A continuous casting method based on digital twin

Abstract

The present invention provides a continuous casting method based on digital twins, comprising: establishing a material performance database of various metal materials under different alloy element compositions; establishing a computer simulation model based on different ingot information, casting information, cold zone information of continuous casting equipment, and environmental information, performing calculations and collating post-processing results; collecting production test data from continuous casting experiments; using a machine learning method to use the result data obtained above as training and validation data sets to train a continuous casting digital twin machine learning model; utilizing the machine-learned digital twin model to automatically calculate and adjust continuous casting process parameters in real time, and during the processing, collecting ingot information to automatically calculate the internal structure, stress, and temperature information of the ingot. The present invention can quickly calculate the required process, and by adjusting a single process, realize the linkage adjustment of the remaining processes, and can quickly calculate the internal structure of the ingot, realizing digital twins.

Description

Technical Field

[0001] The present invention relates to the technical field of continuous casting, and in particular to a continuous casting method based on digital twinning. Background Art

[0002] Continuous casting technology plays a crucial role in the production of preforms for steel and nonferrous metals. During casting, liquid metal, such as steel, iron, copper alloys, aluminum alloys, and magnesium alloys, is continuously poured into a crystallizer filled with cooling water through a continuous casting machine, forming continuous ingots. These ingots undergo subsequent processing to form various metal profiles, which are widely used in construction, automotive, shipbuilding, and home appliances. Continuous casting produces relatively long continuous ingots, some as long as 8,000 mm. To ensure the quality of subsequent molded parts, the ingots must be free of defects such as shrinkage cavities, porosity, and fractures. However, during the ingot molding process, it is impossible to visually observe the solid-liquid phase distribution, stress distribution, shell thickness, and other information within the ingot. This makes continuous casting production control and process design difficult, leading to high continuous casting design costs. Modern computer simulation methods, such as finite element simulation, phase field simulation, and machine learning, can use computer equipment to evaluate the quality of continuous casting ingots (casting stress, solid-liquid phase distribution, and shell thickness) through technical means to select better continuous casting process solutions.

[0003] However, during production, factors such as the cooling water temperature, flow rate, and stability, as well as ambient temperature changes during the casting process, often fail to fully align with the boundary conditions set during computer simulation. This can lead to errors in the actual production process and conditions such as the internal structure of the ingot and stress distribution that cannot be directly observed during actual production. Therefore, before finalizing the casting process, a trial production step is added to determine the final production process. However, when faced with urgent needs to increase production efficiency, such as changes in ingot size or shortened production schedules, this method requires manual redesign and testing of the production process, making it less versatile. Summary of the Invention

[0004] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a continuous casting method based on digital twins. On the basis of traditional continuous casting, in order to solve the problems that the solid-liquid area inside the ingot is difficult to observe and the casting process is difficult to adjust during the continuous casting process, a digital twin continuous casting model is established according to a method combining computer simulation and experiments. Then, a temperature measuring device is added to the continuous casting machine and combined with the existing measuring device of the equipment itself to realize the real data input of the digital twin model. The model is then used to automatically calculate the solid-liquid phase distribution, shell thickness, stress distribution and defect conditions inside the ingot. At the same time, the established continuous casting machine learning model is used to automatically calculate the optimal cooling water flow rate and casting speed and other parameters that match it when increasing the ingot pulling speed and improving production efficiency. The internal structure of the ingot is visualized, the casting process is quickly adjusted, and the production quality and production efficiency of continuous casting are improved.

[0005] Specifically, the present invention provides a continuous casting method based on digital twinning, comprising the following steps:

[0006] S1: Establish a database of material properties of various metal materials with different alloy element compositions;

[0007] S2: Establish a computer simulation model based on different ingot information, casting information, cold zone information of continuous casting equipment, and environmental information, perform calculations, and organize post-processing results;

[0008] S3: Install a temperature detection device in the first cooling zone of the continuous casting mold, a temperature detection device in the second cooling zone, and an alloy composition analysis detection device. Combined with the existing measurement devices on the equipment, the production and test data can be collected.

[0009] S4: Using machine learning methods, the result data obtained from S1, S2, and S3 are used as training and verification data sets to train the continuous casting digital twin machine learning model. Mathematical statistics methods are used to select computer simulation schemes under different processes, conduct experimental verification, collect post-experimental data, calibrate and train the model, and obtain the digital twin model after machine learning;

[0010] S5: Utilize the digital twin model after machine learning to realize the automatic calculation and real-time adjustment of continuous casting process parameters. During the processing, the ingot information is collected and automatically calculated to obtain the internal structure, stress and temperature information of the ingot.

[0011] As a further illustration of the present invention, in step S1, based on different alloy compositions, a material properties database of various metal materials with different alloy element compositions is established by using the first principles of materials or metal phase diagram calculation and material properties simulation methods.

[0012] As a further explanation of the present invention, in step S2, the ingot information includes the ingot shape and the ingot size; the casting information includes the casting temperature and the casting speed; the cold zone information includes the first cold zone information and the second cold zone information of the crystallizer, wherein the first cold zone information and the second cold zone information both include the cooling water flow rate and the cooling water temperature; the environmental information includes the ambient temperature and the mold temperature.

[0013] As a further illustration of the present invention, in step S2, the post-processing results specifically include: statistically analyzing the distribution of solid and liquid phases, stress distribution, temperature distribution, defect distribution, and microstructure in the ingot under different boundary conditions.

[0014] As a further illustration of the present invention, in step S3, the process of collecting production and test data specifically includes:

[0015] S301: Collect ingot information, casting information, cooling information, and environmental information, wherein the ingot information includes ingot type, first cooling zone temperature, second cooling zone temperature, and ingot bottom temperature; the casting information includes casting temperature and casting speed; the cooling information includes cooling water flow rate and cooling water temperature; and the environmental information includes ambient temperature and mold temperature;

[0016] S302: Test the material to obtain the material constitutive parameters;

[0017] S303: The test results obtained in step S301 and step S302 are collected, and the shell thickness of the ingot is obtained by testing the continuous casting equipment. The residual stress, defects, material properties and microstructure of the ingot are tested.

[0018] As a further illustration of the present invention, in step S4, a machine learning model is used to train alloy composition-microstructure-material properties-energy consumption-cost-quality, and a mapping relationship between process-alloy composition-microstructure-material properties-energy consumption-cost-quality is constructed to obtain the optimal matching model.

[0019] As a further illustration of the present invention, in step S5, the digital twin model after machine learning is used to realize the automatic calculation and real-time adjustment of continuous casting process parameters, including:

[0020] Based on the set ingot size, shell thickness, defect requirement parameters and the measured ambient temperature and cooling water temperature, the trained model in step S4 automatically calculates the required ingot pulling speed, casting temperature, casting speed, and cooling water flow rate, and sends the calculated speed and temperature data to the setting end of the continuous casting equipment.

[0021] As a further illustration of the present invention, in step S5, the digital twin model after machine learning is used to realize the automatic calculation and real-time adjustment of continuous casting process parameters, which also includes:

[0022] When the ingot pulling speed changes, the model trained in step S4 will use the changed ingot pulling speed as data input, maintain the set ingot size, shell thickness, defect requirement parameters, and the ambient temperature and cooling water temperature measured by the continuous casting equipment, automatically calculate the required casting temperature, casting speed, and cooling water flow rate, and send the calculated speed and temperature data to the setting end of the continuous casting equipment.

[0023] As a further illustration of the present invention, in step S5, the digital twin model after machine learning is used to collect ingot information during the processing process to automatically calculate the internal structure, stress and temperature information of the ingot, including:

[0024] During continuous casting production, the continuous casting equipment will collect the temperature of the first cooling zone of the ingot, the temperature of the second cooling zone, the temperature of each area at the bottom of the ingot, the casting speed, the casting temperature, the cooling water flow rate, the cooling water temperature, the ambient temperature, and the ingot pulling speed, and use the collected information as the data input of the model trained in step S4, automatically calculate the internal result information and stress distribution of the ingot, and display them in the display window of the continuous casting equipment.

[0025] As a further illustration of the present invention, in step S5, the digital twin model after machine learning is used to collect ingot information during the processing process to automatically calculate the internal structure, stress and temperature information of the ingot, which also includes:

[0026] Based on the metal composition information obtained by detection, the model trained in step S4 will calculate the actual thermal conductivity, specific heat capacity, latent heat, thermal expansion coefficient, plasticity, and elasticity of the material in the current state, and use the above material constitutive properties as corrected parameters and re-import them into the model described in step S4, and correct the calculation results of the model in real time to make the calculated internal information and stress and strain conditions of the ingot more accurate.

[0027] Compared with the prior art, the present invention has the following beneficial technical effects:

[0028] 1. The existing finite element simulation trial production method requires a high time cost in terms of finite element model establishment and calculation. The present invention can quickly calculate the required process, reduce process design costs, and improve efficiency.

[0029] 2. The existing finite element simulation trial production method has low model versatility. The present invention adopts a machine learning model that includes training for multiple ingot sizes. The model has high versatility, reduces pre-processing time, and improves production efficiency.

[0030] 3. The existing finite element simulation trial production method cannot intuitively observe the internal structure of the ingot during the production process. The present invention uses a model trained by machine learning to quickly calculate the internal structure of the ingot based on the real data input by the equipment, realizing digital twin and improving the stability of the continuous casting process.

[0031] 4. The existing finite element simulation trial production method cannot achieve rapid adjustment of the casting process. The present invention can achieve coordinated adjustment of other processes by adjusting a single process while maintaining the original production quality, thereby improving production efficiency.

[0032] Other features and advantages of this technical solution will be described in the subsequent description, and in part will become apparent from the description, or understood by practicing this technical solution. The objectives and other advantages of this technical solution can be achieved and obtained through the structures specifically pointed out in the written description and the accompanying drawings.

[0033] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings are used to provide a further understanding of the present technical solution and constitute a part of the specification. Together with the embodiments of the present technical solution, they are used to explain the present technical solution and do not constitute a limitation of the present technical solution. In the accompanying drawings:

[0035] Figure 1 This is a schematic diagram of the installation of a temperature detection device and an alloy composition analysis detection device on a continuous casting device in an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of the machine learning process in an embodiment of the present invention;

[0037] Figure 3 The temperature fields at different times in the embodiment of the present invention;

[0038] Figure 4 The solid-liquid phase changes at different times in the embodiments of the present invention;

[0039] Figure 5 This is a cloud diagram of the equivalent stress distribution of the ingot in the embodiment of the present invention. DETAILED DESCRIPTION

[0040] The preferred embodiments of the present technical solution are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present technical solution and are not used to limit the present technical solution.

[0041] An embodiment of the present invention provides a continuous casting method based on digital twinning, comprising the following steps:

[0042] S1: Establish a material properties database for various metal materials with different alloy element compositions.

[0043] Specifically, based on different alloy compositions, we can use the first principles of materials or metal phase diagram calculation and material performance simulation methods to establish a material performance database for various metal materials with different alloy element compositions. The material properties mainly include elastic modulus, Poisson's ratio, plasticity, thermal expansion coefficient, specific heat capacity, latent heat of fusion, liquidus temperature, solidus temperature, etc.

[0044] S2: Establish a computer simulation model based on different ingot information, casting information, cold zone information of continuous casting equipment, and environmental information, perform calculations, and organize post-processing results.

[0045] The above-mentioned computer simulation methods include finite element simulation methods, phase field simulation methods, machine learning methods and other methods that use computers to perform simulation calculations.

[0046] Specifically, the ingot information includes the ingot shape and ingot size; the casting information includes the casting temperature and casting speed; the cold zone information includes the first cold zone information and the second cold zone information of the crystallizer, wherein the first cold zone information and the second cold zone information both include the cooling water flow rate and the cooling water temperature; the environmental information includes the ambient temperature and the mold temperature.

[0047] The post-processing results specifically include: statistics on the distribution of solid and liquid phases in the ingot under different boundary conditions, stress distribution, temperature distribution, defect distribution, microstructure, etc.

[0048] S3: Install a temperature detection device in the first cooling zone of the continuous casting mold, a temperature detection device in the second cooling zone, and an alloy composition analysis detection device. Combined with the existing measurement devices on the equipment, the production and test data can be collected.

[0049] The temperature detection device and alloy composition analysis detection device are commercially available products, and the installation position is as follows: Figure 1 shown.

[0050] The above process of collecting production and test data specifically includes:

[0051] S301: Collect ingot information, casting information, cooling information, and environmental information, wherein the ingot information includes ingot type, first cooling zone temperature, second cooling zone temperature, and ingot bottom temperature; the casting information includes casting temperature and casting speed; the cooling information includes cooling water flow rate and cooling water temperature; and the environmental information includes ambient temperature and mold temperature;

[0052] S302: Testing the material to obtain material constitutive parameters. Material constitutive parameters refer to parameters of the material constitutive model. Taking the hyperbolic sine model (also known as the Arrhennius constitutive model, as shown in the following formula) as an example, the constitutive parameters are the parameters in the formula. They can be obtained by statistically analyzing the results of mechanical tests and then processing the result data.

[0053]

[0054] S303: The test results obtained in step S301 and step S302 are collected, and the shell thickness of the ingot is obtained by testing the continuous casting equipment. The residual stress, defects, material properties and microstructure of the ingot are tested.

[0055] S4: As Figure 2 As shown in the figure, the machine learning method is used to take the result data obtained from S1, S2 and S3 as the training and verification data sets to train the continuous casting digital twin machine learning model, and the mathematical statistics method is used to select computer simulation schemes under different processes, conduct experimental verification, collect data after the experiment, calibrate and train the model, and obtain the digital twin model after machine learning.

[0056] Specifically, during the experimental verification process, the orthogonal test method can be used to select different processes to achieve a wider range of results with fewer simulation times.

[0057] It should also be noted that in step S4, a machine learning model is used to train the alloy composition-microstructure-material properties-energy consumption-cost-quality, and a mapping relationship between process-alloy composition-microstructure-material properties-energy consumption-cost-quality is constructed to obtain the optimal matching model.

[0058] S5: Utilize the digital twin model after machine learning to realize the automatic calculation and real-time adjustment of continuous casting process parameters. During the processing, the ingot information is collected and automatically calculated to obtain the internal structure, stress and temperature information of the ingot.

[0059] Specifically, step S5 includes the following parts:

[0060] S501: Based on the set ingot size, shell thickness, defect requirement parameters and the measured ambient temperature and cooling water temperature, the model trained in step S4 automatically calculates the required ingot pulling speed, casting temperature, casting speed, and cooling water flow rate, and sends the calculated speed and temperature data to the setting end of the continuous casting equipment.

[0061] S502: During continuous casting production, the continuous casting equipment will collect the temperature of the first cooling zone of the ingot, the temperature of the second cooling zone, the temperature of each area at the bottom of the ingot, the casting speed, the casting temperature, the cooling water flow rate, the cooling water temperature, the ambient temperature, and the ingot pulling speed, and use the collected information as the data input of the model trained in step S4, automatically calculate the internal result information and stress distribution of the ingot, and display them in the display window of the continuous casting equipment. Figure 3 is the temperature condition of aluminum alloy ingot as an example, Figure 4 is the cloud diagram of the solid-liquid phase distribution inside the ingot, Figure 5 is the stress distribution.

[0062] S503: Based on the metal composition information obtained by detection, the model trained in step S4 will calculate the actual thermal conductivity, specific heat capacity, latent heat, thermal expansion coefficient, plasticity, and elasticity of the material in the current state, and use the above material constitutive properties as corrected parameters and re-import them into the model of step S4 to correct the calculation results of the model in real time, so that the calculated internal information of the ingot and the stress and strain conditions are more accurate.

[0063] S504: When the ingot pulling speed changes, the model trained in step S4 will use the changed ingot pulling speed as data input, maintain the set ingot size, shell thickness, defect requirement parameters, and the ambient temperature and cooling water temperature measured by the continuous casting equipment, automatically calculate the required casting temperature, casting speed, and cooling water flow rate, and send the calculated speed and temperature data to the setting end of the continuous casting equipment.

[0064] The above step S5 can automatically calculate the required process parameters based on the set product requirements using the model obtained by machine learning, and collect ingot information during the processing to automatically calculate the internal structure, stress and temperature information of the ingot. At the same time, if a single process parameter is adjusted, the remaining processes will be adjusted in conjunction, thereby improving design and production efficiency.

[0065] In summary, the present invention adopts a method combining computer simulation with experimental data to train the continuous casting digital twin model. A temperature detection device is added to the first cooling zone of the crystallizer of the continuous casting equipment, and a temperature detection device and an alloy composition analysis detection device are added to the second cooling zone. In combination with the existing measuring devices of the equipment itself, the real data input of the digital twin model is realized. The various detection devices on the continuous casting equipment are used as the data input of the digital twin model to automatically calculate the internal microstructure, material properties, stress distribution, defect distribution, etc. of the ingot in real time. When the ingot pulling speed is changed, the model will automatically calculate the remaining production processes that match the changed ingot pulling speed based on the set production quality parameters, and send the process to the execution end of the continuous casting equipment to achieve rapid response during the casting process.

[0066] Obviously, those skilled in the art may make various changes and modifications to this technical solution without departing from the spirit and scope of this technical solution. Thus, if these modifications and variations of this technical solution fall within the scope of the claims of this technical solution and their equivalents, this technical solution is intended to include these modifications and variations.

Claims

1. A continuous casting method based on digital twin, characterized in that: The steps include: S1: Establish a database of material properties of various metal materials with different alloy element compositions; S2: establishing a computer simulation model based on different ingot information, casting information, cold zone information of the continuous casting equipment, and environmental information, performing calculations and collating post-processing results; the ingot information includes ingot shape and ingot size; the casting information includes casting temperature and casting speed; the cold zone information includes the first cold zone information and the second cold zone information of the crystallizer, wherein the first cold zone information and the second cold zone information both include cooling water flow rate and cooling water temperature; the environmental information includes ambient temperature and mold temperature; the collated post-processing results specifically include: statistical analysis of the distribution of solid and liquid phases, stress distribution, temperature distribution, defect distribution, and microstructure in the ingot under different boundary conditions; S3: Install a temperature detection device in the first cooling zone of the continuous casting mold, a temperature detection device in the second cooling zone, and an alloy composition analysis detection device. Combined with the existing measurement devices on the equipment, the production and test data can be collected. S4: Using machine learning methods, the result data obtained from S1, S2, and S3 are used as training and validation data sets to train the continuous casting digital twin machine learning model: the machine learning model is used to train the alloy composition-microstructure-material properties-energy consumption-cost-quality, and the mapping relationship between process-alloy composition-microstructure-material properties-energy consumption-cost-quality is constructed to obtain the optimal matching model; Use mathematical statistics to select computer simulation schemes under different processes, conduct experimental verification, collect post-experimental data, calibrate and train the model, and obtain a digital twin model after machine learning; S5: Utilize the machine learning digital twin model to automatically calculate and adjust the continuous casting process parameters in real time: Based on the set ingot size, shell thickness, defect requirement parameters, and the measured ambient temperature and cooling water temperature, the model trained in step S4 automatically calculates the required ingot pulling speed, casting temperature, casting speed, and cooling water flow rate, and transmits the calculated speed and temperature data to the setting end of the continuous casting equipment; By using the digital twin model after machine learning, ingot information is collected during the processing to automatically calculate the internal structure, stress and temperature information of the ingot: during continuous casting production, the continuous casting equipment will collect the temperature of the first cooling zone of the ingot, the temperature of the second cooling zone, the temperature of each area at the bottom of the ingot, the casting speed, the casting temperature, the cooling water flow rate, the cooling water temperature, the ambient temperature, and the ingot pulling speed, and use the collected information as the data input of the model trained in step S4, automatically calculate the internal result information and stress distribution of the ingot, and display them in the display window of the continuous casting equipment.

2. The continuous casting method based on digital twinning according to claim 1, characterized in that: In step S1, based on different alloy compositions, a material properties database of various metal materials with different alloy element compositions is established by using the first principles of materials or metal phase diagram calculation and material properties simulation methods.

3. The continuous casting method based on digital twinning according to claim 1, characterized in that: In step S3, the process of collecting production and test data specifically includes: S301: Collect ingot information, casting information, cooling information, and environmental information, wherein the ingot information includes ingot type, first cooling zone temperature, second cooling zone temperature, and ingot bottom temperature; the casting information includes casting temperature and casting speed; the cooling information includes cooling water flow rate and cooling water temperature; and the environmental information includes ambient temperature and mold temperature; S302: Test the material to obtain the material constitutive parameters; S303: The test results obtained in step S301 and step S302 are collected, and the shell thickness of the ingot is obtained by testing the continuous casting equipment. The residual stress, defects, material properties and microstructure of the ingot are tested.

4. The continuous casting method based on digital twinning according to claim 1, characterized in that: In step S5, the digital twin model after machine learning is used to automatically calculate and adjust the continuous casting process parameters in real time, which also includes: When the ingot pulling speed changes, the model trained in step S4 will use the changed ingot pulling speed as data input, maintain the set ingot size, shell thickness, defect requirement parameters, and the ambient temperature and cooling water temperature measured by the continuous casting equipment, automatically calculate the required casting temperature, casting speed, and cooling water flow rate, and send the calculated speed and temperature data to the setting end of the continuous casting equipment.

5. The continuous casting method based on digital twinning according to claim 1, characterized in that: In step S5, the digital twin model after machine learning is used to collect ingot information during the processing process to automatically calculate the internal structure, stress and temperature information of the ingot, which also includes: Based on the metal composition information obtained by detection, the model trained in step S4 will calculate the actual thermal conductivity, specific heat capacity, latent heat, thermal expansion coefficient, plasticity, and elasticity of the material in the current state, and use the above material constitutive properties as corrected parameters and re-import them into the model described in step S4, and correct the calculation results of the model in real time to make the calculated internal information and stress and strain conditions of the ingot more accurate.

Images