Steelmaking process optimization method and system based on digital twin model

Abstract

The invention relates to the field of process optimization, in particular to a steelmaking process optimization method and system based on a digital twin model. A steelmaking process optimization system based on a digital twin model comprises a converter digital twin model construction module, a physical parameter adjustment module, a converter wear model construction module and a steelmaking process optimization module. By establishing the high-precision converter digital twinning model, heat transfer, flow field change and chemical reaction in the steelmaking process can be accurately described, a physical rule basis is provided for analogue simulation, and then component information of input materials and converter process parameters are optimized and adjusted by utilizing analogue simulation, so that the converter process parameters are optimized and adjusted, and the converter process parameters are optimized and adjusted. During the period, the flame image is used as a process quality evaluation index, so that the accuracy and efficiency of the steelmaking process are remarkably improved, and the dependence on experience of operators is reduced.

Description

Technical Field

[0001] The present invention relates to the field of process optimization, and in particular to a steelmaking process optimization method and system based on a digital twin model. Background Art

[0002] In traditional steelmaking processes, the optimization of production efficiency and product quality mainly depends on the experience and intuition of operators. With the development of steelmaking technology and the expansion of production scale, higher requirements are placed on the precise control of steelmaking process parameters. Traditional methods have many limitations in dealing with complex production processes, such as long response time for parameter adjustment, low accuracy, and difficulty in dealing with emergencies in the production process. These problems not only affect production efficiency, but also limit the further improvement of product quality. Summary of the invention

[0003] The present invention performs a series of processing on the converter point cloud data set, and ensures the accuracy of the model by removing outliers in the point cloud data; surface reconstruction restores the actual geometric shape of the converter according to the point cloud data, and converts discrete point cloud data into a continuous surface model, laying an accurate geometric foundation for simulating the internal process of the converter; voxelization processing further discretizes the continuous surface model, providing a computational basis for thermodynamics, fluid mechanics, and chemical reaction analysis. In addition, through thermodynamics, fluid mechanics, and chemical reaction analysis, the established converter digital twin model can accurately describe the heat transfer, flow field changes, and chemical reactions in the steelmaking process, providing a physical rule basis for simulation, and also using The model is preliminarily verified and adjusted using standard environmental data and the working data of the new converter to ensure that the model can accurately reflect the actual process conditions. The digital twin model is updated to reflect the wear status of the converter by collecting the converter working data and the furnace mouth flame image in real time, so as to achieve adaptive adjustment of the model and more realistically reflect the actual status of the converter. Finally, the input material composition information and the converter process parameters are used as variables for optimization and adjustment through simulation, and the flame image is used as an evaluation indicator for the steelmaking process quality to further optimize the input material composition information and the converter process parameters, which greatly reduces the need to rely on the operator's experience to set the process parameters and improves the accuracy and efficiency of the steelmaking process.

[0004] A steelmaking process optimization method based on a digital twin model, comprising: Step S1: acquiring a point cloud data set of a converter by using a laser scanner, performing outlier removal, surface reconstruction and voxelization processing on the point cloud data set of the converter to construct a converter entity model; Step S2: Based on the converter entity model, a thermodynamic analysis, a fluid dynamics analysis, and a chemical reaction analysis are performed on the steelmaking process inside the converter to construct a converter physical model; Step S3: combining the converter entity model and the converter physical model to construct a converter digital twin model, simulating the converter digital twin model according to standard environmental data and standard converter working data when the converter is not worn, and adjusting the physical parameters in the converter physical model in the converter digital twin model based on the standard furnace mouth flame image corresponding to the standard converter working data, and recording the adjusted physical parameters as standard physical parameters; Step S4: Collect the converter working data and the corresponding furnace mouth flame image within any period of time to form a local converter working data time series set, obtain all local converter working data time series sets to form a global converter working data time series set, simulate the converter digital twin model based on the global converter working data time series set to construct a converter wear model, and use the converter wear model to adaptively adjust the converter physical model in the converter digital twin model; Step S5: After setting the environmental data and the length of time the converter is in use, the input material composition information and the converter process parameters are input as input variables into the converter digital twin model for simulation, and the input material composition information and the converter process parameters are optimized based on the target furnace mouth flame image to generate the optimal input material composition information and converter process parameters under different environmental data and converter usage times, so as to optimize the steelmaking process.

[0005] The present invention performs abnormal point removal, surface reconstruction and voxelization on the point cloud data set of the converter, wherein the abnormal point removal can remove abnormal point cloud data in the point cloud data to ensure the accuracy of the digital twin model of the converter, and the surface reconstruction can reconstruct the geometric shape of the converter according to the point cloud data, and convert the discrete point cloud data into a continuous surface model, thereby providing an accurate geometric basis for simulating the internal process of the converter, and the voxelization processing further converts the continuous surface model into a discrete voxel grid, thereby providing a discrete calculation domain for subsequent thermodynamic analysis, fluid mechanics analysis and chemical reaction analysis; by performing thermodynamic analysis, fluid mechanics analysis and chemical reaction analysis on the steelmaking process, the established digital twin model of the converter can more accurately describe the heat conduction, flow field changes and material reactions occurring in the steelmaking process, thereby providing a physical rule basis for subsequent simulation; and The model is preliminarily verified and adjusted through standard environmental data and the working data of unworn converters to ensure that the model can accurately reflect the actual process conditions; and by collecting the real-time working data of the converter and the flame image of the furnace mouth, the converter digital twin model is updated to reflect the wear status of the converter, so that the converter digital twin model can be adaptively adjusted according to the usage time, which can more realistically reflect the actual status of the converter, and further ensure the accuracy when executing steelmaking process optimization; finally, the input material composition information and converter process parameters are used as variables, and the flame image is used as an indicator for steelmaking process quality evaluation, and the input material composition information and converter process parameters are optimized and adjusted, and the optimal input material composition information and converter process parameters under different environmental data and converter usage time are output, without the need for operators to set process parameters based on experience, thereby improving steelmaking efficiency.

[0006] Preferably, the specific content of step S1 is: Step S11: performing maximum bounding box sampling on the point cloud data set of the converter to obtain a downsampled point cloud data set of the converter; Step S12: extracting neighborhood relations for each point cloud data in the downsampled point cloud data set of the converter, and generating a point cloud neighborhood relation matrix; Step S13: performing a graph structure conversion on the downsampled point cloud data set based on the point cloud neighborhood relationship matrix to generate a converter point cloud feature map; Step S14: extracting local structural features of each point cloud data and global structural features of all point cloud data for the converter point cloud feature map, and generating converter point cloud local structural features and converter point cloud global structural features; Step S15: removing abnormal points from the converter point cloud data set based on the local structural features of the converter point cloud and the global structural features of the converter point cloud to generate a denoised point cloud data set; Step S16: reconstructing the surface based on the point cloud data in the denoised point cloud data set to obtain a converter surface model, and globally optimizing the converter surface model under specified topological structures and geometric constraints; Step S17: performing three-dimensional voxel mesh division according to the converter surface model, and mapping the converter surface model into the three-dimensional voxel mesh and performing voxelization processing to obtain a converter solid model.

[0007] The present invention can reduce the complexity of point cloud data by performing maximum bounding box sampling on the point cloud data set of the converter, while retaining sufficient point cloud data information for constructing a converter entity model; by performing neighborhood relationship extraction on each point cloud data in the downsampled point cloud data set of the converter, the features corresponding to each point cloud data, such as relative distance, etc., can be extracted, and then the downsampled point cloud data set is converted into a graph structure through a cloud neighborhood relationship matrix, so that the point cloud data is converted into a graph structure, and the graph structure can better learn the global structural features of the converter point cloud such as point cloud density and distribution in the point cloud data, and finally, based on the local structural features of the converter point cloud and the global structural features of the converter point cloud, the removal of abnormal points can be achieved under a method such as setting a threshold; in the process of surface reconstruction of the point cloud data, global optimization of the converter surface model under the specified topological structure and geometric constraints can ensure the overall consistency of the model; voxelization of the converter surface model can accurately describe the spatial distribution and internal structure in the converter, and can also support more complex calculations and simulation operations.

[0008] Preferably, step S2 specifically includes the following contents: S21: Perform finite element analysis on the steelmaking process in the converter according to the converter entity model, and analyze the temperature field change in the converter through the heat conduction equation to obtain the converter temperature field model, thereby realizing thermodynamic analysis of the steelmaking process inside the converter; S22: Perform finite element analysis on the steelmaking process in the converter according to the converter entity model, and analyze the flow field change in the converter by using the Navier-Stokes equation to obtain the converter flow field model, thereby realizing the fluid mechanics analysis of the steelmaking process inside the converter; S23: performing finite element analysis on the steelmaking process in the converter according to the converter entity model, and analyzing the chemical reaction changes in the converter through a kinetic equation to obtain a converter chemical reaction kinetic model, thereby realizing chemical reaction analysis of the steelmaking process inside the converter; S33: The converter temperature field model, the converter flow field model and the converter chemical reaction kinetics model are coupled by the finite difference method to obtain the converter physical model.

[0009] The present invention realizes accurate analysis and simulation of key physical phenomena in the converter steelmaking process by simulating the temperature field, flow field and chemical reaction dynamics in the converter respectively, and provides a physical rule basis for subsequent simulation.

[0010] Preferably, the specific content of step S3 is: S31: Combine the converter entity model and the converter physical model to construct a converter digital twin model; S32: collecting standard environmental data, standard converter working data when the converter is not worn, and corresponding standard furnace mouth flame images, wherein the standard converter working data includes input material composition information and converter process parameters, and sending the standard converter working data to the converter digital twin model for simulation, and outputting a first simulated furnace mouth flame image; S33: performing spectral analysis on the standard furnace mouth flame image to generate standard furnace mouth flame spectrum data; performing polarized light analysis on the standard furnace mouth flame spectrum data to generate standard furnace mouth flame spectrum response data; performing image enhancement on the standard furnace mouth flame image based on the standard furnace mouth flame spectrum response data to generate a standard furnace mouth flame enhanced image; similarly, performing spectral analysis on the first simulated furnace mouth flame image to generate first simulated furnace mouth flame spectrum data; performing polarized light analysis on the first simulated furnace mouth flame spectrum data to generate first simulated furnace mouth flame spectrum response data; performing image enhancement on the first simulated furnace mouth flame image based on the first simulated furnace mouth flame spectrum response data to generate a first simulated furnace mouth flame enhanced image; S34: performing feature matching on the standard furnace mouth flame enhancement image and the first simulated furnace mouth flame enhancement image, outputting first furnace mouth flame difference data, and performing geometric consistency check during the feature matching process; S35: Using the flame difference data of the first furnace mouth as a constraint condition, the physical parameters in the converter physical model in the converter digital twin model are adjusted through the particle swarm optimization algorithm, and the adjusted physical parameters are recorded as standard physical parameters.

[0011] The present invention sends standard environmental data and working data of an unworn converter into a digital twin model of the converter for simulation, and performs feature matching on the output first simulated furnace mouth flame image and the standard furnace mouth flame image corresponding to the standard converter working data, determines the degree of difference between the first simulated furnace mouth flame image and the standard furnace mouth flame image corresponding to the standard converter working data, and uses this degree of difference as a constraint to adjust the physical parameters in the converter physical model, so that the digital twin model of the converter can more accurately reflect the actual steelmaking process; and using the furnace mouth flame image as a judgment standard is simpler and more convenient than quality inspection of steel products; image enhancement operations are also performed on the standard furnace mouth flame image and the first simulated furnace mouth flame image, and spectral analysis can extract material change information in the steelmaking process, and polarized light analysis can further enhance the image contrast. Performing image enhancement operations on the standard furnace mouth flame image and the first simulated furnace mouth flame image based on the results of spectral analysis and polarized light analysis can enable higher accuracy in subsequent calculations of differences between images.

[0012] Preferably, the specific steps of generating a standard furnace mouth flame enhancement image are: Step S331: performing spectral analysis on the standard furnace mouth flame image to generate spectral feature data of different bands, and splicing the spectral feature data of different bands according to channels to generate standard furnace mouth flame spectral data; Step S332: sampling the standard furnace mouth flame spectrum data at different polarization angles, the polarization angles being 0°, 45°, 90° and 135°, to obtain 0° standard furnace mouth flame spectrum response data, 45° standard furnace mouth flame spectrum response data, 90° standard furnace mouth flame spectrum response data and 135° standard furnace mouth flame spectrum response data; Step S333: combining the 0° standard furnace mouth flame spectral response data, the 45° standard furnace mouth flame spectral response data, the 90° standard furnace mouth flame spectral response data, and the 135° standard furnace mouth flame spectral response data to obtain the standard furnace mouth flame spectral response data; Step S334: linearly stretching the standard furnace mouth flame spectrum data based on the standard furnace mouth flame spectrum response data to obtain standard furnace mouth flame enhancement spectrum data; Step S335: performing image enhancement on the standard furnace mouth flame image based on the standard furnace mouth flame enhancement spectrum data and the attention mechanism to obtain a standard furnace mouth flame enhanced image; And the steps of generating the first simulated furnace mouth flame enhancement image are consistent with the steps of generating the standard furnace mouth flame enhancement image.

[0013] The present invention performs spectral analysis on the furnace mouth flame image. Since the combustion of substances will have different flames under different reaction conditions, spectral analysis can reflect the information of substance reaction changes. Polarized light analysis is performed on the extracted spectral data at different polarization angles to extract the flame spectral responses at different polarization angles, that is, the light intensity changes at different spectral bands. Then, based on the flame spectral response data, linear stretching is performed on the data range of the spectral data to make the image contrast more balanced and saturated. Finally, the standard furnace mouth flame image is enhanced through the attention mechanism and the standard furnace mouth flame enhancement spectral data, which can make the subsequent calculation of the difference between images have higher accuracy.

[0014] Preferably, the specific content of step S4 is as follows: Step S41: collecting converter working data and corresponding standard furnace mouth flame images within any period of time to form a local converter working time series data set, and obtaining all local converter working data time series sets to form a global converter working data time series set; Step S42: sending the local converter working data time series sets in the global converter working data time series set one by one into the converter digital twin model for simulation, and adjusting the physical parameters in the converter physical model in the converter digital twin model based on the furnace mouth flame image corresponding to the converter working data in the local converter working data time series set, and outputting the time series physical parameters corresponding to each local converter working data time series set; Step S43: performing differential operations on the time-series physical parameters corresponding to each local converter operation data time-series set and the standard physical parameters, respectively, to generate time-series physical parameter differential data corresponding to each local converter operation data time-series set; Step S44: using the time series physical parameter differential data corresponding to the time series set of all local converter working data arranged in chronological order as training data, training the time series prediction network model, the time series prediction network model adopts the LSTM model, and generating the converter wear model; Step S45: Predict the converter wear condition based on the converter wear model, generate the physical parameter adjustment coefficients of the converter under different usage times, and adaptively adjust the converter physical model in the converter digital twin model through the physical parameter adjustment coefficients.

[0015] The present invention sends the working data of the converter with different usage time into the digital twin model of the converter for simulation, calculates the physical parameters of the converter under different usage time, and uses the difference between the physical parameters of the converter under different usage time and the standard physical parameters as training data to construct a converter wear model. The converter wear model can be used to describe the changes in physical parameters of the converter physical model in the digital twin model of the converter under different usage time. The physical parameter adjustment coefficients at different usage time output by the converter wear model can realize adaptive adjustment of the converter physical model in the digital twin model of the converter under different usage time.

[0016] Preferably, the specific content of step S5 is as follows: Step S51: When the environmental data and the converter usage time are set, the input material composition information and the converter process parameters are input as input variables into the converter digital twin model for simulation, and a second simulated furnace mouth flame image is generated; Step S52: performing spectral analysis on the target furnace mouth flame image to generate target furnace mouth flame spectral data; performing polarized light analysis on the target furnace mouth flame spectral data to generate target furnace mouth flame spectral response data; performing image enhancement on the target furnace mouth flame image based on the target furnace mouth flame spectral response data to generate a target furnace mouth flame enhanced image; similarly, performing spectral analysis on the second simulated furnace mouth flame image to generate second simulated furnace mouth flame spectral data; performing polarized light analysis on the second simulated furnace mouth flame spectral data to generate second simulated furnace mouth flame spectral response data; performing image enhancement on the second simulated furnace mouth flame image based on the second simulated furnace mouth flame spectral response data to generate a second simulated furnace mouth flame enhanced image; Step S53: performing feature matching on the target furnace mouth flame enhancement image and the first simulated furnace mouth flame enhancement image, outputting second furnace mouth flame difference data, and performing geometric consistency check during the feature matching process; Step S54: Taking the second furnace mouth flame difference data as a constraint condition, the input material composition information and the converter process parameters are optimized by the particle swarm optimization algorithm to generate the optimal input material composition information and converter process parameters under different environmental data and converter usage time.

[0017] The present invention takes input material composition information and converter process parameters as variables, and simulates the steelmaking process corresponding to different input material composition information and converter process parameters through the converter digital twin model, and uses flame images as indicators for steelmaking process quality evaluation to optimize and adjust the input material composition information and converter process parameters, and outputs the optimal input material composition information and converter process parameters under different environmental data and converter usage time.

[0018] A steelmaking process optimization system based on a digital twin model, comprising: The converter digital twin model construction module is used to obtain the converter point cloud data set through a laser scanner, remove abnormal points, reconstruct surfaces and voxelize the converter point cloud data set to build a converter solid model, perform thermodynamic analysis, fluid dynamics analysis and chemical reaction analysis on the steelmaking process inside the converter based on the converter solid model to build a converter physical model, and combine the converter solid model and the converter physical model to build a converter digital twin model; A physical parameter adjustment module is used to simulate the converter digital twin model according to standard environmental data and standard converter working data when the converter is not worn, and adjust the physical parameters in the converter physical model in the converter digital twin model based on the standard furnace mouth flame image corresponding to the standard converter working data, and record the adjusted physical parameters as standard physical parameters; The converter wear model construction module is used to collect the converter working data and the corresponding furnace mouth flame image within any period of time to form a local converter working data time series set, obtain all local converter working data time series sets to form a global converter working data time series set, and simulate the converter digital twin model based on the global converter working data time series set to construct the converter wear model. The converter wear model is used to adaptively adjust the converter physical model in the converter digital twin model; The steelmaking process optimization module is used to input the input material composition information and converter process parameters as input variables into the converter digital twin model for simulation when the environmental data and converter usage time are set, and optimize the input material composition information and converter process parameters based on the target furnace mouth flame image to generate the optimal input material composition information and converter process parameters under different environmental data and converter usage time.

[0019] The present invention has the following advantages: The present invention performs a series of processing on the converter point cloud data set, and ensures the accuracy of the model by removing outliers in the point cloud data; surface reconstruction restores the actual geometric shape of the converter according to the point cloud data, and converts discrete point cloud data into a continuous surface model, laying an accurate geometric foundation for simulating the internal process of the converter; voxelization processing further discretizes the continuous surface model, providing a computational basis for thermodynamics, fluid mechanics, and chemical reaction analysis. In addition, through thermodynamics, fluid mechanics, and chemical reaction analysis, the established converter digital twin model can accurately describe the heat transfer, flow field changes, and chemical reactions in the steelmaking process, providing a physical rule basis for simulation, and also using The model is preliminarily verified and adjusted using standard environmental data and the working data of the new converter to ensure that the model can accurately reflect the actual process conditions. The digital twin model is updated to reflect the wear status of the converter by collecting the converter working data and the furnace mouth flame image in real time, so as to achieve adaptive adjustment of the model and more realistically reflect the actual status of the converter. Finally, the input material composition information and the converter process parameters are used as variables for optimization and adjustment through simulation, and the flame image is used as an evaluation indicator for the steelmaking process quality to further optimize the input material composition information and the converter process parameters, which greatly reduces the need to rely on the operator's experience to set the process parameters and improves the accuracy and efficiency of the steelmaking process. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a structural schematic diagram of a steelmaking process optimization system based on a digital twin model adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Example 1

[0022] A steelmaking process optimization method based on a digital twin model, comprising: Step S1: acquiring a point cloud data set of a converter by using a laser scanner, performing outlier removal, surface reconstruction and voxelization processing on the point cloud data set of the converter to construct a converter entity model; Step S2: Based on the converter entity model, a thermodynamic analysis, a fluid dynamics analysis, and a chemical reaction analysis are performed on the steelmaking process inside the converter to construct a converter physical model; Step S3: combining the converter entity model and the converter physical model to construct a converter digital twin model, and simulating the converter digital twin model according to standard environmental data and standard converter working data when the converter is not worn, that is, inputting the quasi-environmental data and the standard converter working data when the converter is not worn as input data into the simulation software storing the converter digital twin model for operation simulation. The simulation software may adopt ANSYS, and adjusting the physical parameters in the converter physical model in the converter digital twin model based on the standard furnace mouth flame image corresponding to the standard converter working data, and recording the adjusted physical parameters as standard physical parameters; Step S4: Collect the converter working data and the corresponding furnace mouth flame image within any period of time to form a local converter working data time series set, obtain all local converter working data time series sets to form a global converter working data time series set, simulate the converter digital twin model based on the global converter working data time series set to construct a converter wear model, and use the converter wear model to adaptively adjust the converter physical model in the converter digital twin model; Step S5: After setting the environmental data and the length of time the converter is in use, the input material composition information and the converter process parameters are input as input variables into the converter digital twin model for simulation, and the input material composition information and the converter process parameters are optimized based on the target furnace mouth flame image to generate the optimal input material composition information and converter process parameters under different environmental data and converter usage times, so as to optimize the steelmaking process.

[0023] The present invention performs abnormal point removal, surface reconstruction and voxelization on the point cloud data set of the converter, wherein the abnormal point removal can remove abnormal point cloud data in the point cloud data to ensure the accuracy of the digital twin model of the converter, and the surface reconstruction can reconstruct the geometric shape of the converter according to the point cloud data, and convert the discrete point cloud data into a continuous surface model, thereby providing an accurate geometric basis for simulating the internal process of the converter, and the voxelization processing further converts the continuous surface model into a discrete voxel grid, thereby providing a discrete calculation domain for subsequent thermodynamic analysis, fluid mechanics analysis and chemical reaction analysis; by performing thermodynamic analysis, fluid mechanics analysis and chemical reaction analysis on the steelmaking process, the established digital twin model of the converter can more accurately describe the heat conduction, flow field changes and material reactions occurring in the steelmaking process, thereby providing a physical rule basis for subsequent simulation; and The model is preliminarily verified and adjusted through standard environmental data and the working data of unworn converters to ensure that the model can accurately reflect the actual process conditions; and by collecting the real-time working data of the converter and the flame image of the furnace mouth, the converter digital twin model is updated to reflect the wear status of the converter, so that the converter digital twin model can be adaptively adjusted according to the usage time, which can more realistically reflect the actual status of the converter, and further ensure the accuracy when executing steelmaking process optimization; finally, the input material composition information and converter process parameters are used as variables, and the flame image is used as an indicator for steelmaking process quality evaluation, and the input material composition information and converter process parameters are optimized and adjusted, and the optimal input material composition information and converter process parameters under different environmental data and converter usage time are output, without the need for operators to set process parameters based on experience, thereby improving steelmaking efficiency.

[0024] Preferably, the specific content of step S1 is: Step S11: performing maximum bounding box sampling on the point cloud data set of the converter to obtain a downsampled point cloud data set of the converter; In this embodiment, the specific content of step S11 is, based on the collected point cloud data set of the converter, calculating the minimum cube that can surround the point cloud data set of the entire converter, that is, the maximum bounding box, the purpose of this cube is to determine a working area, so as to effectively manage and process the point cloud data; using the maximum bounding box to downsample the point cloud data set of the converter, that is, re-collecting the point cloud data at a predetermined sampling rate within the bounding box, this step can greatly reduce the amount of data, remove redundant information, and retain the key information of the geometric structure of the converter; Step S12: extracting neighborhood relations for each point cloud data in the downsampled point cloud data set of the converter, including the relative position of each point cloud data, etc., and generating a point cloud neighborhood relation matrix; Step S13: performing a graph structure conversion on the downsampled point cloud data set based on the point cloud neighborhood relationship matrix to generate a converter point cloud feature map; Step S14: extracting local structural features of each point cloud data and global structural features of all point cloud data for the converter point cloud feature map, and generating converter point cloud local structural features and converter point cloud global structural features; Step S15: removing abnormal points from the converter point cloud data set based on the local structural features of the converter point cloud and the global structural features of the converter point cloud. A corresponding threshold value may be set for each feature to achieve the removal of abnormal points and generate a denoised point cloud data set. Step S16: reconstructing the surface based on the point cloud data in the denoised point cloud data set to obtain a converter surface model, and globally optimizing the converter surface model under specified topological structures and geometric constraints; Step S17: performing three-dimensional voxel mesh division according to the converter surface model, and mapping the converter surface model into the three-dimensional voxel mesh and performing voxelization processing to obtain a converter solid model.

[0025] The present invention can reduce the complexity of point cloud data by performing maximum bounding box sampling on the point cloud data set of the converter, while retaining sufficient point cloud data information for constructing a converter entity model; by performing neighborhood relationship extraction on each point cloud data in the downsampled point cloud data set of the converter, the features corresponding to each point cloud data, such as relative distance, etc., can be extracted, and then the downsampled point cloud data set is converted into a graph structure through a cloud neighborhood relationship matrix, so that the point cloud data is converted into a graph structure, and the graph structure can better learn the global structural features of the converter point cloud such as point cloud density and distribution in the point cloud data, and finally, based on the local structural features of the converter point cloud and the global structural features of the converter point cloud, the removal of abnormal points can be achieved under a method such as setting a threshold; in the process of surface reconstruction of the point cloud data, global optimization of the converter surface model under the specified topological structure and geometric constraints can ensure the overall consistency of the model; voxelization of the converter surface model can accurately describe the spatial distribution and internal structure in the converter, and can also support more complex calculations and simulation operations.

[0026] Preferably, step S2 specifically includes the following contents: S21: Perform finite element analysis on the steelmaking process in the converter according to the converter entity model, and analyze the temperature field change in the converter through the heat conduction equation to obtain the converter temperature field model, thereby realizing thermodynamic analysis of the steelmaking process inside the converter; S22: Perform finite element analysis on the steelmaking process in the converter according to the converter entity model, and analyze the flow field change in the converter by using the Navier-Stokes equation to obtain the converter flow field model, thereby realizing the fluid mechanics analysis of the steelmaking process inside the converter; S23: performing finite element analysis on the steelmaking process in the converter according to the converter entity model, and analyzing the chemical reaction changes in the converter through a kinetic equation to obtain a converter chemical reaction kinetic model, thereby realizing chemical reaction analysis of the steelmaking process inside the converter; S33: The converter temperature field model, the converter flow field model and the converter chemical reaction kinetics model are coupled by the finite difference method to obtain the converter physical model.

[0027] The present invention realizes accurate analysis and simulation of key physical phenomena in the converter steelmaking process by simulating the temperature field, flow field and chemical reaction dynamics in the converter respectively, and provides a physical rule basis for subsequent simulation.

[0028] Preferably, the specific content of step S3 is: S31: Combine the converter entity model and the converter physical model to construct a converter digital twin model; S32: collecting standard environmental data, standard converter working data when the converter is not worn, and corresponding standard furnace mouth flame images, wherein the standard converter working data includes input material composition information and converter process parameters, and sending the standard converter working data to the converter digital twin model for simulation, and outputting a first simulated furnace mouth flame image; S33: performing spectral analysis on the standard furnace mouth flame image to generate standard furnace mouth flame spectrum data; performing polarized light analysis on the standard furnace mouth flame spectrum data to generate standard furnace mouth flame spectrum response data; performing image enhancement on the standard furnace mouth flame image based on the standard furnace mouth flame spectrum response data to generate a standard furnace mouth flame enhanced image; similarly, performing spectral analysis on the first simulated furnace mouth flame image to generate first simulated furnace mouth flame spectrum data; performing polarized light analysis on the first simulated furnace mouth flame spectrum data to generate first simulated furnace mouth flame spectrum response data; performing image enhancement on the first simulated furnace mouth flame image based on the first simulated furnace mouth flame spectrum response data to generate a first simulated furnace mouth flame enhanced image; S34: performing feature matching on the standard furnace mouth flame enhancement image and the first simulated furnace mouth flame enhancement image, and adopting algorithms such as SIFT (Scale Invariant Feature Transform) or SURF (Speeded Up Robust Feature) to perform feature matching, outputting the first furnace mouth flame difference data, and performing geometric consistency check during the feature matching process; S35: Using the flame difference data of the first furnace mouth as a constraint condition, the physical parameters in the converter physical model in the converter digital twin model are adjusted through the particle swarm optimization algorithm, and the adjusted physical parameters are recorded as standard physical parameters.

[0029] The present invention sends standard environmental data and working data of an unworn converter into a digital twin model of the converter for simulation, and performs feature matching on the output first simulated furnace mouth flame image and the standard furnace mouth flame image corresponding to the standard converter working data, determines the degree of difference between the first simulated furnace mouth flame image and the standard furnace mouth flame image corresponding to the standard converter working data, and uses this degree of difference as a constraint to adjust the physical parameters in the converter physical model, so that the digital twin model of the converter can more accurately reflect the actual steelmaking process; and using the furnace mouth flame image as a judgment standard is simpler and more convenient than quality inspection of steel products; image enhancement operations are also performed on the standard furnace mouth flame image and the first simulated furnace mouth flame image, and spectral analysis can extract material change information in the steelmaking process, and polarized light analysis can further enhance the image contrast. Performing image enhancement operations on the standard furnace mouth flame image and the first simulated furnace mouth flame image based on the results of spectral analysis and polarized light analysis can enable higher accuracy in subsequent calculations of differences between images.

[0030] Preferably, the specific steps of generating a standard furnace mouth flame enhancement image are: Step S331: performing spectral analysis on the standard furnace mouth flame image to generate spectral feature data of different bands, and splicing the spectral feature data of different bands according to channels to generate standard furnace mouth flame spectral data; Step S332: sampling the standard furnace mouth flame spectrum data at different polarization angles, the polarization angles being 0°, 45°, 90° and 135°, to obtain 0° standard furnace mouth flame spectrum response data, 45° standard furnace mouth flame spectrum response data, 90° standard furnace mouth flame spectrum response data and 135° standard furnace mouth flame spectrum response data; Step S333: combining the 0° standard furnace mouth flame spectral response data, the 45° standard furnace mouth flame spectral response data, the 90° standard furnace mouth flame spectral response data, and the 135° standard furnace mouth flame spectral response data to obtain the standard furnace mouth flame spectral response data; Step S334: linearly stretching the standard furnace mouth flame spectrum data based on the standard furnace mouth flame spectrum response data to obtain standard furnace mouth flame enhancement spectrum data; Step S335: performing image enhancement on the standard furnace mouth flame image based on the standard furnace mouth flame enhancement spectrum data and the attention mechanism to obtain a standard furnace mouth flame enhanced image; And the steps of generating the first simulated furnace mouth flame enhancement image are consistent with the steps of generating the standard furnace mouth flame enhancement image.

[0031] The present invention performs spectral analysis on the furnace mouth flame image. Since the combustion of substances will have different flames under different reaction conditions, spectral analysis can reflect the information of substance reaction changes. Polarized light analysis is performed on the extracted spectral data at different polarization angles to extract the flame spectral responses at different polarization angles, that is, the light intensity changes at different spectral bands. Then, based on the flame spectral response data, linear stretching is performed on the data range of the spectral data to make the image contrast more balanced and saturated. Finally, the standard furnace mouth flame image is enhanced through the attention mechanism and the standard furnace mouth flame enhancement spectral data, which can make the subsequent calculation of the difference between images have higher accuracy.

[0032] Preferably, the specific content of step S4 is as follows: Step S41: collecting converter working data and corresponding standard furnace mouth flame images within any period of time to form a local converter working time series data set, and obtaining all local converter working data time series sets to form a global converter working data time series set; Step S42: sending the local converter working data time series sets in the global converter working data time series set one by one into the converter digital twin model for simulation, and adjusting the physical parameters in the converter physical model in the converter digital twin model based on the furnace mouth flame image corresponding to the converter working data in the local converter working data time series set, and outputting the time series physical parameters corresponding to each local converter working data time series set; Step S43: performing differential operations on the time-series physical parameters corresponding to each local converter operation data time-series set and the standard physical parameters, respectively, to generate time-series physical parameter differential data corresponding to each local converter operation data time-series set; Step S44: using the time series physical parameter differential data corresponding to the time series set of all local converter working data arranged in chronological order as training data, training the time series prediction network model, the time series prediction network model adopts the LSTM model, and generating the converter wear model; Step S45: Predict the converter wear condition based on the converter wear model, generate the physical parameter adjustment coefficients of the converter under different usage times, and adaptively adjust the converter physical model in the converter digital twin model through the physical parameter adjustment coefficients.

[0033] The present invention sends the working data of the converter with different usage time into the digital twin model of the converter for simulation, calculates the physical parameters of the converter under different usage time, and uses the difference between the physical parameters of the converter under different usage time and the standard physical parameters as training data to construct a converter wear model. The converter wear model can be used to describe the changes in physical parameters of the converter physical model in the digital twin model of the converter under different usage time. The physical parameter adjustment coefficients at different usage time output by the converter wear model can realize adaptive adjustment of the converter physical model in the digital twin model of the converter under different usage time.

[0034] Preferably, the specific content of step S5 is as follows: Step S51: When the environmental data and the converter usage time are set, the input material composition information and the converter process parameters are input as input variables into the converter digital twin model for simulation, and a second simulated furnace mouth flame image is generated; Step S52: performing spectral analysis on the target furnace mouth flame image to generate target furnace mouth flame spectral data; performing polarized light analysis on the target furnace mouth flame spectral data to generate target furnace mouth flame spectral response data; performing image enhancement on the target furnace mouth flame image based on the target furnace mouth flame spectral response data to generate a target furnace mouth flame enhanced image; similarly, performing spectral analysis on the second simulated furnace mouth flame image to generate second simulated furnace mouth flame spectral data; performing polarized light analysis on the second simulated furnace mouth flame spectral data to generate second simulated furnace mouth flame spectral response data; performing image enhancement on the second simulated furnace mouth flame image based on the second simulated furnace mouth flame spectral response data to generate a second simulated furnace mouth flame enhanced image; Step S53: performing feature matching on the target furnace mouth flame enhancement image and the first simulated furnace mouth flame enhancement image, outputting second furnace mouth flame difference data, and performing geometric consistency check during the feature matching process; Step S54: Taking the second furnace mouth flame difference data as a constraint condition, the input material composition information and the converter process parameters are optimized by the particle swarm optimization algorithm to generate the optimal input material composition information and converter process parameters under different environmental data and converter usage time.

[0035] The present invention takes input material composition information and converter process parameters as variables, and simulates the steelmaking process corresponding to different input material composition information and converter process parameters through the converter digital twin model, and uses flame images as indicators for steelmaking process quality evaluation to optimize and adjust the input material composition information and converter process parameters, and outputs the optimal input material composition information and converter process parameters under different environmental data and converter usage time. Example 2

[0036] A steelmaking process optimization system based on digital twin model, see Figure 1 ,include: The converter digital twin model construction module is used to obtain the converter point cloud data set through a laser scanner, remove abnormal points, reconstruct surfaces and voxelize the converter point cloud data set to build a converter solid model, perform thermodynamic analysis, fluid dynamics analysis and chemical reaction analysis on the steelmaking process inside the converter based on the converter solid model to build a converter physical model, and combine the converter solid model and the converter physical model to build a converter digital twin model; A physical parameter adjustment module is used to simulate the converter digital twin model according to standard environmental data and standard converter working data when the converter is not worn, and adjust the physical parameters in the converter physical model in the converter digital twin model based on the standard furnace mouth flame image corresponding to the standard converter working data, and record the adjusted physical parameters as standard physical parameters; The converter wear model construction module is used to collect the converter working data and the corresponding furnace mouth flame image within any period of time to form a local converter working data time series set, obtain all local converter working data time series sets to form a global converter working data time series set, and simulate the converter digital twin model based on the global converter working data time series set to construct the converter wear model. The converter wear model is used to adaptively adjust the converter physical model in the converter digital twin model; The steelmaking process optimization module is used to input the input material composition information and converter process parameters as input variables into the converter digital twin model for simulation when the environmental data and converter usage time are set, and optimize the input material composition information and converter process parameters based on the target furnace mouth flame image to generate the optimal input material composition information and converter process parameters under different environmental data and converter usage time.

[0037] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A steelmaking process optimization method based on a digital twin model, characterized in that: include: Step S1: acquiring a point cloud data set of a converter by using a laser scanner, performing outlier removal, surface reconstruction and voxelization processing on the point cloud data set of the converter to construct a converter entity model; Step S2: Based on the converter entity model, a thermodynamic analysis, a fluid dynamics analysis, and a chemical reaction analysis are performed on the steelmaking process inside the converter to construct a converter physical model; Step S3: combining the converter entity model and the converter physical model to construct a converter digital twin model, simulating the converter digital twin model according to standard environmental data and standard converter working data when the converter is not worn, and adjusting the physical parameters in the converter physical model in the converter digital twin model based on the standard furnace mouth flame image corresponding to the standard converter working data, and recording the adjusted physical parameters as standard physical parameters; Step S4: Collect the converter working data and the corresponding furnace mouth flame image within any period of time to form a local converter working data time series set, obtain all local converter working data time series sets to form a global converter working data time series set, simulate the converter digital twin model based on the global converter working data time series set to construct a converter wear model, and use the converter wear model to adaptively adjust the converter physical model in the converter digital twin model; Step S5: After setting the environmental data and the length of time the converter is in use, the input material composition information and the converter process parameters are input as input variables into the converter digital twin model for simulation, and the input material composition information and the converter process parameters are optimized based on the target furnace mouth flame image to generate the optimal input material composition information and converter process parameters under different environmental data and converter usage times, so as to optimize the steelmaking process.

2. The steelmaking process optimization method based on the digital twin model according to claim 1 is characterized in that: The specific content of step S1 is: Step S11: performing maximum bounding box sampling on the point cloud data set of the converter to obtain a downsampled point cloud data set of the converter; Step S12: extracting neighborhood relations for each point cloud data in the downsampled point cloud data set of the converter, and generating a point cloud neighborhood relation matrix; Step S13: performing a graph structure conversion on the downsampled point cloud data set based on the point cloud neighborhood relationship matrix to generate a converter point cloud feature map; Step S14: extracting local structural features of each point cloud data and global structural features of all point cloud data for the converter point cloud feature map, and generating converter point cloud local structural features and converter point cloud global structural features; Step S15: removing abnormal points from the converter point cloud data set based on the local structural features of the converter point cloud and the global structural features of the converter point cloud to generate a denoised point cloud data set; Step S16: reconstructing the surface based on the point cloud data in the denoised point cloud data set to obtain a converter surface model, and globally optimizing the converter surface model under specified topological structures and geometric constraints; Step S17: performing three-dimensional voxel mesh division according to the converter surface model, and mapping the converter surface model into the three-dimensional voxel mesh and performing voxelization processing to obtain a converter solid model.

3. The steelmaking process optimization method based on digital twin model according to claim 2 is characterized in that: Step S2 specifically includes the following contents: S21: Perform finite element analysis on the steelmaking process in the converter according to the converter entity model, and analyze the temperature field change in the converter through the heat conduction equation to obtain the converter temperature field model, thereby realizing thermodynamic analysis of the steelmaking process inside the converter; S22: Perform finite element analysis on the steelmaking process in the converter according to the converter entity model, and analyze the flow field change in the converter by using the Navier-Stokes equation to obtain the converter flow field model, thereby realizing the fluid mechanics analysis of the steelmaking process inside the converter; S23: performing finite element analysis on the steelmaking process in the converter according to the converter entity model, and analyzing the chemical reaction changes in the converter through a kinetic equation to obtain a converter chemical reaction kinetic model, thereby realizing chemical reaction analysis of the steelmaking process inside the converter; S33: The converter temperature field model, the converter flow field model and the converter chemical reaction kinetics model are coupled by the finite difference method to obtain the converter physical model.

4. The steelmaking process optimization method based on digital twin model according to claim 3 is characterized in that: The specific content of step S3 is: S31: Combine the converter entity model and the converter physical model to construct a converter digital twin model; S32: collecting standard environmental data, standard converter working data when the converter is not worn, and corresponding standard furnace mouth flame images, wherein the standard converter working data includes input material composition information and converter process parameters, and sending the standard converter working data to the converter digital twin model for simulation, and outputting a first simulated furnace mouth flame image; S33: performing spectral analysis on the standard furnace mouth flame image to generate standard furnace mouth flame spectrum data; performing polarization light analysis on the standard furnace mouth flame spectrum data to generate standard furnace mouth flame spectrum response data; Based on the standard furnace mouth flame spectral response data, the standard furnace mouth flame image is enhanced to generate a standard furnace mouth flame enhanced image; similarly, the first simulated furnace mouth flame image is spectrally analyzed to generate first simulated furnace mouth flame spectral data; the first simulated furnace mouth flame spectral data is subjected to polarization light analysis to generate first simulated furnace mouth flame spectral response data; based on the first simulated furnace mouth flame spectral response data, the first simulated furnace mouth flame image is enhanced to generate a first simulated furnace mouth flame enhanced image; S34: performing feature matching on the standard furnace mouth flame enhancement image and the first simulated furnace mouth flame enhancement image, outputting first furnace mouth flame difference data, and performing geometric consistency check during the feature matching process; S35: Using the flame difference data of the first furnace mouth as a constraint condition, the physical parameters in the converter physical model in the converter digital twin model are adjusted through the particle swarm optimization algorithm, and the adjusted physical parameters are recorded as standard physical parameters.

5. The steelmaking process optimization method based on digital twin model according to claim 4 is characterized in that: The specific steps to generate a standard furnace flame enhancement image are: Step S331: performing spectral analysis on the standard furnace mouth flame image to generate spectral feature data of different bands, and splicing the spectral feature data of different bands according to channels to generate standard furnace mouth flame spectral data; Step S332: sampling the standard furnace mouth flame spectrum data at different polarization angles, the polarization angles being 0°, 45°, 90° and 135°, to obtain 0° standard furnace mouth flame spectrum response data, 45° standard furnace mouth flame spectrum response data, 90° standard furnace mouth flame spectrum response data and 135° standard furnace mouth flame spectrum response data; Step S333: combining the 0° standard furnace mouth flame spectral response data, the 45° standard furnace mouth flame spectral response data, the 90° standard furnace mouth flame spectral response data, and the 135° standard furnace mouth flame spectral response data to obtain the standard furnace mouth flame spectral response data; Step S334: linearly stretching the standard furnace mouth flame spectrum data based on the standard furnace mouth flame spectrum response data to obtain standard furnace mouth flame enhancement spectrum data; Step S335: performing image enhancement on the standard furnace mouth flame image based on the standard furnace mouth flame enhancement spectrum data and the attention mechanism to obtain a standard furnace mouth flame enhanced image; And the steps of generating the first simulated furnace mouth flame enhancement image are consistent with the steps of generating the standard furnace mouth flame enhancement image.

6. The steelmaking process optimization method based on digital twin model according to claim 5 is characterized in that: The specific content of step S4 is as follows: Step S41: collecting converter working data and corresponding standard furnace mouth flame images within any period of time to form a local converter working time series data set, and obtaining all local converter working data time series sets to form a global converter working data time series set; Step S42: sending the local converter working data time series sets in the global converter working data time series set one by one into the converter digital twin model for simulation, and adjusting the physical parameters in the converter physical model in the converter digital twin model based on the furnace mouth flame image corresponding to the converter working data in the local converter working data time series set, and outputting the time series physical parameters corresponding to each local converter working data time series set; Step S43: performing differential operations on the time-series physical parameters corresponding to each local converter operation data time-series set and the standard physical parameters, respectively, to generate time-series physical parameter differential data corresponding to each local converter operation data time-series set; Step S44: using the time series physical parameter differential data corresponding to the time series set of all local converter working data arranged in chronological order as training data, training the time series prediction network model, the time series prediction network model adopts the LSTM model, and generating the converter wear model; Step S45: Predict the converter wear condition based on the converter wear model, generate the physical parameter adjustment coefficients of the converter under different usage times, and adaptively adjust the converter physical model in the converter digital twin model through the physical parameter adjustment coefficients.

7. The steelmaking process optimization method based on digital twin model according to claim 6 is characterized in that: The specific content of step S5 is as follows: Step S51: When the environmental data and the converter usage time are set, the input material composition information and the converter process parameters are input as input variables into the converter digital twin model for simulation, and a second simulated furnace mouth flame image is generated; Step S52: performing spectral analysis on the target furnace mouth flame image to generate target furnace mouth flame spectral data; performing polarized light analysis on the target furnace mouth flame spectral data to generate target furnace mouth flame spectral response data; performing image enhancement on the target furnace mouth flame image based on the target furnace mouth flame spectral response data to generate a target furnace mouth flame enhanced image; similarly, performing spectral analysis on the second simulated furnace mouth flame image to generate second simulated furnace mouth flame spectral data; performing polarized light analysis on the second simulated furnace mouth flame spectral data to generate second simulated furnace mouth flame spectral response data; performing image enhancement on the second simulated furnace mouth flame image based on the second simulated furnace mouth flame spectral response data to generate a second simulated furnace mouth flame enhanced image; Step S53: performing feature matching on the target furnace mouth flame enhancement image and the first simulated furnace mouth flame enhancement image, outputting second furnace mouth flame difference data, and performing geometric consistency check during the feature matching process; Step S54: Taking the second furnace mouth flame difference data as a constraint condition, the input material composition information and the converter process parameters are optimized by the particle swarm optimization algorithm to generate the optimal input material composition information and converter process parameters under different environmental data and converter usage time.

8. A steelmaking process optimization system based on a digital twin model, characterized in that: The system applies a steelmaking process optimization method based on a digital twin model as described in any one of claims 1 to 7, comprising: The converter digital twin model construction module is used to obtain the converter point cloud data set through a laser scanner, remove abnormal points, reconstruct surfaces and voxelize the converter point cloud data set to build a converter solid model, perform thermodynamic analysis, fluid dynamics analysis and chemical reaction analysis on the steelmaking process inside the converter based on the converter solid model to build a converter physical model, and combine the converter solid model and the converter physical model to build a converter digital twin model; A physical parameter adjustment module is used to simulate the converter digital twin model according to standard environmental data and standard converter working data when the converter is not worn, and adjust the physical parameters in the converter physical model in the converter digital twin model based on the standard furnace mouth flame image corresponding to the standard converter working data, and record the adjusted physical parameters as standard physical parameters; The converter wear model construction module is used to collect the converter working data and the corresponding furnace mouth flame image within any period of time to form a local converter working data time series set, obtain all local converter working data time series sets to form a global converter working data time series set, and simulate the converter digital twin model based on the global converter working data time series set to construct the converter wear model. The converter wear model is used to adaptively adjust the converter physical model in the converter digital twin model; The steelmaking process optimization module is used to input the input material composition information and converter process parameters as input variables into the converter digital twin model for simulation when the environmental data and converter usage time are set, and optimize the input material composition information and converter process parameters based on the target furnace mouth flame image to generate the optimal input material composition information and converter process parameters under different environmental data and converter usage time.

Images