CN106597898B - A kind of the Floating Production Process control method and system of Behavior-based control portrait - Google Patents
A kind of the Floating Production Process control method and system of Behavior-based control portrait Download PDFInfo
- Publication number
- CN106597898B CN106597898B CN201611168388.1A CN201611168388A CN106597898B CN 106597898 B CN106597898 B CN 106597898B CN 201611168388 A CN201611168388 A CN 201611168388A CN 106597898 B CN106597898 B CN 106597898B
- Authority
- CN
- China
- Prior art keywords
- operation mode
- grade
- data
- ore pulp
- floatation
- 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
- 238000000034 method Methods 0.000 title claims abstract description 93
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 27
- 238000007667 floating Methods 0.000 title claims description 27
- 239000012141 concentrate Substances 0.000 claims abstract description 75
- 238000005188 flotation Methods 0.000 claims abstract description 74
- 230000003542 behavioural effect Effects 0.000 claims abstract description 44
- 238000012706 support-vector machine Methods 0.000 claims abstract description 43
- 238000010276 construction Methods 0.000 claims abstract description 5
- 238000013500 data storage Methods 0.000 claims abstract 2
- 230000006399 behavior Effects 0.000 claims description 63
- 239000003814 drug Substances 0.000 claims description 39
- 238000005259 measurement Methods 0.000 claims description 25
- 238000012549 training Methods 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 10
- 238000003860 storage Methods 0.000 claims description 9
- 229920002472 Starch Polymers 0.000 claims description 8
- 238000012544 monitoring process Methods 0.000 claims description 8
- 235000019698 starch Nutrition 0.000 claims description 8
- 239000008107 starch Substances 0.000 claims description 8
- HEMHJVSKTPXQMS-UHFFFAOYSA-M Sodium hydroxide Chemical compound [OH-].[Na+] HEMHJVSKTPXQMS-UHFFFAOYSA-M 0.000 claims description 7
- 238000009412 basement excavation Methods 0.000 claims description 7
- 230000031068 symbiosis, encompassing mutualism through parasitism Effects 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 5
- 238000013075 data extraction Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 238000003064 k means clustering Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 abstract 2
- 238000005065 mining Methods 0.000 abstract 1
- 230000000087 stabilizing effect Effects 0.000 abstract 1
- 239000000126 substance Substances 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000012797 qualification Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 2
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 238000011144 upstream manufacturing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/04—Programme control other than numerical control, i.e. in sequence controllers or logic controllers
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Paper (AREA)
Abstract
一种基于行为画像的浮选生产过程控制方法及系统，该方法包括浮选过程行为指标数据存储以及预处理；构建浮选过程行为画像标签体系；挖掘矿浆种类与操作模式的对应关系；为不同矿浆种类找出对应的最佳操作方式；构建多层支持向量机分类器对实时的行为画像确定矿浆种类；给每一矿浆种类构建多级支持向量机分类器，进行操作模式调整。该系统包括行为指标选取及预处理单元、画像标签体系构建单元、画像标签挖掘单元、最佳选取单元、矿浆种类分类单元、操作模式调整单元、阈值判定单元。本发明对选矿石性质、加药量、精矿闸门开度、刮板刮尾厚度建立行为画像和关联关系，达到稳质降尾、降低成本、提高、生产效益的目的。
A flotation production process control method and system based on behavioral portrait, the method includes flotation process behavior index data storage and preprocessing; constructing a flotation process behavior portrait label system; excavating the corresponding relationship between pulp types and operation modes; Find the corresponding best operation mode for the type of pulp; build a multi-layer support vector machine classifier to determine the type of pulp from the real-time behavior profile; build a multi-level support vector machine classifier for each type of pulp to adjust the operation mode. The system includes a behavior index selection and preprocessing unit, a portrait label system construction unit, a portrait label mining unit, an optimal selection unit, a pulp type classification unit, an operation mode adjustment unit, and a threshold determination unit. The invention establishes behavioral profiles and correlations on the properties of beneficiation, the dosage of chemicals, the opening of the concentrate gate, and the thickness of the scraper tail, so as to achieve the purpose of stabilizing the quality and reducing the tail, reducing the cost, improving the production efficiency.
Description
Technical field
The present invention relates to flotation technology fields, and in particular to a kind of Floating Production Process control method of Behavior-based control portrait
And system.
Background technique
Floating Production Process is the important link of ore-dressing technique, currently, Floating Production Process depends on the experience of people
Knowledge and a small amount of measurement data are operated, and since the factor for influencing floating operation is numerous, are selected in ore pulp type, operating upstream
Stability, dosage, liquid level etc., these factors have the characteristics that non-linear, time variation, large time delay, close coupling, therefore completely
Using manual operation, cause post labor intensity of workers big, be easy to cause the quality fluctuation of concentrate product.Portrait brand technology is a kind of
New big data processing mode, for being modeled for user or user behavior.Usage behavior Portrait brand technology portraying and
The behavioural characteristic of Floating Production Process is characterized, and the recommendation operation of production process is provided, advantageously accounts for above-mentioned business problem.It is floating
Selecting production process is the important link of ore-dressing technique, currently, Floating Production Process causes post worker completely using manual operation
Large labor intensity, and it be easy to cause the quality fluctuation of concentrate product.
Summary of the invention
In view of the problems of the existing technology, the present invention provides a kind of Floating Production Process controlling party of Behavior-based control portrait
Method and system are classified, for different ore pulps by building floatation process behavior portrait label system in conjunction with Multistage Support Vector Machine
Type proposes corresponding optimum operation mode, to control Floating Production Process operation, reaches steady matter drop tail, improves and float
Select the purpose of productivity effect.
On the one hand, the present invention provides a kind of Floating Production Process control method of Behavior-based control portrait, comprising:
1) storage of floatation process behavioral indicator data and pretreatment, the floatation process behavioral indicator, comprising: give mine product
Position, various medicament flows, concentrate gatage, scraper plate scrape tail thickness, Floatation Concentrate Grade, flotation tailing grade；
2) the floatation process behavioral indicator of selection building floatation process behavior portrait label system: is subjected to single dimension k-
Means cluster, is clustered into several grades for each floatation process behavioral indicator respectively, and each grade is drawn a portrait as the behavior generated and marked
Label store the form that the data in data warehouse switch to behavior portrait label, construct floatation process behavior portrait label body
The corresponding relationship of system, i.e. floatation process behavioral indicator and behavior portrait label；
3) it excavates the corresponding relationship of ore pulp type and operation mode: being grasped each floatation process behavioral indicator data as one
Operation mode carries out operation mode excavation, and the operation mode in discovery floatation process as much as possible is simultaneously named as ore pulp type
Portrait label；
4) corresponding optimum operation mode is found out for different ore pulp types: using k-means clustering algorithm, by n kind ore pulp
Type, m kind Floatation Concentrate Grade, p kind flotation tailing grade are combined, and symbiosis is at n*m*p=q kind sample, by every kind of sample
Corresponding operation mode is clustered, and it is the ore pulp kind that the cluster centre that the most classification of quantity is chosen in cluster result, which is used as,
Operation mode i.e. optimum operation mode is instructed in class floatation process；
5) building multi-layer SVM classifier carries out determining ore pulp type of classifying to real-time behavior portrait；
6) Multistage Support Vector Machine classifier is constructed to each ore pulp type, flotation is predicted according to real-time operation mode
Concentrate grade, flotation tailing grade, carry out the adjustment of operation mode；
7) when there is Floatation Concentrate Grade that Multistage Support Vector Machine classifier predicts and flotation tailing grade does not all conform to
When the number of lattice is more than threshold value, return step 3).
Further, step 1) includes:
1.1) when floatation process behavioral indicator data store, off-line measurement gives mine grade, Floatation Concentrate Grade, flotation tail
Mine grade data every two hours store once, dosing monitoring data, the concentrate gatage, scraper plate of the dispensing of on-line measurement
It is primary to scrape storage in tail thickness every 10 minutes；
1.2) data of off-line measurement and the data of on-line measurement are subjected to alignment of data；
1.3) nearly 2 years all data are extracted to store into database；
1.4) ETL processing is done to data in the dump procedure from database to data warehouse.
Further, the step 3) includes:
3.1) any operation mode is selected, the Euclidean distance of the operation mode Yu other operation modes is calculated；
3.2) it will be classified as one kind with the smallest preceding 5%~8% operation mode of operation mode Euclidean distance currently selected,
It is named as ore pulp type 1, the portrait label as ore pulp type 1；
3.3) another operation mode is randomly choosed in unnamed operation mode, the operation mode and other operation modes
Euclidean distance；
3.4) it will be classified as one kind with the smallest preceding 5%~8% operation mode of operation mode Euclidean distance currently selected,
It is named as ore pulp type 2, the portrait label as ore pulp type 2；Loop iteration has been classified simultaneously until all operation modes
Name.
Further, step 5) building multi-layer SVM classifier carries out determining mine of classifying to real-time behavior portrait
Starch type, comprising:
5.1) multi-layer SVM classifier is constructed using historical operation mode data, is defeated with operation mode data
Enter, different ore pulp types are output；
5.2) classified using multi-layer SVM classifier to real-time operation mode data, determine behaviour in real time
The affiliated ore pulp type of operation mode data.
Further, the step 6) includes:
6.1) for each ore pulp type one Multistage Support Vector Machine classifier of building: by m kind Floatation Concentrate Grade, p
Kind flotation tailing grade is combined, and m*p=q kind sample is obtained；Utilize the corresponding distance instruction of q kind sample and every kind of sample
Practice several operation modes of time recently as sample training collection, constructs q-1 grades of support vector machine classifiers；
6.2) real-time operation mode is inputted into q-1 grades of support vector machine classifiers, specimen types is exported, according to sample kind
The corresponding relationship of class and Floatation Concentrate Grade, flotation tailing grade predicts Floatation Concentrate Grade, flotation tailing grade；
6.3) according to obtained from specimen types Floatation Concentrate Grade and flotation tailing grade to scene operation mode into
Row adjustment.
On the other hand, the present invention provides a kind of Floating Production Process control system of Behavior-based control portrait, comprising:
Behavioral indicator chooses and pretreatment unit, stores and pre-processes for floatation process behavioral indicator data, described
Floatation process behavioral indicator, comprising: scrape tail thickness, flotation essence to mine grade, various medicament flows, concentrate gatage, scraper plate
Mine grade, flotation tailing grade；
Portrait label system construction unit, for constructing floatation process behavior portrait label system: by the flotation of selection
Journey behavioral indicator carries out single dimension k-means cluster, and each floatation process behavioral indicator is clustered into several grades, each grade respectively
As the behavior portrait label of generation, the form that the data in data warehouse switch to behavior portrait label is stored, is constructed floating
Select process state portrait label system, the i.e. corresponding relationship of floatation process behavioral indicator and behavior portrait label；
Label of drawing a portrait excavates unit, for excavating the corresponding relationship of ore pulp type and operation mode: by each floatation process row
It is achievement data as an operation mode, carries out operation mode excavation, the operation mould in discovery floatation process as much as possible
Formula is simultaneously named as ore pulp type portrait label；
Optimal choice unit, for finding out corresponding optimum operation mode for different ore pulp types: being clustered using k-means
N kind ore pulp type, m kind Floatation Concentrate Grade, p kind flotation tailing grade are combined by algorithm, and symbiosis is at n*m*p=q kind sample
This, the corresponding operation mode of every kind of sample is clustered, and the cluster centre of the most classification of quantity is chosen in cluster result
As to instruct operation mode i.e. optimum operation mode in the ore pulp type floatation process；
Ore pulp type classification unit classifies to real-time behavior portrait for constructing multi-layer SVM classifier
Determine ore pulp type；
Operation mode adjustment unit, for constructing Multistage Support Vector Machine classifier to each ore pulp type, according to real-time
Operation mode predict Floatation Concentrate Grade, flotation tailing grade, carry out the adjustment of operation mode；
There is the Floatation Concentrate Grade and flotation that Multistage Support Vector Machine classifier predicts for working as in threshold determination unit
When all underproof number of tailings grade is more than threshold value, returns and portrait label is called to excavate unit.
Further, the behavioral indicator selection and pretreatment unit, comprising:
Memory module, floatation process behavioral indicator data store when, off-line measurement to mine grade, Floatation Concentrate Grade,
Flotation tailing grade data every two hours store once, and the dosing monitoring data of the dispensing of on-line measurement, concentrate gate are opened
It is primary that degree, scraper plate scrape storage in tail thickness every 10 minutes；
The data of off-line measurement and the data of on-line measurement are carried out alignment of data by alignment of data module；
Data extraction module extracts nearly 2 years all data and stores into database；
ETL processing module does ETL processing to data in the dump procedure from database to data warehouse.
Further, the portrait label excavates unit, comprising:
First computing module selects any operation mode, calculates the Euclidean distance of the operation mode Yu other operation modes；
Label of drawing a portrait names module, by with the smallest preceding 5%~8% operation of operation mode Euclidean distance that currently selects
Mode is classified as one kind, is named as ore pulp type 1, the portrait label as ore pulp type 1；
Second computing module randomly chooses another operation mode in unnamed operation mode, the operation mode and its
The Euclidean distance of his operation mode；
Sort out name module, by with the smallest preceding 5%~8% operation mode of operation mode Euclidean distance that currently selects
It is classified as one kind, is named as ore pulp type 2, the portrait label as ore pulp type 2；Loop iteration until all operation modes
It is classified and names.
Further, the ore pulp type classification unit, comprising:
First classifier constructs module, multi-layer SVM classifier is constructed using historical operation mode data, with behaviour
Operation mode data are input, and different ore pulp types are output；
Ore pulp type determining module divides real-time operation mode data using multi-layer SVM classifier
Class determines the affiliated ore pulp type of real-time operation mode data.
Further, the operation mode adjustment unit, comprising:
Second classifier constructs module, constructs a Multistage Support Vector Machine classifier for each ore pulp type: by m kind
Floatation Concentrate Grade, p kind flotation tailing grade are combined, and m*p=q kind sample is obtained；Utilize q kind sample and every kind of sample
This corresponding several operation mode nearest apart from the training time constructs q-1 grades of support vector machines point as sample training collection
Class device；
Real-time operation mode is inputted q-1 grades of support vector machine classifiers, exports specimen types by prediction module, according to
The corresponding relationship of specimen types and Floatation Concentrate Grade, flotation tailing grade predicts Floatation Concentrate Grade, flotation tailing grade；
Adjust module, according to obtained from specimen types Floatation Concentrate Grade and flotation tailing grade to scene operation mould
Formula is adjusted.
The utility model has the advantages that
The present invention selects ore properties, dosage, concentrate gatage, scraper plate to scrape tail thickness to Floating Production Process, builds
Stood and drawn a portrait than more complete behavior, and established the incidence relation between them, instruct accordingly flotation production station worker into
Row operation has reached steady matter drop tail, has reduced cost, improve the purpose of flotation productivity effect.
Detailed description of the invention
Fig. 1 is the Floating Production Process control method flow chart that Behavior-based control is drawn a portrait in the embodiment of the present invention 1；
Fig. 2 is the flow chart of step 1) in the embodiment of the present invention 1；
Fig. 3 is the flow chart of step 4) in the embodiment of the present invention 1；
Fig. 4 is the flow chart of step 6) in the embodiment of the present invention 1；
Fig. 5 is the flow chart of step 7) in the embodiment of the present invention 1；
Fig. 6 is the Floating Production Process control system block diagram that Behavior-based control is drawn a portrait in the embodiment of the present invention 2；
Fig. 7 is data warehouse unit block diagram in the embodiment of the present invention 2；
Fig. 8 is label excavation unit block diagram of drawing a portrait in the embodiment of the present invention 2；
Fig. 9 is ore pulp type classification unit block diagram in the embodiment of the present invention 2；
Figure 10 is operation mode adjustment unit block diagram in the embodiment of the present invention 2；
Figure 11 is the flow chart that n-1 grades of support vector machines are constructed in the embodiment of the present invention 1.
Specific embodiment
Specific embodiments of the present invention will be described in detail with reference to the accompanying drawing.
Embodiment 1
The present invention provides a kind of Floating Production Process control method of Behavior-based control portrait, as shown in Figure 1, comprising:
1) storage of floatation process behavioral indicator data and pretreatment, the floatation process behavioral indicator, comprising: give mine product
Position, various medicament flows, concentrate gatage, scraper plate scrape tail thickness, Floatation Concentrate Grade, flotation tailing grade；
Various medicament flows, comprising: the thick medicament flow of LKY, the thin medicament flow of LKY, NAOH medicament flow, starch medicament stream
Amount；
2) the floatation process behavioral indicator of selection building floatation process behavior portrait label system: is subjected to single dimension k-
Means cluster, is clustered into several grades for each floatation process behavioral indicator respectively, and each grade is drawn a portrait as the behavior generated and marked
Label store the form that the data in data warehouse switch to behavior portrait label, construct floatation process behavior portrait label body
The corresponding relationship of system, i.e. floatation process behavioral indicator and behavior portrait label；
Wherein the thick medicament flow of LKY, the thin medicament flow of LKY, NAOH medicament flow, starch medicament flow, concentrate gate are opened
Degree, scraper plate scrape tail thickness and are divided into third or five etc., that is, are divided into large, medium and small third or very big, big, medium, small, minimum
Five etc..It is " grade is medium ", " grade is preferable ", " grade is poor " that flotation tailing grade, which generates label, and Floatation Concentrate Grade generates
Label is " qualification rate range ", " higher in qualification rate ", " relatively low in qualification rate "；It artificially provides to include " mine to mine grade label
Shi Yixuan ", " ore is medium optional ", " choosing of ore difficulty ", " the extremely difficult choosing of ore "；Data in data warehouse are switched into behavior portrait
The form of label stores；
3) it excavates the corresponding relationship of ore pulp type and operation mode: being grasped each floatation process behavioral indicator data as one
The thick medicament flow of operation mode, i.e. LKY, the thin medicament flow of LKY, NAOH medicament flow, starch medicament flow, concentrate gatage,
Scraper plate scrapes six behavioral indicators of tail thickness and collectively forms an operation mode, carries out operation mode excavation, discovery as much as possible
Operation mode in floatation process and naming is drawn a portrait label as ore pulp type；
4) corresponding optimum operation mode is found out for different ore pulp types: using k-means clustering algorithm, by n kind ore pulp
Type, m kind Floatation Concentrate Grade, p kind flotation tailing grade are combined, and symbiosis is at n*m*p=q kind sample, by every kind of sample
Corresponding operation mode is clustered, and it is the ore pulp kind that the cluster centre that the most classification of quantity is chosen in cluster result, which is used as,
Operation mode i.e. optimum operation mode is instructed in class floatation process；
5) building multi-layer SVM classifier carries out determining ore pulp type of classifying to real-time behavior portrait；
6) Multistage Support Vector Machine classifier is constructed to each ore pulp type, flotation is predicted according to real-time operation mode
Concentrate grade, flotation tailing grade, carry out the adjustment of operation mode；
7) when there is Floatation Concentrate Grade that Multistage Support Vector Machine classifier predicts and flotation tailing grade does not all conform to
When the number of lattice is more than threshold value, return step 3).
Further, as shown in Fig. 2, step 1) includes:
1.1) when floatation process monitoring index data store, off-line measurement gives mine grade, Floatation Concentrate Grade, flotation tail
Mine grade data every two hours store once, dosing monitoring data, the concentrate gatage, scraper plate of the dispensing of on-line measurement
It is primary to scrape storage in tail thickness every 10 minutes；
1.2) data of off-line measurement and the data of on-line measurement are subjected to alignment of data；
1.3) nearly 2 years all data are extracted to store into database；
1.4) ETL processing is done to data in the dump procedure from database to data warehouse.
Further, as shown in figure 3, the step 3) includes:
3.1) select any operation mode, using formula one calculate the operation mode and other operation modes it is European away from
From；
Wherein, d is the Euclidean distance of two operation modes, and it is one that x and y, which respectively represent two different operation modes,
Six-vector, i=1,2,3,4,5,6 respectively indicate the thick medicament flow of LKY, the thin medicament flow of LKY, NAOH medicament flow, starch
Medicament flow, concentrate gatage, scraper plate scrape tail thickness.
3.2) it will be classified as one kind with the smallest preceding 5%~8% operation mode of operation mode Euclidean distance currently selected,
It is named as ore pulp type 1, the portrait label as ore pulp type 1；
3.3) another operation mode is randomly choosed in unnamed operation mode, the operation mode and other operation modes
Euclidean distance；
3.4) it will be classified as one kind with the smallest preceding 5%~8% operation mode of operation mode Euclidean distance currently selected,
It is named as ore pulp type 2, the portrait label as ore pulp type 2；Loop iteration has been classified simultaneously until all operation modes
Name.
Further, as shown in figure 4, step 5) constructs multi-layer SVM classifier to real-time behavior portrait progress
Classify and determine ore pulp type, comprising:
5.1) multi-layer SVM classifier is constructed using historical operation mode data, is defeated with operation mode data
Enter, different ore pulp types are output；
5.2) classified using multi-layer SVM classifier to real-time operation mode data, determine behaviour in real time
The affiliated ore pulp type of operation mode data.
Further, as shown in figure 5, the step 6) includes:
6.1) a Multistage Support Vector Machine classifier is constructed for each ore pulp type: according to experimental data, by flotation
Concentrate grade is divided into acceptability limit, is higher in qualified, relatively low in qualified three classes, and it is product respectively that flotation tailing grade, which is divided into three classes,
Position is preferably, grade is poor, grade is medium is combined, and 3*3=9 kind sample is obtained；Utilize 9 kinds of samples and every kind of sample pair
15~20 operation modes nearest apart from the training time answered are instructed as sample training collection using algorithm of support vector machine
Practice, input is the sample training collection of six-vector, and each training sample format is x=(the thick medicament flow of LKY, the thin medicament stream of LKY
Amount, NAOH medicament flow, starch medicament flow, concentrate gatage, scraper plate scrape tail thickness), obtained support vector cassification
Device output is specimen types i；
- 1 representative sample 1 of first order support vector machine classifier use, remaining sample 2, sample 3 ... sample q are adopted
It is indicated with+1, the training sample set of input " operation mode ,+1 " and " operation mode, -1 " structure obtains first order support vector machines
Classifier；And so on, i-stage support vector machine classifier use -1 represents i-th kind of sample, represents i+1, i+2 ... with+1
Q kind sample, training obtain i-stage support vector machine classifier, and such q kind sample passes through q-1 support vector machine classifier
Classify, corresponding Floatation Concentrate Grade and flotation tailing grade are obtained according to the type of sample；
6.2) after the completion of support vector machine classifier training, the operation mode x generated in real time is inputted into q-1 grades of supporting vectors
In machine classifier, specimen types are exported, thus according to the corresponding relationship of specimen types and Floatation Concentrate Grade, flotation tailing grade
Predict Floatation Concentrate Grade, flotation tailing grade；
If 6.3) Floatation Concentrate Grade obtained in the specimen types is " relatively low in qualification rate ", flotation tailing grade is
" grade is poor " then instructs operation mode to be adjusted the operation mode at scene according to existing.
Embodiment 2
The present embodiment provides a kind of Floating Production Process control systems of Behavior-based control portrait, as shown in Figure 6, comprising:
Behavioral indicator chooses and pretreatment unit, stores and pre-processes for floatation process behavioral indicator data, described
Floatation process behavioral indicator, comprising: scrape tail thickness, flotation essence to mine grade, various medicament flows, concentrate gatage, scraper plate
Mine grade, flotation tailing grade；
Portrait label system construction unit, for constructing floatation process behavior portrait label system: by the flotation of selection
Journey behavioral indicator carries out single dimension k-means cluster, and each floatation process behavioral indicator is clustered into several grades, each grade respectively
As the behavior portrait label of generation, the form that the data in data warehouse switch to behavior portrait label is stored, is constructed floating
Select process state portrait label system, the i.e. corresponding relationship of floatation process behavioral indicator and behavior portrait label；
Label of drawing a portrait excavates unit, for excavating the corresponding relationship of ore pulp type and operation mode: by each floatation process row
It is achievement data as an operation mode, carries out operation mode excavation, the operation mould in discovery floatation process as much as possible
Formula is simultaneously named as ore pulp type portrait label；
Optimal choice unit, for finding out corresponding optimum operation mode for different ore pulp types: being clustered using k-means
N kind ore pulp type, m kind Floatation Concentrate Grade, p kind flotation tailing grade are combined by algorithm, and symbiosis is at n*m*p=q kind sample
This, the corresponding operation mode of every kind of sample is clustered, and the cluster centre of the most classification of quantity is chosen in cluster result
As to instruct operation mode i.e. optimum operation mode in the ore pulp type floatation process；
Ore pulp type classification unit classifies to real-time behavior portrait for constructing multi-layer SVM classifier
Determine ore pulp type；
Operation mode adjustment unit, for constructing Multistage Support Vector Machine classifier to each ore pulp type, according to real-time
Operation mode predict Floatation Concentrate Grade, flotation tailing grade, carry out the adjustment of operation mode；
There is the Floatation Concentrate Grade and flotation that Multistage Support Vector Machine classifier predicts for working as in threshold determination unit
When all underproof number of tailings grade is more than threshold value, returns and portrait label is called to excavate unit.
Further, as shown in fig. 7, the behavioral indicator chooses and pretreatment unit, comprising:
Memory module, floatation process monitoring index data store when, off-line measurement to mine grade, Floatation Concentrate Grade,
Flotation tailing grade data every two hours store once, and the dosing monitoring data of the dispensing of on-line measurement, concentrate gate are opened
It is primary that degree, scraper plate scrape storage in tail thickness every 10 minutes；
The data of off-line measurement and the data of on-line measurement are carried out alignment of data by alignment of data module；
Data extraction module extracts nearly 2 years all data and stores into database；
ETL processing module does ETL processing to data in the dump procedure from database to data warehouse.
Further, as shown in figure 8, the portrait label excavates unit, comprising:
First computing module selects any operation mode, calculates the Euclidean distance of the operation mode Yu other operation modes；
Label of drawing a portrait names module, by with the smallest preceding 5%~8% operation of operation mode Euclidean distance that currently selects
Mode is classified as one kind, is named as ore pulp type 1, the portrait label as ore pulp type 1；
Second computing module randomly chooses another operation mode in unnamed operation mode, the operation mode and its
The Euclidean distance of his operation mode；
Sort out name module, by with the smallest preceding 5%~8% operation mode of operation mode Euclidean distance that currently selects
It is classified as one kind, is named as ore pulp type 2, the portrait label as ore pulp type 2；Loop iteration until all operation modes
It is classified and names.
Further, as shown in figure 9, the ore pulp type classification unit, comprising:
First classifier constructs module, multi-layer SVM classifier is constructed using historical operation mode data, with behaviour
Operation mode data are input, and different ore pulp types are output；
Ore pulp type determining module divides real-time operation mode data using multi-layer SVM classifier
Class determines the affiliated ore pulp type of real-time operation mode data.
Further, as shown in Figure 10, the operation mode adjustment unit, comprising:
Second classifier constructs module, constructs a Multistage Support Vector Machine classifier for each ore pulp type: by m kind
Floatation Concentrate Grade, p kind flotation tailing grade are combined, and m*p=q kind sample is obtained；Utilize q kind sample and every kind of sample
This corresponding several operation mode nearest apart from the training time constructs q-1 grades of support vector machines point as sample training collection
Class device；
- 1 representative sample 1 of first order support vector machine classifier use, remaining sample 2, sample 3 ... sample q are adopted
It is indicated with+1, the training sample set of input " operation mode ,+1 " and " operation mode, -1 " structure obtains first order support vector machines
Classifier；And so on, i-stage support vector machine classifier use -1 represents i-th kind of sample, represents i+1, i+2 ... with+1
Q kind sample, training obtain i-stage support vector machine classifier, and such q kind sample passes through q-1 support vector machine classifier
Classify, corresponding Floatation Concentrate Grade and flotation tailing grade are obtained according to the type of sample；
Real-time operation mode is inputted q-1 grades of support vector machine classifiers, exports specimen types by prediction module, according to
The corresponding relationship of specimen types and Floatation Concentrate Grade, flotation tailing grade predicts Floatation Concentrate Grade, flotation tailing grade；
Adjust module, according to obtained from specimen types Floatation Concentrate Grade and flotation tailing grade to scene operation mould
Formula is adjusted.
Claims (10)
1. a kind of Floating Production Process control method of Behavior-based control portrait characterized by comprising
1) floatation process behavioral indicator data storage and pretreatment, the floatation process behavioral indicator, comprising: to mine grade,
Various medicament flows, concentrate gatage, scraper plate scrape tail thickness, Floatation Concentrate Grade, flotation tailing grade；Various medicament streams
Amount, comprising: the thick medicament flow of LKY, the thin medicament flow of LKY, NAOH medicament flow, starch medicament flow；
2) the floatation process behavioral indicator of selection building floatation process behavior portrait label system: is subjected to single dimension k-means
Cluster, is clustered into several grades for each floatation process behavioral indicator respectively, and each grade will be counted as the behavior portrait label generated
Floatation process behavior portrait label system is constructed in the form storage for switching to behavior portrait label according to the data in warehouse, i.e., floating
Select the corresponding relationship of process state index and behavior portrait label；
3) corresponding relationship of ore pulp type and operation mode is excavated: using each floatation process behavioral indicator data as an operation mould
Formula carries out operation mode excavation, and the operation mode in discovery floatation process as much as possible is simultaneously named as ore pulp type portrait
Label；
4) find out corresponding optimal operation mode for different ore pulp types: using k-means clustering algorithm, by n kind ore pulp type,
M kind Floatation Concentrate Grade, p kind flotation tailing grade are combined, and symbiosis is at n*m*p kind sample, by the corresponding behaviour of every kind of sample
Operation mode is clustered, and the cluster centre of the most classification of quantity is chosen in cluster result as the ore pulp type floatation process
In instruct operation mode i.e. optimal operation mode；
5) building multi-layer SVM classifier carries out determining ore pulp type of classifying to real-time behavior portrait；
6) Multistage Support Vector Machine classifier is constructed to each ore pulp type, flotation concentrate is predicted according to real-time operation mode
Grade, flotation tailing grade, carry out the adjustment of operation mode；
7) when there is Floatation Concentrate Grade that Multistage Support Vector Machine classifier predicts and flotation tailing grade is all underproof
When number is more than threshold value, return step 3).
2. the Floating Production Process control method of Behavior-based control portrait according to claim 1, which is characterized in that step 1)
Include:
1.1) when floatation process behavioral indicator data store, off-line measurement gives mine grade, Floatation Concentrate Grade, flotation tailing product
Position data every two hours store once, and the medicament flow monitoring data of the dispensing of on-line measurement, concentrate gatage, scraper plate scrape tail
Storage in thickness every 10 minutes is primary；
1.2) data of off-line measurement and the data of on-line measurement are subjected to alignment of data；
1.3) nearly 2 years all data are extracted to store into database；
1.4) ETL processing is done to data in the dump procedure from database to data warehouse.
3. the Floating Production Process control method of Behavior-based control portrait according to claim 1, which is characterized in that the step
It is rapid 3) to include:
3.1) any operation mode is selected, the Euclidean distance of the operation mode Yu other operation modes is calculated；
3.2) it will be classified as one kind with the smallest preceding 5% ~ 8% operation mode of operation mode Euclidean distance currently selected, be named as
Ore pulp type 1, the portrait label as ore pulp type 1；
3.3) another operation mode is randomly choosed in unnamed operation mode, calculates the operation mode and other operation modes
Euclidean distance；
3.4) it will be classified as one kind with the smallest preceding 5% ~ 8% operation mode of operation mode Euclidean distance currently selected, be named as
Ore pulp type 2, the portrait label as ore pulp type 2；Loop iteration is until all operation modes have been classified and are named.
4. the Floating Production Process control method of Behavior-based control portrait according to claim 1, which is characterized in that step 5)
Building multi-layer SVM classifier carries out determining ore pulp type of classifying to real-time behavior portrait, comprising:
5.1) multi-layer SVM classifier is constructed using historical operation mode data, is input with operation mode data, no
It is output with ore pulp type；
5.2) classified using multi-layer SVM classifier to real-time operation mode data, determine operation mould in real time
The affiliated ore pulp type of formula data.
5. the Floating Production Process control method of Behavior-based control portrait according to claim 1, which is characterized in that the step
It is rapid 6) to include:
6.1) a Multistage Support Vector Machine classifier is constructed for each ore pulp type: m kind Floatation Concentrate Grade, p kind are floated
It selects tailings grade to be combined, m*p kind sample is obtained；When distance training corresponding using m*p kind sample and every kind of sample
Between nearest several operation modes be used as sample training collection, m*p-1 grades of support vector machine classifiers of construction；
6.2) real-time operation mode is inputted into m*p-1 grades of support vector machine classifiers, specimen types is exported, according to specimen types
Floatation Concentrate Grade, flotation tailing grade are predicted with the corresponding relationship of Floatation Concentrate Grade, flotation tailing grade；
6.3) according to obtained from specimen types Floatation Concentrate Grade and flotation tailing grade to scene operation mode adjust
It is whole.
6. a kind of Floating Production Process control system of Behavior-based control portrait characterized by comprising
Behavioral indicator chooses and pretreatment unit, stores and pre-processes for floatation process behavioral indicator data, the flotation
Process state index, comprising: scrape tail thickness, flotation concentrate product to mine grade, various medicament flows, concentrate gatage, scraper plate
Position, flotation tailing grade；Various medicament flows, comprising: the thick medicament flow of LKY, the thin medicament flow of LKY, NAOH medicament flow,
Starch medicament flow；
Portrait label system construction unit, for constructing floatation process behavior portrait label system: by the floatation process row of selection
Single dimension k-means cluster is carried out for index, each floatation process behavioral indicator is clustered into several grades, each grade conduct respectively
The behavior portrait label of generation, the form that the data in data warehouse switch to behavior portrait label is stored, flotation is constructed
Cheng Hangwei portrait label system, the i.e. corresponding relationship of floatation process behavioral indicator and behavior portrait label；
Label of drawing a portrait excavates unit, for excavating the corresponding relationship of ore pulp type and operation mode: each floatation process behavior is referred to
Data are marked as an operation mode, carry out operation mode excavation, the operation mode in discovery floatation process as much as possible is simultaneously
Name is as ore pulp type portrait label；
Optimal choice unit, for finding out corresponding optimal operation mode for different ore pulp types: being calculated using k-means cluster
N kind ore pulp type, m kind Floatation Concentrate Grade, p kind flotation tailing grade are combined by method, and symbiosis, will at n*m*p kind sample
The corresponding operation mode of every kind of sample is clustered, and the cluster centre that the most classification of quantity is chosen in cluster result, which is used as, to be somebody's turn to do
Operation mode i.e. optimal operation mode is instructed in ore pulp type floatation process；
Ore pulp type classification unit carries out classification determination to real-time behavior portrait for constructing multi-layer SVM classifier
Ore pulp type；
Operation mode adjustment unit, for constructing Multistage Support Vector Machine classifier to each ore pulp type, according to real-time behaviour
Operation mode predicts Floatation Concentrate Grade, flotation tailing grade, carries out the adjustment of operation mode；
There is the Floatation Concentrate Grade and flotation tailing that Multistage Support Vector Machine classifier predicts for working as in threshold determination unit
When all underproof number of grade is more than threshold value, returns and portrait label is called to excavate unit.
7. system according to claim 6, which is characterized in that the behavioral indicator chooses and pretreatment unit, comprising:
Memory module, when floatation process behavioral indicator data store, off-line measurement gives mine grade, Floatation Concentrate Grade, flotation
Tailings grade data every two hours store once, and the medicament flow monitoring data of the dispensing of on-line measurement, are scraped at concentrate gatage
It is primary that plate scrapes storage in tail thickness every 10 minutes；
The data of off-line measurement and the data of on-line measurement are carried out alignment of data by alignment of data module；
Data extraction module extracts nearly 2 years all data and stores into database；
ETL processing module does ETL processing to data in the dump procedure from database to data warehouse.
8. system according to claim 6, which is characterized in that the portrait label excavates unit, comprising:
First computing module selects any operation mode, calculates the Euclidean distance of the operation mode Yu other operation modes；
Label of drawing a portrait names module, will return with the smallest preceding 5% ~ 8% operation mode of operation mode Euclidean distance currently selected
For one kind, it is named as ore pulp type 1, the portrait label as ore pulp type 1；
Second computing module randomly chooses another operation mode in unnamed operation mode, calculates the operation mode and its
The Euclidean distance of his operation mode；
Sort out name module, one will be classified as with the smallest preceding 5% ~ 8% operation mode of operation mode Euclidean distance currently selected
Class is named as ore pulp type 2, the portrait label as ore pulp type 2；Loop iteration has been classified until all operation modes
And it names.
9. system according to claim 6, which is characterized in that the ore pulp type classification unit, comprising:
First classifier constructs module, multi-layer SVM classifier is constructed using historical operation mode data, to operate mould
Formula data are input, and different ore pulp types are output；
Ore pulp type determining module classifies to real-time operation mode data using multi-layer SVM classifier, really
The fixed real-time affiliated ore pulp type of operation mode data.
10. system according to claim 6, which is characterized in that the operation mode adjustment unit, comprising:
Second classifier constructs module, constructs a Multistage Support Vector Machine classifier for each ore pulp type: by m kind flotation
Concentrate grade, p kind flotation tailing grade are combined, and m*p kind sample is obtained；Utilize m*p kind sample and every kind of sample pair
Several operation modes nearest apart from the training time answered construct m*p-1 grades of support vector cassifications as sample training collection
Device；
Real-time operation mode is inputted m*p-1 grades of support vector machine classifiers, specimen types is exported, according to sample by prediction module
The corresponding relationship of this type and Floatation Concentrate Grade, flotation tailing grade predicts Floatation Concentrate Grade, flotation tailing grade；
Adjust module, according to obtained from specimen types Floatation Concentrate Grade and flotation tailing grade to scene operation mode into
Row adjustment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611168388.1A CN106597898B (en) | 2016-12-16 | 2016-12-16 | A kind of the Floating Production Process control method and system of Behavior-based control portrait |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611168388.1A CN106597898B (en) | 2016-12-16 | 2016-12-16 | A kind of the Floating Production Process control method and system of Behavior-based control portrait |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106597898A CN106597898A (en) | 2017-04-26 |
| CN106597898B true CN106597898B (en) | 2019-05-31 |
Family
ID=58599546
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611168388.1A Active CN106597898B (en) | 2016-12-16 | 2016-12-16 | A kind of the Floating Production Process control method and system of Behavior-based control portrait |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106597898B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119985393B (en) * | 2025-04-17 | 2025-08-01 | 中检易兴元科技(北京)有限公司 | Method for detecting aniline additives in motor gasoline based on near infrared spectrum |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101036904A (en) * | 2007-04-30 | 2007-09-19 | 中南大学 | Flotation froth image recognition device based on machine vision and the mine concentration grade forecast method |
| WO2009086606A1 (en) * | 2008-01-09 | 2009-07-16 | Bhp Billiton Ssm Development Pty Ltd | Processing nickel bearing sulphides |
| CN101511488A (en) * | 2006-08-30 | 2009-08-19 | 奥图泰有限公司 | Equipment and method for flotating and classifying mineral slurry. |
| CN102855492A (en) * | 2012-07-27 | 2013-01-02 | 中南大学 | Classification method based on mineral flotation foam image |
| CN103839057A (en) * | 2014-03-28 | 2014-06-04 | 中南大学 | Antimony floatation working condition recognition method and system |
| CN104299045A (en) * | 2014-09-24 | 2015-01-21 | 东北大学 | System and method for forecasting yield of concentrate in whole mineral dressing process |
| CN105260805A (en) * | 2015-11-16 | 2016-01-20 | 中南大学 | Antimony ore grade soft-measurement method based on selective fusion of heterogeneous classifier |
-
2016
- 2016-12-16 CN CN201611168388.1A patent/CN106597898B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101511488A (en) * | 2006-08-30 | 2009-08-19 | 奥图泰有限公司 | Equipment and method for flotating and classifying mineral slurry. |
| CN101511488B (en) * | 2006-08-30 | 2013-05-08 | 奥图泰有限公司 | Equipment and method for flotating and classifying mineral slurry. |
| CN101036904A (en) * | 2007-04-30 | 2007-09-19 | 中南大学 | Flotation froth image recognition device based on machine vision and the mine concentration grade forecast method |
| WO2009086606A1 (en) * | 2008-01-09 | 2009-07-16 | Bhp Billiton Ssm Development Pty Ltd | Processing nickel bearing sulphides |
| CN102855492A (en) * | 2012-07-27 | 2013-01-02 | 中南大学 | Classification method based on mineral flotation foam image |
| CN103839057A (en) * | 2014-03-28 | 2014-06-04 | 中南大学 | Antimony floatation working condition recognition method and system |
| CN104299045A (en) * | 2014-09-24 | 2015-01-21 | 东北大学 | System and method for forecasting yield of concentrate in whole mineral dressing process |
| CN105260805A (en) * | 2015-11-16 | 2016-01-20 | 中南大学 | Antimony ore grade soft-measurement method based on selective fusion of heterogeneous classifier |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106597898A (en) | 2017-04-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110378799A (en) | Aluminium oxide comprehensive production index decision-making technique based on multiple dimensioned depth convolutional network | |
| CN103871246B (en) | Based on the Short-time Traffic Flow Forecasting Methods of road network spatial relation constraint Lasso | |
| CN108090456A (en) | A kind of Lane detection method and device | |
| CN109190226B (en) | Soft measurement method for overflow granularity index of ore grinding system | |
| CN105467449B (en) | Deep layer Thin oil sandwich quantitatively characterizing method based on earthquake classification Sensitive Attributes fusion | |
| CN117943213B (en) | Real-time monitoring and early warning system and method for micro-bubble flotation machine | |
| Truong et al. | Machine learning-based method for forecasting water levels in irrigation and drainage systems | |
| EP3816935A1 (en) | Display control device, display control method, and display control program | |
| CN103226728B (en) | High density polyethylene polymerization cascade course of reaction Intelligent Measurement and yield optimization method | |
| CN107506862A (en) | A kind of online real-time estimate system and method for grinding particle size based on Internet of Things | |
| CN108985941A (en) | A kind of stock intelligent Forecasting of combination newsletter archive | |
| WO2025112936A1 (en) | Lightweight active online modeling method fusing mechanism and dynamic memory | |
| Di et al. | An automatic and integrated self-diagnosing system for the silting disease of drainage pipelines based on SSAE-TSNE and MS-LSTM | |
| CN106597898B (en) | A kind of the Floating Production Process control method and system of Behavior-based control portrait | |
| Xu et al. | MsFfTsGP: Multi-source features-fused two-stage grade prediction of zinc tailings in lead-zinc flotation process via multi-stream 3D convolution with attention mechanism | |
| CN113095597A (en) | Novel flood process design method suitable for small watershed without data | |
| CN106874568A (en) | A kind of material particular diameter distribution forecasting method of mechanical milling process | |
| Xue et al. | Evolutionary feature manipulation in data mining/big data | |
| CN108734207A (en) | A kind of model prediction method based on double preferred Semi-Supervised Regression algorithms | |
| CN115445785A (en) | Zinc fast and coarse dosing quantity adjusting method based on memory network and prediction controller | |
| CN115482232A (en) | Negative sample guided self-distillation pathological image analysis method | |
| CN118820980B (en) | Iron ore sorting big data analysis and decision-making system | |
| CN115545267A (en) | A Flotation Recovery Prediction Method Based on Significance P Value and XGBoost | |
| CN115293391A (en) | A runoff prediction method and system based on mixed runoff model | |
| CN117443564A (en) | A magnetic separation optimization control method based on knowledge graph |
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 |