CN109234491A - A kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine - Google Patents
A kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine Download PDFInfo
- Publication number
- CN109234491A CN109234491A CN201811382180.9A CN201811382180A CN109234491A CN 109234491 A CN109234491 A CN 109234491A CN 201811382180 A CN201811382180 A CN 201811382180A CN 109234491 A CN109234491 A CN 109234491A
- Authority
- CN
- China
- Prior art keywords
- manganese content
- learning machine
- data
- extreme learning
- bof steelmaking
- 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
Classifications
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21C—PROCESSING OF PIG-IRON, e.g. REFINING, MANUFACTURE OF WROUGHT-IRON OR STEEL; TREATMENT IN MOLTEN STATE OF FERROUS ALLOYS
- C21C5/00—Manufacture of carbon-steel, e.g. plain mild steel, medium carbon steel or cast steel or stainless steel
- C21C5/28—Manufacture of steel in the converter
Abstract
The BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine that the invention discloses a kind of, acquire data involved in pneumatic steelmaking actual production, find the influence factor for influencing BOF Steelmaking Endpoint manganese content, then pretreatment is carried out to selected sample data and determines extreme learning machine input node number according to these data, output node number, later, the training to extreme learning machine will be completed in training dataset input limits learning machine, recently enter remaining sample data, complete the prediction to BOF Steelmaking Endpoint manganese content, BOF Steelmaking Endpoint manganese content is predicted by extreme learning machine, the biasing of the input weight and hidden layer of adjustment network is not needed in the training process, the hidden layer node number of network need to be only set, unique optimal solution can be generated, and the model training speed is fast, prediction essence Degree is high, adaptability is preferable, and then prediction promptly and accurately can be carried out to BOF Steelmaking Endpoint manganese content.
Description
Technical field
The present invention relates to a kind of prediction technique, specially a kind of BOF Steelmaking Endpoint manganese content based on extreme learning machine is pre-
Survey method, belongs to technical field of ferrous metallurgy.
Background technique
Pneumatic steelmaking is current steel-making side most important in the world as link highly important in Steel Production Flow Chart
Method.The purpose of pneumatic steelmaking is to provide chemical component and temperature qualified first steel-making water for next procedure, thus it is guaranteed that blowing
Steady progress and Accurate Prediction and the ingredient and temperature that control smelting endpoint molten steel, be pneumatic steelmaking vital task it
One.In converter steelmaking process, if being able to achieve the accurate and quick prediction to endpoint molten steel manganese content, operation can be improved
The accuracy that personnel judge tapping, and the efficiency of steel alloying operation out, so that production cost is reduced, raising molten steel matter
Amount.In order to realize that steel smelting procedure is precisely controlled, how timely and accurately to forecast that the content of BOF Steelmaking Endpoint molten steel manganese is
Major issue urgently to be resolved.
In order to realize the target of the high hit rate of BOF Steelmaking Endpoint Control, bessemerize from pervious artificial experience operate to
Automatic Steelmaking direction develop.Currently, the prediction technique of BOF Steelmaking Endpoint manganese content mainly include statistical method and and
Non-statistical method, statistical method include: linear regression, nonlinear regression etc., non-statistical method include: expert system,
BP neural network etc..Due to converter steelmaking process molten bath chemical reaction the mechanism of action it is sufficiently complex, influence molten steel manganese content because
It is plain very much, and influence each other between these factors, there is stronger non-linear relation with terminal manganese content.Based on statistics side
Method establish BOF Steelmaking Endpoint manganese content prediction model adaptability and generalization ability it is weaker, and certain statistical methods be
On the basis of sublance sampling analysis molten steel manganese content, is established using thermodynamics and material balance from sublance and be sampled to blowing
The manganese content prediction model of terminal, higher cost, and it is only applicable to the converter equipped with sublance;And based on non-statistical method
BOF Steelmaking Endpoint manganese content prediction model has stronger adaptability and generalization ability, due to influencing steel-making terminal manganese content
Factor is numerous, and has stronger non-linear relation between each influence factor and terminal manganese content, and artificial neural network is stronger
None-linear approximation ability can be very good to solve the problems, such as this, BP neural network be current widely used neural network it
One.Equally, this method also has predicts applied to BOF Steelmaking Endpoint manganese content.But the model based on this method foundation is in training
Process needs to consume a large amount of time, easily fall into local optimum and training process needs to be arranged a large amount of network training ginseng
Number, and precision of prediction is low, it is difficult to BOF Steelmaking Endpoint manganese content is accurately predicted rapidly, in time, is unfavorable for steel
The high-efficiency reform of enterprise's high-quality steel.Therefore, a kind of adaptable, converter that arithmetic speed is fast and predictablity rate is high is developed
Terminal manganese content prediction technique is made steel, is of great significance for improving BOF Steelmaking Endpoint Composition Control level.
Summary of the invention
The object of the invention is that providing a kind of pneumatic steelmaking based on extreme learning machine to solve the above-mentioned problems
Terminal manganese content prediction technique.
The present invention is through the following technical solutions to achieve the above objectives: a kind of BOF Steelmaking Endpoint based on extreme learning machine
Manganese content prediction technique, includes the following steps
Step A, the selection of the input variable of extreme learning machine, according to the influence factor of BOF Steelmaking Endpoint manganese content and end
Point manganese content carries out correlation analysis in conjunction with metallurgical mechanism and with Pearson correlation coefficient, finds influence BOF Steelmaking Endpoint
The influence factor of manganese content;
Step B is acquired the influence factor data of BOF Steelmaking Endpoint manganese content, and is located in advance to these data
Reason, it is ensured that the real effectiveness of data, the final sample data for determining this method and using;
Extreme learning machine input data is normalized in step C;
Step D chooses 4/5ths data therein to train extreme learning machine, chooses to historical data collected
Remaining 1/5th data verify the accuracy of this method, by the training degree of fitting and manganese that comprehensively consider extreme learning machine
Reasonable hidden layer node number and hidden layer activation primitive is arranged in the precision of prediction of content, to guarantee the optimal of network structure
Change.
Preferably, in order to avoid the precision of prediction to model impacts, in the step B, data prediction is to pass through
It rejects and data smoothing technique pre-processes abnormal data.
Preferably, in order to avoid due to data bulk grade, there are bigger differences to the pneumatic steelmaking end based on extreme learning machine
The predictablity rate of point manganese content prediction model has an impact, and can keep the raw information of each variable data, the step C
In, the normalized range of choice of data is [- 1,1].
Preferably, for the ease of the activation primitive of selection suitable hidden layer node number and hidden layer, the step D
In, the foundation of entire model specifically includes that the parameter selection of model training, model verifying and model.
Preferably, in order to generate unique optimal solution, and then BOF Steelmaking Endpoint manganese content can be carried out quasi- in time
True prediction in the step D, does not need the biasing of the input weight and hidden layer of adjustment network, only needs in training process
The hidden layer node number of network is set.
Preferably, it in order to for acquisition in real time, record convertor steelmaking process data, is mentioned for the operation of industrial control computer
For data supporting, described this method is realized using industrial control computer and process database to BOF Steelmaking Endpoint manganese content
Carry out real-time prediction, wherein industrial control computer for predicting BOF Steelmaking Endpoint manganese content, process database and work in real time
Industry control computer is connected.
It is closed the beneficial effects of the present invention are: should be designed based on the BOF Steelmaking Endpoint manganese content prediction technique of extreme learning machine
It manages, in step B, data prediction is pre-processed with data smoothing technique to abnormal data by rejecting, due to influencing to turn
The factor that furnace makes steel terminal manganese content is numerous, and the order of magnitude between each influence factor data has biggish difference, therefore number
Data preprocess, which can be avoided, impacts the precision of prediction of model, in step C, the normalized ranges of choice of data be [-
It 1,1], can be to avoid due to data bulk grade, there are bigger difference is pre- to the BOF Steelmaking Endpoint manganese content based on extreme learning machine
The predictablity rate for surveying model has an impact, and can keep the raw information of each variable data, in step D, entire model
The parameter selection for specifically including that model training, model verifying and model is established, by randomly select sample data 4/5ths
As the training data of model, it is used to verify the validity of model with remaining 1/5th data, comprehensively considers simultaneously
The precision of prediction of the training degree of fitting of extreme learning machine and BOF Steelmaking Endpoint manganese content selects suitable hidden layer node
Several and hidden layer activation primitives, in step D, do not needed in training process adjustment network input weight and hidden layer it is inclined
It sets, the hidden layer node number of network need to be only set, unique optimal solution can be generated, and the model training speed is fast, prediction
Precision is high, adaptability is preferable, pre- compared with based on BOF Steelmaking Endpoints manganese contents such as statistical method, expert system, BP neural networks
The precision of prediction and arithmetic speed for surveying model, which have, to be obviously improved, and then can be carried out to BOF Steelmaking Endpoint manganese content timely
Accurately prediction, this method is realized using industrial control computer and process database carries out BOF Steelmaking Endpoint manganese content
Real-time prediction, wherein industrial control computer for predicting BOF Steelmaking Endpoint manganese content in real time, control by process database and industry
Computer processed is connected, and for acquiring in real time, recording convertor steelmaking process data, provides number for the operation of industrial control computer
According to support.
Detailed description of the invention
Fig. 1 is that the present invention constitutes structural schematic diagram;
Fig. 2 is model construction of the present invention and calculation process structural schematic diagram.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Please refer to Fig. 1~2, a kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine, including it is following
Step
Step A, the selection of the input variable of extreme learning machine, according to the influence factor of BOF Steelmaking Endpoint manganese content and end
Point manganese content carries out correlation analysis in conjunction with metallurgical mechanism and with Pearson correlation coefficient, finds influence BOF Steelmaking Endpoint
The influence factor of manganese content;
Step B is acquired the influence factor data of BOF Steelmaking Endpoint manganese content, and is located in advance to these data
Reason, it is ensured that the real effectiveness of data, the final sample data for determining this method and using;
Extreme learning machine input data is normalized in step C;
Step D chooses 4/5ths data therein to train extreme learning machine, chooses to historical data collected
Remaining 1/5th data verify the accuracy of this method, by the training degree of fitting and manganese that comprehensively consider extreme learning machine
Reasonable hidden layer node number and hidden layer activation primitive is arranged in the precision of prediction of content, to guarantee the optimal of network structure
Change.
In the step B, data prediction is pre-processed with data smoothing technique to abnormal data by rejecting, by
It is numerous in the factor for influencing BOF Steelmaking Endpoint manganese content, and the order of magnitude between each influence factor data has biggish difference
It is different, therefore data prediction can be avoided and impact to the precision of prediction of model, in the step C, at the normalization of data
Managing range of choice is [- 1,1], can be to avoid due to data bulk grade, there are bigger differences to the converter refining based on extreme learning machine
The predictablity rate of steel terminal manganese content prediction model has an impact, and can keep the raw information of each variable data, described
In step D, the foundation of entire model specifically includes that the parameter selection of model training, model verifying and model, by randomly selecting
Training data of 4/5ths of sample data as model is used to verify having for model with remaining 1/5th data
Effect property, while the precision of prediction for the training degree of fitting and BOF Steelmaking Endpoint manganese content for comprehensively considering extreme learning machine selects to close
The activation primitive of suitable hidden layer node number and hidden layer in the step D, does not need the defeated of adjustment network in training process
Enter the biasing of weight and hidden layer, the hidden layer node number of network need to be only set, unique optimal solution can be generated, and should
Model training speed is fast, precision of prediction is high, adaptability is preferable, is relatively turned based on statistical method, expert system, BP neural network etc.
The precision of prediction and arithmetic speed of furnace steel-making terminal manganese content prediction model, which have, to be obviously improved, and then can be to pneumatic steelmaking
Terminal manganese content carries out prediction promptly and accurately, and described this method is realized pair using industrial control computer and process database
BOF Steelmaking Endpoint manganese content carries out real-time prediction, wherein industrial control computer for predicting BOF Steelmaking Endpoint manganese in real time
Content, process database are connected with industrial control computer, are work for acquiring in real time, recording convertor steelmaking process data
The operation that industry controls computer provides data supporting.
Working principle: when using the BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine, firstly, right
Sample data collected is pre-processed, and metallurgical mechanism is combined with correlation analysis, determines the required data of this model;
Secondly, 4/5ths data sample is used to establish the BOF Steelmaking Endpoint manganese content prediction model based on extreme learning machine,
/ 5th data samples to be predicted with established model treatment again;Finally, according to the prediction result and reality of model
The numerical value of measurement, which is compared, draws a conclusion, and the time needed for counting the model calculation.No setting is required for training process largely
Parameters of Neural Network Structure, it is only necessary to selection determines that suitable hidden layer node number can obtain the optimal solution of this model,
Greatly improve the arithmetic speed of model and the accuracy of prediction.
It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie
In the case where without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power
Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims
Variation is included within the present invention.Any reference signs in the claims should not be construed as limiting the involved claims.
In addition, it should be understood that although this specification is described in terms of embodiments, but not each embodiment is only wrapped
Containing an independent technical solution, this description of the specification is merely for the sake of clarity, and those skilled in the art should
It considers the specification as a whole, the technical solutions in the various embodiments may also be suitably combined, forms those skilled in the art
The other embodiments being understood that.
Claims (6)
1. a kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine, it is characterised in that: include the following steps
Step A, the selection of the input variable of extreme learning machine, according to the influence factor of BOF Steelmaking Endpoint manganese content and terminal manganese
Content carries out correlation analysis in conjunction with metallurgical mechanism and with Pearson correlation coefficient, and finding, which influences BOF Steelmaking Endpoint manganese, contains
The influence factor of amount;
Step B is acquired the influence factor data of BOF Steelmaking Endpoint manganese content, and pre-processes to these data,
Ensure the real effectiveness of data, the final sample data for determining this method and using;
Extreme learning machine input data is normalized in step C;
Step D chooses 4/5ths data therein to historical data collected to train extreme learning machine, chooses remaining
1/5th data verify the accuracy of this method, by the training degree of fitting and manganese content that comprehensively consider extreme learning machine
Precision of prediction reasonable hidden layer node number and hidden layer activation primitive are set, to guarantee the optimization of network structure.
2. a kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine according to claim 1, special
Sign is: in the step B, data prediction is pre-processed with data smoothing technique to abnormal data by rejecting.
3. a kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine according to claim 1, special
Sign is: in the step C, the normalized range of choice of data is [- 1,1].
4. a kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine according to claim 1, special
Sign is: in the step D, the foundation of entire model specifically includes that the parameter selection of model training, model verifying and model.
5. a kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine according to claim 1, special
Sign is: in the step D, the biasing of the input weight and hidden layer of adjustment network is not needed in training process, need to only be set
Set the hidden layer node number of network.
6. a kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine according to claim 1, special
Sign is: described this method is realized using industrial control computer and process database carries out BOF Steelmaking Endpoint manganese content
Real-time prediction, wherein industrial control computer for predicting BOF Steelmaking Endpoint manganese content in real time, control by process database and industry
Computer processed is connected.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811382180.9A CN109234491A (en) | 2018-11-20 | 2018-11-20 | A kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811382180.9A CN109234491A (en) | 2018-11-20 | 2018-11-20 | A kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109234491A true CN109234491A (en) | 2019-01-18 |
Family
ID=65076114
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811382180.9A Pending CN109234491A (en) | 2018-11-20 | 2018-11-20 | A kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109234491A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114611844A (en) * | 2022-05-11 | 2022-06-10 | 北京科技大学 | Method and system for determining alloy addition amount in converter tapping process |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1588346A (en) * | 2004-08-30 | 2005-03-02 | 邢台钢铁有限责任公司 | Method for predicting converter terminal point using artificial nurve network technology |
CN104537167A (en) * | 2014-12-23 | 2015-04-22 | 清华大学 | Interval type index forecasting method based on robust interval extreme learning machine |
CN105925750A (en) * | 2016-05-13 | 2016-09-07 | 南阳理工学院 | Steelmaking end point prediction method based on neural networks |
CN108256260A (en) * | 2018-02-05 | 2018-07-06 | 北京科技大学 | A kind of continuous casting billet quality Forecasting Methodology based on extreme learning machine |
CN108676955A (en) * | 2018-05-02 | 2018-10-19 | 中南大学 | A kind of BOF Steelmaking Endpoint carbon content and temprature control method |
-
2018
- 2018-11-20 CN CN201811382180.9A patent/CN109234491A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1588346A (en) * | 2004-08-30 | 2005-03-02 | 邢台钢铁有限责任公司 | Method for predicting converter terminal point using artificial nurve network technology |
CN104537167A (en) * | 2014-12-23 | 2015-04-22 | 清华大学 | Interval type index forecasting method based on robust interval extreme learning machine |
CN105925750A (en) * | 2016-05-13 | 2016-09-07 | 南阳理工学院 | Steelmaking end point prediction method based on neural networks |
CN108256260A (en) * | 2018-02-05 | 2018-07-06 | 北京科技大学 | A kind of continuous casting billet quality Forecasting Methodology based on extreme learning machine |
CN108676955A (en) * | 2018-05-02 | 2018-10-19 | 中南大学 | A kind of BOF Steelmaking Endpoint carbon content and temprature control method |
Non-Patent Citations (1)
Title |
---|
张小晨: ""基于极限学习机的转炉炼钢终点预测模型研究"", 《中国优秀硕士学位论文全文数据库程科技辑》, no. 03, pages 1 - 3 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114611844A (en) * | 2022-05-11 | 2022-06-10 | 北京科技大学 | Method and system for determining alloy addition amount in converter tapping process |
US11900255B2 (en) | 2022-05-11 | 2024-02-13 | University Of Science And Technology Beijing | Multi-mode smelting method based on the classification system of molten iron |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109359723A (en) | Based on the converter terminal manganese content prediction technique for improving regularization extreme learning machine | |
CN109252009A (en) | BOF Steelmaking Endpoint manganese content prediction technique based on regularization extreme learning machine | |
CN103882176B (en) | The online dynamic control method of a kind of convertor steelmaking process based on data-driven | |
CN106345823B (en) | The method of online real-time estimate mechanical performance based on coils of hot-rolled steel production procedure | |
Wang et al. | The Control and Prediction of End‐Point Phosphorus Content during BOF Steelmaking Process | |
CN106198408B (en) | The method and device thereof of alloying component on-line checking and visualizing control in casting | |
CN108256260A (en) | A kind of continuous casting billet quality Forecasting Methodology based on extreme learning machine | |
CN114611844B (en) | Method and system for determining alloy addition amount in converter tapping process | |
CN109935280A (en) | A kind of blast-melted quality prediction system and method based on integrated study | |
CN113919206A (en) | Narrow composition control method for steelmaking alloying | |
CN110885912B (en) | Automatic steelmaking method and system based on data analysis | |
CN106126842A (en) | A kind of emulation mode of the dynamic production run of steelmaking continuous casting workshop | |
CN105807741A (en) | Industrial production flow prediction method | |
CN106834593A (en) | A kind of method that RH refining furnace decarbonization process data are determined with reference heats method | |
CN113192568B (en) | Method and system for forecasting desulfurization end point of refining furnace | |
CN102399933A (en) | Automatic controlling method for converter oxygen lance used for converting low-carbon steel | |
CN112036081A (en) | Method for determining addition amount of silicon-manganese alloy in converter tapping based on yield prediction | |
CN109234491A (en) | A kind of BOF Steelmaking Endpoint manganese content prediction technique based on extreme learning machine | |
CN109581988A (en) | A kind of intelligent steelmaking system based on converter gas analytical technology | |
CN104419799A (en) | Method for online prediction of carbon content of high-carbon steel during converter smelting process | |
CN107630122A (en) | A kind of RH dynamic decarburization optimization methods based on flue gas analysis | |
CN116469481B (en) | LF refined molten steel composition forecasting method based on XGBoost algorithm | |
CN104894328A (en) | Converter endpoint phosphorus content forecasting method | |
Zheng et al. | Method to predict alloy yield based on multiple raw material conditions and a PSO-LSTM network | |
CN201258354Y (en) | Wire galvanization production line on-line zinc layer weight detection apparatus |
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 |