Cold plate design method for balancing heat dissipation and power consumption and cold plate

Abstract

The invention belongs to the related technical field of data center chip heat management, and discloses a cold plate design method for balancing heat dissipation and power consumption and a cold plate, and the method comprises the steps: carrying out the two-dimensional topological optimization design of a flow channel region of the cold plate, and determining a topological optimization flow channel structure; establishing a three-dimensional model of the topological optimization flow channel based on the topological optimization flow channel structure; performing straight micro-channel structure design on a flow channel area of the cold plate to obtain a straight micro-channel structure three-dimensional model; and the three-dimensional model of the topological optimization flow channel and the three-dimensional model of the straight micro-channel structure are integrated, and a three-dimensional model of the cold plate is obtained. The advantages of a straight micro-channel cold plate and the advantages of a topological optimization structure cold plate are integrated, an innovative integrated design scheme is provided, topological optimization and the straight micro-channel structure are integrated, primary and secondary multi-stage flow channels are formed, the complementary advantages are brought into full play, and pump work and energy consumption are reduced while the heat dissipation performance and temperature uniformity are improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field related to data center chip heat management, and more particularly to a cold plate design method balancing heat dissipation and power consumption and a cold plate. BACKGROUND

[0002] With the rapid development of cloud computing, artificial intelligence and 5G communication technology, the power density of high-performance chips has been greatly improved, which puts forward strict requirements on the heat management system. Efficient heat dissipation has become a key factor to ensure the reliability, performance and service life of electronic components, especially in data centers and high-performance computing environments. Traditional air-cooled heat dissipation technology has been unable to meet the heat dissipation needs of high heat flux density chips. Cold plate heat sinks are widely used in engineering due to their compact structure, low cost, low noise and mature technology.

[0003] Common cold plates include S-shaped flow channel cold plates, Y-shaped flow channel cold plates and Z-shaped flow channel cold plates, etc. The flow channel form of the cold plate is generally designed by experience. In order to meet the heat dissipation needs of high heat flux density, the existing flow channel form is usually designed to be complex, so as to realize better heat dissipation effect by prolonging the flow process of the cooling liquid in the flow channel. However, as the flow channel form tends to be complex, the flow power consumption is increased while the heat dissipation performance is improved, which requires greater pump power to drive the cooling liquid to flow smoothly, thereby reducing the practicality of the flow channel structure. SUMMARY

[0004] In view of the above defects or improvement needs of the prior art, the present application provides a cold plate design method balancing heat dissipation and power consumption and a cold plate, which is used to solve the problem that the existing cold plate flow channel design is difficult to balance the heat dissipation performance and pump power consumption, thereby reducing the practicality of the flow channel structure. The design scheme of integrating topology optimization and straight micro-channel structure can meet the heat dissipation needs of high heat flux density chips and electronic equipment in data centers.

[0005] To achieve the above-mentioned purpose, according to one aspect of the present application, a cold plate design method balancing heat dissipation and power consumption is provided, comprising: performing two-dimensional topology optimization design on the flow channel region of the cold plate to determine a topology optimization flow channel structure; establishing a three-dimensional model of the topology optimization flow channel based on the topology optimization flow channel structure; designing a straight micro-channel structure for the flow channel region of the cold plate to obtain a three-dimensional model of the straight micro-channel structure; integrating the three-dimensional model of the topology optimization flow channel and the three-dimensional model of the straight micro-channel structure to obtain a three-dimensional model of the cold plate.

[0006] According to the cold plate design method balancing heat dissipation and power consumption provided by the present application, the two-dimensional topology optimization design on the flow channel region of the cold plate specifically comprises: A two-dimensional design domain is created according to a flow channel region of the cold plate, and mesh division is performed, a plurality of physical field boundary conditions, initial conditions, material interpolation models, density filtering and projection models are set, and the minimum cold plate average temperature and power consumption are selected as the target of the topology optimization; Physical field data are obtained by solving control equations, sensitivity analysis is performed by using the adjoint method, the optimization variables and the topology structure are updated by using the moving asymptote method, and the topology optimization flow channel structure is determined after the convergence condition is met.

[0007] According to the cold plate design method for balancing heat dissipation and power consumption provided by the application, the target of the topology optimization is: ; Among them, the final topology optimization target, the temperature weight factor, the power consumption weight factor, the normalized cold plate average temperature, the normalized cold plate power consumption, and the sum is 1, and 0.5-0.8.

[0008] According to the cold plate design method for balancing heat dissipation and power consumption provided by the application, the straight microchannel structure design of the flow channel region of the cold plate is specifically: A flow convergence zone is arranged on each side of the flow channel region along the flow direction, a plurality of straight microchannels are arranged in parallel between the two flow convergence zones to form the straight microchannel structure, each straight microchannel is arranged along the flow direction and the two ends thereof are communicated with the two flow convergence zones, and the plurality of straight microchannels are arranged at equal intervals in the entire region of the flow channel region.

[0009] According to the cold plate design method for balancing heat dissipation and power consumption provided by the application, the width of each straight microchannel is the same as the spacing between the two adjacent straight microchannels, and the width of each straight microchannel is 10µm-1mm.

[0010] According to the cold plate design method for balancing heat dissipation and power consumption provided by the application, the topology optimization flow channel three-dimensional model and the straight microchannel structure three-dimensional model are integrated, specifically including: The topology optimization flow channel three-dimensional model and the straight microchannel structure three-dimensional model are integrated by performing a union set through Boolean operation to generate a cold plate three-dimensional model.

[0011] According to the cold plate design method for balancing heat dissipation and power consumption provided by the application, further including: establishing a simulation model based on the cold plate three-dimensional model; The heat flux density of the cold plate application object, the cooling liquid inlet temperature of the cold plate, and the cooling liquid flow rate are selected as design variables, sample points are obtained based on an experimental design method, the maximum temperature value of the application object corresponding to each sample point is calculated based on a simulation model through parameterized simulation, and a data set is obtained; A proxy model of the maximum temperature of the application object corresponding to the three-dimensional model of the cold plate is established based on the data set; The design selection of the cold plate working condition parameters is performed based on the proxy model and the actual application working condition.

[0012] According to the cold plate design method for balancing heat dissipation and power consumption provided in the application, the proxy model is established based on the data set using a response surface method, and the expression of the proxy model is as follows: ; Wherein, T max is the maximum temperature of the application object, q represents the heat flux density of the application object, T in is the cooling liquid inlet temperature, v in is the cooling liquid inlet flow rate. According to the cold plate design method for balancing heat dissipation and power consumption provided in the application, within the value range of the three design variables of the heat flux density of the cold plate application object, the cooling liquid inlet temperature of the cold plate, and the cooling liquid flow rate, a plurality of sample points of the combination of the three design variables are obtained through Latin hypercube sampling; wherein the value range of the heat flux density of the cold plate application object is 20-100 W / cm 2 ; the value range of the cooling liquid inlet temperature of the cold plate is 293.15-303.15 K; and the value range of the cooling liquid flow rate is 0.3-0.8 L / min.

[0013] According to another aspect of the application, a cold plate is provided, which is designed based on the cold plate design method for balancing heat dissipation and power consumption according to any one of the above.

[0014] Overall, compared with the prior art, the cold plate design method for balancing heat dissipation and power consumption and the cold plate provided in the application have the following advantages: 1.The present application considers that the straight micro-channel cold plate has excellent flow field uniformity and low pressure drop characteristics, which are crucial for maintaining uniform distribution of cooling liquid throughout the area, in contrast, the topology optimization structure cold plate enhances the local heat dissipation capacity by preferentially guiding the cooling liquid to the hot spot area, thus showing excellent heat dissipation performance, and by integrating the advantages of both, an innovative integrated design scheme is proposed, which integrates topology optimization and straight micro-channel structure to form primary and secondary multi-stage flow channels, so as to fully exert the complementary advantages, which is conducive to improving heat dissipation performance and temperature uniformity while reducing pump power and energy consumption, and provides an expandable and efficient thermal management solution for high heat flux electronic equipment; 2.The straight micro-channel structure can exert the low pressure drop characteristics in the entire flow channel area of the cold plate, and after integration with the three-dimensional model of the topology optimization flow channel, it is conducive to improving the flow field uniformity while ensuring the heat dissipation performance, and effectively reducing the pump power and energy consumption; 3.Further, an application object maximum temperature proxy model is established for the integrated structure, which is conducive to better optimization design of the cold plate working condition parameters when facing actual application conditions, and provides an expandable and efficient thermal management solution for high heat flux electronic equipment; 4.Compared with the traditional straight micro-channel, the integrated design structure can reduce the maximum temperature of the application object such as a chip by 4.28 K, reduce the pressure drop by 319.42 Pa, and at the same time improve the temperature uniformity. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 is a flow chart of the cold plate design method of the present application embodiment balancing heat dissipation and power consumption.

[0016] Figure 2 is a two-dimensional topology optimization problem schematic diagram of the cold plate of the present application embodiment.

[0017] Figure 3 is a modeling flow chart schematic diagram of the integrated topology optimization and micro-channel structure cold plate of the present application embodiment.

[0018] Figure 4 is a three-dimensional model schematic diagram of the cold plate of different design schemes of the present application embodiment.

[0019] Figure 5 is a mesh independence test result schematic diagram of the present application embodiment.

[0020] Figure 6 is a chip maximum temperature and minimum temperature comparison diagram under different cold plate design schemes of the present application embodiment.

[0021] Figure 7 is a temperature standard deviation and cooling thermal resistance comparison diagram under different cold plate design schemes of the present application embodiment.

[0022] Figure 8 The figure is a comparison of the inlet and outlet pressure drop and flow resistance coefficient of different cold plate design schemes of the embodiment of the present application.

[0023] Figure 9 The figure is a comparison of the predicted value and the true value of the maximum temperature of the chip on the test set of the embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical scheme and advantages of the present application clearer, further detailed description will be made to the present application in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0025] Please refer to Figure 1 The embodiment provides a cold plate design method for balancing heat dissipation and power consumption, which comprises the following steps: performing two-dimensional topological optimization design on the flow channel region of the cold plate to determine a topologically optimized flow channel structure; establishing a three-dimensional model of the topologically optimized flow channel based on the topologically optimized flow channel structure; performing straight micro-channel structure design on the flow channel region of the cold plate to obtain a three-dimensional model of the straight micro-channel structure; integrating the three-dimensional model of the topologically optimized flow channel and the three-dimensional model of the straight micro-channel structure to obtain a three-dimensional model of the cold plate.

[0026] In some embodiments, in the topological optimization design of the cold plate, the flow channel region of the cold plate is regarded as a design domain for two-dimensional topological optimization, and the flow channel region is the region corresponding to the application object such as a chip. A two-dimensional topological optimization model of the cold plate can be established based on COMSOL software, and a topologically optimized flow channel structure is obtained through topological optimization. The topological optimization adopts the variable density method, assumes that the material of the design domain is a porous medium, and adjusts the material properties by controlling the value of γ. Figure 2 The figure is a schematic diagram of the two-dimensional topological optimization problem of the cold plate, and the parameters in the two-dimensional topological optimization are as follows: T in The inlet temperature of the cooling liquid is T_in, P in The pressure inlet is P_in, P out The pressure outlet is P_out. The value of γ ranges from 0 to 1, γ=0 represents a solid region, and γ=1 represents a fluid region. In order to balance heat dissipation and power consumption, the target of the topological optimization is set to minimize the average temperature of the cold plate and the power consumption of the cold plate. The temperature weight factor and the power consumption weight factor The two optimization objectives are combined, and the final optimization objective is expressed as follows: ; wherein, is the final topology optimization objective, is a temperature weight factor, is a power consumption weight factor, is a normalized cold plate average temperature, is a normalized cold plate power consumption, and the sum of which is 1, and is 0.5-0.8.

[0027] Further, the two-dimensional topology optimization design of the flow channel region of the cold plate specifically includes: creating a two-dimensional design domain according to the flow channel region of the cold plate and performing mesh division, setting multi-physical field boundary conditions, initial conditions, material interpolation models, density filtering and projection models, and selecting the minimum cold plate average temperature and power consumption as the topology optimization objective; obtaining physical field data by solving control equations, performing sensitivity analysis by using the adjoint method, updating the optimization variables and topology structure by using the moving asymptote method, and determining the topology optimization flow channel structure after meeting the convergence condition. The specific settings of the topology optimization can be made in a conventional manner, and are not limited in particular.

[0028] In some embodiments, the straight micro-channel structure design of the flow channel region of the cold plate specifically includes: setting a flow convergence zone on each side of the flow channel region along the flow direction, and setting a plurality of straight micro-channels in parallel between the two flow convergence zones to form the straight micro-channel structure; each straight micro-channel is arranged along the flow direction and communicates with the two flow convergence zones at both ends; the plurality of straight micro-channels are arranged at equal intervals in the entire region of the flow channel region, as shown in Figure 3 Optionally, the width of each straight micro-channel and the spacing between adjacent two straight micro-channels can be the same; the width of each straight micro-channel is 10 µm - 1 mm.

[0029] Further, the three-dimensional model of the cold plate is constructed as follows: exporting the two-dimensional flow channel obtained by topology optimization to a general CAD format file and importing it into Solidworks software, stretching in the thickness direction of the cold plate based on the assembly relationship and structure and structure parameters, and forming a shell to construct a three-dimensional cold plate model under the topology optimization structure, and establishing a three-dimensional cold plate model under the straight micro-channel, integrating the topology optimization model and the straight micro-channel model through Boolean operation, i.e., embedding the straight micro-channel cold plate into the topology optimization cold plate to form the main channel and the secondary channel. This collaborative design not only realizes efficient flow, but also enhances the heat dissipation capacity, thereby breaking through the limitations of single structure design application, Figure 3A schematic diagram of a modeling process for integrating topological optimization and micro-channel structure cold plate.

[0030] Specifically, the topological optimization flow channel three-dimensional model and the straight micro-channel structure three-dimensional model are integrated, specifically including: the topological optimization flow channel three-dimensional model and the straight micro-channel structure three-dimensional model are integrated by a Boolean operation to obtain a union to generate a cold plate three-dimensional model integrating topological optimization and micro-channel structure.

[0031] The topological optimization flow channel three-dimensional model and the straight micro-channel structure three-dimensional model are integrated to obtain a cold plate three-dimensional model, and further includes: after the topological optimization flow channel three-dimensional model and the straight micro-channel structure three-dimensional model are integrated and merged, the merged three-dimensional model is meshed (for example, the merged three-dimensional model is imported into Workbench for meshing, and the mesh type and size can be set according to experience values), the mesh quality is judged, and the local area whose mesh quality does not meet the set requirement value is adjusted until the quality of all meshes meets the set requirement value, and the final cold plate three-dimensional model is obtained. The local area whose mesh quality does not meet the set requirement value is, for example, a sharp area, and the adjustment includes setting a chamfer adjustment, a deletion setting, or a local adjustment direction setting, etc. After adjustment, meshing can be performed again to judge the quality.

[0032] To prove the excellent performance of the cold plate integrating topological optimization and straight micro-channel structure, this embodiment compares a separate straight micro-channel cold plate and a topological optimization cold plate, as shown in Figure 4 The three-dimensional model schematic diagram of the cold plate of different design schemes is shown. Scheme A is a straight micro-channel structure cold plate constructed in the above embodiment, which has a lower pressure drop due to the simple and regular geometry of the flow channel. The three-dimensional models of schemes B and C are established based on two-dimensional topological optimization results. Scheme B is a topological optimization structure with a temperature weight factor =0.6, and scheme C is a topological optimization structure with a weight factor =0.7. Schemes B and C can more flexibly guide the fluid to the hot spot area, thereby improving the heat dissipation efficiency, but at the cost of increasing the flow resistance. The three-dimensional model of scheme D proposed in this embodiment is designed after integrating topological optimization and micro-channel structure, which not only maintains the efficient flow characteristics of micro-channels, but also utilizes the strong heat dissipation capacity of topological optimization.

[0033] Further, a cold plate design method balancing heat dissipation and power consumption further includes: establishing a simulation model based on the cold plate three-dimensional model; selecting the heat flux of the cold plate application object, the cooling liquid inlet temperature of the cold plate, and the cooling liquid flow rate as design variables, obtaining sample points based on an experimental design method, calculating the maximum temperature value of the application object corresponding to each sample point based on the simulation model through parameterized simulation to obtain a data set; establishing a proxy model of the maximum temperature of the application object corresponding to the three-dimensional model of the cold plate based on the data set; designing and selecting the cold plate working condition parameters based on the proxy model and the actual application working condition. The actual application working condition can determine the heat flux density of the application object in the actual application, and the cold plate working condition parameter can be the cooling liquid inlet temperature and the cooling liquid flow rate of the cold plate. The establishment of the proxy model is beneficial to the rapid optimization and selection of the cold plate working condition parameters.

[0034] Specifically, the construction and result analysis of the three-dimensional numerical simulation model of the cold plate are as follows: The three-dimensional model of the cold plate can be imported into Workbench to establish a simulation model, and three-dimensional numerical simulation is performed based on Fluent. In the parameterized simulation process, the cold plate and the cooling liquid are defined as the material properties of aluminum and water respectively. The cooling liquid is regarded as an incompressible fluid with constant characteristics, and the application object is taken as a chip. The thermal conductivity of the thermal interface material at the contact between the chip and the cold plate is set to 6 W / (m·K). The parameter settings of the numerical simulation model are shown in Table 1. The cooling liquid inlet temperature and flow rate are set to 298.15 K and 0.4 L / min respectively, and the outlet pressure is set to 0 Pa. A constant heat flux of 100 W / cm2 is applied to the chip surface. 2 The remaining exposed surfaces are set as adiabatic surfaces.

[0035] Table 1 Parameter settings of the three-dimensional numerical simulation model

[0036] Based on the three-dimensional model of the cold plate, the simulation model also includes: performing different number of grid division tests on the three-dimensional model of the cold plate, and then performing grid independence tests on different grid numbers, considering the calculation accuracy and cost to determine the final grid number; the grid number can be gradually increased to establish the simulation model, and the numerical simulation is performed under the condition that the model parameters except the grid number are the same to obtain the maximum temperature of the application object, i.e. the heat source, when the maximum temperature of the heat source tends to be stable, the corresponding grid number is the final grid number; thereby ensuring that the setting of the grid number will not affect the numerical simulation.

[0037] In addition, the simulation model is constructed under the final grid number and the boundary condition parameters are set to verify the accuracy of the simulation model and adjust the boundary condition parameters to determine the final simulation model. The accuracy of the simulation model can be verified by using the flow channel model of the existing research results (not limited to the cold plate model provided in the embodiment of the present application), and the simulation simulation is performed under the same boundary condition parameters, and compared with the existing research results. When the result deviation is not large, it means that the boundary condition parameters of the simulation model are effective, and the simulation model is accurate.

[0038] The grid independence test is the basis of ensuring the accuracy of the calculation results and improving the calculation speed. The tetrahedral grid is used when dividing the grid, and the contact area between the solid domain and the fluid domain is refined to avoid the generation of low-quality grids. Figure 5 The variation of the maximum temperature of the chip with the number of grids is shown. Due to the more complex flow channel structure of schemes B, C, and D than scheme A, the number of grids increases significantly after refinement. For scheme A, when the number of grids exceeds 969443, the number of grids has negligible effect on the maximum temperature of the chip. For schemes B, C, and D, when the number of grids exceeds 6521346, 6478465, and 6587782, respectively, the number of grids has almost no effect on the maximum temperature of the chip. To balance the calculation accuracy and speed, the number of grids for schemes A, B, C, and D is 969443, 6521346, 6478465, and 6587782, respectively, when performing three-dimensional numerical simulation.

[0039] To comprehensively evaluate the thermal and hydraulic performance of the integrated topology optimization and micro-channel structure cold plate proposed in this embodiment, key performance indicators are defined, which provide a quantitative basis for evaluating the thermal and hydraulic performance of different schemes. The chip temperature is crucial to the service life and operation reliability of the chip, and the maximum temperature of the chip T max and the minimum temperature T min are defined as follows: ; wherein, T max is the maximum temperature of the chip, T min is the minimum temperature of the chip, T is the temperature of the grid node.

[0040] To evaluate the uniformity of the entire chip temperature distribution, the temperature standard deviation is introduced, and a lower indicates better temperature uniformity and reduces the risk of local overheating. The calculation formula of the temperature standard deviation is as follows: ; wherein, is the temperature standard deviation, is the temperature of the th grid, is the average value of , and n is the number of grids.

[0041] The heat dissipation performance of the cold plate can also be evaluated intuitively through the thermal resistance, and the smaller the cooling thermal resistance, the higher the heat dissipation efficiency. The definition of the thermal resistance is as follows: ; in, T surf,max This is the highest surface temperature of the cold plate. T in This refers to the coolant inlet temperature. Q in This is the power input to the chip.

[0042] The flow characteristics of the cold plate are determined by the pressure drop at the inlet and outlet. and dimensionless flow resistance coefficient Characterization, and The flow resistance indirectly reflected by the internal geometry of the cold plate can be calculated using the following formula: ; ; in, For the pressure drop at the inlet and outlet of the cold plate, Due to import pressure, Due to export pressure, The dimensionless flow resistance coefficient, For coolant density, This represents the coolant flow rate.

[0043] Figure 6 To compare the maximum and minimum chip temperatures under different cold plate designs, scheme A shows a maximum chip temperature of 352.29 K, with a large thermal gradient and excessively high center temperature. Scheme C achieves the lowest peak temperature with a maximum temperature of 346.43 K, indicating superior heat dissipation performance. Scheme D, with its integrated topology optimization and microchannel structure cold plate, achieves a maximum chip temperature of 348.01 K, comparable to scheme B. Figure 7 The figure shows a comparison of temperature standard deviation and cooling thermal resistance under different cold plate design schemes. Scheme A has the highest temperature non-uniformity and temperature standard deviation. The maximum temperature is 1.97 K, while scheme D has the lowest temperature standard deviation. With a temperature as low as 1.20 K, it exhibits excellent thermal uniformity, which is more conducive to extending the lifespan of the chip. Figure 8 A comparison chart of inlet and outlet pressure drop and flow resistance coefficient under different cold plate design schemes, from Figure 8 It can be seen that although schemes B and C achieve efficient heat dissipation under minimum thermal resistance conditions, the inlet and outlet pressure drops increase significantly. Scheme D reduces the inlet and outlet pressure drops to 1415.23 Pa, which is 319.42 Pa lower than that of the straight microchannel cold plate, and has the lowest flow resistance coefficient. =0.88. Although Scheme A has a lower voltage drop, its thermal performance is poor and it is not suitable for high heat flux devices. Overall, Scheme D strikes a balance between heat dissipation efficiency and voltage drop, while also possessing excellent temperature uniformity and the lowest flow resistance coefficient, making it an ideal choice for cooling high heat flux density chips.

[0044] The surrogate model for the maximum temperature of the application object under the integrated topology optimization and microchannel structure cold plate is as follows: To further guide the establishment of the control model, under scheme D integrating topology optimization and microchannel structure cold plate, a surrogate model for the highest temperature of the application object is established based on response surface methodology. Specifically, in the two-dimensional topology optimization design stage, the temperature weighting factor corresponding to the average temperature of the cold plate in the optimization objective is used. =0.6, average fluid volume fraction =0.5, coolant inlet temperature and flow rate are set to 298.15 K and 0.4 L / min respectively, outlet pressure is set to 0 Pa; the design domain is rectangular, where the length-to-width ratio of the design domain is 1-1.5, and the length-to-width ratio of the inlet and outlet is 1.5-2 (see specific examples). Figure 2 As shown, the cold plate has a length of 14.8 mm and a width of 12.4 mm, with inlets and outlets on both sides along the length direction (the dimensions of the inlets and outlets are 2.6 × 1.5 mm). A three-dimensional model of the cold plate in scheme D is obtained by integrating a straight microchannel structure. The cold plate and coolant are defined as aluminum and water, respectively, and the thermal conductivity of the thermal interface material at the contact point between the application object and the cold plate is set to 6 W / (m·K) for numerical simulation.

[0045] Within the range of three design variables—heat flux density of the cold plate application object, coolant inlet temperature of the cold plate, and coolant flow rate—multiple sample points were obtained using Latin hypercube sampling to combine the three design variables. The heat flux density of the cold plate application object ranged from 20 to 100 W / cm². 2 The coolant inlet temperature of the cold plate ranges from 293.15 to 303.15 K, and the coolant flow rate ranges from 0.3 to 0.8 L / min. Through parametric simulation, 80 data points were obtained to establish a maximum temperature proxy model. This model comprehensively considers the interaction of three key variable parameters: the heat flux density of the application object, the coolant inlet temperature, and the coolant inlet flow rate. The maximum temperature proxy model for the application object obtained using 60 sets of data as the training set is as follows: ; in, T max The maximum temperature of the application object. q The heat flux density representing the application object, T in This refers to the coolant inlet temperature.v in is the coolant inlet flow rate. The surrogate model captures both the main effects and the second-order interactions of the variables, and can be used to quickly evaluate the thermal behavior without repeating full-scale simulations, which is suitable for design screening or real-time control applications. The accuracy of the surrogate model is verified by using 20 groups of data in the dataset as the test set. The average absolute error, root mean square error, and coefficient of determination of the model on the test set are 0.375, 0.118, and 0.999, respectively. As shown in FIG. 6, the predicted maximum temperature of the chip on the test set is compared with the actual value, and it can be seen that the predicted temperature is highly consistent with the actual value under various working conditions, and all the predicted values fall within the range of ±1% of the actual temperature data. Overall, the surrogate model can be used as a practical tool for fast thermal prediction and system control strategy development. Figure 9 Figure 9 As shown in FIG. 6, the predicted maximum temperature of the chip on the test set is compared with the actual value, and it can be seen that the predicted temperature is highly consistent with the actual value under various working conditions, and all the predicted values fall within the range of ±1% of the actual temperature data. Overall, the surrogate model can be used as a practical tool for fast thermal prediction and system control strategy development.

[0046] In other embodiments, a cold plate is also provided, which is designed based on the cold plate design method for balancing heat dissipation and power consumption according to any one of the above.

[0047] Those skilled in the art will easily understand that the above description is only the preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.​

Claims

1. A cold plate design method balancing heat dissipation and power consumption, characterized in that, The method comprises the following steps: topology optimization design is performed on the flow channel region of the cold plate to determine a topology optimization flow channel structure; a three-dimensional model of the topology optimization flow channel is established based on the topology optimization flow channel structure; straight micro-channel structure design is performed on the flow channel region of the cold plate to obtain a three-dimensional model of the straight micro-channel structure; the three-dimensional model of the topology optimization flow channel and the three-dimensional model of the straight micro-channel structure are integrated to obtain a three-dimensional model of the cold plate.

2. The cold plate design method of balancing heat dissipation and power consumption of claim 1, wherein, The topology optimization design on the flow channel region of the cold plate specifically comprises the following steps: a two-dimensional design domain is created based on the flow channel region of the cold plate and is meshed, multi-physical field boundary conditions, initial conditions, material interpolation models, density filtering and projection models are set, and the minimum average temperature and power consumption of the cold plate are selected as the target of topology optimization; physical field data are obtained by solving control equations, sensitivity analysis is performed by using the adjoint method, and the optimization variables and topology structure are updated by using the moving asymptote method, and the topology optimization flow channel structure is determined after meeting the convergence condition.

3. The cold plate design method of balancing heat dissipation and power consumption of claim 2, wherein, The target of topology optimization is: ； wherein, is the final topology optimization target, is a temperature weight factor, is a power consumption weight factor, is the normalized cold plate average temperature, is the normalized cold plate power consumption, and sum to 1, and takes a value of 0.5-0.

8.

4. The cold plate design method of balancing heat dissipation and power consumption of claim 1, wherein, The straight micro-channel structure design on the flow channel region of the cold plate specifically comprises the following steps: a plurality of straight micro-channels are arranged in parallel between the two sides of the flow channel region along the flow direction to form the straight micro-channel structure; each straight micro-channel is arranged along the flow direction and communicates with the two sides of the flow channel region at both ends; and the plurality of straight micro-channels are arranged at equal intervals in the entire region of the flow channel region.

5. The cold plate design method of balancing heat dissipation and power consumption of claim 4, wherein, The width of each straight micro-channel is the same as the spacing between adjacent two straight micro-channels; and the width of each straight micro-channel is 10 µm - 1 mm.

6. The cold plate design method of balancing heat dissipation and power consumption of claim 1, wherein, The integration of the three-dimensional model of the topology optimization flow channel and the three-dimensional model of the straight micro-channel structure specifically comprises the following steps: the three-dimensional model of the topology optimization flow channel and the three-dimensional model of the straight micro-channel structure are integrated by performing a Boolean operation to obtain a union to generate a three-dimensional model of the cold plate.

7. The cold plate design method of balancing heat dissipation and power consumption of claim 1, wherein, The method further comprises the following steps: a simulation model is established based on the three-dimensional model of the cold plate; the heat flux density of the application object of the cold plate, the cooling liquid inlet temperature of the cold plate and the cooling liquid flow rate are selected as design variables, sample points are obtained based on an experimental design method, and the maximum temperature value of the application object corresponding to each sample point is calculated by parameterized simulation based on the simulation model to obtain a data set; a proxy model of the maximum temperature of the application object corresponding to the three-dimensional model of the cold plate is established based on the data set; the design selection of the cold plate working condition parameters is performed based on the proxy model and the actual application working condition.

8. The cold plate design method of balancing heat dissipation and power consumption of claim 7, wherein, The proxy model is established based on the data set by using a response surface method, and the expression of the proxy model is specifically as follows: ； wherein, T max is the maximum temperature of the application object, q represents the heat flux density of the application object, T in is the inlet temperature of the cooling liquid, v in is the inlet flow rate of the cooling liquid.

9. The cold plate design method of balancing heat dissipation and power consumption of claim 7, wherein, In the value range of the three design variables of the heat flux density of the cold plate application object, the cooling liquid inlet temperature of the cold plate and the cooling liquid flow rate, a plurality of sample points of the three design variable combinations are obtained through Latin hypercube sampling; wherein the value range of the heat flux density of the cold plate application object is 20-100W / cm 2 ; the value range of the cooling liquid inlet temperature of the cold plate is 293.15-303.15K; and the value range of the cooling liquid flow rate is 0.3-0.8L / min.

10. A cold plate characterized by, The cold plate design method for balancing heat dissipation and power consumption is designed based on any one of the above claims 1-9. The cold plate design method for balancing heat dissipation and power consumption is designed based on any one of the above claims 1-9.

Images