A liquid cooling plate design method based on bayesian optimization and topology optimization

Abstract

The application relates to the technical field of battery heat dissipation, in particular to a liquid cooling plate design method based on Bayesian optimization and topology optimization, which comprises the following steps: step one, obtaining the heat generation characteristics of a battery and establishing an initial physical model of a liquid cooling plate; step two, determining a multi-target optimization function, boundary conditions, constraint conditions and decision variables; step three, solving a fluid-solid coupling equation set by using a variable density method, driving the evolution of materials in a flow channel design domain to a fluid domain or a solid domain, generating an initial topology configuration of a liquid cooling plate flow channel, and outputting performance indexes; step four, constructing a comprehensive evaluation function proxy model through Gaussian process regression, searching for decision variables through function iterative collection, and sending the decision variables determined in the current iteration to a topology optimization execution layer; after iteration is completed, the topology optimization execution layer outputs a liquid cooling plate flow channel topology configuration based on the finally determined decision variables. The application is favorable for improving the design efficiency and comprehensive performance of a liquid cooling plate flow channel structure.

Description

Technical Field

[0001] This invention belongs to the field of battery heat dissipation technology, and in particular relates to a liquid cooling plate design method based on Bayesian optimization and topology optimization. Background Technology

[0002] With the increasing demands for battery energy density in new energy vehicles and energy storage systems, the heat generation intensity of batteries during charging and discharging has significantly increased. Battery thermal safety has become a key factor restricting its development. Because batteries generate a large amount of ohmic heat and polarization heat during charging and discharging, failure to control the battery temperature within a safe range will severely affect battery life and may lead to thermal runaway. Among various thermal management solutions, liquid cooling technology has become the mainstream due to its high heat exchange capacity. The flow channel configuration of the liquid cooling plate directly determines the heat dissipation efficiency. However, existing parallel direct-flow channels or serpentine flow channels often face an inherent contradiction between heat dissipation performance and flow energy consumption. While serpentine flow channels have a long heat exchange path, they also have a huge pressure drop; parallel flow channels have a small pressure drop, but are prone to generating local hot spots due to uneven flow distribution, making it difficult to achieve a balance between temperature uniformity and power consumption under complex operating conditions.

[0003] To overcome the limitations of traditional geometric flow channel structures, topology optimization has been introduced into the design of liquid-cooled plate flow channels. This method can generate complex flow channel configurations with high heat transfer efficiency by optimizing the distribution of fluid and solid domains within the design space. However, in practical applications, the objective function of topology optimization usually needs to balance heat dissipation performance and flow energy consumption through weighting coefficients. Different weighting parameters will directly affect the final flow channel configuration and its overall performance. The setting of existing weighting parameters mostly relies on manual experience or a large number of trial and error calculations, and lacks an optimization mechanism that can automatically search and iteratively update within the preset parameter space, resulting in low design efficiency and difficulty in quickly obtaining a flow channel configuration with better overall performance under the influence of multivariate coupling. Summary of the Invention

[0004] In view of this, the present invention aims to provide a liquid cooling plate design method based on Bayesian optimization and topology optimization. Through feedback iteration of Bayesian optimization and topology optimization, the manual experience parameter tuning process is reduced, which is beneficial to improving the design efficiency and overall performance of the liquid cooling plate flow channel structure.

[0005] To achieve the above objectives, the technical solution created by this invention is implemented as follows: This invention provides a liquid-cooled plate design method based on Bayesian optimization and topology optimization, comprising: Step 1, obtaining the heat generation characteristics of the battery and establishing an initial physical model of the liquid-cooled plate; Step 2, determining the multi-objective optimization function, boundary conditions, constraints, and decision variables of the Bayesian optimization decision layer based on the heat generation characteristics and the initial physical model of the liquid-cooled plate; Step 3, the topology optimization execution layer, based on the heat generation characteristics of the battery, the initial physical model of the liquid-cooled plate, the multi-objective optimization function, boundary conditions, constraints, and the decision variables issued by the Bayesian optimization decision layer, solving the fluid-structure interaction equations using the variable density method, and through sensitive... The degree analysis drives the material within the flow channel design domain to evolve into the fluid or solid domain, automatically generating the initial topology of the liquid-cooled plate flow channel and outputting performance indicators to the Bayesian optimization decision layer. In step four, the Bayesian optimization decision layer constructs a comprehensive evaluation function surrogate model based on the performance indicators through Gaussian process regression, iteratively searches for decision variables through the acquisition function, and sends the decision variables determined in the current iteration to the topology optimization execution layer. The iteration stops when the accuracy of the comprehensive evaluation function surrogate model reaches the preset accuracy and the comprehensive evaluation function meets the preset convergence condition. The topology optimization execution layer outputs the liquid-cooled plate flow channel topology based on the finally determined decision variables.

[0006] Furthermore, in step one, the heat generation characteristics of the battery are calculated using the battery heating rate model based on the battery cell parameters, and an initial physical model of the liquid cooling plate is established based on the battery cell parameters. The battery cell parameters include: rated capacity, geometric dimensions, operating current, discharge rate, battery state of charge range, main heat generation surface range, open circuit voltage characteristics, and internal resistance parameters. The initial physical model of the liquid cooling plate includes the outer contour and dimensions of the liquid cooling plate, the flow channel design domain, the non-optimizable sealing boundary, the position of the liquid inlet, and the position of the liquid outlet.

[0007] Furthermore, in step two, with maximizing the heat dissipation capacity of the liquid cooling plate and minimizing fluid dissipation as the topology optimization objectives, and using the battery's heat generation characteristics as the heat source benchmark, a heat dissipation performance objective function and a fluid flow dissipation objective function are constructed. Based on these objective functions, a multi-objective optimization function is obtained; where the heat dissipation performance objective function... Fluid flow dissipation objective function Multi-objective optimization function ;in, , , ; Indicates the heat dissipation of a solid. This represents the amount of fluid energy dissipation. Indicates the flow channel design domain. Represents the density parameter. Indicates the heat source coefficient. For reference temperature, Indicates the temperature of the flow channel design domain. Dynamic viscosity; For fluid velocity, For the velocity gradient, This represents the normalized objective function for heat dissipation performance. This represents the normalized objective function for fluid flow dissipation. This represents the maximum fluid energy dissipation. This represents the minimum fluid energy dissipation. This represents the minimum heat dissipation of a solid object. This indicates the maximum heat dissipation of the solid element. As a weighting factor for heat dissipation performance, For flow dissipation weighting factor, Within the range of 0.1 to 0.9, Indicates the resistance to permeation. ,in: As a reverse osmosis penalty factor, Represents the reverse osmosis rate of the fluid domain. This represents the reverse osmosis rate of the solid domain.

[0008] Furthermore, in step two, a heat dissipation performance weighting factor is selected. and coolant inlet flow rate As decision variables for the Bayesian optimization decision layer; the boundary conditions and constraints for the topology optimization execution layer are determined based on the initial physical model of the liquid-cooled plate. The boundary conditions for the topology optimization execution layer include: coolant inlet temperature and outlet pressure. The constraints include: the upper limit of the volume ratio of the fluid region to the total flow channel design domain of the liquid-cooled plate, the range of the filtration radius, the range of the projection slope, and the range of the reverse osmosis penalty factor; among these, the heat dissipation performance weight factor... Within the range of 0.1 to 0.9, the coolant inlet velocity is within the range of 0.01 m / s to 0.04 m / s, the upper limit of the volume ratio of the fluid region to the total flow channel design domain of the liquid cooling plate is within the range of 40% to 60%, the coolant inlet temperature is 25℃, the outlet pressure is 0 Pa, the filtration radius is within the range of 1.5 times the grid size to 2 times the grid size, the projection slope is within the range of 8 to 32, and the reverse osmosis penalty factor is within the range of 0.01 to 0.1.

[0009] Furthermore, in step three, solving the fluid-structure interaction equations using the variable density method and driving the material evolution within the flow channel design domain towards the fluid or solid domain through sensitivity analysis includes: dividing the flow channel design domain into multiple finite element meshes, each of which is assigned an initial material density value. , This indicates that the finite element mesh is a fluid domain. This indicates that the finite element mesh is a solid domain, with the initial material density value... The initial material density field of the flow channel design domain is obtained by setting the value to 0.5. Based on the initial material density field, the decision variables issued by the Bayesian optimization decision layer, the coolant inlet temperature, the outlet pressure, and the heat generation characteristics of the battery, the fluid-structure interaction equations are solved using the variable density method to obtain the flow field distribution and temperature field distribution within the flow channel design domain. The fluid-structure interaction equations include the Brinkman equation for solving the flow field distribution and the heat transfer equation for solving the temperature field distribution. Based on the multi-objective optimization function, the flow field distribution, and the temperature field distribution, the adjoint variable method is used to obtain... Sensitivity to material density values ​​within each finite element mesh; an update algorithm is used to update the material density values ​​of each finite element mesh based on the sensitivity; numerical oscillations are suppressed and fluid-structure boundary is sharpened by Helmholtz density filtering and tangent projection based on the filter radius, projection slope, and updated material density field, and the process is iterated until the multi-objective optimization function converges to generate the initial topology configuration of the liquid cooling plate channel.

[0010] Furthermore, in step three, the performance indicators include average temperature, maximum temperature difference, and pressure drop. The output performance indicators include: after generating the initial topology of the liquid cooling plate channel, the global flow field, pressure field, and temperature field are obtained through the fluid-structure interaction equations. Then, temperature statistics are performed on the channel design domain to obtain the average temperature and maximum temperature difference. The pressure drop is obtained by calculating the difference between the inlet pressure and the outlet pressure.

[0011] Furthermore, in step four, the Bayesian optimization decision layer constructs a comprehensive evaluation function surrogate model based on performance indicators using Gaussian process regression, including: constructing a comprehensive evaluation function based on performance indicators. Within the range of decision variable values, Latin hypercube sampling is used to select multiple sets of decision variables. Each set of decision variables is then sent to the topology optimization execution layer to obtain the corresponding performance indicators. Based on these performance indicators, the corresponding comprehensive evaluation function values ​​are obtained, forming an evaluation function sample set containing multiple comprehensive evaluation function values. Based on this evaluation function sample set, Gaussian process regression is used to establish a comprehensive evaluation function surrogate model. The comprehensive evaluation function surrogate model uses weighting factors... With coolant inlet flow rate The nonlinear mapping model is obtained by fitting a Gaussian process regression with input variables and a comprehensive evaluation function constructed with performance indicators as the output response. In step four, the decision variables are iteratively searched through the acquisition function, and the decision variables determined in the current iteration are sent to the topology optimization execution layer. Iteration stops when the accuracy of the comprehensive evaluation function surrogate model reaches the preset accuracy and the comprehensive evaluation function meets the preset convergence condition. This includes: using EI to improve the acquisition function to balance local development and global exploration, iteratively selecting the better decision variables and continuously updating the comprehensive evaluation function surrogate model; repeating the above process until the comprehensive evaluation function meets the preset convergence condition, and the determination coefficient R of the surrogate model reaches the preset convergence condition. 2If the root mean square error (RMSE) meets the preset accuracy requirements, the final decision variables are obtained. The preset convergence conditions include: the change in the current iteration value of the comprehensive evaluation function is less than a preset threshold in 10 consecutive iterations. The preset threshold is 10. -2 Coefficient of determination R 2 Meeting the preset accuracy requirement with the root mean square error (RMSE) means that: R 2 ≥0.95 and RMSE≤0.05.

[0012] Furthermore, the comprehensive evaluation function as follows: ;in, Indicates average temperature. Indicates the maximum temperature difference. Indicates pressure drop. , as well as All are weighting factors. Take 0.3, Take 0.3, Take 0.4, This represents the minimum average temperature. This represents the maximum value of the average temperature. This represents the minimum value of the maximum temperature difference. This represents the maximum value of the maximum temperature difference. Describe the minimum value of the pressure drop. This indicates the maximum pressure drop.

[0013] Furthermore, after step four, the following steps are also included: Step five, based on the topology of the liquid cooling plate flow channel, the material density values ​​of the finite element mesh are binarized to identify the flow channel region and the solid region; morphological opening operation and Gaussian filtering are used to perform smoothing processing, the flow channel contour is extracted and vectorized boundary is generated through cubic spline interpolation, the vector contour in DXF format is exported, and finally the vector contour is imported into the CAD platform, and solid extrusion, Boolean subtraction and interface encapsulation are completed according to the preset thickness to generate a three-dimensional liquid cooling plate physical model in STEP format.

[0014] Furthermore, after step five, step six is ​​also included: after assembling the three-dimensional liquid cooling plate physical model and the battery model, a simulation model is established and simulation analysis is performed. If the simulation model meets the preset performance requirements, the current three-dimensional liquid cooling plate physical model is used as the final three-dimensional liquid cooling plate physical model. If the simulation model does not meet the preset performance requirements, the process returns to step two to adjust the constraints and iterate and optimize again until the final three-dimensional liquid cooling plate physical model that meets the preset performance requirements is obtained.

[0015] Compared with existing technologies, this invention achieves the following beneficial effects: Addressing the problems of traditional battery liquid cooling plate flow channel design struggling to balance heat dissipation uniformity and flow energy consumption, and the excessive reliance on manual experience in setting weight coefficients during topology optimization, making it difficult to efficiently obtain optimal parameter combinations within a large parameter space, this invention provides a liquid cooling plate design method based on Bayesian optimization and topology optimization. It constructs a two-layer optimization architecture combining a Bayesian optimization decision layer and a topology optimization execution layer, enabling automatic optimization of topology optimization parameters, improving design efficiency, and achieving coordinated optimization of heat dissipation performance and flow energy consumption while meeting battery temperature control requirements, resulting in a flow channel structure with engineering manufacturability. Attached Figure Description

[0016] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 A flowchart illustrating the liquid cooling plate design method based on Bayesian optimization and topology optimization as described in the embodiments of the present invention; Figure 2 A logical schematic diagram of the two-layer optimization architecture described in the embodiments of the present invention; Figure 3 A schematic diagram showing the location of thermocouple temperature measuring points on the surface of a battery cell as described in an embodiment of the present invention; Figure 4 A schematic diagram of the flow channel design domain as described in the embodiments of the present invention; Figure 5 A schematic diagram of the comprehensive evaluation function proxy model described in the embodiments of the present invention; Figure 6 A schematic diagram of a liquid-cooled plate flow channel topology configuration based on the output of a final determined decision variable, as described in an embodiment of the present invention; Figure 7 This is a schematic diagram of the structure of a parallel flow channel liquid cooling plate; Figure 8 This is a schematic diagram of the serpentine flow channel liquid cooling plate. Figure 9 This is a comparison chart of temperature and pressure drop data between the topological flow channel liquid cooling plate of the present invention and the conventional cooling plate of the control group; Figure 10 This is a comparison chart showing the maximum temperature difference between the topological flow channel liquid cooling plate of this invention and the conventional cooling plate of the control group. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not constitute a limitation thereof.

[0018] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0019] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0020] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0021] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] refer to Figures 1 to 2This invention provides a liquid-cooled plate design method based on Bayesian optimization and topology optimization, comprising: Step 1, obtaining the heat generation characteristics of the battery and establishing an initial physical model of the liquid-cooled plate; Step 2, determining the multi-objective optimization function, boundary conditions, constraints, and decision variables of the Bayesian optimization decision layer based on the heat generation characteristics and the initial physical model of the liquid-cooled plate; Step 3, the topology optimization execution layer, based on the heat generation characteristics of the battery, the initial physical model of the liquid-cooled plate, the multi-objective optimization function, boundary conditions, constraints, and the decision variables issued by the Bayesian optimization decision layer, solving the fluid-structure interaction equations using the variable density method, and through sensitive... The degree analysis drives the material within the flow channel design domain to evolve into the fluid or solid domain, automatically generating the initial topology of the liquid-cooled plate flow channel and outputting performance indicators to the Bayesian optimization decision layer. In step four, the Bayesian optimization decision layer constructs a comprehensive evaluation function surrogate model based on the performance indicators through Gaussian process regression, iteratively searches for decision variables through the acquisition function, and sends the decision variables determined in the current iteration to the topology optimization execution layer. The iteration stops when the accuracy of the comprehensive evaluation function surrogate model reaches the preset accuracy and the comprehensive evaluation function meets the preset convergence condition. The topology optimization execution layer outputs the liquid-cooled plate flow channel topology based on the finally determined decision variables.

[0023] The liquid cooling plate design method based on Bayesian optimization and topology optimization provided by this invention constructs a two-layer optimization architecture consisting of an upper decision layer (Bayesian optimization decision layer) and a lower execution layer (topology optimization execution layer). The Bayesian optimization decision layer uses the Bayesian optimization algorithm to optimize the decision variables (heat dissipation performance weight factors). and coolant inlet flow rate Within a predefined search space, the topology optimization execution layer performs fluid-structure interaction topology optimization based on the decision variables issued by the Bayesian optimization decision layer, and feeds back performance indicators to the Bayesian optimization decision layer. The Bayesian optimization decision layer and the topology optimization execution layer achieve iterative closed loop through feedback loop.

[0024] Furthermore, in step one, the heat generation characteristics of the battery are calculated using the battery heating rate model based on the battery cell parameters, and an initial physical model of the liquid cooling plate is established based on the battery cell parameters. The battery cell parameters include: rated capacity, geometric dimensions, operating current, discharge rate, battery state of charge range, main heat generation surface range, open circuit voltage characteristics, and internal resistance parameters. The initial physical model of the liquid cooling plate includes the outer contour and dimensions of the liquid cooling plate, the flow channel design domain, the non-optimizable sealing boundary, the position of the liquid inlet, and the position of the liquid outlet.

[0025] In step one, the open-circuit voltage characteristics and internal resistance parameters can be obtained through HPPC (Hybrid Pulse Power Characterization) experiments. The heat generation characteristics of the battery can be calculated using a battery heating rate model, including: calculating the heat generation of the battery at a specific discharge rate using a battery heating rate model (such as the Bernardi model), which serves as the heat generation characteristics of the battery and is used as the heat source parameters for subsequent topology optimization and fluid-structure interaction simulation models; the inlet and outlet positions of the liquid cooling plate and the flow channel design domain can be set according to the geometry of the battery cell and the range of the main heating surface, ensuring that the flow channel design domain covers the main heating surface of the battery, providing a physical basis for setting boundary conditions in the subsequent two-layer optimization architecture.

[0026] In some embodiments, after obtaining the open-circuit voltage characteristics and internal resistance parameters through HPPC experiments in step one, as follows: Figure 3 As shown, five thermocouple temperature measuring points 3 can be arranged on the surface of the battery cell. Local temperature data at different discharge rates can be collected in real time through the five thermocouple temperature measuring points 3. This data is used to experimentally verify the calculation results of the battery heating rate model to ensure that the obtained heat generation characteristics of the battery are reasonable.

[0027] It should be noted that in step one, the area of ​​the flow channel design domain should be smaller than the size of the main heat-generating surface of the actual battery cell, such as... Figure 4 As shown, a non-optimizable solid region 1 of a certain width is reserved outside the flow channel design domain 2. The width of the non-optimizable solid region 1 can be in the range of 3mm to 10mm. The non-optimizable solid region 1 always maintains the solid material properties and does not fill the flow channel during the topology evolution process. The purpose of setting the non-optimizable solid region 1 is to provide sufficient welding sealing interface for the liquid cooling plate cover, while avoiding the flow channel being too close to the edge, which would lead to a decrease in structural strength or processing interference, and ensure the reliability and sealing of the liquid cooling plate in actual production.

[0028] Furthermore, in step two, with maximizing the heat dissipation capacity of the liquid cooling plate and minimizing fluid dissipation as the topology optimization objectives, and using the battery's heat generation characteristics as the heat source benchmark, a heat dissipation performance objective function and a fluid flow dissipation objective function are constructed. Based on these objective functions, a multi-objective optimization function is obtained; where the heat dissipation performance objective function... Fluid flow dissipation objective function Multi-objective optimization function ;in, , , ; Indicates the heat dissipation of a solid. This represents the amount of fluid energy dissipation. Indicates the flow channel design domain. Represents the density parameter. Indicates the heat source coefficient. For reference temperature, Indicates the temperature of the flow channel design domain. Dynamic viscosity; For fluid velocity, For the velocity gradient, This represents the normalized objective function for heat dissipation performance. This represents the normalized objective function for fluid flow dissipation. This represents the maximum fluid energy dissipation. This represents the minimum fluid energy dissipation. This represents the minimum heat dissipation of a solid object. This indicates the maximum heat dissipation of the solid element. As a weighting factor for heat dissipation performance, For flow dissipation weighting factor, Within the range of 0.1 to 0.9, Indicates the resistance to permeation. ,in: As a reverse osmosis penalty factor, Represents the reverse osmosis rate of the fluid domain. This represents the reverse osmosis rate of the solid domain.

[0029] In some examples, =0, With dimensionless numbers related, The calculation formula is as follows: ;in: Darcy number, also known as the Darcy number, is used to describe the relationship between viscous and frictional forces in porous media. ; The characteristic length is taken as the diameter of the flow channel at the coolant inlet.

[0030] Furthermore, in step two, a heat dissipation performance weighting factor is selected. and coolant inlet flow rate As decision variables for the Bayesian optimization decision layer; the boundary conditions and constraints for the topology optimization execution layer are determined based on the initial physical model of the liquid-cooled plate. The boundary conditions for the topology optimization execution layer include: coolant inlet temperature and outlet pressure. The constraints include: the upper limit of the volume ratio of the fluid region to the total flow channel design domain of the liquid-cooled plate, the range of the filtration radius, the range of the projection slope, and the range of the reverse osmosis penalty factor; among these, the heat dissipation performance weight factor... Within the range of 0.1 to 0.9, the coolant inlet velocity is within the range of 0.01 m / s to 0.04 m / s, the upper limit of the volume ratio of the fluid region to the total flow channel design domain of the liquid cooling plate is within the range of 40% to 60%, the coolant inlet temperature is 25℃, the outlet pressure is 0 Pa, the filtration radius is within the range of 1.5 times the grid size to 2 times the grid size, the projection slope is within the range of 8 to 32, and the reverse osmosis penalty factor is within the range of 0.01 to 0.1.

[0031] Among them, the heat dissipation performance weighting factor The coolant inlet velocity is used to balance the objective function of heat dissipation performance and the objective function of fluid flow dissipation. In the topology optimization solution process, the specific value of the inlet velocity boundary condition is automatically determined by the Bayesian optimization decision layer within the preset search space, thereby reducing the reliance on manual experience in parameter tuning and improving the rationality of the topology optimization parameter selection. The upper limit of the volume ratio of the fluid region to the total flow channel design domain of the liquid cooling plate is used to limit the flow channel ratio, the filter radius is used to control the minimum size characteristics of the flow channel, and the projection slope is used to suppress intermediate density units and improve the clarity of the flow channel boundary. This helps to ensure that the optimized flow channel configuration meets the thermodynamic performance requirements and has good geometric topological characteristics and manufacturability.

[0032] Furthermore, in step three, solving the fluid-structure interaction equations using the variable density method and driving the material evolution within the flow channel design domain towards the fluid or solid domain through sensitivity analysis includes: dividing the flow channel design domain into multiple finite element meshes, each of which is assigned an initial material density value. , This indicates that the finite element mesh is a fluid domain. This indicates that the finite element mesh is a solid domain, with the initial material density value... The initial material density field of the flow channel design domain is obtained by setting the value to 0.5. Based on the initial material density field, the decision variables issued by the Bayesian optimization decision layer, the coolant inlet temperature, the outlet pressure, and the heat generation characteristics of the battery, the fluid-structure interaction equations are solved using the variable density method to obtain the flow field distribution and temperature field distribution within the flow channel design domain. The fluid-structure interaction equations include the Brinkman equation for solving the flow field distribution and the heat transfer equation for solving the temperature field distribution. Based on the multi-objective optimization function, the flow field distribution, and the temperature field distribution, the adjoint variable method is used to obtain... Sensitivity to material density values ​​within each finite element mesh; an update algorithm is used to update the material density values ​​of each finite element mesh based on the sensitivity; numerical oscillations are suppressed and fluid-structure boundary is sharpened by Helmholtz density filtering and tangent projection based on the filter radius, projection slope, and updated material density field, and the process is iterated until the multi-objective optimization function converges to generate the initial topology configuration of the liquid cooling plate channel.

[0033] Furthermore, in step three, the performance indicators include average temperature, maximum temperature difference, and pressure drop. The output performance indicators include: after generating the initial topology of the liquid cooling plate channel, the global flow field, pressure field, and temperature field are obtained through the fluid-structure interaction equations. Then, temperature statistics are performed on the channel design domain to obtain the average temperature and maximum temperature difference. The pressure drop is obtained by calculating the difference between the inlet pressure and the outlet pressure.

[0034] Specifically, in step three, the initial physical model of the liquid-cooled plate constructed in step one, the heat generation characteristics of the battery, the multi-objective optimization function set in step two, the boundary conditions, the constraints, and the decision variables issued by the Bayesian optimization decision layer are imported into the optimization solution system corresponding to the topology optimization execution layer. The optimization solution system uses a topology optimization algorithm based on the variable density method to mesh the liquid-cooled plate design domain and introduces a penalty factor to suppress intermediate density elements. During the calculation process, the topology optimization algorithm of the variable density method solves the fluid-structure interaction equations to obtain the flow field distribution and temperature field distribution within the liquid-cooled plate design domain, and performs sensitivity analysis using the adjoint variable method to obtain the density parameters (material density values) of each element within the flow channel design domain. The sensitivity distribution of the multi-objective optimization function provides directional guidance and quantitative basis for the iterative evolution of density parameters, driving the evolution of material density values. Through multiple iterative calculations, the material in the flow channel design domain gradually evolves towards the fluid domain or solid domain under the combined action of boundary conditions and the multi-objective optimization function, that is, the material density value evolves towards 1 or the material density value evolves towards 0, thereby optimizing the distribution shape of the flow channel until the iteration result meets the preset convergence requirements, thus obtaining the initial topology configuration of the liquid cooling plate flow channel.

[0035] After obtaining the initial topology configuration of the liquid cooling plate channel, the topology optimization execution layer feeds back the obtained average temperature (the average temperature of all finite element meshes in the channel design domain), maximum temperature difference (the difference between the highest and lowest temperatures in the channel design domain, used to measure the heat dissipation uniformity; the smaller the value, the more uniform the heat dissipation), and pressure drop (used to measure the flow resistance; the smaller the pressure drop, the lower the pump energy consumption) as response feedback to the upper Bayesian optimization decision layer.

[0036] In some embodiments, topology optimization is based on a variable density method framework. It can achieve the continuous evolution of fluid and solid domain properties by solving fluid-structure interaction equations in simulation platforms such as COMSOL Multiphysics. During iterative calculations, the algorithm discretizes the channel design domain into multiple non-uniform tetrahedral or hexahedral meshes and assigns a virtual density variable (material density value) to each finite element mesh. A virtual density variable of 1 represents the fluid domain, and a virtual density variable of 0 represents the solid domain. By alternately solving the Brinkman equation and the heat transfer equation, the sensitivity of each finite element mesh to the multi-objective optimization function is calculated. To suppress the checkerboard effect and ensure the steady-state evolution of the channel topology, a density filtering algorithm based on the Helmholtz equation is introduced. This algorithm is executed in each iteration of topology optimization, after updating the virtual density variable, and before the next round of physics field solving. A smooth constraint is applied to the density field based on a preset filtering radius to suppress the checkerboard effect and ensure the steady-state evolution of the channel topology. In this invention, the convergence requirement for the multi-objective optimization function is set as follows: the relative change of the multi-objective optimization function between two adjacent iterations is less than 10. -6 Step three of this invention, through adaptive migration of material distribution, helps to reduce high-resistance redundant branches while meeting heat dissipation requirements, thereby achieving coordinated optimization of heat exchange efficiency and pump power loss.

[0037] Furthermore, in step four, the Bayesian optimization decision layer constructs a comprehensive evaluation function surrogate model based on performance indicators using Gaussian process regression, including: constructing a comprehensive evaluation function based on performance indicators. Within the range of decision variable values, Latin hypercube sampling is used to select multiple sets of decision variables. Each set of decision variables is then sent to the topology optimization execution layer to obtain the corresponding performance indicators. Based on these performance indicators, the corresponding comprehensive evaluation function values ​​are obtained, forming an evaluation function sample set containing multiple comprehensive evaluation function values. Based on this evaluation function sample set, Gaussian process regression is used to establish a comprehensive evaluation function surrogate model. The comprehensive evaluation function surrogate model uses weighting factors... With coolant inlet flow rate The nonlinear mapping model is obtained by fitting a Gaussian process regression with input variables and a comprehensive evaluation function constructed with performance indicators as the output response. In step four, the decision variables are iteratively searched through the acquisition function, and the decision variables determined in the current iteration are sent to the topology optimization execution layer. Iteration stops when the accuracy of the comprehensive evaluation function surrogate model reaches the preset accuracy and the comprehensive evaluation function meets the preset convergence condition. This includes: using EI to improve the acquisition function to balance local development and global exploration, iteratively selecting the better decision variables and continuously updating the comprehensive evaluation function surrogate model; repeating the above process until the comprehensive evaluation function meets the preset convergence condition, and the determination coefficient R of the surrogate model reaches the preset convergence condition. 2If the root mean square error (RMSE) meets the preset accuracy requirements, the final decision variables are obtained. The preset convergence conditions include: the change in the current iteration value of the comprehensive evaluation function is less than a preset threshold in 10 consecutive iterations. The preset threshold is 10. -2 Coefficient of determination R 2 Meeting the preset accuracy requirement with the root mean square error (RMSE) means that: R 2 ≥0.95 and RMSE≤0.05.

[0038] Furthermore, the comprehensive evaluation function as follows: ;in, Indicates average temperature. Indicates the maximum temperature difference. Indicates pressure drop. , as well as All are weighting factors. Take 0.3, Take 0.3, Take 0.4, This represents the minimum average temperature. This represents the maximum value of the average temperature. This represents the minimum value of the maximum temperature difference. This represents the maximum value of the maximum temperature difference. Describe the minimum value of the pressure drop. This indicates the maximum pressure drop.

[0039] Comprehensive evaluation function It is a weighted dimensionless index constructed based on multi-objective decision-making theory. , as well as The weighting ratio can be dynamically adjusted according to the actual thermal safety requirements of the battery and the power consumption constraints of the pump. The above values ​​are just examples.

[0040] In step four, heuristic iterative sampling is performed in the search space using a sampling function to search for decision variables. During each iteration, the current decision variable is used as input to trigger the topology optimization solution (solving the initial topology configuration of the liquid cooling plate channel) and performance index calculation in step three. The obtained performance index is fed back to the comprehensive evaluation function surrogate model for correction and update. During the iteration process, the prediction accuracy of the comprehensive evaluation function surrogate model is iteratively evaluated, and the correlation coefficient or error index between the predicted value and the actual solution value obtained in step three is calculated. If the accuracy of the comprehensive evaluation function surrogate model does not meet the requirements, new sampling points are added through the sampling function and the iteration continues to update the comprehensive evaluation function surrogate model. This continues until the accuracy of the comprehensive evaluation function surrogate model meets the requirements, and the algorithm automatically searches for a combination of decision variables that makes the comprehensive evaluation function meet the preset convergence conditions. Thus, the decision variables that meet the preset convergence conditions are determined without relying on manual trial and error. It should be noted that the multi-objective optimization function in step two is used for solving a single topology optimization problem, while the comprehensive evaluation function in step four is used for global comparison and parameter optimization of the topology optimization results corresponding to different decision variables. The two belong to different levels in the two-layer architecture consisting of the topology optimization execution layer and the Bayesian optimization decision layer.

[0041] In step four, the comprehensive evaluation function surrogate model is constructed using an adaptive Gaussian process regression algorithm, and Latin hypercube sampling is combined to initialize the variable space, which helps improve the coverage of the search space by the initial samples. During the iterative optimization process, the expected improvement sampling function is used to evaluate the potential value of each point in the search space. By balancing the mean and variance of the model predictions, the algorithm is guided to prioritize sampling in regions with high performance improvement potential, thereby reducing the number of calls to the physical solution in step three and improving optimization efficiency.

[0042] In step four, the objective of Bayesian optimization is to find a set of decision variables that makes the comprehensive evaluation function... A surrogate model for the comprehensive evaluation function constructed using Gaussian process regression, satisfying the preset convergence conditions, is as follows: Figure 5 As shown, Figure 5 The surface in the equation represents the comprehensive evaluation function. Response relationships within the search space of decision variables The lower the value, the better the overall performance of the corresponding flow channel configuration. The accuracy evaluation mechanism uses the coefficient of determination R. 2 The root mean square error (RMSE) and the root mean square error (RMSE) are used as dual evaluation criteria, with the evaluation criterion set as the coefficient of determination R. 2 A value ≥0.95 and a root mean square error (RMSE) ≤0.05 are beneficial for improving the reliability of the comprehensive evaluation function surrogate model in predicting flow channel performance.

[0043] Specifically, in step four, when the comprehensive evaluation function is found... Once the decision variables meet the preset convergence conditions, the weighting coefficients and inlet flow velocity for the current heat load conditions can be automatically determined, with reference to... Figure 6 This invention demonstrates the generated liquid cooling plate flow channel topology configuration driven by the combination of decision variables. It realizes the automation process of liquid cooling plate design, reduces the dependence on manual experience parameter tuning in traditional liquid cooling plate design, and helps to improve the overall performance of liquid cooling plate flow channel configuration under the constraints of heat dissipation performance, temperature uniformity and flow resistance.

[0044] Furthermore, after step four, the process also includes: Step five, based on the liquid cooling plate flow channel topology obtained in step four, binarizing the material density values ​​of the finite element mesh to identify the flow channel region and the solid region; performing smoothing processing using morphological opening operation and Gaussian filtering, extracting the flow channel contour and generating vectorized boundaries through cubic spline interpolation, exporting the vector contour in DXF format, and finally importing the vector contour into the CAD platform, completing solid extrusion, Boolean subtraction and interface encapsulation according to the preset thickness, and generating a three-dimensional liquid cooling plate physical model in STEP format.

[0045] Specifically, step five is used to perform post-processing of the topology flow channel and 3D geometric reconstruction. Specifically, the liquid cooling plate flow channel topology determined in step four is image post-processed. Threshold segmentation technology is used to extract clear flow channel boundaries, and a smoothing algorithm is used to remove isolated units, sharp corners, and tiny branches that do not meet the processing size requirements in the liquid cooling plate flow channel topology to ensure the continuity of the flow channel structure and reduce local fluid resistance. The processed 2D flow channel boundaries are exported to the 3D modeling platform in vector format. Combined with the thickness of the liquid cooling plate base plate, the cover plate structure, and the geometric parameters of the flow channel inlet and outlet, a solid extrusion and encapsulation reconstruction is performed. Finally, a 3D physical model of the liquid cooling plate containing complete internal flow channel features is established, providing a geometric basis for subsequent simulation verification and engineering manufacturing.

[0046] In step five, during image post-processing, a segmentation method based on a global threshold is used for the material density distribution matrix determined in step four. The material density is binarized with a threshold of 0.5. Regions with a material density less than 0.5 are identified as solid regions, and regions with a material density greater than or equal to 0.5 are identified as flow channel regions. By extracting the binarized flow channel contour, the preliminary geometric boundary of the flow channel topology is established.

[0047] In step five, during the smoothing process, morphological opening operations are used to first erode and then expand. The diameter D of the structural element is set to 1.1 to 1.5 times the minimum processing radius, which helps to reduce slender branches that do not meet the processing limits. At the same time, Gaussian filtering is used to smooth the flow channel edges, transforming the stepped boundaries generated by topology evolution into smoother curves to reduce local resistance losses in the actual flow process.

[0048] In step five, during the 3D reconstruction, smoothed boundary feature points are extracted, and continuous vectorized closed paths are generated using cubic spline interpolation. These vector paths are then imported into the CAD modeling software in DXF format. The thickness of the side covers of the liquid cooling plate is set to be within the range of 1mm to 2mm, and the thickness in the middle is within the range of 3mm to 5mm. Solid extrusion and subtractive Boolean operations are performed based on the vector paths. 3D interface encapsulation is performed based on the inlet and outlet positions determined in step one, ultimately generating a 3D physical model in STEP format, which can then be used for subsequent mesh generation and simulation verification.

[0049] Furthermore, after step five, step six is ​​also included: after assembling the three-dimensional liquid cooling plate physical model and the battery model, a simulation model is established and simulation analysis is performed. If the simulation model meets the preset performance requirements, the current three-dimensional liquid cooling plate physical model is used as the final three-dimensional liquid cooling plate physical model. If the simulation model does not meet the preset performance requirements, the process returns to step two to adjust the constraints and iterate and optimize again until the final three-dimensional liquid cooling plate physical model that meets the preset performance requirements is obtained.

[0050] Step six is ​​used to realize fluid-structure interaction simulation verification and comprehensive performance evaluation of multiple schemes. Specifically, step six includes: assembling the three-dimensional physical model obtained from the reconstruction in step five with the battery to establish a complete fluid-structure interaction simulation model; under the preset heat generation conditions, loading the heat generation characteristics of the battery obtained in step one; dynamically extracting and analyzing the highest temperature, maximum temperature difference, and pressure drop of the liquid cooling system during operation to verify whether the three-dimensional physical model after geometric reconstruction in step five still meets the preset safety threshold and energy consumption standards (the preset safety threshold and energy consumption standards are the preset performance requirements); simultaneously, introducing a parallel flow channel liquid cooling plate with the same heat exchange area and inlet / outlet conditions (such as... Figure 7 (as shown) and serpentine flow channel liquid cooling plate (such as Figure 8 As shown in the figure, the comprehensive technical advantages of the topology flow channel liquid cooling plate generated by the present invention in improving thermal safety and reducing system energy consumption are quantitatively evaluated by comparing various performance indicators horizontally. If the simulation verification results show that the performance does not meet the standard due to the limitation of design space or the initial constraints are too strict, the process returns to step two to redefine the boundary conditions of the design domain and re-execute the two-layer iteration of steps three and four until a topology flow channel liquid cooling plate that meets the preset performance requirements is obtained.

[0051] Specifically, in step six, before establishing the fluid-structure interaction simulation model, mesh generation and mesh independence analysis are required to optimize resource usage while ensuring computational accuracy. The fluid-structure interaction simulation model is established by using an unstructured mesh to divide the solid plate and the fluid domain, and the boundary layer mesh is refined at the fluid-structure interface. At least three boundary layer meshes can be set.

[0052] In step six, when performing a cross-sectional evaluation of multiple options, the principle of controlling variables must be strictly followed: Figure 7 and Figure 8 As shown, the topology flow channel scheme to be evaluated is ensured to have the same inlet and outlet dimensions, the same fluid domain space ratio, and equivalent heat dissipation boundary conditions as the parallel flow channel and serpentine flow channel schemes used as the reference, so as to reduce the impact of geometric scale differences on the evaluation results.

[0053] Specifically, in step six, the simulation boundary conditions are set as follows: the coolant is pure water or a 50% ethylene glycol aqueous solution; both the coolant and ambient temperatures are 25°C; the inlet velocity of the liquid cooling plate is in the range of 0.1 m / s to 0.3 m / s; and the outlet pressure is 0 Pa. It should be noted that the coolant inlet velocity range in step two is mainly used for parameter optimization in the topology optimization stage, while the inlet velocity range in step six is ​​used to verify the fluid-structure interaction performance of the liquid cooling plate under actual working conditions after 3D reconstruction. These two settings correspond to boundary conditions set at different design stages.

[0054] Specifically, in step six, the preset safety threshold can be set according to the thermal safety standards of power batteries. The preset safety threshold may include: the maximum temperature of a single cell. Maximum temperature difference <40℃ <5℃. After the simulation calculation is completed, the highest temperature of the battery cell under each scheme is extracted. Maximum temperature difference and the pressure drop at the inlet and outlet of the liquid cooling plate The topological flow channel liquid cooling plate generated by this invention is compared with the reference scheme. If the topological flow channel liquid cooling plate can not only achieve the same flow rate, it will also... If the temperature is controlled within a preset safety threshold, and the solution exhibits lower maximum temperature and lower flow resistance compared to parallel cold plates or serpentine flow channel cold plates, then the current solution is deemed to have comprehensive technical advantages. For example... Figure 9 and Figure 10 As shown, a data comparison between the topological flow channel liquid cooling plate of the present invention and the conventional cold plate configuration of the control group reveals that, at a 3C discharge rate, the highest temperature of the three cold plates is... and maximum temperature difference All meet safety threshold requirements, and the topological flow channel liquid cooling plate obtained by this invention exhibits superior temperature control performance; simultaneously, the pressure drop of the topological flow channel liquid cooling plate of this invention... The pressure is only 128.16 Pa, which is lower than the 140.25 Pa of the parallel channel and 62.6% lower than the 342.64 Pa of the serpentine channel. This indicates that the liquid cooling plate with the topology channel determined by the liquid cooling plate design method provided by the present invention has good comprehensive performance in terms of both heat dissipation performance and flow resistance.

[0055] In step six, if the simulation analysis results show that the thermal management performance of the liquid cooling plate does not meet the preset index, the automatic feedback adjustment mechanism is triggered. That is, the system feeds back the deviation data to step two, and drives the topology algorithm to re-execute iterative optimization by adjusting the boundary conditions of topology optimization until the final flow channel scheme that meets the preset performance requirements is obtained.

[0056] Furthermore, step six is ​​followed by step seven, which is used to realize the topology optimization cold plate process verification and prototype trial production. Specifically, step seven includes: for the final three-dimensional liquid cooling plate physical model determined in step six, combined with actual manufacturing process conditions, such as stamping, brazing, or additive manufacturing, the final three-dimensional liquid cooling plate physical model is subjected to processing constraint verification. The depth, width, and bending radius of the flow channel are analyzed to see if they are within the capability range of the existing processing equipment. If there is processing interference or over-limit structure, the local parameters of the flow channel are corrected without affecting the heat dissipation performance to ensure the adaptability of the topology configuration to the production process. After the process verification is completed, the prototype can be processed and assembled according to the optimized three-dimensional physical model, thereby realizing the transformation from the topology design result to the engineering product.

[0057] In some embodiments, the wall thickness at the narrowest point of the flow channel and the minimum bending radius at the branch bend can be checked based on the three-dimensional liquid cooling plate physical model output in step five to ensure that it meets the process limits of stamping, brazing or additive manufacturing. After the prototype is manufactured, an airtightness and pressure resistance test with a pressure of not less than 0.5 MPa is performed on it, and it is assembled into a battery module for a measured thermal balance experiment. By comparing the measured temperature rise data with the simulation prediction results in step six, if the deviation between the two is within 10%, it indicates that the heat dissipation reliability and manufacturing feasibility of the topology flow channel scheme in actual industrial applications are good.

[0058] The liquid-cooled plate design method based on Bayesian optimization and topology optimization provided by this invention has the following beneficial effects: It achieves automatic determination of optimization parameters. Specifically, this invention introduces a Bayesian optimization algorithm to optimize key parameters in topology optimization within a preset search space, reducing reliance on manual parameter tuning and improving the automation level and rationality of parameter selection in the design process. By establishing a two-layer collaborative optimization architecture combining upper-layer Bayesian optimization (Bayesian optimization decision layer) and lower-layer topology optimization (topology optimization execution layer), it realizes feedback iteration between the parameter optimization process and the physical field topology optimization process, which is beneficial to improving the solution efficiency of complex multi-objective optimization problems; it is also beneficial to… The system reduces computational costs. Specifically, by using a proxy model to guide the sampling process, the number of calls to the computationally expensive physical solution process is reduced, thereby lowering the overall computational cost and shortening the design cycle. It also improves the manufacturability of the flow channel structure. Specifically, by geometrically reconstructing and smoothing the topology optimization results, the generated flow channel structure meets the processing size constraints and process implementation requirements, which is beneficial for subsequent engineering modeling and manufacturing implementation. Furthermore, it improves the thermal management performance of the battery system. Specifically, the designed liquid cooling plate with flow channels can optimize the coolant distribution, which, under safe operating conditions, helps reduce the temperature difference between battery cells and improves the thermal consistency and operational stability of the battery system.

[0059] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0060] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A liquid cooling plate design method based on Bayesian optimization and topology optimization, characterized in that, include: Step 1: Obtain the heat generation characteristics of the battery and establish the initial physical model of the liquid cooling plate; Step 2: Based on the heat generation characteristics and the initial physical model of the liquid cooling plate, determine the multi-objective optimization function, boundary conditions, constraints, and decision variables of the Bayesian optimization decision layer for the topology optimization execution layer; Step 3: The topology optimization execution layer solves the fluid-structure interaction equations using the variable density method based on the battery's heat generation characteristics, the initial physical model of the liquid cooling plate, the multi-objective optimization function, boundary conditions, constraints, and the decision variables issued by the Bayesian optimization decision layer. Through sensitivity analysis, it drives the material in the flow channel design domain to evolve into the fluid domain or solid domain, automatically generates the initial topology configuration of the liquid cooling plate flow channel, and outputs performance indicators to the Bayesian optimization decision layer. Step four: The Bayesian optimization decision layer constructs a comprehensive evaluation function surrogate model based on the performance index through Gaussian process regression. It iteratively searches for decision variables by collecting data and sends the decision variables determined in the current iteration to the topology optimization execution layer. The iteration stops when the accuracy of the comprehensive evaluation function surrogate model reaches the preset accuracy and the comprehensive evaluation function meets the preset convergence condition. The topology optimization execution layer outputs the liquid cooling plate flow channel topology configuration based on the finally determined decision variables.

2. The liquid cooling plate design method based on Bayesian optimization and topology optimization according to claim 1, characterized in that, In step one, the heat generation characteristics of the battery are calculated using the battery heating rate model based on the battery cell parameters. An initial physical model of the liquid cooling plate is established based on the battery cell parameters. The battery cell parameters include: rated capacity, geometric dimensions, operating current, discharge rate, battery state of charge range, main heat-generating surface range, open-circuit voltage characteristics, and internal resistance parameters. The initial physical model of the liquid cooling plate includes the outer contour and dimensions of the liquid cooling plate, the flow channel design domain, the non-optimizable sealing boundary, the position of the liquid inlet, and the position of the liquid outlet.

3. The liquid cooling plate design method based on Bayesian optimization and topology optimization according to claim 2, characterized in that, In step two, the topology optimization objectives are to maximize the heat dissipation capacity of the liquid cooling plate and minimize fluid dissipation. The heat generation characteristics of the battery are used as the heat source benchmark. The objective functions of heat dissipation performance and fluid flow dissipation are constructed. Based on the objective functions of heat dissipation performance and fluid flow dissipation, a multi-objective optimization function is obtained. Among them, the objective function for heat dissipation performance Fluid flow dissipation objective function Multi-objective optimization function ;in, , , ; Indicates the heat dissipation of a solid. This represents the amount of fluid energy dissipation. Indicates the flow channel design domain. Represents the density parameter. Indicates the heat source coefficient. For reference temperature, Indicates the temperature of the flow channel design domain. Dynamic viscosity; For fluid velocity, For the velocity gradient, This represents the normalized objective function for heat dissipation performance. This represents the normalized objective function for fluid flow dissipation. This represents the maximum fluid energy dissipation. This represents the minimum fluid energy dissipation. This represents the minimum heat dissipation of a solid object. This indicates the maximum heat dissipation of the solid element. As a weighting factor for heat dissipation performance, For flow dissipation weighting factor, Within the range of 0.1 to 0.9, Indicates the resistance to permeation. ,in: As a reverse osmosis penalty factor, Represents the reverse osmosis rate of the fluid domain. This represents the reverse osmosis rate of the solid domain.

4. The liquid cooling plate design method based on Bayesian optimization and topology optimization according to claim 3, characterized in that, In step two, select the heat dissipation performance weighting factor. and coolant inlet flow rate As decision variables in the Bayesian optimization decision layer; Based on the initial physical model of the liquid cooling plate, the boundary conditions and constraints of the topology optimization execution layer are determined. The boundary conditions of the topology optimization execution layer include: coolant inlet temperature and outlet pressure. The constraints include: the upper limit of the volume ratio of the fluid region to the total flow channel design domain of the liquid cooling plate, the range of the filtration radius, the range of the projection slope, and the range of the reverse osmosis penalty factor. Among them, the heat dissipation performance weighting factor Within the range of 0.1 to 0.9, the coolant inlet velocity is within the range of 0.01 m / s to 0.04 m / s, the upper limit of the volume ratio of the fluid region to the total flow channel design domain of the liquid cooling plate is within the range of 40% to 60%, the coolant inlet temperature is 25℃, the outlet pressure is 0 Pa, the filtration radius is within the range of 1.5 times the grid size to 2 times the grid size, the projection slope is within the range of 8 to 32, and the reverse osmosis penalty factor is within the range of 0.01 to 0.

1.

5. The liquid cooling plate design method based on Bayesian optimization and topology optimization according to claim 4, characterized in that, In step three, the fluid-structure interaction equations are solved using the variable density method, and the evolution of the material within the flow channel design domain into the fluid or solid domain is driven by sensitivity analysis, including: The flow channel design domain is divided into multiple finite element meshes, and each finite element mesh is assigned an initial material density value. , This indicates that the finite element mesh is a fluid domain. This indicates that the finite element mesh is a solid domain, with the initial material density value... The value is 0.5, thus obtaining the initial material density field of the flow channel design domain; Based on the initial material density field, the decision variables issued by the Bayesian optimization decision layer, the coolant inlet temperature, the outlet pressure, and the heat generation characteristics of the battery, the fluid-structure interaction equations are solved using the variable density method to obtain the flow field distribution and temperature field distribution within the flow channel design domain. The fluid-structure interaction equations include the Brinkman equation for solving the flow field distribution and the heat transfer equation for solving the temperature field distribution. Based on the aforementioned multi-objective optimization function, flow field distribution, and temperature field distribution, the adjoint variable method is used to obtain... Sensitivity to material density values ​​within each finite element mesh; An update algorithm is used to update the material density value of each finite element mesh based on sensitivity. Based on the filter radius, projection slope, and updated material density field, numerical oscillations are suppressed and fluid-solid boundary is sharpened through Helmholtz density filtering and tangent projection. The process is iterated until the multi-objective optimization function converges, generating the initial topology configuration of the liquid cooling plate channel.

6. The liquid cooling plate design method based on Bayesian optimization and topology optimization according to claim 5, characterized in that, In step three, the performance indicators include average temperature, maximum temperature difference, and pressure drop. The output performance indicators include: after generating the initial topology of the liquid cooling plate channel, the global flow field, pressure field, and temperature field are obtained through the fluid-structure interaction equations. Then, the average temperature and maximum temperature difference are obtained by performing temperature statistics on the channel design domain. The pressure drop is obtained by calculating the difference between the inlet pressure and the outlet pressure.

7. The liquid cooling plate design method based on Bayesian optimization and topology optimization according to claim 6, characterized in that, In step four, the Bayesian optimization decision layer constructs a comprehensive evaluation function surrogate model based on the performance indicators using Gaussian process regression, including: Construct a comprehensive evaluation function based on performance indicators ; Within the range of decision variables, Latin hypercube sampling is used to select multiple sets of decision variables. Each set of decision variables is sent to the topology optimization execution layer to obtain the corresponding performance index. Based on the performance index, the corresponding comprehensive evaluation function value is obtained, forming an evaluation function sample set including multiple comprehensive evaluation function values. Based on the evaluation function sample set, a comprehensive evaluation function surrogate model is established using Gaussian process regression. The comprehensive evaluation function surrogate model uses weight factors... With coolant inlet flow rate A nonlinear mapping model based on Gaussian process regression fitting, with input variables and a comprehensive evaluation function constructed using performance indicators as the output response; In step four, the decision variables are iteratively searched using the acquisition function, and the decision variables determined in the current iteration are distributed to the topology optimization execution layer. Iteration stops when the accuracy of the comprehensive evaluation function surrogate model reaches the preset accuracy and the comprehensive evaluation function meets the preset convergence condition. This includes: using EI to improve the acquisition function to balance local development and global exploration, iteratively selecting the best decision variables and continuously updating the comprehensive evaluation function surrogate model; repeating the above process until the comprehensive evaluation function meets the preset convergence condition, and the determination coefficient R of the surrogate model reaches the preset convergence condition. 2 If the root mean square error (RMSE) meets the preset accuracy requirements, the final decision variables are obtained. The preset convergence conditions include: the change in the current iteration value of the comprehensive evaluation function is less than a preset threshold in 10 consecutive iterations. The preset threshold is 10. -2 Coefficient of determination R 2 Meeting the preset accuracy requirement with the root mean square error (RMSE) means that: R 2 ≥0.95 and RMSE≤0.

05.

8. The liquid cooling plate design method based on Bayesian optimization and topology optimization according to claim 7, characterized in that, Comprehensive evaluation function as follows: ； in, Indicates average temperature. Indicates the maximum temperature difference. Indicates pressure drop. , as well as All are weighting factors. Take 0.3, Take 0.3, Take 0.4, This represents the minimum average temperature. This represents the maximum value of the average temperature. This represents the minimum value of the maximum temperature difference. This represents the maximum value of the maximum temperature difference. Describe the minimum value of the pressure drop. This indicates the maximum pressure drop.

9. The liquid cooling plate design method based on Bayesian optimization and topology optimization according to claim 1, characterized in that, Step four is followed by step five, which involves binarizing the material density values ​​of the finite element mesh based on the liquid cooling plate flow channel topology to identify the flow channel region and the solid region; performing smoothing processing using morphological opening operation and Gaussian filtering; extracting the flow channel contour and generating vectorized boundaries through cubic spline interpolation; exporting the vector contour in DXF format; and finally importing the vector contour into the CAD platform to complete solid extrusion, Boolean subtraction, and interface encapsulation according to the preset thickness to generate a three-dimensional liquid cooling plate physical model in STEP format.

10. The liquid cooling plate design method based on Bayesian optimization and topology optimization according to claim 9, characterized in that, Step 5 is followed by: Step 6, after assembling the 3D liquid cooling plate physical model and the battery model, a simulation model is established and simulation analysis is performed. If the simulation model meets the preset performance requirements, the current 3D liquid cooling plate physical model is used as the final 3D liquid cooling plate physical model. If the simulation model does not meet the preset performance requirements, the process returns to Step 2 to adjust the constraints and iterate and optimize again until the final 3D liquid cooling plate physical model that meets the preset performance requirements is obtained.

Images