Cold plate optimization design method and apparatus

Abstract

The present disclosure relates to the technical field of liquid cooling heat dissipation, and discloses a cold plate optimization design method and device. The method comprises the following steps: generating a cold plate model set according to a cold plate initial design parameter sample set by using a cold plate modeling tool; generating a liquid cooling simulation data set according to a cooling liquid type, a cooling liquid flow rate and the cold plate model set by using a liquid cooling simulation tool; generating an optimized design parameter combination and a cooling liquid flow rate of an optimized design parameter in the corresponding optimized design parameter combination by using a multi-island genetic algorithm; selecting an optimized design parameter meeting a design requirement as a target cold plate design parameter from the optimized design parameter combination, and determining a target cooling liquid flow rate corresponding to the target cold plate design parameter according to the cooling liquid flow rate of the optimized design parameter in the corresponding optimized design parameter combination, thereby optimizing the determined target cold plate design parameter and improving the heat dissipation efficiency of immersion liquid cooling of the cold plate directional concentrated cooling.

Description

Technical Field

[0001] This disclosure relates to the field of liquid cooling technology, specifically to a method and apparatus for optimizing the design of cold plates. Background Technology

[0002] With the acceleration of digital transformation, the demand for computing power in data centers and servers has surged, leading to increasingly prominent energy consumption and heat dissipation issues. Single-phase immersion systems in liquid cooling technology typically use cold plate radiators for directional and centralized heat dissipation, and widely adopt serrated cold plates that directly reuse pure cold plate server designs. These designs are mostly based on the cooling characteristics of low-viscosity liquids (such as deionized water).

[0003] However, the current immersion cooling system uses a high-viscosity coolant (such as fluorinated liquid or oil) to first flow through a cold plate to cool high-power devices such as the CPU and GPU, and then flows out of the cold plate into the chassis to cool other lower-power devices such as hard drives and network cards.

[0004] Therefore, how to improve the heat dissipation efficiency of immersion liquid cooling with directional centralized cooling when optimizing the design of the spade-shaped cold plate in the immersion heat dissipation system has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, the present disclosure provides a cold plate optimization design method and apparatus to solve the problem of how to improve the heat dissipation efficiency of immersion liquid cooling with directional centralized cooling when optimizing the design of the spade-shaped cold plate of the immersion heat dissipation system.

[0006] In a first aspect, this disclosure provides a method for optimizing the design of cold plates, the method comprising: using a cold plate modeling tool to generate a set of cold plate models based on a sample set of initial design parameters of the cold plates;

[0007] Using liquid cooling simulation tools, a liquid cooling simulation dataset is generated based on the coolant type, coolant flow rate, and cold plate model set. The liquid cooling simulation data in the dataset includes: cold plate cooling device temperature, immersion cooling device temperature, and system flow resistance.

[0008] Using a multi-island genetic algorithm, with the system pressure drop not falling below a pressure drop threshold as a constraint, and the tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameters of the cold plate corresponding to the liquid cooling simulation data in the liquid cooling simulation dataset as optimization parameter variables, and minimizing the temperature of the cold plate cooling device and the temperature of the immersion cooling device in the liquid cooling simulation data as the first objective function, an optimized design parameter combination and the coolant flow rate of the corresponding optimized design parameter combination are generated; the system pressure drop is determined based on the system flow resistance.

[0009] From the optimized design parameter combinations, the optimized design parameters that meet the design requirements are selected as the target cold plate design parameters, and the target coolant flow rate of the corresponding target cold plate design parameters is determined based on the coolant flow rate of the optimized design parameters in the corresponding optimized design parameter combinations.

[0010] The cold plate optimization design method provided in this embodiment can accurately evaluate the impact of different cold plate design parameters on the temperature of the cold plate cooling device, the temperature of the immersion cooling device, and the system flow resistance through liquid cooling simulation data generated by a liquid cooling simulation tool. Optimization using a multi-island genetic algorithm can find the optimal combination of design parameters that minimizes the temperatures of the cold plate cooling device and the immersion cooling device, corresponding to the target coolant flow rate. This optimizes the determined target cold plate design parameters, thereby significantly improving the heat dissipation efficiency of immersion liquid cooling with directional centralized cooling of the cold plate, ensuring that the cooled electronic equipment or system maintains a low operating temperature even under high loads.

[0011] Secondly, this disclosure provides a cold plate optimization design device, the device comprising:

[0012] The model generation module is used to generate a set of cold plate models based on the initial design parameter sample set of the cold plate using the cold plate modeling tool.

[0013] The liquid cooling simulation module is used to generate a liquid cooling simulation dataset based on the coolant type, coolant flow rate, and cold plate model set using liquid cooling simulation tools. The liquid cooling simulation data in the dataset includes: the temperature of the cold plate cooling device, the temperature of the immersion cooling device, and the system flow resistance.

[0014] The parameter optimization module utilizes a multi-island genetic algorithm, with the system pressure drop not falling below a pressure drop threshold as a constraint. It uses the tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameters of the cold plate corresponding to the liquid cooling simulation data in the liquid cooling simulation dataset as optimization parameter variables. The first objective function is to minimize the temperature of the cold plate cooling device and the temperature of the immersion cooling device in the liquid cooling simulation data. This generates an optimized design parameter combination and the corresponding coolant flow rate for the optimized design parameters in the optimized design parameter combination. The system pressure drop is determined based on the system flow resistance.

[0015] The parameter selection module is used to select the optimized design parameters that meet the design requirements from the optimized design parameter combination as the target cold plate design parameters, and to determine the target coolant flow rate of the corresponding target cold plate design parameters based on the coolant flow rate of the optimized design parameters in the corresponding optimized design parameter combination.

[0016] Thirdly, this disclosure provides a computer device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the cold plate optimization design method of the first aspect or any corresponding embodiment described above.

[0017] Fourthly, this disclosure provides a computer-readable storage medium storing computer instructions for causing a computer to execute the cold plate optimization design method of the first aspect or any corresponding embodiment described above. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the specific embodiments of this disclosure, the drawings used in the description of the specific embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1a This is an exemplary schematic diagram of a cold plate directional centralized cooling immersion liquid-cooled server according to an embodiment of the present disclosure;

[0020] Figure 1b This is a flowchart illustrating the cold plate optimization design method according to an embodiment of the present disclosure;

[0021] Figure 1c This is a schematic diagram of the modeling and simulation process of the cold plate optimization design method according to the embodiments of this disclosure;

[0022] Figure 1d This is an exemplary schematic diagram of the cold plate spade tooth structure parameters according to an embodiment of the present disclosure;

[0023] Figure 1e This is an exemplary schematic diagram of the combination of optimization design parameters represented in the form of a Pareto curve, obtained by the cold plate optimization design method according to the embodiments of this disclosure.

[0024] Figure 2a This is a flowchart illustrating another cold plate optimization design method according to an embodiment of the present disclosure;

[0025] Figure 2b This is a flowchart illustrating the process of training and optimizing design parameters based on an agent model according to an embodiment of this disclosure.

[0026] Figure 3 This is a structural block diagram of a cold plate optimization design apparatus according to an embodiment of the present disclosure;

[0027] Figure 4 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present disclosure. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0029] In recent years, with the acceleration of digital transformation, the amount of data generated has grown exponentially, leading to a continuous increase in the demand for computing power in data centers and servers. To address the energy consumption and heat dissipation issues arising from high-performance computing tasks and massive data processing, server liquid cooling technology has developed rapidly. Compared to traditional air cooling, liquid cooling offers advantages such as low energy consumption, high heat dissipation, and low noise, making it particularly advantageous in scenarios with high server computing power and high deployment density. The efficient heat dissipation of liquid cooling technology can effectively improve server efficiency and stability, keeping hardware devices within their normal operating temperature range and enhancing server computing efficiency.

[0030] Liquid cooling technology primarily utilizes immersion cooling and cold plate cooling. Depending on whether the coolant undergoes a phase change, both methods can be categorized as single-phase or two-phase. Single-phase immersion and single-phase cold plate technologies are widely used in data centers due to their lower cost, higher reliability, and easier maintenance. For single-phase immersion liquid cooling, to address the issue of uneven heat dissipation, many studies have proposed directional, centralized cooling methods. Specifically, during the liquid circulation process, the cold liquid, after entering the cooling tank, first flows through devices with higher heat dissipation requirements, such as CPUs and GPUs, and then sequentially flows through devices with lower heat dissipation requirements, thereby improving heat dissipation efficiency.

[0031] This design is currently widely used in high-power devices such as CPUs and GPUs, where the coolant first flows through the coolant plate, exits from the plate, and then flows through the remaining devices. Currently, for single-phase coolants, the inline serrated type is widely used due to its simple structure and good heat dissipation performance.

[0032] In a specific example, you can refer to Figure 1a , Figure 1a This diagram illustrates an exemplary embodiment of a liquid-cooled server using a cold plate for directional centralized cooling. The main components of this server system include a CPU, memory, hard drive, and power supply. Figure 1aIn the example shown, because the CPU has high heat dissipation requirements, a cold plate is designed on the CPU. When the cold liquid enters the chassis from the refrigerant heat exchange unit, it first flows through the two cold plates, carrying away the heat from the CPU, and then flows out from the cold plates into the chassis to immerse and cool other components with relatively low heat dissipation requirements, such as memory and hard drives. This achieves the aforementioned goal of directional and concentrated cooling of high-demand heat dissipation components before cooling the other components.

[0033] In the process of using this type of cold plate for directional centralized cooling in a single-phase immersion liquid cooling system, the most common practice at present is to directly reuse the cold plate used in pure cold plate servers. Under certain operating conditions, the flow channel of the cold plate may be fine-tuned to determine the final design scheme while simultaneously meeting the requirements of heat dissipation and flow resistance.

[0034] However, as liquid cooling becomes increasingly sophisticated, the current optimization design methods for immersion cold plates are too crude, which contradicts the development trend of liquid cooling. The reasons are as follows: 1. In the process of designing the cold plate flow channel for pure cold plate servers, the cooling medium used is generally deionized water, propylene glycol, and other liquids. These liquids have low viscosity. Therefore, in the design of the toothed flow channel, in order to improve the heat dissipation capacity of the cold plate, the tooth arrangement in the cold plate flow channel can be designed to be relatively dense. However, this design is not suitable for immersion systems because the coolant used in immersion systems is generally fluorinated liquid or oil. The viscosity of these liquids is much higher than that of deionized water. If this design is reused, there will be a problem of excessive flow resistance. In addition, when the liquid is different, in order to ensure both heat dissipation capacity and flow resistance, different requirements will be put forward for the height and thickness parameters of the tooth. 2. For pure cold plate servers, in the process of cold plate optimization design, the optimization target only considers the temperature of the components attached to the cold plate. After using the cold plate heat sink in the immersion system, in the optimization design process, in addition to considering the temperature of the components attached to it, the temperature of other immersion components in the chassis should also be considered. That is, a new approach needs to be proposed for the entire optimization process and the definition of optimization targets.

[0035] Currently, research conducted by universities and enterprises on applying cold plate heat sinks to single-phase immersion systems for directional centralized heat dissipation primarily utilizes toothed cold plate designs that directly reuse pure cold plate server designs. These designs employ plasma water as the cooling medium, or are based on these with minor adjustments to suit the coolant's properties. There is currently no optimized toothed cold plate design method specifically for high-power devices in immersion systems. The problem with this approach is that while it may achieve cold plate designs that meet heat dissipation and flow resistance requirements, it may not yield the optimal balance between these requirements in a comprehensive immersion-integrated cold plate design. As liquid cooling technology becomes increasingly sophisticated, this crude design approach is no longer adequate to meet the evolving needs of liquid cooling technology. Optimized design methods for toothed cold plates in immersion systems are currently lacking.

[0036] This disclosure provides a method for optimizing the design of a shovel-shaped cold plate in an immersion liquid cooling system to improve the heat dissipation efficiency of the immersion liquid cooling system.

[0037] According to an embodiment of this disclosure, a method for optimizing the design of a cold plate is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0038] This embodiment provides a cold plate optimization design method, which can be used in the above-mentioned computer systems, such as mobile terminals, such as mobile phones, tablets, portable computers, etc., or fixed terminals, such as desktop computers, etc. Figure 1b This is a flowchart of a cold plate optimization design method according to an embodiment of the present disclosure, such as... Figure 1b As shown, the process includes the following steps:

[0039] Step S101: Using the cold plate modeling tool, generate a cold plate model set based on the initial design parameter sample set of the cold plate.

[0040] In this embodiment, the cold plate modeling tool can be thermal analysis software (ANSYS Icepak), thermal management design tools (Boyd SmartCFD), computer-aided design (CAD) software (such as SolidWorks, AutoCAD), and computational fluid dynamics (CFD) software. Cold plate modeling tools typically possess powerful geometric modeling, mesh generation, and physics field solving capabilities. The model framework can be determined considering the design characteristics and functional requirements of the cold plate, and may include geometry, size range, boundary conditions, etc.

[0041] The initial design parameter sample set for cold-rolled steel plates can be collected from various sources (such as design documents, experimental records, CAD drawings, etc.). These parameters may include the shape of the spade teeth (such as angle, width, and height), layout (such as quantity and spacing), material properties (such as thermal conductivity and specific heat capacity), and working environment conditions (such as temperature and pressure).

[0042] In some specific examples, a sample range of design parameters for the cold plate spade teeth can be given, and then sampling can be performed based on the sample range of design parameters to generate an initial design parameter sample set for the cold plate.

[0043] Using the parametric design function in the cold plate modeling tool, cold plate design parameters (such as serration shape, size, and layout) can be set as variable parameters. Then, these parameters can be changed based on the initial cold plate design parameter sample set to generate a series of cold plate models.

[0044] In a specific example, considering the automatic modeling of cold plates based on the initial design parameter samples of each cold plate in the initial design parameter sample set, and the output of a cold plate model set, a macro command script for 3D modeling software can be established. The function to be implemented here is the automatic modeling of the cold plate spade tooth flow channel, and then assembling it into the existing server model.

[0045] Step S102: Using a liquid cooling simulation tool, generate a liquid cooling simulation data set based on the coolant type, coolant flow rate, and cold plate model set.

[0046] In this embodiment, the liquid cooling simulation data in the liquid cooling simulation dataset includes: the temperature of the cold plate cooling device, the temperature of the immersion cooling device, and the system flow resistance.

[0047] The type of coolant can be determined based on the application scenario and requirements of the cold plate. The coolant flow rate can be determined based on the requirements and design specifications of the cooling system, and can be a coolant flow rate range corresponding to the design parameters of the cold plate or more than one coolant flow rate.

[0048] The temperature of a cold plate-cooled device refers to the temperature of a device that performs complex calculations, processes large amounts of data, or requires high power operation. It is also the temperature of devices that use a cold plate for heat dissipation, such as the temperature of a CPU or GPU.

[0049] Immersion cooling device temperature refers to the temperature of other components being cooled besides the cold plate cooling device, that is, the temperature of components immersed in coolant, such as memory and hard drive.

[0050] System flow resistance refers to the resistance encountered by a fluid when flowing through a system. It is an important parameter for measuring the ease or difficulty of fluid flow and can include the inlet and outlet flow resistance of the system. The magnitude of system flow resistance is affected by a variety of factors, including the properties of the fluid (such as viscosity and density), the flow state (laminar or turbulent), the geometry of the pipe or flow channel, the roughness of the inner wall, and the flow velocity.

[0051] Since the cold plate models in the cold plate model set reflect the actual physical structure of the cold plate, liquid cooling simulation tools such as Star CCM+, Solidworks Flow Simulation, and Ansys Fluent can be used to simulate the application of each cold plate model in the liquid cooling system, thereby generating liquid cooling simulation data for each cold plate model and obtaining the liquid cooling simulation data set for the corresponding cold plate model set.

[0052] In a specific example, simulation calculations need to be performed on all sample point models of the initial design parameters of the cold plates, and the simulation results of cooling the server model using the cold plate model need to be output. Therefore, a macro script needs to be established in the liquid cooling simulation software to execute the mesh generation and simulation calculation of the server model, complete the simulation of the sample point models of the initial design parameters of each cold plate, and then extract the parameters of interest from the simulation calculation results. Specifically, based on the configuration of the selected server model, the parameters of interest are CPU temperature, memory temperature, and system current resistance. Here, memory temperature is chosen as the standard for measuring the temperature of the immersed device. The main reason is that among the immersed devices, the memory device has a relatively large heat dissipation and is considered a high-risk item for heat dissipation. Therefore, memory temperature is chosen as the target.

[0053] Using a general-purpose programming language like Python, macro commands from various software components can be integrated and executed step-by-step. This also requires integrating the cold plate initial design parameter sample point matrix mentioned in the previous steps. Based on this matrix, the macro commands mentioned in step S101 are invoked to obtain multiple different cold plate initial design parameter sample point models. Then, the simulation calculation macro script is executed, concluding the automatic modeling and simulation process. For details, please refer to... Figure 1c , Figure 1c A schematic flowchart illustrating the modeling and simulation process of the cold plate optimization design method according to an embodiment of the present disclosure is shown.

[0054] Step S103: Using the multi-island genetic algorithm, with the system pressure drop not lower than the pressure drop threshold as a constraint, the tooth height, tooth width, tooth spacing and coolant flow rate of the initial design parameters of the cold plate corresponding to the liquid cooling simulation data in the liquid cooling simulation data set are used as optimization parameter variables, and minimizing the temperature of the cold plate cooling device and the temperature of the immersion cooling device in the liquid cooling simulation data is used as the first objective function to generate the optimized design parameter combination and the coolant flow rate of the optimized design parameters in the corresponding optimized design parameter combination.

[0055] In this embodiment, the system pressure drop is the pressure consumed by the coolant to overcome the resistance of the pipeline or cooling equipment during the flow process, and can be determined based on the system flow resistance.

[0056] The aforementioned execution entity can utilize the Multi-Island Genetic Algorithm (MIGA), with the system voltage drop not falling below a voltage drop threshold as a constraint, and minimizing the temperatures of the cold plate cooling device and the immersion cooling device in the liquid cooling simulation data as the first objective function. By dividing the population of optimization parameter variables into multiple "islands," the genetic algorithm runs independently on each island, and individuals are periodically exchanged between islands to generate optimization results, thereby improving the diversity and global adaptability of the optimization results.

[0057] In a specific example, optimization methods such as multi-island genetic algorithms can be used to select optimization parameter variables, such as... Figure 1d The schematic diagram of the cold plate tooth structure shows the height h, width s, and spacing l of the teeth, as well as the flow velocity v, as optimization parameters. Based on the server configuration, the optimization targets can be selected as the CPU device temperature cooled by the cold plate and the memory temperature under immersion cooling. The reason is that, for the immersion components in the server, memory has the highest heat flux density and is considered a high-risk item for heat dissipation; therefore, the memory temperature is used as the objective function for the immersion components. Furthermore, since the cooling liquid used for immersion has a high viscosity, flow resistance must be considered while taking into account the heat dissipation effect of the cold plate. Therefore, during the optimization process, the pressure drop parameter of the entire system is used as a constraint variable to ensure that the pressure does not fall below a certain value according to actual needs.

[0058] It should be noted that: there may be 2-way, 4-way or even more CPUs in the server, so there will be multiple CPU temperatures; each memory module will have multiple chips, and each chip will have a temperature, so each memory module will also have multiple temperatures. Therefore, it is necessary to determine which temperature parameter to choose as the optimization target. One possible solution is as follows: (1) choose the average value of each temperature as the benchmark; (2) choose the chip under the worst heat dissipation environment as the benchmark. If the worst case has a large temperature margin, then there will be no heat dissipation risk in other operating conditions.

[0059] Subsequently, dual-objective optimization can be performed on the temperatures of the cold plate cooling device and the immersion cooling device, as well as the system voltage drop, to obtain a Pareto curve. The horizontal axis of the Pareto curve displays the temperature of the cold plate cooling device, and the vertical axis displays the temperature of the immersion cooling device. In... Figure 1e In the Pareto curve diagram shown, the horizontal and vertical axes represent the central processing unit (CPU) temperature and the double data rate synchronous dynamic random access memory (DIMM) temperature, respectively. It can be seen that their temperature trends are opposite; that is, the lower the temperature of the cold plate cooling device, the higher the temperature of the immersion cooling device, or vice versa.

[0060] Therefore, among the optimal design parameter combinations represented by Pareto curves, the design parameters that meet the design requirements are selected as the target cold plate design parameters.

[0061] Step S104: In the optimized design parameter combination, select the optimized design parameter that meets the design requirements as the target cold plate design parameter, and in the coolant flow rate of the optimized design parameter in the corresponding optimized design parameter combination, determine the target coolant flow rate of the target cold plate design parameter.

[0062] In this embodiment, the temperatures of the cold plate cooling device and the immersion cooling device corresponding to the optimized design parameter combination have opposite trends. That is, the lower the temperature of the cold plate cooling device, the higher the temperature of the immersion cooling device, or the higher the temperature of the cold plate cooling device, the lower the temperature of the immersion cooling device.

[0063] Design requirements can be determined based on the user's actual needs. For example, they may include pre-set allowable temperature ranges for cold plate cooling devices and immersion cooling devices.

[0064] Therefore, the aforementioned implementing entity can select the optimized design parameters from the optimized design parameter combination as the target cold plate design parameters according to the design requirements, and determine the target coolant flow rate of the corresponding target cold plate design parameters from the coolant flow rate of the corresponding optimized design parameters.

[0065] The cold plate optimization design method provided in this embodiment can accurately evaluate the impact of different cold plate design parameters on the temperature of the cold plate cooling device, the temperature of the immersion cooling device, and the system flow resistance through liquid cooling simulation data generated by a liquid cooling simulation tool. Optimization using a multi-island genetic algorithm can find the optimal combination of design parameters that minimizes the temperatures of the cold plate cooling device and the immersion cooling device, corresponding to the target coolant flow rate. This optimizes the determined target cold plate design parameters, thereby significantly improving the heat dissipation efficiency of immersion liquid cooling with directional centralized cooling of the cold plate, ensuring that the cooled electronic equipment or system maintains a low operating temperature even under high loads.

[0066] In some optional implementations of this embodiment, the initial design parameter sample set of the cold plate is determined based on the following steps: obtaining the structural parameter range of the cold plate shovel teeth, including the tooth height range, tooth width range, and tooth pitch range; obtaining the coolant inlet flow range; and using an experimental design method, generating the initial design parameter sample set of the cold plate based on the structural parameter range of the cold plate shovel teeth and the coolant inlet flow range, wherein the initial design parameter samples in the initial design parameter sample set of the cold plate include the tooth height, tooth width, tooth pitch, and coolant inlet flow.

[0067] In this implementation, during the optimization design of the cold plate, the selected coolant is first determined, and the optimization variable is selected as follows: Figure 1d The schematic diagram of the cold plate tooth structure parameters shows the height h (in mm), width s (in mm), and spacing l (in mm) of the tooth. Simultaneously, since the heat dissipation effect and flow resistance of the cold plate are also related to the liquid flow rate, and the flow velocity and tooth structure are interrelated, jointly determining the performance of the cold plate, the flow rate m (in L / min) at the liquid inlet pipe is also selected for parameter optimization. A cold plate design parameter sample set is established based on experimental design methods such as optimal Latin square sampling, uniform design, or orthogonal design. It is understood that other experimental design methods can also be used, such as completely randomized design or randomized block design; this disclosure does not limit these methods.

[0068] In a specific example, the range of h can be set to [3, 7], the range of s to [0.2, 0.6], the range of l to [0.2, 0.6], and the range of m to [2, 6]. Based on the optimal Latin square sampling, 40 initial design parameter samples of cold plates with 4 factors are established, as shown in Table 1 below. These correspond to 40 initial design parameter sample points of cold plates, and simulation needs to be completed for these 40 cold plate design parameter sample points.

[0069] Table 1 - Sample set generated based on optimal Latin square sampling:

[0070]

[0071]

[0072] In some optional implementations of this embodiment, a multi-island genetic algorithm is used. Based on the liquid cooling simulation data set, with the constraint that the system pressure drop determined based on the system flow resistance is not lower than the pressure drop threshold, and with the tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameters of the cold plate corresponding to the liquid cooling simulation data as optimization parameter variables, the first objective function is to minimize the temperature of the cold plate cooling device and the temperature of the immersion cooling device in the liquid cooling simulation data. The optimization results are generated by: using the tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameters of the cold plate corresponding to the liquid cooling simulation data as optimization parameter variables, performing an initialization population on the set of optimization parameter variables corresponding to the liquid cooling simulation data set to obtain multiple subpopulations. The optimization parameter variables include: tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameters of the cold plate; and performing independent standard genetic operations and inter-island migration operations on each subpopulation to obtain genetic subpopulations. The standard genetic operations include: selection, crossover, and variation. The process involves minimizing the temperatures of the cold plate cooling device and the immersion cooling device in the liquid cooling simulation data as the first objective function. The fitness of the new optimized parameter variables is calculated based on the corresponding temperatures of the cold plate cooling device and the immersion cooling device in the liquid cooling simulation data. The system pressure drop is determined based on the system flow resistance of the new optimized parameter variables in the liquid cooling simulation data. The following two constraints are used to guide the iteration: fitness greater than a fitness threshold and system pressure drop not lower than a pressure drop threshold. Independent standard genetic operations and inter-island migration operations are performed on each subpopulation. The optimal optimized parameter variables and their corresponding cold plate cooling device and immersion cooling device temperatures, as well as the system pressure drop determined based on the system flow resistance, are recorded for each iteration until a predetermined number of iterations is reached or the objective function meets the convergence condition. This process yields the optimized design parameter combination and the coolant flow rate of the optimized design parameters in the corresponding optimized design parameter combination.

[0073] In this implementation, the weighted sum of the temperatures of the cold plate cooling device and the immersion cooling device can be used as the objective function; the negative of the objective function can be used as the fitness function; and the fitness function is used to calculate the fitness of the cold plate cooling device temperature and the immersion cooling device temperature in the liquid cooling simulation data corresponding to the new optimization parameter variables in the genetic subpopulation.

[0074] In some optional examples, the multi-island genetic algorithm can also be updated. Specifically, the target cold plate can be manufactured using the target cold plate design parameters; immersion cooling tests can be performed using the target cold plate to obtain the actual tested temperatures of the cold plate cooling device and the immersion cooling device; based on the difference between the actual tested temperatures of the cold plate cooling device and the immersion cooling device and the temperatures of the cold plate cooling device and the immersion cooling device in the liquid cooling simulation data corresponding to the target cold plate design parameters, the multi-island genetic algorithm can be updated to optimize the combination of design parameters.

[0075] This embodiment provides a cold plate optimization design method, which can be used in the aforementioned execution entity. Figure 2a This is a flowchart of a cold plate optimization design method according to an embodiment of the present disclosure, such as... Figure 2a As shown, the process includes the following steps:

[0076] Step S201: Using the cold plate modeling tool, generate a cold plate model set based on the initial design parameter sample set of the cold plate.

[0077] In this embodiment, the parametric design function in the cold plate modeling tool can be used to set the design parameters of the cold plate (such as the shape, size, and layout of the spade teeth) as variable parameters. Then, these parameters can be changed according to the initial design parameter sample set of the cold plate to generate a series of cold plate models.

[0078] For details, please refer to step S101 of the embodiment shown in Figure 1, which will not be repeated here.

[0079] Step S202: Using a liquid cooling simulation tool, generate a liquid cooling simulation data set based on the coolant type, coolant flow rate, and cold plate model set.

[0080] In this embodiment, since the cold plate models in the cold plate model set embody the actual physical structure of the cold plate, liquid cooling simulation tools such as Star CCM+, Solidworks Flow Simulation, and Ansys Fluent can be used to simulate the application of each cold plate model in the liquid cooling system, thereby generating liquid cooling simulation data corresponding to each cold plate model and obtaining the liquid cooling simulation data set of the corresponding cold plate model set.

[0081] For details, please refer to step S102 of the embodiment shown in Figure 1, which will not be repeated here.

[0082] Step S203: Using a proxy model, expand the initial design parameter sample of the cold plate and the corresponding liquid cooling simulation data.

[0083] In this embodiment, the surrogate model collects data points by sampling the original model a limited number of times, and then constructs a simpler, less computationally expensive mathematical model based on these data points. This model can quickly predict the output of the original model at unsampled points, thereby enabling efficient search and optimization within the design space.

[0084] The types of surrogate models can include, but are not limited to, multinomial regression, radial basis functions, kriging models, artificial neural networks, and support vector machines. Those skilled in the art can write their own surrogate model training code, or use the surrogate model training module built into commercial software, taking the initial design parameters of the cold plate as input and the liquid cooling simulation data of the corresponding initial design parameters of the cold plate obtained from simulation calculations as output, to train the surrogate model.

[0085] In a specific example, 40 initial design parameter sample points for the cold plate are formed for 4 optimization variables. During the optimization process, the optimization is based on these 40 sample points. Therefore, these 40 sample points are discrete. In some optimization cases, if these sample points are relatively sparse, it will lead to a certain bias in the optimization process. In such cases, if a very accurate optimization result is obtained, the number of sample points needs to be increased. For example, 100 sample points could be designed for these 4 optimization variables. This would increase the simulation workload and take a lot of time for the optimization process. However, by using a surrogate model training method to fit the existing sample points, an accurate sample set can be obtained while significantly reducing the simulation workload and simulation time.

[0086] Step S204: Using the multi-island genetic algorithm, with the system pressure drop not lower than the pressure drop threshold as a constraint, the tooth height, tooth width, tooth spacing and coolant flow rate of the initial design parameters of the cold plate corresponding to the liquid cooling simulation data in the liquid cooling simulation data set are used as optimization parameter variables, and minimizing the temperature of the cold plate cooling device and the temperature of the immersion cooling device in the liquid cooling simulation data is used as the first objective function to generate the optimized design parameter combination and the coolant flow rate of the optimized design parameters in the corresponding optimized design parameter combination.

[0087] In this embodiment, the aforementioned execution entity can utilize a Multi-Island Genetic Algorithm (MIGA), with the system voltage drop not lower than a voltage drop threshold as a constraint, and minimizing the temperatures of the cold plate cooling device and the immersion cooling device in the liquid cooling simulation data as the first objective function. By dividing the population of optimization parameter variables into multiple "islands," the genetic algorithm runs independently on each island, and individuals are periodically exchanged between islands to generate optimization results, thereby improving the diversity and global adaptability of the optimization results.

[0088] For details, please refer to step S103 of the embodiment shown in Figure 1, which will not be repeated here.

[0089] Step S205: Select the optimized design parameters that meet the design requirements from the optimized design parameter combinations as the target cold plate design parameters, and determine the coolant flow rate of the corresponding target cold plate design parameters based on the coolant flow rate of the optimized design parameters in the corresponding optimized design parameter combinations.

[0090] In this embodiment, the design requirements can be determined according to the user's actual needs, such as including pre-set allowable temperature ranges for cold plate cooling devices and immersion cooling devices. Therefore, the execution entity can select optimized design parameters from the optimized design parameter combination as target cold plate design parameters according to the design requirements, and determine the target coolant flow rate corresponding to the target cold plate design parameters from the coolant flow rates corresponding to the optimized design parameters.

[0091] For details, please refer to step S103 of the embodiment shown in Figure 1, which will not be repeated here.

[0092] The cold plate optimization design method provided in this embodiment expands the initial design parameter sample of the cold plate and the corresponding liquid cooling simulation data through a proxy model. Then, it uses a multi-island genetic algorithm for optimization, further expanding the optimal design parameter combination corresponding to minimizing the temperature of the cold plate cooling device and the temperature of the immersion cooling device, and the corresponding target coolant flow rate. This further optimizes the determined target cold plate design parameters, which can significantly improve the heat dissipation efficiency of immersion liquid cooling with directional centralized cooling of the cold plate, and ensure that the cooled electronic equipment or system can maintain a low operating temperature under high load.

[0093] In some optional implementations of this embodiment, the surrogate model can be trained using the following steps: taking the initial design parameter samples of the cold plate in the initial design parameter sample set as input, taking the liquid cooling simulation data corresponding to the initial design parameter samples of the cold plate as output, taking minimizing the difference between the discrete point set corresponding to the initial design parameter sample set of the cold plate and the predicted surface of the surrogate model as the second objective function, training the surrogate model, and outputting the surrogate model when the fitting accuracy between the discrete point set corresponding to the initial design parameter sample set of the cold plate and the predicted surface of the surrogate model reaches the accuracy threshold.

[0094] Here, during the training of the surrogate model, the decision to output the surrogate model is based on the fitting accuracy. If the accuracy is sufficient, the surrogate model is output, which improves the expression accuracy of the surrogate model and thus improves the accuracy of the represented sample points. This allows the subsequent optimization process to be executed within the range of the expanded cold plate initial design parameter sample set and the corresponding liquid cooling simulation dataset obtained by the surrogate model prediction.

[0095] In some optional implementations of this embodiment, the method further includes: using a cold plate modeling tool to generate a target cold plate model based on the target cold plate design parameters; using a liquid cooling simulation tool to generate liquid cooling simulation data of the target cold plate model based on the coolant type, coolant flow rate, and the target cold plate model; using a surrogate model to determine the liquid cooling prediction data corresponding to the target cold plate design parameters; and if the difference between the liquid cooling simulation data and the liquid cooling prediction data of the target cold plate model is less than a difference threshold, then the target cold plate design parameters are output as the optimization result.

[0096] In this implementation, after determining the design parameters of the target cold plate, it is also necessary to perform simulation verification on the design parameters of the target cold plate. If the error between the liquid cooling simulation result and the prediction result is within a certain allowable range, the result is output, thereby optimizing the determined optimization result. This can significantly improve the heat dissipation efficiency of immersion liquid cooling with directional centralized cooling of the cold plate corresponding to the optimization result, and ensure that the cooled electronic equipment or system can maintain a low operating temperature under high load.

[0097] In some optional implementations of this embodiment, the method further includes: if the difference is greater than the difference threshold, updating the target cold plate design parameters and the liquid cooling simulation data corresponding to the target cold plate design parameters to the cold plate initial design parameter sample set and the liquid cooling simulation data set corresponding to the cold plate initial design parameter sample set, and retraining the surrogate model using the updated cold plate initial design parameter sample set and liquid cooling simulation data set.

[0098] In this implementation, if the error between the simulation result and the prediction result is too high, the target cold plate design parameters and the corresponding liquid cooling simulation data of the target cold plate design parameters need to be expanded to the initial design parameter sample set of the cold plate. The process is then returned to the surrogate model training step, and the surrogate model is retrained until the prediction accuracy of the target cold plate design parameters in the output optimized design parameter combination meets the requirements. Then, the final optimization result is output. This approach achieves accurate optimization results with fewer initial cold plate design parameter samples and surrogate model predictions, while reducing simulation workload and saving simulation time. Furthermore, it improves the heat dissipation efficiency of the immersion liquid cooling with directional centralized cooling of the cold plate corresponding to the final optimization result, ensuring that the cooled electronic equipment or system can maintain a low operating temperature even under high load.

[0099] For example, in a specific case, if a surrogate model is used for prediction, with four parameters for optimization, the number of sample points can be reduced to 30 or even 20. Simultaneously, the range of values ​​for the optimization variables can be broadened. This is because even if the sample points are sparsely distributed, as long as the surrogate model's training accuracy is sufficient, it can still fit a smooth spatial surface, achieving precise expansion of the sample points. For details, please refer to [link to relevant documentation]. Figure 2b , Figure 2bA schematic diagram of the process for training and optimizing design parameters based on a proxy model is shown.

[0100] In some optional implementations of this embodiment, the above-mentioned cold plate optimization design method further includes: manufacturing a target cold plate using the target cold plate design parameters; performing an immersion cooling test using the target cold plate to obtain the actual tested cold plate cooling device temperature and the immersion cooling device temperature; and updating the multi-island genetic algorithm to update the combination of optimization design parameters based on the difference between the actual tested cold plate cooling device temperature and the immersion cooling device temperature and the cold plate cooling device temperature and the immersion cooling device temperature of the liquid cooling simulation data corresponding to the target cold plate design parameters.

[0101] In this implementation, to update the multi-island genetic algorithm to optimize the combination of design parameters, cold plates can be manufactured using the target cold plate design parameters, and actual tests can be conducted to verify the effectiveness of the optimization results. If necessary, the optimization algorithm or objective function can be adjusted based on the actual test results to further optimize the cold plate design.

[0102] The cold plate optimization design method in this implementation clarifies the steps for updating the multi-island genetic algorithm, thereby improving the effectiveness of the multi-island genetic algorithm and, consequently, the effectiveness of the optimized design parameter combination updated using the multi-island genetic algorithm.

[0103] This embodiment also provides a cold plate optimization design device, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0104] This embodiment provides a device for optimizing the design of cold-plates, such as... Figure 3 As shown, it includes:

[0105] The model generation module 301 is used to generate a set of cold plate models based on the initial design parameter sample set of the cold plate using the cold plate modeling tool.

[0106] The liquid cooling simulation module 302 is used to generate a liquid cooling simulation data set based on the coolant type, coolant flow rate and cold plate model set using liquid cooling simulation tools. The liquid cooling simulation data in the liquid cooling simulation data set includes: the temperature of the cold plate cooling device, the temperature of the immersion cooling device and the system flow resistance.

[0107] The parameter optimization module 303 uses a multi-island genetic algorithm, with the system pressure drop not lower than a pressure drop threshold as a constraint, and the tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameters of the cold plate corresponding to the liquid cooling simulation data in the liquid cooling simulation dataset as optimization parameter variables. The first objective function is to minimize the temperature of the cold plate cooling device and the temperature of the immersion cooling device in the liquid cooling simulation data. The module generates an optimized design parameter combination and the coolant flow rate of the optimized design parameters in the corresponding optimized design parameter combination. The system pressure drop is determined based on the system flow resistance.

[0108] The parameter selection module 304 is used to select the optimized design parameters that meet the design requirements from the optimized design parameter combination as the target cold plate design parameters, and to determine the target coolant flow rate of the corresponding target cold plate design parameters based on the coolant flow rate of the optimized design parameters in the corresponding optimized design parameter combination.

[0109] In some optional implementations of this embodiment, the initial design parameter sample set of the cold plate in the model generation module 301 is determined based on the following modules:

[0110] The parameter acquisition module is used to acquire the structural parameter range of the cold plate shovel teeth. The structural parameter range includes: tooth height range, tooth width range, and tooth pitch range.

[0111] The flow acquisition module is used to acquire the coolant inlet flow range;

[0112] The parameter generation module is used to generate a sample set of initial design parameters for the cold plate based on the structural parameter range of the cold plate teeth and the coolant inlet flow range using the experimental design device. The initial design parameters of the cold plate in the sample set include: tooth height, tooth width, tooth spacing and coolant inlet flow.

[0113] In some optional implementations of this embodiment, the apparatus further includes:

[0114] The parameter expansion module uses a proxy model to expand the initial design parameter samples of the cold plate and the corresponding liquid cooling simulation data in the initial design parameter sample set of the cold plate.

[0115] In some optional implementations of this embodiment, the proxy model in the parameter augmentation module is trained using the following modules:

[0116] The model training module takes the initial design parameters of the cold plate from the initial design parameter sample set as input and the liquid cooling simulation data corresponding to the initial design parameter samples as output. The second objective function is to minimize the difference between the discrete point set corresponding to the initial design parameter sample set and the predicted surface of the surrogate model. The surrogate model is trained, and when the fitting accuracy between the discrete point set corresponding to the initial design parameter sample set and the predicted surface of the surrogate model reaches the accuracy threshold, the surrogate model is output.

[0117] In some optional implementations of this embodiment, the apparatus further includes:

[0118] The cold plate modeling module is used to generate a target cold plate model based on the target cold plate design parameters using cold plate modeling tools.

[0119] The liquid cooling simulation module is further used to generate liquid cooling simulation data of the target cold plate model based on the coolant type, coolant flow rate and target cold plate model using liquid cooling simulation tools;

[0120] The liquid cooling prediction module is used to determine the liquid cooling prediction data corresponding to the design parameters of the target cold plate using a surrogate model.

[0121] The parameter output module is used to output the target cold plate design parameters as the optimization result if the difference between the liquid cooling simulation data and the liquid cooling prediction data of the target cold plate model is less than the difference threshold.

[0122] In some optional implementations of this embodiment, the apparatus further includes:

[0123] The model optimization module is used to update the target cold plate design parameters and the corresponding liquid cooling simulation data to the cold plate initial design parameter sample set and the corresponding liquid cooling simulation data set if the difference is greater than the difference threshold. The surrogate model is then retrained using the updated cold plate initial design parameter sample set and liquid cooling simulation data set.

[0124] In some optional implementations of this embodiment, the parameter optimization module includes:

[0125] The initialization group module is used to initialize the group based on the tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameter sample of the cold plate corresponding to the liquid cooling simulation data. This initialization group is used to obtain multiple subgroups. The optimization parameter variables include the tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameter sample of the cold plate.

[0126] The subpopulation operation module is used to perform independent standard genetic operations and inter-island migration operations on each subpopulation to obtain genetic subpopulations. The standard genetic operations include selection, crossover, and mutation.

[0127] The fitness calculation module is used to calculate the fitness of the new optimization parameter variables based on the temperatures of the cold plate cooling device and the immersion cooling device in the liquid cooling simulation data, with the first objective function being to minimize the temperatures of the cold plate cooling device and the immersion cooling device in the liquid cooling simulation data corresponding to the new optimization parameter variables in the genetic subpopulation.

[0128] The pressure drop determination module is used to determine the system pressure drop based on the system flow resistance corresponding to the liquid cooling simulation data of the new optimized parameter variables in the genetic subpopulation.

[0129] The results recording module is used to guide the iteration to perform independent standard genetic operations and inter-island migration operations on each subpopulation, using the following two constraints: fitness greater than fitness threshold and system pressure drop not lower than pressure drop threshold. It records the optimal optimization parameter variables and the corresponding cold plate cooling device temperature and immersion cooling device temperature, as well as the system pressure drop determined based on system flow resistance, for each iteration until the predetermined number of iterations is reached or the objective function meets the convergence condition, thus obtaining the optimal design parameter combination and the coolant flow rate of the optimal design parameter in the corresponding optimal design parameter combination.

[0130] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0131] The cold plate optimization design device in this embodiment is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0132] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of this disclosure. The computer device includes one or more processors 10, a memory 20, and interfaces for connecting the various components, including high-speed interfaces and low-speed interfaces. The various components are interconnected via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system).

[0133] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0134] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0135] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0136] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0137] The computer device also includes an input device 30 and an output device 40. The processor 10, memory 20, input device 30, and output device 40 can be connected via a bus or other means.

[0138] Input device 30 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the computer device, such as a touchscreen, keypad, mouse, trackpad, touchpad, joystick, one or more mouse buttons, trackball, joystick, etc. Output device 40 may include display devices, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibration motors). The aforementioned display devices include, but are not limited to, liquid crystal displays, light-emitting diodes, displays, and plasma displays. In some alternative embodiments, the display device may be a touchscreen.

[0139] This disclosure also provides a computer-readable storage medium in which the methods described in this disclosure can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium after being downloaded over a network. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium may be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium may also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code that, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0140] A portion of this disclosure can be applied to computer program products, such as computer program instructions, which, when executed by a computer, can invoke or provide methods and / or technical solutions according to this disclosure through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, and installation package files. Accordingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions; the computer compiling the instructions and then executing the corresponding compiled program; the computer reading and executing the instructions; or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0141] Although embodiments of the present disclosure have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for optimizing the design of cold-plate, characterized in that, The method is applied to the optimized design of a toothed cold plate in an immersion liquid cooling system, wherein the coolant used in the immersion liquid cooling system is a fluorinated liquid or an oil; the method includes: Using a cold plate modeling tool, a set of cold plate models is generated based on a sample set of initial design parameters for the cold plate. Using a liquid cooling simulation tool, a liquid cooling simulation dataset is generated based on the coolant type, coolant flow rate, and the cold plate model set. The liquid cooling simulation data in the dataset includes: cold plate cooling device temperature, immersion cooling device temperature, and system flow resistance. The cold plate cooling device temperature is the temperature of the cold plate cooling device that uses cold plate bonding for heat dissipation. The immersion cooling device temperature is the temperature of the immersion cooling device that is submerged in the coolant. The cold plate cooling device is located upstream of the immersion cooling device. The cold plate cooling device includes at least a central processing unit, and the immersion cooling device includes at least memory. The coolant flow rate is a coolant flow rate range or more coolant flow rates corresponding to the cold plate design parameters. The surrogate model is used to expand the initial design parameter samples of the cold plate and the corresponding liquid cooling simulation data in the initial design parameter sample set of the cold plate; Using a multi-island genetic algorithm, with the system pressure drop not falling below a pressure drop threshold as a constraint, and the tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameters of the cold plate corresponding to the liquid cooling simulation data in the liquid cooling simulation dataset as optimization parameter variables, the first objective function is to minimize the temperature of the cold plate cooling device and the temperature of the immersion cooling device in the liquid cooling simulation data. This generates an optimized design parameter combination and the corresponding coolant flow rate for the optimized design parameters in the optimized design parameter combination. The system pressure drop is determined based on the system flow resistance. When there are multiple central processing units (CPUs), the temperature of the cold plate cooling device is the average temperature of the multiple CPUs, or the temperature of the CPU with the highest temperature. When there are multiple memory modules, the temperature of the immersion cooling device is the average temperature of the multiple memory modules, or the temperature of the memory module with the highest temperature. From the optimized design parameter combination, select the optimized design parameter that meets the design requirements as the target cold plate design parameter, and determine the target coolant flow rate corresponding to the target cold plate design parameter based on the coolant flow rate of the optimized design parameter in the optimized design parameter combination. Using the cold plate modeling tool, a target cold plate model is generated based on the target cold plate design parameters; Using the liquid cooling simulation tool, liquid cooling simulation data of the target cold plate model is generated based on the coolant type, the coolant flow rate, and the target cold plate model. Using the surrogate model, the liquid cooling prediction data corresponding to the design parameters of the target cold plate is determined; If the difference between the liquid cooling simulation data of the target cold plate model and the liquid cooling prediction data is greater than the difference threshold, then the target cold plate design parameters and the liquid cooling simulation data corresponding to the target cold plate design parameters are updated to the cold plate initial design parameter sample set and the liquid cooling simulation data set corresponding to the cold plate initial design parameter sample set, and the surrogate model is retrained using the updated cold plate initial design parameter sample set and liquid cooling simulation data set.

2. The method according to claim 1, characterized in that, The initial design parameter sample set for the cold plate was determined based on the following steps: Obtain the structural parameter range of the cold plate scraper teeth, which includes: tooth height range, tooth width range, and tooth pitch range. Obtain the coolant inlet flow range; Using experimental design methods, a cold plate initial design parameter sample set is generated based on the structural parameter range and coolant inlet flow range of the cold plate teeth. The initial design parameter samples in the cold plate initial design parameter sample set include: tooth height, tooth width, tooth spacing and coolant inlet flow.

3. The method according to claim 1, characterized in that, The proxy model is trained using the following steps: Using the initial design parameter samples of the cold plate in the initial design parameter sample set as input, and the liquid cooling simulation data corresponding to the initial design parameter samples of the cold plate as output, the second objective function is to minimize the difference between the discrete point set corresponding to the initial design parameter sample set of the cold plate and the predicted surface of the surrogate model. The surrogate model is trained, and when the fitting accuracy between the discrete point set corresponding to the initial design parameter sample set of the cold plate and the predicted surface of the surrogate model reaches the accuracy threshold, the surrogate model is output.

4. The method according to any one of claims 1 to 3, characterized in that, The method further includes: If the difference between the liquid cooling simulation data and the liquid cooling prediction data of the target cold plate model is less than the difference threshold, then the design parameters of the target cold plate will be output as the optimization result.

5. The method according to claim 1, characterized in that, The method utilizes a multi-island genetic algorithm, based on the liquid cooling simulation data set, with the constraint that the system pressure drop determined based on the system flow resistance is not lower than the pressure drop threshold, and with the tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameters of the cold plate corresponding to the liquid cooling simulation data as optimization parameter variables. The first objective function is to minimize the temperature of the cold plate cooling device and the temperature of the immersion cooling device in the liquid cooling simulation data, generating optimization results, including: Using the tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameter sample of the cold plate corresponding to the liquid cooling simulation data as optimization parameter variables, an initialization population is performed on the set of optimization parameter variables corresponding to the liquid cooling simulation data set to obtain multiple sub-populations. The optimization parameter variables include: tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameter sample of the cold plate. Each subpopulation is subjected to independent standard genetic operations and inter-island migration operations to obtain genetic subpopulations. The standard genetic operations include selection, crossover, and mutation. With minimizing the temperature of the cold plate cooling device and the temperature of the immersion cooling device in the liquid cooling simulation data as the first objective function, the fitness of the new optimization parameter variable is calculated based on the temperature of the cold plate cooling device and the temperature of the immersion cooling device in the liquid cooling simulation data corresponding to the new optimization parameter variable in the genetic subpopulation. Based on the system flow resistance corresponding to the liquid cooling simulation data of the new optimized parameter variables in the genetic subpopulation, the system pressure drop is determined. The following two constraints are used to guide the iteration: the fitness is greater than the fitness threshold, and the system pressure drop is not lower than the pressure drop threshold. Independent standard genetic operations and inter-island migration operations are performed on each subpopulation. The optimal optimization parameter variables and the corresponding cold plate cooling device temperature and immersion cooling device temperature, as well as the system pressure drop determined based on the system flow resistance, are recorded for each iteration until a predetermined number of iterations is reached or the objective function meets the convergence condition. The optimized design parameter combination and the coolant flow rate corresponding to the optimized design parameters in the optimized design parameter combination are obtained.

6. A cold plate optimization design device, characterized in that, The device is applied in the optimized design scenario of a toothed cold plate in an immersion liquid cooling system, wherein the coolant used in the immersion liquid cooling system is a fluorinated liquid or an oil; the device includes: The model generation module is used to generate a set of cold plate models based on the initial design parameter sample set of the cold plate using the cold plate modeling tool. The liquid cooling simulation module is used to generate a liquid cooling simulation dataset based on the coolant type, coolant flow rate, and the cold plate model set using liquid cooling simulation tools. The liquid cooling simulation data in the dataset includes: cold plate cooling device temperature, immersion cooling device temperature, and system flow resistance. The cold plate cooling device temperature refers to the temperature of the cold plate cooling device using cold plate bonding for heat dissipation, and the immersion cooling device temperature refers to the temperature of the immersion cooling device submerged in coolant. The cold plate cooling device is located upstream of the immersion cooling device. The cold plate cooling device includes at least a central processing unit, and the immersion cooling device includes at least memory. The coolant flow rate is a range of coolant flow rates corresponding to the cold plate design parameters or one or more coolant flow rates. A surrogate model is used to expand the initial design parameter samples of the cold plate and the corresponding liquid cooling simulation data in the initial design parameter sample set. The parameter optimization module utilizes a multi-island genetic algorithm, with the system pressure drop not falling below a pressure drop threshold as a constraint. It uses the tooth height, tooth width, tooth spacing, and coolant flow rate of the initial design parameters of the cold plate corresponding to the liquid cooling simulation data in the liquid cooling simulation dataset as optimization parameter variables. The first objective function is to minimize the temperature of the cold plate cooling device and the temperature of the immersion cooling device in the liquid cooling simulation data. This generates an optimized design parameter combination and the corresponding coolant flow rate for the optimized design parameters in the optimized design parameter combination. The system pressure drop is determined based on the system flow resistance. When there are multiple central processing units (CPUs), the temperature of the cold plate cooling device is the average temperature of the multiple CPUs, or the temperature of the CPU with the highest temperature. When there are multiple memory modules, the temperature of the immersion cooling device is the average temperature of the multiple memory modules, or the temperature of the memory module with the highest temperature. The parameter selection module is used to select optimized design parameters that meet the design requirements from the optimized design parameter combination as target cold plate design parameters, and to determine the target coolant flow rate corresponding to the target cold plate design parameters based on the coolant flow rate of the optimized design parameters in the optimized design parameter combination; to generate a target cold plate model using the cold plate modeling tool based on the target cold plate design parameters; to generate liquid cooling simulation data of the target cold plate model using the liquid cooling simulation tool based on the coolant type, the coolant flow rate, and the target cold plate model; to determine the liquid cooling prediction data corresponding to the target cold plate design parameters using the surrogate model; if the difference between the liquid cooling simulation data of the target cold plate model and the liquid cooling prediction data is greater than a difference threshold, then the target cold plate design parameters and the liquid cooling simulation data corresponding to the target cold plate design parameters are updated to the cold plate initial design parameter sample set and the liquid cooling simulation data set corresponding to the cold plate initial design parameter sample set, and the surrogate model is retrained using the updated cold plate initial design parameter sample set and liquid cooling simulation data set.

7. A computer device, characterized in that, include: A memory and a processor are interconnected, the memory stores computer instructions, and the processor executes the cold plate optimization design method according to any one of claims 1 to 5 by executing the computer instructions.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to execute the cold plate optimization design method according to any one of claims 1 to 5.

Images