CN110490496A - A method of based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process - Google Patents
A method of based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process Download PDFInfo
- Publication number
- CN110490496A CN110490496A CN201910880769.XA CN201910880769A CN110490496A CN 110490496 A CN110490496 A CN 110490496A CN 201910880769 A CN201910880769 A CN 201910880769A CN 110490496 A CN110490496 A CN 110490496A
- Authority
- CN
- China
- Prior art keywords
- variable
- sample
- cosine
- sensitive
- product quality
- 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 56
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 32
- 238000012216 screening Methods 0.000 title claims abstract description 26
- 230000009467 reduction Effects 0.000 title claims abstract description 20
- 239000011159 matrix material Substances 0.000 claims description 19
- 230000035945 sensitivity Effects 0.000 claims description 18
- 239000003245 coal Substances 0.000 claims description 16
- 230000007246 mechanism Effects 0.000 claims description 14
- 238000004458 analytical method Methods 0.000 claims description 10
- 230000008859 change Effects 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 4
- 238000005065 mining Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005303 weighing Methods 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 claims 1
- 238000012423 maintenance Methods 0.000 abstract 1
- 230000008569 process Effects 0.000 description 11
- 238000001914 filtration Methods 0.000 description 6
- 230000000875 corresponding effect Effects 0.000 description 5
- 239000000243 solution Substances 0.000 description 5
- 238000004517 catalytic hydrocracking Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000002596 correlated effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 239000002283 diesel fuel Substances 0.000 description 2
- 238000009776 industrial production Methods 0.000 description 2
- 238000002347 injection Methods 0.000 description 2
- 239000007924 injection Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- RWSOTUBLDIXVET-UHFFFAOYSA-N Dihydrogen sulfide Chemical compound S RWSOTUBLDIXVET-UHFFFAOYSA-N 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000003556 assay Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012854 evaluation process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Educational Administration (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Manufacturing & Machinery (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of methods based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process, belong to soft-measuring technique field, the following steps are included: choosing the auxiliary variable for influencing product quality by expertise and collecting data sample；Comprehensively consider correlation of variables and variable and auxiliary variable sensibility index is calculated to the sensibility that operating condition changes, preliminary screening influences the sensitive variable of leading variable；Building weighting cosine horse field system, accurate screening influence the crucial sensitive variable of product quality.The present invention can accurately response variable correlation and work information, the redundancy of variable is preferably reduced simultaneously, product quality forecast precision not only can be improved, but also prediction model complexity can be effectively reduced, be equally of great significance to the maintenance of soft sensor model.
Description
Technical field
The present invention relates to soft-measuring technique fields, and in particular to a kind of screened based on substep reduction influences the quick of product quality
Feel the method for variable.
Background technique
With the development of advanced manufacturing technology, manufacturing is to production development from quantity and scale enlargement to quality, benefit
More stringent requirements are proposed for promotion with environmental protection.In order to timely and effectively monitor and evaluation process operation conditions, realization system
The quick tracking of the Accurate Diagnosis, product quality of failure of uniting, needs to be measured in real time process key product quality.However by
It is limited to detect abominable, the sky high cost of analysis instrument and the hysteresis quality of assay of environment, at present these critical products
Quality on-line checking relatively difficult to achieve.Therefore the data-driven soft sensor modeling technology of Kernel-based methods feature and process data is extensive
Applied in industrial production.
Data-driven soft-measuring technique organically combines production process knowledge, and Applied Computer Techniques are to being difficult to
Measurement or temporarily immeasurable product quality, select other to be easy the variable of measurement, are closed by constituting certain mathematics
System is to infer or estimate.Since process measurable variable number is big, if these variables all to be regarded as to the auxiliary of soft sensor modeling
Variable not only will increase the complexity of model, reduce calculating speed, causes dimension disaster, reduce the stability and prediction of model
Precision, and the economic cost of data acquisition and storage can be greatly increased.Therefore, one group how is fast and effeciently chosen to be best able to
The auxiliary variable subset of accurate description or interpretation process leading variable seems particularly important.
Variable optimization method is according to variable search and the difference of evaluation method at present, can be classified into filtering type, packaging type and
Embedded three types.Wherein filtering type method is because its calculating speed is fast and does not easily cause over-fitting to be widely applied.It should
Method is elected the main standard of variable with variables reordering technology, generally using data self character or statistical law as point
Analyse foundation.Common analysis foundation has related coefficient, mutual information, Euclidean distance, Bayesian inference etc..Filtering type variables choice side
Method is by change data come adaptive learning algorithm independent of learning algorithm, but this method is easy omitted variables correlation,
Causing selected subset may not be optimal subset.
To solve filtering type Variable Selection variable redundancy issue, carried out not in domestic and international theoretical circles and industrial practice
Few to attempt and study, these researchs can efficiently solve filtering type Variable Selection and be easy correlation and redundancy between omitted variables
Property problem, do not have but process work information description ability.However during actual industrial production, by entrance primary product
Matter fluctuation, processing scheme adjustment, product specification require variation etc. to influence, and production status is in fluctuation status, operating condition different product
Quality can also have a certain difference.It, will be in certain journey if the auxiliary variable filtered out cannot preferably describe the variation of operating condition
The precision of prediction model is reduced on degree.Therefore one kind can react work information but also react leading variable and auxiliary change at research
The sensitive variable selection method for measuring correlation has the meaning of reality.
Summary of the invention
It is a kind of based on shadow in substep reduction screening complex industrial process technical problem to be solved by the present invention lies in providing
The method realization for ringing the sensitive variable of product quality can react work information but also to react leading variable related to auxiliary variable
Property.
A method of based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process, including with
Lower step:
S1. it is based on production process, by Analysis on Mechanism and expertise, tentatively chooses several possible influence product qualities
Auxiliary variable, and several groups auxiliary variable value and corresponding moment product quality value are collected as sample；
S2. comprehensive consideration auxiliary variable and the correlation and auxiliary variable of product quality calculate the sensibility that operating condition changes
Auxiliary variable sensibility index influences the sensitive variable of product quality according to sensibility index preliminary screening:
S21. outlier rejecting, Wavelet Denoising Method and standardization are carried out to the auxiliary variable value sample of collection；
S22. the correlation matrix of auxiliary variable and product quality is calculated using Pearson came relevant function method, and according to institute
The correlation matrix stated calculates the partial correlation coefficient of auxiliary variable and product quality；
S23. mean value, standard deviation and the variance of auxiliary variable are calculated, and then calculates the coefficient of variation of auxiliary variable；
S24. using the product of the coefficient of variation of the partial correlation coefficient and auxiliary variable of auxiliary variable and product quality as auxiliary
The sensibility index for helping variable is calculated according to auxiliary variable partial correlation coefficient described in step S22 and S23 and the coefficient of variation
Auxiliary variable sensitivity indices；
S25. according to production process object and product quality, based on the expertise threshold different to sensibility target setting
It is worth, the auxiliary variable within the scope of selected threshold is as sensitive variable；
S3. building weighting cosine horse field system carries out attribute to the sensitive variable of primary election from distance and two, direction angle
Reduction accurately filters out the sensitive variable for influencing product quality as crucial sensitive variable:
S31, the sensitive variable sample to collection are classified as normal sample and different by Analysis on Mechanism and expertise
Two class of normal sample, and two class samples are standardized, wherein when handling the exceptional sample data normalization
Mean value and standard deviation are equal to normal sample data；
S32, the mahalanobis distance for calculating separately all normal samples；
S33, the included angle cosine value for calculating separately all normal samples, and then calculate separately the cosine phase of all normal samples
Like degree；
The coefficient of variation of S34, the mahalanobis distance for calculating separately normal sample and cosine similarity, according to mahalanobis distance and remaining
The ratio of the string similarity mutation mechanism coefficient total coefficient of variation of Zhan respectively determines cosine mahalanobis distance weight；
S35, the cosine mahalanobis distance building weighting cosine geneva reference space based on normal sample；
S36, orthogonal arrage is designed, corresponding the weighting cosine geneva reference space of every row, calculates in each base in orthogonal arrage
The cosine mahalanobis distance of exceptional sample in quasi- space；
S37, the signal-to-noise ratio for selecting exceptional sample in each reference space of Wogvily Mining Way signal-to-noise ratio computation；
Then S38, the mean value for calculating separately signal-to-noise ratio when using and be not used the sensitive variable calculate the increasing of its signal-to-noise ratio
Amount sets certain threshold value to signal-to-noise ratio increment according to expertise, and all sensitive variables within the scope of selected threshold are crucial quick
Feel variable.
Further, further include following steps after step S35, before S36:
According to the weighting cosine geneva reference space of building, the cosine mahalanobis distance of exceptional sample is calculated, building is verified
The validity of cosine geneva reference space, if the weighting cosine geneva reference space can preferably distinguish normal sample and exception
The cosine mahalanobis distance of sample, then the weighting cosine geneva reference space constructed are effective；Otherwise, S3 is entered step, is rebuild
Weight cosine horse field system.Further, further include step S4 after step S3: being built using local weighted deflected secondary air
Vertical product quality forecast model, verifies the validity and accuracy of the crucial sensitive variable of selection.
Further, step S21, standardization described in S31 is in the following way:
zij=(xij-μi)/si
Wherein, zijJ-th of sample value of i-th of auxiliary variable or sensitive variable after indicating standardization, xijIt indicates
J-th of sample value of i-th of auxiliary variable or sensitive variable, μiIndicate the mean value of i-th of auxiliary variable or sensitive variable, siTable
Show the standard deviation of i-th of auxiliary variable or sensitive variable.
Further, correlation matrix described in step S22 calculates as follows:
Wherein,
The partial correlation coefficient calculates as follows:
Wherein, cikFor the MccInverse matrixMiddle element:
The coefficient of variation of auxiliary variable described in step S23 calculates as follows:
Wherein, μiIndicate the mean value of i-th of variable, siIndicate the standard deviation of i-th of variable, σiIndicate the side of i-th of variable
Difference；
Auxiliary variable sensitivity indices described in step S24 calculate as follows:
Wherein ηikIndicate i-th of auxiliary variable to the sensitivity indices of k-th of leading variable, rikIndicate that i-th of auxiliary becomes
The partial correlation coefficient of amount and k-th of leading variable, μiIndicate the mean value of i-th of variable, siIndicate the standard deviation of i-th of variable, σi
Indicate the variance of i-th of variable.Further, the mahalanobis distance of normal sample described in step S32 calculates as follows:
S321. n normal sample is chosen, it is assumed that there are q initial sensitive variables, then sample space can indicate in sample
Are as follows:
Wherein oij(i=1,2 ..., n；J=1,2 ..., q) indicate i-th of normal sample, j-th of sensitive variable number
According to；
S322. normal sample data are standardized:
WhereinIndicate the normalized number of i-th of normal sample, j-th of auxiliary variable
According to；
S323. mahalanobis distance are as follows:
Wherein S is the correlation matrix of normal sample,
Further, cosine similarity described in step S33 are as follows:
WhereinFor the data of i-th of sample, j-th of auxiliary variable,For the mean value of j-th of auxiliary variable data.
Further, the weighing computation method of cosine mahalanobis distance described in step S34 are as follows:
Wherein ξ1For the coefficient of variation of normal sample mahalanobis distance, sMDiFor the standard deviation of normal sample mahalanobis distance, μMDi
For the mean value of normal sample mahalanobis distance；ξ2For the coefficient of variation of normal sample cosine similarity, sCSiFor normal sample cosine phase
Like the standard deviation of degree, μCSiFor the mean value of normal sample cosine similarity；
Cosine mahalanobis distance described in step S35 calculates as follows:
CMDi=α MDi+βCSi
Wherein MDiThe mahalanobis distance for indicating sample, to describe the similarity of sample distance；CSiIndicate the cosine phase of sample
Like degree, for describing the similarity of sample orientation；α, β are weight coefficient.
Further, signal-noise ratio computation method described in step S37 is as follows:
Wherein CMDpIndicate that the mahalanobis distance of exceptional sample, m indicate the number of exceptional sample；For auxiliary variable,Indicate the mean value using signal-to-noise ratio when the sensitive variable；The mean value of signal-to-noise ratio when indicating that the sensitive variable is not used；
Signal-to-noise ratio increment Δ SN described in step S38jIt indicates:
Further, the complex industrial process is plus splits production process；The product is that boat coal 10% distillates temperature
Degree.
Compared with prior art, the beneficial effects of the present invention are: in the base of clear sensitive variable and crucial sensitive variable
On plinth, the net correlation of the sensibility and auxiliary variable and product quality that are changed according to variable to operating condition calculates sensibility and refers to
Mark, realizes the primary election of sensitive variable；It solves the problems, such as that variable redundancy is big by the weighting cosine horse field system of building again, realizes
Sensitive variable it is selected.Preferably solve traditional filtering formula Variable Selection be easy omitted variables correlation and cannot be quasi-
Really the problem of reaction work information, has many advantages, such as that calculating is simple, do not easily cause over-fitting, redundancy small.
Detailed description of the invention
Fig. 1 is the flow chart of a specific embodiment of the invention.
Fig. 2 is to add rip current journey sensitive variable signal-to-noise ratio increment histogram in a specific embodiment of the invention.
Fig. 3 is to become using local weighted deflected secondary air in a specific embodiment of the invention and using key is sensitive
The result that duration set is predicted.
Fig. 4 is in a specific embodiment of the invention using local weighted deflected secondary air and utilization sensitive variable collection
Close the result predicted.
Fig. 5 is to use local weighted deflected secondary air in a specific embodiment of the invention and screened using mechanism auxiliary
Help the result that variables collection is predicted.
Fig. 6 is to become using local weighted deflected secondary air in a specific embodiment of the invention and using key is sensitive
The relative error result that duration set is predicted.
Fig. 7 is in a specific embodiment of the invention using local weighted deflected secondary air and utilization sensitive variable collection
Close the relative error result predicted.
Fig. 8 is to use local weighted deflected secondary air in a specific embodiment of the invention and screened using mechanism auxiliary
The relative error result for helping variables collection to be predicted.
Specific embodiment
In order to further disclose the present invention, technical solution disclosed in this invention is carried out below in conjunction with Figure of description
Comprehensively, it meticulously describes:
As shown in Figure 1, a kind of screened in complex industrial process based on substep reduction provided by the invention influences product quality
Sensitive variable method, comprising the following steps:
S1. it is based on production process, by Analysis on Mechanism and expertise, tentatively chooses several possible influence product qualities
Auxiliary variable, and several groups auxiliary variable value and corresponding moment product quality value are collected as sample；
S2. comprehensive consideration auxiliary variable and the correlation and auxiliary variable of product quality calculate the sensibility that operating condition changes
Auxiliary variable sensibility index influences the sensitive variable of product quality according to sensibility index preliminary screening.
S3. building weighting cosine horse field system carries out attribute to the sensitive variable of primary election from distance and two, direction angle
Reduction accurately filters out the sensitive variable for influencing product quality as crucial sensitive variable.
As an improvement preceding solution can also include step S4: using local weighted deflected secondary air
Product quality forecast model is established, the validity and accuracy of the crucial sensitive variable of selection are verified.
Step S2 the specific implementation process is as follows:
S21. outlier rejecting, Wavelet Denoising Method and standardization are carried out to the auxiliary variable value sample of collection；
S22. the correlation matrix of auxiliary variable and product quality is calculated using Pearson came relevant function method, and according to institute
The correlation matrix stated calculates the partial correlation coefficient of auxiliary variable and product quality；
S23. mean value, standard deviation and the variance of auxiliary variable are calculated, and then calculates the coefficient of variation of auxiliary variable；
S24. using the product of the coefficient of variation of the partial correlation coefficient and auxiliary variable of auxiliary variable and product quality as auxiliary
The sensibility index for helping variable is calculated according to auxiliary variable partial correlation coefficient described in step S22 and S23 and the coefficient of variation
Auxiliary variable sensitivity indices；
S25. according to production process object and product quality, based on the expertise threshold different to sensibility target setting
It is worth, the auxiliary variable within the scope of selected threshold is as sensitive variable.
Standardization described in preceding solution step S21, S31 is in the following way:
zij=(xij-μi)/si
Wherein, zijJ-th of sample value of i-th of auxiliary variable or sensitive variable after indicating standardization, xijIt indicates
J-th of sample value of i-th of auxiliary variable or sensitive variable, μiIndicate the mean value of i-th of auxiliary variable or sensitive variable, siTable
Show the standard deviation of i-th of auxiliary variable or sensitive variable.
Correlation matrix described in step S22 calculates as follows:
Wherein,
The partial correlation coefficient calculates as follows:
Wherein, cikFor the MccInverse matrixMiddle element:
The coefficient of variation of auxiliary variable described in step S23 calculates as follows:
Wherein, μiIndicate the mean value of i-th of variable, siIndicate the standard deviation of i-th of variable, σiIndicate the side of i-th of variable
Difference；
Auxiliary variable sensitivity indices described in step S24 calculate as follows:
Wherein ηikIndicate i-th of auxiliary variable to the sensitivity indices of k-th of leading variable, rikIndicate that i-th of auxiliary becomes
The partial correlation coefficient of amount and k-th of leading variable, μiIndicate the mean value of i-th of variable, siIndicate the standard deviation of i-th of variable, σi
Indicate the variance of i-th of variable.Step S3 the specific implementation process is as follows:
S31, the sensitive variable sample to collection are classified as normal sample and different by Analysis on Mechanism and expertise
Two class of normal sample, and two class samples are standardized, wherein when handling the exceptional sample data normalization
Mean value and standard deviation are equal to normal sample data；
S32, the mahalanobis distance for calculating separately all normal samples；
S33, the included angle cosine value for calculating separately all normal samples, and then calculate separately the cosine phase of all normal samples
Like degree；
The coefficient of variation of S34, the mahalanobis distance for calculating separately normal sample and cosine similarity, according to mahalanobis distance and remaining
The ratio of the string similarity mutation mechanism coefficient total coefficient of variation of Zhan respectively determines cosine mahalanobis distance weight；
S35, the cosine mahalanobis distance building weighting cosine geneva reference space based on normal sample；
S36, orthogonal arrage is designed, corresponding the weighting cosine geneva reference space of every row, calculates in each base in orthogonal arrage
The cosine mahalanobis distance of exceptional sample in quasi- space；
S37, the signal-to-noise ratio for selecting exceptional sample in each reference space of Wogvily Mining Way signal-to-noise ratio computation；
Then S38, the mean value for calculating separately signal-to-noise ratio when using and be not used the sensitive variable calculate the increasing of its signal-to-noise ratio
Amount sets certain threshold value to signal-to-noise ratio increment according to expertise, and all sensitive variables within the scope of selected threshold are crucial quick
Feel variable.
As a kind of improvement of preceding solution, after step S35, before S36 further include following steps:
According to the weighting cosine geneva reference space of building, the cosine mahalanobis distance of exceptional sample is calculated, building is verified
The validity of cosine geneva reference space, if the weighting cosine geneva reference space can preferably distinguish normal sample and exception
The cosine mahalanobis distance of sample, then the weighting cosine geneva reference space constructed are effective；Otherwise, S3 is entered step, is rebuild
Weight cosine horse field system.The mahalanobis distance of normal sample described in step S32 calculates as follows:
S321. n normal sample is chosen, it is assumed that there are q initial sensitive variables, then sample space can indicate in sample
Are as follows:
Wherein oij(i=1,2 ..., n；J=1,2 ..., q) indicate i-th of normal sample, j-th of sensitive variable number
According to；
S322. normal sample data are standardized:
WhereinIndicate the normalized number of i-th of normal sample, j-th of auxiliary variable
According to；
S323. mahalanobis distance are as follows:
Wherein S is the correlation matrix of normal sample,
Cosine similarity described in step S33 are as follows:
WhereinFor the data of i-th of sample, j-th of auxiliary variable,For the mean value of j-th of auxiliary variable data.
The weighing computation method of cosine mahalanobis distance described in step S34 are as follows:
Wherein ξ1For the coefficient of variation of normal sample mahalanobis distance, sMDiFor the standard deviation of normal sample mahalanobis distance, μMDi
For the mean value of normal sample mahalanobis distance；ξ2For the coefficient of variation of normal sample cosine similarity, sCSiFor normal sample cosine phase
Like the standard deviation of degree, μCSiFor the mean value of normal sample cosine similarity；
Cosine mahalanobis distance described in step S35 calculates as follows:
CMDi=α MDi+βCSi
Wherein MDiThe mahalanobis distance for indicating sample, to describe the similarity of sample distance；CSiIndicate the cosine phase of sample
Like degree, for describing the similarity of sample orientation；α, β are weight coefficient.
Signal-noise ratio computation method described in step S37 is as follows:
Wherein CMDpIndicate that the mahalanobis distance of exceptional sample, m indicate the number of exceptional sample；For auxiliary variable,
Indicate the mean value using signal-to-noise ratio when the sensitive variable；The mean value of signal-to-noise ratio when indicating that the sensitive variable is not used；
Signal-to-noise ratio increment Δ SN described in step S38jIt indicates:
Specific embodiment:
Application of the technical solution disclosed by the invention to hydrocracking process product quality forecast, comprising the following steps:
A, production process is split based on adding, it is preliminary to choose to boat 10% recovered (distilled) temperature of coal according to Analysis on Mechanism and expertise
The influential correlated variables of quality index totally 39, as the input variable of hydrocracking process prediction of quality, it is denoted as x respectively1、
x2、…、x39, it is extracted the sampled data of 966 days quantity-produceds, 39 correlated variables, when being extracted 966 days daily 8 and 20
When offline chemical examination measure boat 10% recovered (distilled) temperature of coal data, totally 1932 groups.The data obtained is divided into two parts and is used to be based on
The 10% recovered (distilled) temperature quality index of boat coal of LWPLS predicts modeling, wherein 1288 groups are made training set, 644 groups are made test set.It uses
The input data that 180 groups of data are selected as sensitive variable in training set, then input matrix are as follows:
xi=[xI, 1, xI, 2..., xI, 39]T, i=1,2 ..., 180
X=[x1,x2..., x180]
B, comprehensively consider the sensibility that correlation of variables and variable change operating condition, calculate auxiliary variable sensibility index,
The sensitive variable of boat 10% recovered (distilled) temperature of coal is influenced according to sensibility index preliminary screening.
The auxiliary variable sample of selection is standardized:
zij=(xij-μi)/si
Wherein zijData value after indicating standardization, xijIndicate j-th of sample value of i-th of variable, μiIndicate i-th of change
The mean value of amount, siIndicate the standard deviation of i-th of variable.
The partial correlation coefficient for calculating auxiliary variable calculates correlation matrix using Pearson came relevant function method:
Wherein,Auxiliary after standardization
Variable ziWith the partial correlation coefficient r of boat 10% recovered (distilled) temperature A10% of coalIA10%Are as follows:
Wherein cIA10%For MccInverse matrixMiddle element
Calculate the coefficient of variation for the auxiliary variable chosen:
Wherein, μiIndicate the mean value of i-th of variable, siIndicate the standard deviation of i-th of variable, σiIndicate the side of i-th of variable
Difference.
Calculate the sensitivity indices of auxiliary variable, i.e. auxiliary variable and the partial correlation coefficient of 10% recovered (distilled) temperature of coal and auxiliary of navigating
Help the product of the coefficient of variation of variable:
Wherein ηIA10%Indicate i-th of auxiliary variable to the sensitivity indices of boat 10% recovered (distilled) temperature A10% of coal, rIA10%Table
Show the partial correlation coefficient of i-th of auxiliary variable and the 10% recovered (distilled) temperature A10% of coal that navigates, μiIndicate the mean value of i-th of variable, siTable
Show the standard deviation of i-th of variable, σiIndicate the variance of i-th of variable.
Sensitivity indices show that more greatly the auxiliary variable is bigger to the influence of boat 10% recovered (distilled) temperature A10% of coal, right
The variation of operating condition is more sensitive.Calculate the dispersion degree of 39 auxiliary variables and the partial correlation system of boat 10% recovered (distilled) temperature A10% of coal
Several and sensitivity indices, the results are shown in Table 1.
Table 1 is hydrocracked process mechanism screening auxiliary variable sensitivity indices
Based on boat 10% recovered (distilled) temperature of coal, the threshold value of auxiliary variable sensibility index is set as according to expertise
0.01, the sensitivity indices of each variable are analyzed it is found that finishing reactor column bottom temperature indicates (12), finishing reactor pressure difference
(13), water injection tank water injection rate (21), depriving hydrogen sulphide stripper overhead regurgitant volume (24), main fractionating tower middle section return temperature (32),
Diesel oil stripper overhead temperature (38), diesel oil stripper bottom temp (39) sensitivity indices are lower.Therefore this 7 sensitivities are removed
Other remaining 32 auxiliary variables are sensitive variable other than the lower variable of sex index.
C, building weighting cosine horse field system carries out attribute about to the sensitive variable of primary election from distance and two, direction angle
Letter, the selected specific steps of crucial sensitive variable for influencing boat 10% recovered (distilled) temperature A10% of coal include:
(1) 32 normal samples are chosen, there are 32 initial sensitive variables in sample, then sample space can indicate are as follows:
Wherein oij(i=1,2 ..., 32；J=1,2 ..., 32) indicate i-th of normal sample, j-th of sensitive variable number
According to.
Normal sample data are standardized:
WhereinIndicate the normalized number of i-th of normal sample, j-th of auxiliary variable
According to.
Calculate the cosine mahalanobis distance of all normal samples:
CMDi=α MDi+βCSi
Wherein MDiThe mahalanobis distance for indicating sample, to describe the similarity of sample distance；CSiIndicate the cosine phase of sample
Like degree, for describing the similarity of sample orientation；α, β are weight coefficient.
Calculate the mahalanobis distance MD of samplei:
Wherein S is the correlation matrix of normal sample,
Calculate the cosine similarity of sample:
WhereinFor the data of i-th of sample, j-th of auxiliary variable,For the mean value of j-th of auxiliary variable data.
Cosine mahalanobis distance is determined according to the mahalanobis distance degree of variation of normal sample and cosine similarity degree of variation
Weight, specific formula is as follows:
Wherein ξ1For the coefficient of variation of normal sample mahalanobis distance, sMDiFor the standard deviation of normal sample mahalanobis distance, μMDi
For the mean value of normal sample mahalanobis distance；ξ2For the coefficient of variation of normal sample cosine similarity, sCSiFor normal sample cosine phase
Like the standard deviation of degree, μCSiFor the mean value of normal sample cosine similarity.Table 2 is that weighting cosine geneva reference space part is shown
As a result.
Table 2 weights cosine geneva reference space
The cosine mahalanobis distance of normal sample fluctuates near 1 substantially as shown in Table 2, mean value 0.9020.
(2) exceptional sample is standardized, then calculates separately the mahalanobis distance of exceptional sample, exceptional sample and normal
The cosine similarity and cosine mahalanobis distance of sample average, the results are shown in Table 3.
3 exceptional sample cosine mahalanobis distance of table
The cosine mahalanobis distance of exceptional sample is much larger than 1 as shown in Table 3, mean value 205.5255, thus construct plus
Power cosine geneva reference space can be very good to distinguish normal sample and exceptional sample.Wherein exceptional sample 3 is specially selected
Outlier exceptional sample, mahalanobis distance 1.6571, if differentiating sample according only to mahalanobis distance according to traditional horse field system,
Sample 3 is normal sample, is not inconsistent with actual conditions；And the cosine similarity of sample 3 is 5.5180, cosine mahalanobis distance is
2.2362, cosine horse field system is weighted at this time by sample 3 and is determined as exceptional sample, compared to traditional horse field system, weights cosine
Horse field system can preferably distinguish normal sample and exceptional sample.
(3) optimize reference space: orthogonal arrage shown in design table 4, level 1 indicate that, using auxiliary variable, level 2 indicates not
Using auxiliary variable, and calculate signal-to-noise ratio.The corresponding reference space of every row, calculates different in each reference space in orthogonal arrage
The cosine mahalanobis distance and signal-to-noise ratio, calculation formula of normal sample are as follows:
Wherein CMDpFor the mahalanobis distance of exceptional sample.For auxiliary variable,When indicating using the sensitive variable
The mean value of signal-to-noise ratio；The mean value of signal-to-noise ratio when indicating that the sensitive variable is not used.Signal-to-noise ratio increment Δ SNjIt indicates.
It navigates 10% recovered (distilled) temperature of coal for hydrocracking process, signal-to-noise ratio delta threshold is set as 0.3 based on expertise,
All sensitive variables within the scope of selected threshold are that (serial number mechanism screens auxiliary variable sensibility to crucial sensitive variable in bracket
Variable serial number in index table).
4 two-level orthogonal array of table and signal-to-noise ratio
The signal-to-noise ratio increment histograms of 32 sensitive variables as shown in Fig. 2, variable 21 (original machine reason screening auxiliary variable 25),
The signal-to-noise ratio increment of 28 (original machine reason screening auxiliary variables 33) and 32 (original machine reason screening auxiliary variables 37) is negative, and illustrates that these are auxiliary
Help variable invalid to modeling；The signal-to-noise ratio increment of variable 26 (original machine reason screening auxiliary variable 30) is smaller, illustrates the auxiliary variable
It is smaller to modeling effect, it can be ignored.Finally obtain 28 crucial sensitive variables that can be used for predicting modeling.
D, prediction model is established using local weighted offset minimum binary (LWPLS) method, the data for modeling are shared
1610 groups, wherein 966 groups are used as training set, 644 are used as test set, respectively by auxiliary variable set according to mechanism selection variables collection
It closes, sensitive variable set is used to model with crucial sensitive variable set and model parameter is identical, prediction result such as Fig. 3-5 institute
Show, as shown in figs 6-8, root-mean-square error RMSE is as shown in table 5 for prediction error.By Fig. 3 and Fig. 6 it is found that utilizing crucial sensitive change
Amount carries out prediction modeling, and prediction result can preferably track boat 10% recovered (distilled) temperature of coal compared with other two kinds of auxiliary variable set
Actual value, prediction error it is smaller, and its prediction root-mean-square error RMSE be 3.0390, compared with other two kinds of auxiliary variable set
5.81% and 3.94% has been respectively increased, has demonstrated the validity of the proposed method of the present invention.
The root-mean-square error RMSE of 53 kinds of variables collection prediction modelings of table
In addition, offset minimum binary (PLS), support vector machines (SVM) and local Weighted Kernel pivot is also respectively adopted in the present invention
Return (LWKPCR) 3 kinds of method validations validity of mentioned method of the invention, the root-mean-square error of three kinds of methods such as 6 institute of table
Show.
The root-mean-square error RMSE of 63 kinds of variables collection difference prediction modelings of table
Finally, above-described embodiment is used for the purpose of clearly demonstrating example, and do not limit the embodiments.
For those of ordinary skill in the art, other various forms of variations can also be made on the basis of the above description
Or it changes.All within the spirits and principles of the present invention, any modification, equivalent replacement, improvement and so on should be included in this
Within the protection scope of invention.
Claims (10)
1. a kind of method based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process, feature exist
In, comprising the following steps:
S1. it is based on production process, by Analysis on Mechanism and expertise, tentatively chooses several possible auxiliary for influencing product quality
Variable, and several groups auxiliary variable value and corresponding moment product quality value are collected as sample；
S2. the correlation and auxiliary variable of comprehensive consideration auxiliary variable and product quality, which calculate the sensibility that operating condition changes, assists
Variable sensibility index influences the sensitive variable of product quality according to sensibility index preliminary screening:
S21. outlier rejecting, Wavelet Denoising Method and standardization are carried out to the auxiliary variable value sample of collection；
S22. the correlation matrix of auxiliary variable and product quality is calculated using Pearson came relevant function method, and according to described
The partial correlation coefficient of correlation matrix calculating auxiliary variable and product quality；
S23. mean value, standard deviation and the variance of auxiliary variable are calculated, and then calculates the coefficient of variation of auxiliary variable；
S24. become using the product of the coefficient of variation of the partial correlation coefficient and auxiliary variable of auxiliary variable and product quality as auxiliary
Auxiliary is calculated according to auxiliary variable partial correlation coefficient described in step S22 and S23 and the coefficient of variation in the sensibility index of amount
Variable sensitivity indices；
S25. according to production process object and product quality, based on the expertise threshold value different to sensibility target setting, choosing
Take the auxiliary variable in threshold range as sensitive variable；
S3. building weighting cosine horse field system carries out attribute reduction to the sensitive variable of primary election from distance and two, direction angle,
The sensitive variable for influencing product quality is accurately filtered out as crucial sensitive variable:
S31. normal sample and abnormal sample are classified as by Analysis on Mechanism and expertise to the sensitive variable sample of collection
This two class, and two class samples are standardized, wherein mean value when handling the exceptional sample data normalization
It is equal to normal sample data with standard deviation；
S32. the mahalanobis distance of all normal samples is calculated separately；
S33. the included angle cosine value of all normal samples is calculated separately, and then the cosine for calculating separately all normal samples is similar
Degree；
S34. the mahalanobis distance of normal sample and the coefficient of variation of cosine similarity are calculated separately, according to mahalanobis distance and cosine phase
Cosine mahalanobis distance weight is determined like the ratio of the degree coefficient of variation total coefficient of variation of Zhan respectively；
S35. the cosine mahalanobis distance building weighting cosine geneva reference space based on normal sample；
S36. orthogonal arrage is designed, the corresponding weighting cosine geneva reference space of every row in orthogonal arrage calculates empty in each benchmark
Between middle exceptional sample cosine mahalanobis distance；
S37. the signal-to-noise ratio of exceptional sample in each reference space of Wogvily Mining Way signal-to-noise ratio computation is selected；
S38. the mean value of signal-to-noise ratio when using and be not used the sensitive variable is calculated separately, its signal-to-noise ratio increment, root are then calculated
Certain threshold value is set to signal-to-noise ratio increment according to expertise, all sensitive variables within the scope of selected threshold are crucial sensitive change
Amount.
2. according to claim 1 based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process
Method, it is characterised in that: after step S35, before S36 further include following steps:
According to the weighting cosine geneva reference space of building, the cosine mahalanobis distance of exceptional sample is calculated, the cosine of building is verified
The validity of geneva reference space, if the weighting cosine geneva reference space can preferably distinguish normal sample and exceptional sample
Cosine mahalanobis distance, then the weighting cosine geneva reference space constructed is effective；Otherwise, S3 is entered step, weighting is rebuild
Cosine horse field system.
3. according to claim 1 based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process
Method, it is characterised in that: further include step S4 after step S3: product matter established using local weighted deflected secondary air
Prediction model is measured, the validity and accuracy of the crucial sensitive variable of selection are verified.
4. according to claim 1 based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process
Method, it is characterised in that: step S21, the standardization described in S31 is in the following way:
zij=(xij-μi)/si
Wherein, zijJ-th of sample value of i-th of auxiliary variable or sensitive variable after indicating standardization, xijIt indicates i-th
J-th of sample value of auxiliary variable or sensitive variable, μiIndicate the mean value of i-th of auxiliary variable or sensitive variable, siIndicate i-th
The standard deviation of a auxiliary variable or sensitive variable.
5. according to claim 1 based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process
Method, it is characterised in that: correlation matrix described in step S22 calculate it is as follows:
Wherein,
The partial correlation coefficient calculates as follows:
Wherein, cikFor the MccInverse matrixMiddle element:
The coefficient of variation of auxiliary variable described in step S23 calculates as follows:
Wherein, μiIndicate the mean value of i-th of variable, siIndicate the standard deviation of i-th of variable, σiIndicate the variance of i-th of variable；
Auxiliary variable sensitivity indices described in step S24 calculate as follows:
Wherein ηikIndicate i-th of auxiliary variable to the sensitivity indices of k-th of leading variable, rikIndicate i-th of auxiliary variable with
The partial correlation coefficient of k-th of leading variable, μiIndicate the mean value of i-th of variable, siIndicate the standard deviation of i-th of variable, σiIt indicates
The variance of i-th of variable.
6. according to claim 1 based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process
Method, it is characterised in that: the mahalanobis distance of normal sample described in step S32 calculates as follows:
S321. n normal sample is chosen, it is assumed that there are q initial sensitive variables, then sample space can indicate in sample are as follows:
Wherein oij(i=1,2 ..., n；J=1,2 ..., q) indicate i-th of normal sample, j-th of sensitive variable data；
S322. normal sample data are standardized:
WhereinIndicate the standardized data of i-th of normal sample, j-th of auxiliary variable；
S323. mahalanobis distance are as follows:
Wherein S is the correlation matrix of normal sample,
7. according to claim 1 based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process
Method, it is characterised in that: cosine similarity described in step S33 are as follows:
WhereinFor the data of i-th of sample, j-th of auxiliary variable,For the mean value of j-th of auxiliary variable data.
8. according to claim 1 based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process
Method, it is characterised in that:
The weighing computation method of cosine mahalanobis distance described in step S34 are as follows:
Wherein ξ1For the coefficient of variation of normal sample mahalanobis distance, sMDiFor the standard deviation of normal sample mahalanobis distance, μMDiIt is normal
The mean value of sample mahalanobis distance；ξ2For the coefficient of variation of normal sample cosine similarity, sCSiFor normal sample cosine similarity
Standard deviation, μCSiFor the mean value of normal sample cosine similarity；
Cosine mahalanobis distance described in step S35 calculates as follows:
CMDi=α MDi+βCSi
Wherein MDiThe mahalanobis distance for indicating sample, to describe the similarity of sample distance；CSiIndicate that the cosine of sample is similar
Degree, for describing the similarity of sample orientation；α, β are weight coefficient.
9. according to claim 1 based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process
Method, it is characterised in that: signal-noise ratio computation method described in step S37 is as follows:
Wherein CMDpIndicate that the mahalanobis distance of exceptional sample, m indicate the number of exceptional sample；For auxiliary variable,It indicates
Use the mean value of signal-to-noise ratio when the sensitive variable；The mean value of signal-to-noise ratio when indicating that the sensitive variable is not used；
Signal-to-noise ratio increment Δ SN described in step S38jIt indicates:
10. according to claim 1 based on the sensitive change for influencing product quality in substep reduction screening complex industrial process
The method of amount, it is characterised in that: the complex industrial process is plus splits production process；The product is that boat coal 10% distillates
Temperature.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2019105299243 | 2019-06-19 | ||
| CN201910529924 | 2019-06-19 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110490496A true CN110490496A (en) | 2019-11-22 |
| CN110490496B CN110490496B (en) | 2022-03-11 |
Family
ID=68558550
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910880769.XA Active CN110490496B (en) | 2019-06-19 | 2019-09-18 | Method for screening sensitive variables influencing product quality in complex industrial process based on stepwise reduction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110490496B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112101458A (en) * | 2020-09-16 | 2020-12-18 | 河海大学常州校区 | Taguchi function-signal-to-noise ratio-based characteristic measurement method and device |
| CN112465044A (en) * | 2020-12-03 | 2021-03-09 | 上海卫星工程研究所 | Satellite working condition identification and segmentation method and system based on sensitive parameter multi-evidence fusion |
| CN114186879A (en) * | 2021-12-15 | 2022-03-15 | 中国地质大学（北京） | Method and equipment for evaluating influence of geological parameters on resource quantity calculation errors |
| CN116306931A (en) * | 2023-05-24 | 2023-06-23 | 典基网络科技(上海)有限公司 | Knowledge graph construction method applied to industrial field |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130200144A1 (en) * | 2010-07-19 | 2013-08-08 | Advanced Track & Trace | Methods and devices for marking and authenticating a product by a consumer |
| US20170055829A1 (en) * | 2015-09-01 | 2017-03-02 | Oregon Health & Science University | Systems and methods of gluacoma diagnosis based on frequency analysis of inner retinal surface profile measured by optical coherence tomography |
| CN106872658A (en) * | 2017-01-22 | 2017-06-20 | 华南理工大学 | A kind of method of the COD of sewage load prediction based on vector time series model |
| CN108776831A (en) * | 2018-05-15 | 2018-11-09 | 中南大学 | A kind of complex industrial process Data Modeling Method based on dynamic convolutional neural networks |
| CN109164794A (en) * | 2018-11-22 | 2019-01-08 | 中国石油大学(华东) | Multivariable industrial process Fault Classification based on inclined F value SELM |
| CN109307749A (en) * | 2018-11-06 | 2019-02-05 | 重庆大学 | A kind of aeration disturbs the correlation analysis of lower black-odor riverway water quality indicator |
| CN110210687A (en) * | 2019-06-13 | 2019-09-06 | 中南大学 | A kind of Nonlinear Dynamic production process product quality prediction technique returned based on local weighted slow feature |
-
2019
- 2019-09-18 CN CN201910880769.XA patent/CN110490496B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130200144A1 (en) * | 2010-07-19 | 2013-08-08 | Advanced Track & Trace | Methods and devices for marking and authenticating a product by a consumer |
| US20170055829A1 (en) * | 2015-09-01 | 2017-03-02 | Oregon Health & Science University | Systems and methods of gluacoma diagnosis based on frequency analysis of inner retinal surface profile measured by optical coherence tomography |
| CN106872658A (en) * | 2017-01-22 | 2017-06-20 | 华南理工大学 | A kind of method of the COD of sewage load prediction based on vector time series model |
| CN108776831A (en) * | 2018-05-15 | 2018-11-09 | 中南大学 | A kind of complex industrial process Data Modeling Method based on dynamic convolutional neural networks |
| CN109307749A (en) * | 2018-11-06 | 2019-02-05 | 重庆大学 | A kind of aeration disturbs the correlation analysis of lower black-odor riverway water quality indicator |
| CN109164794A (en) * | 2018-11-22 | 2019-01-08 | 中国石油大学(华东) | Multivariable industrial process Fault Classification based on inclined F value SELM |
| CN110210687A (en) * | 2019-06-13 | 2019-09-06 | 中南大学 | A kind of Nonlinear Dynamic production process product quality prediction technique returned based on local weighted slow feature |
Non-Patent Citations (1)
| Title |
|---|
| 王一帆: ""东北三省循环经济效率评价及其影响因素分析"", 《中国优秀博士学位论文全文数据库（博士） 经济与管理科学辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112101458A (en) * | 2020-09-16 | 2020-12-18 | 河海大学常州校区 | Taguchi function-signal-to-noise ratio-based characteristic measurement method and device |
| CN112101458B (en) * | 2020-09-16 | 2024-04-19 | 河海大学常州校区 | Characteristic measurement method and device based on field function-signal-to-noise ratio |
| CN112465044A (en) * | 2020-12-03 | 2021-03-09 | 上海卫星工程研究所 | Satellite working condition identification and segmentation method and system based on sensitive parameter multi-evidence fusion |
| CN114186879A (en) * | 2021-12-15 | 2022-03-15 | 中国地质大学（北京） | Method and equipment for evaluating influence of geological parameters on resource quantity calculation errors |
| CN116306931A (en) * | 2023-05-24 | 2023-06-23 | 典基网络科技(上海)有限公司 | Knowledge graph construction method applied to industrial field |
| CN116306931B (en) * | 2023-05-24 | 2023-08-04 | 典基网络科技(上海)有限公司 | Knowledge graph construction method applied to industrial field |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110490496B (en) | 2022-03-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110490496A (en) | A method of based on the sensitive variable for influencing product quality in substep reduction screening complex industrial process | |
| CN101630376B (en) | Soft-sensing modeling method and soft meter of multi-model neural network in biological fermentation process | |
| CN116380445B (en) | Equipment state diagnosis method and related device based on vibration waveform | |
| CN113259331A (en) | Unknown abnormal flow online detection method and system based on incremental learning | |
| CN111914889B (en) | Rectifying tower abnormal state identification method based on simplified kernel principal component analysis | |
| CN116383636A (en) | Coal mill fault early warning method based on PCA and LSTM fusion algorithm | |
| CN112507479B (en) | Oil drilling machine health state assessment method based on manifold learning and softmax | |
| CN113240527A (en) | Bond market default risk early warning method based on interpretable machine learning | |
| CN105868164A (en) | Soft measurement modeling method based on monitored linear dynamic system model | |
| CN111461354A (en) | Machine learning integration classification method and software system for high-dimensional data | |
| Chen et al. | Key performance index estimation based on ensemble locally weighted partial least squares and its application on industrial nonlinear processes | |
| Yang et al. | An intelligent singular value diagnostic method for concrete dam deformation monitoring | |
| Guo et al. | Multiblock adaptive convolution kernel neural network for fault diagnosis in a large-scale industrial process | |
| CN111985158A (en) | Electro-hydraulic servo valve fault diagnosis method and system based on transfer learning | |
| CN116739100A (en) | Vulnerability detection method of quantum neural network and automatic driving vulnerability detection method | |
| Wang et al. | Identification of abnormal conditions for gold flotation process based on multivariate information fusion and double‐channel convolutional neural network | |
| CN116339275A (en) | Multi-scale process fault detection method based on full-structure dynamic autoregressive hidden variable model | |
| CN111931574B (en) | Robust fault diagnosis method for pneumatic regulating valve | |
| CN114021465A (en) | Electric power system robust state estimation method and system based on deep learning | |
| CN115034307A (en) | Vibration data self-confirmation method, system and terminal for multi-parameter self-adaptive fusion | |
| Shang et al. | Sparse slow feature analysis for enhanced control monitoring and fault isolation | |
| CN111859798A (en) | Flow industrial fault diagnosis method based on bidirectional long-time and short-time neural network | |
| Zhang et al. | Programming strategies of sequential incremental-scale subproblems for large scale data reconciliation and parameter estimation with multi-operational conditions | |
| CN110990916A (en) | Evaluation and prediction integration method for long-term operation safety of dam considering hysteresis effect | |
| Na et al. | A soft-sensing model for feedwater flow rate using fuzzy support vector regression |
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 |