CN117612651A - Method for predicting manganese content of converter endpoint - Google Patents
Method for predicting manganese content of converter endpoint Download PDFInfo
- Publication number
- CN117612651A CN117612651A CN202311627914.6A CN202311627914A CN117612651A CN 117612651 A CN117612651 A CN 117612651A CN 202311627914 A CN202311627914 A CN 202311627914A CN 117612651 A CN117612651 A CN 117612651A
- Authority
- CN
- China
- Prior art keywords
- manganese content
- mass fraction
- data
- converter
- molten iron
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000011572 manganese Substances 0.000 title claims abstract description 178
- PWHULOQIROXLJO-UHFFFAOYSA-N Manganese Chemical compound [Mn] PWHULOQIROXLJO-UHFFFAOYSA-N 0.000 title claims abstract description 164
- 229910052748 manganese Inorganic materials 0.000 title claims abstract description 163
- 238000000034 method Methods 0.000 title claims abstract description 70
- 238000004519 manufacturing process Methods 0.000 claims abstract description 139
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 claims abstract description 102
- 229910052742 iron Inorganic materials 0.000 claims abstract description 51
- 229910000831 Steel Inorganic materials 0.000 claims abstract description 29
- 239000010959 steel Substances 0.000 claims abstract description 29
- 230000008569 process Effects 0.000 claims abstract description 28
- 238000012549 training Methods 0.000 claims abstract description 27
- 238000003723 Smelting Methods 0.000 claims abstract description 17
- 239000000956 alloy Substances 0.000 claims abstract description 13
- 229910045601 alloy Inorganic materials 0.000 claims abstract description 12
- 238000009628 steelmaking Methods 0.000 claims abstract description 8
- 238000006477 desulfuration reaction Methods 0.000 claims abstract description 7
- 230000023556 desulfurization Effects 0.000 claims abstract description 7
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 claims abstract description 6
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 claims abstract description 6
- XUIMIQQOPSSXEZ-UHFFFAOYSA-N Silicon Chemical compound [Si] XUIMIQQOPSSXEZ-UHFFFAOYSA-N 0.000 claims abstract description 6
- 229910052799 carbon Inorganic materials 0.000 claims abstract description 6
- 229910052698 phosphorus Inorganic materials 0.000 claims abstract description 6
- 239000011574 phosphorus Substances 0.000 claims abstract description 6
- 229910052710 silicon Inorganic materials 0.000 claims abstract description 6
- 239000010703 silicon Substances 0.000 claims abstract description 6
- NINIDFKCEFEMDL-UHFFFAOYSA-N Sulfur Chemical compound [S] NINIDFKCEFEMDL-UHFFFAOYSA-N 0.000 claims abstract description 5
- 238000004458 analytical method Methods 0.000 claims abstract description 5
- 229910052717 sulfur Inorganic materials 0.000 claims abstract description 5
- 239000011593 sulfur Substances 0.000 claims abstract description 5
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 5
- 238000012360 testing method Methods 0.000 claims description 27
- 229910000514 dolomite Inorganic materials 0.000 claims description 15
- 239000010459 dolomite Substances 0.000 claims description 15
- 238000010079 rubber tapping Methods 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 13
- 235000008733 Citrus aurantifolia Nutrition 0.000 claims description 10
- 235000011941 Tilia x europaea Nutrition 0.000 claims description 10
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 239000004571 lime Substances 0.000 claims description 10
- 238000012216 screening Methods 0.000 claims description 9
- DALUDRGQOYMVLD-UHFFFAOYSA-N iron manganese Chemical compound [Mn].[Fe] DALUDRGQOYMVLD-UHFFFAOYSA-N 0.000 claims description 6
- XWHPIFXRKKHEKR-UHFFFAOYSA-N iron silicon Chemical compound [Si].[Fe] XWHPIFXRKKHEKR-UHFFFAOYSA-N 0.000 claims description 6
- QMQXDJATSGGYDR-UHFFFAOYSA-N methylidyneiron Chemical compound [C].[Fe] QMQXDJATSGGYDR-UHFFFAOYSA-N 0.000 claims description 6
- DPTATFGPDCLUTF-UHFFFAOYSA-N phosphanylidyneiron Chemical compound [Fe]#P DPTATFGPDCLUTF-UHFFFAOYSA-N 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 230000007246 mechanism Effects 0.000 claims description 5
- 238000004140 cleaning Methods 0.000 claims description 4
- 238000001514 detection method Methods 0.000 claims description 4
- 238000009826 distribution Methods 0.000 claims description 3
- MBMLMWLHJBBADN-UHFFFAOYSA-N Ferrous sulfide Chemical compound [Fe]=S MBMLMWLHJBBADN-UHFFFAOYSA-N 0.000 claims description 2
- 102000005298 Iron-Sulfur Proteins Human genes 0.000 claims description 2
- 108010081409 Iron-Sulfur Proteins Proteins 0.000 claims description 2
- 239000000203 mixture Substances 0.000 claims 1
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000005275 alloying Methods 0.000 description 6
- 230000008901 benefit Effects 0.000 description 6
- 229910000851 Alloy steel Inorganic materials 0.000 description 5
- 239000000126 substance Substances 0.000 description 5
- 238000010801 machine learning Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000006467 substitution reaction Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Evolutionary Biology (AREA)
- Mathematical Physics (AREA)
- Carbon Steel Or Casting Steel Manufacturing (AREA)
Abstract
The invention relates to a method for predicting the endpoint manganese content of a converter, which is used for predicting the manganese content in converter steelmaking and comprises the following steps: s1, collecting current production data of a converter to be tested, wherein the current production data comprises the following steps: carbon mass fraction of molten iron, silicon mass fraction of molten iron, manganese mass fraction of molten iron, phosphorus mass fraction of molten iron, sulfur mass fraction of molten iron, water quantity, steel scrap quantity, desulfurization outlet temperature of molten iron, sublance process temperature, sublance process C mass fraction, sublance end temperature and the like; s2, inputting the current production data into a trained manganese content prediction model, and predicting the final manganese content of the converter event to be detected; the manganese content prediction model is obtained by training a pre-constructed initial manganese content prediction model based on historical production data of the converter. The method has the beneficial effects that the problem of accurate control of the manganese content of molten steel caused by the manganese content analysis lag of the smelting end point of the converter is solved, and the accurate control of alloy feeding is further realized.
Description
Technical Field
The invention relates to the field of converter steelmaking, in particular to a method for predicting the endpoint manganese content of a converter.
Background
At present, the steel production in China mainly uses a converter smelting long process, and along with the improvement of the requirements of the market on the quality and the performance of the steel, the converter tapping alloying becomes an important and indispensable process link. The method is influenced by sampling of the smelting end point and analysis lag of chemical components, and the chemical components of the smelting end point are unknown when the alloy addition amount is calculated, so that the accuracy control of the molten steel alloy components is restricted to a certain extent, and the method has important influence on the performance, the production efficiency and the economic benefit of steel. Manganese is the most widely used alloying element in steel materials, and precise control of its content is particularly important.
Disclosure of Invention
First, the technical problem to be solved
In view of the defects and shortcomings of the prior art, the invention provides a method for predicting the manganese content of a converter endpoint, which solves the technical problems of converter smelting endpoint sampling and chemical component analysis hysteresis.
(II) technical scheme
In order to achieve the above purpose, the invention mainly provides a method for predicting the endpoint manganese content of a converter, which mainly comprises the following steps:
s1, collecting current production data of a converter to be tested, wherein the current production data comprises the following steps: the method comprises the following steps of (1) carrying out carbon mass fraction of molten iron, silicon mass fraction of molten iron, manganese mass fraction of molten iron, phosphorus mass fraction of molten iron, sulfur mass fraction of molten iron, water quantity of scrap, desulfurization outlet temperature of molten iron, sublance process temperature, sublance process C mass fraction, sublance end point temperature, sublance end point C mass fraction, sublance end point O mass fraction, target tapping temperature, target end point C mass fraction, target end point Si mass fraction, target end point Mn mass fraction, target end point P mass fraction, target end point S mass fraction, light burned dolomite mass, lime mass and converter end point residual Mn content;
s2, inputting the current production data into a trained manganese content prediction model, and predicting the final manganese content of the converter event to be detected;
the manganese content prediction model is obtained by training a pre-constructed initial manganese content prediction model based on historical production data of the converter.
Optionally, after S2, the method further includes:
and adjusting the current production data based on the predicted manganese content until the predicted manganese content can meet the steel grade end point control target.
Optionally, the step S1 further includes: s0, training a pre-constructed initial manganese content prediction model based on historical production data of the converter to be tested to obtain a manganese content prediction model, wherein the method comprises the following steps of:
a1, collecting historical production data of the converter to be tested, wherein the historical production data is a set of complete production data of each single production of the converter to be tested;
a2, carrying out data preprocessing on the historical production data;
a3, selecting TSO corresponding to each single production of the converter to detect the manganese content as manganese content based on historical production data subjected to data preprocessing, and checking whether each piece of continuous characteristic data input by the converter accords with statistical normal distribution;
a4, judging characteristic data affecting the manganese content in the historical production data by means of a metallurgical principle;
and A5, screening the characteristic data of the historical production data, and inputting the characteristic data of each single production and the manganese content of the current production into a pre-constructed initial manganese content prediction model for training to obtain a manganese content prediction model.
Optionally, in step A1, the category of the complete production data of each single production includes:
smelting number, gun age, production date, steel code, molten iron carbon fraction, molten iron silicon mass fraction, molten iron manganese mass fraction, molten iron phosphorus mass fraction, molten iron sulfur mass fraction, molten iron quantity, scrap steel quantity, molten iron desulfurization outlet temperature, sublance process C mass fraction, sublance endpoint temperature, sublance endpoint C mass fraction, sublance endpoint O mass fraction, target tapping temperature, target endpoint C mass fraction, target endpoint Si mass fraction, target endpoint Mn mass fraction, target endpoint P mass fraction, target endpoint S mass fraction, light burned dolomite mass, lime mass.
Optionally, the A2 includes:
a21, cleaning the historical production data, and analyzing and deleting obvious invalid characteristic values by using a metallurgical mechanism:
the invalid characteristic value includes; smelting number, gun age, production date and steel code;
a22, carrying out outlier replacement on the cleaned historical production data;
the outlier includes: zero value, null value, discrete value.
Optionally, the a22 specifically is:
and replacing the abnormal value by taking an average value of the historical production data which is in the same category as the abnormal value in the historical production data as a replacement value.
Optionally, step A3 includes:
a31, based on the historical production data, selecting the final detection of the sublance, namely, detecting the mass fraction of manganese by TSO as the endpoint manganese content of the converter;
a32, carrying out normal test on each input characteristic data, and if the normal characteristics are not met, directly deleting the group of data with the discrete value.
Optionally, in step A4, the predefined correlation calculation formula is:
n is the total heat in the historical production data, x i Is the actual value of the characteristic data and,is the mean value of the characteristic data, y i Is the actual value of the manganese content,/->Is the average value of manganese content.
Optionally, in step A4, the characteristic data affecting the manganese content includes:
the quality of molten iron comprises the following components of a mass fraction of molten iron carbon, a mass fraction of molten iron silicon, a mass fraction of molten iron manganese, a mass fraction of molten iron phosphorus, a mass fraction of molten iron, a mass fraction of scrap steel, a temperature of a sublance process, a mass fraction of a sublance process C, a temperature of a sublance end point, a mass fraction of a sublance end point C, a mass fraction of a sublance end point O, a target tapping temperature, a mass fraction of a target end point C, a mass fraction of a target end point Si, a mass fraction of a target end point Mn, a mass fraction of a target end point P, a mass of light burned dolomite, a mass of dolomite and a mass of lime.
Optionally, step A5 includes:
a51, screening and marking the historical production data to obtain the characteristic data of each single production and the label corresponding to the characteristic data;
a52, taking the characteristic data of each single production and the manganese content of the current production as sample data, and inputting the sample data into the initial manganese content prediction model; the sample data comprises a training set and a testing set; the test set is used for testing the manganese content prediction model;
a53, optimizing the super parameters of the initial manganese content prediction model;
a54, training the initial manganese content prediction model through the training set to obtain a manganese content prediction model.
Optionally, after S5, the method further includes:
s6, inputting the characteristic data of each single production in the test set into the manganese content prediction model to generate the predicted manganese content of the secondary production;
calculating a determination coefficient and a root mean square error of the predicted manganese content of the current production based on the predicted manganese content of the current production and the manganese content of the same single production in the test set; testing the manganese content prediction model based on the decision coefficient and root mean square error;
determining coefficient R 2 The calculation formula is as follows:
the root mean square error RMSE calculation formula is as follows:
n is the total sample amount of the test set, f i For the predicted value of the model output, y i For the actual values of the feature data samples,is the characteristic data sample mean value.
(III) beneficial effects
According to the method for predicting the manganese content of the converter endpoint, provided by the invention, the current production data of the converter to be detected is input into the trained manganese content prediction model to predict the manganese content, and then the alloy addition is calculated according to the manganese content, so that the precise control of the first alloy addition is realized, the precise control of the molten steel alloy components is further realized, and the production efficiency and economic benefit are improved.
Drawings
FIG. 1 is a schematic flow chart of a method for predicting the endpoint manganese content of a converter based on machine learning according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of feature selection according to an embodiment of the present invention;
FIG. 3 is a line graph of manganese content predicted by a manganese content prediction model according to an embodiment of the present invention.
Detailed Description
The invention will be better explained by the following detailed description of the embodiments with reference to the drawings.
In order that the above-described aspects may be better understood, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
The converter steelmaking is a main long process of steel smelting, and the steelmaking process is completed in the converter by utilizing the physical heat of molten iron and chemical reaction between molten iron components to generate heat under the condition of no external energy source through molten iron, scrap steel and the like. Has the advantages of high production speed, high yield, high single furnace yield, low cost and less investment.
Along with the improvement of the requirements of the market on the quality and the performance of steel, the requirements on the control precision of the alloy content in the steel are higher and higher. The alloying of the converter tapping is an important link of alloy content control, however, the alloying is influenced by sampling of a smelting end point and analysis lag of chemical components, and the chemical components of the smelting end point are unknown when the alloy addition amount is calculated, so that the control of the component precision of molten steel alloy is restricted to a certain extent, and the influences on the quality, the production efficiency and the economic benefit of steel are all different. Manganese metal is an alloying element commonly used in steel, and the content control thereof is particularly important. Therefore, the invention provides a method for predicting the endpoint manganese content of a converter, which is implemented as follows in the embodiment shown in fig. 1:
s1, collecting current production data of a converter to be tested;
s2, inputting the current production data into a trained manganese content prediction model, and predicting the final manganese content of the converter event to be detected.
The manganese content prediction model is obtained by training a pre-constructed initial manganese content prediction model based on historical production data of the converter.
In an embodiment, the current production data may include:
carbon mass fraction of molten iron, silicon mass fraction of molten iron, manganese mass fraction of molten iron, phosphorus mass fraction of molten iron, molten iron amount, scrap steel amount and the like;
in an embodiment, before S1, training the pre-constructed initial manganese content prediction model based on the historical production data of the converter to be tested to obtain a manganese content prediction model includes:
a1, collecting historical production data of the converter to be tested, wherein the historical production data is a set of complete production data of each single production of the converter to be tested.
The collection of the complete production data can avoid the repeated collection of the data produced at the same time in the historical production data, and can ensure the authenticity of the data, the false data which is not kneaded or the finger derivatives given by the manual experience, and avoid the inaccuracy of the data samples.
In this embodiment, the types of the complete production data of each single production include:
the method comprises the following steps of (1) carrying out carbon mass fraction of molten iron, silicon mass fraction of molten iron, manganese mass fraction of molten iron, phosphorus mass fraction of molten iron, sulfur mass fraction of molten iron, water quantity of scrap, desulfurization outlet temperature of molten iron, sublance process temperature, sublance process C mass fraction, sublance end point temperature, sublance end point C mass fraction, sublance end point O mass fraction, target tapping temperature, target end point C mass fraction, target end point Si mass fraction, target end point Mn mass fraction, target end point P mass fraction, target end point S mass fraction, light burned dolomite mass, lime mass and converter end point residual Mn content.
Of course, in actual production, the types of the complete production data can be added and deleted in a small range according to the actual production.
Further, after the historical production data of the converter to be detected is collected, the problems of data deviation caused by misoperation, equipment failure or other reasons of field data collection personnel may exist, and abnormal data with null values, zero values and larger discrete degrees exists in the data of part of the heat. The missing value can reduce the sample information, the excessive abnormal value can cause deviation of the model prediction result, and the repeated value can cause the variance of the data to be reduced and the distribution to be changed, so that the step A2 can be implemented:
and carrying out data preprocessing on the historical production data, and adopting a deletion method, a substitution method, an interpolation method and the like to process the missing values, the repeated values and the abnormal values.
For example, in one embodiment the A2 implementation steps include:
a21, cleaning the historical production data, and analyzing and deleting obvious invalid characteristic values by using a metallurgical mechanism:
the invalid characteristic value may include; characteristics of smelting number, gun age, production date, steel code and the like which are irrelevant to manganese content;
a22, carrying out outlier replacement on the cleaned historical production data;
the outlier includes: zero value, null value, discrete value.
Specifically, the a22 may replace the abnormal value by using, as a replacement value, an average value of the historical production data of the same category as the abnormal value in the historical production data.
In one embodiment, for example,
the characteristics of the molten iron, such as manganese mass fraction, sublance end point O mass fraction, sublance end point temperature and the like, which have a larger relation with manganese content are replaced by an average value.
Further, implementing step A3, and selecting the corresponding end point manganese content of each single production of the converter based on the historical production data subjected to data pretreatment.
In some embodiments, this step may also be referred to as modeling the mechanism by which the converter to be tested makes the converter steelmaking.
The method can be implemented as follows;
a31, based on the historical production data, selecting the final detection of the sublance, namely detecting Mn mass fraction by TSO as the final manganese content of the converter.
Based on the steps, the manganese content corresponding to the complete production data of each single production is selected. The complete production data for each single production and its corresponding manganese content were taken as a sample data. In practical application, complete production data of enough production times should be collected as much as possible and corresponding manganese content is calculated, so that the model training process is prevented from being over-fitted.
For example, in one embodiment, complete data for more than thousand runs is collected.
Further, step A4 is implemented, and characteristic data affecting manganese content in the historical production data are judged by means of metallurgical principles.
In this embodiment, a pearson correlation coefficient method is selected for judgment and selection, and the predefined correlation calculation formula is as follows:
n is the total heat of the historical production data, x i Is the actual value of the characteristic data and,is the mean value of the characteristic data, y i Is the actual value of the manganese content,/->Is the average value of manganese content.
By carrying out correlation screening on various feature data of the historical production data, a feature group with strong interpretation ability on the target variable can be selected from all features, namely, features with larger correlation with the manganese content are selected from all features, so that the model training effect is better, and the accuracy of predicting the manganese content is high.
In other embodiments, other correlation calculations may be selected for calculation, and are not intended to be limiting.
As in the embodiment of fig. 3, all features include at least: the quality of molten iron comprises the following components of a mass fraction of molten iron carbon, a mass fraction of molten iron silicon, a mass fraction of molten iron manganese, a mass fraction of molten iron phosphorus, a mass fraction of molten iron, a mass fraction of scrap steel, a temperature of a sublance process, a mass fraction of a sublance process C, a temperature of a sublance end point, a mass fraction of a sublance end point C, a mass fraction of a sublance end point O, a target tapping temperature, a mass fraction of a target end point C, a mass fraction of a target end point Si, a mass fraction of a target end point Mn, a mass fraction of a target end point P, a mass of light burned dolomite, a mass of dolomite and a mass of lime.
Further, step A5 is implemented, the historical production data are screened according to the related characteristic types, the related characteristic data in each single production data are reserved, the characteristic data of each single production and the manganese content of the current production are input into a pre-built initial manganese content prediction model for training, and the obtained manganese content prediction model is obtained.
In practice, a manganese content prediction model is established based on machine learning, and at least two models should be established for comparison, including but not limited to: SVM, LGBM, catboost, etc., prevent the deviation of the generated manganese content prediction model.
Specifically, in an embodiment, the initial manganese content prediction model is a Catboost manganese content prediction model (Categorical Boosting), and in step A5, the method may include:
a51, screening and marking the historical production data to obtain the characteristic data of each single production and the label corresponding to the characteristic data; the method comprises the following steps:
looking at the new DataFrame using data.head (), ensuring that features and tags are within specified ranges; using data.dropna () to ensure that there is no NaN value in the data; looking at data types and data amounts using df.info (), it may be necessary to convert the data types as needed; the minimum, maximum, mean, median, standard deviation, and quartile range for each column can be known using df.
A52, taking the characteristic data of each single production and the manganese content of the current production as sample data, and inputting the sample data into the Catboost manganese content prediction model; dividing the sample data into a training set and a testing set; the test set is used for testing the manganese content prediction model.
In one embodiment, the sample data is split using the train_test_split function of Scikit-learn.
A53, optimizing the super parameters of the initial manganese content prediction model;
a54, training the initial manganese content prediction model through the training set to obtain a manganese content prediction model.
Further, in some embodiments, step S6 is also implemented, inputting the characteristic data of each single production in the test set to the manganese content prediction model, generating a current production predicted manganese content;
and calculating a determination coefficient and a root mean square error of the predicted manganese content of the current production based on the predicted manganese content of the current production and the manganese content of the same single production in the test set.
And testing the manganese content prediction model based on the decision coefficient and the root mean square error, and judging the effectiveness of the manganese content prediction model.
Determining coefficient (R) 2 ) Representing the fitting degree of the model; the Root Mean Square Error (RMSE) is taken as a standard for measuring the prediction result of the machine learning model, is also called as a standard error, is the arithmetic square root of the mean square error, and is obtained by introducing the root mean square error and introducing the standard error because the root mean square error is completely consistent with the introduced standard error, namely the dimension of the mean square error is different from the dimension of the data, and the degree of dispersion cannot be intuitively reflected.
Determining coefficient R 2 The calculation formula is as follows:
the root mean square error RMSE calculation formula is as follows:
n is the total quantity of characteristic data samples of the test set, f i For the predicted value of the model output, y i For the actual values of the feature data samples,is the characteristic data sample mean value.
According to the converter endpoint manganese content prediction method provided by the embodiments, the current production data of the converter to be detected is input into the trained manganese content prediction model to predict the manganese content, and then the alloy addition amount is calculated according to the manganese content, so that the precise control of the first alloy addition is realized, the precise control of the molten steel alloy components is further realized, and the production efficiency and the economic benefit are improved. The method predicts the manganese content of the alloying of the converter tapping before the converter steelmaking process begins with higher precision, accurately calculates the alloy addition, realizes the accurate control of the first alloy addition, further realizes the accurate control of the molten steel alloy components, and improves the production efficiency and the economic benefit. The quality qualification rate of steel tapping of converter steelmaking is improved, the probability of furnace return is reduced, and the resource waste is reduced.
The model prediction deviation caused by data repetition or errors in training samples is avoided by collecting complete production data, all the features are subjected to relevant screening, the generalization capability is stronger, excessive features and excessive fitting can be avoided, the calculation complexity and the training difficulty are reduced, and the model precision is increased.
And testing the model through the determination coefficient and the root mean square error to determine the accuracy of the model.
In order to better explain the technical solution proposed by the present invention, a more specific embodiment will be used for explanation.
Example 1
The converter to be measured is a 200t converter of a steelworks.
Firstly, a manganese content prediction model of the converter is established.
Step A1 is carried out, historical production data of the 200t converter is collected, the data are field actual data recorded by a steel mill, incomplete data are removed through preliminary screening, the production data of 1500 furnaces are collected as the historical production data, and the data type (also called as data characteristics) is: the method comprises the following steps of (1) carrying out carbon mass fraction of molten iron, silicon mass fraction of molten iron, manganese mass fraction of molten iron, phosphorus mass fraction of molten iron, sulfur mass fraction of molten iron, water quantity of scrap, desulfurization outlet temperature of molten iron, sublance process temperature, sublance process C mass fraction, sublance end point temperature, sublance end point C mass fraction, sublance end point O mass fraction, target tapping temperature, target end point C mass fraction, target end point Si mass fraction, target end point Mn mass fraction, target end point P mass fraction, target end point S mass fraction, light burned dolomite mass, dolomite mass and lime mass.
And (3) further performing step A2, and preprocessing the collected historical data. Features not related to manganese content are deleted, for example: smelting number, gun age, production date, steel code and the like; and according to the rule of data cleaning, processing the abnormal data with null value, zero value and larger discrete degree by adopting a substitution method.
And (3) performing step A3, selecting the last detection of the sublance, namely detecting the manganese mass fraction by TSO as the manganese content of the converter endpoint, performing normal test on each input characteristic data, and if the normal characteristics are not met, directly deleting the group of data in which the discrete value is located.
And (3) carrying out feature selection in the step A4 based on the manganese content and the historical production data, and judging feature data with great influence on the manganese content of the converter endpoint, namely related feature data.
In this embodiment, the model input variables are finally determined: the quality of molten iron comprises the following components of a mass fraction of molten iron carbon, a mass fraction of molten iron silicon, a mass fraction of molten iron manganese, a mass fraction of molten iron phosphorus, a mass fraction of molten iron, a mass fraction of scrap steel, a temperature of a sublance process, a mass fraction of a sublance process C, a temperature of a sublance end point, a mass fraction of a sublance end point C, a mass fraction of a sublance end point O, a target tapping temperature, a mass fraction of a target end point C, a mass fraction of a target end point Si, a mass fraction of a target end point Mn, a mass fraction of a target end point P, a mass of light burned dolomite, a mass of dolomite and a mass of lime. Screening relevant characteristic data of the 1500 heats, and dividing the relevant characteristic data into a training set and a testing set, wherein the ratio is 8:2; in this embodiment, a total of 1500 heats are selected as sample data, 1200 sets of data are randomly selected as training sets, and 300 sets of data are selected as test sets.
Inputting the characteristic data into a pre-constructed initial manganese content prediction model for training, wherein the super-parameter tuning specifically comprises the following steps: the number of trees or the number of iterations, defining the terms = 828; learning_rate: learning rate, defined learning_rate=0.025; depth: the depth of the tree, which may control the overfit, defines depth=3; leaf_arrival_events: calculating the iteration times when the leaf node value is calculated, and defining leaf_computation_iteration=1; l2_leaf_reg: the canonical parameter defines l2_leaf_reg=6.
Further, in order to realize accurate learning prediction capability of the model on training data in the learning process and good prediction capability on new data, the embodiment performsStep A6, selecting a determination coefficient (R 2 ) And Root Mean Square Error (RMSE) assessment model. Table 1 shows the evaluation index of the 200t converter Catboost manganese content prediction model.
Table 1:
as shown in Table 1, the predicted hit rate of Mn content was 94.7% within.+ -. 0.015 error range.
As shown in fig. 3, fig. 3 is a graph of a comparison of predicted and actual manganese contents of the 200t converter.
R of Catboost manganese content prediction model in the embodiment of the invention 2 Reaches above 0.66 and has lower RMSE; the fitting effect of the predicted value to the actual value accords with the expectation, the manganese content is within the range of +/-0.015, and the hit rate reaches more than 94%, so that the predicted value of the Catboost manganese content prediction model is very close to the actual value.
According to the method for predicting the end point manganese content of the converter, firstly, a converter end point manganese content prediction model is established based on converter smelting historical data according to a metallurgical principle. The obtained prediction model can directly calculate the manganese content of a smelting end point in the complex physicochemical reaction process of converter smelting, thereby ensuring the accurate calculation and control of alloy materials in the tapping process of the converter. Proved by verification, the converter endpoint manganese content prediction model can accurately predict the manganese content in the converter process, so that the charging is controlled more accurately.
In the description of the present invention, it should be understood that the terms "first," "second," and the like are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. In the description of the present invention, the meaning of "a plurality" is two or more, unless explicitly defined otherwise.
In the present invention, unless explicitly specified and limited otherwise, the terms "mounted," "connected," "secured," and the like are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally formed; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium; may be a communication between two elements or an interaction between two elements. The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art according to the specific circumstances.
In the description of the present specification, the terms "one embodiment," "some embodiments," "examples," "particular examples," or "some examples," etc., refer to particular features, structures, materials, or characteristics described in connection with the embodiment or example as being included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, the different embodiments or examples described in this specification and the features of the different embodiments or examples may be combined and combined by those skilled in the art without contradiction.
While embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the invention, and that alterations, modifications, substitutions and variations may be made in the above embodiments by those skilled in the art within the scope of the invention.
Claims (10)
1. A method for predicting a converter endpoint manganese content for use in predicting a manganese content at a converter steelmaking endpoint, comprising:
s1, collecting current production data of a converter to be tested, wherein the current production data comprises the following steps: the method comprises the following steps of (1) carrying out carbon mass fraction of molten iron, silicon mass fraction of molten iron, manganese mass fraction of molten iron, phosphorus mass fraction of molten iron, sulfur mass fraction of molten iron, water quantity of scrap, desulfurization outlet temperature of molten iron, sublance process temperature, sublance process C mass fraction, sublance end point temperature, sublance end point C mass fraction, sublance end point O mass fraction, target tapping temperature, target end point C mass fraction, target end point Si mass fraction, target end point Mn mass fraction, target end point P mass fraction, target end point S mass fraction, light burned dolomite mass, lime mass and converter end point residual Mn content;
s2, inputting the current production data into a trained manganese content prediction model, and predicting the final manganese content of the converter event to be detected;
the manganese content prediction model is a manganese content prediction model obtained by training an initial manganese content prediction model based on historical production data of the converter.
2. The converter endpoint manganese content prediction method according to claim 1, further comprising, after S2:
and adjusting the current production data based on the predicted manganese content until the predicted manganese content can meet the control target of the end point alloy composition of the steel grade.
3. The method for predicting the endpoint manganese content of a converter according to claim 1, further comprising, prior to S1:
s0, training a pre-constructed initial manganese content prediction model based on historical production data of the converter to be tested to obtain a manganese content prediction model, wherein the method comprises the following steps of:
a1, collecting historical production data of the converter to be tested, wherein the historical production data is a set of complete production data of each single production of the converter to be tested;
a2, carrying out data preprocessing on the historical production data;
a3, selecting TSO corresponding to each single production of the converter to detect the manganese content as manganese content based on historical production data subjected to data preprocessing, and checking whether each piece of continuous characteristic data input by the converter accords with statistical normal distribution;
a4, judging characteristic data affecting the manganese content in the historical production data by means of metallurgical mechanism analysis and a Pearson correlation coefficient method;
and A5, screening the characteristic data of the historical production data, and inputting the characteristic data of each single production and the manganese content of the current production into a pre-constructed initial manganese content prediction model for training to obtain a manganese content prediction model.
4. The method for predicting the endpoint manganese content of a converter according to claim 3, wherein in the step A1, the category of the complete production data of each single production includes:
smelting number, gun age, production date, steel code, molten iron carbon fraction, molten iron silicon mass fraction, molten iron manganese mass fraction, molten iron phosphorus mass fraction, molten iron sulfur mass fraction, molten iron quantity, scrap steel quantity, molten iron desulfurization outlet temperature, sublance process C mass fraction, sublance endpoint temperature, sublance endpoint C mass fraction, sublance endpoint O mass fraction, target tapping temperature, target endpoint C mass fraction, target endpoint Si mass fraction, target endpoint Mn mass fraction, target endpoint P mass fraction, target endpoint S mass fraction, light burned dolomite mass, lime mass.
5. The method for predicting the endpoint manganese content of a converter according to claim 4, wherein A2 comprises:
a21, cleaning the historical production data, and analyzing and deleting obvious invalid characteristic values by using a metallurgical mechanism:
the invalid characteristic value includes; smelting number, gun age, production date and steel code;
a22, carrying out outlier replacement on the cleaned historical production data;
the outlier includes: zero value, null value, discrete value.
6. The method for predicting the endpoint manganese content of a converter according to claim 5, wherein a22 is specifically:
and replacing the abnormal value by taking an average value of the historical production data which is in the same category as the abnormal value in the historical production data as a replacement value.
7. The method for predicting the endpoint manganese content of a converter according to claim 3, wherein step A3 comprises:
a31, based on the historical production data, selecting the final detection of the sublance, namely, detecting the mass fraction of manganese by TSO as the endpoint manganese content of the converter;
a32, carrying out normal test on each input characteristic data, and if the normal characteristics are not met, directly deleting the group of data with the discrete value.
8. The method for predicting the endpoint manganese content of a converter according to claim 3, wherein in the step A4, the predefined correlation calculation formula is:
n is the total heat of the historical production data, x i Is the actual value of the characteristic data and,is the mean value of the characteristic data, y i Is the actual value of the manganese content,/->Is the average value of manganese content;
the characteristic data influencing the manganese content comprise:
the quality of molten iron comprises the following components of a mass fraction of molten iron carbon, a mass fraction of molten iron silicon, a mass fraction of molten iron manganese, a mass fraction of molten iron phosphorus, a mass fraction of molten iron, a mass fraction of scrap steel, a temperature of a sublance process, a mass fraction of a sublance process C, a temperature of a sublance end point, a mass fraction of a sublance end point C, a mass fraction of a sublance end point O, a target tapping temperature, a mass fraction of a target end point C, a mass fraction of a target end point Si, a mass fraction of a target end point Mn, a mass fraction of a target end point P, a mass of light burned dolomite, a mass of dolomite and a mass of lime.
9. The method for predicting the endpoint manganese content of a converter according to claim 8, wherein step A5 comprises:
a51, screening and marking the historical production data to obtain the characteristic data of each single production and the label corresponding to the characteristic data;
a52, taking the characteristic data of each single production and the manganese content of the current production as sample data, and inputting the sample data into the initial manganese content prediction model; the sample data comprises a training set and a testing set; the test set is used for testing the manganese content prediction model;
a53, optimizing the super parameters of the initial manganese content prediction model;
a54, training the initial manganese content prediction model through the training set to obtain a manganese content prediction model.
10. The converter endpoint manganese content prediction method according to claim 9, further comprising, after S5:
s6, inputting the characteristic data of each single production in the test set into the manganese content prediction model to generate the predicted manganese content of the secondary production;
calculating a determination coefficient and a root mean square error of the predicted manganese content of the current production based on the predicted manganese content of the current production and the manganese content of the same single production in the test set;
testing the manganese content prediction model based on the decision coefficient and root mean square error;
determining coefficient R 2 The calculation formula is as follows:
the root mean square error RMSE calculation formula is as follows:
n is the total sample amount of the test set, f i For the predicted value of the model output, y i For the actual value of the characteristic data sample, y i Is the characteristic data sample mean value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311627914.6A CN117612651A (en) | 2023-11-30 | 2023-11-30 | Method for predicting manganese content of converter endpoint |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311627914.6A CN117612651A (en) | 2023-11-30 | 2023-11-30 | Method for predicting manganese content of converter endpoint |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117612651A true CN117612651A (en) | 2024-02-27 |
Family
ID=89955861
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311627914.6A Pending CN117612651A (en) | 2023-11-30 | 2023-11-30 | Method for predicting manganese content of converter endpoint |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117612651A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117851816A (en) * | 2024-03-07 | 2024-04-09 | 北京科技大学 | Method for predicting steelmaking end point component of converter |
-
2023
- 2023-11-30 CN CN202311627914.6A patent/CN117612651A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117851816A (en) * | 2024-03-07 | 2024-04-09 | 北京科技大学 | Method for predicting steelmaking end point component of converter |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109685289B (en) | Method, device and system for forward prediction of blast furnace conditions | |
| CN110066895B (en) | Stacking-based blast furnace molten iron quality interval prediction method | |
| CN117612651A (en) | Method for predicting manganese content of converter endpoint | |
| CN109935280B (en) | Blast furnace molten iron quality prediction system and method based on ensemble learning | |
| CN114611844B (en) | Method and system for determining alloy addition amount in converter tapping process | |
| US8862438B2 (en) | Method for improving the prediction of polymer properties | |
| CN110400231B (en) | Failure rate estimation method for electric energy metering equipment based on weighted nonlinear Bayes | |
| CN109359723A (en) | Based on the converter terminal manganese content prediction technique for improving regularization extreme learning machine | |
| CN105807741A (en) | Industrial production flow prediction method | |
| CN102090705A (en) | Method for improving tobacco-drying process capability | |
| CN116306272A (en) | Converter heat loss rate prediction method based on big data | |
| Takalo-Mattila et al. | Explainable steel quality prediction system based on gradient boosting decision trees | |
| CN114462550A (en) | Method for predicting water consumption of water supply network users by using random forest and probability density function | |
| CN113517037B (en) | Method and system for predicting sintering ore FeO by fusing data and knowledge | |
| CN107563656B (en) | Method for evaluating running state of gold hydrometallurgy cyaniding leaching process | |
| WO2014167982A1 (en) | Correction device, correction method and steel refining method | |
| CN106153551A (en) | Converter steel-smelting molten steel carbon content based on SVM online Real-time and Dynamic Detection system | |
| CN115927770A (en) | Real-time tracking method for tapping state of blast furnace | |
| CN112862180A (en) | Denitration system inlet NOx concentration prediction method | |
| CN110688309A (en) | General software test case sequence quantitative evaluation method based on requirements | |
| CN117553921B (en) | Converter molten steel temperature prediction method, system, terminal and storage medium | |
| CN117788021B (en) | Method and system for rapidly screening quality of coal carbon emission data of furnace charging of emission control enterprise | |
| CN118013426B (en) | Steel product inclusion causal effect optimization method, system, electronic equipment and medium | |
| CN115034370B (en) | BP network model-based method for predicting hearth activity of blast furnace | |
| RU2804378C1 (en) | Artificial intelligence system and method for mixing raw material during gasification |
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 |