Prediction method and device for heat dissipation model of extremely small curved surface unit cell structure, electronic equipment, medium and program product

Abstract

The invention provides a prediction method and device for a heat dissipation model of a minimum curved surface unit cell structure, electronic equipment, a medium and a program product. The method comprises the steps that an initial population is acquired, and each individual in the initial population represents a heat dissipation model of a minimum curved surface unit cell structure; the values of at least one first parameter of different heat dissipation models are different. And based on the initial population, carrying out iterative solution on individuals in the initial population by adopting a genetic algorithm to obtain an optimal solution set. In the iterative solution process, based on parameter values corresponding to the individuals, a fitness prediction model constructed based on a neural network is adopted to predict fitness function values of the individuals in the initial population; the fitness function value comprises at least two second parameters, and the second parameters are parameters representing flow heat exchange performance indexes of the heat dissipation model. And determining a heat dissipation model with the optimal flow heat exchange performance from the optimal solution set. According to the method, the minimum curved surface unit cell structure prediction efficiency is greatly improved, and the calculation cost is reduced.

Description

Technical Field

[0001] The present application relates to the technical field of heat dissipation of high power density devices, and in particular to a prediction method, device, electronic device, medium and program product for a heat dissipation model of a minimal surface unit cell structure. Background Art

[0002] With the continuous advancement of integrated circuit technology, semiconductor devices such as high-performance processors and chips are rapidly moving towards miniaturization and high power density. This has led to a significant increase in heat generation per unit volume, necessitating the use of efficient heat sinks to ensure stable operation. As a heat exchange device, the main function of a heat sink is to transfer heat from a heat source (such as a chip or high-performance processor) to the surrounding environment or other media, thereby maintaining a stable operating temperature for the product.

[0003] As a heat dissipation structure, minimal surface structures offer strong flow disturbance, large specific surface area, and good connectivity. By incorporating minimal surface structures into heat sink designs, the heat dissipation efficiency of high-performance processors and chips can be significantly improved, effectively reducing their operating temperatures and extending their service life.

[0004] However, minimal surface structures are relatively complex, and how to improve the efficiency and accuracy of designing minimal surface structures that meet flow and heat transfer performance has become an urgent problem to be solved. Summary of the Invention

[0005] The present application provides a method, device, electronic device, medium and program product for predicting the heat dissipation model of a minimal surface unit cell structure, which is used to solve the problem of how to efficiently and accurately design a heat dissipation model of a minimal surface structure that meets the flow heat transfer performance.

[0006] In a first aspect, the present application provides a method for predicting a heat dissipation model of a minimal surface unit cell structure, the method comprising:

[0007] An initial population is obtained, where each individual in the initial population represents a heat dissipation model of a minimal surface unit cell structure; different heat dissipation models have different values ​​of at least one first parameter, the first parameter including: a structural parameter of the minimal surface unit cell structure and a parameter related to a boundary condition of a fluid flowing through the minimal surface unit cell structure;

[0008] Based on the initial population, a genetic algorithm is used to iteratively solve the individuals in the initial population to obtain an optimal solution set; during the iterative solution process, based on the parameter values ​​corresponding to the individuals, a fitness prediction model constructed based on a neural network is used to predict the fitness function values ​​of the individuals in the initial population; the fitness function values ​​include: at least two second parameters, the second parameters being representative of the flow heat transfer performance of the heat dissipation model;

[0009] A heat dissipation model that determines the optimal flow and heat transfer performance from a set of optimal solutions.

[0010] Optionally, based on the initial population, using the genetic algorithm to iteratively solve the individuals in the initial population to obtain an optimal solution set includes:

[0011] Taking the initial population as the parent population, and based on the parameter values ​​corresponding to the individuals in the parent population, using the fitness prediction model, predicting the fitness function values ​​of the individuals in the parent population;

[0012] Based on the fitness function values ​​of the individuals in the parent population, multi-objective optimization is performed on the parent population to obtain a new parent population;

[0013] Continue to perform multi-objective optimization on the new parent population until the maximum number of iterations is reached;

[0014] The latest parent population is taken as the optimal solution set.

[0015] Optionally, performing multi-objective optimization on the parent population based on the fitness function values ​​of the individuals in the parent population to obtain a new parent population includes:

[0016] Based on the fitness function values ​​of the individuals in the parent population, performing non-dominated sorting and crowding calculation on the parent population;

[0017] Based on the non-dominated sorting result and the crowding calculation result, the parent population is sequentially subjected to individual selection, individual crossover and individual mutation to obtain a child population;

[0018] Using the fitness prediction model, predicting the fitness function values ​​of individuals in the offspring population;

[0019] Merging the parent population with the offspring population to obtain a candidate parent population;

[0020] Based on the fitness function values ​​of the individuals in the candidate parent population, performing non-dominated sorting and crowding calculation on the candidate parent population;

[0021] A preset number of individuals are obtained from the candidate parent population according to the non-dominated sorting result and the crowding calculation result of the candidate parent population to form a new parent population.

[0022] Optionally, determining a heat dissipation model with optimal flow heat transfer performance from the optimal solution set includes:

[0023] Preprocessing the second parameter values ​​of the individuals in the optimal solution set;

[0024] Calculating the entropy weight of the individuals in the optimal solution set based on the preprocessed second parameter values ​​of the individuals in the optimal solution set;

[0025] sorting the individuals in the optimal solution set based on the entropy weights of the individuals in the optimal solution set;

[0026] Based on the ranking result, a heat dissipation model with the optimal flow and heat transfer performance is determined.

[0027] Optionally, preprocessing the second parameter values ​​of the individuals in the optimal solution set includes:

[0028] The second parameter values ​​of the individuals in the optimal solution set are subjected to matrix forwarding and standardization processing.

[0029] Optionally, obtaining the initial population includes:

[0030] Based on the population size and the first parameter of the heat dissipation model of the minimal surface unit cell structure and its value range, multiple individuals are randomly generated; the first parameter of each individual is represented by binary code;

[0031] The binary code of the individual is converted into a decimal floating point number to obtain the initial population.

[0032] Optionally, the method further includes:

[0033] Acquire a sample data set, the sample data set comprising: at least one set of sample data, each set of sample data comprising: a mapping relationship between a first parameter of a heat dissipation model of a minimal surface unit cell structure and a fitness function value;

[0034] Based on the sample data set, the neural network model is trained to obtain the fitness prediction model.

[0035] Optionally, obtaining a sample data set includes:

[0036] Obtaining a first parameter of a heat dissipation model of the minimal surface unit cell structure and a value range thereof;

[0037] Determining a first parameter value of a heat dissipation model of at least one minimal surface unit cell structure based on the first parameter and its value range;

[0038] Based on the first parameter value of the heat dissipation model of at least one minimal surface unit cell structure, a fluid simulation of the heat dissipation model of the minimal surface unit cell structure is performed to obtain at least one set of sample data.

[0039] In a second aspect, the present application provides a device for predicting a heat dissipation model of a minimal surface unit cell structure, the device comprising:

[0040] A first acquisition module is configured to acquire an initial population, wherein each individual in the initial population represents a heat dissipation model of a minimal surface unit cell structure; different heat dissipation models have different values ​​of at least one first parameter, the first parameter including: a structural parameter of the minimal surface unit cell structure, and a parameter related to a boundary condition of a fluid flowing through the minimal surface unit cell structure;

[0041] A first processing module is configured to iteratively solve individuals in the initial population using a genetic algorithm based on the initial population to obtain an optimal solution set; during the iterative solution process, based on parameter values ​​corresponding to the individuals, a fitness prediction model constructed based on a neural network is used to predict fitness function values ​​of the individuals in the initial population; the fitness function value includes at least two second parameters, each of which is a parameter representing a flow heat transfer performance index of the heat dissipation model;

[0042] The determination module is used to determine the heat dissipation model with the optimal flow and heat transfer performance from the optimal solution set.

[0043] In a third aspect, the present application provides an electronic device, comprising: a processor, and a memory communicatively connected to the processor;

[0044] The memory stores computer-executable instructions;

[0045] The processor executes the computer-executable instructions stored in the memory to implement the method as described in any one of the first aspects.

[0046] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method as described in any one of the first aspects.

[0047] In a fifth aspect, the present application provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of the first aspects of the claims.

[0048] The present application provides a method, device, electronic device, medium and program product for predicting the heat dissipation model of a minimal surface unit cell structure. By combining a neural network with a genetic algorithm to design a heat dissipation model of a minimal surface structure, the present application provides a method, device, electronic device, medium and program product for predicting the heat dissipation model of a minimal surface unit cell structure. This method can achieve rapid and accurate prediction of the optimal heat dissipation model of a minimal surface unit cell structure that meets the flow and heat transfer performance indicators. This method can greatly shorten the iterative calculation cycle of product design, reduce the time cost of product design, and improve the accuracy of prediction and optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0050] Figure 1 A schematic flow chart of a method for predicting a heat dissipation model of a minimal surface unit cell structure provided in an embodiment of the present application;

[0051] Figure 2 A schematic diagram of a unit cell structure of a Gyroid minimal surface unit cell structure provided in an embodiment of the present application under the condition that the unit cell solid volume fraction varies in the range of 5% to 60%;

[0052] Figure 3 A structural diagram of a neural network model provided in an embodiment of the present application;

[0053] Figure 4 A schematic diagram showing a comparison of the error between a simulation calculation result of a convective heat transfer coefficient and a prediction result of a neural network model provided in an embodiment of the present application;

[0054] Figure 5 A schematic diagram comparing the error between a simulation calculation result of a fluid flow pressure drop and a prediction result of a neural network model provided in an embodiment of the present application;

[0055] Figure 6 A flow chart of a method for obtaining an optimal solution set using a genetic algorithm provided in an embodiment of the present application;

[0056] Figure 7 A schematic diagram of a Pareto optimal solution set obtained through multi-objective optimization provided in an embodiment of the present application;

[0057] Figure 8 A flow chart of a method for determining a heat dissipation model with optimal flow and heat transfer performance from an optimal solution set provided in an embodiment of the present application;

[0058] Figure 9 A schematic diagram of the results of using the TOPSIS method to sort individuals in the optimal solution set provided in an embodiment of the present application;

[0059] Figure 10 A schematic flow chart of a prediction method for a heat dissipation model for determining optimal flow heat transfer performance provided in an embodiment of the present application;

[0060] Figure 11 An exploded diagram of a chip and overall heat dissipation structure provided in an embodiment of the present application;

[0061] Figure 12 A schematic diagram of a device for predicting a heat dissipation model of a minimal surface unit cell structure provided in an embodiment of the present application;

[0062] Figure 13 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0063] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0064] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0065] With the continuous development of current integrated circuit technology, high-performance processors, chips and other related products are gradually developing rapidly towards miniaturization and high power density, which leads to a significant increase in the heat generation per unit volume. Therefore, high-efficiency heat sinks are needed to ensure stable operation.

[0066] As a heat exchange device, a radiator's primary function is to transfer heat from a heat source (such as a chip or high-performance processor) to the surrounding environment or other media, thereby maintaining a stable product operating temperature. Radiators include those that use fluids (liquid or gas) to transfer heat. Common examples of these include liquid-cooled radiators, air-cooled radiators, water-cooled radiators, and heat pipe radiators. These types of radiators are referred to as fluid radiators.

[0067] Minimal surface structures are complex porous structures designed based on minimal surfaces. Minimal surfaces are a special type of geometric surface, defined as surfaces with approximately zero mean curvature under given boundary conditions. This property enables minimal surfaces to minimize surface area locally while exhibiting a high overall surface area. Furthermore, minimal surfaces lack sharp corners or sudden changes, exhibiting a smooth geometry. These unique geometric properties enable minimal surface structures to exhibit excellent heat dissipation capabilities.

[0068] The high specific surface area of ​​minimally curved structures enables more efficient heat exchange with the surrounding environment. Given the same volume, a high specific surface area means a larger heat exchange interface, allowing for faster heat dissipation and significantly improved heat dissipation efficiency. Furthermore, the continuous curvature and smooth surface of minimally curved surfaces help reduce resistance to fluid flow, further enhancing convective cooling. Due to their excellent heat dissipation capabilities, minimally curved structures are widely used in the design of various fluid radiators.

[0069] Minimal surface unit cells are the fundamental building blocks of minimal surface structures. Minimal surface unit cells are repeated in space according to a certain periodic pattern, forming more complex minimal surface polyhedral structures. Minimal surface structures can consist of a single minimal surface unit cell or a structure composed of multiple units.

[0070] Currently, the heat dissipation model of minimal surface structures can be obtained through the following methods.

[0071] 1. Simulation optimization iterative method

[0072] The iterative simulation optimization method is a systematic approach that combines simulation technology with optimization algorithms. When using iterative simulation optimization to obtain a heat dissipation model for minimal surface structures, preprocessing is first required to clarify the simulation optimization requirements for the heat dissipation model of the minimal surface structure and construct an initial geometric model based on these requirements. Next, meshing the initial geometric model is performed, selecting an appropriate mesh type and density to ensure the accuracy and efficiency of the numerical calculations.

[0073] Subsequently, the initial geometric model is analyzed using simulation techniques such as the finite element method to extract key indicators. Based on these indicators, the heat dissipation model of the minimal surface structure is adjusted to generate a new heat dissipation model for the minimal surface structure. The meshing, numerical calculation, and optimization adjustment process is repeated until a heat dissipation model for the minimal surface structure that meets the design requirements is obtained.

[0074] While iterative simulation optimization methods can be used to design heat dissipation models for minimal surface unit cell junctions, they are computationally expensive. The complex geometry of minimal surface structures complicates meshing, often requiring significant time and effort. Furthermore, the iterative process requires repeated simulations, consuming significant computational resources. Consequently, this method has low optimization efficiency.

[0075] 2. Neural Networks

[0076] Currently, there are also solutions using neural network models to achieve structural design in minimal surface structures in non-heat dissipation applications. Specifically, by training the neural network model to learn the mapping relationship between the topological parameters of the minimal surface structure and related parameters and mechanical performance indicators, the neural network model can be used to predict the minimal surface structure with the optimal pressure bearing capacity, providing guidance for design optimization.

[0077] However, the neural network model falls into the problem of local optimal solution during training. Therefore, it is difficult to find the global optimal solution by using the above method to design the minimal surface structure.

[0078] 3. Genetic Algorithm

[0079] Genetic algorithm is an optimization algorithm based on the principles of natural selection and genetics. By simulating the selection, crossover and mutation operations in the biological evolution process, it can effectively avoid local optimal solutions in complex optimization problems and gradually approach the global optimal solution.

[0080] Genetic algorithms are currently widely used in the design of fin-cell structures for engine radiators. This method uses the pyrolysis dissipation function as the fitness function and employs a genetic algorithm to optimize the layout of the fin structure. Specifically, individuals are encoded based on the original fin structure to generate a parent population. The fitness function value of each individual in the initial population is calculated. Based on this, the parent population is iteratively optimized through selection, crossover, and mutation operations. Ultimately, after multiple generations of evolution, the optimized fin structure is obtained.

[0081] The genetic algorithm used in the above solution is a single-objective optimization approach. The method involves a relatively simple fin structure and uses forced air cooling, which cannot meet the heat dissipation requirements of high-power density semiconductor devices. Single-objective optimization refers to the process of finding an optimal solution in a genetic algorithm to maximize or minimize a single objective.

[0082] In addition, although some solutions have adopted multi-objective optimization methods when designing fin cell structures using genetic algorithms, they only optimize the structural parameters of the fin cell structure and do not consider the impact of other factors on heat dissipation performance, resulting in low accuracy of the above prediction methods. Among them, multi-objective optimization is an optimization method that seeks a balance between at least two or more conflicting objectives.

[0083] Furthermore, when the above scheme is designed using a genetic algorithm, it is necessary to pre-construct a fitness function expression. The fitness function expression is often difficult to obtain, resulting in low efficiency in designing the heat dissipation structure using the above method. Therefore, how to efficiently and accurately design a heat dissipation model of a minimal surface unit cell structure that meets the flow and heat transfer performance has become an urgent problem to be solved.

[0084] In view of this, the present application provides a method for constructing a heat dissipation model for a minimal surface unit cell structure. This method combines a neural network with a genetic algorithm, not only taking into account the structural parameters of the minimal surface unit cell structure itself, but also further combining other parameters related to the boundary conditions of the fluid flowing through the minimal surface unit cell structure as variables to be optimized, thereby constructing a heat dissipation model for the minimal surface unit cell structure. Without the need to construct a fitness function expression, it is only necessary to pre-set the variables and ranges to be optimized for the minimal surface unit cell structure, so as to quickly and accurately predict the optimal heat dissipation model of the minimal surface unit cell structure that meets the fluid heat transfer performance index. This can greatly shorten the iterative calculation cycle of product design, reduce the time cost of product design, and improve the accuracy of prediction and optimization.

[0085] It should be noted that the heat dissipation model of the minimal surface unit cell structure predicted by the method of the embodiment of the present application can be used to construct a fluid radiator, which can be used, for example, in the fields of lithium-ion battery heat dissipation, chip / semiconductor heat dissipation, microchannel heat dissipation technology, electronic equipment heat dissipation technology, etc., without limitation to this.

[0086] The prediction method for the heat dissipation model of the minimal surface unit cell structure provided in the embodiment of the present application can be applied to a prediction system or prediction platform, or a system or platform integrated with the prediction function, or an electronic device, etc.

[0087] In some embodiments, the embodiments of the present application can also be applied to applications, websites or applets with the above-mentioned prediction function. On the application, website or applet, the prediction function of the heat dissipation model of the minimal surface unit cell structure is implemented. For example, a website that deploys the method provided by the embodiment of the present application implements the prediction of the heat dissipation model of the minimal surface unit cell structure. For example, a third-party application calls the deployed method through the application programming interface (Application Programming Interface, referred to as API) to implement the prediction of the heat dissipation model of the minimal surface unit cell structure. Optionally, on the basis of realizing the prediction of the heat dissipation model of the minimal surface unit cell structure, the design of the radiator can be further implemented, and this is not limited.

[0088] For ease of description, the following embodiments are all described using a prediction system as an example.

[0089] The following is a detailed description of the method provided in the embodiments of the present application. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0090] Figure 1 A flow chart of a method for predicting a heat dissipation model of a minimal surface unit cell structure provided in an embodiment of the present application. Figure 1As shown, the method includes the following steps:

[0091] S101. Obtain an initial population, where each individual in the initial population represents a heat dissipation model of a minimal surface unit cell structure; different heat dissipation models have different values ​​of at least one first parameter, and the first parameters include: structural parameters of the minimal surface unit cell structure, and parameters related to boundary conditions of a fluid flowing through the minimal surface unit cell structure.

[0092] The initial population refers to the set of candidate solutions generated at the start of the genetic algorithm. Each individual in the initial population represents a heat dissipation model for a minimal surface unit cell. These individuals serve as the starting point for the genetic algorithm's optimization, gradually evolving to better solutions through iteration and selection.

[0093] The first parameters include structural parameters of the minimal surface unit cell structure and parameters related to boundary conditions of the fluid flowing through the minimal surface unit cell structure.

[0094] The structural parameters of the minimal surface unit cell structure may include, for example, geometric parameters and topological parameters, which are not limited in the present embodiment.

[0095] Among them, boundary conditions refer to the constraints imposed on the boundaries of the minimal surface unit cell structure, which are used to define the interaction between the minimal surface unit cell structure and the external environment.

[0096] Parameters related to the boundary conditions of the fluid flowing through the minimal surface unit cell structure may include, for example, the fluid inlet flow rate, the fluid pressure, and the like.

[0097] Fluid inlet flow rate refers to the rate at which fluid enters a minimal surface unit cell. It defines the amount of fluid passing through the inlet cross-section per unit time. Generally, the greater the flow rate, the more heat the fluid removes, and the higher the heat dissipation efficiency.

[0098] Fluid pressure refers to the pressure of a fluid at its inlet or outlet. The pressure of the fluid entering the inlet of a minimally curved surface unit cell structure is called the inlet pressure, while the pressure of the fluid exiting the outlet of the minimally curved surface unit cell structure is called the outlet pressure. The difference between the inlet and outlet pressures is called the pressure drop. Higher inlet pressure generally increases the fluid flow rate and improves the flow and heat transfer efficiency of the minimally curved surface unit cell structure. Lower outlet pressure increases the pressure drop of the fluid passing through the minimally curved surface unit cell structure, resulting in higher overall energy consumption.

[0099] Geometric parameters may include, for example, unit cell size, unit cell solid volume fraction, porosity, and other parameters.

[0100] Unit cell size refers to the physical dimensions of the basic unit that makes up the entire minimal surface unit cell structure. The smaller the unit cell size, the larger the specific surface area of ​​the minimal surface unit cell structure, and the better the flow and heat transfer performance.

[0101] The unit cell solid volume fraction is the ratio of the solid volume within a minimal surface unit cell to its total volume. The unit cell solid volume fraction and porosity are complementary; as the unit cell solid volume fraction increases, the porosity decreases accordingly.

[0102] Porosity refers to the percentage of pore volume in a minimally curved unit cell structure to its total volume. Higher porosity increases the flow space within the minimally curved unit cell, facilitating fluid flow and heat transfer. However, excessively high porosity can also loosen the minimally curved unit cell structure, reducing its thermal conductivity. Therefore, the impact of porosity on flow and heat transfer performance requires comprehensive consideration.

[0103] Topological parameters include parameters such as structure type, which includes Gyroid structure, Diamond structure, I-Werner structure (I-WP) and Primitive structure. Different types of minimal surface unit cell structures have different flow and heat transfer performance.

[0104] It should be noted that the above examples of first parameters are merely illustrative. The present embodiments do not limit the number or range of first parameters used in designing minimal surface unit cell structures, and can be adjusted based on different usage scenarios and heat dissipation requirements. For example, the unit cell solid volume fraction and the fluid inlet flow rate can be used as first parameters to predict the minimal surface unit cell structure with optimal flow and heat transfer performance.

[0105] This application does not limit the method for obtaining the initial population. Reference may be made to the prior art methods for constructing the initial population when using genetic algorithms for prediction. For example, multiple individuals are randomly generated based on the population size and the first parameter and its value range of the heat dissipation model of the minimal surface unit cell structure. The first parameter of each individual is represented using a binary code. The individual's binary code is converted to a decimal floating point number to obtain the initial population. The population size refers to the number of individuals in the genetic algorithm.

[0106] The above conversion of binary codes to decimal floating-point numbers is because in genetic algorithms, the values ​​of the first parameters of individuals are usually expressed in binary code, but in real problems, the values ​​of the first parameters may be decimal floating-point numbers or integers. The decoding process converts the binary codes into actual values.

[0107] S102. Based on the initial population, a genetic algorithm is used to iteratively solve the individuals in the initial population to obtain an optimal solution set. During the iterative solution process, a fitness prediction model constructed based on a neural network is used to predict the fitness function values ​​of the individuals in the initial population based on the parameter values ​​corresponding to the individuals. The fitness function value includes at least two second parameters, each of which is a parameter representing a flow heat transfer performance indicator of the heat dissipation model.

[0108] Iterative problem-solving in a genetic algorithm involves starting with an initial population and continuously performing operations such as selection, crossover, and mutation to generate a new generation of populations. The fitness prediction model then evaluates the performance of each individual in the population, gradually approaching the optimal solution. Each generation represents an improvement to the problem solution, and through multiple generations of iteration, the optimal solution set is ultimately found.

[0109] The fitness function value is directly related to the optimization objective, reflecting the degree to which that objective is achieved. The optimization objective is the ultimate goal to be achieved in a multi-objective optimization problem, typically expressed as maximizing or minimizing the parameters of a flow and heat transfer performance indicator. If the optimization objective is to maximize the flow and heat transfer performance indicator, a larger fitness function value indicates a more robust individual and closer to the optimal solution. If the optimization objective is to minimize the flow and heat transfer performance indicator, a smaller fitness function value indicates a more robust individual and closer to the optimal solution.

[0110] In this embodiment, the fitness function value includes at least two second parameters. That is, the optimization objective includes at least two second parameters. In other words, this embodiment of the application is a multi-objective optimization problem. Multi-objective optimization can simultaneously consider multiple optimization objectives, synthesizing the relationships between the various optimization objectives to obtain the optimal minimal surface unit cell structure within a given constraint range.

[0111] The second parameter characterizes the flow and heat transfer performance of the minimal surface unit cell's heat dissipation model. These parameters, along with the fluid flow index, can be used to assess the flow and heat transfer performance of the minimal surface unit cell's heat dissipation model design. This quantitatively evaluates the performance of the minimal surface unit cell's heat dissipation model in a specific application, providing a clear reference for its design and optimization.

[0112] The second parameter may include, for example, heat dissipation performance parameters, fluid flow parameters, and the like, which are not limited in this embodiment of the present application.

[0113] Heat dissipation performance parameters include convection heat transfer coefficient, cooling fluid loss and temperature uniformity.

[0114] The convective heat transfer coefficient is a physical quantity that describes the convective heat transfer capacity between a fluid and a solid surface. It reflects the amount of heat that can be transferred between the fluid and the wall per unit area and per unit time at a unit temperature difference. A higher convective heat transfer coefficient indicates better flow heat transfer performance.

[0115] Coolant pressure loss refers to the pressure loss caused by flow resistance when the coolant flows through a minimally curved unit cell structure. Higher coolant pressure loss increases energy consumption and reduces flow and heat transfer efficiency.

[0116] Temperature uniformity refers to the uniformity of temperature distribution within a minimal surface unit cell. Good temperature uniformity can reduce the formation of local hot spots and improve the efficiency of flow and heat transfer performance.

[0117] Fluid flow parameters include parameters such as fluid flow rate and pressure drop.

[0118] Fluid flow pressure drop refers to the pressure drop caused by friction and resistance when a fluid flows through a minimal surface unit cell structure. It reflects the amount of resistance encountered by the fluid as it passes through the structure and is an important indicator for evaluating the flow performance of a minimal surface unit cell heat dissipation model. Therefore, the smaller the fluid flow pressure drop, the better the flow and heat transfer performance of the minimal surface unit cell heat dissipation model.

[0119] For ease of understanding, the following examples are described using the convective heat transfer coefficient and fluid flow pressure drop as targets. It should be noted that this application does not limit the number of second parameters used in the design of the heat dissipation model for minimal surface unit cells. These parameters can be adjusted based on different usage scenarios and heat dissipation requirements.

[0120] As mentioned above, the optimization objective is the ultimate goal to be achieved in a multi-objective optimization problem, typically expressed as maximizing or minimizing a flow and heat transfer performance parameter. If the optimization objective requires maximizing certain flow and heat transfer performance parameters while minimizing them, the fitness function can be designed so that certain flow and heat transfer performance parameters are negative compared to the original flow and heat transfer performance parameters.

[0121] In this implementation, taking the convective heat transfer coefficient and the fluid flow pressure drop as examples, the fitness function value may include, for example, the negative of the convective heat transfer coefficient and the fluid flow pressure drop. Alternatively, the fitness function value may include, for example, the convective heat transfer coefficient and the negative of the fluid flow pressure drop.

[0122] In this embodiment, a fitness prediction model that has pre-learned the mapping relationship between the first parameter and the fitness function value is introduced, and the fitness function value is predicted based on the first parameter. In this way, there is no need to construct a fitness function expression based on the first parameter and the fitness function value in advance. The use of the fitness prediction model can avoid the problems of high computational complexity and parameter adjustment difficulty between the first parameter and multiple optimization objectives in multi-objective optimization problems, thereby improving the efficiency of the optimization process.

[0123] For example, the fitness prediction model can be obtained by pre-training a neural network model using the values ​​of the first parameter and the fitness function value. The execution subject of the training process of the model can be the same as the execution subject of the embodiment of the present application, or a different execution subject.

[0124] Exemplarily, first, a sample dataset is obtained. The sample dataset includes at least one set of sample data, each set of sample data including a mapping relationship between a first parameter of a heat dissipation model of a minimal surface unit cell structure and a fitness function value. A neural network model is trained based on the sample dataset to obtain a fitness prediction model.

[0125] The sample dataset is a portion of data extracted from the design space, used to reflect the characteristics of the entire design space. The design space refers to the set of first parameter combinations and their value ranges for all possible minimal surface unit cell heat dissipation models. These parameters and their value ranges define all possible configurations of the fitness prediction model.

[0126] For example, a first parameter of a heat dissipation model of a minimal surface unit cell structure and its value range are obtained. Based on the first parameter and its value range, a value of the first parameter of the heat dissipation model of at least one minimal surface unit cell structure is determined. The value of the first parameter can be any value within the value range.

[0127] A fluid simulation of the heat dissipation model of the minimal surface unit cell structure is performed based on a first parameter value of at least one minimal surface unit cell structure, and a fitness function value under the at least one first parameter is obtained, thereby obtaining at least one set of sample data based on the at least one first parameter and the fitness function value. The set of sample data may include, for example, the first parameter and the corresponding fitness function value.

[0128] Based on the at least one set of sample data, a sample data set for training a neural network model is obtained.

[0129] In a specific embodiment, a first parameter in which the variation range of the unit cell solid volume fraction is 5% to 60% and the variation range of the fluid inlet flow rate is 0.01 to 0.4 m / s is selected as the sample data set.

[0130] Figure 2 A schematic diagram of a Gyroid minimal surface unit cell structure provided in an embodiment of the present application under the condition that the unit cell solid volume fraction varies within a range of 5% to 60%. Here, a is the value of the unit cell solid volume fraction.

[0131] like Figure 2 As shown in the figure, the larger the unit cell solid volume fraction, the smaller the porosity. As shown above, the unit cell solid volume fraction and porosity are complementary to each other. As the unit cell solid volume fraction increases, the porosity decreases accordingly. The higher the porosity, the larger the flow space of the minimal surface unit cell structure, which is conducive to fluid flow and heat transfer. However, excessively high porosity may also cause the minimal surface unit cell structure to become loose, reducing its thermal conductivity. Therefore, when designing a heat dissipation model for a minimal surface unit cell structure, it is necessary to select an appropriate unit cell solid volume fraction.

[0132] Secondly, set the structure of the neural network model and determine the training algorithm, and then train the neural network model.

[0133] Among them, setting the structure of the neural network model may include, for example, determining key elements such as the hierarchy, connection method, and activation function of the neural network.

[0134] The neural network hierarchy includes, for example, an input layer, a hidden layer, and an output layer.

[0135] The input layer, for example, may be the first layer of a neural network model, responsible for receiving input data. The number of neurons in the input layer may, for example, be the same as the number of the first input parameter. The output layer, for example, may be the last layer of a neural network model, responsible for outputting the prediction results of the neural network model. The number of neurons in the output layer may, for example, be the same as the number of fitness function values.

[0136] For example, the unit cell solid volume fraction and the fluid inlet flow rate are used as the first parameters, and the negative of the convective heat transfer coefficient and the fluid flow pressure drop are used as the fitness function values. In this example, the neural network model has two input layer neurons, including two variables: the unit cell solid volume fraction and the fluid inlet flow rate. The output layer also has two neurons, including two variables: the negative of the convective heat transfer coefficient and the fluid flow pressure drop.

[0137] The hidden layer is a crucial component of a neural network, located, for example, between the input and output layers. Neurons in the hidden layer do not interact directly with the external environment. Instead, they receive data from the input layer and process it through operations such as weighting, summation, and activation. The results are then passed to the next layer. The number of neurons in the hidden layer must be determined by comprehensively considering factors such as the number of input data features and the number of output data categories, model performance, and computing resources. The optimal configuration should be found through repeated trials and adjustments.

[0138] The number of neurons in the hidden layer can be determined, for example, according to the following formula (1):

[0139]

[0140] Where c is the number of neurons in the hidden layer, n is the number of neurons in the input layer, and m is the number of neurons in the output layer. l can be a constant between 1 and 10, for example. For example, if the number of neurons in both the input and output layers is 2, the value range of c calculated based on the above formula (1) can be, for example, [3, 12].

[0141] A specific embodiment, Figure 3 A structural diagram of a neural network model provided in an embodiment of the present application, such as Figure 3 As shown in Figure 2, the input layer of the neural network model is the solid volume fraction a, i.e., the unit cell solid volume fraction, and the inlet flow velocity v, i.e., the fluid inlet flow rate. The output layer is the negative of the overall convective heat transfer coefficient h, i.e., the negative of the convective heat transfer coefficient, and the inlet and outlet pressure drop Δp, i.e., the fluid flow pressure drop. The number of neurons in the hidden layer is 10.

[0142] The connection mode refers to how neurons are connected to each other to transmit and process information. The connection mode may include, for example, full connection, local connection, convolution connection, recurrent connection, etc., which is not limited in the embodiments of the present application.

[0143] The activation function is a function that runs on the neurons in the structure of the neural network model. It enables the neural network to learn complex patterns and relationships by introducing nonlinearity. The embodiments of the present application do not limit the choice of activation function.

[0144] It should be noted that the structure of the above-mentioned neural network model is only one possible implementation method, and the embodiments of the present application do not limit the specific setting of the structure of the neural network model.

[0145] The training algorithm is determined to be an optimization method that can efficiently optimize the neural network input parameters to minimize the loss function, such as gradient descent method, Newton algorithm, conjugate gradient method, quasi-Newton method, etc., which is not limited in this embodiment of the present application.

[0146] Then, a portion of data is randomly selected from the sample data set as a training set to train the neural network model. After the training is completed, the mapping relationship between the input layer neurons and the output layer neurons is obtained.

[0147] In one specific embodiment, 70% of the data in the sample dataset is selected as the training set to train the neural network model. After training, the mapping relationship between the unit cell solid volume fraction, the negative of the fluid inlet flow rate and the convective heat transfer coefficient, and the fluid flow pressure drop is obtained.

[0148] Finally, the neural network model is verified after training.

[0149] For example, data other than the training set in the sample data set is extracted as a test set, and the test set is used to verify the trained neural network model.

[0150] For another example, extract part of the data in the sample data set as a test sample data set. Use the test sample data set to simulate the heat dissipation model of the minimal surface unit cell structure using the simulation optimization iterative method. Obtain the flow heat transfer performance of the heat dissipation model of the minimal surface unit cell structure corresponding to different data points, that is, the convective heat transfer coefficient and the fluid flow pressure drop. Subsequently, use the test sample data set to verify the trained neural network model. Compare and verify the results obtained by the neural network model verification with the results obtained by the simulation calculation. If the results obtained by the neural network model verification and the results obtained by the simulation calculation are within a reasonable error range, it proves that the neural network model training is complete and a fitness prediction model is obtained. If the results obtained by the neural network model verification and the results obtained by the simulation calculation exceed the error range, the neural network model is retrained.

[0151] The error range is selected based on actual conditions and is not limited in this embodiment of the present application. For example, if the error between the negative convective heat transfer coefficient and the fluid flow pressure drop predicted by the constructed neural network model and the simulation results is within 5%, or if the coefficient of determination for the negative convective heat transfer coefficient is 0.9969 and the coefficient of determination for the fluid flow pressure drop is 0.9962, the neural network model has been trained.

[0152] A specific implementation method, Figure 4 A schematic diagram comparing the errors between the simulation calculation results of the convective heat transfer coefficient and the prediction results of the neural network model provided in an embodiment of the present application. Figure 5 A schematic diagram comparing the errors of a simulation calculation result of a fluid flow pressure drop and a prediction result of a neural network model provided in an embodiment of the present application.

[0153] Figure 4 (a) with Figure 5 (a) is the direct error diagram, such as Figure 4 (a) with Figure 5 As shown in (a), the black squares are simulation results and the white circles are prediction results. Figure 4 (a) with Figure 5 (a) The horizontal axes are all sample point numbers. These numbers correspond to specific combinations of the first parameters. One number corresponds to a group of first parameters. Figure 4 (a) The vertical axis is the convective heat transfer coefficient h, Figure 5 (a) The vertical axis is the fluid flow pressure drop Δp. Figure 4 (a) with Figure 5 In (a), we can directly see the error between the simulation results and the prediction results when the sample points are the same, that is, when the value of the first parameter is the same.

[0154] Figure 4 (b) with Figure 5 (b) is the diagonal error diagram, such as Figure 4 (b) with Figure 5 As shown in (b), the horizontal axis is the simulation result, and the vertical axis is the prediction result. Figure 4 The black squares in (b) represent the negative convective heat transfer coefficient h when the sample point numbers of the simulation results and the predicted results are the same. Figure 5 The black squares in (b) represent the fluid flow pressure drop Δp when the sample point numbers of the simulation results and the prediction results are the same. Figure 4 (b) with Figure 5 The solid line in (b) represents the ideal case where the predicted results are completely consistent with the simulation results, that is, the line where y = x. Figure 4 (b) with Figure 5 The closer the points in (b) are to this line, the closer the predicted results are to the simulation results, and the better the prediction performance of the model.

[0155] Figure 4 (b) with Figure 5 In (b), two lines representing ±5% error are drawn, as shown in Figure 4 (b) with Figure 5 As shown in (b), all data points fall between the two lines, which indicates that the error is usually within 5% and the prediction has high accuracy.

[0156] The convective heat transfer coefficient and fluid flow pressure drop predicted by the neural network model constructed based on the above method are within 5% of the simulation results, indicating that the neural network training effect is good and the prediction results are accurate.

[0157] By comparing the neural network model verification results with the simulation calculation results, potential problems of the neural network model can be discovered in a timely manner, the accuracy and generalization ability of the neural network model prediction can be improved, and the credibility of the neural network model can be enhanced.

[0158] The embodiments of the present application do not limit the verification method of the neural network model.

[0159] S103. Determine a heat dissipation model with optimal flow and heat transfer performance from the optimal solution set.

[0160] For example, the Technique for Order Preference by Similarity to Ideal Solution with Entropy Weight Method (TOPSIS), hereinafter referred to as the TOPSIS entropy weight method, can be used to determine the heat dissipation model with the optimal flow and heat transfer performance from the optimal solution set. For another example, the heat dissipation model with the optimal flow and heat transfer performance can be determined from the optimal solution set based on a comprehensive evaluation method of the Analytic Hierarchy Process (AHP). For another example, the heat dissipation model with the optimal flow and heat transfer performance can also be confirmed from the optimal solution set based on a Computer-Aided Geometric Design (CAGD) method. For another example, any solution in the optimal solution set can also be randomly selected as the heat dissipation model with the optimal flow and heat transfer performance. The embodiment of the present application does not limit the method for determining the heat dissipation model with the optimal flow and heat transfer performance from the optimal solution set.

[0161] The method provided in this application combines neural networks and genetic algorithms. Through the global search capability of genetic algorithms, it can effectively find the global optimal solution set for the design of the heat dissipation model of the minimal surface unit cell structure, avoiding the problem that the neural network may fall into the local optimal solution in the design of the heat dissipation model of the minimal surface unit cell structure. In addition, this method does not need to obtain the relationship expression between the optimization parameters and the optimization target. It only needs to pre-set the variables and ranges required to optimize the heat dissipation model of the minimal surface unit cell structure, and then use the neural network to construct a fitness prediction model, reducing the computational complexity and the difficulty of parameter adjustment. This method can also adjust the optimized variables and ranges according to different needs and heat dissipation conditions to obtain a heat dissipation model with optimal flow and heat transfer performance within a given constraint range. This method greatly shortens the iterative calculation cycle of product design, reduces the time cost of product design, and improves the accuracy of prediction and optimization.

[0162] The following example illustrates how to use a genetic algorithm to iteratively solve the individuals in the initial population based on the initial population to obtain the optimal solution set. Figure 6 A flow chart of a method for obtaining an optimal solution set using a genetic algorithm is provided in an embodiment of the present application, as shown in FIG. Figure 6 As shown, the method includes the following steps:

[0163] S601: Taking the initial population as the parent population, and based on the parameter values ​​corresponding to the individuals in the parent population, using the fitness prediction model, predict the fitness function values ​​of the individuals in the parent population.

[0164] Among them, the individuals in the parent population will serve as the basis for subsequent genetic operations (such as selection, crossover, and mutation). As shown above, in genetic algorithms, the fitness function value is used to evaluate the quality of individuals. In the embodiment of the present application, the lower the fitness function value, the better the individual is and the closer it is to the optimal solution to the problem.

[0165] Exemplarily, the parameter values ​​corresponding to the individuals in the parent population are input into the fitness prediction model to predict the fitness function values ​​of the individuals in the parent population.

[0166] S602: Based on the fitness function values ​​of the individuals in the parent population, multi-objective optimization is performed on the parent population to obtain a new parent population.

[0167] Exemplarily, first, based on the fitness function values ​​of the individuals in the parent population, a non-dominated sorting and crowding degree calculation is performed on the parent population.

[0168] Among them, non-dominated sorting is to divide the individuals in the parent population into different non-dominated levels. In this process, individuals in the parent population compare with each other based on their fitness function values ​​to determine who dominates whom or who is not dominated by anyone.

[0169] In multi-objective optimization, if individual a outperforms individual b on all objectives, or outperforms individual b on some objectives and is at least as good as individual b on others, then individual a is said to dominate individual b. All individuals not dominated by any other individual are placed in the first non-dominated layer with a Pareto rank of 1. Then, the dominance of all individuals in the first non-dominated layer over the remaining individuals is removed. Among the remaining individuals, find those that are not dominated by any other solution and place them in the second non-dominated layer, and so on.

[0170] Among them, the Pareto rank is an indicator used to measure the degree of dominance of an individual in multi-objective optimization. It defines the rank based on the number of times an individual is dominated by other individuals. The fewer times an individual is dominated, the higher the rank, which helps to identify and select the optimal non-dominated individual.

[0171] The non-dominated sorting method ensures that during the selection process, better individuals have a higher chance of being selected into the next generation. In addition, the same non-dominated level contains multiple individuals, which achieve different trade-offs between multiple optimization objectives and maintain the diversity of the population.

[0172] Crowding is used to further differentiate individuals within the same non-dominated layer, maintaining diversity. It measures the density of other individuals around a given individual. Individuals with high crowding are relatively isolated in the target space, helping to maintain population diversity. The crowding result is used to further differentiate individuals when the non-dominated rankings are the same. In multi-objective optimization, the target space is the space consisting of the values ​​of all optimization objectives.

[0173] The congestion degree calculation method can be expressed by the following formula (2), for example.

[0174]

[0175] Where d(j) represents the crowding distance of the jth individual on the i-th optimization target. k represents the total number of optimization targets. i (j-1)-f i (j+1) represents the optimization target values ​​of the two individuals adjacent to the jth individual on the i-th optimization target. Here, j-1 and j+1 represent the adjacent positions after the optimization target values ​​are sorted. i max and f i min The values ​​d(j) and d(j) represent the maximum and minimum values ​​of the i-th optimization objective over the entire parent population, respectively. They are used to normalize the crowding distance so that the crowding distances between different optimization objectives are comparable. The normalized differences across all optimization objectives are summed to obtain the crowding distance d(j) of individual j.

[0176] In summary, non-dominated sorting and crowding calculation work together in genetic algorithms to achieve efficient and diverse optimal solution sets for multi-objective optimization.

[0177] Secondly, based on the non-dominated sorting results and crowding calculation results, individual selection, individual crossover and individual mutation are performed on the parent population in sequence to obtain the offspring population.

[0178] It should be understood that individuals in higher non-dominated layers and individuals with higher crowding are more likely to be selected.

[0179] For example, individuals are selected from the parent population to form the offspring population based on the fitness function value, the non-dominated sorting result, and the crowding degree. For example, offspring population selection can be performed using methods such as roulette wheel selection, tournament selection, or crowding-based distance selection. The goal of selection is to retain high-quality individuals and maintain population diversity.

[0180] For example, selected individuals undergo a crossover operation to generate new individuals. The proportion of individual pairs that undergo crossover can be determined, for example, by the crossover probability. The crossover operation simulates genetic recombination in biological genetics, generating new genetic combinations by exchanging partial genetic information between two individuals. The crossover operation helps explore new search spaces and improves the algorithm's global search capabilities.

[0181] The crossover probability indicates how many proportions of individual pairs will actually undergo crossover when selecting individual pairs for crossover operation. The crossover probability can be reasonably set and dynamically adjusted when the genetic algorithm is initialized according to actual conditions, and the present application embodiment does not limit this.

[0182] For example, mutation operations are performed on the new individuals generated by crossover to introduce new genetic diversity. The number of individuals that actually mutate can depend on, for example, the mutation probability. Mutation helps the genetic algorithm escape local optima and enhance global search capabilities. The mutation probability represents the probability that each individual in the population will mutate. Generally speaking, the mutation probability is usually set low to avoid destroying useful genetic information. The mutation probability can be reasonably set and dynamically adjusted during the initialization of the genetic algorithm based on actual conditions, and this embodiment of the application does not impose any restrictions on this.

[0183] Then, using the fitness prediction model, the fitness function values ​​of the individuals in the offspring population are predicted. For example, the parent population and the offspring population can be merged to obtain a candidate parent population. Based on the fitness function values ​​of the individuals in the candidate parent population, a non-dominated sorting and crowding calculation are performed on the candidate parent population. Based on the non-dominated sorting and crowding calculation results of the candidate parent population, a preset number of individuals are obtained from the candidate parent population to form a new parent population.

[0184] By predicting the fitness function values ​​of individuals in the offspring population, potential individuals can be selected more accurately for subsequent crossover and mutation operations, thereby accelerating the convergence speed of the algorithm.

[0185] The offspring population with the predicted fitness function value is merged with the original parent population to form a larger candidate parent population. Merging the population can increase the diversity of the candidate parent population, helping the algorithm avoid premature convergence to the local optimal solution and maintain the ability to search for the global optimal solution.

[0186] The above example is merely an illustrative explanation. For performing multi-objective optimization on the parent population based on the fitness function values ​​of the individuals in the parent population to obtain a new parent population, other implementation methods may also be used, and this is not limited.

[0187] S603: Continue multi-objective optimization on the new parent population until the maximum number of iterations is reached.

[0188] Exemplarily, the new parent population is used as the parent population of the next iteration cycle, and the aforementioned method of optimizing the parent population is used to perform multi-objective optimization on the parent population to obtain the parent population of the next iteration cycle, and so on, until the maximum number of iterations is reached.

[0189] The maximum number of iterations is the total number of generations that the genetic algorithm runs and is one of the common termination conditions in genetic algorithms. Setting the maximum number of iterations helps prevent the genetic algorithm from looping infinitely and ensures that the optimal solution set is obtained within a limited time. The maximum number of iterations can be reasonably set and dynamically adjusted based on actual conditions, and this embodiment of the application does not impose any restrictions on this.

[0190] S604: Take the latest parent population as the optimal solution set.

[0191] When the number of iterations reaches the maximum value, the latest parent population at this time is taken as the optimal solution set.

[0192] A specific implementation method is to obtain the optimal solution set of the embodiment of the present application through the above method. Figure 7 Schematic diagram of the Pareto optimal solution set obtained through multi-objective optimization provided in the embodiment of this application. Figure 7 As shown, the horizontal axis is the negative value of the convective heat transfer coefficient, and the vertical axis is the unit cell pressure drop, that is, the positive value of the fluid flow pressure drop. The Pareto preface of multi-objective optimization refers to the optimal solution set. Figure 7 The asterisks in the middle represent the solutions in the optimal solution set, and the breadth of the distribution of asterisks reflects the diversity of solutions in the optimal solution set.

[0193] As shown above, after multiple generations of iterative optimization, the performance of individuals in the latest parent population in the target space has been significantly improved, and they are able to better approach the optimal solution to the problem while maintaining the diversity of the optimal solution set.

[0194] The method provided in this application combines a neural network with a genetic algorithm. Leveraging the genetic algorithm's global search capabilities, it can effectively find the globally optimal solution set for the design of a heat dissipation model for a minimal surface unit cell, thus avoiding the problem of neural networks being trapped in local optimal solutions in this design. Furthermore, this method eliminates the need to express the relationship between optimization parameters and optimization objectives. Instead, the method predefines the variables and ranges to be optimized for the heat dissipation model of the minimal surface unit cell. This allows the neural network to construct a fitness prediction model, reducing computational complexity and the difficulty of parameter adjustment.

[0195] Determining the optimal solution set based on fitness function values, non-dominated sorting results, crowding, and multi-generation iterative calculations ensures good diversity in the optimal solution set. This method significantly improves the efficiency of minimal surface unit cell structure prediction, reduces computational costs, and enhances the adaptability, flexibility, and scalability of the prediction of heat dissipation models for minimal surface unit cells.

[0196] The following is an example of a method for determining a heat dissipation model of flow heat transfer performance from an optimal solution set using the TOPSIS method.

[0197] Figure 8 A flow chart of a method for determining a heat dissipation model with optimal flow heat transfer performance from an optimal solution set provided in an embodiment of the present application is shown as follows: Figure 8 As shown, the method includes the following steps:

[0198] S801. Preprocess the second parameter values ​​of individuals in the optimal solution set.

[0199] Exemplarily, matrix forwarding and standardization processing is performed on the second parameter values ​​of the individuals in the optimal solution set.

[0200] Among them, matrix forward processing is to convert the second parameter into an indicator in the same direction for comparison. Standardization processing is to standardize the value of the second parameter to eliminate the dimensional difference between different parameters. For example, standardization processing can be performed by any method such as minimum-maximum standardization, standard score (Z-score) standardization, etc., and the embodiment of the present application is not limited by the method of standardization processing.

[0201] It should be noted that the embodiment of the present application does not limit the method of preprocessing the value of the second parameter.

[0202] S802: Calculate the entropy weight of the individuals in the optimal solution set based on the pre-processed second parameter values ​​of the individuals in the optimal solution set.

[0203] Among them, entropy weight is a method used to determine the weight of each indicator, which reflects its importance in decision-making by calculating the information entropy of each second parameter.

[0204] Exemplarily, the entropy value of the second parameter of the individual in the optimal solution set is calculated to evaluate the distance between the individual in the optimal solution set and the positive and negative ideal individuals.

[0205] Among them, the positive ideal individual refers to the individual that achieves the maximum value (for maximization) or the minimum value (for minimization) on all objectives, that is, the theoretical optimal solution. The negative ideal individual refers to the individual that achieves the minimum value (for maximization) or the maximum value (for minimization) on all objectives, that is, the theoretical worst solution.

[0206] The smaller the entropy value, the closer the individual is to the positive ideal individual in the target space and the further away from the negative ideal individual, that is, the better the individual performs in the multi-objective optimization problem. The smaller the entropy value, the closer the individual is to the positive ideal individual in the target space and the further away from the negative ideal individual, that is, the better the individual performs in the multi-objective optimization problem.

[0207] S803. Sort the individuals in the optimal solution set based on the entropy weights of the individuals in the optimal solution set.

[0208] For example, the calculated entropy values ​​are used to sort the individuals in the optimal solution set. Individuals with smaller entropy values, i.e., those closest to the positive ideal individual, are considered to have performed better in the multi-objective optimization problem. Therefore, they are ranked higher in the sorting.

[0209] S804: Determine a heat dissipation model with optimal flow and heat transfer performance based on the sorting results.

[0210] Exemplarily, based on the ranking results, the individual ranked first in the ranking is selected as the heat dissipation model with the optimal flow heat transfer performance.

[0211] A specific implementation method, Figure 9 A schematic diagram of the results of using the TOPSIS method to sort individuals in the optimal solution set provided in an embodiment of the present application.

[0212] like Figure 9 As shown in the figure, based on the above method, the optimal inlet velocity after sorting, that is, the fluid inlet flow rate is 0.34m / s, and the unit cell solid volume fraction is 19%. At this time, the convective heat transfer coefficient can reach 20401W / m2·K, and the unit cell pressure drop, that is, the fluid flow pressure drop is 633Pa.

[0213] The method provided in the embodiment of the present application uses the TOPSIS method to determine the heat dissipation model with optimal flow heat transfer performance from the optimal solution set. The method calculates the entropy weight and evaluates the distance between each individual in the optimal solution set and the positive and negative ideal individuals to identify and select the solution that achieves the best recommendation among multiple objectives, thereby optimizing the decision-making process and improving the decision quality.

[0214] The following is a specific example to illustrate how to determine the prediction method of the heat dissipation model for optimal flow heat transfer performance. Figure 10 A flow chart of a prediction method for a heat dissipation model for determining optimal flow heat transfer performance provided in an embodiment of the present application. Figure 10 As shown, the method includes the following steps:

[0215] S1001. Obtain a first parameter and its value range for a heat dissipation model of a minimal surface unit cell structure. Perform a fluid simulation of the minimal surface unit cell structure based on the value of the first parameter, obtain a fitness function value under the first parameter, and obtain at least one set of sample data based on the first parameter and the fitness function value. Based on the at least one set of sample data, obtain a sample dataset for training a neural network model.

[0216] The first parameter is the unit cell solid volume fraction and the fluid inlet flow rate. The fitness function value is the negative of the convective heat transfer coefficient and the fluid flow pressure drop.

[0217] S1002. Based on the sample data set, the neural network model is trained to obtain a fitness prediction model.

[0218] S1003: Obtain initial parameters of the genetic algorithm.

[0219] Among them, the initial parameters include population size, maximum number of iterations, crossover probability, mutation probability, etc.

[0220] S1004: Generate an initial population.

[0221] Each individual in the initial population represents a heat dissipation model of a minimal surface unit cell structure. Different heat dissipation models have different values ​​for at least one first parameter.

[0222] S1005. Use the fitness prediction model to predict the fitness function value of the individuals in the parent population.

[0223] In the first iteration, the parent population is the initial population. Starting from the second iteration, the parent population is the new parent population obtained by multi-objective optimization in the previous iteration.

[0224] S1006. Based on the fitness function values ​​of the individuals in the parent population, perform non-dominated sorting and crowding calculation on the parent population.

[0225] S1007. Based on the fitness function value, non-dominated sorting result and crowding calculation result of the parent population, individual selection, individual crossover and individual mutation are sequentially performed on the parent population to obtain a child population.

[0226] S1008. Use the fitness prediction model to predict the fitness function values ​​of individuals in the offspring population.

[0227] S1009: Merge the parent population and the offspring population to obtain a candidate parent population.

[0228] S1010. Obtain a preset number of individuals from the candidate parent population based on the non-dominated sorting result and the crowding calculation result of the candidate parent population to form a new parent population.

[0229] S1011. Determine whether the maximum number of iterations has been reached.

[0230] If not, execute S1012. If yes, execute S1013.

[0231] S1012, increase the number of iterations by 1, and return to execute S1005.

[0232] S1013. Take the latest parent population as the optimal solution set.

[0233] S1014. Preprocess the second parameter values ​​of the individuals in the optimal solution set.

[0234] S1015. Calculate the entropy weight of the individuals in the optimal solution set based on the preprocessed second parameter values ​​of the individuals in the optimal solution set.

[0235] S1016. Sort the individuals in the optimal solution set based on the entropy weights of the individuals in the optimal solution set.

[0236] S1017. Based on the ranking results, determine a heat dissipation model with optimal flow and heat transfer performance.

[0237] The method provided in this application combines a neural network with a genetic algorithm. Leveraging the genetic algorithm's global search capabilities, it can effectively find the globally optimal solution set for the design of a heat dissipation model for a minimal surface unit cell, thus avoiding the problem of neural networks being trapped in local optimal solutions in this design. Furthermore, this method eliminates the need to express the relationship between optimization parameters and optimization objectives. Instead, the method predefines the variables and ranges to be optimized for the heat dissipation model of the minimal surface unit cell. This allows the neural network to construct a fitness prediction model, reducing computational complexity and the difficulty of parameter adjustment.

[0238] The TOPSIS method is used to determine the heat dissipation model of the optimal minimal surface unit cell structure from the optimal solution set. This method calculates the entropy weight and evaluates the distance between each individual in the optimal solution set and the positive and negative ideal individuals to identify and select the solution that achieves the best recommendation among multiple objectives, thereby optimizing the decision-making process and improving the decision quality.

[0239] The method provided in the embodiment of the present application can obtain the best first parameter combination under multi-objective optimization to obtain the minimal surface unit cell structure with optimal flow and heat transfer performance, and then combine the optimal minimal surface unit cell structures to form an optimal minimal surface multi-cell structure. The combination form of the optimal minimal surface multi-cell structure can be, for example, a uniform combination type, a hybrid combination type, a local gradient type, and other combination forms. The embodiment of the present application does not limit the combination form of the optimal minimal surface multi-cell structure.

[0240] Among them, the uniform combination type refers to a multi-cell type in which minimal surface unit cells are evenly distributed within the minimal surface structure. Each minimal surface unit cell structure. The hybrid combination type is a combination of different types of minimal surface unit cells, combining the advantages of various types of minimal surface unit cells to achieve better overall performance. The local gradient structure refers to a minimal surface structure in which the size, shape, or arrangement of the minimal surface unit cells gradually changes according to specific design requirements.

[0241] The method provided in the embodiments of the present application can enable the radiator to achieve optimal flow heat exchange performance and minimum pressure drop under different working conditions.

[0242] The method provided in the embodiment of the present application can also help thermal design engineers and researchers to quickly and accurately design a heat dissipation model of a minimal surface unit cell structure that meets the corresponding flow and heat transfer performance requirements, thereby improving the design efficiency of the minimal surface unit cell structure while ensuring accuracy.

[0243] As shown above, the heat dissipation model of the minimal surface unit cell structure predicted by the method of the embodiment of the present application can be used to construct a fluid radiator, which can be used, for example, in the fields of lithium-ion battery heat dissipation, chip / semiconductor heat dissipation, microchannel heat dissipation technology, electronic equipment heat dissipation technology, etc.

[0244] Figure 11 This is an exploded diagram of a chip and overall heat dissipation structure provided in an embodiment of the present application. Figure 11 As shown, the chip and the overall heat dissipation structure include a chip 1, a DBC insulation layer and a primary / secondary welding layer 2, a copper base plate 3, a minimal curved surface structure 4, a fluid domain 5, and an aluminum alloy housing 6 including the minimal curved surface structure.

[0245] Among them, the chip is the core component of the entire structure, responsible for performing computing and processing tasks. In the DBC insulation layer and the primary / secondary welding layer 2, the direct bonded copper (DBC) insulation layer is a special material that combines the high thermal conductivity of ceramics and the high electrical conductivity of copper. This structure is generally used to effectively conduct the heat generated by the chip. The primary / secondary welding layer refers to the welding material used to connect the chip and other heat dissipation components (such as the DBC insulation layer) together. These welding layers ensure close contact and efficient heat conduction between components.

[0246] The copper base plate 3 is a large-area heat dissipation component, usually made of copper material with high thermal conductivity. It is located below the DBC insulation layer and is used to further conduct and disperse the heat generated by the chip.

[0247] The minimal curved surface structure 4 improves the heat dissipation efficiency by increasing the heat dissipation area and optimizing the heat conduction path. The minimal curved surface structure 4 used can be predicted using the above method. Figure 11 In the present invention, the minimal surface structure can be used, for example, on heat dissipation components such as copper base plates or aluminum alloy housings.

[0248] The fluid domain 5 is an area where the fluid (such as liquid or gas) used for cooling flows. Figure 11 In the process, the fluid domain can be located under the copper base plate or around the minimal surface structure, and the heat is taken away by the flow of the fluid, thereby achieving heat dissipation.

[0249] The aluminum alloy housing 6, which includes a minimally curved surface structure, protects the entire heat dissipation structure and the chip. Aluminum alloy is characterized by high thermal conductivity and lightweight. The minimally curved surface structure further enhances heat dissipation performance, increasing heat dissipation area and optimizing the heat conduction path to improve heat dissipation efficiency.

[0250] Figure 11 The chip and overall heat dissipation structure is a complex system that integrates multiple high-efficiency heat dissipation technologies. By combining the DBC insulation layer with the primary and secondary solder layers 2, the copper base plate 3, the minimally shaped surface structure 4, the fluid domain 5, and the aluminum alloy housing containing the minimally shaped surface structure 6, the system effectively dissipates heat generated by the chip, ensuring stable operation.

[0251] Figure 12 A schematic diagram of a device for predicting a heat dissipation model of a minimal surface unit cell structure provided in an embodiment of the present application is shown in FIG. Figure 12 As shown, the apparatus includes: a first acquisition module 1201, a first processing module 1202, and a determination module 1203. Optionally, the apparatus may further include, for example, a second acquisition module 1204 and a second processing module 1205.

[0252] The first acquisition module 1201 acquires an initial population, where each individual in the initial population represents a heat dissipation model of a minimal surface unit cell structure; different heat dissipation models have different values ​​of at least one first parameter, and the first parameters include: structural parameters of the minimal surface unit cell structure, and parameters related to the boundary conditions of the fluid flowing through the minimal surface unit cell structure.

[0253] First processing module 1202 is configured to iteratively solve individuals in the initial population using a genetic algorithm to obtain an optimal solution set. During the iterative solution process, a fitness prediction model constructed using a neural network is used to predict the fitness function values ​​of the individuals in the initial population based on the parameter values ​​corresponding to the individuals. The fitness function value includes at least two second parameters, each of which represents the flow and heat transfer performance of the heat dissipation model.

[0254] The determination module 1203 is used to determine the heat dissipation model with the optimal flow and heat transfer performance from the optimal solution set.

[0255] Optionally, the first processing module 1202 is specifically configured to use the initial population as the parent population and, based on the parameter values ​​corresponding to the individuals in the parent population, use a fitness prediction model to predict the fitness function values ​​of the individuals in the parent population. Based on the fitness function values ​​of the individuals in the parent population, a multi-objective optimization is performed on the parent population to obtain a new parent population. The multi-objective optimization is continued on the new parent population until the maximum number of iterations is reached. The latest parent population is then used as the optimal solution set.

[0256] In this implementation, the first processing module 1202 is specifically configured to perform non-dominated sorting and crowding calculation on the parent population based on the fitness function values ​​of the individuals in the parent population. Based on the non-dominated sorting results and crowding calculation results, the parent population is sequentially subjected to individual selection, individual crossover, and individual mutation to obtain a child population. A fitness prediction model is used to predict the fitness function values ​​of the individuals in the child population. The parent population and the child population are merged to obtain a candidate parent population. Based on the fitness function values ​​of the individuals in the candidate parent population, a non-dominated sorting and crowding calculation is performed on the candidate parent population. Based on the non-dominated sorting results and crowding calculation results of the candidate parent population, a preset number of individuals are obtained from the candidate parent population to form a new parent population.

[0257] Optionally, determination module 1203 is specifically configured to preprocess the second parameter values ​​of the individuals in the optimal solution set. Based on the preprocessed second parameter values ​​of the individuals in the optimal solution set, entropy weights of the individuals in the optimal solution set are calculated. Based on the entropy weights of the individuals in the optimal solution set, the individuals in the optimal solution set are ranked. Based on the ranking results, a heat dissipation model with optimal flow and heat transfer performance is determined.

[0258] In this implementation, the determination module 1203 is specifically configured to perform matrix forwarding and standardization processing on the second parameter values ​​of the individuals in the optimal solution set.

[0259] Optionally, the first acquisition module 1201 is specifically configured to randomly generate multiple individuals based on the population size and the first parameter of the heat dissipation model of the minimal surface unit cell structure and its value range. The first parameter of each individual is represented by a binary code. The binary code of the individual is converted into a decimal floating point number to obtain an initial population.

[0260] Optionally, the second acquisition module 1204 is configured to acquire a sample data set, the sample data set including: at least one set of sample data, each set of sample data including: a mapping relationship between a first parameter of a heat dissipation model of a minimal surface unit cell structure and a fitness function value.

[0261] Optionally, the second processing module 1205 is used to train the neural network model based on the sample data set to obtain a fitness prediction model.

[0262] In this implementation, the second acquisition module 1204 is specifically configured to obtain a first parameter of a heat dissipation model of a minimal surface unit cell structure and its value range. Based on the first parameter and its value range, a value of the first parameter of the heat dissipation model of at least one minimal surface unit cell structure is determined. Based on the first parameter value of the heat dissipation model of at least one minimal surface unit cell structure, a fluid simulation of the heat dissipation model of the minimal surface unit cell structure is performed to obtain at least one set of sample data.

[0263] The prediction device for the heat dissipation model of the minimal surface unit cell structure provided in the embodiment of the present application can be used to execute the method for predicting the heat dissipation model of the minimal surface unit cell structure as described above. Its implementation principle, process and beneficial effects can be found in the above embodiment and will not be repeated here.

[0264] Figure 13 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 13 As shown, the electronic device 1300 may include: a memory 1301 and a processor 1302. Optionally, the electronic device may further include a transceiver 1303, wherein the memory 1301 and the processor 1302 communicate with each other; illustratively, the memory 1301, the processor 1302, and the transceiver 1303 may communicate via a communication bus 1306, the memory 1301 is used to store a computer program, and the processor 1302 executes the computer program to implement the method of the above embodiment.

[0265] Optionally, the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), or application-specific integrated circuits (ASIC). A general-purpose processor may be a microprocessor or any conventional processor. The steps in the method embodiments disclosed in this application may be directly implemented by a hardware processor or implemented by a combination of hardware and software modules in the processor.

[0266] An embodiment of the present application further provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, a method in any of the above method embodiments is implemented.

[0267] An embodiment of the present application further provides a computer program product, including a computer program, which implements the method in any of the above method embodiments when the computer program is executed by a processor.

[0268] All or part of the steps of the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a readable memory. When the program is executed, it performs the steps of the above-mentioned method embodiments; and the aforementioned memory (storage medium) includes: read-only memory (ROM), RAM, flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disc, and any combination thereof.

[0269] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processing unit of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0270] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0271] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0272] Obviously, those skilled in the art may make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include such modifications and variations.

[0273] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of the present application and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0274] It should be understood that the present application is not limited to the exact structure described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for predicting a heat dissipation model of a minimal surface unit cell structure, characterized in that: The method comprises: An initial population is obtained, where each individual in the initial population represents a heat dissipation model of a minimal surface unit cell structure; different heat dissipation models have different values ​​of at least one first parameter, the first parameter including: a structural parameter of the minimal surface unit cell structure and a parameter related to a boundary condition of a fluid flowing through the minimal surface unit cell structure; Based on the initial population, a genetic algorithm is used to iteratively solve the individuals in the initial population to obtain an optimal solution set; during the iterative solution process, based on the parameter values ​​corresponding to the individuals, a fitness prediction model constructed based on a neural network is used to predict the fitness function values ​​of the individuals in the initial population; the fitness function values ​​include: at least two second parameters, the second parameters being representative of the flow heat transfer performance of the heat dissipation model; A heat dissipation model that determines the optimal flow and heat transfer performance from a set of optimal solutions.

2. The method according to claim 1, characterized in that The step of iteratively solving the individuals in the initial population using the genetic algorithm to obtain an optimal solution set includes: Taking the initial population as the parent population, and based on the parameter values ​​corresponding to the individuals in the parent population, using the fitness prediction model, predicting the fitness function values ​​of the individuals in the parent population; Based on the fitness function values ​​of the individuals in the parent population, multi-objective optimization is performed on the parent population to obtain a new parent population; Continue to perform multi-objective optimization on the new parent population until the maximum number of iterations is reached; The latest parent population is taken as the optimal solution set.

3. The method according to claim 2, characterized in that The step of performing multi-objective optimization on the parent population based on the fitness function values ​​of the individuals in the parent population to obtain a new parent population includes: Based on the fitness function values ​​of the individuals in the parent population, performing non-dominated sorting and crowding calculation on the parent population; Based on the non-dominated sorting result and the crowding calculation result, the parent population is sequentially subjected to individual selection, individual crossover and individual mutation to obtain a child population; Using the fitness prediction model, predicting the fitness function values ​​of individuals in the offspring population; Merging the parent population with the offspring population to obtain a candidate parent population; Based on the fitness function values ​​of the individuals in the candidate parent population, performing non-dominated sorting and crowding calculation on the candidate parent population; A preset number of individuals are obtained from the candidate parent population according to the non-dominated sorting result and the crowding calculation result of the candidate parent population to form a new parent population.

4. The method according to claim 1, wherein The heat dissipation model for determining the optimal flow heat transfer performance from the optimal solution set includes: Preprocessing the second parameter values ​​of the individuals in the optimal solution set; Calculating the entropy weight of the individuals in the optimal solution set based on the preprocessed second parameter values ​​of the individuals in the optimal solution set; sorting the individuals in the optimal solution set based on the entropy weights of the individuals in the optimal solution set; Based on the ranking result, a heat dissipation model with the optimal flow and heat transfer performance is determined.

5. The method according to claim 4, characterized in that The preprocessing of the second parameter values ​​of the individuals in the optimal solution set includes: The second parameter values ​​of the individuals in the optimal solution set are subjected to matrix forwarding and standardization processing.

6. The method according to claim 1, characterized in that The obtaining of the initial population includes: Based on the population size and the first parameter of the heat dissipation model of the minimal surface unit cell structure and its value range, multiple individuals are randomly generated; the first parameter of each individual is represented by binary code; The binary code of the individual is converted into a decimal floating point number to obtain the initial population.

7. The method according to claim 1, characterized in that The method further comprises: Acquire a sample data set, the sample data set comprising: at least one set of sample data, each set of sample data comprising: a mapping relationship between a first parameter of a heat dissipation model of a minimal surface unit cell structure and a fitness function value; Based on the sample data set, the neural network model is trained to obtain the fitness prediction model.

8. The method according to claim 7, characterized in that The obtaining of the sample data set includes: Obtaining a first parameter of a heat dissipation model of the minimal surface unit cell structure and a value range thereof; Determining a first parameter value of a heat dissipation model of at least one minimal surface unit cell structure based on the first parameter and its value range; Based on the first parameter value of the heat dissipation model of at least one minimal surface unit cell structure, a fluid simulation of the heat dissipation model of the minimal surface unit cell structure is performed to obtain at least one set of sample data.

9. A device for predicting the heat dissipation model of a minimal surface unit cell structure, characterized in that: The device comprises: A first acquisition module is configured to acquire an initial population, wherein each individual in the initial population represents a heat dissipation model of a minimal surface unit cell structure; different heat dissipation models have different values ​​of at least one first parameter, the first parameter including: a structural parameter of the minimal surface unit cell structure, and a parameter related to a boundary condition of a fluid flowing through the minimal surface unit cell structure; A first processing module is configured to iteratively solve individuals in the initial population using a genetic algorithm based on the initial population to obtain an optimal solution set; during the iterative solution process, based on parameter values ​​corresponding to the individuals, a fitness prediction model constructed based on a neural network is used to predict fitness function values ​​of the individuals in the initial population; the fitness function value includes at least two second parameters, each of which is a parameter representing a flow heat transfer performance index of the heat dissipation model; The determination module is used to determine the heat dissipation model with the optimal flow and heat transfer performance from the optimal solution set.

10. An electronic device, characterized in that: The electronic device includes: a processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 8 when executed by a processor.

12. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 8 when executed by a processor.

Images