Steel heat treatment temperature field simulation and feedback control method and system

Abstract

The application discloses a steel heat treatment temperature field simulation and feedback control method and system, and relates to the technical field of industrial intelligent control, and comprises the following steps: heat transfer calculation is carried out by constructing a furnace space grid and boundary conditions, a predicted temperature field is obtained, and a temperature deviation field is formed in combination with a measured temperature; an abnormal area is identified based on a deviation gradient, a reverse tracing model is established by using a multi-time temperature evolution track, a heat source disturbance is inverted, and the temperature field is corrected; a power-temperature dynamic mapping relationship is constructed by power excitation, and a dominant control mode is extracted by dimension reduction; a control vector is solved by decomposing the difference value between the corrected temperature field and a target temperature field on the dominant mode, and a power regulation amount is obtained by reconstruction; an over-limit peak value is inhibited by combining temperature evolution prediction, an execution power instruction is generated, and accurate regulation and optimization of the temperature field are realized.

Description

Technical Field

[0001] This invention relates to the field of industrial intelligent control technology, specifically to a method and system for simulating and controlling the temperature field of steel heat treatment. Background Technology

[0002] In the heat treatment of steel, the uniformity and stability of the temperature field within the heating furnace directly determine the microstructural transformation and final mechanical properties of the material. Traditional heat treatment processes often rely on experience to set heating curves and zoned power control strategies. Temperatures at limited measurement points are typically obtained using thermocouples or infrared thermometers, and the furnace temperature is estimated using a simple heat transfer model. However, due to the complex furnace structure, varied steel loading methods, and significant coupling effects of radiation, convection, and conduction, the actual temperature field exhibits strong nonlinearity and time-varying characteristics, leading to a large deviation between the predicted results based on simplified models and the actual temperature distribution.

[0003] While existing technologies have introduced numerical simulation methods for offline or near-real-time calculation of furnace temperature fields, most rely on fixed boundary conditions and idealized heat source distributions, making it difficult to reflect disturbances caused by furnace door opening, combustion fluctuations, or local loading differences during actual production. Furthermore, the limited number and discrete distribution of temperature measurement points restricts the accuracy of temperature field reconstruction, making it difficult to identify local overheating or underheating areas in a timely manner, thus affecting product quality consistency. When local temperature anomalies occur within the furnace, traditional control methods struggle to locate heat source deviations and make targeted adjustments in a timely manner, often relying on overall power correction, leading to increased energy consumption and insufficient control accuracy. The lack of a predictive and constraining mechanism for temperature evolution makes temperature prone to overshoot or local peaks, further affecting heat treatment quality. There is also a lack of in-depth analysis of the spatial distribution characteristics of temperature deviations, and an ineffective combination of temperature field inversion and control input reconstruction mechanisms. Therefore, how to achieve high-precision temperature field prediction, deviation tracing, and efficient feedback control based on dominant characteristics in complex furnace environments remains a key technical problem urgently needing to be solved in this field. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for simulating and controlling the temperature field of steel heat treatment, aiming to solve at least one of the technical problems existing in the prior art.

[0005] The technical solution of this invention is: a method for simulating and controlling the temperature field of steel heat treatment, comprising the following steps: The predicted temperature field is obtained by heat transfer calculation based on the furnace space grid and boundary conditions, and the temperature deviation field is obtained by comparing the measured temperature with the predicted temperature field. Identify gradient anomaly regions in the temperature deviation field, extract the multi-time temperature evolution trajectory of the gradient anomaly regions, establish a temperature propagation reverse tracing model, solve the heat source disturbance distribution through the reverse tracing model, and compensate the heat source disturbance distribution to the predicted temperature field to obtain the corrected temperature field. A power excitation is applied to the heating area to collect the temperature response, a dynamic mapping relationship between power and temperature is constructed, and the dynamic mapping relationship is decomposed in a dimensionality reduction manner to extract the dominant control mode. The difference between the corrected temperature field and the target temperature field is decomposed and the control vector is solved on the dominant control mode. The control vector is then reconstructed into a power regulation quantity through the dominant control mode. The power regulation input is used to predict the temperature evolution process through dynamic mapping. Peak values ​​exceeding the preset target envelope and corresponding heating regions are identified during the temperature evolution process. The power regulation of the corresponding heating regions is suppressed to obtain the execution power command, which is then applied to the heating regions.

[0006] The predicted temperature field is obtained by heat transfer calculation based on the furnace spatial grid and boundary conditions. The temperature deviation field is obtained by comparing the measured temperature with the predicted temperature field. The furnace space is divided into a spatial grid, the thermal conductivity and specific heat capacity of the steel are obtained, the furnace wall temperature and furnace door opening state are obtained as boundary conditions, the heat transfer differential equation of the spatial grid is established, and the temperature distribution of the spatial grid is obtained by numerically solving the heat transfer differential equation. The temperature distribution constitutes the predicted temperature field. The measured temperature of the monitoring point inside the furnace is collected by a temperature sensor. The corresponding spatial grid is located in the predicted temperature field according to the spatial coordinates of the monitoring point. The temperature value of the corresponding spatial grid is extracted as the predicted temperature. The difference between the measured temperature and the predicted temperature is calculated to obtain the temperature deviation of the monitoring point. The spatial distance between each spatial grid and each monitoring point is calculated. The temperature deviation of each monitoring point is weighted and summed according to the spatial distance. The weighted summation result is assigned to each spatial grid to form a temperature deviation field.

[0007] Identify gradient anomaly regions in the temperature deviation field, extract multi-time temperature evolution trajectories of these regions, establish a temperature propagation reverse tracing model, solve for the heat source disturbance distribution using this model, and compensate the heat source disturbance distribution to the predicted temperature field to obtain the corrected temperature field, including: Calculate the temperature gradient values ​​of each spatial grid in the temperature deviation field, average all temperature gradient values ​​to obtain the gradient reference value, and identify spatial grids with temperature gradient values ​​greater than the gradient reference value as gradient anomaly regions. The temperature deviation values ​​of the gradient anomaly region at multiple consecutive historical moments are obtained and arranged in chronological order to form a multi-moment temperature evolution trajectory; The time change rate of temperature deviation between adjacent time points in the multi-time temperature evolution trajectory is calculated. Based on the propagation law of heat attenuation with spatial distance, a quantitative relationship between the time change rate, spatial distance and heat source intensity is established as a reverse tracing model of temperature propagation. Calculate the spatial distance from each spatial grid to the gradient anomaly region; Substitute the temporal rate of change of the gradient anomaly region and the spatial distance of each spatial grid into the temperature propagation reverse tracing model, and solve the heat source intensity value when each spatial grid is a heat source in reverse. Identify the spatial grid with a heat source intensity value greater than a preset intensity threshold as the heat source location. The heat source location and the heat source intensity value constitute the heat source disturbance distribution. The heat source disturbance distribution is compensated to the corresponding spatial grid in the predicted temperature field to obtain the corrected temperature field.

[0008] A power excitation is applied to the heating region to collect the temperature response, and a dynamic power-temperature mapping relationship is constructed. The dynamic mapping relationship is then decomposed in a dimensionality reduction manner to extract the dominant control modes, including: Multiple power excitation sequences with different amplitudes are applied to the heating area, and the temperature response time sequence corresponding to each power excitation sequence is collected in real time. Fourier transforms are performed on each power excitation sequence and the corresponding temperature response time sequence to obtain the power excitation spectrum and temperature response spectrum. The complex ratio of the temperature response spectrum to the power excitation spectrum at each frequency point is calculated to obtain the frequency response function under each power excitation amplitude. The frequency response function under each power excitation amplitude is organized into a two-dimensional array according to the power excitation amplitude and frequency point, and a dynamic mapping relationship between power and temperature is constructed. Singular value decomposition of dynamic mapping relationships yields singular values, left singular vectors, and right singular vectors; Calculate the energy percentage of the square of each singular value to the sum of the squares of singular values, and select the singular values ​​whose cumulative energy percentage reaches a preset energy threshold, along with their corresponding left and right singular vectors. The temperature response mode coefficients are obtained by performing an inner product operation between the temperature response time series and the left singular vector, and the power excitation mode coefficients are obtained by performing an inner product operation between the power excitation series and the right singular vector. The temperature response mode coefficients and the power excitation mode coefficients are then linearly related through singular values ​​to form the dominant control mode.

[0009] The difference between the corrected temperature field and the target temperature field is decomposed and the control vector is solved on the dominant control mode. The control vector is then reconstructed into the power regulation quantity through the dominant control mode, including: The temperature deviation field is obtained by calculating the temperature difference between the corrected temperature field and the target temperature field at various spatial locations; The temperature deviation field is transformed into the frequency domain space to obtain the temperature deviation spectrum. The temperature deviation spectrum is then projected onto the temperature response subspace formed by the left singular vectors in the dominant control mode to obtain the frequency domain response mode coefficients of the temperature deviation field in each dominant control mode. The frequency domain normalized response mode coefficients are obtained by performing a ratio calculation between the singular values ​​of the frequency domain response mode coefficients and the corresponding dominant control modes. The frequency domain power excitation mode coefficients are obtained by projecting the normalized response mode coefficients in the frequency domain onto the right singular vector in the dominant control mode. The frequency domain power excitation mode coefficients are then transformed from the frequency domain space to the time domain space to obtain the time domain power excitation mode coefficients as the control vector. The control vector and the right singular vector are weighted and combined. The weighted combination result is then mapped to the spatial location of each heating region through the dominant control mode, and reconstructed into the power regulation amount corresponding to each heating region.

[0010] The power regulation input is used to predict the temperature evolution process through a dynamic mapping relationship. Peak values ​​exceeding a preset target envelope and corresponding heating regions are identified during the temperature evolution process. Power regulation in the corresponding heating regions is suppressed to obtain an execution power command, which is then applied to the heating region. By substituting the power adjustment into the dynamic mapping relationship, the temperature response trajectory of each heating zone in the future time period is calculated to obtain the temperature evolution process. The predicted temperature values ​​of each heating region are extracted during the temperature evolution process. The predicted temperature values ​​are compared with the upper boundary of the preset target envelope to identify the temperature peak, the overshoot of the temperature peak exceeding the upper boundary, the heating region where the temperature peak is located, and the time when the temperature peak occurs. Extract the thermal coupling coefficient between the heating region where the temperature peak is located and other heating regions in the dynamic mapping relationship, filter other heating regions with thermal coupling coefficients greater than a preset coupling threshold, and form a set of suppression regions by combining the heating region where the temperature peak is located with the other filtered heating regions. The time interval from the current moment to the moment when the temperature peak occurs is calculated as the suppression time window, and the power change rate is obtained by calculating the ratio of the overshoot to the suppression time window. The power adjustment amount corresponding to each heating region in the set of suppressed regions is linearly decreased according to the power change rate to obtain the power adjustment amount after suppression. The suppressed power adjustment is superimposed on the current power output to obtain the execution power command, which is then applied to each heating zone.

[0011] This invention provides a steel heat treatment temperature field simulation and feedback control system, the system comprising: The temperature field calculation unit is used to perform heat transfer calculations based on the furnace space grid and boundary conditions to obtain the predicted temperature field, and to collect the measured temperature and compare it with the predicted temperature field to obtain the temperature deviation field. The heat source tracing unit is used to identify gradient anomaly regions in the temperature deviation field, extract the multi-time temperature evolution trajectory of the gradient anomaly regions, establish a temperature propagation reverse tracing model, solve the heat source disturbance distribution through the reverse tracing model, and compensate the heat source disturbance distribution to the predicted temperature field to obtain the corrected temperature field. The mapping construction unit is used to apply power excitation to the heating area, collect the temperature response, construct a dynamic mapping relationship between power and temperature, and perform dimensionality reduction decomposition on the dynamic mapping relationship to extract the dominant control mode; The power calculation unit is used to decompose the difference between the corrected temperature field and the target temperature field into a control vector on the dominant control mode, and reconstruct the control vector into a power adjustment quantity through the dominant control mode. The power suppression unit is used to input the power regulation amount into the dynamic mapping relationship to predict the temperature evolution process, identify the peak value and corresponding heating region that exceed the preset target envelope from the temperature evolution process, suppress the power regulation amount of the corresponding heating region, obtain the execution power command, and apply it to the heating region.

[0012] One technical solution provided in this embodiment of the invention is an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0013] One technical solution provided in this embodiment of the invention is a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the steps in any of the aforementioned methods.

[0014] This invention achieves high-precision reconstruction of the furnace temperature field by fusing heat transfer models with measured data, significantly reducing the deviation between model predictions and actual operating conditions. Utilizing temperature deviation gradient identification and reverse tracing mechanisms, it accurately locates the position of heat source disturbances and their impact range, improving anomaly tracing capabilities and response speed. By constructing a power-temperature dynamic mapping and extracting the dominant control mode through dimensionality reduction, it effectively reduces control dimensions and system complexity, enhancing the controllability of multi-region coupled systems. Based on mode decomposition, it achieves precise solution and reconstruction of control quantities, making power regulation more targeted and efficient. Combined with temperature evolution prediction and peak suppression mechanisms, it avoids local overheating or temperature overshoot, improving the uniformity and stability of the temperature field. Overall, it achieves closed-loop optimized control of the heat treatment process, improving the consistency of steel performance while reducing energy consumption and operating costs. Attached Figure Description

[0015] Figure 1 A flowchart of a method for simulating and controlling the temperature field of steel heat treatment provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the steel heat treatment temperature field simulation and feedback control system according to an embodiment of the present invention. Detailed Implementation

[0016] like Figure 1 As shown, Figure 1 This is a flowchart of a method for simulating and controlling the temperature field of steel heat treatment according to an embodiment of the present invention. The method includes the following steps: Step 101: Perform heat transfer calculations based on the furnace space grid and boundary conditions to obtain the predicted temperature field, and collect the measured temperature and compare it with the predicted temperature field to obtain the temperature deviation field.

[0017] In some embodiments of the present invention, step 101 may specifically include the following sub-steps: Sub-step 1011: Divide the furnace space into a spatial grid, obtain the thermal conductivity and specific heat capacity of the steel, obtain the furnace wall temperature and furnace door opening state as boundary conditions, establish the heat transfer differential equation of the spatial grid, and numerically solve the heat transfer differential equation to obtain the temperature distribution of the spatial grid. The temperature distribution constitutes the predicted temperature field. Sub-step 1012: The measured temperature of the monitoring point inside the furnace is collected by the temperature sensor. The corresponding spatial grid is located in the predicted temperature field according to the spatial coordinates of the monitoring point. The temperature value of the corresponding spatial grid is extracted as the predicted temperature. The difference between the measured temperature and the predicted temperature is calculated to obtain the temperature deviation of the monitoring point. Sub-step 1013: Calculate the spatial distance between each spatial grid and each monitoring point, perform a weighted summation of the temperature deviations of each monitoring point based on the spatial distance, and assign the weighted summation result to each spatial grid to form a temperature deviation field.

[0018] In implementing a method for simulating and controlling the temperature field of steel heat treatment, heat transfer calculations based on the furnace space grid and boundary conditions are crucial. This method first divides the furnace space into a three-dimensional grid structure, with each grid cell being a 1cm × 1cm × 1cm cube, thus forming a complete furnace space model. For each grid cell, the physical parameters of the steel, including thermal conductivity and specific heat capacity, need to be obtained, as these parameters are temperature-dependent. For medium carbon steel, the thermal conductivity is approximately 35 W / (m·K), and the specific heat capacity is approximately 650 J / (kg·K) at 750℃. The furnace wall temperature can be measured in real time by temperature sensors installed on the furnace wall, and it typically varies at different locations; for example, the temperature at the top of the furnace can reach 980℃, while the temperature at the bottom may be 950℃. The furnace door opening status is obtained through a door position sensor, with an opening range of 0-100%. An open door will cause a local temperature drop.

[0019] When establishing the differential equation for heat transfer in a spatial grid, a three-dimensional heat conduction equation is used to describe the heat transfer process within the furnace space. This equation can be expressed as: Where ρ represents the density of steel, and c p Let represent the specific heat capacity of steel, T represent temperature, t represent time, k represent thermal conductivity, x, y, z represent spatial coordinates, and Q represent the internal heat source term. When numerically solving this equation, the continuous domain is discretized using the finite difference method.

[0020] At each grid node, the temperature update can be simplified as follows: ,in, The temperature of node i at time step n. The temperature for the next time step. Let N(i) represent the temperature of node j adjacent to node i at time step n, N(i) represent the set of adjacent nodes of node i, α is a calculated parameter related to thermal conductivity, β is a parameter related to the heat source, and Q is a parameter related to the heat source. i Let be the heat source intensity at node i.

[0021] Solving this difference equation requires setting boundary conditions. For the furnace wall boundary, a first-type boundary condition can be used, i.e., the boundary temperature value can be directly specified. For the furnace door, a convection boundary condition is used according to the opening state; the larger the opening, the higher the heat exchange coefficient. Through iterative calculation, the temperature distribution of the entire furnace space at a given time is finally obtained, i.e., the predicted temperature field.

[0022] Temperature monitoring is achieved through temperature sensors placed at key locations within the furnace. Typically, multiple monitoring points are set up inside the furnace, such as thermocouple sensors positioned at the top, middle, and bottom. These sensors collect and transmit temperature data in real time, with a sampling frequency set to 1 time per second. Each monitoring point has defined spatial coordinates, which can be represented by (x...). m ym , z m ) represents the location of the m-th monitoring point.

[0023] Based on the spatial coordinates of the monitoring point, a corresponding spatial grid can be found in the predicted temperature field. Since the coordinates of the monitoring point usually do not fall precisely on the grid nodes, an interpolation method is needed to obtain the predicted temperature of that point. The trilinear interpolation method is suitable for obtaining the temperature value of any point from three-dimensional grid data, that is, it is calculated by weighting the temperature values ​​of the eight nearest grid nodes around the monitoring point. The difference between the calculated predicted temperature and the measured temperature is the temperature deviation of that monitoring point.

[0024] To construct a complete temperature deviation field, the temperature deviation of discrete monitoring points is extended to the entire furnace space. For each spatial grid within the furnace, the Euclidean distance between it and each monitoring point is calculated. The distance calculation formula is the straight-line distance between two points in space, i.e., the distance between grid point (i, j, k) and the m-th monitoring point (x). m y m , z m The distance between the monitoring points is calculated. Based on the calculated distance, the temperature deviation of each monitoring point is weighted and summed using an inverse distance weighting method. The weight is inversely proportional to the distance; the closer the monitoring point, the greater its influence on the current grid. The result of the weighted summation is the temperature deviation value of the spatial grid, and the temperature deviation values ​​of all grid points together constitute the temperature deviation field.

[0025] This invention achieves high-precision simulation of the temperature field during steel heat treatment by constructing an accurate furnace temperature field prediction model and a real-time temperature monitoring system. The introduction of a temperature deviation field effectively compensates for the discrepancy between the theoretical model and actual conditions, improving temperature control accuracy, reducing temperature fluctuations during steel heat treatment, and significantly improving the stability and consistency of steel quality. This method does not rely on a large amount of experimental data, is highly adaptable, and can be quickly applied to different types of heat treatment furnaces, providing effective technical support for the intelligent control of steel heat treatment processes.

[0026] Step 102: Identify gradient anomaly regions in the temperature deviation field, extract the multi-time temperature evolution trajectory of the gradient anomaly regions, establish a temperature propagation reverse tracing model, solve the heat source disturbance distribution through the reverse tracing model, and compensate the heat source disturbance distribution to the predicted temperature field to obtain the corrected temperature field.

[0027] In some embodiments of the present invention, step 102 may specifically include the following sub-steps: Sub-step 1021: Calculate the temperature gradient value of each spatial grid in the temperature deviation field, average all temperature gradient values ​​to obtain the gradient reference value, and identify spatial grids with temperature gradient values ​​greater than the gradient reference value as gradient anomaly regions. Sub-step 1022: Obtain the temperature deviation values ​​of the gradient anomaly region at multiple consecutive historical moments and arrange them in chronological order to form a multi-moment temperature evolution trajectory. Sub-step 1023: Calculate the time change rate of temperature deviation values ​​between adjacent time points in the multi-time temperature evolution trajectory. Based on the propagation law of heat attenuation with spatial distance, establish a quantitative relationship between the time change rate, spatial distance and heat source intensity as a reverse tracing model of temperature propagation. Sub-step 1024: Calculate the spatial distance from each spatial grid to the gradient anomaly region; Sub-step 1025: Substitute the time change rate of the gradient anomaly region and the spatial distance of each spatial grid into the temperature propagation reverse tracing model, reversely solve the heat source intensity value when each spatial grid is a heat source, identify the spatial grid with a heat source intensity value greater than a preset intensity threshold as the heat source location, and the heat source location and heat source intensity value constitute the heat source disturbance distribution. Sub-step 1026: Compensate the heat source disturbance distribution to the corresponding spatial grid in the predicted temperature field to obtain the corrected temperature field.

[0028] After acquiring the temperature deviation field, gradient anomaly regions within the field are identified, and heat source disturbance analysis is performed. Gradient analysis of the temperature deviation field is a crucial step in identifying abnormal heat sources during heat treatment. For each spatial grid, the gradient value of its temperature deviation is calculated, i.e., the rate of change of temperature deviation between adjacent grids. The temperature gradient calculation uses the central difference method. For a grid point (i, j, k) in three-dimensional space, the magnitude of its temperature gradient can be calculated as the square root of the sum of the squares of the gradients in each direction. Specifically, the gradient in the x-direction is the difference in temperature deviation between two adjacent grid points divided by the grid spacing; the calculation methods for the y-direction and z-direction gradients are similar. After calculating the temperature gradient values ​​for all spatial grids, these values ​​are arithmetically averaged to obtain a gradient reference value. The gradient reference value reflects the average gradient level of the entire temperature deviation field, typically within the range of 0.5-2℃ / cm. The temperature gradient value of each spatial grid is compared with the gradient reference value. If the gradient value of a certain grid exceeds the gradient reference value, that grid is marked as a gradient anomaly region. Gradient anomaly regions often indicate the presence of abnormal heat source disturbances at that location.

[0029] For identified gradient anomaly regions, temperature deviation values ​​at multiple points in the past are extracted from the historical database to form a temperature evolution sequence over time. To ensure the temporal continuity of the data, temperature deviation records recorded once per second within the most recent 60 seconds are typically selected. These temperature deviation values, arranged in chronological order, constitute a multi-time temperature evolution trajectory, used to analyze the temporal characteristics of temperature changes in gradient anomaly regions.

[0030] Based on the multi-time temperature evolution trajectory, the rate of change of temperature deviation between adjacent time points is calculated. For the temperature deviation between time points I and I+1, the rate of change is the difference between the temperature deviations at the two time points divided by the time interval. The rate of change reflects the speed at which the temperature deviation changes. According to the basic principle of heat transfer, the speed of heat propagation in space is related to distance; the farther away from the heat source, the longer it takes for the heat to arrive, and the weaker its influence. Based on this principle, a model relating the rate of change, spatial distance, and heat source intensity can be established, namely, a reverse tracing model of temperature propagation: , where P s Let R represent the heat source power, d represent the time rate of change of temperature deviation, k represent the spatial distance, and λ represent the thermal conductivity. This model considers the squared distance decay and exponential decay characteristics during heat propagation.

[0031] To apply the temperature propagation reverse tracing model, it is necessary to calculate the spatial distance from each spatial grid to the gradient anomaly region. The spatial distance is calculated using Euclidean distance, which is the straight-line distance between two points. For multiple grid points within the gradient anomaly region, its center point can be selected as a reference point for distance calculation. If the gradient anomaly region contains multiple discontinuous regions, the distance to each region must be calculated separately, and the minimum value is taken as the final distance.

[0032] Based on the reverse tracing model, the time change rate of the gradient anomaly region and the spatial distance of each spatial grid are substituted into the calculation to obtain the heat source power value when each spatial grid is a potential heat source. The larger the heat source power value, the higher the probability that the grid is a heat source. A preset power threshold of 5W is set, and spatial grids with heat source power values ​​greater than this threshold are identified as heat source locations. These heat source locations and their corresponding heat source power values ​​constitute the heat source disturbance distribution. The heat source disturbance distribution reveals the spatial distribution of unexpected heat sources in the furnace, which may originate from uneven local fuel combustion, inconsistent heat dissipation from the furnace wall, or the chemical reaction heat of the steel itself.

[0033] After obtaining the heat source perturbation distribution, it is used as a correction term to compensate for the predicted temperature field. Specifically, for each grid point identified as a heat source, the heat source power is converted into a temperature effect, which is then added to the predicted temperature of that grid point. The formula for converting heat source power into temperature is: Where ΔT represents the temperature increment, P s The heat source power is represented by Δt, the time step by Δt, and the mesh cell mass by M. p This represents the specific heat capacity. The calculated temperature increment is added to the temperature values ​​at the corresponding grid points in the predicted temperature field to compensate for heat source disturbances, resulting in the corrected temperature field. Compared to the original predicted temperature field, the corrected temperature field more accurately reflects the actual temperature distribution inside the furnace.

[0034] This invention identifies gradient anomaly regions in the temperature deviation field, extracts the temperature evolution trajectory, establishes a reverse tracing model, and calculates the heat source disturbance distribution. This method achieves precise location and quantitative analysis of unexpected heat sources during steel heat treatment. The introduction of a corrected temperature field improves the accuracy of temperature field simulation, laying the foundation for subsequent precise control. It can effectively identify abnormal heat sources during heat treatment, achieve precise temperature field correction, improve the accuracy of heat treatment quality control, reduce energy waste, and ensure the uniformity and stability of the temperature field during steel heat treatment.

[0035] Step 103: Apply power excitation to the heating area and collect the temperature response to construct a dynamic mapping relationship between power and temperature. Perform dimensionality reduction decomposition on the dynamic mapping relationship to extract the dominant control mode.

[0036] In some embodiments of the present invention, step 103 may specifically include the following sub-steps: Sub-step 1031: Apply multiple power excitation sequences with different amplitudes to the heating area and collect the temperature response time sequence corresponding to each power excitation sequence in real time; Sub-step 1032: Perform Fourier transform on each power excitation sequence and the corresponding temperature response time sequence to obtain the power excitation spectrum and temperature response spectrum, and calculate the complex ratio of the temperature response spectrum and the power excitation spectrum at each frequency point to obtain the frequency response function under each power excitation amplitude. Sub-step 1033: Organize the frequency response function under each power excitation amplitude into a two-dimensional array according to the power excitation amplitude and frequency point, and construct a dynamic mapping relationship between power and temperature; Sub-step 1034: Perform singular value decomposition on the dynamic mapping relationship to obtain singular values, left singular vectors, and right singular vectors; Sub-step 1035: Calculate the energy ratio of the square of each singular value to the sum of the squares of singular values, and select the singular values ​​whose cumulative energy ratio reaches a preset energy threshold and their corresponding left and right singular vectors; Sub-step 1036: The temperature response mode coefficients are obtained by performing an inner product operation between the temperature response time series and the left singular vector, and the power excitation mode coefficients are obtained by performing an inner product operation between the power excitation series and the right singular vector. The temperature response mode coefficients and the power excitation mode coefficients are linearly related through singular values ​​to form the dominant control mode.

[0037] In the heat treatment of steel, to establish a dynamic mapping relationship between power input and temperature response, systematic power excitation testing and temperature response data acquisition are required. In practical applications, power excitation sequences of different amplitudes can be applied to the heating zone, including step signals, sinusoidal signals, or pseudo-random binary sequences. For example, five power excitation sequences with different amplitudes can be selected, ranging from 100W to 500W, with each sequence lasting 300 seconds and a sampling interval of 0.5 seconds. Simultaneously with the power excitation, the corresponding temperature response data is acquired in real time using a temperature sensor, forming a temperature response time series and recording the temperature change over time.

[0038] The acquired power excitation sequence and temperature response time series are processed by using a Fast Fourier Transform (FFT) algorithm to convert the time-domain data to the frequency domain, yielding the power excitation spectrum and temperature response spectrum, respectively. Assuming the original number of sampling points is 600, the converted frequency domain data points are also 600, with a frequency range of 0 to 1 Hz. The complex ratio of the temperature response spectrum to the power excitation spectrum at each frequency point is calculated to obtain the frequency response function for each power excitation amplitude. The frequency response function characterizes the amplitude and phase relationship of the temperature response to the power input at different frequencies, reflecting the dynamic characteristics of the system.

[0039] The calculated frequency response functions for each power excitation amplitude are organized into a two-dimensional array according to the power excitation amplitude and frequency points, forming a dynamic mapping relationship between power and temperature. The rows of this two-dimensional array represent different power excitation amplitudes, the columns represent different frequency points, and the array elements are the frequency response function values ​​under the corresponding conditions. Assuming there are 5 different power excitation amplitudes and 300 effective frequency points, a 5×300 complex matrix is ​​formed, which fully describes the dynamic mapping relationship between power and temperature.

[0040] Singular value decomposition (SVD) is performed on the constructed power-temperature dynamic mapping matrix, decomposing the matrix into three parts: singular values, left singular vectors, and right singular vectors. In SVD, if the original matrix has dimensions a×b, then min(a, b) singular values, along with their corresponding left and right singular vectors, are obtained. The singular values ​​are ordered from largest to smallest, reflecting the contribution of the corresponding modes to the system's dynamic characteristics. Modes corresponding to larger singular values ​​play a dominant role in the system, while modes corresponding to smaller singular values ​​contribute less and may represent noise or nonlinear factors in the system.

[0041] The proportion of the square of each singular value to the sum of the squares of all singular values ​​is calculated to obtain the energy percentage of each mode. This energy percentage is then accumulated sequentially until it reaches a preset energy threshold. Typically, the energy threshold is set to 95% or 99%, indicating that the retained modes can explain 95% or 99% of the system's dynamic characteristics. The singular values ​​whose cumulative energy percentage reaches the preset energy threshold, along with their corresponding left and right singular vectors, are selected. These selected singular values ​​and their corresponding singular vectors constitute the dominant control modes of the system.

[0042] The temperature response mode coefficients are obtained by performing an inner product operation between the temperature response time series and the left singular vector. These coefficients characterize the projection intensity of the temperature response onto each dominant mode, reflecting the contribution of each mode to the temperature response. Similarly, the power excitation mode coefficients are obtained by performing an inner product operation between the power excitation series and the right singular vector, characterizing the projection intensity of the power excitation onto each dominant mode. A linear relationship between the temperature response mode coefficients and the power excitation mode coefficients is established using singular values, constituting the dominant control mode of the system.

[0043] This modal analysis method based on singular value decomposition can extract the dominant control modes in a system, greatly simplifying the design complexity of the control system. In temperature field control, effective control of the entire temperature field can be achieved by controlling only the dominant control modes. The use of inner product operations makes the expression of temperature response and power excitation in modal space more concise, facilitating the establishment of linear control relationships.

[0044] This invention establishes a power-temperature dynamic mapping relationship and extracts the dominant control mode using the aforementioned method, achieving an accurate description and effective simplification of the dynamic characteristics of complex temperature fields. It boasts advantages such as high computational efficiency, significant dimensionality reduction, and high control precision, effectively improving the control accuracy and response speed of the temperature field in steel heat treatment, reducing energy consumption, and enhancing the consistency and stability of heat treatment quality.

[0045] Step 104: Decompose the difference between the corrected temperature field and the target temperature field into a control vector on the dominant control mode, and reconstruct the control vector into a power regulation quantity through the dominant control mode.

[0046] In some embodiments of the present invention, step 104 may specifically include the following sub-steps: Sub-step 1041: Calculate the temperature difference between the corrected temperature field and the target temperature field at each spatial location to obtain the temperature deviation field; Sub-step 1042: The temperature deviation field is transformed into the frequency domain space to obtain the temperature deviation spectrum. The temperature deviation spectrum is then projected onto the temperature response subspace formed by the left singular vector in the dominant control mode to obtain the frequency domain response mode coefficients of the temperature deviation field in each dominant control mode. Sub-step 1043: The ratio of the frequency domain response mode coefficients to the singular values ​​of the corresponding dominant control modes is calculated to obtain the frequency domain normalized response mode coefficients; Sub-step 1044: Project the frequency domain normalized response mode coefficients and the right singular vector in the dominant control mode to obtain the frequency domain power excitation mode coefficients; transform the frequency domain power excitation mode coefficients from the frequency domain space to the time domain space to obtain the time domain power excitation mode coefficients as the control vector. Sub-step 1045 involves weighting the control vector and the right singular vector, and mapping the weighted combination result to the spatial location of each heating region through the dominant control mode, thus reconstructing the power adjustment amount corresponding to each heating region.

[0047] During the heat treatment of steel, there is a deviation between the measured temperature field and the target temperature field. The control system needs to eliminate these deviations by adjusting the power input of each heating zone. The control system calculates the temperature difference between the corrected temperature field and the target temperature field at various spatial locations to obtain the temperature deviation field. The corrected temperature field refers to the actual temperature field after correction by the heat diffusion model, while the target temperature field is the desired temperature distribution. The temperature deviation field can be represented as an array of temperature differences at each sampling point, and its dimension depends on the number of sampling points in the temperature field. If a 10×10 grid is used to sample the temperature field, the temperature deviation field is a 100-dimensional vector.

[0048] The temperature deviation field is represented in the time domain, and the established power-temperature dynamic mapping relationship is transformed to the frequency domain. A Fourier transform is performed on the temperature deviation field to obtain the temperature deviation spectrum. Assuming a sampling interval of 0.5 s and a sampling duration of 300 s, the transformed spectrum resolution is 0.0033 Hz, and the frequency range is 0 to 1 Hz. The temperature deviation spectrum contains amplitude and phase information, represented in complex form. Projecting the temperature deviation spectrum onto the temperature response subspace formed by the left singular vectors of the dominant control modes yields the frequency domain response mode coefficients of the temperature deviation field in each dominant control mode. The projection process is essentially calculating the inner product of the temperature deviation spectrum and the left singular vectors. Assuming there are 3 dominant control modes, 3 frequency domain response mode coefficients are obtained, each coefficient representing the magnitude of the temperature deviation field component in the corresponding dominant mode.

[0049] The frequency-domain response mode coefficients are obtained by ratioing the singular values ​​of the corresponding dominant control modes. This step is equivalent to mapping the temperature response from temperature space to power space, preparing for subsequent power regulation. The normalization operation eliminates scale differences between different modes, making the contributions of each mode comparable. If the singular value of a dominant control mode is 5 and the corresponding frequency-domain response mode coefficient is 10, then the normalized mode coefficient is 2. For the three dominant control modes mentioned above, three frequency-domain normalized response mode coefficients are calculated respectively.

[0050] The frequency-domain power excitation mode coefficients are obtained by projecting the normalized response mode coefficients onto the right singular vector in the dominant control mode. The right singular vector represents the basis of the power space, and the projection operation maps the normalized response mode coefficients to the power space. The projection result reflects the power adjustment required to eliminate temperature deviation in the frequency domain. An inverse Fourier transform is then performed on the frequency-domain power excitation mode coefficients to transform them back from the frequency domain to the time domain, yielding the time-domain power excitation mode coefficients, which serve as the control vector. The control vector contains the contributions of each dominant control mode, representing the power adjustment in the mode space.

[0051] The control vector and the right singular vector are weighted and combined to achieve a mapping from modal space to physical space. During the weighted combination process, each element of the control vector serves as a weight, multiplied by its corresponding right singular vector, and then summed. The weighted combination result is mapped to the spatial location of each heating region through the dominant control mode, reconstructing the power adjustment amount corresponding to each heating region. Assuming there are 5 heating regions, 5 power adjustment amounts are obtained, each corresponding to the power adjustment required for each heating region. If the current power of a heating region is 200W, and the calculated power adjustment amount is 20W, then the adjusted power setting value is 220W.

[0052] In practical applications, it may be necessary to constrain the power regulation to ensure it does not exceed the operating range of the heating equipment. Upper and lower power limits can be set to truncate regulation exceeding these limits. To avoid frequent switching of the heating equipment's operating state, a minimum adjustment threshold can be set, allowing adjustment to be performed only when the power regulation exceeds the threshold. To smooth the control process and reduce temperature fluctuations, a low-pass filter can be introduced to smooth the power regulation, or a proportional-integral-derivative (PI-DE) control strategy can be used to adjust the application method of the power regulation.

[0053] The power adjustment calculated using the above method is applied to each heating zone to achieve precise control of the temperature field during steel heat treatment. Based on real-time temperature feedback, the temperature deviation field is continuously updated, and the power adjustment is repeatedly calculated to form a closed-loop control, ensuring that the temperature field gradually converges to the target temperature field. This invention achieves dimensionality reduction and precise control of the temperature field through mode decomposition and reconstruction technology, significantly improving control efficiency and accuracy. The control strategy based on the dominant control mode can effectively handle the spatial distribution characteristics of the temperature field, overcoming the limitations of traditional single-point control methods and achieving coordinated control of the entire temperature field. By combining frequency domain analysis and time domain synthesis, the dynamic characteristics of the system are fully utilized, improving the control system's ability to suppress temperature disturbances and the accuracy of temperature tracking.

[0054] Step 105: Input the power adjustment amount into the dynamic mapping relationship to predict the temperature evolution process, identify the peak value and corresponding heating region that exceed the preset target envelope from the temperature evolution process, suppress the power adjustment amount of the corresponding heating region, obtain the execution power command, and apply it to the heating region.

[0055] In some embodiments of the present invention, step 105 may specifically include the following sub-steps: Sub-step 1051: Substitute the power adjustment amount into the dynamic mapping relationship, calculate the temperature response trajectory of each heating area in the future time period to obtain the temperature evolution process; Sub-step 1052: Extract the temperature prediction values ​​of each heating region from the temperature evolution process, compare the temperature prediction values ​​with the upper boundary of the preset target envelope, and identify the temperature peak, the overshoot of the temperature peak exceeding the upper boundary, the heating region where the temperature peak is located, and the time when the temperature peak occurs. Sub-step 1053: Extract the thermal coupling coefficient between the heating region where the temperature peak is located and other heating regions in the dynamic mapping relationship, filter other heating regions with thermal coupling coefficients greater than the preset coupling threshold, and form a set of suppression regions by combining the heating region where the temperature peak is located with the other filtered heating regions. Sub-step 1054: Calculate the time interval between the current moment and the moment when the temperature peak occurs as the suppression time window, and calculate the ratio of the overshoot to the suppression time window to obtain the power change rate. Sub-step 1055: The power adjustment amount corresponding to each heating region in the set of suppressed regions is linearly decreased according to the power change rate to obtain the power adjustment amount after suppression; Sub-step 1056: The suppressed power adjustment amount is superimposed on the current power output to obtain the execution power command, and the execution power command is applied to each heating area.

[0056] In the temperature field control process of steel heat treatment, based on the power adjustment amount calculated above, temperature prediction and power suppression processing are required to avoid overheating caused by temperature overshoot. The power adjustment amount calculated in the previous steps is substituted into the dynamic mapping relationship to calculate the temperature response trajectory of each heating zone in the future time period, thus obtaining the temperature evolution process. The dynamic mapping relationship refers to the relationship model between power input and temperature response established earlier, which can be used to predict temperature changes caused by power changes. The temperature response trajectory calculation process can employ time-domain simulation or frequency-domain transformation methods. The time-domain simulation method directly uses the time-domain relationship between power input and temperature output for recursive calculation; the frequency-domain transformation method uses the frequency-domain transfer function to transform the power adjustment amount to the frequency domain, multiplies it with the transfer function, and then inversely transforms it back to the time domain. Assuming the predicted future time period is 60 seconds and the sampling interval is 0.5 seconds, the temperature values ​​at 120 sampling points need to be calculated.

[0057] Temperature prediction values ​​for each heating zone are extracted during the temperature evolution process. These prediction values ​​are compared with the upper boundary of a preset target envelope to identify temperature peaks and whether they exceed the upper boundary. The preset target envelope defines the allowable range for temperature control; the upper boundary represents the highest allowable temperature value, and the lower boundary represents the lowest allowable temperature value. Temperatures exceeding the upper boundary can lead to overheating or even burning of the steel, which must be prevented. By comparing the temperature prediction values ​​with the upper boundary, the temperature peak, the overshoot of the temperature peak exceeding the upper boundary, the heating zone where the temperature peak occurs, and the time when the temperature peak occurs can be identified. The temperature peak refers to the highest point in the temperature prediction trajectory, and the overshoot refers to the temperature difference between the temperature peak and the upper boundary. For example, if the predicted temperature peak for a heating zone is 1050℃, and the upper boundary is 1030℃, the overshoot is 20℃. The heating zone number where the temperature peak occurs and the time when the temperature peak occurs are also recorded to provide a basis for subsequent power suppression.

[0058] The thermal coupling coefficient between the heating region containing the temperature peak and other heating regions in the dynamic mapping relationship is extracted. Other heating regions with thermal coupling coefficients greater than a preset coupling threshold are then selected. The thermal coupling coefficient represents the heat transfer intensity between different heating regions and can be extracted from the off-diagonal elements of the dynamic mapping relationship. A larger thermal coupling coefficient indicates a stronger heat exchange between the two heating regions. Assuming a preset coupling threshold of 0.3, if the thermal coupling coefficient between heating region 1 and heating region 2 is 0.4, then heating region 2 is selected; if the thermal coupling coefficient between heating region 1 and heating region 3 is 0.2, then heating region 3 is not selected. The heating region containing the temperature peak and the selected other heating regions are combined into a suppression region set. The heating regions in this set need to undergo synchronous power suppression to collaboratively reduce the temperature peak.

[0059] The time interval from the current moment to the moment the temperature peak occurs is calculated as the suppression time window. The power change rate is obtained by calculating the ratio of the overshoot to the suppression time window. The suppression time window represents the length of time required for power suppression, and the power change rate represents the amount of power that needs to be reduced per unit time. For example, if the current moment is 10 seconds and the temperature peak occurs at 25 seconds, then the suppression time window is 15 seconds; if the overshoot is 20℃, then the power change rate is 1.33℃ / s. The power change rate will be used in subsequent power reduction calculations to ensure that the temperature can smoothly decrease to the target range.

[0060] The power adjustment amount corresponding to each heating region in the suppression region set is linearly decreased according to the calculated power change rate to obtain the suppressed power adjustment amount. Linear decrease means that the power adjustment amount gradually decreases at a fixed rate within the suppression time window. The formula for calculating the suppressed power adjustment amount is: current power adjustment amount minus the product of the power change rate and time. For example, if the current power adjustment amount is 100W, the power change rate is 5W / s, and the suppression time window is 15 seconds, then within the suppression time window, the power adjustment amount will be 95W, 90W, 85W, until it reaches 25W. The same linear decrease process is performed on each heating region in the suppression region set to ensure that the temperature peak can be effectively suppressed.

[0061] The suppressed power adjustment is superimposed on the current power output to obtain the executed power command. This executed power command is then applied to each heating zone. The superposition operation means adding the suppressed power adjustment to the current power output of each heating zone to obtain a new power setpoint. For example, if the current power output of a heating zone is 200W and the suppressed power adjustment is 25W, then the executed power command is 225W. For heating zones not in the suppressed zone set, their power adjustment remains unchanged. The calculated executed power command is applied to each heating zone through the controller to achieve precise control of the heating power. The executed power command can be applied using pulse width modulation (PWM), where the power output is controlled by adjusting the pulse width ratio.

[0062] After the power command is applied, the control system continuously monitors the actual temperature field and updates the power adjustment based on real-time temperature feedback, forming a closed-loop control. By periodically repeating the aforementioned prediction and suppression process, the system ensures that the temperature field remains within the preset target envelope, avoiding overheating or underheating. The control cycle can be set from 0.5 seconds to 2 seconds, with the specific value depending on the response speed and control accuracy requirements of the heat treatment process. A shorter control cycle improves control accuracy but increases the computational burden; a longer control cycle reduces the computational burden but may lead to a decrease in control accuracy.

[0063] This invention achieves proactive control of the steel heat treatment process, anticipating and preventing temperature overshoot, effectively ensuring heat treatment quality and equipment safety. By introducing a thermal coupling coefficient, the influence of heat exchange between different heating zones is fully considered, enabling multi-zone coordinated control. The linear power reduction strategy avoids temperature fluctuations caused by abrupt adjustments, ensuring a smooth transition of the temperature field.

[0064] like Figure 2 As shown, Figure 2 This is a schematic diagram of the structure of a steel heat treatment temperature field simulation and feedback control system provided in an embodiment of the present invention. The system includes: Temperature field calculation unit 201 is used to perform heat transfer calculations based on the furnace space grid and boundary conditions to obtain the predicted temperature field, and to collect the measured temperature and compare it with the predicted temperature field to obtain the temperature deviation field. The heat source tracing unit 202 is used to identify gradient anomaly regions in the temperature deviation field, extract the multi-time temperature evolution trajectory of the gradient anomaly regions, establish a temperature propagation reverse tracing model, solve the heat source disturbance distribution through the reverse tracing model, and compensate the heat source disturbance distribution to the predicted temperature field to obtain the corrected temperature field. The mapping construction unit 203 is used to apply power excitation to the heating area, collect the temperature response, construct a dynamic mapping relationship between power and temperature, and perform dimensionality reduction decomposition on the dynamic mapping relationship to extract the dominant control mode; The power solving unit 204 is used to decompose the difference between the corrected temperature field and the target temperature field on the dominant control mode to solve the control vector, and reconstruct the control vector into a power adjustment quantity through the dominant control mode. The power suppression unit 205 is used to predict the temperature evolution process by inputting the power adjustment amount into the dynamic mapping relationship, identify the peak value and corresponding heating area that exceed the preset target envelope from the temperature evolution process, suppress the power adjustment amount of the corresponding heating area, obtain the execution power command, and apply it to the heating area.

[0065] One technical solution provided in this embodiment of the invention is an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0066] One technical solution provided in this embodiment of the invention is a computer-readable storage medium storing a computer program, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0067] The specific embodiments described above are preferred embodiments of the present invention and are not intended to limit the specific scope of the present invention. The scope of the present invention includes, but is not limited to, these specific embodiments. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A method of simulation and feedback control of temperature field in steel heat treatment, characterized in that, Includes the following steps: The predicted temperature field is obtained by heat transfer calculation based on the furnace space grid and boundary conditions, and the temperature deviation field is obtained by comparing the measured temperature with the predicted temperature field. Identify gradient anomaly regions in the temperature deviation field, extract the multi-time temperature evolution trajectory of the gradient anomaly regions, establish a temperature propagation reverse tracing model, solve the heat source disturbance distribution through the reverse tracing model, and compensate the heat source disturbance distribution to the predicted temperature field to obtain the corrected temperature field. A power excitation is applied to the heating area to collect the temperature response, a dynamic mapping relationship between power and temperature is constructed, and the dynamic mapping relationship is decomposed by dimensionality reduction to extract the dominant control mode; The difference between the corrected temperature field and the target temperature field is decomposed and the control vector is solved on the dominant control mode. The control vector is then reconstructed into a power regulation quantity through the dominant control mode. The power regulation input is used to predict the temperature evolution process through dynamic mapping. Peak values ​​exceeding the preset target envelope and corresponding heating regions are identified during the temperature evolution process. The power regulation of the corresponding heating regions is suppressed to obtain the execution power command, which is then applied to the heating regions.

2. The method of claim 1, wherein, The predicted temperature field is obtained by heat transfer calculation based on the furnace spatial grid and boundary conditions. The temperature deviation field is obtained by comparing the measured temperature with the predicted temperature field. The furnace space is divided into a spatial grid, the thermal conductivity and specific heat capacity of the steel are obtained, the furnace wall temperature and furnace door opening state are obtained as boundary conditions, the heat transfer differential equation of the spatial grid is established, and the temperature distribution of the spatial grid is obtained by numerically solving the heat transfer differential equation. The temperature distribution constitutes the predicted temperature field. The measured temperature of the monitoring point inside the furnace is collected by a temperature sensor. The corresponding spatial grid is located in the predicted temperature field according to the spatial coordinates of the monitoring point. The temperature value of the corresponding spatial grid is extracted as the predicted temperature. The difference between the measured temperature and the predicted temperature is calculated to obtain the temperature deviation of the monitoring point. The spatial distance between each spatial grid and each monitoring point is calculated. The temperature deviation of each monitoring point is weighted and summed according to the spatial distance. The weighted summation result is assigned to each spatial grid to form a temperature deviation field.

3. The method according to claim 1, characterized in that, Identify gradient anomaly regions in the temperature deviation field, extract multi-time temperature evolution trajectories of these regions, establish a temperature propagation reverse tracing model, solve for the heat source disturbance distribution using this model, and compensate the heat source disturbance distribution to the predicted temperature field to obtain the corrected temperature field, including: Calculate the temperature gradient values ​​of each spatial grid in the temperature deviation field, average all temperature gradient values ​​to obtain the gradient reference value, and identify spatial grids with temperature gradient values ​​greater than the gradient reference value as gradient anomaly regions. The temperature deviation values ​​of the gradient anomaly region at multiple consecutive historical moments are obtained and arranged in chronological order to form a multi-moment temperature evolution trajectory; The time change rate of temperature deviation between adjacent time points in the multi-time temperature evolution trajectory is calculated. Based on the propagation law of heat attenuation with spatial distance, a quantitative relationship between the time change rate, spatial distance and heat source intensity is established as a reverse tracing model of temperature propagation. Calculate the spatial distance from each spatial grid to the gradient anomaly region; Substitute the temporal rate of change of the gradient anomaly region and the spatial distance of each spatial grid into the temperature propagation reverse tracing model, and solve the heat source intensity value when each spatial grid is a heat source in reverse. Identify the spatial grid with a heat source intensity value greater than a preset intensity threshold as the heat source location. The heat source location and the heat source intensity value constitute the heat source disturbance distribution. The heat source disturbance distribution is compensated to the corresponding spatial grid in the predicted temperature field to obtain the corrected temperature field.

4. The method according to claim 1, characterized in that, A power excitation is applied to the heating region to collect the temperature response, and a dynamic power-temperature mapping relationship is constructed. The dynamic mapping relationship is then decomposed in a dimensionality reduction manner to extract the dominant control modes, including: Multiple power excitation sequences with different amplitudes are applied to the heating area, and the temperature response time sequence corresponding to each power excitation sequence is collected in real time. Fourier transforms are performed on each power excitation sequence and the corresponding temperature response time sequence to obtain the power excitation spectrum and temperature response spectrum. The complex ratio of the temperature response spectrum to the power excitation spectrum at each frequency point is calculated to obtain the frequency response function under each power excitation amplitude. The frequency response function under each power excitation amplitude is organized into a two-dimensional array according to the power excitation amplitude and frequency point, and a dynamic mapping relationship between power and temperature is constructed. Singular value decomposition of dynamic mapping relationships yields singular values, left singular vectors, and right singular vectors; Calculate the energy percentage of the square of each singular value to the sum of the squares of singular values, and select the singular values ​​whose cumulative energy percentage reaches a preset energy threshold, along with their corresponding left and right singular vectors. The temperature response mode coefficients are obtained by performing an inner product operation between the temperature response time series and the left singular vector, and the power excitation mode coefficients are obtained by performing an inner product operation between the power excitation series and the right singular vector. The temperature response mode coefficients and the power excitation mode coefficients are then linearly related through singular values ​​to form the dominant control mode.

5. The method according to claim 1, characterized in that, The difference between the corrected temperature field and the target temperature field is decomposed and the control vector is solved on the dominant control mode. The control vector is then reconstructed into the power regulation quantity through the dominant control mode, including: The temperature deviation field is obtained by calculating the temperature difference between the corrected temperature field and the target temperature field at various spatial locations; The temperature deviation field is transformed into the frequency domain space to obtain the temperature deviation spectrum. The temperature deviation spectrum is then projected onto the temperature response subspace formed by the left singular vectors in the dominant control mode to obtain the frequency domain response mode coefficients of the temperature deviation field in each dominant control mode. The frequency domain normalized response mode coefficients are obtained by performing a ratio calculation between the singular values ​​of the frequency domain response mode coefficients and the corresponding dominant control modes. The frequency domain power excitation mode coefficients are obtained by projecting the normalized response mode coefficients in the frequency domain onto the right singular vector in the dominant control mode. The frequency domain power excitation mode coefficients are then transformed from the frequency domain space to the time domain space to obtain the time domain power excitation mode coefficients as the control vector. The control vector and the right singular vector are weighted and combined. The weighted combination result is then mapped to the spatial location of each heating region through the dominant control mode, and reconstructed into the power regulation amount corresponding to each heating region.

6. The method according to claim 1, characterized in that, The power regulation input is used to predict the temperature evolution process through a dynamic mapping relationship. Peak values ​​exceeding a preset target envelope and corresponding heating regions are identified during the temperature evolution process. Power regulation in the corresponding heating regions is suppressed to obtain an execution power command, which is then applied to the heating region. By substituting the power adjustment into the dynamic mapping relationship, the temperature response trajectory of each heating zone in the future time period is calculated to obtain the temperature evolution process. The predicted temperature values ​​of each heating region are extracted during the temperature evolution process. The predicted temperature values ​​are compared with the upper boundary of the preset target envelope to identify the temperature peak, the overshoot of the temperature peak exceeding the upper boundary, the heating region where the temperature peak is located, and the time when the temperature peak occurs. Extract the thermal coupling coefficient between the heating region where the temperature peak is located and other heating regions in the dynamic mapping relationship, filter other heating regions with thermal coupling coefficients greater than a preset coupling threshold, and form a set of suppression regions by combining the heating region where the temperature peak is located with the other filtered heating regions. The time interval from the current moment to the moment when the temperature peak occurs is calculated as the suppression time window, and the power change rate is obtained by calculating the ratio of the overshoot to the suppression time window. The power adjustment amount corresponding to each heating region in the set of suppressed regions is linearly decreased according to the power change rate to obtain the power adjustment amount after suppression. The suppressed power adjustment is superimposed on the current power output to obtain the execution power command, which is then applied to each heating zone.

7. A temperature field simulation and feedback control system for heat treatment of steel, used to implement the method described in any one of claims 1-6, characterized in that, The system includes: The temperature field calculation unit is used to perform heat transfer calculations based on the furnace space grid and boundary conditions to obtain the predicted temperature field, and to collect the measured temperature and compare it with the predicted temperature field to obtain the temperature deviation field. The heat source tracing unit is used to identify gradient anomaly regions in the temperature deviation field, extract the multi-time temperature evolution trajectory of the gradient anomaly regions, establish a temperature propagation reverse tracing model, solve the heat source disturbance distribution through the reverse tracing model, and compensate the heat source disturbance distribution to the predicted temperature field to obtain the corrected temperature field. The mapping construction unit is used to apply power excitation to the heating area, collect the temperature response, construct a dynamic mapping relationship between power and temperature, and perform dimensionality reduction decomposition on the dynamic mapping relationship to extract the dominant control mode; The power calculation unit is used to decompose the difference between the corrected temperature field and the target temperature field into a control vector on the dominant control mode, and reconstruct the control vector into a power adjustment quantity through the dominant control mode. The power suppression unit is used to input the power regulation amount into the dynamic mapping relationship to predict the temperature evolution process, identify the peak value and corresponding heating region that exceed the preset target envelope from the temperature evolution process, suppress the power regulation amount of the corresponding heating region, obtain the execution power command, and apply it to the heating region.

8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1 to 6.

Images