Double-flow-channel heat exchanger design method and system based on implicit minimum curved surface optimization

Abstract

The invention provides a double-flow-channel heat exchanger design method and system based on implicit minimum curved surface optimization, and the method comprises the steps: dividing a whole heat exchanger design space, obtaining an undirected graph, and carrying out the weighting of each edge in the undirected graph according to the direction of the edge and the direction of a flow field; dividing the undirected graph into two connected sub-graphs by taking maximization of the sum of weights of edges removed in the division process as a target, and generating a double-channel skeleton; constructing a scalar field according to the double-channel skeleton, and approximating the scalar field by using a neural network model; and carrying out sampling and data enhancement on the double-flow-channel skeleton, training the neural network model by using the sampled and data-enhanced data, and carrying out joint optimization of maximizing the heat exchange surface area and minimizing the pressure loss to generate the extremely-small-curved-surface heat exchanger structure conforming to the skeleton constraint. According to the method, the flow topology is determined through constrained maximum cut optimization, generation of the minimum curved surface is optimized through neural network implicit field modeling, and the optimal balance between the heat exchange efficiency and the pressure loss is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of optimal design of heat exchangers, and particularly relates to a design method and system for a two-channel heat exchanger based on implicit minimal surface optimization. Background Art

[0002] The statements in this part merely provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] Heat exchangers are widely used in fields such as aerospace, automotive industry, chemical energy, etc. Their main goal is to reduce flow resistance while ensuring heat transfer efficiency. However, traditional heat exchanger structures (such as plate-fin type and shell-and-tube type) are restricted by geometric structures when optimizing the balance between heat transfer efficiency and pressure loss.

[0004] In recent years, the three-period minimal surface (TPMS) structure has attracted attention due to its high specific surface area and uniform fluid channel design. However, the topological morphology of the TPMS structure is constrained by mathematical equations, making it difficult to adapt to complex boundary shapes and unable to actively guide the fluid flow direction, which may lead to flow stagnation and large local pressure losses. Summary of the Invention

[0005] In order to solve the above problems, the present invention proposes a design method and system for a two-channel heat exchanger based on implicit minimal surface optimization. The present invention uses constrained maximum cut optimization to determine the flow topology and neural network implicit field modeling to optimize the generation of minimal surfaces, achieving the optimal balance between heat exchange efficiency and pressure loss.

[0006] According to some embodiments, the present invention adopts the following technical solutions:

[0007] A design method for a two-channel heat exchanger based on implicit minimal surface optimization, comprising the following steps:

[0008] Divide the entire heat exchanger design space to obtain an undirected graph, and for each edge in the undirected graph, weight it according to the direction of the edge and the flow field direction;

[0009] Taking the maximization of the sum of the weights of the edges removed during the division process as the goal, divide the undirected graph into two connected subgraphs to generate a two-channel skeleton;

[0010] According to the two-channel skeleton, construct a scalar field and approximate the scalar field using a neural network model;

[0011] Sample and perform data augmentation on the two-channel skeleton, and use the sampled and data-augmented data to train the neural network model for joint optimization of maximizing the heat transfer surface area and minimizing the pressure loss, generating a minimal surface heat exchanger structure that conforms to the skeleton constraints.

[0012] As an alternative implementation, the process of dividing the entire heat exchanger design space to obtain an undirected graph includes: dividing the entire heat exchanger design space using the centroid Voronoi partitioning method to obtain an undirected graph G(V, E), where the point set V is the seed points and the edge set E is generated by Delaunay triangulation.

[0013] As an alternative implementation, for each edge in the undirected graph, the process of weighting according to the direction of the edge and the flow field direction includes: according to the direction d uv of the edge and the flow field direction f uv the included angle θ uv is weighted as follows:

[0014]

[0015] where a > 1 is a penalty factor, and by weighting, it encourages the skeleton to maintain edges aligned with the flow, guiding the fluid channels to flow along the flow direction.

[0016] As an alternative implementation, with the goal of maximizing the sum of the weights of the edges removed during the partitioning process, when dividing the undirected graph into two connected subgraphs, constraints are added, that is, the degree of each vertex in the two connected subgraphs is greater than a set value to prevent dead-end paths in the flow channels.

[0017] As an alternative implementation, with the goal of maximizing the sum of the weights of the edges removed during the partitioning process, the process of dividing the undirected graph into two connected subgraphs includes: first dividing the undirected graph into two initial connected subgraphs that meet the constraints, then traversing each vertex, considering whether it satisfies the constraints when moving to the opposite subgraph. If it satisfies the constraints, record the change in the objective function, otherwise continue to traverse the next vertex. Finally, select the vertex that makes the objective function increase the most, swap it to the opposite subgraph, and repeat the above process until the objective function converges to obtain the final partitioned subgraphs.

[0018] As an alternative implementation, according to the double-channel skeleton, the process of constructing a scalar field and approximating the scalar field using a neural network model includes: constructing a scalar field f: R 3 →R scalar field, where the sign of f(x) determines the classification of any point x belonging to one of the two regions. A positive value of f(x) indicates that x belongs to region A, and a negative value indicates region B; the zero level set Ω0 = {x ∈ R 3 |f(x) = 0} forms the decision boundary separating the regions and is defined as a minimal surface;

[0019] The neural network model is a neural network model combining high-dimensional feature mapping and a multi-layer perceptron.

[0020] As an alternative embodiment, the process of sampling and data augmentation for the double-channel skeleton includes: sampling on two skeletons, introducing Gaussian perturbations to the sampling points of each skeleton, and taking the sum of the sampling points and the perturbation parameters as the perturbed points, which are included in the sampling range.

[0021] As an alternative embodiment, the process of training the neural network model using the sampled and data-augmented data and performing the joint optimization of maximizing the heat transfer surface area and minimizing the pressure loss includes: setting the field value corresponding to the points sampled from the first skeleton to +1, setting the field value corresponding to the points sampled from the second skeleton to -1, and constructing the first loss function;

[0022] Introducing a regularization term to penalize the total variation of the scalar field f(x) to form the second loss function;

[0023] Based on the first loss function, the second loss function, and hyperparameters, forming the final total loss function, training the neural network model, and obtaining the optimization result.

[0024] A double-channel heat exchanger design system based on implicit minimal surface optimization includes:

[0025] A partitioning module configured to partition the entire heat exchanger design space to obtain an undirected graph, and for each edge in the undirected graph, weighting it according to the direction of the edge and the flow field direction;

[0026] A skeleton generation module configured to partition the undirected graph into two connected subgraphs with the goal of maximizing the sum of the weights of the edges removed during the partitioning process, and generate a double-channel skeleton;

[0027] A model construction module configured to construct a scalar field based on the double-channel skeleton and approximate the scalar field using a neural network model;

[0028] An optimization and solution module configured to sample and perform data augmentation on the double-channel skeleton, train the neural network model using the sampled and data-augmented data, perform the joint optimization of maximizing the heat transfer surface area and minimizing the pressure loss, and generate a minimal surface heat exchanger structure that conforms to the skeleton constraints.

[0029] An electronic device includes a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the steps in the above method are completed.

[0030] Compared with the prior art, the beneficial effects of the present invention are:

[0031] The present invention proposes a heat exchanger design method based on implicit double-channel minimal surface optimization. After considering the free geometric design space and the expected flow field of the user, the topological skeleton of the double-channel heat exchanger is first generated, and then the implicit minimal surface modeling optimization is carried out according to the skeleton optimization to generate the minimal surface to obtain the final form of the heat exchanger.

[0032] The present invention breaks through the regularity limitation of TPMS, enabling the heat exchanger structure to adapt to complex free boundary shapes; through the joint optimization of maximizing the heat transfer surface area and minimizing the pressure loss, the overall heat transfer performance is improved; the implicit minimal surface modeling is introduced, eliminating the need to model at a fixed resolution and automatically learning the optimal minimal surface form.

[0033] The present invention provides greater flexibility in the surface topological structure, and at the same material cost, maintains a comparable heat exchange efficiency and achieves a lower pressure drop, demonstrating superior heat exchange performance.

[0034] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following specific preferred embodiments are given, in conjunction with the accompanying drawings, and are described in detail as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The specification drawings forming a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0036] Figure 1 It is a schematic diagram of the overall process of this method, where a represents the initially input flow field and the inlet and outlet positions of the double channels, b represents the original skeleton G(V, E) generated by CVT, c represents the divided connected subgraphs G1(V1, E1), G2(V2, E2), d represents the generated minimal surface heat exchanger structure, and e represents the forms of the two channels divided from the heat exchanger structure;

[0037] Figure 2 It is a schematic diagram of the separation surface and double channels of heat exchangers with different shapes generated by this method;

[0038] Figure 3 It is a result diagram of the temperature difference and pressure drop at different volume fractions when comparing the results of this method with the Gyroid filling under the U-shaped tube shape;

[0039] Figure 4 It is a result diagram of the fluid flow velocity when comparing the result of this method (b therein) with the Gyroid filling result under the U-shaped tube shape;

[0040] Figure 5 It is a result diagram of the temperature difference and pressure drop when comparing the result of this method (b therein) with the Gyroid filling (a therein) under the zigzag shape;

[0041] Figure 6 It is a result graph of the temperature difference pressure drop for comparing the result of this method (b therein) with the result of multi-bend shape filling (a therein) under the zigzag shape;

[0042] Figure 7 It is a result graph of the changes in the skeleton, structure, and field values of this method using Gaussian perturbations with different standard deviations (σ = 0, σ = 0.05, σ = 0.08) for data augmentation;

[0043] Figure 8 It shows the growth of the objective function value of this method with the increase in the number of iterations under different numbers of vertices, proving its convergence;

[0044] Figure 9 It is the neural network architecture diagram designed by this method;

[0045] Figure 10 It is the process schematic diagram of this method. Specific implementation manners

[0046] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0047] It should be noted that the following detailed descriptions are all illustrative and are intended to provide further explanations of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0048] It should be noted that the terms used herein are only for describing specific implementation manners and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0049] Without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

[0050] Embodiment 1

[0051] A heat exchanger design method based on implicit double-channel minimal surface optimization, as Figure 10 shown, includes the following steps:

[0052] Step (1): Generate the overall initial design skeleton and divide it to generate a double-channel skeleton, that is, the process of optimizing the double-channel fluid skeleton.

[0053] Step (2): Apply the dual-channel design framework to the implicit field for minimal surface optimization, i.e., the process of minimal surface optimization.

[0054] Next, each step in this method will be elaborated specifically:

[0055] The said step (1) mainly includes the following steps:

[0056] The goal of this step is to generate and optimize the framework of the fluid channels, making the internal flow channels of the heat exchanger evenly distributed and improving the heat exchange efficiency.

[0057] Step (1-1) Fluid framework initialization:

[0058] To generate fluid channels within a complex free-form domain, first, the entire heat exchanger design space Ω is partitioned using the centroidal Voronoi tessellation (CVT) method to obtain an undirected graph G(V, E), where the point set V is the seed points of the CVT and the edge set E is generated through Delaunay triangulation.

[0059] For each edge (u, v) ∈ E, according to the direction d of the edge uv and the flow field direction f uv the included angle θ uv is weighted as follows:

[0060]

[0061] where a > 1 is a penalty factor, which is empirically set to 5 in our method. This weighting scheme encourages the framework to maintain edges aligned with the flow, guiding the fluid channels to flow along the flow direction.

[0062] Step (1-2) Constrained maximum cut optimization:

[0063] Partition G(V, E) into two connected subgraphs G1(V1, E1) and G2(V2, E2), which represent the topologies of the two flow channels of the heat exchanger. To maximize the contact area between the two flow channels, this method hopes to make G1 and G2 wind around each other as much as possible, i.e., maximize the sum of the weights of the edges removed during the partitioning process. This is a variant of the connected maximum cut problem, and this method extends the classical problem by adding a constraint: ensuring that the degree of each vertex in G1 and G2 is greater than 1, thus preventing dead-end paths in the flow channels. Therefore, this method designs the following heuristic objective function:

[0064]

[0065] where w uvDenote the weight of the edge between node u and node v in the graph. This function represents that the goal of this method is to maximize the sum of the weights of the cut edges.

[0066] Next, it will be specifically shown how to partition the graph G(V, E) into two connected subgraphs G1(V1, E1) and G2(V2, E2) according to the heuristic objective function:

[0067] First, partition G(V, E) into two initial connected subgraphs G1(V1, E1) and G2(V2, E2) that meet the constraints. Then, traverse each vertex and consider whether moving it to the opposite subgraph satisfies constraints such as connectivity and the degree of all vertices being greater than 1. If the constraints are satisfied, record the change in the objective function; otherwise, continue to traverse the next vertex. Finally, select the vertex that makes the objective function increase the most and swap it to the opposite subgraph. Repeat this process until the objective function converges. In this way, the final partitioned subgraphs G1(V1, E1) and G2(V2, E2) are obtained.

[0068] The steps in step (2) mainly include the following steps:

[0069] The goal of this step is to generate a minimal surface heat exchanger structure that conforms to the skeleton constraints of the fluid channel generated in step (1) using a neural implicit field.

[0070] Step (2-1) Construct a neural network model:

[0071] First, construct a scalar field f: R 3 →R scalar field, where the sign of f(x) determines the classification of any point x belonging to one of the two regions of the heat exchanger. A positive value of f(x) indicates that x belongs to region A, and a negative value indicates region B. The zero-level set Ω0 = {x ∈ R 3 |f(x) = 0} forms the decision boundary separating the regions, and this method defines it as a minimal surface. In this way, the minimal surface modeling optimization problem becomes the problem of finding the decision plane for a binary classification task.

[0072] This method designs a neural network that combines high-dimensional feature mapping and a multilayer perceptron (MLP) to approximate this scalar field.

[0073] As Figure 9 shown, specifically, each three-dimensional coordinate (x, y, z) is mapped into a 2048-dimensional feature vector through random Fourier transform to capture high-frequency spatial information. Then, it is processed using an MLP that contains three hidden layers, each hidden layer contains 256 neurons, and the tanh activation function is used. The final output of the network is an implicit field, and its field value represents a real scalar corresponding to the input position, and this real value represents the region to which the point belongs.

[0074] Step (2-2) performs data augmentation on the connected subgraphs G1(V1, E1) and G2(V2, E2):

[0075] Sample on the two skeletons to obtain the input of the network. Introduce Gaussian perturbations to the points sampled from each skeleton. This noise expands the area around each skeleton, allowing the neural network to generalize better by considering the local neighborhoods belonging to the same class. Specifically, given the sampled point x = (x, y, z) on the skeleton, its corresponding perturbed point is generated according to the following formula:

[0076] x ′ = x + η,

[0077] where η ~ N(0, σ 2 I). The variance σ 2 is a hyperparameter used to control the degree of perturbation.

[0078] Step (2-3) trains the neural network model:

[0079] As can be seen from step (2-2), the training data of the network comes from two skeletons. This method stipulates that the field value corresponding to the points sampled from the first skeleton is +1, and the field value corresponding to the points sampled from the second skeleton is -1. Thus, the first part of the loss function of the neural network is obtained:

[0080]

[0081] where |A| and |B| represent the number of sampled points from the two skeletons respectively.

[0082] To ensure that the resulting surface is smooth and has the minimum area, i.e., a minimal surface, we introduce a regularization term to penalize the total variation of f(x). Specifically, the regularization term is as follows:

[0083]

[0084] where Ω represents the set of sampled points in the geometric design domain. Minimizing the total variation results in a decision boundary with the minimum area because of the area formula in set measure theory.

[0085] The total loss function combines these two components as follows:

[0086] L total = L skeletin + λL sMooth ,

[0087] where λ is a hyperparameter used to balance the two loss functions.

[0088] Use the Adam optimizer to train for 5.12×10 -5 at a learning rate of 3×104 For the wheel, 32768 points are sampled in the free domain, and the standard deviation σ of the Gaussian perturbation for data augmentation is 0.02. After data augmentation of the sampled points on the double-skeleton, they are fed into the neural network to train the implicit field. Finally, the Marching Cubes algorithm is used to extract the surface mesh of the 0 isosurface of the implicit field from the neural network implicit field at a resolution of 256 3 to obtain the heat transfer surface of the heat exchanger.

[0089] The above method solves the design problem of the heat exchanger by directly optimizing the separation surface of the two fluids, achieving the best balance between maximizing the surface area and minimizing the pressure drop. The surface area of the heat exchanger is maximized through a constrained maximum cut problem to determine the flow topology of the two fluids; then, the edges of G(V, E) are weighted according to the angle between the edge and the flow field direction, and the interface minimal surface is optimized as the decision boundary of the binary classifier model to minimize the pressure loss.

[0090] Compared with TPMS, the present method provides greater flexibility in the surface topology, and at the same material cost, maintains comparable heat exchange efficiency and achieves lower pressure drop, demonstrating superior heat exchange performance.

[0091] From Figures 1 - 8 it can be seen that the present method has better fluid flow velocity and temperature difference pressure drop compared with the existing methods under various shapes.

[0092] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0093] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more flows or multiple flows and / or blocks Figure 1 one or more blocks or multiple blocks.

[0094] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction means that implements the functions specified in one or more of the processes Figure 1 or steps and / or blocks Figure 1 specified in one or more of the blocks or blocks.

[0095] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one or more of the processes Figure 1 or steps and / or blocks Figure 1 specified in one or more of the blocks or blocks.

[0096] The above is only the preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made by those skilled in the art without creative efforts within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A design method for a double-channel heat exchanger based on implicit minimal surface optimization, characterized in that: The following steps are involved: The entire heat exchanger design space is divided to obtain an undirected graph. Each edge in the undirected graph is weighted according to the direction of the edge and the direction of the flow field. With the goal of maximizing the sum of the weights of the edges removed during the partitioning process, the undirected graph is partitioned into two connected subgraphs to generate a dual-channel skeleton; According to the dual-channel skeleton, a scalar field is constructed, and the scalar field is approximated by a neural network model; The dual-channel skeleton is sampled and data enhanced, and the neural network model is trained using the sampled and data enhanced data to perform joint optimization of maximizing the heat exchange surface area and minimizing the pressure loss, thereby generating a minimal curved surface heat exchanger structure that meets the skeleton constraints.

2. A double-channel heat exchanger design method based on implicit minimal surface optimization according to claim 1, characterized in that: The process of partitioning the entire heat exchanger design space to obtain an undirected graph includes: partitioning the entire heat exchanger design space using a centroid Voronoi partitioning method to obtain an undirected graph G(V,E), wherein the point set V is a seed point and the edge set E is generated by Delaunay triangulation.

3. The double-channel heat exchanger design method based on implicit minimal surface optimization according to claim 1, characterized in that: For each edge in the undirected graph, the process of weighting according to the direction of the edge and the flow field direction includes: uv and flow field direction f uv The angle θ uv The weighting is as follows: where a>1 is a penalty factor that encourages the skeleton to stay aligned with the flow by weighting the edges, guiding the fluid channel to flow along the flow direction.

4. A double-channel heat exchanger design method based on implicit minimal surface optimization according to claim 1, characterized in that: With the goal of maximizing the sum of weights of edges removed during the partitioning process, a constraint is added in the process of partitioning the undirected graph into two connected subgraphs, that is, the degree of each vertex of the two connected subgraphs is greater than a set value to prevent dead-end paths in the flow channel.

5. The dual-channel heat exchanger design method based on implicit minimal surface optimization according to claim 1, characterized in that: With the goal of maximizing the sum of the weights of the edges removed during the partitioning process, the process of partitioning an undirected graph into two connected subgraphs includes: first, partitioning the undirected graph into two initial connected subgraphs that meet the constraints, then traversing each vertex and considering whether it satisfies the constraints when moved to the opposite subgraph. If the constraints are satisfied, record the change in the objective function, otherwise continue to traverse the next vertex, and finally select the vertex that makes the objective function grow the most, swap it to the opposite subgraph, and repeat the above process until the objective function converges to obtain the final partitioned subgraph.

6. The design method of a double-channel heat exchanger based on implicit minimal surface optimization according to claim 1, characterized in that: According to the dual-channel skeleton, a scalar field is constructed, and the process of approximating the scalar field using a neural network model includes: constructing a scalar field f:R 3 →R, where the sign of f(x) determines the classification of any point x as belonging to one of the two regions, a positive value of f(x) indicates that x belongs to region A, and a negative value indicates that it belongs to region B; the zero level set Ω0 = {x∈R 3 |f(x)=0} forms the decision boundary separating the regions, which is defined as a minimal surface; The neural network model is a neural network model that combines high-dimensional feature mapping and a multi-layer perceptron.

7. The dual-channel heat exchanger design method based on implicit minimal surface optimization according to claim 1, characterized in that: The process of sampling and data enhancement of the dual-channel skeleton includes: sampling on two skeletons, introducing Gaussian perturbation to the sampling points of each skeleton, taking the sum of the sampling point and the perturbation parameter as the perturbation point, and incorporating it into the sampling range.

8. The dual-channel heat exchanger design method based on implicit minimal surface optimization according to claim 1, characterized in that: The process of training the neural network model with the sampled and data-enhanced data to jointly optimize the maximum heat exchange surface area and the minimum pressure loss includes: the field value corresponding to the point sampled from the first skeleton is +1, the field value corresponding to the point sampled from the second skeleton is -1, and a first loss function is constructed; A regularization term is introduced to penalize the total change of the scalar field f(x), forming the second loss function; Based on the first loss function, the second loss function and the hyperparameters, a final total loss function is formed, and the neural network model is trained to obtain an optimization result.

9. A dual-channel heat exchanger design system based on implicit minimal surface optimization, characterized in that: include: A partitioning module is configured to partition the entire heat exchanger design space to obtain an undirected graph, and weight each edge in the undirected graph according to the direction of the edge and the direction of the flow field; The skeleton generation module is configured to maximize the sum of weights of edges removed during the partitioning process, partition the undirected graph into two connected subgraphs, and generate a dual-channel skeleton; A model building module is configured to construct a scalar field according to the dual-channel skeleton and approximate the scalar field using a neural network model; The optimization solution module is configured to sample and enhance the data of the dual-channel skeleton, use the sampled and enhanced data to train the neural network model, perform joint optimization of maximizing the heat exchange surface area and minimizing the pressure loss, and generate a minimal surface heat exchanger structure that meets the skeleton constraints.

10. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the steps in the method according to any one of claims 1 to 8 are completed.

Images