CN114093442B - Method for predicting and optimizing fluxed balling performance - Google Patents

Method for predicting and optimizing fluxed balling performance Download PDF

Info

Publication number
CN114093442B
CN114093442B CN202210018911.1A CN202210018911A CN114093442B CN 114093442 B CN114093442 B CN 114093442B CN 202210018911 A CN202210018911 A CN 202210018911A CN 114093442 B CN114093442 B CN 114093442B
Authority
CN
China
Prior art keywords
data
mass fraction
sio
green
alkalinity
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
Application number
CN202210018911.1A
Other languages
Chinese (zh)
Other versions
CN114093442A (en
Inventor
刘卫星
李�杰
闫奥琪
付之珍
陈太龙
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
North China University of Science and Technology
Original Assignee
North China University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by North China University of Science and Technology filed Critical North China University of Science and Technology
Priority to CN202210018911.1A priority Critical patent/CN114093442B/en
Publication of CN114093442A publication Critical patent/CN114093442A/en
Application granted granted Critical
Publication of CN114093442B publication Critical patent/CN114093442B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
    • C—CHEMISTRY; METALLURGY
    • C22—METALLURGY; FERROUS OR NON-FERROUS ALLOYS; TREATMENT OF ALLOYS OR NON-FERROUS METALS
    • C22B—PRODUCTION AND REFINING OF METALS; PRETREATMENT OF RAW MATERIALS
    • C22B1/00—Preliminary treatment of ores or scrap
    • C22B1/14—Agglomerating; Briquetting; Binding; Granulating
    • C22B1/24—Binding; Briquetting ; Granulating
    • C22B1/2406—Binding; Briquetting ; Granulating pelletizing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/30—Prediction of properties of chemical compounds, compositions or mixtures

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Computing Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Physics & Mathematics (AREA)
  • Geochemistry & Mineralogy (AREA)
  • Computational Linguistics (AREA)
  • Manufacturing & Machinery (AREA)
  • General Life Sciences & Earth Sciences (AREA)
  • Materials Engineering (AREA)
  • Mechanical Engineering (AREA)
  • Metallurgy (AREA)
  • Organic Chemistry (AREA)
  • Environmental & Geological Engineering (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Geology (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Crystallography & Structural Chemistry (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Manufacture And Refinement Of Metals (AREA)

Abstract

The invention discloses a method for predicting and optimizing the performance of fluxed spheronization, which is implemented by using SiO as a raw material component2The content, the alkalinity R, MgO content, the mass fraction of less than 0.074mm in the mixture granularity and the four performances of the green ball such as the balling rate, the compressive strength, the falling strength and the bursting temperature are predicted by using a BP neural network to obtain a prediction model based on the components of the pelletizing raw materials and the mass fraction of less than 0.074mm in the pelletizing mixture on the green ball performance.

Description

Method for predicting and optimizing fluxed balling performance
Technical Field
The invention belongs to the technical field of metallurgical iron making, and particularly relates to a method for predicting and optimizing the performance of fluxed green pellets.
Background
The pelletizing stage of the pellet ore is the first process of pellet ore production, and the quality of green pellets greatly influences the quality of finished pellet ore. For example, the size, moisture, mechanical strength, chemical composition, etc. of green pellets can significantly affect the next consolidation process. The quality of the green ball itself depends on the physicochemical properties of the raw materials in addition to the process.
In the prior art, chinese patent application publication No. CN 111041194 a discloses a preparation apparatus for improving the quality of green pellets of mineral powder pellets and a use method thereof, wherein the distribution of the pellets is obtained by shooting the pellets on a conveyor belt in real time and analyzing and processing the shot pictures, and a disc pelletizer is adjusted accordingly, thereby achieving the purpose of optimizing the quality of the green pellets. The process has the disadvantages that: the method starts from the process, utilizes various equipment instruments to record the distribution of pellet ore particles in the pelletizing process in real time, and does not consider the influence of chemical components in the initial stage of green pellet manufacturing.
Chinese patent application publication No. CN 113570557 a discloses a pellet consolidation degree evaluation method based on image recognition, which is based on pellet ore phase structure images, and adopts an intelligent algorithm to realize segmentation and recognition of pellets inside particles, and has important significance for optimizing the roasting process parameters of pellets. The process has the disadvantages that: the method can be applied and embodied only in the roasting process, and in the pellet production process, the quality of green pellets determines whether pellets can be produced smoothly, and the decisive factors are not considered.
At present, a large amount of green pellet manufacturing process research data, production practice data, laboratory-obtained experimental data and the like in the pellet research field are not effectively integrated, efficient utilization cannot be achieved, the characteristics of pellet raw materials are not deeply researched, tested and summarized by a system, the existing traditional research and development mode cannot be changed, and rapid development of pellet field research is achieved. Therefore, the invention uses a mathematical nonlinear regression method to carry out data modeling, constructs a BP neural network model and establishes SiO2The content, the alkalinity R, MgO content, the mass fraction of the mixture with the granularity less than 0.074mm and the green ball performance have important significance for predicting and optimizing the green ball performance.
Disclosure of Invention
The invention aims to provide a method for predicting and optimizing green ball performance, so that the optimal material ratio is obtained, and energy conservation and emission reduction are realized.
To solve the above technical problem, an embodiment of the present invention provides the following solutions: a method for predicting and optimizing fluxed spheronization performance, comprising the following steps of: SiO in the raw material composition2The mass fraction of less than 0.074mm in content, alkalinity R, MgO content and mixture granularity, and the balling rate, compressive strength, falling strength and bursting temperature of green pellets are predicted by a BP neural network to obtain a prediction model of the green pellet performance based on the components of the pelletizing raw materials and the mass fraction of less than 0.074mm in the granulation mixture, and the BP neural network prediction process is characterized by comprising the following steps:
firstly, data pretreatment is carried out to different alkalinity and SiO2The particle size distribution and the mass fraction of each test raw material under the content and the MgO content are subjected to data processing, the particle size data of the mixture is obtained by calculation by using the data that the particle size of the raw material is less than 0.074mm, and the calculation formula is as follows:
Figure 100002_DEST_PATH_IMAGE001
is provided witha i Is as followsiThe mass fraction of the seed material with the granularity of less than 0.074mm,b ji is as followsjUnder the preparation conditionsiPercentage of seed material in the ingredient, z j Z is the mass fraction of the mixture with a particle size of less than 0.074mm under the jth preparation condition j In thatjEach of =1 and 2 … 15 represents:z 1 -z 5 when the alkalinity R is 1.0 and the MgO mass fraction is 1.8 percent, SiO2The mass fraction is 3.5%, 4.0%, 4.5%, 5.0%, 5.5%;z 6 -z 10 is SiO2When the mass fraction is 5.5% and the mass fraction of MgO is 1.8%, the alkalinity R is 0.6, 0.8, 1.0, 1.2 and 1.4;z 11 -z 15 is SiO25.5 percent by mass and 1.0 percent by alkalinity R, and 15 preparation conditions of 1.8 percent by mass, 2.0 percent by mass, 2.2 percent by mass, 2.4 percent by mass and 2.6 percent by mass of MgO, wherein，i=1,2…6，j=1,2…15。
In the change of SiO2After green pellets are prepared according to the mass fractions of less than 0.074mm in content, alkalinity R, MgO content and mixture granularity, four green pellet performances of the green pellets, namely the pelletizing rate, the bursting temperature, the compressive strength and the falling strength, are measured;
secondly, the method comprises the following steps: carrying out rationalization amplification optimization on the existing test data;
and thirdly: the four amplified variables correspond to the four performances of the green ball, a prediction model of 'the four variables correspond to one performance' is constructed, a BP neural network of a 3-layer network is established, logsig is selected as an excitation function between an input layer and a hidden layer, purelin is selected as an excitation function between the hidden layer and an output layer, and a training function is trainlm;
and finally: the prediction model error detection adopts average absolute percentage errorMAPEAs a model error checking criterion.
Further, the pelletizing raw materials comprise temple ditch powder, ground mountain powder, Iran powder, limestone powder, light-burned magnesia powder and bentonite, and the pelletizing raw materials are mixed to prepare different SiO2Content, alkalinity R, MgO content.
Furthermore, a disc pelletizer with the diameter of 500mm multiplied by 150mm is adopted for pelletizing, and green pellet pelletizing is carried out under the conditions that the linear speed is 0.98m/s and the inclination angle is 48 degrees.
Further, the existing test data is subjected to rationalization amplification optimization, random data which accords with standard Gaussian distribution is generated for each group of data, the fluctuation value is added to the original test data, and the amplified original data is in normal distribution
Figure 100002_DEST_PATH_IMAGE002
。
Further, the error detection method of the prediction model is used for averaging absolute percentage errorsMAPEAs a model error checking criterion, the formula is as follows:
Figure DEST_PATH_IMAGE003
wherein, in total, havenThe number of the variables is one,y t is shown astThe actual value of the one or more of the one or,
Figure DEST_PATH_IMAGE004
show firsttThe number of the predicted values is calculated,MAPEthe smaller the value of (a), the more accurate the prediction result of the model is represented.
The invention has the beneficial effects that: the method is based on the existing experimental data and raw material components, and SiO in the raw material components is extracted2BP neural network prediction is carried out on the mass fraction of less than 0.074mm in content, alkalinity R, MgO content and mixture granularity and four performances of green ball forming rate, compressive strength, falling strength and bursting temperature, a green ball performance prediction model based on raw material components and mass fraction is obtained according to training of the BP neural network, actual smelting process requirements are met, and the method can be applied to green ball performance control and prediction.
Drawings
FIG. 1 is a flow chart of a method of predicting and optimizing green ball performance.
FIG. 2 is a schematic diagram of a flux based green ball performance prediction model.
FIG. 3 is a graph of data from a portion of experimental amplification data in an example of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.
An embodiment of the present invention provides a method for predicting and optimizing green ball performance, as shown in fig. 1, the method comprising the steps of:
step one, batching
Detecting and analyzing chemical compositions, granularity compositions and moisture of test raw materials including temple ditch powder, ground mountain powder, Iran powder, limestone powder, light-burned magnesium powder and bentonite, wherein the fixed ground mountain powder and the bentonite are respectively 10.00% and 0.70% in the test raw material proportion, and adjusting the proportion of temple ditch powder and Iran powder to adjust SiO of a pelletizing mixture2Content (c); adjusting the alkalinity R of the pelletizing mixture by adjusting the proportion of the flux limestone powder; adjusting the MgO content of the pelletizing mixture by adjusting the proportion of the light-burned magnesium powderAmounts, part of the data are shown in tables 1-3:
TABLE 1 ingredient adjustmentw(SiO2) /% regimen
Figure DEST_PATH_IMAGE005
TABLE 2 alkalinity adjustment scheme for ingredients
Figure DEST_PATH_IMAGE006
TABLE 3 ingredient adjustmentw(MgO)/% scheme
Figure DEST_PATH_IMAGE007
Step two, preparing the pellet
The raw materials in different proportions are mixed, and a disc pelletizer with the diameter of 500mm multiplied by 150mm is adopted to carry out green pellet fabrication tests under the conditions that the linear speed is 0.98m/s and the inclination angle is 48 degrees.
Step three, green ball performance detection
The green pellets obtained under the above conditions were subjected to green pellet forming rate, compressive strength, drop strength and burst temperature measurement. Because errors are generated in actual tests and manual measurement and calculation, performance measurement results obtained by three groups of experiments with the same green ball preparation conditions are averaged to be used as a final measurement result of the green ball performance under the preparation conditions, and the four green ball performance measurement methods are as follows:
measurement of the balling rate: the balling rate is obtained by comparing the mass of the green balls with the total mass of all the green balls in the diameter range of 8-20 mm.
Green ball compressive strength determination: the compressive strength of the green pellets is measured by adopting an intelligent particle strength tester (DL III type), 20 green pellets with the diameter of 10.0-12.5 mm are randomly selected from the test samples, the test is carried out according to GB/T14201-charge 1993, the arithmetic mean value is taken as the measurement result, and the unit of the compressive strength of the green pellets is 'N/pellet'.
And (3) burst temperature measurement: measuring the temperature by using a bursting temperature measuring furnace, adjusting the air speed of a standard state to be 1.8m/s, heating the heating furnace to a preset temperature, then filling 20 green pellets, and blowing hot air with a certain temperature onto the green pellets from bottom to top for 5 min. After the pellets are taken out, the bursting condition of the pellets is visually observed, and the bursting temperature is determined as the temperature at which 1 pellet bursts.
Determination of the falling Strength: taking 20 green balls with the diameter of 10.0-12.5 mm, freely dropping at the height of 500mm, and recording the times of non-breakage of each ball. The total times of the 20 green balls not cracking are calculated, and the average time of the 20 green balls not cracking is obtained as the falling strength of the green ball, wherein the unit is as follows: sub/ball.
TABLE 4 green ball Performance
Figure DEST_PATH_IMAGE008
Step four, establishing a model by a BP neural network algorithm
Firstly, data pretreatment is carried out to different SiO2The particle size distribution and the mass fraction of each test raw material with the content and the alkalinity R, MgO are subjected to data processing, the data that the particle size of the raw material is less than 0.074mm are applied, the mass fraction that the particle size of the mixture is less than 0.074mm is obtained through calculation, two effective figures are reserved, and the calculation result is as follows:
z j (j=1,2 … 15) is 1.0 for basicity R and 1.8% for MgO mass fraction, respectively, SiO2The mass fraction is 3.5%, 4.0%, 4.5%, 5.0%, 5.5%; SiO 22When the mass fraction is 5.5% and the mass fraction of MgO is 1.8%, the alkalinity R is 0.6, 0.8, 1.0, 1.2 and 1.4; SiO 2215 preparation conditions including 1.8%, 2.0%, 2.2%, 2.4% and 2.6% of MgO by mass fraction of 5.5% and alkalinity R of 1.0,
Figure DEST_PATH_IMAGE009
the mass fraction of the mixture with the granularity of less than 0.074mm is calculated to be evenly distributed at 70-77%. Because of the measurement error in the metallurgical test, the alkalinity R of the 3 groups is 1.0, the MgO mass fraction is 1.8 percent, and the SiO content is2Incomplete phase of data under preparation conditions with mass fraction of 5.5%Meanwhile, a small difference exists, and in order to reduce the training error of the model, the average value of 3 groups of data is taken as the green ball data under the condition.
TABLE 5 unamplified training data
Figure DEST_PATH_IMAGE010
Because the metallurgical test has the problem of small test data amount, in order to train a prediction model to obtain a more accurate prediction result, the existing test data needs to be subjected to rationalization amplification optimization, random data which is in accordance with standard Gaussian distribution is generated for each group of data, the fluctuation value is added to the original test data, and the amplified original data is in normal distribution
Figure DEST_PATH_IMAGE011
。
According to metallurgical test, SiO is mainly selected among factors influencing the change of the fluxed green ball performance2The content, the alkalinity R, MgO content and the mass fraction of the mixture with the granularity less than 0.074mm are used as important influencing factors to construct a fluxed green ball performance change model.
Selecting green balls subjected to data amplification on different SiO2Under the preparation conditions of content, alkalinity R, MgO content and the like, 2500 groups of data of green ball balling rate, bursting temperature, compressive strength, falling strength and the like are used for model training, 500 groups are used for testing, 15 groups are used for verification, the used data set is a common simple data set which does not relate to time sequence or computer vision, the number of hidden layers is 1, namely, a BP neural network of a 3-layer network is established, and the number of hidden layer nodes is established by an empirical formula:
Figure DEST_PATH_IMAGE012
mthe number of nodes of the hidden layer is determined,nthe number of nodes of the input layer is,lthe number of nodes of the output layer,αa constant between 1 and 10.
Traversing the number of the hidden layer nodes between 1 and 20 according to an empirical formula, and comparing to obtain the number of the hidden layer nodes, wherein the number of the hidden layer nodes with different properties of the pellet is shown in the table 6:
TABLE 6 hidden layer node number of pellet with different properties
Figure DEST_PATH_IMAGE013
Excitation function selection logsig between input layer and hidden layer:
Figure DEST_PATH_IMAGE014
the logsig function acts on normalization to [0,1 ]]The gradient of the data in the range is reduced quickly, and the network error is reduced.
Excitation function selection purelin between hidden and output layers, linear function (purelin function):y(x)=xthe purelin function effectively slows down the gradient descending speed between the hidden layer and the output layer, and logsig and purelin process data together, so that the stability of the network is improved.
The results corresponding to different training functions (taking burst temperature as an example) under the condition that the BP network structure and the activation function are the same:
TABLE 7 error of different training functions
Figure DEST_PATH_IMAGE015
The training functions of the balling rate, the bursting temperature, the compressive strength and the falling strength are obtained by comparison and are trainlm.
The test result of the flux-based spheronization performance is reasonably expanded to 3015 groups through data amplification. 2500 groups of data are selected for the model training, 500 groups of data are selected for the model testing, and 15 groups of data are selected for the model verification. The average absolute percentage error is selected from the prediction modelMAPE) As a standard for model error checking.
Mean absolute percentage error (MAPE) The smaller the value of (A), the closer the predicted result of the model is to the true value, i.e. the more accurate the model is, the formula is as follows:
Figure DEST_PATH_IMAGE016
wherein, in total, havenThe number of the variables is one,y t is shown astThe actual value of the one or more of the one or,
Figure 786755DEST_PATH_IMAGE004
is shown astThe number of the predicted values is calculated,MAPEthe smaller the value of (a), the more accurate the prediction result of the model is represented.
In the green ball balling rate prediction model established after data amplification, the prediction modelMAPE= 2.64%. Predicted values obtained from another 500 test data setsMAPEThe values are all between 0 and 10 percent, which shows that the prediction model can accurately predict the balling rate of the green ball. Of predictive models not trained for data augmentationMAPE=3.45% of predictive model obtained by data amplificationMAPEThe values are not very different.
The result of predicting the bursting temperature of the green ball by the amplified data shows that the result is obtained by predicting the test dataMAPE= 5.09%. Data testing was performed using 500 sets of data, of which 464 sets of dataMAPE0-10%, 36 groups of dataMAPEIs 10-20%. Without data amplification, using predictive models built from the original test dataMAPE=6.55%, therefore, the prediction model trained after data amplification is able to more accurately predict the burst temperature of green pellets.
Predicting the error of green ball compression strength prediction model by using test data obtained by data amplificationMAPE= 5.75%. Model tests were performed using 500 sets of amplified data, 442 sets of whichMAPEValues between 0 and 10%, 50 groups of dataMAPEValues between 10-20%, 8 groups of dataMAPEThe value is 20-21%, which shows that the prediction model can accurately predict the compressive strength of the green ball. Of predictive models not trained by data amplificationMAPE=6.35%, the prediction result of the prediction model after data amplification is more accurate.
Error of green ball falling strength prediction model, prediction model tested by test dataMAPE= 11.19%. 393 of the data sets of the model test results for 500 data setsMAPEValues between 0 and 10%, 83 groups of dataMAPEA value of 10-20%, 24 groups of dataMAPEThe value is between 30 and 40%. Of predictive models not trained by data amplificationMAPE=32.19% of training model with amplification dataMAPEThe difference in values is large, and therefore the prediction result of the prediction model after data amplification is more accurate.
The invention provides a method for predicting green ball performance based on a BP neural network, which is based on the existing experimental data and raw material components to predict SiO in the raw material components2The BP neural network model is established according to the content, the alkalinity R, MgO content, the mass fraction of the mixture with the granularity of less than 0.074mm, the green ball forming rate, the compressive strength, the falling strength and the bursting temperature, the experimental data are reasonably amplified by using a Gaussian distribution data amplification method, the green ball performance can be effectively predicted by using the prediction model, the actual smelting process requirements are met, and the BP neural network model has important significance for optimizing the green ball performance and improving the charging proportion.

Claims (5)

1. A method for predicting and optimizing fluxed spheronization performance, comprising the following steps of: SiO in the raw material composition2The mass fraction of less than 0.074mm in content, alkalinity R, MgO content and mixture granularity, and the balling rate, compressive strength, falling strength and bursting temperature of green pellets are predicted by a BP neural network to obtain a prediction model of the green pellet performance based on the components of the pelletizing raw materials and the mass fraction of less than 0.074mm in the granulation mixture, and the BP neural network prediction process is characterized by comprising the following steps:
firstly, data pretreatment is carried out to different SiO2The particle size distribution and the mass fraction of each test raw material under the content and the alkalinity R, MgO content are subjected to data processing, the data that the particle size of the raw material is less than 0.074mm are applied, the data that the particle size of the mixture is less than 0.074mm are obtained through calculation, and the calculation formula is as follows:
Figure DEST_PATH_IMAGE001
is provided witha i Is as followsiThe granularity of the seed material is less than 0.074mmThe mass fraction of the active carbon is,b ji is as followsjUnder the preparation conditionsiPercentage of seed material in the ingredient, z j Is at the firstjUnder the preparation conditions, the mass fraction of the mixture with the granularity less than 0.074mm changes SiO2After green pellets are prepared according to the mass fractions of less than 0.074mm in content, alkalinity R, MgO content and mixture granularity, four green pellet performances of the green pellets, namely the pelletizing rate, the bursting temperature, the compressive strength and the falling strength, are measured;
secondly, the method comprises the following steps: carrying out rationalization amplification optimization on the existing test data;
and thirdly: establishing a prediction model of 'four variables correspond to one performance' by corresponding the four amplified variables to the four performances of the green ball, establishing a BP neural network of a 3-layer network, selecting logsig as an excitation function between an input layer and a hidden layer, selecting purelin as an excitation function between the hidden layer and an output layer, and training the function to be trainlm;
and finally: the prediction model error detection adopts average absolute percentage errorMAPEAs a model error checking criterion.
2. The method of claim 1, wherein the pelletizing raw materials comprise Miao powder, Mosan powder, Iran powder, limestone powder, light-burned magnesium powder, and bentonite, and the pelletizing raw materials are mixed to prepare different SiO2Content, alkalinity R, MgO content.
3. The method of claim 1, wherein pelletizing is carried out using a 500 x 150mm diameter disk pelletizer with a line speed of 0.98m/s and a pitch angle of 48 °.
4. The method of claim 1, wherein the pre-existing experimental data is optimized for rational amplification to produce a normalized, gaussian random data for each set of dataAdding the fluctuation value to the original test data, and normally distributing the amplified original data
Figure DEST_PATH_IMAGE002
。
5. The method of claim 1, wherein z is calculated as j In thatjEach of =1 and 2 … 15 represents:z 1 -z 5 when the alkalinity R is 1.0 and the MgO mass fraction is 1.8 percent, SiO2The mass fraction is 3.5%, 4.0%, 4.5%, 5.0%, 5.5%;z 6 -z 10 is SiO2When the mass fraction is 5.5% and the mass fraction of MgO is 1.8%, the alkalinity R is 0.6, 0.8, 1.0, 1.2 and 1.4;z 11 -z 15 is SiO215 preparation conditions including 5.5% by mass and 1.0% by mass of the basicity R of MgO, 1.8%, 2.0%, 2.2%, 2.4% and 2.6% by mass, wherein,i=1,2…6，j=1,2…15。
CN202210018911.1A 2022-01-10 2022-01-10 Method for predicting and optimizing fluxed balling performance Active CN114093442B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210018911.1A CN114093442B (en) 2022-01-10 2022-01-10 Method for predicting and optimizing fluxed balling performance

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210018911.1A CN114093442B (en) 2022-01-10 2022-01-10 Method for predicting and optimizing fluxed balling performance

Publications (2)

Publication Number Publication Date
CN114093442A CN114093442A (en) 2022-02-25
CN114093442B true CN114093442B (en) 2022-03-29

Family

ID=80308392

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210018911.1A Active CN114093442B (en) 2022-01-10 2022-01-10 Method for predicting and optimizing fluxed balling performance

Country Status (1)

Country Link
CN (1) CN114093442B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104480300A (en) * 2014-11-20 2015-04-01 中南大学 Pellet production method based on prediction of compressive strength of pellets in rotary kiln
CN109242210A (en) * 2018-10-15 2019-01-18 华北理工大学 A kind of auto recommending method of pellet feed optimum proportioning
WO2021001859A2 (en) * 2019-07-04 2021-01-07 Tata Consultancy Services Limited Method and system for optimization of agglomeration of ores

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104480300A (en) * 2014-11-20 2015-04-01 中南大学 Pellet production method based on prediction of compressive strength of pellets in rotary kiln
CN109242210A (en) * 2018-10-15 2019-01-18 华北理工大学 A kind of auto recommending method of pellet feed optimum proportioning
WO2021001859A2 (en) * 2019-07-04 2021-01-07 Tata Consultancy Services Limited Method and system for optimization of agglomeration of ores

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
《Effect of MgO Content on the Properties of Magnesia Fluxed Pellets》;Yuzhu Zhang等;《Annual Meeting & Exhibition Supplemental Proceedings》;20191231;全文 *
《Feature recognition of irregular pellet images by regularized Extreme Learning Machine in combination with fractal theory》;Weixing Liu等;《Future Generation Computer Systems》;20210903;全文 *
《Pellet Roasting Management System Based on Deep Learning and Internet of Things》;Weixing Liu等;《Hindawi Complexity》;20210415;全文 *
《Prediction Model of Carbon-containing Pellet Reduction Metallization Ratio Using Neural Network and Genetic Algorithm》;Wei ZHANG等;《ISIJ International》;20211231;全文 *
《基于BP神经网络钒钛烧结矿低温还原粉化性能预测》;信自成等;《钢铁钒钛》;20170630;全文 *

Also Published As

Publication number Publication date
CN114093442A (en) 2022-02-25

Similar Documents

Publication Publication Date Title
CN109752261B (en) Method for detecting strength of iron ore powder binding phase in sintering process
CN103588391B (en) A kind of hollow glass micropearl and preparation method thereof
CN106521059B (en) Blast furnace charge level ore coke ratio is measured with phased-array radar to control the method for blast furnace air flow method
CN114093442B (en) Method for predicting and optimizing fluxed balling performance
Zhou et al. Effect of coke rate and basicity on computed tomography-measured pore parameters and effective thermal conductivity of iron ore sinter
CN111561963A (en) Novel method for evaluating coking property and coke thermal strength of coking coal and correlating coking coal with coking coal thermal strength
CN114496107A (en) Method for sintering ore blending based on liquid phase fluidity index of uniformly mixed ore
CN111308894B (en) Parameter adjusting method for coupling ore blending and control process in sintering process
CN109776007B (en) Method for determining optimal gradation of artificial filler
CN114842918B (en) Automatic water adding method for sintering mixture based on machine learning
CN111595718A (en) Test method for detecting coke mixing thermal property by using coke reactivity measuring device
CN114410959A (en) Method for calculating fuel distribution in sintering and granulating pellets
CN111638155B (en) Ore blending structure evaluation method based on granulation quasi-particle sintering behavior
CN107782596A (en) The making of Cavitated sample and method of testing after more than 5mm iron ore sinterings
CN112611678A (en) Method for detecting liquidity of iron ore powder liquid phase based on actual sintering production condition
CN114067926A (en) Drum strength prediction method for sintering ore of each material layer of sintering trolley
CN111560515B (en) Control method for pellet production
CN108929950B (en) A kind of Iron Ore Matching in Sintering method based on mixture granulation characteristic
CN111553602A (en) Method for determining mixing amount of curing material in concrete
CN114636625A (en) Detection and evaluation method for strength of sintered and granulated pellets
CN106834662B (en) A kind of CO/CO based on multi-state sintering process2Ratio prediction technique
CN108593523A (en) A method of the characterization pellet degree of consolidation
CN112903689A (en) Method for detecting quality of auxiliary material dolomite in steelmaking
JP4100042B2 (en) Granulation method of sintering raw material
CN113817489B (en) Coal blending method based on coke optical characteristic index

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