Energy-saving and emission-reducing system of annealing furnace

Abstract

The invention discloses an energy conservation and emission reduction system for an annealing furnace, which relates to the technical field of heat treatment, and comprises a target fitting module for performing parameter matching and fitting on a multi-dimensional environment feature vector and an annealing process constraint parameter set to form a heat treatment target function, and performing optimization solution to obtain a target temperature control interval; the control strategy module is used for carrying out multi-target control strategy calculation on the phase change temperature control window information through a heat treatment target function to obtain a dynamic temperature control path in a temperature rise stage; and the execution regulation and control module is used for controlling an annealing furnace heating execution mechanism according to the dynamic temperature control path, adjusting fuel gas input, an air proportion and a flow rate, adopting infrared thermal imaging and a distributed thermocouple to jointly measure the temperature of a temperature zone, obtaining temperature deviation data, carrying out constant-temperature heat preservation and self-adaptive cooling, and generating a multi-section cooling control instruction. Key requirements such as the heating rate, the heat preservation time and the structure property are fully considered, the control precision and the energy efficiency utilization rate of the annealing process are effectively improved, and the energy-saving and emission-reducing effects are remarkable.

Description

Technical Field

[0001] The present invention relates to the technical field of heat treatment, and particularly to an energy-saving and emission-reduction system for an annealing furnace. Background Art

[0002] With the continuous development of processing industries such as steel, aluminum alloy, and electronic materials, the intelligent control level of the annealing process in heat treatment technology has increasingly become an important factor affecting energy conservation and consumption reduction. As a key equipment for realizing the homogenization of material properties and the stabilization of microstructure, the operating energy consumption of an annealing furnace accounts for a significant proportion in the entire heat treatment process. Therefore, how to improve the temperature control accuracy and thermal energy utilization rate during annealing has become a hot issue in current industrial and academic research. In recent years, existing technologies have attempted to introduce a programmable logic controller (PLC) or an industrial personal computer (IPC) combined with temperature sensors for temperature zone feedback regulation. Some systems adjust the heating power based on PID or fuzzy control strategies, and achieve temperature field control by monitoring the temperature curve in real time. Although the above methods have improved the response ability and basic automation level, when facing complex annealing processes (such as those involving phase change processes, multi-stage heating, heat preservation, and cooling control), they still mainly rely on statically set control models and empirical parameters to drive, and it is difficult to adaptively cope with multi-source disturbances or non-linear temperature responses, resulting in low thermal energy utilization efficiency.

[0003] Therefore, the existing temperature control technologies for annealing furnaces generally have problems such as fixed control paths and difficulty in dynamically responding to changes in the environment and material states, and it is difficult to meet the urgent needs of high-performance heat treatment processes in terms of energy conservation and emission reduction. In view of the above problems, providing an energy-saving and emission-reduction system for an annealing furnace that can perform dynamic path regulation by combining multi-source environmental data and material thermal behavior belongs to the technical field of industrial heat treatment energy-saving control. Summary of the Invention

[0004] In view of the existing problems mentioned above, the present invention is proposed.

[0005] Therefore, the present invention provides an energy-saving and emission-reduction system for an annealing furnace to solve the problem of high energy consumption caused by inaccurate temperature control during the existing annealing process.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: The present invention provides an energy-saving and emission-reduction system for an annealing furnace, which includes a feature extraction module that collects multi-source heterogeneous data, preprocesses it, and generates a multi-dimensional environmental feature vector; a target fitting module that matches the multi-dimensional environmental feature vector with an annealing process constraint parameter set, fits and forms a heat treatment objective function, and performs optimization to obtain a target temperature control range; a phase change recognition module that uses the target temperature control range and real-time temperature data in the multi-dimensional environmental feature vector to identify the material phase change characteristic points, extracts the critical temperature range, and generates phase change temperature control window information; a control strategy module that performs multi-objective control strategy calculation on the phase change temperature control window information through the heat treatment objective function to obtain a dynamic temperature control path in the heating stage; and an execution regulation module that controls the annealing furnace heating actuator according to the dynamic temperature control path, adjusts the gas input, air ratio and flow rate, measures the temperature of the temperature zone by combining infrared thermal imaging and distributed thermocouples, obtains temperature deviation data, and performs constant temperature and heat preservation and adaptive cooling to generate multi-segment cooling control instructions.

[0007] As a preferred embodiment of the energy-saving and emission-reduction system for the annealing furnace of the present invention, wherein: the steps of collecting multi-source heterogeneous data, preprocessing it, and generating a multi-dimensional environmental feature vector are as follows: Collect the real-time temperature, gas composition, pressure and flow rate of each working condition point in the annealing furnace to obtain multi-source heterogeneous data; Use the timestamp synchronization algorithm to align multi-channel data, and combine the sliding window anomaly detection and linear interpolation methods to eliminate outliers and fill in missing items to obtain the cleaned multi-source heterogeneous data; Based on the Z-score normalization method, normalize the cleaned multi-source heterogeneous data, and use principal component analysis for compression and dimension reduction to extract the main environmental change factors and generate a preliminary feature vector; Adopt a weighted fusion algorithm to uniformly encode and fuse the preliminary feature vectors of different sensor channels to generate a multi-dimensional environmental feature vector.

[0008] As a preferred embodiment of the energy-saving and emission-reduction system for the annealing furnace of the present invention, wherein: the steps of matching the multi-dimensional environmental feature vector with the annealing process constraint parameter set, fitting and forming a heat treatment objective function, and performing optimization to obtain a target temperature control range are as follows: Based on the multi-source heterogeneous data, extract the heating rate, holding time and phase change temperature information of the current annealing task to generate an annealing process constraint parameter set; Perform mapping analysis on the multi-dimensional environmental feature vector and the annealing process constraint parameter set, and use the Euclidean distance matching and fuzzy clustering methods to identify the working condition mode to obtain the target working condition interval label; Use the weighted least squares method to construct a multi-objective function expression for heat treatment to form a heat treatment objective function; The genetic algorithm is used to solve the heat treatment objective function to obtain the target temperature control interval.

[0009] As a preferred embodiment of the annealing furnace energy-saving and emission-reduction system of the present invention, the following steps are included: a multi-objective function expression for heat treatment is constructed by using the weighted least squares method to form a heat treatment objective function, and the specific steps are as follows. Collect multi-objective process data of heat treatment to obtain an original sample set. Clean and normalize the original sample set to form a unified data expression format, and assign weighted coefficients to each objective according to the current annealing task to generate a process parameter data set. Use the weighted least squares method to fit the functional relationship between the process parameter data set and the target parameters to form a multi-objective function expression for the heat treatment process. Determine the optimal weight coefficients in the multi-objective function expression through optimization calculations to obtain a heat treatment objective function reflecting the balance relationship between the objectives.

[0010] As a preferred embodiment of the annealing furnace energy-saving and emission-reduction system of the present invention, the following steps are included: the target temperature control interval and the real-time temperature data in the multi-dimensional environmental feature vector are used to identify the material phase change characteristic points, extract the critical temperature interval, and generate phase change temperature control window information, and the specific steps are as follows. Extract the real-time temperature data of the annealing furnace from the multi-dimensional environmental feature vector and perform dimensionality reduction processing using the principal component analysis method. Use the mutation detection method to identify the phase change characteristic points of the material during the heating and cooling processes with the target temperature control interval and the dimensionality-reduced real-time temperature data. Use the sliding window fitting and local extreme value search method to extract the critical temperature interval of the material from the phase change characteristic points and perform interval intersection operation with the target temperature control interval to form phase change temperature control window information.

[0011] As a preferred embodiment of the annealing furnace energy-saving and emission-reduction system of the present invention, the following steps are included: use the mutation detection method to identify the phase change characteristic points of the material during the heating and cooling processes with the target temperature control interval and the dimensionality-reduced real-time temperature data, and the specific steps are as follows. Align the target temperature control interval and the dimensionality-reduced real-time temperature data on the same time axis as a temperature sequence, and perform smoothing processing using local weighted regression to obtain a smoothed temperature change curve. Use the numerical differentiation method to calculate the first derivative of the smoothed temperature curve to obtain a derivative sequence of the temperature change rate over time. Use the mutation point detection method to analyze the derivative sequence to identify the derivative change positions and obtain the phase change characteristic points.

[0012] As a preferred solution of the annealing furnace energy conservation and emission reduction system of the present invention, wherein: the multi-objective control strategy is solved for the phase change temperature control window information through the heat treatment objective function to obtain the dynamic temperature control path in the heating stage. The specific steps are as follows: Taking the heat treatment objective function as the optimization objective, converting the phase change temperature control window information into path constraint conditions to form a multi-objective optimization problem; Based on the multi-objective optimization problem, combining the target temperature control interval, the heating rate in the process constraint parameter set and the phase change temperature control window information, generating an initial heating path candidate sequence; Taking the minimization result of the heat treatment objective function as the goal, using the gradient descent method to iteratively solve the optimal heating path under the condition of satisfying the thermodynamic equilibrium constraint conditions within the phase change temperature control window; According to the optimal heating path, screening the temperature control paths that balance multiple objectives and phase change constraints to obtain the dynamic temperature control path in the heating stage.

[0013] As a preferred solution of the annealing furnace energy conservation and emission reduction system of the present invention, wherein: taking the minimization result of the heat treatment objective function as the goal, using the gradient descent method to iteratively solve the optimal heating path under the condition of satisfying the thermodynamic equilibrium constraint conditions within the phase change temperature control window. The specific steps are as follows: Based on the initial heating path, using numerical calculation methods to calculate the heat treatment objective function value and gradient information; Taking the gradient information as the optimization basis, using the gradient descent method to iteratively optimize the initial heating path parameters and dynamically adjusting the path in combination with the constraint optimization method; Iterating repeatedly until the objective function value meets the preset termination conditions to obtain the optimal heating path.

[0014] As a preferred solution of the annealing furnace energy conservation and emission reduction system of the present invention, wherein: controlling the annealing furnace heating actuator according to the dynamic temperature control path, adjusting the gas input, air ratio and flow rate, and using infrared thermal imaging and distributed thermocouples to jointly measure the temperature in the temperature zone to obtain temperature deviation data. The specific steps are as follows: Generating a heating actuator control instruction according to the dynamic temperature control path, and using the PID control algorithm to adjust the gas input, air ratio and flow rate; Using the infrared thermal imaging and distributed thermocouple joint measurement technology to collect the temperature in the annealing furnace temperature zone in real time, and using the error calculation method to compare the real-time temperature data with the dynamic temperature control path to calculate and obtain the temperature deviation data.

[0015] As a preferred solution of the annealing furnace energy conservation and emission reduction system of the present invention, wherein: performing constant temperature heat preservation and adaptive cooling to generate multi-segment cooling control instructions. The specific steps are as follows: Evaluate the conditions of the temperature stabilization stage based on the temperature deviation data, and adopt a constant temperature control algorithm according to the evaluation results to stabilize the temperature of the annealing furnace within the set target range; Combine the temperature change trend and the phase change temperature control window information, and use an adaptive control strategy to adjust the cooling parameters to generate multi-segment cooling control instructions.

[0016] The beneficial effects of the present invention are as follows: Through the matching relationship between multi-dimensional environmental characteristics and annealing process parameters, the working condition mode is identified by using Euclidean distance and fuzzy clustering, the heat treatment control target is determined by combining the weighted least squares method, and the genetic algorithm is used for optimization and solution to realize the adaptive adjustment and multi-objective coordinated control of the temperature control interval under complex working conditions. The key requirements such as heating rate, holding time and tissue performance are fully considered, the control accuracy and energy efficiency utilization rate of the annealing process are effectively improved, the product consistency is enhanced, and the energy conservation and emission reduction effect is remarkable. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only 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 work.

[0018] Figure 1 It is a schematic diagram of the energy conservation and emission reduction system of the annealing furnace in the present invention.

[0019] Figure 2 It is a flow chart of data processing in the present invention.

[0020] Figure 3 It is a flow chart of temperature control execution in the present invention.

[0021] Figure 4 It is a flow chart of dynamic temperature control path optimization in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0022] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention in conjunction with the drawings of the specification.

[0023] Many specific details are set forth in the following description in order to provide a thorough understanding of the present invention. However, the present invention may be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0024] Second, the "one embodiment" or "embodiment" referred to herein means a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it an individual or selectively mutually exclusive embodiment with other embodiments.

[0025] Referring to Figures 1 to 4 , this embodiment provides an energy-saving and emission-reduction system for an annealing furnace, including the following steps: A feature extraction module that collects multi-source heterogeneous data, preprocesses it, and generates a multi-dimensional environmental feature vector.

[0026] Collect the real-time temperature, gas composition, pressure, and flow rate of each working condition point in the annealing furnace to obtain multi-source heterogeneous data.

[0027] Specifically, thermocouples or infrared temperature measurement probes are set in the annealing furnace to collect real-time temperature data at fixed time intervals; a laser gas analyzer or an infrared gas analysis device is used to synchronously collect the concentrations of gas components such as oxygen, nitrogen, and hydrogen in the annealing furnace atmosphere at the same sampling frequency; then the absolute pressure value inside the annealing furnace is obtained in real time through a pressure transmitter, and the sampling time is synchronously recorded; a hot film type or vortex street type flow sensor is used to collect the gas flow rate changes at the working condition points. All collection operations are synchronized based on a unified time axis and are uploaded in real time through a data collection interface, and finally a multi-source heterogeneous data set including time stamps, temperature, gas composition, pressure, and flow rate is formed.

[0028] Use a time stamp synchronization algorithm to align multi-channel data, and combine a sliding window anomaly detection and linear interpolation method to remove outliers and fill in missing items to obtain cleaned multi-source heterogeneous data.

[0029] Specifically, high-precision time stamps in a unified format are added to the real-time temperature, gas composition, pressure, and flow rate data collected from each working condition point in the annealing furnace. For example, the UTC time format accurate to milliseconds is used; the time stamp synchronization algorithm is used to align the multi-channel data according to the time stamps, and a multi-channel data matrix based on a unified time series is constructed to ensure that each time node corresponds to a complete multi-channel measurement value; a sliding window with a fixed width is set. For example, 100 multi-channel data points are set as a window, and it slides in turn and calculates the average value and standard deviation of each channel within the window. The points that deviate from the average value by more than 3 times the standard deviation are marked as outliers and removed; the linear interpolation method is used to fill in the missing multi-channel data points caused by outlier removal or collection failure. The linear interpolation process is based on the numerical linear fitting of the adjacent two valid time points to estimate the missing points in the middle; the processed multi-channel data is re-saved as the cleaned multi-source heterogeneous data.

[0030] It should also be noted that the specific steps for constructing a multi-channel data matrix based on a unified time series are as follows: The original measurement data collected by each channel is scanned and matched according to timestamps through a timestamp synchronization algorithm, and the set of all timestamps that appear in all channels is extracted as the reference benchmark for the unified time series. Each time node in the unified time series is traversed, and it is checked in turn whether there is corresponding valid data at the time node for each channel. If it exists, the measured value is recorded; if it is missing, interpolation is performed in a preset manner or marked as a missing value. Taking the unified time series as the row index and the channel number as the column index, the data of each channel at the corresponding time node is filled in row by row to form a multi-channel data matrix of the unified time series.

[0031] Based on the Z-score normalization method, the cleaned multi-source heterogeneous data is normalized, and principal component analysis is used for compression and dimensionality reduction to extract the main environmental change factors and generate preliminary eigenvectors.

[0032] Specifically, calculate the mean and standard deviation of the data of each channel. The mean is the arithmetic average of the values of all sampling points, and the standard deviation is the arithmetic square root of the sum of the squares of the differences between each sampling point and the mean. Each channel data point is normalized in turn, and the principal component analysis method is used to process the normalized multi-source heterogeneous data. The specific steps include calculating the covariance matrix, solving the eigenvalues and corresponding eigenvectors of the covariance matrix, and selecting several principal components according to the eigenvalue size. Multiply the eigenvectors of the selected principal components by the normalized multi-source heterogeneous data matrix to extract the factors reflecting the main environmental changes and form preliminary eigenvectors.

[0033] It should also be noted that according to the principal component analysis method, multiple eigenvectors are determined. Each eigenvector contains the weight coefficients of each measurement multi-source heterogeneous data point in the corresponding normalized multi-source heterogeneous data. Traverse the normalized multi-source heterogeneous data matrix row by row in chronological order, and perform weighted combination operations on each row of multi-source heterogeneous data with the eigenvectors of each principal component respectively to obtain the projection values in the principal component direction at the corresponding time points. Repeat the operation to process the data of all time points in turn, and finally form a new set of time series data in each principal component direction. Each set of time series data represents a preliminary eigenvector and is used to reflect the main environmental change trend.

[0034] The weighted fusion algorithm is used to uniformly encode and fuse the preliminary eigenvectors of different sensor channels to generate a multi-dimensional environmental eigenvector.

[0035] Specifically, weight coefficients are determined for the preliminary feature vectors of each sensor channel. The weight coefficients are preset or calculated according to indicators such as the importance of the channel or the signal-to-noise ratio. The preliminary feature vectors of each sensor channel are weighted according to the corresponding weights, that is, each dimension value of the feature vector of each channel is multiplied by the corresponding weight. The weighted feature vectors of each channel are summed element by element in the corresponding dimension to obtain the fused feature vectors of each channel. The fused feature vectors of each channel are normalized to keep the fusion result within a unified numerical range, forming a multi-dimensional environmental feature vector.

[0036] The target fitting module matches the multi-dimensional environmental feature vector with the annealing process constraint parameter set, fits and forms a heat treatment target function, and performs optimization to solve for the target temperature control interval.

[0037] Based on multi-source heterogeneous data, the heating rate, holding time, and phase change temperature information of the current annealing task are extracted to generate an annealing process constraint parameter set.

[0038] Specifically, based on multi-source heterogeneous data, the original sensor information related to the current annealing task is screened out from the cleaned multi-source heterogeneous data. The current annealing task is generated by the annealing process planning module and includes requirements such as specific heating rate, holding time, and phase change temperature. The screened data is processed using time series analysis methods. The heating rate is extracted by calculating the rate of change of temperature over time, the holding time is determined by analyzing the duration during which the temperature is maintained within a specific range, and the phase change temperature is identified by recognizing the inflection points or plateau intervals of the temperature curve. The extracted heating rate, holding time, and phase change temperature information are formatted to generate a structured annealing process constraint parameter set.

[0039] It should also be noted that the specific process of determining the holding time by analyzing the duration during which the temperature is maintained within a specific range is as follows: Set the target temperature range for the holding stage, for example, set the specific range as the temperature between 680°C and 700°C. Traverse the temperature time series to identify the data segments where the temperature values are continuously within the range of 680°C to 700°C. Then calculate the time difference between the start timestamp and the end timestamp corresponding to the data segment to obtain the length of time during which the temperature is continuously maintained within the range. Finally, identify the time periods during which the duration is greater than the set target temperature range for the holding stage (for example, the duration is not less than 300 seconds) as valid holding time periods, and extract the corresponding start time, end time, and duration as the holding time.

[0040] Perform mapping analysis on the multi-dimensional environmental feature vector and the annealing process constraint parameter set, and use Euclidean distance matching and fuzzy clustering methods to identify the working condition mode to obtain the target working condition interval label.

[0041] Specifically, for each multi-dimensional environmental feature vector, calculate the Euclidean distance from it to each parameter vector in all annealing process constraint parameter sets to obtain a distance matrix, where each row corresponds to a multi-dimensional environmental feature vector and each column corresponds to a vector of an annealing process constraint parameter set; use the distance matrix as the input, execute the fuzzy clustering algorithm, calculate the membership degree values of each multi-dimensional environmental feature vector to all working condition categories, generate a membership degree matrix, reflecting the possibility of each feature vector belonging to each working condition category; set a membership degree threshold, such as 0.5, and for each multi-dimensional environmental feature vector, select the working condition category with the largest membership degree value as the dominant working condition category; associate the dominant working condition category with the corresponding multi-dimensional environmental feature vector, mark the target working condition interval label of the multi-dimensional environmental feature vector, and realize the mapping relationship between the multi-dimensional environmental feature vector and the annealing process constraint parameter set.

[0042] Use the weighted least squares method to construct the multi-objective function expression of heat treatment to form the heat treatment objective function.

[0043] Furthermore, collect the multi-objective process data of heat treatment to obtain the original sample set.

[0044] Specifically, collect the heating rate data through a temperature sensor respectively, record the temperature change at a preset time interval to obtain a heating rate curve; record the holding time through a timing device to accurately obtain the duration of the heating and holding stage; use an energy consumption measurement device to collect the energy consumption index and measure the electrical energy or fuel consumption value during the heat treatment process; use a metallographic microscope or tissue analysis equipment to detect the sample tissue performance, obtain microscopic structure parameters such as grain size and phase composition, and mark and summarize them according to time and process batches to form an original sample set including heating rate, holding time, energy consumption index and tissue performance.

[0045] For example, the heating rate is calculated by collecting the temperature change per second, and the example value is 20 °C / min; the holding time is recorded by a timing device, and the example value is 120 minutes; the energy consumption index is measured by a power meter, and the example value is 5 kWh; the tissue performance is measured by a microscope for the average grain diameter, and the example value is 15 microns.

[0046] Clean and normalize the original sample set to form a unified data expression format, and assign weighted coefficients to each target according to the current annealing task to generate a process parameter data set.

[0047] Specifically, check the target process data in the original sample set, remove missing values or outliers to ensure data integrity and accuracy; use the Z-score standardization method to perform standardization processing with a mean of zero and a variance of one on the heating rate, holding time, energy consumption index, and tissue properties respectively to eliminate the dimension difference and form a unified data expression format; determine the corresponding weighting coefficients according to the importance of each target in the current annealing task, such as assigning values through historical data; multiply each standardized target data by the corresponding weighting coefficient and summarize to generate a process parameter data set.

[0048] Use the weighted least squares method to fit the functional relationship between the process parameter data set and the target parameters to form a multi-objective function expression for the heat treatment process.

[0049] Specifically, the target parameters are derived from the actual measurement values in the current annealing task, including the heating rate, holding time, and phase transformation temperature; construct a multivariate function model form with the variables in the process parameter data set as independent variables and the target parameters as dependent variables; set weights according to the squared error of each sample point, and the weights can be assigned according to sample importance or confidence; use the weighted least squares method to calculate the parameters of the multivariate function model to minimize the weighted sum of squared errors; obtain the fitting function expression, map the process parameters to the target parameters, and form a multi-objective function expression for the heat treatment process.

[0050] For example, for the sample point assign the weight , where represents the weight of the sample point , represents the measurement standard deviation of the sample point , obtain the fitting coefficients by solving the weighted normal equations, and construct the relationship between the target parameter and the process parameter .

[0051] Determine the optimal weighting coefficient in the multi-objective function expression through optimization calculation to obtain a heat treatment objective function reflecting the balance relationship between the targets.

[0052] Specifically, based on the multi-objective function expression, initialize the weight coefficient vector to be optimized. The initial weight coefficient values can be set as small positive numbers uniformly distributed according to experience (such as 0.1 or 1.0) to ensure the stability of initial calculations. Then, set the error tolerance to determine whether the fitting process converges. Usually, it is set that the change in the objective function value between two consecutive iterations is less than the set threshold (for example, 1e-4). At the same time, set the maximum number of iterations (such as 100 to 1000 times) to avoid infinite loops in the iteration process and ensure a balance between the accuracy and efficiency of the optimization process. Substitute the initial weight coefficient values into the multi-objective function expression, and calculate the sum of the squared residuals between the predicted value and the actual target parameters as the error index. According to the error index, use the gradient descent method to calculate the partial derivative of the error with respect to the current weight coefficients to obtain the gradient vector. Use the gradient vector to adjust the current weight coefficients according to the preset learning rate, and update the current weight coefficient vector. Determine whether the error index is lower than the set error tolerance or the maximum number of iterations is reached. If not satisfied, return to recalculate the sum of the squared residuals and update the current weight coefficients, and iterate until the termination condition is met. Finally, output the converged current weight coefficient vector as the optimal weight coefficients of the multi-objective function expression to form a heat treatment objective function reflecting the balance relationship between the objectives.

[0053] It should also be noted that the specific process of using the gradient vector to adjust the current weight coefficients according to the preset learning rate and update the current weight coefficient vector is as follows: In each iteration, calculate the partial derivative of the current loss function with respect to each weight coefficient to obtain the corresponding gradient vector. Set the step size for each adjustment according to the preset learning rate. For example, the learning rate is set to 0.01. Then, multiply each component in the gradient vector by the learning rate to obtain the adjustment value for each weight coefficient. Subtract the corresponding adjustment value from the corresponding component in the weight coefficient vector to complete one update. For example, if the current value of a certain weight coefficient is 0.5, the corresponding gradient value is 0.2, and the learning rate is 0.01, then the updated weight coefficient is 0.5 - 0.01×0.2 = 0.498. The process is sequentially executed for all components in the weight coefficient vector, and after the update is completed, enter the next iteration.

[0054] The specific steps to determine whether the error index is lower than the set error tolerance are as follows: It is realized by comparing the currently calculated error value with the pre-set error tolerance. When the error value is less than or equal to the error tolerance, it is considered that the fitting reaches a satisfactory accuracy. To determine whether the maximum number of iterations is reached, compare the statistical iteration number with the preset maximum number of iterations. When the iteration number is equal to or exceeds the maximum number of iterations, stop the iteration process. When either condition is met, stop the optimization calculation; otherwise, continue with the next iteration.

[0055] Use the genetic algorithm to solve the heat treatment objective function to obtain the target temperature control range.

[0056] Specifically, according to the preset population size, for example, set to 100, 100 temperature control interval individuals are randomly generated. Each individual uses a fixed-length binary encoding to represent the parameters of the temperature control interval. For example, each individual consists of 16-bit encoding. The first 8 bits represent the upper temperature limit, and the last 8 bits represent the lower temperature limit. The encoding range corresponds to temperature values between 800°C and 1100°C. For each individual, the temperature control interval parameters are decoded and then substituted into the heat treatment objective function to calculate the fitness score, which is used to measure the quality of the heat treatment effect. According to the fitness score, the roulette wheel selection method is used to calculate the selection probability of each individual and make a selection based on the cumulative probability to generate a set of parent individuals. According to the preset crossover probability, for example, set to 0.8, the parent individuals are paired in pairs and a crossover point is randomly selected, and single-point or multi-point crossover is used to generate offspring individuals. With a preset mutation probability, for example, set to 0.05, a bit-by-bit mutation operation is performed on the coding sites of the offspring individuals, for example, changing a certain bit from 1 to 0 or from 0 to 1 to increase the population diversity. The parent and offspring are combined to form a new generation of population, and the fitness score of each individual in the new population is recalculated. It is judged whether the termination condition is satisfied, for example, the fitness reaches the error tolerance of 0.001 or the number of iterations reaches 500 times. If not satisfied, the selection, crossover, mutation, and fitness calculation steps are continued. When the termination condition is satisfied, the individual with the highest fitness score is selected from the population, and the corresponding temperature control interval is used as the final solution result.

[0057] The phase change recognition module uses the target temperature control interval and the real-time temperature data in the multi-dimensional environmental feature vector to identify the phase change characteristic points of the material, extract the critical temperature interval, and generate phase change temperature control window information.

[0058] Extract the real-time temperature data of the annealing furnace from the multi-dimensional environmental feature vector and perform dimensionality reduction processing using the principal component analysis method.

[0059] Specifically, for the specific dimension number corresponding to temperature in the multi-dimensional environmental feature vector, the data of each dimension is read one by one to form a temperature data sequence; the temperature data sequence is merged with the data of other relevant dimensions in the multi-dimensional environmental feature vector to form a multi-dimensional data matrix containing temperature and associated features; the mean value of the multi-dimensional data matrix is calculated, and the mean value of each dimension is calculated and used to centralize the data of the corresponding dimension, that is, each corresponding dimension data is subtracted by the mean value of the corresponding dimension to form a zero-mean data matrix; based on the zero-mean data matrix, the covariance matrix is calculated, and the elements of the covariance matrix are obtained by calculating the covariance between the data of different dimensions; the eigenvalue decomposition is performed on the covariance matrix to obtain all eigenvalues and corresponding eigenvectors; all eigenvalues are sorted in descending order, and the eigenvectors corresponding to the first several largest eigenvalues whose cumulative contribution rate reaches the preset cumulative contribution rate threshold are selected to form the principal component basis for dimensionality reduction; the zero-mean data matrix is projected onto the selected principal component basis to obtain the dimensionality-reduced data set, and the dimensionality-reduced multi-dimensional feature vector is output as the result.

[0060] It should also be noted that the specific steps for the preset cumulative contribution rate threshold: according to the requirement for the degree of information retention after dimensionality reduction, determine the value of the cumulative contribution rate threshold, for example, set it to 90%; sort all eigenvalues in descending order; calculate the cumulative sum of the contribution rates of the sorted eigenvalues in turn; when the cumulative contribution rate reaches or exceeds the preset 90% threshold, stop the accumulation and select the corresponding number of eigenvalues; the eigenvectors corresponding to the selected eigenvalues form the principal component basis for subsequent dimensionality reduction processing. For example, assume that the eigenvalue contribution rates of the original data are 38%, 22%, 15%, 10%, 5%, etc. in sequence. When the cumulative sum reaches 90% in order, select the eigenvectors corresponding to the first five eigenvalues as the principal component basis.

[0061] The target temperature control interval and the dimensionality-reduced real-time temperature data are used to identify the phase change characteristic points of the material during the heating and cooling processes by using the mutation detection method.

[0062] Furthermore, the target temperature control interval and the dimensionality-reduced real-time temperature data are aligned on the synchronous time axis to form a temperature sequence, and local weighted regression is used for smoothing processing to obtain a smoothed temperature change curve.

[0063] Specifically, align the time axes of the target temperature control range data and the downsampled real-time temperature data. Read the timestamp information in the target temperature control range data and the timestamp information in the downsampled real-time temperature data, and obtain the start and end points of the time for both respectively. Compare the start time of the target temperature control range data with the start time of the downsampled real-time temperature data. If they are inconsistent, select the later start time as the unified start time. Compare the end time of the target temperature control range data with the end time of the downsampled real-time temperature data. If they are inconsistent, select the earlier end time as the unified end time. Determine the unified start time and the unified end time as the effective time range. Based on the effective time range, set the sampling interval of the time axis, which is set to 1 second in the example. Resample the target temperature control range data and the downsampled real-time temperature data at the set sampling interval respectively. If the target temperature control range data is missing at a certain sampling time point, use linear interpolation to fill in the missing data value. If the downsampled real-time temperature data is missing at the sampling time point, also use linear interpolation to fill it in. After completing the time resampling and interpolation filling, match the target temperature control range data sequence with the downsampled real-time temperature data sequence one by one according to the sampling time points to form a temperature sequence containing the target temperature control range data and the real-time temperature data. When performing locally weighted regression processing on the temperature sequence, determine the locally weighted regression window width. In the example, select 21 time points as the window width, select the type of weight function, and use a cubic kernel function to calculate the weight coefficients. Traverse each time point in the temperature sequence, take the current time point as the window center, select the adjacent points within the window from the temperature sequence, calculate the distance between the adjacent points and the center point, and calculate the adjacent point weights according to the distance and the weight function. Use the time and value of the adjacent points and the corresponding weights, and use weighted least squares to fit a local first-order or second-order polynomial curve, and calculate the value of the fitted curve at the center point time as the smoothed temperature value. Complete the process point by point to generate a smoothed temperature change curve over the entire time range.

[0064] Use numerical differentiation methods to calculate the first derivative of the smoothed temperature curve to obtain a derivative sequence of the temperature change rate over time.

[0065] Specifically, the numerical differentiation method is used to calculate the first derivative of the smoothed temperature curve, and the temperature values and corresponding time series at each time point of the smoothed temperature change curve are obtained; the time intervals between adjacent time points are calculated according to the time series, and it is confirmed whether the time intervals are uniform. If they are not uniform, the corresponding intervals are calculated separately; for the first time point in the time series, the forward difference method is used to calculate the derivative value, that is, the temperature value of the second time point minus the temperature value of the first time point, and then divided by the interval between the two time points; for the last time point in the time series, the backward difference method is used to calculate the derivative value, that is, the temperature value of the last time point minus the temperature value of the penultimate time point, and then divided by the interval between the two time points; for all the time points in the middle of the time series, the central difference method is used to calculate the derivative value, that is, the temperature value after the current time point minus the previous temperature value, and then divided by the sum of the time intervals of the two time points; the first derivative values corresponding to all time points are arranged in chronological order to form a derivative sequence of the temperature change rate over time.

[0066] It should also be noted that the specific steps to confirm whether the time intervals are uniform are as follows: according to the time stamp sequence in the dimension-reduced real-time temperature data, the time differences between adjacent time points are calculated in turn to obtain a time interval sequence; the time interval sequence is traversed, and the differences between each time interval and the first time interval are compared. If the differences between all time intervals and the first time interval are within the preset allowable error range, it is judged that the time intervals are uniform; if there is a time interval whose difference from the first time interval exceeds the preset allowable error range, it is judged that the time intervals are not uniform; in the case of non-uniform time intervals, the specific values and occurrence frequencies of each different time interval are counted separately.

[0067] The derivative sequence is analyzed using the mutation point detection method to identify the positions where the derivative changes and obtain the phase change characteristic points.

[0068] Specifically, based on the calculated derivative sequence, a data set of time-derivative value correspondence is formed in chronological order; the differences between the derivative values of adjacent time points in the derivative sequence are calculated to form a difference sequence; the statistical characteristics of the difference sequence, such as the mean and standard deviation, are calculated to determine the discrimination threshold for mutation points; the discrimination threshold for mutation points is set, for example, set to the mean of the difference sequence plus several times the standard deviation (for example, twice the standard deviation), and this is used as the detection standard; the difference sequence is traversed, and each difference value is compared with the discrimination threshold for mutation points. When the difference value at a certain point exceeds the discrimination threshold for mutation points, the corresponding time position is marked as a candidate mutation point; the candidate mutation points are processed by merging adjacent points to avoid repeatedly determining adjacent points as mutation points, and the point with the most significant change is retained as the final mutation point; the time position of the final mutation point is output as the phase change characteristic point.

[0069] For example, in the example, the length of the derivative sequence is 1000, the mean of the difference sequence is 0.01, the standard deviation is 0.005, and the mutation point discrimination threshold is set to 0.01 + 2×0.005 = 0.02. Any time point with a difference value greater than 0.02 is marked as a mutation point.

[0070] Using the sliding window fitting and local extreme value search method, the critical temperature range of the material is extracted from the phase change characteristic points, and the interval intersection operation is performed with the target temperature control range to form the phase change temperature control window information.

[0071] Specifically, centered on the phase change characteristic points, a sliding window with a fixed size is set. For example, the window size is set to 10 time points, and it slides point by point along the time axis to cover the temperature data near the phase change characteristic points; the fitting method (such as polynomial fitting) is applied to the temperature data within each sliding window to calculate the fitting curve; local extreme points are searched in the fitting curve, and the temperature values and time positions corresponding to the extreme points are recorded; based on the set of local extreme points obtained by fitting all sliding windows, the upper and lower limits of the critical temperature range are determined, usually the minimum and maximum values of the extreme point temperatures; the determined critical temperature range is subjected to an interval intersection operation with the pre-obtained target temperature control range to obtain the phase change temperature control window information; the phase change temperature control window information, including the starting and ending temperature values and the corresponding time interval, is output for subsequent temperature control adjustment.

[0072] For example, in the example, the sliding window size is 10 time points, the order of the fitting polynomial is 3, the temperature range of the local extreme points is from 650°C to 720°C, the target temperature control range is from 600°C to 750°C, and the phase change temperature control window information after interval intersection is from 650°C to 720°C.

[0073] The control strategy module performs a multi-objective control strategy solution on the phase change temperature control window information through the heat treatment objective function to obtain the dynamic temperature control path in the heating stage.

[0074] Taking the heat treatment objective function as the optimization objective and transforming the phase change temperature control window information into path constraint conditions to form a multi-objective optimization problem.

[0075] Specifically, clarify the expression of the heat treatment objective function, determine the temperature and related process parameters to be optimized, and explain the physical meanings of each parameter; extract the upper and lower limits of the temperature range from the phase transformation temperature control window information, and construct an inequality expression for the path constraint conditions to ensure that the temperature control path always remains within the temperature range during the optimization process; establish a multi-objective optimization mathematical model that includes the heat treatment objective function and path constraint conditions, where the objective function is used to evaluate the heat treatment effect and the path constraint is used to limit the temperature change range; select an appropriate multi-objective optimization algorithm to solve the model and obtain the optimal temperature control parameter combination that satisfies the path constraint; verify the optimization result to confirm that the temperature control path meets the requirements of the phase transformation temperature control window and the heat treatment objective function reaches the predetermined optimization index, thus forming a multi-objective optimization problem.

[0076] It should also be noted that according to the technical requirements of the heat treatment process, determine the physical quantities to be optimized, such as residual stress, hardness, and tissue uniformity. Use the mathematical expression as the objective function, which usually takes temperature, time, or other process parameters as independent variables; collect experimental data or literature materials, analyze the influence relationship of each process parameter on the target physical quantity, and establish the objective function to ensure that the objective function can accurately reflect the heat treatment effect; define the evaluation index of the objective function, such as minimizing the residual stress or maximizing the tissue uniformity, and clarify the optimization direction and target value; formulate the objective function in the form of a mathematical expression with computability and continuity for subsequent optimization algorithms to solve. When establishing a multi-objective optimization mathematical model that includes the heat treatment objective function and path constraint conditions, first determine the objectives to be optimized according to the heat treatment requirements, such as temperature uniformity and energy consumption, and express them as objective functions using mathematical expressions respectively; then clarify the path constraint conditions such as the upper and lower limit temperature ranges of the temperature control interval, heating rate, cooling rate, and holding time, and use inequality or equality forms to limit them; then combine the objective function and the constraint conditions to form a constrained multi-objective optimization problem to ensure that multiple objectives are optimized simultaneously while meeting the path constraint conditions; finally, select a suitable variable coding and mathematical description according to the specific optimization algorithm to complete the establishment of the multi-objective optimization mathematical model.

[0077] Based on the multi-objective optimization problem, combine the target temperature control interval, the heating rate in the process constraint parameter set, and the phase transformation temperature control window information to generate a candidate sequence of the initial heating path.

[0078] Specifically, determine the starting temperature and the ending temperature of the target temperature control range, and divide it into multiple equally spaced temperature nodes. Extract the allowable value range of the heating rate from the process constraint parameter set, and calculate the corresponding heating time step for each temperature range. According to the phase change temperature control window information, judge whether each temperature node is within the phase change temperature control window. Set the heating rate to a smaller value within the phase change temperature control window range, and set the heating rate to a larger value outside the phase change temperature control window range, so as to construct a heating-up path that meets the constraint requirements. Generate multiple candidate sequences of initial heating-up paths through combinations of multiple heating rates.

[0079] It should also be noted that according to the requirements of the heat treatment process, determine the lowest temperature and the highest temperature of the target temperature control range as the starting temperature and the ending temperature, set a fixed temperature interval value, and equally divide the entire target temperature control range according to the temperature interval value to generate multiple temperature nodes in sequence, serving as the basic temperature reference sequence in the process of constructing the heating-up path. For each temperature node, obtain the corresponding temperature value of the temperature node, and compare the temperature value with the upper and lower limits of the temperature of the phase change temperature control window in sequence. The specific steps are as follows: judge whether the temperature value is greater than or equal to the lower limit temperature of the phase change temperature control window. If the judgment result is no, then the temperature node is not within the phase change temperature control window; if the judgment result is yes, then continue to judge whether the temperature value is less than or equal to the upper limit temperature of the phase change temperature control window. If the judgment result is yes, it is determined that the temperature node is within the phase change temperature control window; if the judgment result is no, then the temperature node is not within the phase change temperature control window. This process is repeated for each temperature node to complete the classification judgment of all temperature nodes.

[0080] Taking the minimization result of the heat treatment objective function as the goal, use the gradient descent method to iteratively solve the optimal heating-up path under the condition of satisfying the thermodynamic equilibrium constraint in the phase change temperature control window.

[0081] Furthermore, based on the initial heating-up path, use numerical calculation methods to calculate the heat treatment objective function value and gradient information.

[0082] Specifically, select the heating time points corresponding to each temperature node in the initial heating-up path to construct a temperature-time discrete sequence; substitute the temperature-time discrete sequence into the heat treatment objective function expression, and perform numerical integration or numerical approximation through the finite difference method or the chain rule to calculate the heat treatment objective function value point by point; use numerical gradient calculation methods, such as the forward difference or the central difference method, perturb the temperature value at each temperature node and recalculate the heat treatment objective function to obtain the difference between the function values before and after the perturbation, and divide it by the perturbation amount to obtain the gradient value of the heat treatment objective function at each temperature node; combine the gradient values at all nodes in sequence to form complete gradient information.

[0083] For example, the heat treatment objective function is to maximize the material hardness. The initial heating path has 10 temperature nodes in the range of 600°C to 750°C. When using the central difference method to calculate the gradient, 1°C is used as the perturbation amount. The temperature node values are perturbed upward and downward respectively, and the difference in the objective function values is calculated. Finally, a gradient value sequence corresponding to each node is obtained.

[0084] Based on the gradient information as the optimization basis, the gradient descent method is used to iteratively optimize the parameters of the initial heating path, and the path is dynamically adjusted in combination with the constraint optimization method.

[0085] Specifically, using the gradient information of the heat treatment objective function at each temperature node of the initial heating path, the descent direction of each temperature node is determined, and the learning rate is set as the step coefficient; the temperature values of each temperature node in the initial heating path are updated according to the gradient direction, and the new path parameters after the update are calculated; in combination with the phase change temperature control window information and the heating rate in the process constraint parameter set, the updated path is adjusted by the projection method constraint optimization method, while satisfying the path constraints, the gradient optimization direction is retained; the adjusted path is used as the starting point for the new round of iteration, and the steps of gradient calculation, path update and path adjustment are continued until the change of the heat treatment objective function meets the upper limit of the iteration times; finally, the converged path is output as the optimization result.

[0086] Iterate repeatedly until the objective function value meets the preset termination condition to obtain the optimal heating path.

[0087] Specifically, based on the initial heating path, the temperature node parameters are updated using the gradient descent method according to the gradient direction of the heat treatment objective function, and the updated path is adjusted using the constraint optimization method in combination with the phase change temperature control window information and the heating rate in the process constraint parameter set; the value of the heat treatment objective function corresponding to the updated heating path is recalculated using numerical calculation methods; the current heat treatment objective function value is compared with the heat treatment objective function value of the previous round of iteration to determine whether the difference is less than the convergence threshold in the preset termination condition; if the termination condition is not met, continue with the path update, path adjustment and objective function value calculation operations to form a new round of iteration; when the difference in the heat treatment objective function value between two consecutive iterations is less than the preset convergence threshold, or the maximum iteration number limit is reached, the iteration process is terminated; the current heating path is recorded as the optimal heating path.

[0088] It should also be noted that the process of determining whether the difference is less than the convergence threshold in the preset termination condition: Set the termination condition before the iteration starts, including two parameters, namely the convergence threshold and the maximum number of iterations. The convergence threshold can be set to 0.001, and the maximum number of iterations can be set to 100 times. After each round of iteration, calculate the absolute difference between the current round's heat treatment objective function value and the previous round's heat treatment objective function value; if the difference is less than the convergence threshold, that is, |current objective function value - previous round's objective function value| < 0.001, it is determined that convergence has been achieved and the termination condition is satisfied; otherwise, determine whether the maximum number of iterations has been reached. For example, determine whether the current iteration count has reached 100 times. If not, continue with the operations of path update, path adjustment, and objective function value calculation, and enter a new round of iteration.

[0089] According to the optimal heating-up path, screen the temperature control paths that meet the multi-objective balance and phase change constraint conditions to obtain the dynamic temperature control path during the heating-up stage.

[0090] Specifically, after obtaining the optimal heating-up path, extract the temperature values and corresponding time steps of each temperature node in the optimal heating-up path to form a complete temperature-time comparison sequence; based on the phase change temperature control window information in the process constraint parameter set, determine the temperature range in the phase change sensitive interval during the heating-up process, and set a numerical constraint on the heating rate within this range, for example, set it not to exceed 2 °C / min; then traverse each adjacent temperature node in the optimal heating-up path, calculate the corresponding heating rate, and eliminate the temperature sections that do not meet the rate limit within the phase change temperature control window; based on multiple objective functions set in the multi-objective optimization problem, such as residual stress, microstructure properties, and heating energy consumption, evaluate and sort the remaining path segments that meet the constraints, and retain the path segment combination with the best balance in each objective index; finally, splice the selected temperature nodes in the heating-up order to generate a continuous and executable dynamic temperature control path during the heating-up stage.

[0091] For example, the temperature of the optimal heating-up path rises from room temperature to 780 °C, which contains a total of 60 temperature nodes, and the time interval between adjacent nodes is 60 seconds. The temperature range of 460 °C - 530 °C in the path belongs to the phase change temperature control window, and the corresponding heating rate requirement is not to exceed 2 °C / min. Calculate the rate for every two temperature nodes. For example, the node temperatures are 480 °C and 500 °C, the time interval is 2 minutes, and the rate is 10 °C / min, which does not meet the requirements, so the path segment is eliminated; for the remaining path segments that meet the rate limit, calculate the residual stress, austenite grain size, and heating energy consumption, and screen out the path segments with residual stress below 80 MPa, grain size better than grade 8, and unit energy consumption not exceeding 3.2 kWh / kg, and finally splice them into the dynamic temperature control path during the heating-up stage.

[0092] The execution control module controls the heating actuator of the annealing furnace according to the dynamic temperature control path, adjusts the gas input, air ratio and flow rate, uses the combination of infrared thermal imaging and distributed thermocouples to measure the temperature of the temperature zone, obtains the temperature deviation data, and performs constant temperature insulation and adaptive cooling to generate multi-segment cooling control instructions.

[0093] Generate the control instruction for the heating actuator according to the dynamic temperature control path, and use the PID control algorithm to adjust the gas input, air ratio and flow rate.

[0094] Specifically, according to the set temperature values and time sequence of each temperature node in the dynamic temperature control path, generate the target temperature sequence corresponding to each time point and use it as the control target to input into the PID control algorithm; collect the current temperature of the heating cavity in real time through the thermocouple, compare the collected temperature with the target temperature, and calculate the temperature error value; then calculate the proportional term, integral term and differential term based on the current temperature error, error change rate and cumulative error respectively, and sum the three results after weighting to obtain the PID output control quantity; then map the PID output control quantity to the percentage of the opening of the gas valve, the percentage of the opening of the air regulating valve and the set value of the induced draft fan frequency, which are used to adjust the gas input, air ratio and flow rate respectively; finally, send the corresponding control instructions to the gas solenoid valve, air regulating valve and induced draft fan frequency converter through the industrial control bus to realize the dynamic closed-loop control of the heating process.

[0095] For example, the set temperature sequence of the dynamic temperature control path is 480°C, 500°C, 530°C, the corresponding time points are the 10th minute, the 20th minute, the 30th minute, and the target temperature change rate is 2.5°C / min; the actually collected temperature is 495°C, the target temperature is 500°C, and the error is -5°C; the set PID parameters are proportional coefficient Kp = 1.2, integral time Ti = 30s, differential time Td = 5s, calculate the proportional term as -6.0, the integral term as -2.0, the differential term as +1.5, and the output control quantity as -6.5; map -6.5 to reduce the opening of the gas valve by 3% and increase the opening of the air valve by 1%, and increase the induced draft fan frequency to 45Hz, and send the corresponding control instructions to each execution device according to the MODBUS protocol.

[0096] Adopt the combined measurement technology of infrared thermal imaging and distributed thermocouples to collect the temperature of the annealing furnace temperature zone in real time, and use the error calculation method to compare the real-time temperature data with the dynamic temperature control path to calculate and obtain the temperature deviation data.

[0097] Specifically, distributed thermocouples are arranged in multiple key temperature control regions inside the annealing furnace, and an infrared thermal imaging device is installed above the furnace chamber to non-contact scan and collect the surface temperature of the temperature zones, ensuring that the thermocouple collection points correspond to the infrared thermal imaging scan regions in space; the temperature measurement values of each thermocouple are read at regular intervals, and the temperature distribution data of the corresponding regions in the infrared thermal imaging images are synchronously extracted, and the two types of temperature data are respectively processed by linear interpolation and time alignment; the real-time temperature value of each measurement point at the current moment is compared with the set temperature at the corresponding time point in the dynamic temperature control path, and the temperature deviation is calculated using an error calculation method, and the error calculation method includes subtracting the target temperature from the real-time temperature; the error values of each measurement point are summarized to form a complete temperature deviation data set for the temperature zones of the annealing furnace.

[0098] For example, the dynamic temperature control path of a certain measurement point is set to 510 °C at the 20th minute, the temperature measured by the thermocouple is actually 505 °C, and the surface temperature of the area shown by the infrared thermal imaging is 508 °C. The weighted average of the two is used to obtain the real-time temperature of 506.8 °C, and the temperature deviation calculation result is 506.8 °C - 510 °C = -3.2 °C.

[0099] Evaluate the conditions for the temperature stabilization stage based on the temperature deviation data, and adopt a constant temperature control algorithm according to the evaluation results to stabilize the temperature of the annealing furnace within the set target range.

[0100] Specifically, perform real-time sliding window analysis on the temperature deviation data of each temperature measurement point in the annealing furnace within the set temperature holding interval, and statistically analyze the absolute value of the deviation of each measurement point within the selected time window; set the judgment threshold conditions for the temperature stabilization stage, for example, the absolute value of the deviation of each measurement point continuously remains within the range of ±2 °C for no less than 10 minutes, and statistically analyze the continuous duration of each measurement point meeting the conditions within the current window; when all measurement points meet the judgment threshold conditions for the temperature stabilization stage, it is determined that the annealing furnace enters the temperature stabilization stage; according to the constant temperature control algorithm, output a micro-amplitude dynamic adjustment to each heating actuator to maintain the constancy of the gas input, air ratio and flow rate, so that the temperature of each measurement point is maintained within the range of the set target temperature ± the specified tolerance of the dynamic temperature control path.

[0101] It should also be noted that the judgment conditions for the temperature stabilization stage are based on the preset temperature fluctuation threshold and duration requirements. Specifically, during the operation process, continuously monitor the real-time temperature changes of each temperature measurement point, and calculate the temperature change amplitude within the sliding time window. When the temperature change amplitudes of all measurement points do not exceed the preset fluctuation threshold (for example, ±0.5 °C) within the set continuous time period (for example, 30 seconds), it is determined that the entire annealing furnace enters the temperature stabilization stage. The judgment conditions are derived from the technical standards for the temperature uniformity and control accuracy requirements of the process in the furnace, ensuring that the temperature field has reached a stable state before the next stage of heat treatment operation, and avoiding adverse effects of temperature fluctuations on the performance of the workpiece.

[0102] Combining the temperature change trend with the phase change temperature control window information, an adaptive control strategy is used to adjust the cooling parameters to generate multi-segment cooling control instructions.

[0103] Specifically, the temperature data in the cooling stage is collected in real time, continuous temperature values are obtained through sensors and the temperature difference between adjacent time points is calculated to obtain the temperature change rate; it is determined whether the current temperature is within the temperature range defined by the phase change temperature control window, and the upper and lower limits of the temperature control window are determined and compared with the current temperature; according to the temperature change rate and the degree of temperature deviation within the temperature control window, the cooling parameters are dynamically adjusted according to the adaptive control strategy, specifically including the flow rate adjustment of the cooling medium, the setting of the cooling wind speed and the division of the cooling time length; the adjusted cooling parameters are segmented into several control time periods according to the time axis, and each time period corresponds to a cooling intensity value to form multi-segment cooling control instructions; the cooling control instructions for each time period are output according to the execution order to guide the cooling equipment to complete the dynamic adjustment of the cooling process and ensure that the cooling process accurately matches the temperature control requirements.

[0104] It should also be noted that during the path planning process, to determine whether the current temperature is within the phase change temperature control window, first, the upper and lower temperature limits of the phase change temperature control window need to be obtained, which are usually given by material phase change characteristic tests or prior knowledge. For example, the upper and lower limits are 540°C and 580°C respectively. During the execution, the current temperature value is read in real time and compared with the upper and lower limits of the phase change temperature control window: if the current temperature is greater than or equal to the lower limit and less than or equal to the upper limit, that is, 540°C ≤ current temperature ≤ 580°C, it is determined that the current temperature is within the phase change temperature control window range; otherwise, it is considered to be in the non-phase change section. This judgment result will be used for the rate adjustment of the subsequent heating path to ensure uniform heating at a lower rate within the phase change section and improve the heat treatment quality.

[0105] In summary, the present invention realizes the adaptive adjustment of the temperature control interval and the multi-objective coordinated control under complex working conditions by: the matching relationship between multi-dimensional environmental characteristics and annealing process parameters, using the Euclidean distance and fuzzy clustering to identify the working condition mode, combining the weighted least squares method to determine the heat treatment control target, and using the genetic algorithm for optimization and solution. It fully takes into account the key requirements such as the heating rate, holding time and tissue performance, effectively improves the control accuracy and energy efficiency utilization rate of the annealing process, enhances the product consistency, and has remarkable energy-saving and emission-reduction effects.

[0106] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. An energy-saving and emission-reduction system for an annealing furnace, characterized in that: including a feature extraction module that collects multi-source heterogeneous data, preprocesses it, and generates a multi-dimensional environmental feature vector; a target fitting module that matches the multi-dimensional environmental feature vector with the annealing process constraint parameter set, fits and forms a heat treatment objective function, and performs optimization to obtain the target temperature control range; a phase change recognition module that uses the target temperature control range and the real-time temperature data in the multi-dimensional environmental feature vector to identify the material phase change feature points, extracts the critical temperature range, and generates phase change temperature control window information; a control strategy module that performs multi-objective control strategy calculation on the phase change temperature control window information through the heat treatment objective function to obtain the dynamic temperature control path in the heating stage; an execution and regulation module that controls the annealing furnace heating actuator according to the dynamic temperature control path, adjusts the gas input, air ratio and flow rate, measures the temperature in the temperature zone by combining infrared thermal imaging and distributed thermocouples, obtains temperature deviation data, and performs constant temperature insulation and adaptive cooling to generate multi-stage cooling control instructions.

2. The energy-saving and emission-reduction system for annealing furnace according to claim 1, wherein: The steps of collecting multi-source heterogeneous data, preprocessing it, and generating a multi-dimensional environmental feature vector are as follows: Collect the real-time temperature, gas composition, pressure and flow rate of each working condition point in the annealing furnace to obtain multi-source heterogeneous data; Use the timestamp synchronization algorithm to align multi-channel data, and combine the sliding window anomaly detection and linear interpolation methods to remove outliers and fill in missing items to obtain the cleaned multi-source heterogeneous data; Normalize the cleaned multi-source heterogeneous data based on the Z-score normalization method, and perform compression and dimensionality reduction using principal component analysis to extract the main environmental change factors and generate a preliminary feature vector; Adopt a weighted fusion algorithm to uniformly encode and fuse the preliminary feature vectors of different sensor channels to generate a multi-dimensional environmental feature vector.

3. The annealing furnace energy conservation and emission reduction system according to claim 2, characterized in that: The steps of matching the multi-dimensional environmental feature vector with the annealing process constraint parameter set, fitting and forming a heat treatment objective function, and performing optimization to obtain the target temperature control range are as follows: Based on the multi-source heterogeneous data, extract the heating rate, holding time and phase change temperature information of the current annealing task to generate an annealing process constraint parameter set; Perform mapping analysis on the multi-dimensional environmental feature vector and the annealing process constraint parameter set, and use the Euclidean distance matching and fuzzy clustering methods to identify the working condition mode to obtain the target working condition interval label; Use the weighted least squares method to construct a multi-objective function expression for heat treatment to form a heat treatment objective function; Use the genetic algorithm to solve the heat treatment objective function to obtain the target temperature control range.

4. The energy-saving and emission-reduction system for annealing furnace according to claim 3, characterized in that: The steps of using the weighted least squares method to construct a multi-objective function expression for heat treatment to form a heat treatment objective function are as follows: Collect multi-objective process data of heat treatment to obtain an original sample set; Clean and normalize the original sample set to form a unified data expression format, and assign weighted coefficients to each target according to the current annealing task to generate a process parameter data set; Use the weighted least squares method to fit the functional relationship between the process parameter data set and the target parameters to form a multi-objective function expression for the heat treatment process; Determine the optimal weight coefficient in the multi-objective function expression through optimization calculation to obtain a heat treatment objective function that reflects the balance relationship between the targets.

5. The annealing furnace energy conservation and emission reduction system according to claim 3, characterized in that: Using the target temperature control range and the real-time temperature data in the multi-dimensional environmental feature vector to identify the material phase change characteristic points, extract the critical temperature range, and generate the phase change temperature control window information. The specific steps are as follows: Extract the real-time temperature data of the annealing furnace from the multi-dimensional environmental feature vector, and perform dimensionality reduction processing using the principal component analysis method; Use the mutation detection method to identify the phase change characteristic points of the material during heating and cooling by comparing the target temperature control range with the dimensionality-reduced real-time temperature data; Use the sliding window fitting and local extreme value search method to extract the critical temperature range of the material from the phase change characteristic points, and perform an interval intersection operation with the target temperature control range to form the phase change temperature control window information.

6. The annealing furnace energy conservation and emission reduction system according to claim 5, characterized in that: Use the mutation detection method to identify the phase change characteristic points of the material during heating and cooling by comparing the target temperature control range with the dimensionality-reduced real-time temperature data. The specific steps are as follows: Align the target temperature control range and the dimensionality-reduced real-time temperature data on the same time axis as a temperature sequence, and perform smoothing processing using locally weighted regression to obtain a smoothed temperature change curve; Use the numerical differentiation method to calculate the first derivative of the smoothed temperature curve to obtain a derivative sequence of the temperature change rate over time; Use the mutation point detection method to analyze the derivative sequence and identify the derivative change positions to obtain the phase change characteristic points.

7. The annealing furnace energy conservation and emission reduction system according to claim 5, characterized in that: Perform a multi-objective control strategy solution on the phase change temperature control window information through the heat treatment objective function to obtain the dynamic temperature control path in the heating stage. The specific steps are as follows: Take the heat treatment objective function as the optimization objective, and transform the phase change temperature control window information into path constraint conditions to form a multi-objective optimization problem; Based on the multi-objective optimization problem, combine the target temperature control range, the heating rate in the process constraint parameter set, and the phase change temperature control window information to generate an initial heating path candidate sequence; Take the minimization result of the heat treatment objective function as the goal, and use the gradient descent method to iteratively solve the optimal heating path under the condition of satisfying the thermodynamic equilibrium constraint conditions within the phase change temperature control window; According to the optimal heating path, screen the temperature control paths that meet the multi-objective balance and phase change constraint conditions to obtain the dynamic temperature control path in the heating stage.

8. The annealing furnace energy conservation and emission reduction system according to claim 7, wherein: Take the minimization result of the heat treatment objective function as the goal, and use the gradient descent method to iteratively solve the optimal heating path under the condition of satisfying the thermodynamic equilibrium constraint conditions within the phase change temperature control window. The specific steps are as follows: Based on the initial heating path, use the numerical calculation method to calculate the heat treatment objective function value and gradient information; Take the gradient information as the optimization basis, use the gradient descent method to iteratively optimize the parameters of the initial heating path, and dynamically adjust the path in combination with the constraint optimization method; Iterate repeatedly until the objective function value meets the preset termination condition to obtain the optimal heating path.

9. The annealing furnace energy conservation and emission reduction system according to claim 7, wherein: Control the heating actuator of the annealing furnace according to the dynamic temperature control path, adjust the gas input, air ratio and flow rate, and use infrared thermal imaging and distributed thermocouples to jointly measure the temperature in the temperature zone to obtain temperature deviation data. The specific steps are as follows: Generate heating actuator control instructions according to the dynamic temperature control path, and use the PID control algorithm to adjust the gas input, air ratio and flow rate; Adopt the combined measurement technology of infrared thermal imaging and distributed thermocouples to collect the temperature of the annealing furnace temperature zone in real time, and use the error calculation method to compare the real-time temperature data with the dynamic temperature control path to calculate and obtain the temperature deviation data.

10. The annealing furnace energy conservation and emission reduction system according to claim 9, characterized in that: Perform constant temperature holding and adaptive cooling to generate multi-segment cooling control instructions. The specific steps are as follows: Evaluate the conditions of the temperature stabilization stage according to the temperature deviation data, and use the constant temperature control algorithm according to the evaluation results to stabilize the annealing furnace temperature within the set target range; Combine the temperature change trend and the phase change temperature control window information, and use the adaptive control strategy to adjust the cooling parameters to generate multi-segment cooling control instructions.

Images