A data-driven cross-scale topology optimization design method and device for non-uniform heat dissipation structures
Through the data-driven cross-scale topology optimization design method, a non-uniform heat dissipation structure is generated using the conditional variational autoencoder architecture, solving the problems of high computational complexity and cost in traditional design methods, and achieving efficient and flexible heat dissipation performance optimization.
Patent Information
- Application Number
- CN202410788160.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-19
- Publication Date
- 2025-01-07
- Estimated Expiration
- 2044-06-19
AI Technical Summary
The traditional non-uniform heat dissipation structure design has problems of computational complexity and high cost, resulting in low design flexibility and efficiency, making it difficult to meet the needs of high-performance heat dissipation structures in the fields of aerospace and electronic information.
Using the data-driven cross-scale topology optimization design method, by constructing a cross-scale topology optimization model at the macro scale and establishing a microstructure reverse design model at the micro scale, using the conditional variational autoencoder architecture to generate an efficient non-uniform heat dissipation structure to achieve Inhomogeneous distribution of materials and efficient heat transfer.
It improves design flexibility and accuracy, reduces computing complexity and cost, and realizes efficient design of high-performance non-uniform heat dissipation structures, with large design space and optimized heat dissipation performance.
Smart Images
Figure CN118645189B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field related to thermal metamaterials, and more specifically, relates to a data-driven cross-scale topology optimization design method and device for non-uniform heat dissipation structures. Background Art
[0002] In practical applications in aerospace and electronic information, the design demand for high-performance heat dissipation structures is becoming increasingly urgent due to the continuous reduction in device size and the continuous increase in power density. The cross-scale design of non-uniform heat dissipation structures aims to achieve non-uniform distribution of materials in the design domain by constructing anisotropic microstructure configurations. Compared with the use of isotropic microstructures to design heat dissipation structures, anisotropic microstructures have higher design freedom and can flexibly adjust the heat conduction path to enhance the heat transfer and dispersion effect.
[0003] However, the high-dimensional design space brought by anisotropic microstructures increases the complexity and cost of calculations, resulting in the lack of flexibility and efficiency in design of traditional topology optimization methods. Summary of the invention
[0004] In response to the above defects or improvement needs of the prior art, the present invention provides a data-driven cross-scale topology optimization design method and device for non-uniform heat dissipation structure, which aims to solve the problems of low flexibility and efficiency caused by the complexity and high cost of the non-uniform heat dissipation structure design in the prior art.
[0005] To achieve the above object, according to one aspect of the present invention, a data-driven cross-scale topology optimization design method for non-uniform heat dissipation structure is provided, and the design method comprises the following steps:
[0006] Step 1: at the macroscopic scale, a cross-scale topology optimization model is constructed based on anisotropic hypothetical materials; based on the heat conduction boundary conditions and the thermal performance objective function, the heat conduction tensor distribution required inside the non-uniform heat dissipation structure to be optimized is calculated by solving the cross-scale topology optimization model;
[0007] Step two, at the microscopic scale, a microstructure data set is constructed based on two given component materials and a microstructure reverse design model is established based on data-driven technology, and then a corresponding microstructure configuration is designed based on the microstructure reverse design model and the thermal conductivity tensor distribution, so as to obtain the final non-uniform heat dissipation structure.
[0008] Furthermore, the mathematical expression of the cross-scale topology optimization model is:
[0009] find:ρ={l 1 ,l 2 ,θ}
[0010] min:Θn
[0011] st:KT=P
[0012]
[0013] θ∈[0,π)
[0014] Among them, ρ(l 1 ,l 2 ,θ) contains three design variables, l 1 , l 2 is the geometric parameter of the anisotropic hypothetical material, θ is the rotation angle, f is the volume fraction constraint of the high thermal conductivity material, v e is the volume of high thermal conductivity material in unit e, V 0 is the total volume, Θ n is the objective function, K is the global heat conduction matrix, P is the global heat load matrix, and T is the global temperature matrix.
[0015] Furthermore, the hypothetical material consists of two geometric parameters l 1 and l 2 Control represents the width of the horizontal rod and the vertical rod respectively, where the rod area is filled with high thermal conductivity material, and the remaining area is filled with low thermal conductivity material. The imaginary material is converted from an orthotropic material to an anisotropic material by rotating the angle θ.
[0016] Furthermore, the training of the microstructure reverse design model adopts the conditional variational autoencoder architecture in data-driven technology, which is composed of an encoder and a decoder.
[0017] Furthermore, the encoder will receive two parts of input: microstructure pixel information and microstructure corresponding heat conduction tensor. After these two parts of input information are processed layer by layer by the neural network, the mean vector μ and variance vector σ in the latent space are generated. These two parameters are used to define the Gaussian distribution in the latent space, that is, Where z is the latent variable, x is the original data, which is the microstructure pixel information; c is the conditional information, which is the heat conduction tensor corresponding to the microstructure; similarly, the decoder has a similar structure to the encoder in the opposite direction, and is passed to the decoder through a sampling point in the latent space as input; the decoder receives two parts of input: the sampled latent variable z and the conditional information c, and then through a series of deconvolution layers, the input information is mapped back to the original data, and the reconstructed microstructure is generated in the last layer of the decoder.
[0018] Furthermore, based on the two given component materials, the original microstructure is generated by the topology optimization method and the microstructure dataset is constructed. The corresponding thermal conductivity tensor is calculated as the attribute condition information by the energy homogenization method, and the thermal conductivity tensor is evenly distributed in the attribute space.
[0019] Further, the two given constituent materials are respectively alumina ceramic and polydimethylsiloxane.
[0020] The present invention also provides a non-uniform heat dissipation structure, which is designed using the data-driven cross-scale topology optimization design method for non-uniform heat dissipation structures as described above.
[0021] The present invention also provides a data-driven cross-scale topology optimization design system for non-uniform heat dissipation structures. The system includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it executes the data-driven cross-scale topology optimization design method for non-uniform heat dissipation structures as described above.
[0022] The present invention also provides a computer-readable storage medium, which stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to implement the data-driven cross-scale topology optimization design method for non-uniform heat dissipation structures as described above.
[0023] In general, compared with the prior art, the data-driven cross-scale topology optimization design method and device for non-uniform heat dissipation structure provided by the present invention have the following beneficial effects:
[0024] 1. The design method of the present invention is a data-driven cross-scale topology optimization design method, which realizes the efficient design of high-performance non-uniform heat dissipation structures, effectively solves the problems of high difficulty in designing non-uniform heat dissipation structures, high calculation cost and complex design process, and improves the design flexibility and accuracy.
[0025] 2. The design of the present invention adopts a data-driven method to construct a microstructure reverse design model, which solves the problems of high computational complexity and high computational cost of traditional structural design methods and realizes the efficient design of high-performance non-uniform heat dissipation structures.
[0026] 3. The microstructure reverse design model constructed by the present invention adopts the model architecture of conditional variational autoencoder, establishes the mapping relationship between thermal properties and microstructure configuration, and has strong generalization and robustness.
[0027] 4. The non-uniform heat dissipation structure designed by the present invention adopts anisotropic microstructure filling at the microscopic scale, which has a larger design space compared with the heat dissipation structure designed using isotropic microstructure.
[0028] 5. The non-uniform heat dissipation structure designed by the present invention adopts a cross-scale design to achieve the non-uniform distribution of the internal materials of the heat dissipation structure. Compared with the uniform structure designed by the single-scale method, the design method proposed by the present invention optimizes the heat dissipation performance of the structure to the greatest extent and has a higher degree of design freedom. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 It is a flow chart of a data-driven cross-scale topology optimization design method for non-uniform heat dissipation structure provided by the present invention;
[0030] Figure 2 (a) and (b) are schematic diagrams of imaginary materials constructed by the present invention;
[0031] Figure 3 is a schematic diagram of heat conduction boundary conditions constructed in an embodiment of the present invention;
[0032] Figure 4 (a) is the heat conduction tensor distribution calculated with minimizing heat dissipation compliance as the objective function; (b) is the heat conduction tensor distribution calculated with minimizing average temperature as the objective function; (c) is the heat conduction tensor distribution calculated with minimizing temperature variance as the objective function; (d) is the heat conduction tensor distribution calculated with minimizing maximum temperature as the objective function;
[0033] Figure 5 It is a schematic diagram of a microstructure reverse design model constructed by the present invention;
[0034] Figure 6 (a) is a non-uniform heat dissipation structure generated by minimizing heat dissipation flexibility as the objective function; (b) is a non-uniform heat dissipation structure generated by minimizing average temperature as the objective function; (c) is a non-uniform heat dissipation structure generated by minimizing temperature variance as the objective function; (d) is a non-uniform heat dissipation structure generated by minimizing maximum temperature as the objective function;
[0035] Figure 7 (a) is the equivalent medium simulation result of the non-uniform heat dissipation structure generated with the minimum heat dissipation flexibility as the objective function; (b) is the equivalent medium simulation result of the non-uniform heat dissipation structure generated with the minimum average temperature as the objective function; (c) is the equivalent medium simulation result of the non-uniform heat dissipation structure generated with the minimum temperature variance as the objective function; (d) is the equivalent medium simulation result of the non-uniform heat dissipation structure generated with the minimum maximum temperature as the objective function. DETAILED DESCRIPTION
[0036] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.
[0037] See also Figure 1 The present invention provides a data-driven cross-scale topology optimization design method for non-uniform heat dissipation structures. The design method adopts a data-driven cross-scale topology optimization design method to achieve efficient design of high-performance non-uniform heat dissipation structures, effectively solving the problems of high difficulty in designing non-uniform heat dissipation structures, high computational cost, and complex design process, and has high design flexibility and accuracy.
[0038] The design method mainly includes the following steps:
[0039] Step 1: At the macro scale, a cross-scale topology optimization model is constructed based on anisotropic hypothetical materials; based on the heat conduction boundary conditions and the thermal performance objective function, the heat conduction tensor distribution required inside the non-uniform heat dissipation structure to be optimized is calculated by solving the cross-scale topology optimization model.
[0040] Firstly, a hypothetical material is constructed as the basic unit for topology optimization to calculate the theoretical heat conduction tensor. The hypothetical material consists of two geometric parameters l 1 and l 2 Control, represents the width of the horizontal bar and the vertical bar, respectively, where the bar area is filled with high thermal conductivity material (99.99% alumina, κ Al2O3 =31Wm -1 K -1 ), and the remaining area is filled with low thermal conductivity material (PDMS, κ PDMS =0.16Wm -1 K -1 ). In addition, the hypothetical material can be transformed from an orthotropic material to an anisotropic material by rotating the material through an angle θ. 1 ,l 2 ,θ), the change of the equivalent thermal conductivity tensor of the hypothetical material can cover the anisotropic thermal conductivity tensor space of two given materials.
[0041] Secondly, a cross-scale topology optimization model is constructed based on anisotropic hypothetical materials. The mathematical expression of the cross-scale topology optimization model is:
[0042] find:ρ={l 1 ,l 2 ,θ}
[0043] min:Θ n
[0044] st:KT=P
[0045]
[0046] θ∈[0,π)
[0047] Among them, ρ(l 1 ,l 2 ,θ) contains three design variables, l 1 , l 2 is the geometric parameter of the anisotropic hypothetical material, θ is the rotation angle, f is the volume fraction constraint of the high thermal conductivity material, v e is the volume of high thermal conductivity material in unit e, V 0 is the total volume, Θ n is the objective function, K is the global heat conduction matrix, P is the global heat load matrix, and T is the global temperature matrix.
[0048] Finally, based on the heat conduction boundary conditions and thermal performance objective functions, the heat conduction tensor distribution required inside the non-uniform heat dissipation structure is calculated by solving the cross-scale topology optimization model. The four typical thermal performance objective function expressions are:
[0049] Thermal flexibility:
[0050] Among them, P is the global heat load matrix and T is the global temperature matrix.
[0051] Average temperature:
[0052] Where L is the unit vector and N is the total number of temperature points.
[0053] Temperature variance:
[0054] in, is the average temperature.
[0055] Maximum temperature:
[0056] Because of the non-differentiability of the maximum value function, the p-norm is used to approximate the maximum temperature, and p=15 in the calculation.
[0057] Step two, at the microscopic scale, a microstructure data set is constructed based on two given component materials and a microstructure reverse design model is established based on data-driven technology, and then the corresponding microstructure configuration is designed based on the microstructure reverse design model and the thermal conductivity tensor distribution required inside the non-uniform heat dissipation structure, thereby obtaining the final non-uniform heat dissipation structure.
[0058] First, based on two given composition materials: alumina ceramic (99.99% alumina, thermal conductivity κ Al2O3 =31Wm -1 K -1 ) and polydimethylsiloxane (PDMS, thermal conductivity κ PDMS =0.16Wm -1 K -1 ), the original microstructure is generated by topology optimization method and the microstructure dataset is constructed. The corresponding heat conduction tensor is calculated as the attribute condition information by energy homogenization method, and the heat conduction tensor is evenly distributed in the attribute space.
[0059] Secondly, a microstructure reverse design model based on data-driven technology is constructed. The training of the microstructure reverse design model adopts the conditional variational autoencoder architecture in data-driven technology, which consists of an encoder and a decoder. The encoder will receive two parts of input: original data (microstructure pixel information) and conditional information (microstructure corresponding to the thermal conductivity tensor). After these two parts of input information are processed layer by layer by the neural network, the mean vector μ and variance vector σ in the latent space are generated. These parameters are used to define the Gaussian distribution in the latent space, that is, Where z is the latent variable, x is the original data, and c is the conditional information; similarly, the decoder has a similar structure to the encoder in the opposite direction, passing it to the decoder through a sampling point in the latent space as input. The decoder receives two parts of input: the sampled latent variable z and the conditional information c, and then passes through a series of deconvolution layers to map the input information back to the original data, and generates a reconstructed microstructure in the last layer of the decoder.
[0060] Finally, based on the required heat conduction tensor distribution inside the non-uniform heat dissipation structure, the corresponding microstructure configuration is generated through the microstructure inverse design model.
[0061] The present invention is further described in detail below with a specific embodiment, which specifically includes the following steps:
[0062] Step 1: At the macro scale, a cross-scale topology optimization model is constructed based on anisotropic hypothetical materials. Based on the heat conduction boundary conditions and thermal performance objective function, the heat conduction tensor distribution required inside the non-uniform heat dissipation structure is calculated by solving the cross-scale topology optimization model.
[0063] First, a hypothetical material (such as Figure 2 As shown in Figure 2, the theoretical heat conduction tensor is calculated by topology optimization. The hypothetical material consists of two geometric parameters l 1 and l 2 Control, represents the width of the horizontal bar and the vertical bar, respectively, where the bar area is filled with high thermal conductivity material (99.99% alumina, κ Al2O3=31Wm -1 K -1 ), and the remaining area is filled with low thermal conductivity material (PDMS, κ PDMS =0.16Wm -1 K -1 ). In addition, the hypothetical material can be transformed from an orthotropic material to an anisotropic material by rotating the material through an angle θ. 1 ,l 2 ,θ), the change of the equivalent thermal conductivity tensor of the hypothetical material can cover the anisotropic thermal conductivity tensor space of two given materials.
[0064] Then, a cross-scale topology optimization model is constructed based on the anisotropic hypothetical material. The mathematical expression of the cross-scale topology optimization model is:
[0065] find:ρ={l 1 ,l 2 ,θ}
[0066] min:Θ n
[0067] st: KT = P (1)
[0068]
[0069] θ∈[0,π)
[0070] Among them, ρ(l 1 ,l 2 ,θ) contains three design variables, l 1 , l 2 is the geometric parameter of the anisotropic hypothetical material, θ is the rotation angle, f is the volume fraction constraint of the high thermal conductivity material, v e is the volume of high thermal conductivity material in unit e, V 0 is the total volume, Θ n is the objective function, K is the global heat conduction matrix, P is the global heat load matrix, and T is the global temperature matrix.
[0071] Finally, based on the heat conduction boundary conditions (such as Figure 3 As shown in Figure 2, the thermal performance objective function and the required heat conduction tensor distribution inside the non-uniform heat dissipation structure can be calculated by solving the cross-scale topology optimization model. The expressions of the four typical thermal performance objective functions are:
[0072] Thermal flexibility:
[0073]
[0074] Among them, P is the global heat load matrix and T is the global temperature matrix.
[0075] Average temperature:
[0076]
[0077] Where L is the unit vector and N is the total number of temperature points.
[0078] Temperature variance:
[0079]
[0080] in, is the average temperature.
[0081] Maximum temperature:
[0082]
[0083] Because of the non-differentiability of the maximum value function, the p-norm is used to approximate the maximum temperature, and p=15 in the calculation.
[0084] Based on the above formula, the required heat conduction tensor distribution inside the non-uniform heat dissipation structure with different thermal performance indicators as the objective function is as follows: Figure 4 shown.
[0085] Step 2: At the microscopic scale, a microstructure data set is constructed and a microstructure reverse design model is established based on data-driven technology. Then, based on the microstructure reverse design model and the required thermal conductivity tensor distribution inside the non-uniform heat dissipation structure, a corresponding microstructure configuration is designed to obtain the final non-uniform heat dissipation structure.
[0086] First, based on two given composition materials: alumina ceramic (99.99% alumina, thermal conductivity κ Al2O3 =31Wm -1 K -1 ) and polydimethylsiloxane (PDMS, thermal conductivity κ PDMS =0.16Wm -1 K -1 ), the original microstructure is generated by topology optimization method and the microstructure dataset is constructed. The corresponding heat conduction tensor is calculated as the attribute condition information by energy homogenization method, and the heat conduction tensor is evenly distributed in the attribute space.
[0087] Secondly, a microstructure reverse design model based on data-driven technology is constructed. The training of the microstructure reverse design model adopts the conditional variational autoencoder architecture in data-driven technology, such as Figure 5As shown in the figure, it consists of an encoder and a decoder; the encoder receives two parts of input, the original data (microstructure pixel information) and the conditional information (microstructure corresponding to the heat conduction tensor). After these two parts of input information are processed layer by layer by the neural network, the mean vector μ and variance vector σ in the latent space are generated. These parameters are used to define the Gaussian distribution in the latent space, that is, Where z is the latent variable, x is the original data, and c is the conditional information; similarly, the decoder has a similar structure to the encoder in the opposite direction, passing it to the decoder through a sampling point in the latent space as input. The decoder receives two parts of input: the sampled latent variable z and the conditional information c, and then passes through a series of deconvolution layers to map the input information back to the original data, and generates a reconstructed microstructure in the last layer of the decoder.
[0088] Finally, based on the required heat conduction tensor distribution inside the non-uniform heat dissipation structure, the corresponding microstructure configuration is generated through the microstructure inverse design model, thereby obtaining the final non-uniform heat dissipation structure, such as Figure 6 shown.
[0089] After obtaining the non-uniform heat dissipation structure, the equivalent medium simulation is then carried out in COMSOL Multiphysics software. The heat conduction boundary conditions of the equivalent medium simulation remain the same as those of the continuous medium simulation. After simulation calculation, the final steady-state simulation results are as follows: Figure 7 The heat dissipation performance indicators are shown in Table 1. From the simulation results, it can be found that when the design is performed with minimization of heat dissipation flexibility, minimization of average temperature, minimization of temperature variance and minimization of maximum temperature as the objective function, the heat dissipation performance measurement indicators corresponding to the obtained design results are all optimal, proving the effectiveness of the design method.
[0090] Table 1 Quantitative analysis results of heat dissipation performance
[0091]
[0092] The present invention also provides a non-uniform heat dissipation structure, which is designed using the data-driven cross-scale topology optimization design method for non-uniform heat dissipation structures as described above.
[0093] The present invention also provides a data-driven cross-scale topology optimization design system for non-uniform heat dissipation structures. The system includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it executes the data-driven cross-scale topology optimization design method for non-uniform heat dissipation structures as described above.
[0094] The present invention also provides a computer-readable storage medium, which stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to implement the data-driven cross-scale topology optimization design method for non-uniform heat dissipation structures as described above.
[0095] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A data-driven cross-scale topology optimization design method for non-uniform heat dissipation structures, characterized in that: The design method includes the following steps: Step 1: at the macroscopic scale, a cross-scale topology optimization model is constructed based on anisotropic hypothetical materials; based on the heat conduction boundary conditions and the thermal performance objective function, the heat conduction tensor distribution required inside the non-uniform heat dissipation structure to be optimized is calculated by solving the cross-scale topology optimization model; Step 2: at the microscopic scale, a microstructure data set is constructed based on two given component materials and a microstructure reverse design model is established based on data-driven technology, and then a corresponding microstructure configuration is designed based on the microstructure reverse design model and the heat conduction tensor distribution, so as to obtain a final non-uniform heat dissipation structure; The mathematical expression of the cross-scale topology optimization model is: Among them, ρ(l1,l2,θ) contains three design variables, l1 and l2 are the geometric parameters of the anisotropic hypothetical material, θ is the rotation angle, f is the volume fraction constraint of the high thermal conductivity material, and v e is the volume of high thermal conductivity material in unit e, V0 is the total volume, Θ n is the thermal performance objective function, K is the global heat transfer matrix, P is the global heat load matrix, and T is the global temperature matrix; The training of the microstructure reverse design model adopts the conditional variational autoencoder architecture in data-driven technology, which consists of an encoder and a decoder; the encoder will receive two parts of input: microstructure pixel information and microstructure corresponding heat conduction tensor. After these two parts of input information are processed layer by layer by the neural network, the mean vector μ and variance vector σ in the latent space are generated. These two parameters are used to define the Gaussian distribution in the latent space, that is, Where z is the latent variable, x is the original data, which is the microstructure pixel information; c is the conditional information, which is the heat conduction tensor corresponding to the microstructure; similarly, the decoder has a similar structure to the encoder in the opposite direction, and is passed to the decoder through a sampling point in the latent space as input; the decoder receives two parts of input: the sampled latent variable z and the conditional information c, and then through a series of deconvolution layers, the input information is mapped back to the original data, and the reconstructed microstructure is generated in the last layer of the decoder.
2. The data-driven cross-scale topology optimization design method for non-uniform heat dissipation structure according to claim 1, characterized in that: The imaginary material is controlled by two geometric parameters l1 and l2, which represent the width of the horizontal rod and the vertical rod respectively. The rod area is filled with high thermal conductivity material, and the remaining area is filled with low thermal conductivity material. The imaginary material is converted from an orthotropic material to an anisotropic material by rotating the angle θ.
3. The data-driven cross-scale topology optimization design method for non-uniform heat dissipation structure according to any one of claims 1 to 2, characterized in that: Based on two given component materials, the original microstructure is generated by the topology optimization method and a microstructure dataset is constructed. The corresponding thermal conductivity tensor is calculated as the attribute condition information through the energy homogenization method, and the thermal conductivity tensor is evenly distributed in the attribute space.
4. The data-driven cross-scale topology optimization design method for non-uniform heat dissipation structure according to claim 3, characterized in that: The two given constituent materials are alumina ceramic and polydimethylsiloxane.
5. A non-uniform heat dissipation structure, characterized in that: The non-uniform heat dissipation structure is designed using the data-driven cross-scale topology optimization design method for non-uniform heat dissipation structure as described in any one of claims 1 to 4.
6. A data-driven cross-scale topology optimization design system for non-uniform heat dissipation structures, characterized by: The system includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the data-driven cross-scale topology optimization design method for non-uniform heat dissipation structure according to any one of claims 1 to 4 is executed.
7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by the processor, the machine-executable instructions prompt the processor to implement the data-driven cross-scale topology optimization design method for non-uniform heat dissipation structure according to any one of claims 1-4.