CN109840362A - A kind of integrated instant learning industrial process soft-measuring modeling method based on multiple-objection optimization - Google Patents
A kind of integrated instant learning industrial process soft-measuring modeling method based on multiple-objection optimization Download PDFInfo
- Publication number
- CN109840362A CN109840362A CN201910039438.3A CN201910039438A CN109840362A CN 109840362 A CN109840362 A CN 109840362A CN 201910039438 A CN201910039438 A CN 201910039438A CN 109840362 A CN109840362 A CN 109840362A
- Authority
- CN
- China
- Prior art keywords
- sample
- model
- learning machine
- verifying
- extreme learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000005457 optimization Methods 0.000 title claims abstract description 36
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 32
- 230000008569 process Effects 0.000 claims abstract description 16
- 238000012360 testing method Methods 0.000 claims description 51
- 238000012549 training Methods 0.000 claims description 23
- 239000013256 coordination polymer Substances 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 claims description 4
- 238000009966 trimming Methods 0.000 claims description 3
- 230000004807 localization Effects 0.000 claims description 2
- 239000001273 butane Substances 0.000 description 9
- IJDNQMDRQITEOD-UHFFFAOYSA-N n-butane Chemical compound CCCC IJDNQMDRQITEOD-UHFFFAOYSA-N 0.000 description 9
- OFBQJSOFQDEBGM-UHFFFAOYSA-N n-pentane Natural products CCCCC OFBQJSOFQDEBGM-UHFFFAOYSA-N 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 5
- 238000005259 measurement Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000006477 desulfuration reaction Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 150000001335 aliphatic alkanes Chemical class 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000023556 desulfurization Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000010992 reflux Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of integrated instant learning industrial process soft-measuring modeling method based on multiple-objection optimization, belongs to industrial process soft sensor modeling field.Nonlinear problem of the present invention for industrial data generally existing redundancy and process, the input variable in historical sample data library is optimized with model structure using the method for Evolutionary multiobjective optimization, reject or weak relevant variable uncorrelated to quality variable, database sample quality is improved, simultaneously effective the relationship between balance model complexity and precision of prediction.Furthermore, local extreme learning machine model is constructed from selection part similar with query sample sample in resulting historical sample library is optimized, and integrated the resulting Pareto optimal solution of multiple-objection optimization using selective ensemble strategy, the nonlinear problem of industrial process can be effectively treated.The present invention improves the precision of prediction and computational efficiency of industrial process soft sensor modeling by Optimization Modeling data structure and extreme learning machine model structure.
Description
Technical field
The invention belongs to industrial process soft sensor modeling field more particularly to it is a kind of based on multiple-objection optimization it is integrated immediately
Learn soft-measuring modeling method.
Background technique
In the modern industrial production process, it in order to guarantee that product quality meets increasingly harsh production target, needs to one
A little crucial process variables or quality variable carry out real-time online detection, to realize the real-time control of production process.However, real
The industrial processes on border are wanted sensor performance sometimes in the rugged environments such as high temperature, high pressure, strong corrosive
Ask extremely harsh, and often there is higher cost, be difficult to the problems such as safeguarding in high performance sensor, in addition, the off-line analysis time
Length is to restrict another key factor of production process real-time control.The appearance of hard measurement is that this kind of difficult on-line monitoring for surveying parameter mentions
A kind of effective way is supplied.
The core of soft-measuring technique is to construct auxiliary variable (easily survey variable) and leading variable by certain optiaml ciriterion
Functional relation between (hardly possible surveys variable), and by computer software, realize the On-line Estimation of leading variable.Have benefited from collecting and distributing
The development of the technologies such as control system, database, data-driven soft sensor modeling technology have received widespread attention.In data-driven
It is the partial model of representative due to the thought using " dividing and rule " using instant learning, having accurately to describe in modeling
Local process feature significantly reduces the computation complexity of model, the characteristics such as the non-linear and time variation of process is effectively treated, soft
Measurement modeling field is concerned.
However, instant learning technology tends to rely on the sample in database, the quality of sample is pre- to model in database
Precision is surveyed to play a crucial role.And actual production process is usually since sensor redundancy, sample frequency are inconsistent etc.
Factor, so that input variable in database the defects of there are weak dependences between redundancy, input variable and output variable,
So as to cause the increase of model computation complexity, while also easy the problems such as over-fitting occurs.Therefore, rationally improve historical sample data
The quality in library is vital for the computational efficiency and prediction accuracy that improve soft-sensing model.Further, since model
Often there is positively related relationship between precision of prediction and model complexity, therefore, how to guarantee that model prediction accuracy is certain
In the case where to reduce model complexity as far as possible be another problem for intending to solve of the present invention.Simultaneously, it is contemplated that industrial process is past
Toward complicated process characteristic, such as strong nonlinearity, time variation etc. is presented, it is only capable of obtaining using instant learning soft sensor modeling technology
A series of suboptimum model, these suboptimum models are difficult to fully consider the different characteristics of process, it is intended that by using integrated
The strategy of study merges different partial models, to obtain high performance soft-sensing model.
Summary of the invention
How the technical problem to be solved by the present invention is to improve the matter in historical sample data library in soft sensor modeling technology
The problem of measuring and how reducing model complexity while improving model prediction accuracy.For this purpose, the invention proposes
A kind of integrated instant learning industrial process soft-measuring modeling method based on multiple-objection optimization, comprising the following steps:
(1) by Distributed Control System or the method for offline inspection, industrial process data D is collected, building is used for hard measurement
The database of modeling determines auxiliary variable X relevant to predictive variable y, auxiliary variable by the Analysis on Mechanism to industrial process
X, that is, input variable, X={ x1,x2,...,xM}；
(2) all samples in database are normalized, and are classified as training set Dtrain, verifying collection
DvalidateWith test set Dtest, wherein training set DtrainFor the training of model, verifying collection DvalidateFor the excellent of model parameter
Change, test set DtestAssessment for model performance；
(3) using Evolutionary multiobjective optimization method to training set DtrainIn input variable and extreme learning machine model hide
Node layer number optimizes, reject redundancy or to the weak relevant input variable of output variable, according to optimization obtain S Pareto most
Excellent solution successively updates training set D according to Pareto optimal solutiontrainIn input variable, to obtain the new training sets of S and work
For new modeling sample database；
Multi-objective optimization question is described as follows:
min[f1(x),f2(x)]
f1(x) and f2It (x) is two objective functions to be optimized
f2(x)=Nhidden×M*
f1It (x) is the prediction error of verifying sample, f2It (x) is the complexity of model, NvalCollect D for verifyingvalidateSample
Quantity,Collect D for verifyingvalidateIn i-th of sample predicted value, yval,iCollect D for verifyingvalidateIn i-th sample it is true
Real value, M*For optimization after input variable number,Specific calculating process be, will verifying collection DvalidateIn sample successively take
It is query sample xq,i, according to M*The input variable for updating query sample, according to Euclidean distance similarity from training set DtrainMiddle choosing
P sample building extreme learning machine models similar to query sample before selecting, obtain its prediction output
X is decision variable to be optimized, x={ x1,x2,...,xm...,xM,Nhidden, wherein xmIndicate m-th of sample
Input variable, M indicate input variable number, NhiddenFor extreme learning machine model hidden layer number of nodes；Lb and ub is respectively x's
The inequality constraints item that lower and upper limit constraint, A and b are x；
(4) according to S new modeling sample database update test set DtestInput variable corresponding obtain S test specimens
This collection, the sample concentrated to each test sample successively take and do query sample, according to Euclidean distance similarity from corresponding new
P similar samples are selected to construct local extreme learning machine model in modeling sample database to get S local extreme learning machine is arrived
Model obtains the prediction output of test sample collectionAccording to S new modeling sample databases, successively to test set DtestIn
Input variable be updated, reject uncorrelated or redundant variables, using gained variable as new input variable collection Xtest,new, from
And S test sample collection is obtained, and the sample that each test sample is concentrated successively is taken and does query sample, it is similar according to Euclidean distance
The similar samples of P construct local extreme learning machine models to get to S before degree is selected from corresponding new modeling sample database
A part extreme learning machine model obtains the prediction output of test sample collection Table
Show the S prediction output.
(5) according to S new modeling sample database update verifying collection DvalidateInput variable corresponding obtain S and verify
Sample set successively takes to the sample in each verifying sample set and does query sample, according to Euclidean distance similarity from corresponding new
Modeling sample database in the similar samples of selection P construct local extreme learning machine models and learn to get to the S local limit
Machine model, and S local extreme learning machine model is integrated；Collection is verified according to S new modeling sample database updates
DvalidateInput variable it is corresponding obtain S verifying sample set, each sample verified in sample set is successively taken and makees inquiry sample
This, P similar sample building local poles before being selected from corresponding new modeling sample database according to Euclidean distance similarity
Learning machine model is limited to get S local extreme learning machine model { ELM is arrived1,ELM2,...,ELMS, it is corresponding to obtain S verifying
The prediction of sample exportsThe prediction output of sample will be verifiedAs input,
Its true output constructs PLS model, saves the regression coefficient of PLS model as output.
(6) S obtained in step (5) local extreme learning machine model is trimmed, selects precision of prediction higher
Submodel constructs final integrated model to the test sample collection in step (4).
(a) to the absolute value of the PLS model regression coefficient of verifying sample carry out by size descending arrange to obtain | β1|,|β2
|...,|βS|}；
(b) by calculating the contribution rate CP of local extreme learning machine model, the preferably preceding S of prediction effect is selected*A submodule
Type constructs integrated model, contribution rate CP calculation formula are as follows:
Wherein, | βS| indicate the absolute value of s-th of regression coefficient of PLS model, and meet | β1|≥|β2|
≥…≥|βS|, as CP >=95%, stop model trimming, save submodel index and corresponding regression coefficient；
(c) according to resulting verifying sample submodel index is trimmed, from S local extreme learning machine mould in step (4)
Type selects S*What a part extreme learning machine model progress PLS was integrated arrives integrated model.
Beneficial effects of the present invention
The present invention is by multi-target evolution optimization method to the hiding node layer to input variable and extreme learning machine model
Number optimizes, and according to Euclidean distance similarity, selection constructs extreme learning machine mould with the higher sample of query sample similarity
Type, and integrated learning approach is used, different Pareto are solved into resulting submodel and are integrated, to obtain high performance integrated
Instant learning soft-sensing model.Compared to current other methods, the present invention can not only handle the redundancy in database well
Variable or to the weak relevant input variable of output variable, and while guaranteeing model prediction accuracy reduce model complexity
Degree, is effectively treated the non-linear of industrial process, improves the computational efficiency of modeling and the precision of prediction of model.
Detailed description of the invention
Fig. 1 is the flow chart of the integrated instant learning soft sensor modeling based on multiple-objection optimization in the method for the present invention.
Specific embodiment
The present invention will be further explained below with reference to the attached drawings and specific examples.
Embodiment 1: as shown in Figure 1, a kind of integrated instant learning industrial process soft sensor modeling based on multiple-objection optimization
Method includes the following steps, step 1: collecting industrial process data, structure by Distributed Control System or the method for offline inspection
Build the database for soft sensor modeling.By the Analysis on Mechanism to industrial process, determine that auxiliary relevant to predictive variable becomes
Amount.
Step 2: the sample in database is normalized, and it is classified as training set (Dtrain∈RJ×Q), it tests
Card collection (Dvalidate∈RK×Q) and test set (Dtest∈RT×Q).Wherein training set is used for the training of model, and verifying collection is used for model
The optimization of parameter, test set are used for the assessment of model performance.
Step 3: being carried out using multi-target evolution optimization method (NSGA-II algorithm) to the input variable in training set excellent
Change,
Multi-objective optimization question is described as follows:
min[f1(x),f2(x)]
f1(x) and f2It (x) is two objective functions to be optimized, due to objective function f1(x) and f2(x) conflicting, i.e.,
One solution is best in some target, may be worst, therefore the mesh of multi-objective optimization algorithm in another target
Mark is to find one group of equilibrium solution, makes each sub-goal all being optimal as much as possible.In general, multi-objective optimization question is not
One optimal solution of existence anduniquess, but obtain one group of optimal solution set being made of numerous Pareto optimal solutions.
f2(x)=Nhidden×M*
f1It (x) is the prediction error of verifying sample, f2It (x) is the complexity of model, NvalCollect D for verifyingvalidateSample
Quantity,Collect D for verifyingvalidateIn i-th of sample predicted value, yval,iCollect D for verifyingvalidateIn i-th sample it is true
Real value, M*For optimization after input variable number,Specific calculating process be, will verifying collection DvalidateIn sample successively take
It is query sample xq,i, according to M*The input variable for updating query sample, according to Euclidean distance similarity from training set DtrainMiddle choosing
P sample building extreme learning machine models similar to query sample before selecting, obtain its prediction output
X is decision variable to be optimized, x={ x1,x2,...,xm...,xM,Nhidden, wherein xmIndicate m-th of sample
Input variable, M indicate input variable number, NhiddenFor extreme learning machine model hidden layer number of nodes；Lb and ub is respectively x's
The inequality constraints item that lower and upper limit constraint, A and b are x；
Step 4: according to S new modeling sample database update test set DtestInput variable corresponding obtain S and survey
Sample set is tried, the sample concentrated to each test sample successively takes and does query sample, according to Euclidean distance similarity from corresponding
P similar samples are selected to construct local extreme learning machine model in new modeling sample database to get the S local limit is arrived
Habit machine model obtains the prediction output of test sample collectionAccording to S new modeling sample databases, successively to test set
DtestIn input variable be updated, reject uncorrelated or redundant variables, using gained variable as new input variable collection
Xtest,new, to obtain S test sample collection, the sample that each test sample is concentrated successively is taken and does query sample, according to Europe
P similar samples construct local extreme learning machine mould before family name's Distance conformability degree is selected from corresponding new modeling sample database
Type obtains the prediction output of test sample collection to get to S local extreme learning machine model Indicate the S prediction output.
Step 5: according to S new modeling sample database update verifying collection DvalidateInput variable corresponding obtain S
Sample set is verified, the sample in each verifying sample set is successively taken and does query sample, according to Euclidean distance similarity from correspondence
New modeling sample database in the similar samples of selection P construct local extreme learning machine models to get to the S part limit
Learning machine model, and S local extreme learning machine model is integrated；It is tested according to S new modeling sample database updates
Card collection DvalidateInput variable it is corresponding obtain S verifying sample set, each sample verified in sample set is successively taken and is looked into
Sample is ask, P similar sample building offices before selecting from corresponding new modeling sample database according to Euclidean distance similarity
Portion's extreme learning machine model is to get to S part extreme learning machine model { ELM1,ELM2,...,ELMS, it is corresponding to obtain S
Verify the prediction output of sampleThe prediction output of sample will be verifiedAs
Input, true output construct PLS model as output, save the regression coefficient of PLS model.
Step 6: trimmed to S obtained in step (5) local extreme learning machine model, select precision of prediction compared with
High submodel constructs final integrated model to the test sample collection in step (4).Specifically:
(a) to the absolute value of the PLS model regression coefficient of verifying sample carry out by size descending arrange to obtain | β1|,|β2
|...,|βS|}；
(b) by calculating the contribution rate CP of local extreme learning machine model, the preferably preceding S of prediction effect is selected*A submodule
Type constructs integrated model, contribution rate CP calculation formula are as follows:
Wherein, | βS| indicate the absolute value of s-th of regression coefficient of PLS model, and meet | β1|≥|β2|≥…≥|βS|,
As CP >=95%, stop model trimming, save submodel index and corresponding regression coefficient；
(c) according to resulting verifying sample submodel index is trimmed, from S local extreme learning machine mould in step (4)
Type selects S*What a part extreme learning machine model progress PLS was integrated arrives integrated model.
Euclidean Distance conformability degree in above-mentioned stepsWherein, dnIndicate query sample and similar sample
Between weighted euclidean distance,xnIndicate similar sample, xqIndicate query sample, n ∈
1 ..., P), q ∈ 1 ..., P), σnIt isStandard deviation,It is localization parameter.
Embodiment 2: being further elaborated with below in conjunction with debutanizing tower industrial process, and debutanizing tower is Oil Refinery Industry process
A part of device of middle desulfurization and naphtha separation, target are to minimize the concentration of tower bottom butane.But butane concentration at present
Also it is difficult to realize real-time online detection.On-line prediction is carried out to butane concentration using flexible measurement method, de- fourth can be effectively improved
The desulfuration efficiency of alkane tower.According to Analysis on Mechanism, by x1Tower top temperature；x2Tower top pressure；x3Overhead reflux amount；x4Overhead product stream
Output；x5Layer 6 column plate temperature；x6Tower low temperature 1；x7This 7 monitored parameters of column bottom temperature 2 are used as building soft-sensing model
Auxiliary variable, output variable are butane concentration.The present invention verifies having for mentioned method by building dynamic process soft-sensing model
Effect property constructs dynamic soft sensor model using moving average model(MA model) structure, obtains 2388 groups of sample datas, input variable 49 altogether
A, wherein 1194 samples are as training sample, 597 samples are as verifying sample, and 597 samples are as test sample.
Next combine the detailed process that implementation steps are described in detail:
1. acquiring debutanizing tower industrial process data, data are pre-processed, excluding outlier, missing values.
2. pair 2388 groups of data are normalized.By 1194 samples composing training sample set therein, 597 samples
This composition verifies sample set, and 597 samples constitute test sample collection.
3. the input variable that offline optimization training sample is concentrated.To 49 input variables and extreme learning machine model hidden layer
Number of nodes is encoded, and the decision variable for optimization is obtained, and wherein decision variable number is 50, using NSGA-II algorithm pair
Decision variable optimizes.
4. more new training sample set and the historical sample data library as soft sensor modeling, resulting defeated according to optimizing
Enter the input variable that variable index concentrates verifying sample and test sample to be updated, rejects redundant variables.
5. pair verifying sample, according to the similarity of itself and historical sample in database of Euclidean distance similarity calculation, selection
30 similar samples construct extreme learning machine soft-sensing model, obtain the local prediction value of butane concentration.By these local predictions
Value is used as input variable, verifies the true output of sample as output valve, constructs PLS model, obtain the regression coefficient of model.So
Descending arrangement, the contribution rate CP of continuous superposition calculation submodel, as CP >=95% are carried out by size to model regression coefficient afterwards
Stop calculate, save current submodel index and corresponding regression coefficient.
6. pair test sample, according to the similarity of itself and historical sample in database of Euclidean distance similarity calculation, selection
30 similar samples construct extreme learning machine soft-sensing model, obtain the local prediction value of butane concentration.
7. selecting the corresponding submodule of test sample according to the resulting submodel index of verifying sample and model regression coefficient
Type carries out PLS to it and integrates, to obtain test sample butane concentration prediction value.
8. comparison of the different soft-sensing models to butane concentration prediction precision.The prediction of butane concentration in the case of comparing 2 kinds
Error, i.e., traditional instant learning extreme learning machine soft-sensing model and proposed by the present invention while optimizing historical sample data library
In input variable and extreme learning machine hidden layer number of nodes integrated instant learning soft-measuring modeling method.Distinct methods it is pre-
It is as shown in table 1 to survey error result, wherein prediction smaller its precision of prediction of explanation of error is higher.By table 1 it is found that the present invention proposes
The integrated instant learning industrial process soft-measuring modeling method based on multiple-objection optimization improve soft-sensing model prediction essence
Degree.
Root-mean-square error (RMSE) of 1 distinct methods of table in debutanizing tower
| Method | RMSE |
| Instant learning extreme learning machine soft sensor modeling | 0.0844 |
| Integrated instant learning industrial process soft sensor modeling based on multiple-objection optimization | 0.0500 |
Above-described embodiment is used to illustrate the present invention, rather than limits the invention, in spirit of the invention and
In claims, to any modifications and changes that the present invention makes, protection scope of the present invention is both fallen within.
Claims (6)
1. a kind of integrated instant learning industrial process soft-measuring modeling method based on multiple-objection optimization, which is characterized in that including
Following steps:
(1) industrial process data D is collected, building is used for the database of soft sensor modeling, by the Analysis on Mechanism to industrial process,
Determine auxiliary variable X, auxiliary variable X, that is, input variable, X={ x relevant to predictive variable y1,x2,...,xM}；
(2) all samples in database are normalized, and are classified as training set Dtrain, verifying collection DvalidateWith
Test set Dtest, wherein training set DtrainFor the training of model, verifying collection DvalidateFor the optimization of model parameter, test set
DtestAssessment for model performance；
(3) using Evolutionary multiobjective optimization method to training set DtrainIn input variable and extreme learning machine model hidden layer section
Points optimize, reject redundancy or to the weak relevant input variable of output variable, according to optimization obtain S Pareto it is optimal
Solution, successively updates training set D according to Pareto optimal solutiontrainIn input variable, to obtain the new training sets of S and conduct
New modeling sample database；
(4) according to S new modeling sample database update test set DtestInput variable it is corresponding obtain S test sample collection,
The sample concentrated to each test sample successively takes and does query sample, according to Euclidean distance similarity from corresponding new modeling sample
The similar samples of P are selected to construct local extreme learning machine models to get to S part extreme learning machine model in database,
Obtain the prediction output of test sample collection
(5) according to S new modeling sample database update verifying collection DvalidateInput variable corresponding obtain S verifying sample
Collection, successively takes the sample in each verifying sample set and does query sample, is built according to Euclidean distance similarity from corresponding new
P similar samples are selected to construct local extreme learning machine model in apperance database to get S local extreme learning machine mould is arrived
Type, and S local extreme learning machine model is integrated；
(6) S obtained in step (5) local extreme learning machine model is trimmed, selects the higher submodule of precision of prediction
Type constructs final integrated model to the test sample collection in step (4).
2. the integrated instant learning industrial process soft-measuring modeling method according to claim 1 based on multiple-objection optimization,
It is characterized in that, step (3) detailed process are as follows:
Multi-objective optimization question is described as follows:
min[f1(x),f2(x)]
f1(x) and f2It (x) is two objective functions to be optimized
f2(x)=Nhidden×M*
f1It (x) is the prediction error of verifying sample, f2It (x) is the complexity of model, NvalCollect D for verifyingvalidateSample size,Collect D for verifyingvalidateIn i-th of sample predicted value, yval,iCollect D for verifyingvalidateIn i-th of sample true value,
M*For optimization after input variable number,Specific calculating process be, will verifying collection DvalidateIn sample successively take and look into
Ask sample xq,i, according to M*The input variable for updating query sample, according to Euclidean distance similarity from training set DtrainBefore middle selection
P samples similar to query sample construct extreme learning machine model, obtain its prediction output
X is decision variable to be optimized, x={ x1,x2,...,xm...,xM,Nhidden, wherein xmIndicate m-th of input of sample
Variable, M indicate input variable number, NhiddenFor extreme learning machine model hidden layer number of nodes；Lb and ub is respectively the lower limit of x
It is constrained with the upper limit, the inequality constraints item that A and b are x.
3. the integrated instant learning industrial process soft-measuring modeling method according to claim 1 based on multiple-objection optimization,
It is characterized in that, step (4) detailed process are as follows:
According to S new modeling sample databases, successively to test set DtestIn input variable be updated, reject it is uncorrelated
Or redundant variables, using gained variable as new input variable collection Xtest,new, so that S test sample collection is obtained, to each survey
The sample of this concentration of sample, which successively takes, does query sample, according to Euclidean distance similarity from corresponding new modeling sample database
P similar samples construct local extreme learning machine model to get to S local extreme learning machine model before middle selection, are surveyed
Try the prediction output of sample set Indicate the S prediction output.
4. the integrated instant learning industrial process soft-measuring modeling method according to claim 1 based on multiple-objection optimization,
It is characterized in that, the detailed process of the step (5) are as follows:
According to S new modeling sample database update verifying collection DvalidateInput variable it is corresponding obtain S verifying sample set,
Sample in each verifying sample set is successively taken and does query sample, according to Euclidean distance similarity from corresponding new modeling sample
P similar samples construct local extreme learning machine models to get S local extreme learning machine model is arrived before selection in database
{ELM1,ELM2,...,ELMS, the corresponding prediction output for obtaining S verifying sampleSample will be verified
This prediction outputAs input, true output constructs PLS model as output, saves PLS
The regression coefficient of model.
5. the integrated instant learning industrial process soft-measuring modeling method according to claim 1 based on multiple-objection optimization,
It is characterized in that, the detailed process of the step (6) are as follows:
(a) to the absolute value of the PLS model regression coefficient of verifying sample carry out by size descending arrange to obtain | β1|,|β2|...,
|βS|}；
(b) by calculating the contribution rate CP of local extreme learning machine model, the preferably preceding S* submodel structure of prediction effect is selected
Build integrated model, contribution rate CP calculation formula are as follows:
Wherein, | βS| indicate the absolute value of s-th of regression coefficient of PLS model, and meet | β1|≥|β2|≥…≥|βS|, when CP >=
When 95%, stop model trimming, save submodel index and corresponding regression coefficient；
(c) according to resulting verifying sample submodel index is trimmed, from S local extreme learning machine model choosing in step (4)
Select S* local extreme learning machine model carry out PLS it is integrated to integrated model.
6. the integrated soft survey of instant learning industrial process according to any one of claim 2-4 based on multiple-objection optimization
Measure modeling method, which is characterized in that the Euclidean distance similarityWherein, dnIndicate query sample with
Weighted euclidean distance between similar sample,xnIndicate similar sample, xqIndicate inquiry sample
This, n ∈ 1 ..., P), q ∈ 1 ..., P) and, σnIt isStandard deviation,It is localization parameter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910039438.3A CN109840362B (en) | 2019-01-16 | 2019-01-16 | Multi-objective optimization-based integrated just-in-time learning industrial process soft measurement modeling method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910039438.3A CN109840362B (en) | 2019-01-16 | 2019-01-16 | Multi-objective optimization-based integrated just-in-time learning industrial process soft measurement modeling method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109840362A true CN109840362A (en) | 2019-06-04 |
| CN109840362B CN109840362B (en) | 2022-06-14 |
Family
ID=66883886
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910039438.3A Active CN109840362B (en) | 2019-01-16 | 2019-01-16 | Multi-objective optimization-based integrated just-in-time learning industrial process soft measurement modeling method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109840362B (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110751987A (en) * | 2019-10-17 | 2020-02-04 | 北京化工大学 | Method for establishing soft measurement model of monitoring variable for ethylene preparation through naphtha cracking |
| CN111768000A (en) * | 2020-06-23 | 2020-10-13 | 中南大学 | Industrial process data modeling method for online adaptive fine-tuning deep learning |
| CN111914466A (en) * | 2019-09-07 | 2020-11-10 | 宁波大学 | Chemical process monitoring method based on related variable distributed modeling |
| CN111914492A (en) * | 2020-04-28 | 2020-11-10 | 昆明理工大学 | Evolution optimization-based semi-supervised learning industrial process soft measurement modeling method |
| 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 |
| CN112446520A (en) * | 2019-08-30 | 2021-03-05 | 京东数字科技控股有限公司 | Method and device for predicting product scheduling time |
| CN112596391A (en) * | 2020-12-24 | 2021-04-02 | 中国兵器装备集团自动化研究所 | Deep neural network large time lag system dynamic modeling method based on data driving |
| CN112989635A (en) * | 2021-04-22 | 2021-06-18 | 昆明理工大学 | Integrated learning soft measurement modeling method based on self-encoder diversity generation mechanism |
| CN113377072A (en) * | 2021-06-24 | 2021-09-10 | 华东交通大学 | Rare earth extraction process drug dosage optimization setting method based on instant learning |
| CN114117919A (en) * | 2021-11-29 | 2022-03-01 | 中国石油大学(华东) | Instant learning soft measurement modeling method based on sample collaborative representation |
| CN114384015A (en) * | 2022-01-12 | 2022-04-22 | 中国环境科学研究院 | Water environment monitoring method based on multi-source remote sensing and machine learning |
| CN114626304A (en) * | 2022-03-21 | 2022-06-14 | 齐鲁工业大学 | Soft measurement modeling method for online prediction of copper grade in ore pulp |
| CN115094193A (en) * | 2022-06-27 | 2022-09-23 | 中冶华天南京工程技术有限公司 | Intelligent molten iron pretreatment desulfurization system based on data mining |
| CN116738866A (en) * | 2023-08-11 | 2023-09-12 | 中国石油大学(华东) | Instant learning soft measurement modeling method based on time sequence feature extraction |
| 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 (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106682312A (en) * | 2016-12-28 | 2017-05-17 | 浙江大学 | Industrial process soft-measurement modeling method of local weighing extreme learning machine model |
| CN106951995A (en) * | 2017-03-22 | 2017-07-14 | 华东交通大学 | A kind of EHV transmission electric field extreme learning machine predicts multiple-objection optimization screen method |
| CN107274387A (en) * | 2017-05-19 | 2017-10-20 | 西安电子科技大学 | The end member extraction method of target in hyperspectral remotely sensed image based on Evolutionary multiobjective optimization |
| CN108133286A (en) * | 2017-12-21 | 2018-06-08 | 河海大学 | A kind of underground water multi-objective calculation method based on surface subsidence alternative model |
| US10061300B1 (en) * | 2017-09-29 | 2018-08-28 | Xometry, Inc. | Methods and apparatus for machine learning predictions and multi-objective optimization of manufacturing processes |
| CN108549234A (en) * | 2018-05-11 | 2018-09-18 | 江南大学 | A kind of multiobjective optimization control method based on dynamic variate |
-
2019
- 2019-01-16 CN CN201910039438.3A patent/CN109840362B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106682312A (en) * | 2016-12-28 | 2017-05-17 | 浙江大学 | Industrial process soft-measurement modeling method of local weighing extreme learning machine model |
| CN106951995A (en) * | 2017-03-22 | 2017-07-14 | 华东交通大学 | A kind of EHV transmission electric field extreme learning machine predicts multiple-objection optimization screen method |
| CN107274387A (en) * | 2017-05-19 | 2017-10-20 | 西安电子科技大学 | The end member extraction method of target in hyperspectral remotely sensed image based on Evolutionary multiobjective optimization |
| US10061300B1 (en) * | 2017-09-29 | 2018-08-28 | Xometry, Inc. | Methods and apparatus for machine learning predictions and multi-objective optimization of manufacturing processes |
| CN108133286A (en) * | 2017-12-21 | 2018-06-08 | 河海大学 | A kind of underground water multi-objective calculation method based on surface subsidence alternative model |
| CN108549234A (en) * | 2018-05-11 | 2018-09-18 | 江南大学 | A kind of multiobjective optimization control method based on dynamic variate |
Non-Patent Citations (4)
| Title |
|---|
| HUAIPING JIN 等: ""Ensemble just-in-time learning framework through evolutionary multi-objective optimization for soft sensor development of nonlinear industrial processes"", 《CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS》, vol. 184, 15 January 2019 (2019-01-15), pages 153 - 166, XP055673435, DOI: 10.1016/j.chemolab.2018.12.002 * |
| MINGYANG ZHANG 等: ""Hyperspectral band selection based on multi-objective optimization with high information and low redundancy"", 《APPLIED SOFT COMPUTING》, vol. 70, 30 September 2018 (2018-09-30), pages 604 - 621 * |
| 王礼云 等: ""极限学习机多目标模型选择研究"", 《计算机仿真》, vol. 31, no. 8, 31 August 2014 (2014-08-31), pages 387 - 391 * |
| 赵佳琦: ""进化多目标优化学习算法及其应用"", 《中国优秀博硕士学位论文全文数据库(博士)信息科技辑》, 15 January 2019 (2019-01-15), pages 140 - 29 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112446520A (en) * | 2019-08-30 | 2021-03-05 | 京东数字科技控股有限公司 | Method and device for predicting product scheduling time |
| CN111914466A (en) * | 2019-09-07 | 2020-11-10 | 宁波大学 | Chemical process monitoring method based on related variable distributed modeling |
| CN111914466B (en) * | 2019-09-07 | 2023-10-24 | 宁波大学 | Chemical process monitoring method based on related variable distributed modeling |
| CN110751987A (en) * | 2019-10-17 | 2020-02-04 | 北京化工大学 | Method for establishing soft measurement model of monitoring variable for ethylene preparation through naphtha cracking |
| CN111914492A (en) * | 2020-04-28 | 2020-11-10 | 昆明理工大学 | Evolution optimization-based semi-supervised learning industrial process soft measurement modeling method |
| CN111768000A (en) * | 2020-06-23 | 2020-10-13 | 中南大学 | Industrial process data modeling method for online adaptive fine-tuning deep learning |
| 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 |
| CN112596391B (en) * | 2020-12-24 | 2022-08-02 | 中国兵器装备集团自动化研究所 | Deep neural network large time lag system dynamic modeling method based on data driving |
| CN112596391A (en) * | 2020-12-24 | 2021-04-02 | 中国兵器装备集团自动化研究所 | Deep neural network large time lag system dynamic modeling method based on data driving |
| CN112989635A (en) * | 2021-04-22 | 2021-06-18 | 昆明理工大学 | Integrated learning soft measurement modeling method based on self-encoder diversity generation mechanism |
| CN113377072B (en) * | 2021-06-24 | 2023-02-07 | 华东交通大学 | Rare earth extraction process dosage optimization setting method based on instant learning |
| CN113377072A (en) * | 2021-06-24 | 2021-09-10 | 华东交通大学 | Rare earth extraction process drug dosage optimization setting method based on instant learning |
| CN114117919B (en) * | 2021-11-29 | 2022-08-23 | 中国石油大学(华东) | Instant learning soft measurement modeling method based on sample collaborative representation |
| CN114117919A (en) * | 2021-11-29 | 2022-03-01 | 中国石油大学(华东) | Instant learning soft measurement modeling method based on sample collaborative representation |
| CN114384015A (en) * | 2022-01-12 | 2022-04-22 | 中国环境科学研究院 | Water environment monitoring method based on multi-source remote sensing and machine learning |
| CN114626304A (en) * | 2022-03-21 | 2022-06-14 | 齐鲁工业大学 | Soft measurement modeling method for online prediction of copper grade in ore pulp |
| CN114626304B (en) * | 2022-03-21 | 2024-02-27 | 齐鲁工业大学 | Online prediction soft measurement modeling method for ore pulp copper grade |
| CN115094193A (en) * | 2022-06-27 | 2022-09-23 | 中冶华天南京工程技术有限公司 | Intelligent molten iron pretreatment desulfurization system based on data mining |
| CN116738866A (en) * | 2023-08-11 | 2023-09-12 | 中国石油大学(华东) | Instant learning soft measurement modeling method based on time sequence feature extraction |
| CN116738866B (en) * | 2023-08-11 | 2023-10-27 | 中国石油大学(华东) | Instant learning soft measurement modeling method based on time sequence feature extraction |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109840362B (en) | 2022-06-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109840362A (en) | A kind of integrated instant learning industrial process soft-measuring modeling method based on multiple-objection optimization | |
| Yuan et al. | Hierarchical quality-relevant feature representation for soft sensor modeling: A novel deep learning strategy | |
| CN103927412B (en) | Instant learning debutanizing tower soft-measuring modeling method based on gauss hybrid models | |
| Guo et al. | Research on CBR system based on data mining | |
| CN107451102A (en) | A kind of semi-supervised Gaussian process for improving self-training algorithm returns soft-measuring modeling method | |
| JPH1074188A (en) | Data learning device and plant controller | |
| Sallehuddin et al. | Application of grey relational analysis for multivariate time series | |
| Pani et al. | A survey of data treatment techniques for soft sensor design | |
| CN116859843B (en) | Production equipment fault monitoring method and system based on industrial big data | |
| CN107644297B (en) | Energy-saving calculation and verification method for motor system | |
| Chen et al. | Ridge estimation for uncertain autoregressive model with imprecise observations | |
| CN105260532A (en) | Sequence approximation optimization based thin sheet tension VBHF (Variable Blank Holder Force) uncertainty design method | |
| CN106530082A (en) | Stock predication method and stock predication system based on multi-machine learning | |
| CN113822499B (en) | Train spare part loss prediction method based on model fusion | |
| Buragohain | Adaptive network based fuzzy inference system (ANFIS) as a tool for system identification with special emphasis on training data minimization | |
| CN114548494B (en) | Visual cost data prediction intelligent analysis system | |
| CN113741362A (en) | Method, system, medium, and computing device for optimizing operation of liquefied natural gas receiving station | |
| CN110490496A (en) | A method of based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process | |
| Parfenova et al. | Fuzzy modelling for tasks of management of the agricultural-industrial complex | |
| Taffese | Case-based reasoning and neural networks for real estate valuation. | |
| CN110378035A (en) | Hydrocracking soft measurement modeling method based on deep learning | |
| CN109493921A (en) | A kind of atmospheric distillation process modeling approach based on multi-agent system model | |
| CN112711912A (en) | Air quality monitoring and alarming method, system, device and medium based on cloud computing and machine learning algorithm | |
| Alshboul et al. | Empirical exploration of predictive maintenance in concrete manufacturing: Harnessing machine learning for enhanced equipment reliability in construction project management | |
| Akbari et al. | Similarity-based error prediction approach for real-time inflow forecasting |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |