-12 太原理工大学 Drilling machine health condition evaluation method of composite information layered convolution neural network

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102788696A (en) * 2012-07-21 2012-11-21 辽宁大学 Evaluation method for health degree of bearing on basis of improved BP (Back Propagation) neural network and fuzzy set theory
CN104392143A (en) * 2014-12-09 2015-03-04 北京四方继保自动化股份有限公司 Method for predicting fault trends of steam turbines by aid of adaptive quantum neural networks
WO2016026558A1 (en) * 2014-08-22 2016-02-25 Abb Technology Ag A method for assessing the condition of rotating machinery connected to an electric motor
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
CN109063308A (en) * 2018-07-26 2018-12-21 北京航空航天大学 A kind of health evaluating method based on depth quantum learning
CN109141881A (en) * 2018-07-06 2019-01-04 东南大学 A kind of rotating machinery health evaluating method of depth autoencoder network

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102788696A (en) * 2012-07-21 2012-11-21 辽宁大学 Evaluation method for health degree of bearing on basis of improved BP (Back Propagation) neural network and fuzzy set theory
WO2016026558A1 (en) * 2014-08-22 2016-02-25 Abb Technology Ag A method for assessing the condition of rotating machinery connected to an electric motor
CN104392143A (en) * 2014-12-09 2015-03-04 北京四方继保自动化股份有限公司 Method for predicting fault trends of steam turbines by aid of adaptive quantum 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
CN109141881A (en) * 2018-07-06 2019-01-04 东南大学 A kind of rotating machinery health evaluating method of depth autoencoder network
CN109063308A (en) * 2018-07-26 2018-12-21 北京航空航天大学 A kind of health evaluating method based on depth quantum learning

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
佘道明，贾民平，张菀: "一种新型深度自编码网络的", 《东南大学学报( 自然科学版)》 *
冯宇旭，李裕梅: "深度学习优化器方法及学习率衰减", 《数据挖掘》 *
徐振忠: "关于卷积神经网络损失函数的改进算法", 《高师理科学刊》 *
陈鹏展等: "《个体行为的机器识别与决策协同》", 31 July 2018 *

Cited By (27)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110398370B (en) * 2019-08-20 2021-02-05 贵州大学 Bearing fault diagnosis method based on HTS-CNN model
CN110398370A (en) * 2019-08-20 2019-11-01 贵州大学 A kind of Method for Bearing Fault Diagnosis based on HTS-CNN model
CN110633792A (en) * 2019-10-22 2019-12-31 西安交通大学 End-to-end bearing health index construction method based on convolution cyclic neural network
CN110866314A (en) * 2019-10-22 2020-03-06 东南大学 Method for predicting residual life of rotating machinery of multilayer bidirectional gate control circulation unit network
CN110633792B (en) * 2019-10-22 2022-03-22 西安交通大学 End-to-end bearing health index construction method based on convolution cyclic neural network
CN110866314B (en) * 2019-10-22 2022-11-15 东南大学 Method for predicting residual life of rotating machinery of multilayer bidirectional gate control circulation unit network
CN111562108A (en) * 2020-05-09 2020-08-21 浙江工业大学 Rolling bearing intelligent fault diagnosis method based on CNN and FCMC
CN111695209A (en) * 2020-05-13 2020-09-22 东南大学 Rotary machine small sample health assessment method driven by meta-depth learning
CN111914883A (en) * 2020-06-18 2020-11-10 西安交通大学 Spindle bearing state evaluation method and device based on deep fusion network
CN111898644A (en) * 2020-07-02 2020-11-06 西安交通大学 Intelligent identification method for health state of aerospace liquid engine under fault-free sample
CN111898644B (en) * 2020-07-02 2022-10-25 西安交通大学 Intelligent identification method for health state of aerospace liquid engine under fault-free sample
CN112033463A (en) * 2020-09-02 2020-12-04 哈尔滨工程大学 Nuclear power equipment state evaluation and prediction integrated method and system
CN112033463B (en) * 2020-09-02 2022-09-06 哈尔滨工程大学 Nuclear power equipment state evaluation and prediction integrated method and system
CN112418277A (en) * 2020-11-03 2021-02-26 西安电子科技大学 Method, system, medium, and apparatus for predicting remaining life of rotating machine component
CN112347898B (en) * 2020-11-03 2024-04-09 重庆大学 Rolling bearing health index construction method based on DCAE neural network
CN112418277B (en) * 2020-11-03 2024-04-09 西安电子科技大学 Method, system, medium and equipment for predicting residual life of rotating machine parts
CN112347898A (en) * 2020-11-03 2021-02-09 重庆大学 Rolling bearing health index construction method based on DCAE neural network
CN112345956A (en) * 2020-11-24 2021-02-09 广州橙行智动汽车科技有限公司 Battery pack charge state detection method and device
CN112365186A (en) * 2020-11-27 2021-02-12 中国电建集团海外投资有限公司 Health degree evaluation method and system for electric power information system
CN112528548A (en) * 2020-11-27 2021-03-19 东莞市汇林包装有限公司 Self-adaptive depth coupling convolution self-coding multi-mode data fusion method
CN113553903B (en) * 2021-06-11 2022-11-22 东南大学 Rotary machine health assessment method of deep time convolution network
CN113553903A (en) * 2021-06-11 2021-10-26 东南大学 Rotary machine health assessment method of deep time convolution network
CN113642465B (en) * 2021-08-13 2022-07-08 石家庄铁道大学 Bearing health assessment method based on relational network
CN113642465A (en) * 2021-08-13 2021-11-12 石家庄铁道大学 Bearing health assessment method based on relational network
CN114354184A (en) * 2021-12-28 2022-04-15 哈尔滨工业大学 Deep learning-based method and device for establishing health early warning model of main shaft of large-scale rotating equipment
CN114897007A (en) * 2022-04-26 2022-08-12 太原理工大学 Drilling machine health condition evaluation method of composite information layered convolution neural network
CN114897007B (en) * 2022-04-26 2024-04-19 太原理工大学 Drilling machine health condition assessment method of composite information hierarchical convolutional neural network

Similar Documents

Publication Publication Date Title
CN110108456A (en) A kind of rotating machinery health evaluating method of depth convolutional neural networks
CN110132603B (en) Ship diesel engine fault positioning method based on union confidence rule base and ant colony algorithm
CN112036547B (en) Rolling bearing residual life prediction method combining automatic feature extraction with LSTM
Martinez-Garcia et al. Visually interpretable profile extraction with an autoencoder for health monitoring of industrial systems
CN109992872B (en) Mechanical equipment residual life prediction method based on stacked separation convolution module
CN112665852B (en) Variable working condition planetary gearbox fault diagnosis method and device based on deep learning
CN108709744B (en) Motor bearings method for diagnosing faults under a kind of varying load operating condition
CN113378967B (en) Structural health monitoring multivariate data anomaly diagnosis method based on convolutional neural network and transfer learning
CN112036042A (en) Power equipment abnormality detection method and system based on variational modal decomposition
CN114705432B (en) Method and system for evaluating health state of explosion-proof motor bearing
CN113128338A (en) Intelligent diagnosis method for printing machine roller fault under small sample
CN114755017B (en) Variable-speed bearing fault diagnosis method of cross-domain data driving unsupervised field shared network
CN116793682A (en) Bearing fault diagnosis method based on iCORAL-MMD and anti-migration learning
CN116108346A (en) Bearing increment fault diagnosis life learning method based on generated feature replay
CN113916535B (en) Bearing diagnosis method, system, equipment and medium based on time frequency and CNN
Chou et al. SHM data anomaly classification using machine learning strategies: A comparative study
CN112149730B (en) High-speed train bearing fault prediction method based on oil sample vision
CN116465623B (en) Gearbox service life prediction method based on sparse converter
CN105741184A (en) Transformer state evaluation method and apparatus
Zhang et al. Fault Diagnosis with Bidirectional Guided Convolutional Neural Networks Under Noisy Labels
Li et al. Transformer-based meta learning method for bearing fault identification under multiple small sample conditions
CN114330141A (en) Fan main shaft bearing service life prediction method based on GRU (generalized regression Unit) super-parameter optimization
CN113469066A (en) Unbalanced sample multitask self-optimization rolling bearing fault migration diagnosis method
Zhou et al. Bearing life prediction method based on parallel multichannel recurrent convolutional neural network
Alqatawneh et al. A Developed Convolutional Neural Network Architecture for Condition Monitoring

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

Application publication date: 20190809

RJ01 Rejection of invention patent application after publication