CN108646688B - A kind of process parameter optimizing analysis method based on recurrence learning - Google Patents
A kind of process parameter optimizing analysis method based on recurrence learning Download PDFInfo
- Publication number
- CN108646688B CN108646688B CN201810550965.6A CN201810550965A CN108646688B CN 108646688 B CN108646688 B CN 108646688B CN 201810550965 A CN201810550965 A CN 201810550965A CN 108646688 B CN108646688 B CN 108646688B
- Authority
- CN
- China
- Prior art keywords
- parameter
- data
- running
- running parameter
- performance
- 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.)
- Active
Links
Classifications
-
- 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]
Landscapes
- Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Manufacturing & Machinery (AREA)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
Abstract
The process parameter optimizing analysis method based on recurrence learning that the invention discloses a kind of, comprising: choose one group of technological parameter, seek its running parameter and show the cross-correlation matrix between parameter, obtain the running parameter that relevance is greater than preset value;Relevance is greater than to two groups of the running parameter extraction of preset value, one group is done trained use, and one group is done verifying and used, two groups of performance parameter extraction will verified without relevance, the first performance parameter used as training and the second performance parameter as verifying;Using linear regression method, the data that the data of the first running parameter are showed parameter with first solve to be associated with the initial coefficients matrix of the first running parameter and the first performance parameter as training sequence;It brings the data of the second running parameter into initial coefficients matrix, must predict the data of performance parameter, the root-mean-square error between the data of prediction performance parameter and the data of the second performance parameter is asked to obtain verified coefficient matrix when root-mean-square error is close to 0.
Description
Technical field
The present invention relates to modern engineering technology field more particularly to a kind of process parameter optimizing analyses based on recurrence learning
Method.
Background technique
Very accurate analytic modell analytical model is had no between modern engineering technology field, many technological parameters to be described, and
Seem between them but it is fixed there are certain connections, in this case, we can be by such method to these
Parameter is analyzed, and common application scenarios include: relationship analysis between the running parameter of equipment and performance parameter, producing line matter
Measure the relationship analysis etc. between data and operation technological parameter.And the prior art and one kind can not be effectively provided can be by equipment
Each parameter between opening relationships method, and then engineering can not be effectively estimated.
Summary of the invention
The embodiment of the present invention solves existing by providing a kind of process parameter optimizing analysis method based on recurrence learning
Can not effectively be provided in technology it is a kind of can by the model of opening relationships between each parameter of equipment, and then can not to engineering into
The technical issues of row is effectively estimated.
In order to solve the above-mentioned technical problem, the embodiment of the invention provides a kind of process parameter optimizings based on recurrence learning
Analysis method, comprising:
S1 carries out classification storage according to data mode to technological parameter, and the technological parameter includes running parameter and therewith
Corresponding performance parameter;
S2 chooses the technological parameter in a kind of data mode, calculates its running parameter and shows the cross-correlation between parameter
Matrix obtains the running parameter that relevance is greater than preset value;
The running parameter that relevance is greater than preset value is extracted two groups, the first running parameter and work as training by S3
For the second running parameter of verifying, two groups of performance parameter extraction will verified without relevance, the first table as training
Now parameter and the second performance parameter as verifying;
The data of first running parameter are showed the data of parameter as training with first using linear regression method by S4
Sequence solves the initial coefficients matrix for obtaining and being associated with first running parameter and the first performance parameter;
S5 brings the data of the second running parameter into the coefficient matrix, solves the data for obtaining prediction performance parameter, meter
The root-mean-square error between the data of the prediction performance parameter and the data of the second performance parameter is calculated, in the root-mean-square error
When close to 0, verified coefficient matrix is obtained.
Further, after S5, further includes:
S1~S5 is repeated, continuous iteration obtains more accurate coefficient matrix.
Further, before S1, further includes:
Structuring arrangement is carried out to the data of technological parameter, and improper data is rejected using statistical method.
Further, further includes:
It determines the target performance parameter for wishing to reach, brings into the coefficient matrix, obtain corresponding third running parameter,
The third running parameter is optimal running parameter under setting target.
Further, further includes:
According to the hypothesis running parameter of setting, the hypothesis running parameter is brought into the coefficient matrix, is obtained optimal
Show parameter.
Using one or more technical solution in the present invention, have the following beneficial effects:
Due to the technical solution using the application, a system can be obtained by training study in original sample parameter data
Matrix number, the enough relational models between running parameter and corresponding performance parameter of the coefficient matrix, and then can be by this
Matrix number Optimization Work parameter or prediction target show parameter, realize effective analysis to technological parameter.
Detailed description of the invention
Fig. 1 shows for the step process that the technological parameter based on recurrence learning swims lake analysis method a kind of in the embodiment of the present invention
It is intended to;
Fig. 2 is the table schematic diagram of sample data in the embodiment of the present invention;
Fig. 3 is that data primary visualizes schematic diagram in the embodiment of the present invention;
Fig. 4 is that the distance between sample data and 5 kinds of cluster centers of gravity are drawn in the embodiment of the present invention;
Fig. 5 is to have 5 running parameters and output power to have ratio according to what 0.3 threshold value marked out in the embodiment of the present invention
Biggish associated table schematic diagram;
Fig. 6 is the prediction error of linear regression returns after calculating in the embodiment of the present invention coefficient matrix and coefficient matrix
Analyze result figure;
Fig. 7 is the Q-Q figure of result obtained in the embodiment of the present invention and prediction result.
Specific embodiment
The embodiment of the present invention solves existing by providing a kind of process parameter optimizing analysis method based on recurrence learning
Can not effectively be provided in technology it is a kind of can by the method for opening relationships between each parameter of equipment, and then can not to engineering into
The technical issues of row is effectively estimated.
In order to solve the above-mentioned technical problem, in conjunction with appended figures and specific embodiments to of the invention
Technical solution is described in detail.
The process parameter optimizing analysis method based on recurrence learning that the embodiment of the invention provides a kind of, as shown in Figure 1, packet
It includes: S1 and classification storage is carried out according to data mode to technological parameter, technological parameter includes running parameter and corresponding performance
Parameter;S2 chooses the technological parameter in a kind of data mode, calculates its running parameter and shows the cross-correlation square between parameter
Battle array obtains the running parameter that relevance is greater than preset value;The running parameter that relevance is greater than preset value is extracted two groups, made by S3
To train the first running parameter used and as the second running parameter of verifying, the performance parameter verified without relevance is taken out
Two groups are taken, the first performance parameter used as training and the second performance parameter as verifying;S4, using linear regression side
Method, the data that the data of the first running parameter are showed parameter with first solve as training sequence and obtain association described first
The initial coefficients matrix of running parameter and the first performance parameter;S5 brings the data of the second running parameter into coefficient matrix, solves
The data for obtaining prediction performance parameter calculate the root mean square between the data of prediction performance parameter and the data of the second performance parameter
Error obtains verified coefficient matrix when root-mean-square error is close to 0.
In S1, technological parameter carries out classification storage according to data mode, and data mode here is for different operating
Technological parameter under state, in actual conditions, collected data can not determine be equipment which kind of working condition, therefore, this
In be using cluster method data are sorted out.The data under different working condition are obtained, in this way, making analytic process more
Add clear.
In a particular embodiment, with the running parameter of certain vacuum power device and corresponding performance parameter (output work
Rate) between relationship analysis for.
The output power of the device is response, other have 18 parameters to be in working condition parameter, and total data sample number is
5000, as shown in Figure 2.
Before S1, structuring arrangement is carried out to the data of technological parameter, and improper data is rejected using statistical method,
Specifically, using the pandas packet of Python to the format error data in data, null value etc. is rejected, obtained data
It is recorded and stored with csv file format, after simple visual analyzing, discovery data are likely that there are a variety of working conditions and deposit
The phenomenon that, as shown in Figure 3.
Then, in S1, using clustering, above-mentioned data are divided into 5 kinds of working conditions, according to the mark after cluster
Note, data are classified and are exported, as shown in figure 4, drawing for the distance between sample data and 5 kinds of cluster centers of gravity after cluster
Figure, it is seen that the non-cluster degree of data is very high.
According to the description in S2, the data for selecting second of cluster state therein to be marked carry out next analysis work
Make, firstly, calculating 18 cross-correlation matrixs between parameter and output power therein, specifically, between two parameters mutually
Independent, then its cross-correlation coefficient is 0, perfectly correlated between two parameters, then cross-correlation coefficient is ± 1, and sets cross correlation
Number threshold values are 0.3, according to above-mentioned standard, as shown in figure 5, have 5 running parameters and output power have it is bigger be associated with,
It is identified, then carries out subsequent analysis.
Then, in S3, two groups are extracted in 5 running parameters of above-mentioned acquisition, one group of first work as training
Make parameter, one group of second running parameter as verifying, two groups of performance parameter extraction will verified without relevance, one group of work
For the first performance parameter of training, one group of second performance parameter as verifying.Then, in S4, using linear regression
Method, specifically can be using least square method, ridge regression method, the Lasso Return Law etc., by the data of the first running parameter with
The data of first performance parameter solve the coefficient square for obtaining the first running parameter of association and the first performance parameter as training sequence
Battle array.The coefficient matrix obtained at this time is to the first running parameter and the first performance parameter establishing associated initial coefficients square
Battle array.
The initial coefficients matrix may not be very accurately, therefore, in S5, to bring the data of the second running parameter into this
Initial coefficients matrix solves the data for obtaining prediction performance parameter, calculates the data and the second performance parameter of prediction performance parameter
Data between root-mean-square error obtain verified coefficient matrix when the root-mean-square error is close to 0.
In practical situations, it is sought most of the time using linear regression method function F (), in the present invention, specifically
It is to describe in this way:
Wherein, ω ∈ R is coefficient matrix, and X is running parameter, and y is corresponding performance parameter.
It, can also be by repeating S1~S5 after S5, continuous iteration obtains more accurate coefficient matrix.It is by finding
Matrix number, and learning model is assessed, the Q-Q figure of obtained result and prediction result is as shown in fig. 7, it can be seen that training knot
Fruit is more accurately.Fig. 6 is specifically the prediction error analysis for returning the coefficient matrix and the coefficient matrix that return after calculating
Result figure.
The coefficient matrix is exactly more accurate as a result, i.e. training pattern training is completed, and certainly, if subsequent, there are also new
Supplemental characteristic, above-mentioned S1~S5 can also be brought into and be trained, thus constantly obtain adapt to current data coefficient matrix.
Obtain adapt to parameter current data coefficient matrix after, the coefficient matrix specifically there are two types of application, one is
After can wishing the target reached performance parameter determining one, target performance parameter is brought into the coefficient matrix, acquisition pair
The third running parameter answered, the third running parameter is exactly to set optimal running parameter under target at this time, in this way,
It can be by setting optimal running parameter, to obtain target performance parameter.Another kind is can to join according to the hypothesis work of setting
Number, which is brought into the coefficient matrix, so that optimal performance parameter is obtained, at this point, the optimal performance parameter
It is exactly the optimal response exported according to the running parameter of hypothesis.
Therefore, it is enabled in engineering technology using the coefficient matrix that the program obtains, according to the wish of operator
It realizes optimal output, or is exported according to the target of operator, operation obtains optimal input parameter.So that in the engineering technology
It can be effectively controlled, realize and optimize.
Although preferred embodiments of the present invention have been described, it is created once a person skilled in the art knows basic
Property concept, then additional changes and modifications may be made to these embodiments.So it includes excellent that the following claims are intended to be interpreted as
It selects embodiment and falls into all change and modification of the scope of the invention.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
Mind and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to include these modifications and variations.
Claims (5)
1. a kind of process parameter optimizing analysis method based on recurrence learning, comprising: S1, to technological parameter according to data mode into
Row classification storage, the technological parameter include running parameter and corresponding performance parameter;
S2 chooses the technological parameter in a kind of data mode, calculates its running parameter and shows the cross-correlation matrix between parameter,
Obtain the running parameter that relevance is greater than preset value;It is characterized by further comprising:
The running parameter that relevance is greater than preset value is extracted two groups by S3, as the first running parameter for using of training and as testing
Second running parameter of card, two groups of performance parameter extraction will verified without relevance, the first performance ginseng as training
It counts and shows parameter as the second of verifying;
S4, using linear regression method, using the data of the data of the first running parameter and the first performance parameter as training sequence,
Solve the initial coefficients matrix for obtaining and being associated with first running parameter and the first performance parameter;
S5 brings the data of the second running parameter into the initial coefficients matrix, solves the data for obtaining prediction performance parameter, meter
The root-mean-square error between the data of the prediction performance parameter and the data of the second performance parameter is calculated, in the root-mean-square error
When close to 0, verified coefficient matrix is obtained.
2. the process parameter optimizing analysis method according to claim 1 based on recurrence learning, which is characterized in that S5 it
Afterwards, further includes:
S1 ~ S5 is repeated, continuous iteration obtains more accurate coefficient matrix.
3. the process parameter optimizing analysis method according to claim 1 or 2 based on recurrence learning, which is characterized in that
Before S1, further includes:
Structuring arrangement is carried out to the data of technological parameter, and improper data is rejected using statistical method.
4. the process parameter optimizing analysis method according to claim 1 based on recurrence learning, which is characterized in that also wrap
It includes:
It determines the target performance parameter for wishing to reach, brings into the coefficient matrix, obtain corresponding third running parameter, it is described
Third running parameter is optimal running parameter under setting target.
5. the process parameter optimizing analysis method according to claim 1 based on recurrence learning, which is characterized in that also wrap
It includes:
According to the hypothesis running parameter of setting, the hypothesis running parameter is brought into the coefficient matrix, optimal performance is obtained
Parameter.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810550965.6A CN108646688B (en) | 2018-05-31 | 2018-05-31 | A kind of process parameter optimizing analysis method based on recurrence learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810550965.6A CN108646688B (en) | 2018-05-31 | 2018-05-31 | A kind of process parameter optimizing analysis method based on recurrence learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108646688A CN108646688A (en) | 2018-10-12 |
CN108646688B true CN108646688B (en) | 2019-05-07 |
Family
ID=63758886
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810550965.6A Active CN108646688B (en) | 2018-05-31 | 2018-05-31 | A kind of process parameter optimizing analysis method based on recurrence learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108646688B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110244670A (en) * | 2019-05-31 | 2019-09-17 | 山东中烟工业有限责任公司 | A kind of volume packet procedures technical parameter management-control method and system |
CN110516355B (en) * | 2019-08-27 | 2020-06-12 | 上海交通大学 | Leakage test curve trend prediction method and system based on pre-estimated function model |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102222166A (en) * | 2011-06-08 | 2011-10-19 | 北京交通大学 | Safety prediction method of pantograph-catenary relationship of motor train unit |
CN103473438A (en) * | 2013-08-15 | 2013-12-25 | 国家电网公司 | Method for optimizing and correcting wind power prediction models |
CN103678869A (en) * | 2013-09-17 | 2014-03-26 | 中国人民解放军海军航空工程学院青岛校区 | Prediction and estimation method of flight parameter missing data |
CN104699991A (en) * | 2015-03-30 | 2015-06-10 | 哈尔滨工业大学 | Grey system theory-based method of predicting annual heat supply of urban heating system |
CN104881582A (en) * | 2015-05-28 | 2015-09-02 | 象辑知源(武汉)科技有限公司 | Air quality prediction method and device |
CN105046057A (en) * | 2015-06-24 | 2015-11-11 | 上海大学 | LSSVM (Least Square Support Vector Machine) pulsation wind speed prediction method based on Morlet wavelet kernel |
KR20170017284A (en) * | 2015-08-06 | 2017-02-15 | 광주과학기술원 | Method for identifying of cancer-driver genes based on a partial covariance selection |
CN106682053A (en) * | 2015-11-10 | 2017-05-17 | 网易(杭州)网络有限公司 | Optimization method and device for putting scheme of game top-up sales promotion |
-
2018
- 2018-05-31 CN CN201810550965.6A patent/CN108646688B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102222166A (en) * | 2011-06-08 | 2011-10-19 | 北京交通大学 | Safety prediction method of pantograph-catenary relationship of motor train unit |
CN103473438A (en) * | 2013-08-15 | 2013-12-25 | 国家电网公司 | Method for optimizing and correcting wind power prediction models |
CN103678869A (en) * | 2013-09-17 | 2014-03-26 | 中国人民解放军海军航空工程学院青岛校区 | Prediction and estimation method of flight parameter missing data |
CN104699991A (en) * | 2015-03-30 | 2015-06-10 | 哈尔滨工业大学 | Grey system theory-based method of predicting annual heat supply of urban heating system |
CN104881582A (en) * | 2015-05-28 | 2015-09-02 | 象辑知源(武汉)科技有限公司 | Air quality prediction method and device |
CN105046057A (en) * | 2015-06-24 | 2015-11-11 | 上海大学 | LSSVM (Least Square Support Vector Machine) pulsation wind speed prediction method based on Morlet wavelet kernel |
KR20170017284A (en) * | 2015-08-06 | 2017-02-15 | 광주과학기술원 | Method for identifying of cancer-driver genes based on a partial covariance selection |
CN106682053A (en) * | 2015-11-10 | 2017-05-17 | 网易(杭州)网络有限公司 | Optimization method and device for putting scheme of game top-up sales promotion |
Also Published As
Publication number | Publication date |
---|---|
CN108646688A (en) | 2018-10-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107038167A (en) | Big data excavating analysis system and its analysis method based on model evaluation | |
CN107220977B (en) | The image partition method of Validity Index based on fuzzy clustering | |
CN104200114B (en) | Flow cytometry data rapid analysis method | |
CN105868534B (en) | A kind of hydrological model Uncertainty Analysis Method based on multiple-objection optimization sampling | |
CN109597968A (en) | Paste solder printing Performance Influence Factor analysis method based on SMT big data | |
CN108646688B (en) | A kind of process parameter optimizing analysis method based on recurrence learning | |
CN107132266A (en) | A kind of Classification of water Qualities method and system based on random forest | |
CN113422695B (en) | Optimization method for improving robustness of topological structure of Internet of things | |
CN110956004B (en) | Concrete penetration depth prediction algorithm with fusion of multi-empirical formula and BP neural network model | |
CN105117550A (en) | Product multidimensional correlation-oriented degradation failure modeling method | |
CN103675610A (en) | Method for extracting characteristic factors in online local discharge detection | |
CN103902798B (en) | Data preprocessing method | |
CN114114039A (en) | Method and device for evaluating consistency of single battery cells of battery system | |
CN116423003B (en) | Tin soldering intelligent evaluation method and system based on data mining | |
CN110990784A (en) | Cigarette ventilation rate prediction method based on gradient lifting regression tree | |
CN108830417A (en) | A kind of residential energy consumption prediction technique and system based on ARMA and regression analysis | |
CN110544047A (en) | Bad data identification method | |
CN108055227B (en) | WAF unknown attack defense method based on site self-learning | |
CN115311452A (en) | Precise glue filling evaluation method and system based on machine vision | |
CN107463532A (en) | A kind of mass analysis method of electric power statistics | |
CN104715160B (en) | Soft sensor modeling data exception point detecting method based on KMDB | |
CN114330486A (en) | Power system bad data identification method based on improved Wasserstein GAN | |
CN109685823A (en) | A kind of method for tracking target based on depth forest | |
CN111680740A (en) | Neural network training method and device and electrical load distinguishing method and device | |
CN105654106A (en) | Decision tree generation method and system thereof |
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 |