Temperature control method and system for hot working process and storage medium

Abstract

The invention relates to the technical field of processing temperature control, and discloses a temperature control method and system for a hot processing process and a storage medium. The method comprises the following steps: collecting hot working whole process temperature data through multiple sensors, calculating a temperature change rate, an extreme point and a uniformity index, and obtaining key characteristic parameters; establishing a quantitative relation matrix of the process parameters and the temperature field by using an orthogonal test; calculating an optimal process parameter combination based on the matrix by using a rolling time domain optimization algorithm to generate a control instruction; and performing execution opportunity and duration prediction and correction on the instruction to form a precise target control instruction. The temperature control precision and stability in the hot working process of the large alloy steel component are improved, so that the product quality is improved, the energy consumption is reduced, and the production period is shortened.

Description

Technical Field

[0001] The present application relates to the technical field of processing temperature control, and particularly to a temperature control method, system and storage medium for a hot processing process. Background Art

[0002] In the manufacturing process of heavy equipment alloy steel components, hot processing is one of the key processes, including smelting, casting, forging and heat treatment. The hot processing process is usually carried out in a high-temperature environment, which is a typical long-process manufacturing process with many process types and many process factors affecting the manufacturing quality. The high-temperature working environment of hot processing also makes it difficult to collect process data. Currently, the temperature control of hot processing mainly relies on empirical judgment and limited point measurements. Traditional temperature control methods often rely on discrete temperature measurement points and manual adjustment combined with the experience of process personnel, lacking systematic real-time temperature field monitoring and precise control means. This method can cope with small or conventional components, but as the size and weight of equipment components continue to increase and the performance requirements continue to improve, traditional control methods are no longer able to meet the needs.

[0003] The main problems existing in the current hot processing temperature control are incomplete temperature field monitoring and inaccurate control. Due to limited temperature measurement points, the complete temperature distribution inside and on the surface of the component cannot be obtained, resulting in insufficient temperature field information; temperature control mainly relies on empirical judgment, lacking a quantitative process parameter-temperature field relationship model, and the control accuracy is low; there is a lag in the response of hot processing equipment, and the actual execution effect of control instructions deviates from the expectation, and there is a lack of effective prediction and correction mechanisms. These problems lead to unstable product quality, large temperature differences between the inside and outside of large components, poor tissue uniformity, and easy generation of defects. Especially when large components continuously break through the limits of size, weight and performance, the problems are more obvious. Summary of the Invention

[0004] The present application provides a temperature control method, system and storage medium for a hot processing process, which are used to improve the temperature control accuracy and stability of the hot processing process of large alloy steel components, thereby improving product quality, reducing energy consumption and shortening the production cycle.

[0005] In a first aspect, the present application provides a temperature control method for a hot processing process. The temperature control method for the hot processing process includes: collecting temperature data of each process of smelting, casting, forging, and heat treatment to obtain the spatio-temporal temperature distribution information of the entire hot processing process; calculating the temperature change rate, temperature extreme points, and temperature uniformity index during the hot processing process based on the spatio-temporal temperature distribution information to obtain key temperature characteristic parameters; performing an orthogonal experiment analysis on process parameters based on the key temperature characteristic parameters to establish a quantitative relationship matrix of the influence of process parameters on the temperature field; calculating the optimal combination of heating power, cooling rate, and holding time using a rolling horizon optimization algorithm according to the quantitative relationship matrix to generate a temperature control instruction; predicting the execution timing and execution duration of the temperature control instruction to obtain an execution prediction result, and correcting the control instruction according to the execution prediction result to obtain a target control instruction.

[0006] In a second aspect, the present application provides a temperature control system for a hot processing process. The temperature control system for the hot processing process includes: a collection module for collecting temperature data of each process of smelting, casting, forging, and heat treatment to obtain the spatio-temporal temperature distribution information of the entire hot processing process; a calculation module for calculating the temperature change rate, temperature extreme points, and temperature uniformity index during the hot processing process based on the spatio-temporal temperature distribution information to obtain key temperature characteristic parameters; an analysis module for performing an orthogonal experiment analysis on process parameters based on the key temperature characteristic parameters to establish a quantitative relationship matrix of the influence of process parameters on the temperature field; a generation module for calculating the optimal combination of heating power, cooling rate, and holding time using a rolling horizon optimization algorithm according to the quantitative relationship matrix to generate a temperature control instruction; a prediction module for predicting the execution timing and execution duration of the temperature control instruction to obtain an execution prediction result, and correcting the control instruction according to the execution prediction result to obtain a target control instruction.

[0007] In a third aspect, there is provided a temperature control device for a hot processing process, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory so that the temperature control device for the hot processing process executes the above-mentioned temperature control method for the hot processing process.

[0008] In a fourth aspect, there is provided a computer-readable storage medium, wherein instructions are stored in the computer-readable storage medium, and when it runs on a computer, it causes the computer to execute the above-mentioned temperature control method for the hot processing process.

[0009] In the technical solution provided by this application, temperature data of each process of smelting, casting, forging, and heat treatment are collected through a multi-sensor network, achieving a comprehensive acquisition of the spatio-temporal temperature distribution information of the entire hot working process and overcoming the problem of insufficient temperature field information caused by limited temperature measurement points in traditional methods; based on the obtained spatio-temporal temperature distribution information, key temperature characteristic parameters such as temperature change rate, temperature extreme points, and temperature uniformity index are calculated, providing quantitative indicators for temperature field analysis and enhancing the accuracy and comprehensiveness of temperature field analysis; through orthogonal experiment analysis, a quantitative relationship matrix of the influence of process parameters on the temperature field is established, transforming the traditional temperature control relying on experience into precise control based on data and models, and realizing the two-way mapping from process parameters to the temperature field and from the temperature field to process parameters; the rolling horizon optimization algorithm is used to calculate the optimal combination of heating power, cooling rate, and holding time, generating temperature control instructions, upgrading the temperature control strategy from static optimization to dynamic optimization, being able to respond to working conditions changes in real time, and improving the adaptability and accuracy of control; by predicting the execution time and execution duration of temperature control instructions and correcting the control instructions according to the prediction results, the problems of lag and deviation in the execution of control instructions are solved, and the execution accuracy of control instructions is significantly improved. Brief Description of the Drawings

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0011] Figure 1 It is a schematic diagram of an embodiment of the temperature control method for the hot working process in the embodiment of this application; Figure 2 It is a schematic diagram of an embodiment of the temperature control system for the hot working process in the embodiment of this application; Figure 3 It is a structural schematic block diagram of the temperature control device for the hot working process in the embodiment of the present invention. Detailed Embodiments

[0012] The embodiments of the present application provide a temperature control method, system and storage medium for the hot processing process. The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and the above-mentioned drawings of the present application are used to distinguish similar objects and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order different from that illustrated or described here. In addition, the term "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific process of the embodiments of the present application will be described below. Please refer to Figure 1 , an embodiment of the temperature control method for the hot processing process in the embodiments of the present application includes: Step S101, collect the temperature data of each process of smelting, casting, forging and heat treatment to obtain the spatio-temporal temperature distribution information of the entire hot processing process; Step S102, calculate the temperature change rate, temperature extreme points and temperature uniformity index during the hot processing process according to the spatio-temporal temperature distribution information to obtain key temperature characteristic parameters; Step S103, perform an orthogonal test analysis on the process parameters based on the key temperature characteristic parameters to establish a quantitative relationship matrix of the influence of the process parameters on the temperature field; Step S104, according to the quantitative relationship matrix, use the rolling horizon optimization algorithm to calculate the optimal combination of heating power, cooling rate and holding time, and generate a temperature control instruction; Step S105, predict the execution timing and execution duration of the temperature control instruction to obtain an execution prediction result, and correct the control instruction according to the execution prediction result to obtain a target control instruction.

[0014] It can be understood that the execution subject of the present application can be a temperature control system for the hot processing process, or a terminal or a server. Specifically, it is not limited here. The embodiments of the present application will be described by taking the server as the execution subject as an example.

[0015] Specifically, in the smelting process, the infrared thermometer is mainly installed above the electric arc furnace, with a measurement range of 1000°C - 1800°C and an accuracy of ±0.5%, used to monitor the surface temperature of the molten metal; the thermocouple is embedded in the furnace wall, with a measurement range of 20°C - 1600°C and an accuracy of ±0.25%, used to monitor the temperature inside the furnace; the thermal imager is set up at the edge of the smelting area, with a spatial resolution of 320×240 pixels and a temperature resolution of 0.05°C, used to monitor the overall temperature distribution. In the casting process, the infrared thermometer is installed in the pouring area to measure the pouring temperature; the thermocouple array is buried in the mold wall in a grid pattern with a spacing of 20 cm to achieve three-dimensional monitoring of the mold temperature; the thermal imager records the temperature change of the ingot from pouring to cooling through an automatic tracking system. In the forging process, the infrared thermometer and the thermal imager are respectively fixed around the 7000T and 18500T hydraulic presses to measure the surface temperature of the forgings in real time; the multi-point thermocouple array is arranged at different heights and regions in the heating furnace to form a temperature monitoring network. In the heat treatment process, the multi-point thermocouple array is distributed on the top, bottom and four walls of the heat treatment furnace to monitor the temperature distribution inside the furnace; the infrared thermometer and the thermal imager are set up in the area where the workpiece exits the furnace to record the temperature curve during the cooling process of the workpiece. The data collected by these sensors are transmitted to the data processing system through the industrial Ethernet and the 5G wireless network, with a collection frequency not lower than 100 ms, and the Kalman filter algorithm is used to eliminate measurement noise and errors. This algorithm processes the sensor data through two stages of prediction and correction. First, it predicts the current state based on the previous state, and then corrects the predicted value according to the actual measurement results, so as to obtain a more accurate temperature estimate. The temperature data after filtering is reconstructed into a continuous temperature field distribution through a three-dimensional interpolation algorithm, and finally a complete spatio-temporal temperature distribution information including temperature values, three-dimensional spatial coordinates and timestamps is obtained.

[0016] The temperature distribution information in space and time is segmented according to the process stages. The ratio of the temperature difference between adjacent time points to the time interval at each measurement point is calculated to obtain the temperature change rate curve. For each point on the curve, its first derivative value is calculated. When the derivative changes from positive to negative or from negative to positive, it is marked as an inflection point, and these inflection points represent the key temperature transition moments in the hot processing process. For example, in the heat treatment stage, when the workpiece changes from heating to holding, the temperature change rate drops sharply from a positive value to near zero, and the inflection point at this time is marked as the heating completion point; when the workpiece changes from holding to cooling, the rate changes from near zero to negative, and the inflection point at this time is marked as the cooling start point. A thermal map is constructed based on the temperature distribution information in space and time. The thermal map indicates the temperature level through the shade of color, with red representing the high-temperature area and blue representing the low-temperature area. The peak search algorithm is applied to the thermal map. This algorithm compares the temperature values of each point with those of its neighboring points. When the temperature of a certain point is higher than all neighboring points, it is marked as a local maximum point, and when the temperature of a certain point is lower than all neighboring points, it is marked as a local minimum point, thus obtaining the distribution map of temperature extreme points. The temperature gradient vector between the extreme points is calculated. The direction of the gradient vector points to the direction of the fastest temperature rise, and the magnitude represents the intensity of the temperature change, thereby obtaining the information on the direction and intensity of the temperature change. Statistical parameters are calculated for the spatial temperature field at each time point, including the standard deviation (reflecting the degree of dispersion of the temperature distribution), skewness (reflecting the asymmetry of the distribution), and kurtosis (reflecting the peakedness of the distribution). These parameters together constitute the temperature uniformity index. Finally, the temperature change rate, extreme point distribution, gradient vector, and statistical parameters are fused into a multi-dimensional feature vector to form the key temperature characteristic parameters of the hot processing process.

[0017] For the smelting process, parameters such as arc power, vacuum degree, bottom blowing gas flow rate, and holding time are selected; for the casting process, parameters such as pouring temperature, pouring speed, mold preheating temperature, and riser size are selected; for the forging process, parameters such as heating temperature, forging deformation amount, forging speed, and die temperature are selected; for the heat treatment process, parameters such as heating rate, holding temperature, holding time, and cooling rate are selected. The orthogonal experimental design method is used to reasonably arrange the experimental scheme, significantly reducing the number of experiments. For example, for the heat treatment process, if each of the four parameters has three levels, a full factorial experiment requires 81 tests, while using an orthogonal table only 9 tests are needed to analyze the main effects. In each group of experiments, temperature data is collected through a multi-sensor network, key temperature characteristic parameters are extracted, and a process parameter-temperature characteristic response data set is established. Variance analysis is performed on this data set to calculate the contribution rate and significance level of each process parameter to the temperature characteristic parameters. The contribution rate represents the degree of explanation of the parameter to the variation of the result, and the significance level represents whether the parameter effect has statistical significance. Based on the results of variance analysis, an influence weight coefficient is constructed, and the polynomial regression method is used to fit the non-linear relationship between the process parameters and the temperature characteristic parameters to obtain the process-temperature influence function. The partial derivative of this function is calculated to obtain the Jacobian matrix, and each element in this matrix represents the degree of influence of a small change in the process parameter on a specific characteristic of the temperature field. The Jacobian matrix is normalized to obtain a quantitative relationship matrix, which clearly describes the corresponding relationship between the adjustment amount of the process parameter and the change amount of the temperature field.

[0018] Convert the product quality indicators into temperature control targets. For example, in the forging process, the temperature difference between the inside and outside of the forging is required not to exceed 50°C. In the heat treatment process, the heating rate is required not to exceed 150°C / hour and the temperature fluctuation range is controlled within ±5°C. Based on the quantitative relationship matrix, construct a calculation formula for predicting the temperature field state. This calculation formula can predict the evolution sequence of the temperature field in the future for a certain period of time under given process parameters. The core idea of the rolling horizon optimization algorithm is that in each control cycle, only the control quantity of the first time step is executed, and then the prediction window is slid for re-optimization. Specifically, the algorithm first sets the prediction horizon (such as 30 minutes) and the control horizon (such as 5 minutes). At the beginning of each control cycle, based on the current temperature state and the quantitative relationship matrix, predict the evolution trajectory of the temperature field under different process parameter combinations within the future prediction horizon, and calculate the weighted error from the target temperature curve. Solve the optimal process parameter combination through the nonlinear programming method to minimize the prediction error. Then only execute the parameter values of the first control cycle in the optimal process parameter combination. After the end of this cycle, update the model state according to the actual temperature feedback, and slide the prediction window to re-perform the optimization calculation. Through this rolling optimization method, the algorithm can adapt to the changes in working conditions and model errors, and achieve precise control of the temperature field. Finally, convert the sequence of process parameters such as the optimized heating power, cooling rate, and holding time into a continuous parameter control curve that meets the execution accuracy of the equipment through piecewise linear interpolation calculation, and generate standard temperature control instructions according to the requirements of the equipment control interface.

[0019] Match and analyze the temperature control instructions and historical execution data, and extract the instruction execution delay characteristics, temperature response characteristics, and equipment status characteristics. The execution delay characteristics are the execution delay time and actual duration calculated by recording the issuance time, start execution time, and end execution time of historical instructions. The temperature response characteristics are the temperature change rate, temperature response time, and temperature stability parameters extracted by time-aligning the historical temperature data with the control instructions. The equipment status characteristics are the equipment start-stop status, workload, maintenance cycle, and fault record information extracted from the equipment operation log. Based on these characteristics, a long short-term memory neural network is constructed, which includes an input layer, two long short-term memory hidden layers, a fully connected layer, and an output layer. The input layer receives the instruction feature vector. The first hidden layer contains 128 neurons for extracting temporal features. The second hidden layer contains 64 neurons for feature fusion. The fully connected layer contains 32 neurons. The output layer generates the execution timing offset and predicted values of the execution duration. Convert the temperature control instruction into a feature vector and input it into the neural network, and obtain the execution prediction result through forward propagation. According to the prediction result, perform temporal rearrangement and parameter adjustment on the instruction sequence to eliminate execution conflicts and optimize the execution order. Finally, evaluate the temperature target achievement degree of the corrected instruction, calculate the temperature evolution trajectory using the heat conduction model, and perform compensation adjustment on the part deviating from the target to obtain the final target control instruction.

[0020] In the embodiment of the present application, the temperature data of each process of smelting, casting, forging, and heat treatment are collected through a multi-sensor network, realizing the comprehensive acquisition of the spatio-temporal temperature distribution information of the entire hot processing process, and overcoming the problem of insufficient temperature field information caused by limited temperature measurement points in traditional methods; calculating key temperature characteristic parameters such as temperature change rate, temperature extreme point, and temperature uniformity index based on the obtained spatio-temporal temperature distribution information, providing quantitative indicators for temperature field analysis, and enhancing the accuracy and comprehensiveness of temperature field analysis; establishing a quantitative relationship matrix of the influence of process parameters on the temperature field through orthogonal test analysis, transforming the traditional temperature control relying on experience into precise control based on data and models, and realizing the two-way mapping from process parameters to the temperature field and from the temperature field to process parameters; using a rolling horizon optimization algorithm to calculate the optimal combination of heating power, cooling rate, and holding time, and generating temperature control instructions, upgrading the temperature control strategy from static optimization to dynamic optimization, being able to respond to working conditions changes in real time, and improving the adaptability and accuracy of control; predicting the execution timing and execution duration of temperature control instructions and correcting the control instructions according to the prediction result, solving the problem of lag and deviation in the execution of control instructions, and significantly improving the execution accuracy of control instructions.

[0021] In a specific embodiment, the process of executing step S101 may specifically include the following steps: In the smelting process, an infrared thermometer, a thermocouple, and a thermal imager are set up to measure the surface temperature of the molten metal, the temperature inside the furnace, and the overall temperature distribution, obtaining the temperature data of the smelting process; In the casting process, an infrared thermometer is set up to measure the pouring temperature, a thermocouple array is embedded in the mold to monitor the mold temperature, and a thermal imager is used to track the cooling process of the ingot, obtaining the temperature data of the casting process; In the forging process, an infrared thermometer and a thermal imager are deployed to measure the surface temperature of the forging, and a multi-point thermocouple array is installed to monitor the temperature field of the heating furnace, obtaining the temperature data of the forging process; In the heat treatment process, a multi-point thermocouple array is arranged to monitor the temperature distribution of the heat treatment furnace, and an infrared thermometer and a thermal imager are used to record the cooling curve of the workpiece after it comes out of the furnace, obtaining the temperature data of the heat treatment process; The temperature data of the smelting process, the casting process, the forging process, and the heat treatment process are transmitted to the data processing system through an industrial control network, and the Kalman filtering algorithm is used to eliminate measurement errors, obtaining the fused temperature data; Interpolation calculation and three-dimensional reconstruction are performed on the fused temperature data to construct the continuous spatio-temporal temperature distribution information of the entire hot processing process. The spatio-temporal temperature distribution information includes three dimensions: temperature value, spatial coordinates, and timestamp.

[0022] Specifically, in the smelting process, the infrared thermometer is fixedly installed about 2 meters above the electric arc furnace, adopting a non-contact measurement method to accurately capture the surface temperature of the molten metal. The measurement range is 1000°C - 1800°C, and the sampling frequency is 10Hz; the thermocouple selects a high-temperature-resistant K-type thermocouple, which is embedded in different positions of the furnace wall to form a measurement network for monitoring the temperature distribution inside the furnace. The measurement range is 20°C - 1600°C, and the sampling frequency is 1Hz; the thermal imager is erected at a safe position on the edge of the smelting area, equipped with a cooling system and a dust-proof device. The scanning frequency is 50Hz, and the resolution is 320×240 pixels, used to obtain the overall temperature distribution information. The three sensors work together to form complementary measurements. The infrared thermometer focuses on measuring the surface temperature, the thermocouple measures the internal temperature, and the thermal imager provides the global temperature distribution, jointly constituting a complete smelting temperature monitoring network. In the casting process, the infrared thermometer is fixed above the pouring area, focusing on the pouring stream to measure the pouring temperature in real time, and the sampling frequency is increased to 20Hz to ensure capturing temperature fluctuations; the thermocouple array is embedded in the mold wall according to the computer-optimized layout, and the number depends on the mold size. Generally, 50 - 100 measuring points are arranged for large molds, forming a three-dimensional grid structure with a spacing of 10 - 30cm, penetrating to different depths inside the mold to comprehensively monitor the temperature changes in each area of the mold; the thermal imager, through an automatic tracking system, continuously scans the entire process from the pouring of the ingot to its complete cooling. The scanning range covers the entire surface of the ingot, recording the temperature change curve and the temperature distribution change during the cooling process, providing data support for the analysis of the solidification process.

[0023] In the forging process, infrared thermometers are installed at key locations in the forging operation area, such as around the press, to measure the surface temperature of the forgings, with a sampling frequency of 5Hz; thermal imagers are set up above the forging equipment to provide temperature distribution information on the entire surface of the forgings, with a scanning frequency of 30Hz; multi-point thermocouple arrays are installed on the top, bottom and four walls of the heating furnace to form a three-dimensional temperature measurement grid to monitor the temperature field distribution in the heating furnace. The spacing between thermocouples is 20-50cm, and the number is determined according to the size of the furnace, usually 30-80. The coordinated use of these three sensors ensures the comprehensive collection of temperature data during the forging process, monitoring both the heating process and the temperature changes during the forming process.

[0024] In the heat treatment process, multi-point thermocouple arrays are installed at different positions in the heat treatment furnace according to the optimized layout, covering all areas in the furnace. The number of thermocouples is usually 20-60, which is determined according to the size of the furnace body and arranged in a three-dimensional grid structure to monitor the temperature distribution in the heat treatment furnace; infrared thermometers and thermal imagers are fixed in the furnace discharge area, and the infrared thermometer is aimed at the specific point of the furnace discharge workpiece. The thermal imager provides the surface temperature distribution image of the entire workpiece. The two cooperate to record the temperature change curve of the workpiece from the furnace discharge to the completion of cooling. The sampling frequencies are 2Hz and 20Hz respectively. The temperature data collected in each process is transmitted to the central data processing system through the industrial control network. The industrial control network adopts a two-layer architecture. The field layer uses industrial Ethernet with a transmission rate of 100Mbps, covering the fixed equipment area; the control layer uses a 5G wireless network with a transmission rate of 1Gbps, covering the mobile device area. The data collection frequency is dynamically adjusted according to the characteristics of the process stage. It is increased to 100ms / time at critical moments (such as the start of pouring and the start of forging), and reduced to 1s / time in the stable stage, ensuring that key temperature changes are not missed while the data volume is controllable.

[0025] After the data is transmitted to the processing system, it is first preprocessed, including outlier removal, noise filtering and data standardization. The 3σ criterion is used for outlier removal, that is, when the data deviates from the mean by more than 3 times the standard deviation, it is judged as an outlier and removed. The Kalman filter algorithm is used for noise filtering. This algorithm combines the measured value and the predicted value in a recursive manner to minimize the covariance of the estimation error and obtain the optimal estimate. When processing temperature data, the Kalman filter algorithm first establishes the system state equation and the measurement equation. The state equation describes the evolution of temperature over time, and the measurement equation describes the relationship between the measured value and the true value. Then, through the prediction step, the current state is predicted based on the state at the previous moment; then, through the correction step, the prediction result is corrected using the current measurement result to obtain the optimal estimate of the current state. This prediction-correction iterative process effectively eliminates random errors and systematic errors in sensor measurement and improves the accuracy of temperature data.

[0026] Interpolate and perform three-dimensional reconstruction on the temperature data processed by Kalman filtering to construct continuous spatio-temporal temperature distribution information. Kriging interpolation method is used for interpolation calculation. This method is based on the theory of regionalized variables, considering the spatial position relationship of sample points and the spatial autocorrelation of data, and performs the best linear unbiased estimation on the values of unsampled points. Three-dimensional reconstruction is based on octree space division and marching cubes algorithm, which constructs discrete temperature data points into a continuous three-dimensional temperature field. The finally obtained spatio-temporal temperature distribution information includes three dimensions: temperature value (i.e., the temperature value of each spatial point at a specific moment), spatial coordinates (i.e., the three-dimensional position information of the temperature point, referenced to the device coordinate system), and timestamp (i.e., the acquisition time of the temperature data, accurate to the millisecond level).

[0027] In a specific embodiment, the process of executing step S102 may specifically include the following steps: Perform time series segmentation processing on the spatio-temporal temperature distribution information, calculate the ratio of the temperature difference between adjacent time points to the time interval, and obtain the temperature change rate curve of each measurement point; Apply the derivative extreme value detection algorithm to the temperature change rate curve, mark the inflection points with significant changes in the slope of the rate curve, and obtain the key temperature transition moments during the hot processing process; Construct a heat map based on the spatio-temporal temperature distribution information, use the peak search algorithm to identify the coordinates of local maximum and minimum points in the temperature field, and obtain the temperature extreme point distribution map; Calculate the temperature gradient between extreme points for the temperature extreme point distribution map, construct a temperature gradient vector field, and obtain the temperature change direction and intensity during the hot processing process; According to the spatio-temporal temperature distribution information, calculate the standard deviation, skewness, and kurtosis of the spatial temperature field for each time point, and obtain the statistical parameters characterizing the temperature distribution uniformity; Fuse the temperature change rate curve, key temperature transition moments, temperature extreme point distribution map, temperature gradient vector field, and statistical parameters to construct a multi-dimensional feature vector, and obtain the key temperature characteristic parameters of the hot processing process.

[0028] Specifically, perform time-series segmentation on the obtained spatio-temporal temperature distribution information, that is, divide the temperature data into different time periods according to the process stages. For the smelting process, it is divided into stages such as stock preparation, rough smelting, refining, and vacuum treatment; for the casting process, it is divided into stages such as pouring, solidification, and cooling; for the forging process, it is divided into stages such as heating, forging, and cooling; for the heat treatment process, it is divided into stages such as heating, holding, quenching, and tempering. Within each time period, calculate the ratio of the temperature difference between adjacent time points to the time interval to obtain the temperature change rate. The specific operation is to take the temperature values and time values of two adjacent time points for each measurement point, calculate their difference, and the temperature difference divided by the time difference is the temperature change rate. By performing this calculation for all time points, a temperature change rate curve for each measurement point is generated. This curve intuitively reflects the speed of temperature change during the hot processing, and is an important basis for analyzing process transitions. Apply the derivative extreme value detection algorithm to the temperature change rate curve to identify the inflection points with significant slope changes on the curve. These inflection points usually correspond to the key temperature transition moments during the hot processing. The core of the derivative extreme value detection algorithm is to calculate the change rate of the rate curve, that is, the change trend of the rate. In actual operation, the derivative value is approximated by calculating the difference between each point on the curve and the points before and after it. When the derivative value changes from positive to negative or from negative to positive, and the change amplitude exceeds the preset threshold, this point is marked as an inflection point. These inflection points correspond to the critical moments of process transitions, such as the heating completion point, the holding start point, the cooling start point, etc., providing a time reference for formulating subsequent temperature control strategies.

[0029] Construct a thermal map based on the spatio-temporal temperature distribution information. The thermal map is an intuitive image that represents the temperature level through the shade of color. Usually, red is used to represent the high-temperature area and blue is used to represent the low-temperature area. During the construction process, first divide the three-dimensional space into grid cells, and each cell is assigned the temperature value obtained from the measured points or interpolation calculation. Then, map the temperature value to a predefined color space to generate a color thermal map. Based on the thermal map, use the peak search algorithm to identify the local highest and lowest points in the temperature field. The peak search algorithm determines whether a point is an extreme point by comparing the temperature value of each point with the temperature values of its surrounding points. If the temperature value of a point is higher than all its adjacent points, it is marked as a local highest point; if it is lower than all adjacent points, it is marked as a local lowest point. To avoid noise interference, usually set a minimum temperature difference threshold, and only when the temperature difference exceeds the threshold is it marked as an extreme point. The obtained temperature extreme point distribution map intuitively shows the key areas where heat is concentrated or dissipated during the hot processing.

[0030] Calculate the temperature gradient between the extreme points for the temperature extreme point distribution map, and construct a temperature gradient vector field. The temperature gradient reflects the spatial rate of change and direction of temperature, and is a direct indicator of the intensity and direction of heat conduction. For any two extreme points in space, calculate the temperature difference between them divided by the spatial distance to obtain the magnitude of the temperature gradient, and the direction points from the low-temperature point to the high-temperature point. The specific operation is to calculate the three-dimensional spatial distance between two points (considering the distances on the x, y, and z coordinate axes), then divide the temperature difference between the two points by this distance to obtain the temperature gradient value, and record the direction vector from the low-temperature point to the high-temperature point at the same time. Combine the gradient vectors of all pairs of extreme points to form a temperature gradient vector field, which clearly shows the direction and intensity of heat flow during the hot processing, providing a decision-making basis for the control of temperature field uniformity.

[0031] According to the spatio-temporal temperature distribution information, calculate the statistical parameters of the spatial temperature field for each time point, including standard deviation, skewness, and kurtosis. The standard deviation calculation is to take the square root of the average of the sum of the squares of the differences between the temperature values of all spatial measurement points at the same time point and their average temperature, which reflects the degree of dispersion of the temperature distribution. The skewness calculation involves the cube of the temperature deviation from the average value, which reflects the asymmetry of the temperature distribution. A positive skewness indicates that the high-temperature region is concentrated, and a negative skewness indicates that the low-temperature region is concentrated. The kurtosis calculation involves the fourth power of the temperature deviation from the average value, which reflects the sharpness of the temperature distribution. A high kurtosis indicates a sharp temperature distribution, and a low kurtosis indicates a flat temperature distribution. These three statistical parameters together constitute the temperature uniformity index, comprehensively characterizing the degree of uniformity of the temperature distribution, and providing a quantitative index for the control of temperature field uniformity during the hot processing.

[0032] Perform feature fusion on the temperature change rate curve, key temperature transition moments, temperature extreme point distribution map, temperature gradient vector field, and statistical parameters to construct a multi-dimensional feature vector and obtain the key temperature characteristic parameters of the hot processing. Feature fusion first normalizes various features to convert features with different measurement units to the same range; then assigns different weights to different features according to process requirements to highlight the influence of important features; finally, combines the weighted features into a multi-dimensional vector to form the key temperature characteristic parameters of the hot processing. These characteristic parameters comprehensively reflect the dynamic change law and spatial distribution characteristics of the temperature field during the hot processing, providing a data basis for subsequent process parameter optimization and temperature control strategy formulation.

[0033] In a specific embodiment, the process of executing step S103 may specifically include the following steps: Select key process parameter groups for the smelting, casting, forging, and heat treatment processes respectively, design an orthogonal test table, and obtain an experimental plan for process parameter combinations; Carry out hot processing tests according to the experimental scheme of process parameter combinations, collect the spatio-temporal temperature distribution information of each group of experiments, extract key temperature characteristic parameters, and obtain a process parameter-temperature characteristic response data set; Conduct an analysis of variance on the process parameter-temperature characteristic response data set, calculate the contribution rate and significance level of each process parameter to the temperature characteristic parameters, and obtain the main effect analysis results; Based on the main effect analysis results, construct the influence weight coefficients of process parameters on the temperature field, use polynomial regression to fit the non-linear relationship between process parameters and temperature characteristic parameters, and obtain the process-temperature influence function; Calculate the partial derivatives of the process-temperature influence function with respect to process parameters to generate a Jacobian matrix, and obtain the sensitivity distribution of the response of the temperature field to changes in process parameters; Perform normalization processing on the sensitivity distribution, construct the forward mapping relationship and inverse solution relationship between process parameters and the temperature field, generate a quantitative relationship matrix, and the quantitative relationship matrix contains the corresponding relationship between the process parameter adjustment amount and the temperature field change amount.

[0034] Specifically, select key process parameter groups for the smelting, casting, forging, and heat treatment processes respectively. For the smelting process, the key process parameters include arc power, vacuum degree, bottom blowing gas flow rate, and heat preservation time; the key parameters for the casting process include pouring temperature, pouring speed, mold preheating temperature, and riser size; the key parameters for the forging process include heating temperature, forging deformation amount, forging speed, and die temperature; the key parameters for the heat treatment process include heating rate, heat preservation temperature, heat preservation time, and cooling rate. For these parameters, the orthogonal experimental design method is used for experimental arrangement. The orthogonal experimental design is an efficient experimental method. Through carefully designed experimental combinations, the most information can be obtained with the least number of experiments. The specific operation is to first determine the value levels of each parameter, usually 3-5 levels, and then select a suitable orthogonal table according to the number of parameters and the number of levels. For example, for the 4 parameters of the heat treatment process, if each parameter is set at 3 levels, only 9 experiments are required using the L9 orthogonal table, while a full factor design requires 81 experiments. Each row of the orthogonal table represents a group of experimental schemes, each column represents a parameter, and the numbers in the table represent the levels of the parameter in that experiment. Through the design of the orthogonal table, a comprehensive experiment of different parameter combinations is ensured, while maintaining the economy of the number of experiments.

[0035] Conduct hot processing tests according to the experimental scheme of process parameter combinations, with the focus on collecting spatio-temporal temperature distribution information for each group of experiments. During the tests, set each process parameter according to the parameter combinations specified in the orthogonal table, and then use a multi-sensor temperature monitoring network to record temperature data throughout the process. For smelting experiments, measure the surface temperature of the molten metal, the temperature inside the furnace, and the overall temperature distribution; for casting experiments, measure the pouring temperature, the mold temperature, and the temperature during the cooling process of the ingot; for forging experiments, measure the surface temperature of the forging and the temperature field of the heating furnace; for heat treatment experiments, measure the temperature distribution of the heat treatment furnace and the cooling curve of the workpiece. After preprocessing the collected temperature data, extract key temperature characteristic parameters, including the temperature change rate, the distribution of temperature extreme points, the temperature gradient distribution, and the temperature uniformity index, etc. After each group of tests is completed, pair the set process parameter values with the extracted temperature characteristic parameter values to form a process parameter-temperature characteristic response data set. This data set is a structured data table, containing columns for process parameters and temperature characteristic parameters. Each row represents the results of a group of tests, recording the temperature characteristic parameter values obtained under a specific process parameter combination.

[0036] Conduct an analysis of variance on the process parameter-temperature characteristic response data set to calculate the contribution rate and significance level of each process parameter to the temperature characteristic parameters. Analysis of variance is a statistical method used to determine the degree of influence of different factors on the variation of the results. Specifically, during implementation, first calculate the average value of the temperature characteristic parameters at different levels of each parameter to obtain a parameter-response relationship table; then calculate the sum of squares caused by each parameter, that is, the sum of the squares of the differences between the average values of each level and the overall average value; then calculate the total sum of squares, that is, the sum of the squares of the differences between all experimental data and the overall average value; the contribution rate is equal to the sum of squares of the parameter divided by the total sum of squares, indicating the degree of explanation of the total variation by this parameter; the significance level is determined through an F-test. Calculate the F value (the ratio of the mean square of the parameter to the mean square of the error) and look up the P value in the table. When the P value is less than 0.05, it is considered that this parameter has a significant influence on the temperature characteristic parameters. Through the analysis of variance, obtain the main effect analysis results, clarify the influence order and degree of each process parameter on the temperature characteristics, and provide a basis for subsequent modeling.

[0037] Based on the results of the main effect analysis, the influence weight coefficients of process parameters on the temperature field are constructed. The influence weight coefficients reflect the relative importance of different process parameters on the temperature field. The calculation method is to normalize the contribution rate of the parameters so that the sum of all weights is 1. After obtaining the weight coefficients, the polynomial regression method is used to fit the non-linear relationship between the process parameters and the temperature characteristic parameters. Polynomial regression is a regression method that can describe non-linear relationships. The basic form is that the temperature characteristic parameters are expressed as polynomial functions of the process parameters, including linear terms, quadratic terms, and interaction terms. During the fitting process, first, the order of the polynomial is set, usually the second or third order; then the least squares method is used to calculate the polynomial coefficients to minimize the mean square error between the predicted value and the actual value; finally, the fitting effect is evaluated through the determination coefficient and residual analysis. The polynomial equation obtained by fitting is the process-temperature influence function, which can predict the output temperature characteristic parameter values based on the input process parameter values, providing model support for temperature control.

[0038] The partial derivatives of the process-temperature influence function with respect to the process parameters are calculated to generate the Jacobian matrix. The Jacobian matrix is a matrix that describes the sensitivity of a multivariable function to its respective independent variables. Each element in the matrix represents the partial derivative of the dependent variable with respect to a certain independent variable. In this method, the rows of the Jacobian matrix correspond to different temperature characteristic parameters, and the columns correspond to different process parameters. The matrix elements represent the sensitivity of a specific temperature characteristic parameter to the change of a specific process parameter. The calculation method is to take the partial derivatives of the fitted process-temperature influence function with respect to each process parameter. The larger the partial derivative value, the more sensitive the temperature characteristic parameter is to the change of the process parameter. For example, if the partial derivative of the heat treatment holding temperature with respect to the temperature difference between the inside and outside of the workpiece is large, it means that a small change in the holding temperature will cause a significant change in the temperature difference between the inside and outside of the workpiece, and precise control is required. Through the calculation of the Jacobian matrix, the sensitivity distribution of the process parameter changes to the temperature field response is obtained, providing an accurate quantitative basis for formulating temperature control strategies.

[0039] Normalize the sensitivity distribution to construct the forward mapping relationship and inverse solution relationship between process parameters and the temperature field. Normalization is to convert the elements in the Jacobian matrix to the same scale for easy comparison and calculation. The commonly used normalization method is the maximum-minimum normalization, which maps the element values to the 0-1 interval. The calculation formula is to divide the difference between the element value and the minimum value by the difference between the maximum value and the minimum value. The normalized Jacobian matrix clearly shows the relative influence degree of different process parameters on different temperature characteristics. Based on the normalized matrix, construct the forward mapping relationship between process parameters and the temperature field, that is, given a combination of process parameters, predict the temperature field characteristics; at the same time, construct the inverse solution relationship, that is, given the target temperature field characteristics, reverse infer the required combination of process parameters. The inverse solution usually uses an optimization algorithm, sets the objective function as the difference between the predicted temperature field and the target temperature field, and minimizes the objective function by adjusting the process parameters. The finally generated quantitative relationship matrix contains the corresponding relationship between the process parameter adjustment amount and the temperature field change amount.

[0040] In a specific embodiment, the process of executing step S104 may specifically include the following steps: Parametrically express the requirements of the hot processing process, convert the product quality index into a temperature control target, and obtain the target value and constraint conditions for the optimal solution; Based on the quantitative relationship matrix, construct a calculation formula for predicting the temperature field state, input the temperature characteristic parameters and candidate process parameters at the current moment, calculate the evolution sequence of the temperature field within the prediction time domain, and obtain the temperature response prediction result; Calculate the error between the temperature response prediction result and the target temperature curve, construct a weighted quadratic performance index evaluation method, and obtain the evaluation index for process parameter optimization; Based on the evaluation index, use the nonlinear programming solution method to perform rolling optimization on the process parameters. Only execute the control amount of the first time step each time, and then slide the prediction window for re-optimization to obtain a dynamically updated sequence of process parameters; According to the sequence of process parameters, perform piecewise linear interpolation calculation on the heating power, cooling rate, and holding time to obtain a continuous parameter control curve that meets the equipment execution accuracy; Convert the continuous parameter control curve according to the requirements of the equipment control interface and arrange the timing to generate a standard temperature control instruction including the execution time, execution object, and execution value.

[0041] Specifically, parametric expression refers to the conversion of product quality indicators into quantifiable temperature control targets, including temperature control accuracy, temperature field uniformity, heating and cooling rates, and the control range of key temperature points. In specific implementation, first, collect the quality requirements in the product technical standards, such as the grain size, tissue uniformity, mechanical properties of forgings, etc. Then, through material science theories and practical experience, convert these quality requirements into temperature control parameters. For example, for large alloy steel forgings, if the required grain size is 4 - 6 levels, it can be converted to the heat treatment temperature being controlled within the range of 880 ± 10 °C; if high tissue uniformity is required, it can be converted to the temperature difference between the inside and outside of the workpiece not exceeding 50 °C; if no heat treatment cracks are required, it can be converted to the heating rate not exceeding 150 °C / hour and the cooling rate not exceeding 80 °C / hour. These specific temperature control parameters constitute the target values and constraint conditions for the optimization solution. The target value is usually the temperature control accuracy (such as the furnace temperature fluctuation range of ± 5 °C), and the constraint conditions include the upper and lower temperature limits, the maximum temperature difference, the maximum heating and cooling rates, etc. Based on the quantitative relationship matrix established previously, construct a temperature field state prediction calculation formula to achieve the prediction of the temperature field change in the future time period. The quantitative relationship matrix is the corresponding relationship between the change amount of process parameters and the change amount of the temperature field. Through this matrix, the influence of specific process parameter adjustments on the temperature field can be calculated. The construction process of the temperature field state prediction calculation formula is as follows: First, determine the prediction time step and prediction time domain. The time step is usually 5 - 10 minutes, and the prediction time domain is 30 - 60 minutes. Then, take the temperature characteristic parameters at the current moment as the initial state, substitute the candidate change amounts of process parameters into the quantitative relationship matrix, and calculate the change amount of the temperature field at the next time step. Add the change amount of the temperature field to the current temperature field to obtain the temperature field state at the next time step. And so on, step by step, to obtain the temperature field evolution sequence within the entire prediction time domain. During the prediction calculation process, physical characteristics of the hot working process, such as thermal inertia, heat transfer lag, etc., are considered, and the prediction accuracy is adjusted through correction factors. The finally generated temperature response prediction result is a time series, indicating the temperature field change trend in the future period under the given process parameters.

[0042] Calculate the error between the predicted temperature response result and the target temperature curve, and construct a performance index to evaluate the quality of process parameters. The error calculation first compares the values of the predicted temperature curve and the target temperature curve at the same time point to calculate the temperature difference; then adopts the weighted quadratic performance index evaluation method, that is, multiplies the square of the temperature difference by the weight coefficient at the corresponding time point, and then sums to obtain the total error. The weighted quadratic evaluation method emphasizes the influence of points with larger deviations, and at the same time reflects the importance differences of different time points or different regions through the weight coefficient. The setting principle of the weight coefficient is: high weight for key process transition points, such as the end point of heating and the starting point of cooling; high weight for key parts, such as thick and large parts of the workpiece and hot spot areas; high weight for time periods with high target control accuracy requirements. Through this weighted evaluation method, a comprehensive process parameter optimization evaluation index is obtained. The smaller the value of this index, the closer the combination of process parameters is to the optimal.

[0043] Based on the evaluation index, use the nonlinear programming solution method to perform rolling optimization on the process parameters. Nonlinear programming is a mathematical optimization method used to find the optimal solution of the objective function under nonlinear constraint conditions. In this method, the objective function is the above-mentioned weighted quadratic evaluation index, and the constraint condition is the reasonable adjustment range of the process parameters. The core idea of rolling optimization is that in each control cycle, only the control quantity of the first time step is executed, and then the prediction window is slid for re-optimization. When specifically implemented, first set the prediction time domain (such as 30 minutes) and the control time domain (such as 5 minutes). At the beginning of each control cycle, calculate the optimal process parameter sequence within the entire prediction time domain through the nonlinear programming algorithm; then only execute the parameter values of the first control time domain (5 minutes); after the execution is completed, use the new temperature field state as the initial state, slide the prediction window, and recalculate the optimal parameter sequence. This rolling optimization method can respond to actual temperature changes in a timely manner, adapt to working condition fluctuations and model errors, and achieve precise control of the temperature field. Through continuous rolling optimization, a dynamically updated process parameter sequence is obtained, and this sequence contains the optimal parameter values of each control cycle.

[0044] According to the optimized process parameter sequence, perform piecewise linear interpolation calculations on key control parameters such as heating power, cooling rate, and holding time to obtain a continuous parameter control curve. Piecewise linear interpolation means that between two adjacent control points, assuming that the parameter values change linearly, the parameter values at intermediate time points are obtained through interpolation calculations. During specific implementation, ensure that the time interval of the interpolation points meets the requirements of the device control accuracy, usually at the second or minute level. For example, for the heating power, if the times of two adjacent control points are t1 and t2, and the power values are P1 and P2, then the power value P at any moment t between t1 and t2 can be calculated by linear interpolation: P = P1 + (P2 - P1) * (t - t1) / (t2 - t1). In this way, the discrete process parameter sequence is converted into a continuous control curve to ensure a smooth transition of temperature control and avoid temperature fluctuations caused by parameter mutations. Convert the continuous parameter control curve according to the requirements of the device control interface for protocol conversion and timing arrangement to generate standard temperature control instructions. Protocol conversion means converting the data format of the control curve into a communication protocol format recognizable by a specific device, such as Modbus, OPC UA, TCP / IP, etc. Timing arrangement means determining the execution order and time points of each control instruction to ensure that the instructions are executed according to the predetermined plan. The standard temperature control instructions contain three core elements: execution time (when the instruction should be executed), execution object (which device or control unit the instruction is targeted at), and execution value (the specific set value of the control parameter). For example, a complete control instruction may be: at 10:30:00, set the power of heating zone 1 to 65 kW; at 10:35:00, set the flow rate of the cooling system to 120 L / min. These standardized control instructions are sent to the control systems of each execution device through the industrial control network to achieve automated temperature control.

[0045] In a specific embodiment, the process of executing step S105 may specifically include the following steps: Match and analyze the temperature control instructions with the historical execution data, extract the instruction execution delay characteristics, temperature response characteristics, and device status characteristics to obtain an instruction execution feature set; Construct a long short-term memory neural network based on the instruction execution feature set. The long short-term memory neural network includes an input layer, two long short-term memory hidden layers, a fully connected layer, and an output layer. The input layer receives the instruction feature vector. The first hidden layer contains 128 neurons for extracting timing features. The second hidden layer contains 64 neurons for feature fusion. The fully connected layer contains 32 neurons. The output layer generates prediction values for the execution timing and duration; Convert the temperature control instructions into feature vectors, input them into the long short-term memory neural network, and through forward propagation calculation, obtain the execution timing offset and execution duration prediction values for each instruction; Perform device constraint verification on the execution timing offset and the predicted execution duration value, and correct the prediction result according to the response characteristics and physical limitations of the hot processing equipment to obtain an executable execution prediction result; Based on the execution prediction result, perform timing rearrangement and parameter adjustment on the temperature control instruction sequence to eliminate instruction execution conflicts and optimize the instruction execution order to obtain a preliminarily corrected control instruction; Evaluate the temperature target achievement degree of the preliminarily corrected control instruction, calculate the corrected temperature evolution trajectory using the heat conduction model, and perform compensation adjustment on the part deviating from the target to obtain the final target control instruction.

[0046] Specifically, it is necessary to perform matching analysis on the temperature control instruction and the historical execution data to extract three types of key features. The instruction execution delay feature refers to the time difference feature between the issuance of the control instruction and its actual execution. By performing time tokenization on the historical temperature control instruction database, record the issuance time, start execution time, and end execution time of each instruction, and calculate the execution delay time and the actual duration. The time tokenization method is to set time stamps for each instruction in the database, including the instruction generation time stamp, the instruction start execution time stamp, and the instruction execution completion time stamp, and calculate the execution delay time and the actual duration through the differences between these time stamps. The temperature response feature refers to the feature of temperature change after the execution of the control instruction. By aligning the historical temperature data with the control instruction in time, extract the temperature change rate, temperature response time, and temperature stability parameters before and after the instruction execution. The time alignment method is to synchronize the time series of the temperature data with the time series of the instruction execution to determine which instruction execution stage the temperature value at each time point corresponds to. The device status feature refers to the feature of the impact of the device operation status on the instruction execution. Based on the device operation log, extract the device start / stop status, workload, maintenance cycle, and fault record information, and construct a device status feature vector. These three types of features form a comprehensive feature association table through association mapping, and then through principal component analysis and feature importance ranking, select the feature subset with the highest correlation to generate an instruction execution feature set.

[0047] Based on the extracted instruction execution feature set, a long short-term memory neural network is constructed to predict the instruction execution timing and duration. The long short-term memory neural network is a special type of recurrent neural network, which is good at processing time series data and capturing long-term dependencies. The network structure includes an input layer, two long short-term memory hidden layers, a fully connected layer, and an output layer. The input layer receives the instruction feature vector, which contains information such as instruction type, parameter size, and device status. After normalization, it is input into the network. The first hidden layer contains 128 neurons, which are mainly responsible for extracting time series features and capturing the dependencies between instruction executions at different time points. The long short-term memory unit controls the information flow through three gating structures (input gate, forget gate, and output gate), and can effectively remember long-term time series patterns. The second hidden layer contains 64 neurons, which are mainly responsible for feature fusion, integrating the extracted time series features with the device status features. The fully connected layer contains 32 neurons, which perform dimensionality reduction and non-linear transformation on the features. Finally, the output layer generates the execution timing offset and execution duration prediction values. The network is trained using a historical data set, with the mean squared error as the loss function, and the network parameters are optimized through the backpropagation algorithm until the loss function converges.

[0048] Convert the temperature control instruction into a feature vector and input it into the long short-term memory neural network. Through forward propagation calculation, the execution timing offset and execution duration prediction values of each instruction are obtained. Feature vector conversion encodes the various attributes of the instruction into a numerical form, including instruction type (such as heating power adjustment, cooling rate adjustment), parameter size, device type, current device status, etc. The encoding method combines one-hot encoding and numerical normalization, converting categorical variables into one-hot vectors and normalizing continuous variables to a specific range. After the feature vector is input into the neural network, through a series of forward calculations, the prediction results are finally obtained. The forward propagation calculation process is as follows: the input layer receives the feature vector and passes it to the first long short-term memory layer through the weight matrix; the first long short-term memory layer processes the time series information and passes the output to the second long short-term memory layer; the second long short-term memory layer performs feature fusion and passes the output to the fully connected layer; the fully connected layer performs dimensionality reduction and non-linear transformation, and finally the output layer generates two prediction values: the execution timing offset (the difference between the actual execution time and the planned execution time of the instruction) and the execution duration (the time required for the actual execution of the instruction).

[0049] Perform device constraint verification on the execution timing offset and the predicted execution duration to ensure that the prediction results conform to the actual operating characteristics of the device. Device constraint verification is to check whether the prediction results meet the physical limitations and response characteristics of the hot processing device, preventing the generation of unrealistic control plans. The specific verification content includes: whether the instruction execution timing exceeds the device response time range, such as the power adjustment response time of the smelting furnace is not less than 30 seconds; whether the instruction execution duration conforms to the sustainable working time of the device, such as continuous maximum power heating does not exceed 60 minutes; whether the instruction execution parameters are within the device capacity range, such as the heating power does not exceed the rated power of the device. After verification, correct the prediction results that do not meet the constraint conditions, adjust the execution timing offset and the execution duration to conform to the device characteristics, and obtain executable execution prediction results.

[0050] Based on the execution prediction results, perform timing rearrangement and parameter adjustment on the temperature control instruction sequence to optimize the instruction execution order. Timing rearrangement is to adjust the issuing order of instructions according to the predicted execution timing offset to ensure that each instruction can be executed in the desired time order. The specific method is to construct an instruction dependency graph, analyze the sequential dependency relationship between instructions, and then reorder the instructions without sequential dependency according to the predicted execution timing to eliminate instruction execution conflicts. Parameter adjustment is to fine-tune the control parameters according to the predicted execution duration to ensure that the control effect is not affected by the change in execution time. For example, if it is predicted that the execution of the heating power adjustment instruction is delayed by 5 minutes, then appropriately increase the heating power to compensate for the heat loss caused by the delay. Through timing rearrangement and parameter adjustment, a preliminarily corrected control instruction sequence is obtained.

[0051] Evaluate the temperature target achievement degree of the preliminarily corrected control instructions to ensure that the corrected instruction sequence can achieve the expected temperature control effect. The evaluation method is to use a heat conduction model to calculate the temperature evolution trajectory after the execution of the corrected instruction sequence. The heat conduction model is based on the thermophysical parameters of the material and the finite element method, and can predict the temperature field distribution and changes during the hot processing process. By comparing the calculated temperature trajectory with the target temperature curve, identify the parts that deviate from the target and perform targeted compensation adjustments. The compensation adjustment adopts a feedforward correction method, and adjusts the parameter values or execution timing of relevant instructions according to the magnitude and nature of the deviation to make the temperature trajectory closer to the target curve. After evaluation and compensation adjustment, the final target control instructions are formed for actual temperature control execution.

[0052] In a specific embodiment, the process of performing matching analysis on the temperature control instructions and historical execution data in the execution step may specifically include the following steps: Perform time tokenization on the historical temperature control instruction database, record the issuance time, start execution time, and end execution time of each instruction, calculate the execution delay time and actual duration, and obtain the instruction execution timing characteristics; According to the instruction execution timing characteristics, perform hierarchical clustering analysis on the instruction type, process stage, and equipment load status to obtain the instruction execution delay distribution law under different conditions; Align the historical temperature data with the control instructions in time, extract the temperature change rate, temperature response time, and temperature stability parameters before and after the instruction execution, and obtain the temperature response characteristics; Based on the equipment operation log, extract the equipment start-stop status, workload, maintenance cycle, and fault record information, construct the equipment status feature vector, and obtain the equipment operation status characteristics; Associate and map the instruction execution timing characteristics, instruction execution delay distribution law, temperature response characteristics, and equipment operation status characteristics, establish a multi-dimensional feature association table, and obtain the comprehensive feature association result; Perform principal component analysis and feature importance ranking on the comprehensive feature association result, select the feature subset with the highest correlation, and generate the instruction execution feature set.

[0053] Specifically, time tokenization is performed on the historical temperature control instruction database. This process refers to adding explicit timestamp information to each control instruction. During specific implementation, the records of each control instruction are extracted from the database, including information such as instruction ID, instruction type, parameter values, target device, etc. Then, three key time nodes are added: the instruction issuance time (the moment when the control instruction is generated and sent), the start execution time (the moment when the device actually starts to respond to the instruction), and the end execution time (the moment when the device completes the instruction execution). By calculating the difference between the start execution time and the issuance time, the execution delay time is obtained; by calculating the difference between the end execution time and the start execution time, the actual duration is obtained. These time difference information constitute the timing characteristics of instruction execution, reflecting the response characteristics of the hot processing equipment to different types of control instructions. According to the extracted instruction execution timing characteristics, hierarchical clustering analysis is carried out to identify the distribution laws of instruction execution delays under different conditions. Hierarchical clustering analysis is a bottom-up clustering method. First, each sample is regarded as an independent category, and then similar categories are gradually merged until the termination condition is met. In this method, the clustering analysis is based on three key dimensions: instruction type (such as heating power adjustment, cooling rate adjustment, holding time setting, etc.), process stage (such as the rough smelting stage and refining stage of smelting; the pouring stage and solidification stage of casting; the heating stage and forging stage of forging; the heating stage, holding stage, and cooling stage of heat treatment, etc.), and equipment load status (such as light load, medium load, heavy load, etc.). The clustering process uses the hierarchical aggregation method. First, the distance matrix between samples is calculated (using the Euclidean distance to measure the similarity between samples), and then the closest categories are gradually merged according to the distance matrix to form a dendrogram clustering structure. By setting an appropriate threshold or specifying the number of categories, the final clustering result is obtained. The clustering result shows the distribution laws of instruction execution delays under different condition combinations. For example, in the heating stage of the forging process, when the equipment is in a heavy load state, the execution delay of the heating power increase instruction is significantly higher than that in the light load state.

[0054] Align the historical temperature data with the control instructions in terms of time, and extract the temperature response characteristics. Time alignment is to match the time series of temperature data with the time series of control instruction execution, and determine which instruction execution stage the temperature value at each time point corresponds to. In specific implementation, first arrange the temperature data and control instructions in chronological order, and then divide the temperature data into corresponding segments according to the execution time period of the instructions (from the start of execution to the end of execution). For each segment, extract three types of key characteristics: the temperature change rate (the rate of change of temperature with time during the execution of the instruction, obtained by calculating the ratio of the temperature difference between adjacent time points to the time interval), the temperature response time (the time interval from the start of the instruction execution to the obvious change in temperature, determined by detecting the inflection point of the temperature curve), and the temperature stability parameter (the degree of temperature fluctuation after the instruction execution, expressed by calculating the standard deviation or coefficient of variation of the temperature values). These characteristics together constitute the temperature response characteristics, reflecting the influence patterns of different control instructions on temperature changes.

[0055] Based on the device operation log, extract the device operation status characteristics. The device operation log is a data record that records the changes in the working status of the device, including information such as the device start and stop times, operation parameters, maintenance records, and fault alarms. Extract four types of key information from the log: the device start and stop status (whether the device is in the running state, and the duration of startup or shutdown), the workload (the current workload level of the device, usually expressed by the power load rate or production capacity utilization rate), the maintenance cycle (the time since the last maintenance and the time of the next planned maintenance), and the fault record (the types, frequencies, and handling situations of recent faults). Structurally process this information to form a device status feature vector, which describes the working status of the device at different time points and provides a device-level basis for predicting the execution effect of the instructions.

[0056] Correlate and map the instruction execution timing characteristics, instruction execution delay distribution law, temperature response characteristics, and device operation status characteristics extracted above to establish a multi-dimensional feature correlation table. Correlation mapping is to establish connections between characteristics from different sources according to time and object relationships to form a comprehensive data structure. In specific implementation, first determine the primary key for correlation, usually selecting the instruction ID or timestamp as the correlation benchmark; then match and merge various types of characteristics according to the primary key to ensure that each record contains complete feature information; for the matching in the time dimension, use the time window method to associate the device status characteristics within a specific time window with the instruction execution characteristics. The multi-dimensional feature correlation table is a structured data table, where each row represents an instruction execution instance, and each column represents a feature dimension, including information on multiple dimensions such as instruction type, parameter value, process stage, device status, execution delay, and temperature response.

[0057] Perform principal component analysis and feature importance ranking on the comprehensive feature association results to generate the final instruction execution feature set. Principal component analysis is a dimensionality reduction technique that transforms the original features into a set of mutually orthogonal principal components through linear transformation, retaining the main information of the data while reducing the number of features. The specific operation steps are as follows: First, standardize the data to make features with different dimensions comparable; then calculate the covariance matrix of the features to analyze the correlation between features; next, calculate the eigenvalues and eigenvectors of the covariance matrix, where the size of the eigenvalues represents the importance of the principal components; sort according to the size of the eigenvalues and select the first few principal components whose cumulative contribution rate reaches a preset threshold (usually 85%-95%) as the new feature space. Feature importance ranking is to evaluate the influence degree of each feature on the target variable through machine learning algorithms. Common methods include the feature importance scoring of random forests, the coefficient size of linear models, and the scoring based on information gain, etc. After ranking, select the feature subset with the highest importance to form the final instruction execution feature set. This feature set contains the key information required to predict the instruction execution timing and duration, with a moderate dimension and rich information content.

[0058] The temperature control method for the hot processing process in the embodiments of the present application has been described above. Next, the temperature control system for the hot processing process in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the temperature control system for the hot processing process in the embodiments of the present application includes: An acquisition module 201, configured to acquire the temperature data of each process of smelting, casting, forging, and heat treatment, and obtain the spatio-temporal temperature distribution information of the entire hot processing process; A calculation module 202, configured to calculate the temperature change rate, temperature extreme points, and temperature uniformity index during the hot processing process according to the spatio-temporal temperature distribution information, and obtain key temperature characteristic parameters; An analysis module 203, configured to perform an orthogonal experiment analysis on the process parameters based on the key temperature characteristic parameters, and establish a quantitative relationship matrix of the influence of the process parameters on the temperature field; A generation module 204, configured to calculate the optimal combination of heating power, cooling rate, and heat preservation time according to the quantitative relationship matrix, and generate a temperature control instruction by using a rolling horizon optimization algorithm; A prediction module 205, configured to predict the execution timing and execution duration of the temperature control instruction to obtain an execution prediction result, and correct the control instruction according to the execution prediction result to obtain a target control instruction.

[0059] Through the collaborative cooperation of the above-mentioned various components, temperature data of each process of smelting, casting, forging, and heat treatment are collected through a multi-sensor network, achieving a comprehensive acquisition of the spatio-temporal temperature distribution information of the entire hot processing process and overcoming the problem of insufficient temperature field information caused by limited temperature measurement points in traditional methods; based on the obtained spatio-temporal temperature distribution information, key temperature characteristic parameters such as temperature change rate, temperature extreme points, and temperature uniformity index are calculated, providing quantitative indicators for temperature field analysis and enhancing the accuracy and comprehensiveness of temperature field analysis; through orthogonal experiment analysis, a quantitative relationship matrix of the influence of process parameters on the temperature field is established, transforming the traditional experience-dependent temperature control into precise control based on data and models, and realizing the two-way mapping from process parameters to the temperature field and from the temperature field to process parameters; the rolling horizon optimization algorithm is used to calculate the optimal combination of heating power, cooling rate, and holding time, generating temperature control instructions, upgrading the temperature control strategy from static optimization to dynamic optimization, being able to respond to working conditions changes in real time, and improving the adaptability and accuracy of control; by predicting the execution time and execution duration of temperature control instructions and correcting the control instructions according to the prediction results, the problems of lag and deviation in the execution of control instructions are solved, and the execution accuracy of control instructions is significantly improved.

[0060] Above Figure 2 The temperature control system for hot processing in the embodiments of the present invention is described in detail from the perspective of modular functional entities. Next, the temperature control device for hot processing in the embodiments of the present invention is described in detail from the perspective of hardware processing.

[0061] Figure 3 FIG. is a schematic structural diagram of a temperature control device for hot processing provided by an embodiment of the present invention. The temperature control device 300 for hot processing may vary greatly due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) for storing application programs 333 or data 332. Among them, the memory 320 and the storage media 330 may be transient storage or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the temperature control device 300 for hot processing. Further, the processor 310 may be set to communicate with the storage media 330 and execute a series of instruction operations in the storage media 330 on the temperature control device 300 for hot processing to implement the steps of the above-mentioned temperature control method for hot processing.

[0062] The temperature control device 300 for the hot working process may further include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, and so on. Those skilled in the art can understand that Figure 3 The structure of the temperature control device for the hot working process shown does not constitute a limitation on the temperature control device for the hot working process provided by the present invention, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0063] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are run on a computer, the computer is caused to execute the steps of the temperature control method for the hot working process.

[0064] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, systems, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0065] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions to cause a temperature control device for the hot working process (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0066] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A temperature control method for a hot working process, characterized in that, The method includes: Collecting temperature data of each process of smelting, casting, forging and heat treatment to obtain the spatio-temporal temperature distribution information of the entire hot working process; Calculating the temperature change rate, temperature extreme points and temperature uniformity index during the hot working process according to the spatio-temporal temperature distribution information to obtain key temperature characteristic parameters; Conducting an orthogonal experiment analysis on the process parameters based on the key temperature characteristic parameters to establish a quantitative relationship matrix of the influence of process parameters on the temperature field; According to the quantitative relationship matrix, using a rolling horizon optimization algorithm to calculate the optimal combination of heating power, cooling rate and holding time, and generating a temperature control instruction; Predicting the execution timing and execution duration of the temperature control instruction to obtain an execution prediction result, and correcting the control instruction according to the execution prediction result to obtain a target control instruction.

2. The temperature control method for a hot working process according to claim 1, characterized in that, The collecting temperature data of each process of smelting, casting, forging and heat treatment to obtain the spatio-temporal temperature distribution information of the entire hot working process includes: Setting an infrared thermometer, a thermocouple and a thermal imager for the smelting process to measure the surface temperature of the molten metal, the temperature in the furnace and the overall temperature distribution, and obtaining the temperature data of the smelting process; Setting an infrared thermometer for the casting process to measure the pouring temperature, using a thermocouple array to embed in the mold to monitor the mold temperature, and using a thermal imager to track the cooling process of the ingot to obtain the temperature data of the casting process; Deploying an infrared thermometer and a thermal imager for the forging process to measure the surface temperature of the forging, and installing a multi-point thermocouple array to monitor the temperature field of the heating furnace to obtain the temperature data of the forging process; Arranging a multi-point thermocouple array for the heat treatment process to monitor the temperature distribution of the heat treatment furnace, and using an infrared thermometer and a thermal imager to record the cooling curve of the workpiece after leaving the furnace to obtain the temperature data of the heat treatment process; Transmitting the temperature data of the smelting process, the casting process, the forging process and the heat treatment process to the data processing system through an industrial control network, and using a Kalman filter algorithm to eliminate measurement errors to obtain the fused temperature data; Performing interpolation calculation and three-dimensional reconstruction on the fused temperature data to construct the continuous spatio-temporal temperature distribution information of the entire hot working process, and the spatio-temporal temperature distribution information includes three dimensions of temperature value, spatial coordinate and timestamp.

3. The temperature control method for a hot working process according to claim 1, characterized in that, The calculating the temperature change rate, temperature extreme points and temperature uniformity index during the hot working process according to the spatio-temporal temperature distribution information to obtain key temperature characteristic parameters includes: Performing time series segmentation processing on the spatio-temporal temperature distribution information, and calculating the ratio of the temperature difference between adjacent time points to the time interval to obtain the temperature change rate curve of each measuring point; Applying a derivative extreme value detection algorithm to the temperature change rate curve, and marking the inflection points with significant changes in the slope of the rate curve to obtain the key temperature transition moments during the hot working process; Constructing a heat map based on the spatio-temporal temperature distribution information, and using a peak search algorithm to identify the coordinates of local highest and lowest points in the temperature field to obtain a temperature extreme point distribution map; Calculating the temperature gradient between the extreme points of the temperature extreme point distribution map, and constructing a temperature gradient vector field to obtain the temperature change direction and intensity during the hot working process; According to the spatio-temporal temperature distribution information, calculate the standard deviation, skewness, and kurtosis of the spatial temperature field at each time point to obtain statistical parameters characterizing the temperature distribution uniformity; Fuse the temperature change rate curve, key temperature transition moments, temperature extreme point distribution map, temperature gradient vector field, and statistical parameters to construct a multi-dimensional feature vector, and obtain the key temperature characteristic parameters of the hot processing process.

4. The temperature control method for a hot working process according to claim 1, characterized in that, Based on the key temperature characteristic parameters, conduct an orthogonal experiment analysis on the process parameters to establish a quantitative relationship matrix of the influence of process parameters on the temperature field, including: Select key process parameter groups for the smelting, casting, forging, and heat treatment processes respectively, design an orthogonal experiment table, and obtain an experimental plan for process parameter combinations; Conduct hot processing experiments according to the experimental plan for process parameter combinations, collect spatio-temporal temperature distribution information for each group of experiments, extract the key temperature characteristic parameters, and obtain a process parameter-temperature characteristic response data set; Conduct variance analysis on the process parameter-temperature characteristic response data set, calculate the contribution rate and significance level of each process parameter to the temperature characteristic parameters, and obtain the main effect analysis result; Based on the main effect analysis result, construct the influence weight coefficient of process parameters on the temperature field, use polynomial regression to fit the non-linear relationship between process parameters and temperature characteristic parameters, and obtain a process-temperature influence function; Calculate the partial derivative of the process-temperature influence function with respect to the process parameters to generate a Jacobian matrix, and obtain the sensitivity distribution of the response of the temperature field to changes in process parameters; Perform normalization processing on the sensitivity distribution, construct a forward mapping relationship and an inverse solution relationship between process parameters and the temperature field, generate a quantitative relationship matrix, and the quantitative relationship matrix contains the corresponding relationship between process parameter adjustment amounts and temperature field change amounts.

5. The temperature control method for a hot working process according to claim 1, characterized in that, According to the quantitative relationship matrix, use a rolling horizon optimization algorithm to calculate the optimal combination of heating power, cooling rate, and holding time, and generate temperature control instructions, including: Parametrically express the requirements of the hot processing process, convert the product quality index into a temperature control target, and obtain the target value and constraint conditions for optimization and solution; Based on the quantitative relationship matrix, construct a temperature field state prediction calculation formula, input the temperature characteristic parameters and candidate process parameters at the current moment, calculate the temperature field evolution sequence within the prediction horizon, and obtain the temperature response prediction result; Calculate the error between the temperature response prediction result and the target temperature curve, construct a weighted quadratic performance index evaluation method, and obtain an evaluation index for process parameter optimization; Based on the evaluation index, use a non-linear programming solution method to perform rolling optimization on the process parameters. Only execute the control quantity of the first time step each time, and then slide the prediction window for re-optimization to obtain a dynamically updated process parameter sequence; According to the process parameter sequence, perform piecewise linear interpolation calculation on the heating power, cooling rate, and holding time to obtain a continuous parameter control curve that meets the equipment execution accuracy; Convert the continuous parameter control curve according to the requirements of the equipment control interface and arrange the timing to generate a standard temperature control instruction including the execution time, execution object, and execution value.

6. The temperature control method for a hot working process according to claim 1, wherein, Performing timing and execution duration prediction on the temperature control instruction, obtaining an execution prediction result, and correcting the control instruction according to the execution prediction result to obtain a target control instruction, including: Performing matching analysis on the temperature control instruction and historical execution data, extracting instruction execution delay characteristics, temperature response characteristics, and equipment status characteristics to obtain an instruction execution characteristic set; Constructing a long short-term memory neural network based on the instruction execution characteristic set. The long short-term memory neural network includes an input layer, two long short-term memory hidden layers, a fully connected layer, and an output layer. The input layer receives an instruction feature vector. The first hidden layer contains 128 neurons for extracting temporal features. The second hidden layer contains 64 neurons for feature fusion. The fully connected layer contains 32 neurons. The output layer generates execution timing and duration prediction values; Converting the temperature control instruction into a feature vector, inputting it into the long short-term memory neural network, and calculating through forward propagation to obtain the execution timing offset and execution duration prediction value of each instruction; Performing equipment constraint verification on the execution timing offset and execution duration prediction value, and correcting the prediction result according to the response characteristics and physical limitations of the hot processing equipment to obtain an executable execution prediction result; Based on the execution prediction result, performing temporal rearrangement and parameter adjustment on the temperature control instruction sequence, eliminating instruction execution conflicts, and optimizing the instruction execution order to obtain a preliminarily corrected control instruction; Evaluating the temperature target achievement degree of the preliminarily corrected control instruction, calculating the corrected temperature evolution trajectory using a heat conduction model, and performing compensation adjustment on the part deviating from the target to obtain the final target control instruction.

7. The temperature control method for a hot working process according to claim 6, wherein, The performing matching analysis on the temperature control instruction and historical execution data, extracting instruction execution delay characteristics, temperature response characteristics, and equipment status characteristics to obtain an instruction execution characteristic set, including: Performing time tokenization processing on the historical temperature control instruction database, recording the issuance time, start execution time, and end execution time of each instruction, calculating the execution delay time and actual duration to obtain instruction execution timing characteristics; According to the instruction execution timing characteristics, performing hierarchical clustering analysis on the instruction type, process stage, and equipment load status to obtain the instruction execution delay distribution law under different conditions; Aligning the historical temperature data and control instructions in time, extracting the temperature change rate, temperature response time, and temperature stability parameters before and after instruction execution to obtain temperature response characteristics; Based on the equipment operation log, extracting equipment start-stop status, workload, maintenance cycle, and fault record information, constructing an equipment status feature vector to obtain equipment operation status characteristics; Associating and mapping the instruction execution timing characteristics, instruction execution delay distribution law, temperature response characteristics, and equipment operation status characteristics, establishing a multi-dimensional feature association table to obtain a comprehensive feature association result; Performing principal component analysis and feature importance ranking on the comprehensive feature association result, selecting the feature subset with the highest correlation, and generating an instruction execution characteristic set.

8. A temperature control system for a hot working process, wherein, For implementing the temperature control method for the hot working process as described in any one of claims 1-7, the temperature control system for the hot working process includes: A collection module, configured to collect temperature data of each process of smelting, casting, forging, and heat treatment, and obtain the spatio-temporal temperature distribution information of the entire hot working process; A calculation module, configured to calculate the temperature change rate, temperature extreme points, and temperature uniformity index during the hot working process according to the spatio-temporal temperature distribution information, and obtain key temperature characteristic parameters; An analysis module, configured to perform an orthogonal experiment analysis on process parameters based on the key temperature characteristic parameters, and establish a quantitative relationship matrix of the influence of process parameters on the temperature field; A generation module, configured to calculate the optimal combination of heating power, cooling rate, and holding time according to the quantitative relationship matrix by using a rolling horizon optimization algorithm, and generate a temperature control instruction; A prediction module, configured to predict the execution timing and execution duration of the temperature control instruction, obtain an execution prediction result, and correct the control instruction according to the execution prediction result to obtain a target control instruction.

9. A temperature control device for a hot working process, wherein, It includes a memory and a processor, the memory stores a computer program that can run on the processor, and when the processor executes the computer program, it implements the temperature control method for the hot working process as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, wherein, When the computer program is run by the processor, it causes the processor to execute the temperature control method for the hot working process as described in any one of claims 1 to 7.

Images