CN109445398A - Propylene polymerization production process optimal online forecasting system based on weighted array study - Google Patents
Propylene polymerization production process optimal online forecasting system based on weighted array study Download PDFInfo
- Publication number
- CN109445398A CN109445398A CN201811625115.4A CN201811625115A CN109445398A CN 109445398 A CN109445398 A CN 109445398A CN 201811625115 A CN201811625115 A CN 201811625115A CN 109445398 A CN109445398 A CN 109445398A
- Authority
- CN
- China
- Prior art keywords
- module
- production process
- optimal
- propylene polymerization
- online forecasting
- 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
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
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS], computer integrated manufacturing [CIM]
- G05B19/41875—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS], computer integrated manufacturing [CIM] characterised by quality surveillance of production
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- 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
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/32—Operator till task planning
- G05B2219/32015—Optimize, process management, optimize production line
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Abstract
The invention discloses a kind of propylene polymerization production process optimal online forecasting systems based on weighted array study, for carrying out quality testing to polypropylene production product, the optimal on-line predictive model based on weighted array study includes phase space reconfiguration, PCA principal component analysis module, support vector machines module, model modification module, cuckoo searching algorithm optimization module, Adaboost module.The present invention carries out online forecasting to polypropylene production process important quality index melt index, overcomes traditional chemical instruments time of measuring to lag big, the low disadvantage of measurement accuracy, realizes that on-line measurement, optimization Generalization Ability is strong, noise resistance interference performance is strong, precision is high.
Description
Technical field
The present invention relates to polymerization process quality testing field, data-drivens to model field, more particularly to a kind of based on weighting
The propylene polymerization production process optimal online forecasting system of ensemble learning.
Background technique
Polypropylene is a kind of thermoplastic resin as prepared by propylene polymerization, the most important downstream product of propylene, the world third
The 50% of alkene, the 65% of China's propylene are all for polypropylene processed, are one of five big general-purpose plastics, close with our daily life
Cut phase is closed.Polypropylene is that fastest-rising general thermoplastic resin, total amount are only only second to polyethylene and polyvinyl chloride in the world.For
Make China's polypropylene product that there is the market competitiveness, exploitation rigidity, flows the good crushing-resistant copolymerization product of sexual balance, is random toughness
Copolymerized product, BOPP and CPP film material, fiber, nonwoven cloth, and develop polypropylene in the application of automobile and field of household appliances, all
It is research topic important from now on.
Melt index is that polypropylene product determines one of important quality index of product grade, it determines the difference of product
Purposes, the measurement to melt index are an important links of control of product quality in polypropylene production, to production and scientific research, all
There are very important effect and directive significance.
However, the on-line analysis measurement of melt index is difficult to accomplish at present, it is on the one hand online melt index analysis instrument
Lack, be on the other hand existing in-line analyzer measured often blocking it is inaccurate even can not be caused by normal use
Use on difficulty.Therefore, at present in industrial production MI measurement, mainly obtained by manual sampling, offline assay
, and general every 2-4 hours can only analyze once, and time lag is big, brings to the quality control that propylene polymerization produces tired
Difficulty becomes a bottleneck problem urgently to be solved in production.The online forecasting system and method for polypropylene melt index is studied, from
And become a forward position and the hot spot of academia and industry.
Summary of the invention
In order to overcome not high, vulnerable to human factor the influence of measurement accuracy of existing propylene polymerization production process at present
Deficiency, the purpose of the present invention is to provide a kind of on-line measurement, forecast speed is fast, model automatically updates, strong antijamming capability, essence
Spend the high optimal online forecasting system of propylene polymerization production process melt index based on weighted array study.
The purpose of the present invention is what is be achieved through the following technical solutions: a kind of propylene polymerization based on weighted array study is raw
Process optimum online forecasting system optimal online forecasting is produced, for carrying out quality testing to polypropylene production product, feature exists
In: including phase space reconfiguration module PCA principal component analysis module, support vector machines module, cuckoo search algorithm module,
Adaboost module.Wherein:
(1) phase space reconfiguration module: inputting 9 performance variables for industrial propylene polymerization processes, and respectively first strand third
Alkene feed flow rates, third stock propylene feed flow rate, major catalyst flow rate, cocatalyst flow rate, are stirred second burst of propylene feed flow rate
Mix temperature in the kettle, pressure in kettle, hydrogen volume concentration in liquid level and kettle in kettle.For chaos time sequence, chaos invariant
Calculating, the foundation of chaotic model and prediction are carried out in phase space, which is used for the mould that will be inputted from DCS database
Type input variable is pre-processed, and passes through difference wherein { x (i) } is the measured value of melt index for time series { x (i) }
Delay time T come construct d dimension phase space vector X (i)=(x (i) ..., x (i+ (d-1) τ)), 1≤i≤n- (d-1) τ, delay
Time is obtained by interactive information method, and Embedded dimensions are obtained by falseness closest to a method；
(2) it PCA principal component analysis module: is used for by input variable pre -whitening processing and variable decorrelation, by input
Variable applies a linear transformation and realizes that is, principal component is obtained by C=MU, and wherein M is input variable, and C is principal component scores square
Battle array, U is loading matrix.It, can be by M=CU if initial data is reconstructedTIt calculates, wherein the transposition of subscript T representing matrix.When
When the principal component number of selection is less than the variable number of input variable, M=CUT+ E, wherein E is residual matrix；
(3) support vector machines module: for using support vector machines, completing to be input to output based on Bayesian frame
Mapping, modeling.The training of support vector machines is carried out under Bayesian frame, assigns weight vectors zero by introducing hyper parameter
The Gaussian prior distribution of mean value ensures the sparsity of model, hyper parameter can using the method for maximization marginal likelihood function come
Estimation.The purpose of entire model is to design a system according to sample set and priori knowledge, so that system is predicted new data defeated
Out；
(4) it cuckoo searching algorithm: is optimized for the hyper parameter to support vector machines, comprising:
(4.1) objective function number f (Z), Z=(z1,…,zd)T, function is initialized, and generates n bird at random
The initial position Z of nesti, population scale n, problem dimension d, maximum detection probability P and maximum number of iterations is arranged in i=1,2 ..., n
MaxGen, current iteration number Gen, minimal error ∈；
(4.2) target function value for asking each bird's nest position obtains the optimal function value of current location；
(4.3) previous generation optimal function value is recorded, is utilizedI=1,2 ... n, to other bird's nests
Position and state be updated, whereinFor i-th of bird's nest in t for the position of bird's nest, initializationIt .* is point pair
Point multiplication, α are step size controlling amount, and for controlling the search range of step-length, value Normal Distribution, L (λ) is that Levy is searched at random
Rope path, arbitrary width meet Levy distribution L (s, λ)~s-λ, 1 < λ≤3, s are the arbitrary widths that Levy flies；
(4.4) target function value of existing position is sought, and is compared with the optimal function value of previous generation record, if currently
Target function value preferably then changes current optimal value；
(4.5) it after carrying out location updating, is compared using random number r ∈ [0,1] and maximum detection probability P, if r > P table
Show that the bird's nest is abandoned, it is rightChanged at random, it is on the contrary then think success, without changing at random, finally retain best
One group of bird's nest position
(4.6) if Gen < MaxGen or not up to minimal error requirement, Gen=Gen+1, return step 4.2, otherwise to
It is lower to execute 4.7；
(4.7) global optimum's target function value is exported as a result, terminating current algorithm and returning.
(5) it Adaboost module: for the obtained weak learner of support vector machines to be weighted ensemble learning, obtains
One anti-interference ability is stronger, the higher strong learner of precision, comprising:
(5.1) algorithm initialization is carried out, setting loss function is root-mean-square error MSE, setting initialization weight vector For D1I-th of component, wherein n be training set element number, setting most
Big the number of iterations MaxN, current iteration times N；
(5.2) to weight vector DNTraining dataset learnt by support vector machines, obtain weak learner FN
(x)；
(5.3) training set { (x is found outi,yi) on worst error EN=max | yi-FN(xi) |, i=1 ..., n, wherein
xi、yi, for training set input sample, export sample data, FNIt (x) is the 6.2 weak learners of gained；Find out the phase of each sample
To errorWhereinIt is square of worst error；Seek regression error ratewNiFor
Weight coefficient seeks weak learner coefficient with regression error rateAnd according to the weight system of regression error rate more new samples
Number Wherein ZNIt is normalization factor, meets
(5.4) if N < MaxN, that is, maximum number of iterations, N=N+1 has not yet been reached, otherwise return step 5.2 executes downwards
5.5；(5.5) it exportsWherein f (x) is αNFN(x), N=1, the median of 2 ... MaxN, knot
Beam current algorithm simultaneously returns.
(6) system update module, the propylene polymerization production process optimal online forecasting system based on integrated study is also
The model mismatch of complicated polymerization process is solved the problems, such as the online updating of detection system including system update module, periodically will
Offline analysis data is input to model training concentration, more new detection system.
Technical concept of the invention are as follows: the important quality index melt index of propylene polymerization production process is carried out online most
Excellent forecast, for overcome existing polypropylene melt index measuring instrumentss measurement accuracy it is not high, vulnerable to human factor influence not
Foot introduces weighted array study module and carries out Automatic Optimal to neural network parameter and structure, does not need artificial experience or multiple
Test is to adjust neural network, to obtain the optimal online forecasting system for the melt index forecast function of having optimal.
Beneficial effects of the present invention are mainly manifested in: the optimal online forecasting system pair based on weighted array study
The important quality index melt index of propylene polymerization production process carries out online Optimal predictor, and existing polypropylene fusion is overcome to refer to
The deficiency of number not high, vulnerable to human factor the influence of measuring instrumentss measurement accuracy introduces Adaboost module to by cuckoo
The supporting vector machine model of bird searching algorithm optimization hyper parameter is weighted ensemble learning, to obtain with optimal melting
The optimal online forecasting system of exponent prediction function has on-line measurement, forecast speed fast, strong antijamming capability, with high accuracy
Feature.
Detailed description of the invention
Fig. 1 is the basic knot of the propylene polymerization production process optimal online forecasting system and method based on weighted array study
Structure schematic diagram；
Fig. 2 is the optimal online forecasting system structure diagram based on weighted array study；
Specific embodiment
The present invention is described further with reference to the accompanying drawing.
Embodiment
Referring to Fig.1, a kind of propylene polymerization production process optimal online forecasting system based on weighted array study, including third
Alkene polymerization production process 1, the control station 3 for measuring performance variable, is deposited the field intelligent instrument 2 for measuring easy survey variable
The DCS database 4, the optimal online forecasting system 5 based on weighted array study and melt index forecast value for putting data are shown
Instrument 6, the field intelligent instrument 2, control station 3 are connect with propylene polymerization production process 1, the field intelligent instrument 2, control station
3 connect with DCS database 4, and DCS database 4 described in gunz is defeated with the optimal online forecasting system 5 that is learnt based on weighted array
Enter end connection, the output end and melt index forecast value of the optimal online forecasting system 5 based on weighted array study are shown
Instrument 6 connects.It is analyzed according to reaction mechanism and flow process, it is contemplated that melt index is had an impact in polypropylene production process
Various factors, take in actual production process common nine performance variables and easily survey variable as modeling variable, be respectively as follows: three
Stock propylene feed flow rates, major catalyst flow rate, cocatalyst flow rate, temperature in the kettle, pressure, liquid level, hydrogen volume concentration in kettle.
Table 1 is listed as 9 modeling variables needed for multi-scale self-adaptive intelligently optimal polypropylene quality of production detection system, respectively
For hydrogen volume concentration (Xv), 3 bursts of propylene feed flow rates (in liquid level (L) in pressure (p) in temperature in the kettle (T), kettle, kettle, kettle
One propylene feed flow rates f1, second gang of propylene feed flow rates f2, third stock propylene feed flow rates f3), 2 strands of catalyst charge streams
Rate (major catalyst flow rate f4, cocatalyst flow rate f5).Polymerization reaction in reaction kettle is participated in after reaction mass mixes repeatedly
Reaction, therefore mode input variable is related to the process variable of material using the average value at preceding several moment.Data are adopted in this example
With the average value of previous hour.The offline laboratory values of melt index are as based on improvement gravitation search algorithm optimization Method Using Relevance Vector Machine
The output variable of optimal soft measurement model 5.It is obtained by manual sampling, offline assay, the acquisition of analysis in every 4 hours is primary.
1 multi-scale self-adaptive of table intelligently models variable needed for optimal polypropylene quality of production detection system
| Variable symbol | Variable meaning | Variable symbol | Variable meaning |
| T | Temperature in the kettle | f1 | First burst of propylene feed flow rates |
| p | Pressure in kettle | f2 | Second burst of propylene feed flow rates |
| L | Liquid level in kettle | f3 | Third stock propylene feed flow rates |
| Xv | Hydrogen volume concentration in kettle | f4 | Major catalyst flow rate |
| f5 | Cocatalyst flow rate |
Referring to described in Fig. 2 based on weighted array study optimal online forecasting system include:
(1) phase space reconfiguration module 7: inputting 9 performance variables for industrial propylene polymerization processes, and respectively first strand third
Alkene feed flow rates, third stock propylene feed flow rate, major catalyst flow rate, cocatalyst flow rate, are stirred second burst of propylene feed flow rate
Mix temperature in the kettle, pressure in kettle, hydrogen volume concentration in liquid level and kettle in kettle.For chaos time sequence, chaos invariant
Calculating, the foundation of chaotic model and prediction are carried out in phase space, which is used for the mould that will be inputted from DCS database
Type input variable is pre-processed, and passes through difference wherein { x (i) } is the measured value of melt index for time series { x (i) }
Delay time T come construct d dimension phase space vector X (i)=(x (i) ..., x (i+ (d-1) τ)), 1≤i≤n- (d-1) τ, delay
Time is obtained by interactive information method, and Embedded dimensions are obtained by falseness closest to a method；
(2) PCA principal component analysis module 8 is used for by input variable pre -whitening processing and variable decorrelation, by input
Variable applies a linear transformation and realizes that is, principal component is obtained by C=MU, and wherein M is input variable, and C is principal component scores square
Battle array, U is loading matrix.It, can be by M=CU if initial data is reconstructedTIt calculates, wherein the transposition of subscript T representing matrix.When
When the principal component number of selection is less than the variable number of input variable, M=CUT+ E, wherein E is residual matrix；
(3) support vector machines module 10, for using support vector machines, completing to be input to output based on Bayesian frame
Mapping, modeling.The training of support vector machines is carried out under Bayesian frame, assigns weight vectors by introducing hyper parameter
The Gaussian prior of zero-mean is distributed the sparsity to ensure model, and hyper parameter can be using the method for maximizing marginal likelihood function
To estimate.The purpose of entire model is to design a system according to sample set and priori knowledge, and system is enable to predict new data
Output；
(4) cuckoo search algorithm module 11 is optimized for the hyper parameter to support vector machines, comprising:
(4.1) objective function number f (Z), Z=(z1..., zd)T, function is initialized, and generates n bird at random
The initial position Z of nesti, population scale n, problem dimension d, maximum detection probability P and maximum number of iterations is arranged in i=1,2 ..., n
MaxGen, current iteration number Gen, minimal error ∈；
(4.2) target function value for asking each bird's nest position obtains the optimal function value of current location；
(4.3) previous generation optimal function value is recorded, is utilizedI=1,2 ... n, to other bird's nests
Position and state be updated, whereinFor i-th of bird's nest in t for the position of bird's nest, initializationIt .* is point pair
Point multiplication, α are step size controlling amount, and for controlling the search range of step-length, value Normal Distribution, L (λ) is that Levy is searched at random
Rope path, arbitrary width meet Levy distribution L (s, λ)~s-λ, 1 < λ≤3, s are the arbitrary widths that Levy flies；
(4.4) target function value of existing position is sought, and is compared with the optimal function value of previous generation record, if currently
Target function value preferably then changes current optimal value；
(4.5) it after carrying out location updating, is compared using random number r ∈ [0,1] and maximum detection probability P, if r > P table
Show that the bird's nest is abandoned, it is rightChanged at random, it is on the contrary then think success, without changing at random, finally retain best
One group of bird's nest position
(4.6) if Gen < MaxGen or not up to minimal error requirement, Gen=Gen+1, return step 4.2, otherwise to
It is lower to execute 4.7；
(4.7) global optimum's target function value is exported as a result, terminating current algorithm and returning.
(5) adaboost module 12 is obtained for the obtained weak learner of support vector machines to be weighted ensemble learning
It is stronger to obtain an anti-interference ability, the higher strong learner of precision, comprising:
(5.1) algorithm initialization is carried out, setting loss function is root-mean-square error MSE, setting initialization weight vector For D1I-th of component, wherein n be training set element number, setting most
Big the number of iterations MaxN, current iteration times N；
(5.2) to weight vector DNTraining dataset learnt by support vector machines, obtain weak learner FN
()；
(5.3) training set { (x is found outi, yi) on worst error EN=max | yi-FN(xi) |, i=1 ..., n, wherein
xi、yi, for training set input sample (i.e. nine modeling variable), the data of output sample (melt index), FNIt (x) is 6.2 gained
Weak learner；Find out the relative error of each sampleWhereinIt is square of worst error；Recurrence is asked to miss
RatewNiWeak learner coefficient is asked with regression error rate for weight coefficientAnd according to recurrence
The weight coefficient of error rate more new samplesWherein ZNIt is normalization factor, meets(5.4) if N < MaxN, that is, maximum number of iterations is had not yet been reached, N=N+1, return step 5.2 is no
Then, 5.5 are executed downwards；(5.5) it exportsWherein f (x) is αNFN(x), N=1,2 ... MaxN
Median, terminate current algorithm simultaneously return.
(6) system update module, the propylene polymerization production process optimal online forecasting system based on integrated study is also
The model mismatch of complicated polymerization process is solved the problems, such as the online updating of detection system including system update module, periodically will
Offline analysis data is input to model training concentration, more new detection system.
Above-described embodiment is used to illustrate the present invention, rather than limits the invention, in spirit of the invention and
In scope of protection of the claims, to any modifications and changes that the present invention makes, protection scope of the present invention is both fallen within.
Claims (7)
1. a kind of propylene polymerization production process optimal online forecasting system optimal online forecasting system based on weighted array study,
For carrying out quality testing to polypropylene production product, it is characterised in that: including phase space reconfiguration module PCA principal component analysis mould
Block, support vector machines module, cuckoo search algorithm module, Adaboost module.
2. the propylene polymerization production process optimal online forecasting system according to claim 1 based on weighted array study,
Be characterized in that: the input of the phase space reconfiguration module is 9 performance variables of industrial propylene polymerization processes, respectively first
Stock propylene feed flow rate, second burst of propylene feed flow rate, third stock propylene feed flow rate, major catalyst flow rate, cocatalyst stream
Rate, stirring temperature in the kettle, pressure in kettle, hydrogen volume concentration in liquid level and kettle in kettle.For chaos time sequence, chaos is not
The calculating of variable, the foundation and prediction of chaotic model are carried out in phase space, and the module from DCS database for that will input
Mode input variable pre-processed, time series { x (i) } is passed through wherein { x (i) } is the measured value of melt index
Different delay time Ts ties up phase space vector X (i)=(x (i) ..., x (i+ (d-1) τ)), 1≤i≤n- (d-1) to construct d
τ, delay time are obtained by interactive information method, and Embedded dimensions are obtained by falseness closest to a method；
3. the propylene polymerization production process optimal online forecasting system according to claim 1 based on weighted array study,
Be characterized in that: the PCA principal component analysis module is used for by input variable pre -whitening processing and variable decorrelation, by right
Input variable applies a linear transformation and realizes that is, principal component is obtained by C=MU, and wherein M is input variable, and C obtains for principal component
Sub-matrix, U are loading matrix.It, can be by M=CU if initial data is reconstructedTIt calculates, wherein turn of subscript T representing matrix
It sets.When the principal component number of selection is less than the variable number of input variable, M=CUT+ E, wherein E is residual matrix；
4. the propylene polymerization production process optimal online forecasting system according to claim 1 based on weighted array study,
It is characterized in that: the support vector machines module, it is defeated for completing to be input to using support vector machines, based on Bayesian frame
Mapping, modeling out.The training of support vector machines is carried out under Bayesian frame, by introduce hyper parameter assign weight to
The Gaussian prior of amount zero-mean is distributed the sparsity to ensure model, and hyper parameter can be using the side for maximizing marginal likelihood function
Method is estimated.The purpose of entire model is to design a system according to sample set and priori knowledge, keeps system pre- to new data
Survey output；
5. the propylene polymerization production process optimal online forecasting system according to claim 1 based on weighted array study,
Be characterized in that: the cuckoo searching algorithm is optimized for the hyper parameter to support vector machines, comprising:
(5.1) objective function number f (Z), Z=(z1..., zd)T, function is initialized, and generates n bird's nest at random
Initial position Zi, population scale n, problem dimension d, maximum detection probability P and maximum number of iterations is arranged in i=1,2 ..., n
MaxGen, current iteration number Gen, minimal error ∈；
(5.2) target function value for asking each bird's nest position obtains the optimal function value of current location；
(5.3) previous generation optimal function value is recorded, is utilizedTo the position of other bird's nests
It sets and is updated with state, whereinFor i-th of bird's nest in t for the position of bird's nest, initialization.* multiply to be point-to-point
Method, α are step size controlling amount, and for controlling the search range of step-length, value Normal Distribution, L (λ) is Levy random search road
Diameter, arbitrary width meet Levy distribution L (s, λ)~s-λ, 1 < λ≤3, s is the arbitrary width that Levy flies；
(5.4) target function value of existing position is sought, and is compared with the optimal function value of previous generation record, if current goal
Functional value preferably then changes current optimal value；
(5.5) it after carrying out location updating, is compared using random number r ∈ [0,1] and maximum detection probability P, if r > P is indicated
The bird's nest is abandoned, rightChanged at random, it is on the contrary then think success, without changing at random, finally retain best
One group of bird's nest position
(5.6) if Gen < MaxGen or not up to minimal error requirement, Gen=Gen+1, return step 5.2, otherwise downwards
Execute 5.7；
(5.7) global optimum's target function value is exported as a result, terminating current algorithm and returning.
6. the propylene polymerization production process optimal online forecasting system according to claim 1 based on weighted array study,
It is characterized in that: the Adaboost module, for the obtained weak learner of support vector machines to be weighted ensemble learning,
It is stronger to obtain an anti-interference ability, the higher strong learner of precision, comprising:
(6.1) algorithm initialization is carried out, setting loss function is root-mean-square error MSE, setting initialization weight vector D1=
(w11... w1n),For D1I-th of component, wherein n be training set element number, setting maximum change
Generation number MaxN, current iteration times N；
(6.2) to weight vector DNTraining dataset learnt by support vector machines, obtain weak learner FN(x)；
(6.3) training set { (x is found outi, yi) on worst error EN=max | yi-FN(xi) |, i=1 ..., n, wherein xi、yiFor
Training set input sample, the data for exporting sample, FNIt (x) is the 6.2 weak learners of gained；Find out the relative error of each sampleWhereinIt is square of worst error；Seek regression error ratewNiFor weight system
Number, seeks weak learner coefficient with regression error rateAnd according to the weight coefficient of regression error rate more new samplesWherein ZNIt is normalization factor, meets
(6.4) if N < MaxN, that is, maximum number of iterations, N=N+1 has not yet been reached, otherwise return step 6.2 executes downwards
6.5；
(6.5) it exportsWherein f (x) is αNFN(x), N=1, the median of 2 ... MaxN, knot
Beam current algorithm simultaneously returns.
7. the propylene polymerization production process optimal online forecasting system based on integrated study according to claim 1, feature
Be: the system update module is used for the online updating of detection system, solves the problems, such as the model mismatch of complicated polymerization process, fixed
Offline analysis data is input to model training and concentrated by the phase, more new detection system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811625115.4A CN109445398B (en) | 2018-12-28 | 2018-12-28 | Propylene polymerization production process optimal online forecasting system based on weighted combination learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811625115.4A CN109445398B (en) | 2018-12-28 | 2018-12-28 | Propylene polymerization production process optimal online forecasting system based on weighted combination learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109445398A true CN109445398A (en) | 2019-03-08 |
| CN109445398B CN109445398B (en) | 2021-04-06 |
Family
ID=65538500
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811625115.4A Expired - Fee Related CN109445398B (en) | 2018-12-28 | 2018-12-28 | Propylene polymerization production process optimal online forecasting system based on weighted combination learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109445398B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112604186A (en) * | 2020-12-30 | 2021-04-06 | 佛山科学技术学院 | Respiratory motion prediction method |
| CN113759834A (en) * | 2020-06-05 | 2021-12-07 | 中国石油天然气股份有限公司 | Chaos multi-scale intelligent optimal propylene polymerization process measuring instrument |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103838209A (en) * | 2013-12-09 | 2014-06-04 | 浙江大学 | Self-adaption optimal forecasting system and method in propylene polymerization production process |
| CN108803525A (en) * | 2018-06-28 | 2018-11-13 | 浙江大学 | A kind of quick high-precision propylene polymerization production process optimal soft survey instrument of chaos |
-
2018
- 2018-12-28 CN CN201811625115.4A patent/CN109445398B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103838209A (en) * | 2013-12-09 | 2014-06-04 | 浙江大学 | Self-adaption optimal forecasting system and method in propylene polymerization production process |
| CN108803525A (en) * | 2018-06-28 | 2018-11-13 | 浙江大学 | A kind of quick high-precision propylene polymerization production process optimal soft survey instrument of chaos |
Non-Patent Citations (1)
| Title |
|---|
| 张杰: ""基于布谷鸟算法的优化问题求解"", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113759834A (en) * | 2020-06-05 | 2021-12-07 | 中国石油天然气股份有限公司 | Chaos multi-scale intelligent optimal propylene polymerization process measuring instrument |
| CN112604186A (en) * | 2020-12-30 | 2021-04-06 | 佛山科学技术学院 | Respiratory motion prediction method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109445398B (en) | 2021-04-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101315557B (en) | Propylene polymerization production process optimal soft survey instrument and method based on genetic algorithm optimization BP neural network | |
| CN108804851A (en) | A kind of high-precision propylene polymerization production process optimal soft survey instrument of chaos gunz optimizing | |
| CN101382801B (en) | Optimum soft measuring instrument based on EGA-optimized polymerization of propylene production process and method | |
| CN108803525A (en) | A kind of quick high-precision propylene polymerization production process optimal soft survey instrument of chaos | |
| CN109507888A (en) | Propylene polymerization production process optimal online forecasting system based on integrated study | |
| CN108982576A (en) | A kind of high-precision propylene polymerization production process optimal soft survey instrument of chaos | |
| CN109445398A (en) | Propylene polymerization production process optimal online forecasting system based on weighted array study | |
| CN103823430B (en) | Intelligence weighting propylene polymerization production process optimal soft measuring system and method | |
| CN103675011B (en) | The industrial melt index soft measurement instrument of optimum support vector machine and method | |
| CN103839103B (en) | Propylene polymerization production process BP Optimal predictor system and method | |
| CN103838206B (en) | Optimum BP multimode propylene polymerization production process optimal soft survey instrument and method | |
| CN108958181A (en) | A kind of propylene polymerization production process optimal soft survey instrument of agility | |
| CN103838209B (en) | Propylene polymerization production process adaptive optimal forecast system and method | |
| CN103955170A (en) | Propylene polymerization production process online forecasting system and method based on group intelligent optimization | |
| CN103824121A (en) | Propylene polymerization production process optimal prediction system based on multimode crowd-sourcing and method | |
| CN103838205B (en) | BP global optimum propylene polymerization production process optimal soft survey instrument and method | |
| CN109829197A (en) | The propylene polymerization production process optimal soft survey instrument of improved cuckoo optimizing | |
| CN103838955A (en) | Optimum soft measurement instrument and method for optimum mixing in propylene polymerization production process | |
| CN103838142A (en) | Propylene polymerization production process optimal soft measurement system and method based on mixed optimizing | |
| CN103675012B (en) | The industrial melt index soft measurement instrument of BP particle group optimizing and method | |
| CN103838958A (en) | Vague intelligent optimal soft measuring instrument and method used in propylene polymerization production process | |
| CN109856971A (en) | Propylene polymerization production process optimal online forecasting system based on gunz optimizing | |
| CN103838957A (en) | Propylene polymerization production process radial basis optimum soft measurement instrument and method | |
| CN109507889A (en) | The propylene polymerization production process optimal online forecasting system of convolutional neural networks | |
| CN113759834A (en) | Chaos multi-scale intelligent optimal propylene polymerization process measuring instrument |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210406 Termination date: 20211228 |