CN107505837A - A kind of semi-supervised neural network model and the soft-measuring modeling method based on the model - Google Patents

A kind of semi-supervised neural network model and the soft-measuring modeling method based on the model Download PDF

Info

Publication number
CN107505837A
CN107505837A CN201710551671.0A CN201710551671A CN107505837A CN 107505837 A CN107505837 A CN 107505837A CN 201710551671 A CN201710551671 A CN 201710551671A CN 107505837 A CN107505837 A CN 107505837A
Authority
CN
China
Prior art keywords
layer
network model
neural network
semi
variable
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
CN201710551671.0A
Other languages
Chinese (zh)
Inventor
葛志强
李�浩
宋执环
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.)
Zhejiang University ZJU
Original Assignee
Zhejiang University ZJU
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 Zhejiang University ZJU filed Critical Zhejiang University ZJU
Priority to CN201710551671.0A priority Critical patent/CN107505837A/en
Publication of CN107505837A publication Critical patent/CN107505837A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Data Mining & Analysis (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Automation & Control Theory (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a kind of soft-measuring modeling method based on semi-supervised neural network model, the model is divided into three layers, first layer is input layer, the second layer is hidden layer, third layer is output layer, output layer is divided into self-encoding encoder output layer and neural network model output layer, self-encoding encoder and neural network model share input layer and hidden layer, the modeling method is made up of self-encoding encoder and neutral net, it is few can effectively to have solved exemplar, caused by unlabeled exemplars are more the problem of soft sensor modeling inaccuracy, so as to establish more accurate semi-supervised soft-sensing model, the monitoring of implementation process and corresponding control.

Description

A kind of semi-supervised neural network model and the soft-measuring modeling method based on the model
Technical field
The invention belongs to industrial process prediction and control field, it is related to a kind of semi-supervised neural network model and based on the mould The soft-measuring modeling method of type.
Background technology
In the industrial processes of reality, more or less critical process variables often be present can not realize online inspection Survey, in order to solve this problem, by being easier the variable of detection in gatherer process, according to certain optimal quasi- side, construct A kind of using these variables as input, critical process variables are the mathematical modeling of output, realize and the online of critical process variables is estimated Meter, this is the soft sensor modeling commonly used in industrial process.
The development of statistic processes soft sensor modeling is extremely notable for the demand of large-scale industrial data.However, soft survey Many problems also be present at present in amount modeling.The complexity of system is also increasingly to improve in industrial processes, in process data Non-linear relation is more and more prominent, if establishing soft-sensing model still with traditional linear method, undoubtedly not competent change The task of amount Accurate Prediction is directed to non-linear process characteristic, has the models such as neutral net kernel method, the nerve net in numerous models The adaptability of network model and the capability of fitting of non-linear process are all extremely strong, can accurately complete the variable prediction of industrial process Task.
At the same time, in many cases in Machine Learning Problems have exemplar extremely precious and very rare, no mark Signed-off sample is originally readily available but handmarking's process is again difficult.How fully to extract useful information in no label data with up to To lift scheme performance, then semi-supervised field increasingly obtains the concern and attention of people.
The content of the invention
Exemplar is few, more than unlabeled exemplars and the problems such as process is non-linear serious for having in current industrial process, this Invention proposes a kind of soft-measuring modeling method based on semi-supervised neutral net, and this method is by self-encoding encoder and neutral net mould Type is combined the semi-supervised soft sensor modeling for carrying out industrial process, realizes the accurate On-line Estimation of critical process variables, Concrete technical scheme is as follows：
A kind of semi-supervised neural network model, described model are made up of self-encoding encoder and neutral net, are divided into three layers, and One layer is input layer, and the second layer is hidden layer, and third layer is output layer, self-encoding encoder and neural network model share input layer and Hidden layer, and output layer is divided into self-encoding encoder output layer and neural network model output layer, input layer input variable is x, input Weight and biasing respectively ω of the layer to hidden layer1And b1, the weight of hidden layer to neutral net output layer and it is biased to ωyWith by, the weight of hidden layer to self-encoding encoder output layer and it is biased to ω2And b2, self-encoding encoder output layer output reconstruction value be Neural network model output layer output predicted value be
A kind of soft-measuring modeling method based on above-mentioned semi-supervised neural network model, its step are as follows：
Step 1：Collect the training dataset of the data composition modeling of history industrial process, described training dataset Both included there are label data collection L, L ∈ R comprising leading variable or comprising auxiliary variablen×d, also include only comprising auxiliary variable Without label data collection U, U ∈ RN×M, n indicates the data sample number of label data collection, and d represents process variable number, and R is real Manifold, N indicate the data sample number of no label data collection, and M indicates the number of the auxiliary variable of no label data collection；
Step 2：The training dataset being collected into is standardized, is 0 by process variable chemical conversion average, variance be 1 it is new Data set；
Step 3：After standardization have label data collection and without label data concentrate auxiliary variable xlAnd xuAs mould The input variable x of type, there is the leading variable that label data is concentrated as output variable y using after standardization, carry out semi-supervised god Trained through network model, so as to obtain the predicted value of the leading variable of semi-supervised neural network model output layer outputWith it is self-editing The reconstruction value corresponding to input variable x of code device model output layer outputFurther obtain the semi-supervised neural network model Whole prediction error：
E=σ * Eae+(1-σ)*Enn+λEweight
Wherein, EaeThe reconstructed error of self-encoding encoder is represented,
EnnThe prediction error of neutral net is represented,
Represent the regularization constraint for weight；σ controls EaeAnd EnnBetween it is flat Weighing apparatus,λ is regularization coefficient, takes empirical value；
Step 4：Associated weight and the gradient of biasing are calculated using back-propagation algorithm；
Step 5：Semi-supervised neural network model is constantly trained according to gradient descent method, calculates the optimal ginseng of the model Number, completes semi-supervised neural network model modeling process；
Step 6：Collect new industrial process data, repeat step one to two, and by the industrial process data generation after processing Enter in the semi-supervised neural network model to after optimization, obtain the predicted value of leading variableSo as to implementation process monitoring and Control.
Brief description of the drawings
Fig. 1 is semi-supervised Artificial Neural Network Structures figure；
Fig. 2 is debutanizing tower procedure structure；
Fig. 3 represents sample actual value and semi-supervised Neural Network model predictive value in the case where having label ratio for 5% Design sketch；
Fig. 4 represents the predicted value of sample actual value and traditional neural network model in the case where having label ratio for 5% Design sketch；
Embodiment
The present invention is discussed further with reference to specific embodiment.
A kind of semi-supervised neural network model, described model are made up of self-encoding encoder and neutral net, are divided into three layers, and One layer is input layer, and the second layer is hidden layer, and third layer is output layer, self-encoding encoder and neural network model share input layer and Hidden layer, and output layer is divided into self-encoding encoder output layer and neural network model output layer, input layer input variable is x, input Weight and biasing respectively ω of the layer to hidden layer1And b1, the weight of hidden layer to neutral net output layer and it is biased to ωyWith by, the weight of hidden layer to self-encoding encoder output layer and it is biased to ω2And b2, self-encoding encoder output layer output reconstruction value be Neural network model output layer output predicted value be
A kind of soft-measuring modeling method based on above-mentioned semi-supervised neural network model, its step are as follows：
Step 1：Collect the training dataset of the data composition modeling of history industrial process, described training dataset Both included there are label data collection L, L ∈ R comprising leading variable or comprising auxiliary variablen×d, also include only comprising auxiliary variable Without label data collection U, U ∈ RN×M, n indicates the data sample number of label data collection, and d represents process variable number, and R is real Manifold, N indicate the data sample number of no label data collection, and M indicates the number of the auxiliary variable of no label data collection；
Step 2：The training dataset being collected into is standardized, is 0 by process variable chemical conversion average, variance be 1 it is new Data set；
Step 3：After standardization have label data collection and without label data concentrate auxiliary variable xlAnd xuAs mould The input variable x of type, there is the leading variable that label data is concentrated as output variable y using after standardization, carry out semi-supervised god Trained through network model, so as to obtain the predicted value of the leading variable of semi-supervised neural network model output layer outputWith it is self-editing The reconstruction value corresponding to input variable x of code device model output layer outputFurther obtain the semi-supervised neural network model Whole prediction error：
E=σ * Eae+(1-σ)*Enn+λEweight
Wherein, EaeThe reconstructed error of self-encoding encoder is represented,
EnnThe prediction error of neutral net is represented,
Represent the regularization constraint for weight；σ controls EaeAnd EnnBetween it is flat Weighing apparatus,λ is regularization coefficient, takes empirical value；
Step 4：Associated weight and the gradient of biasing are calculated using back-propagation algorithm；
(1) error parameter of self-encoding encoder output layer is introducedWherein subscript 3 represents output layer, subscript j Represent j-th of neuron of output layer, zjThe weighting input value of j-th of neuron of hidden layer is represented,Its In,The weight on k-th of neuron of input layer to the connection of j-th of neuron of hidden layer is represented,Represent hidden layer jth The biasing of individual neuron, xkRepresent the input of k-th of neuron of input layer；
(2) hidden layer is obtained to the weight of self-encoding encoder output layer and the gradient of biasing according to back-propagation algorithmWhereinRepresent k-th of neuron of hidden layer to j-th of god of output layer of self-encoding encoder Weight in connection through member,Represent the biasing of j-th of neuron of output layer of self-encoding encoder, akRepresent hidden layer k-th The output of neuron；
(3) error of neutral net output layer is introducedWherein, subscript 3 represents output layer, subscript j Represent j-th of neuron of output layer；
(4) hidden layer is obtained to neutral net output layer weight and the gradient of biasing according to back-propagation algorithm WhereinRepresent k-th of neuron of hidden layer to j-th of nerve of output layer of neutral net Weight in the connection of member,Represent the biasing of j-th of neuron of output layer of neutral net, akRepresent k-th of god of hidden layer Output through member；
(5) input layer is calculated to the error of hidden layer：
The loss function used for hidden layer when calculation error is overall prediction error, is calculating hidden layer Error when will in two kinds of situation, one kind is that have label data, and one kind is no label data.
A) error for having label data had both come from the prediction error of neutral net, and the also reconstruct from self-encoding encoder misses Difference, it is calculated as follows：
The error for having label data of j-th of neuron of hidden layer is represented,Represent self-encoding encoder output layer kth The error of individual neuron,The error of neutral net k-th of neuron of output layer is represented,Represent j-th of nerve of hidden layer Member to the weight in the connection of k-th of neuron of output layer of self-encoding encoder,Represent j-th of neuron of hidden layer to nerve net Weight in the connection of k-th of neuron of output layer of network, f'(zj) represent hidden layer neuron activation primitive derivative；
B) error without label data only comes from the reconstructed error of self-encoding encoder：
Represent the error without label data of j-th of neuron of hidden layer；
C) input layer is calculated to hidden layer weight and the gradient of biasing
Wherein, xk,uIndicate the input value of k-th of neuron of input layer of no label data, xk,lIndicate label data The input value of k-th of neuron of input layer；
Step 5：Semi-supervised neural network model is constantly trained according to gradient descent method, calculates the optimal ginseng of the model Number, completes semi-supervised neural network model modeling process.
Step 6：Collect new industrial process data, repeat step one to two, and by the industrial process data generation after processing Enter in the semi-supervised neural network model to after optimization, obtain the predicted value of leading variableSo as to the monitoring and control of implementation process System.
In order to which the structure of semi-supervised neural network model is better described, it is assumed that input variable x, input layer Number is 3, and neuron number is 4 in hidden layer, because self-encoding encoder is reconstruct input variable x, the output god of self-encoding encoder Identical with input through first number, the output neuron number of neural network model is 2, semi-supervised neural network model knot now Structure is as shown in Figure 1.
Illustrate the performance of semi-supervised neural network model below in conjunction with the example of a specific debutanizing tower.Debutanization Tower is a conventional normal industry process platform for being used for soft sensor modeling proof of algorithm.Debutanizing tower is refining process In an important device, structure is as shown in Fig. 2 the purpose of the device is to remove propane and butane in naphtha gases Process debutanizing tower, the butane content of bottom of towe is a highly important key index, in order to improve the control matter of debutanizing tower Amount to bottom of towe butane content, it is necessary to establish soft-sensing model.
Table 1 gives 7 auxiliary variables selected by for Key Quality variable butane content, respectively tower top temperature, Tower top pressure, return flow, next stage flow, the temperature of sensitive plate, column bottom temperature and tower bottom pressure.For the process, continuously Constant duration acquires 2394 process datas, wherein 1197 data are modeled as training sample, and it is corresponding for it Butane content value carry out off-line analysis and mark.1197 data samples gathered in addition are used for verifying this as test sample The validity of the semi-supervised neural network model of invention.During training set and test set is chosen, employ every empty two Individual adjacent sample point includes the mode of the interval sampling of training set and test set respectively.Certain ratio is randomly selected in training set For the data of example as there is exemplar, training set, which removes, has exemplar is remaining to be used as unlabeled exemplars.
Table 1：Input variable explanation
Input variable Variable description
X1 Tower top temperature
X2 Tower top pressure
X3 Capacity of returns
X4 Next stage flow
X5 6th piece of column plate temperature
X6 Column bottom temperature 1
X7 Column bottom temperature 2
In order to evaluate the precision of prediction of semi-supervised neural network model, error criterion root mean square is defined in the conventional mode Error (RMSE), calculation formula is as follows：
Wherein M is test sample number, yjFor the actual value of leading variable,For the semi-supervised neutral net of leading variable Model predication value.
In Fig. 3-Fig. 4, Fig. 3 represents the predicted value of semi-supervised neural network model and the curve of actual value, and Fig. 4 represents to pass The predicted value of neural network model of uniting and the curve of actual value, pass through Fig. 3-Fig. 4, it can be seen that semi-supervised nerve net of the invention The fitting effect of network model is more preferable, while the RMSE=0.16261 of the semi-supervised neural network model of model of the present invention, and traditional The RMSE=0.24076 of neutral net, the precision of prediction of semi-supervised neural network model are higher than traditional neural network model. The model of the present invention is better than traditional neural network model, and precision is also further improved.

Claims (2)

1. a kind of semi-supervised neural network model, described model are made up of self-encoding encoder and neutral net, it is divided into three layers, first Layer be input layer, and the second layer is hidden layer, and third layer is output layer, self-encoding encoder and the shared input layer of neural network model and hidden Layer is hidden, and output layer is divided into self-encoding encoder output layer and neural network model output layer, input layer input variable is x, input layer It is respectively ω to the weight of hidden layer and biasing1And b1, the weight of hidden layer to neutral net output layer and it is biased to ωyAnd by, Hidden layer to self-encoding encoder output layer weight and be biased to ω2And b2, the x reconstruction value of self-encoding encoder output layer output is Neural network model output layer output predicted value be
2. a kind of soft-measuring modeling method of the semi-supervised neural network model based on described in claim 1, its step are as follows：
Step 1：The training dataset of the data composition modeling of history industrial process is collected, described training dataset both wrapped Include also has label data collection L, L ∈ R comprising leading variable comprising auxiliary variablen×d, also include only comprising auxiliary variable without mark Sign data set U, U ∈ RN×M, n indicates the data sample number of label data collection, and d represents process variable number, and R is set of real numbers, N indicates the data sample number of no label data collection, and M indicates the number of the auxiliary variable of no label data collection；
Step 2：The training dataset being collected into is standardized, is 0 by process variable chemical conversion average, variance is 1 new data Collection；
Step 3：After standardization have label data collection and without label data concentrate auxiliary variable xlAnd xuAs model Input variable x, there is the leading variable that label data is concentrated as output variable y using after standardization, carry out semi-supervised nerve net Network model training, so as to obtain the prediction of the leading variable that neural network model output layer exports in semi-supervised neural network model ValueWith the reconstruction value corresponding to input variable x of self-encoding encoder model output layer outputFurther obtain the semi-supervised nerve The whole prediction error of network model：
E=σ * Eae+(1-σ)*Enn+λEweight
Wherein, EaeThe reconstructed error of self-encoding encoder is represented,
EnnThe prediction error of neutral net is represented,
Represent the regularization constraint for weight；σ controls EaeAnd EnnBetween balance,λ is regularization coefficient, takes empirical value.
Step 4：Associated weight and the gradient of biasing are calculated using back-propagation algorithm；
Step 5：Semi-supervised neural network model is constantly trained according to gradient descent method, calculates the optimized parameter of the model, it is complete Into semi-supervised neural network model modeling process；
Step 6：New industrial process data, repeat step one to two are collected, and the industrial process data after processing is updated to In semi-supervised neural network model after optimization, the predicted value of leading variable is obtainedSo as to the monitoring and control of implementation process.
CN201710551671.0A 2017-07-07 2017-07-07 A kind of semi-supervised neural network model and the soft-measuring modeling method based on the model Pending CN107505837A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710551671.0A CN107505837A (en) 2017-07-07 2017-07-07 A kind of semi-supervised neural network model and the soft-measuring modeling method based on the model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710551671.0A CN107505837A (en) 2017-07-07 2017-07-07 A kind of semi-supervised neural network model and the soft-measuring modeling method based on the model

Publications (1)

Publication Number Publication Date
CN107505837A true CN107505837A (en) 2017-12-22

Family

ID=60679786

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710551671.0A Pending CN107505837A (en) 2017-07-07 2017-07-07 A kind of semi-supervised neural network model and the soft-measuring modeling method based on the model

Country Status (1)

Country Link
CN (1) CN107505837A (en)

Cited By (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108334943A (en) * 2018-01-03 2018-07-27 浙江大学 The semi-supervised soft-measuring modeling method of industrial process based on Active Learning neural network model
CN108445752A (en) * 2018-03-02 2018-08-24 北京工业大学 A kind of random weight Artificial neural network ensemble modeling method of adaptively selected depth characteristic
CN108628164A (en) * 2018-03-30 2018-10-09 浙江大学 A kind of semi-supervised flexible measurement method of industrial process based on Recognition with Recurrent Neural Network model
CN108664706A (en) * 2018-04-16 2018-10-16 浙江大学 A kind of synthetic ammonia process primary reformer oxygen content On-line Estimation method based on semi-supervised Bayes's gauss hybrid models
CN110210495A (en) * 2019-05-21 2019-09-06 浙江大学 The XGBoost soft-measuring modeling method extracted based on parallel LSTM self-encoding encoder behavioral characteristics
CN111062118A (en) * 2019-11-18 2020-04-24 华侨大学 Multilayer soft measurement modeling system and method based on neural network prediction layering
CN111222283A (en) * 2019-10-24 2020-06-02 中国人民解放军空军工程大学 Particle size distribution modeling and control method in crystallization process
CN111245673A (en) * 2019-12-30 2020-06-05 浙江工商大学 SDN time delay sensing method based on graph neural network
CN111460606A (en) * 2020-02-21 2020-07-28 东南大学 Beam forming transmitter behavior level modeling system and method based on neural network
CN111912875A (en) * 2020-06-23 2020-11-10 宁波大学 Fractionating tower benzene content soft measurement method based on stack type Elman neural network
CN111914952A (en) * 2020-08-21 2020-11-10 山东省医学影像学研究所 AD characteristic parameter screening method and system based on deep neural network
CN112101445A (en) * 2020-09-09 2020-12-18 浙江大学 Method for forecasting continuous casting billet subsurface slag inclusion defect in real time based on supervised neural network
CN113158473A (en) * 2021-04-27 2021-07-23 昆明理工大学 Semi-supervised integrated instant learning industrial rubber compound Mooney viscosity soft measurement method
CN113485261A (en) * 2021-06-29 2021-10-08 西北师范大学 CAEs-ACNN-based soft measurement modeling method
CN113722973A (en) * 2020-05-25 2021-11-30 中国石油化工股份有限公司 Correction system and correction method of computer simulation model
CN113723650A (en) * 2020-05-25 2021-11-30 中国石油化工股份有限公司 Chemical process abnormity monitoring system based on semi-supervised model and model optimization device
CN113723649A (en) * 2020-05-25 2021-11-30 中国石油化工股份有限公司 Chemical process abnormity monitoring method based on semi-supervised model and model optimization method
CN116821695A (en) * 2023-08-30 2023-09-29 中国石油大学(华东) Semi-supervised neural network soft measurement modeling method

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101706882A (en) * 2009-11-23 2010-05-12 浙江大学 Embedded platform based neural network model online training method
CN105844331A (en) * 2015-01-15 2016-08-10 富士通株式会社 Neural network system and training method thereof
EP3054403A2 (en) * 2015-02-06 2016-08-10 Google, Inc. Recurrent neural networks for data item generation

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101706882A (en) * 2009-11-23 2010-05-12 浙江大学 Embedded platform based neural network model online training method
CN105844331A (en) * 2015-01-15 2016-08-10 富士通株式会社 Neural network system and training method thereof
EP3054403A2 (en) * 2015-02-06 2016-08-10 Google, Inc. Recurrent neural networks for data item generation

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
LE YAO 等: ""Locally Weighted Prediction Methods for Latent Factor Analysis With Supervised and Semisupervised Process Data"", 《IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING》 *
吴海燕: ""基于自动编码器的半监督表示学习与分类学习研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *

Cited By (29)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108334943A (en) * 2018-01-03 2018-07-27 浙江大学 The semi-supervised soft-measuring modeling method of industrial process based on Active Learning neural network model
CN108445752A (en) * 2018-03-02 2018-08-24 北京工业大学 A kind of random weight Artificial neural network ensemble modeling method of adaptively selected depth characteristic
CN108628164A (en) * 2018-03-30 2018-10-09 浙江大学 A kind of semi-supervised flexible measurement method of industrial process based on Recognition with Recurrent Neural Network model
CN108664706A (en) * 2018-04-16 2018-10-16 浙江大学 A kind of synthetic ammonia process primary reformer oxygen content On-line Estimation method based on semi-supervised Bayes's gauss hybrid models
CN108664706B (en) * 2018-04-16 2020-11-03 浙江大学 Semi-supervised Bayesian Gaussian mixture model-based online estimation method for oxygen content of one-stage furnace in ammonia synthesis process
CN110210495A (en) * 2019-05-21 2019-09-06 浙江大学 The XGBoost soft-measuring modeling method extracted based on parallel LSTM self-encoding encoder behavioral characteristics
CN110210495B (en) * 2019-05-21 2021-05-04 浙江大学 XGboost soft measurement modeling method based on parallel LSTM self-encoder dynamic feature extraction
CN111222283B (en) * 2019-10-24 2022-10-14 中国人民解放军空军工程大学 Particle size distribution modeling and control method in crystallization process
CN111222283A (en) * 2019-10-24 2020-06-02 中国人民解放军空军工程大学 Particle size distribution modeling and control method in crystallization process
CN111062118A (en) * 2019-11-18 2020-04-24 华侨大学 Multilayer soft measurement modeling system and method based on neural network prediction layering
CN111245673B (en) * 2019-12-30 2022-03-25 浙江工商大学 SDN time delay sensing method based on graph neural network
CN111245673A (en) * 2019-12-30 2020-06-05 浙江工商大学 SDN time delay sensing method based on graph neural network
CN111460606B (en) * 2020-02-21 2023-10-03 东南大学 Neural network-based beam forming transmitter behavior level modeling system and method thereof
CN111460606A (en) * 2020-02-21 2020-07-28 东南大学 Beam forming transmitter behavior level modeling system and method based on neural network
CN113723650A (en) * 2020-05-25 2021-11-30 中国石油化工股份有限公司 Chemical process abnormity monitoring system based on semi-supervised model and model optimization device
CN113723649A (en) * 2020-05-25 2021-11-30 中国石油化工股份有限公司 Chemical process abnormity monitoring method based on semi-supervised model and model optimization method
CN113722973A (en) * 2020-05-25 2021-11-30 中国石油化工股份有限公司 Correction system and correction method of computer simulation model
CN111912875A (en) * 2020-06-23 2020-11-10 宁波大学 Fractionating tower benzene content soft measurement method based on stack type Elman neural network
CN111912875B (en) * 2020-06-23 2024-02-13 江苏淮河化工有限公司 Fractionation tower benzene content soft measurement method based on stacked Elman neural network
CN111914952B (en) * 2020-08-21 2024-03-08 山东第一医科大学附属省立医院(山东省立医院) AD characteristic parameter screening method and system based on deep neural network
CN111914952A (en) * 2020-08-21 2020-11-10 山东省医学影像学研究所 AD characteristic parameter screening method and system based on deep neural network
CN112101445A (en) * 2020-09-09 2020-12-18 浙江大学 Method for forecasting continuous casting billet subsurface slag inclusion defect in real time based on supervised neural network
CN112101445B (en) * 2020-09-09 2023-11-28 浙江大学 Continuous casting billet subcutaneous slag inclusion defect real-time forecasting method based on supervision neural network
CN113158473A (en) * 2021-04-27 2021-07-23 昆明理工大学 Semi-supervised integrated instant learning industrial rubber compound Mooney viscosity soft measurement method
CN113158473B (en) * 2021-04-27 2022-03-15 昆明理工大学 Semi-supervised integrated instant learning industrial rubber compound Mooney viscosity soft measurement method
CN113485261B (en) * 2021-06-29 2022-06-28 西北师范大学 CAEs-ACNN-based soft measurement modeling method
CN113485261A (en) * 2021-06-29 2021-10-08 西北师范大学 CAEs-ACNN-based soft measurement modeling method
CN116821695A (en) * 2023-08-30 2023-09-29 中国石油大学(华东) Semi-supervised neural network soft measurement modeling method
CN116821695B (en) * 2023-08-30 2023-11-03 中国石油大学(华东) Semi-supervised neural network soft measurement modeling method

Similar Documents

Publication Publication Date Title
CN107505837A (en) A kind of semi-supervised neural network model and the soft-measuring modeling method based on the model
CN112101480B (en) Multivariate clustering and fused time sequence combined prediction method
CN108334943A (en) The semi-supervised soft-measuring modeling method of industrial process based on Active Learning neural network model
Azadeh et al. Improved estimation of electricity demand function by integration of fuzzy system and data mining approach
CN106951695A (en) Plant equipment remaining life computational methods and system under multi-state
CN104914723B (en) Industrial process soft-measuring modeling method based on coorinated training partial least square model
CN108320043A (en) A kind of distribution network equipment state diagnosis prediction method based on electric power big data
CN108921230A (en) Method for diagnosing faults based on class mean value core pivot element analysis and BP neural network
CN102522709B (en) Decision-making method and decision-making system for state overhaul of transformers
CN104537415A (en) Non-linear process industrial fault prediction and identification method based on compressed sensing and DROS-ELM
CN108628164A (en) A kind of semi-supervised flexible measurement method of industrial process based on Recognition with Recurrent Neural Network model
CN109635245A (en) A kind of robust width learning system
CN107798435A (en) A kind of Power Material needing forecasting method based on Text Information Extraction
CN109389238B (en) Ridge regression-based short-term load prediction method and device
CN111160626B (en) Power load time sequence control method based on decomposition fusion
CN104199441B (en) Blast furnace multi-state fault separating method based on sparse contribution plot and system
CN111178585A (en) Fault reporting amount prediction method based on multi-algorithm model fusion
CN107463994A (en) Semi-supervised flexible measurement method based on coorinated training extreme learning machine model
CN115438726A (en) Device life and fault type prediction method and system based on digital twin technology
CN112633556A (en) Short-term power load prediction method based on hybrid model
Yang et al. Remaining useful life prediction based on normalizing flow embedded sequence-to-sequence learning
CN113281229B (en) Multi-model self-adaptive atmosphere PM based on small samples 2.5 Concentration prediction method
Li et al. Instance weighting-based partial domain adaptation for intelligent fault diagnosis of rotating machinery
CN108204997A (en) Normal line oil flash point online soft sensor method
CN114239397A (en) Soft measurement modeling method based on dynamic feature extraction and local weighted deep learning

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: 20171222

RJ01 Rejection of invention patent application after publication