Intelligent regulation and control system and method for stabilizing top-blown smelting melt

Abstract

The invention relates to the technical field of metal smelting, and discloses an intelligent regulation and control system and method for top-blown smelting melt stabilization, and the method comprises the steps: classifying, filtering and integrating real-time molten pool data; carrying out characteristic statistics on the integrated real-time molten pool data, and dynamically adjusting an error; simulating dynamic behaviors in the molten pool; historical molten pool data are called, and a molten pool behavior prediction model is constructed; extracting spatial features from the real-time molten pool data, capturing time dependence of molten pool liquid level fluctuation, and obtaining a liquid level fluctuation prediction result; generating an optimal spray gun operation parameter based on the simulation result and the prediction result; and a regulation and control instruction is generated according to the optimal spray gun operation parameters, real-time spray gun data and regulation and control feedback are displayed by visual equipment, and switching between a manual control mode and an automatic control mode is provided. The system comprises a molten pool simulation module, a molten pool behavior prediction module and a spray gun adjustment module. According to the invention, closed-loop control based on real-time data is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of metal smelting, and particularly to a method and system for intelligent regulation of the stability of top-blown smelting melt. Background Art

[0002] In the metallurgical industry, the top-blown smelting technology, as one of the core smelting processes for high efficiency and energy conservation, is widely used in the smelting of metals such as copper, tin, and nickel. However, in actual production, the top-blown smelting process is often troubled by problems such as severe liquid surface fluctuations, frequent melt splashing, and uneven temperature fields. This not only affects the smelting efficiency and product quality but also poses severe challenges to the safety of equipment and production costs. Traditional top-blown smelting control mainly relies on empirical regulation and is difficult to effectively cope with the complex dynamic behaviors of the molten bath, such as fluid turbulence, gas-liquid interaction, and rapidly changing working conditions in high-temperature environments. With the development of industrial demands towards high efficiency, greenness, and intelligence, the existing technologies can no longer meet the requirements of precise control and dynamic adaptation in modern smelting.

[0003] Prior Art One, a Chinese patent with the patent number 202410139965.2 provides a method and system for simulating the flow field of a steelmaking converter molten bath under the action of a carbon-oxygen reaction, which relates to the field of iron and steel metallurgy. It includes: obtaining the structural parameters of the converter and constructing a three-dimensional geometric model of the converter; performing mesh division on the geometric model; setting the basic assumptions of the model, selecting the calculation model, setting the physical property parameters, boundary conditions, and solution algorithm, and performing initialization operations and iterative calculations to obtain the flow field of the molten bath under combined blowing conditions; obtaining the converter flue gas information, solving the carbon-oxygen reaction ratio of the molten bath, and calculating the amount of carbon-oxygen reaction bubbles in the liquid phase part of the molten bath; constructing the carbon-oxygen reaction bubble generation points, injecting carbon-oxygen reaction bubbles, and compiling the behavior of carbon-oxygen reaction bubbles at the slag-metal interface; coupling the calculations to obtain the simulation results of the steelmaking converter molten bath under the action of the carbon-oxygen reaction, and post-processing to obtain the flow field diagrams of the molten bath under the action of the carbon-oxygen reaction at different blowing times. Although it can analyze the flow field of the molten bath under the action of the carbon-oxygen reaction at different smelting times, it does not optimize the design and layout of the lance, resulting in poor gas-liquid mixing efficiency.

[0004] Prior Art Two, a Chinese patent with the patent number 202411724755.6 provides an on-line monitoring system and method for the molten pool surface liquid level. The system includes a data acquisition module, a data preprocessing module, a model training and optimization module, an angle detection and compensation module, a non-linear data processing module, a system testing and optimization module, and also includes a display module, a communication module and a control module. It can realize the on-line real-time monitoring of the molten pool surface liquid level and improve the production automation level. When the data acquisition module reaches the detection position, the control system triggers a signal through the coil to turn on the camera to capture the images of the liquid levels of the slag surface and the copper solution surface of the current molten pool, realizing the on-line real-time monitoring function of the molten pool surface liquid level. Compared with the traditional manual measurement or intermittent monitoring methods, although it can greatly improve the detection efficiency, reduce the dependence on manual operation and enhance the automation level of the production process, it does not solve the problem of large fluctuations in the molten pool liquid surface, resulting in splashing and the generation of slag materials.

[0005] Prior Art Three, a Chinese patent with the patent number 202410143851.5 discloses an intermediate ladle for vacuum ingot casting and its liquid level stability control method, which relates to the technical field of intermediate ladle ingot casting. When the molten metal in the intermediate ladle flows out of the intermediate ladle through the water outlet, the height difference and pressure difference between the current molten metal liquid surface and the water outlet are obtained. According to the height difference and pressure difference at the current moment, the outlet molten metal flow rate of the water outlet at the current moment is calculated, and the inlet molten metal flow rate of the inlet at the next moment is adjusted according to the outlet molten metal flow rate, so that the liquid level height of the molten metal in the intermediate ladle is maintained at a set height. Although it can calculate the molten metal flow rate at the water outlet of the intermediate ladle, by matching the molten metal flow rates at the inlet and the outlet, and then relying on the eddy current controller at the water outlet to reduce the turbulence degree inside the molten metal, the liquid level fluctuation of the molten metal can be slowed down, and the stability of the molten metal liquid surface can be maintained, so as to ensure uniform filling of the casting during the vacuum casting process. However, the design and layout of the spray gun are not optimized, resulting in poor gas-liquid mixing efficiency.

[0006] Currently, Prior Art One, Prior Art Two and Prior Art Three have problems that the design and layout of the spray gun are not optimized and the large fluctuations in the molten pool liquid surface are not solved, resulting in poor gas-liquid mixing efficiency. To solve the above problems, the present invention provides a top-blown smelting melt stable intelligent control method and system. Summary of the Invention

[0007] The main purpose of the present invention is to provide a top-blown smelting melt stable intelligent control method and system to solve the problems in the prior art that the design and layout of the spray gun are not optimized and the large fluctuations in the molten pool liquid surface are not solved, resulting in poor gas-liquid mixing efficiency.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for intelligent regulation and control of the stability of the top-blown smelting melt, the method for intelligent regulation and control of the stability of the top-blown smelting melt includes:

[0010] Collect real-time bath data of the lance inlet, bath liquid level and internal melt, classify, filter and integrate the real-time bath data; conduct characteristic statistics on the integrated real-time bath data, and dynamically adjust the error; simulate the dynamic behavior inside the bath;

[0011] Retrieve historical bath data and construct a bath behavior prediction model; preprocess the historical bath data and divide it into a training set, a validation set and a test set; use the training set to train the bath behavior prediction model; extract spatial features from the real-time bath data, capture the time dependence of the bath liquid level fluctuation, and obtain the liquid level fluctuation prediction result; generate the optimal lance operation parameters based on the simulation result and the prediction result;

[0012] Generate a regulation instruction according to the optimal lance operation parameters to obtain the adjusted lance parameters; store the adjusted lance parameters; visually display the adjusted lance parameters and the bath monitoring data, and the visualization device displays the real-time lance data and the regulation feedback, providing a manual and automatic control mode switch.

[0013] As a further improvement of the present invention, the process of simulating the dynamic behavior inside the bath includes the following steps:

[0014] Use a flow meter to collect and record the flow rate of the blown gas in real time; use a pressure sensor to monitor the gas pressure at the lance inlet; use a temperature sensor to measure the gas temperature; use a gas concentration analyzer to measure the composition ratio of the mixed gas;

[0015] Dynamically monitor the key characteristics of the bath liquid level and the internal melt, and evaluate the fluctuations during the smelting process; extract the melt sample inside the bath in real time, detect the extracted melt sample inside the bath, and analyze the physical and chemical properties of the melt sample;

[0016] Process the collected real-time data through a filtering algorithm to eliminate environmental interference, calculate the temperature field, flow field, liquid level fluctuation of the bath and the interaction with the lance gas, and simulate the dynamic behavior inside the bath.

[0017] As a further improvement of the present invention, the process of analyzing the physical and chemical properties of the melt sample includes the following steps:

[0018] Place a laser displacement sensor on the top of the bath for non-contact liquid level height monitoring; use a reflector to guide the laser beam into the measurement area; measure the liquid level height and fluctuation amplitude of the bath in real time;

[0019] Fix a high-frequency vibration sensor on the peripheral structure of the molten pool for dynamic monitoring of liquid surface fluctuations; capture liquid surface fluctuations through solid transmission; make mechanical contact with the molten pool wall through a heat-insulating sleeve; monitor the frequency and amplitude of liquid surface fluctuations;

[0020] Install ultrasonic sensors in a multi-point array at the top of the molten pool for real-time measurement of the molten pool liquid surface height and its fluctuation amplitude; equip with a heat-insulating protective cover to isolate from the molten pool environment; adopt a directional beam design to monitor the intensity and frequency of splashing melt and obtain real-time feedback data for lance adjustment.

[0021] As a further improvement of the present invention, the process of simulating the dynamic behavior inside the molten pool includes the following steps:

[0022] Process the collected real-time data through a filtering algorithm to eliminate environmental interference, monitor and exclude abnormal data points caused by sensor failures or extreme conditions, and smooth the data;

[0023] Utilize the statistical characteristics of the input real-time data to dynamically adjust the coefficients of the filter to minimize the estimation error;

[0024] Calculate the temperature field, flow field, and liquid surface fluctuations of the molten pool, and simulate the dynamic behavior inside the molten pool based on the interaction of the lance gas.

[0025] As a further improvement of the present invention, the process based on the interaction of the lance gas includes the following steps:

[0026] Use a turbulence model to simulate the hydrodynamic behavior in the molten pool; wherein, the hydrodynamic behavior includes flow velocity, pressure, and turbulence characteristics; during the calculation, the gas phase and slag phase in the TSL furnace can coexist in any calculation unit; the volume fractions of the two phases vary between 0 and 1, and the sum of the volume fractions of the two phases is always 1;

[0027] Capture the numerical technique of the gas-liquid interface to simulate the liquid surface fluctuations in the molten pool and gas-liquid interaction, and introduce a volume fraction function to represent the proportion of the fluid in the grid unit;

[0028] Carry out simulation calculations on the interfacial tension to obtain the shape of the liquid surface fluctuations in the molten pool and the impact of the lance gas flow on the liquid surface; and simulate the evolution of the temperature field inside the molten pool, and calculate the effects of convection, heat conduction, and volume heat sources.

[0029] As a further improvement of the present invention, the process of obtaining the liquid surface fluctuation prediction result includes the following steps:

[0030] Retrieve historical molten pool data, and perform preprocessing operations such as cleaning, normalization, and discretization on the historical molten pool data; divide the dataset of the historical molten pool data into a training set, a validation set, and a test set; construct a molten pool behavior prediction model based on a convolutional neural network and a recurrent neural network;

[0031] Train the molten pool behavior prediction model using the training set, perform forward propagation of the input data through the network, and calculate the output result; calculate the loss value between the prediction result and the true label; calculate the gradient of each parameter, and update the weights and biases; iterate sequentially until the preset number of training rounds is reached or the stopping condition is satisfied;

[0032] Evaluate the performance of the molten pool behavior prediction model using the validation set and adjust the parameters; evaluate the generalization ability of the molten pool behavior prediction model through the test set, and adjust the molten pool behavior prediction model according to the evaluation results to obtain the final molten pool behavior prediction model; input the real-time data into the molten pool behavior prediction model for prediction to obtain the prediction result.

[0033] As a further improvement of the present invention, the process of obtaining the prediction result includes the following steps:

[0034] Input the real-time data into the molten pool behavior prediction model, extract the spatial features of the real-time data, the input layer receives the real-time data matrix, and each row represents the time series of a sensor; the convolutional layer extracts local features through convolutional operations; generate a feature map with the characteristics of dimensionality reduction;

[0035] Capture the time dependence of the molten pool liquid level fluctuation, the input layer receives the extracted time feature sequence, and the hidden layer calculates the new hidden state based on the current input and the hidden state at the previous moment; the input layer predicts the future trend value of the liquid level fluctuation;

[0036] Generate the optimal adjustment suggestion for the lance parameters according to the future trend value of the liquid level fluctuation and the simulation result, transmit the optimal adjustment suggestion to the control center, generate a control instruction according to the optimal adjustment suggestion, and perform lance adjustment.

[0037] As a further improvement of the present invention, the process of generating the optimal adjustment suggestion for the lance parameters includes the following steps:

[0038] Randomly generate an initial population, and each particle represents a set of lance parameters; the position of each particle represents a set of parameters, and the velocity represents the moving direction and velocity of the particle in the solution space; based on the simulation and test results, evaluate the molten pool liquid level fluctuation and splashing degree under the current parameter setting;

[0039] Find the parameter combination that achieves the minimum liquid level fluctuation and reduced splashing; calculate the fitness value of each particle, and update the individual optimal position and the global optimal position; if the fitness value of the current particle is better than the historical optimal value, update the individual optimal position; if it is better than the global optimal value, update the global optimal position;

[0040] Iterate sequentially until the maximum number of iterations is reached or the fitness value converges to a preset threshold; output the combination of spray gun parameters corresponding to the global optimal position as the optimal adjustment suggestion.

[0041] As a further improvement of the present invention, the process of visually displaying the adjusted spray gun parameters and the molten pool monitoring data includes the following steps:

[0042] Generate a control instruction, record the adjusted spray gun parameters in the database, including parameter values before and after adjustment, adjustment time, and molten pool status information, etc.; perform regular backups of the recorded spray gun parameter data;

[0043] Visually display the adjusted spray gun parameters and the molten pool monitoring data in real time through a visualization device, and provide regulation feedback information on the visualization interface; including the current molten pool status, parameter adjustment effect, and liquid level fluctuation prediction result;

[0044] Provide a switching function between manual and automatic control modes; manually adjust the spray gun parameters in the manual mode; automatically adjust according to the molten pool behavior prediction model, simulation results, and the formed optimal spray gun operation parameters in the automatic mode; monitor the real-time data and compare it with the prediction result, and adjust the molten pool behavior prediction model according to the comparison result.

[0045] To achieve the above object, the present invention also provides the following technical solutions:

[0046] A top-blown smelting melt stable intelligent regulation system, which is applied to the top-blown smelting melt stable intelligent regulation method, and the top-blown smelting melt stable intelligent regulation system includes:

[0047] A molten pool simulation module, which is used to collect real-time molten pool data of the spray gun inlet, molten pool liquid level, and internal melt, classify, filter, and integrate the real-time molten pool data; perform characteristic statistics on the integrated real-time molten pool data, dynamically adjust the error; simulate the dynamic behavior inside the molten pool;

[0048] A molten pool behavior prediction module, which is used to retrieve historical molten pool data, construct a molten pool behavior prediction model; preprocess the historical molten pool data and divide it into a training set, a validation set, and a test set; use the training set to train the molten pool behavior prediction model; extract spatial features from the real-time molten pool data, capture the time dependence of the molten pool liquid level fluctuation, and obtain the liquid level fluctuation prediction result; generate optimal spray gun operation parameters based on the simulation result and the prediction result;

[0049] A spray gun adjustment module, which is used to generate a regulation instruction according to the optimal spray gun operation parameters to obtain the adjusted spray gun parameters; store the adjusted spray gun parameters; visually display the adjusted spray gun parameters and the molten pool monitoring data, and the visualization device displays the real-time spray gun data and regulation feedback, and provides a switching between manual and automatic control modes.

[0050] The present invention receives real-time data (parameters such as gas velocity, pressure, temperature, depth, liquid level fluctuation conditions, etc.), classifies, filters, and integrates it. While providing boundary condition parameters for simulation, it monitors the smelting situation in real time. According to the prediction results of the numerical calculation module, it generates control commands and sends them to the lance module to dynamically adjust the gas flow rate and pressure of the lance, realizing closed-loop control based on real-time data. It saves all key data of the system operation, including historical monitoring data, control commands, and feedback results, which can be used as training data for the artificial intelligence optimization model. It provides a function to query historical molten pool data for long-term trend analysis. Thus, it ensures real-time data interaction between modules. This unit supports multi-protocol compatibility (such as CAN, Ethernet, etc.), has strong compatibility, and adapts to smelting equipment of different scales and types. It provides an intuitive control panel to display real-time data and control feedback, and provides a switch between manual and automatic control modes, facilitating operators to monitor the system operation status and adjust control parameters. Description of the Drawings

[0051] Figure 1 It is a schematic diagram of the step flow of an embodiment of the method for stable intelligent control of the top-blown smelting melt of the present invention;

[0052] Figure 2 It is a schematic diagram of the step flow of an embodiment of simulating the internal dynamic behavior of the molten pool in the method for stable intelligent control of the top-blown smelting melt of the present invention;

[0053] Figure 3 It is a schematic diagram of the step flow of an embodiment of analyzing the physical and chemical properties of the melt sample in the method for stable intelligent control of the top-blown smelting melt of the present invention;

[0054] Figure 4 It is a schematic diagram of the principle of the numerical calculation operation process in the method for stable intelligent control of the top-blown smelting melt of the present invention;

[0055] Figure 5 It is a schematic diagram of the step flow of an embodiment of simulating the internal dynamic behavior of the molten pool in the method for stable intelligent control of the top-blown smelting melt of the present invention;

[0056] Figure 6 It is a schematic diagram of the step flow of an embodiment of calculating the temperature field, flow field, and liquid level fluctuation of the molten pool and the interaction based on the lance gas in the method for stable intelligent control of the top-blown smelting melt of the present invention;

[0057] Figure 7 It is a schematic diagram of the step flow of an embodiment of obtaining the liquid level fluctuation prediction result in the method for stable intelligent control of the top-blown smelting melt of the present invention;

[0058] Figure 8Schematic diagram of the steps for obtaining the prediction results in an embodiment of the top-blown smelting melt stability intelligent regulation method of the present invention;

[0059] Figure 9 Schematic diagram of the prediction process principle of the bath behavior prediction model in an embodiment of the top-blown smelting melt stability intelligent regulation method of the present invention;

[0060] Figure 10 Schematic diagram of the steps for generating the optimal adjustment suggestions for the lance parameters in an embodiment of the top-blown smelting melt stability intelligent regulation method of the present invention;

[0061] Figure 11 Schematic diagram of the steps for visually displaying the adjusted lance parameters and bath monitoring data in an embodiment of the top-blown smelting melt stability intelligent regulation method of the present invention;

[0062] Figure 12 Schematic diagram of the functional modules in an embodiment of the top-blown smelting melt stability intelligent regulation system of the present invention;

[0063] Figure 13 Schematic diagram of the structure of an embodiment of the electronic device of the present invention;

[0064] Figure 14 Schematic diagram of the structure of an embodiment of the storage medium of the present invention. Detailed implementation manners

[0065] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0066] The terms "first", "second", and "third" in the present invention are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first", "second", and "third" may explicitly or implicitly include at least one of such features. In the description of the present invention, the meaning of "a plurality" is at least two, such as two, three, etc., unless otherwise specifically defined. In the embodiments of the present invention, all directional indications (such as up, down, left, right, front, back...) are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the drawings). If the specific posture changes, the directional indications will also change accordingly. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products, or devices.

[0067] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in connection with the embodiments can be included in at least one embodiment of the present invention. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0068] As Figure 1 shown, this embodiment provides an embodiment of the top-blown smelting melt stability intelligent control method. In this embodiment, the top-blown smelting melt stability intelligent control method specifically includes the following steps:

[0069] Step S1: Collect real-time bath data of the lance inlet, bath surface, and internal melt, classify, filter, and integrate the real-time bath data; perform characteristic statistics on the integrated real-time bath data, dynamically adjust the error; simulate the dynamic behavior inside the bath;

[0070] Step S2: Retrieve historical bath data and construct a bath behavior prediction model; preprocess the historical bath data and divide it into a training set, a validation set, and a test set; use the training set to train the bath behavior prediction model; extract spatial features from the real-time bath data, capture the time dependence of the bath surface fluctuation, and obtain the bath surface fluctuation prediction result; generate optimal lance operation parameters based on the simulation result and the prediction result;

[0071] Step S3: Generate a regulation instruction according to the optimal spray gun operation parameters, and perform adjusted spray gun parameters; store the adjusted spray gun parameters; visually display the adjusted spray gun parameters and the molten pool monitoring data, and the visualization device displays real-time spray gun data and regulation feedback, providing a manual and automatic control mode switch.

[0072] Preferably, in this embodiment, real-time data (parameters such as gas velocity, pressure, temperature, depth, liquid level fluctuation, etc.) is received, classified, filtered, and integrated to provide boundary condition parameters for the simulation while monitoring the smelting situation in real time. According to the prediction results of the numerical calculation module, a regulation instruction is generated and sent to the spray gun module to dynamically adjust the gas flow rate and pressure of the spray gun, realizing closed-loop control based on real-time data. Save all key data of the system operation, including historical monitoring data, regulation instructions, and feedback results. It can be used as training data for the artificial intelligence optimization model. Provide a historical molten pool data query function for long-term trend analysis. Thus, ensure real-time data interaction between modules. This unit supports multi-protocol compatibility (such as CAN, Ethernet, etc.), has strong compatibility, and adapts to smelting equipment of different scales and types. Provide an intuitive control panel to display real-time data and regulation feedback, providing a manual and automatic control mode switch, facilitating operators to monitor the system operation status and adjust control parameters.

[0073] Furthermore, as Figure 2 shown, the process of simulating the dynamic behavior inside the molten pool in step S1 specifically includes the following steps:

[0074] Step S11: Use a flow meter to collect and record the blowing gas flow rate in real time; use a pressure sensor to monitor the gas pressure at the spray gun inlet; use a temperature sensor to measure the gas temperature; use a gas concentration analyzer to measure the composition ratio of the mixed gas;

[0075] Step S12: Dynamically monitor the key features of the molten pool liquid level and the internal melt, evaluate the fluctuation situation during the smelting process; extract the melt sample in the molten pool in real time, detect the extracted melt sample in the molten pool, and analyze the physical and chemical properties of the melt sample;

[0076] Step S13: Process the collected real-time data through a filtering algorithm to eliminate environmental interference, calculate the temperature field, flow field, liquid level fluctuation of the molten pool, and the interaction with the spray gun gas, and simulate the dynamic behavior inside the molten pool.

[0077] Preferably, in step S11 of this embodiment, a flowmeter is used to monitor the flow rate of the gas blown by the spray gun in real time to ensure the stability and accuracy of gas supply; a pressure sensor is used to monitor the pressure of the gas at the inlet of the spray gun in real time to ensure that the gas pressure is within the process requirements; a temperature sensor is used to measure the temperature of the gas to ensure that the gas temperature meets the process requirements; a gas concentration analyzer is used to monitor the composition ratio of the mixed gas in real time to ensure the accuracy and consistency of the gas components; by monitoring and adjusting the gas flow rate, pressure, temperature and composition ratio in real time, the control accuracy of the smelting process can be significantly improved, and the process deviation can be reduced; ensuring the stability and accuracy of gas supply helps to improve the smelting efficiency and reduce energy waste; monitoring the gas parameters in real time can timely detect potential safety hazards and avoid accidents. In step S12, through devices such as sensors and cameras, the changes in the molten pool liquid level and the key characteristics of the internal melt, such as temperature, flow rate, etc., are monitored in real time; the melt samples in the molten pool are extracted in real time for physical and chemical property analysis; the extracted melt samples are subjected to detailed physical and chemical property analysis by laboratory analysis equipment; dynamically monitoring the key characteristics of the molten pool liquid level and the internal melt can evaluate the fluctuations in the smelting process in real time and timely adjust the process parameters; by analyzing the physical and chemical properties of the melt samples, the material changes in the smelting process can be accurately grasped and the smelting energy efficiency can be optimized; extracting and analyzing the melt samples in real time can quickly discover problems and make adjustments to improve production efficiency. In step S13, the real-time data collected is processed by a filtering algorithm to eliminate environmental interference and improve the accuracy and reliability of the data; a numerical simulation method is used to calculate the temperature field, flow field, liquid level fluctuation of the molten pool and its interaction with the spray gun gas; through numerical simulation, the dynamic behavior inside the molten pool is simulated to provide a theoretical basis for process optimization; through numerical simulation, the dynamic behavior inside the molten pool can be deeply understood to provide a scientific basis for process optimization; numerical simulation can replace some experiments, reducing the experimental cost and time; guiding the actual production through the simulation results can improve the stability and reliability of the production process. The flowmeter - records the flow rate of the blown gas in real time, providing dynamic boundary conditions for CFD numerical simulation; the pressure sensor - monitors the gas pressure at the inlet of the spray gun; the temperature sensor - measures the gas temperature; the gas concentration analyzer - measures the component ratio of the gas mixture; multi-source data in the smelting process is collected in real time, providing key boundary conditions and operating conditions information for numerical calculation and feedback control. Through high-precision sensing devices and intelligent data processing technologies, the comprehensiveness, real-time and accuracy of data collection are ensured.

[0078] Further, as Figure 3 shown, the process of analyzing the physical and chemical properties of the melt samples in step S12 specifically includes the following steps:

[0079] Step S121: Place a laser displacement sensor at the top of the molten pool for non-contact liquid level height monitoring; use a mirror to guide the laser beam into the measurement area; measure the molten pool liquid level height and its fluctuation amplitude in real time;

[0080] Step S122: Fix a high-frequency vibration sensor on the peripheral structure of the molten pool for dynamic monitoring of liquid level fluctuations; capture liquid level fluctuations through solid transmission; make mechanical contact with the molten pool wall through a heat-insulating sleeve; monitor the frequency and amplitude of liquid level fluctuations;

[0081] Step S123: Install ultrasonic sensors in a multi-point array at the top of the molten pool for real-time measurement of the molten pool liquid level height and its fluctuation amplitude; equip with a heat-insulating protective cover to isolate from the molten pool environment; adopt a directional beam design to monitor the intensity and frequency of splashing melt and obtain real-time feedback data for lance adjustment;

[0082] Extract a liquid melt sample directly from the molten pool through a high-temperature heat-resistant sampler; send the extracted sample to a chemical composition analyzer to detect the key components in the sample in real time and obtain real-time chemical information of the smelting reaction;

[0083] Among them, the high-temperature heat-resistant sampler includes high-temperature resistant alloy or ceramic; the chemical composition analyzer includes spectrum analysis and X-ray fluorescence analysis.

[0084] Preferably, in step S121 of this embodiment, a laser displacement sensor is used for non-contact measurement, avoiding interference with the molten pool; a reflecting mirror is used to guide the laser beam into the measurement area to ensure the accuracy and stability of the laser beam; the height and fluctuation amplitude of the molten pool liquid surface are measured in real time, providing high-precision data support; high-precision liquid surface height measurement is provided, which is suitable for monitoring the molten pool in a high-temperature environment; non-contact measurement reduces interference with the molten pool and improves the reliability and safety of measurement; real-time monitoring of the liquid surface height fluctuation helps to adjust process parameters in a timely manner and optimize the smelting process. In step S122, a high-frequency vibration sensor is fixed on the peripheral structure of the molten pool to capture the liquid surface fluctuation; the liquid surface fluctuation is captured through solid transmission to ensure the stability and accuracy of the data; the heat-insulating sleeve is in mechanical contact with the molten pool wall to protect the sensor from the influence of high temperature; the frequency and amplitude of the liquid surface fluctuation are monitored to provide comprehensive dynamic monitoring data; dynamic monitoring of the liquid surface fluctuation is provided, which helps to analyze the flow characteristics of the molten pool; the high-frequency vibration sensor can capture subtle liquid surface fluctuations and improve the sensitivity of monitoring; real-time monitoring of the liquid surface fluctuation frequency and amplitude helps to optimize the stability and uniformity of the molten pool. In step S123, a multi-point array ultrasonic sensor is installed on the top of the molten pool to dynamically monitor the intensity and frequency of the splashing melt; an insulating protective cover is equipped to isolate it from the molten pool environment to ensure the safe operation of the sensor; a directional beam design is adopted to improve the accuracy and efficiency of monitoring; a high-temperature heat-resistant sampler is used to directly extract liquid melt samples from the molten pool; the samples are sent to a chemical composition analyzer to detect the key components in the samples in real time; this embodiment provides dynamic monitoring of the intensity and frequency of the splashing melt, which helps to optimize the adjustment of the spray gun and process parameters; real-time feedback data for spray gun adjustment is obtained, improving the control accuracy of the smelting process; the high-temperature heat-resistant sampler can safely extract samples from the molten pool to ensure the representativeness and accuracy of the samples; the chemical composition analyzer can detect the key components in the samples in real time, providing real-time chemical information of the smelting reaction, which helps to optimize the smelting process. Laser displacement sensor (non-contact liquid surface height monitoring), this sensor is installed directly above the top of the molten pool, and a reflecting mirror is used to guide the laser beam into the measurement area. This sensor is equipped with a high-temperature protective cover and a cooling system to prevent the internal components of the sensor from being affected by high-temperature radiation. A long-distance laser emitter is used to reduce thermal radiation interference. This sensor measures the height and fluctuation amplitude of the molten pool liquid surface in real time. The non-contact design avoids direct exposure to a high-temperature environment. Boundary condition data is provided for numerical simulation. High-frequency vibration sensor (dynamic monitoring of liquid surface fluctuation), this sensor is fixed on the peripheral structure of the molten pool (at the periphery of the water-cooled wall), and the liquid surface fluctuation is captured through solid transmission. The sensor elements are wrapped with high-temperature resistant materials and are in mechanical contact with the molten pool wall through a heat-insulating sleeve. A built-in thermocouple monitors the temperature of the sensor in real time to prevent overheating damage. This sensor can detect the frequency and amplitude of the liquid surface fluctuation, provide data for evaluating the operation effect of the spray gun, and reduce the influence of environmental interference through mechanical contact.Ultrasonic sensor (for splash dynamic monitoring), the sensors are installed in a multi-point array at the top of the molten pool, covering the splash area. The sensors are equipped with heat insulation protective covers to isolate the sensors from the molten pool environment. A directional sound beam design is adopted to avoid signal attenuation caused by high-temperature airflows. It can monitor the intensity and frequency of the splashing melt, providing a basis for real-time feedback for lance adjustment. Extract the melt sample in the molten pool in real time, analyze its physical and chemical properties, and provide data support for judging the smelting stage and adjusting process parameters. This module has high-temperature adaptability and automated operation capabilities to ensure the safety and accuracy of sample extraction. In this module, first, a high-temperature heat-resistant sampler (made of high-temperature-resistant alloy or ceramic, suitable for high-temperature environments above 1000 °C, and equipped with a cooling protection system to prevent the sampler from overheating and damage) is used to directly extract the liquid melt sample from the molten pool. The extracted sample is sent to a chemical composition analyzer (covering spectral analysis and X-ray fluorescence analysis) to detect the key components in the sample in real time, providing real-time chemical information of the smelting reaction and data support for process optimization. Then, through a data interface and a storage unit, the analysis results are transmitted to the control center, and historical sampling data is stored (for the specific principle, refer to the appendix. Figure 4 )。

[0085] Further, as Figure 5 shown, the process of simulating the dynamic behavior inside the molten pool in step S13 specifically includes the following steps:

[0086] Step S131: Process the collected real-time data through a filtering algorithm to eliminate environmental interference, monitor and exclude abnormal data points caused by sensor failures or extreme conditions, and smooth the data;

[0087] Step S132: Utilize the statistical characteristics of the input real-time data to dynamically adjust the coefficients of the filter to minimize the estimation error; where the filter output is:

[0088]

[0089] The weight update rule is:

[0090] w i (t + 1) = w i (t) + μe(t)x(t - i)

[0091] where e(t) = d(t) - y(t) represents the filtering error, and μ is the learning rate;

[0092] In the formula, y(t) represents the output value of the filter at time t; x(t - i) represents the time series of the input real-time data, where t represents the current moment and i represents the time delay; w i represents the weight coefficient of the filter, which represents the weighted value of the input data at different time delays i; w i(t) represents the weight coefficient at time t; n represents the order of the filter, indicating the number of time delays considered by the filter. n - 1 represents the maximum delay order of the filter; μ represents the learning rate, which is used to control the step size of weight update; e(t) represents the filtering error, indicating the difference between the filter output y(t) and the actual expected value d(t).

[0093] Step S133: Calculate the temperature field, flow field, and liquid surface fluctuation of the molten pool, and simulate the dynamic behavior inside the molten pool based on the interaction of the spray gun gas.

[0094] Preferably, in step S131 of this embodiment, data is collected in real time through sensors to ensure the timeliness and accuracy of the data; a filtering algorithm is used to process the collected data to eliminate environmental interference and noise and improve the data quality. Specific methods include median filtering, moving average filtering, etc.; abnormal data points caused by sensor failures or extreme conditions are detected and excluded through the filtering algorithm; the data is smoothed to reduce the influence of high-frequency noise and improve the continuity and stability of the data; noise and interference are effectively removed through the filtering algorithm to ensure the reliability and stability of the data; the anti-interference ability and robustness of the system are improved by monitoring and excluding abnormal data points; the smoothed data is more suitable for analysis and processing, improving the usability of the data. In step S132, the coefficients of the filter are dynamically adjusted using the statistical characteristics (such as mean, variance, etc.) of the input data; by adjusting the filter coefficients, the estimation error is minimized to improve the filtering effect; the filter coefficients are dynamically adjusted so that the system can adapt to different environments and data characteristics; by minimizing the estimation error, the accuracy and reliability of the filtering result are improved; algorithms such as the recursive least squares method can quickly adjust parameters in real-time operations to adapt to dynamic changes. In step S133, based on the interaction of the spray gun gas, the temperature field, flow field, and liquid surface fluctuation of the molten pool are calculated; numerical simulations are carried out using the finite element method or the finite volume method to analyze the dynamic behavior inside the molten pool; combined with external actions such as electromagnetic stirring and ultrasonic vibration, their effects on the temperature field, flow field, and morphology of the molten pool are studied; through numerical simulations, the laws of the temperature field, flow field, and liquid surface fluctuation inside the molten pool are revealed, providing a theoretical basis for process optimization; the simulation results can be used to guide the adjustment of process parameters in actual production to improve the molten pool forming quality and material properties; through coupled analysis, the influence of external actions on the molten pool behavior is revealed, providing a scientific basis for process improvement. The filtering algorithm has strong adaptability to signals with dynamically changing environmental noise characteristics and can be applied to the real-time data processing of multi-sensor fusion. Through this filtering algorithm, sensor jitter signals can be dynamically eliminated, and accidental abnormal points can be effectively removed to ensure the reliability of relevant data. Calculate the temperature field, flow field, and liquid surface fluctuation of the molten pool and its interaction with the spray gun gas, and simulate the dynamic behavior inside the molten pool.

[0095] Furthermore, as Figure 6As shown, in step S133, calculating the temperature field, flow field, and liquid surface fluctuation of the molten pool, based on the process of the interaction of the lance gas, specifically includes the following steps:

[0096] Step S1331: Use a turbulence model to simulate the hydrodynamic behavior in the molten pool; among them, the hydrodynamic behavior includes flow velocity, pressure, and turbulence characteristics; during the calculation, the gas phase and slag phase in the TSL furnace can coexist in any calculation unit; the volume fractions of the two phases vary between 0 and 1, and the sum of the volume fractions of the two phases is always 1 (∑ i a i = 1);

[0097] Among them, the process of simulating the hydrodynamic behavior in the molten pool is:

[0098]

[0099] Among them, ρ, u, and μ respectively represent the density, velocity, and viscosity of the fluid phase. t and g respectively represent the instantaneous moment and the acceleration due to gravity; the surface tension between each phase in any control unit is represented by the symbol F;

[0100] Step S1332: A numerical technique for capturing the gas-liquid interface, simulating the liquid surface fluctuation of the molten pool and the gas-liquid interaction, introducing a volume fraction function to represent the proportion of the fluid in the grid unit;

[0101] Among them, the calculation of the proportion of the fluid in the grid is:

[0102]

[0103] Among them, C represents the volume fraction of the fluid, C = 1 represents a pure liquid region, C = 0 represents a pure gas, and 0 < C < 1 represents a gas-liquid mixed region;

[0104] Step S1333: Simulate and calculate the interfacial tension to obtain the shape of the liquid surface fluctuation of the molten pool and the impact of the lance gas flow on the liquid surface; and simulate the evolution of the temperature field inside the molten pool, and calculate the effects of convection, heat conduction, and volume heat sources;

[0105] Among them, the process of simulating and calculating the interfacial tension Fs:

[0106] Fs = δκ▽C

[0107] δ is the interfacial tension coefficient, κ is the interfacial curvature, calculated through the gradient of the volume fraction C, and the VOF method combines with the turbulence model to dynamically capture the shape of the liquid surface fluctuation of the molten pool and the impact of the lance gas flow on the liquid surface;

[0108] The process of calculating the effects of convection, heat conduction, and volume heat sources:

[0109]

[0110] In the formula, E represents the total fluid flow rate, and Φ represents the volume heat source term.

[0111] Preferably, in step S1331 of this embodiment, the hydrodynamic behavior in the molten bath is described by a turbulence model, including flow velocity, pressure, and turbulence characteristics. In the calculation process, the gas phase and the slag phase in the TSL furnace can coexist in any calculation unit, and the volume fractions of the two phases vary between 0 and 1; and the sum of the volume fractions of the two phases is always 1 (∑ i a i = 1); the turbulence model can accurately describe the velocity, pressure, and turbulence characteristics of the fluid in the molten bath, providing a detailed physical description of the molten bath flow; effectively tracking the interface between the gas phase and the slag phase to ensure the correct representation of the distribution and interaction of the two phases in the calculation unit; being able to simulate the dynamic behavior of turbulence, improving the accuracy and reliability of the simulation. In step S1332, the volume fraction function C is introduced to represent the proportion of the fluid in the grid unit. C = 1 represents the pure liquid region, C = 0 represents the pure gas, and 0 < C < 1 represents the gas-liquid mixing region. The VOF method describes the distribution of different fluids by calculating the phase fraction of the fluid, which is suitable for dealing with complex interface problems; capturing the interface position by calculating the gradient of the volume fraction C to ensure the correct representation of the gas-liquid interface in the calculation unit; adopting a high-order time stepping algorithm (such as RK4) to ensure the stability of the numerical calculation and avoid numerical errors caused by too large a time step; being able to dynamically capture the fluctuation form of the molten bath liquid surface and simulate the impact of the lance gas flow on the liquid surface. Through the calculation of the volume fraction function C, the interaction of the gas-liquid interface is accurately simulated, improving the physical authenticity and accuracy of the simulation; the high-order time stepping algorithm ensures the stability of the numerical calculation and reduces the numerical errors caused by too large a time step. In step S1333, the interface curvature is calculated through the gradient of the volume fraction C, ensuring the accurate calculation of the interfacial tension; the energy conservation equation is used to describe the evolution of the temperature field inside the molten bath, considering the effects of convection, heat conduction, and volume heat source. The energy conservation equation can describe the transfer and distribution of energy in the molten bath; dynamically capture the fluctuation form of the molten bath liquid surface and the impact of the lance gas flow on the liquid surface; accurately simulate the fluctuation form of the molten bath liquid surface and the impact of the lance gas flow on the liquid surface through the calculation of the interfacial tension coefficient δ and the interface curvature κ; the energy conservation equation can describe the evolution of the temperature field inside the molten bath, considering the effects of convection, heat conduction, and volume heat source, improving the physical authenticity and accuracy of the simulation; the combination of the VOF method and the turbulence model can comprehensively simulate the fluctuation of the molten bath liquid surface, gas-liquid interaction, and temperature field evolution, providing a comprehensive physical description of the molten bath flow and heat transfer. The gas phase and the slag phase in the TSL furnace can coexist in any calculation unit. The volume fractions of the two phases vary between 0 and 1, and the sum of the volume fractions of the two phases is always 1 (∑ i a i= 1). This means that when a grid is occupied by a single phase, the volume fractions of this phase and other phases are equal to 1 and 0 respectively. The properties of the fluid, such as density and viscosity, are also shared by all phases and expressed as volume averages. In other words, the physical properties of the fluid in a single grid can either purely represent one of the fluid phases or be expressed as a multiphase mixture. The dynamic behavior in the molten pool is affected by the coupling of multiple physical fields (such as hydrodynamics, heat transfer, and gas-liquid interaction), and the following physical mechanisms need to be comprehensively considered; the thermal field inside the molten pool is described by the energy conservation equation, which describes the evolution of the temperature field inside the molten pool, considering the effects of convection, heat conduction, and volumetric heat sources.

[0112] Furthermore, as Figure 7 shown, the process of obtaining the liquid surface fluctuation prediction result in step S2 specifically includes the following steps:

[0113] Step S21: Retrieve historical molten pool data and perform preprocessing operations such as cleaning, normalization, and discretization on the historical molten pool data; divide the dataset of historical molten pool data into a training set, a validation set, and a test set; construct a molten pool behavior prediction model based on a convolutional neural network and a recurrent neural network;

[0114] Step S22: Use the training set to train the molten pool behavior prediction model, perform forward propagation of the input data through the network, calculate the output result; calculate the loss value between the prediction result and the true label; calculate the gradient of each parameter and update the weights and biases; iterate sequentially until the preset number of training epochs is reached or the stopping condition is satisfied;

[0115] Step S23: Use the validation set to evaluate the performance of the molten pool behavior prediction model and adjust the parameters; evaluate the generalization ability of the molten pool behavior prediction model through the test set, and adjust the molten pool behavior prediction model according to the evaluation results to obtain the final molten pool behavior prediction model; input the real-time data into the molten pool behavior prediction model for prediction to obtain the prediction result.

[0116] Among them, the molten pool behavior prediction model consists of a convolutional feature extractor, a recurrent time series modeler, and a feature fusion operator; the optimization objective of the molten pool behavior prediction model is expressed as:

[0117]

[0118] Among them, the convolutional feature extractor:

[0119]

[0120] The recurrent time series modeler:

[0121]

[0122] The feature fusion operator:

[0123]

[0124] Wherein, Θ, Φ, and Ψ respectively represent the parameter sets of the convolution module, the fusion module, the prediction module, and the cyclic module; represents the i-th preprocessed molten pool data tensor (time × feature × space); represents the four-dimensional tensor of the k-th convolutional kernel; represents the dilated convolution operator; σ gated represents the gated activation function σ(x) = tanh(x) · sigmoid(x ′ ); represents the k-th layer spatial attention coefficient matrix; represents the temporal convolutional operator; W (h) , U (h) represent the forward / backward weight matrices of the h-th layer bidirectional GRU; ⊕ represents the feature concatenation operation; represents the tensor contraction operation; β-div represents the β-divergence loss function; represents the Schatten-p norm; F-reg represents the Frobenius-Spectral mixed regularization; represents the distribution of dynamically generated adversarial samples; R(·) represents the virtual adversarial training regularization term; λ1, λ2 represent the adaptive regularization coefficients (updated by EMA on the validation set);

[0125] The training process of the molten pool behavior prediction model needs to alternately optimize four parameter subspaces:

[0126]

[0127] The final model determines the optimal parameter combination through the early stopping strategy (patience = 20) on the validation set and the adversarial sample stress test on the test set.

[0128] In this embodiment, multi-level dilated convolution is used to extract the three-dimensional spatio-temporal features of the molten pool, and the attention mechanism is combined to suppress noise and focus on key regions, so as to solve the local-global feature balance problem in molten pool dynamic monitoring. Sequential convolution enhances local trend modeling (such as shock wave propagation), and BiGRU encodes the global molten pool dynamics (such as phase change process), jointly solving the multi-time scale dependence in the welding process. The attention mechanism (Attn) analyzes the spatio-temporal feature correlation (such as the correlation between molten pool fluidity and temperature fluctuation), and the secondary interaction enhances the feature synergy (such as the composite effect of thermal stress and deformation). Aiming at the characteristics of large data noise and unstable distribution in industrial molten pool monitoring, the model complexity is constrained by mixed regularization, and adversarial training is used to resist environmental perturbations (such as arc interference)

[0129] Preferably, in step S21 of this embodiment, invalid data, missing values, and outliers are removed to ensure data quality; the data is scaled to the same scale; continuous data is converted into discrete data; the data set is divided into a training set, a validation set, and a test set; the training set is used to train the model, the validation set is used to adjust parameters, and the test set is used to evaluate the model performance; a method combining a convolutional neural network (CNN) and a recurrent neural network (RNN) is used to construct a molten pool behavior prediction model. CNN is used to process image or spatial data, and RNN is used to process time series data; through cleaning and normalization, the accuracy and consistency of the data are ensured, and the influence of noise on the model is reduced; through reasonable data division, the independence and representativeness of the model in the training, validation, and test stages are ensured, and overfitting or underfitting is avoided; by combining the advantages of CNN and RNN, the spatio-temporal characteristics of the molten pool behavior can be better captured, and the prediction accuracy and generalization ability of the model can be improved. In step S22, the input data undergoes forward propagation through a network structure (such as CNN and RNN) to calculate the output result; the loss value between the prediction result and the true label is calculated, and common loss functions include mean square error (MSE) and cross-entropy loss; the gradient of each parameter is calculated through the backpropagation algorithm, and the weights and biases are updated; the forward propagation, loss calculation, and backpropagation processes are repeated until a preset number of training epochs is reached or a stop condition is satisfied; through the backpropagation algorithm, the weights and biases are gradually adjusted to continuously optimize the model during the training process and improve the prediction accuracy; by setting a stop condition (such as the maximum number of training epochs or the minimum loss value), it is ensured that the model converges within a reasonable time, and overfitting or excessive training time is avoided; through multiple iterative trainings, the model can better fit the training data and improve the prediction accuracy for new data. In step S23, the validation set is used to evaluate the performance of the model, and parameters (such as learning rate, batch size, etc.) are adjusted to optimize the model; the test set is used to evaluate the generalization ability of the model to ensure that the model can also perform well on unseen data; according to the evaluation results of the test set, the model is further adjusted, such as increasing or decreasing the number of layers, changing the activation function, etc.; real-time data is input into the final molten pool behavior prediction model for prediction to obtain the prediction result; through validation set evaluation, the optimal hyperparameter combination is found to improve the performance and stability of the model; through test set evaluation, it is ensured that the model performs well on new data and overfitting is avoided; the model is applied to real-time data prediction to provide accurate prediction results for actual production.

[0130] Furthermore, as Figure 8 shown, the process of obtaining the prediction result in step S21 specifically includes the following steps:

[0131] Step S211: Input the real-time data into the molten pool behavior prediction model, extract the spatial features of the real-time data. The input layer receives the real-time data matrix, and each row represents the time series of a sensor; the convolutional layer extracts local features through convolution operations; a feature map with dimensionality reduction characteristics is generated.

[0132] Among them, the convolutional layer extracts local features through convolution operations, specifically as follows:

[0133]

[0134] Among them, w k,l is the convolutional kernel parameter, b is the bias, and f(i, j) is the convolution output feature.

[0135] Step S212: Capture the time dependence of the molten pool liquid level fluctuation. The input layer receives the extracted time feature sequence, and the hidden layer calculates a new hidden state based on the current input and the hidden state at the previous moment; the input layer predicts the future trend value of the liquid level fluctuation.

[0136] Among them, calculating the new hidden state is specifically as follows:

[0137] h t = σ(W h g t-1 + W x x t + b h )

[0138] In the formula, h t represents the current hidden state, which is the intermediate result calculated by the model at time t; h t-1 represents the hidden state at the previous moment, which is the calculation result of the model at time t-1 and is used to transmit the historical information of the time series; x t represents the input data at the current moment, that is, the extracted time feature sequence; W h is the hidden state weight matrix, which is used to map the hidden state h t-1 at the previous moment to the calculation of the current hidden state h t ; W x is the input weight matrix, which is used to map the current input x t to the calculation of the current hidden state h t ; b h is the bias term of the hidden state, which is used to adjust the calculation result of the hidden state; σ represents the activation function, which is used to introduce a non-linear relationship and enhance the expression ability of the model.

[0139] The input layer predicts the future trend value of the liquid level fluctuation, specifically as follows:

[0140]

[0141] In the formula, represents the predicted value of the liquid level fluctuation at time t, which is the model's estimation of the future liquid level fluctuation trend; h r represents the hidden state h at the current moment t or other relevant hidden states, serving as the basis for prediction; W y is the output weight matrix, used to map the hidden state h r to the predicted value y t in the calculation; b y is the output bias term, used to adjust the prediction result.

[0142] Step S213: Generate the optimal adjustment suggestions for the spray gun parameters based on the future trend value of the liquid level fluctuation and the simulation results, transmit the optimal adjustment suggestions to the control center, generate control instructions according to the optimal adjustment suggestions, and perform spray gun adjustment.

[0143] Preferably, in step S211 of this embodiment, the real-time data is input into the molten pool behavior prediction model to extract spatial features of the real-time data; the real-time data matrix is received, and each row represents the time series of a sensor; local features are extracted through convolution operations to generate a feature map with dimensionality-reduced features; local features are extracted through convolution operations to enhance the recognition ability of the molten pool behavior prediction model for molten pool behavior; the generated feature map has dimensionality-reduced features, reducing the complexity of the data and improving the calculation efficiency of the molten pool behavior prediction model; it can process real-time data to ensure the timeliness of the prediction results. In step S212, the time dependence of the molten pool liquid level fluctuation is captured; the extracted time feature sequence is received; based on the current input and the hidden state at the previous moment, a new hidden state is calculated; the input layer predicts the future trend of the liquid level fluctuation; the time dependence in the time series is captured through the hidden layer, improving the prediction ability of the molten pool behavior prediction model for the liquid level fluctuation trend; it can dynamically predict the future trend of the liquid level fluctuation; through the recursive calculation of the hidden layer, the robustness and adaptability of the molten pool behavior prediction model are enhanced. In step S213, according to the future trend value of the liquid level fluctuation and the simulation results, an optimal adjustment suggestion for the lance parameters is generated, and the optimal adjustment suggestion is transmitted to the control center. A control instruction is generated according to the optimal adjustment suggestion to perform the operation of lance adjustment; by generating the optimal adjustment suggestion, the lance parameters are optimized, improving the control accuracy of the molten pool behavior, realizing the automated process from prediction to control, and improving the production efficiency and product quality; a closed-loop control system is formed, which can adjust the lance parameters in real time to ensure the stability and consistency of the molten pool behavior. It is composed of a neural network model and a big data analysis platform. Based on the convolutional neural network (CNN) and the recurrent neural network (RNN), a molten pool behavior prediction model is constructed, and it is dynamically updated using real-time monitoring data and historical molten pool data to achieve adaptive prediction. The neural network model is trained using historical molten pool data to predict the molten pool liquid level fluctuation trend and adverse working conditions. Assist in optimizing the lance gas parameters. The neural network model combines the advantages of the convolutional neural network (CNN) and the recurrent neural network (RNN) to complete the prediction of the molten pool liquid level fluctuation trend and adverse working conditions. Spatial features are extracted from real-time monitoring data (such as molten pool liquid level height, lance gas flow parameters). It includes an input layer, a convolutional layer, and a pooling layer. Among them, the input layer receives the real-time data matrix X∈R from the monitoring module m×n, each line represents the time series of a sensor. The pooling layer reduces the dimension of the feature matrix, retains the main features, and avoids overfitting. The final result generates a feature map with reduced-dimensional features as the input to the RNN. It captures the time dependence of the molten pool liquid level fluctuations and predicts future trends. It includes an input layer, a hidden layer, and an output layer. Among them, the input layer receives the feature sequence ht extracted by the CNN. The hidden layer calculates a new hidden state based on the current input and the hidden state at the previous moment. Through real-time data input, real-time monitoring and historical molten pool data are input, and then data preprocessing is carried out, including filtering and outlier detection. Then, spatial and temporal modeling is completed through the CNN and RNN, so as to predict the liquid level fluctuations and melt splashing conditions. Finally, the spray gun parameters are optimized according to the predicted values, and the optimized data is transmitted to the control center to adjust the real-time spray gun parameters. (For the specific principle, please refer to Appendix Figure 9 )

[0144] Furthermore, as Figure 10 shown, the process of generating the optimal adjustment suggestions for the spray gun parameters in step S213 specifically includes the following steps:

[0145] Step S2131: Randomly generate an initial population, and each particle represents a set of spray gun parameters; the position of each particle represents a set of parameters, and the velocity represents the moving direction and speed of the particle in the solution space; based on the simulation and test results, evaluate the molten pool liquid level fluctuations and splashing degree under the current parameter settings;

[0146] Step S2132: Find the parameter combination that minimizes the liquid level fluctuations and reduces splashing; calculate the fitness value of each particle, and update the individual optimal position and the global optimal position; if the fitness value of the current particle is better than the historical optimal value, update the individual optimal position; if it is better than the global optimal value, update the global optimal position;

[0147] Step S2133: Iterate sequentially until the maximum number of iterations is reached or the fitness value converges to a preset threshold; output the spray gun parameter combination corresponding to the global optimal position as the optimal adjustment suggestion.

[0148] Preferably, in step S2131 of this embodiment, the initial population is randomly generated, and each particle represents a set of spray gun parameters. This method can ensure that the algorithm starts searching from multiple initial points, increasing the possibility of finding the global optimal solution; the position of each particle represents a set of parameters, and the velocity represents the moving direction and speed of the particle in the solution space. This representation enables the particle to dynamically adjust its position in the solution space to find the optimal solution; based on the simulation and test results, the molten pool liquid level fluctuation and splashing degree under the current parameter settings are evaluated; random initialization can prevent the algorithm from falling into a local optimal solution and increase the possibility of finding the global optimal solution; through the dynamic adjustment of position and velocity, the particle can efficiently search for the optimal solution in the solution space; the fitness evaluation based on the simulation and test results ensures the accuracy and reliability of the optimization process. In step S2132, the fitness value of each particle is calculated to measure the effect of the current parameter combination; if the fitness value of the current particle is better than its historical optimal value, the individual optimal position is updated; if the fitness value of the current particle is better than the global optimal value, the global optimal position is updated; through the update of the individual optimal and global optimal, the particle swarm algorithm can gradually approach the global optimal solution; the update mechanism of the individual optimal and global optimal enables the particle swarm algorithm to dynamically adjust the search direction and speed according to the current search situation; by continuously updating the optimal position, the algorithm can converge to the optimal solution faster. In step S2133, the maximum number of iterations or the fitness value converging to a preset threshold is set as the termination condition; when the termination condition is met, the spray gun parameter combination corresponding to the global optimal position is output; by setting the termination condition, it is ensured that the algorithm can converge to the optimal solution within a limited time; the spray gun parameter combination corresponding to the output global optimal position can be directly applied to actual production to provide optimal adjustment suggestions; through iterative optimization, the algorithm can find the optimal solution in a short time and improve production efficiency. The genetic algorithm or particle swarm optimization algorithm is applied to seek a balance among air flow control, energy utilization, and smelting efficiency, and according to the simulation and prediction results, the optimal adjustment suggestions for the spray gun parameters are generated to dynamically optimize the gas flow rate and pressure, thereby reducing the molten pool liquid level fluctuation and splashing. It consists of an optimization parameter interface and a feedback updater, outputs the optimization result to the control center, uses the closed-loop feedback algorithm (PID controller) to collect actual feedback data, compares it with the prediction result, and dynamically adjusts the model. The continuous correction of the prediction result and the actual working condition is realized.

[0149] Further, as Figure 11 shown, the process of visually displaying the adjusted spray gun parameters and the molten pool monitoring data in step S3 specifically includes the following steps:

[0150] Step S31: Generate a control instruction, record the adjusted spray gun parameters in the database, including information such as the parameter values before and after adjustment, the adjustment time, and the molten pool state; regularly back up the recorded spray gun parameter data;

[0151] Step S32: The adjusted spray gun parameters and the molten pool monitoring data are displayed in real time through a visualization device, and regulatory feedback information is provided on the visualization interface; this includes the current molten pool state, the effect of parameter adjustment, the prediction result of liquid level fluctuation, etc.;

[0152] Step S33: Provide a switching function between manual and automatic control modes; for the manual mode, the spray gun parameters are adjusted manually; for the automatic mode, the spray gun parameters are automatically adjusted according to the molten pool behavior prediction model, the simulation results, and the formed optimal spray gun operation parameters; the data is monitored in real time and compared with the prediction results, and the molten pool behavior prediction model is adjusted according to the comparison results.

[0153] Preferably, in step S31 of this embodiment, according to the spray gun adjustment suggestion, the system automatically generates a control instruction to adjust the spray gun parameters in real time, records the adjusted spray gun parameters in the database, and performs regular backups; through intelligent algorithms and real-time data feedback, the system can accurately adjust the spray gun parameters, reduce manual intervention, and improve the control accuracy; regularly backing up the parameter data ensures the security and integrity of the data. In step S32, the adjusted spray gun parameters and the molten pool monitoring data are displayed in real time on the interface through a visualization device; the interface provides regulatory feedback information, including the current molten pool state, the effect of parameter adjustment, the prediction result of liquid level fluctuation, etc.; the molten pool state is monitored in real time, and the molten pool geometry and state information are visualized; through the visualization interface, the operator can quickly obtain the molten pool state and the effect of parameter adjustment, improving the operation efficiency. In step S33, the operator can manually adjust the spray gun parameters; the system automatically adjusts the spray gun parameters according to the molten pool behavior prediction model and the simulation results, the real-time monitoring data is compared with the prediction results, and the molten pool behavior prediction model is adjusted according to the comparison results; the manual mode provides flexibility and is suitable for special requirements; the automatic mode improves the automation degree and efficiency of the operation; through the comparison of the real-time monitoring data and the prediction results, the system can continuously optimize the molten pool behavior prediction model and improve the prediction accuracy.

[0154] As Figure 12 shown, this embodiment also provides an embodiment of the top-blown smelting melt stability intelligent regulation system. In this embodiment, the top-blown smelting melt stability intelligent regulation system is applied to the top-blown smelting melt stability intelligent regulation method as described in the above embodiment. The top-blown smelting melt stability intelligent regulation system specifically includes:

[0155] A molten pool simulation module 1, which is used to collect real-time molten pool data of the spray gun inlet, the molten pool liquid level, and the internal melt, classify, filter, and integrate the real-time molten pool data; perform characteristic statistics on the integrated real-time molten pool data, dynamically adjust the error; simulate the dynamic behavior inside the molten pool;

[0156] The molten pool behavior prediction module 2 is used to retrieve historical molten pool data and construct a molten pool behavior prediction model; preprocess the historical molten pool data and divide it into a training set, a validation set, and a test set; train the molten pool behavior prediction model using the training set; extract spatial features from the real-time molten pool data, capture the time dependence of the molten pool liquid level fluctuation, and obtain the liquid level fluctuation prediction result; generate optimal lance operation parameters based on the simulation result and the prediction result.

[0157] The lance adjustment module 3 is used to generate a control instruction according to the optimal lance operation parameter and perform the adjusted lance parameter; store the adjusted lance parameter; visually display the adjusted lance parameter and the molten pool monitoring data, and the visualization device displays the real-time lance data and the control feedback, providing a manual and automatic control mode switch.

[0158] Preferably, the molten pool simulation module 1 of this embodiment is used to collect the real-time data of the lance inlet, the molten pool liquid level and the internal melt, and classify, filter and integrate these data. After the integrated data is analyzed for statistical characteristics, the error is dynamically adjusted to improve the reliability of the data and the accuracy of the model. By simulating the dynamic behavior inside the molten pool, the physical process of the molten pool can be better understood, providing a theoretical basis for optimizing process parameters. Through data preprocessing and feature extraction, the accuracy and consistency of the data are ensured, providing a reliable basis for analysis. Dynamically adjusting the error improves the prediction accuracy of the model, providing more accurate data support for predicting the behavior of the molten pool. By simulating the dynamic behavior inside the molten pool, the physical process of the molten pool can be better understood, providing a theoretical basis for optimizing process parameters. The molten pool behavior prediction module 2 is used to retrieve historical molten pool data and perform preprocessing, dividing it into a training set, a validation set and a test set. The training set is used to train the molten pool behavior prediction model, and the performance of the model is evaluated through the validation set and the test set to ensure the generalization ability of the model. The spatial features of the real-time molten pool data are extracted to capture the time dependence of the molten pool liquid level fluctuation, and the prediction result of the liquid level fluctuation is obtained. Based on the simulation results and the prediction results, the optimal lance operation parameters are generated to achieve precise control of the molten pool behavior. Through the preprocessing of historical molten pool data and model training, the accuracy of molten pool behavior prediction is improved and the error is reduced. Through feature extraction and capturing of time dependence, the molten pool liquid level fluctuation can be predicted more accurately, providing a scientific basis for lance adjustment. Generating the optimal lance operation parameters improves the control accuracy of the molten pool behavior and reduces defects in the production process. The lance adjustment module 3 generates a control command according to the optimal lance operation parameters, adjusts the lance parameters, and stores the adjusted parameters; the adjusted lance parameters and the molten pool monitoring data are visually displayed to provide real-time data and control feedback, supporting the switching between manual and automatic control modes. Through the collaborative perception of multi-sensor data, the real-time perception and monitoring of the molten pool state are realized. By generating the optimal lance operation parameters, the control accuracy of the molten pool behavior is improved and defects in the production process are reduced; through visual display and feedback mechanism, the molten pool state is monitored in real time, the process parameters are adjusted in time, and the production efficiency is improved; through the collaborative perception of multi-sensor data, the comprehensive perception and monitoring of the molten pool state are realized, and the reliability and stability of the system are improved.

[0159] As Figure 13 shown, this embodiment provides an embodiment of an electronic device. In this embodiment, the electronic device 4 includes a processor 41 and a memory 42 coupled to the processor 41.

[0160] The memory 42 stores program instructions for implementing the top-blown smelting melt stable intelligent control method of any of the above embodiments.

[0161] The processor 41 is used to execute the program instructions stored in the memory 42 for the layout of the intelligent regulation method for the top-blown smelting melt stability.

[0162] Among them, the processor 41 can also be called a CPU (Central Processing Unit). The processor 41 may be an integrated circuit chip with the ability to process signals. The processor 41 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0163] Furthermore, Figure 14 FIG. is a schematic structural diagram of a storage medium according to an embodiment of the present application. The storage medium 5 of the embodiment of the present application stores program instructions 51 that can implement all the above methods. Among them, the program instructions 51 can be stored in the above storage medium in the form of a software product, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media that can store program codes such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, or a terminal device such as a computer, a server, a mobile phone, or a tablet.

[0164] In several embodiments provided by the present invention, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some interfaces, and the indirect coupling or communication connection of devices or units can be in an electrical, mechanical, or other form.

[0165] In addition, in each embodiment of the present invention, each functional unit may be integrated into one processing unit, may exist separately as individual physical units, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware or in the form of software functional units. The above are only the implementation manners of the present invention and do not limit the patent scope of the present invention. Any equivalent structural or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied to other related technical fields, shall be equally included in the patent protection scope of the present invention.

[0166] The specific implementation manners of the invention have been described in detail above, but they are only examples, and the present invention is not limited to the specific implementation manners described above. For those skilled in the art, any equivalent modification or substitution of the invention is also within the scope of the present invention. Therefore, all equivalent transformations, modifications, improvements, etc. made without departing from the spirit and principles of the present invention should be covered by the scope of the present invention.

Claims

1. A method for stabilizing and intelligently controlling a top-blown smelting melt, characterized in that: The top-blowing smelting melt stable intelligent control method comprises: Collect real-time molten pool data of the spray gun inlet, molten pool liquid surface and internal melt, classify, filter and integrate the real-time molten pool data; perform characteristic statistics on the integrated real-time molten pool data, dynamically adjust the error; simulate the dynamic behavior inside the molten pool; Retrieve historical molten pool data and build a molten pool behavior prediction model; preprocess historical molten pool data and divide them into training set, validation set and test set; use the training set to train the molten pool behavior prediction model; extract spatial features from real-time molten pool data to capture the time dependence of molten pool liquid level fluctuations and obtain liquid level fluctuation prediction results; generate optimal spray gun operating parameters based on simulation results and prediction results; Generate control instructions based on the optimal spray gun operating parameters to adjust the spray gun parameters; store the adjusted spray gun parameters; visualize the adjusted spray gun parameters and molten pool monitoring data, and the visualization device displays real-time spray gun data and control feedback, and provides switching between manual and automatic control modes.

2. The method for stabilizing and intelligently controlling the melt in top-blowing smelting according to claim 1, characterized in that: The process of simulating the dynamic behavior inside the molten pool includes the following steps: Use a flow meter to collect and record the injection gas flow in real time; use a pressure sensor to monitor the gas pressure at the spray gun inlet; use a temperature sensor to measure the gas temperature; use a gas concentration analyzer to measure the composition ratio of the mixed gas; Dynamically monitor the key characteristics of the molten pool liquid level and the internal melt to evaluate fluctuations in the smelting process; extract melt samples from the molten pool in real time, test the extracted melt samples from the molten pool, and analyze the physical and chemical properties of the melt samples; The collected real-time data is processed through a filtering algorithm to eliminate environmental interference, calculate the temperature field, flow field, liquid level fluctuation of the molten pool and the interaction with the spray gun gas, and simulate the dynamic behavior inside the molten pool.

3. The method for stabilizing and intelligently controlling the melt in top-blowing smelting according to claim 2, characterized in that: The process of analyzing the physical and chemical properties of a melt sample includes the following steps: Place a laser displacement sensor on the top of the molten pool for non-contact liquid level monitoring; use a reflector to guide the laser beam into the measurement area; measure the molten pool liquid level and fluctuation amplitude in real time; A high-frequency vibration sensor is fixed on the peripheral structure of the molten pool for dynamic monitoring of liquid level fluctuations; liquid level fluctuations are captured through solid transfer; mechanical contact is made with the molten pool wall through the thermal insulation sleeve; and the frequency and amplitude of liquid level fluctuations are monitored; Ultrasonic sensors are installed in a multi-point array on the top of the molten pool to measure the molten pool liquid level and its fluctuation amplitude in real time; they are equipped with a heat-insulating protective cover to isolate them from the molten pool environment; a directional beam design is adopted to monitor the intensity and frequency of the splashing melt and obtain real-time feedback data for spray gun adjustment.

4. The method for stabilizing and intelligently controlling the melt in top-blowing smelting according to claim 2, characterized in that: The process of simulating the dynamic behavior inside the molten pool includes the following steps: The collected real-time data is processed through filtering algorithms to eliminate environmental interference, monitor and eliminate abnormal data points caused by sensor failure or extreme conditions, and smooth the data; Using the statistical characteristics of the input real-time data, the coefficients of the filter are dynamically adjusted to minimize the estimation error; The temperature field, flow field, and liquid surface fluctuation of the molten pool are calculated, and the dynamic behavior inside the molten pool is simulated based on the interaction of the spray gun gas.

5. The method for stabilizing and intelligently controlling the melt in top-blowing smelting according to claim 4, characterized in that: The process based on the interaction of the spray gun gases includes the following steps: The turbulence model is used to simulate the fluid dynamics behavior in the molten pool; the fluid dynamics behavior includes flow velocity, pressure and turbulence characteristics; during the calculation process, the gas phase and slag phase in the TSL furnace exist simultaneously in any calculation unit; the volume fractions of the two phases vary between 0 and 1, and the sum of the volume fractions of the two phases is always 1; Numerical technology to capture the gas-liquid interface, simulate the molten pool liquid level fluctuation and gas-liquid interaction, and introduce a volume fraction function to represent the proportion of fluid in the grid unit; The interfacial tension is simulated and calculated to obtain the shape of the molten pool liquid surface fluctuation and the impact of the spray gun airflow on the liquid surface; the temperature field evolution inside the molten pool is simulated to calculate the effects of convection, heat conduction and volume heat source.

6. The method for stabilizing and intelligently controlling the melt in top-blowing smelting according to claim 1, characterized in that: The process of obtaining the liquid level fluctuation prediction result includes the following steps: Retrieve historical molten pool data, and perform cleaning, normalization, and discretization preprocessing operations on the historical molten pool data; divide the data set of historical molten pool data into training set, validation set, and test set; build a molten pool behavior prediction model based on convolutional neural network and recurrent neural network; Use the training set to train the melt pool behavior prediction model, propagate the input data forward through the network, and calculate the output result; calculate the loss value between the predicted result and the true label; calculate the gradient of each parameter, and update the weight and bias; iterate in sequence until the preset number of training rounds is reached or the stopping condition is met; The validation set is used to evaluate the performance of the molten pool behavior prediction model and adjust the parameters. The test set is used to evaluate the generalization ability of the molten pool behavior prediction model, and the molten pool behavior prediction model is adjusted according to the evaluation results to obtain the final molten pool behavior prediction model. The real-time data is input into the molten pool behavior prediction model for prediction to obtain the prediction results.

7. The method for stabilizing and intelligently controlling the melt in top-blowing smelting according to claim 6, characterized in that: The process of obtaining the prediction results includes the following steps: The real-time data is input into the molten pool behavior prediction model, and the spatial features of the real-time data are extracted. The input layer receives the real-time data matrix, and each row represents the time series of a sensor; the convolution layer extracts local features through convolution operations; and generates a feature map with dimensionality reduction characteristics; Capturing the time dependency of the molten pool liquid level fluctuation, the input layer receives the extracted time feature sequence, and the hidden layer calculates the new hidden state based on the current input and the hidden state at the previous moment; the input layer predicts the future trend value of the liquid level fluctuation; According to the future trend value of the liquid level fluctuation and the simulation results, the optimal adjustment suggestions for the spray gun parameters are generated, and the optimal adjustment suggestions are transmitted to the control center. According to the optimal adjustment suggestions, control instructions are generated to adjust the spray gun.

8. The method for stabilizing and intelligently controlling the melt in top-blowing smelting according to claim 7, characterized in that: The process of generating the optimal adjustment recommendations for the spray gun parameters includes the following steps: The initial population is randomly generated, and each particle represents a set of spray gun parameters; the position of each particle represents a set of parameters, and the velocity represents the moving direction and speed of the particle in the solution space; based on the simulation and test results, the molten pool liquid level fluctuation and splashing degree under the current parameter settings are evaluated; Find the parameter combination that minimizes liquid surface fluctuation and reduces splashing; calculate the fitness value of each particle, and update the individual optimal position and the global optimal position; if the fitness value of the current particle is better than the historical optimal value, update the individual optimal position; if it is better than the global optimal value, update the global optimal position; Iterate in sequence until the maximum number of iterations is reached or the fitness value converges to the preset threshold; output the spray gun parameter combination corresponding to the global optimal position as the optimal adjustment suggestion.

9. The method for stabilizing and intelligently controlling the melt in top-blowing smelting according to claim 1, characterized in that: The process of visualizing the adjusted gun parameters and the melt pool monitoring data includes the following steps: Generate control instructions and record the adjusted spray gun parameters into the database, including parameter values ​​before and after adjustment, adjustment time, and molten pool status information; regularly back up the recorded spray gun parameter data; The adjusted spray gun parameters and molten pool monitoring data are displayed in real time through visualization equipment, and control feedback information is provided on the visualization interface; including the current molten pool status, parameter adjustment effect, and liquid level fluctuation prediction results; Provides switching function between manual and automatic control modes; manual mode manually adjusts the spray gun parameters; automatic mode automatically adjusts the optimal spray gun operating parameters according to the molten pool behavior prediction model and simulation results as well as the formed parameters; real-time monitoring of data, and comparison with the prediction results, and adjustment of the molten pool behavior prediction model based on the comparison results.

10. A top-blowing smelting melt stabilization intelligent control system, which is applied to the top-blowing smelting melt stabilization intelligent control method according to any one of claims 1 to 9, characterized in that: The top-blowing smelting melt stabilization intelligent control system comprises: The molten pool simulation module is used to collect real-time molten pool data of the spray gun inlet, molten pool liquid surface and internal melt, classify, filter and integrate the real-time molten pool data; perform characteristic statistics on the integrated real-time molten pool data, dynamically adjust the error; and simulate the dynamic behavior inside the molten pool; The molten pool behavior prediction module is used to retrieve historical molten pool data and build a molten pool behavior prediction model; preprocess the historical molten pool data and divide it into a training set, a validation set, and a test set; use the training set to train the molten pool behavior prediction model; extract spatial features from real-time molten pool data, capture the time dependence of molten pool liquid level fluctuations, and obtain liquid level fluctuation prediction results; generate optimal spray gun operating parameters based on simulation results and prediction results; The spray gun adjustment module is used to generate control instructions according to the optimal spray gun operating parameters to adjust the spray gun parameters; store the adjusted spray gun parameters; and visualize the adjusted spray gun parameters and molten pool monitoring data. The visualization device displays real-time spray gun data and control feedback, and provides switching between manual and automatic control modes.

Images