A method to improve the uniformity of homogeneous cooling

Abstract

This invention relates to the field of industrial aluminum alloys and discloses a method for improving the uniformity of homogeneous cooling. The method includes: determining the volume of homogeneous cooling water and the initial opening degree of the homogeneous cooling chamber water valve based on the temperature of the bar; extracting the chaotic state characteristics of the homogeneous cooling water based on chaos theory, performing multi-step prediction and risk assessment, and training a neural network model of the adjustment amount of the homogeneous cooling water temperature and the opening degree of the homogeneous cooling chamber water valve under the constraint of a stable operating range; outputting the opening degree adjustment amount through the neural network model, and using a co-evolutionary mechanism to optimize and decompose the water valve opening degree adjustment amount into independent evolutionary subpopulations of temperature deviation, energy consumption, chaotic stability, and response smoothness, and dynamically weighting and combining them to obtain the globally optimal opening degree adjustment amount; determining the final opening degree of the homogeneous cooling chamber water valve based on the globally optimal opening degree adjustment amount corresponding to the current homogeneous cooling water temperature and the initial opening degree of the homogeneous cooling chamber water valve. This achieves precise, stable, and efficient control of homogeneous cooling.

Description

Technical Field

[0001] This invention relates to the field of industrial aluminum alloys, and more specifically, to a method for improving the uniformity of homogeneous cooling. Background Technology

[0002] The cooling rate after homogenization annealing has a significant impact on the precipitation behavior of wrought aluminum alloys. At a cooling rate of 100℃ / h, the precipitated Mg₂Si particles have a size of 2.0µm; at a cooling rate of 600℃ / h, the precipitated Mg₂Si particles have a size of 0.25µm. That is, different cooling rates will result in Mg₂Si precipitates of different sizes, leading to different mechanical properties in the material. Even with the same cooling water volume, different water temperatures will result in different cooling rates.

[0003] However, due to factors such as air temperature, the temperature difference of the cooling water entering the homogenization cooling chamber is large, resulting in uneven cooling effect of water mist cooling after homogenization of aluminum alloy round ingots, uneven material size, and ultimately uneven mechanical properties of aluminum alloy materials.

[0004] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0005] In view of the problems in the related technologies, the present invention proposes a method to improve the uniformity of homogeneous cooling, so as to enable the material to obtain uniform and consistent mechanical properties and reduce the fluctuation of the material's mechanical properties, thereby overcoming the above-mentioned technical problems existing in the existing related technologies.

[0006] Therefore, the specific technical solution adopted by the present invention is as follows:

[0007] According to a first aspect of the present invention, a method for improving the uniformity of homogeneous cooling is provided, comprising:

[0008] The amount of homogeneous cooling water is determined based on the temperature of the bar stock, and the initial opening of the water valve in the homogeneous cooling chamber is determined based on the amount of homogeneous cooling water.

[0009] Based on chaos theory, the chaotic state characteristics of homogeneous cooling water are extracted, and multi-step prediction and risk assessment are performed. Under the constraint of stable operation range, a neural network model of homogeneous cooling water temperature and the opening adjustment of homogeneous cooling chamber water valve is trained.

[0010] The valve opening adjustment amount is output through a neural network model, and the valve opening adjustment amount is optimized and decomposed into independent evolutionary subpopulations of temperature deviation, energy consumption, chaotic stability and response smoothness using a co-evolutionary mechanism. The global optimal opening adjustment amount is obtained by dynamically weighting the subpopulations.

[0011] Based on the global optimal opening adjustment amount corresponding to the current homogeneous cooling water temperature and the initial opening of the homogeneous cooling chamber water valve, the final opening of the homogeneous cooling chamber water valve is determined to improve the homogeneous cooling uniformity of the bar stock.

[0012] Furthermore, determining the amount of homogeneous cooling water based on the temperature of the bar includes:

[0013] Calculate the temperature difference based on the current temperature and target temperature of the bar; calculate the waste heat energy requirement of the bar based on the specific heat capacity, weight of the bar, and temperature difference.

[0014] Calculate the actual value of heat absorbed by the cooling water based on the waste heat energy demand and heat transfer efficiency of the bar stock.

[0015] Calculate the volume of homogeneous cooling water based on the actual heat absorbed by the cooling water and the heat absorption capacity per unit mass of the cooling water.

[0016] Furthermore, based on chaos theory, the chaotic state characteristics of homogeneous cooling water are extracted, and multi-step prediction and risk assessment are performed, including:

[0017] Obtain the time series of homogeneous cooling water temperature. Based on chaos theory, calculate the maximum Lyapunov exponent, correlation dimension, and fractal dimension of the homogeneous cooling water temperature time series. Then, construct the feature vector of the chaotic state of the homogeneous cooling water using the values ​​of the maximum Lyapunov exponent, correlation dimension, and fractal dimension.

[0018] The chaotic state feature vector is input into a time series-based multi-step prediction model to progressively predict the chaotic state feature vector at multiple future time steps.

[0019] Risk is determined at each prediction time step based on the magnitude of the maximum Lyapunov exponent. If the exponent is greater than the risk threshold, the state is classified as chaotic; if it is less than or equal to the risk threshold, the state is classified as stable.

[0020] Furthermore, the neural network model for training the homogeneous cooling water temperature and the adjustment amount of the homogeneous cooling chamber water valve opening under the constraint of stable operating range includes:

[0021] Training sample pairs are constructed from the temperature time series corresponding to the steady-state time step and the water valve opening adjustment amount under the corresponding time step conditions.

[0022] The artificial neural network model is trained using training samples so that it can predict the mapping relationship between the homogeneous cooling water temperature and the adjustment amount of the homogeneous cooling chamber water valve opening under chaotic state characteristic input conditions.

[0023] Furthermore, to obtain the homogeneous cooling water temperature time series, based on chaos theory, the maximum Lyapunov exponent, correlation dimension, and fractal dimension of the homogeneous cooling water temperature time series are calculated, including:

[0024] Preprocess the continuous homogeneous cooling water temperature data sequence at a fixed sampling time interval to obtain a normalized and denoised homogeneous cooling water temperature time sequence;

[0025] The time delay parameter and embedding dimension are determined, and the homogeneous cooling water temperature time series is converted into a target dimension phase space reconstructed trajectory matrix by combining the delay coordinate reconstruction method;

[0026] Find the two closest points in the reconstructed trajectory matrix in the phase space of the target dimension, calculate the spatial distance between the two closest points as they evolve over time, and take the natural logarithm of the change in spatial distance.

[0027] A linear fit is made between the natural logarithm of the spatial distance change and time, and the slope of the fitted line is used as the maximum Lyapunov exponent.

[0028] In the trajectory matrix reconstructed in the phase space of the target dimension, different spatial radius values ​​are selected, and for each radius, the proportion of times the distance between pairs of trajectory points is less than the radius is calculated to obtain the correlation integral.

[0029] Plot the relationship curve between the correlation integral value and the spatial radius in a double logarithmic coordinate system, and use the slope of the relationship curve as the correlation dimension;

[0030] Within the selected spatial range of the trajectory matrix in the target dimension phase space, cover the entire trajectory space with a cubic grid of a specified side length; calculate the number of cubic grids occupied by the trajectory points;

[0031] In a double logarithmic coordinate system, the curve relating the number of cubic lattice cells to the reciprocal of the lattice side length is fitted, and the slope of the fitted line is used as the fractal dimension.

[0032] Furthermore, the chaotic state feature vector is input into a time-series-based multi-step prediction model to progressively predict the chaotic state feature vectors for multiple future time steps, including:

[0033] Determine the architecture of the multi-step prediction model, and make the input of the multi-step prediction model a chaotic state feature vector and the output a prediction feature vector for a single time step.

[0034] By using a sliding window to extract samples of the time series of chaotic state feature vectors during historical operation, the input window sequence and output window sequence used for model training are obtained.

[0035] A multi-step prediction model is trained based on the input window sequence and the output window sequence, and the weights of the multi-step prediction model are trained using an optimization algorithm to minimize the error index between the predicted value and the feature vector of the real chaotic state, thus obtaining the trained multi-step prediction model.

[0036] Input the current chaotic state feature vector into the trained multi-step prediction model, and output the chaotic state feature vectors for multiple future time steps.

[0037] Furthermore, the valve opening adjustment amount is output through a neural network model, and the valve opening adjustment amount is optimized and decomposed into independent evolutionary subpopulations of temperature deviation, energy consumption, chaotic stability, and response smoothness using a co-evolutionary mechanism. The globally optimal valve opening adjustment amount is obtained through dynamic weighted combination, including:

[0038] Input the current chaotic state feature vector into the neural network model, and obtain the predicted value of the homogeneous cooling chamber water valve opening adjustment under the current chaotic state condition through the forward calculation of the neural network model;

[0039] The optimization problem of water valve opening adjustment is constructed and decomposed into subproblems of minimizing temperature deviation, minimizing energy consumption, improving chaotic stability, and improving water valve response smoothness. Independent evolutionary populations are created for each subproblem.

[0040] Each independent evolutionary population is allowed to evolve independently, and the globally optimal opening adjustment is obtained by dynamically weighting the population.

[0041] Furthermore, each independent evolutionary population is subjected to independent evolution, and the globally optimal opening adjustment is obtained through dynamic weighting, including:

[0042] A genetic algorithm optimizer is assigned to each independent evolutionary population to independently evolve and generate the local optimal valve opening adjustment amount under the corresponding objective function;

[0043] The local optimal opening adjustment is combined by dynamic weighting to form the global water valve opening adjustment.

[0044] Feedback updates are performed under a comprehensive fitness function that evaluates temperature deviation, energy consumption, water valve response smoothness, and chaotic state stability, and the iteration continues until the global fitness converges.

[0045] The formula for calculating the global valve opening adjustment, which combines the local optimal opening adjustments using dynamic weighting, is as follows:

[0046] ;

[0047] In the formula, This is the global water valve opening adjustment amount. The local optimal water valve opening adjustment amount for the temperature deviation target; The local optimal water valve opening adjustment amount for the energy consumption target; The local optimal valve opening adjustment amount is for the target of valve response smoothness. The local optimal valve opening adjustment amount is for the chaotic stability objective; The dynamic weights for the temperature deviation target; The dynamic weights for energy consumption targets; The dynamic weights for the target smoothness of the water valve response; The dynamic weights are the target for chaotic stability.

[0048] Furthermore, based on the globally optimal opening adjustment amount corresponding to the current homogeneous cooling water temperature and the initial opening of the homogeneous cooling chamber water valve, the final opening of the homogeneous cooling chamber water valve is determined, including:

[0049] Based on the current homogeneous cooling water temperature, the corresponding opening adjustment amount is output through the neural network model, and the corresponding global optimal opening adjustment amount is obtained by updating using the co-evolution mechanism.

[0050] The final opening of the homogeneous cooling chamber water valve is obtained by adding the global optimal opening adjustment corresponding to the current homogeneous cooling water temperature to the initial opening of the homogeneous cooling chamber water valve.

[0051] Furthermore, the temperature of the homogeneous cooling water is obtained using a temperature measuring structure, which includes:

[0052] The cooling water inlet pipe of the cooling chamber is located on the side of the outer wall of the homogeneous cooling chamber. A temperature measuring thermocouple is installed on the cooling water inlet pipe. An electronic display screen is installed on the outer wall of the homogeneous cooling chamber. The electronic display screen and the temperature measuring thermocouple are connected by a signal line. The signal line is covered with a steel pipe.

[0053] According to a second aspect of the present invention, a system for improving the uniformity of homogeneous cooling is provided, comprising:

[0054] The initial opening determination module is used to determine the amount of homogeneous cooling water based on the temperature of the bar, and to determine the initial opening of the homogeneous cooling chamber water valve based on the amount of homogeneous cooling water.

[0055] The chaotic feature extraction and model training module is used to extract the chaotic state features of homogeneous cooling water based on chaos theory, and to perform multi-step prediction and risk assessment. Under the constraint of stable operation range, it trains a neural network model of homogeneous cooling water temperature and the opening adjustment of homogeneous cooling chamber water valve.

[0056] The valve opening adjustment quantity co-evolution module is used to output the valve opening adjustment quantity through a neural network model, and use the co-evolution mechanism to optimize and decompose the valve opening adjustment quantity into independent evolution of target subpopulations of temperature deviation, energy consumption, chaotic stability and response smoothness, and obtain the globally optimal valve opening adjustment quantity through dynamic weighted combination.

[0057] The final opening acquisition module is used to determine the final opening of the homogeneous cooling chamber water valve based on the global optimal opening adjustment amount corresponding to the current homogeneous cooling water temperature and the initial opening of the homogeneous cooling chamber water valve, so as to improve the homogeneous cooling uniformity of the bar.

[0058] According to a third aspect of the present invention, a computer device is provided.

[0059] The computer device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the above method.

[0060] According to a fourth aspect of the present invention, a computer-readable storage medium is provided.

[0061] The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the above method.

[0062] The beneficial effects of this invention are as follows:

[0063] 1. This invention combines physical calculations, chaotic predictions, and co-evolutionary optimization to achieve precise, stable, and efficient control of homogeneous cooling.

[0064] 2. By calculating the physical chain of bar temperature, cooling water volume, and initial valve opening, precise water volume matching based on the bar's thermal state is achieved, rather than relying on experience to set the initial opening. This makes the cooling process more in line with the bar's current thermal requirements, significantly reducing over- and under-cooling, improving the cooling uniformity of the bar's cross-section and length, and improving the stability of the final product quality from the source.

[0065] 3. Based on chaos theory, the maximum Lyapunov exponent, correlation dimension, and fractal dimension of the cooling water temperature are extracted to construct a chaotic state feature vector. Through multi-step time series prediction, future chaotic trends can be predicted in advance. Combined with risk assessment, control measures can be taken before risks occur. This advance control avoids the disruption of cooling water distribution caused by chaotic temperature fluctuations, ensuring the system always operates within a stable range.

[0066] 4. The opening adjustment predicted by the neural network is decomposed into four objective subpopulations: temperature deviation, energy consumption, chaotic stability, and water valve response smoothness. These subpopulations evolve independently, and the global optimal solution is obtained through dynamic weight fusion. This achieves a dynamic balance between control objectives, ensuring cooling performance while avoiding mechanical losses caused by frequent and large-amplitude water valve movements, and reducing energy consumption, enabling the control strategy to be executed long-term. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 This is a flowchart of a method for improving the uniformity of homogeneous cooling according to an embodiment of the present invention;

[0069] Figure 2 This is a schematic diagram of a temperature measuring structure according to an embodiment of the present invention;

[0070] Figure 3 This is a schematic diagram of a hook bracket according to an embodiment of the present invention;

[0071] Figure 4 This is a schematic diagram of a hook according to an embodiment of the present invention;

[0072] Figure 5 This is a schematic diagram of a connecting mechanism according to an embodiment of the present invention;

[0073] Figure 6 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention.

[0074] In the picture:

[0075] 1. Cooling water inlet pipe for cooling chamber; 2. Temperature measuring thermocouple; 3. Signal line; 4. Electronic display screen; 5. Steel pipe; 6. Outer wall of homogeneous cooling chamber; 7. Hook bracket; 8. Hook. Detailed Implementation

[0076] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0077] Figure 1 An embodiment of a method for improving the uniformity of homogeneous cooling according to the present invention is shown, which enables the material to obtain uniform and consistent mechanical properties and reduces fluctuations in the material's mechanical properties.

[0078] In this optional embodiment, a method for improving the uniformity of homogeneous cooling includes:

[0079] S1. Determine the amount of homogeneous cooling water based on the temperature of the bar, and determine the initial opening of the homogeneous cooling chamber water valve based on the amount of homogeneous cooling water.

[0080] S2. Based on chaos theory, extract the chaotic state characteristics of homogeneous cooling water, perform multi-step prediction and risk assessment, and train a neural network model of homogeneous cooling water temperature and the opening adjustment of homogeneous cooling chamber water valve under the constraint of stable operation range.

[0081] S3. The valve opening adjustment amount is output through a neural network model, and the valve opening adjustment amount is optimized and decomposed into independent evolutionary subpopulations of temperature deviation, energy consumption, chaotic stability and response smoothness using a co-evolutionary mechanism. The global optimal opening adjustment amount is obtained by dynamically weighting and combining these subpopulations.

[0082] S4. Based on the global optimal opening adjustment amount corresponding to the current homogeneous cooling water temperature and the initial opening of the homogeneous cooling chamber water valve, determine the final opening of the homogeneous cooling chamber water valve to improve the homogeneous cooling uniformity of the bar.

[0083] In this optional embodiment, determining the amount of homogeneous cooling water based on the temperature of the bar includes:

[0084] Calculate the temperature difference based on the current and target temperatures of the bar; calculate the waste heat energy requirement of the bar based on its specific heat capacity, weight, and temperature difference; calculate the actual heat absorbed by the cooling water based on the waste heat energy requirement and heat transfer efficiency; and calculate the volume of homogeneous cooling water based on the actual heat absorbed by the cooling water and its heat absorption capacity per unit mass.

[0085] Utilizing the principle of heat conservation and heat transfer theory, the required residual heat energy is calculated by combining the difference between the current and target temperatures of the bar with its specific heat capacity and weight. Then, using the heat transfer efficiency coefficient, the heat to be released is converted into the actual heat absorbed by the cooling water. Finally, the required heat volume is calculated based on the heat absorption capacity per unit mass of the cooling water. This heat balance-based calculation of cooling water volume automatically matches the optimal cooling water volume for bars with different temperatures, masses, and thermophysical parameters, avoiding over- or under-cooling and significantly improving the uniformity of homogeneous cooling and product quality stability.

[0086] In this optional embodiment, the chaotic state characteristics of homogeneous cooling water are extracted based on chaos theory, and multi-step prediction and risk assessment are performed, including:

[0087] A homogeneous cooling water temperature time series is obtained. Based on chaos theory, the maximum Lyapunov exponent, correlation dimension, and fractal dimension of the homogeneous cooling water temperature time series are calculated. The values ​​of the maximum Lyapunov exponent, correlation dimension, and fractal dimension are used to construct a chaotic state feature vector of the homogeneous cooling water. The chaotic state feature vector is input into a multi-step prediction model based on the time series to predict the chaotic state feature vectors of multiple future time steps step by step. In each prediction time step, the risk is determined based on the magnitude of the maximum Lyapunov exponent. If it is greater than the risk threshold, it is determined to be a chaotic risk state; if it is less than or equal to the risk threshold, it is determined to be a stable state.

[0088] In this optional embodiment, obtaining the homogeneous cooling water temperature time series and calculating the maximum Lyapunov exponent, correlation dimension, and fractal dimension of the homogeneous cooling water temperature time series based on chaos theory includes:

[0089] Preprocessing of continuous homogeneous cooling water temperature data sequences at fixed sampling time intervals yields normalized and denoised homogeneous cooling water temperature time sequences. The time delay parameter and embedding dimension are determined, and the homogeneous cooling water temperature time sequence is converted into a target-dimensional phase space reconstruction trajectory matrix using a delay coordinate reconstruction method. The two closest points in the target-dimensional phase space reconstruction trajectory matrix are found, and the spatial distance between these two points over time is calculated, with the natural logarithm of the spatial distance change taken. A linear fit is performed between the natural logarithm of the spatial distance change and time, and the slope of the fitted line is used as the maximum Lyapunov exponent. In the target dimension... Different spatial radius values ​​are selected in the phase space reconstructed trajectory matrix, and for each radius, the proportion of times the distance between pairs of trajectory points is less than the radius is calculated to obtain the correlation integral; the relationship curve between the correlation integral value and the spatial radius is plotted in a double logarithmic coordinate system, and the slope of the relationship curve is used as the correlation dimension; within the selected spatial range of the target dimension phase space reconstructed trajectory matrix, the entire trajectory space is covered with a cubic grid with a specified side length; the number of cubic grids occupied by the trajectory points is calculated; in a double logarithmic coordinate system, the relationship curve between the number of cubic grids and the reciprocal of the grid side length is fitted, and the slope of the fitted line is used as the fractal dimension.

[0090] In this optional embodiment, inputting the chaotic state feature vector into a time series-based multi-step prediction model to progressively predict the chaotic state feature vector at multiple future time steps includes:

[0091] The architecture of the multi-step prediction model is determined, with the input being chaotic state feature vectors and the output being predicted feature vectors for a single time step. A sliding window is used to extract samples of the time series of chaotic state feature vectors during historical operation, yielding input and output window sequences for model training. The multi-step prediction model is trained based on these sequences, and an optimization algorithm is used to train the weights to minimize the error index between the predicted values ​​and the actual chaotic state feature vectors, resulting in a trained multi-step prediction model. The current chaotic state feature vector is input into the trained multi-step prediction model, which outputs chaotic state feature vectors for multiple future time steps.

[0092] In this method, one-dimensional temperature data is embedded into a high-dimensional phase space using a delayed coordinate reconstruction method to recover the system's dynamic trajectory. Subsequently, the maximum Lyapunov exponent (measuring the system's sensitivity to initial conditions), correlation dimension (characterizing the complexity of the system's trajectory), and fractal dimension (characterizing the phase space filling characteristics) are calculated, and these three types of values ​​are combined into a chaotic state feature vector. This feature vector is then input into a multi-step time series prediction model, which can predict the chaotic evolution trend at multiple future time steps in advance, enabling early assessment of chaotic risk based on risk thresholds.

[0093] During the homogeneous cooling process of a certain bar stock, a sequence of cooling water temperature data was collected: [36.8℃, 37.2℃, 38.5℃, 39.1℃, 39.8℃…]. After delayed coordinate reconstruction calculation, the maximum Lyapunov exponent was found to be 0.284, the correlation dimension was 2.07, and the fractal dimension was 2.03, forming a chaotic state feature vector (0.284, 2.07, 2.03). The multi-step prediction model predicted that the maximum Lyapunov exponent would rise to 0.512 and 0.586 in the 4th and 5th steps of the next 5 time steps, respectively, which is higher than the set risk threshold of 0.50. Therefore, it was determined that the stock was about to enter the chaotic risk range. With the water valve opening limit of the homogeneous cooling chamber ≤ +40°, the opening was adjusted from 35° to the upper limit of 40° in advance, and the water distribution was optimized, thereby effectively suppressing temperature fluctuations and maintaining cooling uniformity.

[0094] In this optional embodiment, training the neural network model for the homogeneous cooling water temperature and the adjustment amount of the homogeneous cooling chamber water valve under the constraint of a stable operating range includes:

[0095] Training sample pairs are constructed from the temperature time series corresponding to the steady-state time step and the water valve opening adjustment amount under the corresponding time step conditions. The training sample pairs are used to train the artificial neural network model so that the artificial neural network model can predict the mapping relationship between the homogeneous cooling water temperature and the homogeneous cooling chamber water valve opening adjustment amount under the chaotic state feature input conditions.

[0096] Leveraging the nonlinear mapping capabilities of artificial neural networks, the system collects homogeneous cooling water temperature and corresponding valve opening adjustments within a stable operating range. Training samples are constructed by pairing the temperature value and opening adjustment at each time step, thereby capturing the stable relationship between temperature changes and control variables. Because the temperature and opening data are relatively stable within the stable range, the network can learn more accurate and reliable input-output mappings. In subsequent operations, even with chaotic input conditions, the model can rely on the learned stable mapping to provide valve adjustments matching the target stable state, achieving rapid temperature stabilization.

[0097] In this optional embodiment, the valve opening adjustment amount is output through a neural network model, and the valve opening adjustment amount is optimized and decomposed into independent evolutionary subpopulations of temperature deviation, energy consumption, chaotic stability, and response smoothness using a co-evolutionary mechanism. The globally optimal opening adjustment amount is obtained by dynamically weighting and combining these subpopulations.

[0098] Input the current chaotic state feature vector into the neural network model, and obtain the predicted value of the homogeneous cooling chamber water valve opening adjustment under the current chaotic state conditions through the forward computation of the neural network model; construct the water valve opening adjustment optimization problem, and decompose the water valve opening adjustment optimization problem into a temperature deviation minimization subproblem, an energy consumption minimization subproblem, a chaotic stability improvement subproblem, and a water valve response smoothness improvement subproblem, and create independent evolutionary populations for each subproblem; perform independent evolution for each independent evolutionary population, and obtain the globally optimal opening adjustment amount through dynamic weighting.

[0099] In this optional embodiment, each independent evolutionary population undergoes independent evolution, and the globally optimal opening degree adjustment is obtained through dynamic weighting, including:

[0100] A genetic algorithm optimizer is assigned to each independent evolutionary population to independently evolve and generate locally optimal valve opening adjustment values ​​under the corresponding objective function. These locally optimal opening adjustment values ​​are then dynamically weighted and combined to form the global valve opening adjustment value. Feedback updates are performed under a comprehensive fitness function that evaluates temperature deviation, energy consumption, valve response smoothness, and chaotic state stability, iterating until the global fitness converges. The formula for calculating the global valve opening adjustment value by dynamically weighting and combining the locally optimal opening adjustment values ​​is as follows:

[0101] ;

[0102] In the formula, This is the global water valve opening adjustment amount. The local optimal water valve opening adjustment amount for the temperature deviation target; The local optimal water valve opening adjustment amount for the energy consumption target; The local optimal valve opening adjustment amount is for the target of valve response smoothness. The local optimal valve opening adjustment amount is for the chaotic stability objective; The dynamic weights for the temperature deviation target; The dynamic weights for energy consumption targets; The dynamic weights for the target smoothness of the water valve response; The dynamic weights are the target for chaotic stability.

[0103] A neural network model is used to perform forward computation on the feature vector of the current chaotic state to obtain a preliminary predicted value for the water valve opening adjustment. A co-evolutionary mechanism is introduced to decompose the global optimization problem into four sub-problems: minimizing temperature deviation, minimizing energy consumption, improving chaotic stability, and improving response smoothness. Each sub-problem is assigned an independent population and a genetic algorithm optimizer for independent evolution. The local optima of each subpopulation are dynamically weighted and combined to form the global opening adjustment. The weights change in real time with the system operating state and the importance of each objective, and are iteratively updated under the comprehensive fitness function until convergence. This preserves the independent optimization effect of each objective while achieving a globally optimal control scheme through co-evolution.

[0104] In this optional embodiment, determining the final opening of the homogeneous cooling chamber water valve based on the globally optimal opening adjustment amount corresponding to the current homogeneous cooling water temperature and the initial opening of the homogeneous cooling chamber water valve includes:

[0105] Based on the current homogeneous cooling water temperature, the corresponding opening adjustment amount is output by the neural network model, and the corresponding global optimal opening adjustment amount is obtained by updating using the co-evolution mechanism. The global optimal opening adjustment amount corresponding to the current homogeneous cooling water temperature is added to the initial opening of the homogeneous cooling chamber water valve to obtain the final opening of the homogeneous cooling chamber water valve.

[0106] In this optional embodiment, the temperature of the homogeneous cooling water is obtained using a temperature measuring structure, and the temperature measuring structure includes:

[0107] The cooling water inlet pipe 1 of the cooling chamber is located on the side of the outer wall of the homogeneous cooling chamber. A temperature measuring thermocouple 2 is installed on the cooling water inlet pipe 1. An electronic display screen 4 is installed on the outer wall of the homogeneous cooling chamber. The electronic display screen 4 and the temperature measuring thermocouple 2 are connected by a signal line 3. A steel pipe 5 is sleeved on the outside of the signal line 3.

[0108] The present invention also provides an embodiment of a system for improving the uniformity of homogeneous cooling.

[0109] In this optional embodiment, a system for improving the uniformity of homogeneous cooling includes:

[0110] The initial opening determination module is used to determine the amount of homogeneous cooling water based on the temperature of the bar, and to determine the initial opening of the homogeneous cooling chamber water valve based on the amount of homogeneous cooling water.

[0111] The Chaotic Feature Extraction and Model Training Module is used to extract the chaotic state features of homogeneous cooling water based on chaos theory, and to perform multi-step prediction and risk assessment. Under the constraint of stable operation range, it trains a neural network model of homogeneous cooling water temperature and the opening adjustment of homogeneous cooling chamber water valve.

[0112] The valve opening adjustment quantity co-evolution module is used to output the valve opening adjustment quantity through a neural network model, and to optimize and decompose the valve opening adjustment quantity into independent evolutionary subpopulations of temperature deviation, energy consumption, chaotic stability and response smoothness using a co-evolution mechanism, and to obtain the globally optimal valve opening adjustment quantity through dynamic weighted combination.

[0113] The final opening acquisition module is used to determine the final opening of the homogeneous cooling chamber water valve based on the global optimal opening adjustment amount corresponding to the current homogeneous cooling water temperature and the initial opening of the homogeneous cooling chamber water valve, so as to improve the homogeneous cooling uniformity of the bar.

[0114] This invention adjusts the opening degree of the cooling water valve in the homogeneous cooling chamber, setting different opening degrees according to different water temperatures. See Table 1.

[0115] Table 1 Correspondence between Cooling Water Temperature Valve Opening in Homogeneous Cooling Chamber

[0116]

[0117]

[0118] A thermometer is installed on the water pipe leading into the cooling chamber. The thermometer consists of a thermocouple and a temperature display screen. For example... Figures 2-5 As shown. 1 is the cooling water inlet pipe of the cooling chamber. 2 is the temperature-measuring thermocouple. 3 is the signal line. 4 is the electronic display screen. 5 is the steel pipe. 6 is the outer wall of the homogeneous cooling chamber. The electronic display screen shows the temperature value. A hook bracket 7 is located on the outer wall of the cooling chamber, and a hook 8 is mounted on the electronic display screen 4. The hook 8 is inserted into the hook bracket 7 for fixation. The signal line 4 passes through the steel pipe 5 or a rubber hose, which protects and positions it. The end of the temperature-measuring thermocouple 2 extends to the middle of the inlet pipe to measure the water temperature. Working principle: The temperature-measuring thermocouple 2 is equipped with a protective mechanism, which is connected to the water pipe. A stable connection mechanism is provided between the wire and the adjacent pipe or fixed rod. Figure 5 As shown, Figure 5 The letter A in the diagram indicates welding between two steel pipes.

[0119] Thermometer Temperature Measurement Principle: The thermometer sensor probe is inserted into the cooling chamber's water inlet pipe, with a connecting rod fixedly connected to the top of the probe. A mounting base is installed on the outside of the connecting rod, allowing it to move up and down. The depth of the connecting rod insertion into the water inlet pipe is adjusted until the rod head is positioned at half the depth of the pipe, then it is fixed in place. The thermocouple end is inserted into the top of the lower end of the sensor probe's inner hole. The thermocouple wires extend from the protective mechanism, entering the connecting hose and steel pipe, and connecting to another section of the connecting hose and the electronic display screen 4. The temperature measuring rod of the thermocouple 2 extends to the middle of the water inlet pipe to measure the water temperature. The measurement data is transmitted via a signal line to the electronic display screen 4, which is mounted on the outer wall of the cooling chamber, displaying the measured temperature value.

[0120] This invention employs different water valve openings for different homogeneous cooling water temperatures, including:

[0121] 1. Install a thermocouple 2 on the inlet pipe of the homogeneous cooling chamber, and then suspend an electronic temperature display screen (i.e., electronic display screen 4) on the furnace wall of the cooling chamber. Connect the thermocouple 2 and the electronic display screen 4 with a wire (i.e., signal line 3). 2. Compile a table corresponding to the homogeneous cooling water temperature and the opening degree of the homogeneous cooling valve.

[0122] The homogeneous cooling water temperature was measured, with a maximum temperature of 36-40℃ in summer and a minimum of 10-14℃ in winter. Before the improvement, the water valve opening in the homogeneous cooling chamber was always +40°. Theoretically, to achieve a consistent cooling effect, the cooling water flow rate needs to be controlled according to the bar temperature to achieve closed-loop control. The cooling water flow rate is related to both flow rate and water temperature. Higher water temperatures require a larger flow rate, and lower temperatures require a smaller flow rate. The water flow rate is controlled by the water valve opening. When the maximum water temperature is set at 36-40℃, the water valve opening is +40°. When the water temperature decreases, the water valve opening should decrease accordingly to ensure the same cooling power output.

[0123] Using a neural network algorithm, the valve opening is calculated when the water temperature is below the maximum temperature, with each temperature increment representing a 1°C interval. Theoretically, the valve opening can be reduced by 0.5, 0.2, etc., for each temperature drop in the cooling water. A production process is established for trial operation to test the mechanical properties of the materials and observe their consistency. Based on the actual measured homogeneous cooling water temperature, the corresponding opening is set in the furnace top drain valve setting section of the homogeneous cooling control panel interface (see Table 1). Cooling operation is then initiated. After homogeneous cooling of the cast rod according to the process requirements in Table 2, the material's mechanical properties are tested. Using a co-evolutionary algorithm, the optimization target is determined, and the valve opening is further optimized.

[0124] This invention involves performance testing of 10 batches of Al-Mg-Si-Cu alloy after ingot extrusion following semi-continuous casting (required tensile strength ≥360MPa, yield strength ≥330MPa, elongation ≥4%). Actual extrusion performance data are shown in Table 2.

[0125] Table 2 Actual test performance data of 10 batches of materials

[0126]

[0127] As shown in Table 2, the performance deviation is within 10 MPa and the elongation deviation is within 1%, both meeting customer standard requirements. Table 2 also shows that the homogeneous cooling process satisfies the material consistency requirements.

[0128] Building a neural network model includes: 1. Data preparation: First, collect and prepare data for training and testing. In the water valve opening prediction case, the dataset contains N samples, each with multiple feature values ​​and one target value. 2. Data preprocessing: Data preprocessing includes handling missing values, data standardization, and feature encoding. 3. Model building: Use a deep learning framework (such as TensorFlow or PyTorch) to build the neural network model. The model includes an input layer, hidden layers, and an output layer. The number of neurons and activation functions in the hidden layers can be adjusted according to the specific problem. 4. Model training: Train the model using backpropagation and a gradient-based optimizer. During training, the model learns the function that maps input features to the target value. 5. Model evaluation: Evaluate the model's performance using a test set, typically by calculating the model's loss function and accuracy. Model evaluation helps us understand the function that maps to the target value. 6. Result prediction: The trained model can then predict new data. In the water valve opening case, the model can predict the opening value given the feature values.

[0129] Co-evolution decomposes a complex problem into several subproblems. The solution to each subproblem evolves through individual individuals or populations, and then they are combined to form a global solution. In co-evolution, a genetic algorithm is deployed as the evolutionary engine for each subpopulation, searching for local optima against the objective function. These local solutions are then combined using a dynamic weighting formula to form a global solution, and the fitness is fed back for the next iteration. A genetic algorithm is a method that searches for optimal solutions by simulating the natural evolutionary process. It primarily uses selection, crossover, and mutation operations to gradually approach the optimal solution. The following is a description of a genetic algorithm:

[0130] 1. Population Initialization: First, an initial population is randomly generated, with each individual representing the genetic code of its chromosome. Each individual in the population is essentially a physical entity with a distinctive chromosome. 2. Fitness Assessment: While determining the fitness of an individual, it is also checked whether the fitness meets the optimization criteria. The fitness function measures the degree to which a species adapts to its environment. If the fitness meets the optimization criteria, the best individual and the optimal solution it represents are output. 3. Selection Operation: Individuals are selected based on the principle of survival of the fittest. Individuals with high fitness have a higher probability of being selected, while those with low fitness are eliminated. Common selection methods include roulette wheel selection, random competitive selection, and optimal retention selection. 4. Crossover Operation: Based on known crossover methods and probabilities, chromosome crossover is performed to generate new individuals. Common crossover methods include single-point crossover, two-point crossover, and uniform crossover. 5. Mutation Operation: The offspring chromosomes are mutated by flipping the value of a bit through mutation probability, for example, changing 1 to 0 and 0 to 1. Mutation operations can increase population diversity and prevent getting trapped in local optima. 6. Generate a new generation population: The new generation population generated through crossover and mutation is returned to the fitness evaluation step, and the above process is repeated until the optimal solution is found. The advantage of genetic algorithms is that they can automatically acquire and guide the optimization search space without requiring predetermined rules, adaptively adjusting the search direction. Although genetic algorithms do not guarantee finding the optimal solution to a problem, they can find near-optimal solutions in many cases.

[0131] This invention provides a computer device, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores static and dynamic information data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the above method embodiments.

[0132] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device to which the present invention is applied. A specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0133] In addition, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0134] In addition, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0135] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0136] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for improving the uniformity of homogeneous cooling, characterized in that, include: The amount of homogeneous cooling water is determined based on the temperature of the bar stock, and the initial opening of the water valve in the homogeneous cooling chamber is determined based on the amount of homogeneous cooling water. Based on chaos theory, the chaotic state characteristics of homogeneous cooling water are extracted, and multi-step prediction and risk assessment are performed. Under the constraint of stable operation range, a neural network model of homogeneous cooling water temperature and the opening adjustment of homogeneous cooling chamber water valve is trained. The valve opening adjustment amount is output through a neural network model, and the valve opening adjustment amount is optimized and decomposed into independent evolutionary subpopulations of temperature deviation, energy consumption, chaotic stability and response smoothness using a co-evolutionary mechanism. The global optimal opening adjustment amount is obtained by dynamically weighting the subpopulations. Based on the global optimal opening adjustment amount corresponding to the current homogeneous cooling water temperature and the initial opening of the homogeneous cooling chamber water valve, the final opening of the homogeneous cooling chamber water valve is determined to improve the homogeneous cooling uniformity of the bar. The process of extracting chaotic state characteristics of homogeneous cooling water based on chaos theory and performing multi-step prediction and risk assessment includes: Obtain the time series of homogeneous cooling water temperature. Based on chaos theory, calculate the maximum Lyapunov exponent, correlation dimension, and fractal dimension of the homogeneous cooling water temperature time series. Then, construct the feature vector of the chaotic state of the homogeneous cooling water using the values ​​of the maximum Lyapunov exponent, correlation dimension, and fractal dimension. The chaotic state feature vector is input into a time series-based multi-step prediction model to progressively predict the chaotic state feature vector at multiple future time steps. Risk is determined based on the magnitude of the maximum Lyapunov exponent at each prediction time step. If the exponent is greater than the risk threshold, it is determined to be a chaotic risk state; if it is less than or equal to the risk threshold, it is determined to be a stable state. The process involves outputting the valve opening adjustment amount through a neural network model and using a co-evolutionary mechanism to optimize and decompose the valve opening adjustment amount into independent evolutionary subpopulations of target factors such as temperature deviation, energy consumption, chaotic stability, and response smoothness. The resulting globally optimal opening adjustment amount is obtained through dynamic weighted combination. Input the current chaotic state feature vector into the neural network model, and obtain the predicted value of the homogeneous cooling chamber water valve opening adjustment under the current chaotic state condition through the forward calculation of the neural network model; The optimization problem of water valve opening adjustment is constructed and decomposed into subproblems of minimizing temperature deviation, minimizing energy consumption, improving chaotic stability, and improving water valve response smoothness. Independent evolutionary populations are created for each subproblem. Each independent evolutionary population is allowed to evolve independently, and the globally optimal opening adjustment is obtained by dynamically weighting the population.

2. The method for improving the uniformity of homogeneous cooling according to claim 1, characterized in that, The determination of the homogeneous cooling water volume based on the temperature of the bar includes: Calculate the temperature difference based on the current temperature and target temperature of the bar; calculate the waste heat energy requirement of the bar based on the specific heat capacity, weight of the bar, and temperature difference. Calculate the actual value of heat absorbed by the cooling water based on the waste heat energy demand and heat transfer efficiency of the bar stock. Calculate the volume of homogeneous cooling water based on the actual heat absorbed by the cooling water and the heat absorption capacity per unit mass of the cooling water.

3. The method for improving the uniformity of homogeneous cooling according to claim 1, characterized in that, The neural network model for training the homogeneous cooling water temperature and the opening adjustment of the homogeneous cooling chamber water valve under the constraint of stable operating range includes: Training sample pairs are constructed from the temperature time series corresponding to the steady-state time step and the water valve opening adjustment amount under the corresponding time step conditions. The artificial neural network model is trained using training samples so that it can predict the mapping relationship between the homogeneous cooling water temperature and the adjustment amount of the homogeneous cooling chamber water valve opening under chaotic state characteristic input conditions.

4. The method for improving the uniformity of homogeneous cooling according to claim 1, characterized in that, The acquisition of the homogeneous cooling water temperature time series, based on chaos theory, includes calculating the maximum Lyapunov exponent, correlation dimension, and fractal dimension of the homogeneous cooling water temperature time series, including: Preprocess the continuous homogeneous cooling water temperature data sequence at a fixed sampling time interval to obtain a normalized and denoised homogeneous cooling water temperature time sequence; The time delay parameter and embedding dimension are determined, and the homogeneous cooling water temperature time series is converted into a target dimension phase space reconstructed trajectory matrix by combining the delay coordinate reconstruction method; Find the two closest points in the reconstructed trajectory matrix in the phase space of the target dimension, calculate the spatial distance between the two closest points as they evolve over time, and take the natural logarithm of the change in spatial distance. A linear fit is made between the natural logarithm of the spatial distance change and time, and the slope of the fitted line is used as the maximum Lyapunov exponent. In the trajectory matrix reconstructed in the phase space of the target dimension, different spatial radius values ​​are selected, and for each radius, the proportion of times the distance between pairs of trajectory points is less than the radius is calculated to obtain the correlation integral. Plot the relationship curve between the correlation integral value and the spatial radius in a double logarithmic coordinate system, and use the slope of the relationship curve as the correlation dimension; Within the selected spatial range of the trajectory matrix in the target dimension phase space, cover the entire trajectory space with a cubic grid of a specified side length; calculate the number of cubic grids occupied by the trajectory points; In a double logarithmic coordinate system, the curve relating the number of cubic lattice cells to the reciprocal of the lattice side length is fitted, and the slope of the fitted line is used as the fractal dimension.

5. The method for improving the uniformity of homogeneous cooling according to claim 1, characterized in that, The step of inputting the chaotic state feature vector into a time series-based multi-step prediction model to progressively predict the chaotic state feature vector at multiple future time steps includes: Determine the architecture of the multi-step prediction model, and make the input of the multi-step prediction model a chaotic state feature vector and the output a prediction feature vector for a single time step. By using a sliding window to extract samples of the time series of chaotic state feature vectors during historical operation, the input window sequence and output window sequence used for model training are obtained. A multi-step prediction model is trained based on the input window sequence and the output window sequence, and the weights of the multi-step prediction model are trained using an optimization algorithm to minimize the error index between the predicted value and the feature vector of the real chaotic state, thus obtaining the trained multi-step prediction model. Input the current chaotic state feature vector into the trained multi-step prediction model, and output the chaotic state feature vectors for multiple future time steps.

6. The method for improving the uniformity of homogeneous cooling according to claim 1, characterized in that, The process of independently evolving each independent population and obtaining the globally optimal opening adjustment amount through dynamic weighting includes: A genetic algorithm optimizer is assigned to each independent evolutionary population to independently evolve and generate the local optimal valve opening adjustment amount under the corresponding objective function; The local optimal opening adjustment is combined by dynamic weighting to form the global water valve opening adjustment. Feedback updates are performed under a comprehensive fitness function that evaluates temperature deviation, energy consumption, water valve response smoothness, and chaotic state stability, and the iteration continues until the global fitness converges. The formula for calculating the global valve opening adjustment, which combines the local optimal opening adjustments using dynamic weighting, is as follows: ； In the formula, This is the global water valve opening adjustment amount. The local optimal water valve opening adjustment amount for the temperature deviation target; The local optimal water valve opening adjustment amount for the energy consumption target; The local optimal valve opening adjustment amount is for the target of valve response smoothness. The local optimal valve opening adjustment amount is for the chaotic stability objective; The dynamic weights for the temperature deviation target; The dynamic weights for energy consumption targets; The dynamic weights for the target smoothness of the water valve response; The dynamic weights are the target for chaotic stability.

7. The method for improving the uniformity of homogeneous cooling according to claim 1, characterized in that, The determination of the final opening of the homogeneous cooling chamber water valve based on the globally optimal opening adjustment amount corresponding to the current homogeneous cooling water temperature and the initial opening of the homogeneous cooling chamber water valve includes: Based on the current homogeneous cooling water temperature, the corresponding opening adjustment amount is output through the neural network model, and the corresponding global optimal opening adjustment amount is obtained by updating using the co-evolution mechanism. The final opening of the homogeneous cooling chamber water valve is obtained by adding the global optimal opening adjustment corresponding to the current homogeneous cooling water temperature to the initial opening of the homogeneous cooling chamber water valve.

8. The method for improving the uniformity of homogeneous cooling according to claim 1, characterized in that, The temperature of the homogeneous cooling water is obtained using a temperature measuring structure, which includes: A cooling water inlet pipe for the cooling chamber is located on the side of the outer wall of the homogeneous cooling chamber, and a temperature measuring thermocouple is installed on the cooling water inlet pipe for the cooling chamber. An electronic display screen is installed on the outer wall of the homogeneous cooling chamber. The electronic display screen is connected to the temperature measuring thermocouple via a signal line, and a steel pipe is sleeved around the signal line.

Images