CN111676365A - LSTM-based method for predicting annealing temperature set value of transition steel coil - Google Patents
LSTM-based method for predicting annealing temperature set value of transition steel coil Download PDFInfo
- Publication number
- CN111676365A CN111676365A CN202010519714.9A CN202010519714A CN111676365A CN 111676365 A CN111676365 A CN 111676365A CN 202010519714 A CN202010519714 A CN 202010519714A CN 111676365 A CN111676365 A CN 111676365A
- Authority
- CN
- China
- Prior art keywords
- lstm
- steel coil
- transition
- data
- model
- 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
Classifications
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21D—MODIFYING THE PHYSICAL STRUCTURE OF FERROUS METALS; GENERAL DEVICES FOR HEAT TREATMENT OF FERROUS OR NON-FERROUS METALS OR ALLOYS; MAKING METAL MALLEABLE, e.g. BY DECARBURISATION OR TEMPERING
- C21D11/00—Process control or regulation for heat treatments
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21D—MODIFYING THE PHYSICAL STRUCTURE OF FERROUS METALS; GENERAL DEVICES FOR HEAT TREATMENT OF FERROUS OR NON-FERROUS METALS OR ALLOYS; MAKING METAL MALLEABLE, e.g. BY DECARBURISATION OR TEMPERING
- C21D1/00—General methods or devices for heat treatment, e.g. annealing, hardening, quenching or tempering
- C21D1/26—Methods of annealing
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Chemical & Material Sciences (AREA)
- Theoretical Computer Science (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Thermal Sciences (AREA)
- Crystallography & Structural Chemistry (AREA)
- Mechanical Engineering (AREA)
- Materials Engineering (AREA)
- Metallurgy (AREA)
- Organic Chemistry (AREA)
- Control Of Heat Treatment Processes (AREA)
Abstract
The invention relates to a method for predicting an annealing temperature set value of a transition steel coil based on LSTM, which comprises the following steps: acquiring historical production data of the unit; carrying out data preprocessing; constructing an LSTM model, then introducing the preprocessed data serving as input data into the LSTM model for training to obtain an optimal model parameter which enables the overall error of a training sample to be minimum, and taking the optimal model parameter as a final prediction model; predicting the set temperature of the transition steel coil of the current unit by using a prediction model: and (3) forming multidimensional characteristics by using the parameters from the current coil to the subsequent 3-6 steel coils as input variables, inputting the input variables into an LSTM model, and outputting a predicted value as a transition temperature set value of each coil. The method is applied to a continuous annealing unit and a continuous hot galvanizing unit, and has good prediction effect on a plurality of steel coils with continuous temperature changes, especially on the complex situation that a plurality of parameters such as specification, process target temperature and the like change simultaneously.
Description
Technical Field
The application belongs to the technical field of automatic control, and particularly relates to a prediction method of a transition steel coil annealing temperature set value based on an LSTM long-time memory deep learning network.
Background
With the increasing diversification of cold-rolled or galvanized automobile sheet products, for an annealing furnace, the diversified product performance and specification requirements mean that frequent temperature transition is required, and how to achieve stable temperature transition is a key factor influencing the stable operation of the annealing furnace and the product quality.
In the prior art, the transition setting of the strip steel temperature of the continuous annealing furnace basically has two modes, namely a computer automatic control mode and a manual control mode. The computer automatic control mode requires that the temperature difference of the front transition strip steel and the rear transition strip steel is less than 20 ℃, and only the parameters of the next coil of strip steel can be automatically adjusted. At present, the transition temperature required in actual production is basically more than 20 ℃, and the optimal transition temperature is determined after the process temperature of a plurality of steel coils to be produced is analyzed. If the current coil process target temperature is 730 ℃, the second coil process target temperature and the third coil process target temperature are 700 ℃ and 750 ℃ respectively, and the fourth coil process target temperature is 850 ℃, the conventional computer model cannot judge the complex change condition, and can only finish the annealing temperature setting of the whole transition steel coil according to the step temperature rising route of 730-.
The selection of the temperature transition time point is inaccurate due to insufficient experience of operators, or the mechanical property of the product is unqualified due to the fact that the actual temperature of the steel coil does not reach the process required temperature, and then waste products or degraded products are generated; or the actual temperature of the steel coil is higher than the process required temperature, so that the gas consumption is higher and the energy is wasted. Therefore, how to develop an accurate annealing furnace strip steel temperature transition model is always a technical problem to be solved urgently by technical personnel in the field, and is also one of the key technologies of unattended full-automatic control of the annealing furnace of the continuous annealing unit or the hot galvanizing unit.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a method for predicting the annealing temperature set value of the transition coil of the continuous annealing furnace based on the long-time and short-time memory deep learning network LSTM, which can judge the optimal transition set temperature according to a plurality of steel coil parameters to be produced.
In order to solve the problems, the technical scheme adopted by the invention is as follows:
a prediction method of a transition steel coil annealing temperature set value based on LSTM comprises the following steps:
step S1, obtaining historical production data of the unit;
step S2, preprocessing historical production data to be used as a training sample;
s3, constructing a long-time memory network (LSTM) -based deep neural network model, which is called the LSTM model for short, and then transmitting the data preprocessed in the S2 as input data into the LSTM model for training to obtain an optimal model parameter which enables the overall error of a training sample to be minimum, wherein the optimal model parameter is used as a final prediction model;
step S4, predicting the set temperature of the current set of transition steel coil by using the prediction model (i.e. the trained LSTM model): and (3) forming multidimensional characteristics by using the parameters from the current coil to the subsequent 3-6 steel coils as input variables, inputting the input variables into an LSTM model, and outputting a predicted value as a transition temperature set value of each coil.
The technical scheme of the invention is further improved as follows: the historical production data in the step S1 includes the grade of the steel coil, the process target temperature, the upper limit of the process target temperature, the lower limit of the process target temperature, the thickness of the steel coil, the width of the steel coil, the length of the steel coil, the reflectivity coefficient of the steel grade, the actual annealing temperature, the actual production speed and the annealing temperature transition mark, and the historical production data is sorted in descending order according to the production time.
The technical scheme of the invention is further improved as follows: in step S2, the preprocessing means: and removing the steel coil information with abnormal production line speed and degraded surface grade, and filling missing data, wherein the filling method comprises the step of taking the average value of the same steel grade series or the empirical value of the filling process.
The technical scheme of the invention is further improved as follows: and performing one-hot single coding on the steel coil mark.
The technical scheme of the invention is further improved as follows: in step S3, a sliding window with a time step of 1 and a window size of 3-6 is designed in the LSTM neural network to segment the steel coil data set, convert the two-dimensional data structure into a three-dimensional data structure, then normalize the data, finally transmit the processed data into the LSTM model for training, and obtain the optimal model parameters that minimize the overall error of the current training data through appropriate parameter adjustment.
The technical scheme of the invention is further improved as follows: the loss function is MSE and the optimizer is RMSprop or Adam.
Due to the adoption of the technical scheme, the invention has the beneficial effects that: the method is mainly applied to a continuous annealing unit and a continuous hot galvanizing unit, has good prediction effect on a plurality of steel coils with continuous temperature changes, especially on the complex situation that a plurality of parameters such as specification, process target temperature and the like change simultaneously, has high prediction precision, and has extremely high popularization and application values.
Drawings
Fig. 1 is a comparison graph of the actual effect of the present invention.
Detailed Description
The present invention will be described in further detail with reference to examples.
The invention discloses a prediction method of a transition steel coil annealing temperature set value based on LSTM, which comprises the following steps:
step S1, acquiring historical production actual performance data of the unit; if the annual steel coil production data of the unit is inquired from the production management system, each piece of steel coil information comprises the following parameters: the method comprises the following steps of steel coil mark, process target temperature upper limit, process target temperature lower limit, steel coil thickness, steel coil width, steel coil length, reflectivity coefficient, annealing temperature transition mark, actual annealing temperature, production speed and production time, wherein the steel coil data sets are sorted in a descending order according to the production time.
And step S2, performing necessary preprocessing on the historical data, including removing the steel coil information with abnormal production line speed and degraded surface grade, and filling missing data. If the steel coil information with the production speed lower than 45m/min is removed, the steel coil information with the surface grade of non-first-grade products is removed. And performing one-hot single coding on the steel coil mark. And taking the grade of the steel coil, the process target temperature, the upper limit of the process target temperature, the lower limit of the process target temperature, the thickness of the steel coil, the width of the steel coil, the length of the steel coil, the reflectivity coefficient and the production speed as the X input characteristic of the LSTM neural network, and taking the actual annealing temperature as the Y input characteristic for the training of the LSTM neural network.
And S3, constructing a deep neural network model based on the long-time memory network LSTM, and then transmitting the X and Y data in the step S2 as input data into the LSTM model for training to obtain the optimal model parameter which enables the overall error of the training sample to be minimum, and taking the optimal model parameter as a final prediction model. In the step, a sliding window with the time step of 1 and the window size of 3-6 is designed in an LSTM neural network to divide a steel coil data set, a two-dimensional data structure is converted into a three-dimensional data structure, then data is normalized, finally the processed data is transmitted into an LSTM model to be trained, a loss function is an MSE mean square error function, an optimizer is RMSprop or Adam, and through proper parameter adjustment, the proper parameter adjustment refers to the adjustment of the number of hidden layers of the model, the number of neurons of the hidden layers, the learning rate, the number of training batches, batch size, the training period epoch and other parameters, so that a model parameter which enables the overall error of the current training data to be minimum is obtained.
Step S4, predicting the target annealing temperature of the transition coil to be produced entering the unit at present by using the trained LSTM model: and (3) forming multidimensional characteristics by using the parameters from the current coil to the subsequent 3-6 steel coils as input variables, inputting the input variables into an LSTM model, and outputting a predicted value as the target annealing temperature of the next coil.
After the prediction method is used, the prediction effect is good, the prediction precision is high, and the prediction effect graph is shown in figure 1. Fig. 1 predicts the annealing temperature settings for the first 6 coils (simulating the 6 coils to be produced entering the production train). Fig. 1 shows three lines of process target temperature, LSTM predicted transition temperature, and manual set temperature. Because the steel coil process temperature changes greatly, the annealing temperature is low before and high after, the annealing furnace automatic control model cannot be used in the conventional production, and an operator needs to manually adjust the annealing set temperature of each transition steel coil. The prediction result of the annealing temperature set value of the transition steel coil based on the LSTM is similar to that of manual setting, is closer to the process target temperature of the steel coil than the manual set value, has a more reasonable and more scientific temperature transition route, and realizes the function of automatically and accurately setting the transition temperature. The model can replace manual operation, and solves the key technology of full-automatic unattended control of the annealing furnace of the continuous annealing unit or the hot galvanizing unit.
Claims (6)
1. A prediction method of a transition steel coil annealing temperature set value based on LSTM is characterized by comprising the following steps: the method comprises the following steps:
step S1, obtaining historical production data of the unit;
step S2, preprocessing historical production data to be used as a training sample;
s3, constructing a long-time memory network (LSTM) -based deep neural network model, which is called the LSTM model for short, and then transmitting the data preprocessed in the S2 as input data into the LSTM model for training to obtain an optimal model parameter which enables the overall error of a training sample to be minimum, wherein the optimal model parameter is used as a final prediction model;
step S4, predicting the set temperature of the transition steel coil of the current unit by using a prediction model: and (3) forming multidimensional characteristics by using the parameters from the current coil to the subsequent 3-6 steel coils as input variables, inputting the input variables into an LSTM model, and outputting a predicted value as a transition temperature set value of each coil.
2. The method for predicting the annealing temperature set value of the transition steel coil based on the LSTM according to claim 1, wherein the method comprises the following steps: the historical production data in the step S1 includes the grade of the steel coil, the process target temperature, the upper limit of the process target temperature, the lower limit of the process target temperature, the thickness of the steel coil, the width of the steel coil, the length of the steel coil, the reflectivity coefficient of the steel grade, the actual annealing temperature, the actual production speed and the annealing temperature transition mark, and the historical production data is sorted in descending order according to the production time.
3. The method for predicting the annealing temperature set value of the transition steel coil based on the LSTM according to claim 1, wherein the method comprises the following steps: in step S2, the preprocessing means: and removing the steel coil information with abnormal production line speed and degraded surface grade, and filling missing data.
4. The method for predicting the annealing temperature set value of the transition steel coil based on the LSTM according to claim 3, wherein the method comprises the following steps: and performing one-hot single coding on the steel coil mark.
5. The method for predicting the annealing temperature set value of the transition steel coil based on the LSTM according to claim 1, wherein the method comprises the following steps: in step S3, a sliding window with a time step of 1 and a window size of 3-6 is designed in the LSTM neural network to segment the steel coil data set, convert the two-dimensional data structure into a three-dimensional data structure, then normalize the data, finally transmit the processed data into the LSTM model for training, and obtain the optimal model parameters that minimize the overall error of the current training data through parameter adjustment.
6. The method for predicting the annealing temperature set value of the transition steel coil based on the LSTM according to claim 5, wherein the method comprises the following steps: the loss function is MSE and the optimizer is RMSprop or Adam.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010519714.9A CN111676365A (en) | 2020-06-09 | 2020-06-09 | LSTM-based method for predicting annealing temperature set value of transition steel coil |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010519714.9A CN111676365A (en) | 2020-06-09 | 2020-06-09 | LSTM-based method for predicting annealing temperature set value of transition steel coil |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111676365A true CN111676365A (en) | 2020-09-18 |
Family
ID=72455009
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010519714.9A Pending CN111676365A (en) | 2020-06-09 | 2020-06-09 | LSTM-based method for predicting annealing temperature set value of transition steel coil |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111676365A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112532717A (en) * | 2020-11-25 | 2021-03-19 | 四川易诚智讯科技有限公司 | Production process safety monitoring method based on STM32 single chip microcomputer and long-short time memory network |
| CN113033974A (en) * | 2021-03-08 | 2021-06-25 | 华院计算技术(上海)股份有限公司 | Digital steel coil characteristic quality prediction and process parameter dynamic adjustment method and system based on improved LSTM network |
| CN114880951A (en) * | 2022-06-06 | 2022-08-09 | 浙江理工大学 | Fabric flaw prediction method based on digital twinning |
| CN116636423A (en) * | 2023-07-26 | 2023-08-25 | 云南农业大学 | Efficient cultivation method of poria cocos strain |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108251591A (en) * | 2018-01-15 | 2018-07-06 | 上海大学 | Utilize the top bottom blowing converter producing process control method of LSTM systems |
| CN108672504A (en) * | 2018-05-25 | 2018-10-19 | 中冶南方工程技术有限公司 | A kind of cold-strip steel sensing heating coil of strip transition temperature control method |
| CN108803576A (en) * | 2018-07-24 | 2018-11-13 | 广东工业大学 | A kind of fault early warning method and relevant apparatus of temperature control system |
| US20190093186A1 (en) * | 2017-09-27 | 2019-03-28 | International Business Machines Corporation | Manufacturing process control with deep learning-based predictive model for hot metal temperature of blast furnace |
| CN110205427A (en) * | 2019-06-20 | 2019-09-06 | 东北大学 | A kind of intelligence hot-blast stove Optimal Control System and method |
| KR20190136571A (en) * | 2018-05-31 | 2019-12-10 | 주식회사 포스코 | Prediction apparatus for iron loss reduction of electric steel sheet |
| CN110598958A (en) * | 2019-10-10 | 2019-12-20 | 武汉科技大学 | Steel ladle grading management analysis method and system |
-
2020
- 2020-06-09 CN CN202010519714.9A patent/CN111676365A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190093186A1 (en) * | 2017-09-27 | 2019-03-28 | International Business Machines Corporation | Manufacturing process control with deep learning-based predictive model for hot metal temperature of blast furnace |
| CN108251591A (en) * | 2018-01-15 | 2018-07-06 | 上海大学 | Utilize the top bottom blowing converter producing process control method of LSTM systems |
| CN108672504A (en) * | 2018-05-25 | 2018-10-19 | 中冶南方工程技术有限公司 | A kind of cold-strip steel sensing heating coil of strip transition temperature control method |
| KR20190136571A (en) * | 2018-05-31 | 2019-12-10 | 주식회사 포스코 | Prediction apparatus for iron loss reduction of electric steel sheet |
| CN108803576A (en) * | 2018-07-24 | 2018-11-13 | 广东工业大学 | A kind of fault early warning method and relevant apparatus of temperature control system |
| CN110205427A (en) * | 2019-06-20 | 2019-09-06 | 东北大学 | A kind of intelligence hot-blast stove Optimal Control System and method |
| CN110598958A (en) * | 2019-10-10 | 2019-12-20 | 武汉科技大学 | Steel ladle grading management analysis method and system |
Non-Patent Citations (1)
| Title |
|---|
| 宋志超: ""热镀锌机组锌层厚度自动控制技术"", 《河北冶金》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112532717A (en) * | 2020-11-25 | 2021-03-19 | 四川易诚智讯科技有限公司 | Production process safety monitoring method based on STM32 single chip microcomputer and long-short time memory network |
| CN113033974A (en) * | 2021-03-08 | 2021-06-25 | 华院计算技术(上海)股份有限公司 | Digital steel coil characteristic quality prediction and process parameter dynamic adjustment method and system based on improved LSTM network |
| CN113033974B (en) * | 2021-03-08 | 2022-02-18 | 华院计算技术(上海)股份有限公司 | Digital steel coil characteristic quality prediction and process parameter dynamic adjustment method and system based on improved LSTM network |
| CN114880951A (en) * | 2022-06-06 | 2022-08-09 | 浙江理工大学 | Fabric flaw prediction method based on digital twinning |
| CN114880951B (en) * | 2022-06-06 | 2023-04-07 | 浙江理工大学 | Fabric flaw prediction method based on digital twinning |
| CN116636423A (en) * | 2023-07-26 | 2023-08-25 | 云南农业大学 | Efficient cultivation method of poria cocos strain |
| CN116636423B (en) * | 2023-07-26 | 2023-09-26 | 云南农业大学 | Efficient cultivation method of poria cocos strain |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111676365A (en) | LSTM-based method for predicting annealing temperature set value of transition steel coil | |
| CN111353656B (en) | Steel enterprise oxygen load prediction method based on production plan | |
| CN104846306B (en) | A kind of Zinc Coating Thickness control system and method | |
| CN110814050B (en) | Rolling mill model control method based on BP neural network | |
| CN110789128B (en) | Additive manufacturing part forming quality prediction and control system and method | |
| CN109047683B (en) | Continuous casting billet intelligence basis weight scale on-line control system | |
| CN102540879A (en) | Multi-target evaluation optimization method based on group decision making retrieval strategy | |
| CN100445901C (en) | Dynamic cost control method for industrial process of procedure based on AR(p)model | |
| CN110991916A (en) | Casting blank quality judgment system and method | |
| CN113849020B (en) | Billet heating curve design method and device based on artificial intelligence algorithm | |
| CN111008477A (en) | Method for adjusting technological parameters based on mechanical properties of cold-rolled galvanized strip steel | |
| CN110263494A (en) | A kind of alumina producing operation optimizing system and method based on the collaboration of cloud side | |
| CN108265157A (en) | Intelligent arc furnace steelmaking system | |
| CN114888092B (en) | Cold rolling deformation resistance prediction method based on cross-process data platform | |
| CN106874568B (en) | A kind of material particular diameter distribution forecasting method of mechanical milling process | |
| CN113325819B (en) | Continuous annealing unit fault diagnosis method and system based on deep learning algorithm | |
| CN111258984A (en) | Product quality end-edge-cloud collaborative forecasting method under industrial big data environment | |
| CN109240203B (en) | Multi-model-based continuous casting billet constant weight control method | |
| CN116703254B (en) | Production information management system for mechanical parts of die | |
| CN112845610A (en) | Steel rolling power consumption parameter recommendation method and system | |
| CN110837959A (en) | Method for balancing and fixing welding dispatching working hours based on welding quantity and operation mode | |
| CN110543942A (en) | Multi-space-time long and short memory depth network accurate prediction method | |
| CN116984380A (en) | Method for forecasting and controlling width narrowing quantity of cold-rolled full-process strip steel | |
| CN116705211B (en) | Digital twin-based online prediction method and system for copper loss rate of oxygen-enriched copper molten pool | |
| CN117193223B (en) | Plastic product production control system |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200918 |
|
| RJ01 | Rejection of invention patent application after publication |