Optimization Method and System for Thermal Equilibrium Distribution of Die Casting Molds Based on Energy Model

Abstract

The present invention provides an optimization method and system for the thermal balance distribution of a die-casting mold based on an energy model, which relates to the technical field of die-casting molds. The method includes collecting thermal distribution data and establishing a mapping relationship of heat energy transfer; extracting spatio-temporal correlation features by using a graph attention network; constructing a dynamic evolution model of thermal distribution; constructing a temperature field predictor based on a deep belief network; training the predictor by using a multi-agent co-evolution algorithm; constructing a temperature field optimization strategy network; generating an optimal temperature field control sequence; establishing a sub-region adaptive control strategy; constructing a temperature field controller based on deep reinforcement learning; and adjusting cooling control parameters by using a dynamic fuzzy immune optimization algorithm to realize intelligent optimized control of the thermal distribution of the die-casting mold.

Description

Technical Field

[0001] The present invention relates to the technical field of die-casting molds, and in particular, to an optimization method and system for the thermal balance distribution of die-casting molds based on an energy model. Background Art

[0002] Die-casting forming is an efficient metal part manufacturing process, and the uniformity and stability of the mold temperature field directly affect the quality of die-cast parts. The heat transfer process of die-casting molds involves complex thermo-mechanical coupling effects, including various heat transfer methods such as heat conduction, convective heat transfer, and radiative heat transfer between different regions of the mold. With the continuous improvement of die-casting process requirements, higher requirements are put forward for the control of the mold temperature field, and the traditional empirical temperature control method is difficult to meet the needs of high-quality die-cast part production.

[0003] In the prior art, the die-casting mold temperature field control technology mainly has the following deficiencies: The traditional thermodynamic model is too simplified, resulting in insufficient modeling accuracy of heat energy transfer; There is a lack of effective means for extracting and characterizing the spatio-temporal evolution characteristics of the temperature field, making it difficult to accurately describe the dynamic interaction relationship of temperature changes in different regions of the mold; It is difficult to handle the coupling relationship between multiple regions and unable to achieve the balanced control of the global temperature field; Using fixed control rules and lacking the ability of adaptive adjustment, it is difficult to cope with process fluctuations and uncertain disturbances in the production process.

[0004] In summary, aiming at the problems existing in the prior art, there is an urgent need for an optimization method for the thermal balance distribution of die-casting molds based on an energy model to achieve accurate modeling of the heat energy transfer process of die-casting molds, provide an effective method for extracting the spatio-temporal characteristics of the temperature field, achieve the collaborative optimization control of the multi-region temperature field, and establish a real-time control strategy with adaptive ability. The present invention can solve the problems in the prior art. Summary of the Invention

[0005] The embodiments of the present invention provide an optimization method and system for the thermal balance distribution of die-casting molds based on an energy model, which can solve the problems in the prior art.

[0006] In the first aspect of the embodiments of the present invention,

[0007] There is provided an optimization method for the thermal balance distribution of die-casting molds based on an energy model, including:

[0008] Collect the thermal distribution data of the die-casting mold, establish the thermal energy transfer mapping relationship of the die-casting mold, convert the thermal energy transfer mapping relationship into a feature vector matrix, input the feature vector matrix into a graph attention network based on unsupervised adversarial learning for feature extraction, and obtain the spatio-temporal correlation features of the die-casting mold; establish a non-linear state equation of the thermal distribution of the die-casting mold based on the spatio-temporal correlation features, and use a hybrid particle swarm quantum genetic algorithm to solve the non-linear state equation to construct a dynamic evolution model of the thermal distribution of the die-casting mold;

[0009] Input the dynamic evolution model of the thermal distribution into a deep belief network based on knowledge distillation to construct a temperature field predictor for the die-casting mold; use a multi-agent cooperative evolution algorithm to train the temperature field predictor to determine the hierarchical optimization function of the temperature field of the die-casting mold; construct a temperature field optimization strategy network for the die-casting mold based on the hierarchical optimization function, and use an adaptive ant colony optimization algorithm based on tabu search to generate an optimal temperature field control sequence for the die-casting mold; establish a sub-region adaptive control strategy for the die-casting mold according to the optimal temperature field control sequence;

[0010] Input the sub-region adaptive control strategy into the dynamic evolution model of the thermal distribution to obtain the temperature field response characteristics of the die-casting mold; construct a temperature field controller based on deep reinforcement learning according to the temperature field response characteristics, and use the temperature field change parameters of the die-casting mold obtained in real time; input the temperature field change parameters into the temperature field predictor for online correction, and use a dynamic fuzzy immune optimization algorithm to adjust the cooling control parameters to realize the intelligent optimization control of the thermal distribution of the die-casting mold.

[0011] In an alternative embodiment,

[0012] Collect the thermal distribution data of the die-casting mold, establish the thermal energy transfer mapping relationship of the die-casting mold, convert the thermal energy transfer mapping relationship into a feature vector matrix, input the feature vector matrix into a graph attention network based on unsupervised adversarial learning for feature extraction, and the spatio-temporal correlation features of the die-casting mold obtained include:

[0013] Construct a temperature sensor grid array on the surface of the die-casting mold. The temperature sensor grid array includes a preset number of temperature acquisition points, records the temperature value, acquisition timestamp, and spatial coordinate information; continuously collect temperature data based on the temperature sensor grid array at a preset sampling frequency, and organize the temperature data into a time series dataset; calculate the temperature gradient vector between adjacent temperature acquisition points in the time series dataset, calculate the heat flux density, and establish a thermal energy transfer mapping relationship according to the temperature gradient vector and the heat flux density;

[0014] Perform Fourier transform on the time series data set to obtain the frequency domain features of the temperature field, combine the frequency domain features of the temperature field with the temperature gradient vector, and construct the first mapping feature vector; calculate the thermal conductivity distribution based on the heat transfer mapping relationship, and combine the thermal conductivity distribution with the first mapping feature vector to construct the second mapping feature vector; arrange the second mapping feature vector according to the spatial distribution of the temperature acquisition points to construct a feature vector matrix;

[0015] Construct a graph structure based on the spatial distribution of the temperature sensor grid array, use the temperature acquisition points as the graph structure nodes, and use the thermal conductivity matrix as the graph structure edge weights; construct an unsupervised adversarial learning network, where the generator of the unsupervised adversarial learning network generates simulated heat transfer data based on the feature vector matrix, and the corresponding discriminator compares and trains the simulated heat transfer data with the time series data set;

[0016] Based on the graph structure, calculate the attention weight coefficient between nodes, and perform weighted fusion of the attention weight coefficient and the heat transfer mapping relationship to obtain the heat conduction weight; reconstruct the feature vector matrix according to the heat conduction weight to obtain the spatio-temporal correlation feature.

[0017] In an alternative embodiment,

[0018] Based on the spatio-temporal correlation feature, establish a non-linear state equation for the thermal distribution of the die-casting mold, and use the hybrid particle swarm quantum genetic algorithm to solve the non-linear state equation. The construction of the dynamic evolution model of the thermal distribution of the die-casting mold includes:

[0019] Based on the spatio-temporal correlation feature, extract and calculate the heat spatial diffusion rate, the heat time cumulative change rate, and the heat transfer coefficient of the mold material, and combine them to construct a state variable vector; generate a control variable vector based on the heating power parameter and the cooling rate parameter of the die-casting mold; use the state variable vector and the control variable vector to establish a non-linear state equation describing the heat transfer law inside the die-casting mold;

[0020] Convert the solution parameters of the non-linear state equation into quantum state encoding, construct an initial particle swarm including quantum position genes and quantum velocity genes, substitute the quantum position genes and quantum velocity genes of the initial particle swarm into the non-linear state equation, calculate the fitness value, and determine the individual optimal solution and the global optimal solution of the initial particle swarm based on the fitness value;

[0021] Based on the individual optimal solution and the global optimal solution, calculate the quantum rotation angle. Based on the quantum rotation angle, perform a quantum rotation gate operation on the quantum position gene to obtain the crossed quantum position gene, and perform a quantum NOT gate operation on the quantum velocity gene to obtain the mutated quantum velocity gene; combine the crossed quantum position gene and the mutated quantum velocity gene to construct an optimized particle swarm;

[0022] Substitute the quantum position gene and the quantum velocity gene of the optimized particle swarm into the non-linear state equation to calculate the new fitness value, update the individual optimal solution and the global optimal solution according to the new fitness value, and repeat the iterative update until the change rate of the new fitness value is less than the preset convergence threshold to obtain the optimal solution;

[0023] Construct the temperature state transition matrix of the die-casting mold under discrete time series according to the optimal solution, establish the heat transfer response function between regions of the die-casting mold based on the temperature state transition matrix, and combine the temperature state transition matrix and the heat transfer response function to construct a dynamic evolution model of the thermal distribution of the die-casting mold.

[0024] In an alternative embodiment,

[0025] Input the dynamic evolution model of the thermal distribution into a deep belief network based on knowledge distillation to construct a die-casting mold temperature field predictor; use a multi-agent cooperative co-evolution algorithm to train the temperature field predictor, and determine that the hierarchical optimization function of the die-casting mold temperature field includes:

[0026] Convert the temperature state transition matrix in the dynamic evolution model of the thermal distribution into a feature map, convert the heat transfer response function in the dynamic evolution model of the thermal distribution into a time series vector, and perform normalization processing on the feature map and the time series vector to obtain the input data of the die-casting mold temperature field predictor;

[0027] Construct a knowledge distillation framework composed of a teacher network and a student network. The teacher network adopts the structure of a deep residual network, and the student network adopts the structure of a three-layer deep belief network. Input the input data into the teacher network to obtain the soft label of the temperature field distribution, use the soft label of the temperature field distribution as the training target of the student network, pre-train each layer of the student network, use the contrastive divergence algorithm to train each layer of the restricted Boltzmann machine, and use the trained student network as the basic network structure of the die-casting mold temperature field predictor to obtain the die-casting mold temperature field predictor;

[0028] Divide the die-casting mold temperature field predictor into multiple sub-predictors. Each sub-predictor is responsible for predicting the temperature of its respective independent corresponding area. An information transmission channel is established between the sub-predictors through an attention mechanism; use a multi-agent cooperative evolution algorithm to train the sub-predictors, construct a fitness function including local prediction error and prediction consistency, evaluate the prediction performance of the sub-predictors based on the fitness function, use the prediction performance as an evolution index, and optimize the network parameters of each sub-predictor based on the evolution index to obtain an optimized temperature field prediction performance index; construct a hierarchical optimization function for the die-casting mold temperature field according to the temperature field prediction performance index.

[0029] In an alternative embodiment,

[0030] Construct a temperature field optimization strategy network for the die-casting mold based on the hierarchical optimization function. The optimal temperature field control sequence of the die-casting mold generated by using an adaptive ant colony optimization algorithm based on tabu search includes:

[0031] Construct a temperature field optimization strategy network for the die-casting mold based on the hierarchical optimization function. Divide the temperature field optimization strategy network into a temperature field uniformity layer, a cooling efficiency layer, and an energy consumption control layer, and determine the evaluation index function;

[0032] Initialize the positions of the ant colony. Encode the temperature field control parameters of the die-casting mold as ant position information, determine the temperature field control sequence, input the temperature field control sequence corresponding to each ant position into the temperature field optimization strategy network, calculate the path evaluation value through the evaluation index function, and update the pheromone concentration on the path according to the path evaluation value, where the pheromone increment is positively correlated with the path evaluation value;

[0033] Construct a dynamic tabu list to record the searched temperature field control sequences. Calculate the similarity between the newly generated temperature field control sequence and the temperature field control sequences in the tabu list, and add the temperature field control sequences with a similarity higher than the preset threshold to the dynamic tabu list;

[0034] Calculate the state transition probability based on the pheromone concentration on the path and the dynamic tabu list to guide the search direction of the ant colony;

[0035] Near the current position of each ant, construct a variable neighborhood search structure. The variable neighborhood search structure includes a single-parameter adjustment space and a multi-parameter linkage adjustment space, and independently adjust and combine the parameters in the temperature field control sequence respectively; input the candidate solutions corresponding to the temperature field control sequences generated in the variable neighborhood search structure into the temperature field optimization strategy network, calculate the path evaluation values of each candidate solution, and select the temperature field control sequence with the highest path evaluation value to update the ant position;

[0036] Repeat the iteration until the preset maximum number of iterations is reached to obtain the optimal temperature field control sequence.

[0037] In an alternative embodiment,

[0038] Input the sub-region adaptive control strategy into the thermal distribution dynamic evolution model to obtain the temperature field response characteristics of the die-casting mold; according to the temperature field response characteristics, construct a temperature field controller based on deep reinforcement learning, and adopt real-time acquisition of the temperature field change parameters of the die-casting mold; input the temperature field change parameters into the temperature field predictor for online correction, and adopt a dynamic fuzzy immune optimization algorithm to adjust the cooling control parameters to realize the intelligent optimization control of the thermal distribution of the die-casting mold, including:

[0039] Input the sub-region adaptive control strategy into the thermal distribution dynamic evolution model, establish the temperature field state equation of the die-casting mold based on the law of conservation of energy, use the Kalman filter algorithm to perform state estimation on the temperature field state equation, and iteratively update through the state estimation error covariance matrix and the Kalman gain matrix to obtain the temperature field response characteristics of the die-casting mold;

[0040] Construct a temperature field controller based on deep reinforcement learning according to the temperature field response characteristics. The temperature field controller adopts a dual network architecture, including a policy network for generating control actions and a value network for evaluating state values; the state spaces of the policy network and the value network include the current temperature distribution, the temperature change rate, and the control input history, and the action space includes the control parameter adjustment amounts of each region; use the Advantage Actor-Critic algorithm to train the policy network and the value network, construct an immediate reward function based on the temperature uniformity value and the control stability value, and obtain the temperature field change parameters of the die-casting mold through the optimization of the policy loss function and the value loss function;

[0041] Input the temperature field change parameters into the temperature field predictor for online correction, construct a dynamic fuzzy rule base according to the corrected prediction results, and design fuzzy control rules based on the temperature deviation and the temperature change rate; use the dynamic fuzzy immune optimization algorithm to generate a cooling parameter adjustment strategy set, construct the adjustment strategy set into an antibody group, calculate the antibody affinity based on the error of temperature control, and optimize the antibody group through cloning and mutation operations to finally obtain the optimal cooling control parameters and intelligently adjust the thermal distribution of the die-casting mold.

[0042] In an alternative embodiment,

[0043] Use the dynamic fuzzy immune optimization algorithm to generate a cooling parameter adjustment strategy set, construct the adjustment strategy set into an antibody group, calculate the antibody affinity based on the error of temperature control, and optimize the antibody group through cloning and mutation operations to finally obtain the optimal cooling control parameters and intelligently adjust the thermal distribution of the die-casting mold, including:

[0044] Obtain the cooling control parameters of the die-casting mold, encode the cooling control parameters to generate an initial antibody population, and each antibody individual contains a parameter vector corresponding to the cooling control parameters;

[0045] Collect the real-time temperature data of each measuring point of the die-casting mold, calculate the temperature deviation and the rate of change of temperature deviation according to the real-time temperature data, use the temperature deviation and the rate of change of temperature deviation as input variables, establish a dynamic fuzzy rule mapping relationship, and construct a fuzzy rule base for temperature field control;

[0046] Based on the sum of squares of the deviation between the real-time temperature data and the target temperature, construct a temperature control error index, convert the temperature control error index into an antibody affinity value through an exponential function, the exponential function includes an adjustment coefficient for controlling the selection pressure, and according to the antibody affinity value, sort the antibody individuals in the antibody population from high to low; according to the preset decreasing rule of the clone number, assign the clone number to each antibody individual in turn to generate a cloned antibody population, perform Gaussian mutation operation on the cloned antibody population, and the mutation step size of the Gaussian mutation is inversely proportional to the antibody affinity value to obtain a mutated antibody population;

[0047] Calculate the affinity values of each antibody individual in the mutated antibody population, determine candidate mutated antibody individuals from the mutated antibody population and candidate antibody individuals from the initial antibody population according to the preset affinity selection threshold, merge the candidate mutated antibody individuals and the candidate antibody individuals to generate a new generation of antibody population, and at the same time, introduce randomly generated new antibody individuals into the new generation of antibody population according to the preset number;

[0048] Repeat the iteration to generate a new generation of antibody population until the preset number of iterations is reached, select the antibody individual corresponding to the maximum affinity value, determine the optimal antibody individual, decode the optimal antibody individual to obtain the optimal cooling control parameters, and adjust the thermal distribution of the die-casting mold.

[0049] In the second aspect of the embodiments of the present invention,

[0050] Provide an optimization system for the thermal balance distribution of a die-casting mold based on an energy model, including:

[0051] The first unit is used to collect the thermal distribution data of the die-casting mold, establish a thermal energy transfer mapping relationship of the die-casting mold, convert the thermal energy transfer mapping relationship into a feature vector matrix, input the feature vector matrix into a graph attention network based on unsupervised adversarial learning for feature extraction to obtain the spatio-temporal correlation features of the die-casting mold; establish a non-linear state equation for the thermal distribution of the die-casting mold based on the spatio-temporal correlation features, and use a hybrid particle swarm quantum genetic algorithm to solve the non-linear state equation to construct a dynamic evolution model of the thermal distribution of the die-casting mold;

[0052] A second unit, configured to input the thermal distribution dynamic evolution model into a deep belief network based on knowledge distillation to construct a temperature field predictor for a die-casting mold; train the temperature field predictor by using a multi-agent co-evolution algorithm to determine a hierarchical optimization function for the temperature field of the die-casting mold; construct a temperature field optimization strategy network for the die-casting mold based on the hierarchical optimization function, and generate an optimal temperature field control sequence for the die-casting mold by using an adaptive ant colony optimization algorithm based on tabu search; establish a sub-region adaptive control strategy for the die-casting mold according to the optimal temperature field control sequence.

[0053] A third unit, configured to input the sub-region adaptive control strategy into the thermal distribution dynamic evolution model to obtain the temperature field response characteristics of the die-casting mold; construct a temperature field controller based on deep reinforcement learning according to the temperature field response characteristics, and obtain the temperature field change parameters of the die-casting mold in real time; input the temperature field change parameters into the temperature field predictor for online correction, and adjust the cooling control parameters by using a dynamic fuzzy immune optimization algorithm to realize the intelligent optimization control of the thermal distribution of the die-casting mold.

[0054] In a third aspect of the embodiments of the present invention,

[0055] there is provided an electronic device, including:

[0056] a processor;

[0057] a memory for storing instructions executable by the processor;

[0058] wherein the processor is configured to call the instructions stored in the memory to execute the method described above.

[0059] In a fourth aspect of the embodiments of the present invention,

[0060] there is provided a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0061] In the embodiments of the present invention, through spatio-temporal correlation feature extraction and non-linear modeling, the thermal distribution can be predicted more accurately; the dynamic evolution model describes the spatio-temporal changes of the thermal distribution, making the system more adaptable when adjusting control strategies; the model supports precise control of the temperature field, reduces energy consumption, extends the die life, and improves production efficiency; by constructing a temperature field predictor based on a deep belief network with knowledge distillation, the dynamic changes of the die temperature field can be predicted more precisely; through a multi-agent co-evolution algorithm and a hierarchical optimization function, the temperature field optimization strategy network is trained to improve the accuracy and adaptability of temperature control; an adaptive ant colony tabu search algorithm is used to generate an optimal temperature control sequence, and based on this, a sub-region adaptive control strategy is formulated to achieve precise thermal management of the die, further improving the die-casting quality; real-time monitoring and feedback adjustment significantly improve the control accuracy of the die temperature field, ensure uniform temperature distribution in each region, and effectively avoid die loss and product defects caused by thermal non-uniformity; the temperature field response characteristics and the correction function of real-time data input to the predictor enable the control system to adapt to temperature changes during the die-casting process, improving the system's response speed and robustness to fluctuations; the dynamic fuzzy immune optimization algorithm automatically adjusts the cooling control parameters according to real-time data, realizing the intelligence and high efficiency of the cooling process, and reducing the cooling time and energy consumption. Description of the Drawings

[0062] Figure 1 It is a schematic flow chart of an optimization method for the thermal balance distribution of a die-casting die based on an energy model according to an embodiment of the present invention;

[0063] Figure 2 It is a schematic structural diagram of an optimization system for the thermal balance distribution of a die-casting die based on an energy model according to an embodiment of the present invention. Detailed Embodiments

[0064] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0065] The technical solutions of the present invention will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0066] Figure 1 It is a schematic flow chart of an optimization method for the thermal balance distribution of a die-casting die based on an energy model according to an embodiment of the present invention, as Figure 1 shown, and the method includes:

[0067] S101. Collect the thermal distribution data of the die-casting mold, establish the mapping relationship of heat energy transfer of the die-casting mold, convert the heat energy transfer mapping relationship into a feature vector matrix, input the feature vector matrix into the graph attention network based on unsupervised adversarial learning for feature extraction, and obtain the spatio-temporal correlation features of the die-casting mold; establish a non-linear state equation of the thermal distribution of the die-casting mold based on the spatio-temporal correlation features, and use the hybrid particle swarm quantum genetic algorithm to solve the non-linear state equation to construct a dynamic evolution model of the thermal distribution of the die-casting mold;

[0068] Collect the thermal distribution data on the surface and inside of the die-casting mold, capture the temperature change information of the mold under different working conditions, and ensure the integrity and accuracy of the data; based on the collected thermal distribution data, establish the mapping relationship of heat energy transfer inside and outside the mold, and describe the transfer laws of heat flow and temperature in different regions; perform mathematical modeling on the heat energy transfer mapping relationship and convert it into a feature vector matrix so that it can be effectively input into the subsequent feature extraction network.

[0069] Input the feature vector matrix into the graph attention network based on unsupervised adversarial learning to extract spatio-temporal correlation features, capture the dynamic changes and regional correlations of the mold's thermal distribution; based on the extracted spatio-temporal correlation features, establish a non-linear state equation to describe the thermal distribution state of the mold at different times and regions for subsequent solution and prediction; use the hybrid particle swarm quantum genetic algorithm to solve the non-linear state equation to construct a dynamic evolution model to accurately describe the change of the mold's thermal distribution over time; construct a dynamic evolution model of the thermal distribution: use the solution results to construct a dynamic evolution model of the thermal distribution, comprehensively describe the spatio-temporal change law of the internal thermal distribution of the mold, and provide a basis for the subsequent temperature control strategy.

[0070] In this embodiment, through spatio-temporal correlation feature extraction and non-linear modeling, the thermal distribution can be predicted more accurately; the dynamic evolution model describes the spatio-temporal changes of the thermal distribution, making the system more adaptable when adjusting the control strategy; the model supports precise control of the temperature field, reduces energy consumption, extends the mold life, and improves production efficiency.

[0071] In an alternative embodiment, collecting the thermal distribution data of the die-casting mold, establishing the mapping relationship of heat energy transfer of the die-casting mold, converting the heat energy transfer mapping relationship into a feature vector matrix, and inputting the feature vector matrix into the graph attention network based on unsupervised adversarial learning for feature extraction to obtain the spatio-temporal correlation features of the die-casting mold includes:

[0072] Construct a temperature sensor grid array on the surface of the die-casting mold. The temperature sensor grid array includes a preset number of temperature acquisition points, and records temperature values, acquisition timestamps, and spatial coordinate information. Continuously acquire temperature data based on the temperature sensor grid array at a preset sampling frequency, and organize the temperature data into a time-series data set. Calculate the temperature gradient vectors between adjacent temperature acquisition points in the time-series data set, calculate the heat flux density, and establish a heat energy transfer mapping relationship based on the temperature gradient vectors and the heat flux density.

[0073] Perform Fourier transform on the time-series data set to obtain the frequency-domain characteristics of the temperature field. Combine the frequency-domain characteristics of the temperature field with the temperature gradient vectors to construct a first mapping feature vector. Calculate the distribution of the heat conduction coefficient based on the heat energy transfer mapping relationship, and combine the distribution of the heat conduction coefficient with the first mapping feature vector to construct a second mapping feature vector. Arrange the second mapping feature vectors according to the spatial distribution of the temperature acquisition points to construct a feature vector matrix.

[0074] Construct a graph structure based on the spatial distribution of the temperature sensor grid array. Use the temperature acquisition points as the graph structure nodes and the heat conduction coefficient matrix as the graph structure edge weights. Construct an unsupervised adversarial learning network. The generator of the unsupervised adversarial learning network generates simulated heat transfer data based on the feature vector matrix, and the corresponding discriminator compares and trains the simulated heat transfer data with the time-series data set.

[0075] Calculate the attention weight coefficients between nodes based on the graph structure, and perform weighted fusion of the attention weight coefficients and the heat energy transfer mapping relationship to obtain the heat conduction weight. Reconstruct the feature vector matrix according to the heat conduction weight to obtain the spatio-temporal correlation features.

[0076] Specifically, construct a temperature sensor grid array on the surface of the die-casting mold. The temperature sensor grid array adopts a 12x12 matrix layout, with an adjacent sensor spacing of 50 millimeters. Each sensor node includes a temperature acquisition unit, a time synchronization unit, and a three-dimensional coordinate positioning unit. The measurement range of the temperature acquisition unit is 0 - 800 degrees Celsius, and the measurement accuracy is ±0.1 degrees Celsius. The time synchronization unit records millisecond-level timestamps. The three-dimensional coordinate positioning unit records the spatial position coordinates of the sensor on the mold surface, with a coordinate accuracy of 0.1 millimeter.

[0077] Continuously collect temperature data at a sampling frequency of 100 Hz. Each sensor node collects 1000 temperature data points within a die-casting cycle, forming a multi-dimensional time-series data set containing timestamps, spatial coordinates, and temperature values. Calculate the temperature differences between adjacent temperature acquisition points to obtain the temperature gradient components in the x and y directions, and combine the sensor spacing to obtain the temperature gradient vector. Calculate the heat flux density based on the temperature gradient vector, and establish a thermal energy transfer mapping function through the ratio relationship between the temperature gradient vector and the heat flux density.

[0078] Perform fast Fourier transform processing on the time-series data set to extract the frequency-domain features of the temperature field, including the main frequency component, harmonic components, and phase information. Correspond the frequency-domain features with the temperature gradient vectors one by one according to the sensor node positions to construct a 144-dimensional first mapping feature vector. Calculate a 144x144 thermal conductivity distribution matrix based on the thermal energy transfer mapping function, and combine the thermal conductivity distribution matrix with the first mapping feature vector to construct a second mapping feature vector containing thermal conduction characteristics. Rearrange the second mapping feature vector according to the 12x12 spatial distribution of the temperature sensors to form a feature vector matrix.

[0079] Construct a graph structure based on the temperature sensor grid array. 144 sensor nodes serve as the vertices of the graph, and connection edges are established between adjacent nodes. The weight of the edge is set to the corresponding thermal conductivity. Construct an adversarial learning network architecture containing a generator and a discriminator. The generator inputs the feature vector matrix and generates a simulated heat transfer data sequence through a multi-layer convolutional neural network. The discriminator compares the generated simulated data with the actually collected time-series data set, calculates the discriminant loss, and backpropagates to optimize the network parameters.

[0080] In the graph structure, calculate the attention scores between each pair of connected nodes to obtain a 144x144 attention weight coefficient matrix. Perform weighted combination of the attention weight coefficient and the thermal energy transfer mapping function with weight coefficients of 0.6 and 0.4 to obtain a comprehensive thermal conduction weight. Use the thermal conduction weight to perform weighted reconstruction on each element in the feature vector matrix to generate a spatio-temporal feature representation that fuses time evolution features and spatial correlation features.

[0081] In practical applications, for a typical die-casting mold, a single die-casting cycle is 30 seconds, and a 144x1000 temperature time-series data matrix is collected. After Fourier transform, the main frequency component is 0.033 Hz, and it contains 3 - 5 significant harmonic components. The amplitude range of the temperature gradient vector is 0.5 - 2.0 degrees Celsius per millimeter, and the numerical range of the heat flux density is 50 - 200 kilowatts per square meter. The finally reconstructed spatio-temporal features can accurately characterize the dynamic evolution law of the temperature field during die-casting, providing a basis for precise control of the mold temperature field.

[0082] In this embodiment, the sensor grid array and high-precision data acquisition ensure the detailed monitoring of the temperature field, capturing the subtle changes in the surface temperature of the die-casting mold; through the calculation of the temperature gradient and heat flux density, the heat transfer mapping is accurately established, improving the accuracy of the temperature field model; the combination of frequency-domain analysis and adversarial learning network with graph attention mechanism generates spatio-temporal feature representations, which helps to better understand the dynamic changes of the temperature field in time and space; the generated high-dimensional feature vector matrix provides comprehensive spatio-temporal features, providing an important reference basis for the precise control and regulation of the temperature field, improving the mold quality and extending the mold life; through fast Fourier transform and eigenvector rearrangement, the system can efficiently process a large amount of temperature data, extract the main frequency components and harmonic components, and effectively analyze the dynamic changes of the temperature field.

[0083] In an alternative embodiment, a nonlinear state equation for the thermal distribution of the die-casting mold is established based on the spatio-temporal correlation features, and a hybrid particle swarm quantum genetic algorithm is used to solve the nonlinear state equation. The construction of the dynamic evolution model of the thermal distribution of the die-casting mold includes:

[0084] Based on the spatio-temporal correlation features, the heat space diffusion rate, the heat time cumulative change rate, and the heat transfer coefficient of the mold material are extracted and calculated, and a state variable vector is constructed by combination; based on the heating power parameter and the cooling rate parameter of the die-casting mold, a control variable vector is generated; a nonlinear state equation describing the internal heat transfer law of the die-casting mold is established using the state variable vector and the control variable vector;

[0085] The solution parameters of the nonlinear state equation are converted into quantum state encoding, an initial particle swarm containing quantum position genes and quantum velocity genes is constructed, the quantum position genes and quantum velocity genes of the initial particle swarm are substituted into the nonlinear state equation, the fitness value is calculated, and the individual optimal solution and the global optimal solution of the initial particle swarm are determined based on the fitness value;

[0086] Based on the individual optimal solution and the global optimal solution, the quantum rotation angle is calculated. Based on the quantum rotation angle, a quantum rotation gate operation is performed on the quantum position genes to obtain the crossed quantum position genes, and a quantum NOT gate operation is performed on the quantum velocity genes to obtain the mutated quantum velocity genes; the crossed quantum position genes and the mutated quantum velocity genes are combined to construct an optimized particle swarm;

[0087] The quantum position genes and quantum velocity genes of the optimized particle swarm are substituted into the nonlinear state equation to calculate the new fitness value, the individual optimal solution and the global optimal solution are updated according to the new fitness value, and the iteration update is repeated until the change rate of the new fitness value is less than the preset convergence threshold to obtain the optimal solution;

[0088] Construct the temperature state transition matrix of the die-casting mold under discrete time series according to the optimal solution, establish the heat transfer response function between regions of the die-casting mold based on the temperature state transition matrix, and combine the temperature state transition matrix with the heat transfer response function to construct a dynamic evolution model of the thermal distribution of the die-casting mold.

[0089] Extract heat transfer characteristic parameters from spatio-temporal correlation features, including heat spatial diffusion rate, heat time cumulative change rate, and mold material heat transfer coefficient. The heat spatial diffusion rate reflects the speed of heat propagation between regions of the mold, and its value range is 0.5 - 2.0 m / s; the heat time cumulative change rate characterizes the degree of heat accumulation per unit time, and its value range is 20 - 100 kJ / s; the mold material heat transfer coefficient describes the thermal conductivity of the material, and the typical value is 30 - 50 W / (m·K). Combine these three parameters to form a nine-dimensional state variable vector, which is used to characterize the thermodynamic state of the mold.

[0090] Obtain the process control parameters of the die-casting mold, including heating power parameters and cooling rate parameters. The heating power parameters include heater power, heating time, and heating area distribution, the power range is 10 - 50 kW, and the heating time is 5 - 15 s; the cooling rate parameters include cooling medium flow rate, cooling channel size, and cooling time, the flow rate range is 10 - 30 L / min, and the cooling time is 10 - 30 s. Integrate these parameters into a six-dimensional control variable vector. Based on the state variable vector and the control variable vector, establish a nonlinear state equation that describes the internal heat transfer law of the die-casting mold.

[0091] Use quantum coding to represent the solution parameters of the nonlinear state equation, and each parameter is encoded with 8-bit quantum bits. Construct an initial particle swarm containing 100 particles, and each particle contains a quantum position gene and a quantum velocity gene. The quantum position gene represents the parameter value, and the quantum velocity gene represents the parameter change direction and step size. Substitute the gene information of the initial particle swarm into the nonlinear state equation, calculate the fitness value of each particle, and the fitness value is the reciprocal of the temperature field prediction error. Determine the individual historical optimal position of each particle and the global optimal position of the entire population based on the fitness value.

[0092] Calculate the quantum rotation angle according to the difference between the individual optimal position and the global optimal position, and the angle range is 0.01π to 0.05π. Perform a quantum rotation gate operation on the quantum position gene to achieve gene crossover by adjusting the phase of the quantum bit; perform a quantum NOT gate operation on the quantum velocity gene to achieve gene mutation by flipping the quantum bit. Recombine the crossed quantum position gene and the mutated quantum velocity gene to construct an optimized particle swarm.

[0093] Substitute the gene information of the optimized particle swarm into the non - linear state equation to calculate the new fitness value, and update the individual optimal position and the global optimal position. When the relative change rate of the fitness value is less than 0.1% in 50 consecutive generations of iteration, it is considered that the optimization process converges and the optimal solution is obtained. The typical optimization process requires 200 - 300 generations of iteration to converge.

[0094] Based on the optimal solution, construct a 100x100 temperature state transition matrix, where the matrix elements represent the temperature transfer relationship between different regions of the die at adjacent times. The eigenvalue distribution range of the temperature state transition matrix is 0.85 - 1.15, which reflects the stability of the system. Use the temperature state transition matrix to establish the heat transfer response function between regions. The rise time of the response function is 1 - 3 seconds, and the adjustment time is 5 - 8 seconds. Combine the temperature state transition matrix and the heat transfer response function to obtain a mathematical model that can accurately predict the dynamic evolution of the temperature field of the die - casting die.

[0095] In the actual application verification, the prediction error of the model for the die temperature field is less than 3%, which can effectively capture the dynamic characteristics of heat transfer and provide a reliable theoretical basis for realizing precise temperature field control. The predicted temperature field distribution by the model is in good agreement with the actual measurement results, verifying the effectiveness and practicality of the modeling method.

[0096] In this embodiment, through the application of quantum coding optimization and the temperature state transition matrix, the prediction error of the model for the temperature field is less than 3%, ensuring the accuracy of the prediction results; by extracting key parameters such as the heat spatial diffusion rate, the time - cumulative change rate, and the material heat transfer coefficient, the model can effectively characterize the spatio - temporal heat transfer law in the die; using quantum rotation gates and quantum NOT gates to optimize the particle swarm, the solution efficiency is improved, enabling the optimization to converge within 200 - 300 generations; the temperature state transition matrix and the heat transfer response function reflect the system stability and provide a stable response for heat transfer between regions, which helps to maintain the uniformity of the temperature distribution during the die - casting process; the model can predict the dynamic evolution process of the temperature field, providing a reliable basis for the precise control of the temperature field of the die - casting die and effectively improving the processing quality and service life of the die.

[0097] S102. Input the thermal distribution dynamic evolution model into a deep belief network based on knowledge distillation to construct a die - casting die temperature field predictor; use a multi - agent collaborative evolution algorithm to train the temperature field predictor to determine the hierarchical optimization function of the die - casting die temperature field; based on the hierarchical optimization function, construct a temperature field optimization strategy network for the die - casting die, and use an adaptive ant colony optimization algorithm based on tabu search to generate the optimal temperature field control sequence for the die - casting die; establish a sub - region adaptive control strategy for the die - casting die according to the optimal temperature field control sequence;

[0098] In this embodiment, a temperature field predictor is constructed through a deep belief network based on knowledge distillation, which can more accurately predict the dynamic changes of the die temperature field; through a multi-agent co-evolution algorithm and a hierarchical optimization function, the temperature field optimization strategy network is trained to improve the accuracy and adaptability of temperature control; an adaptive ant colony tabu search algorithm is used to generate an optimal temperature control sequence, and a sub-region adaptive control strategy is formulated accordingly to achieve precise thermal management of the die, further improving the die casting quality; sub-region control reduces the temperature adjustment response time, helps to quickly adapt to temperature changes during the die casting process, and improves production efficiency and die life.

[0099] In an alternative embodiment, the thermal distribution dynamic evolution model is input into a deep belief network based on knowledge distillation to construct a die casting die temperature field predictor; a multi-agent co-evolution algorithm is used to train the temperature field predictor, and the hierarchical optimization function of the die casting die temperature field is determined to include:

[0100] The temperature state transition matrix in the thermal distribution dynamic evolution model is converted into a feature map, and the heat transfer response function in the thermal distribution dynamic evolution model is converted into a time series vector. The feature map and the time series vector are normalized to obtain the input data of the die casting die temperature field predictor.

[0101] A knowledge distillation framework composed of a teacher network and a student network is constructed. The teacher network adopts the structure of a deep residual network, and the student network adopts the structure of a three-layer deep belief network. The input data is input into the teacher network to obtain the soft label of the temperature field distribution. Taking the soft label of the temperature field distribution as the training target of the student network, each layer of the student network is pre-trained, and the contrastive divergence algorithm is used to train each layer of the restricted Boltzmann machine. The trained student network is used as the basic network structure of the die casting die temperature field predictor to obtain the die casting die temperature field predictor.

[0102] The die casting die temperature field predictor is divided into multiple sub-predictors, and each sub-predictor is responsible for predicting the temperature of its respective independent corresponding region. An information transmission channel is established between the sub-predictors through an attention mechanism; a multi-agent co-evolution algorithm is used to train the sub-predictors, a fitness function including local prediction error and prediction consistency is constructed, the prediction performance of the sub-predictors is evaluated based on the fitness function, and the prediction performance is used as an evolution index. Based on the evolution index, the network parameters of each sub-predictor are optimized to obtain an optimized temperature field prediction performance index; a hierarchical optimization function of the die casting die temperature field is constructed according to the temperature field prediction performance index.

[0103] Specifically, the 100x100 temperature state transition matrix in the dynamic evolution model of heat distribution is processed and converted into a feature map with 64 channels through a convolution operation, and the size of each channel is 32x32. The heat transfer response function is sampled at 1000 points in the time dimension to form a time series vector. The feature map and the time series vector are processed by min-max normalization to uniformly scale the numerical range to between 0 and 1, which is used as the standardized input data for the temperature field predictor.

[0104] Construct a temperature field prediction network based on the knowledge distillation framework. The teacher network adopts a 50-layer deep residual network structure, including 5 residual blocks, and each residual block contains 10 convolutional layers with a kernel size of 3x3, and the number of channels increases from 64 to 512. The student network adopts a three-layer deep belief network structure, with the first layer containing 1024 neurons, the second layer containing 512 neurons, and the third layer containing 256 neurons. The normalized input data is fed into the teacher network to obtain the soft label prediction result of the temperature field distribution.

[0105] The student network adopts a layer-by-layer pre-training strategy, and a restricted Boltzmann machine structure is constructed for each layer. The number of visible layer units of the first-layer restricted Boltzmann machine is the same as the input data dimension, and the number of hidden layer units is 1024; the number of visible layer units of the second layer is 1024, and the number of hidden layer units is 512; the number of visible layer units of the third layer is 512, and the number of hidden layer units is 256. The contrastive divergence algorithm is used to train each layer of the restricted Boltzmann machine, with the learning rate set to 0.01, the momentum factor to 0.9, and the number of training rounds to 100 rounds. The three trained networks are stacked and combined to form the basic network structure of the die-casting mold temperature field predictor.

[0106] The surface of the die-casting mold is divided into 16 prediction regions, and each region corresponds to a sub-predictor. The sub-predictor inherits the basic network structure, and the input is the local features of the corresponding region. An information transfer channel with a multi-head attention mechanism is constructed between adjacent sub-predictors, with 8 attention heads, and the dimension of each attention head is 32, which is used to capture the temperature field correlation between regions.

[0107] The multi-agent co-evolution algorithm is used to train 16 sub-predictors, with the population size set to 50 and the number of evolutionary generations to 200 generations. A fitness function is constructed to evaluate the performance of the sub-predictors. The fitness function consists of two components: the proportion of local prediction error is 60%, and the proportion of prediction consistency is 40%. The local prediction error is calculated using the mean absolute error, and the prediction consistency is calculated through the correlation coefficient of the predicted values in adjacent regions.

[0108] During the evolution process, a crossover operation and a mutation operation are performed on the network parameters of the sub-predictor every 10 generations. The uniform crossover method is adopted for the crossover operation, and the crossover probability is 0.8; the Gaussian mutation method is adopted for the mutation operation, the mutation probability is 0.1, and the mutation intensity is 0.05. Based on the fitness function, the prediction performance after crossover and mutation is evaluated, and individuals with excellent performance are retained to enter the next generation of evolution.

[0109] After 200 generations of evolutionary optimization, the average prediction error of the sub-predictor is reduced to less than 1.5%, and the correlation coefficient of the predicted values in adjacent regions is increased to more than 0.95. The optimized prediction performance indicators are integrated into a hierarchical optimization function to guide the precise control of the temperature field of the die-casting mold. The hierarchical optimization function reflects the overall temperature field distribution characteristics at the macroscopic level and the local temperature gradient change law at the microscopic level.

[0110] In this embodiment, through the combination of the knowledge distillation framework, the sub-predictors for different regions, and the multi-agent co-evolution algorithm, the prediction error is reduced to less than 1.5%, significantly improving the accuracy of temperature field prediction; the multi-head attention mechanism is used to capture the temperature correlation in adjacent prediction regions, making the prediction consistency correlation coefficient in adjacent regions reach more than 0.95, which helps to achieve the coordinated control of the overall temperature field; the optimization of the number of evolution generations and the crossover and mutation operations make the training process more efficient, ensuring rapid convergence to the optimal solution and improving the adaptability and stability of the sub-predictor in a dynamic environment; the hierarchical optimization function combines the macroscopic temperature field distribution and the local temperature gradient details, providing multi-level guidance for the precise temperature field control of the die-casting mold and improving the quality and stability of the die-casting process.

[0111] In an alternative embodiment, a temperature field optimization strategy network for the die-casting mold is constructed based on the hierarchical optimization function, and an adaptive ant colony optimization algorithm based on tabu search is used to generate the optimal temperature field control sequence for the die-casting mold, including:

[0112] A temperature field optimization strategy network for the die-casting mold is constructed based on the hierarchical optimization function, the temperature field optimization strategy network is divided into a temperature field uniformity layer, a cooling efficiency layer, and an energy consumption control layer, and an evaluation index function is determined;

[0113] The positions of the ant colony are initialized, the temperature field control parameters of the die-casting mold are encoded as ant position information, the temperature field control sequence is determined, the temperature field control sequence corresponding to each ant position is input into the temperature field optimization strategy network, and through the evaluation index function, the path evaluation value is calculated, and the pheromone concentration on the path is updated according to the path evaluation value, where the pheromone increment is positively correlated with the path evaluation value;

[0114] Construct a dynamic tabu list to record the searched temperature field control sequences, calculate the similarity between the newly generated temperature field control sequence and the temperature field control sequences in the tabu list, and add the temperature field control sequences with similarity higher than the preset threshold to the dynamic tabu list;

[0115] Based on the pheromone concentration on the path and the dynamic tabu list, calculate the state transition probability to guide the search direction of the ant colony;

[0116] Near the current position of each ant, construct a variable neighborhood search structure. The variable neighborhood search structure includes a single-parameter adjustment space and a multi-parameter linkage adjustment space, and independently adjust and combinatorially adjust the parameters in the temperature field control sequence respectively; input the candidate solutions corresponding to the temperature field control sequences generated in the variable neighborhood search structure into the temperature field optimization strategy network, calculate the path evaluation values of each candidate solution, select the temperature field control sequence with the highest path evaluation value, and update the ant position;

[0117] Repeat the iteration until the preset maximum number of iterations is reached to obtain the optimal temperature field control sequence.

[0118] Specifically, construct a temperature field optimization strategy network with a three-layer structure. The temperature field uniformity layer focuses on the consistency of the temperature distribution on the mold surface, and uses the standard deviation of the temperatures at 16 temperature measurement points as the evaluation index, with the standard deviation threshold set at 3 degrees Celsius; the cooling efficiency layer focuses on the cooling rate and cooling time, requiring the cooling rate not to exceed 50 degrees Celsius per second, and the cooling time to be controlled within the range of 15 - 30 seconds; the energy consumption control layer focuses on the heating power and the cooling water flow rate, with the heating power limited within 40 kilowatts and the cooling water flow rate limited within 25 liters per minute. Combine the three-layer evaluation indexes according to the weight ratio of 4:3:3 to construct a comprehensive evaluation index function.

[0119] Create an ant colony containing 100 ants. The position information of each ant includes 10 temperature field control parameters: heater power, heating time, heating area distribution coefficient, cooling water flow rate, cooling channel opening degree, cooling time, temperature set value, control period, proportional coefficient, integral time. Encode these parameters into a 200-bit binary string to form a temperature field control sequence. At initialization, randomly generate the ant positions within the parameter feasible region, and the feasible region boundary is determined according to the process requirements.

[0120] Input the temperature field control sequence corresponding to each ant into the optimization strategy network to calculate the path evaluation value. The higher the path evaluation value, the better the control effect. Update the pheromone concentration on the path according to the path evaluation value. The pheromone evaporation coefficient is set at 0.1, and the pheromone increment is proportional to the path evaluation value, with the proportional coefficient being 0.2. The update period of the pheromone concentration is 10 iterations.

[0121] The dynamic tabu list adopts a circular queue structure with a maximum capacity of 1000 control sequences. For the newly generated control sequence, calculate the Hamming distance from the existing sequences in the tabu list. If the minimum Hamming distance is less than 5% of the control sequence length, add this sequence to the tabu list. When the tabu list reaches its maximum capacity, delete the earliest added sequence. Update the tabu list every 50 iterations.

[0122] Based on the path pheromone concentration and the tabu list information, calculate the state transition probability. The influence weight of the pheromone concentration on the transition probability is 0.7, and the influence weight of the tabu list information is 0.3. The greater the transition probability of a direction, the greater the possibility that the ant chooses this direction for search. In each iteration, 20% of the ants perform random search to increase the exploration ability of the algorithm.

[0123] Construct a variable neighborhood search structure around the current position of the ant. For the single-parameter adjustment space, for each control parameter, it fluctuates up and down around the current value with a step size of 5%; for the multi-parameter coupled adjustment space, select strongly correlated parameter combinations, such as heating power and heating time, cooling water flow rate and cooling time, etc., and adjust these parameters simultaneously. Randomly generate 50 candidate solutions from the variable neighborhood structure, calculate their path evaluation values, and select the candidate solution with the highest evaluation value to update the ant's position.

[0124] The algorithm terminates after 300 iterations to obtain the optimal temperature field control sequence.

[0125] In this embodiment, by setting the temperature standard deviation threshold and temperature control parameters, the consistency of the die surface temperature distribution is ensured, and the quality problems caused by temperature differences are reduced; the rate and time limit of the cooling efficiency layer ensure a stable and fast cooling process, meet the process requirements and avoid material stress problems caused by excessive cooling; the energy consumption control layer effectively limits the heating and cooling power, realizes energy-saving operation, and at the same time ensures the temperature control accuracy, optimizing the resource use; the ant colony algorithm combines tabu search and dynamic pheromone concentration update, prevents local optimum, ensures that the temperature control parameters gradually tend to the optimal solution after multiple iterations, and improves the stability of the optimization result; through the random search and variable neighborhood search strategies, both the global exploration ability of the algorithm is increased and the local adjustment is refined, which helps to obtain a more comprehensive and optimized temperature field control sequence.

[0126] S103. Input the sub-region adaptive control strategy into the thermal distribution dynamic evolution model to obtain the temperature field response characteristics of the die-casting die; according to the temperature field response characteristics, construct a temperature field controller based on deep reinforcement learning, and use it to obtain the temperature field change parameters of the die-casting die in real time; input the temperature field change parameters into the temperature field predictor for online correction, and use the dynamic fuzzy immune optimization algorithm to adjust the cooling control parameters to realize the intelligent optimization control of the thermal distribution of the die-casting die.

[0127] In this embodiment, real-time monitoring and feedback adjustment significantly improve the control accuracy of the die temperature field, ensure uniform temperature distribution in each area, and effectively avoid die loss and product defects caused by thermal non-uniformity; the temperature field response characteristics and the correction function of the real-time data input predictor enable the control system to adapt to temperature changes during die-casting, improving the system's response speed and robustness to fluctuations; the dynamic fuzzy immune optimization algorithm automatically adjusts the cooling control parameters according to real-time data, realizing the intelligentization and high efficiency of the cooling process, reducing the cooling time and energy consumption; the optimized temperature field distribution control reduces the thermal fatigue of the die, reduces the thermal stress, and improves the die service life and production stability.

[0128] In an alternative embodiment, input the sub-region adaptive control strategy into the thermal distribution dynamic evolution model to obtain the temperature field response characteristics of the die-casting die; according to the temperature field response characteristics, construct a temperature field controller based on deep reinforcement learning, and use real-time acquisition of the temperature field change parameters of the die-casting die; input the temperature field change parameters into the temperature field predictor for online correction, and use the dynamic fuzzy immune optimization algorithm to adjust the cooling control parameters to realize the intelligent optimization control of the thermal distribution of the die-casting die, including:

[0129] Input the sub-region adaptive control strategy into the thermal distribution dynamic evolution model, establish the temperature field state equation of the die-casting die based on the law of conservation of energy, use the Kalman filter algorithm to perform state estimation on the temperature field state equation, and iteratively update through the state estimation error covariance matrix and the Kalman gain matrix to obtain the temperature field response characteristics of the die-casting die;

[0130] Construct a temperature field controller based on deep reinforcement learning according to the temperature field response characteristics. The temperature field controller adopts a dual-network architecture, including a policy network for generating control actions and a value network for evaluating state values; the state space of the policy network and the value network includes the current temperature distribution, the temperature change rate, and the control input history, and the action space includes the control parameter adjustment amounts for each area; use the advantage actor-critic algorithm to train the policy network and the value network, construct an immediate reward function based on the temperature uniformity value and the control stability value, and obtain the temperature field change parameters of the die-casting die through the optimization of the policy loss function and the value loss function;

[0131] Input the temperature field change parameters into a temperature field predictor for online correction, construct a dynamic fuzzy rule base according to the corrected prediction results, and design fuzzy control rules based on temperature deviation and temperature change rate; adopt a dynamic fuzzy immune optimization algorithm to generate a set of cooling parameter adjustment strategies, construct the adjustment strategy set into an antibody population, calculate the antibody affinity based on the error of temperature control, and optimize the antibody population through cloning and mutation operations to finally obtain the optimal cooling control parameters and intelligently adjust the thermal distribution of the die-casting mold.

[0132] Divide the surface of the die-casting mold into 16 control regions, and configure independent heating and cooling channels for each region to achieve regional temperature control. Establish a temperature field state equation based on the law of conservation of energy, and the state variables include the temperature values of each region, heat flux density, and boundary heat exchange coefficient. Use the Kalman filter algorithm for state estimation, and the observed data comes from the real-time acquisition values of temperature sensors buried on the mold surface, with a sampling period of 100 milliseconds.

[0133] Obtain the temperature field response characteristics by iteratively calculating the state estimation error covariance matrix and the Kalman gain matrix. The initial value of the error covariance matrix is set to the identity matrix, and the system noise covariance matrix and the measurement noise covariance matrix are set to 0.01 and 0.05 times the identity matrix respectively. After 1000 iterations, the root mean square error of the state estimation is reduced to within 2 degrees Celsius.

[0134] Construct a deep reinforcement learning controller with a dual network architecture, including a policy network and a value network. The policy network adopts a four-layer fully connected neural network structure: the input layer has 1024 neurons, the two hidden layers have 512 and 256 neurons respectively, and the output layer has 48 neurons corresponding to the heating power, cooling water flow rate, and channel opening adjustment amounts of 16 regions. The value network adopts a three-layer structure: the input layer has 1024 neurons, the hidden layer has 512 neurons, and the output layer has 1 neuron for state value evaluation.

[0135] The state space dimension is 1024, including the current temperature values of 16 regions, the temperature change rates of the previous three moments, and the historical control input of the last 10 times. The action space dimension is 48, corresponding to the adjustment amounts of three control parameters for each region. The heating power adjustment range is plus or minus 5 kW, the cooling water flow rate adjustment range is plus or minus 2 L / min, and the channel opening adjustment range is plus or minus 10%.

[0136] Train the dual network using the Advantage Actor-Critic algorithm, and the immediate reward function is constructed based on temperature uniformity and control stability. Temperature uniformity is characterized by the standard deviation of the temperature values of 16 regions, and the smaller the standard deviation, the greater the reward; control stability is characterized by the sum of the squares of the control parameter adjustment amounts, and the smaller the adjustment amount, the greater the reward. The reward weight ratio is 7:3.

[0137] During the training process of the Advantage Actor-Critic algorithm, it includes two core components: the policy network and the value network. The policy network is responsible for generating control actions, and the value network is responsible for evaluating the state value. The two networks interact and cooperate to complete the training process.

[0138] The construction of the immediate reward function considers two key metrics: the temperature uniformity value and the control stability value. The temperature uniformity value is characterized by calculating the temperature deviation of 16 control regions. The specific method is to sum the absolute values of the differences between the actual temperature and the target temperature in each region. The smaller this value, the more uniform the temperature field. The control stability value is characterized by calculating the change amplitude of three consecutive control actions. The specific method is to sum the absolute values of the differences between the adjustment amounts of the control parameters before and after. The smaller this value, the more stable the control.

[0139] In practical applications, the target threshold for the temperature uniformity value is set to 5 degrees Celsius, and the target threshold for the control stability value is set to 10% of the adjustment amount. When the temperature uniformity value is less than 5 degrees Celsius and the control stability value is less than 10%, a positive reward value of 1.0 is given; when one of them exceeds the threshold, a negative reward value of -0.5 is given; when both exceed the threshold, a negative reward value of -1.0 is given.

[0140] The policy loss function is used to optimize the policy network. Its design idea is to make the control actions generated by the network maximize the future cumulative reward. Specifically, when implementing, first calculate the advantage value obtained by executing a certain action in the current state. The advantage value is equal to the actually obtained cumulative reward minus the state value predicted by the value network. If the advantage value is positive, it indicates that this action is better than the average level, and the output probability of this action is increased during training; if the advantage value is negative, the output probability of this action is decreased.

[0141] The value loss function is used to optimize the value network. Its design idea is to make the network accurately predict the value of each state. Specifically, when implementing, calculate the deviation between the predicted value output by the value network and the actually obtained cumulative reward, and optimize the network parameters by reducing this deviation.

[0142] During the training process, an alternating optimization method is adopted: first fix the value network and optimize the policy network for several steps; then fix the policy network and optimize the value network for several steps. The optimization step size is dynamically adjusted according to the training progress. A larger step size such as 0.001 is adopted at the initial stage of training, and it is gradually reduced to 0.0001 at the later stage of training to ensure convergence stability.

[0143] Specifically, during each round of training, 1000 state-action-reward samples are collected and randomly divided into 10 batches. For each batch, first calculate the state value using the value network, then calculate the advantage value in combination with the actually obtained reward, and finally update the parameters of the policy network and the value network respectively. Repeat this process until convergence, which generally requires 5000 - 10000 rounds of training.

[0144] To avoid local optima, an exploration mechanism is introduced during the training process: a random perturbation is added to the control actions generated by the policy network, and the perturbation intensity gradually decreases during training, from an initial 20% to a final 5%. At the same time, an experience replay mechanism is adopted to store historical training samples, and historical samples are randomly selected and mixed with new samples for training each time to improve the sample utilization efficiency and training stability.

[0145] After training and optimization, the policy network can output appropriate control actions according to the current temperature field state to achieve precise regulation of the temperature field. Under typical working conditions, the accuracy rate of the control actions can reach over 85%, and the temperature field adjustment time is shortened to within 15 seconds, meeting the actual production requirements.

[0146] The experience replay mechanism is adopted for training. The capacity of the experience pool is 10,000, and the size of each sampling batch is 128. The learning rates of the policy network and the value network are set to 0.0001 and 0.0002 respectively, and the discount factor is 0.99. The training lasts for 10,000 rounds, and the training stops when the average reward value change is less than 1% for 100 consecutive rounds.

[0147] The temperature field change parameters obtained from training are input into the temperature field predictor for online correction. A dynamic rule base containing 49 fuzzy rules is constructed, and control rules are designed based on the temperature deviation and the temperature change rate. The temperature deviation is divided into seven levels, the temperature change rate is divided into seven levels, and the control output is divided into seven levels.

[0148] The dynamic fuzzy immune algorithm is used to optimize the cooling control parameters. 100 antibodies are initialized, and each antibody contains 48 gene positions corresponding to the cooling control parameters of 16 regions. The antibody affinity is calculated by the reciprocal of the temperature control error, and the smaller the error, the higher the affinity. The 20 antibodies with the highest affinity are selected for cloning and amplification, and the number of clones is proportional to the affinity.

[0149] Gaussian mutation operation is performed on the cloned antibodies, with a mutation probability of 0.1, and the mutation intensity is inversely proportional to the affinity. After the mutation, the antibody population undergoes suppression calculation to retain the antibodies with high affinity and low similarity. The cloning, mutation, and suppression operations are repeated for 50 generations to obtain the optimal combination of cooling control parameters.

[0150] In this embodiment, based on deep reinforcement learning and regional control, the temperature of each region can be automatically adjusted under a complex temperature field, and finally the temperature uniformity target can be achieved, with the temperature difference controlled within 5 degrees Celsius. The Kalman filter estimates the temperature state in real time, improving the response speed of the temperature field control, and the adjustment time is shortened to within 15 seconds, meeting the requirements of the production beat. The dynamic fuzzy immune algorithm flexibly adjusts the control parameters in the optimization of the cooling parameters, ensuring the stability of the cooling process, reducing energy consumption, and shortening the cooling time and energy consumption. This control strategy can adapt to the temperature field fluctuations, correct the predictor in the online state, and maintain high control accuracy and strong robustness under complex working conditions.

[0151] In an alternative embodiment, a dynamic fuzzy immune optimization algorithm is used to generate a set of cooling parameter adjustment strategies, the set of adjustment strategies is constructed as an antibody population, and the antibody affinity is calculated based on the error of temperature control. By means of cloning and mutation operations, the antibody population is optimized, and finally the optimal cooling control parameters are obtained. The intelligent adjustment of the thermal distribution of the die-casting mold includes:

[0152] Obtain the cooling control parameters of the die-casting mold, encode the cooling control parameters to generate an initial antibody population, and each antibody individual contains a parameter vector corresponding to the cooling control parameters;

[0153] Collect the real-time temperature data of each measuring point of the die-casting mold, calculate the temperature deviation and the rate of change of the temperature deviation according to the real-time temperature data, use the temperature deviation and the rate of change of the temperature deviation as input variables, establish a dynamic fuzzy rule mapping relationship, and construct a fuzzy rule base for temperature field control;

[0154] Based on the sum of squares of the deviation between the real-time temperature data and the target temperature, construct a temperature control error index, convert the temperature control error index into an antibody affinity value through an exponential function, the exponential function includes an adjustment coefficient for controlling the selection pressure, and according to the antibody affinity value, sort the antibody individuals in the antibody population from high to low; according to a preset decreasing rule of the cloning number, allocate the cloning number to each antibody individual in turn to generate a cloned antibody population, perform Gaussian mutation operation on the cloned antibody population, and the mutation step size of the Gaussian mutation is inversely proportional to the antibody affinity value to obtain a mutated antibody population;

[0155] Calculate the affinity values of each antibody individual in the mutated antibody population, determine candidate mutated antibody individuals from the mutated antibody population and candidate antibody individuals from the initial antibody population according to a preset affinity selection threshold, merge the candidate mutated antibody individuals and the candidate antibody individuals to generate a new generation of antibody population, and at the same time, introduce randomly generated new antibody individuals into the new generation of antibody population according to a preset number;

[0156] Iteratively generate a new generation of antibody populations until the preset number of iterations is reached. Select the antibody individual corresponding to the maximum affinity value to determine the optimal antibody individual. Decode the optimal antibody individual to obtain the optimal cooling control parameters and adjust the thermal distribution of the die-casting mold.

[0157] First, obtain the initial cooling control parameters of the die-casting mold, including three types of parameters: the cooling water flow rate, the opening degree of the cooling channels, and the cooling time in 16 control regions. Encode these parameters into a 48-dimensional vector, with each parameter using 8-bit binary encoding. Initialize an antibody population consisting of 100 antibody individuals, with each antibody corresponding to a complete configuration scheme of cooling control parameters.

[0158] Pre-bury 32 temperature sensors on the mold surface, with a sampling frequency of 10 Hz. The temperature data collected in real-time is used to calculate the temperature deviation and the rate of change of temperature deviation. The temperature deviation is defined as the difference between the measured temperature and the target temperature, and the rate of change of temperature deviation is the ratio of the difference between the temperature deviations of two adjacent samplings to the sampling time interval.

[0159] According to actual production experience, divide the temperature deviation into seven fuzzy subsets: negative large, negative medium, negative small, zero, positive small, positive medium, positive large, with the corresponding specific intervals being: less than -10 degrees Celsius, -10 to -5 degrees Celsius, -5 to -2 degrees Celsius, -2 to 2 degrees Celsius, 2 to 5 degrees Celsius, 5 to 10 degrees Celsius, greater than 10 degrees Celsius. The rate of change of temperature deviation is also divided into seven fuzzy subsets, with the specific intervals being: less than -2 degrees Celsius per second, -2 to -1 degrees Celsius per second, -1 to -0.3 degrees Celsius per second, -0.3 to 0.3 degrees Celsius per second, 0.3 to 1 degrees Celsius per second, 1 to 2 degrees Celsius per second, greater than 2 degrees Celsius per second.

[0160] Based on the above fuzzy subsets, construct a rule base containing 49 fuzzy rules. For example: when the temperature deviation is positive large and the rate of change of temperature deviation is positive large, the cooling intensity is adjusted to the maximum enhancement; when the temperature deviation is negative large and the rate of change of temperature deviation is negative large, the cooling intensity is adjusted to the maximum weakening. In this way, the dynamic fuzzy control of the temperature field is achieved.

[0161] Construct an antibody affinity calculation criterion using an exponential function. The smaller the temperature control error, the larger the corresponding affinity value. Set the adjustment coefficient of the control selection pressure to 0.5 to adjust the uniformity of the fitness distribution. Calculate the affinity values of 100 antibody individuals and sort them. The antibody with the highest affinity obtains the largest number of clones, specifically using a linear decreasing method: the first place clones 20, the second place clones 19, and so on.

[0162] Perform Gaussian mutation operation on the antibody population generated by cloning. The mutation step size is inversely proportional to the antibody affinity value. For antibodies with an affinity value above 0.9, the mutation step size is set to 0.01; for antibodies with an affinity value between 0.7 and 0.9, the mutation step size is set to 0.05; for antibodies with an affinity value between 0.5 and 0.7, the mutation step size is set to 0.1; for antibodies with an affinity value below 0.5, the mutation step size is set to 0.2. This adaptive mutation mechanism can increase the diversity of the search space while maintaining high-quality solutions.

[0163] Recalculate the affinity values of each antibody after mutation, and set the affinity selection threshold to 0.6. Select antibodies with an affinity greater than 0.6 from the mutated antibody population as candidate mutated antibodies, and select antibodies with an affinity greater than 0.8 from the initial antibody population as candidate antibodies. Combine these two parts of antibodies to form a new generation population, and randomly generate 10 new antibodies and inject them into the population to keep the population size at 100.

[0164] Repeat the above evolutionary process, and set the number of iterations to 50 generations. In practical applications, when the change range of the optimal affinity value for 10 consecutive generations is less than 0.1%, the iteration can be terminated in advance. Finally, select the antibody individual with the highest affinity value, and decode it to obtain the optimal cooling control parameters.

[0165] Verification under typical working conditions shows that the optimal control parameters obtained by this method are: cooling water flow rate of 18 - 25 liters per minute, cooling channel opening of 60% - 85%, and cooling time of 8 - 12 seconds. Precise control of the die temperature field is achieved, with a temperature control accuracy better than ±1.5 degrees Celsius, and the temperature field uniformity controlled within 3 degrees Celsius, significantly improving the quality stability of die-cast parts.

[0166] In this embodiment, the optimized adjustment of the cooling control parameters enables the control accuracy of the die temperature field to reach ±1.5 degrees Celsius, meeting the requirements of high-precision temperature regulation; through the dynamic adjustment of the cooling water flow rate, channel opening, and cooling time in multiple regions, the temperature field uniformity is controlled within 3 degrees Celsius, which helps to improve the quality stability of die-cast parts; based on the fuzzy rule base, intelligent response to real-time temperature deviation and change rate is achieved, and the cooling intensity can be adaptively adjusted to ensure stable control effects under different production conditions; the antibody affinity adaptive mutation mechanism increases the diversity of the search space while maintaining high-affinity solutions, accelerating the convergence of the optimal control parameters and significantly improving the parameter search efficiency; through the continuous optimization and evolution of the cooling control parameters, the method has good adaptability, is applicable to various die-casting working conditions, and ultimately improves the quality consistency and production efficiency of die-cast parts.

[0167] Figure 2 For the structural schematic diagram of the die-casting mold thermal balance distribution optimization system based on the energy model in the embodiment of the present invention, as Figure 2As shown, the system includes:

[0168] A first unit, configured to collect heat distribution data of a die-casting mold, establish a heat transfer mapping relationship of the die-casting mold, convert the heat transfer mapping relationship into a feature vector matrix, input the feature vector matrix into a graph attention network based on unsupervised adversarial learning for feature extraction, and obtain spatio-temporal correlation features of the die-casting mold; establish a non-linear state equation of the heat distribution of the die-casting mold based on the spatio-temporal correlation features, solve the non-linear state equation using a hybrid particle swarm quantum genetic algorithm, and construct a dynamic evolution model of the heat distribution of the die-casting mold;

[0169] A second unit, configured to input the dynamic evolution model of the heat distribution into a deep belief network based on knowledge distillation to construct a temperature field predictor for the die-casting mold; train the temperature field predictor using a multi-agent cooperative co-evolution algorithm to determine a hierarchical optimization function of the temperature field of the die-casting mold; construct a temperature field optimization strategy network for the die-casting mold based on the hierarchical optimization function, generate an optimal temperature field control sequence for the die-casting mold using an adaptive ant colony optimization algorithm based on tabu search; establish a sub-region adaptive control strategy for the die-casting mold according to the optimal temperature field control sequence;

[0170] A third unit, configured to input the sub-region adaptive control strategy into the dynamic evolution model of the heat distribution to obtain the temperature field response characteristics of the die-casting mold; construct a temperature field controller based on deep reinforcement learning according to the temperature field response characteristics, and acquire the temperature field change parameters of the die-casting mold in real time; input the temperature field change parameters into the temperature field predictor for online correction, and adjust the cooling control parameters using a dynamic fuzzy immune optimization algorithm to realize intelligent optimization control of the heat distribution of the die-casting mold. In the third aspect of the embodiments of the present invention,

[0171] There is provided an electronic device, including:

[0172] A processor;

[0173] A memory for storing instructions executable by the processor;

[0174] Wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0175] In the fourth aspect of the embodiments of the present invention,

[0176] There is provided a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0177] The present invention may be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having thereon computer-readable program instructions for performing various aspects of the present invention.

[0178] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An optimization method for the thermal equilibrium distribution of die-casting molds based on an energy model, characterized in that, Including: Collecting the thermal distribution data of the die-casting mold, establishing the thermal energy transfer mapping relationship of the die-casting mold, converting the thermal energy transfer mapping relationship into a feature vector matrix, inputting the feature vector matrix into a graph attention network based on unsupervised adversarial learning for feature extraction, and obtaining the spatio-temporal correlation features of the die-casting mold; Based on the spatio-temporal correlation features, establishing a non-linear state equation for the thermal distribution of the die-casting mold, solving the non-linear state equation by using a hybrid particle swarm quantum genetic algorithm, and constructing a dynamic evolution model for the thermal distribution of the die-casting mold; Inputting the dynamic evolution model of the thermal distribution into a deep belief network based on knowledge distillation to construct a temperature field predictor for the die-casting mold; Training the temperature field predictor by using a multi-agent cooperative co-evolution algorithm to determine the hierarchical optimization function of the temperature field of the die-casting mold; Based on the hierarchical optimization function, constructing a temperature field optimization strategy network for the die-casting mold, generating an optimal temperature field control sequence for the die-casting mold by using an adaptive ant colony optimization algorithm based on tabu search; establishing a sub-region adaptive control strategy for the die-casting mold according to the optimal temperature field control sequence; Inputting the sub-region adaptive control strategy into the dynamic evolution model of the thermal distribution to obtain the temperature field response characteristics of the die-casting mold; According to the temperature field response characteristics, constructing a temperature field controller based on deep reinforcement learning, and acquiring the temperature field change parameters of the die-casting mold in real time; inputting the temperature field change parameters into the temperature field predictor for online correction, and adjusting the cooling control parameters by using a dynamic fuzzy immune optimization algorithm to realize the intelligent optimization control of the thermal distribution of the die-casting mold; Constructing a temperature field optimization strategy network for the die-casting mold based on the hierarchical optimization function, and generating an optimal temperature field control sequence for the die-casting mold by using an adaptive ant colony optimization algorithm based on tabu search includes: Constructing a temperature field optimization strategy network for the die-casting mold based on the hierarchical optimization function, dividing the temperature field optimization strategy network into a temperature field uniformity layer, a cooling efficiency layer and an energy consumption control layer, and determining the evaluation index function; Initializing the positions of the ant colony, encoding the temperature field control parameters of the die-casting mold as ant position information, determining the temperature field control sequence, inputting the temperature field control sequence corresponding to each ant position into the temperature field optimization strategy network, calculating the path evaluation value through the evaluation index function, and updating the pheromone concentration on the path according to the path evaluation value, where the pheromone increment is positively correlated with the path evaluation value; Constructing a dynamic tabu list to record the searched temperature field control sequences, calculating the similarity between the newly generated temperature field control sequence and the temperature field control sequences in the tabu list, and adding the temperature field control sequences with similarity higher than the preset threshold to the dynamic tabu list; Calculating the state transition probability based on the pheromone concentration on the path and the dynamic tabu list to guide the search direction of the ant colony; Construct a variable neighborhood search structure near the current position of each ant. The variable neighborhood search structure includes a single-parameter adjustment space and a multi-parameter linkage adjustment space, and independently adjusts and combines the parameters in the temperature field control sequence respectively; input the candidate solutions corresponding to the temperature field control sequences generated in the variable neighborhood search structure into the temperature field optimization strategy network, calculate the path evaluation values of each candidate solution, select the temperature field control sequence with the highest path evaluation value, and update the ant positions; Repeat the iteration until the preset maximum number of iterations is reached to obtain the optimal temperature field control sequence; Input the sub-region adaptive control strategy into the thermal distribution dynamic evolution model to obtain the temperature field response characteristics of the die-casting mold; construct a temperature field controller based on deep reinforcement learning according to the temperature field response characteristics, and use it to obtain the temperature field change parameters of the die-casting mold in real time; input the temperature field change parameters into the temperature field predictor for online correction, and use the dynamic fuzzy immune optimization algorithm to adjust the cooling control parameters to achieve intelligent optimization control of the thermal distribution of the die-casting mold, including: Input the sub-region adaptive control strategy into the thermal distribution dynamic evolution model, establish the temperature field state equation of the die-casting mold based on the law of conservation of energy, use the Kalman filter algorithm to perform state estimation on the temperature field state equation, and iteratively update through the state estimation error covariance matrix and the Kalman gain matrix to obtain the temperature field response characteristics of the die-casting mold; Construct a temperature field controller based on deep reinforcement learning according to the temperature field response characteristics. The temperature field controller adopts a dual-network architecture, including a policy network for generating control actions and a value network for evaluating state values; the state spaces of the policy network and the value network include the current temperature distribution, temperature change rate, and control input history, and the action space includes the control parameter adjustment amounts of each region; use the Advantage Actor-Critic algorithm to train the policy network and the value network, construct an immediate reward function based on the temperature uniformity value and the control stability value, and obtain the temperature field change parameters of the die-casting mold through the optimization of the policy loss function and the value loss function; Input the temperature field change parameters into the temperature field predictor for online correction, construct a dynamic fuzzy rule base according to the corrected prediction results, and design fuzzy control rules based on the temperature deviation and the temperature change rate; use the dynamic fuzzy immune optimization algorithm to generate a set of cooling parameter adjustment strategies, construct the adjustment strategy set into an antibody population, calculate the antibody affinity based on the error of temperature control, and optimize the antibody population through cloning and mutation operations to finally obtain the optimal cooling control parameters and intelligently adjust the thermal distribution of the die-casting mold; Use the dynamic fuzzy immune optimization algorithm to generate a set of cooling parameter adjustment strategies, construct the adjustment strategy set into an antibody population, calculate the antibody affinity based on the error of temperature control, and optimize the antibody population through cloning and mutation operations to finally obtain the optimal cooling control parameters and intelligently adjust the thermal distribution of the die-casting mold, including: Obtain the cooling control parameters of the die-casting mold, encode the cooling control parameters to generate an initial antibody population, and each antibody individual contains a parameter vector corresponding to the cooling control parameters; Collect the real-time temperature data of each measuring point of the die-casting mold, calculate the temperature deviation and the rate of change of the temperature deviation according to the real-time temperature data, use the temperature deviation and the rate of change of the temperature deviation as input variables, establish a dynamic fuzzy rule mapping relationship, and construct a fuzzy rule base for temperature field control; Based on the sum of squares of the deviations between the real-time temperature data and the target temperature, construct a temperature control error index, convert the temperature control error index into an antibody affinity value through an exponential function, the exponential function includes an adjustment coefficient for controlling the selection pressure, and according to the antibody affinity value, sort the antibody individuals in the antibody population from high to low; According to the preset decreasing rule of the clone number, assign the clone number to each antibody individual in turn to generate a cloned antibody population, perform Gaussian mutation operation on the cloned antibody population, and the mutation step size of the Gaussian mutation is inversely proportional to the antibody affinity value to obtain a mutated antibody population; Calculate the affinity values of each antibody individual in the mutated antibody population, determine candidate mutated antibody individuals from the mutated antibody population and candidate antibody individuals from the initial antibody population according to the preset affinity selection threshold, merge the candidate mutated antibody individuals and the candidate antibody individuals to generate a new generation of antibody population, and at the same time, introduce randomly generated new antibody individuals into the new generation of antibody population according to the preset number; Repeat the iteration to generate a new generation of antibody population until the preset number of iterations is reached, select the antibody individual corresponding to the maximum affinity value, determine the optimal antibody individual, decode the optimal antibody individual to obtain the optimal cooling control parameters, and adjust the thermal distribution of the die-casting mold.

2. The method according to claim 1, wherein Collect the thermal distribution data of the die-casting mold, establish a thermal energy transfer mapping relationship of the die-casting mold, convert the thermal energy transfer mapping relationship into a feature vector matrix, and input the feature vector matrix into a graph attention network based on unsupervised adversarial learning for feature extraction to obtain the spatio-temporal correlation features of the die-casting mold, including: Construct a temperature sensor grid array on the surface of the die-casting mold, the temperature sensor grid array includes a preset number of temperature acquisition points, record the temperature value, acquisition timestamp, and spatial coordinate information; Continuously collect temperature data based on the temperature sensor grid array at a preset sampling frequency, and organize the temperature data into a time series dataset; Calculate the temperature gradient vector between adjacent temperature acquisition points in the time series dataset, calculate the heat flux density, and establish a thermal energy transfer mapping relationship according to the temperature gradient vector and the heat flux density; Perform Fourier transform on the time series dataset to obtain the frequency domain characteristics of the temperature field, combine the frequency domain characteristics of the temperature field with the temperature gradient vector to construct a first mapping feature vector; Calculate the thermal conductivity distribution based on the thermal energy transfer mapping relationship, and combine the thermal conductivity distribution with the first mapping feature vector to construct a second mapping feature vector; Arrange the second mapping feature vectors according to the spatial distribution of the temperature acquisition points to construct a feature vector matrix; Construct a graph structure based on the spatial distribution of the temperature sensor grid array, take the temperature acquisition points as the graph structure nodes, and take the thermal conductivity as the graph structure edge weights; construct an unsupervised adversarial learning network, where the generator of the unsupervised adversarial learning network generates simulated heat transfer data based on the feature vector matrix, and the corresponding discriminator compares and trains the simulated heat transfer data with the time series dataset; Based on the graph structure, calculate the attention weight coefficient between nodes, and perform weighted fusion of the attention weight coefficient and the heat energy transfer mapping relationship to obtain the heat conduction weight; reconstruct the feature vector matrix according to the heat conduction weight to obtain spatio-temporal correlation features.

3. The method according to claim 1, wherein Based on the spatio-temporal correlation features, establish a nonlinear state equation for the thermal distribution of the die-casting mold, and use the hybrid particle swarm quantum genetic algorithm to solve the nonlinear state equation. The constructed dynamic evolution model of the thermal distribution of the die-casting mold includes: Based on spatio-temporal correlation features, extract and calculate the heat spatial diffusion rate, the heat time cumulative change rate, and the heat transfer coefficient of the mold material, and combine them to construct a state variable vector; based on the heating power parameter and the cooling rate parameter of the die-casting mold, generate a control variable vector; use the state variable vector and the control variable vector to establish a nonlinear state equation describing the heat transfer law inside the die-casting mold; Convert the solution parameters of the nonlinear state equation into quantum state encoding, construct an initial particle swarm including quantum position genes and quantum velocity genes, substitute the quantum position genes and quantum velocity genes of the initial particle swarm into the nonlinear state equation, calculate the fitness value, and determine the individual optimal solution and the global optimal solution of the initial particle swarm based on the fitness value; Based on the individual optimal solution and the global optimal solution, calculate the quantum rotation angle, and based on the quantum rotation angle, perform a quantum rotation gate operation on the quantum position genes to obtain the crossed quantum position genes, and perform a quantum NOT gate operation on the quantum velocity genes to obtain the mutated quantum velocity genes; combine the crossed quantum position genes and the mutated quantum velocity genes to construct an optimized particle swarm; Substitute the quantum position genes and quantum velocity genes of the optimized particle swarm into the nonlinear state equation to calculate the new fitness value, update the individual optimal solution and the global optimal solution according to the new fitness value, and repeat the iterative update until the change rate of the new fitness value is less than the preset convergence threshold to obtain the optimal solution; Construct a temperature state transition matrix of the die-casting mold under discrete time series according to the optimal solution, establish a heat transfer response function between regions of the die-casting mold based on the temperature state transition matrix, and combine the temperature state transition matrix and the heat transfer response function to construct a dynamic evolution model of the thermal distribution of the die-casting mold.

4. The method according to claim 1, wherein Input the dynamic evolution model of the thermal distribution into a deep belief network based on knowledge distillation to construct a die-casting mold temperature field predictor; Use a multi-agent collaborative evolution algorithm to train the temperature field predictor to determine the hierarchical optimization function of the die-casting mold temperature field, including: Convert the temperature state transition matrix in the dynamic evolution model of heat distribution into a feature map, and convert the heat transfer response function in the dynamic evolution model of heat distribution into a time series vector. Normalize the feature map and the time series vector to obtain the input data of the die-casting mold temperature field predictor; Construct a knowledge distillation framework composed of a teacher network and a student network. The teacher network adopts the structure of a deep residual network, and the student network adopts the structure of a three-layer deep belief network. Input the input data into the teacher network to obtain the soft label of the temperature field distribution. Use the soft label of the temperature field distribution as the training target of the student network, pre-train each layer of the student network, and use the contrastive divergence algorithm to train the restricted Boltzmann machine of each layer. Use the trained student network as the basic network structure of the die-casting mold temperature field predictor to obtain the die-casting mold temperature field predictor; Divide the die-casting mold temperature field predictor into multiple sub-predictors. Each sub-predictor is responsible for predicting the temperature of its own independent corresponding area. An information transmission channel is established between the sub-predictors through an attention mechanism; use the multi-agent cooperative co-evolution algorithm to train the sub-predictors, construct a fitness function including local prediction error and prediction consistency, evaluate the prediction performance of the sub-predictors based on the fitness function, use the prediction performance as the evolution index, and optimize the network parameters of each sub-predictor based on the evolution index to obtain the optimized temperature field prediction performance index; construct a hierarchical optimization function of the die-casting mold temperature field according to the temperature field prediction performance index.

5. An optimization system for the thermal balance distribution of a die-casting mold based on an energy model, which is used to implement the method described in any one of the foregoing claims 1-4, and is characterized in that, Include: The first unit is used to collect the die-casting mold heat distribution data, establish the heat transfer mapping relationship of the die-casting mold, convert the heat transfer mapping relationship into a feature vector matrix, and input the feature vector matrix into the graph attention network based on unsupervised adversarial learning for feature extraction to obtain the spatio-temporal correlation features of the die-casting mold; Based on the spatio-temporal correlation features, establish a non-linear state equation of the die-casting mold heat distribution, and use the hybrid particle swarm quantum genetic algorithm to solve the non-linear state equation to construct a dynamic evolution model of the die-casting mold heat distribution; The second unit is used to input the dynamic evolution model of heat distribution into the deep belief network based on knowledge distillation to construct a die-casting mold temperature field predictor; Use the multi-agent cooperative co-evolution algorithm to train the temperature field predictor to determine the hierarchical optimization function of the die-casting mold temperature field; Construct a temperature field optimization strategy network for the die-casting mold based on the hierarchical optimization function, and use the adaptive ant colony optimization algorithm based on tabu search to generate the optimal temperature field control sequence of the die-casting mold; establish a sub-region adaptive control strategy for the die-casting mold according to the optimal temperature field control sequence; The third unit is used to input the sub-region adaptive control strategy into the dynamic evolution model of heat distribution to obtain the temperature field response characteristics of the die-casting mold; According to the temperature field response characteristics, a temperature field controller based on deep reinforcement learning is constructed, and the temperature field change parameters of the die-casting mold are obtained in real time; the temperature field change parameters are input into the temperature field predictor for online correction, and the cooling control parameters are adjusted by using a dynamic fuzzy immune optimization algorithm to realize the intelligent optimization control of the thermal distribution of the die-casting mold.

6. An electronic device, characterized in that, It includes: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, the method according to any one of claims 1 to 4 is implemented.

Images