A method for accurately controlling the amount of temperature control agent added in the TSC stage of a large converter

Abstract

The present invention relates to a method for accurately controlling the amount of temperature regulating agent added in the TSC stage of a large converter, and belongs to the technical field of metallurgical control methods. The technical scheme of the present invention is: after the smelting plan is issued, the temperature regulating agent control coefficient is calculated, and the predicted value of the temperature regulating agent addition amount of the planned production furnace is calculated according to the BP neural network under the guidance of the heat balance model, and the temperature regulating agent addition amount is finally obtained according to the temperature regulating agent coefficient and the BP neural network under the guidance of the heat balance model. The beneficial effects of the present invention are: constructing the internal connection between the various influencing factors of the converter, and realizing the prediction of the temperature regulating agent addition amount of the TSC stage of the large converter under different material conditions, solving the problem of temperature control in the TSC stage of the large converter under unstable conditions where the material type and material composition vary greatly and the process operation level of the operators is different, reducing quality accidents and production accidents, and reducing smelting costs.

Description

Technical Field

[0001] The invention relates to a method for accurately controlling the amount of a temperature regulating agent added in a TSC stage of a large converter, and belongs to the technical field of metallurgical control methods. Background Art

[0002] For the converter smelting process using auxiliary gun control, the smelting process is divided into two stages, with the auxiliary gun using TSC probe measurement as the dividing point, and the smelting stage before the auxiliary gun measurement is called the TSC stage. Temperature control is the main task of the TSC stage. Reasonable temperature control in this stage facilitates the stable control of the converter terminal temperature control, terminal carbon content control, terminal phosphorus content and terminal oxygen content of molten steel, which is conducive to the stable control of the overall converter process, the stability of product quality, and the reduction of smelting costs.

[0003] The temperature of the converter TSC stage is mainly regulated by a temperature regulator. When the scrap steel is quantitatively loaded into the converter, the heat is generally sufficient, and ore is mainly used as a coolant to regulate the smelting temperature. However, under the condition of a large scrap steel ratio, insufficient heat will occur, and ferrosilicon is used as a temperature-raising agent.

[0004] The type and amount of temperature control agent added are affected by the smelting level of the team, the age of the converter, the amount of scrap steel added and the stability of the material. It involves the heat of dissolution of different materials, the heat of chemical reaction of different elements, the heat loss of splashing during the smelting process, the heat loss of flue gas and other issues. There are many uncertain factors and it is impossible to predict them using ordinary linear laws. Especially under the increasingly stringent environmental protection requirements, steelmaking processes generally use steel slag materials. The composition and temperature drop effect of this material are unstable, and its exact effect on temperature cannot be predicted by theoretical calculation. When the amount of this material added increases, it will not only cause abnormal temperature control in the TSC stage, but also cause converter spraying in severe cases.

[0005] At the same time, since domestic steel mills are purchasing large quantities of scrap steel, the amount of scrap steel resources for smelting is unstable, the amount of scrap steel added to the converter fluctuates, and the ordinary model calculation adaptability is poor, resulting in the inability to accurately predict the amount of temperature control agent added in the TSC stage, which ultimately leads to deviations in the converter endpoint control.

[0006] The patent application number 201810410253.4 discloses "a method for controlling the final carbon content and temperature of converter steelmaking", in which the final carbon content and temperature are controlled in the TSC stage based on the characteristic parameters of the initial state of the molten iron and the requirements of the target steel grade to be smelted, based on the oxygen blowing amount prediction model and the final carbon content prediction model.

[0007] Patent application number 201611020342.5 discloses "a method and server for predicting converter steelmaking temperature". This method calculates the first molten steel temperature based on a theoretical model established according to the principles of material balance and heat balance, and calculates the second molten steel temperature based on the current furnace information and a first preset number of historical furnace information, and finally evaluates the endpoint temperature by combining the first temperature and the second temperature.

[0008] Reference [Liu Dongmei, Zou Zongshu, Yu Aibing. Fuzzy neural network converter terminal temperature prediction model based on subtractive clustering, Journal of Materials and Metallurgy, 2006] established a prediction model through molten iron carbon, lime addition, furnace oxygen consumption, ore addition, molten iron initial temperature and scrap steel ratio.

[0009] The above literature only considers the influence of coolant, i.e. ore, on the converter smelting temperature, and does not involve the calculation of heating agent addition under the condition of large scrap ratio, the influence of unstable materials such as slag on temperature, and the influence of different steel grades, different furnace age conditions, and different operator technical levels on temperature control. It is impossible to accurately control the TSC stage temperature of a large converter.

[0010] Therefore, it is necessary to find a reasonable method to accurately control the amount of temperature regulator added under different processes and different material conditions, and ultimately solve the temperature control problem in the TSC stage. Summary of the invention

[0011] The purpose of the present invention is to provide a method for accurately controlling the amount of temperature regulating agent added in the TSC stage of a large-scale converter. A temperature regulating agent addition model with a temperature adjustment coefficient A is adopted. Through heat balance calculation and BP neural network intelligent algorithm, the internal connection between various influencing factors of the converter is constructed, so as to achieve the prediction of the amount of temperature regulating agent added in the TSC stage of a large-scale converter under different material conditions. The problem of temperature control in the TSC stage of a large-scale converter under unstable conditions where the material type and material composition vary greatly and the process operation level of operators is different is solved, quality accidents and production accidents are reduced, smelting costs are reduced, and the above-mentioned problems existing in the background technology are effectively solved.

[0012] The technical solution of the present invention is: a method for accurately controlling the amount of temperature control agent added in the TSC stage of a large converter, comprising the following steps:

[0013] (1) After the smelting plan is issued, the temperature control coefficient is calculated. The calculation formula is: j ≤60min, When 60min<T j ≤120min, When 120min<T j ≤240min, When T j >240min,

[0014] Where: A is the temperature control coefficient, T j is the production time interval between the previous furnace and the predicted furnace; Ms is the actual amount of temperature control agent added; My is the predicted amount of temperature control agent added;

[0015] (2) The predicted value of the amount of temperature control agent added for the planned production batch is calculated using the BP neural network under the guidance of the heat balance model, wherein the heat balance model formula for the amount of temperature control agent added is as follows:

[0016] Q s =Q stw +Q sty +Q y

[0017] Where: Q s is the heat income, Q stw is the physical heat of molten iron, Q sty is the elemental heat of molten iron, Q y for smoke and dust fever;

[0018] Q z =Q zg +Q f +Q lz +Q lq +Q zy +Q zt +Q pj +Q ss

[0019] Where: Q z is the heat expenditure, Q zg is the physical heat of molten steel, Q f Heat for scrap steel, Q lz Heat dissipated for slag, Q lq Heat for furnace gas, Q zy Heat is expended for smoke and dust, Q zt The slag iron beads eject heat, Q pj To dissipate heat for splashing, Q ss It is the heat loss of smelting;

[0020] M=(|Q s -Q z |) / k

[0021] In the formula: M is the amount of temperature control agent added, k is the calorific value of the temperature control agent;

[0022] (3) According to the temperature control agent coefficient and the BP neural network calculation guided by the heat balance model, the temperature control agent addition amount is finally obtained. The temperature control agent addition amount calculation formula is as follows:

[0023] Mt=Mm×(1-A×500)

[0024] Where: Mt is the final predicted amount of temperature control agent added; Mm is the amount of temperature control agent added predicted by the temperature control agent model; A is the temperature control agent control coefficient.

[0025] In the step (1), the training data involved in the calculation of the temperature control coefficient A is the first three furnaces of the actual production batch, and the training data screening rules are as follows: ① the calculation data is the first three furnaces of the planned production, and they are produced in the same furnace seat; ② when there is |A i -A|≥35, the furnace will be deleted from the training data, and the 4th furnace before the predicted furnace will be added to the training data to participate in the prediction, and so on; ③ When the 8th furnace is screened, if there are still less than 3 furnaces, stop screening and select the coefficient used for the previous furnace for calculation.

[0026] In the step (2), when Q s >Q z When adding ore cooling agent, k takes the value of 5235313; when Q s <Q z When ferrosilicon heating agent is added, the value of k is 3641289.

[0027] In the step (2), the BP neural network model performs prediction, and the main structure is that the input is 17-dimensional data and the output is 1-dimensional data. The 17-dimensional input data are respectively the amount of temperature regulating agent added in the mechanism model, the team, the furnace seat, the converter life, scrap steel, molten iron temperature, molten iron weight, molten iron manganese, molten iron silicon, molten iron phosphorus, slag, dolomite addition, lime addition, TSC phosphorus, TSC manganese, TSC carbon and TSC temperature, and the 1-dimensional output data is the amount of temperature regulating agent added; the input data is standardized based on the mechanism model, and the processing method is as follows:

[0028] ① Based on the historical data set, obtain the data characteristics of the historical data and the necessary physical metallurgical principles to define the maximum value y max and the minimum value y min .

[0029] ②When encountering new data, normalize it according to historical data alignment:

[0030]

[0031] y is the parameter value to be standardized. max is the maximum value of this set of parameters, y min is the customized minimum value of this set of parameters;

[0032] The neural network model structure is a 3-hidden layer structure. The number of nodes in the first hidden layer is 5, the number of nodes in the second hidden layer is 20, and the number of nodes in the third hidden layer is 15.

[0033] In order to improve data accuracy, the predicted training data is screened. The screening rules are as follows:

[0034] ① The time interval between the molten iron temperature measurement and the molten iron loading into the converter is greater than 3 hours, and the training data group is selected;

[0035] ② For the heats where the hot metal is desulfurized by hot metal injection, if the temperature of the hot metal is missing after pretreatment, the hot metal temperature minus 20°C is used in the training model;

[0036] ③ Calculate the oxygen blowing time by the oxygen blowing amount and oxygen flow rate of the oxygen lance, and compare it with the refining time of the first refining. If the time difference is greater than ±5min, select the training data group.

[0037] The beneficial effects of the present invention are: adopting a temperature regulating agent addition model with a temperature regulating coefficient A, through heat balance calculation and BP neural network intelligent algorithm, the internal connection between various influencing factors of the converter is constructed, and the amount of temperature regulating agent added in the TSC stage of a large converter is predicted under different material conditions. This solves the temperature control problem of the TSC stage of a large converter under unstable conditions where the material type and material composition vary greatly and the process operation levels of operators are different, thereby reducing quality accidents and production accidents and reducing smelting costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a work flow chart of the present invention;

[0039] Figure 2 It is the BP neural network architecture diagram of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the invention implementation cases clearer, the technical solutions in the implementation cases of the present invention will be clearly and completely described below in conjunction with the drawings in the implementation cases. Obviously, the implementation cases described are only a small part of the implementation cases of the present invention, rather than all the implementation cases. Based on the implementation cases in the present invention, all other implementation cases obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0041] A method for accurately controlling the amount of temperature control agent added in the TSC stage of a large converter comprises the following steps:

[0042] (1) After the smelting plan is issued, the temperature control coefficient is calculated. The calculation formula is: j ≤60min, When 60min<T j ≤120min, When 120min<T j ≤240min, When T j >240min,

[0043] Where: A is the temperature control coefficient, T j is the production time interval between the previous furnace and the predicted furnace; Ms is the actual amount of temperature control agent added; My is the predicted amount of temperature control agent added;

[0044] (2) The predicted value of the amount of temperature control agent added for the planned production batch is calculated using the BP neural network under the guidance of the heat balance model, wherein the heat balance model formula for the amount of temperature control agent added is as follows:

[0045] Q s =Q stw +Q sty +Q y

[0046] Where: Q s is the heat income, Q stw is the physical heat of molten iron, Q sty is the elemental heat of molten iron, Q y for smoke and dust fever;

[0047] Q z =Q zg +Q f +Q lz +Q lq +Q zy +Q zt +Q pj +Q ss

[0048] Where: Q z is the heat expenditure, Q zg is the physical heat of molten steel, Q f Heat for scrap steel, Q lz Heat dissipated for slag, Q lq Heat for furnace gas, Q zy Heat is expended for smoke and dust, Q zt The slag iron beads eject heat, Q pj To dissipate heat for splashing, Q ss It is the heat loss of smelting;

[0049] M=(|Q s -Q z |) / k

[0050] In the formula: M is the amount of temperature control agent added, k is the calorific value of the temperature control agent;

[0051] (3) According to the temperature control agent coefficient and the BP neural network calculation guided by the heat balance model, the temperature control agent addition amount is finally obtained. The temperature control agent addition amount calculation formula is as follows:

[0052] Mt=Mm×(1-A×500)

[0053] Where: Mt is the final predicted amount of temperature control agent added; Mm is the amount of temperature control agent added predicted by the temperature control agent model; A is the temperature control agent control coefficient.

[0054] In the step (1), the training data involved in the calculation of the temperature control coefficient A is the first three furnaces of the actual production batch, and the training data screening rules are as follows: ① the calculation data is the first three furnaces of the planned production, and they are produced in the same furnace seat; ② when there is |A i -A|≥35, the furnace will be deleted from the training data, and the 4th furnace before the predicted furnace will be added to the training data to participate in the prediction, and so on; ③ When the 8th furnace is screened, if there are still less than 3 furnaces, stop screening and select the coefficient used for the previous furnace for calculation.

[0055] In the step (2), when Q s >Q z When adding ore cooling agent, k takes the value of 5235313; when Q s <Q z When ferrosilicon heating agent is added, the value of k is 3641289.

[0056] In the step (2), the BP neural network model performs prediction, and the main structure is that the input is 17-dimensional data and the output is 1-dimensional data. The 17-dimensional input data are respectively the amount of temperature regulating agent added in the mechanism model, the team, the furnace seat, the converter life, scrap steel, molten iron temperature, molten iron weight, molten iron manganese, molten iron silicon, molten iron phosphorus, slag, dolomite addition, lime addition, TSC phosphorus, TSC manganese, TSC carbon and TSC temperature, and the 1-dimensional output data is the amount of temperature regulating agent added; the input data is standardized based on the mechanism model, and the processing method is as follows:

[0057] ① Based on the historical data set, obtain the data characteristics of the historical data and the necessary physical metallurgical principles to define the maximum value y max and the minimum value y min .

[0058] ②When encountering new data, normalize it according to historical data alignment:

[0059]

[0060] y is the parameter value to be standardized. max is the maximum value of this set of parameters, y min is the customized minimum value of this set of parameters;

[0061] The neural network model structure is a 3-hidden layer structure. The number of nodes in the first hidden layer is 5, the number of nodes in the second hidden layer is 20, and the number of nodes in the third hidden layer is 15.

[0062] In order to improve data accuracy, the predicted training data is screened. The screening rules are as follows:

[0063] ① The time interval between the molten iron temperature measurement and the molten iron loading into the converter is greater than 3 hours, and the training data group is selected;

[0064] ② For the heats where the hot metal is desulfurized by hot metal injection, if the temperature of the hot metal is missing after pretreatment, the hot metal temperature minus 20°C is used in the training model;

[0065] ③ Calculate the oxygen blowing time by the oxygen blowing amount and oxygen flow rate of the oxygen lance, and compare it with the refining time of the first refining. If the time difference is greater than ±5min, select the training data group.

[0066] Example:

[0067] The implementation unit has three 260t top and bottom blown converters, namely 1, 2, and 3. All of them are controlled by auxiliary guns, and the operators are in four teams, 1, 2, 3, and 4. The raw material information is as follows:

[0068]

[0069] The amount of auxiliary materials added in the TSC stage, TSC temperature, TSC carbon and temperature regulator added are as follows:

[0070]

[0071] From the prediction results of the embodiment, it can be observed that under different furnace ages, different shifts, and different loading conditions, the deviation between the predicted amount of ore added and the actual amount added is within 2t, the deviation of ferrosilicon is within 200kg, and the deviation of the actual temperature from the set temperature is within ±15°C, which meets the requirements for on-site use.

Claims

1. A method for accurately controlling the amount of temperature control agent added in the TSC stage of a large converter, characterized in that The following steps are involved: (1) After the smelting plan is issued, the temperature control coefficient is calculated. The calculation formula is: j ≤60min, When 60min<T j ≤120min, When 120min<T j ≤240min, When T j >240min, Where: A is the temperature control coefficient, T j is the production time interval between the previous furnace and the predicted furnace; Ms is the actual amount of temperature control agent added; My is the predicted amount of temperature control agent added; (2) The predicted value of the amount of temperature control agent added for the planned production batch is calculated using the BP neural network under the guidance of the heat balance model, wherein the heat balance model formula for the amount of temperature control agent added is as follows: Q s =Q stw +Q sty +Q y Where: Q s is the heat income, Q stw is the physical heat of molten iron, Q sty The elemental heat of molten iron, Q y for smoke and dust fever; Q z =Q zg +Q f +Q lz +Q lq +Q zy +Q zt +Q pj +Q ss Where: Q z is the heat expenditure, Q zg is the physical heat of molten steel, Q f Heat for scrap steel, Q lz Heat dissipated for slag, Q lq Heat for furnace gas, Q zy Heat is expended for smoke and dust, Q zt The slag iron beads eject heat, Q pj To dissipate heat for splashing, Q ss It is the heat loss of smelting; M=(|Q s -Q z |) / k In the formula: M is the amount of temperature control agent added, k is the calorific value of the temperature control agent; (3) According to the temperature control agent coefficient and the BP neural network calculation guided by the heat balance model, the temperature control agent addition amount is finally obtained. The temperature control agent addition amount calculation formula is as follows: Mt=Mm×(1-A×500) Where: Mt is the final predicted amount of temperature control agent added; Mm is the amount of temperature control agent added predicted by the temperature control agent model; A is the temperature control agent control coefficient.

2. The method for accurately controlling the amount of temperature control agent added in the TSC stage of a large converter according to claim 1, characterized in that: In the step (1), the training data involved in the calculation of the temperature regulating agent control coefficient A are the first three furnaces of the actual production furnace, and the training data screening rules are as follows: ① the calculation data are the first three furnaces of the planned production, and are produced by the same furnace seat; ② when |Ai-A|≥35, the furnace is deleted from the training data, and the fourth furnace before the predicted furnace is added to the training data to participate in the prediction, and so on; ③ when the eighth furnace is screened, if there are still less than 3 furnaces, the screening is stopped and the coefficient used for the previous furnace is selected for calculation.

3. The method for accurately controlling the amount of temperature control agent added in the TSC stage of a large converter according to claim 1, characterized in that: In step (2), when Q s >Q z When adding ore cooling agent, k takes the value of 5235313; when Q s <Q z When ferrosilicon heating agent is added, the value of k is 3641289.

4. The method for accurately controlling the amount of temperature control agent added in the TSC stage of a large converter according to claim 1, characterized in that: In the step (2), the BP neural network model performs prediction. The BP neural network model architecture is that the input is 17-dimensional data and the output is 1-dimensional data. The 17-dimensional input data are respectively the amount of temperature control agent added in the mechanism model, the team, the furnace seat, the converter life, scrap steel, molten iron temperature, molten iron weight, molten iron manganese, molten iron silicon, molten iron phosphorus, slag, dolomite addition, lime addition, TSC phosphorus, TSC manganese, TSC carbon and TSC temperature, and the 1-dimensional output data is the amount of temperature control agent added; the input data is standardized based on the mechanism model, and the processing method is as follows: ① Based on the historical data set, obtain the data characteristics of the historical data and the necessary physical metallurgical principles to define the maximum value y max and the minimum value y min ; ②When encountering new data, normalize it according to historical data alignment: y is the parameter value to be standardized. max is the maximum value of this set of parameters, y min It is the customized minimum value of this set of parameters; The neural network model structure is a 3-hidden layer structure. The number of nodes in the first hidden layer is 5, the number of nodes in the second hidden layer is 20, and the number of nodes in the third hidden layer is 15.

5. The method for accurately controlling the amount of temperature control agent added in the TSC stage of a large converter according to claim 1, characterized in that: In order to improve data accuracy, the predicted training data is screened. The screening rules are as follows: ① The time interval between the molten iron temperature measurement and the molten iron loading into the converter is greater than 3 hours, and the training data group is selected; ② For the heats where the hot metal is desulfurized by hot metal injection, if the temperature of the hot metal is missing after pretreatment, the hot metal temperature minus 20°C is used in the training model; ③ Calculate the oxygen blowing time by the oxygen blowing amount and oxygen flow rate of the oxygen lance, and compare it with the refining time of the first refining. If the time difference is greater than ±5min, select the training data group.

Images