CN106407550A - Soft sensor modeling method in industrial process - Google Patents

Soft sensor modeling method in industrial process Download PDF

Info

Publication number
CN106407550A
CN106407550A CN201610817977.1A CN201610817977A CN106407550A CN 106407550 A CN106407550 A CN 106407550A CN 201610817977 A CN201610817977 A CN 201610817977A CN 106407550 A CN106407550 A CN 106407550A
Authority
CN
China
Prior art keywords
data
industrial process
modeling
soft
soft sensor
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
CN201610817977.1A
Other languages
Chinese (zh)
Inventor
田慧欣
刘玉栋
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.)
Tianjin Polytechnic University
Original Assignee
Tianjin Polytechnic 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 Tianjin Polytechnic University filed Critical Tianjin Polytechnic University
Priority to CN201610817977.1A priority Critical patent/CN106407550A/en
Publication of CN106407550A publication Critical patent/CN106407550A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • 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

Landscapes

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

Abstract

Embodiments of the present invention disclose a soft sensor modeling method in an industrial process. The method comprises the steps of obtaining modeling data used for the soft sensor modeling method in the industrial process as to-be-processed data, selecting training data according to the weight, establishing a soft sensor sub-model by using a single intelligent method, and obtaining a soft sensor model of the soft sensor modeling method in the industrial process by using the soft sensor sub-model. According to the method disclosed by the embodiments of the present invention, the problem that the conventional soft sensor modeling method in the industrial process leads to inaccurate modeling and difficult operation for large industrial noise data is solved, and the soft sensor modeling method in the industrial process disclosed by the embodiments of the present invention focuses on the training of poorly trained samples, so that while the prediction accuracy is improved, the interference caused by large noise data can be overcome.

Description

A kind of industrial process soft-measuring modeling method
Technical field
The present embodiments relate to a kind of artificial intelligence technology, more particularly, to a kind of industrial process soft-measuring modeling method.
Background technology
The fast development of soft-measuring technique in recent years, the various flexible measurement methods based on artificial intelligence technology are in commercial production In be widely used, but many industrial processes contain the physical-chemical reaction process of various complexity, or Carry out under the adverse circumstances such as High Temperature High Pressure, therefore each influence factor and measured between exist complexity nonlinear dependence How system, make soft-measuring technique can improve certainty of measurement, can simplify practical operation again, be that soft-measuring technique is applied to industry The bottleneck producing.
The main thought of integrated study is to be merged many sub- learning machines to improve the performance of whole learning system, from And overcome the shortcomings of single intelligent algorithm generalization ability in soft-measuring technique is poor, model accuracy is relatively low, built with obtaining hard measurement The optimal discreet value of output variable in mould.At present, integrated study study in classification problem more, for the research of regression problem Less.And existing AdaBoost algorithm is required for greatly setting more parameter, the selection of parameter determines the performance of algorithm, but It is that the selection of these parameters does not have industrial mechanism as guidance, there is blindness, need repeatedly to test and just can draw appropriate ginseng Number, these complicated parameter settings increased practical operation difficulty in use, is that hard measurement application in the industry brings Very big difficulty.
Content of the invention
In view of this, the embodiment of the present invention provides a kind of industrial process soft-measuring modeling method, solves hard measurement in work In industry application, operation easier is big, the problem of prediction effect difference.
The embodiment of the present invention provides a kind of industrial process soft-measuring modeling method, including：
S101, obtain M modeling data for described industrial process soft-measuring modeling method as pending data；
S102, the slack variable of the described industrial process soft-measuring modeling method of setting；
S103, to described modeling data distribute weight；
S104, setting maximum iteration time T simultaneously initialize iterationses t=1；
S105, extract m described pending data as training data according to the size of described modeling data weight, its In, m≤M；
The single intelligent method of S106, selection one trains described training data to obtain a hard measurement submodel；
S107, the absolute error calculating described training data, relative error and root-mean-square error；
S108, the described pending data weight of renewal and described iterationses t+1；
If S109 described iterationses t is less than or equal to described maximum iteration time T, return to step S105；
S110, obtain the model of described industrial process soft-measuring modeling method according to T described hard measurement submodel.
In the above-mentioned methods it is preferred that described pending data also includes：
For testing the test data of soft-sensing model.
In the above-mentioned methods it is preferred that described slack variable is equal to the maximum absolute error that described industrial process allows.
In the above-mentioned methods it is preferred that single intelligent method can be BP neural network method.
Brief description
Fig. 1 is a kind of industrial process soft-measuring modeling method schematic flow sheet that the embodiment of the present invention one provides
Brief description
Further illustrate technical scheme below in conjunction with the accompanying drawings and by specific embodiment.May be appreciated It is that specific embodiment described herein is only used for explaining the present invention, rather than limitation of the invention.Further need exist for explanation It is, for the ease of description, in accompanying drawing, to illustrate only part related to the present invention rather than entire infrastructure.
Specific embodiment one
Fig. 1 is a kind of industrial process soft-measuring modeling method schematic flow sheet that the embodiment of the present invention one provides.As Fig. 1 institute Show, industrial process soft-measuring modeling method includes：
S101, obtain M modeling data for industrial process soft-measuring modeling method as pending data.
Specifically, pending data is the creation data collecting from a kind of industrial processes, for example, can be, The creation data of LF stove ladle refining production process, determines and sets up the output variable that soft-sensing model obtains, for example, can be, LF Stove ladle refining production process Molten Steel End Point, LF stove ladle refining production process Molten Steel End Point described in analyzing influence Factor, can be for example, initial temperature that molten steel enters the station, ladle state, the consumption of electric energy, refining cycle, the gross mass of molten steel, Ladle lining temperature when the heat absorption of alloy and slag charge and heat release, average BOTTOM ARGON BLOWING flow and tapping.M LF is gathered according to above-mentioned parameter The creation data of stove ladle refining production process is as modeling data the data_train=[(x setting up soft-sensing model1, y1), (x2, y2) ... (xM, yM)] and K described LF stove ladle refining production process creation data as test soft-sensing model Test data data_test=[(x1, y1), (x2, y2) ... (xK, yK)], wherein x is the influence factor of output variable, permissible Become input variable, y is output variable.Gathering modeling data number can be for example, M=200, test data number, for example Can be, K=50.
S102, the slack variable of the described industrial process soft-measuring modeling method of setting；
Specifically, described slack variable ε is equal to the maximum absolute error that described industrial process allows, and LF stove ladle refining is given birth to Product process molten steel emphasis maximum temperature absolute error is 10 DEG C, and slack variable can be set as 10 DEG C, i.e. ε=10.
S103, to described modeling data distribute weight.
Specifically, to the modeling data data_train distribution initial weight collecting：D1(i)=1/M.
Wherein D1I () is the initial weight of i-th sample, M is modeling data number.
S104, setting maximum iteration time T simultaneously initialize iterationses t=1.
Specifically, according to iteration it needs to be determined that maximum iteration time T, for example, can be T=20, initialization iteration time simultaneously Number t=1.
S105, extract m described modeling data as training data according to the size of described pending data weight, its In, m≤M.
Specifically, select a kind of sampling approach to extract and extract m modeling data from M modeling data data_train, Can be for example to extract 150 data from 200 modeling data data_train as training number by the use of layered sampling method According to the chance that weight is pumped to greatly increases, and the little chance being pumped to of weight reduces, and weight is equal, is randomly drawed.
The single intelligent method of S106, selection one trains described training data to obtain a hard measurement submodel.
Specifically, select a single intelligent method, for example, can be, BP neural network, built using MATLAB emulation Vertical BP neural network model, wherein BP neural network parameter is：Input layer number is 8, and output layer nodes are 1, hidden layer Nodes are 15, and the hard measurement submodel after foundation is：ft(x)→y.
S107, the absolute error calculating described training data, relative error and root-mean-square error.
Specifically, absolute error formula, the phase of the hard measurement submodel that each training data is set up are calculated by current iteration To error and root-mean-square error；
Absolute error is：errort(i)=| ft(xi)-yi|；
Relative error is：
Root-mean-square error is：
Wherein, ft(xi) it is that during the t time iteration, i-th training data is become by the output that hard measurement submodel calculates Amount, yiFor the output variable of i-th training data, errortI () is the absolute error of i-th training data during the t time iteration, AREtI () is the relative error of i-th training data during the t time iteration, RMSEtFor being obtained by hard measurement submodule during the t time iteration The root-mean-square error collecting data with corresponding industry of the output variable going out.
S108, the described pending data weight of renewal and described iterationses t+1；
Specifically, if the absolute error of training data is less than slack variable ε, reduced according to the corresponding weight of formula (1), such as Fruit absolute error is more than slack variable ε, is increased according to the corresponding weight of formula (1), the corresponding weight of not selected modeling data Constant；
Wherein Dt+1I () is the corresponding weight of i-th modeling data, error during next iterationtI () is the t time iteration When i-th modeling data absolute error, AREtI () is the relative error of i-th modeling data during the t time iteration；
The corresponding weight of modeling data, iterationses t+1 are updated.
If S109 described iterationses t is less than or equal to described maximum iteration time T, return to step S105.
Specifically, if iterationses t≤T return to step S105, if iterationses t ＞ T execution step S110.
S110, obtain the model of described industrial process soft-measuring modeling method according to T described hard measurement submodel.
Specifically, the soft-sensing model according to the integrated industrial process flexible measurement method of formula (2).
Wherein, the industrial soft-sensing model that F (x) sets up for the present embodiment, ftX () is the hard measurement of the t time iteration foundation Submodel, for example, can be the hard measurement submodel set up by BP neural network, RMSEtFor during the t time iteration by soft survey The root-mean-square error collecting data with corresponding industry of the output variable that quantum mould draws.
Note, above are only presently preferred embodiments of the present invention and institute's application technology principle.It will be appreciated by those skilled in the art that The invention is not restricted to specific embodiment described here, can carry out for a person skilled in the art various obvious changes, Readjust and substitute without departing from protection scope of the present invention.Therefore although being carried out to the present invention by above example It is described in further detail, but the present invention is not limited only to above example, without departing from the inventive concept, also Other Equivalent embodiments more can be included, and the scope of the present invention is determined by scope of the appended claims.

Claims (4)

1. a kind of industrial process soft-measuring modeling method is it is characterised in that include：
S101, obtain M modeling data for described industrial process soft-measuring modeling method as pending data；
S102, the slack variable of the described industrial process soft-measuring modeling method of setting；
S103, to described modeling data distribute weight；
S104, setting maximum iteration time T simultaneously initialize iterationses t=1；
S105, extract m described modeling data as training data, wherein, m≤M according to the size of described modeling data weight；
The single intelligent method of S106, selection one trains described training data to obtain a hard measurement submodel；
S107, the absolute error calculating described training data, relative error and root-mean-square error；
S108, the described pending data weight of renewal and described iterationses t+1；
If S109 described iterationses t is less than or equal to described maximum iteration time T, return to step S105；
S110, obtain the model of described industrial process soft-measuring modeling method according to T described hard measurement submodel.
2. method according to claim 1, wherein, described pending data also includes the survey for testing soft-sensing model Examination data.
3. method according to claim 1, wherein, described slack variable is equal to the maximum absolute of described industrial process permission Error.
4. method according to claim 1, wherein, single intelligent method can be BP neural network method.
CN201610817977.1A 2016-09-07 2016-09-07 Soft sensor modeling method in industrial process Pending CN106407550A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610817977.1A CN106407550A (en) 2016-09-07 2016-09-07 Soft sensor modeling method in industrial process

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610817977.1A CN106407550A (en) 2016-09-07 2016-09-07 Soft sensor modeling method in industrial process

Publications (1)

Publication Number Publication Date
CN106407550A true CN106407550A (en) 2017-02-15

Family

ID=57999563

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610817977.1A Pending CN106407550A (en) 2016-09-07 2016-09-07 Soft sensor modeling method in industrial process

Country Status (1)

Country Link
CN (1) CN106407550A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108334999A (en) * 2018-05-09 2018-07-27 山东交通学院 The failure prediction method and system of fume hot-water type BrLi chiller
CN111291657A (en) * 2020-01-21 2020-06-16 同济大学 Crowd counting model training method based on difficult case mining and application
CN112016241A (en) * 2020-07-23 2020-12-01 武汉数字化设计与制造创新中心有限公司 Soft measurement modeling method based on mLASSO-MLP model

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102175345A (en) * 2011-01-06 2011-09-07 华东理工大学 Soft measurement method for fire box temperature of multi-nozzle opposed coal water slurry gasification furnace
CN102944653A (en) * 2012-12-10 2013-02-27 天津工业大学 Soft measurement method for sizing percentage in sizing process based on Learn++
CN103018426A (en) * 2012-11-26 2013-04-03 天津工业大学 Soft measurement method for sizing percentage during yarn-sizing process based on Bagging
CN104102837A (en) * 2014-07-11 2014-10-15 天津工业大学 Incremental learning integrated algorithm for soft measurement modeling
CN104462797A (en) * 2014-11-24 2015-03-25 天津工业大学 Increment integration algorithm used for procedure parameter online testing

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102175345A (en) * 2011-01-06 2011-09-07 华东理工大学 Soft measurement method for fire box temperature of multi-nozzle opposed coal water slurry gasification furnace
CN103018426A (en) * 2012-11-26 2013-04-03 天津工业大学 Soft measurement method for sizing percentage during yarn-sizing process based on Bagging
CN102944653A (en) * 2012-12-10 2013-02-27 天津工业大学 Soft measurement method for sizing percentage in sizing process based on Learn++
CN104102837A (en) * 2014-07-11 2014-10-15 天津工业大学 Incremental learning integrated algorithm for soft measurement modeling
CN104462797A (en) * 2014-11-24 2015-03-25 天津工业大学 Increment integration algorithm used for procedure parameter online testing

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108334999A (en) * 2018-05-09 2018-07-27 山东交通学院 The failure prediction method and system of fume hot-water type BrLi chiller
CN108334999B (en) * 2018-05-09 2024-02-27 山东交通学院 Fault prediction method and system for flue gas hot water type lithium bromide refrigerating unit
CN111291657A (en) * 2020-01-21 2020-06-16 同济大学 Crowd counting model training method based on difficult case mining and application
CN111291657B (en) * 2020-01-21 2022-09-16 同济大学 Crowd counting model training method based on difficult case mining and application
CN112016241A (en) * 2020-07-23 2020-12-01 武汉数字化设计与制造创新中心有限公司 Soft measurement modeling method based on mLASSO-MLP model
CN112016241B (en) * 2020-07-23 2023-09-29 武汉数字化设计与制造创新中心有限公司 Soft measurement modeling method based on mLASSO-MLP model

Similar Documents

Publication Publication Date Title
CN104699894B (en) Gaussian process based on real-time learning returns multi-model Fusion Modeling Method
Zhang et al. Local parameter optimization of LSSVM for industrial soft sensing with big data and cloud implementation
CN110378044B (en) Multi-time scale convolution neural network soft measurement method based on attention mechanism
CN103302777B (en) Neural-network-based energy consumption anomaly detection method of tire vulcanizing machine
CN107590565A (en) A kind of method and device for building building energy consumption forecast model
CN106407550A (en) Soft sensor modeling method in industrial process
CN110739031B (en) Supervised prediction method and device for metallurgical sintering process and storage medium
CN107944349A (en) GIS partial discharge type-scheme recognition methods based on particle swarm support vector machine
CN108667069B (en) Short-term wind power prediction method based on partial least squares regression
CN105182755B (en) A kind of fractional order predictive functional control algorithm of industry heating furnace system
CN102601881B (en) Method for monitoring on-line quality and updating prediction model of rubber hardness
CN106156434A (en) Sliding window time difference Gaussian process regression modeling method based on the low and deep structure of local time
CN107092743A (en) The blast furnace ironmaking process data-driven modeling method of rule-based auxiliary
CN103876736B (en) Complexity spectrum electric signal classification and identification method based on power spectrum division
CN113379164B (en) Load prediction method and system based on deep self-attention network
CN102722103A (en) Method for optimizing multi-objective parameters based on fuzzy analysis and visualization technology
CN107273633A (en) Varying delay method of estimation and flow time lag method of estimation is hydrocracked between multiple operation
CN103745271A (en) Forecasting method for induction thermal deposition calcium-phosphate coating process on basis of neural network
CN106200984A (en) Mental imagery brain-computer interface model modelling approach
Zhang et al. Industrial cyber-physical system driven intelligent prediction model for converter end carbon content in steelmaking plants
CN103885867B (en) Online evaluation method of performance of analog circuit
CN103778466B (en) Converter steel-making temperature modeling forecasting method based on vector error and system thereof
CN104050380A (en) LF furnace final temperature forecasting method based on Adaboost-PLS-ELM
CN102621953B (en) Automatic online quality monitoring and prediction model updating method for rubber hardness
CN106547899B (en) Intermittent process time interval division method based on multi-scale time-varying clustering center change

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20170215