Intelligent control method and device for metallurgical reactor based on real-time simulation and monitoring

Abstract

The present invention provides an intelligent control method and device for a metallurgical reactor based on real-time simulation and monitoring, and relates to the field of metallurgical intelligent control technology. The method includes: obtaining image data, geometric parameters, melt physical parameters, process parameters, and monitoring data of the smelting process of the metallurgical reactor, and forming a data set with the obtained data and the results of the simulation; constructing an image parameter recognition model A, a clearance height prediction model B, a multi-physical field prediction model C, and a process parameter optimization model D for the metallurgical reactor; obtaining image data of the smelting process of the current heat metallurgical reactor, and intelligently controlling the process parameters of the metallurgical reactor based on the above-established models. The present invention provides an efficient and intelligent control method for a metallurgical reactor based on real-time simulation monitoring and capable of accurately reflecting the physical field information within the reactor.

Description

Technical Field

[0001] The present invention relates to the technical field of metallurgical intelligent control, and in particular to an intelligent control method and device for a metallurgical reactor based on real-time simulation and monitoring. Background Art

[0002] Metallurgical reactors, such as blast furnaces, agitators for hot metal desulfurization (Kanbara Reactor, KR), converters, crystallizers, and electrolytic cells, serve as chemical reaction carriers and containers in the metallurgical industry and play a vital role. Controlling the efficient and stable production of metallurgical reactors is crucial for reducing costs and increasing efficiency in the industry. Currently, production control during the production of metallurgical reactors is often done manually based on real-time on-site monitoring of information such as temperature and oxygen flow rates from various sensors. However, existing sensors capture very limited information, capable of only monitoring a small amount of physical information within a small area and unable to fully monitor the detailed physical field within the entire metallurgical reactor. This limited amount of monitoring information makes it difficult to accurately reflect the overall operating status of the metallurgical reactor. Consequently, manual control based on this limited information makes it difficult to achieve stable and efficient utilization of the metallurgical reactor's performance.

[0003] Simulation is a method that uses a computer to solve partial differential equations for various physical quantities within a metallurgical reactor to obtain information about various physical fields. By coupling multiple physical fields, high-precision simulations of metallurgical reactors can be achieved, and the results obtained play an important guiding role in the regulation of metallurgical reactors. Compared to sensor monitoring, this method can obtain detailed physical field information within the entire metallurgical reactor. However, the disadvantage is that simulation efficiency is low, and when the production process of the metallurgical reactor changes, it is impossible to quickly solve the physical field information within the metallurgical reactor after the change.

[0004] The calculation model of the metallurgical reactor is complex to change, and it is difficult to consider the impact of the increase in the age of the metallurgical reactor on the geometric model. The results obtained by simulating the initial geometric model can accurately reflect the physical field information inside the metallurgical reactor in the early stage, but cannot accurately reflect the physical field information inside the metallurgical reactor at different ages.

[0005] In the existing technology, there is a lack of an efficient and intelligent control method for metallurgical reactors based on real-time simulation monitoring and capable of accurately reflecting the physical field information within the reactor. Summary of the Invention

[0006] In order to solve the technical problem that the existing technology can only monitor a small amount of physical information in a small area during the metallurgical reactor production process, and cannot accurately reflect the physical field information inside the metallurgical reactor, resulting in the difficulty of manually controlling and achieving stable and efficient utilization of the metallurgical reactor performance, the embodiment of the present invention provides an intelligent control method and device for metallurgical reactors based on real-time simulation and monitoring. The technical solution is as follows:

[0007] In one aspect, a method for intelligent control of a metallurgical reactor based on real-time simulation and monitoring is provided, the method being implemented by an intelligent control device, the method comprising:

[0008] During the smelting process, the metallurgical reactor is monitored by N max Heat data collection, obtaining first image data of the metallurgical melt before it is charged into the metallurgical reactor, first geometric parameters of the metallurgical reactor before it is charged into the metallurgical reactor, second image data of the metallurgical melt after it is charged into the metallurgical reactor, second geometric parameters of the metallurgical reactor after it is charged into the metallurgical reactor, melt physical property parameters, first process parameters, and physical field monitoring data;

[0009] Using the first image data and N of the first geometric parameters a Heat data to build a data set a; according to the second image data, the second geometric parameters and the melt physical parameters N a The heat data constructs dataset b;

[0010] Based on the data set a, a model is constructed using the 3DGS algorithm to obtain an image parameter recognition model A; based on the data set b, a model is constructed based on OpenCV to obtain a clearance height prediction model B;

[0011] Based on the image parameter recognition model A and the headroom prediction model B, calculations are performed according to the first image data and the second image data to obtain a melt headroom height, a topological scheme of a reactor geometric structure, a reactor geometric model, and a structured grid of the reactor geometric model;

[0012] Based on the reactor geometric model, the structured grid, and preset metallurgical industry specifications, and according to the melt physical property parameters, the first process parameters, and the physical field monitoring data, CFD software is used to perform simulation to obtain multi-physical field simulation data;

[0013] N is constructed according to the multi-physics field simulation data, the reactor geometry model, the melt physical property parameters, the melt clearance height and the first process parameters. a The dataset c and N of the group data maxA data set d of group data; using the data set c, training a first deep learning prediction model to obtain a multi-physics field prediction model C;

[0014] Inputting the data set d into the multi-physics field prediction model C for data prediction to obtain multi-physics field prediction data; constructing a data set e based on the multi-physics field prediction data, the reactor geometry model, the melt physical properties, the melt clearance height, and the first process parameters; using the data set e, training a second deep learning prediction model to obtain a process parameter optimization model D;

[0015] Through the monitoring equipment, the third image data of the current heat of metallurgical melt before being charged into the metallurgical reactor, the fourth image data after the metallurgical melt is charged into the metallurgical reactor, the physical field data of the current heat and the target physical field data are obtained; based on the preset physical field threshold range, according to the third image data of the current heat of metallurgical melt before being charged into the metallurgical reactor, the fourth image data after the metallurgical melt is charged into the metallurgical reactor, the melt physical properties, the physical field data of the current heat and the target physical field data, the reactor process parameters are controlled through the image parameter recognition model A, the clearance height prediction model B, the multi-physical field prediction model C and the process parameter optimization model D to obtain the second process parameters; the metallurgical reactor is controlled according to the second process parameters.

[0016] On the other hand, an intelligent control device for a metallurgical reactor based on real-time simulation and monitoring is provided. The device is applied to an intelligent control method for a metallurgical reactor based on real-time simulation and monitoring. The device comprises:

[0017] The training data acquisition module is used to monitor the metallurgical reactor during the smelting process through monitoring equipment. max Heat data collection, obtaining first image data of the metallurgical melt before it is charged into the metallurgical reactor, first geometric parameters of the metallurgical reactor before it is charged into the metallurgical reactor, second image data of the metallurgical melt after it is charged into the metallurgical reactor, second geometric parameters of the metallurgical reactor after it is charged into the metallurgical reactor, melt physical property parameters, first process parameters, and physical field monitoring data;

[0018] A data set construction module is configured to use the first image data and N of the first geometric parameters to construct a a Heat data to build a data set a; according to the second image data, the second geometric parameters and the melt physical parameters N a Heat data constructs data set b; the N a The maximum number of smelting furnaces N in the current age of the metallurgical reactor max one-third;

[0019] The first model training module is used to construct a model based on the data set a using the 3DGS algorithm to obtain an image parameter recognition model A; and to construct a model based on the data set b based on OpenCV to obtain a clearance height prediction model B;

[0020] a topology solution construction module, configured to calculate, based on the image parameter recognition model A and the headroom prediction model B, the melt headroom, a topology solution of the reactor geometry, a reactor geometry model, and a structured grid of the reactor geometry model according to the first image data and the second image data;

[0021] a multi-physics field simulation module, configured to perform simulation using CFD software based on the reactor geometry model, the structured grid, and preset metallurgical industry specifications, according to the melt physical property parameters, the first process parameters, and the physical field monitoring data, to obtain multi-physics field simulation data;

[0022] The second model training module is used to construct N according to the multi-physics field simulation data, the reactor geometry model, the melt physical property parameters, the melt clearance height and the first process parameters. a The dataset c and N of the group data max Group data set d; using the data set c, the first deep learning prediction model is trained to obtain a multi-physics field prediction model C;

[0023] a third model training module, configured to input the data set d into the multi-physics field prediction model C for data prediction to obtain multi-physics field prediction data; construct a data set e based on the multi-physics field prediction data, the reactor geometry model, the melt physical properties, the melt clearance height, and the first process parameters; and use the data set e to train a second deep learning prediction model to obtain a process parameter optimization model D;

[0024] A metallurgical reactor control module is used to obtain, through monitoring equipment, third image data of the current heat of metallurgical melt before it is loaded into the metallurgical reactor, fourth image data after the metallurgical melt is loaded into the metallurgical reactor, physical field data of the current heat, and target physical field data; based on a preset physical field threshold range, according to the third image data of the current heat of metallurgical melt before it is loaded into the metallurgical reactor, fourth image data after the metallurgical melt is loaded into the metallurgical reactor, the melt physical properties, physical field data of the current heat, and target physical field data, the reactor process parameters are controlled through the image parameter recognition model A, the clearance height prediction model B, the multi-physical field prediction model C, and the process parameter optimization model D to obtain a second process parameter; and the metallurgical reactor is controlled according to the second process parameter.

[0025] On the other hand, an intelligent control device is provided, which includes: a processor; a memory, on which computer-readable instructions are stored, and when the computer-readable instructions are executed by the processor, any one of the above-mentioned intelligent control methods of metallurgical reactors based on real-time simulation and monitoring is implemented.

[0026] On the other hand, a computer-readable storage medium is provided, in which at least one instruction is stored. The at least one instruction is loaded and executed by a processor to implement any one of the above-mentioned intelligent control methods of metallurgical reactors based on real-time simulation and monitoring.

[0027] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0028] The present invention proposes an intelligent control method for metallurgical reactors based on real-time simulation and monitoring. It combines image recognition, machine learning and simulation methods, uses geometric image recognition technology and parametric automatic modeling and meshing technology to obtain a large amount of simulation data; utilizes machine learning algorithms to achieve real-time simulation and prediction of metallurgical reactors, and on this basis completes dynamic monitoring of multiple physical fields inside the metallurgical reactor; establishes a deep learning optimization prediction model to achieve intelligent control of metallurgical reactors. The present invention can achieve real-time simulation of multiple physical fields in metallurgical reactors and intelligent control of metallurgical reactors, solving the problems of difficulty in obtaining monitoring information of the physical fields in metallurgical reactors during production, low simulation efficiency and inability to obtain information in real time, and dependence on experience and difficulty in controlling metallurgical reactors. The present invention is an efficient and intelligent control method for metallurgical reactors based on real-time simulation and monitoring that can accurately reflect the physical field information inside the reactor. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0030] Figure 1 This is a flow chart of an intelligent control method for a metallurgical reactor based on real-time simulation and monitoring provided by an embodiment of the present invention;

[0031] Figure 2 This is a block diagram of an intelligent control device for a metallurgical reactor based on real-time simulation and monitoring provided by an embodiment of the present invention;

[0032] Figure 3 It is a structural diagram of an intelligent control device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0033] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0034] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as an "exemplary" in the present invention should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner. Furthermore, in the embodiments of the present invention, "and / or" can mean both or either of the two.

[0035] In the embodiments of the present invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same. The terms "of," "corresponding," and "corresponding" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same.

[0036] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0037] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0038] The embodiment of the present invention provides an intelligent control method for a metallurgical reactor based on real-time simulation and monitoring. The method can be implemented by an intelligent control device, which can be a terminal or a server. Figure 1 The flowchart of the intelligent control method of a metallurgical reactor based on real-time simulation and monitoring is shown. The processing flow of the method may include the following steps:

[0039] S1. During the smelting process, the metallurgical reactor is monitored by monitoring equipment. max Heat data collection obtains the first image data of the metallurgical melt before it is charged into the metallurgical reactor, the first geometric parameters of the metallurgical reactor before it is charged into the metallurgical reactor, the second image data after the metallurgical melt is charged into the metallurgical reactor, the second geometric parameters of the metallurgical reactor after it is charged into the metallurgical reactor, the melt physical properties, the first process parameters and the physical field monitoring data.

[0040] In a feasible implementation mode, the present invention collects N (N=0,1,2…N max(Maximum age of metallurgical reactor)) After each heat, the image data of the metallurgical reactor before and after the metallurgical melt is loaded, the three-dimensional structure and geometric parameters of the metallurgical reactor, the physical properties of the metallurgical melt in the reactor, the process parameters used in production, and the monitoring data monitored by the monitoring equipment under the process.

[0041] Metallurgical reactors include but are not limited to blast furnaces, KR stirrers, converters, ladles, tundishes, crystallizers and other metallurgical reactors. The geometric parameters of the metallurgical reactor should include the length, width, height, etc. of the metallurgical reactor, and the parameter data necessary to construct the mathematical model of the metallurgical reactor. The acquired image data should cover but not be limited to the three views of the interior of the metallurgical reactor before and after the metallurgical melt is loaded; the acquired melt physical parameters should include but are not limited to the density, temperature, viscosity, etc. of the metallurgical melt. The first process parameters acquired should include: such as oxygen lance height, nozzle immersion depth, etc. The parameters monitored on site should include but are not limited to the liquid level fluctuation inside the metallurgical reactor, the metallurgical melt temperature, the magnetic field distribution and intensity, etc.

[0042] Among them, N max It is the maximum value of smelting heat N in the current age of the metallurgical reactor.

[0043] In a feasible implementation, monitoring data is collected using multiple types of monitoring equipment, including: images, surveillance cameras: measuring the level of metallurgical melts, ultrasonic level meters; temperature measurement, thermocouples, infrared thermometers; pressure transmitters: used to measure the density, pressure and flow of liquids and gases; oxygen sensors, measuring the oxygen concentration in the production process, etc., totaling N max Group data.

[0044] S2, using the first image data and N of the first geometric parameters a Heat data is used to construct a data set a; according to the second image data, the second geometric parameters and the melt physical parameters N a The heat data constructs dataset b.

[0045] Among them, N a The maximum number of smelting furnaces N in the current age of the metallurgical reactor max One third of .

[0046] In a feasible embodiment, the N used in this step is a The number of furnaces is the total number of furnaces N max One third of a Heat data comes from N max The early data, mid-term data and late data of the heat data are randomly selected from the early, mid-term and late stages of the total heat data to construct corresponding data sets a and b.

[0047] Based on the image data and geometric parameters of the metallurgical reactor before being loaded with metallurgical melt, a dataset a is constructed and divided into a training set and a test set in a ratio of 8:2.

[0048] Based on the image data and geometric parameters of the metallurgical reactor after the metallurgical melt is loaded, the melt clearance height is calculated according to the melt physical properties, and a dataset b is constructed. The dataset b is divided into a training set and a test set in a ratio of 8:2.

[0049] S3. Based on data set a, the 3DGS algorithm is used to build a model to obtain image parameter recognition model A. Based on OpenCV, the model is built according to data set b to obtain clearance height prediction model B.

[0050] In one feasible implementation, an image parameter recognition model A is established using an algorithm for 3D reconstruction and rendering (3D Gaussian Splatting, 3DGS). Model A is trained based on a training set from dataset a. The trained model A is used to extract the internal geometric parameters of the metallurgical reactor. The 3D structure of the metallurgical reactor is then constructed based on these geometric parameters. A deep learning prediction model B is then established based on a deep learning algorithm. The trained model B is used to predict the headroom height of the melt within the metallurgical reactor.

[0051] The test sets in data sets a and b are used to verify the accuracy of image parameter recognition model A and clearance height prediction model B respectively.

[0052] For model A: The three-dimensional structure and geometric parameters obtained by model A are used to build a mathematical model. In the same coordinate system, 100 points are randomly selected in the test case. The relative error A of the coordinate values ​​in the x, y, and z directions is x , A y , A z Less than or equal to 5%, and more than 95% of each model in the test set data meets A x , A y , A z If the error is less than or equal to 5%, model A is considered accurate; otherwise, model A is considered inaccurate and model A should be readjusted and trained, and the accuracy of the model should be verified again until the model meets the requirements. For model B: if the relative error B1 between the predicted clearance height of the melt and the actual clearance height of more than 95% of the image data in the test set is less than or equal to 5%, the predicted model B is considered accurate; otherwise, it should be readjusted and trained as above until it meets the requirements.

[0053] Among them, the accuracy rate A x , A y , A z , the calculation formula of B1 is as follows (1), (2), (3), (4):

[0054] (1);

[0055] (2);

[0056] (3);

[0057] (4);

[0058] S4. Based on the image parameter recognition model A and the clearance height prediction model B, calculations are performed according to the first image data and the second image data to obtain the melt clearance height, the topological scheme of the reactor geometric structure, the reactor geometric model, and the structured grid of the reactor geometric model.

[0059] Optionally, based on the image parameter recognition model A and the headroom prediction model B, calculations are performed according to the first image data and the second image data to obtain the melt headroom height, the topological scheme of the reactor geometry, the reactor geometry model, and the structured grid of the reactor geometry model, including:

[0060] Inputting the first image data into the image parameter recognition model A to generate parameters and obtain third geometric parameters;

[0061] Inputting the second image data into the headroom height prediction model B to predict the melt headroom height, thereby obtaining the melt headroom height;

[0062] Constructing a topology scheme of the metallurgical reactor geometry model according to the third geometric parameter and the melt clearance height;

[0063] Based on Python and OpenFOAM, we wrote a mesh parameterization code for the metallurgical reactor geometry according to the topology scheme to obtain a pre-processing program.

[0064] The third geometric parameter is input into a pre-processing program to generate a mesh, thereby obtaining a reactor geometric model and a structured mesh of the reactor geometric model.

[0065] In a feasible implementation method, model A is used to extract the third geometric parameters of all furnaces within the service life of the metallurgical reactor, and combined with model B to predict the melt clearance height in the metallurgical reactor, all three-dimensional structures are approximated by simple geometric bodies (cylinders, spheres, cuboids, etc.) to obtain a topological solution that meets the requirements of the metallurgical reactor model after all times of use.

[0066] Use Python combined with OpenFOAM to write a pre-processing program for the automatic generation of metallurgical reactor models and their structured mesh parameterization.

[0067] For metallurgical reactors after using different heats, by acquiring images before loading the metallurgical melt, the three-dimensional structure and geometric parameters of the metallurgical reactor are obtained using model A. The obtained geometric parameters are input into the above-mentioned pre-processing program to quickly batch generate metallurgical reactor models and their structured grids after using different heats.

[0068] S5. Based on the reactor geometry model, structured grid and preset metallurgical industry specifications, CFD software is used to perform simulation according to the melt physical properties, the first process parameters and the physical field monitoring data to obtain multi-physical field simulation data.

[0069] Optionally, based on the reactor geometry model, structured grid, and preset metallurgical industry specifications, CFD software is used to perform simulation according to the melt physical property parameters, the first process parameters, and physical field monitoring data to obtain multi-physics field simulation data, including:

[0070] Based on the preset metallurgical industry specifications, the operating range is determined according to the primary process parameters, melt physical properties and physical field monitoring data; a simulation plan is formulated according to the gradient based on the operating range;

[0071] Based on the simulation scheme, CFD software is used to perform simulation according to the reactor geometric model and structured grid to obtain multi-physics field simulation data.

[0072] In a feasible implementation method, multiple groups of simulation plans are formulated with reference to the operating range determined by the metallurgical industry specifications, according to actual production technical conditions and reasonable gradient conditions.

[0073] Use CFD software to read the generated reactor geometry models and structured meshes for different heats and perform calculations according to the established simulation plan. Use CFD software such as Fluent and OpenFOAM and pre-processing software to generate metallurgical reactor models and meshes. Calculate according to the established simulation plan, set boundary conditions and initial values.

[0074] The simulation results are verified using the physical field monitoring data obtained by the on-site monitor. If the accuracy of the simulation results does not meet the use requirements, the calculation model and grid are readjusted until the accuracy of the simulation results meets the requirements. The method is characterized in that the simulation calculation results need to be verified. The verification method is to compare the values ​​monitored by the reactor production on-site monitor (such as the melt temperature of the ladle at a certain moment; the end temperature of the converter; the S content of the KR stirrer end, etc.) with the corresponding simulation calculation values. The verification method is to use the same production plan and arbitrarily take multiple groups of simulation results and corresponding monitoring results under different furnace ages. If the relative error M1 of more than 95% of the results is less than 5%, the simulation results are considered correct. Otherwise, the simulation results are considered inaccurate and the calculation model and the number of grids need to be readjusted and recalculated until the accuracy meets the requirements. If the accuracy still does not meet the requirements after adjusting the calculation model and the number of grids, it is necessary to re-optimize the topology structure and write the pre-processing software to improve the grid quality and ensure the calculation accuracy. The calculation formula of the relative error is as follows (5):

[0075] (5)

[0076] S6, construct N according to the multi-physics field simulation data, the reactor geometry model, the melt physical properties, the melt clearance height and the first process parameters a The dataset c and N of the group data max The first deep learning prediction model is trained using the dataset c to obtain the multi-physics field prediction model C.

[0077] In a feasible embodiment, the present invention uses post-processing software (such as Tecplot) to post-process the calculated results, extract the data of multiple physical fields such as temperature field, flow field, solute field, magnetic field, etc., and compare the extracted data with the corresponding N a The reactor geometry model, melt physical parameters, melt clearance and N max The first process parameter of the group is used to construct the data set c, and the data set c is divided into a training set and a test set at a ratio of 9:1. Similarly, the data set d is constructed and divided into a training set and a test set at a ratio of 9:1. The data in the data set c is N a Heat group data, the data in data set d is N max Total heat group data.

[0078] The prediction model C is verified using all the data in the test set of the dataset c. The verification standard is to randomly select the prediction results of 100 points inside the metallurgical container. If the relative error C1 between the prediction results (temperature field, solute field, etc.) of the 100 points and the simulation calculation results of the corresponding 100 points in the test set in c is less than or equal to 5% and more than 95% of the test data in the test set meet the requirements, the recognition model C is considered accurate. Otherwise, the recognition model C is considered inaccurate and the recognition model C should be readjusted and trained, and the accuracy of the model should be verified again until the model meets the requirements. The calculation formula of the relative error C1 is as follows (6):

[0079] (6);

[0080] S7. Input the data set d into the multi-physics field prediction model C for data prediction to obtain multi-physics field prediction data; construct a data set e based on the multi-physics field prediction data, the reactor geometry model, the melt physical properties, the melt clearance height and the first process parameters; use the data set e to train the second deep learning prediction model to obtain the process parameter optimization model D.

[0081] In a feasible implementation, the data in the test set in the data set e is used to obtain the optimized second process parameters through the prediction model D, and the second process parameters are passed into the model C. If the relative error D1 between the multi-physical field prediction result of the model C and the target physical field is less than 5% and more than 95% of the data in the test set meet the relative error D1 less than 5%, then the model D is considered accurate. Otherwise, the prediction model D is considered inaccurate, and the recognition model D should be readjusted and trained, and the accuracy of the model should be verified again until the model meets the requirements; arbitrarily take 100 points at the same position in the target physical field and the predicted physical field.

[0082] The calculation formula of the relative error D1 is as follows (7):

[0083] (7);

[0084] Among them, the model structure of the first deep learning prediction model and the second deep learning prediction model is a convolutional neural network, a long short-term memory neural network or a bidirectional gated recurrent neural network.

[0085] In one feasible implementation, different models and approaches are used for different containers and operating conditions. Commonly used deep learning prediction models include convolutional neural networks, long-short-term memory neural networks, principal component analysis, multi-head attention mechanisms, and bidirectional gated recurrent algorithms. Depending on the complexity of the model, a combination of multiple neural network models can be used, and hyperparameters can be adjusted based on the model's characteristics.

[0086] S8. Obtain, through monitoring equipment, the third image data of the current heat of metallurgical melt before it is loaded into the metallurgical reactor, the fourth image data after the metallurgical melt is loaded into the metallurgical reactor, the physical field data of the current heat, and the target physical field data; based on a preset physical field threshold range, and according to the third image data of the current heat of metallurgical melt before it is loaded into the metallurgical reactor, the fourth image data after the metallurgical melt is loaded into the metallurgical reactor, the melt physical properties, the physical field data of the current heat, and the target physical field data, control the reactor process parameters through the image parameter recognition model A, the clearance height prediction model B, the multi-physical field prediction model C, and the process parameter optimization model D to obtain the second process parameters; control the metallurgical reactor according to the second process parameters.

[0087] Optionally, based on a preset physical field threshold range, according to the third image data of the current heat of metallurgical melt before it is charged into the metallurgical reactor, the fourth image data of the metallurgical melt after it is charged into the metallurgical reactor, the melt physical property parameters, the current heat physical field data and the target physical field data, the reactor process parameter is controlled by the image parameter recognition model A, the clearance height prediction model B, the multi-physical field prediction model C and the process parameter optimization model D to obtain the second process parameter, including:

[0088] Inputting the third image data into the image parameter recognition model A to obtain the current heat geometric parameters; performing data processing through the pre-processing program according to the current heat geometric parameters to obtain the current reactor geometric model;

[0089] Inputting the fourth image data into the headroom height prediction model B to obtain the headroom height of the melt of the current furnace;

[0090] Input the current reactor geometry model, the current furnace melt clearance height, the current furnace physical field data, the target physical field data and the melt physical property parameters into the process parameter optimization model D to obtain the predicted process parameters;

[0091] Input the current reactor geometry model, the current furnace melt clearance height, melt physical properties and predicted process parameters into the multi-physics prediction model C to obtain predicted multi-physics data;

[0092] Verify the predicted multi-physics field data according to the preset physics field threshold range to obtain the physics field verification result;

[0093] When the physical field verification result is passed, the predicted process parameter is determined as the second process parameter.

[0094] In a feasible implementation manner, images of a metallurgical reactor before and after a certain heat of metallurgical melt is loaded in actual production, and physical properties of the metallurgical melt in the reactor are obtained, a geometric model of the metallurgical reactor is obtained using model A and pre-processing software, and the headroom height of the melt in the metallurgical reactor is obtained using model B. Based on the above results and the initial physical field / physical field to be optimized and the target physical field monitored on site, they are input into model D, and the optimized process parameters for achieving the target physical field are output. The prediction results of model D are used as input to model C, and model C is run in parallel with actual production to obtain prediction results of multiple physical fields in the metallurgical reactor under the production process through real-time monitoring;

[0095] The maximum production capacity of a metallurgical reactor is constrained by the equipment's maximum load capacity. The threshold ranges for relevant physical field values ​​are determined by the reactor's power and size, as well as the refractory material and furnace age. This invention sets threshold ranges for relevant physical fields, such as temperature and flow fields, based on the reactor's specific production conditions.

[0096] After the actual production starts, if the physical field results output by model C exceed the preset physical field threshold range at a certain moment, the physical field monitoring information and physical field threshold information at this moment are input into model D to obtain the process parameters that can achieve the target physical field and enable stable production again, and perform physical field threshold range verification until the process parameters meet the production requirements under the target physical field.

[0097] In a feasible embodiment, the present invention is based on the application of the KR stirrer as follows:

[0098] After 300 heats of use, a total of 300 sets of image data were obtained, including the KR stirrer's three-dimensional structure and geometric parameters, the physical properties and weight of the molten iron, the process parameters used in each heat production (stirrer immersion depth, stirrer speed), the temperature measured under the process, and the S content obtained by manual sampling.

[0099] The images of the KR stirrer before being loaded with molten iron after different heats and the geometric parameters of the stirrer were obtained to construct a dataset a, and the 3D Gaussian Splatting (3DGS) algorithm was used to establish model A.

[0100] A total of 90 sets of data were extracted, including furnaces 1-10, 40-50, 80-90, 100-110, 140-150, 180-190, 200-210, 240-250, and 280-290. The image data and geometric parameters of the KR stirrer before the stirrer is loaded with molten iron. The image data include the three-view images of the KR stirring head and the three-view images of the KR stirring tank. The geometric parameters include the geometric parameters of the KR stirring head and the geometric parameters of the KR stirring tank. The constructed data a is divided into training set and test set in a ratio of 8:2. The 3DGS algorithm is used to establish the 3D model A of the KR stirrer. The model can quickly identify the 3D structure and geometric parameters of the KR stirrer by inputting the three-view images of the stirring head and the stirring tank. The test set data in the data set a are used to verify the model A. The relative error A1 of the 18 groups of results constructed by the 3D structure and geometric parameters identified using the image data of a total of 18 heats in the test set and the model constructed with the actual size in the same coordinate system is less than 5%, proving that the model is accurate.

[0101] Images of the KR stirrer after being loaded with molten iron after different heats, as well as the stirrer's geometric parameters, tonnage, and density of the loaded molten iron, were obtained. The clearance height of the molten pool in the KR stirrer (i.e., the initial liquid level) was calculated based on the obtained stirrer geometric parameters, molten iron tonnage, and density. The calculated clearance height and images of the KR stirrer after being loaded with molten iron were used to construct dataset b. A deep learning algorithm was used to establish model B, and model B was verified.

[0102] In this step, a total of 90 groups of data were extracted, including furnaces 1-10, 40-50, 80-90, 100-110, 140-150, 180-190, 200-210, 240-250, and 280-290. The geometric parameters of the KR stirrer after the stirrer is loaded with molten iron, the image data after the molten iron is loaded, and the tonnage and density of the loaded molten iron are obtained. The image data should be able to clearly reflect the relative position of the molten iron and the top of the stirring tank. The geometric parameters include the geometric parameters of the KR stirring head and the geometric parameters of the KR stirring tank. The clearance height of the molten pool is calculated by obtaining the geometric parameters of the stirring tank and the tonnage and density of the molten iron. The images after the molten iron is loaded and the calculated clearance height of the molten pool are used to construct a dataset b and divide the training set and test set into a ratio of 8:2. The deep learning algorithm is used to establish a prediction model B for the clearance height of the molten pool. The model can predict the clearance height of the molten pool in the current furnace by inputting the image after the molten iron is loaded. The test set data in the dataset b are used to test the model B. The relative error B1 between the molten pool clearance height predicted by the 18 sets of image data in the test set and the actual clearance height is less than 5%, which proves that the model is accurate.

[0103] The established model A was used to identify the image data of 300 heats of KR agitator, and the three-dimensional structure of the KR agitator in all heats was obtained. A topological method for automatic generation of structured grids was developed based on the obtained three-dimensional structure. According to the developed topological scheme, the Python programming language was combined with the open source software OpenFOAM to write the pre-processing software for automatic integrated generation of the model and structured grid parameterization.

[0104] The software developed in this step can quickly generate a KR stirrer model and structured mesh by simply inputting the KR stirrer model parameters. The generated mesh is high-quality and takes only seconds to generate the model and mesh. Furthermore, as the KR stirrer ages and the KR stirring head and stirring tank become worn and eroded, the corresponding geometric model and structured mesh can be regenerated in seconds by modifying the corresponding geometric parameters.

[0105] Based on the acquired on-site KR production process data, including agitator speed, agitator head immersion depth, desulfurizer composition, and physical parameters of the desulfurizer, such as mass, initial molten iron temperature, density, and viscosity, multiple orthogonal simulation schemes were developed according to reasonable gradient conditions. The pre-processing software was used to quickly generate 90 sets of models and structured grids for furnace ages 1-10, 40-50, 80-90, 100-110, 140-150, 180-190, 200-210, 240-250, and 280-290. Simulation calculations were performed according to the developed simulation schemes, and the results obtained using on-site monitoring were verified.

[0106] The number of KR agitator grids generated in this step is between 800,000 and 1,000,000. The simulation software used is Fluent 2023R1. The calculation models used are the RNG k-ε turbulence model, the VOF two-phase flow model, the DPM discrete phase model and the unreacted nuclear chemical reaction model. The time step is 0.005 seconds, and each case is calculated to 1000 seconds. A total of 1,800 orthogonal simulation schemes were used in 90 models. The calculation results were verified using the molten iron temperature monitored on site and the S content of the molten iron in the stirring tank at different times. The molten iron temperature and the S content of the molten iron in the stirring tank at different times were collected from 90 groups of models with furnace ages of 1-10, 40-50, 80-90, 100-110, 140-150, 180-190, 200-210, 240-250, and 280-290, and compared with the monitoring values ​​and measured values ​​under the same process parameters on site. After comparative calculation, it was found that the relative error M1 of 88 of the 90 simulation results was below 5%, proving that the simulation results were accurate and reliable.

[0107] The KR production process, molten iron physical properties, KR agitator geometric model and simulation calculation results are used to construct a data set c. A deep learning algorithm is used to establish a multi-physics field model C, and model C is verified.

[0108] In this step, the KR production process, including agitator speed, agitator head immersion depth, desulfurizer composition and quality, and molten iron physical properties, including initial hot metal temperature and sulfur content, were used. KR agitator geometric models for 90 heats were also included. The geometric models for each heat were automatically generated using a preprocessor parameterization program. The 900 simulation results (temperature, flow, and solute fields) from step 4 were used to construct dataset c, which was divided into training and test sets at a ratio of 9:1. A deep learning algorithm was then used to establish a multi-physics prediction model C. This prediction model, by inputting the KR production process, molten iron physical properties, and agitator geometric model, can generate full-time predictions for the multi-physics fields associated with the model and production technology. By modifying the production process at a specific moment, the multi-physics prediction results for the modified process can be obtained. Model c was tested using 90 test data sets from dataset c. For 88 of these results, the relative error (C1) between the solute and temperature fields and the actual headroom monitoring values ​​was less than 5%, demonstrating the accuracy of the model.

[0109] The maximum load capacity of the on-site KR stirrer was obtained, and model A and pre-processing software were used to generate geometric models for 300 sets of data. Based on the prediction model C, 300 sets of KR stirrer geometric models from different furnaces were input. Each set of models was paired with 20 sets of production processes and 10 sets of different molten pool depths and molten iron physical properties, totaling 60,000 sets of input values. 60,000 sets of multi-physics field full-time prediction results were obtained, and the input and output data were constructed into a dataset d. A deep learning algorithm was used to establish a prediction model D for the production process, and model D was verified.

[0110] The maximum load capacity of the KR agitator in this step includes the maximum agitator speed, the depth of the agitator tank, and the temperature tolerance of the refractory material. The KR production process includes the agitator speed, the immersion depth of the agitator head, the composition and quality of the desulfurizer, and the production process. The physical properties of the molten iron include the initial molten iron temperature and sulfur content, as well as the geometric model of the KR agitator for the corresponding heat. The geometric models for different heats are automatically generated using a parameterized program. The simulation prediction results (temperature field, flow field, solute field) in step 7 include data exceeding the maximum load capacity of the equipment. A dataset d is constructed and divided into training and test sets in a ratio of 9:1. A multi-physics prediction model D is established using a deep learning algorithm. This prediction model can predict the production process that achieves the target physical field under the current geometric model by inputting the molten iron physical properties, the molten pool depth, the agitator geometric model, the physical field at the initial or time to be optimized, and the target physical field. Model D was validated using the test data from Dataset D. The model was tested using 6,000 sets of data from the test set. The resulting production (tuning) technical solutions from 5,973 of these results were fed back into Model C. The relative errors (D1) between the target physical fields (temperature, flow, and solute fields) were all less than 5%, demonstrating the accuracy of the model. The interfaces of the previously established Models A, B, C, and D, as well as the preprocessing programs, were unified to enable data transmission between them.

[0111] In a feasible implementation method, images of the 83rd furnace of the on-site KR stirrer before and after the metallurgical melt is loaded are obtained, and the images before the 83rd furnace is loaded with the metallurgical melt are transmitted to model A to output the real three-dimensional structure and geometric parameters of the current KR stirrer. The images after the molten iron is loaded are transmitted to model B to output the clearance height of the current molten pool. The geometric parameter results output by model A are transmitted to the prepared software to obtain the geometric mathematical model of the current KR stirrer. The obtained geometric model is automatically transmitted to model D in step 7. The temperature information of 1300°C and the initial S content information of 400ppm monitored by the on-site monitor are converted into temperature field and solute field and transmitted to model D. At the same time, the target temperature field of 1250°C or above, the solute field S content of 8ppm or less, the maximum load capacity of the equipment, the stirring tank depth of 4640mm, the maximum speed of the stirrer of 120r / min, and the physical property parameters of the loaded molten iron density of 6700 kg∙m are input into model D. −3 Viscosity 6.486×10 −3 kg∙m −1 ∙s −1. Model D outputs a production process that meets the maximum capacity of the equipment under this goal: the immersion depth of the stirring paddle is 1700mm, the stirring speed is 110r / min, the desulfurizer is 2370kg, and the stirring time is 700s. The technical solution output by model D, as well as the stirrer model, melt physical parameters and initial physical field information are automatically transferred to model C. Model C outputs the physical field results of the stirrer at all times under this solution. The results show that the melt temperature did not fall below 1250℃ within 700s, and the melt did not splash out of the stirring tank. Moreover, the sulfur content of the molten iron was 7ppm after 700 seconds, achieving the production goal. The technical solution output by model D was used in on-site production and achieved corresponding production results.

[0112] Images of the 137th furnace of the on-site KR agitator before and after molten iron was loaded were obtained. The images before molten iron was loaded were transferred to Model A to output the true three-dimensional structure and geometric parameters of the current reactor. The images after molten metal was loaded were transferred to Model B to output the headroom height of the molten metal. The geometric parameter results output by Model A were transferred to the prepared software to obtain the geometric mathematical model of the current KR agitator. The obtained geometric model was automatically transferred to Model D in step 7. The temperature information of 1300°C and the initial S content of 420 ppm detected by the on-site monitor were converted into temperature and solute fields and transferred to Model D. At the same time, the target temperature field of 1250°C or above, the solute field S content of 8 ppm or less, the maximum load capacity of the equipment, the stirring tank depth of 4640 mm, the maximum speed of the agitator of 120 r / min, and the physical property parameters of the loaded molten iron density of 6700 kg∙m were input into Model D. −3 Viscosity 6.486×10 −3 kg∙m −1 ∙s −1. Model D outputs the production process that meets the maximum capacity of the equipment under this goal: the stirring paddle immersion depth is 1700mm, the stirring speed is 115r / min, the desulfurizer is 2370kg, and the stirring time is 700s. The technical solution output by Model D, as well as the stirring model, melt physical parameters and initial physical field information are automatically transmitted to Model C. Model C outputs the physical field results of the stirrer at all times under this solution. The results show that the melt temperature did not fall below 1250℃ within 700s, but it shows that molten iron flew out of the stirring tank at 430s. The abnormal result at 430s is transmitted to Model D, and the target phase field distribution result is also transmitted. Model D outputs the corresponding optimized production process, adjusts the stirring paddle immersion depth to 2000mm, reduces the speed to 100r / min, and extends the stirring time to 720s. The output technical solution was automatically transferred to Model C. At 425 seconds, when no molten iron splashing occurred, the technical solution was adjusted to obtain the corresponding prediction results. The results showed that molten iron splashing occurred at subsequent moments. In addition, at 720 seconds, the sulfur content of the molten iron was 8ppm and the temperature was 1258℃, which met the generation requirements. This solution was used in on-site production and achieved production results consistent with the prediction results.

[0113] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0114] The present invention proposes an intelligent control method for metallurgical reactors based on real-time simulation and monitoring. It combines image recognition, machine learning and simulation methods, uses geometric image recognition technology and parametric automatic modeling and meshing technology to obtain a large amount of simulation data; utilizes machine learning algorithms to achieve real-time simulation and prediction of metallurgical reactors, and on this basis completes dynamic monitoring of multiple physical fields inside the metallurgical reactor; establishes a deep learning optimization prediction model to achieve intelligent control of metallurgical reactors. The present invention can achieve real-time simulation of multiple physical fields in metallurgical reactors and intelligent control of metallurgical reactors, solving the problems of difficulty in obtaining monitoring information of the physical fields in metallurgical reactors during production, low simulation efficiency and inability to obtain information in real time, and dependence on experience and difficulty in controlling metallurgical reactors. The present invention is an efficient and intelligent control method for metallurgical reactors based on real-time simulation and monitoring that can accurately reflect the physical field information inside the reactor.

[0115] Figure 2 This is a block diagram of an intelligent control device for a metallurgical reactor based on real-time simulation and monitoring according to an exemplary embodiment. The device is used in an intelligent control method for a metallurgical reactor based on real-time simulation and monitoring. Figure 2The device includes a training data acquisition module 210, a data set construction module 220, a first model training module 230, a topology solution construction module 240, a multi-physics field simulation module 250, a second model training module 260, a third model training module 270, and a metallurgical reactor control module 280. Among them:

[0116] The training data acquisition module 210 is used to monitor the metallurgical reactor during the smelting process through monitoring equipment. max Heat data collection, obtaining first image data of the metallurgical melt before it is charged into the metallurgical reactor, first geometric parameters of the metallurgical reactor before it is charged into the metallurgical reactor, second image data of the metallurgical melt after it is charged into the metallurgical reactor, second geometric parameters of the metallurgical reactor after it is charged into the metallurgical reactor, melt physical property parameters, first process parameters, and physical field monitoring data;

[0117] The data set construction module 220 is configured to use the first image data and N of the first geometric parameters a Heat data is used to construct a data set a; according to the second image data, the second geometric parameters and the melt physical parameters N a Heat data build data set b; N a The maximum number of smelting furnaces N in the current age of the metallurgical reactor max one-third;

[0118] The first model training module 230 is used to construct a model based on the data set a using the 3DGS algorithm to obtain an image parameter recognition model A; and to construct a model based on the data set b based on OpenCV to obtain a clearance height prediction model B;

[0119] A topology solution construction module 240 is configured to calculate, based on the image parameter recognition model A and the headroom prediction model B, the melt headroom, a topology solution for the reactor geometry, a reactor geometry model, and a structured mesh of the reactor geometry model according to the first image data and the second image data;

[0120] The multi-physics simulation module 250 is used to perform simulation using CFD software based on the reactor geometry model, the structured grid, and preset metallurgical industry specifications, according to the melt physical properties, the first process parameters, and the physical field monitoring data, to obtain multi-physics simulation data;

[0121] The second model training module 260 is used to construct N according to the multi-physics field simulation data, the reactor geometry model, the melt physical properties, the melt clearance height and the first process parameters. a The dataset c and N of the group data max Data set d of group data; using data set c, the first deep learning prediction model is trained to obtain a multi-physics field prediction model C;

[0122] The third model training module 270 is configured to input the data set d into the multi-physics field prediction model C for data prediction to obtain multi-physics field prediction data; construct a data set e based on the multi-physics field prediction data, the reactor geometry model, the melt physical properties, the melt headroom, and the first process parameters; and use the data set e to train the second deep learning prediction model to obtain a process parameter optimization model D.

[0123] The metallurgical reactor control module 280 is used to obtain, through monitoring equipment, the third image data of the current heat of metallurgical melt before it is loaded into the metallurgical reactor, the fourth image data after the metallurgical melt is loaded into the metallurgical reactor, the physical field data of the current heat, and the target physical field data; based on a preset physical field threshold range, according to the third image data of the current heat of metallurgical melt before it is loaded into the metallurgical reactor, the fourth image data after the metallurgical melt is loaded into the metallurgical reactor, the melt physical properties, the physical field data of the current heat, and the target physical field data, the reactor process parameters are controlled through the image parameter recognition model A, the clearance height prediction model B, the multi-physical field prediction model C, and the process parameter optimization model D to obtain the second process parameters; and the metallurgical reactor is controlled according to the second process parameters.

[0124] Among them, N max It is the maximum value of smelting heat N in the current age of the metallurgical reactor.

[0125] Among them, N a The maximum number of smelting furnaces N in the current age of the metallurgical reactor max One third of .

[0126] Optionally, the topology solution building module 240 is further configured to:

[0127] Inputting the first image data into the image parameter recognition model A to generate parameters and obtain third geometric parameters;

[0128] Inputting the second image data into the headroom height prediction model B to predict the melt headroom height, thereby obtaining the melt headroom height;

[0129] Constructing a topology scheme of the metallurgical reactor geometry model according to the third geometric parameter and the melt clearance height;

[0130] Based on Python and OpenFOAM, we wrote a mesh parameterization code for the metallurgical reactor geometry according to the topology scheme to obtain a pre-processing program.

[0131] The third geometric parameter is input into a pre-processing program to generate a mesh, thereby obtaining a reactor geometric model and a structured mesh of the reactor geometric model.

[0132] Optionally, the multi-physics simulation module 250 is further configured to:

[0133] Based on the preset metallurgical industry specifications, the operating range is determined according to the primary process parameters, melt physical properties and physical field monitoring data; a simulation plan is formulated according to the gradient based on the operating range;

[0134] Based on the simulation scheme, CFD software is used to perform simulation according to the reactor geometric model and structured grid to obtain multi-physics field simulation data.

[0135] Among them, the model structure of the first deep learning prediction model and the second deep learning prediction model is a convolutional neural network, a long short-term memory neural network or a bidirectional gated recurrent neural network.

[0136] Optionally, the metallurgical reactor control module 280 is further configured to:

[0137] Inputting the third image data into the image parameter recognition model A to obtain the current heat geometric parameters; performing data processing through a pre-processing program based on the current heat geometric parameters to obtain the current reactor geometric model;

[0138] Inputting the fourth image data into the headroom height prediction model B to obtain the headroom height of the melt of the current furnace;

[0139] Input the current reactor geometry model, the current furnace melt clearance height, the current furnace physical field data, the target physical field data and the melt physical property parameters into the process parameter optimization model D to obtain the predicted process parameters;

[0140] Input the current reactor geometry model, the current furnace melt clearance height, melt physical properties and predicted process parameters into the multi-physics prediction model C to obtain predicted multi-physics data;

[0141] Verify the predicted multi-physics field data according to the preset physics field threshold range to obtain the physics field verification result;

[0142] When the physical field verification result is passed, the predicted process parameter is determined as the second process parameter.

[0143] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0144] The present invention proposes an intelligent control method for metallurgical reactors based on real-time simulation and monitoring. It combines image recognition, machine learning and simulation methods, uses geometric image recognition technology and parametric automatic modeling and meshing technology to obtain a large amount of simulation data; utilizes machine learning algorithms to achieve real-time simulation and prediction of metallurgical reactors, and on this basis completes dynamic monitoring of multiple physical fields inside the metallurgical reactor; establishes a deep learning optimization prediction model to achieve intelligent control of metallurgical reactors. The present invention can achieve real-time simulation of multiple physical fields in metallurgical reactors and intelligent control of metallurgical reactors, solving the problems of difficulty in obtaining monitoring information of the physical fields in metallurgical reactors during production, low simulation efficiency and inability to obtain information in real time, and dependence on experience and difficulty in controlling metallurgical reactors. The present invention is an efficient and intelligent control method for metallurgical reactors based on real-time simulation and monitoring that can accurately reflect the physical field information inside the reactor.

[0145] Figure 3 This is a schematic diagram of the structure of an intelligent control device provided by an embodiment of the present invention. Figure 3 As shown, the intelligent control device may include the above Figure 2 The intelligent control device of the metallurgical reactor based on real-time simulation and monitoring is shown. Optionally, the intelligent control device 310 may include a first processor 2001.

[0146] Optionally, the intelligent control device 310 may further include a memory 2002 and a transceiver 2003 .

[0147] The first processor 2001, the memory 2002 and the transceiver 2003 may be connected via a communication bus.

[0148] The following combination Figure 3 The components of the intelligent control device 310 are described in detail:

[0149] The first processor 2001 is the control center of the intelligent control device 310 and can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs) or one or more field programmable gate arrays (FPGAs).

[0150] Optionally, the first processor 2001 can execute various functions of the intelligent control device 310 by running or executing a software program stored in the memory 2002 and calling data stored in the memory 2002.

[0151] In a specific implementation, as an embodiment, the first processor 2001 may include one or more CPUs, such as Figure 3 CPU0 and CPU1 are shown in FIG.

[0152] In a specific implementation, as an embodiment, the intelligent control device 310 may also include multiple processors, such as Figure 3 1 and 2. The first processor 2001 and the second processor 2004 are shown in FIG. Each of these processors can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). A processor herein can refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0153] The memory 2002 is used to store the software program for executing the solution of the present invention, and is controlled by the first processor 2001 for execution. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0154] Alternatively, the memory 2002 may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compact disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 may be integrated with the first processor 2001 or exist independently and accessed through the interface circuit ( Figure 3 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0155] The transceiver 2003 is used to communicate with a network device or a terminal device.

[0156] Optionally, the transceiver 2003 may include a receiver and a transmitter ( Figure 3 The receiver is used to implement a receiving function, and the transmitter is used to implement a sending function.

[0157] Optionally, the transceiver 2003 may be integrated with the first processor 2001, or may exist independently and communicate with the first processor 2001 through the interface circuit ( Figure 3 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0158] It should be noted that Figure 3 The structure of the intelligent control device 310 shown in the figure does not constitute a limitation on the router. The actual knowledge structure recognition device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0159] In addition, the technical effects of the intelligent control device 310 can refer to the technical effects of the intelligent control method of the metallurgical reactor based on real-time simulation and monitoring described in the above method embodiment, and will not be repeated here.

[0160] It should be understood that the first processor 2001 in the embodiment of the present invention may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, etc.

[0161] It should also be understood that the memory in the embodiments of the present invention may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory may be random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0162] The above embodiments can be implemented in whole or in part via software, hardware (e.g., circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in accordance with the embodiments of the present invention are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0163] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0164] In this disclosure, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, "at least one of a, b, or c" can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or plural.

[0165] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0166] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0167] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0168] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interface, indirect coupling or communication connection of the device or unit, which can be electrical, mechanical or other forms.

[0169] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0170] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0171] If the functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical disks.

[0172] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An intelligent control method for metallurgical reactors based on real-time simulation and monitoring, characterized in that: The method comprises: During the smelting process, the metallurgical reactor is monitored by N max Heat data collection, obtaining first image data of the metallurgical melt before it is charged into the metallurgical reactor, first geometric parameters of the metallurgical reactor before it is charged into the metallurgical reactor, second image data of the metallurgical melt after it is charged into the metallurgical reactor, second geometric parameters of the metallurgical reactor after it is charged into the metallurgical reactor, melt physical property parameters, first process parameters, and physical field monitoring data; Using the first image data and N of the first geometric parameters a Heat data to build a data set a; according to the second image data, the second geometric parameters and the melt physical parameters N a The heat data constructs dataset b; Based on the data set a, a model is constructed using the 3DGS algorithm to obtain an image parameter recognition model A; based on the data set b, a model is constructed based on OpenCV to obtain a clearance height prediction model B; Based on the image parameter recognition model A and the headroom prediction model B, calculations are performed according to the first image data and the second image data to obtain a melt headroom height, a topological scheme of a reactor geometric structure, a reactor geometric model, and a structured grid of the reactor geometric model; Based on the reactor geometric model, the structured grid, and preset metallurgical industry specifications, and according to the melt physical property parameters, the first process parameters, and the physical field monitoring data, CFD software is used to perform simulation to obtain multi-physical field simulation data; N is constructed according to the multi-physics field simulation data, the reactor geometry model, the melt physical property parameters, the melt clearance height and the first process parameters. a The dataset c and N of the group data max A data set d of group data; using the data set c, training a first deep learning prediction model to obtain a multi-physics field prediction model C; Inputting the data set d into the multi-physics field prediction model C for data prediction to obtain multi-physics field prediction data; constructing a data set e based on the multi-physics field prediction data, the reactor geometry model, the melt physical properties, the melt clearance height, and the first process parameters; using the data set e, training a second deep learning prediction model to obtain a process parameter optimization model D; Through the monitoring equipment, the third image data of the current heat of metallurgical melt before being charged into the metallurgical reactor, the fourth image data after the metallurgical melt is charged into the metallurgical reactor, the physical field data of the current heat and the target physical field data are obtained; based on the preset physical field threshold range, according to the third image data of the current heat of metallurgical melt before being charged into the metallurgical reactor, the fourth image data after the metallurgical melt is charged into the metallurgical reactor, the melt physical properties, the physical field data of the current heat and the target physical field data, the reactor process parameters are controlled through the image parameter recognition model A, the clearance height prediction model B, the multi-physical field prediction model C and the process parameter optimization model D to obtain the second process parameters; the metallurgical reactor is controlled according to the second process parameters.

2. The intelligent control method of a metallurgical reactor based on real-time simulation and monitoring according to claim 1, characterized in that: The N max It is the maximum value of smelting heat N in the current age of the metallurgical reactor.

3. The intelligent control method of a metallurgical reactor based on real-time simulation and monitoring according to claim 1, characterized in that: The N a The maximum number of smelting furnaces N in the current age of the metallurgical reactor max One third of .

4. The intelligent control method of a metallurgical reactor based on real-time simulation and monitoring according to claim 1, characterized in that: The method of calculating based on the image parameter recognition model A and the headroom height prediction model B according to the first image data and the second image data to obtain the melt headroom height, the topological scheme of the reactor geometric structure, the reactor geometric model, and the structured grid of the reactor geometric model includes: Inputting the first image data into the image parameter recognition model A to generate parameters to obtain third geometric parameters; Inputting the second image data into the headroom height prediction model B to predict the melt headroom height to obtain the melt headroom height; Constructing a metallurgical reactor geometric model and a topology scheme of a structured grid of the reactor geometric model according to the third geometric parameter and the melt clearance height; Based on Python and OpenFOAM, a mesh parameterization code was written for the metallurgical reactor geometry according to the topology scheme to obtain a pre-processing program; The third geometric parameter is input into the pre-processing program to generate a grid, thereby obtaining a reactor geometric model and a structured grid of the reactor geometric model.

5. The intelligent control method of a metallurgical reactor based on real-time simulation and monitoring according to claim 1, characterized in that: The method includes performing simulation using CFD software based on the reactor geometry model, the structured grid, and preset metallurgical industry specifications, according to the melt physical property parameters, the first process parameters, and the physical field monitoring data, to obtain multi-physical field simulation data, including: Based on preset metallurgical industry specifications, determining an operating condition range according to the first process parameters, the melt physical property parameters, and the physical field monitoring data; formulating a simulation plan according to a gradient based on the operating condition range; Based on the simulation scheme, according to the reactor geometric model and the structured grid, CFD software is used to perform simulation to obtain multi-physics field simulation data.

6. The intelligent control method of a metallurgical reactor based on real-time simulation and monitoring according to claim 1, characterized in that: The model structures of the first deep learning prediction model and the second deep learning prediction model include convolutional neural networks, long short-term memory neural networks, or bidirectional gated recurrent neural networks.

7. The intelligent control method of a metallurgical reactor based on real-time simulation and monitoring according to claim 4, characterized in that: The method includes: based on a preset physical field threshold range, according to the third image data of the current heat of metallurgical melt before being charged into the metallurgical reactor, the fourth image data of the metallurgical melt after being charged into the metallurgical reactor, the melt physical property parameters, the current heat physical field data, and the target physical field data, performing reactor process parameter regulation through the image parameter recognition model A, the clearance height prediction model B, the multi-physical field prediction model C, and the process parameter optimization model D to obtain the second process parameter, including: Inputting the third image data into the image parameter recognition model A to obtain the current heat geometric parameters; performing data processing through the pre-processing program according to the current heat geometric parameters to obtain the current reactor geometric model; Inputting the fourth image data into the headroom height prediction model B to obtain the headroom height of the melt of the current furnace; Inputting the current reactor geometry model, the current heat melt clearance height, the current heat physical field data, the target physical field data and the melt physical property parameters into the process parameter optimization model D to obtain predicted process parameters; Inputting the current reactor geometry model, the current heat melt clearance height, the melt physical property parameters and the predicted process parameters into the multi-physics field prediction model C to obtain predicted multi-physics field data; Verifying the predicted multi-physical field data according to a preset physical field threshold range to obtain a physical field verification result; When the physical field verification result is passed, the predicted process parameter is determined as the second process parameter.

8. An intelligent control device for a metallurgical reactor based on real-time simulation and monitoring, wherein the intelligent control device for a metallurgical reactor based on real-time simulation and monitoring is used to implement the intelligent control method for a metallurgical reactor based on real-time simulation and monitoring according to any one of claims 1 to 7, characterized in that: The device comprises: The training data acquisition module is used to monitor the metallurgical reactor during the smelting process through monitoring equipment. max Heat data collection, obtaining first image data of the metallurgical melt before it is charged into the metallurgical reactor, first geometric parameters of the metallurgical reactor before it is charged into the metallurgical reactor, second image data of the metallurgical melt after it is charged into the metallurgical reactor, second geometric parameters of the metallurgical reactor after it is charged into the metallurgical reactor, melt physical property parameters, first process parameters, and physical field monitoring data; A data set construction module is configured to use the first image data and N of the first geometric parameters to construct a a Heat data to build a data set a; according to the second image data, the second geometric parameters and the melt physical parameters N a Heat data constructs data set b; the N a The maximum number of smelting furnaces N in the current age of the metallurgical reactor max one-third; The first model training module is used to construct a model based on the data set a using the 3DGS algorithm to obtain an image parameter recognition model A; and to construct a model based on the data set b based on OpenCV to obtain a clearance height prediction model B; a topology solution construction module, configured to calculate, based on the image parameter recognition model A and the headroom prediction model B, the melt headroom, a topology solution of the reactor geometry, a reactor geometry model, and a structured grid of the reactor geometry model according to the first image data and the second image data; a multi-physics field simulation module, configured to perform simulation using CFD software based on the reactor geometry model, the structured grid, and preset metallurgical industry specifications, according to the melt physical property parameters, the first process parameters, and the physical field monitoring data, to obtain multi-physics field simulation data; The second model training module is used to construct N according to the multi-physics field simulation data, the reactor geometry model, the melt physical property parameters, the melt clearance height and the first process parameters. a The dataset c and N of the group data max Group data set d; using the data set c, the first deep learning prediction model is trained to obtain a multi-physics field prediction model C; a third model training module, configured to input the data set d into the multi-physics field prediction model C for data prediction to obtain multi-physics field prediction data; construct a data set e based on the multi-physics field prediction data, the reactor geometry model, the melt physical properties, the melt clearance height, and the first process parameters; and use the data set e to train a second deep learning prediction model to obtain a process parameter optimization model D; A metallurgical reactor control module is used to obtain, through monitoring equipment, third image data of the current heat of metallurgical melt before it is loaded into the metallurgical reactor, fourth image data after the metallurgical melt is loaded into the metallurgical reactor, physical field data of the current heat, and target physical field data; based on a preset physical field threshold range, according to the third image data of the current heat of metallurgical melt before it is loaded into the metallurgical reactor, fourth image data after the metallurgical melt is loaded into the metallurgical reactor, the melt physical properties, physical field data of the current heat, and target physical field data, the reactor process parameters are controlled through the image parameter recognition model A, the clearance height prediction model B, the multi-physical field prediction model C, and the process parameter optimization model D to obtain a second process parameter; and the metallurgical reactor is controlled according to the second process parameter.

9. An intelligent control device, characterized in that: The intelligent control device includes: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program code, which can be called by a processor to execute the method according to any one of claims 1 to 7.

Images