CN110287456A - Large coil rolling surface defect analysis method based on data mining - Google Patents
Large coil rolling surface defect analysis method based on data mining Download PDFInfo
- Publication number
- CN110287456A CN110287456A CN201910582711.7A CN201910582711A CN110287456A CN 110287456 A CN110287456 A CN 110287456A CN 201910582711 A CN201910582711 A CN 201910582711A CN 110287456 A CN110287456 A CN 110287456A
- Authority
- CN
- China
- Prior art keywords
- group
- data
- defect
- analysis
- surface defect
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000007547 defect Effects 0.000 title claims abstract description 125
- 238000004458 analytical method Methods 0.000 title claims abstract description 55
- 238000005096 rolling process Methods 0.000 title claims abstract description 50
- 238000007418 data mining Methods 0.000 title claims abstract description 25
- 238000000034 method Methods 0.000 claims abstract description 40
- 238000007477 logistic regression Methods 0.000 claims abstract description 23
- 230000008569 process Effects 0.000 claims abstract description 22
- 238000004519 manufacturing process Methods 0.000 claims abstract description 13
- 238000004886 process control Methods 0.000 claims abstract description 10
- 238000005457 optimization Methods 0.000 claims abstract description 8
- 238000013480 data collection Methods 0.000 claims abstract description 5
- 229910000831 Steel Inorganic materials 0.000 claims description 33
- 239000010959 steel Substances 0.000 claims description 33
- 230000002950 deficient Effects 0.000 claims description 27
- 239000000047 product Substances 0.000 claims description 25
- 238000012797 qualification Methods 0.000 claims description 9
- 206010039509 Scab Diseases 0.000 claims description 8
- 208000037656 Respiratory Sounds Diseases 0.000 claims description 7
- 230000002159 abnormal effect Effects 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 6
- 239000006227 byproduct Substances 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 5
- 238000004804 winding Methods 0.000 claims description 5
- 238000010438 heat treatment Methods 0.000 claims description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 4
- 238000004513 sizing Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 238000007689 inspection Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000001543 one-way ANOVA Methods 0.000 claims description 2
- 238000003908 quality control method Methods 0.000 abstract description 5
- 238000007405 data analysis Methods 0.000 abstract description 4
- 238000007781 pre-processing Methods 0.000 abstract 1
- 230000007812 deficiency Effects 0.000 description 5
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 4
- 238000012216 screening Methods 0.000 description 3
- 238000002791 soaking Methods 0.000 description 3
- 238000010273 cold forging Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 229910052742 iron Inorganic materials 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Mathematics (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Physics (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Operations Research (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Factory Administration (AREA)
Abstract
The invention provides a large coil rolling surface defect analysis method based on data mining, which comprises the following steps of: 1) and data collection and sorting: collecting rolling process data and surface defect judgment data of a large coil product; 2) data preprocessing; 3) and system clustering: classifying the defect groups by adopting a system clustering algorithm; 4) interpretation of clustering results; 5) and logistic regression analysis: establishing a regression model for the clustering result based on a multi-classification logistic regression analysis method; 6) and analyzing results: analyzing key process factors causing each defect group of the large coil product according to the model result; 7) optimization of process control: according to the analysis result, the process control is optimized, the defect judgment proportion is reduced, and the product percent of pass is improved. The method can predict the possible defects, enhance the surface quality control of the large coiled product and improve the delivery quality of the product. The invention has great practical value in actual production based on data analysis.
Description
Technical field
The invention belongs to the field of quality control of steel products, and in particular to a kind of bulk lots volume rolling based on data mining
Surface deficiency analysis method.
Background technique
The leading products of large coil wire production line are cold-forging steel, require height to product surface quality, surface defect is to cause greatly
Coil the low main factor of cold-forging steel qualification rate.Since bulk lots volume product specification is more, surface defect generate link is more, reason
Complexity, in the actual production process, it is difficult to accurately analyze and position the producing cause of certain defect.Although Bang Xian factory is on surface
Certain experience and method are had accumulated in terms of the analysis and Control of defect, but this method using artificial judgment is carrying out defect
It is more demanding to the technical level of worker when analysis, it is built upon worker and controls Product Process and quality very familiar base
On plinth, mainly according to past operating experience, the accuracy of Resolving probiems and spend the time uncertain, and to not yet running into
Surface deficiency analysis in can it is helpless, affect the production efficiency of enterprise to a certain extent.
In recent years, popularizing with data mining technology provides the foundation for steel products quality analysis control.Data
Digging technology has been succeeded application in external certain iron companies, and domestic major iron and steel enterprise also takes up data and digs
The application in steel products quality analysis is dug, the research of product quality has been carried out using big data analysis and controlled, some
Achieve certain achievement.In terms of surface deficiency analysis, there is scholar successively to utilize data digging method to hot-rolled coil and cold
The surface quality defect for rolling product carried out analysis.But surface is being carried out to gren rod and bulk lots volume using data mining
In terms of defect analysis, domestic related patents and document not yet were introduced, and did not propose detailed method.
Summary of the invention
In view of this, the invention proposes a kind of bulk lots volume rolled surface defect analysis method based on data mining, purport
In the method analyzed by using data, data and operation of rolling data, which carry out data, to be determined to the surface defect of bulk lots volume product
It excavates and analyzes, the critical process factor found out the rolled surface defect distribution of bulk lots volume product and it is had a major impact,
Clearly to reduce and eliminating the control measure that defect should be taken, as the auxiliary of artificial defect analysis, help to reinforce to big
Coil the quality control on the surface of product.
To achieve the above object, the invention provides the following technical scheme:
The embodiment of the present application discloses a kind of bulk lots volume rolled surface defect analysis method based on data mining, feature
It is, step includes:
1), data collection and arrangement: the data and surface defect for collecting the bulk lots volume operation of rolling determine data, while with big
It is key variables two groups of data sources of series connection that lot number is rolled in coiling, rejects the batch containing abnormal data, calculates the ratio of all kinds of surface defects
Example, screens the initial data of analysis and modeling；
2) it, data prediction: is grouped by product specification, the classified variable in the data of the bulk lots volume operation of rolling is assigned
Value, is standardized continuous variable；
3), Hierarchical Clustering: being divided into zero defect group and defective group for pretreated data, then to defective group of use
System clustering algorithm is classified to sample as target variable using the major defect filtered out, and is examined to cluster result
It tests, determines cluster optimal solution in conjunction with Cluster tendency, cluster result collection is exported according to cluster optimal solution；
4), the explanation of cluster result: being analyzed and explained to the data after Hierarchical Clustering, and carry out packet marking, will
Zero defect group echo is A group, and defective group after cluster is successively labeled as B group, C group, D in alphabetical order respectively
Group ...；
5), logistic regression analysis: being grouped with specification, Multinomial Logistic Regression method is based on, with intact
Falling into group is that A group is established logistic using successive Regression mode and returned for reference group to defective group of B group, C group, D group ...
Model；
6), interpretation of result: according to logistic regression model as a result, analysis causes the key of each defect group of bulk lots volume product
Process factors, clearly to reduce and eliminate the control measure that defect should be taken；
7), the optimization of process control: based on the analysis results, specific aim measure is taken, the control of optimization process parameter is reduced
Bulk lots volume surface defect sentences time ratio, promotes product qualification rate.
Further, the data of the bulk lots volume operation of rolling are from heating, de-scaling, the number for being rolled down to collection volume each process
According to including at least workshop, teams and groups, heating temperature, time inside furnace, high pressure water dephosphorization pressure, start rolling temperature, entering to subtract sizing temperature
Degree, winding temperature, each section roll rear size, mill speed, each section of groove steel transportation amount, the surface defect determines that data include at least
Total decision content and folding, crackle, pit, scab at scribing line.
Further, in the bulk lots volume rolled surface defect analysis method above-mentioned based on data mining, in step 1)
The abnormal data be the data for not meeting processing range requirement and working specification.
Further, in the bulk lots volume rolled surface defect analysis method above-mentioned based on data mining,
Data normalization processing in step 2) is handled using Z-score, and conversion formula is as follows::
Wherein, x* is certain variable treated value, and x is the variable original value, and μ is the mean value of all data of the variable, and σ is
The standard deviation of all data of the variable.
Further, in the bulk lots volume rolled surface defect analysis method above-mentioned based on data mining, in step 3)
The zero defect group be without any surface defect, i.e., total surface defective proportion be zero batch composed by set, have scarce
Falling into group is at least containing a kind of surface defect, i.e. set composed by batch of the total surface defective proportion greater than zero.
Further, in the bulk lots volume rolled surface defect analysis method above-mentioned based on data mining, the step
3) one-way analysis of variance method is used to the inspection of cluster result in.
Further, in the bulk lots volume rolled surface defect analysis method above-mentioned based on data mining, the step
5) the logistic regression analysis model established in, expression formula are as follows:
……
Wherein, PA、PB、PC、PDThe probability that-A group, B group, C group, D group result occur；
αB、αC、αD- constant term；
X1、X2、…、Xp- defect occurs there is the process variable significantly affected；
β11、β21、β31, β12、β22、β32..., β1p、β2p、β3p- regression coefficient
Compared with the prior art, the advantages of the present invention are as follows: the clustering algorithm in data mining is applied to bulk lots volume
In surface deficiency analysis, classify from the angle of data itself, eliminates artificially to the intervention of data, there is data processing more accurately
Effect；With logistic regression analysis model, the pass having a major impact to bulk lots volume rolled surface defect is grasped in terms of quantitative
The drawbacks of key process factors can be used as the auxiliary of manual analysis defect, avoid empirical overabundance of data, so that analysis result is more
With convincingness；According to model result, the defect that certain batch is likely to occur can be predicted, reduce the outflow of defective product, reinforce
To the quality control on the surface of bulk lots volume product, the quality of finished product is improved.The present invention bases oneself upon data analysis, has in actual production
Very big practical value.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The some embodiments recorded in application, for those of ordinary skill in the art, without creative efforts,
It is also possible to obtain other drawings based on these drawings.
Fig. 1 is the flow chart of data prediction in specifically example of the invention.
Fig. 2 is the flow chart of Hierarchical Clustering in specifically example of the invention.
Fig. 3 is the flow chart of logistic algorithm with regress analysis method in specifically example of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out detailed retouch
It states, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Based on the present invention
In embodiment, those of ordinary skill in the art's every other implementation obtained without making creative work
Example, shall fall within the protection scope of the present invention.
Embodiment one
By taking ML08Al bulk lots volume as an example, join shown in Fig. 1 to Fig. 3, the deep bid rolling based on data mining in the present embodiment
Control surface defect analysis method, step include:
1) every operation of rolling of ML08Al bulk lots volume in certain time, is collected from Rolling Production scene and control system
Data, the surface defect that corresponding batch is collected from MES system determine data, to roll lot number as key variables two groups of numbers of series connection
According to source, reject the batch containing abnormal data, calculate the ratio of all kinds of surface defects, select major defect scribing line, folding, crackle,
Scab, pit etc. is as analysis object, using the data after screening as the initial data of analysis and modeling, altogether containing sample 1098
Item, product specification are Φ 16mm- Φ 38mm, and data mode is shown in Table 1,
ML08Al bulk lots volume rolling defect and process data in certain time of table 1
；
2) it, data prediction: is grouped by product specification, to the steelshop in operation of rolling data, mill, teams and groups
Equal variables carry out assignment, are standardized, are shown in Table 1 to continuous variable；
3), Hierarchical Clustering: by treated, data are divided into zero defect group and defective group, and wherein zero defect group 567, has
Defect group 531, system clustering algorithm then is used to defective group, is crossed, folded, split with the major defect filtered out
Line, scab, the ratio of the five class defects such as pit classifies to sample as target variable, and tests to cluster result,
It determines that cluster optimal solution is 5 groups in conjunction with Cluster tendency, cluster result collection is exported according to cluster optimal solution；
4), the explanation of cluster result: analyzing cluster result and explained, and carries out packet marking, by zero defect group
Labeled as A group, defective group after cluster is respectively labeled as B group, C group, D group, E group, F group, is shown in Table 2:
2 ML08Al bulk lots volume defect cluster result of table is explained
Within this time it can be seen from 2 cluster result of table, ML08Al bulk lots volume major defect type is scribing line and splits
The defects of line, a small amount of batch contains folding, pit and scabs；
5) it, is grouped with specification, is based on Multinomial Logistic Regression method, it is right using zero defect group A group as reference group
Defective group of B group, C group, D group, E group, F group establish logistic regression model using successive Regression mode, larger with output
And organize and be illustrated for the more complete Φ 26mm specification and Φ 30mm specification of class, model result is as follows:
Φ 26mm specification:
Φ 30mm specification:
Variable in formula is the variable after standardization；
6), interpretation of result: according to model result, analysis causes ML08Al bulk lots volume Φ 26mm specification and Φ 30mm specification is each
The critical process factor of defect group mainly has: ruler after bringing-up section furnace temperature, soaking zone furnace temperature, start rolling temperature, winding temperature, roughing
Very little, finish to gauge speed, it is thick in offset equal steel transportation amount, pre- finish rolling is averaged steel transportation amount etc..Influence degree (regression coefficient) is shown in Table 3 and table 4
It is shown:
3 ML08Al bulk lots volume Φ 26mm specification logistic regression coefficient of table
4 ML08Al bulk lots volume Φ 30mm specification logistic regression coefficient of table
As shown in Table 3, for Φ 26mm specification ML08Al bulk lots volume, bringing-up section furnace temperature, start rolling temperature are to each defect group
Regression coefficient is negative value, size after winding temperature, roughing, finish to gauge speed, it is thick in offset equal steel transportation amount, pre- finish rolling and averagely cross steel
Amount is positive value to the regression coefficient of each defect group.In negative value, bringing-up section furnace temperature is larger to the coefficient of D group, to reduce defect
Occur, Ying Tigao bringing-up section furnace temperature；In positive value, it is thick in offset equal steel transportation amount larger to the coefficient of C group and F group, for reduction defect
Generation, should reduce in thick and offset equal steel transportation amount；
As shown in Table 4, for Φ 30mm specification ML08Al bulk lots volume, bringing-up section furnace temperature is to D group and F group, start rolling temperature to B
Group and C group, it is thick in offset equal steel transportation amount be negative value to the regression coefficient of D group, remaining is positive value.In negative value, bringing-up section furnace
Temperature is larger to the coefficient of D group, for the generation for reducing defect, Ying Tigao bringing-up section furnace temperature；In positive value, size is to E group, thick after roughing
In to offset equal steel transportation amount larger to the coefficient of F group to C group and F group, pre- the finish rolling steel transportation amount that is averaged, for the generation for reducing defect, answer
Reduce size after roughing, and reduces to roll in thick and be averaged steel transportation amount with pre- finish rolling；
7) it, the optimization of process control: based on the analysis results, during subsequent Rolling Production, for Φ 26mm specification, improves
Bringing-up section furnace temperature reduces in thick and offsets equal steel transportation amount, for Φ 30mm specification, size after improving bringing-up section furnace temperature, reducing roughing,
It reduces to roll in thick and be averaged steel transportation amount with pre- finish rolling；
By optimizing process control, the surface defect for reducing ML08Al bulk lots volume Φ 26mm specification and Φ 30mm specification is sentenced
Secondary ratio, improves qualification rate, makes the qualification rate of Φ 26mm specification and Φ 30mm specification respectively by 80.7% original He
78.2% has been increased to 91.6% and 88.3%.
Embodiment two
SWRCH35K bulk lots volume rolled surface defect is analyzed based on data mining, comprising:
1), data collection and arrangement: from collecting in Rolling Production scene and control system, SWRCH35K in certain time is big
Every operation of rolling data of coiling, the surface defect that corresponding batch is collected from MES system determine data, using roll lot number as
Key variables two groups of data sources of series connection, reject the batch containing abnormal data, calculate the ratio of all kinds of surface defects, select main lack
Scribing line is fallen into, crackle, folds, scab etc. as analysis object, using the data after screening as the initial data of analysis and modeling, is total to
Containing 965, sample, product specification is Φ 16mm- Φ 38mm, and data mode is shown in Table 5,
SWRCH35K bulk lots volume rolling defect and process data in certain time of table 5
；2) it, data prediction: is grouped by product specification, to the steelshop in operation of rolling data, mill, class
The variables such as group carry out assignment, are standardized, are shown in Table 5 to continuous variable；
3), Hierarchical Clustering: by treated, data are divided into zero defect group and defective group, and wherein zero defect group 485, has
Defect group 480, system clustering algorithm then is used to defective group, is crossed with the major defect filtered out, crackle, folding
The ratio of the four class defects such as folding, scab classifies to sample as target variable, and tests to cluster result, in conjunction with poly-
Class pedigree chart determines that cluster optimal solution is 5 groups, exports cluster result collection according to cluster optimal solution；
4), the explanation of cluster result: analyzing cluster result and explained, and carries out packet marking, by zero defect group
Labeled as A group, defective group after cluster is respectively labeled as B group, C group, D group, E group, F group, is shown in Table 6:
6 SWRCH35K bulk lots volume defect cluster result of table is explained
Within this time it can be seen from 6 cluster result of table, SWRCH35K bulk lots volume major defect type is scribing line and splits
The defects of line, a small amount of batch is containing folding and scabbing；
5), logistic regression analysis: being grouped with specification, Multinomial Logistic Regression method is based on, with intact
Falling into group A group is reference group, and defective group of B group, C group, D group, E group, F group are established logistic using successive Regression mode and returned
Model, larger by output and be illustrated for organizing the more complete Φ 20mm specification and Φ 28mm specification of class, model result is as follows:
Φ 20mm specification:
Φ 28mm specification:
Variable in formula is the variable after standardization；
6), interpretation of result: according to model result, analysis causes SWRCH35K bulk lots volume Φ 20mm specification and Φ 28mm specification
The critical process factor of each defect group mainly has: bringing-up section furnace temperature, start rolling temperature enter to subtract sizing temperature, winding temperature, in furnace
Between, size after roughing, high pressure water dephosphorization pressure, it is thick in offset equal steel transportation amount, pre- finish rolling is averaged steel transportation amount etc..Influence degree (is returned
Return coefficient) it is shown in Table shown in 7 and table 8:
7 SWRCH35K bulk lots volume Φ 20mm specification logistic regression coefficient of table
8 SWRCH35K bulk lots volume Φ 28mm specification logistic regression coefficient of table
As shown in Table 7, for Φ 20mm specification SWRCH35K bulk lots volume, start rolling temperature and time inside furnace are to B group, bringing-up section
Furnace temperature and start rolling temperature are negative value to the regression coefficient of D group, remaining is positive value.In negative value, coefficient of the start rolling temperature to D group
It is larger, for the generation for reducing defect, Ying Tigao start rolling temperature；In positive value, after roughing size and it is thick in offset equal steel transportation amount difference
It is larger to the coefficient of E group and C group, for the generation for reducing defect, should reduce after roughing size and it is thick in offset equal steel transportation amount；
As shown in Table 8, for Φ 28mm specification SWRCH35K bulk lots volume, bringing-up section furnace temperature is to B group and D group, start rolling temperature
It is negative value to the regression coefficient of F group to D group, high pressure water dephosphorization pressure, remaining is positive value.In negative value, bringing-up section furnace temperature is to D
The coefficient of group is larger, for the generation for reducing defect, Ying Tigao bringing-up section furnace temperature；In positive value, after roughing size to E group, it is thick in roll
Average steel transportation amount is larger to the coefficient of C group, for the generation for reducing defect, should reduce after roughing size and it is thick in offset steel
Amount；
7) it, the optimization of process control: based on the analysis results, during subsequent Rolling Production, for Φ 20mm specification, improves
Start rolling temperature, reduce after roughing size and it is thick in offset equal steel transportation amount, for Φ 28mm specification, improve bringing-up section furnace temperature, reduce
After roughing size and it is thick in offset equal steel transportation amount；
By optimizing process control, reduce the surface defect of SWRCH35K bulk lots volume Φ 20mm specification and Φ 28mm specification
Sentence time ratio, improves qualification rate, make the qualification rate of Φ 20mm specification and Φ 28mm specification respectively by 78.2% original He
76.5% has been increased to 87.4% and 84.8%.
Embodiment three
51B20 bulk lots volume rolled surface defect is analyzed based on data mining, comprising:
1), data collection and arrangement: from collecting 51B20 bulk lots volume in certain time in Rolling Production scene and control system
Every operation of rolling data, the surface defect that corresponding batch is collected from MES system determines data, to roll lot number as crucial
Variable two groups of data sources of series connection, reject the batch containing abnormal data, calculate the ratio of all kinds of surface defects, select major defect and draw
Line, crackle, pit, folding etc. are as analysis object, using the data after screening as the initial data of analysis and modeling, contain sample altogether
This 713, product specification is Φ 20mm- Φ 34mm, and data mode is shown in Table 9；
51B20 bulk lots volume rolling defect and process data in certain time of table 9
；
2) it, data prediction: is grouped by product specification, to the steelshop in operation of rolling data, mill, teams and groups
Equal variables carry out assignment, are standardized, are shown in Table 9 to continuous variable；
3), Hierarchical Clustering: by treated, data are divided into zero defect group and defective group, and wherein zero defect group 351, has
Defect group 362, system clustering algorithm then is used to defective group, is crossed with the major defect filtered out, is crackle, recessed
The ratio of the four class defects such as hole, folding classifies to sample as target variable, and tests to cluster result, in conjunction with poly-
Class pedigree chart determines that cluster optimal solution is 5 groups, exports cluster result collection according to cluster optimal solution；
4), the explanation of cluster result: analyzing cluster result and explained, and carries out packet marking, by zero defect group
Labeled as A group, defective group after cluster is respectively labeled as B group, C group, D group, E group, F group, is shown in Table 10:
10 51B20 bulk lots volume defect cluster result of table is explained
Within this time it can be seen from 10 cluster result of table, 51B20 bulk lots volume major defect type is scribing line and splits
The defects of line, a small amount of batch contains pit and folds；
5), logistic regression analysis: being grouped with specification, Multinomial Logistic Regression method is based on, with intact
Falling into group A group is reference group, and defective group of B group, C group, D group, E group, F group are established logistic using successive Regression mode and returned
Model, larger by output and be illustrated for organizing the more complete Φ 24mm specification of class, model result is as follows:
Variable in formula is the variable after standardization；
6), interpretation of result: according to model result, analysis causes the key of each defect group of 51B20 bulk lots volume Φ 24mm specification
Process factors mainly have: size after bringing-up section furnace temperature, soaking zone furnace temperature, start rolling temperature, roughing, it is thick in offset equal steel transportation amount, pre-
Finish rolling is averaged steel transportation amount etc..Influence degree (regression coefficient) is shown in Table 11:
11 51B20 bulk lots volume Φ 24mm specification logistic regression coefficient of table
As shown in Table 11, for Φ 24mm specification 51B20 bulk lots volume, bringing-up section furnace temperature is to D group and F group, soaking zone furnace temperature
It is negative value to the regression coefficient of D group to C group, start rolling temperature, remaining is positive value.In negative value, bringing-up section furnace temperature is to D group
Number is larger, for the generation for reducing defect, Ying Tigao bringing-up section furnace temperature；In positive value, it is thick in offset equal steel transportation amount to the coefficient of C group compared with
Greatly, it is the generation for reducing defect, should reduces in thick and offset equal steel transportation amount；
7) it, the optimization of process control: based on the analysis results, during subsequent Rolling Production, for Φ 24mm specification, improves
Equal steel transportation amount etc. is offseted during bringing-up section furnace temperature, reduction are thick；
By optimizing process control, the surface defect for reducing 51B20 bulk lots volume Φ 24mm specification is sentenced time ratio, is improved
Qualification rate makes the qualification rate of Φ 24mm specification 51B20 be increased to 85.6% by original 74.3%.
In conclusion the clustering algorithm in data mining is applied in the surface deficiency analysis of bulk lots volume by the present invention, from
The angle of data itself is classified, and is eliminated artificially to the intervention of data, is made data processing more accurate and effective；It is returned with logistic
Return analysis model, the critical process factor having a major impact to bulk lots volume rolled surface defect is grasped in terms of quantitative, can be used as
The auxiliary of manual analysis defect, the drawbacks of avoiding empirical overabundance of data, so that analysis result more has convincingness；According to mould
Type reduces the outflow of defective product as a result, the defect that certain batch is likely to occur can be predicted, reinforces the surface to bulk lots volume product
Quality control, improves the quality of finished product.The present invention bases oneself upon data analysis, there is very big practical value in actual production.
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including the element.
The above is only the specific embodiment of the application, it is noted that for the ordinary skill people of the art
For member, under the premise of not departing from the application principle, several improvements and modifications can also be made, these improvements and modifications are also answered
It is considered as the protection scope of the application.
Claims (7)
1. a kind of bulk lots volume rolled surface defect analysis method based on data mining, which is characterized in that step includes:
1), data collection and arrangement: the data and surface defect for collecting the bulk lots volume operation of rolling determine data, while with bulk lots volume
Rolling lot number is key variables two groups of data sources of series connection, rejects the batch containing abnormal data, calculates the ratio of all kinds of surface defects, sieves
Select the initial data of analysis and modeling；
2), data prediction: being grouped by product specification, carries out assignment to the classified variable in the data of the bulk lots volume operation of rolling,
Continuous variable is standardized；
3), Hierarchical Clustering: being divided into zero defect group and defective group for pretreated data, then uses system to defective group
Clustering algorithm classifies to sample as target variable using the major defect filtered out, and tests to cluster result, knot
It closes Cluster tendency and determines cluster optimal solution, cluster result collection is exported according to cluster optimal solution；
4), the explanation of cluster result: being analyzed and explained to the data after Hierarchical Clustering, and carry out packet marking, will be intact
Falling into group echo is A group, and defective group after cluster is successively labeled as B group, C group, D group ... in alphabetical order respectively；
5), logistic regression analysis: being grouped with specification, Multinomial Logistic Regression method is based on, with zero defect group
I.e. A group is that reference group establishes logistic regression model using successive Regression mode to defective group of B group, C group, D group ...；
6), interpretation of result: according to logistic regression model as a result, analysis causes the critical process of each defect group of bulk lots volume product
Factor, clearly to reduce and eliminate the control measure that defect should be taken；
7), the optimization of process control: based on the analysis results, specific aim measure is taken, the control of optimization process parameter reduces deep bid
Volume surface defect sentences time ratio, promotes product qualification rate.
2. the bulk lots volume rolled surface defect analysis method according to claim 1 based on data mining, it is characterised in that:
The data of the bulk lots volume operation of rolling are to include at least production vehicle from heating, de-scaling, the data for being rolled down to collection volume each process
Between, teams and groups, heating temperature, time inside furnace, high pressure water dephosphorization pressure, start rolling temperature, enter to subtract sizing temperature, winding temperature, each section
Roll rear size, mill speed, each section of groove steel transportation amount, the surface defect determine data include at least total decision content and folding,
Scribing line, pit, scabs at crackle.
3. the bulk lots volume rolled surface defect analysis method according to claim 1 based on data mining, it is characterised in that:
The abnormal data in step 1) is the data for not meeting processing range requirement and working specification.
4. the bulk lots volume rolled surface defect analysis method according to claim 1 based on data mining, it is characterised in that:
Data normalization processing in step 2) is handled using Z-score, and conversion formula is as follows:
Wherein, x* be certain variable treated value, x be the variable original value, μ be all data of the variable mean value, σ be the change
Measure the standard deviation of all data.
5. the bulk lots volume rolled surface defect analysis method according to claim 1 based on data mining, which is characterized in that
The zero defect group in step 3) is to collect composed by the batch that is, total surface defective proportion is zero without any surface defect
It closes, defective group is at least containing a kind of surface defect, i.e. set composed by batch of the total surface defective proportion greater than zero.
6. the bulk lots volume rolled surface defect analysis method according to claim 1 based on data mining, which is characterized in that
One-way analysis of variance method is used to the inspection of cluster result in the step 3).
7. the bulk lots volume rolled surface defect analysis method according to claim 1 based on data mining, which is characterized in that
The logistic regression analysis model established in the step 5), expression formula are as follows:
……
Wherein, PA、PB、PC、PDThe probability that-A group, B group, C group, D group result occur；
αB、αC、αD- constant term；
X1、X2、…、Xp- defect occurs there is the process variable significantly affected；
β11、β21、β31, β12、β22、β32..., β1p、β2p、β3p- regression coefficient.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910582711.7A CN110287456A (en) | 2019-06-30 | 2019-06-30 | Large coil rolling surface defect analysis method based on data mining |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910582711.7A CN110287456A (en) | 2019-06-30 | 2019-06-30 | Large coil rolling surface defect analysis method based on data mining |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110287456A true CN110287456A (en) | 2019-09-27 |
Family
ID=68021398
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910582711.7A Pending CN110287456A (en) | 2019-06-30 | 2019-06-30 | Large coil rolling surface defect analysis method based on data mining |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110287456A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111062487A (en) * | 2019-11-28 | 2020-04-24 | 支付宝(杭州)信息技术有限公司 | Machine learning model feature screening method and device based on data privacy protection |
| CN111948994A (en) * | 2020-08-12 | 2020-11-17 | 杭州百子尖科技股份有限公司 | Industrial production line closed-loop automatic quality control method based on data integration and correlation analysis |
| CN112114571A (en) * | 2020-09-24 | 2020-12-22 | 中冶赛迪重庆信息技术有限公司 | Industrial data processing method, system and equipment |
| CN112760561A (en) * | 2020-12-21 | 2021-05-07 | 江苏省沙钢钢铁研究院有限公司 | Wire rod for hand tool and preparation method thereof |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104484828A (en) * | 2014-11-14 | 2015-04-01 | 中国南方电网有限责任公司超高压输电公司检修试验中心 | High-voltage electrical appliance operation quality evaluation method and system based on logistic regression |
| US20150302042A1 (en) * | 2012-11-20 | 2015-10-22 | Hitachi, Ltd. | Data analysis apparatus and data analysis method |
| CN107442576A (en) * | 2016-05-31 | 2017-12-08 | 本钢板材股份有限公司 | A kind of control method of cold-rolled products dimensional accuracy |
| CN107633349A (en) * | 2017-08-28 | 2018-01-26 | 中国西电电气股份有限公司 | Fault impact factor quantitative analysis method based on high-voltage switch gear |
| CN108074025A (en) * | 2016-11-10 | 2018-05-25 | 上海宝信软件股份有限公司 | Coil of strip surface defect determination method based on surface defect distribution characteristics |
-
2019
- 2019-06-30 CN CN201910582711.7A patent/CN110287456A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150302042A1 (en) * | 2012-11-20 | 2015-10-22 | Hitachi, Ltd. | Data analysis apparatus and data analysis method |
| CN104484828A (en) * | 2014-11-14 | 2015-04-01 | 中国南方电网有限责任公司超高压输电公司检修试验中心 | High-voltage electrical appliance operation quality evaluation method and system based on logistic regression |
| CN107442576A (en) * | 2016-05-31 | 2017-12-08 | 本钢板材股份有限公司 | A kind of control method of cold-rolled products dimensional accuracy |
| CN108074025A (en) * | 2016-11-10 | 2018-05-25 | 上海宝信软件股份有限公司 | Coil of strip surface defect determination method based on surface defect distribution characteristics |
| CN107633349A (en) * | 2017-08-28 | 2018-01-26 | 中国西电电气股份有限公司 | Fault impact factor quantitative analysis method based on high-voltage switch gear |
Non-Patent Citations (1)
| Title |
|---|
| 郝阳等: "冶炼关键参数对于IF钢轧材夹杂缺陷的影响", 《工程科学学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111062487A (en) * | 2019-11-28 | 2020-04-24 | 支付宝(杭州)信息技术有限公司 | Machine learning model feature screening method and device based on data privacy protection |
| CN111062487B (en) * | 2019-11-28 | 2021-04-20 | 支付宝(杭州)信息技术有限公司 | Machine learning model feature screening method and device based on data privacy protection |
| CN111948994A (en) * | 2020-08-12 | 2020-11-17 | 杭州百子尖科技股份有限公司 | Industrial production line closed-loop automatic quality control method based on data integration and correlation analysis |
| CN112114571A (en) * | 2020-09-24 | 2020-12-22 | 中冶赛迪重庆信息技术有限公司 | Industrial data processing method, system and equipment |
| CN112114571B (en) * | 2020-09-24 | 2021-11-30 | 中冶赛迪重庆信息技术有限公司 | Industrial data processing method, system and equipment |
| CN112760561A (en) * | 2020-12-21 | 2021-05-07 | 江苏省沙钢钢铁研究院有限公司 | Wire rod for hand tool and preparation method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110287456A (en) | Large coil rolling surface defect analysis method based on data mining | |
| CN104794491B (en) | Based on the fuzzy clustering Surface Defects in Steel Plate detection method presorted | |
| CN104573740B (en) | A kind of equipment fault diagnosis method based on svm classifier model | |
| CN101996405B (en) | Method and device for rapidly detecting and classifying defects of glass image | |
| CN110135486B (en) | Chopstick image classification method based on adaptive convolutional neural network | |
| CN103604809B (en) | A kind of online visible detection method of pattern cloth flaw | |
| CN108595414B (en) | Soil heavy metal enterprise pollution source identification method based on source-sink space variable reasoning | |
| CN112547807B (en) | Decision tree algorithm-based hot rolled strip quality accurate judgment method | |
| Shi et al. | An improved faster R-CNN for steel surface defect detection | |
| CN111325410A (en) | General fault early warning system based on sample distribution and early warning method thereof | |
| CN111369341A (en) | Intelligent risk scoring system for clients before automobile financial loan | |
| CN112231971A (en) | Blast furnace fault diagnosis method based on relative integral trend diffusion fault sample generation | |
| CN115860211A (en) | Casting blank quality prediction method based on local online modeling | |
| CN109636029B (en) | Power distribution network medium-short term voltage out-of-limit early warning method based on big data | |
| CN115034437A (en) | Hot rolled plate convexity prediction method based on improved XGboost | |
| CN109469919A (en) | A kind of stifled grey monitoring method of power station air preheater based on weight cluster | |
| CN112598026B (en) | Continuous annealing production process fault diagnosis method based on unbalanced data deep learning | |
| CN107403391A (en) | A kind of optimal screening method for oil field development pre-arranged project | |
| CN116561692A (en) | Dynamic update real-time measurement data detection method | |
| CN106777237B (en) | A kind of analysis method of surface defect | |
| CN106959668A (en) | A kind of mechanical production devices cutting state identification and data processing method method | |
| Brosas et al. | Adaptive Neuro-Fuzzy Approach for Cacao Bean Grading Classification Process | |
| CN112036403B (en) | Intelligent detection method for missing of bolt pin of power transmission tower based on attention mechanism | |
| CN107516313B (en) | Forging surface defect based on integrated study and Density Clustering is in position detecting method | |
| CN115114968A (en) | Intelligent casting blank crack prediction method based on random forest |
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 |