CN117540663A - Blast furnace hearth high-temperature melt viscosity prediction method and system based on neural network - Google Patents
Blast furnace hearth high-temperature melt viscosity prediction method and system based on neural network Download PDFInfo
- Publication number
- CN117540663A CN117540663A CN202410031144.7A CN202410031144A CN117540663A CN 117540663 A CN117540663 A CN 117540663A CN 202410031144 A CN202410031144 A CN 202410031144A CN 117540663 A CN117540663 A CN 117540663A
- Authority
- CN
- China
- Prior art keywords
- neural network
- temperature melt
- blast furnace
- temperature
- viscosity
- 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.)
- Granted
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000003062 neural network model Methods 0.000 claims abstract description 57
- 238000012549 training Methods 0.000 claims abstract description 50
- 238000011156 evaluation Methods 0.000 claims abstract description 24
- 230000000694 effects Effects 0.000 claims abstract description 21
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 14
- 239000007788 liquid Substances 0.000 claims abstract description 14
- 239000007790 solid phase Substances 0.000 claims abstract description 14
- 238000012360 testing method Methods 0.000 claims abstract description 13
- 238000012795 verification Methods 0.000 claims abstract description 12
- 238000001556 precipitation Methods 0.000 claims abstract description 10
- 230000006870 function Effects 0.000 claims description 8
- 239000007791 liquid phase Substances 0.000 claims description 8
- 238000010200 validation analysis Methods 0.000 claims description 8
- 238000002790 cross-validation Methods 0.000 claims description 5
- 239000000203 mixture Substances 0.000 claims description 5
- 210000002569 neuron Anatomy 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 2
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 16
- 239000000155 melt Substances 0.000 description 14
- 229910052742 iron Inorganic materials 0.000 description 8
- 229910052751 metal Inorganic materials 0.000 description 7
- 239000002184 metal Substances 0.000 description 5
- 238000003723 Smelting Methods 0.000 description 4
- 229910045601 alloy Inorganic materials 0.000 description 4
- 239000000956 alloy Substances 0.000 description 4
- 229910052755 nonmetal Inorganic materials 0.000 description 4
- 239000002893 slag Substances 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000003628 erosive effect Effects 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000011478 gradient descent method Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 229910000838 Al alloy Inorganic materials 0.000 description 1
- GYHNNYVSQQEPJS-UHFFFAOYSA-N Gallium Chemical compound [Ga] GYHNNYVSQQEPJS-UHFFFAOYSA-N 0.000 description 1
- 229910000831 Steel Inorganic materials 0.000 description 1
- ATJFFYVFTNAWJD-UHFFFAOYSA-N Tin Chemical compound [Sn] ATJFFYVFTNAWJD-UHFFFAOYSA-N 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 239000010883 coal ash Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000005496 eutectics Effects 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 229910052733 gallium Inorganic materials 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000001764 infiltration Methods 0.000 description 1
- 230000008595 infiltration Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000009533 lab test Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000008018 melting Effects 0.000 description 1
- 238000002844 melting Methods 0.000 description 1
- 238000000329 molecular dynamics simulation Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000000704 physical effect Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21B—MANUFACTURE OF IRON OR STEEL
- C21B5/00—Making pig-iron in the blast furnace
- C21B5/006—Automatically controlling the process
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21B—MANUFACTURE OF IRON OR STEEL
- C21B7/00—Blast furnaces
- C21B7/24—Test rods or other checking devices
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0499—Feedforward networks
-
- 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
-
- 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
- G16C20/70—Machine learning, data mining or chemometrics
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21B—MANUFACTURE OF IRON OR STEEL
- C21B2300/00—Process aspects
- C21B2300/04—Modeling of the process, e.g. for control purposes; CII
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/08—Thermal analysis or thermal optimisation
Abstract
The invention belongs to the technical field of high-temperature melt property prediction, in particular to a method and a system for predicting the viscosity of a high-temperature melt of a blast furnace hearth based on a neural network, wherein the temperature, components, solid phase precipitation and liquid state structure of the high-temperature melt of a historical data set are taken as input variables, and the viscosity of the high-temperature melt is taken as output variables; selecting an SGD algorithm as an optimizer, dividing a neural network training set, a verification set and a test set, and constructing a neural network model comprising an input layer, a hidden layer and an output layer; adjusting the number M of hidden layers and the number N of network nodes, and training a neural network model; and selecting the relative error, the absolute error and the decision coefficient as evaluation indexes, and obtaining a neural network model with good training effect based on the discrimination standard of each evaluation index to realize the prediction of the high-temperature melt viscosity of the hearth of the blast furnace to be tested. The invention solves the problem that the existing semi-empirical model can not well simulate the viscosity of the multicomponent melt, and provides a new idea for predicting the viscosity at high temperature.
Description
Technical Field
The invention relates to the technical field of high-temperature melt property prediction, in particular to a blast furnace hearth high-temperature melt viscosity prediction method and system based on a neural network.
Background
The high-temperature melt is an important product in the steel smelting process, and with the continuous rising of smelting intensity, serious accidents such as burning through of a plurality of blast furnace hearth and the like occur at home and abroad in recent years, so that great economic loss and even casualties are caused. In the anatomical investigation of different vertical furnaces, the infiltration erosion and the circulation erosion of the high-temperature melt in the hearth to the refractory are found to be the main reasons for increasing the erosion of the refractory and reducing the service life of the hearth. However, when the melt viscosity is too high, slag and iron are difficult to separate, the coupling flow of slag and iron and the mass transfer of elements between interfaces are limited, the yield is reduced, and the fuel consumption is increased. Therefore, the prediction and regulation of the high-temperature melt viscosity of the blast furnace hearth are key to realizing safe low-carbon smelting.
Currently, there are three main ways to study the viscosity behavior of high temperature melts: 1. laboratory viscosity measurement mainly comprises a rotation method, a vibration method, a falling body method, a parallel plate method, a capillary method and the like; 2. software calculation, namely performing software simulation by using FactSage, thermalco thermodynamic software and performing simulation by using calculation means such as molecular dynamics; 3. an online prediction model is usually established by a theoretical equation to describe the changes in physical properties of the melt caused by temperature, composition, etc. The disadvantage of the above embodiment 1 is that: no matter what way is adopted to simulate the flow behavior of the melt, the flow behavior of the melt cannot be completely consistent with the state of the melt in the high-temperature smelting container, and in addition, the interference of environmental factors is difficult to be eliminated in the laboratory experiment process, so that the experiment precision is difficult to control. Mode 2 does not need to carry out experiments, but the prediction result is limited by the range of a software database and the selection of original parameters, so that the prediction of the melt viscosity under the complex influence of multiple factors can not be fully reproduced well. The prediction model of partial melt viscosity in the mode 3 can relate to certain thermodynamic parameters of a system or certain physical and chemical properties of pure components, can keep higher consistency with experimental results when being used for predicting alloy melt, and can be suitable for various alloy systems such as eutectic, metacrystalloid and the like. However, these models are mostly built for alloy melts of metal-metal systems, and when the models are applied to viscosity prediction of high-temperature melts containing semi-metal or non-metal components, the calculated values of the models which are originally good in applicability deviate from the true values greatly, so that the accurate prediction of the viscosity of the high-temperature melts containing the semi-metal or non-metal components cannot be realized. In view of the foregoing, it is desirable to find a method for predicting the high temperature melt viscosity of a blast furnace hearth that is compatible with the actual high temperature melt flow behavior and is easy to operate.
Disclosure of Invention
Aiming at the characteristics of the high-temperature melt containing semi-metal or nonmetal components, the invention comprehensively considers various factors such as melt temperature, components, solid phase particles, liquid state structures and the like, and provides a new thought for predicting and controlling the performance of the high-temperature melt.
In order to solve the technical problems, according to one aspect of the present invention, the following technical solutions are provided:
a blast furnace hearth high-temperature melt viscosity prediction method based on a neural network comprises the following steps:
s1, taking the temperature, the components, the solid phase precipitation and the liquid state structure of a high-temperature melt in a historical data set as input variables and the viscosity of the high-temperature melt as output variables; dividing a neural network training set, a verification set and a test set, and constructing a neural network model comprising an input layer, a hidden layer and an output layer;
s2, adjusting the number M of hidden layers and the number N of network nodes, and training a neural network model; selecting Relative Error (RE), absolute error (MAE), determining coefficient (R 2 ) As an evaluation index, acquiring a neural network model with good training effect based on the discrimination standard of each evaluation index;
and S3, predicting the high-temperature melt viscosity of the hearth of the blast furnace to be tested by adopting a neural network model with good training effect.
As a preferable scheme of the blast furnace hearth high-temperature melt viscosity prediction method based on the neural network, the invention comprises the following steps: in the step S1, the temperature, composition, solid phase precipitation and liquid structure of the high-temperature melt include: c content, si content, mn content, P content, S content, ti content, melt temperature, solid-liquid phase ratio, free volume ratio, atomic cluster volume ratio, number of atomic clusters and liquidus temperature.
As a preferable scheme of the blast furnace hearth high-temperature melt viscosity prediction method based on the neural network, the invention comprises the following steps: in step S1, the training set, the validation set and the test set respectively account for 75%, 25% and 25% of the entire data set.
As a preferable scheme of the blast furnace hearth high-temperature melt viscosity prediction method based on the neural network, the invention comprises the following steps: in the step S1, the input variable is normalized, two hidden layers are designed, and the first and second hidden layer activation functions select ReLU (Rectified Linear Units) functions, i.e. rectifying linear unit functions.
As a preferable scheme of the blast furnace hearth high-temperature melt viscosity prediction method based on the neural network, the invention comprises the following steps: in the step S1, a K-fold Cross Validation method (Cross Validation) is selected to set a Validation set, and a random gradient descent method (SGD, stochastic gradient descent) is selected to perform optimization adjustment on the model super-parameter setting (hyperparameter compile).
As a preferable scheme of the blast furnace hearth high-temperature melt viscosity prediction method based on the neural network, the invention comprises the following steps: in the step S1, the model hyper-parameters include the number of hidden neurons and the number of iterations.
As a preferable scheme of the blast furnace hearth high-temperature melt viscosity prediction method based on the neural network, the invention comprises the following steps: in the step S2, each evaluation index is calculated in the following manner:
wherein RE is the relative error; MAE is absolute error; r is R 2 To determine coefficients; SS (support System) res Is the sum of squares of the differences between the true and predicted values; η (eta) Predictive value Is the predicted value of the viscosity of the high-temperature melt, eta True value Is the true value of the viscosity of the high-temperature melt, SS tot Is the square difference between the true value and the predicted value.
As a preferable scheme of the blast furnace hearth high-temperature melt viscosity prediction method based on the neural network, the invention comprises the following steps: in the step S2, the MAE converges and is lower than 0.4, RE is less than 10%, R 2 And 0.9, the neural network model is considered well trained.
In order to solve the above technical problems, according to another aspect of the present invention, the following technical solutions are provided:
a neural network-based blast furnace hearth high temperature melt viscosity prediction system, comprising:
the neural network model building module is used for taking the components, temperature, liquid phase parameters and solid phase parameters of the high-temperature melt in the historical data set as input variables and taking the viscosity of the high-temperature melt as output variables; dividing a neural network training set, a verification set and a test set, and constructing a neural network model comprising an input layer, a hidden layer and an output layer;
the neural network model training module is used for adjusting the number M of hidden layers and the number N of network nodes and training a neural network model; selecting Relative Error (RE), absolute error (MAE), determining coefficient (R 2 ) As an evaluation index, acquiring a neural network model with good training effect based on the discrimination standard of each evaluation index;
and the high-temperature melt viscosity prediction module is used for predicting the high-temperature melt viscosity of the hearth of the blast furnace to be detected by adopting a neural network model with good training effect.
The beneficial effects of the invention are as follows:
the invention provides a blast furnace hearth high-temperature melt viscosity prediction method and a system based on a neural network, wherein the temperature, the composition, the solid phase precipitation and the liquid state structure of a high-temperature melt in a historical data set are used as input variables, and the viscosity of the high-temperature melt is used as output variables; selecting an SGD algorithm as an optimizer, dividing a neural network training set, a verification set and a test set, and constructing a neural network model comprising an input layer, a hidden layer and an output layer; adjusting the number M of hidden layers and the number N of network nodes, and training a neural network model; the relative error, the absolute error and the decision coefficient are selected as evaluation indexes, and a neural network model with good training effect is obtained based on the discrimination standard of each evaluation index; and a neural network model with good training effect is adopted to realize the prediction of the high-temperature melt viscosity of the hearth of the blast furnace to be tested. The invention solves the problem that the existing semi-empirical model can not well simulate the viscosity of the multi-component melt, provides a new thought for predicting the viscosity of the high-temperature melt, and provides an important basis for controlling the performance of the high-temperature melt.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to the structures shown in these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of the neural network algorithm of the present invention.
FIG. 2 is a schematic diagram of the verification result of the high temperature melt viscosity neural network model of the present invention.
FIG. 3 is a schematic diagram showing the comparison of the predicted value and the actual value of the viscosity of the high-temperature melt of the blast furnace to be tested.
The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
The following description will be made clearly and fully with reference to the technical solutions in the embodiments, and it is apparent that the described embodiments are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
With the development of computer technology and artificial intelligence technology, many excellent machine learning algorithms are widely used in the field of metallurgical engineering research. The BP neural network model is the most commonly applied, and is used as the most traditional classical neural network model in the artificial intelligence algorithm, and the algorithm is easy to be combined with other algorithms such as a genetic algorithm and the like to form the harvester. The BP neural network is suitable for the characteristics of multiple characteristic parameters, multiple coupling relations and multiple data sets, is very suitable for grasping the uncertain rules in the metallurgical technology and predicting the multi-factor guiding result, and has been applied to the prediction of molten iron components of a blast furnace, the prediction of the melting characteristic temperature of coal ash, the prediction of molten iron quality and temperature, the prediction of slag component range, the prediction of slag viscosity and the like. At present, the viscosity of alloy melts such as liquid aluminum alloy, liquid gallium, liquid tin base and the like is predicted, but a high-temperature melt system of a blast furnace hearth is a melt containing semi-metal or non-metal components, and the existing model cannot be suitable for the high-temperature melt of the blast furnace hearth. Therefore, the invention provides a blast furnace hearth high-temperature melt viscosity prediction method and a blast furnace hearth high-temperature melt viscosity prediction system based on a neural network, which have the following advantages:
(1) The method for predicting the high-temperature melt viscosity of the blast furnace hearth based on the neural network algorithm has the advantages of a traditional viscosity prediction model, can accurately and truly predict the flow condition of the high-temperature melt by extracting the physical parameters of the existing melt, avoids complicated deduction of various parameters in the traditional viscosity model, and has high prediction precision and high calculation speed.
(2) The invention can solve the problem that the precision of the viscosity model is tested when the existing model is replaced aiming at a high-temperature melt component system, and solves the problem of viscosity prediction of the melt containing the semi-metallic element and the nonmetallic element simultaneously by taking the temperature, the component, the solid phase precipitation and the liquid structure (namely, the C content, the Si content, the Mn content, the P content, the S content, the Ti content, the melt temperature, the solid-liquid phase ratio, the free volume ratio, the atomic cluster number and the liquidus temperature) of the high-temperature melt of the historical data set as input variables.
(3) According to the method, the training effect of the neural network model is evaluated by calculating a plurality of evaluation indexes, the neural network model with good training effect is obtained according to the evaluation standards of the indexes, the neural network model with good training effect is adopted to realize the prediction of the high-temperature melt viscosity of the hearth of the blast furnace to be tested, and the requirement of good prediction precision can be ensured.
As shown in fig. 1, the embodiment of the invention provides a blast furnace hearth high-temperature melt viscosity prediction method based on a neural network, which comprises the following steps:
s1, taking the temperature, the components, the solid phase precipitation and the liquid state structure of a high-temperature melt in a historical data set as input variables and the viscosity of the high-temperature melt as output variables; dividing a neural network training set, a verification set and a test set, and constructing a neural network model comprising an input layer, a hidden layer and an output layer;
s2, adjusting hidingThe number of layers M and the number of network nodes N are used for training a neural network model through a four-layer feed-forward algorithm by utilizing a training set; using the validation set, a Relative Error (RE), an absolute error (MAE), a decision coefficient (R 2 ) As an evaluation index, acquiring a neural network model with good training effect based on the discrimination standard of each evaluation index;
and S3, predicting the high-temperature melt viscosity of the hearth of the blast furnace to be tested by adopting a neural network model with good training effect.
In one embodiment of the present invention, in the step S1, the high-temperature melt is molten iron.
In one embodiment of the present invention, in the step S1, the temperature, composition, solid phase precipitation, and liquid structure of the high temperature melt include: c content, si content, mn content, P content, S content, ti content, melt temperature, solid-liquid phase ratio, free volume ratio (i.e., the volume occupied by atoms in the high temperature melt that do not form clusters), cluster volume ratio, number of clusters, and liquidus temperature.
In one embodiment of the present invention, in the step S1, the training set and the test set respectively account for 75% and 25% of the entire data set, and the validation set is one third, i.e., 25%, of the random decimation from the training set.
In one embodiment of the present invention, in the step S1, the input variable is normalized, and two hidden layers are designed, and the first and second hidden layer activation functions select ReLU (Rectified Linear Units) functions, i.e., rectifying linear unit functions.
In one embodiment of the present invention, in the step S1, a K-fold Cross Validation method (Cross Validation) is selected to set the Validation set, and a random gradient descent method (SGD, stochastic gradient descent) is selected to optimally adjust the model super-parameter setting (hyperparameter compile).
In one embodiment of the present invention, in the step S1, the model hyper-parameters include the number of hidden neurons and the number of iterations. In this embodiment, the number of hidden neurons is determined to be 64, and the number of iterations is 1000.
In one embodiment of the present invention, in the step S2, each evaluation index is calculated as follows:
wherein RE is the relative error; MAE is absolute error; r is R 2 To determine coefficients; SS (support System) res Is the sum of squares of the differences between the true and predicted values; η (eta) Predictive value Is the predicted value of the viscosity of the high-temperature melt, eta True value Is the true value of the viscosity of the high-temperature melt, SS tot Is the square difference between the true value and the predicted value.
In one embodiment of the present invention, in the step S2, the MAE converges and is lower than 0.4, RE < 10%, R 2 And 0.9, the neural network model is considered well trained.
In one embodiment of the present invention, in the step S3, when the neural network model with good training effect is used to implement viscosity prediction of the high-temperature melt of the blast furnace hearth, the physical parameters of the high-temperature melt of the blast furnace hearth to be measured may be all input variables or part of the input variables during construction of the neural network model.
The embodiment of the invention also provides a blast furnace hearth high-temperature melt viscosity prediction system based on the neural network, which comprises the following steps:
the neural network model building module is used for taking the components, temperature, liquid phase parameters and solid phase parameters of the high-temperature melt in the historical data set as input variables and taking the viscosity of the high-temperature melt as output variables; dividing a neural network training set, a verification set and a test set, and constructing a neural network model comprising an input layer, a hidden layer and an output layer;
neural network model training module for adjusting hidingThe number of layers M and the number of network nodes N, and training a neural network model through a four-layer feed-forward algorithm; selecting Relative Error (RE), absolute error (MAE), determining coefficient (R 2 ) As an evaluation index, acquiring a neural network model with good training effect based on the discrimination standard of each evaluation index;
and the high-temperature melt viscosity prediction module is used for predicting the high-temperature melt viscosity of the hearth of the blast furnace to be detected by adopting a neural network model with good training effect.
The neural network-based blast furnace hearth high-temperature melt viscosity prediction method and system are used for viscosity prediction of molten iron containing molten iron in a blast furnace hearth of a certain factory, and a historical data set is used as input variables (C content, si content, mn content, P content, S content, ti content, melt temperature, solid-liquid phase ratio, free volume ratio, atomic cluster number and liquidus temperature) and is shown in the following table:
based on the input variables in the table above, taking the viscosity of the high temperature melt as the output variable; dividing a neural network training set, a verification set and a test set, and constructing a neural network model comprising an input layer, a hidden layer and an output layer;
based on the neural network model constructed by the blast furnace hearth high-temperature melt viscosity prediction method based on the neural network, adjusting the number M of hidden layers and the number N of network nodes, and training the neural network model by using a training set through a four-layer forward feedback algorithm; using the validation set, a Relative Error (RE), an absolute error (MAE), a decision coefficient (R 2 ) Verification as evaluation indexAs shown in FIG. 2, wherein MAE converges and approaches 0.25, RE < 10% is 100%, RE < 5% is 82%, RE < 3% is 73%, and in the range of 5-12 mPa.s, the predicted viscosity value and the true viscosity value have good correlation (R 2 =0.95), showing that the blast furnace hearth high-temperature melt viscosity neural network model established in the above embodiment of the present invention has good prediction accuracy.
The molten iron samples of three iron-making blast furnace hearths are respectively selected as research objects, namely a 1# sample, a 2# sample and a 3# sample, and part of physical parameters are shown in the following table:
viscosity predictions are performed on the 1# sample, the 2# sample and the 3# sample based on the well-trained neural network model, and viscosity predicted values and true values of the samples are respectively compared at 1200-1450 ℃, and the results are shown in fig. 3. The maximum predicted deviation of the No. 1 sample at 1200 ℃ is 9.9%, the predicted deviation of the No. 2 sample is not more than 4.2%, and the predicted deviation of the No. 3 sample at 1350 ℃ is only 1.3%. The relative prediction error of about 56% is less than 5%, wherein the data point proportion of the relative prediction error of less than 3% is 39%, and the relative overall prediction error of the data is less than 10%, so that the prediction value and the true value of the prediction method have higher consistency, namely the accuracy of the high-temperature melt viscosity prediction method for the blast furnace hearth based on the neural network is high.
The invention takes the temperature, the components, the solid phase precipitation and the liquid structure of the high-temperature melt as input variables and the viscosity of the high-temperature melt as output variables; selecting an SGD algorithm as an optimizer, dividing a neural network training set, a verification set and a test set, and constructing a neural network model comprising an input layer, a hidden layer and an output layer; adjusting the number M of hidden layers and the number N of network nodes, and training a neural network model; the relative error, the absolute error and the decision coefficient are selected as evaluation indexes, and a neural network model with good training effect is obtained based on the discrimination standard of each evaluation index; and a neural network model with good training effect is adopted to realize the prediction of the high-temperature melt viscosity of the blast furnace hearth. The invention solves the problem that the existing semi-empirical model can not well simulate the viscosity of a multi-component melt (containing semi-metallic elements and nonmetallic elements), provides a new thought for predicting the viscosity of a high-temperature melt, and provides an important basis for controlling the performance of the high-temperature melt.
The foregoing description is only of the preferred embodiments of the present invention and is not intended to limit the scope of the invention, and all equivalent structural changes made by the content of the present invention or direct/indirect application in other related technical fields are included in the scope of the present invention.
Claims (9)
1. A blast furnace hearth high-temperature melt viscosity prediction method based on a neural network is characterized by comprising the following steps:
s1, taking the temperature, the components, the solid phase precipitation and the liquid state structure of a high-temperature melt in a historical data set as input variables and the viscosity of the high-temperature melt as output variables; dividing a neural network training set, a verification set and a test set, and constructing a neural network model comprising an input layer, a hidden layer and an output layer;
s2, adjusting the number M of hidden layers and the number N of network nodes, and training a neural network model; the relative error, the absolute error and the decision coefficient are selected as evaluation indexes, and a neural network model with good training effect is obtained based on the discrimination standard of each evaluation index;
and S3, predicting the high-temperature melt viscosity of the hearth of the blast furnace to be tested by adopting a neural network model with good training effect.
2. The neural network-based blast furnace hearth high-temperature melt viscosity prediction method according to claim 1, wherein in the step S1, the temperature, composition, solid phase precipitation, and liquid structure of the high-temperature melt include: c content, si content, mn content, P content, S content, ti content, melt temperature, solid-liquid phase ratio, free volume ratio, atomic cluster volume ratio, number of atomic clusters and liquidus temperature.
3. The neural network-based blast furnace hearth high temperature melt viscosity prediction method according to claim 1, wherein in the step S1, the training set, the validation set and the test set respectively account for 75%, 25% and 25% of the entire data set.
4. The method for predicting the high-temperature melt viscosity of a blast furnace hearth based on a neural network according to claim 1, wherein in the step S1, the input variables are normalized, two hidden layers are designed, and the first and second hidden layer activation functions select the ReLU function.
5. The neural network-based blast furnace hearth high-temperature melt viscosity prediction method according to claim 1, wherein in the step S1, a K-fold cross validation method is selected to set a validation set, and an SGD algorithm is selected to perform optimization adjustment on model super-parameter setting.
6. The method for predicting the high-temperature melt viscosity of a blast furnace hearth based on a neural network according to claim 5, wherein in the step S1, the model hyper-parameters include the number of hidden neurons and the number of iterations.
7. The neural network-based blast furnace hearth high-temperature melt viscosity prediction method according to claim 1, wherein in the step S2, each evaluation index is calculated by the following method:
wherein RE is the relative error; MAE is absolute error; r is R 2 To determine coefficients; SS (support System) res Is the sum of squares of the differences between the true and predicted values; η (eta) Predictive value Is the predicted value of the viscosity of the high-temperature melt, eta True value Is the true value of the viscosity of the high-temperature melt, SS tot Is the square difference between the true value and the predicted value.
8. The neural network-based blast furnace hearth high temperature melt viscosity prediction method according to claim 1, wherein in the step S2, MAE converges and falls below 0.4, re < 10%, R 2 And 0.9, the neural network model is considered well trained.
9. A neural network-based blast furnace hearth high temperature melt viscosity prediction system for implementing the neural network-based blast furnace hearth high temperature melt viscosity prediction method according to any one of claims 1 to 8, comprising:
the neural network model building module is used for taking the components, temperature, liquid phase parameters and solid phase parameters of the high-temperature melt in the historical data set as input variables and taking the viscosity of the high-temperature melt as output variables; dividing a neural network training set, a verification set and a test set, and constructing a neural network model comprising an input layer, a hidden layer and an output layer;
the neural network model training module is used for adjusting the number M of hidden layers and the number N of network nodes and training a neural network model; the relative error, the absolute error and the decision coefficient are selected as evaluation indexes, and a neural network model with good training effect is obtained based on the discrimination standard of each evaluation index;
and the high-temperature melt viscosity prediction module is used for predicting the high-temperature melt viscosity of the hearth of the blast furnace to be detected by adopting a neural network model with good training effect.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410031144.7A CN117540663B (en) | 2024-01-09 | 2024-01-09 | Blast furnace hearth high-temperature melt viscosity prediction method and system based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410031144.7A CN117540663B (en) | 2024-01-09 | 2024-01-09 | Blast furnace hearth high-temperature melt viscosity prediction method and system based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117540663A true CN117540663A (en) | 2024-02-09 |
| CN117540663B CN117540663B (en) | 2024-04-23 |
Family
ID=89794305
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410031144.7A Active CN117540663B (en) | 2024-01-09 | 2024-01-09 | Blast furnace hearth high-temperature melt viscosity prediction method and system based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117540663B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107526927A (en) * | 2017-08-10 | 2017-12-29 | 东北大学 | A kind of online robust flexible measurement method of blast-melted quality |
| CN112163380A (en) * | 2020-09-24 | 2021-01-01 | 哈尔滨锅炉厂有限责任公司 | System and method for predicting furnace oxygen concentration based on numerical simulation neural network |
| US20210365784A1 (en) * | 2020-05-19 | 2021-11-25 | Zhejiang University | Method for deriving fault diagnosis rules of blast furnace based on deep neural network |
| CN116842402A (en) * | 2023-09-01 | 2023-10-03 | 北京科技大学 | Blast furnace abnormal furnace condition detection method based on stable characteristic extraction of twin neural network |
-
2024
- 2024-01-09 CN CN202410031144.7A patent/CN117540663B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107526927A (en) * | 2017-08-10 | 2017-12-29 | 东北大学 | A kind of online robust flexible measurement method of blast-melted quality |
| US20210365784A1 (en) * | 2020-05-19 | 2021-11-25 | Zhejiang University | Method for deriving fault diagnosis rules of blast furnace based on deep neural network |
| CN112163380A (en) * | 2020-09-24 | 2021-01-01 | 哈尔滨锅炉厂有限责任公司 | System and method for predicting furnace oxygen concentration based on numerical simulation neural network |
| CN116842402A (en) * | 2023-09-01 | 2023-10-03 | 北京科技大学 | Blast furnace abnormal furnace condition detection method based on stable characteristic extraction of twin neural network |
Non-Patent Citations (1)
| Title |
|---|
| 高善超: "高炉炉缸含钛铁基熔体结构与粘度关系研究", CNKI, 1 June 2023 (2023-06-01), pages 131 - 149 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117540663B (en) | 2024-04-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Radhakrishnan et al. | Neural networks for the identification and control of blast furnace hot metal quality | |
| Hashimoto et al. | Transient model-based operation guidance on blast furnace | |
| CN113239482B (en) | Dynamic prediction method and device for converter post-blowing carbon content | |
| CN110378064A (en) | A method of the prediction dead stock column temperature of blast furnace furnace core | |
| Yuan et al. | Modeling and optimization of coal blending and coking costs using coal petrography | |
| Feng et al. | Case-based reasoning method based on mechanistic model correction for predicting endpoint sulphur content of molten iron in KR desulphurization | |
| Zhou et al. | Data-driven dynamic modeling for prediction of molten iron silicon content using ELM with self-feedback | |
| Cardoso et al. | Prediction of silicon content in the hot metal using Bayesian networks and probabilistic reasoning | |
| Fei et al. | Prediction of oxygen-blowing volume in BOF steelmaking process based on BP neural network and incremental learning | |
| Cardoso et al. | A novel committee machine to predict the quantity of impurities in hot metal produced in blast furnace | |
| CN117540663B (en) | Blast furnace hearth high-temperature melt viscosity prediction method and system based on neural network | |
| Spirin et al. | Software for the raw-materials management system in blast-furnace smelting | |
| Peet et al. | Neural network modelling of hot deformation of austenite | |
| Jun et al. | Prediction of hot metal temperature based on data mining | |
| Jiang et al. | Prediction of FeO content in sintering process based on heat transfer mechanism and data-driven model | |
| Mitra et al. | Evolutionary algorithms in ironmaking applications | |
| Qiang et al. | The combination prediction of BTP in sintering process based on Bayesian framework and LS-SVM | |
| JPS6324044B2 (en) | ||
| Han et al. | Cascade model for continuous prediction of silicon content of molten iron with coupled state variable nodes | |
| Niu et al. | Prediction of sinter chemical composition based on ensemble learning algorithms | |
| Ojeda et al. | Holistic indices for productivity control assessment, applied to the comparative analysis of PID and fuzzy controllers within an isasmelt furnace | |
| Tian et al. | Hybrid modeling for soft sensing of molten steel temperature in LF | |
| Huang et al. | Greedy kernel components acting on ANFIS to predict BOF steelmaking endpoint | |
| CN111679584B (en) | Regulating and controlling method and device for blast furnace smelting | |
| CN116127345B (en) | Converter steelmaking process mode design method based on deep clustering generation countermeasure network |
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 |