CN114330137A - Sinter quality prediction method based on width learning - Google Patents
Sinter quality prediction method based on width learning Download PDFInfo
- Publication number
- CN114330137A CN114330137A CN202111672099.6A CN202111672099A CN114330137A CN 114330137 A CN114330137 A CN 114330137A CN 202111672099 A CN202111672099 A CN 202111672099A CN 114330137 A CN114330137 A CN 114330137A
- Authority
- CN
- China
- Prior art keywords
- data
- nodes
- width
- neural network
- sinter
- 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
Images
Landscapes
- Manufacture And Refinement Of Metals (AREA)
Abstract
The invention provides a width learning-based sintered mineral content prediction method, which comprises the steps of determining a plurality of factors with high correlation degree with prediction indexes by preprocessing historical data and analyzing relevance, inputting the factors into an improved width learning neural network for training, obtaining optimal weight and bias by using a matrix pseudo-inverse mode, testing the accuracy of the optimal weight and bias, increasing the width of the neural network when the optimal weight and bias are not reached to an expected value, and training on the basis of the existing model until an optimal sintered mineral content prediction model is obtained. The invention can also add new data in an incremental learning mode, train on the basis of the historical data model and ensure the timely update of the model. The method improves the training speed of the model by combining the historical data and the newly added data, and realizes the rapid construction and real-time update of the model, thereby accurately and rapidly predicting the performance and the quality of the sinter and providing a new thought and a new method for guiding the production of the sinter and optimizing the sintering process.
Description
Technical Field
The invention relates to the technical field of sinter performance prediction, in particular to a sinter quality prediction method based on width learning.
Background
The sintering process is an energy-consuming big household in the whole iron and steel industry, the sintered ore generated in the sintering process is a main raw material for blast furnace ironmaking, and the quality of the sintered ore has a decisive influence on the blast furnace operation and the quality of the produced molten iron. In the actual sintering production process, the detection mode of the quality of the sintered minerals is generally that after the finished sintered minerals are produced, the finished minerals are sampled and detected every two hours, and operators adjust sintering process parameters according to experience according to inspection results. The traditional detection method has great hysteresis in controlling the quantity of the sintered minerals, and the proportion of the raw materials cannot be adjusted in time according to the quantity of the sintered minerals. Therefore, establishing a proper sinter quality prediction model has important significance for improving the sinter quality.
The patent with publication number CN 110070217A discloses a method for predicting the quality of a sinter based on process parameters, which adopts BP neural network mapping nonlinear function to establish a model for predicting the quality of the sinter based on the process parameters; the sintering process parameters collected in real time are input into the sintered mineral quantity prediction model as input variables, and the output variables are the predicted sintered mineral quantity parameters, so that the purpose of predicting the sintered mineral quantity in advance is achieved. The patent with publication number CN 108549791A discloses a model parameter self-adaptive sinter property prediction method, which fits a relation function between mixed ore physicochemical indexes and sinter properties through an RBF neural network algorithm, can quickly predict the sinter properties, realizes automatic update of model parameters, and provides a basis for formulating an ore blending strategy.
However, the above methods all adopt a deep learning method, and although the methods such as the BP neural network and the RBF neural network have the capability of fitting a nonlinear function, the methods all adopt a mode of increasing the number of network layers and solving the gradient layer by layer to update the weight and the bias, which will cause that the model is easy to suffer from the problems of local optimum, gradient disappearance, gradient explosion, slow modeling speed and the like. In addition, when the training data is updated, it is time-consuming to retrain the deep neural network again, and if the network cannot be updated in time, the user experience is poor, and the prediction result of the quality of the sinter is delayed.
In view of the above, there is a need to design an improved method for predicting quality of sintered ore based on width learning to solve the above problems.
Disclosure of Invention
The invention aims to provide a sinter quality prediction method based on width learning, which comprises the steps of preprocessing historical data, obtaining a plurality of factors with the highest correlation degree with prediction indexes by adopting a correlation analysis method, and establishing a sinter quality prediction model by taking an analysis result as an input parameter of an improved width neural network; the accuracy of the parameters in the sintering process is checked, and the parameters are adjusted in time, so that the purpose of improving the performance and quality of the sintered ore is achieved.
In order to achieve the above object, the present invention provides a method for predicting agglomerate quality based on width learning, comprising the steps of:
s1, preprocessing data
Acquiring historical data in sinter production, and removing large noise and repeatability data in the historical data; carrying out normalization processing on the data to obtain preprocessed data;
s2, determining input parameters
Performing relevance analysis on the preprocessed data and the prediction index obtained in the step S1, and screening m factors with high relevance with the prediction index as input parameters;
s3, establishing a sinter quality prediction model
S31, improving the width learning neural network to reduce overfitting of the neural network;
s32, carrying out initialization setting on the hyper-parameters, setting the number of groups of enhanced nodes or feature nodes, and setting the number of nodes in each group; inputting data of factors with high correlation degree with the prediction index into the width learning neural network as input parameters for training, wherein the output parameters are the correlation data of the prediction index;
s33, solving the optimal bias and weight of the width learning neural network by using a matrix pseudo-inverse mode; testing by using the test set, calculating the accuracy, and comparing the accuracy with a preset expected accuracy value; when the width is larger than a preset expected value, obtaining a model for learning with the optimal width; when the width of the neural network is smaller than a preset expected value, the width of the neural network is increased, and the pseudo-inverse matrix is recalculated by using the calculation result of the previous model through a ridge regression algorithm until an optimal sinter quality prediction model based on width learning is obtained;
s4 prediction of performance indexes of sintered ore
And normalizing the data to be detected, inputting the data to be detected into the sinter quality prediction model, and performing inverse normalization on the output parameters to obtain a prediction result.
As a further improvement of the invention, the method can also add new data in an incremental learning mode, and train on the basis of a model of historical data by using the ridge regression algorithm to update the sinter quality prediction model.
As a further improvement of the present invention, in step S33, the method of increasing the width of the neural network includes increasing the number of groups of feature nodes, increasing the number of groups of enhanced nodes, or increasing the number of groups of feature nodes and enhanced nodes at the same time.
As a further improvement of the present invention, in step S33, the method for obtaining the optimal bias and weight of the width learning neural network by using a matrix pseudo-inverse method includes the following steps:
s1, setting the weight and bias of the input parameters to the mapping nodes through a sparse self-encoder, and randomly generating the weight and bias from the characteristic nodes to the enhanced nodes;
s2, performing matrix splicing on the characteristic nodes and the enhanced nodes according to the columns of the matrix to obtain a matrix H, and calculating a pseudo-inverse matrix T of the matrix H;
and S3, multiplying the pseudo-inverse matrix T by the label of the data to obtain the optimal bias and weight of the output node.
As a further improvement of the present invention, in step S2, the criteria for screening out the factors with high correlation with the prediction index is to arrange the correlation of different factors from large to small, and then select m factors characterized by correlation before the sequence as the factors with high correlation with the prediction index, where the number of m is set according to the user requirement.
As a further improvement of the present invention, in step S31, the method for improving the width learning neural network includes an algorithm for randomly adding a dropout layer, and randomly deactivating nodes of an intermediate layer and an input layer of the width learning neural network.
As a further improvement of the present invention, in step S2, the method of the correlation analysis is a gray correlation analysis method or a principal component analysis algorithm.
As a further improvement of the present invention, in step S1, the formula of the normalization process includes a min-max normalization formula or a Z-score normalization formula.
As a further improvement of the present invention, in step S32, the number of groups of the enhanced nodes or feature nodes and the number of nodes of each group are set according to user requirements.
As a further improvement of the invention, the prediction model of the quality of the sintered ore can also be used for predicting the performance index of the pellet ore or coke in other fields of metallurgical production.
The invention has the beneficial effects that:
(1) the invention provides a sinter quality prediction method based on width learning, which is characterized in that based on historical data of actual production of a sintering plant, preprocessing and correlation analysis with a prediction index are carried out on the historical data, data of factors with high correlation degree with the prediction index are determined to be input parameters, an improved width learning neural network is input for training, and an optimal sinter quality prediction model is established; and newly added data can be added in an incremental learning mode, and training is performed by using a ridge regression algorithm on the basis of a model of historical data so as to ensure that the sintered mineral quality prediction model is updated in time. The method is simple, convenient and quick to operate, can adapt to different environments, fully utilizes newly added data, and trains on the basis of the existing model; the method has the advantages that a traditional gradient back propagation method is not adopted, an incremental learning method is adopted, the training speed is greatly increased, the rapid construction and real-time updating of the model are realized, the performance and the quality of the sinter are accurately and rapidly predicted, and a new thought and a new method are provided for guiding the production of the sinter and optimizing the sintering process.
(2) The width neural network directly obtains the optimal weight and bias of the network by using a matrix pseudo-inverse mode, and when the precision of the neural network does not meet an expected value, the width of the neural network can be increased by increasing characteristic nodes and enhancing nodes, so that the performance of the neural network is improved; the width of the neural network is increased, meanwhile, an inverse matrix after network nodes are added is calculated by adopting a ridge regression mode on the basis of the existing model weight results, the calculation amount is greatly reduced, the requirement on calculation power is low, historical data can be fully utilized, and the rapid modeling of the neural network is realized. In addition, the method also keeps the characteristic of deep learning, avoids the error of manual operation and ensures the precision of the prediction index; and the prediction of the sinter quality and the performance index is realized, and a foundation is laid for the optimized regulation and control of the sinter.
(3) When a sintered mineral quality prediction model is established, the method improves the width learning neural network by adding an algorithm, randomly inactivates nodes of the middle layer and the input layer, and avoids the problem of neural network overfitting; the problems that the existing deep neural network is easy to fall into local optimization, gradient disappears or gradient explodes and the modeling speed is low in the gradient search process due to the fact that the number of layers of the network is increased in order to improve the network performance are solved.
Drawings
Fig. 1 is a schematic flow chart of a sinter quality prediction method based on width learning according to the present invention.
FIG. 2 is a flow chart of a sintered ore according to example 1 of the present invention.
FIG. 3 is a diagram showing the predicted TFe content of the sintered ore in example 2 of the present invention.
FIG. 4 is a graph showing the prediction results of the FeO content in the sintered ore in example 2 of the present invention.
FIG. 5 is a graph showing the result of predicting the CaO content in the sintered ore in example 2 of the present invention.
FIG. 6 shows a sintered SiO ore of example 2 of the present invention2Content prediction result graph.
FIG. 7 is a graph showing the result of predicting the MgO content of the sintered ore in example 2 of the present invention.
FIG. 8 shows a sintered ore K in example 2 of the present invention2Prediction result graph of O element content
FIG. 9 shows sintered ore Na in example 2 of the present invention2Prediction result graph of O element content
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in detail with reference to the accompanying drawings and specific embodiments.
It should be noted that, in order to avoid obscuring the present invention with unnecessary details, only the structures and/or processing steps closely related to the aspects of the present invention are shown in the drawings, and other details not closely related to the present invention are omitted.
In addition, it is also to be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
Referring to fig. 1, a method for predicting the quality of a sinter based on width learning includes the following steps:
s1, preprocessing data
Acquiring historical data in sinter production, and removing large noise and repeatability data in the historical data; carrying out normalization processing on the data to obtain preprocessed data; the historical data comprises sintering process parameters in a sintering process and corresponding sinter quality parameters (prediction indexes) of each sintering, the sintering process parameters can be sinter ingredient components, alkalinity, sintering temperature, ignition temperature, negative pressure and the like, and the normalization formula is a min-max standardization formula which is as follows:
wherein X' is data after normalization, X is data before normalization, and X isminIs the minimum value of the data sequence, XmaxIs the maximum value of the data sequence; the normalization processing formula can also be a Z-score normalization formula;
s2, determining input parameters
Performing relevance analysis on the preprocessed data and the prediction index obtained in the step S1, and screening m factors with high relevance with the prediction index as input parameters;
the standard for screening the factors with high correlation with the prediction index is to arrange the correlation of different factors from large to small, and then select m factors characterized by the correlation in front of the sequence as the factors with high correlation with the prediction index, wherein the number of m is set according to the user requirement, and m is more than or equal to 3 and less than or equal to 15 in the specific embodiment. The correlation analysis method is a grey correlation degree analysis method or a principal component analysis algorithm; the analysis method of the priority grey correlation degree comprises the following analysis formulas:
in the formula: s-degree of association of different sequences,
y (k) -comparing the sequences,
p-is the resolution factor, 0 < p < 1, usually 0.5,
n-number of distinct sequences;
s3, establishing a sinter quality prediction model
S31, improving the width learning neural network to reduce overfitting of the neural network;
s32, carrying out initialization setting on the hyper-parameters, setting the number of groups of enhanced nodes or feature nodes, and setting the number of nodes in each group; taking data of factors with high correlation degree with the prediction index as input, inputting parameters into the width learning neural network for training, and outputting the parameters as the correlation data of the prediction index;
s33, solving the optimal bias and weight of the width learning neural network by using a matrix pseudo-inverse mode; testing by using the test set, calculating the accuracy, and comparing the accuracy with a preset expected accuracy value; when the width is larger than a preset expected value, obtaining a model for learning with the optimal width; when the width of the neural network is smaller than a preset expected value, the width of the neural network is increased, and the pseudo-inverse matrix is recalculated by using the calculation result of the previous model through a ridge regression algorithm until an optimal sinter quality prediction model based on width learning is obtained;
s4 prediction of performance indexes of sintered ore
And normalizing the data to be detected, inputting the data into a sintered mineral quality prediction model, and performing inverse normalization on output parameters to obtain a prediction result.
Specifically, in step S31, the method for improving the width learning neural network includes an algorithm for randomly adding a dropout layer, and randomly deactivating nodes of the middle layer and the input layer of the width learning neural network. In step S32, the number of groups of enhanced nodes or feature nodes and the number of nodes per group are set according to user requirements. In step S33, the method for increasing the width of the neural network includes increasing the number of groups of feature nodes, increasing the number of groups of enhanced nodes, or simultaneously increasing the number of groups of feature nodes and augmented nodes. The method for solving the optimal bias and weight of the width learning neural network by utilizing a matrix pseudo-inverse mode comprises the following steps of:
s1, setting the weight and bias of the input parameters to the mapping nodes through a sparse self-encoder, and randomly generating the weight and bias from the characteristic nodes to the enhanced nodes;
s2, performing matrix splicing on the characteristic nodes and the enhanced nodes according to the columns of the matrix to obtain a matrix H, and calculating a pseudo-inverse matrix T of the matrix H;
and S3, multiplying the pseudo-inverse matrix T by the label of the data to obtain the optimal bias and weight of the output node.
In particular, the invention can also add new data in an incremental learning mode, and train by utilizing the ridge regression algorithm on the basis of a model of historical data so as to ensure the timely update of the sintered mineral quality prediction model. And continuously updating the model by combining historical data and newly added data to improve the performance of the sinter prediction system. The established sinter prediction model can adapt to different environments, newly added data are fully utilized, training is carried out on the basis of the existing model, the training speed is greatly improved, and the model is quickly constructed and updated in real time, so that the quality prediction of the performance of the sinter is accurately and quickly carried out, and a new thought and a new method are provided for guiding the production of the sinter and optimizing the sintering process.
Specifically, the prediction model of the quality of the sintered ore can be used for predicting the performance indexes of pellets or coke in other fields of metallurgical production.
Example 1
Referring to fig. 2, example 1 provides a method for predicting agglomerate quality based on width learning, and performs accuracy test on agglomerate material data, where the data of prediction parameters includes TFe%, FeO%, and SiO2Percent, CaO and MgO; the method specifically comprises the following steps:
s1, preprocessing data
Acquiring historical data of ingredients in sinter production, and removing large noise and repeatability data in the historical data; carrying out normalization processing on the data to obtain preprocessed data;
s2, determining input parameters
Performing relevance analysis on the preprocessed data and the prediction index obtained in the step S1, and screening 6 factors with high relevance with the prediction index as input parameters, wherein the 6 factors comprise the iron-containing raw material ratio, the quicklime addition amount, the limestone addition amount, the dolomite addition amount, the anthracite addition amount and the coke powder addition amount;
s3, establishing a sinter quality prediction model
S31, improving the width learning neural network to reduce overfitting of the neural network;
s32, carrying out initial setting on hyper-parameters, setting the number of groups of enhanced nodes or feature nodes to be 10 groups, and setting the number of nodes in each group to be 10; inputting the input parameters into a width learning neural network for training;
s33, solving the optimal bias and weight of the width learning neural network by using a matrix pseudo-inverse mode; testing by using the test set, calculating the accuracy, and comparing with a preset expected accuracy value; when the width is larger than a preset expected value, obtaining a model for learning with the optimal width; when the weight is smaller than a preset expected value, increasing the width of the neural network, and recalculating the pseudo-inverse matrix by using the weight result through a ridge regression algorithm until an optimal sinter quality prediction model based on width learning is obtained;
s4 prediction of performance indexes of sintered ore
The data to be detected adopts the sintering ore ingredient data of No. 1 month from 10 months to 9 months per month in 2020 to 2021 of a sintering factory, normalization processing is carried out on the data, a sintering mineral quality prediction model is input, and the output parameters are subjected to inverse normalization to obtain a prediction result.
Example 15 indices TFe%, FeO%, SiO2Actual conditions and errors of% CaO, MgO are shown in tables 1 to 5; wherein, the 1-9 groups of data in each table respectively correspond to the test data of No. 1 monthly from 1 to 9 months in 2021, and the 10-12 groups respectively correspond to the test data of No. 1 monthly from 10 to 12 months in 2020.
TABLE 1 examination of TFe% content test data
TABLE 2 examination of FeO% content test data
TABLE 3 examination of SiO 2% content test data
TABLE 4 examination of CaO% content test data
TABLE 5 examination of MgO% content test data
As can be seen from tables 1-5, the predicted TFe%, FeO%, and SiO values of the test using the breadth learning neural network and incremental learning with the addition of enhanced nodes2The total difference between the actual conditions of percent, CaO percent and MgO percent and the predicted values is small, and the error is small. The method can make full use of historical data, and realizes rapid modeling of the network; the error of manual operation is avoided, and the accuracy of the prediction index is ensured.
Example 2
This example provides a method for predicting the amount of agglomerate based on width learning, which is different from example 1 in that the data of the prediction parameters include TFe%, FeO%, and SiO2％、CaO％、MgO％、Na2O% and K2O%, the data to be detected in step S4 is 55-day production data of 2020 in a certain ironworks; it is substantially the same as embodiment 1, and will not be described herein again.
Example 2 the 7 measured indices TFe%, FeO%, SiO2％、CaO％、MgO％、Na2O% and K2The predicted results of O% are shown in FIGS. 3 to 9; the credential error values for the predicted results are shown in the following table:
TABLE 6 average error values of predicted results for sinter
It can be seen from fig. 3-9 and table 6 that the error between the actual value and the predicted value of the TFe% of the sintered ore is large, and after the width of the neural network is increased, the model is built more accurately by recalculation in a ridge regression manner based on the existing data. FeO and SiO in sintered ore2％、CaO％、MgO％、K2O% and Na2The average error of the actual value and the predicted value of O% is smaller, which shows that the accuracy of the model prediction index is higher.
In summary, the invention provides a sintered mineral quality prediction method based on width learning, which determines a plurality of factors with high correlation with prediction indexes as input parameters through historical data preprocessing and relevance analysis; inputting an improved width learning neural network for training, obtaining the optimal weight and bias of the neural network by using a matrix pseudo-inverse mode, and testing the accuracy of the neural network; when the accuracy reaches the requirement, a sinter quality prediction model is established, when the accuracy does not reach the requirement, the width of the network can be increased by increasing feature nodes and enhancing the nodes, and the pseudo-inverse matrix is recalculated by using the existing model through a ridge regression algorithm until the optimal sinter quality prediction model is obtained. The method can also add new data in an incremental learning mode, and train by utilizing a ridge regression algorithm on the basis of a model of historical data so as to ensure the timely update of the sintered mineral quality prediction model. The method is simple, convenient and quick to operate, can adapt to different environments, fully utilizes newly added data, trains on the basis of the existing model, does not adopt a traditional gradient back propagation method but adopts an incremental learning method, greatly improves the training speed, realizes quick construction and real-time updating of the model, accurately and quickly predicts the performance and quality of the sinter, and provides a new idea and method for guiding the production of the sinter and optimizing the sintering process.
Although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted for elements thereof without departing from the spirit and scope of the present invention.
Claims (10)
1. A sinter quality prediction method based on width learning is characterized by comprising the following steps:
s1, preprocessing data
Acquiring historical data in sinter production, and removing large noise and repeatability data in the historical data; carrying out normalization processing on the data to obtain preprocessed data;
s2, determining input parameters
Performing relevance analysis on the preprocessed data and the prediction index obtained in the step S1, and screening m factors with high relevance with the prediction index as input parameters;
s3, establishing a sinter quality prediction model
S31, improving the width learning neural network to reduce overfitting of the neural network;
s32, carrying out initialization setting on the hyper-parameters, setting the number of groups of enhanced nodes or feature nodes, and setting the number of nodes in each group; inputting data of factors with high correlation degree with the prediction index into the width learning neural network as input parameters for training, wherein the output parameters are the correlation data of the prediction index;
s33, solving the optimal bias and weight of the width learning neural network by using a matrix pseudo-inverse mode; testing by using the test set, calculating the accuracy, and comparing the accuracy with a preset expected accuracy value; when the width is larger than a preset expected value, obtaining a model for learning with the optimal width; when the width of the neural network is smaller than a preset expected value, the width of the neural network is increased, and the pseudo-inverse matrix is recalculated by using the calculation result of the previous model through a ridge regression algorithm until an optimal sinter quality prediction model based on width learning is obtained;
s4 prediction of sinter performance
And normalizing the data to be detected, inputting the data to be detected into the sinter quality prediction model, and performing inverse normalization on the output parameters to obtain a prediction result.
2. The method for predicting the quality of the sintering ore based on the width learning as claimed in claim 1, wherein the method can further add new data by means of incremental learning, and train on the basis of a model of historical data by using the ridge regression algorithm to update the sintering ore quality prediction model.
3. The method for predicting quality of a sintering ore based on width learning according to claim 1, wherein in step S33, the method for increasing the width of the neural network comprises increasing the number of groups of feature nodes, increasing the number of groups of enhanced nodes or increasing the number of groups of feature nodes and augmented nodes simultaneously.
4. The method for predicting the quality of the sintering ore based on the width learning according to claim 1, wherein in step S33, the method for obtaining the optimal bias and weight of the width learning neural network by using the matrix pseudo-inverse method comprises the following steps:
s1, setting the weight and bias of the input parameters to the mapping nodes through a sparse self-encoder, and randomly generating the weight and bias from the characteristic nodes to the enhanced nodes;
s2, performing matrix splicing on the characteristic nodes and the enhanced nodes according to the columns of the matrix to obtain a matrix H, and calculating a pseudo-inverse matrix T of the matrix H;
and S3, multiplying the pseudo-inverse matrix T by the label of the data to obtain the optimal bias and weight of the output node.
5. The method according to claim 1, wherein in step S2, the criteria for screening out the factors with high correlation with the prediction index is to arrange the correlation of different factors from large to small, and then select m factors characterized by correlation before the sequence as the factors with high correlation with the prediction index, wherein the number of m is set according to the user' S requirement.
6. The method for predicting the quality of the sinter based on the width learning of claim 1, wherein in step S31, the method for improving the width learning neural network includes an algorithm for randomly adding a dropout layer, and nodes of an intermediate layer and an input layer of the width learning neural network are randomly inactivated.
7. The method for predicting the quality of a sintering ore based on the width learning according to claim 1, wherein in step S2, the correlation analysis method is a gray correlation analysis method or a principal component analysis algorithm.
8. The width learning-based sinter quality prediction method according to claim 1, wherein in step S1, the formula of the normalization process includes a min-max normalization formula or a Z-score normalization formula.
9. The method for predicting quality of a sintered ore based on width learning according to claim 1, wherein the number of groups of the enhanced nodes or the characteristic nodes and the number of nodes of each group are set according to user requirements in step S32.
10. The method for predicting the quality of the sinter based on the width learning as claimed in claim 1, wherein the model for predicting the quality of the sinter can be used for predicting the performance index of the pellet ore or the coke in other fields of metallurgical production.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111672099.6A CN114330137A (en) | 2021-12-31 | 2021-12-31 | Sinter quality prediction method based on width learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111672099.6A CN114330137A (en) | 2021-12-31 | 2021-12-31 | Sinter quality prediction method based on width learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114330137A true CN114330137A (en) | 2022-04-12 |
Family
ID=81021553
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111672099.6A Pending CN114330137A (en) | 2021-12-31 | 2021-12-31 | Sinter quality prediction method based on width learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114330137A (en) |
-
2021
- 2021-12-31 CN CN202111672099.6A patent/CN114330137A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112699613B (en) | Multi-target integrated burdening optimization method, system, equipment and medium for iron making | |
| Wang et al. | Applying input variables selection technique on input weighted support vector machine modeling for BOF endpoint prediction | |
| CN114611844B (en) | Method and system for determining alloy addition amount in converter tapping process | |
| CN111915080B (en) | Raw fuel cost optimal proportioning method based on molten iron quality constraint | |
| CN107368125B (en) | A kind of blast furnace temperature control system and method based on CBR Yu the parallel mixed inference of RBR | |
| CN107299170B (en) | A kind of blast-melted quality robust flexible measurement method | |
| CN104778361B (en) | The method of modified EMD Elman neural network prediction molten iron silicon contents | |
| Jiang et al. | Real-time moisture control in sintering process using offline–online NARX neural networks | |
| CN116469481B (en) | LF refined molten steel composition forecasting method based on XGBoost algorithm | |
| CN110502781B (en) | Prior knowledge-based ferroalloy production ingredient optimization method | |
| CN109934421B (en) | Blast furnace molten iron silicon content prediction and compensation method for fluctuating furnace condition | |
| CN107808221A (en) | Blast furnace material distribution Parameter Decision Making method based on case matching | |
| CN114662763A (en) | Method and system for evaluating cost performance of single coal for coking coal blending | |
| CN113177364B (en) | Soft measurement modeling method for temperature of blast furnace tuyere convolution zone | |
| Li et al. | A novel multiple-input–multiple-output random vector functional-link networks for predicting molten iron quality indexes in blast furnace | |
| CN110849149A (en) | Energy perception-based sintering batching scheme cascade optimization obtaining method and device | |
| CN102156405B (en) | Sintered ore chemical component prediction and intelligent control system under small sample poor information | |
| CN114330137A (en) | Sinter quality prediction method based on width learning | |
| CN110491454B (en) | Blast furnace smelting cost management method and system and computer-storable medium | |
| CN110619931B (en) | Sintering process carbon efficiency optimization method based on multi-time scale optimization | |
| CN111798023A (en) | Method for predicting comprehensive coke ratio in steelmaking sintering production | |
| CN114525372B (en) | Multi-mode fusion based blast furnace state monitoring method and device | |
| CN115359851A (en) | Multi-objective prediction optimization method for sintering burdening based on extreme random tree-NSGA-II | |
| CN107798433B (en) | A kind of method and apparatus that raw material proportioning is determined based on cost in manufacture course of products | |
| CN112287283B (en) | Blast furnace running state evaluation method and device and storage medium |
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 |