A dynamic soft measurement method and system for arc furnace molten pool temperature

By combining numerical simulation and physical information neural network, dynamic soft measurement and real-time prediction of arc furnace melting pool temperature are achieved, which solves the problem that traditional methods are difficult to reflect temperature changes in real time, improves prediction accuracy and operation control efficiency, and adapts to complex processes Changes reduce costs.

CN118734659BActive Publication Date: 2024-11-12NORTHEASTERN UNIV CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411217533.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-02
Publication Date
2024-11-12
Estimated Expiration
2044-09-02

AI Technical Summary

Technical Problem

Traditional arc furnace melting pool temperature measurement methods are difficult to reflect the dynamic temperature changes in real time, and the numerical simulation prediction accuracy is limited by the model complexity and computing resources, making it difficult to adapt to complex process conditions.

Method used

Combining numerical simulation, physical information neural network and real-time data acquisition, by establishing a numerical simulation model and a time series neural network model of the arc furnace melting pool, dynamic soft measurement and real-time prediction of the arc furnace melting pool temperature, and dynamically adjust the power supply and scrap steel added to control the temperature.

Benefits of technology

It improves the accuracy and real-time performance of melt pool temperature prediction, optimizes the operation control of arc furnaces, adapts to complex process changes, reduces energy consumption and production costs, and provides comprehensive temperature field information and intelligent control means.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118734659B_ABST
    Figure CN118734659B_ABST
Patent Text Reader

Abstract

The present invention belongs to the technical field of intelligent metallurgical engineering, and discloses a dynamic soft measurement method and system for the temperature of an electric arc furnace molten pool. A numerical simulation model of flat molten pool smelting of an electric arc furnace is established according to the size and production data of the electric arc furnace, covering electric arc, bottom blowing, oxygen blowing, carbon injection, scrap steel feeding and foamy slag. The model is corrected by actual molten pool temperature data, and the Monte Carlo method generates a combination of boundary conditions to establish a numerical simulation database. Then, an energy conservation equation and a physical information neural network model are established, and a time series neural network is used for training to obtain a model for real-time prediction. The system collects real-time data, inputs the trained model to predict the molten pool temperature, and adjusts the operation of the electric arc furnace according to the prediction results. The system displays the prediction results, operating parameters and control suggestions through a user interface module. The method and system can effectively improve the accuracy of the prediction of the molten pool temperature of the electric arc furnace, optimize the production process, improve the quality of molten steel, and reduce production costs.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent metallurgical engineering, and in particular to a dynamic soft measurement method and system for arc furnace molten pool temperature. Background Art

[0002] Electric arc furnaces are commonly used smelting equipment in the steel industry. They generate high temperatures through arc discharge to melt scrap steel and other raw materials. However, accurate control of the temperature of the arc furnace molten pool is crucial to improving smelting quality and energy efficiency. Traditional temperature measurement methods mainly rely on thermocouples, which are limited by the limitations of measurement points and measurement frequency, and it is difficult to reflect the dynamic changes of the molten pool temperature in real time.

[0003] Numerical simulation technology is increasingly used in metallurgical engineering. By establishing a numerical simulation model of an electric arc furnace, the temperature of the molten pool can be predicted. However, the prediction accuracy of pure numerical simulation is limited by the complexity of the model and computing resources. In addition, relying solely on numerical simulation is difficult to adapt to the complex and changeable process conditions in the actual production process. Physical information neural network, as an emerging machine learning method, can improve the accuracy and robustness of predictions while ensuring physical consistency by embedding physical equations into the neural network model. By combining physical information neural network with numerical simulation and actual production data, dynamic soft measurement of the temperature of the electric arc furnace molten pool can be achieved, providing support for the optimization control of the smelting process. Summary of the invention

[0004] The purpose of the present invention is to provide a dynamic soft measurement method and system for the temperature of the arc furnace molten pool, which can accurately predict and control the temperature of the arc furnace molten pool by combining numerical simulation, physical information neural network and real-time data acquisition, thereby improving the efficiency of the smelting process and product quality.

[0005] In order to achieve the above object, the present invention provides a dynamic soft measurement method and system for the temperature of an arc furnace molten pool.

[0006] The technical solution of the present invention is as follows: A dynamic soft measurement method for arc furnace molten pool temperature comprises the following steps:

[0007] A1: A numerical simulation model of the arc furnace molten pool is established based on the actual arc furnace size and production data; the magnetohydrodynamic model is used to simulate the arc, the VOF multiphase flow model is used to simulate the bottom blowing, the Euler-Euler multiphase flow model is used to simulate the oxygen blowing, the DPM model is used to simulate the carbon injection, the VOF multiphase flow model is used to simulate the continuous feeding of scrap steel, and the Euler-Euler multiphase flow model is used to simulate the foamed slag;

[0008] Define the arc furnace geometry model, set material properties, define boundary conditions, establish the control equations, and solve them using the finite element method;

[0009] A2: Obtain the actual molten pool temperature of 10 to 20 groups of arc furnaces through thermocouples, and modify the numerical simulation model based on the actual molten pool temperature data;

[0010] A3: According to the production process of the electric arc furnace, the range of variation of the operating process is set as the boundary conditions, including the oxygen blowing amount, carbon injection amount, lime amount, power consumption, scrap steel addition amount and furnace weight; the Monte Carlo method is used to randomly generate n or more sets of boundary condition combinations, which are input into the numerical simulation model corrected in step A2 to obtain the temperature field of the entire process of flat molten pool smelting in the electric arc furnace, and the temperature data of the center point of the molten pool temperature field are extracted;

[0011] A4: Collect the boundary conditions of the numerical simulation in step A3 and the temperature data of the center point of the molten pool temperature field, and establish a numerical simulation database;

[0012] A5: Establish the energy conservation equation for the whole process of flat pool smelting in an electric arc furnace, and construct a physical information neural network model with a time series neural network structure, and embed the energy conservation equation into the physical information neural network model as a constraint condition; read the numerical simulation database in step A4, use the boundary conditions as input variables, and the temperature data of the center point of the molten pool temperature field as output variables, and pass them into the physical information neural network model for training; minimize the error between the model prediction value and the actual measurement value by adjusting the parameters of the physical information neural network model; the trained physical information neural network model is used for real-time soft measurement in flat pool smelting in an electric arc furnace;

[0013] A6: Collect the input variables at the current moment and input them into the physical information neural network model trained in step A5 to obtain the prediction result of the molten pool temperature at the current moment.

[0014] When the magnetohydrodynamics model, VOF multiphase flow model, Euler-Euler multiphase flow model, and DPM model have the same parameter requirements, the simulation results of each model and the set value with the smallest actual error average error are used as parameters.

[0015] The arc furnace geometric model described in step A1 includes the furnace shell structure, furnace lining structure, graphite electrode structure and its arrangement, charging port structure and its arrangement, initial molten pool shape and size, arc column geometry and action area, oxygen lance and bottom lance structure and their arrangement; the material properties include the thermophysical parameters of molten steel, furnace lining, graphite electrode, gas, scrap steel, lime and foam slag; the boundary conditions include arc power and distribution, heat transfer coefficient between arc and molten pool surface, furnace wall heat conduction, furnace wall cooling system heat dissipation, molten pool surface heat dissipation, charging temperature, charging rate, charging position, solid-liquid phase change latent heat, and chemical reaction heat.

[0016] The time series neural network structure described in step A5 adopts the long short-term memory network LSTM or the gated recurrent unit model GRU; the training of the physical information neural network model adopts the back propagation algorithm; the energy conservation equation in the arc furnace is described as follows: ,in: Indicates the mass of the molten pool, unit: kg; Represents the specific heat capacity of the molten pool, unit: J / [kg·K]; Indicates the rate of change of the molten pool temperature over time, unit: K / s; energy input item Including arc input energy 、Oxygen combustion releases heat 、Carbon combustion releases heat , the heat added by scrap steel , energy output term Including heat loss from furnace wall , Radiant heat loss , chemical reaction absorbs heat 、Slag system heat ; The above energy input items and energy output items are obtained through the results of the numerical simulation model;

[0017] After the physical information neural network model in step A5 is established, the numerical simulation model in step A1 is corrected according to the actual temperature measurement data every time t actual temperature measurements are accumulated, and steps A2 to A5 are repeated to update the real-time soft measurement model in flat pool smelting in electric arc furnaces.

[0018] The input variables of step A6 at the current moment include the real-time power consumption, real-time oxygen blowing amount, real-time carbon injection amount, real-time lime amount, real-time scrap steel addition amount, and real-time furnace body weight of the electric arc furnace; the real-time power consumption is acquired through the electric energy meter or power analyzer in the power monitoring system, the real-time oxygen blowing amount is acquired through the gas flow meter and the control unit of the oxygen supply system, the real-time carbon injection amount is acquired through the control unit and weight sensor of the carbon powder injection system, the real-time lime amount is acquired through the control unit and silo weighing system of the lime feeding system, the real-time scrap steel addition amount is acquired through the control unit and hopper weighing system of the scrap steel feeding system, and the real-time furnace body weight is acquired through the load sensor in the furnace body weighing system.

[0019] When smelting is close to the end point, the arc furnace parameters are adjusted according to the comparison between the current molten pool temperature and the target steel tapping temperature;

[0020] If the target steel-tapping temperature is higher than the current molten pool temperature, the power supply of the arc furnace is increased to quickly reach the steel-tapping temperature; if the target steel-tapping temperature is lower than the current molten pool temperature, the power supply of the arc furnace is reduced to lower the molten pool temperature to quickly reach the steel-tapping temperature; when the predicted molten pool temperature reaches the target temperature, prepare for the steel-tapping operation.

[0021] A dynamic soft measurement system for arc furnace molten pool temperature, comprising:

[0022] Numerical simulation module, used to establish a numerical simulation model of the arc furnace molten pool according to the actual arc furnace size and production data;

[0023] A temperature acquisition module, including a thermocouple, is used to obtain actual molten pool temperature data of the electric arc furnace;

[0024] The boundary condition generation module is used to set the variation range of the operation process according to the production process of the electric arc furnace, and randomly generate boundary condition combinations using the Monte Carlo method;

[0025] Database module, used to store the boundary conditions of numerical simulation and the temperature data of the center point of the molten pool temperature field;

[0026] Physical information neural network module, including time series neural network structure and energy conservation equation constraints, is used to establish and train the prediction model of the whole process of flat pool smelting in electric arc furnace;

[0027] Real-time data acquisition module, used to collect input variables at the current moment;

[0028] The temperature prediction module is used to predict the current molten pool temperature based on the input variables collected in real time and the trained physical information neural network model;

[0029] A control module, used to adjust the power supply to the electric arc furnace or add scrap steel according to the predicted molten pool temperature and target tapping temperature;

[0030] The model updating module is used to regularly update the numerical simulation model and the physical information neural network model according to the accumulated actual temperature measurement data.

[0031] The numerical simulation module comprises:

[0032] Arc simulation unit, using magnetohydrodynamic model;

[0033] Bottom blowing simulation unit, using VOF multiphase flow model;

[0034] Oxygen blowing simulation unit, using Euler-Euler multiphase flow model;

[0035] Carbon injection simulation unit, using DPM model;

[0036] The scrap steel feeding simulation unit uses the VOF multiphase flow model;

[0037] The foam slag simulation unit uses the Euler-Euler multiphase flow model;

[0038] A geometric model definition unit is used to define the arc furnace geometric model;

[0039] Material property setting unit, used to set material properties;

[0040] Boundary condition definition unit, used to define boundary conditions;

[0041] The control equation solving unit uses the finite element method to solve the control equations.

[0042] The physical information neural network module adopts the long short-term memory network LSTM or the gated recurrent unit GRU model as the time series neural network structure and uses the back propagation algorithm for training; the real-time data acquisition module includes:

[0043] Power monitoring systems, including energy meters or power analyzers, to collect real-time power consumption;

[0044] Gas flow meter and oxygen supply system control unit, used to collect real-time oxygen blowing volume;

[0045] Carbon powder injection system control unit and weight sensor, used to collect real-time carbon injection amount;

[0046] Lime feeding system control unit and silo weighing system for collecting real-time lime quantity;

[0047] The scrap steel feeding system control unit and hopper weighing system are used to collect the real-time scrap steel feeding amount;

[0048] The furnace weighing system includes a load sensor for collecting real-time furnace weight.

[0049] A user interface module is also included to display the predicted melt pool temperature, current operating parameters and control suggestions, and to allow operators to perform human-machine interaction and manual adjustments.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. Improve the accuracy of melt pool temperature prediction

[0052] The present invention adopts a method combining numerical simulation and physical information neural network to predict the temperature of the arc furnace molten pool in real time. The complex physical process of the arc furnace molten pool is simulated in detail through the numerical simulation model, and the model is corrected using actual measurement data to ensure the accuracy of the model. In addition, the physical information neural network model combines the time series neural network structure and the energy conservation equation to further improve the accuracy and reliability of temperature prediction.

[0053] 2. Real-time soft measurement

[0054] The present invention realizes real-time prediction of the molten pool temperature by collecting the operating parameters of the electric arc furnace in real time and inputting them into the trained physical information neural network model. This real-time soft measurement method overcomes the measurement lag and discontinuity problems existing in traditional temperature measurement methods (such as thermocouples) and provides timely and accurate temperature information for the operation of the electric arc furnace.

[0055] 3. Optimizing arc furnace operation control

[0056] By predicting the molten pool temperature in real time and combining it with the target steel-out temperature, the present invention can dynamically adjust the power supply of the electric arc furnace or the amount of scrap steel added to achieve rapid temperature regulation. This optimization control strategy not only improves smelting efficiency and reduces energy consumption, but also improves molten steel quality and reduces production costs.

[0057] 4. Adapt to complex production process changes

[0058] The present invention generates a large number of different boundary condition combinations through the Monte Carlo method and constructs a widely covered numerical simulation database. In this way, even when the production process changes greatly, the physical information neural network model can accurately predict the molten pool temperature based on the operating parameters collected in real time, and has strong adaptability and robustness.

[0059] 5. Improve the adaptability of the model

[0060] After the physical information neural network model is established, the present invention regularly corrects the numerical simulation model according to the actual temperature measurement data and updates the neural network model parameters to maintain the accuracy and real-time performance of the model. This dynamic update mechanism enables the model to adapt to changes in the production process and improves the long-term stability and reliability of the prediction.

[0061] 6. Provide comprehensive temperature field information

[0062] Through the numerical simulation model, the present invention can provide detailed distribution information of the temperature field in the molten pool, rather than just the temperature of a certain point in the temperature field. This comprehensive temperature field information provides a richer reference for the operation of the electric arc furnace and helps to optimize all aspects of the smelting process.

[0063] 7. Provide intelligent control means

[0064] The present invention combines advanced neural network technology and physical models to predict and control the arc furnace molten pool temperature through intelligent means. Compared with the traditional method that relies on experience and manual adjustment, the intelligent control method has higher accuracy and efficiency, and reduces the impact of human factors on the production process.

[0065] 8. Implemented modular design of the system

[0066] The system design of the present invention includes a numerical simulation module, a temperature acquisition module, a boundary condition generation module, a database module, a physical information neural network module, a real-time data acquisition module, a temperature prediction module, a control module and a model update module. Each module is independent of each other but closely related, has good modularity and expansibility, and is convenient for system maintenance and upgrading.

[0067] In summary, the present invention has significant beneficial effects in improving the accuracy of molten pool temperature prediction, realizing real-time soft measurement, optimizing arc furnace operation control, adapting to complex production process changes, improving model adaptability, providing comprehensive temperature field information, providing intelligent control means, and realizing system modular design, and is significantly superior to the existing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a flow chart of the method for predicting the real-time molten pool temperature after the arc furnace is purged according to the present invention.

[0069] Figure 2 It is a schematic diagram of the system structure for predicting the real-time molten pool temperature after the arc furnace is melted and cleared according to the present invention. DETAILED DESCRIPTION

[0070] The specific implementation of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0071] The present invention realizes a large number of numerical simulations through numerical simulation models to obtain a numerical simulation database with sufficient numerical values; and when establishing a physical information neural network model of a time series neural network structure, the energy conservation equation is embedded into the physical information neural network model as a constraint condition, which is a first of its kind in the present invention.

[0072] Example 1: Method for predicting real-time molten pool temperature after arc furnace cleaning

[0073] As attached Figure 1 As shown in the figure, the method flow for predicting the real-time molten pool temperature after the arc furnace is melted is as follows:

[0074] Step A1: Establishing numerical simulation model

[0075] According to the size and production data of the actual electric arc furnace, a numerical simulation model of the arc furnace molten pool is established. The specific steps are as follows:

[0076] 1. Arc simulation: The magnetohydrodynamic model is used to simulate the arc, taking into account the arc's thermal effects and flow characteristics. The magnetohydrodynamic model can simulate the arc's behavior in the molten pool, including the arc's thermal effects and flow characteristics.

[0077] 2. Bottom blowing simulation: The VOF (Volume of Fluid) multiphase flow model is used to simulate the bottom blowing process and handle the multiphase flow of bottom blowing gas in the molten pool.

[0078] 3. Oxygen blowing simulation: The Euler-Euler multiphase flow model is used to simulate the oxygen blowing process and the flow and reaction of oxygen in the molten pool. The Euler-Euler model is suitable for simulating the flow and reaction of oxygen in the molten pool.

[0079] 4. Carbon injection simulation: The DPM (Discrete Phase Model) model is used to simulate the carbon injection process and deal with the movement and reaction of carbon particles in the molten pool.

[0080] 5. Scrap steel feeding simulation: The VOF multiphase flow model is used to simulate the continuous feeding process of scrap steel and process the multiphase flow behavior of scrap steel in the molten pool.

[0081] 6. Foamy slag simulation: The Euler-Euler multiphase flow model is used to simulate the behavior of foamy slag.

[0082] 7. Define the geometric model: Define the geometric model of the electric arc furnace, including the furnace shell structure, furnace lining structure, graphite electrode structure and layout, charging port structure and layout, initial molten pool shape and size, arc column geometry and active area, oxygen gun and bottom gun structure and layout.

[0083] 8. Set material properties: Set the thermal physical parameters of various materials in the electric arc furnace, including the thermal physical parameters of molten steel, furnace lining, graphite electrode, gas, scrap steel, lime and foam slag.

[0084] 9. Define boundary conditions: including arc power and distribution, heat transfer coefficient between arc and molten pool surface, heat conduction of furnace wall, heat dissipation of furnace wall cooling system, heat dissipation of molten pool surface, feeding temperature, feeding rate, feeding position, latent heat of solid-liquid phase change, chemical reaction heat, etc.

[0085] 10. Establish a group of control equations: Establish a group of control equations including conservation of mass, conservation of momentum and conservation of energy.

[0086] 11. Solving the equations: The finite element method is used to solve the above control equations to obtain the temperature field distribution in the molten pool.

[0087] Step A2: Numerical simulation model modification

[0088] Thermocouples are used to obtain 10 to 20 sets of actual arc furnace molten pool temperature data. The numerical simulation model is corrected based on these actual data and the model parameters are adjusted to make the simulation results closer to the actual measured data.

[0089] Step A3: Generate boundary condition combinations

[0090] According to the production process of the electric arc furnace, the range of variation of the operating process is set as the boundary condition. It mainly includes the oxygen blowing amount, carbon injection amount, lime amount, power consumption, scrap steel addition amount and furnace weight. The Monte Carlo method is used to randomly generate more than 10,000 sets of boundary condition combinations, and input them into the modified numerical simulation model to obtain the temperature field of the entire arc furnace molten pool smelting process, and extract the temperature data of the center point of the molten pool temperature field.

[0091] Step A4: Establishing numerical simulation database

[0092] The boundary conditions of the numerical simulation in step A3 and the temperature data of the center point of the corresponding molten pool temperature field are collected to establish a numerical simulation database. The database content includes each group of boundary condition combinations and the temperature data of the center point of the corresponding molten pool temperature field.

[0093] Step A5: Training the physical information neural network model

[0094] The energy conservation equation for the entire process of flat pool smelting in an electric arc furnace is established, and a physical information neural network model with a time series neural network structure is constructed, and the energy conservation equation is embedded into the physical information neural network model as a constraint condition. The specific process is as follows:

[0095] 1. Energy conservation equation: Establish the energy conservation equation for the entire process of electric arc furnace flat molten pool smelting:

[0096]

[0097] in, Indicates the mass of the molten pool (unit: kg), represents the specific heat capacity of the molten pool (unit: J / [kg·K]), Indicates the rate of change of the molten pool temperature over time (unit: K / s); the energy input item includes the arc input energy 、Oxygen combustion releases heat 、Carbon combustion releases heat , the heat added by scrap steel ; Energy output includes heat loss from furnace wall , Radiant heat loss , chemical reaction absorbs heat 、Slag system heat .

[0098] 2. Physical information neural network model: Construct a time series neural network structure (such as LSTM or GRU) and embed the energy conservation equation as a constraint into the neural network model.

[0099] 3. Training the neural network model: Read the data in the numerical simulation database, use the boundary conditions as input variables, and the temperature data of the center point of the molten pool temperature field as output variables to train the model. By adjusting the model parameters, the error between the model prediction value and the actual measurement value is minimized.

[0100] Step A6: Real-time prediction of melt pool temperature

[0101] During the smelting process, the input variables at the current moment (such as the real-time power consumption of the arc furnace, the real-time oxygen blowing amount, the real-time carbon injection amount, the real-time lime amount, the real-time scrap steel addition amount, and the real-time furnace body weight) are collected in real time, and these variables are input into the trained physical information neural network model to obtain the prediction result of the molten pool temperature at the current moment.

[0102] Step A7: Control the arc furnace temperature

[0103] Temperature control is performed based on the predicted molten pool temperature and target tapping temperature:

[0104] 1. Temperature is higher than the target temperature: If the target tapping temperature is higher than the current molten pool temperature, the power supply of the arc furnace is increased to quickly reach the tapping temperature.

[0105] 2. Temperature is lower than the target temperature: If the target tapping temperature is lower than the current molten pool temperature, the power supply to the arc furnace is reduced to lower the molten pool temperature so as to quickly reach the tapping temperature.

[0106] 3. When the temperature reaches the target temperature: perform subsequent steel tapping operations.

[0107] Step A8: Model Update

[0108] After the physical information neural network model is established, the numerical simulation model is revised according to the actual temperature measurement data every 50 times, and steps A2 to A5 are repeated to update the real-time soft measurement model in the flat pool smelting of the electric arc furnace. The accuracy and real-time performance of the prediction model are ensured by continuously updating the model.

[0109] Further, in a specific embodiment, the step A2 is specifically as follows:

[0110] 1. Compare the actual measured temperature data with the output of the numerical simulation model.

[0111] 2. Adjust the parameters of the numerical simulation model to make the simulation results closer to the actual measured data and ensure the accuracy of the numerical model.

[0112] Further, in a specific embodiment, the step A3 is specifically as follows:

[0113] 1. Set the range of variation of boundary conditions: Determine the range of variation of each operating parameter based on the actual production process.

[0114] 2. Monte Carlo method to generate combinations: The Monte Carlo method is used to randomly generate a large number (more than 10,000 groups) of boundary condition combinations.

[0115] 3. Numerical simulation: The generated boundary condition combination is input into the modified numerical simulation model to obtain the temperature field of the entire arc furnace molten pool smelting process, and the temperature data of the center point of the molten pool temperature field is extracted.

[0116] Example 2: System for predicting real-time molten pool temperature after arc furnace cleaning

[0117] As attached Figure 2 As shown, the present invention provides a system for predicting the real-time molten pool temperature after the arc furnace is purged, comprising the following modules:

[0118] 1. Numerical simulation module: used to establish a numerical simulation model of the arc furnace molten pool according to the size and production data of the actual arc furnace. It includes arc simulation unit, bottom blowing simulation unit, oxygen blowing simulation unit, carbon injection simulation unit, scrap steel feeding simulation unit, foam slag simulation unit, geometric model definition unit, material property setting unit, boundary condition definition unit and control equation solving unit.

[0119] 2. Temperature acquisition module: including thermocouples, used to obtain the actual molten pool temperature data of the arc furnace.

[0120] 3. Boundary condition generation module: used to set the range of variation of the operating process according to the production process of the electric arc furnace, and randomly generate boundary condition combinations using the Monte Carlo method.

[0121] 4. Database module: used to store the boundary conditions of numerical simulation and the temperature data of the center point of the molten pool temperature field.

[0122] 5. Physical information neural network module: including time series neural network structure and energy conservation equation constraints, used to establish and train the prediction model of the whole process of electric arc furnace flat pool smelting.

[0123] 6. Real-time data acquisition module: used to collect input variables at the current moment, including power monitoring system, gas flow meter and oxygen supply system control unit, carbon powder injection system control unit and weight sensor, lime feeding system control unit and silo weighing system, scrap steel feeding system control unit and hopper weighing system, furnace weighing system.

[0124] 7. Temperature prediction module: used to predict the current molten pool temperature based on the real-time collected input variables and the trained physical information neural network model.

[0125] 8. Control module: used to adjust the power supply of the electric arc furnace or add scrap steel according to the predicted molten pool temperature and target steel tapping temperature.

[0126] 9. Model update module: used to regularly update the numerical simulation model and physical information neural network model based on the accumulated actual temperature measurement data.

[0127] 10. User interface module: used to display the predicted molten pool temperature, current operating parameters and control suggestions, and allow operators to perform human-computer interaction and manual adjustments.

[0128] Through the above-mentioned implementation mode, the present invention can realize real-time and accurate prediction of the arc furnace molten pool temperature, and make dynamic adjustments according to the prediction results, thereby improving the efficiency of arc furnace operation and the quality of molten steel, reducing production costs, and having significant practical value and promotion prospects.

Claims

1. A dynamic soft measurement method for arc furnace molten pool temperature, characterized in that: The following steps are involved: A1: A numerical simulation model of the arc furnace molten pool is established based on the actual arc furnace size and production data; the magnetohydrodynamic model is used to simulate the arc, the VOF multiphase flow model is used to simulate the bottom blowing, the Euler-Euler multiphase flow model is used to simulate the oxygen blowing, the DPM model is used to simulate the carbon injection, the VOF multiphase flow model is used to simulate the continuous feeding of scrap steel, and the Euler-Euler multiphase flow model is used to simulate the foamed slag; Define the arc furnace geometry model, set material properties, define boundary conditions, establish the control equations, and solve them using the finite element method; A2: Obtain the actual molten pool temperature of 10 to 20 groups of arc furnaces through thermocouples, and modify the numerical simulation model based on the actual molten pool temperature data; A3: According to the production process of the electric arc furnace, the range of variation of the operating process is set as the boundary conditions, including the oxygen blowing amount, carbon injection amount, lime amount, power consumption, scrap steel addition amount and furnace weight; the Monte Carlo method is used to randomly generate n or more sets of boundary condition combinations, which are input into the numerical simulation model corrected in step A2 to obtain the temperature field of the entire process of flat molten pool smelting in the electric arc furnace, and the temperature data of the center point of the molten pool temperature field are extracted; A4: Collect the boundary conditions of the numerical simulation in step A3 and the temperature data of the center point of the molten pool temperature field, and establish a numerical simulation database; A5: Establish the energy conservation equation for the whole process of flat molten pool smelting in an electric arc furnace, and construct a physical information neural network model with a time series neural network structure, and embed the energy conservation equation into the physical information neural network model as a constraint condition; read the numerical simulation database in step A4, use the boundary conditions as input variables, and the temperature data of the center point of the molten pool temperature field as the output variable, and pass them into the physical information neural network model for training; minimize the error between the model prediction value and the actual measurement value by adjusting the parameters of the physical information neural network model; The trained physical information neural network model is used for real-time soft measurement in flat pool smelting of electric arc furnace; A6: Collect the input variables at the current moment and input them into the physical information neural network model trained in step A5 to obtain the prediction result of the molten pool temperature at the current moment.

2. The dynamic soft measurement method of arc furnace molten pool temperature according to claim 1, characterized in that: When the magnetohydrodynamics model, VOF multiphase flow model, Euler-Euler multiphase flow model, and DPM model have the same parameter requirements, the simulation results of each model and the set value with the smallest actual error average error are used as parameters.

3. The dynamic soft measurement method of arc furnace molten pool temperature according to claim 2, characterized in that: The arc furnace geometric model described in step A1 includes the furnace shell structure, furnace lining structure, graphite electrode structure and its arrangement, charging port structure and its arrangement, initial molten pool shape and size, arc column geometry and action area, oxygen lance and bottom lance structure and their arrangement; the material properties include the thermophysical parameters of molten steel, furnace lining, graphite electrode, gas, scrap steel, lime and foam slag; the boundary conditions include arc power and distribution, heat transfer coefficient between arc and molten pool surface, furnace wall heat conduction, furnace wall cooling system heat dissipation, molten pool surface heat dissipation, charging temperature, charging rate, charging position, solid-liquid phase change latent heat, and chemical reaction heat.

4. The dynamic soft measurement method of arc furnace molten pool temperature according to claim 3, characterized in that: The time series neural network structure described in step A5 adopts the long short-term memory network LSTM or the gated recurrent unit model GRU; the training of the physical information neural network model adopts the back propagation algorithm; the energy conservation equation in the arc furnace is described as follows: ,in: Indicates the mass of the molten pool, unit: kg; Represents the specific heat capacity of the molten pool, unit: J / [kg·K]; Indicates the rate of change of the molten pool temperature over time, unit: K / s; energy input item Including arc input energy 、Oxygen combustion releases heat 、Carbon combustion releases heat , the heat added by scrap steel , energy output term Including heat loss from furnace wall , Radiant heat loss , chemical reaction absorbs heat 、Slag system heat ; The above energy input items and energy output items are obtained through the results of the numerical simulation model; After the physical information neural network model in step A5 is established, the numerical simulation model in step A1 is corrected according to the actual temperature measurement data every time t actual temperature measurements are accumulated, and steps A2 to A5 are repeated to update the real-time soft measurement model in flat pool smelting in electric arc furnace.

5. The dynamic soft measurement method of arc furnace molten pool temperature according to claim 4, characterized in that: The input variables of step A6 at the current moment include the real-time power consumption, real-time oxygen blowing amount, real-time carbon injection amount, real-time lime amount, real-time scrap steel addition amount, and real-time furnace body weight of the electric arc furnace; the real-time power consumption is acquired through the electric energy meter or power analyzer in the power monitoring system, the real-time oxygen blowing amount is acquired through the gas flow meter and the control unit of the oxygen supply system, the real-time carbon injection amount is acquired through the control unit and weight sensor of the carbon powder injection system, the real-time lime amount is acquired through the control unit and silo weighing system of the lime feeding system, the real-time scrap steel addition amount is acquired through the control unit and hopper weighing system of the scrap steel feeding system, and the real-time furnace body weight is acquired through the load sensor in the furnace body weighing system.

6. The dynamic soft measurement method for arc furnace molten pool temperature according to any one of claims 1 to 5, characterized in that: When smelting is close to the end point, the arc furnace parameters are adjusted according to the comparison between the current molten pool temperature and the target steel tapping temperature; If the target steel-tapping temperature is higher than the current molten pool temperature, the power supply of the arc furnace is increased to quickly reach the steel-tapping temperature; if the target steel-tapping temperature is lower than the current molten pool temperature, the power supply of the arc furnace is reduced to lower the molten pool temperature to quickly reach the steel-tapping temperature; when the predicted molten pool temperature reaches the target temperature, prepare for the steel-tapping operation.

7. A dynamic soft measurement system for arc furnace molten pool temperature, characterized in that: include: Numerical simulation module, used to establish a numerical simulation model of the arc furnace molten pool according to the actual arc furnace size and production data; A temperature acquisition module, including a thermocouple, is used to obtain actual molten pool temperature data of the electric arc furnace; The boundary condition generation module is used to set the variation range of the operation process according to the production process of the electric arc furnace, and randomly generate boundary condition combinations using the Monte Carlo method; Database module, used to store the boundary conditions of numerical simulation and the temperature data of the center point of the molten pool temperature field; Physical information neural network module, including time series neural network structure and energy conservation equation constraints, is used to establish and train the prediction model of the whole process of electric arc furnace flat pool smelting; Real-time data acquisition module, used to collect input variables at the current moment; The temperature prediction module is used to predict the current molten pool temperature based on the input variables collected in real time and the trained physical information neural network model; A control module, used to adjust the power supply to the electric arc furnace or add scrap steel according to the predicted molten pool temperature and target tapping temperature; The model updating module is used to regularly update the numerical simulation model and the physical information neural network model according to the accumulated actual temperature measurement data.

8. The dynamic soft measurement system for arc furnace molten pool temperature according to claim 7, characterized in that: The numerical simulation module comprises: Arc simulation unit, using magnetohydrodynamic model; Bottom blowing simulation unit, using VOF multiphase flow model; Oxygen blowing simulation unit, using Euler-Euler multiphase flow model; Carbon injection simulation unit, using DPM model; The scrap steel feeding simulation unit uses the VOF multiphase flow model; The foam slag simulation unit uses the Euler-Euler multiphase flow model; A geometric model definition unit is used to define the arc furnace geometric model; Material property setting unit, used to set material properties; Boundary condition definition unit, used to define boundary conditions; The control equation solving unit uses the finite element method to solve the control equations.

9. The dynamic soft measurement system for arc furnace molten pool temperature according to claim 7, characterized in that: The physical information neural network module adopts a long short-term memory network LSTM or a gated recurrent unit GRU model as a time series neural network structure and uses a back propagation algorithm for training; The real-time data acquisition module comprises: Power monitoring systems, including energy meters or power analyzers, to collect real-time power consumption; Gas flow meter and oxygen supply system control unit, used to collect real-time oxygen blowing volume; Carbon powder injection system control unit and weight sensor, used to collect real-time carbon injection amount; Lime feeding system control unit and silo weighing system for collecting real-time lime quantity; The scrap steel feeding system control unit and hopper weighing system are used to collect the real-time scrap steel feeding amount; The furnace weighing system includes a load sensor for collecting real-time furnace weight.

10. The dynamic soft measurement system for arc furnace molten pool temperature according to any one of claims 7 to 9, characterized in that: A user interface module is also included to display the predicted melt pool temperature, current operating parameters and control suggestions, and to allow operators to perform human-machine interaction and manual adjustments.